a 1774 (2007) 323–333www.elsevier.com/locate/bbapap

Biochimica et Biophysica Act

Aggregation as the basis for complex behaviour of cutinasein different denaturants

Daniel E. Otzen a,⁎, Lise Giehm a, Ricardo P. Baptista c, Søren R. Kristensen b,Eduardo P. Melo c,d, Shona Pedersen a,b

a Department of Life Sciences, Aalborg University, Sohngaardsholmsvej 49, Aalborg DK- 9000, Denmarkb Department of Clinical Biochemistry, Aalborg University Hospital, Hobrovej 18-22, DK-9100 Aalborg, Denmark

c Centro de Engenharia Biológica e Química, Instituto Superior Técnico, Av. Rovisco Pais, 1049-001 Lisboa, Portugald Centro de Biomedicina Molecular e Estrutural, Universidade do Algarve, Campus de Gambelas, 8005-139 Faro, Portugal

Received 4 October 2006; received in revised form 12 November 2006; accepted 27 November 2006Available online 30 November 2006

Abstract

We have previously described the complexity of the folding of the lipolytic enzyme cutinase from F. solani pisi in guanidinium chloride. Herewe extend the refolding analysis by refolding from the pH-denatured state and analyze the folding behaviour in the presence of the weakerdenaturant urea and the stronger denaturant guanidinium thiocyanate. In urea there is excellent consistency between equilibrium and kinetic data,and the intermediate accumulating at low denaturant concentrations is off-pathway. However, in GdmCl, refolding rates, and consequently thestability of the native state, vary significantly depending on whether refolding takes place from the pH- or GdmCl-denatured state, possibly due totransient formation of aggregates during folding from the GdmCl-denatured state. In GdmSCN, stability is reduced by several kcal/mol withsignificant aggregation in the unfolding transition region. The basis for the large variation in folding behaviour may be the denaturants' differentialability to support formation of exposed hydrophobic regions and consequent changes in aggregative properties during refolding.© 2006 Elsevier B.V. All rights reserved.

Keywords: Cutinase; Folding scheme; Denaturant; Guanidinium thiocyanate; Aggregation

1. Introduction

Since Tanford's classic studies in the 1960s and 1970s [1–3],chemical denaturants such as urea and guanidinium chloride(GdmCl) have been very popular as tools to elucidate proteinfolding mechanisms. The key to their use lies in the empiricallyobserved linear relationship between free equilibrium andactivation energies of folding/unfolding versus denaturantconcentration. Such linear relationships make it possible notonly to extrapolate free energy values to zero molar denaturant,where unfolding cannot be measured directly, but also topropose kinetic folding models based on the variation of

Abbreviations: ANS, 8-anilino-1-naphthalenesulphonic acid; DLS,Dynamiclight scattering⁎ Corresponding author. Tel.: +45 96358525; fax: +45 98141808.E-mail address: dao@bio.aau.dk (D.E. Otzen).

1570-9639/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.bbapap.2006.11.012

observed rate constants with denaturant concentration [4–7].On a molar basis, GdmCl is approximately twice as potent asurea but also introduces somewhat more complex behaviour,due to its charge. This can apparently lead to stabilization ofstates which go unobserved in urea. Thus, GdmCl has beenreported to induce the accumulation of partially folded states ofprotein disulfide isomerase [8] and carbonic anhydrase [9],which are not seen when unfolding is carried out in urea;however, when chloride ions in the form of NaCl are added tourea, an unfolding intermediate is in fact observed. An evenmore potent denaturant is guanidinium thiocyanate which hasbeen used to denature e.g. exceptionally stable circularizedvariants of chymotrypsin inhibitor 2 [10] as well as compareeffects of different denaturants [11,12].

Stable intermediates are typically inferred from equilibriumexperiments when more than one spectroscopic transition isobserved, or when transitions recorded by different techniques

324 D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

do not coincide [13]. Although folding intermediates may bemissed in equilibrium experiments because they do notaccumulate as stable entities, they can be detected as transientlypopulated states through stopped flow techniques [14].Typically the log of the measured rate constant kobs is plottedas a function of denaturant concentration in a so-called chevronplot. Proteins folding according to a simple two-state system(D↔N) show a V-shaped chevron plot, in which the log ofrefolding rate constant declines linearly with denaturantconcentration and vice versa for the unfolding rate constant.In contrast, a tell-tale signature of an intermediate is the “roll-over”, in which the log of kobs starts to deviate negatively fromlinearity at low denaturant concentrations, typically reaching aplateau level. This is an indication that the ground state fromwhich folding occurs switches from the unfolded state to theintermediate state [4,6], leading to a slower refolding rate.However, such a roll-over can occur both through the formationof a productive (“on-pathway”) as well as an unproductive(“off-pathway”) intermediate (see Schemes 1 and 3 in Materialsand methods) [4]. If the roll-over plateau flips over and assumesa positive slope, this is added evidence for an off-pathwayintermediate which will have to unfold back to the denaturedstate to allow folding to occur [15]. These intermediates can alsobe regarded as physiological versions of the denatured state, i.e.the protein's response to the change in solvent conditions whenit is rapidly diluted to lower concentrations of denaturant [6,16–18]. Due to their important but equivocal roles in folding,partially folded states have been the subject of intenseinvestigation for many years.

The lipolytic enzyme cutinase is a 197-residue protein with apI between 7.6 and 8.0 [19], which belongs to the superfamilyof a/b hydrolase folds [20]. It shows complex folding andunfolding behaviour which together with its use in the food andchemical industry has attracted attention from several researchgroups. At pH 4.5, near-UVabsorbance difference spectroscopyand fluorescence spectroscopy differ by ca. 0.1 M in themidpoints of their transitions in GdmCl, suggesting thepresence of an intermediate state around the transition regionof unfolding (1 M GdmCl) [21], which seems to have highaffinity for the hydrophobic probe ANS. In addition, anunfolding intermediate also appears to accumulate transientlyduring unfolding at denaturant concentrations more than 1.5 Mabove the transition midpoint [22]. In the same study weobserved a roll-over during refolding at low denaturantconcentrations. The roll-over was very pronounced and seemedto be on the cusp of changing the sign of its slope, but as wewere unable to follow unfolding all the way down to zero molardenaturant, we were not able to determine more unequivocallywhether an off-pathway intermediate is formed. This subject isaddressed in the present study, where we have extended therefolding concentration regime down to zero molar denaturantby measuring refolding from the pH-denatured state, obviatingthe need to dilute out denaturant to start refolding. We have alsoincluded the use of stabilizing salts to distinguish between theaccumulation of on- and off-pathway intermediates. In addition,we compare the folding of cutinase in different denaturants(urea, GdmCl and guanidinium thiocyanate). This has allowed

us to distinguish the different states that are populated duringthe folding process as well as revealing how sensitive thefolding mechanism of a protein can be to the solvent.

2. Materials and methods

2.1. Materials

Ultrapure Guanidinium chloride (GdmCl) was from Life Technologies Inc.(Rockville, MA), guanidinium thiocyanate (GdmSCN) grade SigmaUltra(>99%) and all other chemicals were from Sigma-Aldrich (St. Louis, MO).All chemicals were of analytical or biological grade. An ANS stock solution wasprepared by dissolving 30 mg ANS in 1 ml DMSO (dimethylsulfoxide) andadding 1 ml 25 mM acetate buffer pH 4.5 to give a final ANS stock solution of50 mM. 197-residue cutinase from Fusarium solani pisi, cloned into the vectorpFCEX1, was expressed in E. coli BL21 (DE3), purified to 99% purity, dialyzedagainst water and lyophilized [23]. The protein concentration was determinedusing ε280=13,500 M−1 cm−1.

2.2. Equilibrium denaturation experiments

All experiments were performed in 25 mM acetate buffer, pH 4.5 at 25 °C.This pH-value is far from the isoelectric point (pI=8.0) and thus reduces thepotential for aggregation during folding or unfolding [24]. 8 M urea stocksolutions were prepared fresh on a daily basis.

Equilibrium fluorescence studies at different denaturant concentrationswere carried out using 5 μM cutinase with excitation of the single Trp at298 nm and emission at 350 nm on an RTC2000 spectrometer (PhotonTechnology International, Lawrenceville, NJ) using slits of 5 nm for bothexcitation and emission. 298 nm was used to avoid contributions from tyrosineresidues. When ANS was included, it was present at 40 μM; excitation was at350 nm and emission was recorded at 475 nm. At each denaturantconcentration, five scans were averaged to yield the final spectrum. Nosignificant amount of photo-bleaching was observed due to the low slit valuesand the low absorbance at 298 nm. Equilibrium circular dichroism studieswere performed on a Jasco J-715 spectropolarimeter (Jasco Spectroscopic Co.,Hachioji City, Japan) with a Jasco PTC-348W1 temperature control unit. Sixscans were averaged to yield the final spectrum. Ellipticity values at 223 nmwere used, since ellipticity at this wavelength showed the greatest changeupon denaturation. Absorption by urea prevented us from going below215 nm.

The ellipticity and fluorescence values are analyzed using the linearextrapolation method [3,7], which is based on the relationship:

logKureaD�N ¼ logKwater

D�N þ mD�N½urea� ð1Þ

where KD–N=[D]/[N] is the equilibrium constant of unfolding (in water or ata specified urea concentration) and mD–N is a parameter which reflects thedegree of surface area buried in the native state, N, relative to the denaturedstate, D [3]. The larger the m-value, the greater the difference between N andD in exposed surface area. The advantage of this standard linear extrapolationapproach is that it allows us to compare equilibrium and kinetic data, andthus test the consistency of kinetic interpretations (see below). Assuming alinear dependence of the pre- and post-transition baselines on ureaconcentration [7,25], we obtain the following equation for urea denaturationcurves:

Yobs ¼ aN þ bN½urea� þ ðaD þ bD½urea�Þ*10mD�Nð½urea��½urea50% �Þ

1þ 10mD�Nð½urea��½urea50%�Þ ð2Þ

where Yobs is the observed signal, αN and αD denote the signal at 0 M ureafor the native and denatured states respectively, βN and βD are the slopes ofthe baselines of the native and denatured states, respectively and [urea50%] isthe urea concentration where 50% of the protein is denatured. As fluorescencesignal we use the ratio of the emission at 350 and 310 nm, which increasesby a factor of 6–8 upon unfolding. Non-linear least-squares regressionanalysis was carried out with the program Kaleidagraph, version 3.5 (Synergy

Scheme 2. On-pathway unfolding intermediate I and off-pathway folding

Scheme 3. Three-state folding with on-pathway intermediate I according toScheme 3 [4] (GdmSCN).

325D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

Software, Reading, PA). From the fit to Eq. (2), the stability of cutinase in theabsence of denaturant is calculated as follows [7,25]:

DGD�N ¼ �RT lnð10Þ*mD�N*½urea50%� ð3Þ

2.3. Stopped-flow folding and unfolding experiments

Stopped-flow fluorescence measurements were performed on an SX18MVstopped-flow apparatus (Applied Photophysics, Leatherhead, UK). The proteinsolution and the denaturant solutions were mixed at a 1:10 ratio to give a finalprotein concentration of 2 μM and the desired denaturant concentrations.Unfolding was induced by mixing the cutinase in buffer with the appropriateconcentration of denaturants. Refolding in different denaturants was induced byunfolding the cutinase in 1.65 M GdmSCN, 1.5 M GdmCl or 3.3 M urea at pH4.5 and mixing with the appropriate denaturant concentrations. For urea andGdmCl, refolding was also initiated from the pH-denatured state (pH 1.5 in HCl)and refolded by mixing 1:10 with denaturant at pH 4.5 (the final pH was within0.1 pH unit of 4.5). At each denaturant concentration, 4–7 experiments werepreformed sequentially and the kinetics traces averaged before fitting the data.The temperature was held constant at 25 °C with a circulating water bath.Cutinase was excited at 280 nm and the fluorescence was monitored using a320 nm cut-off filter.

In experiments involving ANS, this probe was present in the denaturantsolution at 44μMto give a final concentration of 40μMaftermixing. Fluorescencewas monitored using a 435 nm cut-off filter with excitation at 385 nm.

The refolding kinetic traces of cutinase in all three denaturants were bestfitted using a double exponential equation:

Yobs ¼ A1de�k1dt þ A2de

�k2d t þ c ð4Þwhere Yobs is the observed fluorescence signal, A1 and A2 are the amplitudes ofthe two phases, k1 and k2 are the rate constants of the two phases, and c isbaseline value. The fast phase corresponds to a direct folding phase while theslow phase has been attributed to cis–trans peptidyl-proline isomerization [26].Unfolding proceeds via one phase and can accordingly be fitted to a singleexponential. When refolding or unfolding in the presence of 0.15–0.65 MGdmSCN, a pronounced aggregation signal followed the initial refolding orunfolding signal (see Results) and was excluded from the kinetic analysis.

2.3.1. Analysis of kinetic dataAnalogous to stability parameters, the logarithms of the microscopic rate

constant for folding and unfolding (kf and ku) as well as the equilibriumconstants for the formation of on- and off-pathway intermediates are assumed todepend linearly on denaturant concentration [3]. We operate with the followingfolding schemes:

Scheme 1. Folding with an off-pathway folding intermediate C and a movingtransition state with curvature in the unfolding limb [27] (urea).

where KC=[C]/[D]. Here the observed rate constant kobs=kf*fD+ku where fD=

[D]/([D]+[C]), leading to the equation:

logkobs ¼ log10ðlogk

waterf þmf ½den�Þ

1þ 10ðlogKwaterC þmC ½den�Þ þ 10ðlogk

wateru þmu ½den�þm∗

u ½den�2Þ !

ð9Þ

where mu* is a constant which describes the curvature in the plot of log ku versus

urea concentration and the other m-values describe the linear dependence of thelog of the preceding rate or equilibrium constant on [denaturant].

where KI = [I]/[N] and KD–N=KIku/kf.

logkobs ¼ log10ðlogk

waterf þmf ½den�Þ

1þ 10ðlogKwaterC þmC ½den�Þ þ

10ðlogkwateru þmu ½den�Þ

1þ 10�ðlogKwaterI þmi ½den�Þ

!ð10Þ

intermediate C according to Scheme 2 (urea, GdmCl).

logkobs ¼ log 10ðlogkwaterf þmf ½den�Þ þ 10ðlogk

wateru þmu ½den�Þ

1þ 10�ðlogKwaterI þmI ½den�Þ

!ð9Þ

Here kobs=kf+ku*fI where fI = [I]/([I]+ [N]), KI = [I]/[N] and KD–N=KIku/kf.

3. Results

3.1. Equilibrium denaturation of cutinase in differentdenaturants

It is important to be able to compare stability data fromequilibrium and kinetic studies in order to make sure that ourkinetic analysis provides a complete description of the foldingpathway [6]. Therefore, the first step in the analysis of thebehaviour of cutinase in different denaturants is to carry outequilibrium denaturation in the different denaturants (Fig. 1;data summarized in Table 1). As expected, cutinase denatures ata higher denaturant concentration in the weak denaturant urea(midpoint 2.36±0.02 M) than in GdmCl (midpoint 1.09±0.02 M). When unfolding of urea is followed in the presence ofANS, ANS fluorescence starts out at a relatively high level, dueto its ability to bind to the native state [24]. It peaks just aroundthe midpoint of denaturation as monitored by Trp fluorescence,after which it subsides to a level slightly higher than the startingpoint (Fig. 1A). In GdmCl (Fig. 1B), the ANS peak alsocoincides with the midpoint of Trp-monitored denaturation[28]. In GdmSCN, the Trp fluorescence data suggest adenaturation midpoint of 0.82±0.03 M with a plateau reachedaround 1.5 M GdmSCN, whereas when unfolding is monitoredby ANS fluorescence, there is a top around 0.2 M GdmSCN(Fig. 1C).

Table 1Equilibrium denaturation data for cutinase a

Denaturant [Denaturant]50% b mD–N (M−1) b ΔGD–N (kcal/mol) b

Urea 2.34±0.02 2.40±0.18 7.69±0.57Urea 1 M NaCl 4.00±0.03 1.72±0.07 9.34±0.39GdmCl 1.18±0.07 4.93±0.15 7.90±0.20GdmSCN 0.82±0.03 3.07±0.47 3.42±0.54a All experiments were performed in 25 mM acetate buffer, pH 4.5 at 25 °C.b Parameters obtained from fitting equilibrium data to Eq. (2).

Fig. 2. (A) Far-UV CD spectra of cutinase at pH 4.5 and pH 1.5 (stippled line).(B) Denaturation of cutinase at low pH followed by the ellipticity at 222 nm.

Fig. 1. Equilibrium denaturation of cutinase in (A) urea, (B) GdmCl and (C)GdmSCN followed by changes in Trp fluorescence. In (A) and (C), ANSfluorescence data are also included.

326 D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

3.2. Kinetic analysis

In order to be able to carry out a complete analysis of thefolding behaviour of cutinase in denaturants and monitor the

formation of potential off-pathway intermediates, it wasnecessary to be able to refold cutinase at zero molar denaturant.The simplest approach is to denature the protein at low pH andbring it back to native conditions by mixing with buffer at therequired pH [29]. We first need to establish the pH range wherecutinase is denatured. There is a clear transition between foldedand unfolded cutinase as the pH is reduced from 3 to 2 (Fig. 2).Far-UV CD spectroscopy reveals that cutinase has clear α-helical structure at pH 3 and above, whereas at pH 2 and below,the protein has unfolded to a random coil (Fig. 2). Cutinase can

ig. 3. Chevron plots for folding and unfolding of cutinase in differentenaturants. (A) Urea in the absence (●) and presence (○) of 1 M NaCl. Dataitted to Scheme 2. Arrows mark the transition from refolding from the pH-enatured to the urea-denatured state. (B) GdmCl. Refolding occurs from thedmCl-denatured state (●) and the pH-denatured state (○). Data fitted tocheme 2 using only pH-denatured state. (C) GdmSCN. Data fitted to Scheme 3.

327D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

be unfolded by reducing the pH to 1.5; transferring the proteinback to pH 4.5 completely restores the native structure (datanot shown). Thus we can measure refolding rates at any givendenaturant concentration by mixing the pH-denatured protein1:10 with buffer pH 4.5 and the desired denaturant concentra-tion. We have therefore measured refolding and unfolding ratesas a function of the concentration of urea, GdmCl andGdmSCN over a broad concentration range. In the followingwe describe and analyze the chevron plots that result from ourdata.

3.2.1. UreaIn urea, there is a smooth continuity between the refolding

rates obtained from the pH-denatured state and the urea-denatured state (Fig. 3A). There is a roll-over below ca. 1.5 Murea which however does not lead to a complete change inslope; the unfolding limb shows a mild degree of curvaturewhich can be fitted as a polynomial (often interpreted as movingtransition state [27,30]) or an unfolding intermediate [31] withequal success. In order to expand the region in which the roll-over occurs and thus obtain a more reliable fitting to our foldingscheme, we also carried out refolding and unfolding experi-ments in urea in the presence of 1 M NaCl (Fig. 3A). In thiscase, the roll-over occurred below 2.5 M urea. In addition, thecurvature in the unfolding limb became more marked and wasbest fitted as an unfolding intermediate. Data are summarized inTable 2.

3.2.2. GdmClThe situation is significantly different for folding and

unfolding in GdmCl (Fig. 3B). There is a clear differencebetween refolding rates when refolding takes place from the pH-denatured state (kobs≈0.3 s−1) as opposed to refolding from theGdmCl-unfolded state (kobs≈1.2 s−1). The slope of therefolding plot based on the GdmCl-unfolded state is alsosignificantly steeper than that from the pH-denatured state. Thedata summarized in Table 2 indicate that the kinetic data derivedfrom the pH-denatured state lead to a much better correspon-dence with the value of ΔGD–N from urea experiments than thedata derived from the GdmCl-denatured state. Similarly, thekinetic mD–N-values correspond much better to the equilibriummD–N-values when data from the pH-denatured state areemployed.

3.2.3. GdmSCNRefolding and unfolding in GdmSCN is complicated by

significant aggregation (see below). However, it is still possibleto obtain refolding and unfolding rates over the interval 0.15–2.5 M GdmSCN (Fig. 3C). GdmSCN is so destabilizing underkinetic conditions that we can only measure refolding rates up toaround 0.5 M GdmSCN, and an independent refolding limb isconspicuously absent from the chevron plot. In the absence ofan independent folding limb, we can only fit the data assuming arefolding rate which is independent of denaturant concentration.There is a clear roll-over in the unfolding limb, indicative of aconventional unfolding intermediate. However, the data inTable 2 predict an extremely low stability of 3.42 kcal/mol,

FdfdGS

Table 2Summary of kinetic parameters for folding and unfolding of cutinase in different denaturants at pH 4.5 and 25 °C a

Parameter Urea Urea 1 M NaCl GdmCl b GdmCl c GdmSCN

Scheme Scheme 2 Scheme 2 Scheme 3 Scheme 2 Scheme 3

log kf (s−1) 0.095±0.28 2.92±0.26 0.97±0.19 4.90±0.57 0.07±0.07

mf (M−1) −1.58±0.15 −1.28±0.09 −3.09±0.22 −6.93±0.62 – d

log KC 1.42±0.28 2.51±0.25 1.40±0.19 4.54±0.54 – d

mC (M−1) −1.44±0.10 −1.26±0.07 −3.83±0.15 −5.83±0.54 – d

log ku (s−1) −1.73±0.81 −0.08±0.29 1.08±0.04 1.08±0.05 1.63±0.07

mu (M−1) 0.25±0.08 0.13±0.03 0.09±0.01 0.08±0.01 0.49±0.04

log KI −2.82±0.68 −4.01±0.21 −5.12±0.05 −5.11±0.06 −2.64±0.18mI (M

−1) 0.39±0.05 0.60±0.03 1.98±0.03 1.98±0.03 4.73±0.49mD–Nkin (M−1) e 2.22±0.14 2.01±0.10 5.16±0.22 8.99±0.62 5.22±0.50

[den]50% (M) f 2.30±0.20 3.49±0.31 0.99±0.09 1.01±0.09 0.21±0.02ΔGD–N

kinetic (kcal/mol) g 8.00±0.32 9.53±0.53 6.81±0.27 12.13±0.78 1.47±0.28ΔGD–N

equilibrium (kcal/mol) g 7.69±0.566 9.34±0.39 7.90±0.20 [31] 7.90±0.20 [31] 3.42±0.54a All experiments were performed in 25 mM acetate buffer, pH 4.5 at 25 °C.b Refolding data from the pH-denatured state.c Refolding data from the GdmCl-denatured state.d Due to very restricted refolding range, it was not possible to obtain these data.e mD–N

kin =−mf+mu+mI.f Denaturant concentration where [D]=[N].g ΔGD–N=−RT*ln(10)*log KD–N. KD–N for the different schemes is calculated as described in Materials and methods.

328 D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

around half of the value obtained from the urea- and GdmCl-derived data. This may reflect changes in the unfoldingmechanism of cutinase in the presence of GdmSCN, whichwe explore further in the next section.

3.2.4. Aggregation during unfolding and refolding in GdmSCNIn addition to the rapid transients due to proper folding and

unfolding reactions, the fluorescence signals for refolding andunfolding in the region 0.3–0.7 M GdmSCN are dominated by alarge signal increase after a characteristic lag time (Fig. 4A).The amplitudes are typically an order of magnitude larger thanconventional folding signals, indicating that they do not involvefluorescence changes due to conformational rearrangements atthe monomeric level, but presumably derive from lightscattering. Aggregation is also accompanied by changes in thebinding of ANS over time (Fig. 4B). The lag time can beestimated visually from the intersection of straight linesextrapolated from the lag phase period and the growth phase.The light scattering signal can be fitted to a simple exponentialtime profile when the lag phase is cut out from the dataanalysis. The corresponding amplitudes and rate constants peakaround 0.45 M and 0.25 M GdmSCN, respectively (Fig. 4C).We observe a linear correlation between rate constant and lagtime (Fig. 4D), indicating that the shorter the lag phase, thefaster the ensuing aggregation. In turn, such a simplecorrelation suggests that the same aggregation process isoccurring throughout the reaction at this particular proteinconcentration. When aggregation is measured as a function ofprotein concentration for refolding in 0.25 M GdmSCN, there isa linear correlation between the log of the half life ofaggregation t½ (estimated as illustrated in Fig. 4A) and thelog of the protein concentration (Fig. 4E). If aggregation isassumed to involve a pre-equilibrium corresponding to theformation of an oligomeric nucleus, followed by irreversiblepolymerization, then the first derivative in this log–log plot

equals −n*/2, where n* is the nucleus size [32]. The simplestapproach is to approximate the data to a linear fit, which yieldsa slope of −1.18±0.10, suggesting that the nucleus is a dimer.However, the curvature in the data points prompted us to fit thedata to a second-order polynomial which matched the datamore closely. In the resulting curve, the slope variescontinuously from −1.8±0.28 at the lowest concentration(suggesting a 3–4mer nucleus size under these conditions) to−0.49±0.28 at the highest (suggesting a monomeric nucleus).Thus it is possible that the molecular details of the aggregationprocess will vary with protein concentrations. Most of ourstopped-flow experiments were done at 2 μM cutinase, wherethe slope of −1.68±0.28 suggests a trimer.

Dynamic light scattering experiments indicate the formationof large and stable aggregates in the range of GdmSCNconcentrations where the lag phase effect is observed; theseaggregates are much less pronounced in the presence of GdmCland completely absent in the presence of urea (data not shown).We would have liked to analyze the secondary structure of theseaggregates by FTIR; unfortunately the presence of theguanidinium group prohibits the recording of spectra due tostrong spectral overlap in the amide II region. The strong far-UV absorbance of GdmSCN and the contribution from light-scattering artifacts made it unfeasible to interpret CD data of theaggregates in GdmSCN.

3.2.5. Off-pathway stateTo probe the nature of the species accumulating at low

denaturations of GdmCl, we carried out a more detailed analysisof refolding rates from the pH-denatured state below 0.3 MGdmCl. This clearly reveals a negative roll-over (Fig. 5A) thatis consistent with folding via an off-pathway intermediate [15]and is a good indication that folding and unfolding in thepresence of GdmCl follow Scheme 2. However, in the case ofurea, the roll-over does not lead to a positive slope at low

Fig. 4. Aggregation of cutinase in GdmSCN. (A) Time profile of Trp fluorescence for refolding from the pH-denatured state into 0.25–0.75 M GdmSCN. Theintersection of the two stippled lines for the 0.45 M time profile provides an estimate of the lag time. (B) Changes in ANS fluorescence upon refolding from the pH-denatured state into 0.2 and 0.4 M GdmSCN in the presence of 40 μM ANS. (C) Changes in amplitude (●) and rate constant (○) of aggregation upon refolding ofcutinase from the pH-denatured state into 0.15–0.65 M GdmSCN. (D) Correlation between lag time and aggregate rate constant at 0.15–0.65 M GdmSCN. The fittedline follows the equation 1.88*×−0.86 and has a correlation coefficient R=0.996. (E) Aggregation half life as a function of cutinase concentration. The solid line is thebest linear fit with a slope of 1.19±0.10 and a correlation coefficient R=0.98. The stippled line indicates the best fit to a second order polynomium. See text for details.

329D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

denaturant concentrations (Fig. 3A). To provide furtherevidence that an off-pathway intermediate accumulates at lowurea concentrations, we have used the stabilizing salt sodiumsulfate. Inorganic salts such as sodium sulfate favour compactprotein conformations because they are preferentially excludedfrom the protein surface [33]. In addition to stabilizing thenative state, these salts will also induce partially structured

states to accumulate to a larger extent during the refoldingprocess. Accumulation of a folding intermediate will have aprofoundly different effect on folding rates, depending onwhether it is on-pathway or off-pathway. In the off-pathwayscenario, the folding rate is limited by the fraction of the proteinwhich is in the denatured state D (fD=[D]/(D]+[C]), and themore C is stabilized by salts relative to D, the slower the

Fig. 5. (A) Refolding rate constants for refolding from the pH-denatured stateinto 0-0.6 M GdmCl. Data fitted to Scheme 2 excluding unfolding. (B)Refolding from the pH-denatured state into different concentrations of Na2SO4

in the presence of 0 (●), 0.5 (○) and 1.5 M (×) urea. Data fitted as in A.

330 D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

observed folding rate. In contrast, the on-pathway scenario leadsto the folding rate being limited by the fraction fI= [I]/([I]+ [D]),and the more I is stabilized, the faster the rate should be (sincethe transition state for folding is generally stabilized even morethan I). Under conditions where folding occurs directly from thedenatured state and the intermediate does not accumulate,folding rates should be accelerated by salt. We have previouslyshown that intermediates, which accumulate when the ribo-somal protein S6 [15] and the allergen Bet v 1 [34] fold in thepresence of Na2SO4, are off-pathway, since they slow downfolding.

For cutinase, we have measured the effect of salt under threedifferent circumstances: (1) when the intermediate is thedominant species in the ground state before refolding starts(0 M urea, leading to ca. 96% intermediate and 4% denaturedstate according to data in Table 2), (2) when there is a largerproportion of denatured state (0.5 M urea, 16% denatured state),(3) when the unfolded state is the dominant species (1.5 M urea,85% denatured state). There is a uniform decline in the refolding

rate in 0 M urea with increasing salt concentration (Fig. 5B),consistent with the increased stabilization of an off-pathwaystate C whose accumulation impedes folding and which isalready the dominant species in the absence of salt. However, in1.5 M urea, there is initially a rise, followed by a decline; thelatter can be explained by the increasing dominance of C as theground state from which folding occurs as the salt concentrationincreases. Folding in 0.5 M urea shows an in-between situation,where there is only an increase in folding rates over a very shortsalt interval before C accumulates to such an extent that itreduces the refolding rate. Data at 0.5 and 1.5 M urea can befitted to an off-pathway scenario (C↔D→N), where both logkf and log KC depend linearly on [Na2SO4] (with positive m-values). Thus the refolding data in salt clearly support ourprevious suggestion that an off-pathway intermediate accumu-lates at low denaturant concentrations. This intermediate occurswhen refolding from the pH-denatured state, both in urea and inGdmCl.

4. Discussion

4.1. Existence of two distinct intermediate states in cutinasefolding: an unfolding intermediate and an off-pathway foldingintermediate

In this study, we have investigated the kinetic behaviour ofcutinase in the presence of three denaturants with increasingpotency, namely urea, GdmCl and GdmSCN. We will start bydiscussing the urea denaturation data, which provide thesimplest and most internally consistent data. Although theurea chevron plot recorded in the absence of salt fit equally wellto models involving unfolding intermediates or movingtransition states in the unfolding limb, the inclusion of 1 MNaCl removes this ambiguity. The data can only be fittedsatisfactorily to a model with an unfolding intermediate Iaccumulating in the dead-time of the unfolding reaction, in linewith our previous studies using GdmCl [22]. Thus NaClstabilizes I, although the effect is modest; based on data in Table2, in 10 M urea, ΔGI is around 1.1 kcal/mol in the absence ofsalt and only 2 kcal/mol in 1 M NaCl. It seems surprising that astabilizing salt can lead to the accumulation of I at the expenseof N; since N is more compact than I, N should be preferentiallystabilized with increasing salt concentrations due to theexclusion of salt from the protein surface [33]. If I were morecompact than N, mass action would dictate that it should notaccumulate at high denaturant concentrations, in contrast towhat is observed. In contrast, it is entirely to be expected thatsalts can favour a folding intermediate C over the denaturedstate D, due to the increased compaction of C vis-à-vis D. Apossible explanation for the discrepancy is that the presence ofNaCl affects the denaturant activity in a non-linear fashion.Clarke and co-workers have been able to convert denaturantconcentration to activity based on the solubility of freetryptophan at different concentrations [35] and find that thismakes it possible to superimpose chevron plots recorded atdifferent salt concentrations [36]. When a chevron plot isreplotted using activity rather than concentration, the effect is to

331D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

stretch the refolding limb and compress the unfolding limb. Thisactually describes the change in the urea chevron plot caused bythe addition of NaCl. However, the limited experimental datawithin activity–concentration relationships makes it difficult tosolve this issue conclusively.

Having established that cutinase unfolds through an interme-diate, the question arises whether this intermediate also plays arole in the refolding reaction. A priori it is possible for the sameintermediate state to accumulate in the deadtime of refolding andunfolding in specific denaturant concentration ranges. Theintermediate merely has to have a stability in between that of Dand N under both these conditions, and this can easily beachieved using appropriate m- and G-values in simple energydiagrams (based on linear relationships between denaturantconcentration and the stability of the various states). Fortunatelywe are in a position to clarify this, based on the ability to monitorrefolding down to zero molar denaturant which makes it possibleto probe folding under conditions which are favourable to theaccumulation of intermediates. Two lines of evidence suggestthat the intermediate accumulating at low denaturant concentra-tions is in fact an off-pathway state C. Firstly, there is acharacteristic decrease in refolding rates as the GdmClconcentration approaches 0 M. Secondly, the effect of thestabilizing salt sodium sulfate on the refolding rates under variousconditions is consistent with an off-pathway scenario. This makesit very likely that the intermediate states accumulating underrefolding and unfolding conditions are separate species, barringthe unlikely scenario that the unfolding intermediate itselfhappened to be off-pathway. There is little precedence for off-pathway unfolding intermediates, not least because it is difficultto establish their off-pathway character. For example, unfoldingat different salt concentrations (in analogy to Fig. 5B) would leadto a uniform decline in unfolding rate constants whether I wasoff-pathway or not, simply because the ground state from whichunfolding occurs (N or I) in both cases would be stabilized bysalt. Although increasing salt concentrations might disfavour theaccumulation of an off-pathway intermediate (because of itsdecreased compactness compared to N) and thus in principleaccelerate unfolding, this would be counterbalanced by theincreased stabilization of N compared to the transition state ofunfolding and ensuing decrease in unfolding rates.

4.2. The discrepancy between data from different denaturantsmay involve aggregate formation

The complexity of cutinase's folding behaviour prompted usto compare its behaviour in different denaturants in the hope ofobtaining consistent results from a variety of sources.Denaturant potency (on a molar basis) can be estimated throughe.g. the ability to solubilize amino acids [35,37,38] in order toobtain transfer energies from water to denaturant. These transferenergies scale linearly with the accessible amino acid surfacearea and the steeper the slope of the correlation, the stronger thepotency. Alternatively, one can directly measure thermalstability as a function of denaturant concentration [39]. Suchstudies suggest that the potency increases roughly two-foldfrom urea to GdmCl, and an additional two-fold from GdmCl to

GdmSCN. Nevertheless, different denaturants should ideallystill lead to the same value for protein stability in the absence ofdenaturant, as seen for e.g. barnase [25] and CI2 [10]. However,we find that the GdmCl-based data are only consistent with theurea data, as well as internally consistent between kinetic andequilibrium data, if we use refolding rate constants obtainedusing the pH-denatured state, rather than the GdmCl-denaturedstate. This remains puzzling, since the equilibrium data werebased on the GdmCl-denatured state. A number of explanationscan be evoked: Firstly, the unfolded ground state from whichrefolding occurs could be different. The ensemble of denaturedstates is very sensitive to solvent conditions in terms of overallcompactness, presence of residual structure and tendency toaggregate [40–42]. Generally, the GdmCl-denatured stateapproaches the random-coil approximation to a larger extentthan the pH-denatured state [1]. This should not affect the actualrefolding reaction, in which the denatured state is rapidlytransferred to conditions that lead to folding and these groundstate conditions (low denaturant concentration and neutral pH)are the same whether the protein starts from the GdmCl- or pH-denatured state. It is assumed that the different ensembles(GdmCl- or pH-denatured) rapidly equilibrate to the sameground state ensemble from which folding can occur. Only ifthis assumption is invalid can we expect deviations. A priorithis might occur if for example cutinase was trapped in transientaggregates formed in the dead-time of mixing in the “deviant”GdmCl-denatured state. While transient aggregation coulddefinitely not be ruled out at the low GdmCl end-concentrationsused for refolding, this would be expected to lead to adeceleration rather than acceleration of folding [43], unlessthese aggregates had native-like elements of structure that mightlower the activation barrier to folding. The aggregates wouldhave to be very small, perhaps even dimers or trimers (whichappear to accumulate during the initial stages of aggregation inGdmSCN), to avoid unproductive entanglement. Their forma-tion might even be aided by the intrinsic tendency of cutinase toassociate via sticky patches in the active site, whose formationwould be linked to native structure [24]. We do not see a changein refolding rates from the GdmCl-denatured state at increasingcutinase concentrations (spanning the 50-fold concentrationdifference from 0.4 to 20 μM) or different GdmCl concentra-tions (data not shown). However, the dimerization scenario isdifficult to prove or disprove by simple protein concentrationdependence experiments, since dimerization-folding couldcompete with higher-order aggregation reactions. Whilespeculative, the oligomerization scenario is consistent with theinsights provided by refolding behaviour in the presence ofGdmSCN. Aggregation behaviour as monitored by lightscattering is paradoxically linked to denaturation potency,with aggregation being absent in urea, observed to a smalldegree in GdmCl and very pronounced in GdmSCN. It ispossible that intermediate concentrations of GdmSCN favourthe formation of exposed hydrophobic regions in cutinaseduring the refolding process and thus traps them in a state thatcannot continue productive folding. This suggestion issupported by two separate observations, namely the peak inANS binding around 0.2 M GdmSCN (Fig. 1D) and the change

332 D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

in ANS fluorescence over time when refolding cutinase in 0.2–0.5 M GdmSCN (Fig. 4B). In contrast to GdmSCN, ANSfluorescence for cutinase in urea and GdmCl closely mirrors thechange in Trp-fluorescence, indicating that aggregative com-plications are particularly pronounced in GdmSCN. Thevariation in ANS time profiles at different GdmSCN concentra-tions no doubt reflect the extent to which the initial dimersprogress to larger aggregates with altered affinities for ANS.The rather complex effect of ANS on cutinase aggregation seenin other contexts [24], however, makes it difficult to make adetailed comparison between the Trp- and ANS-based timeprofiles of refolding. The strong interference in the folding ofcutinase would also explain the marked deviation betweenstability data obtained from GdmSCN versus urea (Table 2).

In conclusion, it is possible that the three denaturants'differential ability to support formation of exposed hydrophobicregions and consequent changes in aggregative properties duringthe refolding of cutinase is the basis for the very large variation infolding behaviour. This could be a general phenomenon withinprotein folding; different denaturants may thus bring differentpartially folded or aggregated states out into the open [44]. Inaddition, it cannot be excluded that it has some connection to thefunctional properties of cutinase. Aggregation may be favouredby the existence of sticky patches on the surface of cutinasewhich are required for the binding of its hydrophobic substrates.We have previously reported that cutinase and other proteinswith hydrophobic ligands or substrates have a much morepronounced tendency to aggregate during thermal scanningaround their pI than enzymes with more hydrophilic enzymes[24]. Cutinase is thus an interesting example of how functionalproperties can have bearing on a protein's folding andaggregation properties. On a more general note, we note thataggregation propensities can be enhanced somewhat counter-intuitively by increasing the denaturing potency of thedenaturant. This illustrates how sensitive the protein associationenergy landscape can be to solvent conditions, analogous to ourobservation that fibrils of the peptide hormone glucagon formedunder different conditions vary remarkably in their structure andstability [45–47].

Acknowledgements

D.E.O. is supported by the Danish National ResearchFoundation (inSPIN), the Villum Kann Rasmussen Foundation(BioNET) and EMBO's Young Investigator Programme.

References

[1] C. Tanford, Protein denaturation: Part A. Characterization of the denaturedstate, Adv. Protein Chem. 23 (1968) 121–217.

[2] C. Tanford, Protein denaturation: Part B. The transition from native todenatured state, Adv. Protein Chem. 23 (1968) 218–282.

[3] C. Tanford, Protein denaturation. Part C. Theoretical models for themechanism of denaturation, Adv. Protein Chem. 24 (1970) 1–95.

[4] R. Baldwin, On-pathway versus off-pathway folding intermediates, Fold.Des. 1 (1996) R1–R8.

[5] R.L. Baldwin, How Hofmeister ion interactions affect protein stability,Biophys. J. 71 (1996) 2056–2063.

[6] A.R. Fersht, Structure and Mechanism in Protein Science. A Guide toEnzyme Catalysis and Protein Folding, Freeman & Co., New York, 1999.

[7] C.N. Pace, Determination and analysis of urea and guanidine hydrochlo-ride denaturation curves, Methods Enzymol. 131 (1986) 266–279.

[8] N.A. Morjana, B.J. McKeone, H.F. Gilbert, Guanidine hydrochloridestabilization of a partially unfolded intermediate during the reversibledenaturation of protein disulfide isomerase, Proc. Natl. Acad. Sci. U. S. A.90 (1993) 2107–2111.

[9] K. Boren, H. Grankvist, P. Hammarstrom, U. Carlsson, Reshaping thefolding energy landscape by chloride salt: impact on molten-globuleformation and aggregation behavior of carbonic anhydrase, FEBS Lett.566 (2004) 95–99.

[10] D.E. Otzen, A.R. Fersht, Folding of circular and permuted ChymotrypsinInhibitor 2: retention of the folding nucleus, Biochemistry 37 (1998)8139–8146.

[11] R. Perez-Jimenez, R. Godoy-Ruiz, B. Ibarra-Molero, J.M. Sanchez-Ruiz,The efficiency of different salts to screen charge interactions in proteins: aHofmeister effect? Biophys. J. 86 (2004) 2414–2429.

[12] M. Zeeb, G. Lipps, H. Lilie, J. Balbach, Folding and association of anextremely stable dimeric protein from Sulfolobus islandicus, J. Mol. Biol.336 (2004) 227.240.

[13] J.M. Sanz, A.R. Fersht, Rationally designing the accumulation of a foldingintermediate of barnase by protein engineering, Biochemistry 32 (1993)13584–13592.

[14] A. Matouschek, J.T. Kellis, L. Serrano, M. Bycroft, A.R. Fersht, Transientfolding intermediates characterized by protein engineering, Nature 346(1990) 440–445.

[15] D.E. Otzen, M. Oliveberg, Salt-induced detour through compact regions ofthe protein folding landscape, Proc. Natl. Acad. Sci. U. S. A. 96 (1999)11746–11751.

[16] T.R. Sosnick, L. Mayne, S.W. Englander, Molecular collapse: the rate-limiting step in two-state cytochrome c folding, Proteins 24 (1996) 413–426.

[17] B.A. Krantz, L. Mayne, J. Rumbley, S.W. Englander, T.R. Sosnick, Fastand slow intermediate accumulation and the initial barrier mechanism inprotein folding, J. Mol. Biol. 324 (2002) 359–371.

[18] W.A. Eaton, P.A. Thompson, C.K. Chan, S.J. Hagen, J. Hofrichter, Fastevents in protein folding, Structure 4 (1996) 1133–1139.

[19] S. Longhi, C. Cambillau, Structure–activity of cutinase, a small lipolyticenzyme, Biochim. Biophys. Acta 1441 (1999) 185–196.

[20] D.L. Ollis, E. Cheah, M. Cygler, B. Dijkstra, F. Frolow, S.M. Franken, M.Harel, S.J. Remington, I. Silman, J. Schrag, et al., The alpha/beta hydrolasefold, Protein Eng. 5 (1992) 197–211.

[21] E.P. Melo, T.Q. Faria, L.O. Martins, A.M. Goncalves, J.M. Cabral,Cutinase unfolding and stabilization by trehalose and mannosylglycerate,Proteins 42 (2001) 542–552.

[22] E.P. Melo, L. Chen, J.M. Cabral, P. Fojan, S.B. Petersen, D.E. Otzen,Trehalose favors a cutinase compact intermediate off-folding pathway,Biochemistry 42 (2003) 7611–7617.

[23] K.R. Poulsen, T.K. Sørensen, L. Duroux, E.I. Petersen, S.B. Petersen, R.Wimmer, The interaction of Fusarium solani pisi cutinase with long chainspin label esters, Biochemistry 45 (2006) 9163–9171.

[24] S. Pedersen, L.W. Nesgaard, R.P. Baptista, E.P. Melo, S.R. Kristensen,D.E. Otzen, pH-dependent aggregation of cutinase is efficientlysuppressed by 1,8-ANS, Biopolymers 83 (2006) 619–629.

[25] J. Clarke, A.R. Fersht, Engineered disulfide bonds as probes of the foldingpathway of barnase: increasing the stability of proteins against the rate ofdenaturation, Biochemistry 32 (1993) 4322–4329.

[26] E.P. Melo, L. Chen, J.M. Cabral, P. Fojan, S.B. Petersen, D.E. Otzen,Trehalose favors a cutinase compact intermediate off-folding pathway,Biochemistry 42 (2003) 7611–7617.

[27] D.E. Otzen, O. Kristensen, M. Proctor, O. Oliveberg, Structural changes inthe transition state of protein folding: an alternative interpretation of curvedchevron plots, Biochemistry 38 (1999) 6499–6511.

[28] E.P. Melo, T.Q. Faria, L.O. Martins, A.M. Goncalves, J.M. Cabral,Cutinase unfolding and stabilization by trehalose and mannosylglycerate,Proteins: Struct. Funct. Genet. 42 (2001) 542–552.

[29] S.E. Jackson, A.R. Fersht, Folding of chymotrypsin inhibitor 2. 1.Evidence for a two-state transition, Biochemistry 30 (1991) 10428–10435.

333D.E. Otzen et al. / Biochimica et Biophysica Acta 1774 (2007) 323–333

[30] M. Silow, M. Oliveberg, High-energy channelling in protein folding,Biochemistry 36 (1997) 7633–7637.

[31] D.E. Otzen, Antagonism, non-native interactions and non-two-statefolding in S6 revealed by double-mutant cycle analysis, Prot. Eng. Des.Sel. 18 (2005) 547–557.

[32] R.F. Goldstein, L. Stryer, Cooperative polymerization reactions. Analyticalapproximations, numerical examples, and experimental strategy, Biophys.J. 50 (1986) 583–599.

[33] S. Timasheff, Protein hydration, thermodynamic binding, and preferentialhydration, Biochemistry 41 (2002) 13473–13482.

[34] J.E. Mogensen, H. Ibsen, J. Lund, D.E. Otzen, Elimination of an off-pathway folding intermediate by a single point mutation, Biochemistry 43(2004) 3357–3367.

[35] M.J. Parker, J. Spencer, A.R. Clarke, An Integrated kinetic analysis ofintermediates and transition states in protein folding reactions, J. Mol. Biol.253 (1995) 771–786.

[36] M.J. Parker, C.E. Dempsey, M. Lorch, A.R. Clarke, Acquisition of nativebeta-strand topology during the rapid collapse phase of protein folding,Biochemistry 36 (1997) 13396–13405.

[37] Y. Nozaki, C. Tanford, The solubility of amino acids, diglycine, andtriglycine in aqueous guanidine hydrochloride solutions, J. Biol. Chem.245 (1970) 1648–1652.

[38] T.E. Creighton, Electrophoretic analysis of the unfolding of proteins byurea, J. Mol. Biol. 129 (1979) 235–264.

[39] P.H. von Hippel, K.Y. Wong, On the conformational stability of globularproteins. The effects of various electrolytes and nonelectrolytes on thethermal ribonuclease transition, J. Biol. Chem. 240 (1965) 3909–3923.

[40] K.A. Dill, D. Shortle, Denatured states of proteins, Ann. Rev. Biochem. 60(1991) 795–825.

[41] Z. Shi, K. Chen, Z. Liu, N.R. Kallenbach, Conformation of the backbonein unfolded proteins, Chem. Rev. 106 (2006) 1877–1897.

[42] L.Y. Chen, G.J.M. Cabrita, D.E. Otzen, E. Pinho e Melo, Stabilization ofthe ribosomal protein S6 by trehalose is counterbalanced by the formationof a putative off-pathway species, J. Mol. Biol. 351 (2005) 402–416.

[43] D.E. Otzen, P. Kristensen, M. Oliveberg, Designed protein tetramer zippedtogether with an Alzheimer sequence: a structural clue to amyloidassembly, Proc. Natl. Acad. Sci. U. S. A. 97 (2000) 9907–9912.

[44] I. Gitlin, K.L. Gudiksen, G.M. Whitesides, Effects of surface charge ondenaturation of bovine carbonic anhydrase, ChemBioChem 7 (2006)1241–1250.

[45] J.S. Pedersen, D. Dikov, D.E. Otzen, N- and C-terminal hydrophobicpatches are involved in fibrillation of glucagon, Biochemistry 45 (2006)14503–14512.

[46] J.S. Pedersen, D. Dikov, J.L. Flink, D.E. Otzen, Sulfates dramatically stabilizea salt dependent type of glucagon fibrils, Biophys. J. 90 (2006) 4181–4194.

[47] J.S. Pedersen, D. Dikov, J.L. Flink, H.A. Hjuler, G. Christiansen, D.E.Otzen, The changing face of glucagon fibrillation: structural polymor-phism and conformational imprinting, J. Mol. Biol. 355 (2006) 501–523.