Probing the “Annealing” Mechanism of GroEL Minichaperone using Molecular Dynamics Simulations

doi:10.1016/j.jmb.2005.05.012 J. Mol. Biol. (2005) 350, 817–829

Probing the “Annealing” Mechanism of GroEL Mini-chaperone using Molecular Dynamics Simulations

George Stan1, Bernard R. Brooks1 and D. Thirumalai2,3*

1Laboratory of ComputationalBiology, National Heart, Lungand Blood Institute, NationalInstitutes of Health, BethesdaMD 20892, USA

2Biophysics Program, Institutefor Physical Science andTechnology, University ofMaryland, College Park, MD20742, USA

3Department of Chemistry andBiochemistry, University ofMaryland, College Park, MD20742, USA

0022-2836/$ - see front matter q 2005 E

Abbreviations used: SP, substratebinding peptide; WBP, weakly bindtransient binding release; CyPA, cy

E-mail address of the [email protected]

Although the intact chaperonin machinery is needed to rescue naturalsubstrate proteins (SPs) under non-permissive conditions the “mini-chaperone” alone, containing only the isolated apical domain of GroEL,can assist folding of a certain class of proteins. To understand the annealingfunction of the minichaperone, we have carried out molecular dynamicssimulations in the NPT ensemble totaling 300 ns for four systems; namely,the isolated strongly binding peptide (SBP), the minichaperone, andthe SBP and a weakly binding peptide (WBP) in complex with theminichaperone. The SBP, which is structureless in isolation, adopts ab-hairpin conformation in complex with the minichaperone suggestingthat favorable non-specific interactions of the SPs confined to helices H andI of the apical domains can induce local secondary structures. Comparisonof the dynamical fluctuations of the apo and the liganded forms of theminichaperone shows that the stability (needed for SP capture) involvesfavorable hydrophobic interactions and hydrogen bond network formationbetween the SBP and WBP, and helices H and I. The release of the SP, whichis required for the annealing action, involves water-mediated interactionsof the charged residues at the ends of H and I helices. The simulationresults are consistent with a transient binding release (TBR) model for theannealing action of the minichaperone. According to the TBR model, SPannealing occurs in two stages. In the first stage the SP is captured by theapical domain. This is followed by SP release (by thermal fluctuations) thatplaces it in a different region of the energy landscape from which it canpartition rapidly to the native state with probability F or be trapped inanother misfolded state. The process of binding and release can result inenhancement of the native state yield. The TBR model suggests “that anycofactor that can repeatedly bind and release SPs can be effective inassisting protein folding.” By comparing the structures of the non-chaperone a-casein (which has no sequence similarity with the apicaldomain) and the minichaperone and the hydrophobicity profiles we showthat a-casein has a pair of helices that have similar sequence and structuralprofiles as H and I. Based on this comparison we identify residues thatstabilize (destabilize) a-casein–protein complexes. This suggests thata-casein assists folding by the TBR mechanism.

q 2005 Elsevier Ltd. All rights reserved.

Keywords: GroEL minichaperone; transient binding release mechanism;protein–protein interactions; MD simulations
*Corresponding author
Introduction

The GroEL/GroES nanomachine undergoes a

lsevier Ltd. All rights reserve

protein; SBP, stronglying peptide; TBR,clophilin A.ing author:

series of concerted allosteric transitions aided byATP binding and hydrolysis to rescue substrateproteins (SPs) that may otherwise be destined foraggregation.1 For stringent substrates under non-permissive conditions (i.e. spontaneous refoldingwith sufficient yield of the native state is unlikely)the intact GroEL machinery is required. Never-theless, it is interesting to wonder if there is aredundancy in the oligomeric structure of GroEL sothat a simpler construct is sufficient for its annealing

d.

818 GroEL Minichaperone Annealing Mechanism

ability. Yoshida and co-workers reported that the307-residue proteolytic GroEL fragment (GroEL150–456) enables refolding of denatured rhoda-nese.2 More surprisingly, Fersht and co-workersfound that the monomeric apical domain (GroEL191–345), referred to as “minichaperone”, hasGroEL-like activity at least for the SPs rhodaneseand cyclophilin A.3 The ability of minichaperones toassist in the folding of rhodanese (yield between20% and 40%) is particularly intriguing becauseprevious experiments had shown that this SPrequires GroES and ATP.

Weissman and co-workers also showed that theminichaperone can fold rhodanese (to some extent)under optimal conditions (excess concentration ofGroEL (191–345) and TZ25 8C).4 More importantly,it was shown that not only does minichaperoneassist folding of cyclophilin A, but also a non-chaperone protein a-casein was equally efficient inenhancing the yield of class II proteins (those thatare not stringent substrates). This observation andthe inability of minichaperone to assist folding ofthe stringent proteins such as rhodanese underphysiological conditions prompted Wang et al. toconclude that the full machinery of GroEL-GroES isrequired for in vivo assisted folding.4 Despite thediffering emphasis, both reports showed that evenfor SPs that undergo spontaneous folding the yieldof the reaction is substantially increased (seeFigure 2(b) of Wang et al.4) in the presence ofminichaperones.3,4 Therefore, it is important tounderstand, in molecular detail, the annealingaction of the GroEL minichaperone.

The molecular mechanism of assisted folding byminichaperone has to be different from thatenvisioned for the intact GroEL–GroES system.5 Inthe latter case, the change in the GroEL–SPinteraction, which occurs as a result of a series ofallosteric interactions GroEL undergoes in responseto ATP and GroES binding, is the cause for itsannealing activity.6 The role of the minichaperone(in the intact GroEL nanomachine) is to capturesubstrate proteins with exposed hydrophobic resi-

Table 1. Details of MD simulations in the NPT ensemble

System Type Initial conformation

Unliganded minichaperone Wild-type PDBc

E257ALiganded minichaperone Wild-type PDBd

E257AWBP PDBe

Isolated peptideg Wild-type Hairpinf

Randomh

a Each trajectory is 10 ns long.b Initial linear size of the rhombic dodecahedral box.c the PDB structure 1KID (GroEL 191–376 and N-terminal tag) wad The PDB structure 1DKD (SBP-GroEL 191–336) was used as inite The starting structure is modeled using the minichaperone–SBPf The starting structure is taken as the b-hairpin conformation in tg Using the one-letter code for amino acids, the sequence of the sth The initial structures are generated by randomizing the dihedra

dues through favorable SP–apical domain inter-actions. Thus, at a first glance, it is surprising thatthe minichaperone can help protein folding at all.The experimental findings3,4 can be rationalizedusing the results of a previous study by Betancourt& Thirumalai (BT),7 who showed that a hydro-phobic cavity can assist folding under non-permissive conditions provided the strength of theSP–cavity interaction is modest.8,9 The rationale forthe BT proposal is that modest SP–cavity inter-actions enable a stable enough complex from whichpeptide dissociation can also take place. We showedthat modest SP–cavity interaction has the effect oftransient trapping followed by the release of the SP,which results in its partial unfolding. As a result,upon release by the minichaperone, the SP findsitself in a different (perhaps favorable) part of thefree energy landscape from which refolding canstart anew. Here, we argue, using a series ofmolecular dynamics simulations of a peptideinteracting with the apical domain, that theminichaperone can mimic the binding–unbindingof SP just as observed in simple model systems.7,10

We find that an interplay between the interactionsof the charged and hydrophobic residues in thehelices H and I of the apical domain with the SPdetermines the annealing efficiency. Comparison ofa putative structure of a-casein and the minichaper-one structure also readily explains the similarity inthe annealing mechanisms of the two proteins. Theonly fundamental requirement for assisted foldingby the minichaperone appears to be a transientinteraction of the SP with the hydrophobic surfacesfollowed by its release by thermal fluctuations.

Results

Binding to the minichaperone induces a randomcoil to b-hairpin transition in the stronglybinding peptide

The strongly binding peptide (SBP), which is one

No. trajectoriesa Sizeb (A) No. water molecules

8 65 480215 52 237331 52 23742 35 26074 30 1623, 1605, 1611, 16173 35 2751, 2745, 27333 40 4209, 4203, 4203

s used as initial conformation.ial conformation.PDB complex as a template.he SBP–GroEL (191–336) complex.rongly binding peptide is SWMTTPWGFLHP.l angles.

GroEL Minichaperone Annealing Mechanism 819

of the library of peptides considered by Chen &Sigler, is structureless in isolation.11 To monitor thestructural transformation of the SBP we firstsimulated (see Table 1 for simulation details) theisolated neutral SBP in water.

Molecular dynamics simulations show that theisolated SBP is a random coil in water

To draw statistically significant conclusions it isnecessary to run several independent trajectoriesstarting from distinct initial conditions. Crystalstructures show that in the GroEL (191–336)–SBPcomplex the peptide adopts a b-hairpin structure.11

The distribution of P(Rg) of the radius of gyration(Rg) of the SBP in complex with the apical domain isnarrowly peaked around the value (Z6.5 A) that itadopts in the crystal structure (Figure 1). If thestarting conformation of the peptide in the bulk isthe b-hairpin extracted from the GroEL (191–336)–SBP complex then the typical time for the transitionto random coil would be at least 1 ms, which isthe minimum unfolding time for a preformedb-hairpin. Thus, in the course of relatively shortMD simulations the SBP is likely to remain in thestructured conformations. In accord with thisexpectation, we find that P(Rg) for the SBP (Figure 1)remains close to the b-hairpin structure throughoutthe two 10 ns trajectories if the starting conditionsfor the SBP are taken from the SBP–minichaperonecomplex. The lack of population of the randomstructures starting from the initial b-hairpin struc-ture, and the two peaks in Figure 1, are due to thenon-ergodicity of the simulations.

To obtain the structural characteristics of the SBPin the bulk we generated a total of 100 ns trajectories(see Table 1) starting from initial conformations thatrandomize the dihedral angles. We find that the SBPquickly becomes random coil-like and undergoes

large fluctuations during the course of the simu-lation. The distribution P(Rg) of SBP is broad andhas the characteristics of a random coil withconsiderable conformational flexibility. The simu-lated P(Rg), at large values of Rg, nearly coincideswith the rigorous theoretical predictions for a self-avoiding walk (polymer in a good solvent)(Figure 1). These results also suggest that the SBP(or more generally substrate proteins in the intactmachinery) undergoes conformational transitionsupon capture by the apical domain. The structuresSPs adopt when interacting with GroEL need notcoincide with what is found in the native state.

Peptide binding and the stability of the mini-chaperone–peptide complex

The findings by BT7 suggest that transientbinding with a hydrophobic surface can assist thefolding of some SPs provided the interactionstrength does not significantly exceed kBT.8,10 Weuse molecular dynamics simulations of the solvatedapical domain, which contrast the dynamics of theunliganded and the liganded minichaperone, toidentify residues that ensure the stability of theminichaperone–SBP complex. In addition, wepinpoint putative regions responsible for SPunbinding based on the unstable interactionsfound during dynamics.

SBP binding suppresses fluctuations in helices Hand I

The finding by Chen & Sigler11 that four regions,helices H (residues 234–243) and I (residues 257–268), and the loop regions 207–211 and 301–311 (seeFigure 5 in Chen & Sigler11), are perturbed by SBPbinding suggests that they lock into favorableconformations upon complex formation of the

Figure 1. Distribution of theradius of gyration of the SBP incomplex with the minichaperone(blue). The green curve gives P(Rg)for the hairpin in the bulk with theinitial hairpin structure, whilethe red curve is obtained with arandom starting structure. Themean radius of gyration, hR2

gi1=2,

of the SBP in the three cases are:bound, 6.6 A; free hairpin, 6.9 A;and free random coil, 8.3 A. Thebroken line represents PðRgÞwexpðKx1=ð1KnÞÞðnz0:6; xZRg=hR

2gi

1=2Þ fora polymer in a good solvent.29 Theexcellent agreement between P(Rg)for a self-avoiding walk and thered curve for xO1 shows that theisolated SBP is a random coil.

Figure 2. Structural details of hydrogen bonds betweenhelices H and I. Hydrogen bonds are formed between


complex. In our MD simulations, rms fluctuationsare reduced only in the helices H and I (data notshown). The mobility of segments 207–211 and301–311 in solution is largely unaffected uponcomplex formation because the SBP binding-induced relaxation times of residues in the loopsexceed the simulation time-scale. Because theminichaperone–peptide complex is stable duringthe simulations, we conclude that the suppressionof the fluctuations of the two loop regions (207–211and 301–311) must occur after SBP binding. Forshort peptides that bind to the minichaperone,helices H and I appear to be the only regions thatadopt the required conformations to accommodatethe ligand. The flexible loop regions may play anessential role in the annealing action of theminichaperone through thermal fluctuations thatultimately result in peptide unbinding.

OE1 or OE2 of Glu255 with NZ (with hydrogen bondingsites HZ1, HZ2 or HZ3) of Lys226, CO (Asp253)-NH(Ile227), CO (Ile227)-NH (Glu255), CO (Ser228)-NH(Ala258).

Movement of helices H and I, upon binding SBP,maintains the integrity of the binding sites

The relative motion of helices H and I, analyzedusing the helix-to-helix packing procedure,12 showsthat the relative separation between the two helicesincreases by only about 2 A upon SBP binding (datanot shown). Fluctuations of the interhelical sepa-ration are reduced in the liganded case compared tothe unliganded case, which is consistent with thereduced flexibility of helices H and I upon peptidebinding. The importance of the hydrophobicgroove, presented by the binding site, is highlightedby the observed parallel relative orientation of thetwo helices of the minichaperone or the mini-chaperone–SBP complex. The motion of the side-chains is synchronous with the backbonefluctuations (data not shown), which implies thatthe constituent structural elements move as rigidunits even at the local structural level.

Peptide binding does not significantly perturbinterhelical interactions

The stability of the binding site is maintainedthrough a set of contacts that are concentrated nearthe N terminus of helices H and I. These contacts arenot significantly affected by peptide binding. Thestrongest interhelical interactions in the mini-chaperone–peptide complex involve residuesLys226–Glu252, Lys226–Asp253, Lys226–Glu255,Ile227–Val254, Ser228–Glu255, Asn229–Glu257,and Ile230–Glu257. Interactions involving Lys226are more stable in the complex. The contact Lys226–Glu255 becomes stronger by about 12 kcal/mol(data not shown) at the expense of the Lys266–Glu252 and Lys226–Asp253 contacts.

The stable hydrogen bond between Lys226 andGlu255, which results in a salt-bridge, is present inboth the free minichaperone and in the complex.The hydrogen bond between these side-chains(Figure 2) fluctuates significantly among the hydro-gen bonding sites of Lys226 (hydrogen atoms HZ1,HZ2, HZ3) and the acceptor atoms of Glu255 (OE1

and OE2). The enhanced stability and strength ofthe Lys226–Glu255 interaction in the mini-chaperone–peptide complex suggests that thissalt-bridge must play a major role in the overallstability of the binding sites. Two additionalhydrogen bonds (Figure 2) contribute significantlyto the stability of the binding sites. These are strong(xK4 kcal/mol), stable, bonds formed between thebackbone amide and carbonyl groups of Ile227 andAsp253, and Glu255 and Ile227, respectively (datanot shown). However, the overall energy contri-bution of these contacts to the stability of thebinding site is only xK2.5 kcal/mol for Ile227–Asp253 (Figure 2) and xK0.75 kcal/mol for Ile227–Glu255 (data not shown). In these cases, the stronghydrogen bonding energy contribution is counteredby unfavorable interactions between the hydro-phobic side-chain of Ile227 and the charged side-chains of Asp253 and Glu255.

In a bioinformatic study, we showed that thechemical nature (hydrophobic, polar, positively ornegatively charged) of the residues Lys226, Ile227,Ser228, Asp253, Val254, Glu255, and Val271 isstrongly conserved.13 The present simulationssuggest that the charged and polar residues in thisset also contribute to the polypeptide bindingfunction by forming a network of residues thatstabilize the SBP–minichaperone complex.We predict that mutations at these residues,particularly those that alter the hydrophobiccharacter of contacts between Ile227–Val254 andthe salt-bridge Lys226–Glu255, will adversely affectpolypeptide binding by the minichaperone.

Protein–protein interactions ensure the stability ofthe minichaperone–SBP complex

The crystal structure of the SBP–minichaperone

Figure 3. Details of the minichaperone–SBP interface.(a) Bulky hydrophobic residues Trp7 and Phe9 of the SBPare located in the binding pocket of the minichaperone.Interactions of Trp7 and Phe9 with minichaperoneresidues, along with the hydrogen bond network,established between the minichaperone and SBP ensurethe stability of the complex. (b) Hydrogen bonds at theinterface between the minichaperone and the peptide.Strong and stable hydrogen bonds are formed betweenAsn265 of the minichaperone and Leu10 of SBP, whilerelatively unstable hydrogen bonds are formed betweenArg231 and Pro12.


complex shows that Trp7 and Phe9 are buried in thehydrophobic pocket (Figure 3(a)). A hydrogen bondnetwork is also formed between the backbone of thepeptide and the polar surface of the apical domain.In the process of forming the SBP–apical domaininterface, the peptide buries considerable surfacearea. The combined surface area of helices H and Iexposed to the solvent in free solution is 1952 A2,whereas, in the bound state, it is 1486 A2. Theminichaperone–SBP interface is stabilized by vander Waals contacts involving minichaperone residuesLeu234, Leu237, Glu238, Ala241, Thr261, Asn265, andIle270, and hydrogen bonds between Asn265 andLeu10, Arg231 and Pro12, and Arg268 and Gly8.

Most of the crystallographically identified con-tacts are stable during our molecular dynamicssimulations and make significant contributions tothe attractive interaction between the minichaper-one and SBP. Hydrogen bonds formed betweenAsn265 and Leu10 (Figure 3(b)) are particularlystrong and stable (Figure 4). The hydrogen bondsformed by Arg231 with Pro12 are unstable duringour simulations, which is in accord with the Chen &Sigler suggestions. Both oxygen atoms (OT1 andOT2) from the carboxylic group of Pro12 competefor an NH2 hydrogen atom of the Arg231 side-chain(Figure 3(b)), which results in fluctuations in thesehydrogen bonds. The hydroxyl oxygen atom ofPro12 (OT2) was not resolved in the crystalstructure. The hydrogen bond between Arg268and Gly8, proposed by Chen & Sigler, was notobserved in the simulations. This hydrogen bond israpidly destroyed in solution, which suggests thatthis interaction is not needed to stabilize thecomplex.

Peptide binding alters water structure at the proteinsolvent interface

Simulations are particularly useful in examiningthe role water plays in stabilizing the complex.The overall number of hydrogen bonds formed bythe two helices with water molecules is larger in theunliganded state (Figure 5(a)), which ensuresgreater mobility of the protein binding sites.Hydrogen bonding sites at Thr261, Asn265, andThr266 are nearly completely dehydrated as a resultof the formation of stable contacts with SBP. Thedramatic reduction in the mobility of these side-chains upon complex formation is illustrated by thelong lifetimes (w100 ps) (Figure 5(b)) of watermolecules at these hydrogen bonding sites. Thelong lifetimes should be compared with the typicalhydrogen bond lifetime of w1 ps for liquid water.14

The enhanced stability of hydrogen bonds betweenwater molecules and protein residues suggests astructural role of water near these sites. Theapparent paradox of reduced hydration and long-lived water hydrogen bonds near these sites is dueto the rigidity of side-chains involved in peptidebinding.

Hydrogen bonds of Arg268 with water moleculesare also more stable in the liganded state than in

Figure 4. Average energies(circles) over all simulation trajec-tories of hydrogen bonds at theinterface between the minichaper-one and the peptide. Error barsrepresent average rms fluctu-ations and dots indicate theenergy values corresponding tocoordinates in the crystal struc-ture. Lines are drawn as a guide tothe eye. The hydrogen bonds arelabeled according to the heavyatom of the minichaperone andthe heavy atom of the peptideparticipating in each bond. (Here,OT1 denotes the C terminus carbo-nyl oxygen atom and OT2 the Cterminus hydroxyl oxygen atom.)


the unliganded structure. This is reflected insignificantly increased lifetimes of water hydrogenbonds at its sites (Figure 5(b)) in the minichaperone–SBP complex, even as the fractional occupancy ofthese sites increases (Figure 5(a)). This shows thatthe side-chain of Arg268 is stabilized in theliganded state by formation of hydrogen bondswith water molecules. Moreover, the Arg268–Gly8hydrogen bond, present in the crystal structure, isnot among the interactions that stabilize theminichaperone–SBP complex. The minichaperoneresidues involved in the interface with the peptidehave also been identified in the crystal structure ofthe N-terminal tag of a neighboring minichaperone.15

Based on these interactions, Buckle et al. argue thatthe residues involved in the polypeptide bindingare Ile230, Glu238, Ala241, Glu257, Ala260, Thr261,Asn265, Arg268, Ile270, Val271. Our results suggestan important role for Ile230, Glu238, Thr261,Asn265, Ile270, and Val271 in polypeptide binding.Residues Ala241 and Ala260 do not appear to be asrelevant for peptide binding, while Glu257 (asdiscussed in the next section) provides destabilizinginteractions that might trigger release of the bound SP.The simulations and the static crystal structures showthat a network of hydrogen bonds (involving thecomplex and water) as well as several side-chaincontacts stabilize the apical domain–peptide complex.

Water-mediated interactions involving chargedresidues trigger peptide release

For efficient annealing, the minichaperone mustnot only form a stable (but not hyperstable)complex with SP but also must trigger the releaseof the transiently bound SP.7 In the intact systembinding to the apical domain is followed by GroES-induced encapsulation of the SP. During the powerstroke of the cycle the SP unfolds perhaps due to a

ATP and GroES-binding-induced force generatedduring the GroEL domain movement.16,17 Incontrast, the release mechanism in the mini-chaperone must be mediated by thermal fluctu-ations7 involving residues near helices H and I. Wepropose that residues that most directly render theSP–minichaperone complex unstable serve as atrigger for SP release.

Interhelical contacts are weakened upon SBPbinding

The increased spacing between the helices andthe formation of contacts between Leu237, Asn265,and Glu257 with SBP residues result in weakerinteractions especially between Ser228 and Gly256,Ser228 and Ala258, Asn229 and Ala258, Ile230 andGlu257, Met233 and Ala262. Following SBP bind-ing, the hydrogen bond between Ala258 and Ser228is strongly perturbed. The energetic contribution ofthe bond between the amide group of Ala258 andthe carbonyl group of Ser228 is xK2.5 kcal/mol inthe free minichaperone. In the minichaperone–peptide complex, the larger separation betweenhelices H and I strains this bond, reducing itscontribution to about K1 kcal/mol. The amidegroup of Ala258 partially compensates the loss ofthis hydrogen bond through the formation ofhydrogen bonds with water (Figure 5). Disruptionof this hydrogen bond is part of the rearrangementin the helical region that leads to accommodation ofthe peptide. Ser228, which is involved in the switchof the chemical nature of the GroEL cavity, is moreexposed to solvent in the transition from theuncomplexed GroEL (T state) to GroEL-GroES-(ADP)7 (R00 state).13 For Ser228, the solvent-accessiblesurface area changes from 12 A to 44 A.13 Thus,breaking the Ser228–Ala258 hydrogen bond in theminichaperone-assisted folding is reminiscent of

Figure 5. Water hydrogen bonds formed by residues in helices H and I in the free minichaperone (open bars) and in theminichaperone-peptide complex (filled bars). (a) Fractional occupancy of hydrogen bonding sites. (b) Protein–waterhydrogen bond lifetimes at each site. The lifetimes are extracted using equation (5).


the structural transformations that take place in theGroEL cycle.

Instabilities in the minichaperone–SBP complexinvolve charged residues at the ends of H and Ihelices

The boundaries of the polypeptide binding site

terminate at the long charged side-chains (Arg231,Lys242, Glu257, and Arg268) located at the ends ofhelices H and I. Destabilizing interactions atresidues Arg231, Lys242, and Glu257 side-chainsdue to formation of hydrogen bonds with water arelikely to result in SP unfolding at long time-scales.Although hydration decreases at both the NEand NH2 hydrogen bonding sites of Arg231


(Figure 5(a)), the lifetime of water hydrogen bondsat these sites increases in the liganded state. Thisbehavior results in destabilizing hydrogen bondsbetween Arg231 and Pro12. The hydrogen bondingsites of Lys242 also experience a reduced hydrationupon complex formation, while the lifetime ofwater hydrogen bonds at these sites increases.This suggests an unfavorable reduction of flexibilityof the Lys242 side-chain.

A significant reduction of the water hydrogenbond lifetime occurs in the liganded state at theGlu257 amide site (Figure 5(a)). By contrast, thehydration of this site is only slightly reduced. Theseobservations suggest that water can freely exchangeat Glu257. The solvent exposure of Glu257 alsoresults in access of water at the amide site of Ala258,which is unfavorable. Taken together these resultssuggest that Glu257 in concert with Ala258 candestabilize the complex resulting in the SP release.It appears that the SBP unbinding proceeds fromboth ends of the hairpin and the turn. In view oftheir destabilizing behavior we propose that,Arg231 and Glu257 should be key factors in therelease binding mechanism of the SBP. The simu-lation results are in accord with the bioinformaticanalysis that showed that Glu257 is stronglyconserved as the chemical type, and Arg231, whilenot conserved, is replaced mostly by Gln and Lys.13

The mutation E257A effectively suppressesthe interaction between residue 257 and SBP.The overall geometry of the binding site in theunliganded state is largely unaffected over thetimescale of our molecular dynamics. However, inthe liganded state the separation between helices isreduced compared to the wild-type case. In theliganded state of the mutant E257A, a salt-bridge istransiently formed between Lys226 and Asp253.As a result, the minichaperone–SBP complex isweakened and the binding function of the mini-chaperone could be affected. A surprising effect ofthe E257A mutation is found at the opposite end ofhelix I. The hydrogen bond formed between Arg268and Gly8 in the crystal structure is slightly morestable than in the wild-type. A second trajectory ofthe mutated minichaperone was started from aconfiguration with this hydrogen bond present.However, the hydrogen bond is disrupted afternearly 2 ns.

The unbinding of a weakly binding peptide (WBP) istriggered by destabilization of interactions thatinvolve charged residues at the ends of helices Hand I

The high affinity of SBP for the apical domain(dissociation constant KDZ2 mM at 20 8C11) resultsin only rare peptide release events by thermalfluctuations on the simulation time-scales. Toascertain if instabilities at similar sites are observed,we performed MD simulations for a weaklybinding peptide (WBP) from the Chen–Siglerpeptide library whose KD exceeds 200 mM. Theinitial structure for the minichaperone–WBP

complex was generated using the minichaperone–SBP structure as a template (see Materials andMethods). During a 10 ns molecular dynamicstrajectory of the minichaperone–WBP complex, theWBP partially unbinds from helices H and I(Figure 6). The partial unfolding of WBP during arelatively short simulation is consistent with thepositive value of DDG. A lower bound to DDG, thedifference in the free energy of unbinding SBP andWBP, is:

DDGZ kBT lnðKðWÞD =KðSÞ

D ÞO2:8 kcal=mol (1)

As noted in the case of the minichaperone–SBPcomplex, hydrogen bonds between Arg231 and thepeptide are unstable. In the minichaperone–WBPcomplex, these bonds are Arg231 (NE site)–Thr12and Arg231 (NH2 site)–Thr12 (Figure 6). Formationof hydrogen bonds between Arg231 and watermolecules results in the complete loss of Arg231–WBP hydrogen bonds and the initiation of partialunbinding of WBP from the minichaperone. Waterpenetration at the amide site of Ala258 due todestabilizing behavior of Glu257 is even morepronounced for WBP than for SBP. The similardestabilizing interactions present in both theminichaperone–SBP and minichaperone–WBPsuggest that the unbinding occurs as a result ofthe loss of hydrogen bonds between Arg231 and thepeptide. Despite sequence variation in SBP andWBP these simulations lead to similar conclusions.The weaker binding explains the ease of unbindingof WBP and allows us to probe the regions causingthe instability of the WBP–minichaperone complex.

Discussion

Dual requirements for annealing action ofminichaperone

For the apical domain of GroEL to facilitate thefolding of the class II proteins there are twoopposing requirements. The helices H and I shouldensnare the non-native protein to render theminichaperone–SBP complex transiently stable.The bound SP should be released by a combinationof thermal fluctuations and water-mediatedinstability of certain regions of the minichaperone.The stability requirement is most easily satisfied bystrong interaction of the SP with the apical domain,whereas the need to release the SP is bestaccommodated by a weak SP–minichaperonecomplex. Thus, the annealing by the minichaperonecan occur only for a moderate value of the SP–apicaldomain interaction as proposed by BT7 as arequirement for efficient folding in a hydrophobiccavity. It is possible that, in vivo, the interactionsbetween the apical domain and the SPs such asrhodanese are too strong or too weak for mini-chaperones to be effective.

The dual opposing requirements are satisfied byfavorable interactions of the residues in the SBP

Figure 6. Partial unbinding of WBP (orange) from the minichaperone helices H and I (green) during 10 ns moleculardynamics simulations. Snapshots are taken (after equilibration) at: (a) 1.1 ns; (b) 4.7 ns; (c) 5.1 ns; and (d) 10 ns. Thehydrogen bonds between Arg231 and Thr12 are shown in (a).


with the bulky hydrophobic residues in helices Hand I, and by unfavorable solvent-induced insta-bilities in the charged residues at the ends of thesehelices. These opposing tendencies result in optimalstability values for the SP–minichaperone complexfor annealing to occur for class II proteins. For theintact system, the optimal stability for SP–GroELmust be in the range of a few kBT. Because of thelack of forced-unfolding of the SPs, the range ofstability must be considerably less than for SPinteraction with the intact nanomachine. As a result,stringent substrates cannot be rescued by theminichaperone alone.

Transient binding release (TBR) model forminichaperone annealing

The hallmark of the iterative annealing mechan-ism5 for the intact chaperonin system is that multiplerounds of capture, encapsulation, and release of theSP can result in sufficient yield of the native state.These steps are built into the architecture of the ring-like structure of the GroEL particles, which enablesSPs to unfold (at least partially) when the nature ofSP–GroEL interaction changes during the hemicycle.The present simulations allow us to propose avariant of the iterative annealing mechanism for the

Figure 7. Transient binding release (TBR) model for theminichaperone annealing action. The minichaperone isrepresented as an ellipse, highlighting the binding regionformed by helices H and I, and the polypeptide as a chain.In the first stage (I) the minichaperone binds the unfoldedpolypeptide, and during the second stage (II) theunfolded polypeptide is released in a different freeenergy state. Subsequently, the polypeptide partitionsrapidly either into the native state (which is no longerrecognized by the minichaperone), or into an unfoldedstate (which can undergo further processing by theminichaperone). Hydrophilic (hydrophobic) regions aredescribed by red (blue) color.


minichaperone action (Figure 7). According to thismodel, annealing, during one cycle of binding andrelease, occurs in two stages. The first stage issimilar to the capture of the SP by the GroEL.

Because the minichaperone has only one bindingregion the stability of the resulting complex incompromised. In addition to the favorable hydro-phobic interaction between SP and the mini-chaperone the complex is also entropicallystabilized because the non-native SP can adopt amultitude of conformations. The second stageinvolves release of the SP that puts it in a differentregion of the energy landscape from which it canpartition to the native state with probability F or gettrapped in another conformation with 1KF proba-bility (Figure 7). Upon repeating this process the yieldof the native state is enhanced. Just as in the iterativeannealing mechanism, the yield of the folded state inthe TBR isJNZ1K(1KF)n, where n is the number ofiterations. The release (stage II) is facilitated bysolvent-induced instabilities of the minichaperone.

The TBR mechanism also suggests that any SPthat can transiently interact with the intact nano-machine can be chaperoned to the native state evenif it cannot be encapsulated within the GroEL cavity.The model suggests that the combination of TBRand iterative annealing mechanism might beoperative in the rescue of SPs (aconitase, forexample18) that are too large to be fully encapsu-lated in the interior of GroEL.

Figure 8. Casein structure predicted by the Robettamethod. The two helices, having similarities to helices Hand I, are exposed to solvent. The putative binding site isshown in yellow. Solvent-exposed hydrophobic side-chains are shown in blue and charged/polar side-chainsat the binding site boundaries are shown in red.

Chaperone-like activity can also be mimicked bya-casein

Polypeptide binding and unbinding are crucialsteps for the annealing action of the GroEL–GroESchaperonin system or other cofactors that mimic

chaperone activity. Although the minichaperonefragment (191–345) differs considerably fromGroEL, this monomeric fragment is able to assistthe folding of rhodanese and cyclophilin A (CyPA).This suggests that cofactors that can mimic the TBRmechanism are sufficient for annealing someproteins. The molecular details of the TBR model,which have been elucidated through the presentsimulations, show that any protein (or othermacromolecules) that can transiently bind andrelease SP can have some chaperone-like activity.Indeed, Weissman and co-workers showed that thenon-chaperone a-casein has a similar effect onrhodanese and CyPA folding as the minichaper-one.4 They suggested that this may be due torecognition of the solvent-exposed hydrophobicsurfaces of the SP by a-casein. We quantified thelocal hydrophobicity of the minichaperone anda-casein using a Kyte–Doolittle type plot19 with awindow of seven residues and hydrophobicityvalues from Eisenberg et al.20 These results arecorrelated with the solvent-accessible surface areaper residue to distinguish exposed hydrophobicregions. Helices H and I are among the regions thathave both hydrophobic peaks in the Kyte–Doolittleplots and large solvent-accessible surface areas.


The structure of a-casein is not known. BLASTanalysis,21 using default options, shows thata-casein and the minichaperone share no sequencesimilarity. To offer a molecular explanation for theability of a-casein to assist folding we determinedits structure using the Robetta method.22 Amongthe ten structures, seven predict that helices in theregions 104–116 and 119–135 are exposed to thesolvent. These regions also have large hydrophobicresidues, which can serve to capture the SP.Interesting similarities exist between the putativebinding sites of a-casein and the GroEL apicaldomain. The pair of proposed (for a-casein) peptidebinding helices has a parallel packing with thecharged side-chains that are located at their endsjust as in the minichaperone (Figure 8). Both Robettaand PHD23 predict that a-casein should be helical.The analysis shows that a-casein must assist foldingof class II proteins by the TBR mechanism. Thesimilarity of the proposed SP binding helices ina-casein and the apical domain shows that even themolecular details of the annealing mechanism maybe identical. In particular, the dual requirements forSP binding and release are satisfied by thehydrophobic core that terminates at the chargedresidues just as in the apical domain of GroEL(Figure 8). The structural similarity betweena-casein and the minichaperone suggests that themolecular regions of binding and release havesimilar chemical character for the amino acids. Itshould be emphasized that any cofactor thattransiently binds and releases SPs can facilitatefolding of non-stringent substrates. In this sense,neither a-casein nor minichaperone is special.

Materials and Methods

Molecular dynamics simulations of the GroEL mini-chaperone with and without the SP in aqueous solutionwere performed using the CHARMM program24 and theCHARMM22 force field.25 The substrate proteins are thestrongly binding peptide (SBP) and a weakly bindingpeptide (WBP) from the peptide library of Chen &Sigler.11 We were interested in the conformationaltransitions that accompany the association of the peptidewith the minichaperone, but also in the alterations(compared to the bulk) in the structure of the peptidewhen it binds to the minichaperone. To dissect thesechanges, at the atomic level, we simulated four systems;namely, the minichaperone, a minichaperone–SBP (WBP)complex, and the isolated SBP.

Initial conditions

For the minichaperone, the coordinates of the crystalstructure corresponding to one subunit of a mini-chaperone fused with a 17-residue N-terminal tag(Protein Data Bank access code 1KID, containing GroELresidues 191–376 and residues K7 to K1, the part of theN-terminal tag that was resolved in the X-ray structure)15

were taken as the starting structure in the moleculardynamics simulations. The initial conformation for theminichaperone–SBP complex was taken from the PDBstructure 1DKD, representing GroEL residues 191–336

and a 12-mer synthetic polypeptide.11 To obtain thestructures of SBP in isolation we started our simulationsfrom several distinct initial conformations. The simu-lations of the isolated neutral SBP in aqueous solutionwere started either from the hairpin structure adopted bySBP in the minichaperone–SBP complex (taken from the1DKD X-ray structure), or from extended conformationsobtained by randomizing the dihedral angles (f,j) alongthe backbone (except for proline residues due to theirspecial side-chain). Steric clashes in the random confor-mations were avoided by rejecting conformations withlarge repulsive van der Waals interactions (O106 kcal/mol).

In order to study the WBP in complex with theminichaperone we chose a peptide with significantlylower affinity for the apical domain than SBP (peptide 3,with sequence SSPWWLVSFTST).11 Because the structureof this complex is not known, we built a model structurefor the minichaperone–WBP complex using the crystalstructure of the minichaperone–SBP complex as atemplate.

We prepared a net neutral system by adding 6 (5)sodium ions in the minichaperone (minichaperone–peptide complex) simulations. A sample of an equili-brated TIP3P26 water box was superimposed over theprotein/peptide systems, and water molecules within1.8 A of protein atoms or ions were deleted. A rhombicdodecahedral box was centered on the protein and all thewater molecules outside it were also deleted. Three-dimensional periodic boundary conditions wereimposed. We chose the simulation cell size so that theprotein was completely inside this box and the minimumdistance between the protein and any of its images(measured in terms of Ca–Ca distances) was at least 16 A.The total number of water molecules was 4802 and 2373(2374) for the minichaperone and minichaperone–SBP(WBP) complex simulations respectively. The linear sizeof the initial boxes, as determined by the length of the unitvectors, was 65 A (52 A). The solvation and initial size ofthe periodic cell in the isolated peptide simulations arelisted in Table 1.

Simulation details

To maintain the equilibrium volume of the simulationcell, we performed the simulations in the isobaric–isothermal (NPT) ensemble. Electrostatic interactionswere treated using the Ewald summation with a particlemesh algorithm.27 We integrated the classical equations ofmotion using the Verlet leapfrog algorithm with a 1 fstime-step. Water was allowed to adjust to the successivelyless restrained minichaperone through minimization andNPT dynamics at 300 K. Initially, we imposed harmonicrestraints with a force constant of 1 kcal/(mol A2) on allthe heavy atoms of the minichaperone (peptide in theisolated peptide case) and 50 steps of steepest descentminimization were performed followed by 400 steps ofadopted basis Newton–Raphson method. Next, hydrogenbonds were constrained to their ideal values usingSHAKE with an accuracy of 10K9 ps and 10 ps of NPTdynamics were performed. The restraints were removedin three stages: first the force constant was reduced to0.1 kcal/(mol A2) and 20 ps of NPT dynamics wereperformed, then the force constant was set to 0.01 kcal/(mol A2) and 10 ps of dynamics were run, and, finally, theforce constant was set to 0 kcal/(mol A2) for the last 20 ps.A further 50 ps of NPT dynamics was performed with norestraints. For each system we generated a number of10 ns trajectories (Table 1). In each trajectory, the first 1 ns


was not included in the data analysis, to allow the systemto fully equilibrate. For analysis purposes, we obtained atotal of 300 ns simulations. Molecular graphics imageswere produced using VMD30 and PDV-Ray†.

Data analysis

Radius of gyration

The probability density of the radius of gyration iscomputed using:

PðRgÞZ1

N

XNiZ1

1

ti

ðds dðRðiÞ

g ðsÞKRgÞ (2)

where N is the number of trajectories simulated and t isthe duration of each trajectory. The mean-square of theradius of gyration is hR2

giZÐR2

gPðRgÞdRg.

Hydrogen bond

We assume hydrogen bonds are formed if the distancebetween donor and acceptor atoms is less than 3.2 A, thedonor-H-acceptor angle is larger than 1308, and thehydrogen bond energy is lower than K2 kcal/mol.During the simulations the hydrogen bond energy is notincluded as an explicit term. For analysis purposes, wecalculate the energy of the hydrogen bond, involving theacceptor (A), hydrogen (H), and donor (D) atoms,according to the formula:24

Ehb ZA0

r6AD

KB0

r4AD

� �cos2ðqAKHKDÞ

! fswðr2AD; r

2on; r

2offÞfsw½cos2ðqAKHKDÞ; cos2ðqonÞ; cos2ðqoffÞ�

(3)

where rAD is the A–D distance, and qA–H–D is the A–H–Dangle. The switching function, fsw(x,xon,xoff), is 1 forx%xon, (xoffKx2)(xoffC2xK3xon)/(xoffKxon)3 for xon!x%xoff, and 0 for xOxoff. We chose ronZ3.2 A, roffZ3.4 A, qonZ1308, and qoffZ1108.

The fractional occupancy of hydrogen bonding sitesis:28

fH Z1

N

XNiZ1

NðiÞH

ti(4)

where NðiÞH is the total number of hydrogen bonds found in

the conformations during the trajectory i.To obtain the dynamics of hydrogen bond fluctuations

we also computed correlation functions of hydrogenbonds between water molecules and protein residuesusing:

ChðtÞZ1

N

XNiZ1

1

NðiÞDW

XNðiÞDW

kZ1

hpðiÞk ðtÞpðiÞk ð0ÞiK hpðiÞk pðiÞk i

sðiÞk s

ðiÞk

(5)

where NDW is the total number of different watermolecules making hydrogen bonds with the given site.The occupation number of the hydrogen bonding site attime t, is pkðtÞZ1=½1CexpðrkðtÞKRÞ� if QA–H–DO1308, and0 otherwise. Here, rk(t) is the A–D distance at time t andRZ3.2 A is the hydrogen bond cut-off distance. Inequation (5), skZ ½hðpkK hpkiÞ

2i�1=2 is the rms occupation

† http://www.povray.org

number and averages refer to time averages over theduration of each trajectory. Distinguishing between watermolecules is necessary because of the heterogeneity of theprotein–solvent interface. We determined the protein–water hydrogen bond lifetimes by fitting the resultingcorrelation curves using two exponential functions andextracting the longer (over-picosecond) time-scale.

Acknowledgements

This work was supported, in part, by a grant fromthe National Institutes of Health (number1R01GM067851-01) to D.T.

References

1. Horovitz, A., Fridmann, Y., Kafri, G. & Yifrach, O.(2001). Allostery in chaperonins. J. Struct. Biol. 135,104–114.

2. Makino, Y., Taguchi, H. & Yoshida, M. (1993).Truncated GroEL monomer has the ability to promotefolding of rhodanese without GroES and ATP. FEBSLetters, 336, 363–367.

3. Zahn, R., Buckle, A. M., Perrett, S., Johnson, C. M.,Corrales, F. J., Golbik, R. & Fersht, A. R. (1996).Chaperone activity and structure of monomericpolypeptide binding domains of GroEL. Proc. NatlAcad. Sci. USA, 93, 15024–15029.

4. Wang, J. D., Michelitsch, M. D. & Weissman, J. S.(1998). GroEL–GroES-mediated protein foldingrequires an intact central cavity. Proc. Natl Acad. Sci.USA, 95, 12163–12168.

5. Todd, M. J., Lorimer, G. H. & Thirumalai, D. (1996).Chaperonin-facilitated protein folding: optimizationof rate and yield by an iterative annealing mechanism.Proc. Natl Acad. Sci. USA, 93, 4030–4035.

6. Thirumalai, D. & Lorimer, G. H. (2001). Chaperonin-mediated protein folding. Annu. Rev. Biophys. Biomol.Struct. 30, 245–269.

7. Betancourt, M. R. & Thirumalai, D. (1999). Exploringthe kinetic requirements for enhancement of proteinfolding rates in the GroEL cavity. J. Mol. Biol. 287,627–644.

8. Thirumalai, D. (1994). Theoretical perspectives onin vitro and in vivo protein folding. In StatisticalMechanics, Protein Structure, and Protein–SubstrateInteractions (Doniach, S., ed.), pp. 115–134, PlenumPress, New York.

9. Orland, H. & Thirumalai, D. (1997). A kinetic modelfor chaperonin assisted protein folding. J. Phys. 7,533–560.

10. Jewett, A. I., Baumketner, A. & Shea, J. E. (2004).Accelerated folding in the weak hydrophobicenvironment of a chaperonin cavity: creation of analternate fast folding pathway. Proc. Natl Acad. Sci.USA, 101, 13192–13197.

11. Chen, L. & Sigler, P. B. (1999). The crystal structure of aGroEL/peptide complex: plasticity as a basis forsubstrate diversity. Cell, 99, 757–768.

12. Chothia, C., Levitt, M. & Richardson, D. (1981). Helixto helix packing in proteins. J. Mol. Biol. 145, 215–250.

13. Stan, G., Thirumalai, D., Lorimer, G. H. & Brooks,


B. R. (2003). Annealing function of GroEL: structuraland bioinformatic analysis. Biophys. Chem. 100,453–467.

14. Mountain, R. D. & Thirumalai, D. (1991). Ergodicmeasures for the simulation for dielectric propertiesof water. Comput. Phys. Commun. 62, 352–359.

15. Buckle, A. M., Zahn, R. & Fersht, A. R. (1997).A structural model for GroEL–polypeptide recog-nition. Proc. Natl Acad. Sci. USA, 94, 3571–3575.

16. Shtilerman, M., Lorimer, G. H. & Englander, S. W.(1999). Chaperonin function: folding by forcedunfolding. Science, 284, 822–825.

17. van der Vaart, A., Ma, J. & Karplus, M. (2004). Theunfolding action of GroEL on a protein substrate.Biophys. J. 87, 562–573.

18. Chaudhuri, T. K., Farr, G. W., Fenton, W. A., Rospert,S. & Horwich, A. L. (2001). GroEL/GroES-mediatedfolding of a protein too large to be encapsulated. Cell,107, 223–233.

19. Kyte, J. & Doolittle, R. F. (1982). A simple method fordisplaying the hydropathic character of a protein.J. Mol. Biol. 157, 105–132.

20. Eisenberg, D., Schwartz, E., Komaromy, M. & Wall, R.(1984). Analysis of membrane and surface proteinsequences with the hydrophobic moment plot. J. Mol.Biol. 179, 125–142.

21. Tatusova, T. A. & Madden, T. L. (1999). BLAST 2sequences, a new tool for comparing protein andnucleotide sequences. FEMS Microbiol. Letters, 174,247–250.

22. Chivian, D., Kim, D. E., Malmstrom, L., Bradley, P.,

Robertson, T., Murphy, P. et al. (2003). Automatedprediction of CASP-5 structures using the Robettaserver. Proteins: Struct. Funct. Genet. 53, 524–533.

23. Rost, B. & Sander, C. (1993). Prediction of proteinstructure at better than 70% accuracy. J. Mol. Biol. 232,584–599.

24. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States,D. J., Swaminathan, S. & Karplus, M. (1983).CHARMM: a program for macromolecular energy,minimization and dynamics calculations. J. Comput.Chem. 4, 187–217.

25. MacKerell, A. D., Jr, Bashford, D., Bellott, M.,Dunbrack, R. L., Jr, Evanseck, J., Field, M. J. et al.(1998). All-hydrogen empirical potential for molecu-lar modeling and dynamics studies of proteins usingthe CHARMM22 force field. J. Phys. Chem. B, 102,3586–3616.

26. Jorgensen, W. L., Chandrasekhar, J., Madura, J. D.,Impey, R. W. & Klein, M. L. (1983). Comparison ofsimple potential functions for simulating liquid water.J. Chem. Phys. 79, 926–935.

27. Essmann, U., Perera, L., Berkowitz, M. L., Darden, T.,Lee, H. & Pedersen, L. G. (1995). A smooth particlemesh Ewald method. J. Chem. Phys. 103, 8577–8593.

28. Berendsen, H. J. C., van Gunsteren, W. F.,Zwinderman, H. R. J. & Geurtsen, R. G. (1986).Simulations of proteins in water. Ann. New York Acad.Sci. 482, 268–286.

29. Fisher, M. E. (1966). Shape of a self-avoiding walk orpolymer chain. J. Chem. Phys. 44, 616–622.

30. Humphrey, W., Dalke, A. & Schulten, K. (1996).VMD–visual molecular dynamics. J. Mol. Graph. 14,33–38.

Edited by M. Levitt

(Received 19 January 2005; received in revised form 3 May 2005; accepted 4 May 2005)

Documents

Probing the “Annealing” Mechanism of GroEL Minichaperone using Molecular Dynamics Simulations