Dynamic Features of camp-dependent Protein Kinase Revealed by Apoenzyme Crystal Structure

Transcription

1 doi: /s (02) J. Mol. Biol. (2003) 327, Dynamic Features of camp-dependent Protein Kinase Revealed by Apoenzyme Crystal Structure Pearl Akamine 1, Madhusudan 1, Jian Wu 1, Nguyen-Huu Xuong 2 Lynn F. Ten Eyck 3 and Susan S. Taylor 1,4 * 1 Department of Chemistry and Biochemistry, University of California, 9500 Gilman Drive San Diego, La Jolla, CA , USA 2 Departments of Biology Chemistry and Biochemistry and Physics, University of California, 9500 Gilman Drive San Diego, La Jolla, CA , USA 3 Departments of Chemistry and Biochemistry, Pharmacology and the San Diego Supercomputer Center University of California, 9500 Gilman Drive, San Diego, La Jolla, CA , USA 4 Howard Hughes Medical Institute, University of California, 9500 Gilman Drive San Diego, La Jolla, CA , USA *Corresponding author To better understand the mechanism of ligand binding and ligandinduced conformational change, the crystal structure of apoenzyme catalytic (C) subunit of adenosine-3 0,5 0 -cyclic monophosphate (camp)- dependent protein kinase (PKA) was solved. The apoenzyme structure (Apo) provides a snapshot of the enzyme in the first step of the catalytic cycle, and in this unliganded form the PKA C subunit adopts an open conformation. A hydrophobic junction is formed by residues from the small and large lobes that come into close contact. This greasy patch may lubricate the shearing motion associated with domain rotation, and the opening and closing of the active-site cleft. Although Apo appears to be quite dynamic, many important residues for MgATP binding and phosphoryl transfer in the active site are preformed. Residues around the adenine ring of ATP and residues involved in phosphoryl transfer from the large lobe are mostly preformed, whereas residues involved in ribose binding and in the Gly-rich loop are not. Prior to ligand binding, Lys72 and the C-terminal tail, two important ATP-binding elements are also disordered. The surface created in the active site is contoured to bind ATP, but not GTP, and appears to be held in place by a stable hydrophobic core, which includes helices C, E, and F, and b strand 6. This core seems to provide a network for communicating from the active site, where nucleotide binds, to the peripheral peptide-binding F-to-G helix loop, exemplified by Phe239. Two potential lines of communication are the D helix and the F helix. The conserved Trp222-Phe238 network, which lies adjacent to the F-to-G helix loop, suggests that this network would exist in other protein kinases and may be a conserved means of communicating ATP binding from the active site to the distal peptide-binding ledge. q 2003 Elsevier Science Ltd. All rights reserved Keywords: crystal structure; apoenzyme; camp-dependent protein kinase; catalytic subunit; preformed active site Introduction Protein kinases are a large and diverse family of enzymes that play a critical role in eukaryotic Abbreviations used: camp, cyclic adenosine monophosphate; C subunit, catalytic subunit; PKA, camp-dependent protein kinase; Apo, apoenzyme; ApoA, apoenzyme structure, molecule A; ApoB, apoenzyme structure, molecule B; R subunit, regulatory subunit; IP20 p, iodinated inhibitor peptide substrate; SP20, substrate peptide; mc, mammalian C subunit; rc, recombinant C subunit; RMSD, root-mean-squared deviation; MPD, 2-methyl-2,4-pentanediol; b-me, b- mercaptoethanol; SIM, signal integration motif. address of the corresponding author: staylor@ucsd.edu signal transduction. 1 They catalyze the phosphoryl transfer of the g-phosphate group from adenosine triphosphate (ATP) to the hydroxyl group of a recipient protein substrate, upon receiving a cue from an upstream signaling protein. Depending on their state of phosphorylation, protein substrates, including protein kinases themselves, are toggled between active and inactive conformational states. Cyclic adenosine monophosphate (camp)- dependent protein kinase (PKA), one of the simpler, better-characterized members of the protein kinase family, serves as a model system. PKA exists as a heterotetramer in the inactive state, with a dimeric regulatory (R) subunit and two catalytic (C) subunits. Upon binding camp, /03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved

2 160 camp-dependent Protein Kinase Apoenzyme Structure Table 1. Summary of apoenzyme data and refinement Space group P2 1 Unit cell dimensions a (Å) 48.8 b (Å) c (Å) 62.8 b (deg.) Resolution (Å) a 2.9 No. of reflections Total 102,700 Unique 18,800 Completeness (%) b 91.8 (82.8/87.1) I/s 23 (2.5/3.6) R sym (%) c 5.5 (49.3/35.6) R refinement (%) 25.7 R free (%) d 29.1 Average B (Å 2 ) All 73 Main-chain 74 Side-chain 71 MPD 75 ApoA Small lobe 78 Large lobe 68 ApoB Small lobe 82 Large lobe 66 RMSD Bond (Å) 0.01 Angle (deg.) 1.38 Ramachandran plot Most favored region (%) 83.3 a Data were processed to 2.8 Å, but for refinement, used only to 2.9 Å. b First number in parentheses is for the highest shell Å, the second is for the second highest shell, Å. c R ¼ ( P ABS(I 2, I. ))/( P I). d 5% of data are in the test set (823 reflections). the C subunits are released from the R subunits to interact with protein substrates. The catalytic core of the C subunit (residues of PKA) is conserved among family members and contains the basic features for binding ATP and protein substrate and for catalyzing phosphoryl transfer. 2 Conformational change is a common mechanism for kinases to orient a phosphoryl leaving group and to eliminate water from the active site. 3,4 Although it is generally agreed that there is a conformational change upon ligand binding, the details of conformational change and the timing of events are less clear. Previously solved crystal structures 5 9 and biochemical and biophysical data show that the dynamic C subunit most likely exists as an ensemble of many structures in solution. Several conformational states of PKA C subunit have been captured in the crystal lattice in order to obtain a detailed picture of the steps of catalysis. Three different conformations have been selected here, an open, an intermediate, and a closed conformation, to represent the conformational dynamics of the active enzyme, which reflects the various parts of the catalytic cycle. Prior to ligand binding, the C subunit is thought to be in its most open conformation, which maximizes accessibility of the active site to solvent and to ATP. The mammalian C (mc) subunit complexed with an iodinated inhibitor peptide (C:IP20 p ) 5,7 is the current structural model, which shows that the two lobes forming the active-site cleft are further rotated away from one another than in other C subunit structures. To further increase the active site accessibility for loading ATP, a segment of the C-terminal tail (residues ), the gate to the active site, is disordered. 8 Upon adenosine binding, the C subunit adopts an intermediate conformation, as shown by the crystal structure of the binary complex (C:Ade), 8 where the two lobes are closer together than in the open state and the C-terminal tail is structured such that the active site is more protected from solvent. ATP plus peptide binding induces a closed conformation, as seen in several crystal structures: C subunit complexed to ATP and IP20, a 20 amino acid residue inhibitor peptide (C:ATP:IP20); 6 C subunit complexed with AMP-PNP, a non-hydrolyzable ATP analog, and IP20 (C:AMPPNP:IP20); 17 and the recently solved transition state mimic of the C subunit complexed with ADP, aluminum fluoride (AlF), and SP20, a 20 amino acid residue peptide substrate (C:AlF:SP20). 9 In C:AlF:SP20, which represents a snapshot at the heart of the catalytic cycle, the two lobes are closest together, and ATP and the substrate hydroxyl group are engulfed. Hydrophobic residues cluster around the substrate site of phosphorylation (the P site) to ensure that the g-phosphate groupis transferred to the correct residue and to exclude water. Until now, an open conformation has been represented by C:IP20 p. While this may suffice for comparison of overall domain rotation, it is not clear how the bound peptide affects the structure of specific residues. The recently solved unliganded Saccharomyces cerevisiae C subunit (TPK1) has the first 79 residues deleted, which could significantly impact the structural findings. 18 To provide a true snapshot of the unliganded recombinant murine C subunit (rc), the apoenzyme structure (Apo) was solved to 2.9 Å resolution by X-ray crystallography. The Apo structure highlights the very different dynamic properties of the two lobes and shows unambiguously that the ATP-binding site, on the surface of the large lobe, is largely preformed and independent of ligand binding. This is in contrast to the Gly-rich loop and Lys72 in the small lobe as well as the C-terminal tail, which are disordered. Finally, the different dynamic properties of the two lobes allows us to appreciate the importance of the hydrophobic core of the large lobe and how this hydrophobic core might provide a mechanism of communication for events at the active site to the distal peptide-binding ledge.

3 camp-dependent Protein Kinase Apoenzyme Structure 161 Figure 1. (A) ApoA and ApoB superimposed. The two molecules in the asymmetric unit are superimposed to show that they differ in overall domain rotation. ApoA is black and ApoB is gold. Broken lines are distances of representative areas of the small lobe. The distance between the Ser53 C a atoms, in the Gly-rich loop, is 1.9 Å. The distance between the C a atoms of Ser339 is 3.4 Å. An MPD molecule and a covalently attached b-me group were seen in both structures. The ApoA MPD and b-me modified Cys199 are pink. Residues were superimposed. (B) The F o 2 F c omit map of Cys199 and the covalently attached b-me, contoured at 3s. Oxygen, red; nitrogen, blue; carbon, gray; sulfur, green. Results Overall structure The Apo crystals of the rc subunit are in the monoclinic space group (Table 1), similar to those seen with the porcine heart C subunit (mammalian C, mc) apoenzyme, 19 which diffracted to only 4 Å resolution. Previous crystals of mc apoenzyme in the cubic space group were used to create a preliminary model to 3.9 Å resolution, 7 but that model showed that the apoenzyme had a conformation similar to that of C:IP20 p, and was never solved to completion. Due to improvements in crystallographic methods, the Apo resolution was Figure 2. (A) Closed conformation and ApoA. Superposition of ApoA (black) and C:AlF:SP20 (green), 9 in the open and closed conformations, respectively. Broken lines show the distances of two representative parts of the small lobe; the Gly-rich loop (Ser53 C a ) and the C-terminal tail (Ser339 C a ). Residues were superimposed. (B) Intralobe hydrophobic contacts. The hydrophobic patch between the small and large lobes, which may provide the grease for the shearing motion associated with domain rotation, is shown. Glu91, a conserved residue in the C helix (C), which is important for orienting the phosphate groups of ATP during phosphoryl transfer, is preformed and is within hydrogenbonding distance from the amide hydrogen atom of Phe185 in the large lobe. From the small lobe (gold ribbon) are residues Glu91, Ile94, Val98, Phe100, Phe102, Leu103, and Val104 (side-chains, blue). From the large lobe, the residues shown are Thr153, Tyr156, Leu162, Tyr179, Gln181, and Phe185 (side-chains, pink). Residues that come into close contact are Ile94-Leu162, Phe185; Val98, Phe100 and Leu103-Tyr156, Leu103-Phe185, and Val104-Gln181. Other hydrogen bonding pairs are: Asn99 amide group to Tyr156 hydroxyl group, and Val104 amide hydrogen atom to Val182 carbonyl oxygen atom. The gray ribbon represents the E helix (E) and the black ribbon includes the Mg-positioning loop (Mg), both from the large lobe. Helix A (A) is shown, since Phe18, Ala22, and Phe26 contribute, peripherally. extended to 2.9 Å 20 and the structure is described here. The Apo crystals have two molecules in the asymmetric unit, ApoA and ApoB, with a solvent content of 51% (v/v) (assuming a molecular density of 1.3 g ml 21 ). The root-mean-squared

4 162 camp-dependent Protein Kinase Apoenzyme Structure Figure 3. (A) Overall average main-chain temperature factors (B-factors). A plot is shown of residue number versus average main-chain B-factor for ApoA (ApoB has a similar profile). Bars represent the various motifs in the C subunit. Residues with an average main-chain temperature factor below 50 Å 2 and a side-chain with an average temperature factor below 60 Å 2 were considered part of the stable core. Regions I, II, and III comprise the stable core. (B) A three-dimensional representation of Apo main-chain temperature factors mapped onto the crystal structures of ApoA (left) and ApoB (right): 0 50 Å 2, red; Å 2, black; Å 2, gold. The core regions I, II, and III correspond to regions in the C helix, the E helix and b strand 6, and the F helix, respectively. Many conserved residues are within the stable core. Balls represent the endpoints of the main-chain breaks. Specific residues in the stable core are 93, 94, 97, and 98 in the C helix, in the E helix, in b strand 6, and , in the F helix. The A helix and the Mg-positioning loop are also stable regions of the protein. deviation (RMSD) between ApoA and ApoB is 0.6 Å overall; however, it is only 0.2 Å for just the small lobe (residues ) or just the large lobe (residues ). This indicates that although the structures of the lobes are very similar, there is a slightly larger domain rotation in ApoA (Figure 1(A)). The higher R-value, 25.8%, may be due to the inherent flexibility of the protein and the disorder of the first eight and ten residues at the N terminus of ApoA and ApoB, respectively, and residues (ApoA) and (ApoB) in the C-terminal tail. In general, the first residues are disordered in the rc subunit structures, regardless of their ligands. In mc, however, the myristic acid moiety stabilizes the N terminus and confers enhanced thermostability, 21 even though mc and rc are biochemically identical. 22 The high average temperature factors (B-factors) of ApoA and ApoB, 74 Å 2, also suggest a dynamic molecule. ApoB adopts a conformation very similar to that of the existing open state model, C:IP20 p. ApoA, on the other hand, is slightly more open than C:IP20 p and all other available structures. The overall RMSD of ApoA and C:IP20 p is 0.8 Å, and for both the small and the large lobes it is 0.5 Å. When the two large lobes of C:IP20 p and ApoA are superimposed (residues ), even though the tip of the Gly-rich loop is displaced in C:IP20 p by 1.9 Å relative to ApoA, the disposition of the Glyrich loop in the two structures is the same relative to the large lobe (not shown). The rest of the small lobe, excluding the Gly-rich loop, is rotated slightly further away from the large lobe in ApoA when compared to C:IP20 p. In the C-terminal tail, there is a 3.3 Å displacement of Ser338 C a, which is representative of the displacement of the small lobe b-sheet and B helix. This supports the idea that the Gly-rich loop acts separately from the rest of the small lobe, 8,23 with the base of the Gly-rich loop as the pivot point. While the differences found in ApoA and ApoB may be due to crystal contacts, ApoA is demonstrative of a conformation that is slightly more open than those that have been seen previously. In comparison to C:AlF:SP20, a closed state structure, Apo adopts an open conformation (Figure 2(A)). When residues from the small and large lobes that are within 4 Å of each other are highlighted (see Figure 2(B) for a list of specific residues), there appears to be a hydrophobic patch in Apo. This grease may provide the necessary glide for domain rotation to occur upon ATP binding and ADP release. Through those interactions, the highly conserved Glu91, in the C helix, appears to be held in place by a hydrogen bonding interaction with the large lobe, which is not seen in the ternary complex structures. Thus, the portion of the small lobe, Glu91 and the C helix are poised to salt-bridge with the conserved Lys72, which orients the a and b-phosphate groups of ATP, as seen in ternary complex structures. The Apo structure contains two components from the crystallization mixture. 2-methyl-2,4-pentanediol (MPD), the precipitating agent in crystal growth, occupies the myristic acid binding site found in mc 7 (Figure 1(A)). An MPD molecule was seen in C:AlF:SP20 in the same position. 9 To account for positive 3s density in the F o 2 F c maps adjacent to Cys199, a covalently attached b-mercaptoethanol (b-me) molecule was built. Although a sulfenic acid group, among other oxidized forms of Cys, was tried, b-me best fit the shape of the density (Figure 1(B)). This modification is likely due to the age of the crystals, six to 12 months, and the reactivity of the Cys. 24 Recently, PKA at Cys199 was shown to be readily glutathionylated. 25 It has been shown with other proteins, by mass spectrometry, that Cys can form

5 camp-dependent Protein Kinase Apoenzyme Structure 163 Figure 4. Stereoview of the ATP-binding site. The ApoA ATP-binding region is shown, which is representative of both molecules. Residues that have been highlighted previously for adenine binding: Leu49, Val57, Ala70, Val104, Met120, Glu121, Tyr122, Val123, Leu173, Thr183, and Phe327 (disordered). Blue C a atoms and C b atoms signify residues for adenine binding that are preformed. Residues that create the ribose subsite are: Leu49, Gly50, Thr51, Val57, Glu127, Glu170, Leu173, and Phe327. Many of these residues move upon ligand binding. Residues important for a and b-phosphate orientation and for Mg 2þ coordination are displayed: Gly52, Ser53, Phe54, Gly55, Lys72, Asn171, Asp184, and Glu91 indirectly through its salt-bridge with Lys72. Green C a and C b atoms signify residues for Mg 2þ and phosphate coordination that are preformed. White C a and C b atoms signify phosphoryl transfer residues that are preformed, Asp166 and Lys168. Many of the ATP-binding residues are preformed. ADP, Mg, and AlF from C:AlF:SP20 are pink. Residues were superimpose from ApoA and C:AlF:SP20. The continuous line from N6 to the carbonyl oxygen atom of Glu121 represents the potential for a hydrogen bond to form upon adenosine binding. Colors are the same as before, where residues with an average main-chain B-factor from 0 50 Å 2 are shown in red, from Å 2 in black, from Å 2 in yellow. mixed disulfide adducts with reducing agents such as b-me. 26 Crystal contacts found in Apo are different from other structures of the C subunit. In particular, Arg133 and Trp196, which are important substrate-binding residues, have been noted in all other structures solved so far. 27 In ApoA, Arg133 contacts symmetry-related ApoB Tyr235, but Arg133 in ApoB does not have any crystal contact. Trp196 in ApoB interacts with symmetry-related ApoA Lys345 and Thr348, but ApoA Trp196 is not making crystal contacts. Table 2. Summary of adenosine-induced changes to ligand binding residues Subsite Adenine Ribose Residues with ligand-dependent ordering upon adenosine binding Phosphates or Mg 2þ SP20 V57 G50 G52-G55 T51 M120 T51 K72 E127 a L173 V57 R133 a F327 E127 F239 a Y330 L173 F327 Preformed residues L49 b L49 b E91 b F129 in Apo A70 b E170 D166 b,c E170 V104 b K168 b L198 E121 N171 b G200 Y122 b D184 b P202 V123 b E203 T183 b P205 E230 Y235 P236 a Highlighted in Figure 5. b Highlighted in Figure 4. c D166, a highly conserved residue, does not contact ligands directly. Substrate-induced conformational changes The effects of MgATP and peptide binding are determined by the comparison of Apo with C:Ade 8 and C:AlF:SP20. 9 To avoid misinterpreting structural changes caused by crystal contacts, only residues in the identical position in ApoA and ApoB are discussed. Side-chain residues present in one molecule, but not in the other, are presumed to be not well structured. Stable core Despite high, overall B-factors, there are three major regions of Apo that have average mainchain B-factors below 50 Å 2 and side-chain residues below 60 Å 2 (Figure 3(A)), the stable core region of the protein. These trough regions define a stable core. When mapped onto the crystal structures of ApoA and ApoB, these regions correspond to the C, E, and F helices, and b strand 6 (Figure 3(B)). Many of these residues are hydrophobic (81%), where Arg93 is considered hydrophobic, since it makes stacking interactions with Trp30. His158 and Tyr164 are considered hydrophobic, since their side-chain polar groups make hydrogen bonding interactions with Asp220 and with the Thr183 carbonyl oxygen atom, while the rest of their side-chain is sandwiched between hydrophobic residues. The small lobe is more malleable as suggested by higher B-factors (Table 1), whereas the large lobe has a stable, hydrophobic core. Of the 27 residues in the core region, 85% are from the large lobe. The two anti-parallel hydrophobic E and F helices were first noted in the first structure solved of PKA. 28 In Apo, these two helices appear to serve as a scaffold, possibly also a folding nucleus, 29 that maintains the tertiary

6 164 camp-dependent Protein Kinase Apoenzyme Structure Figure 5. Substrate-binding ledge of apoenzyme. Residues that interact with the Phe that is 11 amino acid residues N-terminal to the P site (P-11 residue) of SP20 are Tyr235, Pro236, Phe239, Arg133 from the C subunit. Glu203 interacts with the P-6 residue. Glu127, Phe129, and Thr51 together interact with the P-3 Arg of the substrate. Glu170 and Glu230 are important for the recognition of the P-2 Arg, while Asp166, Lys168 and Gly200 are important for orienting the P site Ser during phosphoryl transfer. Pro202, Pro205, and Leu198 create the P þ 1 site, which is important for the recognition of the residue one position C-terminal to the P site. (Left) Shown as sticks are Tyr235, Glu203, Phe129, and Glu230. Asp220 is just below the catalytic loop and may play a role in communicating with the peptide-binding region when ATP is bound in the active site. Residues are shown as a ribbon, while balls represent important substrate-binding residues that become oriented upon binding adenosine; Phe239, Arg133, and Glu127 (Right) Shown rendered is a close-up of Glu127, Arg133, and Phe239 (red). Shown are ApoA, ApoB, C:Ade, 20 and C:AlF:SP20. 9 In C:AlF:SP20, the P-11 Phe is shown (green), which makes hydrophobic interactions with Arg133 and Phe239. The gray ribbon is the protein backbone and in beige are the E, F and G helices. Residues were superimposed to create this Figure. structure of important nucleotide and peptidebinding residues. Interestingly, many conserved residues are embedded within the hydrophobic core, and their positions in the active site are mostly preformed. Polar side-chains within the hydrophobic core make important contacts for structure (Glu155) or are important for catalysis (Arg165 and Asp166). Although it did not make the strict cut-off for the stable core region, the side-chains from a segment of the A helix and from the Mg-positioning loop, the DFG motif, are preformed and relatively stable (side-chain B-factors less than 65 Å 2 ). Adenosine-binding subsite Upon binding adenosine, the C subunit assumes an intermediate conformation, where both the small lobe and the Gly-rich loop are closer to the large lobe than in Apo. 8 When the adenine-binding residues are superimposed in Apo and C:Ade, a number of residues in both structures occupy similar positions (Figure 4, blue C a ) except for Val57, Met120, Leu173, and Phe327, suggesting that the adenine-binding site is largely preformed (Table 2). Val57 and Leu173 side-chains are not seen in one of the two Apo molecules. The side-chain of Met120, which is important for providing a hydrophobic environment for the adenine ring, is pointed away from the active site in Apo, whereas in C:Ade, Met120 points into the active site. As seen in C:IP20 p, Phe327 and this segment of the C-terminal tail are disordered. Following adenosine binding, represented by C:Ade, Phe327 is structured and creates one side of the adenine ring hydrophobic pocket. 8 The structuring of Phe327 is demonstrative of the disorder-to-order transition of the C-terminal tail gate, which helps to regulate access to the active site 30 upon binding adenosine and presumably ATP. This dynamic segment includes Tyr330, the latch (residues ) of the C-terminal tail that interacts with the nucleotide, substrate, and the small and large lobes via an ordered water molecule. 31 Of the preformed residues, the linker backbone contacts are probably the most important for discriminating against GTP and ITP binding in the active site. The C subunit binds ATP with 10 mm binding affinity, whereas it binds GTP (14 mm) and ITP (4 mm) considerably more weakly. 32,33 The primary difference between the nucleotides is the nitrogen atom at the N6 position of ATP, which is occupied by an oxygen atom in GTP and ITP; the electron-accepting group is changed to an electron-donating group. In C:AlF:SP20, the N6 of ATP hydrogen bonds to the carbonyl oxygen atom of Glu121 in the linker on one side and on the other is complemented by the side-chain of Met120. When the nitrogen atom is changed to an

7 camp-dependent Protein Kinase Apoenzyme Structure 165 Figure 6. The hydrophobic core stabilizes the active-site residues. A, Close-up of the catalytic loop and the stable core. This Figure highlights the amphipathic nature of the preformed catalytic loop. In between important catalytic residues are hydrophobic residues, Leu167, Leu172 and, to lesser extent, Pro169 (not shown), which make stabilizing contacts to the hydrophobic core below. Gold spheres represent different regions of the catalytic subunit. Arg165 contacts the Thr197 phosphate group to position the catalytic loop for ATP and substrate binding. Asp166 selects the correct rotomer of the P-site hydroxyl group for efficient phosphoryl transfer. Lys 168 accompanies the g-phosphate group throughout the phosphoryl transfer and neutralizes the charge of ATP. Glu170 forms a salt-bridge with the P-2 Arg from the substrate. Asn171 helps to coordinate the Mg 2þ for charge neutralization. The F-helix is shown below, with Asp220 at one end and Glu230 at the other. The important hydrophobic contacts to Leu167, which help to stabilize the catalytic loop, are shown as broken green lines. B, Tyr164 stabilizes catalytic and Mg-positioning loops. Shown is a close-up of Tyr164, which is nestled under the catalytic loop and the Mg-positioning loop. All polar groups of Tyr164 are hydrogen bonded, which may enhance the hydrophobic environment. The hydroxyl group of Tyr164 hydrogen bonds to the Thr183 carbonyl group the amide nitrogen atom hydrogen bonds to the Glu 220 side-chain, the carbonyl oxygen atom more than likely hydrogen bonds to a water molecule, as observed in the C:AlF:SP20 structure (broken lines). PhosphoThr197, represented by a gold sphere, is critical for the correct structure of the catalytic loop and for catalysis. oxygen atom in GTP and ITP, it can no longer make these favorable interactions. 32,33 Val104 makes a hydrophobic interaction and the hydroxyl group of Thr183 hydrogen bonds to N7. Together, these residues create a contoured surface, which selects ATP binding over GTP or ITP binding, and complements the adenine ring of ATP (Figure 4). Ribose-binding subsite Eight residues important for creating the ribosebinding pocket are listed in Figure 4. The couple of residues that are unique to ribose binding are displaced in Apo relative to the C:Ade structure; Gly50 and Thr51. Glu127, a residue involved in ribose binding, is involved in peptide binding, and will be discussed further in the next section. The distance from Asn171 C a to Gly50 C a, which represents the breadth of this subsite, is 10.1 Å in the closed C:AlF:SP20 structure, 10.4 Å in C:Ade binary structure, and 16.3 Å in ApoA. Glu170 maintains its position regardless of ligands and lends its carbonyl oxygen atom for hydrogen bonding to the ribose 3 0 hydroxyl group, while the sidechain is poised to interact with substrate. Triphosphate-binding subsite Many residues that are important for initially binding and orienting the phosphate groups of ATP are disordered in the Apo structure. Gly52 to Gly55, in the Gly-rich loop, have very high B-factors and the side-chain of Lys72 is not seen in either Apo structure. Adenosine binding alone, even though the phosphate groups are missing, orders Lys72 and closes the Gly-rich loop to an intermediate conformation, as seen in C:Ade. As noted before, Glu91 is preformed in Apo. Asn171 and Asp184, which are important for Mg 2þ coordination, are in the same position, regardless of the presence of ligand. With regard to the g-phosphate-binding site, the residues from the small lobe are significantly flexible, while the residues from the large lobe are nestled in a stable conformation that does not change in the presence or in the absence of ligand. This region, residues (Figure 4, white C a ), includes the highly conserved Asp166, and Lys168, that interacts directly with the g-phosphate group throughout the phosphoryl transfer process. Inactivation experiments showed that Lys168 was not reactive with acetic anhydride in the apoenzyme form, 34 which may be explained by hydrogen bond between Lys168 and Thr201 in C:AlF:SP20 and in Apo. SP20 binding site The C subunit SP20-binding residues that are preformed are Glu203, Phe129, Glu170, Glu230,

8 166 camp-dependent Protein Kinase Apoenzyme Structure Table 3. Multiple functions of the catalytic loop Residue Function Tyr164 p Bridge between the catalytic loop and the Mgpositioning loop Arg165 Bridge to activation loop Bridge to Mg-positioning loop Bridge to F helix via Asp220 Asp166 Catalytic base Leu167 p Hydrophobic anchor to large lobe Lys168 Bridge to g-phosphate group of ATP Pro169 Hydrophobic anchor to large lobe Glu170 Recognition site for P-2 Arg Asn171 Coordinate 2 Mg 2þ Leu172 p Hydrophobic anchor to large lobe Residues in bold are conserved among protein kinase family members. Asterisks ( p ) indicate residues that anchor the catalytic loop to the large lobe. Asp166, and Lys168, and the residues that create the pocket for binding the residue that is one position C-terminal to the P site (the P þ 1 site) (Figure 5, left). Prior to binding ligand, in Apo, three important peptide-binding residues (Glu127, Arg133, and Phe239) are disordered or not oriented properly. Adenosine binding alone, however, is enough to orient these residues in the productive substrate-binding conformation, as seen in the C:AlF:SP20 structure (Figure 5, right). Glu 127 is probably oriented by direct interaction with the ribose 2 0 hydroxyl group (O2 0 ). Since Glu127 also contacts the peptide substrate in the ternary C:AlF:SP20 structure, it may be an important residue for coordinating ATP binding to substrate binding. In Apo the side-chain of Glu127 is not seen; in C:Ade, Glu127 OE2 is 3.7 Å away from the ribose O2 0 ; and in C:AlF:SP20, OE2 hydrogen bonds to O2 0 of ribose and to the Arg of SP20, which is three positions N-terminal to the P site (P-3 Arg). This is suggestive of a mechanism where ATP binding orients Glu127 in a conformation that is favorable to binding peptide. Similarly, the Gly-rich loop adopts an intermediate conformation upon adenosine binding. This conformational change orients the carbonyl oxygen atom of Thr51 close to its position in C:AlF:SP20, where it also hydrogen bonds to the P-3 Arg. How Arg133 is ordered upon adenosine binding is less clear. This residue may be oriented through the D helix, since the amide hydrogen atom of Met128 at the N terminus of the D helix hydrogen bonds to the carbonyl oxygen atom of Leu172, which is a part of the active site. In C:Ade and C:AlF:SP20, Arg133 makes hydrogen-bonding interactions with Glu230, at the C terminus of the F helix. This interaction, which is not present in Apo, creates part of the P-2 site and the P-11 site. Phe239, in the loop between the F and G helices (F-to-G helix loop), is an essential part of the P-11 substrate-specificity site. In Apo, this residue is not oriented properly for substrate binding. However, in C:Ade, Phe239 is oriented properly, despite the C a of Phe239 being 20.5 Å away from the ribose O3 0. On the basis of this observation, it appears that MgATP binding at the active site can be communicated distally. Hydrophobic core network After examining adenosine-induced changes relative to the hydrophobic core, it appears that there are two roles for the hydrophobic core. It first serves as a scaffold that holds important active-site residues in place to create a platform for MgATP binding. A second role of the hydrophobic core may be to communicate ATP binding in the active site to the remote F-to-G helix loop. A stable core region seems to be centered around Leu167 of the catalytic loop (Figure 6(A)), which is reasonably conserved among protein kinases. Many hydrophobic residues converge to create this stable core, which appears to rigidify critical residues for phosphoryl transfer (i.e. Asp166 and Lys168). Another network of interactions extends through Tyr164 in b-strand 6 (Figure 6(B)). This residue forms multiple hydrophobic interactions with the Mg-positioning loop, via Val182 and Phe185, and the E helix through Phe154. The hydrophobic interactions may be further enhanced by the complex network of hydrogen bonds around Tyr164, which is best seen in C:AlF:SP20, where nearly all polar side-chains and all backbone amide hydrogen atoms and carbonyl oxygen atoms are hydrogen bonding to other residues or to ordered water molecules. Tyr164 is highly conserved in other protein kinases as either His or Tyr. It seems largely responsible for organizing and stabilizing the catalytic loop and the Mgpositioning loop by a substrate-independent mechanism. These two layers of interactions, from the catalytic loop to the F helix and from the catalytic loop to the Mg-positioning loop, appear to create a stable MgATP-binding platform. While it has been noted previously that each polar residue of the catalytic loop plays an important role (Figure 6), here, the hydrophobic residues of the catalytic loop are highlighted. Leu167, Leu172, and Tyr164 (from b strand 6) in particular, appear to stabilize the MgATP platform, which has not been reported previously (Table 3). In communicating adenosine binding from the active site to the peptide-binding residues, there appear to be two lines of communication. One is from the active site to the D helix via Glu127, which contacts the ribose moiety directly when ATP is bound. While Glu127 is at one end of the D helix, Arg133, at the other end also becomes ordered upon adenosine binding. In C:AlF:SP20, Arg133 and Glu230 hydrogen bond and create the P-2 site. How Arg133 selects the correct rotomer is not clear. A second possible link from the active site to the peptide-binding ledge is through the F helix, in

9 camp-dependent Protein Kinase Apoenzyme Structure 167 Figure 7. (Left) The F helix communication line. One potential communication line through the hydrophobic core of the large lobe for transmitting ATP-binding from the active site to the peptide-binding region and vice versa. The catalytic loop (beige ribbon) is intimately involved in coordinating phosphoryl transfer. Just below is the F helix (gray ribbon), which may be networked to the catalytic loop via Asp220. Shown from the catalytic loop are: Arg165, thought to be important in structuring the catalytic loop, by salt-bridging to the phosphate group of Thr197 (orange sphere); Asp166 (red); a highly conserved residue that selects the correct rotomer of the P-site Ser; and Glu170, which hydrogen bonds to the P-2 Arg (green sphere) from the substrate. In gold are F helix residues that comprise the network. Asp220 hydrogen bonds to the catalytic loop through the amide hydrogen atoms of Tyr164 and Arg165. Trp222 makes hydrophobic interactions (red arrows) with the conserved Phe238 from the loop between the F and G helix, and with Arg280, which make a highly conserved salt-bridge. Glu230 and Tyr235 make contacts with the substrate P-2 and P-11 residues (green spheres). Tyr204 makes hydrogen-bonding interactions with Glu230 and has been shown to be important in substrate binding. Trp221, a less conserved residue, contributes to the hydrophobic core region, extending in the region of the E helix (not shown). Red sphere, Phe239, represents a distal site in Apo that becomes oriented upon adenosine binding. (Right) Sequence alignment of stable core with other protein kinases. Sequence alignment of PKA (A), CDK2 (C), ERK2 (E), Src (S), and EGFR (G) 2. PKA, CDK2, ERK2 are Ser/Thr kinases, whereas Src and EGFR are Tyr kinases. A number of the hydrophobic residues within the core region are highly conserved among different kinase family members (gold), suggesting that the F helix communication network may be conserved among protein kinase family members. two ways. As a rigid body, the F helix may communicate ATP binding from Asp220, a conserved residue that is directly below the catalytic loop, to Glu230, which is important for substrate binding. In this way, the F helix would act like a seesaw that can sense a change at one end and transmit information to the other end. A second, and more likely, network, is through hydrophobic interactions between Phe238 and the conserved Trp222 (Figure 7, left). In C:AlF:SP20, Lys168, which is two positions downstream from Asp166, makes salt-bridge interactions with AlF, a mimic of the phosphoryl group in the transition state. Directly below Asp166, Asp220 hydrogen bonds to the amide groups of Tyr164 and Arg165. Through a turn of the F helix, this network is relayed from Asp220 to Phe238 in the F-to-G helix loop by hydrophobic interactions with Trp222. On the other side, Trp222 is buttressed by the Arg280- Glu208 conserved ion pair, which seems to stabilize the activation loop, but may help to organize this network. Finally, Phe239, also in the F-to-G helix loop, seems to be oriented for substrate binding upon binding adenosine and presumably ATP through this extensive network from the active site. The second model is appealing, because it would justify many of the highly conserved residues in this region; namely, Asp220, Trp222, Phe238, and the Glu208-Arg280 ion pair (Figure 7, right). The F helix (residues ), as a signal integration motif (SIM), does not contact substrate or peptide directly, but it appears to contribute to ATP binding and substrate binding. 35 Trp222, in particular, stabilizes the activation loop, which is important for ATP and substrate binding, and appears to communicate ATP binding to the peptide binding F-to-G helix loop. Since the family of kinases has a conserved feature of binding nucleotide and binding substrate, albeit different substrate sequences, it is likely that the mechanism of signaling ATP binding from the active site to the substratebinding site, will be conserved. The conformational changes that occur in peptide-binding residues upon adenosine binding could explain the synergistic, high-affinity binding of the protein kinase inhibitor (PKI) by ATP binding. 36 Likewise, these structural changes would support the idea that ATP binding precedes substrate binding, since adenosine would facilitate the ordering of substrate-binding residues. 37,38 Additionally, the network may be

10 168 camp-dependent Protein Kinase Apoenzyme Structure able to communicate in the opposite direction to enhance ATP binding when peptide is present. Discussion The active form of the catalytic subunit is a highly dynamic protein, as demonstrated by a variety of solution methods including fluorescence anisotropy, 39 fluorescence energy transfer, 13 chemical mapping, 10 small-angle scattering, 16 and hydrogen/deuterium exchange. 11,15 To appreciate the molecular features of the C subunit in various stages of catalysis, we have attempted to capture some of these dynamic features in a crystal lattice. The Apo structure provides a snapshot of the C subunit in the unliganded state. The dramatic differences between the two lobes highlights an active site that is partially preformed, and a unique hydrophobic network that permeates the large lobe and allows for communication of binding event that can then be translated long distances to mediate conformational change. In terms of the catalytic cycle, Apo represents the enzyme prior to any ligand binding. The structural results here are completely consistent with solution experiments that show Apo exists in an open, solvent-exposed conformation. 10,13 In terms of the open and closed states, Apo represents the most open form reported to date. Domain closure has been characterized as having hinge motions or shear motions. 40 In adenylate kinase, the domains rotate 908 relative to one another by a hinge motion upon substrate binding. On the other hand, a shear motion is utilized for closing the active sites of alcohol dehydrogenase, aspartate aminotransferase, and citrate synthase. Hinge motions are typically used when the two domains are not in contact with one another except by a linker, while shear motions are employed when domains have close-packed segments, usually involving helices. In the past, the PKA C subunit, based on the concerted motion of the small lobe to close the cleft, suggested that the hinge bending was transmitted through residues 30 40, the base of the Gly-rich loop, and the linker (residues ). 8 Smallangle scattering suggested that a Gly in the linker served as a hinge for the domain rotation. 16 Molecular dynamics calculations suggested that the hinge was not a simple one and involved residues 106 and 107 as well as The PKA C subunit appears to utilize both hinge and shear motions. The linker serves as the hinge, while the hydrophobic patch between the two lobes, described here for the first time in a comprehensive way, appears to provide the glide for the shear motion. The inherent flexibility of Apo helps to understand the wide range of competitive inhibitors of ATP that PKA can bind, including staurosporine, the isoquinolinesulfonyl family of inhibitors, and balanol The Apo structure provides a more comprehensive understanding of the dynamic properties of this enzyme. Apo provides a picture of what ligands see prior to binding to the C subunit and may aid in the design of specific PKA inhibitors. A stable core, the C, E and F helices, and b sheet 6, appears to rigidify residues in the active site and in the peptide-binding ledge, despite their exposure to solvent. The MgATP-binding residues that are preformed and stable in Apo are Glu91, Val104, Thr183, Glu170, Asn171, Asp184, and Lys168 (Figure 4, Table 2). This preformed ATP docking platform creates a surface that can differentiate between GTP and ATP. The comparison of Apo with the adenosine complex (C:Ade) reveals that adenosine binding alone is sufficient to order both active-site residues and peptide-binding residues. Specifically, upon adenosine binding, in addition to domain rotation, Met120 points into the active site and Lys72 and Glu127 become ordered and poised to bind the ribose and phosphate groups of ATP and peptide. When adenosine is bound, the C-terminal tail becomes structured over the active site, as represented by Phe327 and Tyr330, and the Gly-rich loop adopts an intermediate conformation. 8 The preformed residues important for substrate binding are Asp166 at the core and peripheral to that are Glu203, Phe129, Glu170, Glu230, Pro202, Pro205, Leu198, regions that complement substrate binding from the P-6 Arg site to the P þ 1 site. In addition to domain rotation and the structuring of ATP-binding residues, important substrate-binding residues in the D helix and the loop between the F and G helices (Glu127, Arg133, and Phe239) are oriented upon binding adenosine (Figure 5). Adenosine binding at the active site appears to communicate to the peptide-binding ledge through the D and F helices, which converge to create the P-11 site. The F helix, as a SIM, stabilizes the activation loop, and appears to serve as a conduit to communicate MgATP binding from the active site to the distal peptide-binding site via the Trp222-Phe238 network of interactions (Figure 7, left). This hydrophobic network is comprised of mostly conserved residues, some which have not been fully appreciated. Because of the sequence conservation in this network (Figure 7, right), it may be a common mechanism for protein kinases to communicate MgATP-binding to the peptidebinding ledge. Materials and Methods Protein expression Recombinant murine PKA C a (40.8 kda) was purified as described. 44 Briefly, the C subunit was expressed in BL21 Escherichia coli cells grown for six to eight hours in LB medium prior to induction of protein expression with IPTG. Protein was purified on a P11 cellulose column. The different phosphorylated species of PKA were separated by ion-exchange using a salt gradient. Protein purity was confirmed by SDS-PAGE and

11 camp-dependent Protein Kinase Apoenzyme Structure 169 isoelectric focusing gel, both stained with Coomassie brilliant blue. Peak II, autophosphorylated on Ser10, Thr197, and Ser338, was used for crystallization. 44 For crystallization, purified protein was dialyzed into a 50 mm Bicine buffer (ph 8.0) with 150 mm ammonium acetate and 10 mm b-me overnight, then concentrated to 8 10 mg ml 21 using a Millipore concentrator. The protein concentration was determined by a Bradford assay using bovine serum albumin (BSA) as the standard. Crystallization, data collection and processing, structure determination, and model analysis 2-Methyl-2,4-pentanediol (MPD) and ammonium acetate were purchased from Aldrich. Crystals grew at 4 8C by the vapor-diffusion, hanging-drop method with a 1 ml reservoir solution, which contained 0.1 mm Tris HCl (ph 7.5) and 10 20% (v/v) MPD. The drop (20 ml) contained 3% MPD, 0.1 mm Bicine (ph 8.0), and 7.5 mg ml 21 of protein. 20 All solutions were sterile filtered. Crystals were temperature-sensitive and took six months to one year to grow. Crystals were cryoprotected with quick dips in a gradient of increasing concentration of MPD in 0.1M Bicine (ph 8.0), starting with 15% and ending with 30% in increments of 5%. Later it was shown that a single dip in the final concentration of MPD was sufficient for cryoprotection. Crystals were mounted in nylon loops and flash-frozen in the liquid nitrogen stream at 97K. Data were collected at the Stanford Synchrotron Radiation Laboratory (SSRL) in Palo Alto, CA, on beamline 7-1 (l ¼ 1.08 Å) using a Mar image plate. All data were collected from a single crystal. Although crystals diffracted to 2.8 Å, data were used only to 2.9 Å (Table 1). In all, 318 images were processed using Denzo and Scalepack. 45 CNS was used for molecular replacement and refinement using a maximum likelihood target function. Molecular replacement was used to calculate phases using C:IP20 p as the starting model. 5 The rotation and translation function were determined with data from 20 to 4 Å. After an initial rotation (two highest solutions) and translation were found, all residues were changed to Ala to avoid model bias and the B-factors were set to 50 Å 2. The first rigid body refinement resulted in an R-factor of 47.8%. The quality of the model was primarily judged by the decrease in R free. Restrained non-crystallographic symmetry was used for the whole molecule initially. Residues that were clearly different in the electron density maps of ApoA and ApoB, were no longer restrained. In the end, the residues included in the NCS symmetry definition were 12 14, 16 18, 26 36, 38 43, 45 52, 55 56, 62 66, 69 72, 80 86, , , , , , , , , , , , , , , , , , , , , , , and 350. One round of high-temperature (4000 K) simulated annealing was done to further remove model bias. In the initial stages of refinement, no B-factor correction was used. In subsequent rounds of refinement, energy minimization, an initial anisotropic B-factor correction, isotropic B-factor refinement, and bulk solvent correction (with a cut-off of 6 Å) were employed. For refinement, a 2s cut off for structure factors was used, but no cut-off was applied in calculating the maps. All data were used for the calculation of the R-factor. TurboFrodo and Tom were used for model building. The stereochemical parameters were good, with no residues in the disallowed region of the Ramachandran plot, as determined by PROCHECK. 46,47 In ApoA, 74 residues were modeled as Ala, since their side-chains were not seen in the electron density maps; there were 83 in ApoB. Cys b-me adducts were confirmed by simulated-annealing omit maps. Hydrogen bonding was determined by the hydrogen bond function in InsightII (Molecular Simulations Inc, San Diego, CA). All Figures were made with InsightII, except for Figure 1(B), which was made in Bobscript v Protein Data Bank accession code The coordinates are deposited with the RCS Protein Data Bank with accession code 1J3H. Acknowledgements Special thanks to Narendra Narayana for determination of crystallization conditions; to Chung Wong and Jie Yang for insightful discussions; to Nick Nguyen, who maintains the X-ray facility at University of California, San Diego; to Chris Neilson for computer support; to Cindy Gribskov and Teresa Clifford, who purified the C subunit protein; to Mary Ellen Perry and Genna Yuasa- Acosta for help with manuscript preparation; and to the NIH for financial support. References 1. Plowman, G. D., Sudarsanam, S., Bingham, J., Whyte, D. & Hunter, T. (1999). The protein kinases of Caenorhabditis elegans: a model for signal transduction in multicellular organisms. Proc. Natl Acad. Sci. USA, 96, Hanks, S. K. & Hunter, T. (1995). The eukaryotic protein kinase superfamily: kinase (catalytic) domain structure and classification. FASEB J. 9, Blake, C. C. F. (1978). The trouble with kinase crystals. Nature 271, Anderson, C. M., Zucker, F. H. & Steitz, T. A. (1979). Space-filling models of kinase clefts and conformation changes. Science, 204, Karlsson, R., Zheng, J. H., Xuong, N. H., Taylor, S. S. & Sowadski, J. M. (1993). Structure of the mammalian catalytic subunit of camp-dependent protein kinase and an inhibitor peptide displays an open conformation. Acta Crystallog. sect. D, 49, Zheng, J., Knighton, D. R., ten Eyck, L. F., Karlsson, R., Xuong, N., Taylor, S. S. & Sowadski, J. M. (1993). Crystal structure of the catalytic subunit of campdependent protein kinase complexed with MgATP and peptide inhibitor. Biochemistry, 32, Zheng, J., Knighton, D. R., Xuong, N. H., Taylor, S. S., Sowadski, J. M. & Ten Eyck, L. F. (1993). Crystal structures of the myristylated catalytic subunit of camp-dependent protein kinase reveal open and closed conformations. Protein Sci. 2, Narayana, N., Cox, S., Nguyen-huu, X., Ten Eyck, L. F. & Taylor, S. S. (1997). A binary complex of the catalytic subunit of camp-dependent protein kinase