Kinetic folding studies of the P22 tailspike beta-helix domain reveal multiple unfolded states

Abstract

The beta-helix is an important protein fold in many pathogens, and is a challenging system for folding pathway prediction because it primarily is stabilized by non-local interactions along the primary sequence. A useful experimental model of this fold is a monomeric truncation of P22 tailspike protein, the beta-helix domain (bhx). This report describes a systematic in vitro study of the chemical denaturation and refolding of bhx. Results from equilibrium chemical denaturation experiments were consistent with a two-state folding mechanism, but showed only partial reversibility. Stopped-flow fluorescence studies showed a single unfolding step, but two refolding steps. The slow refolding step could be partly attributed to proline isomerization, based on an increased rate during refolding in the presence of PPIase and an increased relative amplitude of this step with increasing delay time in double-jump refolding experiments observed over delays of 5–100 s. However, double-jump refolding experiments with delay times longer than 100 s along with size exclusion chromatography and dynamic light scattering of refolding samples showed that the overall refolding yield decreased as bhx was unfolded for longer periods of time. Furthermore, the losses resulted from aggregate formation during refolding. This suggests that a change occurs over time in the unfolded or denatured state ensemble that increases the propensity for aggregation upon the shift to more native-favoring conditions. Alternatively aggregate nuclei may be able to form even under high denaturant conditions, and these subsequently grow and consume monomer when placed under native-favoring conditions.

1. Introduction

In protein folding, the formation of beta-sheet structure often involves interactions between residues that are far from each other in the primary sequence. For this reason, beta-sheet formation is often more difficult to predict than alpha-helix formation. However, the formation of beta-sheet structures during folding has gained recent attention, partially due to the prevalence of that structure in amyloid and other protein aggregates present in disease states [1]. One of the proposed model structures for amyloid is the parallel beta-helix [2,3], a coiled structure in which each rung of the helix contributes a beta-strand to each of three parallel beta-sheets that run along the helix surface. The beta-helix structure is being increasingly discovered as a native protein fold, particularly in proteins associated with infectious disease such as viral adhesin and tail proteins [4]. Based on sequence information, more than 97% of the passenger domains from autotransporters – secreted virulence proteins of gram-negative pathogenic bacteria – are predicted to adopt a beta-helix structure even though they show little sequence homology [5]. A detailed understanding of beta-helix formation is obviously important given this fold’s utility as a model for protein aggregates in human disease, association with pathogenic systems, and ability to form robustly from diverse and non-repetitive protein sequences.

The current work focuses on the tailspike protein from the P22 bacteriophage. Fig. 1 shows the crystal structure of the tailspike trimer [6–8]. Tailspike has been used as a model system to study the folding of large, multimeric proteins, and its folding pathway has been extensively characterized both in vitro and in vivo [9–17]. Although tailspike is thermally stable and resistant to aggregation at temperatures up to 80 °C in its mature, trimeric form [18], it is highly susceptible to aggregation during folding, particularly in vitro, at elevated temperature, or in amino acid variants [12]. The aggregation is believed to occur through folding intermediates, which are destabilized under these conditions [12]. During refolding, tailspike adopts the majority of its secondary structure quickly, and in a monomeric state. It is believed that the successful formation of the beta-helix structure in the monomeric state is the critical step in determining the partitioning between a productive folding pathway or misfolding and aggregation [19]. Unfortunately, the complexity and irreversibility of the tailspike folding and assembly pathway, including competing aggregation events, make it difficult to study the beta-helix formation as an isolated event. For this reason, we are studying the in vitro folding of a truncated form of tailspike, comprising the beta-helix domain of the protein (bhx). The approximate location of the truncation points (residues 109 and 544) in the full-length structure is shown in Fig. 1. The bhx truncation has a molecular weight of 47.2 kDa, and exists primarily as a monomer at low concentrations [20]. Preliminary studies have demonstrated that bhx is spectroscopically similar to full length tailspike, retains partial enzymatic activity, has folding reversibility [20], and readily forms fibrous aggregates [21]. These observations indicate that the bhx truncation is structurally similar to the helix domain in the full-length protein and can act as a good model for a structured, monomeric folding intermediate.

Fig. 1.

P22 tailspike protein structure. Ribbon diagram of full-length, trimeric tailspike protein. Arrows indicate the approximate locations of residues 109 and 544, the truncation points for bhx.

2. Experimental

2.1. Materials

All chemicals used were obtained from major commercial suppliers. The cyclophilin enzyme was purchased from Sigma. Ultrapure urea was obtained from MP Biomedicals and used in all spectroscopic experiments.

2.2. Protein expression

Chemically competent E. coli BL21(DE3) cells (Novagen) were transformed with a pET11a plasmid containing the DNA encoding residues 109–544 of P22 tailspike (1tyr: pdb.org) under control of the T7 promoter [22]. Cells were selected on LB-Ampicillin (100 μg/mL) plates. For protein expression, an individual colony was selected and grown to an OD₆₀₀ of approximately 0.6 at 20 °C before expression was induced by addition of 1 mM IPTG. Following induction for 18 h at 20 °C, cells were harvested by centrifugation at 4000 ×g for 20 min, resuspended in lysis buffer (50 mM Tris, pH 7.6, 25 mM NaCl, 2 mM EDTA, 20 mM MgSO₄, 20 μg/ml DNase, 100 μg/ml lysozyme, 0.1% Triton-X 100) and subjected to two freeze/thaw cycles (−80 °C/20 °C). Cellular debris was removed by centrifugation at 13,000 ×g for 30 min and the resulting supernatant was used for subsequent purification.

2.3. Protein purification

Purification of beta helix protein (bhx) was performed as previously described for the full length tailspike protein [23]. Purified bhx was applied to a Superdex 75 size exclusion column (GE Healthcare) and eluted isocratically in 100 mM phosphate buffer, pH 7, to remove any larger multimers formed during the purification prior to folding experiments. Protein concentration was determined by absorbance at 280 nm (1 OD₂₈₀=1.33 mg/mL) [22]. Protein purity was N99% as determined by SDS–PAGE with silver staining.

2.4. Equilibrium folding/unfolding

Bhx was diluted to a concentration of 0.35 mg/mL in 100 mM phosphate buffer, pH 7, containing either 0 or 6 M urea and incubated in a 10 °C water bath for 1 h. The bhx samples were then diluted ten-fold into final urea concentrations between 0 and 6 M urea with a final protein concentration of 0.035 mg/mL and incubated for 1 h before measuring the fluorescence spectra. Samples incubated for longer times (up to 20 h) did not show a change in behavior compared to the 1 h samples. Spectra were measured at approximately 16 °C in a PC1 spectrofluorometer (ISS, Illinois) using excitation and emission slit widths of 0.5 and 1.0 mm, respectively, and an excitation wavelength of 280 nm. The emission was measured from 300 to 400 nm, and the spectra were quantified using the spectral center of mass (COM) [24], defined as

$$\text{COM}={\left(\frac{\sum I_{\nu }\cdot \nu }{\sum I_{\nu }}\right)}^{-1}$$ (1)

where ν is the wavenumber, and I_ν is the intensity at a given wavenumber, ν.

2.5. Stopped-flow folding/unfolding

Kinetic folding and unfolding experiments were performed in a model SX.18MV-R stopped-flow fluorescence instrument (Applied Photophysics, Leatherhead, U.K.). Unfolding and refolding were initiated by rapid dilution of native or denatured bhx into varying final urea concentrations. All experiments were performed at 15 °C in 100 mM phosphate buffer, pH 7. Intrinsic fluorescence was measured with an excitation wavelength of 280 nm, and the integrated emission signal was measured above a 305 nm cutoff filter. Kinetic traces were fit to the equation:

$$I\left(t\right)=I_{\infty }+\sum _i{A}_i\cdot e^{-k_i\cdot t}$$ (2)

where I is the fluorescence intensity, A_i is the amplitude of the ith phase, k_i is the rate constant of the ith phase, t is time in seconds, and I_∞ is the signal at infinite time. Quality of fit was evaluated by magnitude and randomness of residuals.

2.6. Double-jump refolding assay

For double-jump (sequential mixing) refolding experiments, bhx samples were unfolded by dilution in 7 M urea, and were refolded by stopped-flow mixing after a measured delay time. For delay times shorter than 100 s, both mixing steps were performed using stopped-flow, and the final solution conditions were 0.05 mg/mL bhx in 1.17 M urea. For delay times longer than 100 s, the unfolding step was performed by manual dilution, and the final solution conditions were 0.05 mg/mL bhx in 0.64 M urea. Refolding data were fit to Eq. (2) as described for single-mixing stopped-flow experiments.

2.7. Size exclusion chromatography

Size exclusion chromatography was performed using a Superdex 200 column (GE Healthcare) in order to determine the oligomeric state of refolded samples. Samples were eluted isocratically with 50 mM Tris, pH 7.6, 250 mM NaCl, 2 mM EDTA. The high salt concentration in the mobile phase was used to minimize non-specific interactions between the resin and the sample. Peak identity in refolded samples was based on comparison with monomeric samples as well as the known void elution volume of the column.

2.8. Light scattering

Oligomeric states of refolding samples were also assayed using laser light scattering. Dynamic light scattering experiments were performed using a Brookhaven (Holtsville, NY) BI9000AT correlator and a BI200SM goniometer, and a Lexel (Fremont, CA) model 95 argon ion laser operating at 488 nm. The autocorrelation function was measured at 90°, and analyzed using a method of cumulants as described by Weiss et al. [25], to give values for the average hydrodynamic radius (R_h) and polydispersity index (p₂).

3. Results

3.1. Folding of bhx is not fully reversible

Equilibrium chemical denaturation experiments were used to assess the folding behavior of beta helix protein (bhx). The intrinsic fluorescence spectra of bhx in 0 and 6 M urea are shown in Fig. 2a. Upon unfolding, the spectrum shows a decrease in intensity as well as a red shift. Although the red shift upon denaturation is small, it is consistent with that observed for the fluorescence spectra of full length tailspike and results from high surface exposure of the tryptophans in the native structure. Equilibrium unfolding and refolding experiments were performed by dilution of bhx samples beginning from 0 and 6 M urea, respectively. All samples were incubated at the final urea concentration for at least 1 h prior to recording the spectra, and no change was observed upon further incubation.

Fig. 2.

Equilibrium folding of bhx. (a) Intrinsic fluorescence spectra of bhx in 0 M (_) and 6 M (…) urea. Denatured protein shows a decrease in intensity along with a small red shift. Spectral center of mass (b) and emission intensity at 340 nm (c) are shown for bhx unfolding (○) and refolding (●) as a function of final urea concentration.

Denaturation curves are shown for both the spectral center of mass (Fig. 2b) and the fluorescence intensity at 340 nm (Fig. 2c). Although the curves show two-state behavior consistent with earlier reports [20], the fluorescence intensity was not fully recovered upon refolding, indicating partial misfolding or a loss in folding yield. Spectral center of mass data were fit to a two-state pseudo-equilibrium model, and the resulting parameters are listed in Table 1. It should be emphasized that these are only apparent equilibrium parameters, as the unfolding is clearly not fully reversible.

Table 1.

Apparent thermodynamic^a and kinetic^b parameters for bhx unfolding.

Parameter	Calculated valuec
$\Delta G_{\text{app}}^0\left(\text{kcal\hspace{0.17em}}{\text{mol}}^{-1}\right)$	7.4 ±1.3
meq (kcal mol−1 M−1)	−2.9 ± 0.5
cm (M urea)	2.6 ± 0.04
$k_{\text{u}}^0\times {10}^6\left({\text{s}}^{-1}\right)$	2.2 [0.2, 18]
$m_{\text{u}}^1\left({\text{M}}^{-1}\right)$	4.3 ± 0.99
$m_{\text{u}}^2\left({\text{M}}^{-1}\right)$	−0.35 ± 0.11

^aApparent free energy of unfolding in water($\Delta G_{\text{app}}^0$), denaturant dependence (m_eq), and midpoint unfolding concentration (c_m) calculated from regression of equilibrium COM data to a standard two-state model [46].

^bUnfolding rate constant in water ($k_{\text{u}}^0$) and denaturant dependences ($m_{\text{u}}^1,m_{\text{u}}^2$) calculated from regression of kinetic unfolding data to Eq. (3).

^cUncertainties represent 95% confidence intervals from non-linear regression. For the unfolding rate constant, upper and lower bounds are given in brackets, as the confidence interval is not symmetric about the mean.

3.2. Kinetics of bhx folding and unfolding

The time-dependent folding behavior of bhx was further investigated through stopped-flow fluorescence experiments. For all conditions observed, unfolding data could be well fit to a single-exponential model (Fig. 3a), which suggests a single unfolding step. In contrast, refolding traces required two summed exponentials to accurately fit the data, as can be seen from the residuals of the one and two phase fits, shown in Fig. 3b and c, respectively. Inability to fit folding traces with a single exponential term indicates non-two-state behavior and could signify the presence of parallel folding paths or of folding intermediates along a sequential path. No improvement in quality of fit was observed by increasing the number of exponentials in the summation.

Fig. 3.

Kinetics of (un)folding. Representative kinetic traces for unfolding (a) and refolding (b), (c) of bhx. Experimental data (○) and fits to a summed exponential expression (_) with one (a), (b) or two (c) terms are shown in the main plot. Residuals for the shown fit are plotted below. For clarity, plots show every third experimental point.

A chevron plot representation of the data from different (un) folding endpoint conditions is shown in Fig. 4. The unfolding rate constant shows a non-linear dependence on denaturant concentration. This type of chevron plot curvature has been observed for many protein systems including creatine kinase [26] and alpha-spectrin [27]. Suggested interpretations of chevron curvature include accumulation of kinetic intermediates and changes in the unfolding transition state as a function of denaturant concentration [28–30]. Unfolding data was fit with the following quadratic expression to extrapolate the value of k_u in the absence of denaturant:

$$\text{ln}\left(k_{\text{u}}\right)=\text{ln}\left(k_{\text{u}}^0\right)+m_{\text{u}}^1\left[c\right]+m_{\text{u}}^2{\left[c\right]}^2$$ (3)

where k_u is the unfolding rate constant, $k_{\text{u}}^0$ is the unfolding rate constant without urea, and c is the concentration of urea (Table 1).

Fig. 4.

Chevron plot representation of kinetic folding data. The urea dependence of the rate constants for unfolding (○) as well as the slow (■) and fast (▼) folding phases of bhx. Error bars represent the standard deviation from at least three experiments, and are smaller than the size of the symbol where not shown.

Perhaps more significantly, the rate constants of both folding phases show little dependence on denaturant concentration. The two phases have relaxation times of approximately 20 s and 100 s. As in the equilibrium experiments, the fluorescence intensity was not fully recovered upon refolding.

3.3. Refolding in presence of PPIase

Slow steps in protein folding can often be attributed to cis–trans isomerization around prolyl-peptide bonds [31]. In unordered polypeptides, the trans conformation is favored, and the ratio of trans:cis residues is approximately 70:30. In contrast, proline residues in folded protein structures typically adopt a specific isomeric state, based on steric constraints. In the bhx domain of tailspike, there are 16 proline residues, all of which adopt the trans conformation in the native state, according to the crystal structure [6–8].

To test for the significance of proline isomerization during the in vitro folding of bhx, time-dependent folding experiments were performed in the presence of the human peptidyl-prolyl isomerase, cyclophilin (PPIase). Representative refolding traces of bhx with and without the presence of 0.5 μM PPIase in the refolding buffer are shown in Fig. 5. A fit of the data to Eq. (2) showed a 30% increase in the rate of the slower folding phase, consistent with catalysis of an isomerization step by the PPIase. No statistically significant change was observed in the rate constant of the fast folding phase.

Fig. 5.

Effect of PPIase. Refolding kinetic traces for bhx with (♦) and without (+) PPIase.

3.4. Double-jump refolding

Double-jump refolding experiments were used to test for changes in the unfolded state over time. Because proline isomerization rates are slow relative to protein unfolding rates, the proline residues in a newly unfolded protein will initially be in the same conformation they had in the native state. Over time, an equilibrium will be established between cis and trans isomers in the unfolded protein. If proline isomerization is responsible for a slow folding step, this step should become more pronounced as a larger fraction of the unfolded protein population adopts the non-native isomer.

Double-jump refolding experiments were performed for both short delay times (5–100 s) and long delay times (N100 s). In both cases, the rate constants for the two folding phases did not change significantly over time (Fig. 6a and c) indicating that the mechanism is independent of delay time. With increasing delay times up to 100 s, the amplitude of the slow phase increased as that of the fast phase decreased (Fig. 6b). This is consistent with a slow phase linked to proline isomerization, where the population of the slow-folding isomer increases during the time the protein is unfolded. Interestingly, at delay times longer than 100 s the amplitudes of both phases decreased with increasing delay time (Fig. 6d) while the ratio of the amplitudes from the two phases remained constant. Thus the total amount of fluorescence recovery, which correlates with protein refolding yield, decreased over time. Additionally, the partitioning of unfolded protein into fast and slow folding populations was established by 100 s, and remained constant over longer times.

Fig. 6.

Sequential-mixing refolding. Samples of bhx were allowed to unfold in high concentration urea solutions for varying delay times, before being refolded by dilution into low urea conditions. The rate constants (a,c) and relative amplitudes (b,d) of the fast (▼) and slow (■) refolding phases of bhx are plotted as a function of the refolding delay time. Error bars for data from long delay times (c,d) represent the 95% confidence intervals from the non-linear regression. For data with short delay times, where variability between runs was significant, error bars represent the standard deviation in the parameters from fits of at least 5 traces for each time point.

3.5. Irreversible loss due to aggregation

One key question is whether the loss of recovery of fluorescence intensity is due to misfolding of bhx to a state with lower fluorescence, or to a loss in overall refolding yield. To address this question, refolding samples were analyzed by size exclusion chromatography (Fig. 7). The chromatogram shows that the refolded sample consists of a mixture of monomer and a second species, which elutes at an earlier volume and is likely a small, soluble oligomer. The total peak area across both species remained constant, indicating that no protein is lost (i.e. becomes insoluble). The oligomer formed relatively quickly upon refolding, as it was observed in samples loaded onto the SEC column immediately following dilution into native-favoring conditions. In double-jump experiments analyzed by SEC, the monomer peak area decreased as a function of delay time, with a corresponding increase in the oligomer peak. If refolding was initiated following unfolding overnight, the resulting sample consisted entirely of oligomer, with no monomer observed in the chromatogram.

Fig. 7.

Size-exclusion chromatogram of native bhx (_) and a refolded bhx sample (…). The chromatogram of bhx refolded after denaturation for approximately 2 h shows an additional peak with an earlier elution time, indicating formation of an oligomer.

Dynamic light scattering (DLS) was performed to determine the approximate size and heterogeneity of the oligomeric species. The light scattering measurement was performed on a sample of native monomer (never unfolded) and a refolded sample consisting of only oligomer as confirmed by SEC. Analysis using the method of cumulants [25] gave a calculated hydrodynamic radius of 4.04±0.07 nm for the monomer and 9.97±0.15 nm for the oligomer. The polydispersity indexes for the monomer and oligomer were 0.15±0.04 and 0.21±0.03, respectively, indicating that both samples have relatively low polydispersity. From the combined SEC and DLS data, it is clear that the loss in fluorescence intensity upon refolding results from a loss of refolding yield due to formation of a small, misfolded oligomer.

4. Discussion

In this study, we examined the folding behavior of a truncation of P22 tailspike protein comprising the beta-helix domain as a model for beta-helix formation in the full-length protein. We found that although equilibrium unfolding occurred via a two-state pathway, refolding was not a simple two-state process. A proposed folding scheme for bhx that is consistent with all the data is shown in Fig. 8. The parentheses indicate steps that occur primarily in the unfolded state. The proposed mechanism shows an equilibrium between two unfolded species, U_trans and U_cis, which refold through parallel paths. U_trans refolds to the native species, N, in one step with rate constant k₁, while U_cis refolds in two steps through an intermediate, I. The first step, with rate constant k₃, is the major folding step, while the second, slower step, with rate constant k₂, represents an isomerization event. The scheme also shows an aggregation-prone unfolded state, U*, which forms during time in denatured-favoring conditions with rate constant, k₄. Aggregates or oligomers which are formed from U* are labeled A.

Fig. 8.

Proposed folding mechanism for bhx showing a parallel refolding pathway which results from proline isomerization, as well as formation of an off-pathway oligomer. Parentheses indicate folding steps which occur primarily in the unfolded state.

The ability of PPIase to catalyze the slow folding step of bhx is a strong indication that this step involves proline isomerization. Although the 30% increase in refolding rate is not dramatic, the activity of PPIase is often low, particularly in non-optimal solution conditions or in cases where the prolines are not easily accessible to the enzyme [31,32].

Changes in proline isomerization that occur in the unfolded state are generally not accompanied by a change in tryptophan fluorescence, and are therefore not directly observable using intrinsic fluorescence spectroscopy. Sequential mixing, or double jump experiments provide an indirect method to observe changes in the unfolded population based on observed differences in behavior when refolding is induced. Our double jump experiments with delay times up to 100 s showed the characteristic proline isomerization behavior, in which the amplitude of the slow folding phase increases relative to the fast phase with increasing delay times. The time scale at which proline isomerization effects were observed in double jump experiments varies. For some systems, changes in relative amplitude are observed at delay times of several minutes or longer [28], while for other systems the effect is primarily seen at shorter delay times [33]. For bhx, an equilibrium between fast and slow folding isomers was established within 100 s and no further change in the relative amplitudes of the two folding phases was observed after that time.

Experimentally, only two distinct kinetic folding phases are observed, although there are three total folding steps in the proposed mechanism. There are two mathematically possible explanations for this discrepancy: k₂ is very fast compared to k₃ such that the conversion of U_cis to I is rate limiting, or k₁ and k₃ have very similar values such that they are observed as one combined phase in the kinetic experiments. The first possibility is not consistent with our knowledge of the physical system, since k₂, the rate constant of the isomerization step, should be slow. In contrast, the latter possibility is easily explained, since k₁ and k₃ are both rate constants of major structural rearrangement steps in folding, and it is therefore reasonable for them to have similar values.

Proline isomerization has been identified as the slow folding step in another beta-helical protein, pectate lyase C (pelC) from Erwinia chrysanthemi [28,34]. There are 12 prolines in pelC; one of them has the cis confirmation in the native structure and is responsible for the slowest folding phase present. A second slow phase was also attributed to proline isomerization, although a comprehensive mutagenesis study was unable to identify the specific residue responsible for this phase [34]. The authors concluded that multiple proline residues contribute to this phase, and bhx folding behavior can likely be similarly explained.

The loss of overall folding yield with increased residence time in the denatured state is an unusual behavior in protein folding. The oligomer is seen in SEC samples injected immediately after dilution into native favoring conditions; therefore it forms relatively quickly. The development of aggregation propensity in the unfolded ensemble, however, is significantly slower than other folding steps, with a time constant of approximately 50 min. Attempts to refold bhx after long denaturation periods (overnight) resulted in formation of only the aggregated species with no detectable monomer, indicating that the formation of the aggregation prone species is effectively irreversible.

The formation of an aggregation prone unfolded state of bhx is likely to have implications for the folding of full-length tailspike. It has been shown that the early in vitro folding states of tailspike are significantly different from those observed in vivo, where the beta-helix structure is formed while the nascent polypeptide is still attached to the ribosome [35]. Decreased efficiency in formation of the beta-helix structure from polypeptides that have been unfolded in denaturant for significant periods of time may explain the significant loss to aggregation observed during in vitro refolding of tailspike. The current study conclusively demonstrates that folding of the monomeric beta-helix structure from tailspike, a critical first step in folding and assembly, is not a simple two-state process and may be responsible for the differences in folding efficiency in vitro and in vivo.

One possible physical explanation for the aggregation-prone unfolded state of bhx is isomerization of multiple proline residues. Perhaps a single proline in the cis conformation merely creates a slow folding form of bhx requiring an additional isomerization step to reach the native form, but multiple prolines in the cis conformation make it nearly impossible for the helix to form at all. It is also possible that some proline residues are critical for folding, while others are less important. Given the large number of proline residues in the bhx sequence, it is likely that both the number and position of cis prolines in the unfolded polypeptide contribute to its ability to productively refold. The resulting partially folded or misfolded intermediate species would be vulnerable to aggregation due to exposure of hydrophobic residues that are normally buried in the native structure. Refolding mechanisms with kinetic folding traps involving proline isomerization have also been reported for single chain antibodies [36,37].

Aggregation due to kinetic folding traps is consistent with the common view that protein aggregation occurs through partially unfolded, molten-globule monomers, where early association steps result from relatively non-specific interactions [38,39]. In some cases, however, aggregation is driven by specific interactions. The presence of a specific conformation of protein, even in a small fraction of the population, can act as a seed for aggregation. These aggregate seeds or nuclei, which may consist of monomers or oligomers, are able to recruit additional monomer leading to formation of aggregates. An important example of this behavior is observed in prion diseases, where the aggregation-prone species is a conformationally altered monomeric form of prion protein [40,41]. Perhaps bhx adopts a similar, specific conformation under denaturing conditions that leads to the rapid formation of aggregate upon dilution into native favoring conditions.

The denatured state ensemble of proteins, once viewed as a random coil, is now recognized to contain significant structure and is being characterized with a variety of techniques [42–44]. Here, low levels of dimer (~3–10%) were observed in samples of unfolded bhx that had been chemically crosslinked using bis(maleimido)hexane prior to separation by SDS–PAGE (data not shown). It is possible that either these dimers or a specific conformation of monomers act as nuclei or seeds for further aggregation upon dilution into native favoring conditions. The low polydispersity of the aggregates formed during refolding of bhx is consistent with a mechanism in which fast growth occurs from preformed nuclei [45]. Further study of the denatured state of bhx could provide valuable insight into the nature of the aggregate nuclei.