Retraction

CHEMISTRYRetraction for “Probing osmolyte participation in the unfoldingtransition state of a protein,” by Lorna Dougan, Georgi Z.Genchev, Hui Lu, and Julio M. Fernandez, which appeared inissue 24, June 14, 2011, of Proc Natl Acad Sci USA (108:9759–9764; first published May 25, 2011; 10.1073/pnas.1101934108).The authors note the following: “We wish to retract our article

because interpretation of the experimental data was based onincorrect calibration of the apparatus in viscous solutions andthere is no basis for the major conclusions of the study on thestructure of the transition state for mechanical unfolding.”

www.pnas.org/cgi/doi/10.1073/pnas.1118432108

20850 | PNAS | December 20, 2011 | vol. 108 | no. 51 www.pnas.org/cgi/doi/10.1073/pnas.1118432108

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1 D

ownl

oade

d by

gue

st o

n F

ebru

ary

15, 2

021

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1 D

ownl

oade

d by

gue

st o

n F

ebru

ary

15, 2

021

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1 D

ownl

oade

d by

gue

st o

n F

ebru

ary

15, 2

021

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 15

, 202

1 D

ownl

oade

d by

gue

st o

n F

ebru

ary

15, 2

021

Probing osmolyte participation in theunfolding transition state of a proteinLorna Dougana,1, Georgi Z. Genchevb, Hui Lub,c,1, and Julio M. Fernandezd

aSchool of Physics and Astronomy, University of Leeds, Leeds LS2 9JT, United Kingdom; bDepartment of Bioengineering, University of Illinois, Chicago, IL60607; cShanghai Institute of Medical Genetics, Children’s Hospital of Shanghai, Shanghai 200040, China; and dDepartment of Biological Sciences,Columbia University, New York, NY 10027

Edited by George N. Somero, Stanford University, Pacific Grove, CA, and approved April 26, 2011 (received for review February 7, 2011)

Understanding the molecular mechanisms of osmolyte protectionin protein stability has proved to be challenging. In particular, littleis known about the role of osmolytes in the structure of the unfold-ing transition state of a protein, the main determinant of itsdynamics. We have developed an experimental protocol to directlyprobe the transition state of a protein in a range of osmolyteenvironments. We use an atomic force microscope in force-clampmode to apply mechanical forces to the protein I27 and obtainforce-dependent rate constants of protein unfolding. We measurethe distance to the unfolding transition state, Δxu, along a 1Dreaction coordinate imposed by mechanical force. We find thatfor the small osmolytes, ethylene glycol, propylene glycol, andglycerol, Δxu scales with the size of the molecule, whereas for lar-ger osmolytes, sorbitol and sucrose, Δxu remains the same as thatmeasured in water. These results are in agreement with steeredmolecular dynamics simulations that show that small osmolytesact as solvent bridges in the unfolding transition state structure,whereas only water molecules act as solvent bridges in large osmo-lyte environments. These results demonstrate that novel forceprotocols combined with solvent substitution can directly probeangstrom changes in unfolding transition state structure. Thisapproach creates new opportunities to gain molecular level under-standing of the action of osmolytes in biomolecular processes.

protein folding ∣ single molecule ∣ mechanical protein

Solvent composition is actively modulated in vivo providing adiverse and optimized environment for biological processes

(1). Solvent molecules facilitate necessary structural and dynamicarrangements, permit rapid conformational changes, catalyzechemical reactions, and mediate the self-assembly of biologicalmolecules (2–4). There has been much effort to understand therole of the solvent environment on the behavior of proteins (5–9).In particular, studies have focused on understanding the functionof osmolytes, small organic compounds that affect protein stabi-lity and are ubiquitous in living systems (5–8, 10, 11). Despitemuch effort, studies continue to attempt to find a universal mo-lecular theory that can explain the mechanism by which osmolytesinteract with proteins to affect protein stability (11). Althoughexperiments have revealed a wealth of information regardingthe thermodynamics of protein folding, very little is knownabout the role that solvent molecules play on the structure ofthe folding/unfolding transition state of a protein, which is themain determinant of protein dynamics (12).

Single-molecule force spectroscopy has emerged as a powerfultool to probe transition states in a protein (13). This techniqueis used to apply a mechanical force to a single protein, causingthe protein to unfold and extend along a well-defined reactioncoordinate: the end-to-end length of the protein (14). Along thisunfolding pathway, a mechanically resistant transition state deter-mines the force-dependent rate of unfolding, kuðFÞ, easily mea-sured with force spectroscopy techniques. The force dependencyof the unfolding rate is typically fit with an Arrhenius term thatmeasures properties of the unfolding transition state. In itssimplest representation, the unfolding transition state is charac-

terized by two parameters: the size of its activation energy barrier,ΔGu, and the elongation of the protein necessary to reach thetransition state, Δxu. Of particular interest are the force spectro-scopy measurements of Δxu that provide a direct measure of thelength scales of a transition state, which were hitherto unknown.

We have previously demonstrated that force spectroscopy canbe used to measure the distance to the unfolding transition stateof a protein Δxu (13, 15). Completing single-molecule proteinunfolding experiments on a number of different proteins inwater, we measured values of Δxu in the range of 1.7–2.5 Å(13, 15, 16). These values of Δxu are comparable to the size ofa water molecule (17), suggesting that water molecules are inte-gral components of the unfolding transition state. Steered mole-cular dynamics (SMD) simulations have complemented ourobservations by providing a detailed atomic picture of stretchingand unfolding of individual protein domains (18). The simula-tions are carried out by fixing one terminus of the protein andapplying external forces to the other terminus. Earlier SMDsimulations of forced unfolding of the 27th immunoglobulin-like domain of cardiac titin (I27) protein suggested that whena stretching force is applied between the protein’s termini, resis-tance to unfolding originated from a set of hydrogen bondsbetween two parallel β-strands (A′ andG) of the protein structure(19, 20). These β-strands provide a “mechanical clamp” that mustbe broken before unfolding can occur. Because the hydrogenbonds in the mechanical clamp region are perpendicular to theaxis of extension, they must rupture simultaneously to allowrelative movement of the two termini. Earlier SMD simulationsshowed that the breakage of interstrand hydrogen bonds could befollowed by bonding to water molecules that then formed bridgesbetween the two separating strands (19, 20). Given that forcespectroscopy experiments measure a Δxu comparable to the sizeof a water molecule, one way to interpret the experimental resultsis that the transition state structure is formed by water moleculesbridging the gap between separating β-strands and taking theplace of some of the broken interstrand hydrogen bonds (19).Indeed, recent SMD simulations on the protein ubiquitin havedemonstrated that this protein also contains a mechanical clampregion, and water molecules play an integral role in the protein’sunfolding transition state structure. Furthermore, these studiesshowed that hydrophobic interactions in the surface residuesof the mechanical clamp region regulated the insertion of watermolecules prior to hydrogen bond breakage and subsequentunfolding of the protein (21). These earlier studies point to theimportance of water in the force-induced unfolding of a protein

Author contributions: L.D., H.L., and J.M.F. designed research; L.D. and G.Z.G. performedresearch; L.D., G.Z.G., and H.L. contributed new reagents/analytic tools; L.D., G.Z.G., andH.L. analyzed data; and L.D. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence may be addressed. E-mail: L.Dougan@leeds.ac.uk orhuilu.bioinfo@gmail.com.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1101934108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1101934108 PNAS ∣ June 14, 2011 ∣ vol. 108 ∣ no. 24 ∣ 9759–9764

CHEM

ISTR

Y

See Retraction Published , 2011 December 12

and suggests a “solvent bridging” mechanism in the unfoldingtransition state structure of the protein.

We have recently used force-clamp spectroscopy to test thissolvent bridging mechanism by measuring Δxu of the I27 proteinin the presence of the protecting osmolyte, glycerol, and deuter-ium oxide (13, 15). Glycerol is a good hydrogen bonding moleculethat is ubiquitous in living systems, known to enhance protein sta-bility, and is larger in molecular size than water (22). Deuteriumoxide forms stronger hydrogen bonds than water while having asimilar size (23). Our experiments showed that upon replacementof water by increasing amounts of the larger glycerol molecules,Δxu increased rapidly and reached a plateau at its maximum valueof 4.4 Å (13). In contrast, replacement of water by the similarlysized deuterium oxide did not change the value ofΔxu (15). Theseexperiments, combined with SMD simulations, directly demon-strated that solvent molecules form part of the structure of themechanical transition state of the I27 protein, acting as “solventbridges.”

To establish whether solvent bridging is a unique feature of theosmolyte glycerol and to further understand the unfolding transi-tion state structure of the I27 protein, we have greatly expandedour studies. We have completed an extensive series of single-molecule experiments of protein unfolding on the I27 protein ina range of osmolyte solutions: ethylene glycol, propylene glycol,sorbitol, and sucrose. These osmolytes (Fig. 1) are capable offorming hydrogen bonds and vary in size from 4.8 Å (ethyleneglycol) to 8.8 Å (sucrose), making them excellent candidates totest our solvent bridging hypothesis. These osmolytes also providea toolbox of molecules of varying size to probe the unfoldingtransition state structure of the I27 protein.

ResultsUsing the experimental protocol of our previous studies (24),we constructed and expressed a polyprotein that consisted ofeight identical repeats of the I27 protein, ðI27Þ8, the mechanicalproperties of which have been extensively characterized experi-mentally (24, 25) and also in silico using molecular dynamics tech-niques (18, 19). The use of polyproteins is advantageous in thatthey provide a clear mechanical fingerprint of our system ofinterest to distinguish them against a background of spuriousinteractions. They also provide us with a larger number of eventsper recording than otherwise possible with monomers (26). Wemeasure the properties of the mechanical unfolding transitionstate of the I27 protein by measuring the force dependency ofthe unfolding rate of single I278 polyproteins. When a proteinis subjected to an external force, its unfolding rate, ku, is well de-scribed by an Arrhenius term of the form kuðFÞ ¼ ku0 expðFΔxu∕kBTÞ, where ku0 is the unfolding rate in the absence of externalforces, F is the applied force, and Δxu is the distance from thenative state to the transition state along the pulling direction(27). By measuring how the unfolding rate changes with anapplied force, we can readily obtain estimates for the values ofboth, ku0 and Δxu (13, 27). Given that ku0 ¼ A expðΔGu∕kBTÞ,we can estimate the size of the activation energy barrier ofunfolding ΔGu. From transition state theory, we assume a valueof 1013 s−1 for the prefactor, as has previously been used formechanical unfolding of the I27 protein (24). The distance totransition state, Δxu, determines the sensitivity of the unfoldingrate to the pulling force and measures the elongation of the pro-tein at the transition state of unfolding. Given that k0u and Δxureflect properties of the unfolding transition state, we expectedthat these variables would be strongly dependent on the solventenvironment (13).

Under force-clamp conditions, stretching a polyprotein resultsin a well-defined series of step increases in length, marking theunfolding and extension of the individual modules in the chain(13). The size of the observed steps corresponds to the numberof amino acids released by each unfolding event (28). Stretchinga single I278 polyprotein in aqueous solution at a constant forceof 180 pN results in a series of step increases in length of24 nm (Fig. 1A). The time course of these events is a directmeasure of the unfolding rate at 180 pN. To probe the role ofthe solvent environment in setting the structure of the unfoldingtransition state, we studied the effect of solvent substitution onthe force dependency of the unfolding rate. We completed forceunfolding experiments on the I27 protein in a range of osmolytesolutions: ethylene glycol, propylene glycol, sorbitol, and sucrose(Fig. 1 B–F). These simple molecules are all capable of forminghydrogen bonds, and their thermodynamic properties have beenextensively characterized (10, 11, 29, 30). Importantly, theseosmolytes vary in size, making them excellent candidates to testour solvent bridging hypothesis and to probe the unfolding tran-sition state structure of the I27 protein (11).

For each osmolyte solution we measure the unfolding rateof the I27 protein by fitting a single exponential to an averageof 30 traces similar to the ones shown in Fig. 1 (27). Previous workon the protein ubiquitin has shown that a single exponential fitcaptures 81% of the unfolding events of the protein and thusrepresents a reasonable measure of the unfolding rate (31, 32).The summation and normalization of unfolding traces are shownin Fig. 2 A–D for I27 protein unfolding in a solution of ethyleneglycol, propylene glycol, sorbitol, and sucrose at a range offorces from 140–200 pN. We define the unfolding rate as kuðFÞ ¼1∕ζðFÞ, where ζðFÞ is the time constant of the exponential fits tothe averaged unfolding traces, shown in Fig. 2 A–D (27). Further-more, we obtain an estimate of the standard error of kuðFÞ, usingthe bootstrapping technique (33, 34) (see SI Text). We repeatedthese measurements and obtained the force dependency of the

Fig. 1. A molecular toolbox for determining the role of osmolyte moleculesin the unfolding transition state of a protein. Using an array of osmolytes ofvarying size and hydrogen bonding abilities, we test the importance of sol-vent bridging in the I27 protein. Force-clamp protein unfolding traces for theðI27Þ8 polyprotein at a constant force of 180 pN in (A) water and a range ofosmolyte solutions at a concentration of 1M (B) ethylene glycol, (C) propyleneglycol, (D) glycerol, (E) sorbitol, and (F) sucrose. Each unfolding trace showsthe characteristic staircase of unfolding events, with each step of 24 nmcorresponding to the unfolding of one I27 module of the polyprotein.

9760 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1101934108 Dougan et al.

See Retraction Published , 2011 December 12

unfolding rate for each osmolyte solution: ethylene glycol, propy-lene glycol, sorbitol, and sucrose at a range of concentrations.

In Fig. 3 we show the natural logarithm of ku as a functionof applied force for each osmolyte for a range of concentrationsup to the solubility limit of the osmolyte. We fit the Arrheniusrate equation to the data (Fig. 3) to obtain ku0 and Δxu. It isapparent from Fig. 3 A–D that the introduction of increasingamounts of osmolyte decreases the value of ku0 compared withthat of water (black data points and line in Fig. 3). In additionto the observed changes in ku0, there are striking differences

in the slope (Δxu∕kBT) of the force dependency of unfoldingin Fig. 3. In the case of ethylene glycol (Fig. 3A) and pro-pylene glycol (Fig. 3B) the slope increased, resulting in anincrease in the distance to transition state from Δxu ¼ 2.5 Å inaqueous solution and up to 3.65� 0.1 Å in 62% by weight ethy-lene glycol and 4.1� 0.1 Å in 60% by weight propylene glycol,respectively. Interestingly, for the larger osmolytes, sorbitoland sucrose, the slope of the force dependency of the unfoldingrate showed no change compared with that measured in water(Fig. 3 C and D).

In Fig. 4 we show the full picture of the effect of osmolytes onΔGu (Fig. 4A) and Δxu; (Fig. 4B), where ΔGu is calculated usingku0 ¼ A expðΔGu∕kBTÞ. For all the osmolytes studied, ΔGu isfound to increase with increasing weight fraction of osmolytecompared with that measured in water (Fig. 4A). Thus, theosmolytes used in this study stabilize the protein I27 against thedenaturation induced by mechanical force. In the case of thesmall osmolytes, ethylene glycol and propylene glycol, ΔGuincreases significantly from 23.11 kcal mol−1 in aqueous solutionto 27.08 kcal mol−1 in aqueous ethylene glycol and 28.80 kcalmol−1 in aqueous propylene glycol. For the larger osmolytes ΔGuis also observed to increase, but to a lesser extent with a ΔGu of24.67 kcal mol−1 in aqueous sorbitol and 24.21 kcal mol−1 inaqueous sucrose. Thus, small osmolytes seem to be more effectivein stabilizing the I27 protein against mechanical denaturation. InFig. 4B we show the measured value of Δxu for each osmolytesolution for a range of different concentrations, in weight fractionof osmolyte. Strikingly, for the small osmolytes, ethylene glycoland propylene glycol, we observe that the value of Δxu increasesin a sharply nonlinear manner. For ethylene glycol (squares), Δxuincreases rapidly and then saturates at a value of 3.65 Å, whereasfor propylene glycol Δxu increases rapidly and saturates at a valueof 4.0 Å. This is in good agreement with our earlier experimentalobservations with aqueous glycerol where we measured a sharplynonlinear increase in Δxu (13). Contrary to the behavior of thesmall osmolytes, our experiments of aqueous solutions of sorbitoland sucrose show a very different trend in the measured value ofΔxu. As before, we measured the value of Δxu in sorbitol and su-crose solutions by fitting an Arrhenius term to the force depen-dency of the unfolding rate in Fig. 3 C and D. Our measurementsshowed that Δxu remained unchanged across the entire concen-tration range measured (Fig. 4B). Indeed, the measured value ofΔxu is 2.5 Å for both sorbitol (triangle) and sucrose (circles) at all

Fig. 2. The average time course of unfolding was obtained by summation and normalization of unfolding traces. Multiple trace averages (n > 30 in each trace)of unfolding events measured for I27 in (A) 44% by weight ethylene solution at constant forces of 220, 200, 180, and 160 pN; (B) 30% by weight propyleneglycol solution at constant forces of 200, 180, 160, and 140 pN; (C) 35% by weight sorbitol solution at constant forces of 200, 180, 160, and 140 pN; (D) 34% byweight sucrose solution at constant forces of 220, 200, 180, and 160 pN.

Fig. 3. Semilogarithmic plot of the rate of unfolding of I27 as a function ofpulling force in (A) 6% (squares), 32% (upward triangles), 44% (downwardtriangles), and 66% (diamonds) by weight ethylene glycol solution; (B) 10%(squares), 20% (upward triangles), 36% (downward triangles), and 60% (dia-monds) by weight propylene glycol solution; (C) 15% (squares), 35% (upwardtriangles), 52% (downward triangles) by weight sorbitol solution; (D) 14%(squares), 34% (upward triangles), and 54% (downward triangles) by weightsucrose solution. For comparison, the force dependency of unfolding I27 inwater is shown on each graph (circle).

Dougan et al. PNAS ∣ June 14, 2011 ∣ vol. 108 ∣ no. 24 ∣ 9761

CHEM

ISTR

Y

See Retraction Published

November 10, 2010

, 2011December 12

concentrations. This is the same value ofΔxu that is measured in awater solution (13).

In Fig. 5A (squares) the maximum ΔxU measured for eachosmolyte is shown along with the molecular size of each osmolyte.For the smaller osmolytes, ethylene glycol, propylene glycol andfrom our earlier studies, glycerol (13), ΔxU increases from thatmeasured in water. Furthermore there is a correlation betweenthe experimentally measured value of ΔxU and osmolyte molecu-lar size. For the larger osmolytes, sorbitol and sucrose, no suchcorrelation between ΔxU and molecular size is observed. We

complemented our experimental measurements by exploringthe mechanisms of solvent molecule participation in the unfold-ing transition state using SMD simulations. These SMD simula-tions provide a detailed atomic picture of stretching andunfolding of individual I27 protein domains (18, 19). Our simula-tions of forced unfolding of the I27 protein in ethylene glycol andsorbitol solutions showed that the resistance to unfolding origi-nates from the hydrogen bonds between the A′ and G β-strands(Fig. 5B). During unfolding, the solvent molecules attempt tobreak and bridge the backbone hydrogen bonds between theβ-strands. In the simulations of ethylene glycol solutions, thelarger size of the cosolvent and the solvent bridge it forms leadsto a greater gap separating the β-strands (Fig. 5C). The simula-tions showed that the separation between the two strands in-creased due to the insertion of the ethylene glycol solventbridge. Our earlier SMD simulations of forced unfolding ofthe I27 protein in glycerol solutions showed that the larger sizeof this cosolvent and its ability to act as a solvent bridge could leadto a greater distance between the neighboring A′—G β-strands(13). In the case of sorbitol solutions, there was no evidencein the SMD simulations of a sorbitol molecule acting in the sameway as ethylene glycol. There were cases when the end hydroxylgroup could hydrogen bond with the A′ andG β-strands (Fig. 5D).However, we found no evidence for a full sorbitol molecule sol-vent bridging between the A′ and G β-strands of the I27 protein.Instead, the simulations predominately identified water mole-cules bridging the A′ and G β-strands. Thus for the sorbitolSMD simulations we did not observe an increased separationbetween the β-strands.

Given that there is a multitude of possible transition statestructures formed by water, osmolyte, and the protein backbone,there is no straightforward way to link a wider gap between theβ-strands A′ and G in the simulations and the experimentallymeasured values of ΔxU . In Fig. 5 C and D, we define the pullingcoordinate for the separating β-strands as the distance betweenthe first amino acid of strand A′ (shown as a large blue sphere)and the last amino acid of strand G (also shown as a large bluesphere). This distance gets longer as the two β-strands separateunder a constant force, filling the gap with solvent molecules untila transition state is reached. In the SMD simulations the elonga-tion of this separation of the β-strands up to the transition state isdefined as the distance to the transition stateΔxA0-G. The crossingof the transition state is marked by an abrupt increase in the

Fig. 5. The maximum Δxu measured using the experimental data for propylene glycol, ethylene glycol, sorbitol, and sucrose (squares) and ΔxA0-G measuredfrom the SMD simulations for ethylene glycol and sorbitol (circles). For comparison Δxu and ΔxA0-G are also shown for glycerol and water (13). For the smallerosmolytes, ethylene glycol, propylene glycol, and glycerol, there is a clear correlation between measured Δxu, ΔxA0-G and molecular size. For the largerosmolytes, sorbitol and sucrose, no such correlation between Δxu or ΔxA0-G and molecular size is observed and the value is Δxu is the same as that of water.(B) The 27th immunoglobulin-like domain of cardiac titin I27 protein (1TIT.pdb). Resistance to mechanical unfolding originates from six hydrogen bondsbetween two parallel β-strands A′ and G (19). (C) Snapshot from an SMD simulation of forced unfolding of the I27 protein in ethylene glycol solution. Osmolytebridging between the two parallel β-strands A′ and G is observed leading to an increased separation of the two β-strands. (D) Snapshot from an SMD simulationof forced unfolding of the I27 protein in sorbitol solution. The sorbitol molecule does not insert between the two parallel β-strands but instead forms ahydrogen bind from the terminal hydroxyl. No increased separation of the two β-strands is observed.

Fig. 4. Osmolyte participation in the unfolding transition state of protein.(A) Unfolding energy barrier ΔGu and (B) distance to the unfolding transitionstate Δxu for a range of ethylene glycol (squares), propylene glycol (dia-monds), sorbitol (triangles), and sucrose (circles) solutions of varying weightpercentage (osmolyte) obtained from force-clamp experiments. The dashedline in A shows monotonic increase of ΔGu with concentration of osmolyte.The dashed line in B shows exponential increase ofΔxu with concentration ofethylene glycol and propylene glycol and monotonic behavior with concen-tration of sorbitol and sucrose

9762 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1101934108 Dougan et al.

See Retraction Published

,

2011

December 12

separation length, which leads to complete unraveling of theprotein (Fig. S1). We measured ΔxA0-G for I27 unfolding at a con-stant force in water, ethylene glycol, and sorbitol solutions(Fig. 5A, circles). We found that ΔxA0-G increased from 2.84 Åin water to 3.72 Å in ethylene glycol, although in the simulationsof I27 unfolding in sorbitol solutions, ΔxA0-G remained relativelyunchanged at 2.97 Å. In Fig. 5A it can be seen that the values ofΔxA0-G from the simulations are in good qualitative agreementwith the experimental values of Δxu providing support that ourexperimental measurement of Δxu is a measurement of the elon-gation of A′ and G β-strands in the protein before the transitionstate is reached. The increase in Δxu and ΔxA0-G measured forethylene glycol, propylene glycol, and previously for glycerol (13)is smaller than the full elongation expected if Δxu and ΔxA0-Gsimply followed the molecular size of the osmolyte. Nonetheless,the near doubling of Δxu is highly significant because it directlypoints to an integral structural role of these osmolytes in theunfolding transition state of the I27 protein.

DiscussionRecent studies have shown that the mechanical stability of aprotein can be modulated by the presence of cosolvents in thesurrounding environment. Mechanical unfolding experiments ofthe protein I27 in aqueous glycerol solutions measured a consid-erable increase in the mechanical stability of the protein, withthe average unfolding force increasing by approximately 50%(13). Similarly, in the presence of 30% dextran, a polysaccharidemolecule, the average unfolding force of the protein ubiquitinincreases by approximately 11% (35). Conversely, denaturingosmolytes such as guanidinium chloride have been shown todecrease the mechanical stability of the small protein GB1 (theB1 immunoglobulin-binding domain of protein G from Strepto-coccus) by approximately 80% in 2.25 M GdmCl compared withan aqueous solution (36). Denaturing and protecting osmolytestherefore offer an attractive route to modulate the mechanicalproperties of a protein. Indeed, recent single-molecule atomicforce microscopy (AFM) experiments have shown that naturallyoccurring protecting and denaturing osmolytes have profoundeffects on the mechanical folding pathways of polycystic kidneydisease (PKD) domains (37). This study demonstrated thatprotecting osmolytes such as sorbitol and trimethylamine N oxideare efficient in counteracting the effect of the denaturing osmo-lyte urea on the mechanical stability of PKD domains.

In the present study we have tested the origin of enhancedmechanical stability in a protein by developing an experimentalapproach that provides insight into the role of osmolytes in theunfolding transition state structure of a protein, the main deter-minant of protein kinetics. The unfolding transition state of aprotein under force can be defined as that structure that, takenas a starting point, leads to either full unraveling or to a stablefold, with equal probability (38). Transition state structures areextremely short-lived, typically requiring femtosecond laser spec-troscopy for their capture (12). However, under a constantstretching force, the effect of mechanical work on the energylandscape of the unfolding protein is felt by the transition statestructure, regardless of its lifetime. Thus, our finding that thedistance to transition state of protein unfolding is sensitive tothe size of the osmolyte suggests that the osmolyte molecules,ethylene glycol, propylene glycol, and glycerol, are part of thisshort-lived structure. These results lend strong support to asolvent-bridging mechanism in the unfolding transition statestructure of the I27 protein (13). These experiments suggest thatalthough solvent bridging is possible for small osmolytes, it is notfeasible for the larger osmolytes. Our experimental protocol hasthus demonstrated the successful use of cosolvents as molecularprobes to determine features of the unfolding transition stateof the I27 protein. There have been many previous attemptsto understand the molecular mechanism of protein stabilization

by osmolytes such as glycerol (11, 39, 40). Our experimentalresults have identified that small osmolytes directly participatein the unfolding transition state structure of a protein and stabi-lize the protein. This lends support to a direct mechanism ofprotein stabilization by these osmolytes. On the other hand, largeosmolytes do not participate in the unfolding transition statestructure of a protein and yet stabilize the protein, albeit to alesser extent. This could lend support to an indirect mechanismof protein stabilization by these osmolytes. Our results thereforesuggest that osmolyte size is important in determining whichmolecular mechanism is responsible for protein stabilization.

It is interesting to note that although ΔxU increased for onlythe small osmolytes, an increase in ΔGU was measured for allosmolytes studied (Fig 4A). The experimental measurementsshowed that stabilization of the I27 protein (increase in ΔGU)is stronger for the smaller osmolytes than for the larger osmo-lytes. These results suggest it is energetically favorable whensmall osmolytes penetrate the mechanical clamp region of the I27protein and act as solvent bridges in the unfolding transition statestructure of this protein. The small osmolytes glycerol, ethyleneglycol, and propylene glycol are all efficient at hydrogen bondingand are able to form more hydrogen bonds than a water moleculein the unfolding transition state of the protein. Indeed, our SMDstudies of I27 unfolding in ethylene glycol solution and in glycerolsolution found evidence for multiple hydrogen bonds between theosmolyte and the β-strands A′ and G. As well as hydrogen bondability, molecule accessibility is an important factor in determin-ing the stability of a protein. Previous studies have shown thathindering solvent accessibility can increase the mechanical stabi-lity of a protein (19, 41, 42). We postulate that although smallosmolytes are efficient at hydrogen bonding, their ability to pe-netrate the mechanical clamp region may be hindered, comparedto that of water, due to geometrical constraints as well as shield-ing from surrounding β-strands. Therefore we propose that osmo-lyte hydrogen bonding ability and accessibility to the mechanicalclamp region in the protein is the critical molecular determinantfor enhancing the mechanical stability of a protein.

Given that cellular solvent composition is made of a wide vari-ety of osmolytes (43), molecular size may offer an attractive routefor proteins to vary the structural architecture of the unfoldingtransition state, thus controlling the dynamics of important mo-lecular mechanisms. Furthermore, regulation of cellular solventcomposition, by membrane channel proteins such as GlpF, maybe an important mechanism under conditions of mechanicalstress where sustained mechanical forces applied to tissues maytrigger widespread protein unfolding (44, 45). Phase separationand active selection of specific osmolytes may allow the controlof protein dynamics through modulation of the unfolding transi-tion state (46).

MethodsProtein Engineering and Purification. We constructed an eight-domain N–C-linked polyprotein of I27, the 27th Ig-like domain of cardiac titin, throughsuccessive cloning in modified pT7Blue vectors and then expressed the geneusing vector pQE30 in Escherichia coli strain BLR(DE3) (24). The polyproteinconstruct was finally purified by histidine metal-affinity chromatographywith Talon resin (Clontech) and by gel filtration using Superdex 200 HRcolumn (GE BioSciences).

Force Spectroscopy. Force-clamp AFM experiments were completed atroom temperature using a homemade setup under force-clamp conditionsdescribed elsewhere (14). Experiments were carried out in a sodium phos-phate buffer solution [specifically, 50 mM sodium phosphate (Na2HPO4 andNaH2PO4), and 150 mM NaCl, pH ¼ 7.4] with the desired weight percentageof the cosolvent. Samples of propylene glycol (99%), ethylene glycol (99%),glycerol (99%), sorbitol (99%), and sucrose (99%) were obtained from Sigma-Aldrich and used without additional purification.

Steered Molecular Dynamics Simulations. We completed SMD simulationsof I27 unfolding in ethylene glycol and sorbitol using a method described

Dougan et al. PNAS ∣ June 14, 2011 ∣ vol. 108 ∣ no. 24 ∣ 9763

CHEM

ISTR

Y

See Retraction Published

, 2011December 12

previously (13). Each simulation started from the equilibrated proteinstructure with the corresponding solvent. Constant forces are added to theC termini of the I27 protein, described in Table S1 and in SI Text. The coordi-nate PDB file of titin’s I27 protein (1TIT.pdb) was used. A molecular structurefile was generated for the full system using visual molecular dynamics (VMD)(47), and the solvent environment was modeled explicitly.

ACKNOWLEDGMENTS. This work was supported by a grant from theEngineering and Physical Sciences Research Council (EP/H020616/1) (toL.D.), National Institutes of Health Grants (HL66030 and HL61228) (toJ.M.F.), and China National Basic Research Program (2007CB947800) (to H.L.).It was also supported by the National Science Foundation through Teragridresources provided by TACC-Lonestar (TG-MCB090005) (to H.L.) andT32HL007692 from the National Heart, Lung, and Blood Institute.

1. Hénin J, Tajkhorshid E, Schulten K, Chipot C (2008) Diffusion of glycerol throughEscherichia coli aquaglyceroporin GlpF. Biophys J 94:832–839.

2. Ball P (2008) Water as a biomolecule. Chemphyschem 9:2677–2685.3. Chaplin M (2006) Opinion—Do we underestimate the importance of water in cell

biology? Nat Rev Mol Cell Biol 7:861–866.4. Levy Y, Onuchic JN (2006) Water mediation in protein folding and molecular recogni-

tion. Annu Rev Biophys Biomol Struct 35:389–415.5. Canchi DR, Paschek D, Garcia AE (2010) Equilibrium study of protein denaturation by

urea. J Am Chem Soc 132:2338–2344.6. Zhang YJ, Cremer PS (2010) Chemistry of Hofmeister anions and osmolytes. Annu Rev

Phys Chem 61:63–83.7. Rossky PJ (2008) Protein denaturation by urea: Slash and bond. Proc Natl Acad Sci USA

105:16825–16826.8. Holthauzen LMF, Rosgen J, Bolen DW (2010) Hydrogen bonding progressively

strengthens upon transfer of the protein urea-denatured state to water and protect-ing osmolytes. Biochemistry 49:1310–1318.

9. O’Brien EP, Ziv G, Haran G, Brooks BR, Thirumalai D (2008) Effects of denaturantsand osmolytes on proteins are accurately predicted by the molecular transfer model.Proc Natl Acad Sci USA 105:13403–13408.

10. Auton M, Bolen DW (2005) Predicting the energetics of osmolyte-induced proteinfolding/unfolding. Proc Natl Acad Sci USA 102:15065–15068.

11. Street TO, Bolen DW, Rose GD (2006) A molecular mechanism for osmolyte-inducedprotein stability. Proc Natl Acad Sci USA 103:13997–14002.

12. Zewail AH (2000) Femtochemistry: Atomic-scale dynamics of the chemical bond usingultrafast lasers—(Nobel lecture). Angew Chem Int Ed Engl 39:2586–2631.

13. Dougan L, Fang G, Lu H, Fernandez JM (2008) Solventmolecules bridge themechanicalunfolding transition state of a protein. Proc Natl Acad Sci USA 105:3185–3190.

14. Fernandez JM, Li H (2004) Force-clamp spectroscopy monitors the folding trajectory ofa single protein. Science 303:1674–1678.

15. Dougan L, Koti ASR, Genchev G, Lu H, Fernandez JM (2008) A single-molecule perspec-tive on the role of solvent hydrogen bonds in protein folding and chemical reactions.Chemphyschem 9:2836–2847.

16. Garcia-Manyes S, Dougan L, Fernandez JM (2009) Osmolyte-induced separation of themechanical folding phases of ubiquitin. Proc Natl Acad Sci USA 106:10540–10545.

17. Soper AK (2002) Water and ice. Science 297:1288–1289.18. Lu H, Schulten K (1999) Steered molecular dynamics simulations of force-induced

protein domain unfolding. Proteins 35:453–463.19. Lu H, Schulten K (2000) The key event in force-induced unfolding of titin’s immuno-

globulin domains. Biophys J 79:51–65.20. Lu H, Isralewitz B, Krammer A, Vogel V, Schulten K (1998) Unfolding of titin immu-

noglobulin domains by steered molecular dynamics simulation. Biophys J 75:662–671.21. Li JY, Fernandez JM, Berne BJ (2010) Water’s role in the force-induced unfolding of

ubiquitin. Proc Natl Acad Sci USA 107:19284–19289.22. Yancey PH, Clark ME, Hand SC, Bowlus RD, Somero GN (1982) Living with water-stress

—Evolution of osmolyte systems. Science 217:1214–1222.23. Scheiner S, Cuma M (1996) Relative stability of hydrogen and deuterium bonds. J Am

Chem Soc 118:1511–1521.24. Carrion-Vazquez M, et al. (1999) Mechanical and chemical unfolding of a single

protein: A comparison. Proc Natl Acad Sci USA 96:3694–3699.25. Li HB, et al. (2002) Reverse engineering of the giant muscle protein titin. Nature

418:998–1002.26. Garcia-Manyes S, Brujic J, Fernandez JM (2007) Force-clamp spectroscopy of single

protein monomers reveals the individual unfolding and folding pathways of 127and ubiquitin. Biophys J 93:2436–2446.

27. Schlierf M, Li H, Fernandez JM (2004) The unfolding kinetics of ubiquitin capturedwith single-molecule force-clamp techniques. Proc Natl Acad Sci USA 101:7299–7304.

28. Garcia-Manyes S, Dougan L, Badilla CL, Brujic J, Fernandez JM (2009) Direct observa-tion of an ensemble of stable collapsed states in the mechanical folding of ubiquitin.Proc Natl Acad Sci USAmerica 106:10534–10539.

29. Kiyosawa K (1991) Volumetric properties of polyols (ethylene-glycol, glycerol, meso-erythritol, xylitol and mannitol) in relation to their membrane-permeability—Groupadditivity and estimation of the maximum radius of their molecules. Biochim BiophysActa 1064:251–255.

30. Liu YF, Bolen DW (1995) The peptide backbone plays a dominant role in proteinstabilization by naturally-occurring osmolytes. Biochemistry 34:12884–12891.

31. Brujic J, Hermans RI, Walther KA, Fernandez JM (2006) Single-molecule force spectro-scopy reveals signatures of glassy dynamics in the energy landscape of ubiquitin.Nature Phys 2:282–286.

32. Brujic J, Hermans RIZ, Garcia-Manyes S, Walther KA, Fernandez JM (2007) Dwell-timedistribution analysis of polyprotein unfolding using force-clamp spectroscopy. BiophysJ 92:2896–2903.

33. Efron B (1982) The Jackknife, the Bootstrap, and other Resampling Plans, , CBMS-NSFMonographs (Society of Industrial and AppliedMathematics, Philadelphia, PA), Vol 38.

34. Wiita AP, et al. (2007) Probing the chemistry of thioredoxin catalysis with force.Nature450:124–127.

35. Yuan JM, et al. (2008) The effects of macromolecular crowding on the mechanicalstability of protein molecules. Protein Sci 17:2156–2166.

36. Cao Y, Li H (2008) How do chemical denaturants affect the mechanical folding andunfolding of proteins? J Mol Biol 375:316–324.

37. Ma L, Xu MX, Oberhauser AF (2010) Naturally occurring osmolytes modulate thenanomechanical properties of polycystic kidney disease domains. J Biol Chem285:38438–38443.

38. Li L, Shakhnovich EI (2001) Constructing, verifying, and dissecting the foldingtransition state of chymotrypsin inhibitor 2 with all-atom simulations. Proc Natl AcadSci USA 98:13014–13018.

39. Hua L, Zhou RH, Thirumalai D, Berne BJ (2008) Urea denaturation by strongerdispersion interactions with proteins than water implies a 2-stage unfolding. Proc NatlAcad Sci USA 105:16928–16933.

40. Bennion BJ, Daggett V (2003) The molecular basis for the chemical denaturation ofproteins by urea. Proc Natl Acad Sci USA 100:5142–5147.

41. Berne BJ, Li JY, Fernandez JM (2010) Water’s role in the force-induced unfolding ofubiquitin. Proc Natl Acad Sci USA 107:19284–19289.

42. Guan ZB, Guzman DL, Randall A, Baldi P (2010) Computational and single-moleculeforce studies of a macro domain protein reveal a key molecular determinant formechanical stability. Proc Natl Acad Sci USA 107:1989–1994.

43. Walter H, Brooks DE, Srere PA (2000)Microcompartmentation and Phase Separation inCytoplasm (Academic, San Diego).

44. de Groot BL, Grubmuller H (2001) Water permeation across biological membranes:Mechanism and dynamics of aquaporin-1 and GlpF. Science 294:2353–2357.

45. Wang Y, Schulten K, Tajkhorshid E (2005) What makes an aquaporin a glycerolchannel? A comparative study of AqpZ and GlpF. Structure 13:1107–1118.

46. Helfrich MR, Mangeney-Slavin LK, Long MS, Djoko KY, Keating CD (2002) Aqueousphase separation in giant vesicles. J Am Chem Soc 124:13374–13375.

47. Humphrey W, Dalke A, Schulten K (1996) VMD: Visual molecular dynamics. J MolGraph 14:33–38.

9764 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1101934108 Dougan et al.

See Retraction Published , 2011 December 12