Structure, Vol. 10, 1085–1096, August, 2002, 2002 Elsevier Science Ltd. All rights reserved. PII S0969-2126(02)00808-0

Pathways and Intermediates in ForcedUnfolding of Spectrin Repeats

in experiments, in agreement with simulation. Spectrinrepeats may thus function as elastic elements, ex-tendable to intermediate states at various lengths.

Stephan M. Altmann,1,8 Raik G. Grunberg,2,8

Pierre-Francois Lenne,3 Jari Ylanne,4

Arnt Raae,5 Kristina Herbert,1,9 Matti Saraste,6

Michael Nilges,2,7 and J.K. Heinrich Horber1,7,10

1European Molecular Biology Laboratory IntroductionCell Biology and Biophysics ProgramMeyerhofstr. 1 Cells and tissues are commonly subjected to external

mechanical stress. The extracellular matrix and the69117 HeidelbergGermany membrane skeleton have evolved to tolerate such stress

while still providing the necessary elasticity. Much of2 Institut PasteurUnite de Bioinformatique Structurale the elasticity is probably supplied via the flexibility of

the molecular network in these matrices through interac-25-28 Rue du docteur Roux75015 Paris tions between proteins and, in the case of the extracellu-

lar matrix, carbohydrates. On the other hand, some in-France3 Institut Fresnel herited diseases are due to the weakening of tissues

caused by mutations in single proteins [1, 2]. Thus, theENSPMDomaine Universitaire de Saint Jerome elasticity of single molecules is obviously also involved.

In red blood cells, which have to resist particularly13 397 Marseille Cedex 20France large deformations, the rigid actin filament-based cy-

toskeleton is missing completely. In this particular well-4 University of Oulu and Biocenter OuluDepartment of Biochemistry studied case, shape integrity and elasticity are conferred

by other components of the membrane skeleton aloneLinnanmaa90570 Oulu [3]. The predominant building block of this two-dimen-

sional network is spectrin, a large tetrameric multido-Finland5 University of Bergen main protein composed of two � and two � subunits.

Its most prominent domain is the spectrin repeat, whichDepartment of Molecular BiologyHIB typically contains 106 amino acids. Each spectrin repeat

consists of three antiparallel � helices, which are sepa-P.O. Box 78005020 Bergen rated by loops and fold into a left-handed coiled coil [4,

5, 6]. About 500 spectrin repeats can be found in theNorway6 European Molecular Biology Laboratory human genome, most commonly in cytoskeletal actin

filament-associated proteins, such as spectrins, dys-Structural and Computational Biology ProgramMeyerhofstr. 1 trophin, utrophin, and �-actinins. These proteins usually

contain multiple, 4–40, spectrin repeats. Interestingly,69117 HeidelbergGermany many proteins containing spectrin repeats are found

in locations that are regularly subjected to mechanicalstress, for instance, the muscle Z band, muscle-base-ment membrane contacts, and the membrane skeletonSummaryin erythrocytes [7, 8] as well as the outer hair cells ofthe ear [9]. Genetic diseases caused by mutations inSpectrin repeats are triple-helical coiled-coil domainsspectrin and related proteins include elliptocytosis,found in many proteins that are regularly subjected tomuscular dystrophies, and myopathies [1, 2].mechanical stress. We used atomic force microscopy

Investigations of the mechanical properties of pro-technique and steered molecular dynamics simula-teins are progressing rapidly since the introduction oftions to study the behavior of a wild-type spectrinsingle-molecule techniques. Atomic force microscopesrepeat and two mutants. The experiments indicate that(AFM) [10], optical and magnetic tweezers [11, 12], andspectrin repeats can form stable unfolding intermedi-related techniques are being used to investigate howates when subjected to external forces. In the simula-macromolecular biopolymers with higher structural or-tions the unfolding proceeded via a variety of path-der react under an externally applied force. Recent ex-ways. Stable intermediates were associated to kinkingperiments have revealed many details of the nature ofof the central helix close to a proline residue. A mutantsequential unfolding of natural as well as engineeredstabilizing the central helix showed no intermediatesconstructs of modular proteins consisting of a numberof globular domains [13, 14, 15, 16]. Since the original

7 Correspondence: hhoerber@med.wayne.edu (J.K.H.H.), nilges@ work by Rief et al. in 1997 on titin, it has become widelypasteur.fr (M.N.) accepted that single domains of such linear proteins8 These authors contributed equally to this work.

will unfold in an all or none fashion when stretched9 Present address: Biophysics Program, Stanford University, Stan-ford, California 94305.10 Present address: Department of Physiology, School of Medicine, Key words: atomic force microscopy; molecular dynamics simula-

tion; spectrin; unfolding intermediate; elasticityWayne State University, Detroit, Michigan 48201.

Structure1086

between a surface and an AFM tip [13]. Most of the AFM the other hand, alternative topologies could regain me-chanical stability at much longer extensions. These dis-studies [17, 18, 19, 20] have focused on two domains

that have to sustain high forces in their physiological tinct properties may be important in the elastic behaviorof cellular structures containing this domain.environment. Both immunoglobulin (Ig) type I domains

and the fibronectin type III (FnIII) domain are all-� pro-teins and have a similar (Ig-like) topology. Both domains

Resultsare found in the giant muscle protein titin. Another well-studied example, the extracellular matrix protein tenas-

Forced Unfolding of Wild-Type Spectrin Repeatscin [21], consists mainly of FnIII domains. Molecularby AFMdynamics (MD) simulations performed on these � sand-When a single protein is attached to a surface and thewich structures of Ig-like domains have reinforced thetip of an AFM cantilever, it can be stretched, and thenotion that globular protein domains support only littleelongations as well as the forces necessary for the defor-deformation under external force. A catastrophic eventmation can be measured in real time, down to the milli-then transforms the proteins rapidly into an unorderedsecond timescale. A peak in the AFM force extensionpolypeptide peptide chain with little or no higher struc-profile marks the breakup of a stable conformation ofture left [22, 23, 24, 25]. This view was slightly revisedthe peptide chain. In a protein construct with multiplein one study [26], which suggested that Ig type I domainsidentical domains, the repeated sawtooth pattern in themay display a transient kinetic intermediate during con-AFM force extension curve is an indication of repeatedstant-speed forced unfolding, based on experimentalunfolding events specific for the domains contained indata supported by MD simulations. Since this intermedi-the construct. The spacing between peaks correspondsate occurred immediately before the catastrophic eventto the gain in length due to the unfolding of a part ofafter only a few angstroms of deformation of the domain,the protein chain.it did not bring the notion of all or none unfolding into

In earlier experiments we demonstrated that, in con-question [25]. On the other hand, Best et al. demon-trast to all other work published up to then, forced un-strated, in a recent study on forced unfolding of barnase,folding of engineered spectrin constructs may also showthat such regular all or none behavior cannot necessarilyelongations significantly shorter than those of com-be expected from “nonmechanical” proteins [27]. More-pletely unfolded domains [30]. Further measurementsover, simulations by Paci and Karplus showed that theon the wild-type spectrin repeat confirmed our original� sandwich as well as � helix structures could theoretic-findings. Our peak to peak distance results are summa-ally create stable forced-unfolding intermediates atrized in the elongation histogram (Figure 1C). Our forcemuch larger extensions [28, 29]. These results supporteddistributions were quite broad and peaked between 50our earlier studies on forced unfolding of spectrinand 80 pN (data not shown). Examples of force curvesrepeats [30].are shown in Figures 1A and 1B. One can observe con-In our previous study, we subjected an engineeredsecutive unfolding events with various peak to peakconstruct of four identical copies of the 16th repeat ofdistances. We have indicated several peak to peak dis-� spectrin to forced unfolding by AFM [30]. In contrasttances between unfolding events. We did observe ato all prior AFM work, our study reported two distinctsizable fraction of more regular curves of the type shownpopulations of unfolding events. Because the distribu-in Figure 1A, representative of the full-length unfoldingtions of unfolding forces and elongations were quitemaximum at 32 nm shown in the histogram in Figurebroad in comparison to those from studies on titin and1C. However, the majority of experiments showed quitedue to the lack of further insights from simulations atsome variation of peak to peak distances between un-the time, we only tentatively suggested that unfoldingfolding events. Two representative examples are givenevents at elongations around 15 nm could represent ain Figure 1B. The unmarked leftmost peak on the firstforced-unfolding intermediate. In order to either dismisscurve is not used as a reference because it appears toor support this suggestion, we here relate experimentalbe part of an adhesion event, sometimes observed atmeasurements and MD simulations of forced unfoldingthe beginning of force curves. In the second curve weof the wild-type spectrin repeat and two designed mu-count five unfolding events, i.e., one more than the num-tants. In one mutant, we crosslinked helices A and B byber of domains in the construct, while the total lengthintroducing a disulfide bridge in order to stop unfoldingof unfolding is less than the maximum stretched lengthat an elongation close to that of the proposed intermedi-of the construct. As we reported already in [30] and asate. AFM results with this mutant indicated that we couldillustrated in Figure 1B, shorter peak to peak distancesclearly distinguish the shorter unfolding events from theare often detected in the beginning of the experiments.full elongation of the spectrin repeat, thus validating our

The features of these representative unfolding curvesprevious measurements. Our simulations also showedindicate a complex unfolding behavior of the tetramericintermediates of appropriate length along one unfoldingspectrin construct. A distribution of peak to peak dis-pathway, in which nonnative topologies formed due totances developed during the stretching process in thea pronounced kink of helix B in the center of the mole-majority of experiments. Such behavior cannot be ex-cule. We engineered a second mutant with a strength-plained by independent all or none unfolding of the com-ened helix B and demonstrated that this mutation pre-plete spectrin domain. On the contrary, not only do sin-vented the occurrence of unfolding intermediates.gle force curves, as shown in Figure 1B, already exhibitThe spectrin repeat appears optimized for an elasticvarious distances, which cannot be attributed to purelyresponse to force in two ways: on the one hand, the

native fold sustains considerable elongations, and, on statistical variations from experiment to experiment

Unfolding Intermediates of Spectrin Repeats1087

which also received some support from a theoreticalstudy [29].

Simulations of Forced Unfolding of Wild-TypeSpectrin RepeatsWe simulated mechanical unfolding of the wild-typespectrin repeat with a constant extension rate of 0.2 A �ps�1 (2 � 107 nm � ms�1) by molecular dynamics. Themolecule was extended to a distance rNC between theN and C termini of about 24 nm, corresponding to anelongation of the native structure by 20 nm. This valueis smaller than the molecule’s maximal extension, aswe were mainly interested in the rupture of the tertiarystructure. We calculated 11 trajectories of 1 ns each.Unfolding was enforced by a time-dependent harmonicdistance restraint between the N terminus and the Cterminus. The molecule’s resistance toward unfoldingis illustrated in force extension profiles (Figure 2). Inanalogy to the displacement of an experimental cantile-ver, restoring forces were calculated from the deviationbetween rNC and the target value to which it was re-strained.

The maximum force recorded in each trajectoryshowed a broad variation from 400 to 645 pN, with anaverage of 475 pN (�65 pN). In contrast to other systemsstudied by AFM and MD [24, 25, 28], the extension atwhich this peak force occurred was not well defined butvaried between 6.5 and 16 nm rNC. Moreover, severalpeaks of similar force were often observed at differentextensions. In Figure 2, we arbitrarily marked all forceswithin 10% of each trajectory’s maximum to illustratethis fact. The varied positioning of such peaks at exten-sions up to 11 nm apart has important implications forthe variation of unfolding lengths in AFM experiments.

Unfolding was typically initiated by a gradual stretch-ing of the two outer helices, A and C, with varying re-storing forces. Hence, the native triple-helical coiled-coil topology sustained considerable elongations. Thisphase culminated in the disruption of the native fold,marked by a drop in the force extension profile. Severalunfolding trajectories proceeded without further pro-nounced resistance (simulations d, i, l, and m; see Figure2). However, six simulations exhibited distinct forcepeaks after the initial unfolding event and in two cases(e.g., simulation h) such peaks constituted the maximumforce of the whole simulation. These peaks were causedby compact nonnative folds that provided interim me-chanical stability at various extensions.

In order to describe the origin of these potential inter-Figure 1. Forced-Unfolding Curves of Spectrin Constructs Made of mediates, we focus on shifts in the molecule’s topologyFour Identical Wild-Type Repeats

rather than analyze the complex atomic detail of our(A) A sizable fraction of the curves display regular unfolding lengths simulations. The native spectrin repeat consists of threearound 32 nm, but (B) the majority of curves show varying unfolding

secondary structure elements, helices A, B, and C,lengths displaying intermediate and also full-length peak to peakwhich are connected by two loops, AB and BC (Figuredistances.3). These loops are the two obvious hinge points in the(C) Distribution of elongation lengths from forced-unfolding experi-

ments on a tetramer of R16 wild-type spectrin repeats (n � 250). repeat’s topology that need to open up to completelyunfold the molecule. Before that, however, our simula-tions showed a limited melting of helix B near Pro62, in

alone, but, even including the averaging effect of the the center of the molecule. Increasing a naturally oc-histogram of peak to peak distances, Figure 1C also indi- curring bend in this position, this helix was effectivelycates a significant fraction of partial unfolding events. In divided into two. Due to this additional hinge, the repeat[30], we could only speculate that this might possibly could evade complete unfolding beyond lengths of

about 13 nm.point to the formation of forced-unfolding intermediates,

Structure1088

Figure 2. Force-Extension Profiles from Unfolding Simulations with the Wild-Type Spectrin Repeat

Three profiles are shown in detail. The complete data recorded at 0.5 ps resolution are drawn in blue. Black is a running average, with a 50ps window size. The running averages of all simulations are presented in the lower-right plot (note the shifted force scale). Force peaks within10% of a trajectory’s maximum force are marked with circles and are included for all simulations in the lower-right plot. They illustrate thevariance of extensions sustained by the repeat.

In Figure 4 we describe the status of each of these Two main routes emerge from a variety of pathsthrough this topology landscape. Along those routes wethree hinges by a separate axis. Each conformation ofidentified four regions of distinct topologies. Examplea given trajectory corresponds to a point in this three-structures for each of these topologies are marked withdimensional graph. Transitions along axis AB indicateletters and shown in Figure 5. The first region (illustratedthe opening of loop AB, transitions along axis BC showby snapshots a, b, and c) depicts the native fold ofthe disruption of loop BC, and the third axis, B, describesthe spectrin repeat. This native topology exhibited highthe closing and opening of the additional hinge in helixforce resistance in a wide range of elongations. A promi-B. We can thus follow the topology of the repeat in thenent second region of force-resistant topologies is de-course of each trajectory. In addition, we color codedpicted by snapshots d and g. It features the sharp kinkthe smoothed force profile from Figure 2 onto each tra-in helix B and, hence, appears shifted along axis B injectory trace (blue for lowest forces, red for highestFigure 4. Due to this kink, compact structures formedforces). From the resulting plot one can easily pick outfrom the remaining part of helix C, the two halves of(1) topologies that exhibited pronounced force resis-helix B, and a distorted, but still persisting, loop AB. Intance and (2) how often these topologies were visitedterms of extension this region partly overlaps with thein the 11 simulations.longest-observed native folds. Snapshots c and d pro-vide an example of this extension range, where highunfolding resistance could stem from a still-native-liketopology in one simulation. Similar resistance wasevoked by the nonnative arrangement in five other simu-lation runs.

The two remaining regions of prominent topology be-long to the two dominant unfolding pathways. In threetrajectories loop BC was disrupted first, and the kink inhelix B was straightened. Loop AB persisted over aconsiderable range of extensions (snapshot j), but thisconformation displayed only moderate resistance to fur-ther unfolding. In the other, most frequent, pathway,loop AB opened up first. The kink in helix B allowed forpronounced force peaks from a helix-loop-helix element(snapshot e) consisting of the remaining parts of helicesFigure 3. Structure of the Spectrin Repeat R16B and C locked together by the native BC loop. The

C� atoms of the two residues substituted in the AA mutant arelongest intermediate observed originated from thishighlighted in green. The disulfide bond introduced into the CCstructure and had its maximum unfolding resistance atmutant is marked in blue. Blue lines describe the three angles that

were used to represent hinge movements in Figures 4 and 9. 21 nm rNC. In this case, the helix-loop-helix element was

Unfolding Intermediates of Spectrin Repeats1089

Figure 4. Topology and Force Resistance of the Unfolding WT Spectrin Repeat

Trajectories are described by three hinge movements (see Figure 3). The complete recording is shown on the left, where trajectories arecolored as in Figure 2. For the right plot, angles and forces were averaged over a 50 ps window, and forces were then color coded onto thesmoothed topology traces. Conformations exhibiting high restoring forces are prominent around the native structure (a) and the completelystretched endpoint (k). Moreover, we observed two force-resistant nonnative topologies (d and g and e and f). Structure snapshots for eachtopology region are marked with letters and given in Figure 5.

tightly folded back onto the N-terminal half of helix B, followed neither of the two pathways strictly but left theinitially chosen main route by a premature opening ofand the resulting fold resembled spectrin’s native triple-

helical coiled coil (snapshot f�). The detailed force profile loop BC.Two simulations were continued to complete exten-of this simulation (j) is given in Figure 2. Three trajectories

Figure 5. Characteristic Topologies of the Unfolding Spectrin Repeat

Snapshots were chosen to describe regions of topologies emerging from Figure 4. Structures marked with an asterisk (*) directly correspondto force peaks. Simulations started from the NMR structure a; snapshots b and c were taken from simulation g; snapshots f, f�, and k weretaken from simulation j; snapshots g and h were taken from simulation h. The detailed force profiles of these three simulations are given inFigure 2. The structures have the following lengths (rNC in A): a, 42; b, 71; c, 135; d, 126; e, 173; f, 191; f�, 205; g, 158; h, 175; i, 143; j, 192;k, 241.

Structure1090

Figure 6. Equilibrium Properties of WT and Mutant Spectrin Re-peats

(A) CD spectra of all spectrin repeats in micromolar concentrationat room temperature: wild-type (WT), alanine-alanine (AA), and thecysteine-cysteine mutants without (CC) and with 10 mM DTT (CCDTT).(B) Percentage of unfolded values during a temperature ramp moni-tored at 222 nm for these repeats. The thermal denaturation wasnot reversible.

sion of the molecule, marked by a steep linear increaseof restoring forces (data not shown). In experimentalreality, unraveling domains can only be extended up tothe disruption of the next folded structure, the weakestelement of the chain. In our simulations we assumedthis to be the case if a running average of restoringforces (as described in Figure 2) surpassed a threshold

Figure 7. Forced-Unfolding Curves of Spectrin Constructs Made ofof 300 pN. The resulting maximal distances between theFour Mutated Cysteine-Cysteine Repeats in Oxidized and ReducedN and C termini were 35.8 and 35.9 nm, with the actualFormnumbers being rather insensitive to the choice of defini-Probability distribution of elongation events after unfolding eventstion or force threshold. Due to the initial distance be-of cysteine-cysteine constructs in oxidized (C) and reduced formtween the N and C termini (4 nm), this would correspond(D) (n � 96). The lines in (C) and (D) each correspond to a fit by a

to a length gain of 32 nm in AFM experiments. normalized sum of two Gauss functions.

Forced Unfolding of Mutant Spectrin Repeatsby AFM that in the wild-type experiments. The forces were again

broadly distributed with a wide maximum, ranging fromThe principal use of the first mutant (CC) was to provethat our AFM measurements were sensitive enough to 40 pN to 60 pN. Thus, the forces are similar to those

measured during the unfolding of the wild-type spectrin-detect partial unfolding events. We inserted two cys-teine residues at positions 17 and 67 on helices A and B, repeat tetramers. The distribution of elongations shown

in Figure 7C, on the other hand, revealed that, by far,respectively, because the side chains of these residueshave approximately the correct distance and conforma- the most probable event (�85%) was an elongation of

�14 nm. We only rarely detected unfolding events oftion in the wild-type to make the formation of a disulfidebridge likely. Correct folding of the mutant was indicated the entire domain, giving an elongation of 32 nm (�15%),

which we ascribed to unformed disulfide bonds. Weby the circular dichroism (CD) spectra of Figure 6A.We recorded force curves during stretching of mu- then added 10 �M DTT to reduce the disulfide bond.

Figure 7B shows some typical curves after chemicaltated constructs, as shown in Figure 7A. During retrac-tion of the tip from the surface, a maximum of four well- reduction. The most probable elongation events were

32 nm, equivalent to the total unfolding length of thedefined peaks were detected. However, the form of thecurves near force peaks appeared more irregular than wild-type, but we could also detect a significant number

Unfolding Intermediates of Spectrin Repeats1091

of partial unfolding events around 15 nm. The resultsare again summarized in the histogram of Figure 7D.The crosslink of helices A and B should reduce theunfolding length of the spectrin repeat by 18 nm. Ourmeasurements of the CC mutant reproduced this valuewith high accuracy, despite the broad variation of partic-ular unfolding lengths. The reduced number of forcedintermediates may be explained by an overall destabili-zation effect, as the two cysteines replaced a hydropho-bic cluster in the protein. This is supported by the CCmutant being thermodynamically less stable than thenative spectrin repeat, evident from the lower transitiontemperature for thermal unfolding observed by CD spec-troscopy in Figure 6B. In solution, only a fraction ofpotential disulfides in the CC mutant was formed, butthe formation of disulfides could be enhanced in suitableoxidizing conditions (data not shown). The AFM results,however, suggest that, during adsorption on the metalsurface, most cysteines formed disulfides.

Figure 8. Probability Distribution of Elongations after UnfoldingThe AA mutant was suggested by our MD simulations,Events of Mutated Alanine-Alanine Spectrin Constructs (n � 120)which indicated a hinge hidden in helix B and the moder-The peak at 15 nm, present for the wild-type in Figure 1C, has beenately conserved occurrence of proline and glycine resi-almost completely diminished. A single Gauss fit gives a maximum

dues toward the middle of this helix in sequence align- peak at 31 nm. The majority of force curves are mostly similar toments of spectrin repeats. It seemed plausible to try to the two examples for the wild-type in Figure 1A, which show onlystrengthen this central part of the B helix by replacing full length peak to peak distances. Curves with unfolding lengths

distinctly shorter than 32 nm, like those in Figure 1B, occurred mucha proline residue at position 62 and a glycine residue atless frequently.position 66 with two alanine residues.

The unfolding forces measured in the AFM were dis-tributed broadly, with a maximum between 40 and 70

mutations on the protein’s backbone conformation. Dur-pN (data not shown) and were thus again similar to thoseing equilibration helix B straightened, and unfoldingfor the wild-type. However, significantly fewer curvesstarted, on average, with helix B angled at 134.5 (�3.8),displayed shorter peak to peak distances, such as thosecompared to 122.4 (�4.9) in the case of the nativeshown in Figure 1B. The majority of peak to peak dis-repeat. The maximum force observed during unfoldingtances corresponded to complete unfolding events inwas, on average, 483 � 22 pN, similar to the nativethe wild-type measurements, as shown in Figure 1A.spectrin repeat. However, the trajectories exhibitedThe resulting histogram of peak to peak distances isfewer force peaks, and, with one exception, all of themgiven in Figure 8. The distribution of elongations peaksoccurred before disruption of the native fold.around 31 nm, similar to the second peak in Figure 1C.

In terms of topology, all trajectories closely followedIn contrast to that in the wild-type, the peak to peakone of the two main unfolding pathways that had alreadydistribution does not contain a statistically significantemerged from the wild-type simulations. However, Fig-peak at 15 nm. The AA mutant also proved thermallyure 9 reveals that the relative importance of the twomore stable than the native spectrin repeat, as is evidentroutes has reversed. In all but one trajectory, loop BCfrom the higher transition temperature during CD-moni-was disrupted first, and unfolding proceeded via thetored thermal unfolding (Figure 6B).topology depicted by snapshot j in Figures 4 and 5. Inone case, such a topology caused high restoring forcesaround 22 nm rNC, a possibility that we had not observedSimulations of Forced Unfolding of Mutantbefore. By contrast, wild-type simulations had shownSpectrin Repeatstwo other regions of nonnative topology with high resis-We derived a model for the CC mutant from the water-tance against unfolding. Only one trajectory of the AArefined NMR structure and calculated five unfolding tra-mutant followed the formerly most important route andjectories, each covering 0.75 ns. The connection of heli-visited these topologies, which, however, unraveledces A and B limits unfolding to a maximal extension rNC

without pronounced resistance. In summary, the twoof about 19.5 nm. Employing the same conditions asmutations in helix B appeared to disfavor structures thatbefore, we observed a linear force increase at exten-depended on the proposed hinge. Consequently, theysions beyond 17 nm, and our simulations were stoppedsteered unfolding away from the topologies that hadat 19 nm. However, even before this, a variety of forceproduced all long-forced intermediates in the wild-typepeaks appeared. Maximum forces (before 16.5 nm rNC)simulations.were on average 498 � 39 pN. The maximal extension,

determined as described for the wild-type, was, on aver-age, 18.1 � 0.1 nm. Discussion

The AA mutant was modeled from the same NMRstructure and subjected to five unfolding simulations Variation in Unfolding Lengths and Forces

Peaks, or, rather, the sudden drop of force in AFM exper-employing the identical conditions as for the wild-type.No assumptions were made about the influence of these iments, mark the breakup of a single domain [31]. In a

Structure1092

Figure 9. Unfolding Pathways of Mutated Spectrin Repeats

Hinge movements are defined in Figure 3. Restoring forces are color coded onto the smoothed angle traces, as described in Figure 4. Right,AA mutant; left, CC mutant. The hinge movement of helix B is restricted in the AA mutant, and unfolding is steered away from the pathwaythat was most frequent in simulations of the wild-type repeat.

chain consisting of identical repeats, all of the repeats rupture behavior under a vectorial force: the distributionof unfolding lengths in the AFM data is too broad to beshould likewise be on the verge of breaking at the time

of an unfolding event. Hence, the distance between two explained by experimental error alone, and the initialforce peaks in MD are spread over a significant rangeforce peaks in the experiment corresponds to the gain

of length due to unraveling of the folded content of a of unfolding distances. Because of this rather undefinedbehavior, spectrin repeats have been regarded as “forceprestretched domain (as pointed out in [24]).

A common feature of all our AFM experiments is the compliant” by authors comparing them to Ig-like do-mains [29, 31]. This kind of behavior of a single proteinbroad distribution of those peak to peak distances. Such

variation might be, in part, attributed to the low force domain can also be called elasticity. We propose thatthis property has an important role in the physiologicalneeded for a single unfolding event (50–80 pN). This is

much lower than the force measured for the titin Ig-like behavior of the domain in proteins that are subjectedto mechanical stress.domain and required us to work close to the sensitiv-

ity limit of the AFM instruments. Nevertheless, theshortening of the crosslinked CC mutant by 18 nm wasaccurately measured by the average elongation, and Translating Simulation to Experiment

There are certain difficulties in comparing the simula-chemical disruption of the disulfide bond reproducedthe wild-type average elongation of 32 nm. Thus, in tions and AFM experiments. First, because of the vari-

ability encountered, 11 simulations for the wild-type re-spite of the broad variation, we could determine thedifferences in unfolding lengths with confidence. In the peat seem insufficient to obtain statistics comparable

to the experiment. Second, we only simulated a singleMD simulations, on the other hand, initial force peakswere distributed seemingly at random between elonga- domain, and, in a polymer of four spectrin repeats used

in the AFM experiments, the statuses of the completelytions (from the NMR structure) of 1.5–11 nm. This sug-gests that the broad variation of unfolding lengths or partially folded, but prestretched, domains are unde-

fined at the time of single-domain rupture. Furthermore,observed in the experiment not only stem from experi-mental error, but also reflect variable prestretching of it is unknown to what extent the partially distorted do-

mains refold when the force drops. Third, similar to thosethe domain. The precise unfolding length is, in this sce-nario, obscured by the undefined prestretching of the in other studies [22–25, 27, 32], elongation speeds of

experiment and simulation differ by almost eight ordersunfolding domain itself and its replicas in the proteinchain. of magnitude (3 � 10�7 versus 20 m � s�1). Despite

this discouraging gap between the experimental andIn both experiment and simulation, the diverse elonga-tions were, furthermore, accompanied by a broad distri- computationally accessible timescale, MD simulations

have been instrumental for the interpretation of un-bution of unfolding forces. The absolute force valuesare much lower than those observed for Ig-like proteins. folding experiments and, in various instances, provided

results consistent with AFM data [18, 26]. In general,Thermal contributions are certainly playing a more pro-nounced role in this force range and could act as a an unfolding scenario is reproduced regardless of the

simulation details (pulling speed, pulling force, and for-random generator for selecting the actual prestretchingand force resistance of a particular domain. It becomes mulation of the restraint) [24, 28, 33]. Even quantitative

statements about the height of energy barriers can beclear, even without relying on these assumptions, thatthe spectrin repeat does not show a single catastrophic made [33] if one takes into account that simulations

Unfolding Intermediates of Spectrin Repeats1093

operate much further away from equilibrium [34] than 15 nm are, in summary, supported by several observa-tions. (1) AFM measurements of wild-type repeats fea-do AFM experiments.

Lu and Schulten criticized the application of implicit ture a second prominent unfolding event around a 15.5nm gain of length. (2) Force-resistant structures of ap-solvent models (such as the generalized Born continu-

ous solvent model that we employed in our study) to propriate length form in MD simulations, thanks to ahyphenation point in the center of the molecule. (3) Twounfolding simulations, since they cannot describe the

detailed mechanism of (concurrent) hydrogen bond point mutations eliminate the short unfolding event bystiffening this hyphenation point. (4) More than fourbreaking [25] However, it has been argued that artifacts

due to high unfolding speeds in the simulation are allevi- peaks were observed in some force curves, but withinthe expected maximum stretched length (compare Fig-ated by the fact that implicit solvent models mimic equili-

brated water at each step of the simulation [28]. Apart ure 1B), and “half” unfolding events occur mainly inearlier extension states at, on average, lower forces [30].from practical considerations (the size of the water bath

necessary to simulate the complete unfolding of a spec-trin repeat and the necessary CPU time) the approxima-

Comparison with Other Studiestion may well be an advantage, since a slow step in theRegular peak spacing and nearly constant ruptureunfolding (equilibration of water around the unfoldingforces (200 pN at 1 �m � ms�1) found in many studiesprotein) is not simulated. The generalized Born modelon Ig-like domains from titin and fibronectin have estab-has been demonstrated to be a good approximationlished a standard for the type of results commonly ex-of explicit solvent simulations of proteins (see, forpected from such experiments [13, 17, 21]. In theseexample, [35]).studies, force curves were fitted to the worm-like chainIn summary, despite of the difficulties in direct com-model [13, 36], implying a singular well-defined ruptureparison between AFM experiments and MD simulations,that transforms the native fold at once into a completelythe broad distribution of the position of force peaksdisordered protein chain held together by purely en-could be seen in both. Moreover, MD simulations gavetropic effects. MD studies provided consistent mecha-valuable insight in the unfolding pathways and providednisms that convincingly supported such a defined allus with a way to experimentally test these pathways.or none unfolding of domain I27 from titin [26]. SimilarThis is further discussed below.predictions were initially made for FnIII domains, al-though initial prestretching by 1–2 nm has now beensuggested from more recent simulations [32], and two-Unfolding Pathways of Spectrin Repeats

The fluctuating topology of the spectrin repeat in the step unfolding was observed in another simulationstudy [28].simulation depended mainly on the status of three hinge

regions: loop AB, loop BC, and a potential kink in helix Our experimental and theoretical results clearly showthat spectrin repeats react to an external force veryB. These three parameters provide a concise view of

the trajectories (Figure 4). One nonnative force-resistant differently than the Ig-like domains: (1) spectrin repeatscan be stretched to varying extensions before the rup-topology partly occurred at extensions that were still

sustained by native-like folds in other simulations. ture of the triple-helical fold, and (2) several unfoldingpathways exist, and some of them may lead to force-Hence, one should not observe intermediates of this

type in AFM experiments. resistant nonnative intermediate folds. For both rea-sons, the worm-like chain model and, especially, theHowever, other force-resistant topologies appeared

at extensions that could well explain the additional ex- filtering of AFM results for regular peak spacing werenot appropriate to describe the unfolding behavior. Ex-perimental peak (snapshots e and f). Disruption of these

topologies would result in length gains of 15 nm or more. perimental artifacts, such as pickup of multiple moleculesor surface interactions, could be ruled out, due to theAll force-resistant topologies (Figures 4 and 5, d and g

and e and f) critically depended on the additional hinge special experimental procedures described earlier [30].Two experimental as well as two computational stud-inside helix B.

To test the role of B helix bending in the unfolding ies previously examined the forced unfolding of spectrinrepeats, and, on both sides, there remained disagree-pathways, we introduced two mutations that stabilized

helix B at its potential hyphenation point. In the experi- ment about whether two-step unfolding (i.e., forced in-termediates) can occur or not. Rief et al. were the firstment, elongations of this mutant peaked around a single

value identical to the long unfolding event of the wild- to study unfolding of spectrin repeats with AFM [16].They used a hexameric construct of nonidentical re-type repeat. Simulations explained this lack of interme-

diates with a switch to the unfolding path without the peats and analyzed their data with the worm-like chainmodel. Concerned about pickup of multiple molecules,late bending of helix B. Hence, the effect of the AA

mutation supports the suggested unfolding pathways they only considered force curves with evenly spacedpeaks. Consequently, they concluded that the repeatwith the proposed hinge in helix B, although it does not

exclude alternative roles of the replaced proline in a has a wider unfolding barrier than Ig-like domains but,nevertheless, unravels in a single step. By contrast,particular intermediate. However, a more general pertur-

bation of the wild-type would not explain our results, Lenne et al. studied the described tetrameric constructof identical repeats and used different criteria to filtersince the mutations actually stabilized the native struc-

ture thermodynamically (compare Figure 6) and did not out multiple pickups [30]. They found both complete andhalf unfolding events and also reproduced this bimodalalter the observed rupture forces.

Forced intermediates around an average extension of distribution with Rief et al.’s hexameric construct.

Structure1094

On the theoretical side, Paci and Karplus [29] sub- Biological Implicationsjected the spectrin repeat to MD unfolding simulations.They used a different force field model, a rather different The spectrin membrane skeleton provides mechanical

strength and elasticity to the cell. It is well studied informulation of the pulling restraint, a faster unfoldingregime, and a more-approximate model of solvation erythrocytes, where spectrin filaments connect a multi-

tude of actin nodes in a two dimensional network under-than we did in our work. Nevertheless, their simulationsseem to resemble our results in that they lack a singular neath the membrane [3, 40]. Experimental evidence and

theoretical calculations indicate that the spectrin �2�2well-defined force peak. The published average forceprofile also indicates high initial restoring forces distrib- tetramers are the elastic part of the network. On the

other hand, the network itself and modest shifts in itsuted between 5 and �15 nm rNC. They suggested anintermediate to sometimes occur before this destruction connectivity may also explain part of the elasticity [41].

Changes in the super coiling of its � and � subunitsof the original helix arrangement. Although we could notreproduce their suggested � hairpin structure, we did could provide elasticity on the level of the single-spec-

trin tetramer [42]. Changes in the interactions betweenobserve similar force-resistant structures at the de-scribed length. Figure 5C shows such a state that particular domains are very likely contributing another

level of elasticity, and there is recent evidence of stabiliz-caused a near-maximum force peak at 13.5 nm rNC. Paciand Karplus did not report later intermediates, which, ing interactions between neighboring spectrin repeats

[43]. Moreover, a crystal structure of repeats 16 and 17,however, are necessary to explain the observation ofpeaks separated by about 15 nm. obtained by Grum et al., shows an alternative location

of loop BC leading to an overlap and, hence, shorteningKlimov and Thirumalai obtained predictions from amuch more simplified computational model. They com- of the two domains [5]. Thus, there is experimental evi-

dence for structural changes within single repeats. Ourpared the native interactions between complete second-ary structure blocks (in this case, helices A, B, and C), results add yet another level to this picture and suggest

that partial unfolding of single spectrin repeats mightwhich they assumed to each unfold in all or none fashion[37]. According to this model a spectrin repeat would be an additional source of elasticity. We detected half-

extended, yet stable, intermediates that form thanks tounravel starting with helix A, without exhibiting any inter-mediates. They assumed that it is mainly the native to- a hyphenation point in helix B. Spectrin repeats often

feature proline and glycine residues at this position orpology that defines a protein’s reaction to externalforce—a picture supported by previous simulations of at neighboring positions, and the helix is kinked in six out

of seven known structures of spectrin repeats. Recentforced [24, 29] and thermal [38] unfolding. Their path-ways predicted for the unfolding of two Ig-like domains experiments on intact spectrin networks of erythrocytes

tentatively suggested unfolding of single repeats [41],agree with MD simulations done on those systems. Forspectrin, however, their predictions contradict our ex- and forces measured in our AFM experiments are in

the physiological range observed on erythrocyte ghostsperiments and simulations. As a model is often mostinteresting when it fails, this one illustrates the key find- [44]. Our combined data from single-molecule experi-

ments and MD simulations provide an intriguing glimpseings of our study: in the case of spectrin, helix B cannotbe considered as a single structural block, and forced of a molecule that might be optimized for elasticity on

all scales of its architecture: from the structure of its cell-intermediates can, in fact, arise from nonnative topol-ogies. spanning network down to predetermined hyphenation

points in its single domains. The mechanical propertiesof this protein are the link between structure and

Implications for Folding? function.Can we expect parallels between forced unfolding andfolding pathways? We had demonstrated multiple re- Experimental Proceduresfolding of spectrin repeats in previous AFM experiments

Engineering of Mutants[30], and one could speculate that refolding domainsSpectrin repeat 16 from Gallus gallus nonerythroid � spectrin (acces-might face N- to C-terminal extensions in the cellular con-sion number P07751; Wasenius et al., 1988) was used for the studies.text and follow routes reversing their forced-unfoldingConstructs containing four identical repeats were generated as de-

pathways. Completely unrestrained folding is likely to scribed earlier [30]. All the constructs contained the N-terminal se-proceed via rather different ways, especially considering quence MKHHHHHHPMSDYDIPTTENLYFQGAMEMSAA followed

by residues 1763–1877 of the chicken � spectrin. In this paper wethe variability we already suggested for forced unfoldingare using amino acid residues numbering 1–116 for the chickenalong only one dimension. Nevertheless, locally stabi-� spectrin residues 1762–1877, respectively. The linker sequencelized structures in the later stages of our simulationsIMEMSAA was added between the repeats of the construct, andcould also be involved in early folding steps. Classicsequence IMTCC was added to the C terminus. Compared to the

ensemble methods, like thermal unfolding, or earlier ex- structure of the R16-R17 [5], our constructs have 17 extra residuesperiments [39] did not detect unfolding intermediates. as a linker between the domains. This was engineered to avoid

interaction between the adjacent domains during the AFM experi-Likewise, direct averaging of force curves from our ex-ments. The mutagenesis was done with the QuikChange mutagene-periments or simulations would clearly diminish any in-sis kit (Stratagene) to the single domains. After mutagenesis theformation about forced intermediates. Single-moleculeconstructs were verified by DNA sequencing. Polymerization, pro-techniques thus reveal details that are not directly ac-tein expression, and purification were done as described earlier [30].

cessible to the more-classical ensemble methods, and The double-cysteine mutant (CC mutant) was engineered suchexperimental advances in this direction promise further that the amino acids methylene and valine at positions 17 and 67 on

helices A and B, respectively, were changed into cysteine residues.insights into the unfolding and folding of proteins.

Unfolding Intermediates of Spectrin Repeats1095

According to the wild-type spectrin repeat’s structure, this would refined structure without changes to the backbone geometry. In thecase of the CC mutant, WhatIf [48] was used to suggest rotamersallow the formation of a disulfide bond between helices A and B.

The double-alanine mutant (AA mutant) was engineered such that for the introduced Cys side chains. The altered structures weresubjected to the same protocol as was the wild-type, but equilibra-the amino acids proline and glycine at positions 62 and 66 on the

central B helix were both changed into alanine residues. tion was extended to 500 ps before the generation of the first startingconfiguration for unfolding.

Trajectories were inspected with VMD [49]. Angles and distancesCircular Dichroism Measurementswere extracted with ptraj and visualized in MatLab (The MathWorks).Measurements of circular dichroism were performed in a Jasco-710Structure figures were prepared with Molscript [50] using secondaryspectropolarimeter at room temperature. The path length of the cellstructure assignments from Rasmol [51].was 2 mm. Mean residue ellipticity at 222 nm was monitored during

thermal unfolding at a scan rate of 90C/hr and converted into afraction of unfolded values to facilitate comparison of the different Acknowledgmentsdenaturation curves as proposed in [43]. The protein concentrationof the tetrameric constructs was 20 �g/ml for the wild-type and 50 Jens P. Linge and Roger Abseher provided initial molecular dynam-

ics calculations. We thank Othmar Marti for helpful discussions and�g/ml for the mutants, each in 140 mM NaCl and 10 mM sodiumphosphate (pH 7.3). Johan Leckner for critical inspection of the manuscript. Part of the

project was funded by the Deutsche Forschungsgemeinschaft (DFG;grant Ni-499-5/1). S.M.A. was supported by Cusanuswerk. R.G. wasAFM Experimentssupported by the Boehringer Ingelheim Fonds. J.Y. was partly sup-We probed the effect of the two mutations on the mechanical prop-ported by the Academy of Finland. K.H. was supported by a Fulbrighterties of the spectrin repeat by forced unfolding of single moleculesFellowship.using a novel sensor-stabilization system in our apparatus. A de-

Matti Saraste, who had devoted a large part of his scientific worktailed description of our AFM can be found in [45]. The instrumentto the structure and function of spectrin, made this collaborationfeatures local stabilization of the distance between the equilibriumpossible and was always an essential driving force behind it. Heposition of the tip and the surface and allows for direct measure-died unexpectedly in May 2001.ments of elongation distances by local probe-based linearization

of the piezo movement. This provided the stability and resolutionimprovements in our custom-built instrument to work in the low- Received: January 24, 2002

Revised: May 21, 2002force-unfolding regime using an otherwise standard AFM setup. Theexact spring constants, k, of the C (k � 10 pN/nm) and D (k � Accepted: May 24, 200230 pN/nm) cantilevers of Park Scientific, MLCT-AUHW levers weredetermined by the equipartition theorem. The scattered values had Referencesthe same order of magnitude as the values given by the manufac-turer. We used a constant pulling speed of 0.3 nm/ms. The proteins 1. Delaunay, J. (1995). Genetic disorders of the red cell membrane.were stored in buffer, 50 mM Tris-HCl (pH 8.0) and 50 mM NaCl. A FEBS Lett. 369, 34–37.40 �l drop of protein solution (50 �g/ml) was put on freshly sputtered 2. Toniolo, D., and Minetti, C. (1999). Muscular dystrophies: alter-gold surfaces to which the cysteine residues at the C terminus of ations in a limited number of cellular pathways. Curr. Opin.our constructs can bind covalently. The proteins were allowed to Genet. Dev. 9, 275–282.adsorb for 5–15 min. The sample was then washed with buffer and 3. Discher, D.E., and Carl, P. (2001). New insights into red cellbrought into the fluid chamber of the AFM. Success in the experi- network structure and spectrin unfolding—a current review.ments was positively correlated with lower humidity levels in the Cell. Mol. Biol. Lett. 6, 593–606.lab. Data acquisition was done using a PCI-M-I/O 16E-4 A/D board 4. Pascual, J., Pfuhl, M., Walther, D., Saraste, M., and Nilges, M.(National Instruments, Texas). Force curves were analyzed with our (1997). Solution structure of the spectrin repeat: a left-handedown custom programs as well as built-in functions of Igor (Wave- antiparallel triple-helical coiled-coil. J. Mol. Biol. 273, 740–751.metrics, Oregon). Forced-unfolding extensions were measured be- 5. Grum, V.L., Li, D., MacDonald, R.I., and Mondragon, A. (1999).tween two consecutive and distinctive force peaks only. Adhesion Structures of two repeats of spectrin suggests models for flexi-peaks, sometimes present in the beginning of force curves, were bility. Cell 98, 523–535.not used as reference points. The last peak of a curve was only 6. Djinovic-Carugo, K., Young, P., Gautel, M., and Saraste, M.used if its force was less than 100 pN. (1999). Structure of the alpha-actinin rod: molecular basis for

cross-linking of actin-filaments. Cell 98, 537–546.7. Elgsaeter, A., Stokke, B.T., Mikkelsen, A., and Branton, D. (1986).MD Simulations

Simulations were performed with the Amber 6.0 program package The molecular basis of erythrocyte shape. Science 234, 1217–1223.using the modified Cornell et al. all-atom force field parm96 [46, 47].

Bond lengths involving hydrogen atoms were fixed with the SHAKE 8. Hansen, J.C., Skalak, R., Chien, S., and Hoger, A. (1996). Anelastic network model based on the structure of the red bloodalgorithm, allowing for an integration time step of 2 fs. A cutoff

of 15 A was applied to nonbonded interactions. Temperature was cell membrane skeleton. Biophys. J. 70, 146–166.9. Raphael, R.M., Popel, A.S., and Brownell, W.E. (2000). A mem-controlled with the Berendsen coupling algorithm using a “coupling

constant” of 5 ps. Effects of solvation were emulated with the gener- brane bending model of outer hair cell electromotility. Biophys.J. 78, 2844–2862.alized Born model and a tension term proportional to the molecule’s

surface area, both implemented in Amber 6. 10. Binnig, G., Quate, C.F., and Gerber, C. (1986). Atomic forcemicroscope. Phys. Rev. Lett. 56, 930–933.The simulations were based on the solution structure of the 16th

repeat of chicken brain � spectrin [4], which had been subjected to 11. Ashkin, A. (1987). Optical trapping and manipulation of virusesand bacteria. Science 235, 1517–1520.an additional refinement in explicit water. The molecule was mini-

mized and heated to 300 K, with a linear temperature increase over 12. Smith, S.B., Finzi, L., and Bustamente, C. (1992). Direct mechan-ical measurements of the elasticity of single DNA molecules by30 ps, while atomic velocities were reassigned from a Maxwell distri-

bution every 2.5 ps. An equilibration for 250 ps was followed by the using magnetic beads. Science 258, 1122–1126.13. Rief, M., Gautel, M., Oesterhelt, F., Fernandez, F., and Gaub,1 ns production MD, which yielded 11 restart files spaced 100 ps

apart. Unfolding was initiated from these restart files by imposing H.E. (1997). Reversible unfolding of individual titin immuno-globin domains by AFM. Science 276, 1109–1112.a harmonic distance restraint (force constant 5.9616 kcal �

mol�1 A�2) on the system, which forced the N and C termini apart 14. Kellermayer, M.S.K., Smith, S.B., Granzier, H.L., and Busta-mente, C. (1997). Folding-unfolding transitions in single titinat a constant velocity of 0.2 A � ps�1. This approach was chosen

for its ease of implementation and because the molecule can be molecules characterized with laser tweezers. Science 276,1112–1116.assumed to align to an external force vector prior to any recorded

unfolding events [25]. Mutations were introduced into the water- 15. Tskhovrebova, L., Trinick, J., Sleep, J.A., and Simmons, R.M.

Structure1096

(1997). Elasticity and unfolding of single molecules of the giant mines force-induced unfolding pathways in globular proteins.Proc. Natl. Acad. Sci. USA 97, 7254–7259.muscle protein titin. Nature 387, 308–312.

16. Rief, M., Pascual, J., Saraste, M., and Gaub, H.E. (1999). Single 38. Gsponer, J., and Caflisch, A. (2001). Role of native topologyinvestigated by multiple unfolding simulations of four SH3 do-molecule force spectroscopy of spectrin repeats: low unfolding

forces in helix bundles. J. Mol. Biol. 286, 553–561. mains. J. Mol. Biol. 309, 285–298.39. DeSilva, T.M., Harper, S.L., Kotula, L., Hensley, P., Curtis, P.J.,17. Carrion-Vazquez, M., Marszalek, P.E., Oberhauser, A.F., and

Fernandez, J.M. (1999). Atomic force microscopy captures Otvos, L., Jr., and Speicher, D.W. (1997). Physical properties ofa single-motif erythrocyte spectrin peptide. Biochemistry 36,length phenotypes in single proteins. Proc. Natl. Acad. Sci. USA

96, 11288–11292. 3991–3997.40. Byers, T.J., and Branton, D. (1985). Visualization of the protein18. Li, H., Oberhauser, A.F., Fowler, S., Clarke, J., and Fernandez,

J.M. (2000). Atomic force microscopy reveals the mechanical associations in the erythrocyte membrane skeleton. Proc. Natl.Acad. Sci. USA 82, 6153–6157.design of a modular protein. Proc. Natl. Acad. Sci. USA 97,

6527–6531. 41. Lee, J.C., and Discher, D.E. (2001). Deformation-enhanced fluc-tuations in the red cell skeleton with theoretical relations to19. Oberhauser, A.F., Hansma, P.K., Carrion-Vazquez, M., and Fer-

nandez, J.M. (2001). Stepwise unfolding of titin under force- elasticity, connectivity, and spectrin unfolding. Biophys. J. 81,clamp atomic force microscopy. Proc. Natl. Acad. Sci. USA 98, 3178–3192.468–472. 42. McGough, A. (1999). How to build a molecular shock absorber.

20. Carrion-Vazquez, M., Oberhauser, A.F., Fowler, S.B., Marszalek, Curr. Biol. 9, R887–R889.P.E., Broedel, S.E., Clarke, J., and Fernandez, J.M. (1999). Me- 43. MacDonald, R.I., and Pozharski, E.V. (2001). Free energies ofchanical and chemical unfolding of a single protein: a compari- urea and of thermal unfolding show that two tandem repeatsson. Proc. Natl. Acad. Sci. USA 96, 3694–3699. of spectrin are thermodynamically more stable than a single

21. Oberhauser, A.F., Marszalek, P.E., Erickson, H.P., and Fernan- repeat. Biochemistry 40, 3974–3984.dez, J.M. (1998). The molecular elasticity of the extracellular 44. Sleep, J., Wilson, D., Simmons, R., and Gratzer, W. (1999). Elas-matrix protein tenascin. Nature 393, 181–185. ticity of the red cell membrane and its relation to hemolytic

22. Krammer, A., Lu, H., Isralewitz, B., Schulten, K., and Vogel, V. disorders: an optical tweezers study. Biophys. J. 77, 3085–3095.(1999). Forced unfolding of the fibronectin type III module re- 45. Altmann, S.M., Lenne, P.-F., and Horber, J.K.H. (2001). Multipleveals a tensile molecular recognition switch. Proc. Natl. Acad. sensor stabilization system for local probe microscopes. Rev.Sci. USA 96, 1351–1356. Sci. Instrum. 72, 142–149.

23. Lu, H., Isralewitz, B., Krammer, A., Vogel, V., and Schulten, K. 46. Cornell, W.D., Cieplak, P., Bayly, C.I., Gould, I.R., Merz, K.M.,(1998). Unfolding of titin immunoglobulin domains by steered Ferguson, D.M., Spellmeyer, D.C., Fox, T., Caldwell, J.W., andmolecular dynamics simulation. Biophys. J. 75, 662–671. Kollman, P.A. (1995). A second generation force field for the

24. Lu, H., and Schulten, K. (1999). Steered molecular dynamics simulation of proteins, nucleic acids, and organic molecules. J.simulations of force induced protein domain unfolding. Proteins Am. Chem. Soc. 117, 5179–5197.35, 453–463. 47. Kollman, P., Dixon, R., Cornell, W., Fox, T., Chipot, C., and

25. Lu, H., and Schulten, K. (2000). The key event in force-induced Pohorille, A. (1991). The development/application of a ‘mini-unfolding of titin’s immunoglobulin domains. Biophys. J. 79, malist’ organic/biochemical molecular mechanics force field us-51–65. ing a combination of ab initio calculations and experimental

26. Marszalek, P.E., Lu, H., Li, H., Carrion-Vazquez, M., Oberhauser, data. In Computer Simulation of Biomolecular Systems, VolumeA.F., Schulten, K., and Fernandez, J.M. (1999). Mechanical un- 3, W.F. Van Gunsteren and P.K. Weiner, eds. (Amsterdam: Else-folding intermediates in titin modules. Nature 402, 100–103. vier), pp. 83–96.

27. Best, R.B., Li, B., Steward, A., Daggett, V., and Clarke, J. (2001). 48. Vriend, G. (1990). WHAT IF: a molecular modeling and drugCan non-mechanical proteins withstand force? Stretching bar- design program. J. Mol. Graph. 8, 52–66, 29.nase by atomic force microscopy and molecular dynamics sim- 49. Humphrey, W., Dalke, A., and Schulten, K. (1996). VMD: visualulation. Biophys. J. 81, 2344–2356. molecular dynamics. J. Mol. Graph. 14, 33–38.

28. Paci, E., and Karplus, M. (1999). Forced unfolding of fibronectin 50. Kraulis, P. (1991). MOLSCRIPT: a program to produce bothtype 3 modules: an analysis by biased molecular dynamics sim- detailed and schematic plots of proteins. J. Appl. Crystallogr.ulations. J. Mol. Biol. 288, 441–459. 24, 946–950.

29. Paci, E., and Karplus, M. (2000). Unfolding proteins by external 51. Sayle, R.A., and Milner-White, E.J. (1995). RASMOL: biomolecu-forces and temperature: the importance of topology and ener- lar graphics for all. Trends Biochem. Sci. 20, 374.getics. Proc. Natl. Acad. Sci. USA 97, 6521–6526.

30. Lenne, P.F., Raae, A.J., Altmann, S.M., Saraste, M., and Horber,J.K.H. (2000). States and transitions during forced unfolding ofa single spectrin repeat. FEBS Lett. 476, 124–128.

31. Carrion-Vazquez, M., Oberhauser, A.F., Fisher, T.E., Marszalek,P.E., Li, H., and Fernandez, J.M. (2000). Mechanical design ofproteins studied by single-molecule force spectroscopy andprotein engineering. Prog. Biophys. Mol. Biol. 74, 63–91.

32. Craig, D., Krammer, A., Schulten, K., and Vogel, V. (2001). Com-parison of the early stages of forced unfolding for fibronectintype III modules. Proc. Natl. Acad. Sci. USA 98, 5590–5595.

33. Izrailev, S., Stepaniants, S., Balsera, M., Oono, Y., and Schulten,K. (1997). Molecular dynamics study of unbinding of the avidin-biotin complex. Biophys. J. 72, 1568–1581.

34. Balsera, M., Stepaniants, S., Izrailev, S., Oono, Y., and Schulten,K. (1997). Reconstructing potential energy functions from simu-lated force-induced unbinding processes. Biophys. J. 73, 1281–1287.

35. Cornell, W., Abseher, R., Nilges, M., and Case, D.A. (2001). Con-tinuum solvent molecular dynamics study of flexibility in in-terleukin-8. J. Mol. Graph. Model. 19, 136–145.

36. Rief, M., Fernandez, J.M., and Gaub, H.E. (1998). Elasticallycoupled two-level systems as a model for biopolymer extensibil-ity. Phys. Rev. Lett. 81, 4764–4767.

37. Klimov, D.K., and Thirumalai, D. (2000). Native topology deter-