Protein purification

Active, diphosphorylated (2P) ERK1 was produced from plasmid pET-His-ERK1-R4F, containing wild-type human His₆-ERK1 expressed in tandem with constitutively active mutant MKK1 (R4F: ΔN3/S218E/S222D) [16]. Inactive, unphosphorylated (0P) human His₆-ERK1 was produced using plasmid NpT7-ERK1 [17]. Each kinase form was expressed in E. coli strain BL21(DE3)-pLysS, and purified using Ni-NTA agarose (Qiagen) chromatography, desalted by gel filtration (PD10, GE Healthcare), and further purified by MonoQ FPLC. Proteins were dialyzed overnight into 50 mM KPO₄ (pH 7.4), 100 mM KCl and 5 mM dithiothreitol, and stored in aliquots at −80 °C. The 0P-ERK1 protein was found to be mono-phosphorylated to 0.13 mol/mol, and was therefore dephosphorylated using λ phosphatase (New England Biolabs, 100 U/μg, 3 h, 30 C) before Mono Q separation. The activation lip phosphorylation stoichiometries of the final kinase preparations showed 0% phosphorylation in 0P-ERK1 at the regulatory Thr and Tyr residues in the activation lip, and 97% diphosphorylation in 2P-ERK1. Intact ESI-MS of 0P-ERK1 and 2P-ERK1 showed full length masses expected of the recombinant ERK1 sequences, in unphosphorylated and diphosphorylated forms (Suppl. Fig. 1).

HX-MS measurements

Data were collected on a QStar Pulsar QqTOF mass spectrometer (ABI) interfaced with an Agilent Cap1100 HPLC (0.5 mm i.d. × 10 cm column, packed with POROS R1 20 resin), and measurements of weighted average mass were performed as described [18,19]. Proteins (4 μg) were incubated in 90% D₂O (Sigma) at 10 °C, allowing in-exchange reactions to take place between 8 s – 4 h. Reactions were quenched with 90 μL 25 mM succinic acid, 25 mM citric acid (pH 2.4), and cooled rapidly to 0 °C. Proteins were digested by adding 10 μL pepsin (4 μg) and analyzed immediately by LC-MS. Time-zero measurements were performed by quenching the reaction before adding D₂O.

Peptides in pepsin digests of ERK1 were identified by LC-MS/MS, analyzing replicate digests on two instruments in order to maximize the probability that a peptide would be selected for MS/MS. Aliquots (4 μg) of proteolyzed ERK1 were analyzed on the QStar mass spectrometer, with m/z window 400–1600 Da, duty cycle 15.5 s, and three MS/MS per cycle. Aliquots (60 ng) of ERK1 were analyzed on an LTQ-Orbitrap mass spectrometer interfaced with an Eksigent 2DLC HPLC (75 μm i.d.× 150 mm column, Zorbax C18 resin), with m/z window 300–2000 Da, duty cycle 4-6 s (10–14 cycles/min), and five MS/MS per cycle. MS/MS were converted to .mgf files and searched against the human ERK1 sequence using MASCOT (v. 1.9), with no enzyme specified. Mass tolerances were 2.5 Da (parent) and 1.2 Da (fragment) for QStar datasets, and 1.2 Da (parent) and 0.8 Da (fragment) for LTQ-Orbitrap datasets. Peptides with MOWSE score greater than 20 were validated by manual analysis of MS/MS spectra. LC-MS of ERK1 was then carried out in water on the QStar, and only peptides with validated MS/MS that were also observed in the LC-MS run were reported.

HX-MS measurements of protection by Mg⁺²-AMP-PNP were carried out as above, with the following modifications. Prior to the HX reaction, 8 μL (5 μg) of 0P-ERK1 or 2P-ERK1 was incubated on ice for 10 min with either 7 μL of 21 mM MgCl₂ + 2.1 mM AMP-PNP or 7 μL of 21 mM MgCl₂. At t=0, exchange was initiated by adding 85 μL of a D₂O/solute mixture to the kinase solution to reach a final buffer composition of 85% (v/v) D₂O, 25 mM KPO₄ (pH 7.4), 50 mM KCl, and 0.25 mM DTT, 10 mM MgCl₂, and ± 1 mM AMP-PNP. Samples were incubated between 8 s and 110 min at 10 °C to allow in-exchange, then quenched by acidification with 90 μL 25 mM sodium succinate, 25 mM sodium citrate (pH 2.4), and cooled to 0 °C. Proteins were digested by addition of pepsin, and peptides were analyzed on the QStar mass spectrometer. Time-zero controls were performed by quenching the sample prior to addition of the D₂O/solute mixture. To avoid systematic variations, time points and the order of sample analysis ± AMP-PNP were randomized.

Weighted average mass measurements were facilitated by using in-house HX-Analyzer software [15]. Samples were randomized to control for systematic variations, and replicate runs were performed at 0 s and 60 s in order to quantify the average standard deviation of weighted average mass (0.05 Da) and maximum standard deviation (0.13 Da). Estimates of artifactual in-exchange and back-exchange were obtained as described [18,19]. Back-exchange was estimated in two ways. First, back-exchange was calculated for each peptide using an empirical formula which assumes 1% back-exchange per minute after quenching [Eqn. (4) in reference 18]. Second, back-exchange was estimated by desalting and then lyophilizing peptides generated from pepsin digestion of 2P-ERK1, incubating with 90% D₂O for 90 min at 90 °C to allow complete in-exchange, and measuring peptide mass by LC-MS under HX experimental conditions [18,20]. When counting the number of exchangeable amides for a given peptide, we included the first amide proton and did not assume rapid back-exchange on the timescale of the experiment. The two methods for estimating back-exchange were in reasonable agreement, where the measured values yielded an average of 28 ± 7 % back-exchange for 21 observable peptides, while the empirical calculation yielded an average of 22 ± 2 % back-exchange for the same peptides. Therefore we applied the second method involving back-exchange calculations to all peptides.

After correcting weighted average masses for artifactual in-exchange and back-exchange, deuteration time courses were fitted by nonlinear least squares to the equation Y = N − Ae^−k1t − Be^−k2t − Ce^−k3t, where Y is the number of deuterons exchanged at time t; A, B, and C are the numbers of backbone amides exchanging at rates k₁, k₂, and k₃, respectively; and N is the maximal deuteration over the experimental time period (N=A+B+C). Subtracting N from the total number of exchangeable backbone amides yields NE, the number of amides that are non-exchanging over the experimental time period. Nonlinear least squares calculations and graphics of time courses were generated using Sigmaplot v9.0 (SPSS Inc.).

Hydrogen exchange measurements

Hydrogen-deuterium in-exchange measurements of ERK1 were obtained by LC-MS as described in Materials and Methods. Fifty-five peptides were identified after pepsin proteolysis, ranging in length between 6 and 36 residues (Table S1). These accounted for 98 % coverage of the primary sequence (370 of 379 residues), and 94 % coverage of exchangeable amides (334 of 356 amides). The peptides are annotated onto the primary sequence of His₆-ERK1 in Fig. 1.

Information about the degree of HX occurring within localized regions of 0P-ERK1 was assessed by measuring the extent of deuteration into different peptides over the 4 h timescale of the experiment, and mapping these against an X-ray structure determined for ERK1 [5] (Fig. 2A, Table S2). Slow exchange, leading to deuteration of less than 25% of exchangeable amides, was observed within core regions of the kinase corresponding to helix αE and strand β6 within the C-terminal lobe, the latter which contains the catalytic base within the active site. Slow exchange was also observed in the loop between helix αC and β4 as well as in strand β5, which contain residues that create a hydrophobic core within the N-terminal lobe. In ERK2, these residues form intramolecular connectivities with distal residues that in turn prevent autophosphorylation and autoactivation of the enzyme by constraining flexibility at the activation lip [21]. Fast exchange, leading to >75% deuteration of exchangeable amides, was observed within loop regions, including part of the activation lip and loops connecting the MAP kinase insert with C-terminal helices αG and αH. Deuteration exceeding 50% of exchangeable amides was also observed within N-terminal loop regions, the P+1 substrate binding loop, and the MAP kinase insert.

The extent of deuteration in ERK1 revealed many similarities with previous HX measurements of ERK2 [9,11] (Fig. 2A,B). Both enzymes showed low levels of exchange within the core regions of the N- and C-terminal domains, including the active site, helices αE and αF in the C-terminal domain, and the N-terminal core containing β3–αC–β4–β5. In both enzymes, high levels of exchange occurred within loops and peripheral secondary structure including the N-terminal strands β1LO-β2LO and β1-β2, intervening loop regions between β4-β5 and αF-αG, and the MAPK insert. Interestingly, some dissimilarities in HX behavior were observed between ERK1 and ERK2. Several prominent regions that underwent nearly complete in-exchange in ERK2 showed only partial exchange in ERK1; these included the activation lip and C-terminal helix αL16, which showed less than 50% exchange in ERK1, but greater than 90% exchange in ERK2. These differences in HX suggested lower conformational mobility in ERK1 compared to ERK2 within regions of the molecule linked to catalytic function.

Regulation of HX by ERK1 activation

Time courses comparing 0P-ERK1 to 2P-ERK1 (Suppl. Fig. 2) revealed significant changes in hydrogen exchange behavior following activation of ERK1 by dual phosphorylation. Sites of proteolytic cleavage were nearly identical between the 0P and 2P forms, with the exception of the activation lip, where cleavage between residues Glu203-Tyr204 (numbered as in Fig. 1) was observed only in the unphosphorylated form of ERK1. In total, eight peptides exhibited altered in-exchange behavior between 0P-ERK1 and 2P-ERK1 (Fig. 1, Figs. 3A-H). In 7 of 8 cases, kinase activation led to decreased HX rates; these included peptide 180-187 (LKICDFGL), containing the conserved DFG motif; peptide 210-216 (YRAPEIM), containing the P+1 substrate binding loop; peptides 218-232 (NSKGYTKSIDIWSVG) and 236-254 (AEMLSNRPIFPGKHYLDQL), containing αF-αG; peptides 255-271 (NHILGILGSPSQEDLNC) and 280-306 (YLQSLPSKTKVAWAKLFPKSDSKALDL), containing the αL14 MAP kinase insert and αH; and peptide 349-365 (AMELDDLPKERLKIELIF), containing the C-terminal helix αL16. These could either reflect structural changes leading to solvent inaccessibility, or else decreased flexibility and internal motions, following activation lip phosphorylation. Only one peptide exhibited increased HX upon activation; this region corresponded to the β7–β8 loop (peptide 173-180, LINTTCDL), which forms part of a docking motif binding site (Fig. 3A), and might reflect increased backbone flexibility within a binding pocket for high affinity interactions with substrates, MKKs, and other effectors.

The localized regions which showed HX alterations upon kinase activation were mapped against the backbone structure of ERK1, and compared to those previously reported for ERK2 (Figs. 2C,D). ERK1 and ERK2 showed some similarities in the patterns of HX that were correlated with activity. Both enzymes showed activation-induced changes in HX within the activation lip, a region which undergoes significant conformational remodeling in ERK2, which would also be expected to occur in ERK1. Decreased HX was observed in both enzymes at the C-terminal αL16 helix and surrounding regions, which in ERK2 has been shown to form an interface for dimerization. Similarly, decreased HX was observed in both enzymes within the P+1 substrate interaction loop, as well as helices αF and αG, which comprise the likely interface for extended substrate binding interactions. ERK2 showed few structural changes within this region that might account for the observed changes in HX, suggesting that reduced hydrogen exchange rates reflected decreased conformational mobility following kinase activation. Decreased HX had also been observed in this region following activation of p38α MAP kinase [15], suggesting that decreased flexibility in the substrate binding interface is a feature common to several MAP kinases.

ERK1 and ERK2 also showed key differences in their patterns of HX regulated by activation. For example, ERK1 showed no increased HX in the MAP kinase insert, which had previously been observed in ERK2 and ascribed to disrupted hydrogen bonding with activation lip residues upon kinase activation [9]. Likewise, ERK1 did not recapitulate the increased HX observed in the glycine-rich loop upon activation of ERK2, a region implicated in nucleotide binding. Importantly, differences were noted between the two enzymes in the hinge region localized between the N- and C-terminal domains, which in cAMP-dependent protein kinase corresponds to the pivot point for rigid body rotation and interdomain closure needed for formation of the catalytic site. In ERK2, this region of the hinge (METDL, Suppl. Fig. 3) showed significant enhancement of HX upon activation, which correlated with altered internal motions, allowing the kinase to switch from an open to closed solution conformation [9-11]. In contrast, peptide 125-141 (METDLYKLLKSQQLSND), peptide 125-145 (METDLYKLLKSQQLSNDHICY), and the overlapping peptide 130-145 (YKLLKSQQLSNDHICY) exhibited no change in HX behavior between active and inactive states in ERK1, from which we conclude that no change in HX occurs within the corresponding METDL hinge sequence in ERK1 (Fig. 3I,J, Suppl. Fig. 3). This suggested that hinge flexibility is not regulated by activation of ERK1 as it is in ERK2, and that the two enzymes differ with respect to the regulation of interdomain closure.

Domain closure in ERK1 can be monitored by HX-MS

The results above suggested that unlike ERK2, the hinge flexibility of ERK1 might not be controlled by activity state, and thus might allow interdomain closure in both inactive and active forms. In order to test this hypothesis, we conducted an HX-MS experiment which measured the steric protection from HX upon AMP-PNP binding to ERK1. Prior studies with ERK2 have shown that AMP-PNP protects N-terminal and hinge regions in the inactive and the active kinase, as expected from the known interactions between kinases and ATP, and that active ERK2 shows additional protection of the conserved DFG motif in the C-terminal domain. Thus, steric protection of the DFG motif reflects interdomain closure between N- and C-terminal lobes in solution. If our hypothesis were correct, we would expect to see protection of the DFG motif comparable in both forms of ERK1.

We measured HX time courses in the presence or absence of 1 mM Mg²⁺-AMP-PNP, comparing inactive vs active ERK1 (Suppl. Fig. 4). Only twenty of the 54 peptides were observed in these experiments between each of the four conditions, which nevertheless included most of the ERK1 sequence. Of these, four peptides showed significant reduction in HX rates upon AMP-PNP binding (Fig. 4A-D). Three of these corresponded to regions which are known to form contacts with nucleotide within the N-terminal domain. These included peptide 46-69 (QYIGEGAYGMVSSA), containing the glycine loop which forms hydrogen bond interactions with β and γ-phosphate groups of ATP and forms a flexible gate for nucleotide binding; peptide 60-88 (YDHVRKTRVAIKKISPFEHQTYCQRTLRE), containing conserved Lys and Glu residues in strand β3 and helix αC, respectively, which form ion-pair interactions with ATP α and β-phosphate groups; and peptide 115-124 (MRDVYIVQDL), containing residues in strand β5 which form hydrogen bonding interactions with nucleotide (Fig. 4A-C). The fourth peptide, peptide 181-187 (KICDFGL), contains the conserved DFG motif (Fig. 4D). Here, we observed comparable protection upon AMP-PNP binding to both inactive and active kinase forms. This was in contrast to the corresponding DFG region in ERK2, where previous measurements revealed 10-fold greater nucleotide protection from HX in the active kinase compared to the inactive form (Suppl. Fig. 5) [11]. Our measurements of ERK1 indicate that that nucleotide interacts productively with both the N-and C-terminal domains in 0P- and 2P-ERK1, and more productively with the C-terminal domain in 0P-ERK1 than 0P-ERK2. Taken together, the findings suggest that both inactive and active forms of ERK1 are able to adopt interdomain interactions leading to a closed conformation in solution.