Supporting Information Masterson et al. 10.1073/pnas.1102701108 SI Materials and Methods Production of Synthetic Peptides. Peptides were synthesized using standard Fmoc chemistry on a microwave synthesizer (CEM Corp.), starting with N-α-Fmoc-L-glutamic acid γ-t-butyl ester on polyethylene glycol-polystyrene resin (0.4 g, 0.5 mmol g). Side-chain protecting groups were 2,2,5,7,8-pentamethylchroman-6-sulfonyl for arginine, N ω -triphenylmethyl for asparagine and glutamine, tert-butyl ester for glutamic acid, and tert-butyl ethers for serine, threonine, and tyrosine. Deprotection of the resin-bound peptide was done using Reagent K (82.5% TFA, 5% phenol, 5% thioanisole, 2.5% 1,2-ethandiol, and 5% water) for 3 h at 298 K. The resin mixture was washed three times (2 ml each) using the same cocktail, and filtrate was collected. The peptide was precipitated overnight at 273 K in 80 ml of diethyl ether, then collected by centrifugation, and washed three times with 30 ml of diethyl ether. The crude peptide was then purified by preparative HPLC using a Waters C18 reversed-phase cartridge (2.5 10 cm, 15 μm, 300 Å) with 0.1% TFA and CH 3 CN as eluents with detection at 220 nm. A linear gradient of 100 0 to 70 30 over 30 min at 5 ml min was used. Purities of pooled fractions were >97% as determined by analytical HPLC using a Vydac C18 column (0.46 25 cm) and confirmed by electrospray ionization time-of-flight mass spectrometry; for protein kinase inhibitor peptide PKI 5 24, a calculated m z of 2,222.4, m z 2,222.1 found. Stock aliquots were prepared in water and lyophilized. Concentrations of these solutions were verified by amino acid analysis. NMR Spectroscopy. NMR buffer consisted of 90 mm KCl, 20 mm KH 2 PO 4 (ph ¼ 6.5, uncorrected for deuterium istope effects), 10 mm octanoyl-n-methylglucamide, 12 mm adenosine 5 - (β,γ-imido)triphosphate (AMP-PNP), 20 mm DTT, and 5% D 2 O. Mapping of slow conformational interconversion rates (R ex ) along the amide backbone was performed using methods established by Wang et al. (1). Briefly, a modified transverse relaxation-optimized spectroscopy sequence allowed the detection of the α, β, or the longitudinal two-spin order magnetization (I zz ) during a Hahn echo period set to 2 J NH (10.8 ms). The detection of these states allowed the measurement of the cross-correlated relaxation rates between dipole dipole relaxation and chemical shift anisotropy (η xy ), and the in-phase nitrogen amide transverse relaxation rate (R 2α ), which can be used to measure R ex using the following relationship (2): R ex C zz lnðρ zz ÞþC β lnðρ β Þ; [S1] where C zz ¼ð2τÞ 1, C β ¼ðκ 1Þð4τÞ 1, κ ¼ 1 2 ln ρ zz ln ρ β, ρ zz ¼ I zz I α, and ρ β ¼ I β I α. Three 2D experiments were recorded in duplicate and in an interleaved manner to obtain the intensities I zz, I α, and I β. The value hκi was obtained from the mean of all amide resonances within one standard deviation (trimmed mean) and did not exhibit chemical exchange. Due to site to site variations in magnitude and orientation of the chemical shift anisotropy tensor, fluctuations of R ex distributed around zero are typical for nonexchanging residues (1). However, this method does not require the high pulsing rates inherent to Carr Purcell Meiboom Gill-based sequences which have been shown to cause instabilities when used on cryogenic probes for samples containing high salt concentrations (3). Molecular Dynamics Simulations. Several different simulations were performed, including the catalytic subunit of protein kinase A (PKA-C) in its apo form, binary form with ATP, ternary form with the inhibitor PKI 5 25, and ternary form with the substrate (corresponding to the cytoplasmic domain of phospholamban, PLN 1 20 ). To study the effects of Mg 2þ, PKA-C with one Mg 2þ bound was also simulated, including binary and ternary complexes. The starting structure for the apo form came from the crystal structure Protein Data Bank ID (PDB) 1CMK, from which a PKI peptide was removed. The binary complex and PKI-bound complex came from PDB ID 1ATP. The ternary complex with PLN was obtained from an NMR docking study, which was obtained from chemical shift perturbations. During the docking, the PKA-C structure was taken from PDB ID 1ATP and limited flexibility was allowed for the backbone atoms. The backbone rmsd between 1CMK and 1ATP is around 0.3 Å, whereas the PLN-bound PKA structure has an rmsd of 0.6 Å with respect to 1ATP. In total, seven different simulations were performed. Each system was solvated in a cubic box of 80 80 80 Å 3. Counter ions (K þ and Cl ) were added to reach neutrality and an ionic strength of approximately 150 mm. Mg 2þ ions were not added to the solvent because negligible free Mg 2þ would be present under low Mg 2þ conditions used for NMR analysis (Mg 2þ :AMP-PNP was 1 1.2). All simulations were set up using CHARMM c36a1 (4) and production dynamics were performed with parallel and scalable program NAMD 2.7b1 (5). CHARMM27 force field (with 2D dihedral energy grid correction map) was used to describe all the molecular interactions with the TIP3P water model. After initial minimization, the system was gradually heated from 10 to 310 K every 30 K using 10 ps NPT simulations at each temperature. During minimization and heating, harmonic restraints were kept on nonwater and noncounter ions and gradually decreased from 25 to 3 kcal mol Å 2. The harmonic restraints force constant (3 kcal mol Å 2 ) was applied to all heavy atoms of proteins during the initial 5-ns simulations in NAMD and removed afterward. All simulations in NAMD were performed at 310 K employing Langevin dynamics with a damping constant of 1.0 ps 1 and at 1 atm with Nosé-Hoover Langevin piston pressure control. Long-range electrostatic interactions were accounted by particle-mesh Ewald summation. A cutoff of 11 Å was used when evaluating nonbond interactions. Shake algorithms were applied to all bonds involving hydrogen atoms. The equations of motions were integrated using a reversible reference system propagator algorithm multiple step scheme with a time step of 2 fs. Each system was simulated for 75 ns and harmonic restraints were applied to backbone atoms in the first 5 ns. We analyzed the rmsd of individual components of the systems as a function of time (Fig. S4A). The initial structure after heating up from each simulation was used as a reference. To study the fluctuations of the PKA along the sequence, we used root mean square fluctuations (rmsf), which is the time average of rmsd for each residue (with reference to the average structure) during the simulation (Eq. S2). The last 40-ns trajectory (saved every 1 ps) was used for rmsf analysis. rmsfðiþ ¼ 1 T qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðx T i ðtjþ x i;avg Þ 2 ; [S2] t j ¼1 where T is the total number of structures used and x iavg is the average coordinates for residue i after superimposing all backbone atoms. 1of7
The rmsd reports the overall fluctuations compared to the starting structure. The rmsds of each simulation are shown in Fig. S4A. The backbone rmsd of PKA-C oscillated approximately 2.2 Å for apo form and slightly decreased to below 2 Å in the binary and ternary forms. The inhibitor PKI-bound ternary complex had a lower rmsd (approximately 1 Å) compared with the substrate PLN bound (approximately 2 Å) during the simulation. The rmsd for the ligand ATP is less than 1 Å in both the binary and ternary PKI forms and increased slightly in the ternary PLN case. The rmsd for ATP increases in all cases in the presence of one Mg 2þ. The PLN backbone rmsd (approximately 3 Å) is much smaller after Mg 2þ removal, indicating reduced fluctuations. The other rmsd values are similar. The backbone fluctuations of each residue around its average position are characterized by its rmsf value, shown in Fig. S4 B and C. The rmsf values are independent of the reference structure used in the calculation. Using the starting structure or the last snapshot as a reference gives almost identical rmsf values. In Fig. S4 B and C, we compared the fluctuations of the apo form, the binary and ternary forms (both PKI- and PLN-bound) in presence of two Mg 2þ in the simulations. Within each simulation, the rmsf values were generally large for unstructured loops compared with α-helices and β-sheets. Overall, the apo form of PKA-C has the largest fluctuations and the inhibitor (PKI) bound form of PKA-C is most stable (lowest rmsf values). ATP-bound and PLN-bound forms of PKA-C are in the middle with similar fluctuations. Apo form remains the most flexible, whereas binary, PKI-bound, and PLN-bound have similar fluctuations due to increase in PKI-bound PKA-C. Although difference can be found between different binding partners, the relative order is dependent on the sequence positions. In order to extract the large-scale motions, principal component analysis (PCA) was applied to PKA-C backbone atoms (N, Cα, C, and O) from the molecular dynamics (MD) trajectory. It has been widely used to study conformational space sampling of proteins using MD simulations trajectories, X-ray structure, or NMR structures (6, 7). We chose MD simulations of PKA-C bound to ATP PLN 1 20 in the presence of two Mg 2þ ions to perform PCA. Snapshots saved every 10 ps (7,500 snapshots) were used to construct a covariance matrix C, as in Eq. S3: C ij ¼hðx i hx i iþðx j hx j iþi: [S3] Prior to analysis, translational and rotational motions were removed by overlaying backbone atoms of two helical regions in the big lobe, including helix E (residue 140 160) and helix F (residue 217 233). In addition, the N- and C-terminal residues were excluded in the PCA analysis and only backbone atoms from residue 50 to 300 were used. By diagonalizing the covariance matrix C, a set of eigenvalues P and the corresponding eigenvectors V are obtained. P i corresponds to mean displacement along directions v i. The first several principal components (PCs), ranked according to their eigenvalues, account for a large percentage of the fluctuations. The first two PCs are shown in Fig. S4D, accounting for approximately 60% of the overall fluctuations. PC1 corresponds to the open and close motion between the small lobe and big lobe. PC2, instead, depicts a shearing motion of the small lobe orthogonal to the open close motion. The projection ProjðM;PC i Þ of any structure (trajectory) M onto the ith PC was calculated using Eq. S4: Proj½M;V i Š¼M bb V i ; [S4] where M bb is the backbone coordinates after overlaying M with the reference structure using the same two helix. We used the R and its Bio3d library for the PCA analysis (8). Because we used the large lobe helices as a reference, the large lobe did not show large motions in each PC. When all backbone atoms are used for PCA and the covariance prepared by superimposing all backbone atoms, the PCA results are much noisier. The first five PCs accounts for nearly equal amount of variance (approximately 10%), with the first 10 PCs accounting for about 55% (Fig. S4D). Although visual inspection clearly showed the existence of open close between the small and big lobe, the N terminus showed large fluctuations in the first several PCs. A similar choice of references for PCA analysis has also been reported in the literature (6). 1. Wang C, Rance M, Palmer AG, 3rd (2003) Mapping chemical exchange in proteins with MW > 50 kd. J Am Chem Soc 125:8968 8969. 2. Tatulian SA, Jones LR, Reddy LG, Stokes DL, Tamm, LK (1995) Secondary structure and orientation of phospholamban reconstituted in supported bilayers from polarized attenuated total reflection FTIR spectroscopy. Biochemistry 34:4448 4456. 3. Tamm LK, Liang B (2006) NMR of membrane proteins in solution. Prog Nucl Magn Reson Spectrosc 48:201 210. 4. Brooks BR, et al. (2009) CHARMM: The biomolecular simulation program. J Comput Chem 30:1545 1614. 5. Phillips JC, et al. (2005) Scalable molecular dynamics with NAMD. J Comput Chem 26:1781 1802. 6. Henzler-Wildman KA, et al. (2007) A hierarchy of timescales in protein dynamics is linked to enzyme catalysis. Nature 450:913 916. 7. Yang L, Song G, Carriquiry A, Jernigan RL (2008) Close correspondence between the motions from principal component analysis of multiple HIV-1 protease structures and elastic network modes. Structure 16:321 330. 8. Grant BJ, Rodrigues AP, El Sawy KM, McCammon JA, Caves LS (2006) Bio3d: An R package for the comparative analysis of protein structures. Bioinformatics 22:2695 2696. Fig. S1. Sequences of PLN 1 20 and PKI 5 24. The seven residue canonical recognition sequence (X-X-R-R-X-S/T-Φ, where X is any amino acid and Φ has hydrophobic character) is highlighted. 2of7
Fig. S2. (A C) Overlay of normalized displacements of C α atoms (derived from X-ray crystallography) and NMR chemical shift perturbations. The overall agreement in the trends from both indicate that the NMR data, in part, probes the structural transitions observed by X-ray crystallography. The perturbations probed are (A) apo nucleotide bound, (B) nucleotide bound nucleotide PLN 1 20, and (C) nucleotide bound nucleotide PKI 5 24 bound. (D G) Overlay of NMR spectra of the PKA-C in different forms. The overlays are done such that they can be used to compare the transition of apo nucleotide bound nucleotide substrate (or inhibitor) bound. 3of7
Fig. S3. (A and B) NMR analysis of picosecond to nanosecond dynamics in the PKA-C as function of Mg 2þ ions. Heteronuclear NOE (H-X NOE) values show a systematic increase in the presence of high Mg 2þ (black data points), which is accompanied by longer T 2 times. The enzyme also becomes more compact and tumbles faster in solution, as reflected by the decrease T 1 T 2 values. (C E) Inverse peak heights of PKA-C when bound to (C) AMP-PNP PLN 1 19,(D) AMP-PNP PKI 5 24 with low Mg 2þ, and (E) AMP-PNP PKI 5 24 with high Mg 2þ. The trend in slow dynamics decreases from C to E. 4of7
Fig. S4. (A) Backbone and all-atom rmsd of PKA-C, ATP, PKI, and PLN during nine different simulations. In each simulation, the number of Mg 2þ ions during the simulation is specified. The backbone atoms include heavy atoms N, Ca, C, and O. For ATP, all atoms are used to calculate the rmsd values. (B and C) Rmsf of backbone atoms for each residue in simulations bound to different ligands with (B) 1 Mg 2þ or (C) 2 Mg 2þ. The rmsf values for each residue are averaged over the last 40 ns of the trajectory. (D F) PCA of the MD trajectories of PKA-C bound to ATP PLN 1 20 in the presence of two Mg 2þ ions. (D) Plot describing the contribution to the overall motions based on each eigenvalue. (E and F) Domain motions described by the first two principle components correspond to opening and closing of the active site (PC1) and shearing between the two lobes (PC2) of the enzyme. The crystal structures corresponding to open (1CMK), intermediate (1BX6), and closed (1ATP) states are shown in black, red, and green, respectively. 5of7
Table S1. Thermodynamic parameters of PKA-C binding to the substrate, PLN 1 19, or the inhibitor, PKI 5 24 ΔH, kcal mol ΔS, kcal mol T TΔS, kcal mol Kd, μm PLN 1 20 to apo-pka 0.6 ± 0.1 0.017 ± 0.001 5.1 ± 0.2 49 ± 8 PLN 1 20 to ADP:PKA 1.2 ± 0.1 0.019 ± 0.001 5.7 ± 0.1 10 ± 4 PKI 5 24 to apo-pka 12.2 ± 0.3 0.0145 ± 0.0003 4.3 ± 0.1 1.8 ± 0.4 PKI 5 24 to ADP:PKA 12.3 ± 0.5 0.0089 ± 0.0003 2.4 ± 0.1 6 10 2 3 10 2 Parameter Table S2. Nuclear spin relaxation parameters on PKA-C in the free-state and bound to ligands Apo AMP-PNP bound (low Mg 2þ ) AMP-PNP PLN 1 20 bound (low Mg 2þ ) AMP-PNP PKI 5 24 bound (low Mg 2þ ) AMP-PNP PKI 5 24 bound (high Mg 2þ ) T 1, s 1.55 ± 0.18 1.47 ± 0.19 1.35 ± 0.38 1.35 ± 0.48 1.19 ± 0.21 Trimmed T 1, s 1.58 ± 0.14 1.49 ± 0.15 1.41 ± 0.32 1.33 ± 0.25 1.21 ± 0.10 T 2, s 0.036 ± 0.002 0.029 ± 0.005 0.033 ± 0.003 0.031 ± 0.001 0.038 ± 0.006 Trimmed T 2, s 0.019 ± 0.001 0.020 ± 0.003 0.025 ± 0.002 0.024 ± 0.009 0.029 ± 0.007 Trimmed T 1 T 2 82.05 ± 0.10 74.50 ± 0.18 57.09 ± 0.24 55.42 ± 0.42 41.75 ± 0.31 τ c, ns 19.8 ± 0.7 15.6 ± 0.5 16.1 ± 0.6 14.2 ± 0.8 H-X NOE 0.71 ± 0.32 0.74 ± 0.28 0.70 ± 0.32 0.75 ± 0.28 0.79 ± 0.23 Trimmed H-X NOE 0.76 ± 0.13 0.78 ± 0.13 0.74 ± 0.16 0.79 ± 0.13 0.81 ± 0.10 % of residues with R ex > 8 Hz 5 16 13 11 5 H-X NOE, steady-state 1H/15N heteronuclear NOE. Table S3. PDB accession numbers for the structures used to define the d S53-G186 conformational space State Open Closed Intermediate PDB ID 1CMK 1CTP 1J3H 1SYK 2QVS 1APM 1ATP 1CDK 1FMO 1JBP 1L3R 1RDQ 1U7E 2CPK 1BKX 1BX6 1JLU 1RE8 1REJ 1REK 1STC 1YDR 1YDS 1YDT 3IDB 3IDC 3DND 3DNE 6of7
Table S4. Average and standard deviation of backbone heavy atom rmsds during the last 20 ns of MD simulations for each form of PKA-C Open Intermediate Closed Apo no Mg 2þ 2.18±0.25 2.72±0.23 2.83±0.26 ATP bound 1 Mg 2þ 2.52±0.13 1.71±0.18 1.98±0.20 ATP bound 2 Mg 2þ 2.61±0.11 1.87±0.11 2.03±0.14 ATP/PLN bound 1 Mg 2þ 2.26±0.12 1.81±0.14 1.97±0.17 ATP/PLN bound 2 Mg 2þ 2.41±0.09 1.79±0.13 1.90±0.16 ATP/PKI bound 1 Mg 2þ 2.39±0.09 1.93±0.15 2.14±0.17 ATP/PKI bound 2 Mg 2þ 3.13±0.12 1.74±0.10 1.61±0.08 Representative crystal structures from each form of PKA-C, are used as reference state to measure these values. Values in bold indicate the forms with the best fit. 7of7