- Email: [email protected]
Mutants of Protein Kinase A that Mimic the ATP-binding Site of Protein Kinase B (AKT)
doi:10.1016/S0022-2836(03)00518-7
J. Mol. Biol. (2003) 329, 1021–1034
Mutants of Protein Kinase A that Mimic the ATP-binding Site of Protein Kinase B (AKT) Michael Gaßel1, Christine B. Breitenlechner2, Petra Ru¨ger3 Ute Jucknischke3, Thorsten Schneider1, Robert Huber2 Dirk Bossemeyer1* and Richard A. Engh2,3* 1
German Cancer Research Center (DKFZ), Division of Pathochemistry, Im Neuenheimer Feld 280, 69120 Heidelberg, Germany 2
Max Planck Institut fu¨r Biochemie, Abteilung Strukturforschung, Am Klopferspitz 18a, 82152 Martinsried, Germany
3 Roche Diagnostics GmbH Pharmaceutical Research Nonnenwald 2, 82377 Penzberg Germany
The mutation of well behaved enzymes in order to simulate less manageable cognates is the obvious approach to study specific features of the recalcitrant target. Accordingly, the prototypical protein kinase PKA serves as a model for many kinases, including the closely related PKB, an AGC family protein kinase now implicated as oncogenic in several cancers. Two residues that differ between the a isoforms of PKA and PKB at the adenine-binding site generate differing shapes of the binding surface and are likely to play a role in ligand selectivity. As the corresponding mutations in PKA, V123A would enlarge the adenine pocket, while L173M would alter both the shape and its electronic character of the adenine-binding surface. We have determined the structures of the corresponding double mutant (PKAB2: PKAa V123A, L173M) in apo and MgATP-bound states, and observed structural alterations of a residue not previously involved in ATP-binding interactions: the side-chain of Q181, which in native PKA points away from the ATP-binding site, adopts in apo double mutant protein a new rotamer conformation, which places the polar groups at the hinge region in the ATP pocket. MgATP binding forces Q181 back to the position seen in native PKA. The crystal structure shows that ATP binding geometry is identical with that in native PKA but in this case was determined under conditions with only a single Mg ion ligand. Surface plasmon resonance spectroscopy studies show that significant energy is required for this ligand-induced transition. An additional PKA/PKB mutation, Q181K, corrects the defect, as shown both by the crystal structure of triple mutant PKAB3 (PKAa V123A, L173M, Q181K) and by surface plasmon resonance spectroscopy binding studies with ATP and three isoquinoline inhibitors. Thus, the triple mutant serves well as an easily crystallizable model for PKB inhibitor interactions. Further, the phenomenon of Q181 shows how crystallographic analysis should accompany mutant studies to monitor possible spurious structural effects. q 2003 Elsevier Science Ltd. All rights reserved
*Corresponding authors
Keywords: PKA; PKB; protein kinase; crystal structure; point mutation
Introduction Abbreviations used: PKA, cAMP-dependent protein kinase; PKB, protein kinase B (also called Akt); PKAB2, double mutant PKA V123A/L173M; PKAB3, triple mutant PKA V123A/L173M/Q181K; PKI, protein kinase inhibitor; H7, 1-(5-isoquinolinesulfonyl)-2-methylpiperazine; H9, N-(2-aminoethyl)-5-isoquinolinesulfonamide; HA1077, (Fasudil), (5-isoquinolinesulfonyl)homopiperazine. E-mail address of the corresponding authors: [email protected]; [email protected]
The protein kinases represent one of the largest families of human proteins, with more than 500 protein kinases identified by the human genome project. Considering variations arising from population heterogeneity, differential splicing, and enhanced mutation rates caused e.g. by cancer, many thousands of distinct sequences will occur in the human population alone. Despite the variations in sequence, the fold of the active protein kinase catalytic domain is well conserved; inactive
0022-2836/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved
1022
PKA-PKB Hybrid Mutants
Figure 1 (legend opposite)
protein kinase structures differ but cluster into protein kinase subfamilies. As a consequence of the conservation of the active structure, many properties can be analysed with respect to relatively few sequence positions which define the property, for example, selectivity of binding at the ATP-binding pocket. The large number of kinases fuels concern that
new drugs designed against certain protein kinase targets will fail in clinical tests because of unexpected activity against other protein kinases. Compounds STI571 (Glivec) and HA1077 (Fasudil) have demonstrated that protein kinase inhibition is a commercially viable therapeutic strategy, but represent the only protein kinase inhibitors on the market.1 Thus, kinase inhibitor selectivity will
PKA-PKB Hybrid Mutants
1023
Figure 1. (a) Alignment of human PKAa and PKB isoforms. Amino acid residues highlighted with blue background show positions that are not identical between PKAa and PKBa. Amino acid residues marked in yellow are exchanged in this study. Amino acid residues highlighted in green (P101, L103, E121, D175, I180) form the conserved surface hydrogen bonding network that binds Q181 in PKA. The charge of R308 (orange) in this network is conserved but not the residue identity (K in PKB). T183 (red) is subject to ligand induced rotation. See the text for details. (b) Stereo view of the solvent-accessible surface of PKAa colored according to identity with PKBa. PKA surface elements that differ from PKB are colored blue, while atoms from identical residues are colored gray. Cyan ball-and-stick atoms show the positions of substrate-mimic inhibitor PKI(5-24); multicolored ball-and-stick atoms depict AMP-PNP. Differences are distributed over the surface rather uniformly, but are limited to a few residues in the ATP-binding site. (c) Close-up of the solvent-accessible surface in the ATP pocket. Surfaces from two non-conserved side-chains closest to AMP-PNP are colored violet. These positions, V123 and L173, were first mutated to A and M, respectively, to mimic the PKB-binding pocket. The blue surface arises from V104, which is the isosteric threonine residue in PKB isoforms. ATP hydrogen bonding interactions are shown in green.
remain a crucial issue for the foreseeable future, intensifying the critical role that mutagenesis and protein crystallography possess in identifying selectivity mechanisms. The kinases share a common bi-lobal catalytic domain structure with the ATP-binding site in the interlobe cleft. ATP is bound to the kinase via interactions with some 15 residues of the protein, including about ten side-chain interactions that therefore represent especially important potential selectivity determinants. Thus, the essential binding properties of ATP and other ATP-site ligands should, in many cases, be simulated for a protein kinase by a limited set of point mutations of a closely related protein kinase. The construction of such hybrids has been demonstrated for kinases.2 Even when subtle differences lead to quite different overall reaction kinetics, for practical ligand design purposes most or all protein – ligand interactions are likely to be identical for both the target kinase and its model. That a single residue can be identified as the principal selectivity determinant for an inhibitor type by mutation of a series of kinases verifies this approach.3 Along these lines, we have chosen the cAMPdependent protein kinase (PKA) as surrogate kinase for another closely related member of the
AGC kinase family, protein kinase B or Akt. PKA responds via cAMP to a great variety of stimulatory agents in the regulation of ion conduction, memory, gene transcription, muscle contraction, and metabolism.4,5 PKA has historically enjoyed a central role as prototypical protein kinase for functional studies,6 – 8 in particular because recombinant expression, purification, and crystallization is well established. PKB is involved in insulin signaling pathways and plays a central role in apoptosis; disregulation is associated with several cancers including non-small cell lung carcinoma,9 prostate carcinoma,10 pancreatic carcinoma11 and colorectal carcinoma.12 Thus, as an important and emerging drug target, with a well characterized surrogate (PKA), PKB is an ideal candidate as a target kinase in our approach. Further, the very similar structures of the active forms of PKA and PKB13 support the expectation that side-chain differences on highly similar backbone scaffolds will be responsible for much of their relative ATP-site selectivities. Since inhibitor design requires facile cocrystallization, which has not been demonstrated for PKB, we remain interested in PKA-based models for PKB binding, and enjoy, additionally, the validation offered
1024
by the PKB structure. As a first step toward generation of a PKA-based model for the PKB ATP-binding site, we produced the double mutant PKA (V123A, L173M), or PKAB2. These two residues in the ATP-binding site represent the differences between isoforms PKAa and PKBa that form contacts with ATP in PKA structures and that must alter the shape of the binding pocket. Both are at the adenine-binding site. One of these, V123A, enlarges the pocket, while the other, L173M, possesses a largely unchanged side-chain volume but changes the shape of the adenine-binding surface and introduces a sulfur atom for potential adenine interactions. These mutants were characterized by means of X-ray crystallography and surface plasmon resonance spectroscopy binding assays. The ligand-free structure of the double mutant (PKAB2) shows an unanticipated structural change. The side-chain of Q181, which is normally exposed to solvent and points away from the ATP-binding site, adopts instead a new x1 rotamer conformation, which places the side-chain at the adenine-binding site. In this orientation, ATP binding is partially blocked, and both the shape and polar character of the binding pocket are altered. The structure of PKAB2 complexed with ATP, however, shows that ligand binding restores Q181 into its position known from PKA. Presumably, ligand binding in the ATP site is weakened by the energy required for the ligand-induced effect. In PKB, the residue corresponding to Q181 of PKA is lysine. We therefore constructed the triple mutant V123A, L173M, Q181K (PKAB3). This triple mutant in the unliganded form shows K181 “restored” to a position distant from the catalytic site and embedded in a surface hydrogen-bond/salt-bridge network. The mutated residues in the adenine-binding site of PKAB3 adopt conformers identical with the corresponding residues of PKB. Binding affinity studies with ATP and several isoquinoline inhibitors using surface plasmon spectroscopy support the conclusions that the double mutation introduces a defect into the kinase that weakens ligand binding, and that the triple mutant restores native-like behavior. We describe these results and discuss their relevance for PKB inhibition and inhibitor design.
Results and Discussion Sequence similarity of PKA and PKB An alignment of the primary sequences (Figure 1(a)) between PKAa and PKB isoforms illustrates the affiliation of these kinases, with amino acid identities between e.g. PKAa and PKBa of 45% over 294 amino acid residues. This high degree of similarity gives confidence in the quality of homology models of the kinase domain
PKA-PKB Hybrid Mutants
of PKB based on the structure of PKA.14 Further, the recently published structures of activated forms of PKBb13 verify the validity of homology modeling for the active form. The unconserved amino acid side-chains are rather dispersed and occur especially at the protein surface. Their occurrence at the peptide substrate-binding sites plays a direct role in peptide substrate selectivity (Figure 1(a) and (b)). The ATP-binding site is highly conserved, with only three amino acid substitutions in the adenine pocket (V123A, L173M and V104T; Figure 1(c)). Of these three, substitutions V123A and L173M alter the shape of the pocket, while V104T would possibly introduce a hydrogen bonding partner, although its orientation in the PKB structure shows only hydrophobic interactions with ATP. L173M potentially alters the electronic character of the adenine-binding site, especially with the addition of a sulfur atom. To test PKA mutants as improved models for PKB we first made the double mutant V123A, L173M (PKAB2) and solved crystal structures of the ATPfree and ATP-bound structures. After observing that the conformation of Q181 is modified by these mutations (see below), we made the triple mutant V123A, L173M, Q181K and solved its crystal structure. Overall structures The published structures of mammalian PKAs include sequences from mouse (Mus musculus), pig (Sus scrofa), and cow (Bos taurus), which differ from the human (Homo sapiens) sequence at seven, four, and two positions, respectively. These structures have clarified many PKA properties, including mechanisms of phosphotransfer, binding interactions with the peptide inhibitor PKI-(5-24), substrates and pseudosubstrates based thereon, with Mn2AMP-PNP, AlFADP, and with a variety of natural and synthetic inhibitors.15 – 19 In addition, dynamic properties have been studied by fluorescence depolarization,20 and 31P NMR.21 The crystal structures of PKAB2 (PKA V123A, L173M) and PKAB3 (PKA V123A, L173M, Q181K) as binary complexes with PKI, and of PKAB2 as the MgATP þ PKI ternary complex possess the general structural features of PKA known from crystal structures. The individual N and C-terminal lobes are largely superimposable, and the structures are in a closed conformation. This is consistent with crystallization in the P212121 space group with cell dimensions a ¼ 73:5 A; b ¼ 76:7 A; 22 However, the PKAB3 structure c ¼ 80:5 A: crystallized in a new crystal packing arrange ment (P212121 with cell dimensions a ¼ 59:9 A; and c ¼ 100:9 A), possibly as a result b ¼ 79:4 A; of the change in charge distribution at the surface of the protein behind the hinge region. It retains the closed conformation in this packing arrangement.
PKA-PKB Hybrid Mutants
1025
Figure 2. Sim-weighted electron density (2mFo 2 dFc, blue at 1s) and difference electron density (mFo 2 dFc, white at 2s and red at 2 2s) maps showing the rotation of Q181. The maps were calculated from a model of PKA (orange sticks) after refinement but prior to rotation of the side-chain. Green sticks depict the refined structure of the double mutant PKAB2. The density shows unambiguously how Q181 rotates to occupy the cavity near the V123A position and near the adenine-binding interaction sites.
The Q181 switch Q181 adopts an “internalized” rotamer conformation The unliganded PKAB2 structure shows a surprising new rotamer conformation of Q181, with the side-chain amide group not at the surface of the protein as in PKA, but moved toward and
partially occluding the ATP-binding site (Figure 2). This conformation is verified unambiguously by the electron densities calculated from the PKA molecular replacement model. The rotation is apparently enabled by the creation of the cavity associated with the amino acid exchange V123A. Superposition of the unliganded PKA and PKAB2 structures show that close contacts with V123 prohibit this
Figure 3. Comparison of the different positions of Q181 in unliganded PKAB2 (orange sticks), in the ATP– PKAB2 complex (multicolored by atom type with white carbon atoms), and in the AMP-PNP – PKA complex (green sticks, 1CDK16). The distance measurement between Q181 of unliganded PKAB2 and V123 of native PKA shows how the rotation of Q181 is prevented in the wild-type by V123. Furthermore, it can be seen that in PKAB2-ATP as well as in PKA-AMPPNP, Q181 occupies the same position.
1026
PKA-PKB Hybrid Mutants
Figure 4. (a) Stereo view of Q181 distance measurements of the unliganded PKAB2 and comparative measurements in an overlay of the PKAB2-ATP complex. Binding of ATP does not obviously conflict with Q181 in this geometry. However, ATP binding may require a rotation of T183 that disrupts the space-filling packing of Q181, V104 and T183. This leads to the exit of Q181 to form more energetically favorable polar interactions at the protein surface. (b) Stereo view of the rotation of T183 between the liganded and unliganded PKAB2 structures. The distances and geometries show how the rotation of T183 is essential to form the contact between T183:OG2 and ATP:N7.
conformation of Q181 in wild-type PKA (Figure 3). Other changes associated with the “internal” glutamine rotamer include a x1 rotation of T183 that places ˚ from Q181:OE2, and a displaceCG2 at about 3.4 A ment of Val104:CG1 (Figure 4(a)). Structural flexibility, particularly of these two residues, has been reported.18,23 In the PKAB2 structure, these residues form a triangle of close non-polar (or weakly polar ˚ ) between Q181:OE1, C–H–-O) interactions (,3.4 A V104:CG1 and T183:CG2 (Figure 4(a)). The binding of ATP restores Q181 Observation of this change intensified our efforts to cocrystallize the PKAB2 mutant with ATP. The successful structure determination showed that
MgATP binding forces the residue Q181 back in its original position (Figure 3). The structure of Q181 in PKAB2-MgATP is indistinguishable from its structure in PKA. Binding of MgATP abolishes the interactions between V104, T183, and Q181, because Q181 is restored to its wild-type position, ˚, and because V104:CG1 is displaced by some 1.6 A and because T183 undergoes a rotamer conversion (Figure 4(a)). In its position at the hinge, the glutamine residue is stabilized by packing contacts but does not block ATP binding completely. On the contrary, an energetically favorable contact ˚ ) might between Q181:OE2 and ATP:N6 (at 2.8 A seem possible with this Q181 rotamer (Figure 4(a); red dotted line). However, the rotamer of T183 in the PKAB2 crystal form does not enable the typical
PKA-PKB Hybrid Mutants
adenine anchoring link contact between ATP:N7 and T183:OG1 (Figure 4(b)). Taken together, the hinge position of Q181 would seemingly still allow binding of MgATP, but the rotamer conversion of T183 apparently to enable the T183:OG1 ATP:N7 contact removes T183:CG2 from the T183V104-Q181 triangle and disrupts this stable packing arrangement so that Q181 exits the pocket. An intermediate structure that has Q181 at the hinge position but T183 in an ATP-binding geometry might hypothetically exist under some conditions or with certain ligands, but we have not observed it. The propensity of Q181 to switch between binding at the adenine site near the kinase hinge, or at the outer surface of the protein, may not play a physiological role for PKA, since it is observed only in conjunction with the V123A mutation. However, among the hundreds of natural kinases, appropriate sequences may exist that allow a similar transition. What makes this possibility particularly intriguing is the fact that this property represents a possible mechanism to communicate the occupation state of the ATP pocket to a potential co-factor or conversely to alter ATP-site ligand-binding properties via cofactor binding. Should such a mechanism occur in a natural kinase, it would offer a means for selective inhibition as well. MgATP binding in PKAB2 The position of ATP in the double mutant PKAB2 is very similar to that observed in the
1027 wild-type enzyme (Figure 5; 1CDK;16 1ATP24). A major exception is that a single Mg ion is present in the complex, at the position coordinating the b and g phosphate groups of ATP. The crystallization conditions included concentrations of 3 mM Mg2þ and 1.5 mM ATP. Kinetic studies of PKA have shown that lowering the concentration of free Mg2þ from 10 mM to the more physiological concentration of 0.5 mM causes the occupancy of the second site to shift from 80% to 20%.25 Thus, although unusual for PKA – ATP complexes, the observation of a single bound metal ion is consistent with expected binding strengths: the apparent millimolar binding constant for the second metal-binding site determined from kinetic studies is roughly the concentration of metal ion in the crystallization solution. This millimolar binding constant may be weakened further by crystallization conditions and/or mutations. The Mg ion in PKAB2 –MgATP is in the position of the activating16 metal ion. Figure 5 shows an apparent lengthening of the coordination distance between Asp184 and Mg, but this is not within the resolution of the structures and may arise from the use of different refinement parameters. A ˚ crystal structure of PKA– Mg2ATP has been 2.7 A reported26 (coordinates not deposited) but the coordination geometry was not described in detail. Interestingly, Zheng et al.26 reported low occupancies of the second metal-binding site at concentrations of Mg2þ below a 2.5-fold excess over ATP.
Figure 5. Stereo view of an overlay of the MgATP– PKAB2 complex structure (multicolored by atom type) and the MnAMP-PNP– PKA complex structure (green sticks, PDB code 1CDK16). The 2Fo 2 Fc density (1s) belongs to PKAB2 liganded with ATP. The PKAB2 structure is the first Mg ATP – PKA structure and shows only a single bound metal ion.
1028
PKA-PKB Hybrid Mutants
Figure 6. Stereo view of the interactions of M173 SD in the ATP-liganded PKAB2 structure.
Sulfur interactions The L173M mutation introduces the possibility of structured sulfur –ligand interactions. Such interactions have been the subject of recent theoretical and statistical analyses,27,28 including observations of non-random distributions of e.g. methionine around tryptophan.† Earlier studies predicted that nucleophiles (e.g. oxygen, aromatic pi molecular orbital clouds) would tend to approach the sulfur ATOM along the posterior extensions of the CE –SD or CG –SD bonds, while electrophiles (e.g. metal ions) would approach SD from a direction roughly (within 408) perpendicular to the CG –SD – CE plane.29 In the PKAB2 –ATP complex, the sulfur atom of Met173 makes two corresponding interactions (Figure 6), one with the carbonyl oxygen atom of G126 and one with the aromat of ATP. The backbone carbonyl oxygen atom of G126 approaches the Met173 SD atom at a ˚ and with a nearly 908 formed by distance of 3.78 A C – O – SD. The second sulfur contact is formed with the aromat edge at N3 with a distance of ˚ . The contact distances are consistent with 3.85 A specific sulfur interactions. However, since the resolution of the structure does not establish the position of the Met CE atom unambiguously, details of the interaction geometry and interaction type are not possible to analyze here. Further structural studies are required for this and to establish the degree of flexibility of the methionine residue. Methionine is conserved at this position in all PKB isoforms, and in some PKC isoforms and other kinases including S6K, Kin1/2. Another methionine residue, M120 at the hinge region of the adenine-binding site (conserved also among PKB isoforms and other AGC kinases) provides † Internet Atlas of Protein Side-Chain Interactions: http://www.biochem.ucl.ac.uk/bsm/sidechains
additional sulfur – ATP aromatic interactions. These are edge (co-planar), and occur between Met120 and N6 and N7 of ATP with distances ˚ to approximately 4 A ˚. varying from 3.6 A Triple mutant corrects 181 side-chain and restores the pocket In all its isoforms, PKB possesses lysine at the position corresponding to Q181 (Figure 1(a)). The observation of the Q181 rotation in PKAB2 suggested the introduction of the additional mutation Q181K to explore the characteristics of this unexpected feature. We therefore created the triple mutant V123A, L173M, Q181K (PKAB3) and determined its crystal structure in the absence of ligand in the ATP-binding site. The structure shows that K181, like Q181 in native PKA, is bound at the surface of the protein, and that the hinge-binding region of the ATP pocket has the well known PKA-like geometry. Comparison of surface polar networks A comparison of the interactions of lysine and glutamine in their protein surface-bound conformations suggests that glutamine is, in general, more susceptible to this potential rearrangement. In both, ATP-bound PKAB2 and PKA, only a single H-bond, to the backbone oxygen atom of I180, anchors Q181 into its external position. A second ˚ ) with R308:NH2 might provide contact (at 3.4 A an additional tether, but only if Q181 is partially disordered such that the amide oxygen atom may be partially oriented toward R308 (Figure 7(a)). (R308 is fixed in a salt-bridge interaction with D175:OD2 and with a hydrogen bond to I180:CO, so the proximity to Q181 may result from these other contacts and even have repulsive character.) In contrast to the weakly bound Q181 in PKA, the newly introduced lysine residue in the PKAB3
PKA-PKB Hybrid Mutants
1029
Figure 7. (a) Stereo view of Q181 distance measurements of the liganded PKAB2. The sole strong H-bond of Q181 (with I180:CO) and the contact with R308:NH2 are highlighted in yellow. All other potential hydrogen bond partners are too distant to form H-bonds. (b) Stereo view of polar interactions involving and surrounding K181 in unliganded PKAB3. The hydrogen bonding network is highlighted in yellow. Important interactions include tight binding to the backbone oxygen atom of P101 and indirect interactions to the backbone oxygen atom of L103 and the side-chain oxygen atom of E121:OE1 via a strongly fixed water molecule (X60). Weaker indirect contacts via a water molecule (X61) to the backbone oxygen atom of I180 and side-chain oxygen atom D180:OE1 are visible.
structure is anchored tightly between two complementary charges via an extended H-bond network involving water molecules and several residues ˚ direct hydrogen (Figure 7(b)). K181 shares a 2.9 A bond interaction with the backbone oxygen atom of P101. A second strong H-bond to a tightly bound water molecule links lysine further to the backbone oxygen atom of L103 and the complementary charge of E121:OE1. Furthermore, there are weaker indirect contacts via water X61 to the
backbone oxygen atom of I180 and side-chain oxygen atom D175:OD2. Significantly, the two charged residues E121 and D175 that are hydrogen bonded indirectly with Lys181 are absolutely conserved among all isoforms of PKA, PKB and PKC (Figure 1(a), highlighted in green for PKB isoforms and PKAa). Additionally, the hydrophobic residues whose main-chain carbonyl groups form direct or indirect hydrogen bonds with K181 are absolutely or highly conserved among the AGC
1030
PKA-PKB Hybrid Mutants
group kinases. The entire side-chain (CG onwards) of R308, which forms a salt-bridge with Asp175 and is near Q181 in PKA, is disordered in the triple mutant PKAB3. It is not clear whether the mutation Q181K leads to disordering of R308 or whether this is an effect of the new crystal packing of PKAB3. The side-chain of K181 is partially fixed by hydrophobic contacts as it emerges to the protein surface, so it may compete successfully with R308 for ordering interactions with D175 as a salt-bridge partner. However, E121 is available also as a complementary charge and the introduction of lysine at position 181 neutralizes the otherwise apparent negative surface charge. The crystal structure of PKB13 shows that the lysine residue corresponding to position Q181 is indeed fixed by a salt-bridge with the equivalent of E121 of PKA and by a hydrogen bond to the equivalent of P101:CO. The lysine equivalent of R308 in PKB is ordered and present as a counterion but at a distance too large to be considered a saltbridge.
inhibitors H7, H9 and HA1077 show roughly the same binding energy loss when comparing PKAB2 to PKAB3; this represents in the first approximation the energy required to expel Q181 from the hinge region. Similarly, they show roughly the same binding energy loss when comparing PKAB3 to native PKA (Table 1). Again in a first approximation, this energy loss should reflect differences in binding energies that arise from the two mutations V123A and L171M and their effects on either inhibitor or solvent interactions. The data may indicate an anomalously weakened binding for H7 to PKAB3 and thus reflect PKA selectivity over PKB. Further, a relatively lowered sensitivity of HA1077 binding strengths to both PKAB2 and PKAB3 mutations may appear anomalous. Both apparent anomalies are, however, on the margins of statistical significance. A precise analysis of these effects will require experiments with a variety of inhibitors with a varied PKA and PKB selectivity profile.
Biacore binding results
Summary
The surface plasmon resonance spectroscopy binding assays confirms the weakening of ligand binding to PKAB2. This is evidently due to the energy required for the ligand to induce the movement of Q181 away from the binding pocket into the position seen in the PKAB2 – ATP complex structure (Table 1). The ligand-induced transition might be visualized in detail as follows. During ATP binding, a rotation of T183 takes place to create the interaction of T183:OG1 and ATP:N7. This rotation disrupts the packing of Q181, V104, and T183, so that the surface structure of Q181 becomes the energetically preferred geometry. The energy costs associated with this rearrangement leads to an increase of the Km value of ATP by a factor of 10. Most of the apparent energy cost is eliminated with the restoration of the adenine pocket by the introduction of the Q181K mutation in PKAB3. A similar effect of lowered binding affinities can be observed by binding of isoquinoline inhibitors to PKAB2 in comparison to PKA and PKAB3. Although the isoquinolines do not possess an equivalent to N7 of ATP, an equivalent energy cost is associated with their binding. Thus, to within experimental error, ATP and all three
The utility of a surrogate protein for structurebased selectivity studies lies in the ability to observe individual protein –ligand interactions and draw conclusions relevant to the actual target. Mutations of the surrogate protein in order to better approximate the target clearly facilitates this. Ideally then, the geometries of ligand – sidechain interactions for both the target and a (mutated) surrogate should be identical to within the accuracy relevant to modelling. For most practical modelling tasks, a structure that shows the set of polar interactions and extent of buried surface areas is sufficient for planning new chemical syntheses. When more subtle effects become important, such as when ligand-induced conformation changes occur, or when the need for selectivity against very similar targets is important, structural details including, for example, the quality of hydrogen bonding, become important and the relevance of the surrogate protein must be correspondingly high. Further, it would be desirable that the surrogate reflect the kinetic properties of the target as well. This would enhance confidence that the observed structures are relevant in all details, for example, or may offer advantages in screening. For PKB, PKA is already a good surrogate for many purposes. This study shows that the double mutant PKA (V123A, L173M) could represent a better PKB model than PKA, but only for ligands whose binding properties depend significantly on the two substitutions, and then only if the ligands share identical costs of ejecting Q181 from the hinge-binding site (as the isoquinoline inhibitors H7, H9 and HA1077 appear to do). With the triple mutant V123A, L173M, Q181K, this defect is removed. The crystal structure determination, however, was necessary to diagnose the cause of
Table 1. Ki/Km values determined from surface plasmon resonance spectroscopy Inhibitor/substrate (mM) H7 H9 HA1077 ATP
PKA
PKAB2
PKAB3
6.8 ^ 2.2 3.3 ^ 2.2 5.7 ^ 1.2 9.6 ^ 2.1
94.8 ^ 0.7 44.0 ^ 8.7 34.9 ^ 1.5 115.4 ^ 18.6
36.6 ^ 0.9 7.7 ^ 1.7 10.0 ^ 3.3 24.1 ^ 3.7
H7, 1-(5-isoquinolinesulfonyl)-2-methylpiperazine. H9, N-(2aminoethyl)-5-isoquinolinesulfonamide. HA1077 (Fasudil), (5isoquinolinesulfonyl)homopiperazine.
1031
PKA-PKB Hybrid Mutants
Figure 8. Overlay of the ATP-binding sites of PKB and PKAB3 (stereo view). The conformers of the mutated residues of PKAB3 are very similar to PKB, as is the ATP-binding site as a whole. The greatest differences occur at the b-turn regions of the respective glycine loops (here depicted as ribbons and apparently disordered in PKB) and the conformer of flexible residue 104.
the weakened ligand binding; without the structure it would have been natural to assume that weakened interactions with either A123 or M173 would have been the cause. The procedure here has been analogous to the evolutionary occurrence of a second site-suppressor mutation (Q181K) to correct a defect caused by an earlier mutation (V123A). A superposition of the structures of PKAB3 and PKB highlights the similarity of conformers of A123, M173 and K181 (Figure 8). Furthermore, the equivalence of the ATP-binding site as a whole is evident. Differences occur at the b turn region of the glycine loop, which is apparently partially disordered in the PKB structure, and at residue 104, which is known to be flexible among PKA structures.30 In general, therefore, the PKB structure validates the surrogate approach; PKAB3 provides a good model for the adenine-binding site and further provides a tool useful to analyze the influence of amino acid residue variations on ligand-binding properties. While the accrual of 3D structures, point mutations, and activity data for protein kinases attests to their complexity and apparent unpredictability, this same catalog of data increasingly allows systematization of structure–function relationships. The relative N and C-terminal domain orientations, phosphorylation-induced refolding of the activation loop, and co-factordependent restructuring of structural elements into active configurations are examples of recurring phenomena. Similarly, the roles of individual residues can be described for catalytic, selectivity and other properties. The propensity of Q181 to switch between binding at the ATP pocket or at the outer surface of the protein seems to be an artifact of mutation, but may find a counterpart in a natural kinase, in particular because this is a possible mechanism to communicate the occupation state of the ATP pocket to a potential co-factor or conversely to alter ATP site ligandbinding properties via co-factor binding.
Materials and Methods Mutagenesis of PKA Site-directed mutagenesis was performed using pT7PKA as template and the Stratagene (La Jolla) Quick Change Kit following the manual of the supplier including the design of the corresponding primer pairs. All constructs using for protein expression were verified by DNA sequencing.
Expression and purification of PKA, PKAB2 and PKAB3 Expression vectors pT7-7 and pET28b, both carrying a bacteriophage T7 promoter, were used to express bovine PKA Ca31 and PKAB2 and PKAB3 in Escherichia coli strain BL21(DE3), respectively. Two positions distinguish bovine (N32, M63) from human PKA (S32, K63), both at sites distant from substrate and ATP-binding sites. Cells were grown in LB medium supplemented with either 100 mg/ml of carbenicillin or 75 mg/ml of kanamycin, depending on the appropriate vector. After reaching an A610 of 0.7, cells were induced with 0.4 mM isopropyl-bD -galactopyranoside (IPTG) for 16 hours at 24 8C. Cells were collected by centrifugation, resuspended in 30 mM Mes, 50 mM KCl, 1 mM EDTA, 1 mM dithiothreitol and sonicated. For affinity purification,32,33 the supernatant of a one hour, 30,000g centrifugation was supplemented with 3 mM MgCl2, 2 mM ATP and applied to immobilized protein kinase inhibitor peptide PKI-(5-24) in batch for one hour at room temperature. The gravitypacked column was washed with five volumes each of TMN 50 and TMN250 (20 mM Tris – HCl (pH 7.4), 2 mM MgCl2, 50/250 mM NaCl 0,4 mM ATP). About 5 mg of heterogenously phosphorylated PKA or the abovementioned derivatives per liter of cell culture was eluted with a buffer solution of 200 mM arginine, 50 mM NaCl, 1 mM EDTA, 20 mM Tris – HCl (pH 7.4). The enzymes were concentrated via Centripreps 30 (Millipore) and diluted fivefold with 25 mM BisTris – propane (pH 8.5). Homogenously phosphorylated samples were obtained by applying these samples to a Mono-S 10/10 column and eluting with a LiCl gradient up to 300 mM in 25 mM BisTris – propane at pH 8.5. The identity of the
1032
PKA-PKB Hybrid Mutants
Table 2. X-ray data and refinement statistics
Space group Cell constants ˚) a (A ˚) b (A ˚) c (A ˚) Data resolution (A Rsym (Rsym last shell) Completeness 2s (last shell) Rcryst-work, Rcryst-free Weighted Rwork, Rfree RMS deviation ˚) Bonds (A Angles (deg.)
samples was spectrometry.
confirmed
by
PKAB2
PKAB2-ATP
PKAB3
P212121
P212121
P212121
73.5 76.7 80.5 2.6 0.12 (0.34) 0.96 (0.91) 0.204, 0.256 0.186, 0.235
73.3 75.2 80.3 2.6 0.10 (0.39) 0.86 (0.72) 0.197, 0.241 0.161, 0.191
59.9 79.4 100.9 2.2 0.14 (0.34) 0.94 (0.94) 0.218, 0.259 0.200, 0.236
0.012 1.3
0.010 1.3
0.010 1.2
electrospray
mass
Biacore sensor chip preparation Proteins used for Biacore analysis were dialyzed twice with the 400-fold volume of 50 mM Mops (pH 6.8), 10 mM MgCl2, 50 mM KCl. Coupling of PKA, PKAB2 and PKAB3 to CM5 Biasensor chip via amine group linkage was achieved using standard coupling procedures.34 Briefly, CM5 sensor chips were activated by injecting 35 ml of a 1:1 (v/v) mixture of N-ethyl-N0 -[dimethylamino]carbodiimide/N-hydroxysuccinimide at 5 ml/ minute. After diluting the proteins in 10 mM sodium acetate (pH 5.5), PKA, PKAB2, PKAB3 and HSA as a control were coupled to the CM5 sensor chip by injecting a 50 mM solution of the protein selected with a flow-rate of 5 ml/minute until 11,000 RU was reached. Generation of kinetic binding data Kinetic studies with a range of analyte concentrations were determined at a flow-rate of 10 ml/minute by allowing 300 seconds for association and 900 seconds for dissociation. Analyte concentrations used were of (256), 128, 64, 32, 16, 8, 4, 2, (1) mM concentrations (depending on the analyte), diluted in MilliQ water or running buffer (50 mM Mops (pH 7.4), 50 mM KCl, 10 mM MgCl2). Kinetic data were analyzed with BIA evaluation 3.0 software. For each binding curve, the response obtained using the HSA cell as control was subtracted. Due to the small signals (up to 40 RU) the steady-state affinity model was used to determine the Km/Ki of the different small molecular mass compounds. Goodness of fit measured as a x2 value was less than 4 for binding of the low molecular mass compounds. All binding experiments were repeated either two or, more usually, three times, and biosensor chips coupled at different times yielded surfaces with identical binding affinities. The binding affinities of ATP, H735 H9,35 HA1077,36 all three purchased from Calbiochem, to PKA were similar to the Km/Ki values reported in different studies using enzymatic assays.35 – 38,15 X-ray crystallography Hanging droplets containing 18 mg/ml of protein, 25 mM Mes-BisTris, 75 mM LiCl, 0.1 mM MgCl2, 1 mM dithiothreitol,15 1.5 mM octanoyl-N-methylglucamide,39 1 mM PKI-(5-24), pH 6.3 – 6.6, 1.5 mM ATP and 3 mM
MgCl2 in the corresponding trials, were equilibrated at 5 8C against 12 – 18% (v/v) methanol. In each case, one crystal was sufficient to obtain a complete data set. Diffraction data were measured at 4 8C in a sealed capillary using either on a Bruker X1000 wire array detector (PKAB2-ATP) or a MAR image plate (PKAB2, PKAB3) on a copper target rotating anode X-ray generator (Rigaku RU200) with a graphite monochromator. Data evaluation was performed with SAINT (PKAB2-ATP) or MOSFLM and SCALA (PKAB2, PKAB3). Structure solution and refinement was done with REFMAC5 and other programs of the CCP4 crystallography package. (Data and model quality parameters are given in Table 2.)
Acknowledgements We are grateful to Dr Wolf Lehman and Dr Matthias Wind for expert mass spectrometry analysis and we thank Professor Volker Kinzel for constant support. This work was supported, in part, by the Bayerische Wirtschaftsministerium.
References 1. Cohen, P. (2002). Protein kinases—the major drug targets of the twenty-first century? Nature Rev. Drug Discov. 1, 309–315. 2. Fox, T., Coll, J. T., Xie, X., Ford, P. J., Germann, U. A., Porter, M. D. et al. (1998). A single amino acid substitution makes ERK2 susceptible to pyridinyl imidazole inhibitors of p38 MAP kinase. Protein Sci. 7, 2249– 2255. 3. Liu, Y., Bishop, A., Witucki, L., Kraybill, B., Shimizu, E., Tsien, J. et al. (1999). Structural basis for selective inhibition of Src family kinases by PP1. Chem. Biol. 6, 671– 678. 4. Shabb, J. B. (2001). Physiological substrates of cAMPdependent protein kinase. Chem. Rev. 10, 12381– 12411. 5. Montminy, M. (1997). Transcriptional regulation by cyclic AMP. Annu. Rev. Biochem. 66, 807– 822. 6. Bossemeyer, D. (1994). The glycine-rich sequence of protein kinases: a multifunctional element. Trends Biochem. Sci. 19, 201– 205. 7. Walsh, D. A. & Van Patten, S. M. (1994). Multiple
PKA-PKB Hybrid Mutants
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
pathway signal transduction by the cAMP-dependent protein kinase. FASEB J. 8, 1227– 1236. Taylor, S. S., Buecher, J. A. & Yonemoto, W. (1990). cAMP dependent protein kinase: framework for a diverse family of regulatory enzymes. Annu. Rev. Biochem. 59, 84 – 86. Brognard, J., Clark, A. S., Ni, Y. & Dennis, P. A. (2001). Akt/protein kinase B is constitutively active in non-small cell lung cancer cells and promotes cellular survival and resistance to chemotherapy and radiation. Cancer Res. 61, 3986– 3997. Kulik, G., Carson, J. P., Vomastek, T., Overman, K., Gooch, B. D., Srinivasula, S. et al. (2001). Tumor necrosis factor alpha induces BID cleavage and bypasses antiapoptotic signals in prostate cancer LNCaP cells. Cancer Res. 61, 2713– 2719. Ng, S. S. W., Tsao, M. S., Chow, S. & Hedley, D. W. (2000). Inhibition of phosphatidylinositide 3-kinase enhances gemcitabine-induced apoptosis in human pancreatic cancer cells. Cancer Res. 60, 5451– 5455. Itoh, N., Semba, S., Ito, M., Takeda, H., Kawata, S. & Yamakawa, M. (2002). Phosphorylation of Akt/PKB is required for suppression of cancer cell apoptosis and tumor progression in human colorectal carcinoma. Cancer, 94, 3127 –3134. Yang, J., Cron, P., Thompson, V., Good, V. M., Thompson, V., Hemmings, B. A. & Barford, D. (2002). Crystal structure of an activated Akt/Protein Kinase B ternary complex with GSK3-peptide and AMP-PNP. Nature Struct. Biol. 9, 940– 944. Kumar, C. C., Diao, R., Yin, Z., Liu, Y., Samatar, A. A., Madison, V. & Xiao, L. (2001). Expression, purification, characterization and homology modeling of active Akt/PKB, a key enzyme involved in cell survival signaling. Biochim. Biophys. Acta, 1526, 257–268. Engh, R. A., Girod, A., Kinzel, V., Huber, R. & Bossemeyer, D. (1996). Crystal structures of catalytic subunit of cAMP-dependent protein kinase in complex with isoquinolinesulfonyl protein kinase inhibitors H7, H8, and H89. Structural implications for selectivity. J. Biol. Chem. 271, 26157– 26164. Bossemeyer, D., Engh, R. A., Kinzel, V., Ponstingl, H. & Huber, R. (1993). Phosphotransferase and substrate binding mechanism of the cAMP-dependent protein kinase catalytic subunit from porcine heart as deduced from the 2.0 A structure of the complex with Mn2þ adenylyl imidodiphosphate and inhibitor peptide PKI(5-24). EMBO J. 12, 849– 859. Madhusudan, A. P., Xuong, N. H. & Taylor, S. S. (2002). Crystal structure of a transition state mimic of the catalytic subunit of cAMP-dependent protein kinase. Nature Struct. Biol. 9, 273– 277. Prade, L., Engh, R. A., Girod, A., Kinzel, V., Huber, R. & Bossemeyer, D. (1997). Staurosporine-induced conformational changes of cAMP-dependent protein kinase catalytic subunit explain inhibitory potential. Structure, 5, 1627– 1637. Narayana, N., Diller, T. C., Koide, K., Bunnage, M. E., Nicolaou, K. C., Brunton, L. L. et al. (1999). Crystal structure of the potent natural product inhibitor balanol in complex with the catalytic subunit of cAMP-dependent protein kinase. Biochemistry, 38, 2367–2376. Lew, J., Coruh, N., Tsigelny, I., Garrod, S. & Taylor, S. S. (1997). Synergistic binding of nucleotides and inhibitors to cAMP-dependent protein kinase examined by acrylodan fluorescence spectroscopy. J. Biol. Chem. 272, 1507– 1513.
1033
21. Seifert, M. H., Breitenlechner, C. B., Bossemeyer, D., Huber, R., Holak, T. A. & Engh, R. A. (2002). Phosphorylation and flexibility of cyclic-AMP-dependent protein kinase (PKA) using (31)P NMR spectroscopy. Biochemistry, 41, 5968–5977. 22. Karlsson, R., Madhusudan, Taylor, S. S. & Sowadski, J. M. (1994). Intermolecular contacts in various crystal forms related to the open and closed conformational states of the catalytic subunit of campdependent protein kinase. Acta Crystallog. sect. D, 50, 657 –662. 23. Engh, R. A. & Bossemeyer, D. (2002). Structural aspects of protein kinase control—role of conformational flexibility. Pharmacol. Ther. 93, 99 – 111. 24. Zheng, J. H., Trafny, E. A., Knighton, D. R., Xuong, N. H., Taylor, S. S., Teneyck, L. F. & Sowadski, J. M. (1993). 2.2-Angstrom refined crystal structure of the catalytic subunit of cAMP-dependent protein-kinase complexed with MnATP and peptide inhibitor. Acta Crystallog. sect. D, 49, 362– 365. 25. Shaffer, J. & Adams, J. A. (1999). An ATP-linked structural change in protein kinase A precedes phosphoryl transfer under physiological magnesium concentrations. Biochemistry, 38, 5572– 5581. 26. 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, 2154– 2161. 27. Pal, D. & Chakrabarti, P. (2001). Non-hydrogen bond interactions involving the methionine sulfur atom. J. Biomol. Struct. Dynam. 19, 115 – 128. 28. Meyer, E. A., Castellano, R. K. & Diederich, F. (2003). Interactions with Aromatic Rings in Chemical and Biological Recognition. Angew. Chem. Int. Ed. 42, 1210 –1250. 29. Rosenfield, R. E., Jr, Parthasarathy, R. & Dunitz, J. D. (1977). Directional preferences of nonbonded atomic contacts with divalent sulfur. 1. Electrophiles and nucleophiles. J. Am. Chem. Soc. 99, 4860– 4862. 30. Engh, R. A. & Bossemeyer, D. (2002). Structural aspects of protein kinase control—role of conformational flexibility. Pharmacol. Ther. 93, 99 – 111. 31. Wiemann, S., Kinzel, V. & Pyerin, W. (1992). Cloning of the C alpha catalytic subunit of the bovine cAMP-dependent protein kinase. Biochim. Biophys. Acta, 1171, 93 – 96. 32. Ohlsen, S. R. & Uhler, M. D. (1989). Affinity purification of C alpha and C beta isoforms of the catalytic subunit of cAMP-dependend protein kinase. J. Biol. Chem. 264, 18662– 18666. 33. Girod, A., Kinzel, V. & Bossemeyer, D. (1995). In vivo activation of recombinant cAPK catalytic subunit active site mutants by coexpression of the wild-type enzyme, evidence for intermolecular cotranslational phosphorylation. FEBS Letters, 391, 121– 125. 34. Lo¨fa˚s, S. & Johnsson, B. (1990). A novel hydrogel matrix on gold surfaces in surface plasmon resonance sensors for fast and efficient covalent immobilization of ligands. J. Chem. Soc. Chem. Commun. 21, 1526 –1528. 35. Hidaka, H., Inagaki, M., Kawamoto, S. & Sasaki, Y. (1984). Isoquinolinesulfonamides, novel and potent inhibitors of cAMP dependent protein kinase and protein kinase C. Biochemistry, 23, 5036– 5041. 36. Asano, T., Suzuki, T., Tsuchiya, M., Satoh, S., Ikegaki, I., Shibuya, M. et al. (1989). Vasodilator actions of HA1077 in vitro and in vivo putatively mediated by
1034
PKA-PKB Hybrid Mutants
the inhibition of protein kinase. Br. J. Pharmacol. 98, 1091– 1100. 37. Whitehouse, S., Feramisco, J. R., Casnellie, J. E., Krebs, E. G. & Walsh, D. A. (1983). Studies on the kinetic mechanism of the catalytic subunit of c-amp dependent protein kinase. J. Biol. Chem. 258, 3693– 3701. 38. Whitehouse, S. & Walsh, D. A. (1983). MgATPdependent interactions of the inhibitor protein of
the cAMP dependent protein kinase with the catalytic subunit. J. Biol. Chem. 258, 3682– 3692. 39. Knighton, D. R., Zheng, J. H., Ten Eyck, L. F., Ashford, V. A., Xuong, N. H., Taylor, S. S. & Sowadski, J. M. (1991). Crystal structure of the catalytic subunit of cyclic adenosine monophosphate-dependent protein kinase. Science, 253, 407– 414.
Edited by K. Nagai (Received 12 December 2002; received in revised form 3 April 2003; accepted 15 April 2003)
J. Mol. Biol. (2003) 329, 1021–1034
Mutants of Protein Kinase A that Mimic the ATP-binding Site of Protein Kinase B (AKT) Michael Gaßel1, Christine B. Breitenlechner2, Petra Ru¨ger3 Ute Jucknischke3, Thorsten Schneider1, Robert Huber2 Dirk Bossemeyer1* and Richard A. Engh2,3* 1
German Cancer Research Center (DKFZ), Division of Pathochemistry, Im Neuenheimer Feld 280, 69120 Heidelberg, Germany 2
Max Planck Institut fu¨r Biochemie, Abteilung Strukturforschung, Am Klopferspitz 18a, 82152 Martinsried, Germany
3 Roche Diagnostics GmbH Pharmaceutical Research Nonnenwald 2, 82377 Penzberg Germany
The mutation of well behaved enzymes in order to simulate less manageable cognates is the obvious approach to study specific features of the recalcitrant target. Accordingly, the prototypical protein kinase PKA serves as a model for many kinases, including the closely related PKB, an AGC family protein kinase now implicated as oncogenic in several cancers. Two residues that differ between the a isoforms of PKA and PKB at the adenine-binding site generate differing shapes of the binding surface and are likely to play a role in ligand selectivity. As the corresponding mutations in PKA, V123A would enlarge the adenine pocket, while L173M would alter both the shape and its electronic character of the adenine-binding surface. We have determined the structures of the corresponding double mutant (PKAB2: PKAa V123A, L173M) in apo and MgATP-bound states, and observed structural alterations of a residue not previously involved in ATP-binding interactions: the side-chain of Q181, which in native PKA points away from the ATP-binding site, adopts in apo double mutant protein a new rotamer conformation, which places the polar groups at the hinge region in the ATP pocket. MgATP binding forces Q181 back to the position seen in native PKA. The crystal structure shows that ATP binding geometry is identical with that in native PKA but in this case was determined under conditions with only a single Mg ion ligand. Surface plasmon resonance spectroscopy studies show that significant energy is required for this ligand-induced transition. An additional PKA/PKB mutation, Q181K, corrects the defect, as shown both by the crystal structure of triple mutant PKAB3 (PKAa V123A, L173M, Q181K) and by surface plasmon resonance spectroscopy binding studies with ATP and three isoquinoline inhibitors. Thus, the triple mutant serves well as an easily crystallizable model for PKB inhibitor interactions. Further, the phenomenon of Q181 shows how crystallographic analysis should accompany mutant studies to monitor possible spurious structural effects. q 2003 Elsevier Science Ltd. All rights reserved
*Corresponding authors
Keywords: PKA; PKB; protein kinase; crystal structure; point mutation
Introduction Abbreviations used: PKA, cAMP-dependent protein kinase; PKB, protein kinase B (also called Akt); PKAB2, double mutant PKA V123A/L173M; PKAB3, triple mutant PKA V123A/L173M/Q181K; PKI, protein kinase inhibitor; H7, 1-(5-isoquinolinesulfonyl)-2-methylpiperazine; H9, N-(2-aminoethyl)-5-isoquinolinesulfonamide; HA1077, (Fasudil), (5-isoquinolinesulfonyl)homopiperazine. E-mail address of the corresponding authors: [email protected]; [email protected]
The protein kinases represent one of the largest families of human proteins, with more than 500 protein kinases identified by the human genome project. Considering variations arising from population heterogeneity, differential splicing, and enhanced mutation rates caused e.g. by cancer, many thousands of distinct sequences will occur in the human population alone. Despite the variations in sequence, the fold of the active protein kinase catalytic domain is well conserved; inactive
0022-2836/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved
1022
PKA-PKB Hybrid Mutants
Figure 1 (legend opposite)
protein kinase structures differ but cluster into protein kinase subfamilies. As a consequence of the conservation of the active structure, many properties can be analysed with respect to relatively few sequence positions which define the property, for example, selectivity of binding at the ATP-binding pocket. The large number of kinases fuels concern that
new drugs designed against certain protein kinase targets will fail in clinical tests because of unexpected activity against other protein kinases. Compounds STI571 (Glivec) and HA1077 (Fasudil) have demonstrated that protein kinase inhibition is a commercially viable therapeutic strategy, but represent the only protein kinase inhibitors on the market.1 Thus, kinase inhibitor selectivity will
PKA-PKB Hybrid Mutants
1023
Figure 1. (a) Alignment of human PKAa and PKB isoforms. Amino acid residues highlighted with blue background show positions that are not identical between PKAa and PKBa. Amino acid residues marked in yellow are exchanged in this study. Amino acid residues highlighted in green (P101, L103, E121, D175, I180) form the conserved surface hydrogen bonding network that binds Q181 in PKA. The charge of R308 (orange) in this network is conserved but not the residue identity (K in PKB). T183 (red) is subject to ligand induced rotation. See the text for details. (b) Stereo view of the solvent-accessible surface of PKAa colored according to identity with PKBa. PKA surface elements that differ from PKB are colored blue, while atoms from identical residues are colored gray. Cyan ball-and-stick atoms show the positions of substrate-mimic inhibitor PKI(5-24); multicolored ball-and-stick atoms depict AMP-PNP. Differences are distributed over the surface rather uniformly, but are limited to a few residues in the ATP-binding site. (c) Close-up of the solvent-accessible surface in the ATP pocket. Surfaces from two non-conserved side-chains closest to AMP-PNP are colored violet. These positions, V123 and L173, were first mutated to A and M, respectively, to mimic the PKB-binding pocket. The blue surface arises from V104, which is the isosteric threonine residue in PKB isoforms. ATP hydrogen bonding interactions are shown in green.
remain a crucial issue for the foreseeable future, intensifying the critical role that mutagenesis and protein crystallography possess in identifying selectivity mechanisms. The kinases share a common bi-lobal catalytic domain structure with the ATP-binding site in the interlobe cleft. ATP is bound to the kinase via interactions with some 15 residues of the protein, including about ten side-chain interactions that therefore represent especially important potential selectivity determinants. Thus, the essential binding properties of ATP and other ATP-site ligands should, in many cases, be simulated for a protein kinase by a limited set of point mutations of a closely related protein kinase. The construction of such hybrids has been demonstrated for kinases.2 Even when subtle differences lead to quite different overall reaction kinetics, for practical ligand design purposes most or all protein – ligand interactions are likely to be identical for both the target kinase and its model. That a single residue can be identified as the principal selectivity determinant for an inhibitor type by mutation of a series of kinases verifies this approach.3 Along these lines, we have chosen the cAMPdependent protein kinase (PKA) as surrogate kinase for another closely related member of the
AGC kinase family, protein kinase B or Akt. PKA responds via cAMP to a great variety of stimulatory agents in the regulation of ion conduction, memory, gene transcription, muscle contraction, and metabolism.4,5 PKA has historically enjoyed a central role as prototypical protein kinase for functional studies,6 – 8 in particular because recombinant expression, purification, and crystallization is well established. PKB is involved in insulin signaling pathways and plays a central role in apoptosis; disregulation is associated with several cancers including non-small cell lung carcinoma,9 prostate carcinoma,10 pancreatic carcinoma11 and colorectal carcinoma.12 Thus, as an important and emerging drug target, with a well characterized surrogate (PKA), PKB is an ideal candidate as a target kinase in our approach. Further, the very similar structures of the active forms of PKA and PKB13 support the expectation that side-chain differences on highly similar backbone scaffolds will be responsible for much of their relative ATP-site selectivities. Since inhibitor design requires facile cocrystallization, which has not been demonstrated for PKB, we remain interested in PKA-based models for PKB binding, and enjoy, additionally, the validation offered
1024
by the PKB structure. As a first step toward generation of a PKA-based model for the PKB ATP-binding site, we produced the double mutant PKA (V123A, L173M), or PKAB2. These two residues in the ATP-binding site represent the differences between isoforms PKAa and PKBa that form contacts with ATP in PKA structures and that must alter the shape of the binding pocket. Both are at the adenine-binding site. One of these, V123A, enlarges the pocket, while the other, L173M, possesses a largely unchanged side-chain volume but changes the shape of the adenine-binding surface and introduces a sulfur atom for potential adenine interactions. These mutants were characterized by means of X-ray crystallography and surface plasmon resonance spectroscopy binding assays. The ligand-free structure of the double mutant (PKAB2) shows an unanticipated structural change. The side-chain of Q181, which is normally exposed to solvent and points away from the ATP-binding site, adopts instead a new x1 rotamer conformation, which places the side-chain at the adenine-binding site. In this orientation, ATP binding is partially blocked, and both the shape and polar character of the binding pocket are altered. The structure of PKAB2 complexed with ATP, however, shows that ligand binding restores Q181 into its position known from PKA. Presumably, ligand binding in the ATP site is weakened by the energy required for the ligand-induced effect. In PKB, the residue corresponding to Q181 of PKA is lysine. We therefore constructed the triple mutant V123A, L173M, Q181K (PKAB3). This triple mutant in the unliganded form shows K181 “restored” to a position distant from the catalytic site and embedded in a surface hydrogen-bond/salt-bridge network. The mutated residues in the adenine-binding site of PKAB3 adopt conformers identical with the corresponding residues of PKB. Binding affinity studies with ATP and several isoquinoline inhibitors using surface plasmon spectroscopy support the conclusions that the double mutation introduces a defect into the kinase that weakens ligand binding, and that the triple mutant restores native-like behavior. We describe these results and discuss their relevance for PKB inhibition and inhibitor design.
Results and Discussion Sequence similarity of PKA and PKB An alignment of the primary sequences (Figure 1(a)) between PKAa and PKB isoforms illustrates the affiliation of these kinases, with amino acid identities between e.g. PKAa and PKBa of 45% over 294 amino acid residues. This high degree of similarity gives confidence in the quality of homology models of the kinase domain
PKA-PKB Hybrid Mutants
of PKB based on the structure of PKA.14 Further, the recently published structures of activated forms of PKBb13 verify the validity of homology modeling for the active form. The unconserved amino acid side-chains are rather dispersed and occur especially at the protein surface. Their occurrence at the peptide substrate-binding sites plays a direct role in peptide substrate selectivity (Figure 1(a) and (b)). The ATP-binding site is highly conserved, with only three amino acid substitutions in the adenine pocket (V123A, L173M and V104T; Figure 1(c)). Of these three, substitutions V123A and L173M alter the shape of the pocket, while V104T would possibly introduce a hydrogen bonding partner, although its orientation in the PKB structure shows only hydrophobic interactions with ATP. L173M potentially alters the electronic character of the adenine-binding site, especially with the addition of a sulfur atom. To test PKA mutants as improved models for PKB we first made the double mutant V123A, L173M (PKAB2) and solved crystal structures of the ATPfree and ATP-bound structures. After observing that the conformation of Q181 is modified by these mutations (see below), we made the triple mutant V123A, L173M, Q181K and solved its crystal structure. Overall structures The published structures of mammalian PKAs include sequences from mouse (Mus musculus), pig (Sus scrofa), and cow (Bos taurus), which differ from the human (Homo sapiens) sequence at seven, four, and two positions, respectively. These structures have clarified many PKA properties, including mechanisms of phosphotransfer, binding interactions with the peptide inhibitor PKI-(5-24), substrates and pseudosubstrates based thereon, with Mn2AMP-PNP, AlFADP, and with a variety of natural and synthetic inhibitors.15 – 19 In addition, dynamic properties have been studied by fluorescence depolarization,20 and 31P NMR.21 The crystal structures of PKAB2 (PKA V123A, L173M) and PKAB3 (PKA V123A, L173M, Q181K) as binary complexes with PKI, and of PKAB2 as the MgATP þ PKI ternary complex possess the general structural features of PKA known from crystal structures. The individual N and C-terminal lobes are largely superimposable, and the structures are in a closed conformation. This is consistent with crystallization in the P212121 space group with cell dimensions a ¼ 73:5 A; b ¼ 76:7 A; 22 However, the PKAB3 structure c ¼ 80:5 A: crystallized in a new crystal packing arrange ment (P212121 with cell dimensions a ¼ 59:9 A; and c ¼ 100:9 A), possibly as a result b ¼ 79:4 A; of the change in charge distribution at the surface of the protein behind the hinge region. It retains the closed conformation in this packing arrangement.
PKA-PKB Hybrid Mutants
1025
Figure 2. Sim-weighted electron density (2mFo 2 dFc, blue at 1s) and difference electron density (mFo 2 dFc, white at 2s and red at 2 2s) maps showing the rotation of Q181. The maps were calculated from a model of PKA (orange sticks) after refinement but prior to rotation of the side-chain. Green sticks depict the refined structure of the double mutant PKAB2. The density shows unambiguously how Q181 rotates to occupy the cavity near the V123A position and near the adenine-binding interaction sites.
The Q181 switch Q181 adopts an “internalized” rotamer conformation The unliganded PKAB2 structure shows a surprising new rotamer conformation of Q181, with the side-chain amide group not at the surface of the protein as in PKA, but moved toward and
partially occluding the ATP-binding site (Figure 2). This conformation is verified unambiguously by the electron densities calculated from the PKA molecular replacement model. The rotation is apparently enabled by the creation of the cavity associated with the amino acid exchange V123A. Superposition of the unliganded PKA and PKAB2 structures show that close contacts with V123 prohibit this
Figure 3. Comparison of the different positions of Q181 in unliganded PKAB2 (orange sticks), in the ATP– PKAB2 complex (multicolored by atom type with white carbon atoms), and in the AMP-PNP – PKA complex (green sticks, 1CDK16). The distance measurement between Q181 of unliganded PKAB2 and V123 of native PKA shows how the rotation of Q181 is prevented in the wild-type by V123. Furthermore, it can be seen that in PKAB2-ATP as well as in PKA-AMPPNP, Q181 occupies the same position.
1026
PKA-PKB Hybrid Mutants
Figure 4. (a) Stereo view of Q181 distance measurements of the unliganded PKAB2 and comparative measurements in an overlay of the PKAB2-ATP complex. Binding of ATP does not obviously conflict with Q181 in this geometry. However, ATP binding may require a rotation of T183 that disrupts the space-filling packing of Q181, V104 and T183. This leads to the exit of Q181 to form more energetically favorable polar interactions at the protein surface. (b) Stereo view of the rotation of T183 between the liganded and unliganded PKAB2 structures. The distances and geometries show how the rotation of T183 is essential to form the contact between T183:OG2 and ATP:N7.
conformation of Q181 in wild-type PKA (Figure 3). Other changes associated with the “internal” glutamine rotamer include a x1 rotation of T183 that places ˚ from Q181:OE2, and a displaceCG2 at about 3.4 A ment of Val104:CG1 (Figure 4(a)). Structural flexibility, particularly of these two residues, has been reported.18,23 In the PKAB2 structure, these residues form a triangle of close non-polar (or weakly polar ˚ ) between Q181:OE1, C–H–-O) interactions (,3.4 A V104:CG1 and T183:CG2 (Figure 4(a)). The binding of ATP restores Q181 Observation of this change intensified our efforts to cocrystallize the PKAB2 mutant with ATP. The successful structure determination showed that
MgATP binding forces the residue Q181 back in its original position (Figure 3). The structure of Q181 in PKAB2-MgATP is indistinguishable from its structure in PKA. Binding of MgATP abolishes the interactions between V104, T183, and Q181, because Q181 is restored to its wild-type position, ˚, and because V104:CG1 is displaced by some 1.6 A and because T183 undergoes a rotamer conversion (Figure 4(a)). In its position at the hinge, the glutamine residue is stabilized by packing contacts but does not block ATP binding completely. On the contrary, an energetically favorable contact ˚ ) might between Q181:OE2 and ATP:N6 (at 2.8 A seem possible with this Q181 rotamer (Figure 4(a); red dotted line). However, the rotamer of T183 in the PKAB2 crystal form does not enable the typical
PKA-PKB Hybrid Mutants
adenine anchoring link contact between ATP:N7 and T183:OG1 (Figure 4(b)). Taken together, the hinge position of Q181 would seemingly still allow binding of MgATP, but the rotamer conversion of T183 apparently to enable the T183:OG1 ATP:N7 contact removes T183:CG2 from the T183V104-Q181 triangle and disrupts this stable packing arrangement so that Q181 exits the pocket. An intermediate structure that has Q181 at the hinge position but T183 in an ATP-binding geometry might hypothetically exist under some conditions or with certain ligands, but we have not observed it. The propensity of Q181 to switch between binding at the adenine site near the kinase hinge, or at the outer surface of the protein, may not play a physiological role for PKA, since it is observed only in conjunction with the V123A mutation. However, among the hundreds of natural kinases, appropriate sequences may exist that allow a similar transition. What makes this possibility particularly intriguing is the fact that this property represents a possible mechanism to communicate the occupation state of the ATP pocket to a potential co-factor or conversely to alter ATP-site ligand-binding properties via cofactor binding. Should such a mechanism occur in a natural kinase, it would offer a means for selective inhibition as well. MgATP binding in PKAB2 The position of ATP in the double mutant PKAB2 is very similar to that observed in the
1027 wild-type enzyme (Figure 5; 1CDK;16 1ATP24). A major exception is that a single Mg ion is present in the complex, at the position coordinating the b and g phosphate groups of ATP. The crystallization conditions included concentrations of 3 mM Mg2þ and 1.5 mM ATP. Kinetic studies of PKA have shown that lowering the concentration of free Mg2þ from 10 mM to the more physiological concentration of 0.5 mM causes the occupancy of the second site to shift from 80% to 20%.25 Thus, although unusual for PKA – ATP complexes, the observation of a single bound metal ion is consistent with expected binding strengths: the apparent millimolar binding constant for the second metal-binding site determined from kinetic studies is roughly the concentration of metal ion in the crystallization solution. This millimolar binding constant may be weakened further by crystallization conditions and/or mutations. The Mg ion in PKAB2 –MgATP is in the position of the activating16 metal ion. Figure 5 shows an apparent lengthening of the coordination distance between Asp184 and Mg, but this is not within the resolution of the structures and may arise from the use of different refinement parameters. A ˚ crystal structure of PKA– Mg2ATP has been 2.7 A reported26 (coordinates not deposited) but the coordination geometry was not described in detail. Interestingly, Zheng et al.26 reported low occupancies of the second metal-binding site at concentrations of Mg2þ below a 2.5-fold excess over ATP.
Figure 5. Stereo view of an overlay of the MgATP– PKAB2 complex structure (multicolored by atom type) and the MnAMP-PNP– PKA complex structure (green sticks, PDB code 1CDK16). The 2Fo 2 Fc density (1s) belongs to PKAB2 liganded with ATP. The PKAB2 structure is the first Mg ATP – PKA structure and shows only a single bound metal ion.
1028
PKA-PKB Hybrid Mutants
Figure 6. Stereo view of the interactions of M173 SD in the ATP-liganded PKAB2 structure.
Sulfur interactions The L173M mutation introduces the possibility of structured sulfur –ligand interactions. Such interactions have been the subject of recent theoretical and statistical analyses,27,28 including observations of non-random distributions of e.g. methionine around tryptophan.† Earlier studies predicted that nucleophiles (e.g. oxygen, aromatic pi molecular orbital clouds) would tend to approach the sulfur ATOM along the posterior extensions of the CE –SD or CG –SD bonds, while electrophiles (e.g. metal ions) would approach SD from a direction roughly (within 408) perpendicular to the CG –SD – CE plane.29 In the PKAB2 –ATP complex, the sulfur atom of Met173 makes two corresponding interactions (Figure 6), one with the carbonyl oxygen atom of G126 and one with the aromat of ATP. The backbone carbonyl oxygen atom of G126 approaches the Met173 SD atom at a ˚ and with a nearly 908 formed by distance of 3.78 A C – O – SD. The second sulfur contact is formed with the aromat edge at N3 with a distance of ˚ . The contact distances are consistent with 3.85 A specific sulfur interactions. However, since the resolution of the structure does not establish the position of the Met CE atom unambiguously, details of the interaction geometry and interaction type are not possible to analyze here. Further structural studies are required for this and to establish the degree of flexibility of the methionine residue. Methionine is conserved at this position in all PKB isoforms, and in some PKC isoforms and other kinases including S6K, Kin1/2. Another methionine residue, M120 at the hinge region of the adenine-binding site (conserved also among PKB isoforms and other AGC kinases) provides † Internet Atlas of Protein Side-Chain Interactions: http://www.biochem.ucl.ac.uk/bsm/sidechains
additional sulfur – ATP aromatic interactions. These are edge (co-planar), and occur between Met120 and N6 and N7 of ATP with distances ˚ to approximately 4 A ˚. varying from 3.6 A Triple mutant corrects 181 side-chain and restores the pocket In all its isoforms, PKB possesses lysine at the position corresponding to Q181 (Figure 1(a)). The observation of the Q181 rotation in PKAB2 suggested the introduction of the additional mutation Q181K to explore the characteristics of this unexpected feature. We therefore created the triple mutant V123A, L173M, Q181K (PKAB3) and determined its crystal structure in the absence of ligand in the ATP-binding site. The structure shows that K181, like Q181 in native PKA, is bound at the surface of the protein, and that the hinge-binding region of the ATP pocket has the well known PKA-like geometry. Comparison of surface polar networks A comparison of the interactions of lysine and glutamine in their protein surface-bound conformations suggests that glutamine is, in general, more susceptible to this potential rearrangement. In both, ATP-bound PKAB2 and PKA, only a single H-bond, to the backbone oxygen atom of I180, anchors Q181 into its external position. A second ˚ ) with R308:NH2 might provide contact (at 3.4 A an additional tether, but only if Q181 is partially disordered such that the amide oxygen atom may be partially oriented toward R308 (Figure 7(a)). (R308 is fixed in a salt-bridge interaction with D175:OD2 and with a hydrogen bond to I180:CO, so the proximity to Q181 may result from these other contacts and even have repulsive character.) In contrast to the weakly bound Q181 in PKA, the newly introduced lysine residue in the PKAB3
PKA-PKB Hybrid Mutants
1029
Figure 7. (a) Stereo view of Q181 distance measurements of the liganded PKAB2. The sole strong H-bond of Q181 (with I180:CO) and the contact with R308:NH2 are highlighted in yellow. All other potential hydrogen bond partners are too distant to form H-bonds. (b) Stereo view of polar interactions involving and surrounding K181 in unliganded PKAB3. The hydrogen bonding network is highlighted in yellow. Important interactions include tight binding to the backbone oxygen atom of P101 and indirect interactions to the backbone oxygen atom of L103 and the side-chain oxygen atom of E121:OE1 via a strongly fixed water molecule (X60). Weaker indirect contacts via a water molecule (X61) to the backbone oxygen atom of I180 and side-chain oxygen atom D180:OE1 are visible.
structure is anchored tightly between two complementary charges via an extended H-bond network involving water molecules and several residues ˚ direct hydrogen (Figure 7(b)). K181 shares a 2.9 A bond interaction with the backbone oxygen atom of P101. A second strong H-bond to a tightly bound water molecule links lysine further to the backbone oxygen atom of L103 and the complementary charge of E121:OE1. Furthermore, there are weaker indirect contacts via water X61 to the
backbone oxygen atom of I180 and side-chain oxygen atom D175:OD2. Significantly, the two charged residues E121 and D175 that are hydrogen bonded indirectly with Lys181 are absolutely conserved among all isoforms of PKA, PKB and PKC (Figure 1(a), highlighted in green for PKB isoforms and PKAa). Additionally, the hydrophobic residues whose main-chain carbonyl groups form direct or indirect hydrogen bonds with K181 are absolutely or highly conserved among the AGC
1030
PKA-PKB Hybrid Mutants
group kinases. The entire side-chain (CG onwards) of R308, which forms a salt-bridge with Asp175 and is near Q181 in PKA, is disordered in the triple mutant PKAB3. It is not clear whether the mutation Q181K leads to disordering of R308 or whether this is an effect of the new crystal packing of PKAB3. The side-chain of K181 is partially fixed by hydrophobic contacts as it emerges to the protein surface, so it may compete successfully with R308 for ordering interactions with D175 as a salt-bridge partner. However, E121 is available also as a complementary charge and the introduction of lysine at position 181 neutralizes the otherwise apparent negative surface charge. The crystal structure of PKB13 shows that the lysine residue corresponding to position Q181 is indeed fixed by a salt-bridge with the equivalent of E121 of PKA and by a hydrogen bond to the equivalent of P101:CO. The lysine equivalent of R308 in PKB is ordered and present as a counterion but at a distance too large to be considered a saltbridge.
inhibitors H7, H9 and HA1077 show roughly the same binding energy loss when comparing PKAB2 to PKAB3; this represents in the first approximation the energy required to expel Q181 from the hinge region. Similarly, they show roughly the same binding energy loss when comparing PKAB3 to native PKA (Table 1). Again in a first approximation, this energy loss should reflect differences in binding energies that arise from the two mutations V123A and L171M and their effects on either inhibitor or solvent interactions. The data may indicate an anomalously weakened binding for H7 to PKAB3 and thus reflect PKA selectivity over PKB. Further, a relatively lowered sensitivity of HA1077 binding strengths to both PKAB2 and PKAB3 mutations may appear anomalous. Both apparent anomalies are, however, on the margins of statistical significance. A precise analysis of these effects will require experiments with a variety of inhibitors with a varied PKA and PKB selectivity profile.
Biacore binding results
Summary
The surface plasmon resonance spectroscopy binding assays confirms the weakening of ligand binding to PKAB2. This is evidently due to the energy required for the ligand to induce the movement of Q181 away from the binding pocket into the position seen in the PKAB2 – ATP complex structure (Table 1). The ligand-induced transition might be visualized in detail as follows. During ATP binding, a rotation of T183 takes place to create the interaction of T183:OG1 and ATP:N7. This rotation disrupts the packing of Q181, V104, and T183, so that the surface structure of Q181 becomes the energetically preferred geometry. The energy costs associated with this rearrangement leads to an increase of the Km value of ATP by a factor of 10. Most of the apparent energy cost is eliminated with the restoration of the adenine pocket by the introduction of the Q181K mutation in PKAB3. A similar effect of lowered binding affinities can be observed by binding of isoquinoline inhibitors to PKAB2 in comparison to PKA and PKAB3. Although the isoquinolines do not possess an equivalent to N7 of ATP, an equivalent energy cost is associated with their binding. Thus, to within experimental error, ATP and all three
The utility of a surrogate protein for structurebased selectivity studies lies in the ability to observe individual protein –ligand interactions and draw conclusions relevant to the actual target. Mutations of the surrogate protein in order to better approximate the target clearly facilitates this. Ideally then, the geometries of ligand – sidechain interactions for both the target and a (mutated) surrogate should be identical to within the accuracy relevant to modelling. For most practical modelling tasks, a structure that shows the set of polar interactions and extent of buried surface areas is sufficient for planning new chemical syntheses. When more subtle effects become important, such as when ligand-induced conformation changes occur, or when the need for selectivity against very similar targets is important, structural details including, for example, the quality of hydrogen bonding, become important and the relevance of the surrogate protein must be correspondingly high. Further, it would be desirable that the surrogate reflect the kinetic properties of the target as well. This would enhance confidence that the observed structures are relevant in all details, for example, or may offer advantages in screening. For PKB, PKA is already a good surrogate for many purposes. This study shows that the double mutant PKA (V123A, L173M) could represent a better PKB model than PKA, but only for ligands whose binding properties depend significantly on the two substitutions, and then only if the ligands share identical costs of ejecting Q181 from the hinge-binding site (as the isoquinoline inhibitors H7, H9 and HA1077 appear to do). With the triple mutant V123A, L173M, Q181K, this defect is removed. The crystal structure determination, however, was necessary to diagnose the cause of
Table 1. Ki/Km values determined from surface plasmon resonance spectroscopy Inhibitor/substrate (mM) H7 H9 HA1077 ATP
PKA
PKAB2
PKAB3
6.8 ^ 2.2 3.3 ^ 2.2 5.7 ^ 1.2 9.6 ^ 2.1
94.8 ^ 0.7 44.0 ^ 8.7 34.9 ^ 1.5 115.4 ^ 18.6
36.6 ^ 0.9 7.7 ^ 1.7 10.0 ^ 3.3 24.1 ^ 3.7
H7, 1-(5-isoquinolinesulfonyl)-2-methylpiperazine. H9, N-(2aminoethyl)-5-isoquinolinesulfonamide. HA1077 (Fasudil), (5isoquinolinesulfonyl)homopiperazine.
1031
PKA-PKB Hybrid Mutants
Figure 8. Overlay of the ATP-binding sites of PKB and PKAB3 (stereo view). The conformers of the mutated residues of PKAB3 are very similar to PKB, as is the ATP-binding site as a whole. The greatest differences occur at the b-turn regions of the respective glycine loops (here depicted as ribbons and apparently disordered in PKB) and the conformer of flexible residue 104.
the weakened ligand binding; without the structure it would have been natural to assume that weakened interactions with either A123 or M173 would have been the cause. The procedure here has been analogous to the evolutionary occurrence of a second site-suppressor mutation (Q181K) to correct a defect caused by an earlier mutation (V123A). A superposition of the structures of PKAB3 and PKB highlights the similarity of conformers of A123, M173 and K181 (Figure 8). Furthermore, the equivalence of the ATP-binding site as a whole is evident. Differences occur at the b turn region of the glycine loop, which is apparently partially disordered in the PKB structure, and at residue 104, which is known to be flexible among PKA structures.30 In general, therefore, the PKB structure validates the surrogate approach; PKAB3 provides a good model for the adenine-binding site and further provides a tool useful to analyze the influence of amino acid residue variations on ligand-binding properties. While the accrual of 3D structures, point mutations, and activity data for protein kinases attests to their complexity and apparent unpredictability, this same catalog of data increasingly allows systematization of structure–function relationships. The relative N and C-terminal domain orientations, phosphorylation-induced refolding of the activation loop, and co-factordependent restructuring of structural elements into active configurations are examples of recurring phenomena. Similarly, the roles of individual residues can be described for catalytic, selectivity and other properties. The propensity of Q181 to switch between binding at the ATP pocket or at the outer surface of the protein seems to be an artifact of mutation, but may find a counterpart in a natural kinase, in particular because this is a possible mechanism to communicate the occupation state of the ATP pocket to a potential co-factor or conversely to alter ATP site ligandbinding properties via co-factor binding.
Materials and Methods Mutagenesis of PKA Site-directed mutagenesis was performed using pT7PKA as template and the Stratagene (La Jolla) Quick Change Kit following the manual of the supplier including the design of the corresponding primer pairs. All constructs using for protein expression were verified by DNA sequencing.
Expression and purification of PKA, PKAB2 and PKAB3 Expression vectors pT7-7 and pET28b, both carrying a bacteriophage T7 promoter, were used to express bovine PKA Ca31 and PKAB2 and PKAB3 in Escherichia coli strain BL21(DE3), respectively. Two positions distinguish bovine (N32, M63) from human PKA (S32, K63), both at sites distant from substrate and ATP-binding sites. Cells were grown in LB medium supplemented with either 100 mg/ml of carbenicillin or 75 mg/ml of kanamycin, depending on the appropriate vector. After reaching an A610 of 0.7, cells were induced with 0.4 mM isopropyl-bD -galactopyranoside (IPTG) for 16 hours at 24 8C. Cells were collected by centrifugation, resuspended in 30 mM Mes, 50 mM KCl, 1 mM EDTA, 1 mM dithiothreitol and sonicated. For affinity purification,32,33 the supernatant of a one hour, 30,000g centrifugation was supplemented with 3 mM MgCl2, 2 mM ATP and applied to immobilized protein kinase inhibitor peptide PKI-(5-24) in batch for one hour at room temperature. The gravitypacked column was washed with five volumes each of TMN 50 and TMN250 (20 mM Tris – HCl (pH 7.4), 2 mM MgCl2, 50/250 mM NaCl 0,4 mM ATP). About 5 mg of heterogenously phosphorylated PKA or the abovementioned derivatives per liter of cell culture was eluted with a buffer solution of 200 mM arginine, 50 mM NaCl, 1 mM EDTA, 20 mM Tris – HCl (pH 7.4). The enzymes were concentrated via Centripreps 30 (Millipore) and diluted fivefold with 25 mM BisTris – propane (pH 8.5). Homogenously phosphorylated samples were obtained by applying these samples to a Mono-S 10/10 column and eluting with a LiCl gradient up to 300 mM in 25 mM BisTris – propane at pH 8.5. The identity of the
1032
PKA-PKB Hybrid Mutants
Table 2. X-ray data and refinement statistics
Space group Cell constants ˚) a (A ˚) b (A ˚) c (A ˚) Data resolution (A Rsym (Rsym last shell) Completeness 2s (last shell) Rcryst-work, Rcryst-free Weighted Rwork, Rfree RMS deviation ˚) Bonds (A Angles (deg.)
samples was spectrometry.
confirmed
by
PKAB2
PKAB2-ATP
PKAB3
P212121
P212121
P212121
73.5 76.7 80.5 2.6 0.12 (0.34) 0.96 (0.91) 0.204, 0.256 0.186, 0.235
73.3 75.2 80.3 2.6 0.10 (0.39) 0.86 (0.72) 0.197, 0.241 0.161, 0.191
59.9 79.4 100.9 2.2 0.14 (0.34) 0.94 (0.94) 0.218, 0.259 0.200, 0.236
0.012 1.3
0.010 1.3
0.010 1.2
electrospray
mass
Biacore sensor chip preparation Proteins used for Biacore analysis were dialyzed twice with the 400-fold volume of 50 mM Mops (pH 6.8), 10 mM MgCl2, 50 mM KCl. Coupling of PKA, PKAB2 and PKAB3 to CM5 Biasensor chip via amine group linkage was achieved using standard coupling procedures.34 Briefly, CM5 sensor chips were activated by injecting 35 ml of a 1:1 (v/v) mixture of N-ethyl-N0 -[dimethylamino]carbodiimide/N-hydroxysuccinimide at 5 ml/ minute. After diluting the proteins in 10 mM sodium acetate (pH 5.5), PKA, PKAB2, PKAB3 and HSA as a control were coupled to the CM5 sensor chip by injecting a 50 mM solution of the protein selected with a flow-rate of 5 ml/minute until 11,000 RU was reached. Generation of kinetic binding data Kinetic studies with a range of analyte concentrations were determined at a flow-rate of 10 ml/minute by allowing 300 seconds for association and 900 seconds for dissociation. Analyte concentrations used were of (256), 128, 64, 32, 16, 8, 4, 2, (1) mM concentrations (depending on the analyte), diluted in MilliQ water or running buffer (50 mM Mops (pH 7.4), 50 mM KCl, 10 mM MgCl2). Kinetic data were analyzed with BIA evaluation 3.0 software. For each binding curve, the response obtained using the HSA cell as control was subtracted. Due to the small signals (up to 40 RU) the steady-state affinity model was used to determine the Km/Ki of the different small molecular mass compounds. Goodness of fit measured as a x2 value was less than 4 for binding of the low molecular mass compounds. All binding experiments were repeated either two or, more usually, three times, and biosensor chips coupled at different times yielded surfaces with identical binding affinities. The binding affinities of ATP, H735 H9,35 HA1077,36 all three purchased from Calbiochem, to PKA were similar to the Km/Ki values reported in different studies using enzymatic assays.35 – 38,15 X-ray crystallography Hanging droplets containing 18 mg/ml of protein, 25 mM Mes-BisTris, 75 mM LiCl, 0.1 mM MgCl2, 1 mM dithiothreitol,15 1.5 mM octanoyl-N-methylglucamide,39 1 mM PKI-(5-24), pH 6.3 – 6.6, 1.5 mM ATP and 3 mM
MgCl2 in the corresponding trials, were equilibrated at 5 8C against 12 – 18% (v/v) methanol. In each case, one crystal was sufficient to obtain a complete data set. Diffraction data were measured at 4 8C in a sealed capillary using either on a Bruker X1000 wire array detector (PKAB2-ATP) or a MAR image plate (PKAB2, PKAB3) on a copper target rotating anode X-ray generator (Rigaku RU200) with a graphite monochromator. Data evaluation was performed with SAINT (PKAB2-ATP) or MOSFLM and SCALA (PKAB2, PKAB3). Structure solution and refinement was done with REFMAC5 and other programs of the CCP4 crystallography package. (Data and model quality parameters are given in Table 2.)
Acknowledgements We are grateful to Dr Wolf Lehman and Dr Matthias Wind for expert mass spectrometry analysis and we thank Professor Volker Kinzel for constant support. This work was supported, in part, by the Bayerische Wirtschaftsministerium.
References 1. Cohen, P. (2002). Protein kinases—the major drug targets of the twenty-first century? Nature Rev. Drug Discov. 1, 309–315. 2. Fox, T., Coll, J. T., Xie, X., Ford, P. J., Germann, U. A., Porter, M. D. et al. (1998). A single amino acid substitution makes ERK2 susceptible to pyridinyl imidazole inhibitors of p38 MAP kinase. Protein Sci. 7, 2249– 2255. 3. Liu, Y., Bishop, A., Witucki, L., Kraybill, B., Shimizu, E., Tsien, J. et al. (1999). Structural basis for selective inhibition of Src family kinases by PP1. Chem. Biol. 6, 671– 678. 4. Shabb, J. B. (2001). Physiological substrates of cAMPdependent protein kinase. Chem. Rev. 10, 12381– 12411. 5. Montminy, M. (1997). Transcriptional regulation by cyclic AMP. Annu. Rev. Biochem. 66, 807– 822. 6. Bossemeyer, D. (1994). The glycine-rich sequence of protein kinases: a multifunctional element. Trends Biochem. Sci. 19, 201– 205. 7. Walsh, D. A. & Van Patten, S. M. (1994). Multiple
PKA-PKB Hybrid Mutants
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
pathway signal transduction by the cAMP-dependent protein kinase. FASEB J. 8, 1227– 1236. Taylor, S. S., Buecher, J. A. & Yonemoto, W. (1990). cAMP dependent protein kinase: framework for a diverse family of regulatory enzymes. Annu. Rev. Biochem. 59, 84 – 86. Brognard, J., Clark, A. S., Ni, Y. & Dennis, P. A. (2001). Akt/protein kinase B is constitutively active in non-small cell lung cancer cells and promotes cellular survival and resistance to chemotherapy and radiation. Cancer Res. 61, 3986– 3997. Kulik, G., Carson, J. P., Vomastek, T., Overman, K., Gooch, B. D., Srinivasula, S. et al. (2001). Tumor necrosis factor alpha induces BID cleavage and bypasses antiapoptotic signals in prostate cancer LNCaP cells. Cancer Res. 61, 2713– 2719. Ng, S. S. W., Tsao, M. S., Chow, S. & Hedley, D. W. (2000). Inhibition of phosphatidylinositide 3-kinase enhances gemcitabine-induced apoptosis in human pancreatic cancer cells. Cancer Res. 60, 5451– 5455. Itoh, N., Semba, S., Ito, M., Takeda, H., Kawata, S. & Yamakawa, M. (2002). Phosphorylation of Akt/PKB is required for suppression of cancer cell apoptosis and tumor progression in human colorectal carcinoma. Cancer, 94, 3127 –3134. Yang, J., Cron, P., Thompson, V., Good, V. M., Thompson, V., Hemmings, B. A. & Barford, D. (2002). Crystal structure of an activated Akt/Protein Kinase B ternary complex with GSK3-peptide and AMP-PNP. Nature Struct. Biol. 9, 940– 944. Kumar, C. C., Diao, R., Yin, Z., Liu, Y., Samatar, A. A., Madison, V. & Xiao, L. (2001). Expression, purification, characterization and homology modeling of active Akt/PKB, a key enzyme involved in cell survival signaling. Biochim. Biophys. Acta, 1526, 257–268. Engh, R. A., Girod, A., Kinzel, V., Huber, R. & Bossemeyer, D. (1996). Crystal structures of catalytic subunit of cAMP-dependent protein kinase in complex with isoquinolinesulfonyl protein kinase inhibitors H7, H8, and H89. Structural implications for selectivity. J. Biol. Chem. 271, 26157– 26164. Bossemeyer, D., Engh, R. A., Kinzel, V., Ponstingl, H. & Huber, R. (1993). Phosphotransferase and substrate binding mechanism of the cAMP-dependent protein kinase catalytic subunit from porcine heart as deduced from the 2.0 A structure of the complex with Mn2þ adenylyl imidodiphosphate and inhibitor peptide PKI(5-24). EMBO J. 12, 849– 859. Madhusudan, A. P., Xuong, N. H. & Taylor, S. S. (2002). Crystal structure of a transition state mimic of the catalytic subunit of cAMP-dependent protein kinase. Nature Struct. Biol. 9, 273– 277. Prade, L., Engh, R. A., Girod, A., Kinzel, V., Huber, R. & Bossemeyer, D. (1997). Staurosporine-induced conformational changes of cAMP-dependent protein kinase catalytic subunit explain inhibitory potential. Structure, 5, 1627– 1637. Narayana, N., Diller, T. C., Koide, K., Bunnage, M. E., Nicolaou, K. C., Brunton, L. L. et al. (1999). Crystal structure of the potent natural product inhibitor balanol in complex with the catalytic subunit of cAMP-dependent protein kinase. Biochemistry, 38, 2367–2376. Lew, J., Coruh, N., Tsigelny, I., Garrod, S. & Taylor, S. S. (1997). Synergistic binding of nucleotides and inhibitors to cAMP-dependent protein kinase examined by acrylodan fluorescence spectroscopy. J. Biol. Chem. 272, 1507– 1513.
1033
21. Seifert, M. H., Breitenlechner, C. B., Bossemeyer, D., Huber, R., Holak, T. A. & Engh, R. A. (2002). Phosphorylation and flexibility of cyclic-AMP-dependent protein kinase (PKA) using (31)P NMR spectroscopy. Biochemistry, 41, 5968–5977. 22. Karlsson, R., Madhusudan, Taylor, S. S. & Sowadski, J. M. (1994). Intermolecular contacts in various crystal forms related to the open and closed conformational states of the catalytic subunit of campdependent protein kinase. Acta Crystallog. sect. D, 50, 657 –662. 23. Engh, R. A. & Bossemeyer, D. (2002). Structural aspects of protein kinase control—role of conformational flexibility. Pharmacol. Ther. 93, 99 – 111. 24. Zheng, J. H., Trafny, E. A., Knighton, D. R., Xuong, N. H., Taylor, S. S., Teneyck, L. F. & Sowadski, J. M. (1993). 2.2-Angstrom refined crystal structure of the catalytic subunit of cAMP-dependent protein-kinase complexed with MnATP and peptide inhibitor. Acta Crystallog. sect. D, 49, 362– 365. 25. Shaffer, J. & Adams, J. A. (1999). An ATP-linked structural change in protein kinase A precedes phosphoryl transfer under physiological magnesium concentrations. Biochemistry, 38, 5572– 5581. 26. 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, 2154– 2161. 27. Pal, D. & Chakrabarti, P. (2001). Non-hydrogen bond interactions involving the methionine sulfur atom. J. Biomol. Struct. Dynam. 19, 115 – 128. 28. Meyer, E. A., Castellano, R. K. & Diederich, F. (2003). Interactions with Aromatic Rings in Chemical and Biological Recognition. Angew. Chem. Int. Ed. 42, 1210 –1250. 29. Rosenfield, R. E., Jr, Parthasarathy, R. & Dunitz, J. D. (1977). Directional preferences of nonbonded atomic contacts with divalent sulfur. 1. Electrophiles and nucleophiles. J. Am. Chem. Soc. 99, 4860– 4862. 30. Engh, R. A. & Bossemeyer, D. (2002). Structural aspects of protein kinase control—role of conformational flexibility. Pharmacol. Ther. 93, 99 – 111. 31. Wiemann, S., Kinzel, V. & Pyerin, W. (1992). Cloning of the C alpha catalytic subunit of the bovine cAMP-dependent protein kinase. Biochim. Biophys. Acta, 1171, 93 – 96. 32. Ohlsen, S. R. & Uhler, M. D. (1989). Affinity purification of C alpha and C beta isoforms of the catalytic subunit of cAMP-dependend protein kinase. J. Biol. Chem. 264, 18662– 18666. 33. Girod, A., Kinzel, V. & Bossemeyer, D. (1995). In vivo activation of recombinant cAPK catalytic subunit active site mutants by coexpression of the wild-type enzyme, evidence for intermolecular cotranslational phosphorylation. FEBS Letters, 391, 121– 125. 34. Lo¨fa˚s, S. & Johnsson, B. (1990). A novel hydrogel matrix on gold surfaces in surface plasmon resonance sensors for fast and efficient covalent immobilization of ligands. J. Chem. Soc. Chem. Commun. 21, 1526 –1528. 35. Hidaka, H., Inagaki, M., Kawamoto, S. & Sasaki, Y. (1984). Isoquinolinesulfonamides, novel and potent inhibitors of cAMP dependent protein kinase and protein kinase C. Biochemistry, 23, 5036– 5041. 36. Asano, T., Suzuki, T., Tsuchiya, M., Satoh, S., Ikegaki, I., Shibuya, M. et al. (1989). Vasodilator actions of HA1077 in vitro and in vivo putatively mediated by
1034
PKA-PKB Hybrid Mutants
the inhibition of protein kinase. Br. J. Pharmacol. 98, 1091– 1100. 37. Whitehouse, S., Feramisco, J. R., Casnellie, J. E., Krebs, E. G. & Walsh, D. A. (1983). Studies on the kinetic mechanism of the catalytic subunit of c-amp dependent protein kinase. J. Biol. Chem. 258, 3693– 3701. 38. Whitehouse, S. & Walsh, D. A. (1983). MgATPdependent interactions of the inhibitor protein of
the cAMP dependent protein kinase with the catalytic subunit. J. Biol. Chem. 258, 3682– 3692. 39. Knighton, D. R., Zheng, J. H., Ten Eyck, L. F., Ashford, V. A., Xuong, N. H., Taylor, S. S. & Sowadski, J. M. (1991). Crystal structure of the catalytic subunit of cyclic adenosine monophosphate-dependent protein kinase. Science, 253, 407– 414.
Edited by K. Nagai (Received 12 December 2002; received in revised form 3 April 2003; accepted 15 April 2003)