Decision letter: A dynamic mechanism for allosteric activation of Aurora kinase A by activation loop phosphorylation

Abstract

Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods References Decision letter Author response Article and author information Metrics Abstract Many eukaryotic protein kinases are activated by phosphorylation on a specific conserved residue in the regulatory activation loop, a post-translational modification thought to stabilize the active DFG-In state of the catalytic domain. Here we use a battery of spectroscopic methods that track different catalytic elements of the kinase domain to show that the ~100 fold activation of the mitotic kinase Aurora A (AurA) by phosphorylation occurs without a population shift from the DFG-Out to the DFG-In state, and that the activation loop of the activated kinase remains highly dynamic. Instead, molecular dynamics simulations and electron paramagnetic resonance experiments show that phosphorylation triggers a switch within the DFG-In subpopulation from an autoinhibited DFG-In substate to an active DFG-In substate, leading to catalytic activation. This mechanism raises new questions about the functional role of the DFG-Out state in protein kinases. https://doi.org/10.7554/eLife.32766.001 eLife digest The transfer of phosphate groups onto proteins (protein phosphorylation) is one of the most important methods used to send signals inside cells. The enzymes that catalyze this process, called protein kinases, are themselves controlled by the phosphorylation of a flexible region called the activation loop. For many years it had been thought that the purpose of activation loop phosphorylation was to clamp the otherwise flexible activation loop in an active state that allows molecules that need to be phosphorylated to bind to the kinase. This assumption was based on static pictures of protein kinases obtained by X-ray crystallography, in which individual states are trapped and visualized in a crystal lattice. However, new methods and approaches now mean it is possible to visualize how the position of the activation loop changes as it moves in solution. By applying these techniques, Ruff et al. show that the static model is incorrect in a protein kinase called Aurora A. In this enzyme, the phosphorylated activation loop continues to switch back and forth between active and inactive states. Phosphorylation instead enhances the catalytic activity of the active state. Aurora A regulates several important steps in cell division, and plays important roles in several kinds of cancer. The discovery that activated forms of Aurora A can have different dynamic properties raises the possibility that inhibitor molecules could be designed to exploit these differences and block specific activities of Aurora A in cancer cells. To realize this goal we need to better understand how a kinase switching between active and inactive states affects the ability of inhibitors to interact with it. https://doi.org/10.7554/eLife.32766.002 Introduction Stringent regulatory control of protein kinases is critically important for the integrity of cellular signal transduction. The catalytic activity of protein kinases is regulated by finely-tuned allosteric mechanisms that reversibly switch the kinase domain between active and inactive conformational states (Huse and Kuriyan, 2002). Disruption of these mechanisms, leading to constitutive kinase activity, is a major cause of human cancer, and small molecules that inhibit specific disease-associated kinases are an increasingly important component of many modern cancer therapies (Zhang et al., 2009). Phosphorylation on a specific site in the activation loop of the kinase domain is the most widely conserved regulatory mechanism in kinases (Johnson et al., 1996). X-ray structures show that ionic interactions between the phosphate moiety and a surrounding pocket of basic residues stabilize the activation loop in a conserved active conformation (Knighton et al., 1991; Yamaguchi and Hendrickson, 1996; Steichen et al., 2012). In this active state, a catalytic asp-phe-gly (DFG) motif at the N-terminal end of the activation loop adopts an active ‘DFG-In’ conformation, with the aspartate residue of the DFG motif pointing into the active site to coordinate Mg-ATP, and the C-terminal segment of the activation loop positioned to bind peptide substrates. The DFG-In state is stabilized by the assembly of a network of hydrophobic residues, termed the regulatory spine, that lock together the N-terminal lobe, the αC-helix, the DFG motif and the C-terminal lobe of the kinase (Kornev et al., 2006; Kornev et al., 2008). In the absence of phosphorylation on the activation loop, kinase activity is usually restrained by rearrangements of the activation loop and DFG motif into specific autoinhibited conformations. An important autoinhibited state, called ‘DFG-Out’, involves a flip in the backbone torsion angles of the DFG motif, reorienting the DFG aspartate out of the active site to prevent magnesium coordination, repositioning the activation loop to block peptide binding, and disassembling the regulatory spine (Hubbard et al., 1994; Nagar et al., 2003; Mol et al., 2004). This conformational change dramatically alters the chemical makeup of the active site, and selective recognition of the DFG-Out state underlies the mechanism of action of many small-molecule kinase inhibitors (Liu and Gray, 2006). The serine/threonine kinase Aurora A (AurA) is an essential mitotic protein that controls a variety of cellular processes including mitotic spindle assembly, centrosome maturation, and mitotic entry (Glover et al., 1995; Hannak et al., 2001; Berdnik and Knoblich, 2002; Macůrek et al., 2008; Seki et al., 2008). These functions of AurA are driven by two distinct activation mechanisms of the kinase operating in different spatiotemporal contexts. At the centrosome, AurA must first be activated by autophosphorylation on the activation loop threonine residue T288 in order to carry out its centrosomal functions. At the mitotic spindle, AurA is instead activated by binding to the spindle assembly factor Tpx2 (Kufer et al., 2002). This spindle-associated pool of AurA must be maintained in the unphosphorylated state by the phosphatase PP6 in order for spindle assembly to proceed faithfully (Zeng et al., 2010; Toya et al., 2011). Extensive in vitro studies have confirmed that Tpx2 and phosphorylation can act independently to increase AurA kinase activity by up to several hundred-fold (Zorba et al., 2014; Dodson and Bayliss, 2012). We recently showed that activation of AurA by Tpx2 is driven by a population shift from a DFG-Out to the DFG-In state (Cyphers et al., 2017). Since crystal structures of phosphorylated AurA bound to Tpx2 show the T288 phosphothreonine residue forming the canonical ionic interactions thought to stabilize the DFG-In state (Bayliss et al., 2003; Zhao et al., 2008; Clark et al., 2009), it has been assumed that phosphorylation also triggers a transition from the DFG-Out to the DFG-In state. In this paper, we show that phosphorylation on T288 in fact activates AurA through a completely different mechanism than Tpx2. Three complementary spectroscopic methods, infrared spectroscopy, Förster resonance energy transfer, and double electron-electron resonance, all show that phosphorylation does not trigger a switch to the DFG-In state, and that phosphorylated AurA continually samples both DFG-In and DFG-Out conformational states. Instead, phosphorylation triggers a conformational switch from a previously unknown inactive DFG-In substate to a fully activated DFG-In substate, enhancing the catalytic activity of the DFG-In subpopulation within a dynamic conformational ensemble. Results Phosphorylation of AurA on T288 does not switch the kinase into the DFG-In state We set out to explain how phosphorylation of AurA on T288 leads to a ~100 fold increase in catalytic activity (Figure 1—figure supplement 1a) (Zorba et al., 2014; Dodson and Bayliss, 2012). We previously used an infrared (IR) probe that tracks the DFG motif of AurA to show that Tpx2 binding triggers a conformational change from the DFG-Out to the DFG-In state (Cyphers et al., 2017), resulting in the assembly of the active site and the regulatory spine (Figure 1—figure supplement 2). In this method, a cysteine residue is introduced at position Q185 at the back of the active site of AurA, and chemical labeling is used to introduce a nitrile infrared probe at this position (Fafarman et al., 2006). To test whether phosphorylation of AurA also causes a conformational shift of the DFG motif, we prepared samples of AurA Q185C phosphorylated on T288. Homogeneous phosphorylation and nitrile labeling were verified by western blotting and mass spectrometry (Figure 1—figure supplement 1). IR spectra of nitrile-labeled phosphorylated AurA showed predominantly a single absorbance band centered at 2158 cm−1 (Figure 1a, solid black line). We previously assigned this peak in IR spectra of unphosphorylated AurA to the DFG-Out form of the kinase, in which the nitrile probe is buried in a hydrophobic pocket (Figure 1b, lower panel) (Cyphers et al., 2017). Addition of saturating amounts of Tpx2 peptide (residues 1–43 of human Tpx2) to the IR samples caused a dramatic spectral change wherein the central peak at 2158 cm−1 is largely replaced by two new peaks at 2149 cm−1 and 2164 cm−1 (Figure 1a, dashed black line). These changes are indicative of a shift to the DFG-In state, in which water molecules coordinated to the DFG motif form hydrogen bonds to the nitrile probe, causing pronounced spectral shifts (Figure 1b, upper right panel) (Cyphers et al., 2017). To confirm that the peak at 2158 cm−1 arises from the DFG-Out state, we mutated residue W277, which is positioned directly against the IR probe in the DFG-Out state, but is displaced away from it in the DFG-in state, to alanine (Figure 1b). IR spectra of the W277A mutant showed a clear spectral shift of the 2158 cm−1 peak (Figure 1c), consistent with this peak arising from the DFG-Out state. Figure 1 with 2 supplements see all Download asset Open asset Phosphorylation on T288 does not switch AurA into the DFG-In state. (a) IR spectra of nitrile-labeled phosphorylated AurA. The apo sample (solid black line), and the sample bound to Tpx2 (dashed black line), were measured at 5°C, and the kinase bound to ADP (colored lines) was measured at the indicated temperatures. Arrows indicate peaks assigned to the DFG-In and DFG-Out states. The inset shows the same experiments performed with unphosphorylated AurA. Single representative spectra are shown, normalized to peak maxima. (b) Overview of the structure of AurA in the active conformation bound to ADP (yellow) and Tpx2 (beige), with enlarged views of the DFG-In (right, PDB ID: 1OL5) and DFG-Out (bottom, PDB ID: 5L8K) states with the nitrile probe (Q185CN) modeled into the structures. (c) Second derivatives of IR spectra of apo WT and W277A AurA, showing the ~1.5 cm−1 spectral shift of the 2158 cm−1 peak (arrow). https://doi.org/10.7554/eLife.32766.003 The addition of ADP to apo AurA resulted in the appearance of a DFG-In subpopulation, apparent in the IR spectra as small shoulders on either side of the main 2158 cm−1 peak. Experiments performed over a range of temperatures showed that this DFG-In subpopulation increases at higher temperature (Figure 1a, colored lines), but does not reach the level observed in the presence of Tpx2. A similar DFG-In subpopulation was also detected in unphosphorylated AurA bound to ADP (Cyphers et al., 2017) (Figure 1a, inset), highlighting that although nucleotide binding shifts the DFG equilibrium towards the DFG-In state, phosphorylation does not seem to enhance this effect. These IR results suggested that phosphorylation alone does not substantially change the DFG-In/Out equilibrium of AurA, unlike Tpx2 binding. However, as replacement of the Q185 residue with the nitrile probe was found to alter the activation properties of AurA (Cyphers et al., 2017), it was necessary to confirm this interpretation using an alternative method. The phosphorylated activation loop adopts a range of conformations in solution, and only becomes highly ordered upon Tpx2 binding We used intramolecular FRET to track movements of the activation loop of AurA with and without phosphorylation on T288, using a construct with a native Q185 residue (Cyphers et al., 2017). Donor (D) and acceptor (A) fluorophores (Alexa 488 and Alexa 568, respectively) were incorporated on the activation loop (S284C) and αD helix (L225C) using maleimide chemistry (Cyphers et al., 2017). These labeling positions were chosen to track the movement of the activation loop across the active-site cleft as the kinase switches from the DFG-Out to the DFG-In state, with the dyes predicted to be further apart in the DFG-In state (Figure 2a). Phosphorylation of the protein on T288 was confirmed by tryptic mass spectrometry (Figure 2—figure supplement 1), and the labeled phosphorylated sample exhibited robust catalytic activity in the absence of Tpx2, and was further activated ~4 fold by the addition of Tpx2 (Figure 2—figure supplement 1c) (Zorba et al., 2014; Dodson and Bayliss, 2012). Figure 2 with 2 supplements see all Download asset Open asset The phosphorylated activation loop remains flexible and shifts to a more active conformation upon Tpx2 binding. (a) (left) Schematics showing the labeling scheme used to detect the DFG-In/Out transition by FRET. (right) Structures of the DFG-Out (top) and DFG-In (bottom) states of AurA, highlighting the β-sheet hydrogen bonds constraining the N- and C-terminal segments of the activation loop. The S284C labeling site is shown as a sphere. (b) (top) Time-resolved fluorescence waveforms for D-only (dashed lines) and D + A (solid lines) phosphorylated AurA in the presence and absence of 125 μM Tpx2 and 1 mM AMPPNP. Data are for a single representative experiment, normalized to the fluorescence peak. (bottom) Comparison of single-Gaussian distance distribution fits to fluorescence lifetime data obtained with unphosphorylated (left) and phosphorylated AurA (right). (c) Binding constants of ADP (top panels) and Tpx2 (bottom panels) for phosphorylated (blue) and unphosphorylated (red) AurA determined with and without the other ligand pre-bound to the kinase. Data represent mean values ± s.d.; n = 3. https://doi.org/10.7554/eLife.32766.006 Steady-state fluorescence emission spectra were measured for D- and D + A labeled forms of both unphosphorylated and phosphorylated AurA. In either phosphorylation state, titrating ADP or Tpx2 onto the kinase resulted in enhanced fluorescence emission from the donor dye and reduced emission from the acceptor, indicating a decrease in FRET efficiency (Figure 2—figure supplement 2) consistent with a shift towards the DFG-In state. To gain more insight into the conformation of the activation loop and how it is altered by phosphorylation and ligand binding, we performed time-resolved (TR) FRET experiments to quantify energy transfer through its effect on the fluorescence lifetime of the donor dye. TR fluorescence decays were recorded using time-correlated single-photon counting (TCSPC) (Figure 2b, top panel), and were then fit to a structural model consisting of a Gaussian distribution of inter-fluorophore distances (Muretta et al., 2013; Agafonov et al., 2009; Nesmelov et al., 2011) to represent the ensemble of conformations sampled in solution (Figure 2b, bottom panels). The distance distributions measured for the phosphorylated and unphosphorylated kinase in the absence of ligands are strikingly similar (Figure 2b, bottom panels). In both cases, a broad distribution centered around ~30 angstroms is observed for apo AurA, indicating that the activation loop is highly flexible regardless of the phosphorylation state (Figure 2b, black). This broad distribution is consistent with the DFG-Out state, in which the C-terminal half of the activation loop lacks contacts with the rest of the kinase domain, and is typically disordered in X-ray structures (Wu et al., 2013; Coumar et al., 2009; Fancelli et al., 2006) (Figure 2a, top panel). In contrast, when the phosphorylated and unphosphorylated samples were saturated with both Tpx2 peptide and nucleotide (either ADP or the non-hydrolysable ATP-analog AMPPNP), narrow distributions were observed that were shifted to ~54 angstroms (Figure 2b, blue). This indicates adoption of a well-defined structure consistent with the DFG-In state, in which the segment of the loop containing the labeling site is anchored to the C-terminal lobe of the kinase by flanking β-sheet interactions (Bayliss et al., 2003) (Figure 2a, bottom panel). In the presence of ADP or AMPPNP alone the observed distance distributions were intermediate in both distance and width between the other samples, consistent with nucleotide binding driving unphosphorylated and phosphorylated AurA into a similar equilibrium between DFG-Out and DFG-In states (Figure 2b, red), as was observed in the IR experiments. Phosphorylation and Tpx2 have synergistic effects on AurA conformation and nucleotide binding We used steady-state fluorescence to measure the equilibrium dissociation constants of ADP and Tpx2 for unphosphorylated and phosphorylated AurA (Figure 2c). Importantly, ADP bound to phosphorylated and unphosphorylated AurA with similar affinities (Figure 2c, top left panel), indicating that the interaction of the kinase with nucleotide is not substantially affected by phosphorylation on the activation loop. This is consistent with our IR and TR-FRET experiments, which show that phosphorylation by itself fails to trigger the long-range conformational change presumably required to couple the phosphorylation site on the activation loop to the distant ATP-binding site. In contrast, Tpx2, which does trigger a conformational change from the DFG-Out to the DFG-In state in both unphosphorylated and phosphorylated AurA, also enhances the binding affinity of nucleotides in both cases (Figure 2c, compare top panels). While phosphorylation does not affect nucleotide binding to apo AurA, we found that it does substantially enhance the binding of Tpx2 to the exterior surface of the kinase, increasing the affinity by a factor of ~20 (Figure 2c, bottom panels). This is remarkable considering that phosphorylation does not appear to stabilize the DFG-In state, and that there are no direct contacts between the T288 residue and Tpx2 (Bayliss et al., 2003). In addition, once Tpx2 is bound, phosphorylation does lead to an enhancement of nucleotide affinity (Figure 2c, top right panel, compare red and blue), indicating that allosteric coupling between the phosphorylation site and the active site, missing in apo AurA, is established in the AurA:Tpx2 complex. These trends in the affinity data are in good agreement with previous enzyme kinetics measurements (Dodson and Bayliss, 2012). Interestingly, the synergy observed between Tpx2 and phosphorylation is also reflected in our TR-FRET experiments (Figure 2b). A comparison between the unphosphorylated and phosphorylated samples bound to Tpx2 shows that while the unphosphorylated sample requires nucleotide to fully shift to the active state, Tpx2 alone is sufficient to achieve this in phosphorylated AurA, and the further addition of nucleotide has little effect (Figure 2b, compare yellow and blue). The same trend was observed in steady-state FRET experiments (Figure 2—figure supplement 2c, double-headed arrows). Together these data suggest a model in which the allosteric effects of phosphorylation are somehow masked in apo AurA, and only become apparent when Tpx2 switches the kinase to the DFG-In state, at which point phosphorylation further stabilizes this state. Phosphorylation promotes a single functional conformation in the DFG-In state While our results reveal synergy between phosphorylation and Tpx2, they do not answer the key question of how phosphorylation itself activates AurA. Indeed, the IR and FRET data clearly show that phosphorylation on T288 by itself does not cause a substantial shift towards the DFG-In state, and that the phosphorylated kinase, like the unphosphorylated enzyme, instead samples a range of different conformations spanning the DFG-In and DFG-Out states. We hypothesized that phosphorylation must instead drive catalytic activation of AurA by altering the structure and dynamics of the DFG-In subpopulation, presumably allowing it to populate catalytically competent geometries. To provide insight into how phosphorylation alters the structure and dynamics of the DFG-In state, we performed molecular dynamics simulations of the wild-type kinase. Simulations were initiated from the X-ray structure of DFG-In AurA bound to ADP and Tpx2 (PDB ID: 1OL5) (Bayliss et al., 2003), and were run in the presence and absence of Tpx2 and with and without phosphorylation on T288. For each of these four biochemical states, 250 trajectories up to 500 nanoseconds in length were obtained on the distributed computing platform Folding@home, for a total of over 100 microseconds of aggregate simulation time for each biochemical state. Analysis of the DFG conformation revealed that the simulations remained predominantly in their initial DFG-In state (Figure 3—figure supplement 1), suggesting that the simulation time was insufficient to capture the slow conformational change to the DFG-Out state. The simulations can thus be regarded as probing the conformational dynamics of the DFG-In kinase. The T288 phosphorylation site lies in the C-terminal segment of the activation loop, the correct positioning of which is essential for the binding of peptide substrates (Figure 3a). In the crystal structure used to initiate the simulations, this segment of the loop appears to be stabilized by interactions between the pT288-phosphate moiety and three arginine residues: R180 from the αC helix, R286 from the activation loop, and the highly conserved R255 from the catalytic loop ‘HRD motif’ (Figure 3a) (Bayliss et al., 2003). To probe the integrity of these interactions in the simulations, and to investigate loop dynamics in their absence, we examined the distribution of distances between the Cζ atoms of either R180 or R255 and the Cα atoms of T288 following equilibration within the DFG-In state (Figure 3—figure supplement 1b). We also tracked the distance between the L225 and S284 Cα atoms (the sites used for incorporating spectroscopic probes) to capture movements of the activation loop along a roughly orthogonal axis across the active site cleft. Figure 3 with 2 supplements see all Download asset Open asset Molecular dynamics simulations of AurA show that phosphorylation disfavors an autoinhibited DFG-In substate and promotes a fully-activated configuration of the activation loop. (a) Structure of active, phosphorylated AurA bound to Tpx2 and ADP (PDB ID: 1OL5) showing the interactions between pT288 and the surrounding arginine residues. The S284 and L225 Cα atoms are shown as black spheres. (b) Contour plots showing the L225 Cα - S284 Cα distances plotted against the T288 Cα - R255 Cζ distances for all four biochemical conditions. The active and autoinhibited DFG-In states observed for the unphosphorylated kinase in the absence of Tpx2 (red), and the shift in the L225-S284 distance between them, are indicated. (c) Simulation snapshot showing the helical turn in the activation loop and the position of the T288 sidechain at the C-terminal end of the helix. (d) The L225 - S284 distance is plotted against the dihedral angle defined by the Cα atoms of residues 283–286 (pseudodihedral). The helical conformation in the autoinhibited state is indicated. (e) Kinase activity (shown as ATP turnover per second) for phosphorylated WT (blue) and phosphorylated R180A (purple) AurA unlabeled FRET constructs in the presence and absence of 10 μM Tpx2. The decrease in the activity in the absence of Tpx2 due to the R180A mutation is highlighted by the arrow. Data represent mean values ± s.d.; n = 3. https://doi.org/10.7554/eLife.32766.009 We plotted the simulated L225-S284 distances as a function of the R255-T288 distance to assess how Tpx2 and phosphorylation affect the conformation of the activation loop (Figure 3b). As expected, the simulations of unphosphorylated AurA without Tpx2 show that the activation loop is highly dynamic, reflected as relatively broad distributions of L225-S284 and R255-T288 distances (Figure 3b, bottom right panel). The N-terminal lobe of the kinase was particularly dynamic in these simulations, and local unfolding occurred within the αC-helix in many of the trajectories, as seen previously in simulations of the epidermal growth factor receptor (Shan et al., 2012) as well as in X-ray structures of the related AGC-family kinase Akt in the unphosphorylated state (Yang et al., 2002a). In striking contrast, the simulations showed that phosphorylated AurA is locked into a single conformation with a long L225-S284 distance (42 Å, cf. 41 Å in the 1OL5 x-ray structure) and short R255-T288 distance (~5 Å, cf. 5.5 Å in 1OL5), indicative of a stable active state in which the loop is fully ordered and the phosphothreonine residue forms ion-pairing interactions with R255 and R180 (Figure 3b, bottom left panel, and Figure 3—figure supplement 1b). Interestingly, phosphorylation alone is almost as effective at constraining the loop in the active state as phosphorylation and Tpx2 together (Figure 3b, left panels). In contrast, the simulations show that the activation loop of unphosphorylated AurA bound to Tpx2 remains somewhat dynamic (Figure 3b, top right panel), and additional phosphorylation significantly stabilizes the loop. Phosphorylation disrupts an autoinhibitory DFG-In substate Although unphosphorylated AurA is highly dynamic in the absence of Tpx2, the activation loop is not in fact disordered in the simulations. Instead, two discrete subpopulations are observed: one subpopulation corresponding to the active-like state with a long L225-S284 distance (~43 Å), and another with a much shorter distance (~38 Å), representing a DFG-In conformation in which the activation loop is not correctly positioned for catalytic function (Figure 3b, lower right panel). Manual inspection of the trajectories revealed that in this subpopulation the tip of the activation loop folds into a short helical turn spanning residues P282-R286, with the P282 proline residue serving as the N-terminal capping residue (Kumar and Bansal, 1998) in most of the trajectories (Figure 3c). Calculating the pseudodihedral angle for the Cα atoms of S283-R286 across all trajectories confirmed that the inactive subpopulation possessed well-defined helical pseudodihedral values of 50–75° (Figure 3d). Although this conformation has not been observed in X-ray structures of AurA, the formation of short helices in the activation loop is a common feature of the inactive states of other protein kinases (Sicheri et al., 1997; Wood et al., 2004; Lee et al., 2010; De Bondt et al., 1993). An interesting feature of the autoinhibited DFG-In substate observed in the simulations is that the T288 residue, which immediately follows the helical segment in the protein sequence, is positioned close to the C-terminal end of the helix in almost all of the trajectories (Figure 3c), with the sidechain hydroxyl forming hydrogen bonds to the backbone carbonyls of residues R285 and R286 in many of the simulation snapshots. We reasoned that upon phosphorylation of T288, the proximity of the phosphate group to the negatively-charged end of the helix dipole (Hol et al., 1978) would destabilize this autoinhibited substate, promoting the refolding of the activation loop to the active conformation. We wondered why the helical conformation of the activation loop has not been observed in X-ray structures of AurA. In fact, the activation loop adopts the active conformation in only a small subset of AurA structures, specifically those determined either in the presence of Tpx2 (Bayliss et al., 2003; Zhao et al., 2008; Clark et al., 2009) or other protein factors that stabilize the active state (Richards et al., 2016). Instead, almost all of the structures of AurA in the DFG-In state (76 structures out of 138 total structures of AurA in the PDB) were determined in the same hexagonal crystal form in which the kinase adopts an inactive conformation with the activation loop misaligned and the peptide binding site disassembled. Upon examination of the crystal lattice we noticed that this conformational state of the activation loop appears to be induced by a crystal contact between the peptide binding site and a neighboring molecule in the lattice (Figure 3—figure supplement 2). This apparent crystallographic artifact may have prevented previous observation of the helical autoinhibited DFG-In substate visualized in our simulations, which model the kinase in solution rather than in the crystallographic context. Our MD simulations, which represent over a millisecond of simulation data, predict that phosphorylation has profound effects on the activation-loop conformation of AurA within the DFG-In state, both disrupting an autoinhibited substate and promoting an active substate that is primed for catalytic function. In an attempt to confirm the simulation result that the T288 phosphothreonine residue of phosphorylated AurA is correctly coordinated in the active state, we mutated its ion-pairing partner R180 to an alanine residue and measured the effect on kinase activity. The R180A mutant possessed 4-fold lower activity in the absence of Tpx2, whereas the