The gatekeeper residue controls autoactivation of ERK2 via a pathway of intramolecular connectivity

Abstract

Studies of protein kinases have identified a “gatekeeper” residue, which confers selectivity for binding nucleotides and small-molecule inhibitors. We report that, in the MAP kinase ERK2, mutations at the gatekeeper residue unexpectedly lead to autoactivation due to enhanced autophosphorylation of regulatory Tyr and Thr sites within the activation lip that control kinase activity. This occurs through an intramolecular mechanism, indicating that the gatekeeper residue indirectly constrains flexibility at the activation lip, precluding access of the phosphoacceptor residues to the catalytic base. Other residues that interact with the gatekeeper site to form a hydrophobic cluster in the N-terminal domain also cause autoactivation when mutated. Hydrogen-exchange studies of a mutant within this cluster reveal perturbations in the conserved DFG motif, predicting a route for side chain connectivity from the hydrophobic cluster to the activation lip. Mutations of residues along this route support this model, explaining how information about the gatekeeper residue identity can be transmitted to the activation lip. Thus, an N-terminal hydrophobic cluster that includes the gatekeeper forms a novel structural unit, which functions to maintain the “off” state of ERK2 before cell signal activation.

The MAP kinase ERK2 is activated in response to many stimuli and elicits diverse cellular responses by phosphorylating cytoplasmic and nuclear proteins (1, 2). ERK2 is activated by dual phosphorylation at Thr and Tyr residues in a conserved region called the activation lip, catalyzed by MAP kinase kinase (MKK) 1/2. Insight into the activation mechanism has been gained from x-ray structures of ERK2 in its active and inactive forms (3, 4). Dual phosphorylation causes remodeling of the activation lip conformation to reorganize active site residues, recognize substrate specificity determinants, and expose a hydrophobic pocket for substrate docking motif binding (3, 5).

In ERK, autophosphorylation at regulatory Tyr/Thr residues occurs intramolecularly and must be maintained at low levels to prevent kinase autoactivation (6–8). It has been shown that low but chronic activation of MAP kinases is sufficient for signaling (9–12), and mutants of MKK and ERK that bypass phosphorylation cause activation of transcription, differentiation, and cell transformation (12–16). Thus, MAP kinases must constrain autophosphorylation to prevent dysregulated signaling before cell stimulation. Mechanisms that confer such constraints are poorly understood.

Structure–function studies of protein kinases have identified a “gatekeeper” residue, which regulates binding of nucleotide and small-molecule inhibitors (17–22). This residue is located on a conserved β5 strand, distal to the active site and adjacent to a hinge region that connects the N- and C-terminal domains of the enzyme. In most protein kinases the gatekeeper residue side chain is large, although a few kinases have small side chains at this position, which has been exploited in pharmaceutical inhibitor design (22–24). Substituting a bulky side chain at the gatekeeper site with a small side chain confers selective inhibition by molecules that target the empty space, such as the p38 inhibitor, SB203580, and C3-substituted analogues of the Src inhibitor PP1 (19, 20). Such mutants also allow preferential recognition of ATP analogs with bulky groups at the N₆ ring position (17, 18, 20). As expected, mutation of the gatekeeper residue in ERK2 to smaller residues leads to recognition and binding of PP1 derivatives and N₆-cyclopentyl ATP (25, 26).

Here we report a previously undescribed response in ERK2 to gatekeeper residue mutations, which have the unexpected effect of elevating kinase-specific activity by enhancing autophosphorylation of the regulatory Tyr/Thr sites within the activation lip. The mechanism is intramolecular, indicating that the gatekeeper residue somehow prevents interactions between activation lip residues and the catalytic base (D147) needed for phosphoryl transfer. Based on evidence from hydrogen-exchange (HX) MS, steady-state kinetics, and mutagenesis, we formulate a model in which the gatekeeper and neighboring residues form a structural unit that constrains backbone flexibility at the activation lip through an intramolecular pathway of connectivity. This provides a mechanism to retain ERK2 in an inactive state before activation by MKK1/2.

Results

Mutation of ERK2 at Q103 Increases Basal Kinase Activity.

To examine the gatekeeper residue in ERK2, Q103 was mutated to Gly or Ala, and the basal specific activity (i.e., without further phosphorylation by MKK1/2) was measured against myelin basic protein (MBP) as substrate. Unexpectedly, mutations Q103G and Q103A led to autoactivation, with specific activities that were respectively 10- and 35-fold greater than ERK2-WT (Fig. 1A). Because Q103 is adjacent to the hinge region of ERK2, other hinge residues were tested for effects on activity. Double mutants substituting both Q103 and residues immediately adjacent (V102G/A + Q103G/A = “¹⁰²VQ/GG or AA”; Q103G/A + D104G/A = “¹⁰³QD/GG or AA”) exhibited 6- to 36-fold increased specific activity over ERK2-WT (Fig. 1A); however, mutation of other residues within the hinge showed no effect (Fig. 7, which is published as supporting information on the PNAS web site). Thus, mutations at Q103 alone were sufficient for autoactivation.

Fig. 1.

Mutation of ERK2 at Q103 increases basal activity. (A) Kinase assays measuring basal rates of MBP phosphorylation for ERK2 gatekeeper mutants, normalized to the activity of ERK2-WT (20 ± 2 pmol/min·mg). (B) Kinase assays after ERK2 activation by active MKK1. (C) ERK2 mutants show slower SDS/PAGE mobility after phosphorylation by MKK1. (D) Basal kinase activities of WT and mutant ERK2 toward MBP in the presence of SB203580 or 1NM-PP1, normalized to activities without inhibitor (WT, 14 ± 1; Q103A, 239 ± 24, L73P/S151D, 301 ± 5 pmol/min·mg).

To test whether the gatekeeper residue affected intrinsic kinase activity, each mutant was phosphorylated in vitro with active MKK1 to produce fully active enzyme (“ppERK2”) (Fig. 1B). Specific activities of ppERK2-Q103G and Q103A toward MBP were indistinguishable from ppERK2-WT, and ¹⁰²VQ and ¹⁰³QD double mutants retained at least 50% of maximal activity. Furthermore, all ERK2 mutants showed complete SDS/PAGE mobility retardation indicating high stoichiometry of phosphorylation after reaction with MKK1 (Fig. 1C). This demonstrated proper folding of each mutant, because structural integrity of ERK2 is required for recognition by MKK1.

Previous studies showed that ppERK2-Q103A was susceptible to inhibitors SB203580 and PP1 analogues (21, 26). We therefore asked whether the enhanced basal specific activity of ERK2-Q103A was also sensitive to these inhibitors. MBP kinase assays showed that 1 μM SB203580 or 1NM-PP1 suppressed the basal activity of ERK2-Q103A, whereas WT was unaffected (Fig. 1D). Thus, the ERK2 gatekeeper mutation retained the expected inhibitor selectivity observed in other kinases, despite the unusual elevation of basal specific activity unique to ERK2.

Q103 Mutations Enhance Autophosphorylation at the Activation Lip.

We next asked whether ERK2 autoactivation depended on the phosphorylation of regulatory sites, T183 and Y185. ERK2 mutants in their basal states showed gel-shifted forms for Q103G/A, ¹⁰²VQ/GG or AA, and ¹⁰³QD/GG or AA that correlated with basal activity (Fig. 2A). To test whether the gel shifts indeed reflected phosphorylation of activation lip residues, proteins were examined by Western blotting using antibodies specific for phosphorylated ERK (Fig. 2B). Immunoreactivity with phosphospecific antibodies was elevated in mutants of Q103 and the ¹⁰²VQ and ¹⁰³QD double mutants, confirming enhanced autophosphorylation at the activation lip.

Fig. 2.

The gatekeeper mutation enhances activation lip autophosphorylation. (A) ERK2 mutants show SDS/PAGE shifts correlating with activity. (B) Western blots of ERK2 probed with antibodies recognizing pTpY-ERK, pY-ERK, or ERK2. “pERK2” denotes monophosphorylated or diphosphorylated ERK2, and “ERK2” denotes unphosphorylated ERK2. (C) LC/MS showing extract ions of MH₂²⁺, MH₃³⁺, and MH₄⁴⁺ charge states corresponding to unphosphorylated, monophosphorylated (+80 Da), and diphosphorylated (+160 Da) forms of the tryptic peptide ¹⁷¹VADPDHDHTGFLTEYVATR. (D) Phosphorylation of T183 and Y185 is required for autoactivation. ERK2-Q103A and ERK2-QD/AA or ppERK2-WT (phosphorylated by active MKK1) were pretreated with PP2A or CD45, with or without okadaic acid/microcystin (OA/MC) or Na₃VO₄ (VAN), followed by kinase assays toward MBP.

LC-MS was performed to quantify the phosphorylation stoichiometry (27). Unphosphorylated, monophosphorylated, and diphosphorylated forms of a peptide containing both Thr and Tyr phosphorylation sites were quantified by summing over integrated peak areas for each observed charge state. Phosphorylation at T183 and Y185 was elevated in ERK2-Q103A and ¹⁰³QD/AA compared with ERK2-WT, with greatest occupancy at pY185 (Fig. 2C). Stoichiometries were 0.38 mol/mol for ERK2-Q103A (1% 2P, 36% 1P-pY185, 63% 0P) and 0.64 mol/mol for ERK2-¹⁰³QD/AA (1% 2P, 62% 1P-pY185, 37% 0P), compared with 0.13 mol/mol for ERK2-WT (0% 2P, 11% 1P-pY185, 2% 1P-T183, 87% 0P). The extent of autophosphorylation reported by MS agreed well with the relative immunoreactivity by Western blotting (Fig. 2B).

Activation Lip Phosphorylation Causes Autoactivation.

The level of diphosphorylation observed in ERK2-Q103A and ¹⁰³QD/AA, although low, correlated well with the increased basal specific activity. To prove that phosphorylation of both pT183 and pY185 was required for autoactivation, we tested the sensitivity of each mutant to treatment with phosphatases PP2A or CD45, which each inhibited ppERK2-WT (Fig. 2D Right), and selectively dephosphorylated ppERK2 at pT183 and pY185, respectively (LC-MS data not shown). We found that PP2A and CD45 each ablated the basal activity of gatekeeper mutant kinases (Fig. 2D Left). These results demonstrate that, although mutations at the gatekeeper residue primarily enhance phosphorylation at pY185, they also affect phosphorylation at pT183 and that phosphorylation of both residues is required for autoactivation.

Autophosphorylation Occurs Intramolecularly.

The mechanism of autophosphorylation was investigated by measuring phosphoryl transfer to GST-ERK2-K52R by His₆-ERK2-Q103A; because GST-ERK2-K52R is catalytically inactive, its autophosphorylation should occur in trans (Fig. 3A). Increasing the amount of mutant His₆-ERK2-Q103A with fixed levels of GST-ERK2-K52R led to increased ³²P incorporation into His₆-ERK2 but not GST-ERK2 (Fig. 3A Lower) and no effect on gel mobility of GST-ERK2 (Fig. 3A Upper). Similar results were observed upon mixing His₆-ERK2-¹⁰³QD/AA with GST-ERK2-K52R (Fig. 3A). The absence of trans-phosphorylation implied that ERK2-Q103A and ¹⁰³QD/AA autophosphorylate through an intramolecular mechanism. To exclude the possibility that GST-ERK2-K52R was somehow incapable of serving as an acceptor for trans-phosphorylation, we alternatively measured autophosphorylation initial rates vs. ERK2 concentration; rates of trans-phosphorylation requiring interactions between two kinase subunits should increase as the square of ERK2 concentration, whereas rates of cis-phosphorylation should vary linearly with kinase. The results showed that autophosphorylation rates were linear with concentration of both ERK2-Q103A and WT (Fig. 3B), confirming an intramolecular mechanism.

Fig. 3.

Autophosphorylation occurs through an intramolecular mechanism. (A) GST-ERK2-K52R (0.5 μg, lanes 2–7 and 9–12) incubated with or without His₆-ERK2-Q103A or His₆-ERK2-QD/AA (500, 5, 50, 250, and 500 ng in lanes 1–5 and 8–12) in the presence of Mg²⁺ [γ-³²P]ATP. GST-ERK2-K52R gel shifts after phosphorylation by active MKK1 are shown in lane 7. (B) Initial rates of autophosphorylation were measured by phosphoryl transfer from [γ-³²P]ATP at varying concentrations of His₆-ERK2-WT, His₆-ERK2-Q103A, or His₆-ERK2-L73P/S151D. Linearity of rate vs. concentration with slope = 1 confirms an intramolecular mechanism.

A Cluster of N-Terminal Domain Residues Regulate Activation Lip Phosphorylation.

The characteristics of the gatekeeper mutations resembled previous findings of ERK2 mutations L73P and S151D, which increased basal specific activity via intramolecular autophosphorylation, and enhanced mammalian cell signaling (12). ERK2-L73P/S151D shares characteristics with the gatekeeper mutations, by its increased reactivity with anti-pY-ERK2 and anti-pYpT-ERK2 (Fig. 2B), elevated occupancy of pY185 and pT183 (Fig. 2C), and linear dependence of autophosphorylation rate on kinase concentration (Fig. 3B), although it is resistant to inhibition by SB203580 or 1NM-PP1 (Fig. 1D). Combining mutations at Q103 with L73P/S151D showed no further increase in activity over Q103 mutants alone, and lower activity compared with L73P/S151D alone (Fig. 4A). This nonadditive effect indicated that the mutations are not independent with respect to autoactivation. We hypothesized that they might form a structural motif, controlling autophosphorylation through a shared mechanism.

Fig. 4.

Residues mediating ERK2 activation form a hydrophobic cluster in the N-terminal domain. (A) L73 and Q103 mutations in ERK2 show no additivity, by SDS/PAGE shifts or basal kinase activities measured toward MBP, normalized to WT (20 ± 2 pmol/min·mg). (B) Ribbon diagram of inactive ERK2 (1ERK), with space-filled residues Q103, I82, I84, and L73 in the hydrophobic cluster. Stick representations show T183, Y185, and D147. (C) Kinase activities toward MBP normalized to ERK2-WT (54 ± 9 pmol/min·mg) show autoactivation, and SDS/PAGE shows reduced mobility of ERK2-I84A. As a control, kinase assays performed after activation by MKK1 showed that ppERK2-I84A, Q103A, ¹⁰³QD/AA, and L73P/S151D retained >55% of the activity observed with ppERK2-WT, whereas ppERK2-I82A retained <25% activity (Fig. 8, which is published as supporting information on the PNAS web site), indicating that I82A interferes with kinase function. (D) Initial rates of autophosphorylation at varying concentrations of His₆-ERK2-I84A.

Inspection of the ERK2 structure showed that L73 and Q103 form hydrophobic side chain interactions within the N-terminal domain (Fig. 4B). If our hypothesis were correct, we predicted that other residues which interact with Q103 might also affect autophosphorylation and kinase activity. The ERK2 structure showed two other residues within van der Waals distances of L73 and Q103, located at positions I82 and I84 (Fig. 4B). Side chain interactions between these four residues comprised a “hydrophobic cluster” in the N terminus, and except for I82, all appeared inaccessible to solvent in space filling models (not shown).

To test the importance of this hydrophobic cluster, we engineered mutations at I82 and I84 to see whether reducing side chain size would recapitulate the effects of Q103 and L73 mutations on basal kinase activity and lip phosphorylation. I84A enhanced basal specific activity by 80-fold over WT (Fig. 4C), whereas I82A showed no effect on basal activity. Again, the mechanism of I84A autoactivation appeared to be due to intramolecular autophosphorylation, as shown by increased level of the phosphorylated gel-shifted form and linearity of initial rates of autophosphorylation vs. enzyme concentration (Fig. 4 C and D). These results support a model in which these hydrophobic residues are part of a single structural unit that contains the gatekeeper residue, and functions to constrain intramolecular autophosphorylation in ERK2.

A Model for Long-Distance Perturbation of Activation Lip Flexibility.

How might residues Q103, L73, and I84 in the hydrophobic cluster normally act to constrain autophosphorylation within the activation lip? The x-ray structure of inactive ERK2 shows that Y185 is buried within a pocket in the C-terminal domain, in a conformation stabilized by interactions between residues in the lip and surrounding residues in the C-terminal domain (Fig. 4B). Intramolecular autophosphorylation requires movement of the hydroxyl group of Y185 to a position within hydrogen bonding distance of the catalytic base, D147. Mutations which enhance autophosphorylation must somehow alter the flexibility of the activation lip, to allow this interaction. However, the Q103 side chain atoms are >20 Å away from the nearest Y185 atoms, raising the question of how perturbations within the hydrophobic cluster influence autophosphorylation at Y185 over long distances.

We hypothesized that this information is transmitted through side chain interactions that control backbone flexibility at the lip through an intramolecular route in the protein. HX has been used to monitor internal motions within folded proteins, where exchange predominantly occurs through low energy fluctuations in protein structure which enable transient solvent accessibility (28, 29). We therefore compared HX behavior of ERK2-WT to ERK2-I84A using HX-MS, which measures rates of deuteration at backbone amides with spatial resolution of 5–20 residues in ERK2 in peptides generated through pepsin cleavage (30).

HX-MS of ERK2-WT and I84A was compared, monitoring 37 peptides with 91% coverage of amide hydrogens in each protein (Fig. 9, which is published as supporting information on the PNAS web site). To eliminate heterogeneity in the preparations, each form was pretreated with phosphatase to produce uniformly dephosphorylated protein. The I84A mutation did not grossly alter HX behavior, indicating that its overall solution structure was similar to WT kinase. However, HX perturbations introduced by I84A were observed in four specific regions (Fig. 5). In peptide 16 containing the catalytic D147, decreased HX was observed in the I84A mutant, consistent with expected activation lip interactions that promote autophosphorylation and also protect this region from solvent. Increased HX was also observed in peptides containing I84 (peptide L/8), and Q103 (peptide 12), consistent with the decreased side chain volume after the Ile/Ala substitution, which was likely due to increased local backbone flexibility around the interacting residues. Importantly, an unexpected increase in HX was observed within the conserved DFG motif (residues 165–167, peptide 18/19) suggesting that communication between I84 and the activation lip involves their mutual connections to this conserved region.

Fig. 5.

ERK2-I84A perturbs conformation and/or mobility in the conserved DFG motif. (A) HX measured for WT (black) and I84A (red and green) forms of ERK2. Increased HX was observed in regions containing I84 (peptide L/8) and Q103 (peptide 12), consistent with side chain perturbations introduced by the I84A mutation. Decreased HX was observed around D147 (peptide 16), consistent with activation lip interactions that promote autophosphorylation. Increased HX was observed in the DFG region (peptide 18/19), suggesting its involvement in communication between I84 and the lip. (B) Each region is indicated on the ERK2 ribbon diagram.

Upon reexamining the ERK2 structure, we noted that Q103, L73 and I84 formed hydrophobic interactions with conserved residue F166, located within the DFG motif (Fig. 6A and B). The effect of I84A on HX suggested that mutations within the hydrophobic cluster perturb backbone conformation and/or mobility in proximity to F166. F166 is linked through the backbone to L168, which in turn forms side chain interactions with V186, located in the lip. This pathway of connectivity links the hydrophobic core to the activation lip. In addition, the hydroxyl side chain of lip residue T188 forms hydrogen bond interactions with the catalytic base, D147, and a conserved interacting residue, K149. Together, these residues form a network of contiguous side chain interactions that tether the activation lip in a stable conformation. This would constrain the accessibility of Y185/T183 to D147, and restrict autophosphorylation in ERK2-WT.

Fig. 6.

A model for control of activation lip flexibility by residues in the hydrophobic cluster. (A and B) The ERK2 structure (1ERK) reveals side chain and backbone connectivities between residues in the hydrophobic cluster and the activation lip, suggesting a route for control of activation lip flexibility. Distances are indicated in angstroms. (C and D) Although ERK2 mutants L168A, V186S, and T188A cause reduced activity after phosphorylation by active MKK1 (C), they result in increased rates of autophosphorylation (D) comparable to Q103A and I84A. Values normalized to the rate of ERK2-WT autophosphorylation.

Connectivity Between the Activation Lip and N-Terminal Hydrophobic Cluster.

We hypothesized that substitution of Q103, L73, or I84 with smaller side chains might form a cavity in the hydrophobic cluster, enabling movement of F166 and L168 in the DFG region, and disrupting side chain interactions between L168-V186 and T188–D147/K149 which stabilize the lip. This model was tested by mutating residues in the activation lip (V186, T188) and residues connecting the hydrophobic core and lip (F166, L168), and measuring effects on autophosphorylation. Many of these mutants interfered with ERK2 activity after MKK1 phosphorylation, most likely by disrupting stability of the catalytic or substrate binding sites (Fig. 6C). Nevertheless, autophosphorylation was increased in mutants L168A, V186S, and T188A compared with WT, mimicking the behavior of I84A and Q103A (Fig. 6D). LC/MS of ERK2-V186S and ERK2-T188A confirmed increased phosphorylation of Y185 and T183 (Fig. 10, which is published as supporting information on the PNAS web site). Mutant F166A did not affect autophosphorylation, but in combination with I84A suppressed the enhanced autophosphorylation seen with I84A alone, suggesting the importance of F166 in mediating autoactivation by I84A (Fig. 6D). Thus, mutations that disrupted contiguous interactions between the hydrophobic core and the activation lip were able to control autophosphorylation, consistent with a mechanism for long distance communication involving specific side chain connectivities.

As a negative control for these studies, we tested an alternative route for connecting Q103/I84/L73 to the activation lip, involving a contiguous pathway of hydrophobic side chain and backbone interactions within helix αC and the C-terminal L16 loop (Fig. 11 A and B, which is published as supporting information on the PNAS web site). Mutations at several residues in this pathway neither increased autophosphorylation, nor interfered with the extent of autoactivation by I84A (Fig. 11 C and D), indicating that the intramolecular interactions involving the DFG motif provide the primary route for control of ERK2 autophosphorylation.

Discussion

The main finding of this study is that the gatekeeper residue in ERK2 forms part of a novel N-terminal structural unit that functions to impede autoactivation. Starting with the unexpected property of gatekeeper mutants in promoting kinase activation by increasing autophosphorylation of activation lip residues, we identified a cluster of hydrophobic residues that interact with the gatekeeper and show similar behavior. Using HX and 3D structure analysis, we formulated a plausible model for intramolecular connectivity to the activation lip and validated it by showing that mutations at connecting residues in this pathway increased autophosphorylation. In contrast, mutations at residues forming an alternative route for connectivity between the hydrophobic cluster and lip had little effect. The results reveal insight into how constraints are built into the structure of ERK2 to prevent autophosphorylation and autoactivation before MKK-catalyzed phosphorylation. Without such constraints, constitutive activation of ERK2 would occur in the absence of upstream signaling, leading to dysregulated pathway function, uncontrolled proliferation, and transformation.

Autophosphorylation occurs through an intramolecular reaction, indicating that the gatekeeper mutations alter conformation or flexibility within the kinase in a manner that promotes interaction between the phosphoacceptor sites and the catalytic base. In inactive ERK2, Y185 and T183 hydroxyl groups are ≈8 Å and ≈17 Å from the D147 carboxyl group. These distances, along with constraints on lip backbone flexibility, must block ERK2 autophosphorylation under normal conditions, thus preventing spontaneous activation. Activities of each ERK2 mutant after MKK1 phosphorylation are comparable to ppERK2-WT, indicating that the gatekeeper mutations do not significantly distort the geometry of residues in the catalytic site and that major conformational changes involving D147 do not account for the enhanced autophosphorylation. Furthermore, the phosphorylation stoichiometry of each mutant is moderate (0.38–0.64 mol/mol), suggesting that conformers allowing intramolecular phosphotransfer are achieved only a fraction of the time. For these reasons, we favor a mechanism in which mutations near the gatekeeper residue increase flexibility within the activation lip, enabling transient sampling of conformers productive for intramolecular autophosphorylation and elevating kinase-specific activity.

The fact that the gatekeeper residue is located 20 Å from Y185 in ERK2 means that activation lip flexibility can be controlled over long distances. Understanding how this occurs in detail will require extensive analysis of structure and dynamics in ERK2. Nevertheless, a clue was provided by the L73P and S151D mutations, which like Q103A, elevate specific activity by enhancing intramolecular autophosphorylation. Interestingly, L73 and Q103 side chains are located within 4.2 Å of each other, suggesting the importance of hydrophobic interactions between these residues. Combining the Q103 mutations with L73P/S151D did not further enhance autophosphorylation, suggesting that these mutations converge on a shared mechanism for controlling lip flexibility. The hydrophobic cluster of side chains involving Q103 and L73 predicted that reduced side chain volumes at other interacting residues would mimic effects of the gatekeeper mutations. This prediction was confirmed with I84A, revealing the shared importance of residues in the hydrophobic cluster for controlling phosphorylation.

The ERK2 structure suggests a model to explain how residues in this hydrophobic cluster, distal to the active site, control activation lip flexibility. The side chains contact F166, which forms part of the highly conserved DFG motif, a central structural unit in kinases. The HX-MS measurements show that the I84A mutation increases HX in the DFG region, revealing an importance of this motif to the autoactivation mechanism. Such behavior suggests that I84A leads to movement of F166 and enhancement of backbone mobility surrounding the DFG motif. The Asp-Phe residues are often found in two thermodynamically accessible conformers (31, 32). In several kinases these have been associated with transitions between active forms, where Asp faces the catalytic pocket and Phe is buried in the N-terminal domain, and inactive forms, which show an inverted conformation with Phe facing the catalytic pocket. However, in both inactive and active ERK2, a single conformer is found, with Asp facing the catalytic pocket and Phe facing the hydrophobic cluster. We propose that this configuration is stabilized by interactions of F166 with L73, Q103 and I84 in the inactive form, to restrain autoactivation.

Together, the data leads us to a model in which L73, Q103, or I84 mutations create a cavity in the hydrophobic cluster, leading to movement of F166. Coupling of F166 to L168 through the backbone then causes movement of L168, which in turn disrupts hydrophobic interactions between L168-V186 and hydrogen bonding interactions between T188–D147/K149. Disruption of these interactions results in increased activation lip flexibility, allowing intramolecular phosphoryl transfer from ATP to Y185 and T183. Consistent with this, Ala mutations chosen to disrupt interactions between L168 and V186 or between T188 and D147/K149 were found to enhance the rate of ERK2 autophosphorylation. The model also explains how the S151D mutation enhances autophosphorylation, by forming a D151–K149 ion pair which further disrupts the T188–K149 interaction (12) (Fig. 6B).

In summary, our studies demonstrate a novel structural unit containing the gatekeeper residue, which plays an important role in restraining ERK2 autoactivation in the absence of upstream signaling. The autoinhibitory mechanism involves regulation of protein flexibility over long distances, as distinct from other examples of autoinhibition in kinases which often involve local control of activation lip conformation, e.g., through steric inhibition by domains outside the kinase core. It is interesting to note that mutations at the gatekeeper site in v-Src also affect autophosphorylation, in this case decreasing autophosphorylation with reduced side chain size (18). Intramolecular pathways that confer coupling of conformational and motional changes over long distances underlie enzyme allostery but are in general poorly defined. The involvement of the DFG motif in this pathway provides a novel function for this conserved central structural region in transmitting information about the gatekeeper site identity to the activation lip.

Materials and Methods

ERK2 Purification.

ERK2 mutations were generated in plasmids NpT7-5 His₆-ratERK2 (6) and pGEX-KG GST-ratERK2 and confirmed by sequencing. His₆-ERK2 plasmids were transformed into BL21 Star Escherichia coli (Invitrogen, Carlsbad, CA), and proteins were prepared from 50-ml cultures as described (12). Purified ERK2 was eluted from Ni²⁺-NTA agarose (40 μl; Qiagen, Valencia, CA) in 200 μl of 50 mM KH₂PO₄ (pH 8.0)/0.3 M NaCl/0.1% 2-mercaptoethanol/0.3 M imidazole (pH 8.0), dialyzed into storage buffer (10 mM Hepes, pH 7.5/0.1 M NaCl/1 mM DTT), and frozen in single-use aliquots at −80°C. GST-ERK2-K52R was induced in E. coli, purified with glutathione-Sepharose (40 μl), eluted with storage buffer plus 20 mM glutathione (pH 8.0), dialyzed into storage buffer, and frozen in aliquots.

For HX-MS, His₆-ERK2-WT and His₆-ERK2-I84A were each induced in 1-liter culture of E. coli and purified by Ni²⁺-NTA agarose (1 ml) as above. Lambda phosphatase (50,000 units; New England Biolabs, Ipswich, MA) was added to 1 mg of ERK2 and incubated for 3 h at 30°C before MonoQ FPLC to resolve ERK2 from phosphatase. Phosphatase treatment of ERK2 reduced phosphorylation to <1% mol/mol by LC-MS.

Enzyme Assays.

Kinase autophosphorylation was assayed by phosphoryl transfer to inactive GST-ERK2-K52R catalyzed by His₆-ERK2 for 2 h at 30°C in assay buffer {25 mM Tris·HCl, pH 8.0/2 mM DTT/0.1 mM ATP/10 mM MgCl₂/2.5 μCi (1 Ci = 37 GBq) of [γ-³²P]ATP}. Alternatively, the autophosphorylation of His₆-ERK2 was measured at 5, 10, 30, and 60 min, from which initial rates were calculated (20 mM Hepes, pH 7.4/50 mM NaCl/1 mM DTT/0.02% Triton X-100/10 mM MgCl₂/0.2 mM ATP/2.5 μCi of [γ-³²P]ATP). Reactions were quenched, proteins were resolved by SDS/PAGE, and phosphate incorporation was quantified by PhosphorImager (Molecular Dynamics, Sunnyvale, CA).

Kinase activity of ERK2 (0.5 μg) or diphosphorylated ERK2 (0.01 μg) was measured against 5 μg of rabbit MBP (Sigma, St. Louis, MO) for 1 or 10 min at 30°C in assay buffer in the presence or absence of SB203580 (Calbiochem, San Diego, CA) or 1NM-PP1 (19). MBP phosphorylation was quantified by PhosphorImager. Diphosphorylation of ERK2 was performed with 0.05 μg of constitutively active His₆-MKK1(ΔN4/S218E/M219D/N221D/S222D) (33), yielding ≈1.7 mol/mol by LC-MS.

Phosphatase treatments were performed by incubating ERK2 (WT or mutant) or diphosphorylated ERK2-WT (0.5 μg) with 0.1 μg of PP2A or 0.2 μg (4U) of CD45 (Calbiochem) for 30 min at 30°C in 20 mM Tris·HCl (pH 7.5), 1 mM DTT, in the presence or absence of 3 μM okadaic acid plus 3 μM microcystin or 3 mM Na₃VO₄. Phosphatase was stopped by adding okadaic acid/microcystin to PP2A or Na₃VO₄ to CD45, and kinase assays were run for 1 min at 30°C with MBP substrate.

LC-MS.

Phosphorylation stoichiometry was measured by LC-MS (27). ERK2 (10 μg) was digested overnight at 37°C in 0.2 M Tris·HCl (pH 8.0), 1 mM CaCl₂, 1.6 M urea, and 0.5 μg of porcine trypsin (Wako, Richmond, VA), and peptides were separated by RP-LC-MS (Phenomenex Jupiter C18, 250-μm i.d. × 20 cm, 0–40% acetonitrile in 0.1% formic acid, ABI QStar Pulsar). HX-MS measurements were carried out as described (5, 30) in reactions with 5 μg of ERK2, 90% (vol/vol) D₂O, 5 mM K₂HPO₄ (pH_read7.4), 10 mM KCl, and 0.25 mM DTT. Samples were incubated at 10°C, quenched by acidification, and digested with 10 μl (5 μg) of pepsin (Sigma). Peptides were analyzed by RP-LC-MS (POROS 20 R1; 250-μm i.d. × 10 cm). Time-zero controls were acid-quenched before D₂O, and weighted average masses were corrected for back-exchange and artifactual in-exchange. Time points and sample order were randomized, and deuterons incorporated were determined from three independent measurements at 1 min and two measurements each at 4 and 10 min.

Immunoblotting.

ERK2 proteins were probed with rabbit anti-ERK2 (1:4,000) (C14; Santa Cruz Biotechnology, Santa Cruz, CA), mouse anti-pY-ERK (1:10,000), which recognizes pY185 monophospho-ERK1/2, and mouse anti-pTpY-ERK1/2 (1:2,000) (Sigma), which recognizes diphospho-ERK1/2.

Abbreviations

HX

hydrogen exchange

MKK

MAP kinase kinase

MBP

myelin basic protein.