Turning a protein kinase on or off from a single allosteric site via disulfide trapping

Abstract

There is significant interest in identifying and characterizing allosteric sites in enzymes such as protein kinases both for understanding allosteric mechanisms as well as for drug discovery. Here, we apply a site-directed technology, disulfide trapping, to interrogate structurally and functionally how an allosteric site on the Ser/Thr kinase, 3-phosphoinositide-dependent kinase 1 (PDK1)—the PDK1-interacting-fragment (PIF) pocket—is engaged by an activating peptide motif on downstream substrate kinases (PIFtides) and by small molecule fragments. By monitoring pairwise disulfide conjugation between PIFtide and PDK1 cysteine mutants, we defined the PIFtide binding orientation in the PIF pocket of PDK1 and assessed subtle relationships between PIFtide positioning and kinase activation. We also discovered a variety of small molecule fragment disulfides (< 300 Da) that could either activate or inhibit PDK1 by conjugation to the PIF pocket, thus displaying greater functional diversity than is displayed by PIFtides conjugated to the same sites. Biochemical data and three crystal structures provided insight into the mechanism of action of the best fragment activators and inhibitors. These studies show that disulfide trapping is useful for characterizing allosteric sites on kinases and that a single allosteric site on a protein kinase can be exploited for both activation and inhibition by small molecules.

Allosteric sites on protein kinases are difficult to find, characterize, and target with small molecules. Whereas the development of active site inhibitors has benefited greatly from the existence of a natural small molecule substrate, ATP, there are virtually no natural small molecule starting points from which allosteric kinase modulators can be built. Thus, most allosteric modulators have been discovered serendipitously, generally through high-throughput screening (HTS), which is inherently inefficient. Confirming the binding sites and mechanisms of action of allosteric modulators also tends to be more laborious relative to ATP-mimetic inhibitors. Prior to initiating HTS campaigns, it would be beneficial to have a simple method that permits directed interrogation of kinase allosteric sites with small compounds and facile assessment of the effects of binding on protein structure and function.

We envisioned that a site-directed approach developed previously by our laboratory, disulfide trapping (or tethering), would allow more straightforward characterization and interrogation of allosteric sites on protein kinases with small molecules. Disulfide trapping involves screening disulfide-containing compounds for their ability to form a mixed disulfide with a natural or engineered cysteine residue near a site of interest on a protein target (1, 2). Because the approach is site-directed, disulfide trapping has proven valuable as a tool to validate known, suspected, and orphan allosteric sites on protein surfaces as small molecule targets (3–6), but has not yet been applied to kinase allosteric sites.

Here, we use disulfide trapping to target an allosteric site on the surface of 3-phosphoinositide-dependent kinase 1 (PDK1)—the so-called PDK1-interacting fragment (PIF) pocket. PDK1 is a member of the AGC family of Ser/Thr kinases, which includes protein kinase A (PKA), B (PKB/Akt), and C (PKC) isozymes. In PKA, PKB, and PKC the site analogous to the PIF pocket binds to a specific hydrophobic motif (HM) at the C terminus of these kinases (7). In PDK1, which lacks its own HM segment, the PIF pocket serves a dual purpose: It recruits downstream substrate kinases that possess an HM segment (e.g., PKC, S6K, RSK, and others) and, upon HM binding, allosterically stimulates PDK1 activity (8–11). Short peptides (12–25 residues) derived from the HM of several kinases (PIFtides) can activate PDK1 by about four- to sevenfold, although their mode of interaction with PDK1 has not yet been established (8, 9). Recently, small molecules that mimic PIFtides and similarly activate PDK1 by binding the PIF pocket have been discovered and characterized structurally and biochemically (12).

We used disulfide trapping to (i) better elucidate how a PIFtide binds to and activates PDK1, (ii) identify new small molecule fragments targeting the PIF pocket, and (iii) begin to characterize their allosteric mechanism. Most remarkably, some of the disulfide-trapped fragments activate and others inhibit the kinase. Biochemical and X-ray structural studies provided insight into the binding modes of allosteric modulators and protein conformational changes that may propagate their allosteric effects. These studies provide a framework for systematically assessing functional and structural effects of small molecules directed at specific allosteric sites on protein kinases.

Results

Trapping of Peptides at the PIF Pocket on PDK1.

Our first goal was to use disulfide trapping to better characterize the binding of PIFtides to PDK1. Because no crystal structures of PIFtide/PDK1 complexes have been reported, we have generated an energy-minimized homology model based upon the known structure of a chimeric Akt construct, in which the endogenous C-terminal HM was replaced by the HM of the kinase PRK2 (13) (Fig. S1A).

To test our model, we prepared a panel of recombinant PDK1 Cys mutants in which residues at six positions around the PIF pocket were mutated individually to a cysteine (K115C, I119C, V124C, R131C, T148C, and Q150C). All of these Cys mutants displayed comparable catalytic activity, but were somewhat less active than the wild-type enzyme (by approximately 20–50%; Fig. S2). We next synthesized a panel of 11 PIFtide variants derived from the HM of PRK2 in which a disulfide-capped cysteine replaced residues along the sequence (Fig. 1A). We allowed each of the six Cys-mutant kinases (at 1 μM) to form a disulfide adduct with each of the 11 Cys-PIFtides (at 100 μM). The concentration of β-mercaptoethanol (β-ME) was varied (100 μM or 1 mM) to adjust for differences in intrinsic reactivity toward disulfides of the introduced cysteines on PDK1, as gauged by promiscuity in binding to the library of Cys-PIFtides. The level of conjugation was determined by mass spectrometry.

Fig. 1.

Trapping of Cys-PIFtides to PIF pocket on PDK1. (A) Sequence of parent PIFtide 15-mer and Cys-scanned region. (B) Conjugation strength of each PDK1 Cys mutant to each Cys position in Cys-PIFtides, presented as a heat map on a homology model of the PIFtide/PDK1 complex.

The Cys-PIFtides showed pronounced selectivity among the six PDK1 Cys mutants (Fig. 1B). Furthermore, the extent of conjugation tracked well with the predicted proximity between the cysteines on the protein and the peptide according to our calculated model for the PIFtide/PDK1 complex. For example, PDK1^K115C conjugated extensively to the Cys at position 3 in PIFtide^Q3C, and the extent of conjugation to other PIFtides tapered off with distance from residue 115 in PDK1. Similar effects were seen with conjugation to Cys at four other positions: 119, 124, 131, and 148. One of the PDK1 mutants, PDK1^Q150C, was significantly more promiscuous in its reactivity although this mutant did show strongest conjugation to the residue on the PIFtide (D10C) nearest to it in our model.

We tested if the conjugation data could be used to further refine our initial homology model. We performed a conformational search using a conservative restraint of 7.0 Å on the Cβ–Cβ distances between PIFtide residues and residues in PDK1 that showed the greatest conjugation in disulfide-trapping experiments. We followed up with unrestrained molecular dynamics (MD) simulations with explicit aqueous solvent. The PIFtide remained bound to the pocket throughout the simulations and did not deviate greatly in conformation relative to the initial constrained model. By contrast, models derived from MD simulations on an initially unrestrained complex showed significantly greater heterogeneity and sampled conformations very different from that of the starting structure (Fig. S1 and Movie S1).

We next assessed the effects of disulfide-trapped PIFtides on the catalytic activity of their associated PDK1 Cys mutants (Fig. 2). These assays were performed with peptides and PDK1 mutants for which we could achieve stoichiometric conjugation at the lowest β-ME concentration tested (100 μM). Kinase activities (initial rates) were measured with a standard assay using a peptide substrate taken from the activation loop of Akt and radiolabeled ATP.

Fig. 2.

Modulation of PDK1 mutants by disulfide-bound PIFtides. (A) Effects of nearest-neighbor Cys–Cys cross-links on PDK1 activity. Values shown are fold catalytic activities (relative to a DMSO-treated control) of the indicated PDK1 mutant conjugated to various Cys-PIFtides. (B) Effects of nearby and distant cross-links on activity of PDK1^Q150C.

All of the PDK1 Cys mutants tested had at least one Cys-PIFtide that when conjugated increased kinase activity relative to unconjugated PDK1 (Fig. 2A). The catalytic activity of the most activated complex, PIFtide^D10C/PDK1^Q150C (5.2-fold activation), even exceeded that which we measured for the noncovalent complex between the parent PIFtide and wild-type PDK1 (3.9-fold activation at saturation). Interestingly, although several PIFtide disulfides could be conjugated stoichiometrically to PDK1^Q150C (Fig. 1B), PIFtide^D10C was clearly more effective than the others at activating this kinase (Fig. 2B). The average errors in activation or inhibition potency did not exceed 15% of the values shown (see Experimental Procedures).

Trapping of Small Molecules at the PIF Pocket.

We next evaluated whether small molecule disulfides would mimic effects seen for the larger Cys-PIFtides (molecular mass of approximately 2,000 Da). Thus, we screened each of the six PDK1 Cys mutants against a library of 480 disulfide-bearing fragments (molecular mass 200–500) that was synthesized in the Small Molecule Discovery Center at University of California, San Francisco. These compounds represent a diversity of scaffolds common in drugs and drug candidates (e.g., heterocycles, piperazines, etc.) and most obey the “Rule of 3,” which was established to judge fragments as suitable starting points for drug discovery efforts (mass < 300 Da, predicted cLogP ≤ 3, number of hydrogen-bond donors/acceptors ≤ 3) (14). Preliminary experiments revealed that the disulfide fragments reacted much more readily with the PDK1 Cys mutants than did the Cys-PIFtides; at 100 μM β-ME and 100 μM compound, most small molecule disulfides reacted quantitatively with every Cys mutant. Therefore, in screening disulfide fragments for binding, we increased the assay stringency by raising the concentration of β-ME to 10 mM.

Under our assay conditions, the hit rate for disulfide fragment binding varied from 1.3% to 13.8% for the six PDK1 Cys mutants (using 20% conjugation to define a hit). The least hit-rich PDK1 mutant, PDK1^I119C, trapped only six compounds, whereas the most hit-rich mutant, PDK1^K115C, trapped 66 compounds. Interestingly, the mutants that were the most promiscuous in binding to small molecules (K115C, R131C, and T148C) were the most selective in binding to Cys-PIFtides (Fig. 1B). About 3% of the disulfide fragments in the library (15 out of 480) were trapped to > 50% conjugation to one and only one of the six PDK1 Cys mutants. The wild-type PDK1 catalytic domain has four endogenous cysteines, which were retained in our mutant constructs. Nevertheless, none of the fragment disulfide hits conjugated to wild-type PDK1, indicating specific binding to the Cys incorporated at the PIF pocket.

Whereas Cys-PIFtides activated PDK1 mutants, the disulfide fragment hits comprised both activators and inhibitors, as well as several inert ligands that neither activated nor inhibited the enzyme (Fig. 3 A and B and Fig. S3). Every PDK1 Cys mutant produced all three outcomes that varied depending on the fragment bound. Some mutants (e.g., K115C and Q150C) seemed to trap inhibitors or activators preferentially, although the sample size underlying these trends is small. The mutants showing the greatest range in activity on stoichiometrically trapping small molecule disulfides were PDK1^R131C and PDK1^T148C where activity varied about 14-fold from the most inhibited to the most activated complex. The effects of compound concentration on percent conjugation (Fig. 3C) and kinase activity (Fig. 3D) were compared for the most activating and inhibiting fragments targeting PDK1^T148C—2A2 and 1F8, respectively. The concentrations required for 50% conjugation to this mutant (DR₅₀ values) were 5.6 μM for 2A2 and 4.6 μM for 1F8 (at 1 mM β-ME). These values tracked closely the concentrations required for half-maximal effect on catalytic activity (EC₅₀ values)—7.1 and 7.2 μM for 2A2 and 1F8, respectively. None of the most activating or inhibiting disulfide fragments had a significant impact on the activity of wild-type PDK1 and all were approximately equally effective toward mutants in the absence and presence of the detergent CHAPS, which is known to disperse small molecule aggregates that can lead to spurious enzyme activation or inhibition (15–17) (Fig. 3B and Fig. S4). Overall, these results indicate that specific disulfide trapping of compounds at the PIF pocket is critical for the functional effects observed.

Fig. 3.

Activation and inhibition of PDK1 by small molecule fragment disulfides. (A) Activity data (relative to DMSO control) for each PDK1 mutant conjugated to disulfide fragments. (B) Structures of PIF pocket activators 2A2 and JS30, inert ligand 1H9, and inhibitor 1F8 for PDK1^T148C and their effects on the activity of PDK1^T148C in CHAPS-containing buffer. (C and D) Dose-response curves measuring (C) conjugation or (D) kinase activity for the best activator, 2A2, and inhibitor, 1F8, against PDK1^T148C and comparison to the effect on wild-type PDK1. Error values calculated from duplicate measurements (see SI Text).

Optimizing a Disulfide Fragment Activator.

To assess relationships between structure and activity of disulfide activators, we synthesized and tested 30 analogs of 2A2 in which the N-chlorophenylpiperazine moiety was replaced with various N-substituted piperazines. Compounds were then assayed for binding to and activation of PDK1^T148C (Table S1). Only N-phenyl and N-benzyl piperazines conjugated detectably to this mutant, whereas similar N-acyl and N-benzoyl derivatives did not (at 100 μM compound, 10 mM β-ME). Substitution of the aryl ring within these derivatives significantly affected both conjugation strength and activation potency and these properties did not seem to be correlated. The best compound of the series (“JS30”; Fig. 3B) activated PDK1^T148C by 6.3-fold, representing a significant improvement over 2A2 (3.9-fold).

Biochemical Characterization of Allosteric PDK1 Modulators.

To explore the mechanism behind activation and inhibition of PDK1 by PIF pocket ligands, we first wanted to assess whether binding at the PIF pocket is coupled to substrate binding at the active site. The weak affinity of the Akt-derived substrate peptide for PDK1 (K_m approximately 10 mM) precluded measurement of Michaelis–Menten parameters (8). Thus, we chose to assess interplay between the ATP pocket and the PIF pocket using disulfide trapping. Dose-response curves were measured for binding of the best disulfide fragment activator (JS30) and the best inhibitor (1F8) to PDK1^T148C in the presence and in the absence of a saturating concentration of staurosporine (100 μM), a tightly binding, ATP-competitive inhibitor of PDK1 (Fig. 4A). These curves showed only minimal effects of staurosporine on the efficiency of disulfide-trapping for either JS30 or 1F8.

Fig. 4.

Biochemical characterization of the PDK1^T148C activator, JS30, and inhibitor, 1F8. (A) Disulfide trapping in the presence (filled symbols) or absence (open symbols) of staurosporine (STS). (B) Proteolytic stability assays for PDK1^T148C conjugated to JS30, 1F8, or the inert disulfide fragment 1H9.

We next probed whether the allosteric PDK1 modulators affected local or global protein stability. Park and Marqusee have shown that susceptibility of a protein/ligand complex to proteolysis by a nonspecific protease (e.g., thermolysin) is inversely proportional to the energetic stability of the complex (18). We compared thermolysin-catalyzed degradation of PDK1^T148C alone and bound stoichiometrically to one of our best allosteric small molecule modulators (JS30 or 1F8) or an inert allosteric ligand (1H9) as visualized by SDS-PAGE (Fig. 4B). All of these ligands stabilized PDK1^T148C to degradation; the half-life of the protein was 1.9 min in the absence of these compounds and 6.7–9.5 min in their presence, representing 3.5- to 5-fold stabilization. The half-life of wild-type PDK1 (2.5 min) was, by contrast, insignificantly affected by the disulfide fragments (Fig. S5).

We also performed hydrogen-deuterium (HD) exchange experiments to assess stabilization of PDK1^T148C bound to allosteric modulators (Fig. S6). Upon stabilization, one expects hydrogen-bonding interactions involving backbone amide protons to become stronger, limiting exchange with deuterium from D₂O (19). We exposed unliganded PDK1^T148C and PDK1^T148C bound stoichiometrically to JS30, 1F8, or 1H9 in H₂O to D₂O and monitored deuterium incorporation by mass spectrometry. The mass of the protein alone shifted upward by 76 Da upon exposure to D₂O for 1 min, whereas the mass shift was +53, +52, and +51 Da for PDK1^T148C bound to JS30, 1F8, or 1H9, respectively. Thus, 23–25 amides, or approximately 30% of the total number of exchangeable amides, were protected from exchange in the bound versus unbound protein.

Structural Characterization of Allosteric PDK1 Modulators.

To gain deeper insight into the binding mode and possible mechanism behind allosteric modulation of PDK1 by PIF pocket ligands, we solved high-resolution X-ray crystal structures of three complexes, two between PDK1^T148C and a disulfide fragment activator, 2A2 or JS30, and one between this mutant and a disulfide inhibitor, 1F8 (to 2.0, 2.1, and 2.2 Å resolution, respectively; Table S2). Both activated complexes were crystallized in the presence of the known ATP-competitive ligand bisindolylmaleimide-II (BIM-II), whereas the inhibited complex was crystallized in the absence of an ATP-site ligand.

Continuous electron density could be traced in all three disulfide-bound PDK1^T148C structures from Cys148 to the appended fragment, 2A2, JS30, or 1F8, allowing unambiguous determination of the position of these ligands in the PIF pocket (Fig. 5 A–C). All three fragments target the same general area in the PIF pocket of PDK1, filling a hydrophobic cavity lined by the side chains of residues Lys115, Ile118, and Ile119 on helix B; Val124, Val127, Thr128, and Arg131 on helix C; and Gln150, Leu155, and Phe157 on β-strand 4. However, the shape of this cavity differs in the three structures due primarily to movements of the B and C helices (see below).

Fig. 5.

Crystal structures of PDK1^T148C conjugated to allosteric disulfide fragments. Close-up views of the PIF pocket in complex with activators (A) 2A2 or (B) JS30 or (C) inhibitor 1F8, showing residues on the B helix (pink), C helix (yellow), and β-strand 4 (cyan) that form the interaction surface for these ligands. (D) Effects of small molecule modulator binding at the PIF pocket on positioning of the B and C helices (arrows), Tyr126, and other PDK1 active site residues. Structures shown in D are various mutants of PDK1 bound to PIF pocket ligands JS30 (pink), PS48 (orange) (12), 2A2 (yellow), nothing (i.e., no ligand; gray) (22), or 1F8 (dark/light blue).

In all three structures, PDK1^T148C adopts an active-like conformation overall, as characterized by specific interactions among conserved catalytic residues (Fig. 5D and Fig. S7). For example, salt-bridge interactions are formed between Glu130 on the C helix and Lys111, residues that are critical for positioning ATP for phosphoryl transfer. Salt-bridge interactions are also apparent between phosphorylated Ser241 (pS241) on the activation loop and two key Arg residues, Arg204 of the so-called HRD motif and Arg129 on the C helix, which are responsible for positioning the activation loop to accept a protein/peptide substrate (20).

Despite the similarities among the inhibited and activated PDK1^T148C structures, several differences are also apparent (Fig. 5D and Fig. S7). For example, the conformation of the activation segment (residues 223–252) varies. In the two activated structures, this segment is largely disordered and displays poor electron density between residues 232 and 240. A poorly-defined activation loop was also observed for one of the two conformers found in the asymmetric unit of the inhibited 1F8 complex, whereas in the other conformer, the activation segment could be almost completely refined, forming a short α-helical turn between residues 231 and 236. The positions of the B and C helices are also closer to the activation loop and ATP-binding site in the 2A2 and JS30 complexes relative to the inhibited 1F8 complex. Residues on the C helix adopt different conformations in the three structures. For example, in the most activated complex (with JS30) the Tyr126 hydroxyl makes putative H-bonding interactions with the side chains of Asp223 (in the DFG motif) and Glu130 and with the backbone amide of Gly225 (in the DFG motif). In the 2A2 complex, the Tyr126 hydroxyl is positioned slightly further away from these residues and makes a hydrogen bond to a water molecule that instead makes the above contacts with the DFG motif and Glu130. In the inhibited 1F8 complex, the Tyr126 side chain adopts two conformations in the crystal, one in the direction of the active site and one forming an H-bonding interaction with pSer241, but both too far from the DFG motif or Glu130 to form contacts with either.

We mutated Tyr126 in PDK1^T148C to Phe to investigate the role of this residue in activation/inhibition by PIF pocket disulfide compounds (Fig. S8). The resulting Y126F/T148C double mutant had approximately twofold lower kinase activity relative to the T148C single mutant. The double-mutant was less well activated by JS30 than was PDK1^T148C (5.2- vs. 6.0-fold activation, respectively), but was inhibited slightly better by 1F8 relative to the single mutant (3.8-fold vs. 3.1-fold inhibition).

Discussion

Unlike conventional screening approaches, disulfide trapping is well suited for hypothesis-driven ligand discovery efforts of the type described here in which one asks whether targeting a specific protein site with small molecules or peptides can produce a desired functional and/or structural effect. Here, we have used the technology to gain a better understanding of how PIF peptides interact with and allosterically activate PDK1 and to probe the PIF site with small molecule fragments.

Molecular modeling simulations provided a structural hypothesis for how PIFtides bind to PDK1 and we subsequently used disulfide trapping in solution to test this hypothesis. Taken together, our conjugation data strongly support a binding orientation of the PRK2-derived PIFtide in the PIF pocket on PDK1 that is similar to that adopted by hydrophobic motifs of AGC kinases bound intramolecularly to their own analogous (HM) pocket (7). Comparison of the catalytic activities of different cross-linked peptide/PDK1 pairs showed that the extent of PDK1 activation is sensitive to PIFtide positioning. For example, several Cys-PIFtide/PDK1 mutant cross-links could be formed between residues that were nearby in our model, but these were nonequivalent in terms of PDK1 activation. That the most activated complex, PDK1^Q150C/PIFtide^D10C, was more active than the corresponding wild-type (noncovalent) PIFtide/PDK1 complex suggests that the full activation potential of wild-type PDK1 has not been realized in experiments with isolated PIFtides. It is possible that the entire PIF-containing protein (i.e., PRK2) is required for inducing maximal activation of wild-type PDK1.

Screening of a small library of 480 disulfide fragments yielded a number of ligands with diverse structures and functional impacts on PDK1. Each Cys site on PDK1 trapped a distinct panel of disulfide fragments that were selective for each mutant over other mutants or the wild-type enzyme, suggesting specific binding modes in each case. The best disulfide fragment activators from our initial screen were nearly as potent as the best Cys-PIFtide activators; chemical analogs produced even more potent fragment activators (reaching 6.3-fold activation).

Perhaps most surprising was the fact that the fragments, unlike PIFtides, included both PDK1 activators and inhibitors. Both binding and extent of activation or inhibition were highly dependent on chemical structure of the bound fragment and positioning of the fragment in the PIF pocket. The activity of PDK1^T148C in complex with small fragments was variable over the largest range (approximately 19-fold), from the most inhibited (in complex with 1F8) to the most activated (in complex with JS30). Although the fragments are only about 1/10th the size of PIFtides, their effects on PDK1 activity were more extreme (i.e., greater activation and inhibition). It is possible that the activation of PDK1 observed by PIFtides represents the net sum of interactions between a number of subsites, some of which can be activating and some inhibiting. Perhaps the small fragments can occupy individual subsites and thereby generate more potent activating or inhibiting effects on PDK1. Previous disulfide-trapping experiments on the C5a receptor, a G-protein coupled receptor (GPCR), produced both agonists and antagonists at the same site (4, 5). It is intriguing that a single allosteric site on a kinase, the PIF pocket on PDK1, can similarly generate a continuum of agonistic and antagonistic effects, by analogy to single ligand binding sites in GPCRs.

Biochemical data and high-resolution crystal structures for activating and inhibiting fragments bound stoichiometrically to PDK1^T148C showed that these ligands stabilize the protein globally and induce conformational changes that may account for their effects on activity. In crystallizing our most strongly inhibited complex, 1F8/T148C, in the absence of an ATP-site ligand and the most strongly activated complex, JS30/T148C, in the presence of an ATP-site ligand (BIM-II), we expected to observe widely divergent PDK1 conformations. Surprisingly, the structures of the 1F8/PDK1^T148C and JS30/PDK1^T148C complexes were quite similar, with all key catalytic residues (e.g., Lys111, Glu130, and Asp223 in the ATP-binding cleft; Arg204 of the HRD motif; and Arg129 on the C helix) poised in canonically “active” conformations (20). However, clear changes were observed in the conformation of the activation loop and the positions of the B and C helices, which move approximately 5 Å away from the ATP-binding site in the most inhibited versus the most activated complex (Movie S2). Additionally, in both activated complexes, the side chain of Tyr126 in the C helix is directed inward toward the ATP-binding site. In complex with JS30, the Tyr126 hydroxyl on PDK1^T148C forms a hydrogen bond with the side chains of Asp223 in the DFG motif and Glu130 on the C helix, residues that are strictly conserved and critically important in all kinases for positioning ATP during catalysis (21). This type of interaction has not, to our knowledge, been observed in other kinase structures. In all PDK1 structures solved to date, Tyr126 is positioned outward, forming an H bond to pSer241. In PKA and PKB, the residue corresponding to Tyr126 in PDK1 is a His residue, which forms an electrostatic interaction with an analogous phosphorylated residue on the activation loop. It is possible that the interaction of Tyr126 with the DFG Asp and/or Glu130 in the JS30/PDK1^T148C complex plays a role in allosteric activation. Indeed, mutation of Tyr126 to Phe had a modest, but significant negative impact on activation by JS30. Interestingly, this mutation also makes inhibition of PDK1^T148C by 1F8 slightly more effective, suggesting that Tyr126 may also destabilize the inhibited state of the kinase. Prior to phosphoryl transfer in protein kinases, a Mg²⁺ ion bridges the Asp of the DFG motif and the gamma phosphate group of bound ATP (20). It seems unlikely, therefore, that Tyr126 in PDK1 would be able to form a close interaction with the DFG Asp of PDK1 in the Mg²⁺/ATP-bound state. However, it is possible that Tyr126 assists in release of ADP following phosphoryl transfer, a step which is rate limiting for many kinases.

Comparisons among the biochemical and structural data reported here for disulfide-bound modulators of PDK1 and elsewhere for PDK1, alone and in complex with a known small molecule activator that binds to the PIF pocket, PS48 (12), provide additional clues regarding the mechanism(s) behind allosteric modulation of this kinase. For example, PS48 was reported to stabilize PDK1, but was paradoxically competitive with ATP binding (12). We also observe kinase stabilization by PIF pocket-directed disulfide fragments, but do not observe significant communication with the ATP site by PIF pocket activators, suggesting that this feature is not necessary in all cases for activation. Our disulfide fragments also bind to the PIF pocket in slightly different ways relative to PS48. Like PIFtides, PS48 possesses two aromatic groups that become buried in the hydrophobic PIF pocket and a carboxylate that forms an electrostatic interaction with Arg131 on the C helix (12). Only the hydrophobic components are reproduced in the disulfide activators presented here, yet activation is observed. The carboxylate/Arg131 contact between PIFtides/PS48 and PDK1 may therefore be important for binding to the PIF pocket, but may not be important for activation. Finally, a comparison of the structures here of PDK1^T148C bound to inhibitory and activating disulfide fragments and those previously reported of PDK1 absent a PIF pocket ligand (22) and in complex with PS48 (12) reveal a compelling correlation between C-helix position and kinase activity (Fig. 5D and Movie S2). The positioning of the C helix relative to the ATP site for PDK1 bound to various PIF pocket ligands follows the trend, from nearest to farthest, JS30, PS48, 2A2, none (i.e., bound only to an ATP-site ligand), and 1F8, which tracks well with the activities of the corresponding complexes relative to the unliganded proteins: 630%, 400% (12), 390%, 100%, and 32%, respectively. This correlation supports previous hypotheses implicating the C helix in allosteric activation of PDK1 (12), but shows additionally that the C helix may also be positioned to inhibit the kinase. We cannot, however, discount possible effects of crystal packing differences in the various structures as causes for the structural changes we describe.

Many kinases possess a functional allosteric pocket analogous to the PIF pocket on PDK1 that should be targetable using the approach we have described. Within the AGC kinase family, PKA, PKB (Akt), PKC isozymes, RSK, R6K, SGK, PRK1/2, and GRK all have a C-terminal HM segment that docks to a pocket analogous to the PIF pocket and cis activates these kinases (7). In Src, a nonreceptor tyrosine kinase, the analogous pocket is occupied by residues on the SH2-kinase domain linker N-terminal to the catalytic domain, which stabilize an inactive conformation (23). Other kinases are regulated by interactions of separate protein domains with a PIF-like pocket (e.g., TPX2/Aurora A, cyclin/CDKs, and the EGFR/EGFR dimer) (24–26). Based on our success with PDK1, we expect that disulfide trapping will allow us to study and validate such sites in an efficient and systematic fashion while simultaneously providing chemical starting points for developing bioactive kinase modulators.

Experimental Procedures

Peptide Synthesis.

Peptides were synthesized using standard procedures (see SI Text).

Fragment Disulfide Synthesis.

Two equivalents of a commercially available amine or carboxylic acid (Sigma-Aldrich) were reacted with one equivalent of either a succinimidyl ester form of the appropriate carboxylic acid (4,4′-dithiodibutyric acid or 3,3′-dithiopropionic acid) or cystamine and diisopropylcarbodiimide, respectively, in a mixture of dimethylformamide and diisopropylethylamine (∼10∶1) to generate a disulfide-linked dimer. The dimer was then reacted with either cystamine or N,N,N′,N′-tetramethylcystamine in the presence of a catalytic amount of cysteamine to give the final asymmetric disulfide-capped fragment, which was purified by HPLC. The identities and purities of all reaction products were verified by liquid chromatography (LC)/MS.

Plasmid Construction and Protein Preparation.

The catalytic domain of PDK1 (residues 51–359) was cloned into a pFastBacHTa vector and expressed in Sf21 insect cells. All mutants were generated by Quikchange mutagenesis (Stratagene) using standard protocols (see SI Text).

Disulfide-Trapping Conjugation Assays.

Immediately prior to disulfide-trapping experiments, PDK1 was exchanged into 25 mM Tris pH 7.5 using a Nap-5 size-exclusion column (GE Amersham) and fresh β-ME (sealed vials from Thermo Scientific) was added to the desired concentration (see Results). Proteins were incubated with disulfide compounds in 96-well plates at room temperature for 1 h prior to mass spectrometric analysis using a LCT-Premier LC/electrospray ionization-MS instrument (Waters). Protein masses were deconvoluted using the Max-Ent algorithm within the MassLynx software. Conjugation strengths were measured by comparing peak areas for the conjugated versus unconjugated protein.

Kinase Activity Assays.

A radioactivity-based kinase assay using as substrates [γ-³²P]-ATP and a peptide derived from the activation loop of Akt (KTFAGTPEYLAPEVRR) was used to monitor effects of disulfide compounds on PDK1 activity (see SI Text).

Crystallography.

Stoichiometric disulfide/PDK1 conjugates were prepared immediately prior to crystallization by incubating partially purified protein with an excess of disulfide compound in the presence of β-ME until conjugation was complete (as indicated by LC/MS). The conjugates were purified by size-exclusion chromatography and concentrated to 8.4, 8.4, or 16 mg/mL for the 2A2, JS30, and 1F8 complexes, respectively. To the JS30/T148C and 2A2/T148C complexes were added BIM-II at an approximate twofold molar excess. All protein complexes were crystallized by hanging drop vapor diffusion by mixing the protein with reservoir solution. Orange, hexagonal rod-shaped crystals of PDK1^T148C/2A2/BIM-II and PDK1^T148C/JS30/BIM-II were obtained at 20 °C using the following reservoir solutions: for 2A2, 0.1 M sodium citrate pH 5.4, 0.1 M potassium sodium tartrate, 1.8 M ammonium sulfate; and for JS30, 0.1 M sodium citrate pH 5.6, 0.2 M NDSB-211, 1.7 M ammonium sulfate. Square rod-shaped crystals of PDK1^T148C/1F8 were obtained at 4 °C using the following reservoir solution: 0.1 M Bis-Tris pH 6.5, 0.2 M lithium sulfate, 25% (wt/vol) PEG 3350. Crystals were immersed briefly in cryoprotectant—for 2A2 and 1F8 complexes, a mixture of Paratone-N and paraffin oil, or for JS30, 0.1 M sodium citrate pH 5.6, 1.7 M ammonium sulfate, 0.2 M NDSB-211, 20% glycerol—and flash-frozen in liquid nitrogen. Diffraction data were collected at the Advanced Light Source, Berkeley, California (Beamline 8.3.1) and scaled with HKL2000. All structures were solved by molecular replacement (in CCP4) using, as a search model, the ATP-bound structure of PDK1 (Protein Data Bank ID 1H1W) (27). Structures were subsequently refined iteratively using WinCoot (28) and Phenix (29) (Table S2).