Dynamically committed, uncommitted, andquenched states encoded in protein kinaseA revealed by NMR spectroscopyLarry R. Mastersona,1, Lei Shib,1, Emily Metcalfeb, Jiali Gaob, Susan S. Taylorc,2, and Gianluigi Vegliaa,b,2

aDepartments of Biochemistry, Molecular Biology, and Biophysics, and bChemistry, University of Minnesota, Minneapolis, MN 55455-0431; andcDepartment of Chemistry and Biochemistry, University of San Diego, San Diego CA 92093-0654

Contributed by Susan S. Taylor, February 18, 2011 (sent for review October 19, 2010)

Protein kinase A (PKA) is a ubiquitous phosphoryl transferase thatmediates hundreds of cell signaling events. During turnover, itscatalytic subunit (PKA-C) interconverts between three major con-formational states (open, intermediate, and closed) that are dyna-mically and allosterically activated by nucleotide binding. We showthat the structural transitions between these conformational statesare minimal and allosteric dynamics encode the motions from onestate to the next. NMR and molecular dynamics simulations definethe energy landscape of PKA-C, with the substrate allowing the en-zyme to adopt a broad distribution of conformations (dynamicallycommitted state) and the inhibitors (high magnesium and pseudo-substrate) locking it into discrete minima (dynamically quenchedstate), thereby reducing the motions that allow turnover. Theseresults unveil the role of internal dynamics in both kinase functionand regulation.

allostery ∣ cooperativity ∣ phospholamban ∣ substrate recognition ∣intrinsically disordered proteins

Posttranslational phosphorylation is among the most commonmechanisms of cell signaling both in eukaryotes and prokar-

yotes (1). Phosphorylation is orchestrated by kinases, which areinvolved in many vital cellular functions including metabolism,growth, and cell differentiation, and they target substrates loca-lized in several compartments, including cytoplasm, mitochon-dria, plasma membrane, sarcoplasmic reticulum membrane,nucleus, microtubules, and actin filaments (2). The human ki-nome (the collection of all human protein kinases) accounts forapproximately 2% of the entire genome and therefore encom-passes the largest family of enzymes (3).

In addition to differential expression in various cellular sites,kinases are activated or deactivated and localized by cofactorsand ancillary regulatory proteins to achieve precise control overtime and space. Protein kinase A (PKA) is considered the pro-totype for the protein kinase family (4). PKA exists as an inactiveheterotetrameric assembly with two catalytic subunits (PKA-C)bound to a dimer of regulatory (R) subunits. PKA-C is unleashedupon β-adrenergic stimulation, which disassembles the heterote-tramer. The R subunits are responsible for the localization ofPKA-C via interactions with A kinase anchoring proteins toachieve spatial control (5).

The heat stable protein kinase inhibitor (PKI) also controlsPKA activity and localization (5). This small protein comprisesa high-affinity pseudosubstrate region that binds competitivelyto the substrate binding groove, as well as a nuclear export signal,directing PKA to locations outside the nucleus (6). Binding ofPKI to PKA-C forms a complex in which the enzyme is poisedfor catalysis (correct positioning of atoms), but remains lockeddue to the lack of a hydroxyl group acceptor (3).

The bean-shaped fold of PKA-C is highly conserved (4), withtwo lobes (small and large) undergoing structural rearrangementsduring the substrate recognition and product release steps ofcatalysis (3). Three major conformational states of PKA-C havebeen identified by X-ray crystallography along various stages

of the catalytic cycle (3): open (apo form), intermediate (binaryform), and closed (ternary complex). NMR dynamic measure-ments (7) linked the major conformational states along the reac-tion coordinates, showing the critical role of the enzyme’s internaldynamics (the equilibrium fluctuations that allow the explorationof the free energy landscape) for protein kinase function. Speci-fically, the rates of conformational fluctuations are correlatedwith the transition from closed to open conformations. Thesefluctuations are synchronous with the enzyme rate-limiting step(product release), underscoring the prominent role of conforma-tional dynamics in substrate recognition and catalysis (7).

Understanding how PKA-C interacts with both substrates andinhibitors from both a structural and dynamic perspective willdefine general criteria for activation and deactivation of proteinkinases, with obvious repercussion in the design of new drugs.Here, we examined the effects of inhibitors on the conforma-tional dynamics of the enzyme: a competitive peptide inhibitorcorresponding to the inhibitory region of PKI (PKI5–24) (8) andmagnesium, which under high concentrations behaves kineticallyas a noncompetitive inhibitor of PKA-C (9, 10). The combinationof high Mg2þ concentrations with PKI has implications for thecellular control of PKA-C activity to arrest transcription duringmitosis (11, 12). By comparing the results with those obtained inthe presence of a peptide substrate (7) which competes withPKI5–24 at the binding groove of PKA-C, we found that bothPKI5–24 and excess Mg2þ restrict the enzyme dynamics on a fast(picosecond to nanosecond) and slow (microsecond to millise-cond) NMR timescale without drastically changing the conforma-tion of the ternary complex. Inhibitor binding modifies the energylandscape by restricting the motions of the enzyme backbone(13). These findings unveil a relatively unexplored role of mag-nesium in protein kinase regulation and establish a paradigm forthe design of protein kinase inhibitors.

ResultsThermodynamics of Binding and Enzyme Stability. To comparethe effects of binding substrates and inhibitors to PKA-C, wesynthesized two peptides: the first peptide corresponding to thecytoplasmic domain of phospholamban (PLN1–20), an endogen-ous inhibitor for the sarcoplasmic reticulum Ca-ATPase andnative substrate of PKA-C in cardiac muscle (14); and the secondpeptide corresponding to PKI5–24 (Fig. S1). Crystallographic dataand enzyme kinetic assays indicate that these peptides competefor the same binding site in PKA-C (4, 7). Enzyme thermostability

Author contributions: L.R.M., S.S.T., and G.V. designed research; L.R.M., L.S., and E.M.performed research; L.R.M., J.G., S.S.T., and G.V. analyzed data; and L.R.M., L.S., E.M.,S.S.T., and G.V. wrote the paper.

The authors declare no conflict of interest.1L.R.M. and L.S. contributed equally to this work.2To whom correspondence may be addressed. E-mail: vegli001@umn.edu or staylor@ucsd.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1102701108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1102701108 PNAS ∣ April 26, 2011 ∣ vol. 108 ∣ no. 17 ∣ 6969–6974

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

1

and binding thermodynamics experiments were carried out at lowand high concentrations of MgCl2. At low concentrations ofMgCl2, PKA-C binds one cation, the primary Mg2þ ion, essentialfor enzyme activity (9). At high concentrations of MgCl2, PKA-Cbinds a secondary Mg2þ ion, which inhibits catalytic activity in anoncompetitive manner (9).

To measure the dissociation constants for PKA-C bound tosubstrate or inhibitor peptides, we employed isothermal titrationcalorimetry (ITC). The ITC measurements reveal that PKI5–24binds to PKA-C 25 times more tightly than PLN1–20 (Fig. 1A andTable S1). This difference in affinity is amplified by ADP to 40times, reflecting the strong cooperative effect between PKI andnucleotide (15, 16). Remarkably, PKI5–24 and PLN1–20 havedifferent thermodynamics of binding (Fig. 1A). PLN1–20 bindingto PKA-C is dominated by a favorable overall entropy to thebinding free energy (ΔG). The presence of ADP increases thefavorable entropy of binding, resulting in an enhanced bindingaffinity, revealing a positive cooperativity similar to that ofKemptide (17). In contrast, the binding of PKI5–24 to PKA-C isenthalpically driven, overcoming an unfavorable entropic contri-bution to ΔG. Although the B factors from crystal structures ofPKA-C complexes containing either PLN1–20 or PKI5–24 suggestmore favorable enthalpy for binding the inhibitor (more favor-able intermolecular interactions), the structures alone are notadequate enough to predict these opposite thermodynamic driv-ing forces to binding. The presence of ADP reduces the unfavor-able entropy of binding PKI5–24, leading to a higher bindingaffinity and greater positive cooperativity (approximately 800times).

Thermostability measurements revealed that PKI5–24 con-ferred significantly greater stability to PKA-C than the substrate.Thermal melting of PKA-C bound to nucleotide, substrate, andinhibitor was monitored with CD spectroscopy (Fig. 1B) at lowand high Mg2þ concentrations. Apo-PKA-C melted at 47 °C,whereas under low Mg2þ, the nonhydrolyzable nucleotide, ade-nosine 5′-(β,γ-imido)triphosphate (AMP-PNP), shifted the melt-ing temperature (ΔTm) by approximately 1 °C and the ternarycomplex with PKI5–24 shifted ΔTm by approximately 2 °C, consis-tent with previous reports using adenosine 5'-[γ-thio]triphosphate(18, 19). However, unlike the increased stability for PKI5–24,addition of PLN1–20 to the binary complex had a negligible effecton ΔTm. High Mg2þ shifted ΔTm by approximately 1 °C for boththe nucleotide bound form and ternary complex with PLN1–20. Incontrast, ΔTm was shifted the furthest in the presence of PKI5–24

and high Mg2þ (ΔTm ∼ 4 °C), indicating a synergistic effect ofstabilizing the enzyme when both inhibitors are present.

Taken together, the calorimetry and thermostability measure-ments indicate that PKI5–24 binds to PKA-C with an oppositeenthalpic/entropic balance than PLN1–20. Whereas substratebinding is entropically driven and confers little thermostability,inhibitor binding is enthalpically driven and significantly en-hances the thermostability of the enzyme.

Chemical Shift Perturbations Map the Transitions from Open to ClosedState. The residue-specific changes in the amide backbone ofPKA-C upon binding AMP-PNP, and followed by substrate orinhibitor peptides at low magnesium were monitored by 1H∕15N-transverse relaxation-optimized spectroscopy (TROSY)-hetero-nuclear single quantum coherence NMR spectroscopy. As de-monstrated with the seven-residue peptide, Kemptide (17), theoverall trend in chemical shift changes from these titrationscorrelate with the displacement of Cα atoms observed by X-raycrystallography (Fig. S2 A–C). However, the chemical shift per-turbations were generally small (hΔδi ∼ 0.04 ppm), with AMP-PNP binding accounting for the majority of the differences(Fig. 1C and Fig. S2 D–G). Both PLN1–20 and PKI5–24 bindingto the binary complex gave similar Δδ relative to the apo state(hΔδi ∼ 0.04 ppm; Fig. 1 C and D). Binding of a secondaryMg2þ ion to the PKA-C∕AMP-PNP∕PKI5–24 complex resultedin no appreciable effects in the enzyme fingerprint as demon-strated by the correlation plot in Fig. 1E. These data indicate that(i) once the nucleotide is bound to the enzyme, only minimalstructural changes occur for the transitions from intermediateto closed state, and (ii) the conformation of the enzyme boundto the peptide inhibitor closely resembles the substrate boundconformation. However, small but significant chemical shiftdifferences (Δδ ∼ 0.01–0.04 ppm) are present in catalyticallyimportant regions of the enzyme (Fig. 1 F and G), such as theglycine-rich, activation, DFG (Asp184-Phe185-Gly186), and peptidepositioning loops, which are conserved throughout all proteinkinases (3). Strikingly, chemical shift changes were linear fromapo-PKA-C to binary and ternary complexes with substrate orinhibitor peptide, and finally to the ternary complex with inhibitorpeptide at high magnesium (Fig. 1F). This linearity betweenthe different complexes indicates that the enzyme undergoes fastexchange between the major conformations and that the popula-tions are shifted by ligand binding. Nucleotide binding shifts theconformational ensemble from the open to the intermediate

Fig. 1. Thermodynamic and NMR analysis of PKA-C. (A) Thermodynamics of PKA-C binding to substrate and inhibitor, with or without ADP. The binding ofPLN1–20 is dominated by favorable entropy, whereas PKI5–24 is enthalpy driven, overcoming an entropic penalty. (B) Melting measurements showed that PKI5–24confers the greatest thermostability to PKA-C as represented by the shift in Tm (relative to apo-PKA-C, ΔTm). HighMgþ2 confers slightly higher stability to eachcomplex. (C–E) Correlation of chemical shift perturbations (Δδ) in PKA-C between the different forms. The majority of perturbation occurs upon nucleotidebinding (C), whereas formation of the ternary complexes were quite similar to one another (D and E). (F) Linearity of chemical shifts between the apo form andthe ternary form (AMP-PNP∕PKI5–24 with high Mg2þ) is observed, indicating that the enzyme opens and closes on a fast timescale. (G) Enzymes views from theactive site surface formed by the small and large lobes.

6970 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1102701108 Masterson et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

1

state. Subsequent binding of substrate or inhibitor skews thepopulation toward the closed states. Although there is a cleardistinction between the apo and nucleotide bound forms, the dif-ferences in chemical shifts between various ternary complexes areminimal (Fig. 1F). The latter indicates that both substrate andinhibitors shift the population toward closed states, which arestructurally quite similar.

Inhibitor Binding Quenches the Dynamics in the Enzyme Backbone.The combination of thermodynamics and chemical shift pertur-bations show that the ternary complexes formed by the enzymewith substrate or inhibitor differ in thermostability, but notsignificantly in structure. To probe the internal dynamics of theenzyme, we used nuclear spin relaxation measurements for fast,picosecond to nanosecond [T1, T2, and heteronuclear (H-X)NOE], and slow, microsecond to millisecond (Rex), dynamicson the NMR timescale for each complex (20). Our data show thatthe PKA-C∕AMP-PNP∕PKI5–24 complex with two Mg2þ ionsbound is the most compact with the fastest overall tumbling rate(T1∕T2 ratios) and least flexible (increased in H-X NOEs andlonger T2 values; Fig. 2, Fig. S3 A and B, and Table S2). This find-ing is in agreement with previous fluorescence anisotropy studies(21, 22). More importantly, a comparison of the ternary com-plexes with one Mg2þ ion reveals that the PKA-C∕AMP-PNP∕PKI5–24 complex has substantially decreased picosecond to nano-second dynamics with respect to the PKA-C∕AMP-PNP∕PLN1–20

complex (Fig. 2 and Table S2). This decrease becomes moreapparent under high Mg2þ concentrations. The quenched fastdynamics for the PKA-C∕AMP-PNP∕PKI5–24 complex indicatesa decrease in conformational entropy of the enzyme. Althoughquenched dynamics cannot be directly compared to the macro-scopic methods of thermodynamics, a decrease in conformationalentropy agrees qualitatively with the trend of unfavorable overallentropy of binding and the enhanced enzyme thermostability.

The analysis of inverse peak heights at temperatures rangingbetween 22–33 °C suggests a marked decrease in conforma-tional exchange (i.e., slow dynamics on the NMR timescale) forPKA-C upon inhibitor binding (Fig. S3 C–E). We quantifiedthe microsecond to millisecond conformational exchange rates(Rex) across the enzyme backbone and found diminished valuesfor the ternary complex with PKI5–24 compared to the ternarycomplex with substrate (7) (see Materials and Methods, Fig. 2,and Table S2). Addition of the inhibitory Mg2þ ion to the ternarycomplex with PKI5–24 resulted in nearly absent microsecond tomillisecond dynamics (Fig. 2). Quenched dynamics are evidentfor regions surrounding the conserved loops at the active siteof PKA-C.

PKA-C Energy Landscape.To define the energy landscape and inter-pret the dynamics for each form of PKA-C along stages of thecatalytic cycle, we carried out molecular dynamics (MD) simula-tions of PKA-C in water using the apo, binary (nucleotide bound),and ternary complexes (containing PLN1–20 or PKI5–24, see SIMaterials and Methods for experimental setup and data analysis).In each case, the calculations were carried out in the presence ofeither one or two bound Mg2þ ions to model the low and highMg2þ conditions used experimentally (3). Rmsd plots vs. timeshowed that PKA-C structure reaches a fairly stable minimumafter approximately 20 ns (Fig. S4A), whereas root mean squarefluctuations (rmsf) confirmed that the most dynamic regions ofthe enzyme reside in the proximity of the catalytically importantloops (3, 7) (glycine-rich, activation, DFG, and peptide position-ing loops), as well as some of the structural elements such as theB helix, H helix, F helix, with significant fluctuations detected forthe N and C termini (Fig. S4 B and C).

The conformational interconversion of the enzyme throughdifferent states identified by X-ray crystallography (Table S3)were monitored using principal component analysis (PCA) (23).

The PCA results show the directionality and amplitude of proteinmotions, in which the first several principal components arecorrelated with large conformational changes. PCA analysishas been widely to interpret conformational variations observedexperimentally (24–26). Moreover, PCA analysis provides a rea-sonable method to extrapolate a relatively short MD trajectoryto provide a qualitative description of motions occurring overa longer timescale (27), such as the opening and closing eventsprobed by NMR spectroscopy (7).

We calculated the PCA for the PKA-C complex with PLN1–20,and found that the first two components account for ap-proximately 60% of variance in coordinates during the MD simu-lations (Fig. S4D). The first principal component (PC1) describesthe opening and closing of the two lobes of the enzyme, whereasthe second component (PC2) describes shearing between thelobes (Fig. 3A and Fig. S4 E and F). To probe the opening andclosing of the active site cleft (4, 7), we also monitored the intera-tomic distance between Ser53 and Gly186 (dS53-G186) (3) duringMD trajectories (Fig. 3A) with respect to PC1. This 2D plot(Fig. 3B) describes relative motion between the two lobes andthe opening and closing of the active site. As points of reference,the crystallographic structures were plotted on this graph. Thecomparison of the conformational ensemble sampled by theMD simulations and the X-ray structures representative of thethree conformational states are reported in Table S4.

The apoenzyme (orange trace) samples a broad distribution ofconformations identified by X-ray crystallography (Fig. 3B),whereas those sampled by the binary form with one Mg2þ havedistributions similar to open and intermediate conformations.Addition of the substrate (ternary complex, blue trace) does notchange PC1 observed for the binary form, but changes the distri-bution of dS53-G186 toward the closed state. In contrast, the dis-tribution of conformations for the ternary complex with PKI5–24(red trace) is shifted, where PC1 encompasses conformations si-milar to intermediate and open states. However, the dS53-G186 dis-tance distribution indicates that the enzyme’s active site is notcompletely open. This distribution between conformations is inagreement with the NMR data showing residual dynamics forthe PKA-C∕AMP-PNP∕PKI5–24 complex at low Mg2þ concen-trations.

Calculations with the secondMg2þ ion resolves the degeneracyof the conformational states and defines localized minima(Fig. 3B): The binary form of the enzyme populates inter-

Fig. 2. Backbone dynamics of PKA-C in different ternary complexes. Map-ping of (A) fast and (B) slow backbone dynamics show that, upon inhibitionwith PKI5–24 and with high Mg2þ, a decrease of picosecond to milliseconddynamics occurs throughout the backbone. For the comparison, the pre-viously published dynamics of PKA-C with the substrate PLN1–20 is shown (7).

Masterson et al. PNAS ∣ April 26, 2011 ∣ vol. 108 ∣ no. 17 ∣ 6971

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

1

mediate conformational states, the ternary complex with the con-formational space defined by the corresponding substrate hasdS53-G186 values clustered near the intermediate and closed states,and, strikingly, the ternary complex with inhibitor overlaps withthe crystallographic structures (Fig. 3B, Bottom Right).

The results from MD simulations corroborate the NMR andthermodynamics data, providing a framework to interpret theexperimental results in terms of the energy landscape. Althoughthe linchpin role of Mg2þ ions in both stabilization of the pyro-phosphate group of ATP and local rigidification of the glycine-rich loop has been previously reported (28), our simulationsreveal that the effect of Mg2þ binding is a global phenomenon,affecting the overall dynamics of the enzyme and trapping it ininert states.

DiscussionThe transitions between open and closed states in PKA-C are dri-ven by internal dynamics, which plays a major role for substraterecognition and turnover (7). The nucleotide acts as a dynamicand allosteric effector, causing a population shift of PKA-C fromopen to closed states and selecting for dynamically committedstates—which are defined as nearly isoenergetic conformationscompatible with catalysis that interconvert rapidly. The rates ofthe structural fluctuations between these states are sensitive tonucleotide binding, which increases the substrate binding coop-eratively via a conformational selection mechanism (7). Moreimportantly, the rates of the enzyme conformational fluctuationsare synchronous with the enzyme rate-limiting step (i.e., ADPrelease), underscoring the prominent role of conformationaldynamics in substrate recognition and catalysis (7). The uniqueaspect about the current study is the link between restrictedconformational dynamics and enzyme inhibition, establishing aparadigm for controlling protein kinase activity.

Based on our previous data (7, 17) and this work, we proposethat the energy landscape of PKA-C comprises dynamicallyuncommitted, committed, and quenched states. Ligand binding(via nucleotide, substrates, or inhibitors) drives the enzyme to

select these different states (Fig. 4). The apoenzyme is character-ized by a dynamically uncommitted state—it can explore theenergy landscape to access open and closed conformations, butdoes not transition between these conformations on a timescalerelevant for catalysis (7). The dynamically committed states of theprotein kinase occupy conformational space in which the activesite cleft opens and closes on a timescale optimal for turnover.These fluctuations are induced by nucleotide binding, which actsas a dynamic and allosteric activator (7), coupling the small andlarge lobes, completing the catalytic spine of the enzyme (29, 30),and preparing the active site for substrate binding. This event dif-ferentiates protein kinases from small molecule kinases, such asadenylate kinase, because the assembly of the spine architectureis a prerequisite for protein kinase activation (29, 30). Substratebound PKA-C retains sufficient conformational motions, main-taining the enzyme in a dynamically committed state for productrelease and turnover (7). Although the dynamics are on a slowertimescale than the chemical step (phosphoryl transfer), theyposition and prepare the substrate for catalysis. In contrast,dynamically quenched states of inhibited PKA-C are character-ized by narrow, deep wells with high-energy barriers betweenconformations and hinder opening/closing of the active cleft.The two inhibitors analyzed here (PKI5–24 and excess Mg2þ)raise the energy barriers of the conformational landscape, gener-ating discrete minima with quenched dynamics. This phenomen-on may explain why, in general, crystal structures of proteinkinases have been more accessible in their inhibited forms whilesubstrate bound forms have been elusive to such analyses.

More importantly, these results support a possible role ofmagnesium for regulation and localization of PKA-C in the cell.Reports indicate that Mg2þ concentrations can change depend-ing on the cell compartment (31) and phase of the cell cycle(11, 32). Specifically, nuclear concentrations of Mg2þ increasesubstantially during mitosis (11). This increase triggers bindingof PKA-C to PKI with high affinity (10), causing PKA-C to beexported from the nucleus, and arresting transcription (12). Sucha mechanism of inhibition and localization exploits the enhancedaffinity of PKI in the presence of high Mg2þ (10). Interestingly,enhanced binding affinity in the presence of excess Mg2þ wasalso reported for the type I (but not the type II) regulatorysubunit of PKA (10). Because the regulatory subunits are loca-lized differently within the cell where Mg2þ concentrations mayvary (33, 34), gradients in Mg2þ concentration would offer an-other form of localization and control. Therefore, Mg2þ canpotentially act as a rheostat for the strength of PKA regulation,dictate its compartmental localization, and influence the overalleffect on biological function.

Allosteric enzymes often exist in ensembles of conformations,where catalytic efficiency is achieved by excited states able tocross inherently low-energy barriers between major conforma-tions (35–44). Although the dynamics measured by NMR donot influence the chemical step of catalysis directly (45), it hasbeen shown that intrinsic protein dynamics is the driving forcefor crossing potential energy barriers with ligands, such as cofac-tors or substrates, activating or deactivating enzyme dynamics tomodulate biological function (35, 41, 46–55).

Although our findings are limited to the system studied here,the mechanism of inhibition via quenched dynamics may playa role in other systems. In fact, recent studies on dihydrofolatereductase (56), triosephosphate isomerase (57), ribonuclease A(58), HIV-1 reverse transcriptase (59), and imidazole glycerolphosphate synthase (60) show that slow microsecond to millise-cond motions are severely dampened when these enzymes areinhibited (56, 59, 60) or when catalytically hindering mutationsare introduced (57, 58).

Our results demonstrate that the inhibition of PKA-C ismarked by changes in intrinsic dynamics. These findings aretimely, as they suggest a mode by which inhibitors could be

Fig. 3. Comparison of MD simulations for PKA-C. (A) Global motions sug-gested by PCA analysis of MD trajectories correspond to opening and closingof the active site (PC1), which compared well with the distances betweenresidues S53 and G186 in crystal structures of open (1CMK), intermediate(1BX6), and closed (1ATP) conformations. (B) A map of the interatomicdistances vs. the PC1 from MD simulations indicate that PKA-C accessedthe major crystallographic conformations frequently, except in the presenceof inhibitors.

6972 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1102701108 Masterson et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

1

designed to modulate PKA activity. Protein kinase inhibition is animportant therapeutic avenue for treating diseases (61–63).Although many therapeutic agents targeting protein kinases arecompetitors for the nucleotide binding site, there is a growinginterest in developing inhibitors that bind at remote sites (i.e.,Abl kinase), working through an allosteric mechanism of inhibi-tion (63, 64). Although we have previously described an allostericnetwork of communication in PKA-C modulated by conforma-tional dynamics (7, 17), the current results reveal a role of thisnetwork in the enzyme inhibition. Gaining control of PKA-Cthrough competitive or allosteric modulation of dynamics is anexciting possibility for further research.

Materials and MethodsProtein Expression and Purification. PKA-C was expressed and purified fromEscherichia coli according to procedures previously published (17, 65). Theenzyme concentration was determined spectrophotometrically by A280

(A0.1% ¼ 1.2) and its activity was tested using the standard substrate Kemp-tide. Peptide synthesis of PKI5–24 was performed on a microwave synthesizerand purified using reverse-phase high-pressure liquid chromatography.

ITC Measurements. ITC data were acquired on a microcalorimeter (MicroCalInc.). Stock solutions of PKA-C, PLN1–20, and PKI5–24 were dissolved in 20 mMphosphate buffer (pH 6.5) containing 180 mM KCl and 4 mM MgCl2, anddegassed. Titrations were conducted at 27 °C using 0.1 mM PKA-C in theabsence or presence of 6 mM AMP-PNP and with a stock of synthetic peptide(1.8 mM). The samples were stirred at 410 rpm. Twenty injections were sepa-rated by 300 s of equilibration (5 μL for the first, followed by 10 μL for eachof the remaining). A one-site binding model was assumed and the data werefit using MicroCal Origin software (version 5.0).

Circular Dichroism. Concentrated stocks of PKA-C were diluted to 5 μM in CDbuffer [10 mM piperazine-N-N′-bis(2-ethanesulfonic acid), pH 7.0, 150 mMNaCl], in the presence of 0.48 (low Mg2þ, 1∶1.2 Mg2þ:nucleotide) or3.0 mM MgCl2 (high Mg2þ, 5∶1 Mg2þ:nucleotide), 0.60 mM AMP-PNP(selected samples), and 15 μM peptide (selected samples). Samples wereincubated over a range of 25–75 °C at 1 °C∕min in a rectangular quartzcuvette in a Jasco J-815 spectropolarimeter. Spectra were acquired at222 nm following an equilibration time of 10 s. A blank consisted of allreaction components except PKA-C and was subtracted from each spectrum.The data were fitted to a two-state sigmoidal unfolding model using Origin8.0 (Microcal) using Tm as the midpoint. Errors were derived from nonlinearleast squares fitting of the data.

Acquisition of NMR Data. NMR samples consisted of approximately 500 μMPKA-C, dissolved in buffer. Low (1∶1.2 Mg2þ:nucleotide) or high (5∶1Mg2þ:nucleotide) Mg2þ conditions were done under 10 or 60 mM Mg2þ, re-spectively. Experiments were carried out on a Varian instrument operatingat 800.29 MHz 1H Larmor frequency at 33 °C. The data were processedwith NMRPipe (66) and analyzed with SPARKY (67). Relaxation experimentswere described previously (68), with TROSY-detection (69) and a spectralwidth of 10,500 Hz (2,200 Hz) for the 1H (15N) dimension. R1ρ measurementsused a 1,500 Hz spin-lock field strength centered at the 15N carrier fre-quency. R1ρ values were converted into R2 as described (70).

Rex was measured as described previously (71). Briefly, the detection of α,β, or longitudinal two-spin order magnetization (Izz) during a Hahn echoperiod (2∕JNH ¼ 10.8 ms) was used in the following relationship (72):

Rex ≈ Czz lnðρzzÞ þ Cβ lnðρβÞ; [1]

where Czz ¼ ð2τÞ−1, Cβ ¼ ðhκi − 1Þð4τÞ−1, κ ¼ 1 − 2� ln ρzz∕ ln ρβ, ρzz ¼ Izz∕Iα,and ρβ ¼ Iβ∕Iα. Experiments were recorded in triplicate and in an interleavedmanner to obtain Izz, Iα, and Iβ . The value hκiwas obtained from the trimmedmean of all 15N resonances that did not exhibit chemical exchange.

Verification of Rex was done by measuring inverse peak heights (I) atdecreasing temperatures (73), as described in SI Text. Resonances thatexperienced conformational exchange diminished in peak intensity moresignificantly with decreasing temperatures (73) (Fig. S3 C–E).

MD Simulations. MD simulations were set up using CHARMM c36a1 and runwith NAMD using the crystal structures described in the SI Materials andMethods. All structures were solvated in a TIP3 water box with Kþ andCl− added as counter ions to reach an ionic strength of approximately150 mM. Following an initial equilibration, 75 ns MD simulations for eachsystem were performed at constant temperature and pressure. Rmsd, rmsf,and PCA for all simulations were performed as described in the SI Materialsand Methods.

ACKNOWLEDGMENTS.We are grateful for the fruitful discussions about datainterpretation with A. Kornev and G. Melacini. This work was supportedby the National Institutes of Health (NIH) (GM072701 and HL080081 toG.V.; T32DE007288 to L.R.M.; GM46736 to J.G.; and GM19301 to S.S.T.;University of Minnesota Graduate School Doctoral Dissertation Fellowshipto L.S.,) and the American Heart Association (09PRE2080017 to E.M.). NMRdata were collected at the National Magnetic Resonance Facility at Madi-son [NIH: P41RR02301, P41GM66326, RR02781, and RR08438; NationalScience Foundation (NSF): DMB-8415048, OIA-9977486, and BIR-9214394]and the University of Minnesota NMR Facility (NSF BIR-961477). This workwas carried out using hardware and software provided by the University ofMinnesota Supercomputing Institute.

1. Walsh DA, Van Patten SM (1994) Multiple pathway signal transduction by thecAMP-dependent protein kinase. FASEB J 8:1227–1236.

2. Shabb JB (2001) Physiological substrates of cAMP-dependent protein kinase. ChemRev101:2381–2411.

3. Johnson DA, Akamine P, Radzio-Andzelm E, Madhusudan M, Taylor SS (2001)Dynamics of cAMP-dependent protein kinase. Chem Rev 101:2243–2270.

4. Taylor SS, et al. (2004) PKA: A portrait of protein kinase dynamics. Biochim BiophysActa 1697:259–269.

5. Walsh DA, Van Patten SM (1994) Multiple pathway signal transduction by thecAMP-dependent protein kinase. FASEB J 8:1227–1236.

6. Dalton GD, Dewey WL (2006) Protein kinase inhibitor peptide (PKI): A family of

endogenous neuropeptides that modulate neuronal cAMP-dependent protein

kinase function. Neuropeptides 40:23–34.

7. Masterson LR, et al. (2010) Dynamics connect substrate recognition to catalysis in

protein kinase A. Nat Chem Biol 6:821–828.

8. Glass DB, Cheng HC, Kemp BE, Walsh DA (1986) Differential and common recognition

of the catalytic sites of the cGMP-dependent and cAMP-dependent protein kinases by

inhibitory peptides derived from the heat-stable inhibitor protein. J Biol Chem

261:12166–12171.

Fig. 4. The energy landscape of PKA-C is modulated by ligandbinding. The apo state is dynamically uncommitted, having dynamics whichare not tuned to turnover. Nucleotide binding induces motions which aresynchronized to turnover (dynamically committed) and are persistent inthe ternary complex with substrate. However, PKI5–24, or excess Mg2þ andPKI5–24, induces favorable enthalpy which lowers the energy of one or moreconformational states and raises the energy barriers in the conformationallandscape. This phenomenon hinders conformational fluctuations, inhibitsturnover, and results in a dynamically quenched enzyme.

Masterson et al. PNAS ∣ April 26, 2011 ∣ vol. 108 ∣ no. 17 ∣ 6973

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

1

9. Cook PF, Neville ME, Vrana KE, Hartl FT, Roskoshi R (1982) Adenosine cyclic 3′5′-mono-phosphate dependent protein kinase: Kinetic mechanism for the bovine skeletalmuscle catalytic subunit. Biochemistry 21:5794–5799.

10. Zimmermann B, Schweinsberg S, Drewianka S, Herberg FW (2008) Effect of metal ionson high-affinity binding of pseudosubstrate inhibitors to PKA. Biochem J 413:93–101.

11. Strick R, Strissel PL, Gavrilov K, Levi-Setti R (2001) Cation-chromatin binding as shownby ion microscopy is essential for the structural integrity of chromosomes. J Cell Biol155:899–910.

12. Stoykova AS, Dabeva MD, Dimova RN, Hadjiolov AA (1985) Ribosome biogenesis andnucleolar ultrastructure in neuronal and oligodendroglial rat brain cells. J Neurochem45:1667–1676.

13. Mauldin RV, Carroll MJ, Lee AL (2009) Dynamic dysfunction in dihydrofolate reductaseresults from antifolate drug binding:Modulation of dynamics within a structural state.Structure 17:386–394.

14. Traaseth NJ, et al. (2008) Structural and dynamic basis of phospholamban andsarcolipin inhibition of ca(2+)-ATPase. Biochemistry 47:3–13.

15. Whitehouse S, Walsh DA (1983) Mg X ATP2-dependent interaction of the inhibitorprotein of the cAMP-dependent protein kinase with the catalytic subunit. J Biol Chem258:3682–3692.

16. Lew J, Coruh N, Tsigelny I, Garrod S, Taylor SS (1997) Synergistic binding of nucleotidesand inhibitors to cAMP-dependent protein kinase examined by acrylodan fluores-cence spectroscopy. J Biol Chem 272:1507–1513.

17. Masterson LR, Mascioni A, Traaseth NJ, Taylor SS, Veglia G (2008) Allosteric coopera-tivity in protein kinase A. Proc Natl Acad Sci USA, 105 pp:506–511.

18. Herberg FW, Doyle ML, Cox S, Taylor SS (1999) Dissection of the nucleotide andmetal-phosphate binding sites in cAMP-dependent protein kinase. Biochemistry38:6352–6360.

19. Yang J, et al. (2005) Allosteric network of cAMP-dependent protein kinase revealedby mutation of Tyr204 in the P þ 1 loop. J Mol Biol 346:191–201.

20. Palmer AG, 3rd (2001) Nmr probes of molecular dynamics: Overview and comparisonwith other techniques. Annu Rev Biophys Biomol Struct 30:129–155.

21. Gangal M, et al. (1998) Backbone flexibility of five sites on the catalytic subunit ofcAMP-dependent protein kinase in the open and closed conformations. Biochemistry37:13728–13735.

22. Li F, Juliano C, Gorfain E, Taylor SS, Johnson DA (2002) Evidence for an internal entropycontributin to phosphoryl transfer: A study of domain clossure, backbone flexibility,and the catalytic cycle of cAMP-dependent protein kinase. J Mol Biol 315:459–469.

23. Karplus M, Kushick JN (1981) Method for estimating the configurational entropy ofmacromolecules. Macromolecules 14:325–332.

24. Miyashita O, Onuchic JN, Wolynes PG (2003) Nonlinear elasticity, proteinquakes, andthe energy landscapes of functional transitions in proteins. Proc Natl Acad Sci USA100:12570–12575.

25. Maragakis P, Karplus M (2005) Large amplitude conformational change in proteinsexplored with a plastic network model: Adenylate kinase. J Mol Biol 352:807–822.

26. Henzler-Wildman KA, et al. (2007) A hierarchy of timescales in protein dynamics islinked to enzyme catalysis. Nature 450:913–916.

27. Cheng X, Ivanov I, Wang H, Sine SM, McCammon JA (2007) Nanosecond-timescaleconformational dynamics of the human alpha7 nicotinic acetylcholine receptor.Biophys J 93:2622–2634.

28. Khavrutskii IV, Grant B, Taylor SS, McCammon JA (2009) A transition path ensemblestudy reveals a linchpin role for mg(2+) during rate-limiting ADP release from proteinkinase A. Biochemistry 48:11532–11545.

29. Kornev AP, Taylor SS, Ten Eyck LF (2008) A helix scaffold for the assembly of activeprotein kinases. Proc Natl Acad Sci USA 105:14377–14382.

30. Kornev AP, Taylor SS (2009) Defining the conserved internal architecture of a proteinkinase. Biochim Biophys Acta 1804:440–444.

31. Fatholahi M, LaNoue K, Romani A, Scarpa A (2000) Relationship between totaland free cellular mg(2+) during metabolic stimulation of rat cardiac myocytes andperfused hearts. Arch Biochem Biophys 374:395–401.

32. Murphy E (2000) Mysteries of magnesium homeostasis. Circ Res 86:245–248.33. Mauban JR, O’Donnell M, Warrier S, Manni S, Bond M (2009) AKAP-scaffolding

proteins and regulation of cardiac physiology. Physiology (Bethesda) 24:78–87.34. Alto N, Carlisle Michel JJ, Dodge KL, Langeberg LK, Scott JD (2002) Intracellular

targeting of protein kinases and phosphatases. Diabetes 51(Suppl 3):385S–388.35. Hammes-Schiffer S, Benkovic SJ (2006) Relating protein motion to catalysis. Annu Rev

Biochem 75:519–541.36. Boehr DD, McElheny D, Dyson HJ, Wright PE (2006) The dynamic energy landscape of

dihydrofolate reductase catalysis. Science 313:1638–1642.37. Frederick KK, Marlow MS, Valentine KG, Wand AJ (2007) Conformational entropy in

molecular recognition by proteins. Nature 448:325–329.38. Yao X, Rosen MK, Gardner KH (2008) Estimation of the available free energy in a

LOV2-J alpha photoswitch. Nat Chem Biol 4:491–497.39. Gsponer J, et al. (2008) A coupled equilibrium shift mechanism in calmodulin-

mediated signal transduction. Structure 16:736–746.

40. Mittag T, Kay LE, Forman-Kay JD (2010) Protein dynamics and conformational disorderin molecular recognition. J Mol Recognit 23(3):105–116.

41. Wright PE, Dyson HJ (2009) Linking folding and binding. Curr Opin Struct Biol19:31–38.

42. Smock RG, Gierasch LM (2009) Sending signals dynamically. Science 324:198–203.43. Tzeng SR, Kalodimos CG (2009) Dynamic activation of an allosteric regulatory protein.

Nature 462:368–372.44. Ma B, Nussinov R (2010) Enzyme dynamics point to stepwise conformational selection

in catalysis. Curr Opin Chem Biol 14:652–659.45. Pisliakov AV, Cao J, Kamerlin SC, Warshel A (2009) Enzymemillisecond conformational

dynamics do not catalyze the chemical step. Proc Natl Acad Sci USA 106:17359–17364.46. Freire E (1999) The propagation of binding interactions to remote sites in proteins:

Analysis of the binding of the monoclonal antibody D1.3 to lysozyme. Proc Natl AcadSci USA 96:10118–10122.

47. Kumar S, Ma B, Tsai CJ, Sinha N, Nussinov R (2000) Folding and binding cascades:Dynamic landscapes and population shifts. Protein Sci 9:10–19.

48. Hammes GG (2002) Multiple conformational changes in enzyme catalysis. Biochemis-try 41:8221–8228.

49. Kern DZ, Zuiderweg ER (2003) The role of dynamics in allosteric regulation. Curr OpinStruct Biol 13:748–757.

50. Pufall MA, et al. (2005) Variable control of ets-1 DNA binding by multiple phosphatesin an unstructured region. Science 309:142–145.

51. Swain JF, Gierasch LM (2006) The changing landscape of protein allostery. Curr OpinStruct Biol 16:102–108.

52. Popovych N, Sun S, Ebright RH, Kalodimos CG (2006) Dynamically driven proteinallostery. Nat Struct Mol Biol 13:831–838.

53. Li P, Martins IR, Amarasinghe GK, Rosen MK (2008) Internal dynamics control activa-tion and activity of the autoinhibited vav DH domain. Nat Struct Mol Biol 15:613–618.

54. Boehr DD, Nussinov R, Wright PE (2009) The role of dynamic conformational ensem-bles in biomolecular recognition. Nat Chem Biol 5:789–796.

55. Das R, et al. (2009) Dynamically driven ligand selectivity in cyclic nucleotide bindingdomains. J Biol Chem 284:23682–23696.

56. Mauldin RV, Lee AL (2010) Nuclear magnetic resonance study of the role of M42 inthe solution dynamics of Escherichia coli dihydrofolate reductase. Biochemistry49:1606–1615.

57. Berlow RB, Igumenova TI, Loria JP (2007) Value of a hydrogen bond in triosephosphateisomerase loop motion. Biochemistry 46:6001–6010.

58. Doucet N, Watt ED, Loria JP (2009) The flexibility of a distant loop modulates activesite motion and product release in ribonuclease A. Biochemistry 48:7160–7168.

59. Seckler JM, Barkley MD, Wintrode PL (2011) Allosteric suppression of HIV-1 reversetranscriptase structural dynamics upon inhibitor binding. Biophys J 100:144–153.

60. Lipchock JM, Loria JP (2010) Nanometer propagation of millisecond motions in V-typeallostery. Structure 18:1596–1607.

61. Taylor SS, Radzio-Andzelm E (1997) Protein kinase inhibition: Natural and syntheticvariations on a theme. Curr Opin Chem Biol 1:219–226.

62. Cohen P (2002) Protein kinases—the major drug targets of the twenty-first century?Nat Rev Drug Discov 1:309–315.

63. Engh RA, Bossemeyer D (2002) Structural aspects of protein kinase control-role ofconformational flexibility. Pharmacol Ther 93:99–111.

64. Zhang J, et al. (2010) Targeting bcr-abl by combining allosteric with ATP-binding-siteinhibitors. Nature 463:501–506.

65. Malmendal A, Halpain S, Chazin WJ (2003) Nascent structure in the kinase anchoringdomain of microtubule-associated protein 2. Biochem Biophys Res Commun301:136–142.

66. Delaglio F, et al. (1995) NMRPipe: Amultidimensional spectral processing system basedon UNIX pipes. J Biomol NMR 6:277–293.

67. Goddard TD, Kneller DG (1999) SPARKY 3. (University of California, San Francisco).68. Farrow NA, et al. (1994) Backbone dynamics of a free and phosphopeptide-complexed

src homology 2 domain studied by 15N NMR relaxation. Biochemistry 33:5984–6003.69. Pervushin K, Riek R, Wider G, Wuthrich K (1997) Attenuated T2 relaxation by mutual

cancellation of dipole-dipole coupling and chemical shift anisotropy indicates anavenue to NMR structures of very large biological macromolecules in solution. ProcNatl Acad Sci USA 94:12366–12371.

70. Tjandra N, Wingfield P, Stahl S, Bax A (1996) Anisotropic rotational diffusion ofperdeuterated HIV protease from 15N NMR relaxation measurements at two mag-netic fields. J Biomol NMR 8:273–284.

71. Wang C, Rance M, Palmer AG, 3rd (2003) Mapping chemical exchange in proteinswith MW > 50 kD. J Am Chem Soc 125:8968–8969.

72. Tatulian SA, Jones LR, Reddy LG, Stokes DL, Tamm LK (1995) Secondary structureand orientation of phospholamban reconstituted in supported bilayers from polarizedattenuated total reflection FTIR spectroscopy. Biochemistry 34:4448–4456.

73. Fenwick MK, Oswald RE (2008) NMR spectroscopy of the ligand-binding core ofionotropic glutamate receptor 2 bound to 5-substituted willardiine partial agonists.J Mol Biol 378:673–685.

6974 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1102701108 Masterson et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

1