Conserved spatial patterns across the protein kinase family

Transcription

1 Available online at Biochimica et Biophysica Acta 1784 (2008) Conserved spatial patterns across the protein kinase family Lynn F. Ten Eyck a,b,c,, Susan S. Taylor c,d, Alexandr P. Kornev b a Conway Institute of Biomolecular and Biomedical Research, University College Dublin, Belfield, Dublin 4, Ireland b San Diego Supercomputer Center, University of California at San Diego, 9500 Gilman Drive, La Jolla, California , USA c Department of Chemistry and Biochemistry, University of California at San Diego, 9500 Gilman Drive, La Jolla, California , USA d Howard Hughes Medical Institute, University of California at San Diego, 9500 Gilman Drive, La Jolla, California , USA Received 8 October 2007; received in revised form 2 November 2007; accepted 2 November 2007 Available online 17 November 2007 Abstract Protein kinases are a large family of enzymes heavily involved in signal transduction, regulation of metabolism, and control of cell growth and differentiation. These functions require precise recognition of widely diverse signals and substrates, and very detailed control of protein kinase activity. Large molecules interact primarily through recognition of surface features. Comparison of surfaces is complicated by both sequence diversity and conformational variability, including multiple possible rotameric states of side chains. We used a recently developed method of protein surface comparison to compare different serine/threonine and tyrosine kinases. As we have shown, two hydrophobic cores inside a protein kinase molecule are connected by a unique formation, called the spine. It exists only in the active conformation of protein kinases and is dynamically disassembled during the inactivation process. Detection of such structures by any other method was not possible as the residues which comprise the spine do not form any sequence or 3D motifs in a traditional sense Elsevier B.V. All rights reserved. Keywords: Protein kinase; PKA; Regulation; Activation mechanism; Protein surface; Graph-theory method 1. Introduction Protein kinases function primarily as regulators of metabolism, growth, and differentiation of cells, through specific modification of other proteins. Kinases themselves are also highly regulated, often by other protein kinases, which gives rise to complex regulatory networks [1,2]. The analogy with electronic control mechanisms, in which transistors act as switches to control other transistors, is tempting and useful up to a point. The very different physical realization of these control networks, however, has profound implications with respect to stability and specificity of the networks. Electronic networks are precisely delimited by specific Abbreviations: camp, Cyclic-3',5' -adenosine monophosphate; PKA, camp-dependent protein kinase; ECC, Edge Comparison and Combination; CNB domain, Cyclic Nucleotide Binding domain; PBC, phosphate binding cassette; CAP, catabolite gene activator protein; HCN channel, hyperpolarization-activated, cyclic nucleotide-modulated channel; C-subunit, catalytic subunit of PKA; R-subunit, regulatory subunit of PKA; MAP kinase, mitogen activated protein kinase Corresponding author. Tel.: ; fax address: lynn.teneyck@ucd.ie (L.F. Ten Eyck). physical connections between logical elements. In biological networks, transmission of signals depends on diffusive transport or transport through localized channels or carrier proteins. The actual connections in biological systems depend on stochastic molecular recognition processes, and thus have a larger noise level and more crosstalk between states than electronic mechanisms. Despite their versatility in terms of substrates, regulation and function, protein kinases have a high degree of structural and sequence similarity among the family members [3,4]. Regulation of protein kinase activity occurs through allostery, competitive inhibition, and non-competitive inhibition mechanisms [1,2]. Allosteric regulation is modification of activity through non-covalent binding of other molecules at a site different from the active site. It is presumed to function through conformational changes that affect substrate binding and/or catalysis. Non-competitive inhibition is caused by covalent changes to the enzyme structure that make the kinase inactive. The most common non-competitive inhibition of protein kinases is through addition of a phosphate group (by a protein kinase) or removal of phosphate groups (by a protein phosphatase). Competitive inhibition occurs when a non-substrate binds at the active site and blocks access to the active site by the substrate /$ - see front matter 2007 Elsevier B.V. All rights reserved. doi: /j.bbapap

2 L.F. Ten Eyck et al. / Biochimica et Biophysica Acta 1784 (2008) These mechanisms of regulation are not mutually exclusive. For example, PKA is primarily regulated by competitive inhibition by the R-subunit, which contains a pseudo-substrate sequence. The conformation of the R-subunit is allosterically regulated by camp, which alters the conformation of the subunit and disrupts the binding to the C-subunit of PKA. A second level of regulation of PKA is provided through the covalent modification at Thr197 by addition or removal of phosphate. Protein protein recognition, which is the fundamental first process in protein kinase chemistry, depends on non-covalent chemical binding of one molecular surface to another. Characterization of protein binding surfaces is complex due to the mobility of the surfaces, which can occur at two levels. First, the side chains of those amino acids exposed to the solvent (and hence available for recognition) are often free to rotate, and are perhaps best represented as ensembles of rotameric states instead of as discrete conformations. This is clearly shown by the fact that these side chains are often invisible in crystallographic electron density maps, which usually means that the density is distributed over a sufficiently large volume of space that it is indistinguishable from the background noise in the map. Second, the underlying molecular shape may change due to changes in the conformation of the polypeptide backbone caused by such factors as binding to other molecules, changes in environmental conditions, or simple flexibility of structure. All of these factors have been observed in protein kinase structures and are important aspects of the regulation of protein kinase activity. 2. Procedures employed The ECC algorithm for comparison of local spatial environments [5,6] was developed to address these issues. The basic Fig. 1. The PKA RIIβ regulatory subunit A domain is shown as a typical example of a cyclic mononucleotide binding domain. The PBC is colored red. The camp is shown edge-on to the adenosine ring; the sugar phosphate rings are in the plane of the paper. The side chain volumes show hydrophobic residues, conserved in position but not in sequence. In this case the capping residue, to the upper left of the adenosine ring, is the hydrophobic side chain of R 381, which is positioned by the bending of the B/C helix; additional structural elements are omitted for clarity. When the camp is absent, the B/C helix straightens to one long helix that binds to the catalytic subunit. These hydrophobic residues were not recognized as common structural elements until the ECC algorithm was applied to the structural comparison.

3 240 L.F. Ten Eyck et al. / Biochimica et Biophysica Acta 1784 (2008) notion is that the local spatial environment can be characterized by the position and orientation of the Cα Cβ bonds, which are the major determinants of the volume that can be occupied by the corresponding side chains. Such approach allowed us to develop a measure of similarity that depends on the range of conformational possibilities available to side chains, rather than directly comparing coordinate sets. The ECC method finds sets of residues that have compatible side chains occupying similar relative positions in space. A pair of residues in one structure is considered to be similar to a pair of residues in another structure if the Cα Cβ bonds are in similar relative orientations and the residue types are similar. This means that if the Cα and Cβ atoms are superposed, the remainder of the side chain atoms can occupy corresponding volumes in space, and thus are at least potentially presenting similar binding sites. Similarity is determined through a userdefined lookup table, which can specify a similarity score appropriate to the application. We have used evolutionary similarity tables and tables based on similarity of chemical properties of side chains. The work described here generally uses chemical similarity (acidic, basic, polar, and hydrophobic). The lists of similar pairs are converted into a graph by linking all of the pairs that have residues in common, provided they occur in both lists. For example, if the first structure contains pairs (A, B), (B, C) and (A, C) which are similar to pairs (P, Q), (Q, R) and (P, R) in the second structure, the triangle pattern (A B C) will be considered similar to (P Q R). The graphs are extended to produce the largest common subgraph, which represents the largest similar motif. Full details of the method and characterization of the parameters controlling the comparisons are given in the Supplementary Material deposited with the Proceedings of the National Academy of Sciences of the United States of America [6]. An additional concept, the involvement score, was introduced in [6]. This score is used in the comparison of a set of graphs, and simply counts the number of edges (i.e. the number of conserved relationships) for each residue in the graph. A large involvement score means that a residue and its local geometry are strongly conserved. When two structural states are being compared, as in the comparison of active and inactive states of protein kinases, the difference between the involvement scores for the two states quickly and concisely shows which Fig. 2. A roadmap through the protein kinase catalytic core, showing common structural interactions between conserved motifs present in families studied. Sequence positions are given for the protein kinase A motifs. Interactions shown by the blue arrows are present in the conserved catalytic core; those shown in red are to mobile elements that can be displaced in inactive forms but are present in the active forms. The red pentagon represents ATP, and the purple circle represents the substrate.

4 L.F. Ten Eyck et al. / Biochimica et Biophysica Acta 1784 (2008) portions of the molecule change in a conserved way when the state of the molecule is altered. 3. Overall findings The ECC method has been used to find additional significant structural features in camp binding proteins [5] (some of which regulate protein kinases), to show differences between active and inactive forms of protein kinases [6], and to find significant new elements of the architecture of the protein kinase catalytic core [6] CNB domains study A comparison of cyclic nucleotide binding (CNB) domains in the Protein Data Bank gave insight into the allosteric mechanism by which camp binding modulates protein structural states in cyclic mononucleotide regulated protein kinases, CAP, and HCN channels [5]. Each domain has a helical subdomain and an eight stranded β-sandwich. The key docking site for camp, the Phosphate Binding Cassette (PBC) is embedded within the β-subdomain. In addition to the expected similarities in the PBC, a conserved hydrophobic capping residue covering the distal side of the adenine ring was identified, and a set of five conserved hydrophobic residues was found using the ECC method in the region including strands β2 through β4. Each regulatory subunit of PKA has two tandem CNB domains (A and B) at the C-terminus. The capping residue is usually positioned by the C helix, which is bent or kinked in all cases studied for which camp was bound; in the unbound state, the C helix becomes fully extended, opening and exposing the camp binding site. The five conserved hydrophobic residues form a second layer on the inner side of the camp binding site that protects the hydrogen bonds in the phosphate binding cassette from water and the camp from phosphodiesterases. Fig. 1 shows the A domain of the RIIβ regulatory subunit of PKA as a typical example Protein kinase activation mechanism Activation of protein kinases is a complex process involving significant rearrangements of the catalytic core. This rearrangement can be triggered by a number of distinct events, depending on the particular kinase and the regulatory requirements of the system. At the present time there appears to be only one way to put together a protein kinase active site, but many ways to inactivate the site. A schematic map of the significant structural fragments and motifs of the catalytic core of a protein kinase is given in Fig. 2, which uses the PKA numbering for reference. This numbering is used throughout this paper. To study the activation process we compared active and inactive conformations of different serine/threonine and tyrosine kinases. Although many protein kinase structures are presently known, few are known in both active and inactive conformations. In this work we found two features that characterize all of the active conformations. The first is the conformation around G 186 of the DFG motif in the magnesium binding portion of the activation segment and the following two residues (see Fig. 3); in the active form there is a hydrogen bond between the side chain of D 184 and the amide nitrogen of G 186.D 184 coordinates all three phosphates of the ATP, either directly or through the Mg + ions. There is also a hydrogen bond between the carbonyl oxygen of F 185 and the DFG + 2 amide nitrogen, forming a 3-turn. Phenylalanine F 185 forms hydrophobic contacts with the C helix and the nearby HRD motif of the activation loop (Y 164 RD in PKA, but almost universally called the HRD motif). A simple twist of the peptide bond between F 185 and G 186 is sufficient to alter the conformation of the D 184 side chain and disrupt its Fig. 3. A portion of the active site of PKA shows the conserved elements identified by the methods of this paper in 23 active protein kinase structures. camp, the DFG motif and phosphorous atoms are colored yellow. The rest of the activation loop is colored grey. The HRD motif from the catalytic loop is colored tan. D 184 is positioned to support the ATP phosphate tail by a hydrogen bond to the main chain of the D 184 FG motif. The same peptide bond is part of a 3-turn to the DFG+2 residue. This orients the peptide bond between F 185 and A 188 so that it stabilizes the position of R 165 which in turn is the primary interaction partner of the phosphorylated T 197. This provides a path for the activating phosphorylation to stabilize ATP in a favorable position for transfer of the phosphate to a substrate.

5 242 L.F. Ten Eyck et al. / Biochimica et Biophysica Acta 1784 (2008) catalytically essential ability to position the ATP. The same twist also disrupts interactions between the DFG motif and the HRD motif (Y 164 RD in PKA), which position R 165 to interact with the activating phosphothreonine at position T 197. There is thus a chain of interactions leading directly from the activating phosphorylation site down into the catalytic site, all of which depend on the conformation at a single glycine. We regarded this as particularly noteworthy because G 186 is among the most highly conserved residues in all kinases, and was found in essentially the same conserved spatial relationship with many other residues in all active kinases examined, but was not found in conserved relationships in any of the active inactive comparisons studied. The change of conformation would not be possible for any residue other than glycine due to a steric hindrance from the Cβ, which is a plausible explanation for the total conservation of glycine at this position in the sequence of all protein kinases Dynamically assembled hydrophobic structure A second notable feature of the active structures was found when we looked at conserved hydrophobic residues. We found a spine of four residues (Y 164,F 185,L 95,andL 106 )thatisformed during the activation process and which links together the N and C lobes of the protein. L 95 and F 185 are both positioned during the activation process, F 185 through the interactions described above, and L 95 by the movement of the C helix. This forms a direct connection between the rigid N and C lobes of the catalytic subunit (Fig.4). These four residues are found in a set of 23 protein kinases representing five subfamilies of the eukaryotic protein kinases as well as two more distantly related prokaryotic kinases. Two recent experimental studies support the significance of the spine residues. Mutation of the DFG-phenylalanine (F 185 in PKA) in p38 MAP kinase to alanine, arginine or glycine totally abolished catalytic activity of the kinase [7]. Mutation of the phenylalanine to tyrosine led to substantial reduction of the p38 activity (by two orders of magnitude). Clearly, tyrosine can substitute for phenylalanine to form the spine, but the presence of the hydroxyl group will decrease its stability. Another study was done on the MAP kinase ERK2 [8]. The authors showed that mutation of I 84 (analog of L 106 in PKA) to alanine increased hydrogen/deuterium exchange of the activation loop and increased intramolecular autophosphorylation. Similar results were obtained after mutation of L 73 (analog of L 95 ) to proline. The explanation of these effects is rather obvious: Destabilization of the spine leads to a general loosening of the kinase structure and the activation loop in particular (via the DFG-phenylalanine). Increased flexibility of the activation segment would be expected to favor autophosphorylation. 4. Concluding remarks and perspectives Fig. 4. The hydrophobic spine formed on activation, shown in red, consists of Y 164,F 185,L 106, and L 95. The position of the central two residues is dependent on the active configuration shown in Fig. 3. The spine links the hydrophobic cores of the two lobes (N-lobe is colored blue; C-lobe is colored yellow) of the catalytic core, and thus establishes the relative positioning of portions of the active site contributed by the two lobes. The ATP position is shown to establish a point of reference. The allosteric mechanism proposed for the binding of cyclic mononucleotides through a potentially large structural change triggered by release of the capping residue was demonstrated in a crystal structure determination of the complex between the PKA RIα A domain and the catalytic subunit [9]. In the presence of camp the PKA regulatory subunits are in a compact configuration with the C helix kinked and held in place by the binding of the capping residue to the hydrophobic face of the adenosine ring. Release of camp allows the C helix to straighten out, causing large movements of whatever is connected to it. In the case of the RIα subunit, this will move the entire B domain. The extended helix is part of the interface that binds to the catalytic subunit. The two conformations are clearly in equilibrium; the excess catalytic subunit can displace camp, and the excess camp can displace the catalytic subunit. With structures now available for two very different conformational states of the CNB domains for RIα [9,10] and RIIα [11] we are now using ECC method to analyze these conformations. The analysis of the protein kinase catalytic cores is still in progress. The features described here, and in more detail in [6], do not answer the question of what drives the conversions between inactive and active conformations. It adds precision to the questions to be asked experimentally. Clearly the conformation of the peptide bond between the DFG-phenylalanine and glycine is critical to the structure of the 23 observed active sites. The connections shown in Fig. 3 provide a direct linkage from the activating phosphorylation on T 197 through R 165 and the

6 L.F. Ten Eyck et al. / Biochimica et Biophysica Acta 1784 (2008) peptide bond between the DFG +1 and DFG +2 residues to the FG bond, which in turn directly positions D 184 to orient the triphosphate tail of ATP for transfer of the γ-phosphate [10]. The hydrophobic spine shown in Fig. 4 is formed only when the activation segment and the C helix have reorganized. Under these conditions F 185 in the activation segment and L 106 in the C helix are aligned to create the spine. The spine supports the correct relative orientation of the small and large lobes of the kinase core for activity, positioning different portions of the active site and the ATP binding site. Another possible role for the spine is to provide a flexible hinge for the open/closed transition required for binding ATP and releasing ADP, since it is directly perpendicular to a line of residues identified as close to the axis of this motion [12]. InFig. 4 we show not only the spine but also sets of hydrophobic residues that interface with both ends of the spine. The methods described here have proven very powerful at extracting significant features from structural comparisons. The kinase comparisons [6] ranged over five classes of eukaryotic kinases and two prokaryotic kinases. The CNB studies [5] covered kinase regulatory subunits, CAP, and a camp-gated ion channel. No sequence alignments are required; this method will rapidly produce an alignment based on local structural features. Since the comparisons are local, movements of related motifs are detected regardless of position, which makes the procedure sufficiently robust to detect conserved motifs in the kinases despite the range of open and closed, and active or inactive conformations. The study of CNB domains also demonstrated a limitation of the method. The capping residues were not identified by ECC. They are generally hydrophobic or aromatic, but in one case (illustrated) the capping residue is arginine, with the long hydrophobic portion of the side chain interacting with the adenine ring. When residues have multiple significant properties, it is difficult to come up with a similarity score that is appropriate for all cases. Further, the capping residues are spatially conserved only in that they lie across the aromatic ring of the nucleotide base, and that their position depends on the distortion of the C helix. This illustrates the necessity and value of human examination of the results. ECC is remarkably powerful at identifying areas of importance, but interpretation remains a task for the investigator. Acknowledgments This research was supported by grants from the National Institutes of Health GM70996 and an E. T. S. Walton Visitor Award from the Science Foundation of Ireland (to L.F.T.E.), and from the National Institutes of Health GM19301 and the National Science Foundation DBI (to S.S.T.). Figures were rendered with PyMOL (DeLano Scientific LLC; References [1] S.S. Taylor, G. Ghosh, Protein kinases: catalysis and regulation, Curr. Opin. Struck. Biol. 16 (6) (2006) [2] J.A. Ubersax, J.E. Ferrell Jr., Mechanisms of specificity in protein phosphorylation, Nat. Rev., Mol. Cell Biol. 8 (7) (2007) [3] S.K. Hanks, T. Hunter, Protein kinases 6. The eukaryotic protein kinase superfamily: kinase (catalytic) domain structure and classification, Faseb J. 9 (8) (1995) [4] D.R. Robinson, Y.M. Wu, S.F. Lin, The protein tyrosine kinase family of the human genome, Oncogene 19 (49) (2000) [5] H.M. Berman, et al., The camp binding domain: an ancient signaling module, Proc. Natl. Acad. Sci. U. S. A. 102 (1) (2005) [6] A.P. Kornev, et al., Surface comparison of active and inactive protein kinases identifies a conserved activation mechanism, Proc. Natl. Acad. Sci. U. S. A. 103 (47) (2006) [7] M. Bukhtiyarova, et al., Mutagenesis of p38alpha MAP kinase establishes key roles of Phe169 in function and structural dynamics and reveals a novel DFG-OUT state, Biochemistry 46 (19) (2007) [8] M.A. Emrick, et al., The gatekeeper residue controls autoactivation of ERK2 via a pathway of intramolecular connectivity, Proc. Natl. Acad. Sci. U. S. A. 103 (48) (2006) [9] C. Kim, N.H. Xuong, S.S. Taylor, Crystal structure of a complex between the catalytic and regulatory (RIalpha) subunits of PKA, Science 307 (5710) (2005) [10] C. Kim, et al., PKA-I holoenzyme structure reveals a mechanism for camp-dependent activation, Cell 130 (6) (2007) [11] J. Wu, et al., PKA type IIalpha holoenzyme reveals a combinatorial strategy for isoform diversity, Science 318 (5848) (2007) [12] B. Lu, C.F. Wong, J.A. McCammon, Release of ADP from the catalytic subunit of protein kinase A: a molecular dynamics simulation study, Protein Sci. 14 (1) (2005)