Very Fast Folding and Association of a Trimerization Domain from Bacteriophage T4 Fibritin

Very Fast Folding and Association of a TrimerizationDomain from Bacteriophage T4 Fibritin

Sarah Guthe1†, Larisa Kapinos1†, Andreas Moglich1†Sebastian Meier2, Stephan Grzesiek2 and Thomas Kiefhaber1*

1Division of BiophysicalChemistry, Biozentrum derUniversitat BaselKlingelbergstrasse 70, CH-4056Basel, Switzerland

2Division of Structural BiologyBiozentrum der UniversitatBasel, Klingelbergstrasse 70CH-4056 Basel, Switzerland

The foldon domain constitutes the C-terminal 30 amino acid residues ofthe trimeric protein fibritin from bacteriophage T4. Its function is topromote folding and trimerization of fibritin. We investigated structure,stability and folding mechanism of the isolated foldon domain. Thedomain folds into the same trimeric b-propeller structure as in fibritinand undergoes a two-state unfolding transition from folded trimer tounfolded monomers. The folding kinetics involve several consecutivereactions. Structure formation in the region of the single b-hairpin ofeach monomer occurs on the submillisecond timescale. This reaction isfollowed by two consecutive association steps with rate constants of1.9(^0.5) £ 106 M21 s21 and 5.4(^0.3) £ 106 M21 s21 at 0.58 M GdmCl,respectively. This is similar to the fastest reported bimolecular associationreactions for folding of dimeric proteins. At low concentrations of protein,folding shows apparent third-order kinetics. At high concentrations ofprotein, the reaction becomes almost independent of protein concen-trations with a half-time of about 3 ms, indicating that a first-order foldingstep from a partially folded trimer to the native protein (k ¼ 210ð^20Þ s21)becomes rate-limiting. Our results suggest that all steps on the folding/trimerization pathway of the foldon domain are evolutionarily optimizedfor rapid and specific initiation of trimer formation during fibritinassembly. The results further show that b-hairpins allow efficient andrapid protein–protein interactions during folding.

q 2004 Elsevier Ltd. All rights reserved.

Keywords: protein folding; protein association; trimeric proteins; prolylisomerization; fast folding*Corresponding author

Introduction

Fibritin is a rod-like structural protein ofbacteriophage T4, which is attached to the neckof the virion via its N-terminal domain to formthe collar structures (“whiskers”). Fibritinconsists of an N-terminal anchor domain (residues1–46), a large central coiled-coil part (residues47–456) and a small C-terminal globular domain(residues 457–486).1 The 30 amino acid residueC-terminal domain was termed foldon, since it was

shown to be essential for fibritin trimerizationand folding in vivo and in vitro.1 – 3 Each subunitof the foldon domain consists of a single b-hairpin,which assemble into a b-propeller-like structurein the trimer.1 The trimer is stabilized by hydro-phobic interactions involving Trp476 of eachsubunit, intermolecular salt-bridges betweenGlu461 and Arg471, and intermolecularbackbone hydrogen bonds between Tyr469 andArg471 (Figure 1). Expression of the isolatedfoldon domain (residues 457–483) yields astable trimer, which shows a cooperative two-state thermal unfolding transition.4 Residues484–486 were omitted from this study, since thisregion is unordered in the X-ray structure offibritin.1

The foldon domain was proposed to be anevolutionarily optimized trimerization/folding

0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.

† S.Gu., L.K. and A.M. contributed equally to this work.

E-mail address of the corresponding author:[email protected]

Abbreviations used: F-moc, N-(9-fluorenyl)-methoxycarbonyl; GdmCl, guanidinium chloride; RDC,residual dipolar coupling.

doi:10.1016/j.jmb.2004.02.020 J. Mol. Biol. (2004) 337, 905–915

motif, as its only known function is to promotefolding of fibritin.3 The small size of its structuredpart (27 amino acid residues) and its simple foldmake the foldon domain a perfect system for adetailed study on the mechanism of a foldingreaction linked to intermolecular association steps.All previous folding studies on trimeric proteinsinvestigated large filamentous proteins, whichshow extremely slow and complex folding kinetics,usually accompanied by irreversible aggregationreactions.5

We expressed the foldon domain in Escherichiacoli and synthesized it by solid-phase N-(9-fluorenyl)methoxycarbonyl (F-moc) chemistry toinvestigate its structure, stability and foldingmechanism. For clarity, we are numbering the

foldon sequence from residues 1 to 27

corresponding to residues 457–483 in fibritin. Allkinetic and stability data presented here wereobtained using the chemically synthesized foldondomain, whereas the recombinant E. coli productwas used for structural analysis. The E. coliproduct and the synthetic foldon domain showedidentical stability and folding behavior. Further,the additional C-terminal amino acid residues Ser-Pro-Ala, which are present in the wild-type fibritinsequence, do not affect any thermodynamic orkinetic properties of the foldon domain.

Figure 1. A, Stereo view of a bundle of the 20 lowest-energy structures of the trimeric foldon domain determined byNMR spectroscopy. Each subunit is displayed in a different color. B, Side view of the foldon structure with the singleTrp residues at position 20 of each chain highlighted in green and the two prolyl residues at positions 4 and 7 high-lighted in red. C, Topology of the interactions of the three b-hairpins in the native foldon domain. The figures in Aand B were prepared using the program MOLMOL37 and rendered with PovRay.

906 Fast Folding of a Trimerization Domain

Results and Discussion

Structure and stability of the foldon domain

To test whether the 27 amino acid residue foldondomain adopts the same fold as in fibritin, wesolved its solution structure to a backbone rmsd of0.31 A with 28 experimental restraints per residue(Figure 1 and Table 1). The structure largelyresembles the crystal structure in constructs carry-ing the 75 and 120 C-terminal amino acid residuesof fibritin, in which the foldon domain constitutesonly a minor part of the total construct.1,6 Only atthe immediate N terminus (residues 1–3) theisolated foldon domain assumes a slightly differentand presumably more relaxed structure compared

to fibritin. The trimer consists of an N-terminalhydrophobic stretch in left-handed polyproline IIhelix conformation between Pro4 and Pro7, whichis connected to a b-hairpin (residues 12–23) andforms a hydrophobic cap of the hairpin on theN-terminal side (Figure 1). The hairpin terminatesin a 310 helix at the C terminus with homophilicinteractions of hydrophobic residues (Tyr2, Ile3,Val14, Leu23, Leu27) between the monomersalong the symmetry axis. Large-scale nanoseconddynamics as evidenced by 15N relaxation occuronly at the most N-terminal residue, Tyr2. Allother residues in the highly rigid foldon domainexhibit order parameters S2 . 0:77 at 25 8C asdetermined by the program TENSOR.7

The equilibrium unfolding properties of thefoldon domain were measured by guanidiniumchloride (GdmCl)-induced unfolding transitions atvarious concentrations of protein. Figure 2 showsunfolding curves at monomer concentrations of5 mM and 30 mM. The coincidence of fluorescenceand CD-monitored transition curves demonstratesthat the trimer unfolds in a cooperative two-statetransition at both concentrations of protein. Two-state unfolding is observed for all concentrationsof protein between 2 mM and 100 mM. The sensi-tivity of the unfolding transitions to changes inprotein concentration is expected for unfolding ofa native trimer (N) to unfolded monomers (U):

N O 3U ð1Þ

The transitions at 5 mM and 30 mM can be fitglobally to equation (1) (continuous line inFigure 2) by using:

Keq ¼3f 3

U½M�201 2 fU

ð2Þ

where ½M�0 indicates the total monomer concen-tration ð½M�0 ¼ ½U� þ 3½N�Þ, fu is the fraction of

Table 1. Statistics of the foldon NMR structure

rmsd from experimental distance constraints (A)All (607)a 0.046 ^ 0.002

rmsd from NMR dataNMR quality factor Qb 0.199 ^ 0.0062rmsd (Hz) between measured and calcu-

lated dipolar couplings (81)c

1.91 ^ 0.07

Experimental dihedral constraints (deg.)d (43) 1.73 ^ 0.293JHNHA coupling constants (Hz) (22) 0.89 ^ 0.05Total number of restraints per monomer 753

Deviation from the idealized covalent geometryBonds (A) 0.0079 ^ 0.0004Angles (deg.) 0.92 ^ 0.03Improperse (deg.) 0.75 ^ 0.06

Coordinate precisionf (A)Backbone non-hydrogen atoms 0.273All non-hydrogen atoms 0.636

Non-Gly, non-Pro residues in Ramachandran regionsg

Most favored (%) 91.7Allowed (%) 8.3Generously allowed (%) 0.0Disallowed (%) 0.0

The statistics were obtained from a subset of the 40 bestenergy structures out of 100 following a standard simulatedannealing protocol with dipolar restraints incorporated. Individ-ual simulated annealing structures are fitted to each other usingresidues 2–27 of all subunits. The number of the various con-straints per monomer is given in parentheses.

a Distance restraints comprise: 111 intraresidual NOEs; 139sequential NOEs ðli 2 jl ¼ 1Þ; 73 short range NOEs ð1 , li 2 jl #5Þ; 128 long-range NOEs ðli 2 jl # 5Þ; 156 intermolecular NOEs;11 H-bonds (eight intramolecular, three intermolecular). Foreach backbone hydrogen bond constraint, there are two distancerestraints: rNH– O; 1.7–2.5 A, rN– O; 2.3–3.5 A.

b The NMR quality factor Q is defined as the ratio of thermsd between observed and calculated couplings and the rmsdof the observed couplings.35

c The 81 RDCs comprise 22 1DHN, 20 1DHaCa, 12 1DCaCb, 131DNC0 (0.231), 14 1DCH3. Ramping the force constant for RDCs inthe structure calculation from 0.001 kcal mol21 Hz22 to0.5 kcal mol21 Hz22 was determined as optimal.

d The dihedral angle constraints comprise 69 f and 60 cangles.

e The improper torsion restraints serve to maintain planarityand chirality.

f The coordinate precision is defined as the average rmsdifference between the individual simulated annealingstructures and the mean coordinates. Values are reported forresidues 2–27.

g These values are calculated with the program PROCHECK-NMR.36 Values are reported for all residues.

Figure 2. GdmCl-induced unfolding transition of thefoldon domain at pH 7.1, 20 8C. Transitions at 5 mM(W,X) and 30 mM (O,K) total monomer concentrationð½M�0Þ were measured by changes in Trp fluorescence(X,O) and in far-UV CD at 228 nm (W,K). The data werenormalized to fraction of native molecules using theresult of a global fit of all data according to equations(14a) and (14b) (continuous lines).

Fast Folding of a Trimerization Domain 907

unfolded monomer ð fU ¼ ½U�=½M0�Þ and Keq is theequilibrium constant (for details, see Materialsand Methods). The global fit yields a free energyof unfolding of DG0ðH2OÞ ¼ 89:2ð^0:6Þ kJ/mol,which is unusually high compared to stabilities ofsmall, single-domain proteins of similar size. How-ever, this value applies to standard conditions of1 M total monomer concentration. At typicalphysiological protein concentrations around 5 mMthis corresponds to DG ¼ 29:7 kJ, which is similarto the stabilities found for small monomericproteins. The change in free energy with GdmClðmeq ¼ ›DG0=›½GdmCl�Þ is 210.4(^0.2) (kJ/mol)/M, which is the value expected for a monomericglobular protein of the size of the folded trimer.8

This shows that native foldon has properties com-parable to those of small monomeric proteins witha compact hydrophobic core and a cooperativetwo-state unfolding transition.

Burst phase fluorescence changes

To investigate the folding kinetics of the foldondomain we performed stopped-flow refoldingexperiments starting from GdmCl-unfoldedprotein. Figure 3A shows a refolding trace at aresidual denaturant concentration of 0.58 M and½M�0 of 5 mM. The kinetics were monitored by thechange in intrinsic tryptophan fluorescence above320 nm. Within the first millisecond of refolding, amajor burst phase reaction occurs, which leads toa significant increase in fluorescence intensityabove the signals of both the unfolded and thenative protein. This indicates very rapid structuralchanges in the dead-time of stopped-flow mixing(about 1 ms). The fluorescence intensity decreasesslowly and reaches the value of the native proteinafter about 300 seconds. The burst phase increasein fluorescence is observed for all measuredconcentrations of protein (0.5 mM to 200 mM),indicating that the reaction occurs within themonomer. Even the fastest, diffusion-controlledassociation reaction to a partially folded dimercould not be complete within 1 ms at a monomerconcentration of 1 mM and below, if we assume amaximum second-order rate constant9 of about1 £ 109 M21 s21. To further investigate the structuralchanges occurring in the burst phase, we moni-tored the folding kinetics at single wavelengthsbetween 290 nm and 430 nm. Extrapolating thekinetic traces at the individual wavelengths totime zero allows the determination of the fluor-escence spectrum of the burst phase intermediate(Figure 3B). Comparison of the fluorescencespectra of native and unfolded protein with thezero timepoint spectrum shows that the burstphase intermediate has a fluorescence emissionmaximum around 330 nm, which is between theemission maximum of the native protein(lmax ¼ 317 nm) and the unfolded state(lmax ¼ 345 nm). The significantly blue-shiftedfluorescence maximum and the largely increasedfluorescence intensity in the intermediate relative

to the unfolded state suggest that the burst phaseintermediate has a significantly more hydrophobicenvironment around the single tryptophan residueat position 20 in each b-hairpin (see Figure 1). Theabsence of a tyrosine fluorescence band at 303 nmin the burst phase intermediate further indicatessignificant chain compaction, which allowsefficient energy transfer from the two tyrosine resi-dues at positions 2 and 13 to Trp20. Similarfluorescence properties are observed for an acid-induced monomeric state (A-state) of the foldondomain, which shows virtually the same fluor-escence emission spectrum as the burst phaseintermediate but with reduced fluorescenceintensity (Figure 3).

Fast and slow steps during association of thefoldon domain

To determine the nature of the rate-limiting steps

Figure 3. A, Refolding of the foldon domain in 0.58 MGdmCl, pH 7.1 (½M�0 5 mM) measured by the change inTrp fluorescence using a 320 nm emission cut-off filter.The broken lines represent the signals of the native andof the unfolded state at 0.58 M GdmCl, as indicated.The signal of the unfolded state is extrapolated fromthe unfolded baseline at high concentrations of GdmClto 0.58 M GdmCl (see Materials and Methods). B, Com-parison of the fluorescence spectrum of the kinetic burstphase intermediate (I) with the spectra of native (N)protein in 0.58 M GdmCl, the unfolded protein (U) in8.2 M GdmCl and the monomeric A-state formed at pH2 (A). The spectrum of I was determined in single-wavelength detection stopped-flow experiments. Thefluorescence intensity extrapolated to t ¼ 0 is shown.½M�0 was 5 mM for all spectra.


during folding and association of the foldondomain, we analyzed the concentration-dependence of the refolding kinetics at final con-centrations of protein between 0.5 mM and200 mM. Figure 4A shows that the kinetics arestrongly concentration-dependent. A very slowreaction on the hundreds of seconds timescale isobserved at low concentrations of protein. Athigher concentrations of protein (½M�0 . 10 mM)a concentration-dependent faster reaction with ahalf-time of about 4 ms at ½M�0 ¼ 102 mM and aconcentration-independent slow reaction (t ¼ 50 s)are observed. The foldon domain contains twoprolyl residues (Pro4 and Pro7) per subunit, whichare in the trans conformation in the native state.This makes prolyl isomerization reactions apossible source for the slow, concentration-independent kinetics. Figure 5 shows that theslow reaction is catalyzed efficiently by humancyclophilin 18, a peptidyl-prolyl cis– transisomerase,10 which identifies this reaction as acis– trans isomerization at one or both of the twoXaa-Pro peptide bonds per monomer. The faster reaction is not affected by the presence of cyclo-

philin (data not shown).We tested whether the faster, concentration-

dependent reaction produces native protein or afolding intermediate by performing interruptedrefolding experiments.11 – 13 In these experiments,the protein is allowed to refold for a certain timeðtiÞ: The protein is then transferred to unfoldingconditions and the resulting unfolding kinetics aremonitored. The native state has a characteristicstability and barrier for unfolding, which resultsin a characteristic rate constant for its unfoldingreaction. This distinguishes it from partially foldedintermediates. Interrupted refolding experimentsmeasure the increase in amplitude of the unfoldingreaction of the native protein as a function of therefolding time, ti: This corresponds to the time-course of formation of the native state. Thus, inter-rupted refolding experiments can distinguishwhether a folding reaction produces native proteinor a folding intermediate.12,13 Figure 4B shows thatnative molecules are formed in a fast and a slowreaction, which occur on the same timescale as thetwo fluorescence-detected folding reactions(Figure 4B). Obviously, both the faster, concen-tration-dependent process and the prolyl-isomeri-zation limited process produce native protein.This suggests that the fast reaction reflects for-mation of the native trimer for molecules with allprolyl peptide bonds in the native trans confor-mation (fast-folding molecules, UF). The slowprocess is due to folding of molecules with at leastone non-native cis isomer (slow-folding molecules,US), as shown by its catalysis by cyclophilin(Figure 5).

Mechanism of folding and association

The assignment of the different kinetic phases todirect folding and prolyl isomerization stepsenables us to characterize the folding/association

Figure 4. A, Fluorescence detected folding kinetics inthe presence of 0.58 M GdmCl (pH 7.1), 20 8C at the indi-cated values of ½M�0 measured after stopped-flow mixingat an emission wavelength of 320 nm. B, Time-course offormation of native molecules measured in interruptedrefolding experiments at pH 7.1, 20 8C, ½M�0 ¼ 10 mM.The continuous lines in both panels represent the resultsfrom a global fit of the data. For global fitting, ten fluor-escence-detected refolding traces at ½M�0 between0.5 mM and 102 mM and the time-course of formation ofnative molecules shown in B, were fitted simultaneouslyto the kinetic model shown in equation (4).

Figure 5. Effect of human cyclophilin 18 on the slowrefolding reaction of the foldon domain. The plot com-pares refolding at pH 7.1, 0.58 M GdmCl, 20 8C in theabsence and in the presence of 3.3 mM cyclophilin. Thepresence of cyclophilin increases the rate constant forthe slow reaction from 7 £ 1023 s21 to 4 £ 1022 s21.


process of the foldon domain in more detail. Theanalysis of the concentration-dependence of thehalf-lives ðt1=2Þ of the folding reaction startingfrom UF allows us to determine the apparent reac-tion order for the fast-folding pathway accordingto:

log t1=2 ¼ c 2 ðn 2 1Þlog ½M�0 ð3Þ

where n is the reaction order and c is a reaction-specific constant.14 Analysis of the concentration-dependence of folding thus yields information oncontributions from concentration-dependentassociation reactions and from concentration-independent folding steps. We determined thehalf-time of the folding reaction from the concen-tration-dependence of the fluorescence-detectedkinetics shown in Figure 4A. At high concen-trations of protein, where folding and prolylisomerization are well separated, we evaluated thehalf-lives of the fast reactions in order to obtaininformation on the direct folding reaction. Figure 6shows that the slope of logt1=2 versus log½M�0

changes with protein concentration. At lowconcentrations of protein the slope is22.06 ^ 0.05, which corresponds to apparentthird-order kinetics. At monomer concentrationsabove 5 mM the slope decreases significantly, indi-

cating a change in the apparent reaction order.Between 100 mM and 200 mM initial monomerconcentration there is only little effect of proteinconcentration on the folding kinetics. This showsthat the fast-folding reaction approaches the first-order limit at high concentrations of protein.

The change from apparent third-order kinetics tofirst-order kinetics shows that folding is limited byassociation steps at low concentrations of proteinand by a unimolecular folding reaction at high con-centrations of protein. It is very unlikely thatapparent third-order kinetics for a reaction in sol-ution arise from a true trimolecular reaction.15

Third-order reactions are usually caused by arapid monomer–dimer pre-equilibrium followedby a second bimolecular association step to formthe trimer. This mechanism gives rise to apparentthird-order kinetics if the dimer is populated toonly very low levels. The observation of apparentthird-order kinetics at low concentrations ofprotein confirms the finding that the rapidcollapse observed within the first millisecond ofrefolding is due to a conformational change in themonomer.

The weak concentration-dependence of thefluorescence-detected kinetics at high concen-trations of protein implies that a unimolecular fold-ing step becomes rate-limiting when all associationreactions are fast. The half-time for formation ofthe native state of about 3–4 ms at high concen-trations of protein indicates that the rate constantof the unimolecular step is around 200–300 s21. Itis reasonable to assume that this unimolecularfolding process is due to a structural rearrange-ment in the trimer and represents a late step in theformation of the native structure, and might besimilar to the final steps during folding of asingle-domain protein. However, we cannotexclude completely the possibility that the first-order reaction, which becomes rate-limiting athigh concentrations of protein, occurs at the levelof the burst phase intermediate or a partiallyfolded dimer. These considerations lead to theminimal folding model for the foldon domain:

In this mechanism, U denotes the completelyunfolded monomer, I is the monomeric burst phaseintermediate, D is a partially folded dimer, T is apartially folded trimer and N the native trimer. Thesubscripts c and t indicate monomers with a least

Figure 6. Effect of total monomer concentration ð½M�0Þon the half-time of the fast-refolding reaction of thefoldon domain at 0.58 M GdmCl (pH 7.1), 20 8C. The con-tinuous line represents a fit of the data between 0.5 mMand 4 mM to equation (3). The fit gives a slope of22.06 ^ 0.05 indicating an apparent reaction order of 3at low concentrations of protein.

ð4Þ


one cis Xaa-Pro peptide bond and monomers with alltrans peptide bonds, respectively. km and k2m rep-resent rate constants for formation and unfolding ofthe burst phase intermediate, respectively, whichoccurs on the submillisecond timescale. To testwhether this folding mechanism can describe allexperimental data quantitatively, we fitted the fold-ing kinetics at different concentrations of protein(Figure 4A) together with the time-course of for-mation of native foldon at 10 mM (Figure 4B) globallyto the mechanism shown in equation (4). The resultsshow that both the concentration-dependence offluorescence kinetics and the time-course of for-mation of native molecules are well described bythe model (continuous lines in Figure 4A and B).The global fit allows the determination of the rateconstants for all reactions that occur after the sub-millisecond burst phase, since the apparent reactionorder changes from third-order to almost first-order.The results of the global fit are given in equation (4).In agreement with the lower limit of the estimatefrom the limiting value of the half-lives for foldingat high concentrations of protein (Figure 6) the fitgives a rate constant ðkfÞ of structural rearrangementin the trimer of 210(^20) s21. This is in the samerange as the rate constants observed for the fastestfolding monomeric proteins of similar size.16 Thebimolecular rate constants for formation of thedimer and of the trimer are 1.9(^0.5) £ 106 M21 s21

and 5.4(^0.3) £ 106 M21 s21, respectively, which isabout 100 to 200-fold slower than the expecteddiffusion limit for bimolecular reactions of chains ofthis size.9 However, both association reactions onthe foldon trimerization pathway are significantlyfaster than most bimolecular steps during folding ofdimeric proteins like the well-characterized GCN417

leucine zipper and of many globular dimericproteins.18 Rate constants similar to the twobimolecular steps during trimerization of foldonhave been reported for wild-type arc repressor,9 thefastest folding naturally occurring dimeric proteinknown, and for some designed leucine zippers.19

However, an engineered arc repressor variant20 anda designed fragment of trp repressor21 showbimolecular rate constants of about 3 £ 108 M21 s21,which are the fastest association reactions reportedfor folding of small dimeric proteins to date. In thecase of the engineered arc repressor, the rate-enhancement relative to the wild-type protein wasachieved by replacing the intermolecular salt-bridge/hydrogen bonding network by hydrophobicresidues.20 The fast-folding trp repressor fragment isstabilized mainly by intermolecular hydrophobicinteractions. Similar to wild-type arc repressor, thefoldon domain is stabilized by intermolecular hydro-gen bonds and by an intermolecular salt-bridge,which contributes substantially (17 kJ/mol) to trimerstability (S.M. et al., unpublished results). In thisrespect, it is interesting to note that the associationsteps in the foldon domain have virtually the samerate constants as the dimerization step of wild-typearc repressor.

The unfolding rate constants of foldon under

native conditions (0.58 M GdmCl) obtained fromthe fit reveal that the major barrier for unfoldingis represented by the reaction from the nativetrimer (N) to the partially folded trimer(k ¼ 4:2ð^0:5Þ £ 1024 s21). This value correspondswell to the rate constant for unfolding of nativefoldon measured at high concentrations ofdenaturant and extrapolated to 0.58 M GdmCl(S.Gu. & T.K., unpublished results). Unfolding ofthe partially folded trimer (T) and the dimer (D)are significantly faster with rate constants of110(^40) s21 and 59(^4) s21, respectively, obtainedfrom the fit.

In the folding mechanism shown in equation (4),we assumed that only the monomers with nativetrans prolyl isomers can form productive dimersand trimers. According to studies on model pep-tides the Ile3-Pro4 (12% cis) and Ala6-Pro7 (8% cis)peptide bonds in foldon should lead to 19%unfolded monomers with at least one cis prolylpeptide bond.22 This agrees well with 20(^1)% ofslow-folding molecules observed in the interruptedrefolding experiments (equation (4) and Figure 4B),if we assume that only monomers with both prolylpeptide bonds in trans can enter the productivefolding pathway. The presence of 20(^1)% slow-folding molecules would, however, be observed ifonly the Ala6–Pro7 peptide bond, which is in ahighly structured region of foldon (Figure 1), wasessential for folding and if a cis bond at thisposition could be incorporated into partially foldeddimers and trimers. For this mechanism, thepresence of 22% slow-folding molecules would beexpected. Our data do not allow us to discriminatebetween these mechanisms. However, the rateconstants for the fast-folding pathway are notinfluenced significantly by the folding mechanismof the slow-folding molecules.

The determination of all rate constants on thefolding/association pathway of the foldon domainenables us to calculate the population of eachspecies during folding at various concentrations ofprotein (Figure 7). At an initial concentration of1 mM monomer, only the burst phase intermediate(I) and the native state become populated signifi-cantly (.10%) during folding. At concentrationsof protein above 5 mM, the dimer becomes popu-lated transiently, in agreement with a change inreaction order around this concentration of protein(Figure 6). As a consequence, the fast-folding path-way changes from apparent three-state to apparentfour-state with the burst phase intermediate andthe dimer populated to significant amounts.Above a concentration of 50 mM monomer, alsothe trimeric intermediate becomes populatedsignificantly. The significant population of dimericand trimeric intermediates explains the low appar-ent half-time of the reaction at high concentrationsof protein, which would suggest a unimolecularfolding reaction faster than the 210(^20) s21

obtained from the global fit. Since the apparenthalf-time was determined from the fluorescencemeasurements (Figure 6), it will be influenced by


the kinetics of formation of dimers and trimerswhen these become populated significantly. Thesereactions are fast at high concentrations of protein,which results in apparently smaller and slightlyconcentration-dependent half-times for the foldingreaction at high concentrations of protein. How-ever, these half-times do not represent the truehalf-times for the first-order reactions.

During phage assembly in the cell, the estimatedconcentration of monomeric fibritin molecules isbetween 1 mM and 5 mM. Figure 7 shows thatunder these conditions neither the dimeric (D) northe trimeric (T) intermediate becomes populatedto significant amounts. At in vivo concentrations,

the half-time for folding is between one and tenseconds, which is very fast compared to thegeneration time of the phage of around 30 minutes.

Fast folding of multimeric proteins comparedto monomeric proteins

The folding mechanism of the foldon domain hasseveral interesting differences compared to fast-fold-ing reactions of small monomeric proteins. Smallmonomeric proteins usually do not populate par-tially folded states to significant amounts duringfolding, although high-energy intermediates wereshown to be obligatory for folding of many apparenttwo-state folders.23,24 Obviously, small monomericproteins are able to avoid transient high concen-trations of intermediates, which minimizes the prob-ability of aggregation side-reactions and optimizesthe shape of the free energy barriers for rapidfolding.25 In the case of the foldon domain, however,rapid formation of intermolecular interactions isessential for efficient folding. Formation of the burstphase intermediate (I) leads (i) to pronounced fluor-escence changes of the single Trp residue, which ispart of the b-hairpin, (ii) to chain compaction as indi-cated by efficient energy transfer from the twotyrosine residues to the tryptophan and (iii) to sig-nificant changes in the far-UV CD signal (data notshown). This suggests that the monomeric inter-mediate involves structure formation of the b-hair-pin, which is in agreement with the results of NMRstudies on a monomeric state of the foldon domain(A-state) observed in equilibrium at low pH. In thisstate, the b-hairpin forms essentially the same struc-ture as in the folded trimer (S.M. et al., unpublishedresults). This shows that the monomeric foldondomain has a high propensity to form a b-hairpineven in the absence of intermolecular interactions.The similarity between the burst phase intermediateand the A-state is supported by the identicalfluorescence emission maxima of the two states(Figure 3B). Rapid formation of the b-hairpins in themonomers probably facilitates the subsequentassociation reactions, since the hairpins providehighly specific surfaces that allow fast formation ofintermolecular interactions. This model is in agree-ment with the folding mechanisms of the dimericGCN4 fragment, for which rapid formation of a-heli-cal structure was shown to accelerate the subsequentbimolecular association step.26 Obviously, optimizedenergy landscapes for folding of monomeric andoligomeric proteins are different. Monomericproteins fold fastest when the intermediates aremarginally less stable than the unfolded state,25

whereas populated intermediates promote the sub-sequent concentration-dependent association reac-tions in oligomeric proteins. The rapid formation ofa partially folded state in the 27 residue foldonmonomer shows that small, single-domain proteinsshould be able to rapidly form partially folded inter-mediates, provided this was beneficial for efficientfolding.

Figure 7. Simulation of the time-course of all kineticspecies during folding of the foldon domain at 0.58 MGdmCl (pH 7.1), 20 8C. The results from the global fit ofall kinetic data (see Figure 4) to the kinetic modelshown in equation (4) were used to calculate the concen-trations of the different species at the indicated values of½M�0:


Conclusions

Our folding studies on the foldon domain providethe first detailed information on the folding mechan-ism of a fast-folding trimeric protein. Previousstudies on trimeric proteins focused mainly on largeproteins, which fold very slowly and areaccompanied by irreversible aggregation reactions5

or on collagens, which are limited in folding byprolyl cis–trans isomerization reactions.27 Our resultssuggest that in the small 27 amino acid residuefoldon domain, all folding and association steps areoptimized for rapid formation of a stable trimer. Thetwo consecutive association reactions showbimolecular rate-constants similar to the valuesfound in the fastest-folding small dimeric proteins.Further, the first-order folding reaction, whichbecomes rate limiting at high concentrations ofprotein, has a rate constant comparable to those offast-folding, small, single-domain proteins. In con-trast to previously studied small multimeric proteinsor protein fragments, where association involvedmainly helix–helix interactions, the foldon domaintrimerizes via b-hairpins. The high bimolecular rateconstants for both association steps in the foldondomain demonstrate that b-hairpins allow veryrapid and specific formation of intermolecular inter-actions during protein assembly.

Materials and Methods

Protein synthesis and purification

The foldon domain was synthesized chemically withan ABIMED economy peptide synthesizer EPS 221(Abimed, Germany) using standard F-moc chemistry.Pre-coupled resin was purchased from Novabiochem,Switzerland. Amino acids were from AlexisBiochemicals, USA or from Iris Biotech, Marktredwitz,Germany. Solvents and other chemicals were from Fluka(Buchs, Switzerland). The protein was purified usingHPLC with a C-8 reverse-phase preparative column(Hibar, LiChrosorbw100 RP-8 from Merck, Darmstadt,Germany). 13C,15N-labeled foldon for NMR spectroscopywas expressed in E. coli BL21(DE3) cells grown in13C,15N-enriched M9 minimal medium. Protein puritywas confirmed by nanospray mass spectrometry andanalytical HPLC. Expression and purification of thefoldon domain in E. coli was performed as described.4

NMR spectroscopy

NMR samples ([U-15N]foldon 95% H2O/5% 2H2O;[U-13C,15N]foldon 95% H2O/5% 2H2O; [U-13C,15N]foldon100% 2H2O) of 300 ml volume (Shigemi NMR microtubes)were prepared as 0.3 mM protein solutions at pH 7.1 in5 mM sodium phosphate, 0.02% (w/v) NaN3 withoutany further addition of salt. Residual alignment of[U-13C,15N]foldon for the measurement of residualdipolar couplings was introduced by lamellar ether/n-hexanol phases.28 A set of standard triple and double-resonance assignment, quantitative J-coupling, nuclearOverhauser effect (NOE) spectroscopy (NOESY) and 15Nrelaxation experiments similar to those described29 were

performed on a Bruker DRX 600 spectrometer at 25 8C.Standard data processing and analysis was carried outas described.29

Structure calculation

Experimental NOE distance, torsion angle andresidual dipolar restraints derived from the NMR dataare listed in Table 1. Structure calculations were carriedout with a simulated annealing protocol using theprogram CNS.30 The structural statistics for the best 40structures are given in Table 1.

Protein Data Bank accession code

The structural statistics have been deposited in theBrookhaven Protein Data Bank with PDB accession code1RFO.

Denaturant-induced equilibrium transitions

Denaturant-induced equilibrium transitions wererecorded in an AMINCO Bowman series 2 spectrofluori-meter (SLM Aminco, USA) and with an Aviv 62ADSspectropolarimeter (Aviv, USA). All transitions weremeasured at 10 mM sodium phosphate (pH 7.1 at20 8C). For fluorescence measurements at concentrationsof total monomer ð½M�0Þ of 5 mM and 30 mM, theexcitation wavelengths were 278 nm and 298 nm,respectively, at 2 nm bandwidth. Emission was recordedat 320 nm (2 nm bandwidth). CD measurements wereperformed at 228 nm with 0.5 cm (5 mM) or 0.1 cm(30 mM) path-lengths.

Equilibrium transition curves were analyzedassuming a two-state transition from native trimer (N)to unfolded monomer (U):

N O 3U ð5Þ

The resulting equilibrium constant is given by:

Keq ¼½U�3

½N�ð6Þ

and the total monomer concentration ð½M�0Þ can beexpressed as:

½M�0 ¼ ½U� þ 3½N� ð7Þ

The fractions of monomers in the unfolded state ðfUÞ andin the native state ðfNÞ are given by:

fU ¼½U�

½U� þ 3½N�¼

½U�

½M�0ð8aÞ

fN ¼3½N�

½U� þ 3½N�¼

3½N�

½M�0¼ 1 2 fU ð8bÞ

Thus, equation (6) becomes:

Keq ¼½U�3

½N�¼

3f 3u ½M�20

1 2 fuð9Þ

To fit the equilibrium transitions to equation (9) weexpressed the fraction of unfolded protein at a givendenaturant concentration ðxÞ as:31

fUðxÞ ¼SNðxÞ2 SðxÞ

ðSNðxÞ2 SðxÞÞ þ ðSðxÞ2 SUðxÞÞð10Þ

where SðxÞ corresponds to the measured signal at thegiven denaturant concentration, x. SN and SU representthe spectroscopic signals of the native and unfoldedstate, respectively, obtained from linear extrapolation ofthe baselines according to:32

SNðxÞ ¼ SNðH2OÞ þ mNx ð11aÞ


SUðxÞ ¼ SUðH2OÞ þ mUx ð11bÞ

Using:

DG0 ¼ 2RT ln Keq ð12Þ

and a linear denaturant-dependence of DG0 :32,33

DG0ðxÞ ¼ DG0ðH2OÞ þ mx ð13Þ

we can obtain the free energy for unfolding in waterDG0ðH2OÞ from fitting the transition curve to equations(9)–(13) in a single step, similar to the proceduredescribed by Santoro & Bolen for a two-state transitionof monomeric proteins.34 This yields the followingequation for a two-state unfolding transition from anative trimer to unfolded monomers:

SðxÞ ¼ SNðxÞ2SNðxÞ2 SUðxÞ

3½M�0

�

0BB@

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffie2DG0ðH2OÞþmx

RT

�9

2½M�0 þ

ffiffiffiffiD

p�

3

s

þ

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffie2DG0ðH2OÞþmx

RT

�9

2½M�0 2

ffiffiffiffiD

p�

3

s 1CCA ð14aÞ

with:

D ¼9

2

� �2

½M�20 þ e2DG0ðH2OÞþmx

RT ð14bÞ

Equations (14a) and (14b) were used to fit the individualtransition curves monitored by fluorescence or CD andto globally fit all data obtained by measuring variousspectroscopic probes at different concentrations of totalmonomer ð½M�0Þ:

Kinetic experiments

All stopped-flow experiments were performed with aSX18.MV stopped-flow instrument from Applied Photo-physics (Leatherhead, UK) equipped with a HamamatsuR1104 photomultiplier tube for single-wavelengthkinetics and a Hamamatsu R6095 photomultiplier tubefor all other measurements. For experiments using acut-off filter ($320 nm), the excitation wavelength was278 nm (2 nm bandwidth). For single-wavelength refold-ing experiments, the excitation wavelength was 280 nm(4.5 nm bandwidth) for final concentrations of proteinup to 10 mM and 295 nm (4.5 nm bandwidth) for higherconcentrations. The emission bandwidth was 12 nm. Forrefolding experiments, the protein was allowed to unfoldfor at least ten hours in 6.38 M GdmCl, 10 mM sodiumphosphate (pH 7.1) before the measurements. At finalconcentrations of protein above 10 mM monomer, theprotein was unfolded at the same concentration ofdenaturant at pH 2. Unless stated otherwise, allmeasurements were performed in 10 mM sodiumphosphate (pH 7.1 at 20 8C). Refolding was initiated by11-fold dilution to a final concentration of GdmCl of0.58 M. The fluorescence intensity of the unfolded stateunder refolding conditions (Figure 3) was determinedby extrapolating the fluorescence signal of the unfoldedfoldon domain (measured in the stopped-flow instru-ment) to 0.58 M GdmCl. The effect of human cyclophilin18 on the slow-refolding reaction was measured bymanual mixing fluorescence measurements in 10 mMsodium cacodylate (pH 7.1) at ½M�0 ¼ 5 mM foldondomain in the absence and in the presence of 3.3 mM

cyclophilin. The use of sodium cacodylate instead ofsodium phosphate had no effect on the folding kinetics.

Stopped-flow interrupted refolding experiments wereused to monitor the formation of native moleculesduring refolding in 0.58 M GdmCl, 20 mM sodium phos-phate (pH 7.0) at 20.0 8C. Completely unfolded foldon (in3.4 M GdmCl, 20 mM sodium phosphate (pH 1.7),½M�0 ¼ 60 mM) was diluted sixfold into final conditionsof 10 mM protein, 0.58 M GdmCl, 20 mM sodium phos-phate (pH 7.0) to initiate refolding. After various timesðtiÞ; refolding was interrupted by transferring thesolution into final conditions of 6.7 M GdmCl, 20 mMsodium phosphate (pH 7.0), final protein concentration1.7 mM. Native foldon unfolds with double-exponentialkinetics under these conditions (S.Gu. & T.K., unpub-lished results) with t1 ¼ 7:8ð^0:5Þ seconds (80% ampli-tude) and t2 ¼ 0:24ð^0:02Þ seconds (20% amplitude).The relative amplitudes of the two reactions is indepen-dent of the refolding time. The amplitude of the major,slow-unfolding reaction (t1 ¼ 7:8ð^0:5Þ seconds) wasused as a measure for the amount of native protein thatwas present after the time ti; when refolding was inter-rupted. The observed unfolding amplitudes after varioustimes of refolding were normalized against the ampli-tude of completely refolded foldon to yield the fractionof native molecules that were present after ti:

Data fitting and simulations

Data evaluation was carried out using the programsProFit (Quantumsoft, Zurich, Switzerland) and Matlab(The MathWorks, Natick, MA, USA). Interrupted refold-ing experiments at ½M�0 ¼ 10 mM and ten direct fluor-escence-detected refolding traces with ½M�0 rangingfrom 0.5 mM to 102 mM were analyzed globally by non-linear, least-squares curve fitting. The experimental datawere fit to the numerical solution of the kinetic schemedepicted in equation (4). Rate constants and relative sig-nal amplitudes of the different kinetic species were fittedas global parameters. The equilibrium constant betweennative and unfolded protein determined by a global fitof the equilibrium unfolding transitions measured byfluorescence and CD at various concentrations of protein(Figure 2) was used as an additional constraint for the fit.To ensure that the fit converged to the global minimum,it was repeated 60 times with randomly chosen startingvalues for the fitting parameters.

References

1. Tao, Y., Strelkov, S. V., Mesyanzhinov, V. V. &Rossmann, M. G. (1997). Structure of bacteriophageT4 fibritin: a segmented coiled coil and the role ofthe C-terminal domain. Structure, 5, 789–798.

2. Letarov, A. V., Londer, Y. Y., Boudko, S. P. &Mesyanzhinov, V. V. (1999). The carboxy-terminaldomain initiates trimerization of bacteriophage T4fibritin. Biochemistry (Moscow), 64, 817–823.

3. Boudko, S. P., Londer, Y. Y., Letarov, A. V., Sernova,N. V., Engel, J. & Mesyanzhinov, V. V. (2002). Domainorganization, folding and stability of bacteriophageT4 fibritin, a segmented coiled-coil protein. Eur.J. Biochem. 269, 833–841.

4. Frank, S., Kammerer, R. A., Mechling, D., Schulthess,T., Landwehr, R. J. B., Guo, Y. et al. (2001). Stabilizationof short collagen-like triple helices by protein engineer-ing. J. Mol. Biol. 308, 1081–1089.

5. Seckler, R. (2000). Assembly of multi-subunit


structures. In Mechanisms of Protein Folding (Pain, R.,ed.), pp. 279–308, Oxford University Press, Oxford.

6. Stetefeld, J., Frank, S., Jenny, M., Schulthess, T.,Kammerer, R. A., Boudko, S. P. et al. (2003). Collagenstabilization at atomic level: crystal structure ofdesigned (GlyProPro)10-foldon. Structure, 11, 339–346.

7. Dosset, P., Hus, J. C., Blackledge, M. & Marion, D.(2000). Efficient analysis of macromolecularrotational diffusion from heteronuclear relaxationdata. J. Biomol. NMR, 16, 23–28.

8. Myers, J. K., Pace, C. N. & Scholtz, J. M. (1995).Denaturant m values and heat capacity changes:relation to changes in accessible surface areas ofprotein unfolding. Protein Sci. 4, 2138–2148.

9. Milla, M. E. & Sauer, R. T. (1994). P22 arc repressor:folding kinetics of a single-domain, dimeric protein.Biochemistry, 33, 1125–1133.

10. Fischer, G., Wittmann, L. B., Lang, K., Kiefhaber, T. &Schmid, F. X. (1989). Cyclophilin and peptidyl-prolylcis– trans isomerase are probably identical proteins.Nature, 337, 476–478.

11. Schmid, F. X. (1983). Mechanism of folding of ribo-nuclease A. Slow refolding is a sequential reaction viastructural intermediates. Biochemistry, 22, 4690–4696.

12. Kiefhaber, T. (1995). Protein folding kinetics. InMethods in Molecular Biology (Shirley, B. A., ed.), vol.40, pp. 313–341, Humana Press, Totowa, NJ.

13. Bieri, O. & Kiefhaber, T. (2000). Kinetic models inprotein folding. In Protein Folding: Frontiers in MolecularBiology (Pain, R., ed.), 2nd edit., pp. 34–64, OxfordUniversity Press, Oxford.

14. Moore, J. W. & Pearson, R. G. (1981). Kinetics andMechanisms, Wiley, New York.

15. Gutfreund, H. (1995). Kinetics for the Life Sciences,Cambridge University Press, Cambridge.

16. Jackson, S. E. (1998). How do small single-domainproteins fold? Fold. Des. 3, R81–R91.

17. Zitzewitz, J. A., Bilsel, O., Luo, J., Jones, B. E. &Matthews, C. R. (1995). Probing the folding mechanismof a leucine zipper peptide by stopped-flow circulardichroism spectroscopy. Biochemistry, 34, 12812–12819.

18. Jaenicke, R. (1987). Folding and association ofproteins. Prog. Biophys. Mol. Biol. 49, 117–237.

19. Wendt, H., Berger, C., Baici, A., Thomas, R. M. &Bosshard, H. R. (1995). Kinetics of folding of leucinezipper domains. Biochemistry, 34, 4097–4107.

20. Waldburger, C. D., Jonsson, T. & Sauer, R. T. (1996).Barriers to protein folding: formation of buriedpolar interactions is a slow step in acquisition ofstructure. Proc. Natl Acad. Sci. USA, 93, 2629–2634.

21. Gloss, L. M. & Matthews, C. R. (1998). Mechanism offolding of the dimeric core domain of Escherichia coliTrp repressor: a nearly diffusion-limited reactionleads to the formation of an on-pathway inter-mediate. Biochemistry, 37, 15990–15999.

22. Reimer, U., Scherer, G., Drewello, M., Kruber, S.,Schutkowski, M. & Fischer, G. (1998). Side-chaineffects on peptidyl-prolyl cis/trans isomerization.J. Mol. Biol. 279, 449–460.

23. Bachmann, A. & Kiefhaber, T. (2001). Apparent two-

state tendamistat folding is a sequential processalong a defined route. J. Mol. Biol. 306, 375–386.

24. Sanchez, I. E. & Kiefhaber, T. (2003). Evidence forsequential barriers and obligatory intermediates inapparent two-state protein folding. J. Mol. Biol. 325,367–376.

25. Wagner, C. & Kiefhaber, T. (1999). Intermediates canaccelerate protein folding. Proc. Natl Acad. Sci. USA,96, 6716–6721.

26. Zitzewitz, J. A., Ibarra-Molero, B., Fishel, D. R., Terry,K. L. & Matthews, C. R. (2000). Preformed secondarystructure drives the association reaction of GCN4-p1, amodel coiled-coil system. J. Mol. Biol. 296, 1105–1116.

27. Boudko, S. P., Frank, S., Kammerer, R. A., Stetefeld,J., Schulthess, T., Landwehr, R. et al. (2002). Nuclea-tion and propagation of the collagen triple helix insingle-chain and trimerized peptides: transitionfrom third to first order kinetics. J. Mol. Biol. 317,459–470.

28. Ruckert, M. & Otting, G. (2000). Alignment of bio-logical macromolecules in novel non-ionic liquidcrystalline media for NMR experiments. J. Am.Chem. Soc. 122, 7793–7797.

29. Kahmann, J. D., Sass, H. J., Allan, M. G., Seto, H.,Thompson, C. J. & Grzesiek, S. (2003). Structuralbasis for antibiotic recognition by the TipA class ofmultidrug-resistance transcriptional regulators.EMBO J. 22, 1824–1834.

30. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano,W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998).Crystallography & NMR system: a new softwaresuite for macromolecular structure determination.Acta Crystallog. sect. D, 54, 905–921.

31. Tanford, C. (1968). Protein denaturation. Part B. Thetransition from native to denatured state. Advan.Protein Chem. 23, 218–282.

32. Pace, C. N. (1986). Determination and analysis ofurea and guanidine hydrochloride denaturationcurves. Methods Enzymol. 131, 266–280.

33. Greene, R. F. J. & Pace, C. N. (1974). Urea and guani-dine–hydrochloride denaturation of ribonuclease,lysozyme, alpha-chyomtrypsinn and beta-lacto-globulin. J. Biol. Chem. 249, 5388–5393.

34. Santoro, M. M. & Bolen, D. W. (1988). Unfolding freeenergy changes determined by the linear extra-polation method. 1. Unfolding of phenylmethane-sulfonyl alpha-chymotrypsin using differentdenaturants. Biochemistry, 27, 8063–8068.

35. Cornilescu, G., Marquardt, J. L., Ottiger, M. & Bax, A.(1998). Validation of protein structure from aniso-tropic carbonyl chemical shifts in a dilute liquidcrystalline phase. J. Am. Chem. Soc., 6836–6837.

36. Laskowski, R. A., Rullmannn, J. A., MacArthur,M. W., Kaptein, R. & Thornton, J. M. (1996). AQUAand PROCHECK-NMR programs for checking thequality of protein structures solved by NMR.J. Biomol. NMR, 8, 477–486.

37. Koradi, R., Billeter, M. & Wuthrich, K. (1996).MOLMOL: a program for display and analysis ofmacromolecular structures. J. Mol. Graph. 14, 51–55.

Edited by F. Schmid

(Received 24 November 2003; received in revised form 30 January 2004; accepted 5 February 2004)


Very Fast Folding and Association of a Trimerization Domain from Bacteriophage T4 Fibritin

Documents