Conformational Transitions of Adenylate Kinase: Switching by Cracking

  • Published on
    26-Oct-2016

  • View
    212

  • Download
    0

Embed Size (px)

Transcript

  • oa2

    1041 E. Lowell Street, Tucson,AZ 85721, USA3

    eroresucostofetwpa

    nacompeting native interactions from the open and closed form can accountfor the large conformational transitions in AKE. We further characterize theconformational transitions with a new measure Func, and demonstrate

    asger

    rta

    which complement the static three-dimensional ducing proteins, such as kinases. Signal transduc-

    doi:10.1016/j.jmb.2006.11.085structures provided by X-ray crystallography andNMR measurements, are essential to understandprotein functions.1

    Protein flexibility and plasticity allow proteins tobind ligands, form oligomers, aggregate, and per-form mechanical work. Therefore, the ability to alterprotein dynamics may enable quantitative control ofprotein functionality. While this form of functionalcontrol is very important to biology, it is not well

    ing proteins are mobile and communicate withreceptor proteins, which then produce specificreactions. The activity of many signal transducingproteins is associated with large conformationalchanges. For example, C-terminal Src Kinaseprotein,2 the Cyclin Dependent Kinase family,3 theProtein Kinase C family4 and Adenylate Kinase(AKE)5 have stable inactive conformations, in addi-tion to active forms. Since the balance between*Corresponding author

    Introduction

    Flexibility and conformationacknowledged to be indispenproteins. New experiments usinnology and detailed computbegun to reveal the motionswhich encompass a rich repeon various length and time scunderstood from either a theoretbasis. Thus, the question arises:titatively connect conformation

    E-mail address of the correspondijonuchic@ctbp.ucsd.edu

    0022-2836/$ - see front matter 2006 EKeywords: conformational change; energy landscape theory; cracking; strain

    l changes are wellable properties ofultrafast laser tech-simulations haveof these proteins,oire of movementsles. These motions,

    biomolecular recognition and function? To addressthis question, we propose a structure-basedmodel tostudy the dynamical properties of proteins, specifi-cally, conformational rearrangement.Large conformational changes in proteins are

    important in many cellular signaling pathways,which can be generally described by the followingsteps. First, a signaling protein becomes activated,which then activates, or deactivates, signal trans-Biology,Weizmann Institute ofScience, Rehovot 76100, Israelinteractions. 2006 Elsevier Ltd. All rights reserved.Department of Structural that local unfolding may be due, in part, to competing intra-proteinConformational TransitionsSwitching by Cracking

    Paul C. Whitford1, Osamu Miyashitand Jos N. Onuchic11Center for TheoreticalBiological Physics andDepartment of Physics,University of California atSan Diego, 9500 Gilman Drive,La Jolla, 92093, USA2Department of Biochemistryand Molecular Biophysics,University of Arizona,

    Conformational hettheir function. We pbetween protein strallosteric or non-alldetailed mechanismlate Kinase (AKE) bthat is crucial to thestrain energy whichduring the functioical or experimentalHow can we quan-al dynamics with

    ng author:

    lsevier Ltd. All rights reservef Adenylate Kinase:

    , Yaakov Levy3

    geneity in proteins is known to often be the key toent a coarse grained model to explore the interplayture, folding and function which is applicable toeric proteins. We employ the model to study thethe reversible conformational transition of Adeny-een the open to the closed conformation, a reactionrotein's catalytic function. We directly observe highppears to be correlated with localized unfoldingl transition. This work also demonstrates that

    J. Mol. Biol. (2007) 366, 16611671conformations regulates protein activity, conforma-tional transitions play important roles in the ma-chinery of the cell.6Functional conformational transitions require a

    biomolecule to have at least a pair of conformationalstates of nearly equal free energy. The energylandscapes of these proteins have several basins ofattraction and the transitions between basins dic-

    d.

  • tates the conformational dynamics.7 Despite thebiological significance, the details of these processesare not fully understood. With a complete under-standing of conformational changes we hope topredict which proteins have multiple conforma-tions, predict these alternate conformations, deter-mine the properties of the conformational transitionensemble, explain how proteins have evolved tohave these properties and eventually design novelmacromolecular machines which can execute anygiven biological function. To work towards theseobjectives, we explore the relationship between thestructure, folding and function of AKE.While many studies have investigated the rela-

    tionship between protein structure and folding,fewer have focused on the relationship betweenstructure and function, and even fewer haveexplored the interplay between protein structure,folding mechanism and function. Current experi-mental methods, including NMR, X-ray crystallog-raphy and fluorescence spectroscopy have beensuccessful in describing the structural properties ofindividual states. These methods sometimes alsomanage to capture the chain flexibility.8,9 Nonethe-less, experimental techniques have not been able toprovide the molecular details necessary to fullyunderstand the mechanism of conformationalchanges. Due to these limitations, there has beensignificant effort to develop a theoretical frameworkfor describing functional transitions in proteins.1015

    With a developed framework, one may study theenergetic barriers associated with conformationaltransitions, their coupling to folding/unfolding(cracking), the role of ligands, and the role ofenergetic heterogeneity and frustration in confor-mational transitions.10,11 In this work we propose astructure-based model that has a clear physicalinterpretation. Our model demonstrates that intra-protein contacts formed in the ligand boundstructure of AKE can be responsible for the observedfunctional conformational changes. There has beensuccess in applying simplified models to conforma-tional changes, but our model provides a newphysical interpretation that has not been proposedelsewhere.The simplest model to describe functional transi-

    tions is based on landscape hopping and crackingbetween elastic networks.10,11 To lowest order ap-proximation, all interactions about a minimum areharmonic. Thus, this approach uses the mostsimplified approximation to the landscape abouttwo energetic basins. From this model, the energet-ics of transitions are determined. This approach hasbeen successful in demonstrating the physicalrelationship between protein fluctuations (lowfrequency normal modes) and protein function(conformational transition), and thus serves as abenchmark for further work.To elucidate the relationship between protein

    structure, folding and function, functional transi-12,14

    1662tions have beenmodeled as a result of hoppingbetween structure-based energy surfaces. Thesestructure-based potentials, which were inspired bythe work of Go,16 have had great success in ex-plaining the interplay between protein structureand protein folding.17,18 A limitation of thesemodels is that the two structure-based energysurfaces have many nearly redundant contribu-tions, since the conformations of interest havestructural overlap. When applying these models toentire proteins, these near-redundancies may, ormay not, contribute to the conformational changes.These redundancies add a degree of uncertaintyto the physical interpretation of the system.Therefore, here, redundant interactions havebeen removed and replaced by single contactsfor both structures.Inspired by the successes of minimalist structure-

    based models in advancing our understanding ofprotein folding and molecular recognition,1726 ourapproach begins with the established theoreticalframework of protein folding. As described below,we extend this framework to account for largeconformational changes.It is well established that protein folding is the

    result of a globally funneled, minimally frustratedenergy landscape.2729 The application of the prin-ciple of minimal frustration via structure-basedpotentials with single native basins has had consid-erable success in explaining the physics of proteinfolding. To now explain large functional transitions,there is a need for multiple basins. Thus, wegeneralize the minimally frustrated energy land-scape of protein folding studies to incorporatebiologically functional motions. We propose proteinstructure dominates functional behavior, as well asprotein folding. Thus, we begin with a structure-based potential and add gradual perturbations,based on an alternate structure, to produce multipleminima. Our model implies, as does our previousmodel,10 that the transition ensemble can bedetermined from information of the conformationsof interest. This is an implication of structure-basedmodels in general (folding transition states can bedetermined by information of the native state).Using our model, we also show that multiple stableconformations may be due to amplified roughnessin the global energetic landscape upon ligandbinding.Some proteins undergo large conformational

    changes without the aid of a co-factor. In allostericproteins, however, such as Calmodulin and AKE,large conformational changes are associated witha co-factor, often an ion or a small biological mole-cule. Our model is general enough to be applied toboth allosteric and non-allosteric conformationalchanges.The model protein used in this study is E. Coli

    Adenylate Kinase. AKE is a 214 residue 3 domainprotein (Figure 1) that catalyzes the reaction

    ATPAMPAKE

    2ADP 1

    Switching by Cracking: Adenylate Kinasewhile undergoing large conformational changeswhich are believed to be the rate limiting steps of

  • or closed), interresidue distance distributions andB-factors to experimental results. Four proposed

    1663Switching by Cracking: Adenylate Kinasethe reaction.30 This protein was chosen mainlybecause it is well established that its multiplestructures are catalytically relevant and becausethere is evidence that the conformational changesare rate limiting. Moreover, AKE is a good proteinsystem to study the physics of conformationalswitching because there is a large amount experi-mental and theoretical data available on thisprocess.

    Results

    Hamiltonian determination and implications

    First, we developed several potentials and deter-mined which reproduces the structural properties ofthe open and closed forms of AKE. Second, weemployed the superior potential to study theconformational transitions of AKE. Analogous toprotein folding models where information of thenative state is used to model the folding properties,this work uses information about two stable forms ofAKE to infer conformational transition properties. Todetermine which potential most accurately accountsfor the structural properties of AKE's conformations,we compared conformational preference (i.e., open

    Figure 1. Functionally Relevant Conformations ofAKE. Structure of the open (blue)42 and closed (orange)5

    forms of AKE, with the CORE domain spatially aligned(grey). ATP binds in the pocket formed by the LID andCORE domains. AMP binds in the pocket formed bythe NMP and CORE domains. Figure prepared withVisual Molecular Dynamics.43Hamiltonians were compared: Hopen-Copen-D (open struc-

    ture potential), Hclosed-Copen-D (open/closed mixed struc-

    ture potential), Hopen-Cclosed-D (closed/open structure

    potential) and Hclosed-Cclosed-D (closed structure-based po-

    tential, see Models and Methods). Each potentialstabilizes the contacts native to the open or closedform (denoted by C-open and C-closed) and the di-hedral angles found in the open or closed form(D-open and D-closed). According to the abovecriteria, Hopen-C

    open-D reproduces experimental resultsmost accurately (explained below).The first experimentally known property of AKE

    that our potential must reproduce is that theunligated protein must be predominantly in theopen form. Since we later propose ligand bindingcan be represented by introducing contacts uniqueto the closed form (which are scaled by 2; seeModels and Methods), the simulated AKE withoutthe contacts unique to the closed form (i.e. 2=0.0)must also be in the open form. Hopen-C

    open-D and Hopen-Cclosed-D

    have this property. Hclosed-Copen-D and Hclosed-C

    closed-D do notexhibit this property under any conditions (data notshown). Since the open state is not an energeticminimum (global or local) for Hclosed-C

    open-D and Hclosed-Cclosed-D

    these are not appropriate potentials for our investi-gation, and were not further considered. This resultsuggests that the open state is not purely aconsequence of entropy, but energetic contributionsare important as well.Distance distributions P(r) of residues A55 and

    V169 (located in the NMP and LID domains,respectively) have been determined experimentallyfor unligated and ligated AKE31 and were comparedto the values obtained for the remaining twoHamiltonians: Hopen-C

    open-D and Hopen-Cclosed-D. Rmax is the

    value of r at which P(r) is a maximum. In simula-tions, Rmax does not vary significantly for T

  • Switching by Cracking: Adenylate Kinaseand Hopen-Copen-D is 0.68. The correlation coefficient

    between B-factors from the crystal structure andHopen-Cclosed-D is 0.56. Both potentials' B-factors were

    poorly correlated for the closed conformation(r20 kcal/mol), thus wepredicted that localized regions of the proteinunfold during conformational transitions, as amechanism to reduce strain and enhance catalyticefficiency. The simulations reported here supportthe high strain energy and unfolding hypothesis(Figure 5).Figure 5 (top left) shows the average strain energy

    (defined as the total potential energy) by residue.There are clear peaks near residues 6070, 120125and to a lesser extent residues 1020, 3035, 8090 and170180. These finding are in excellent agreementwith normal mode predictions of high strain in resi-dues 10, 110125, 150 and 160170.10 In normal modestudies, residues 30, 60 and 80 have only weak peaksat the later stage of the conformational transition. Thisis likely due to the previous datamainly reporting thestrain associated with LID closure, and not NMPclosure.10

    We believe that the high strain energy in AKE is

    the result of competing energetic contributions.

    Residues 129, 68117 and 161214.

  • for the two energetic barriers is shown inFigure 5.

    1665Switching by Cracking: Adenylate KinaseSince competing energetic terms can not besatisfied simultaneously, internal strain must re-sult. Some regions of strain drive the protein'sopening transition and other regions of straindrive the closing transition (see Functional -valuesbelow). Thus, the balance between competing

    Figure 2. Contacts Native to the Closed ConformationCan Account for Large Conformational Changes. Freeenergy as a function of the distance between center ofmass of the LID domain and CORE domain (RLID-CORE

    CM )for 2=0.51.2 (incremented by 0.1, colored black topurple). 2 is the interaction stre...