Folding of Prion Protein to Its Native α-Helical Conformation Is under Kinetic Control*

The recombinant mouse prion protein (MoPrP) can be folded either to a monomeric α-helical or oligomeric β-sheet-rich isoform. By using circular dichroism spectroscopy and size-exclusion chromatography, we show that the β-rich isoform of MoPrP is thermodynamically more stable than the native α-helical isoform. The conformational transition from the α-helical to β-rich isoform is separated by a large energetic barrier that is associated with unfolding and with a higher order kinetic process related to oligomerization. Under partially denaturing acidic conditions, MoPrP avoids the kinetic trap posed by the α-helical isoform and folds directly to the thermodynamically more stable β-rich isoform. Our data demonstrate that the folding of the prion protein to its native α-helical monomeric conformation is under kinetic control.

Although protein folding is commonly thought to be controlled by thermodynamic preferences, it has been understood by many, including Anfinsen and others (1,2), that kinetic issues can alter the folding landscape. Whereas most small globular proteins will refold spontaneously in vitro to a native conformation, in vivo folding often exploits auxiliary molecules and defined subcellular compartments to avoid the deposit of misfolded forms (3). Increasingly, a role for protein misfolding in a variety of neurodegenerative diseases has emerged. A common thread joining prion-based diseases and Alzheimer's disease, and possibly Parkinson's disease and frontotemporal dementia, is the conversion of a normal, cellular, monomeric isoform of a protein into a ␤-sheet-rich, polymeric form (4 -6). When the deposited polymeric form is sufficiently ordered to bind Congo red and exhibit birefringence to polarized light, the pathologic term amyloid is used to cluster these and other maladies (7).
Recent studies by Dobson and others (8 -12) have demonstrated that a broad variety of proteins that rapidly fold into monomeric or oligomeric cellular forms under native-like conditions can also be refolded into ␤-rich, amyloid forms under conditions that destabilize the native state. So far, these proteins have not been associated with human deposition diseases. This finding has led to the suggestion that the ability to adopt alternative ␤-rich folds capable of forming amyloid is not a unique property of specific proteins associated with conformational diseases but reflects a general property of polypeptide chains (13). The interplay between protein concentration and the conformational preferences of the monomeric chain in driving the transition to a ␤-rich multimeric isoform remains to be more fully explored.
Glockshuber and colleagues (14) have shown that a fragment of the mouse prion protein folds very rapidly into the ␣-helixrich conformation with a half-life of 170 s as measured at 4°C. Here, we report that a ␤-sheet-rich conformation of the mouse prion protein (MoPrP) 1 is thermodynamically more stable than its native ␣-helix-rich conformation. The conformational transition from the ␣-helical to a ␤-sheet-rich isoform is controlled by a large energetic barrier that is associated with partial unfolding and oligomerization of an intermediate state. Under partially denaturing conditions, it is possible to avoid the kinetic trap that leads to the normal cellular isoform, PrP C , and fold the prion protein directly to a thermodynamically more stable, non-native ␤-isoform. Our data demonstrate that folding the prion protein to its native ␣-conformation is under kinetic, not thermodynamic, control.

EXPERIMENTAL PROCEDURES
Protein Preparation-The expression and purification of recombinant MoPrP 89 -231 was performed as described by Mehlhorn et al. (15).
Circular Dichroism-CD spectra were recorded with a J-720 CD spectrometer (Jasco, Easton, MD) scanning at 20 nm/min, with a bandwidth of 1 nm and data spacing of 0.5 nm using a 0.1-cm cuvette. Each spectrum represents the accumulation of three individual scans after subtracting the background spectra. To monitor the refolding curves, MoPrP was diluted from 10 M to various concentrations of urea in 20 mM sodium acetate in the absence or in the presence of 0.2 M NaCl, pH 3.6, and then incubated at room temperature for different periods of time. No change in pH value was detected during the time course of incubation. To monitor the kinetic trace of the conformational transition, ␣-MoPrP was rapidly mixed with 10 M urea in a 1:1 volume ratio, whereas to monitor the kinetics of refolding to the ␤-MoPrP, MoPrP unfolded in 10 M urea was mixed with buffer, again at a 1:1 volume ratio. All kinetic experiments were carried out in 20 mM sodium acetate and 0.2 M NaCl, pH 3.6.
Analysis of the Kinetic Data-The rate constant and apparent rate order of refolding were calculated from Equation 1, in which C 0 is concentration of the monomer MoPrP at zero time, C is concentration of monomer MoPrP at time t, and n is apparent order of the process. ⌬E ‡ was calculated from the Arrhenius relation (1) with k obs measured experimentally and k 0 determined from the equation for diffusion-controlled reaction, assuming that the reaction follows fifthorder kinetics. Size-exclusion Chromatography-All separations were performed at 23°C with a flow rate of 1 ml/min using TSK-3000 high pressure liquid * This work was supported in part by grants from the National Institutes of Health (AG0Z132, AG10770, and NS14069), as well as by a gift from the G. Harold and Leila Y. Mathers Foundation. I.B. was supported by the John Douglas French Foundation for Alzheimer's Disease Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Thioflavin T Assay-To follow the kinetics of amyloid formation, 0.64 mg/ml ␤-MoPrP was incubated in 20 mM sodium acetate and 0.2 M NaCl, pH 5.5, constantly shaken at 36°C. In the time course of incubation, aliquots of MoPrP were diluted 8 times by phosphate-buffered saline, pH 7.0, and the fluorescence was measured using a LS50B fluorimeter (PerkinElmer Life Sciences) at 482 nm (excitation at 450 nm, excitation slit is 5 nm, emission slit is 10 nm, 0.4-cm rectangular cuvettes) with 5 M thioflavin T.
Congo Red Binding-Congo red (Sigma) was dissolved in 5 mM potassium phosphate, 150 mM NaCl, filtered 5 times with a 0.22-mm filter (Millipore, Bedford, MA), and adjusted to 0.2 mM. The difference spectra were obtained by subtracting the Congo red spectra in the absence of MoPrP from the Congo red spectra in the presence of 1.5 M MoPrP amyloid, corrected for MoPrP scattering.
Electron Microscopy-Samples were absorbed on carbon-coated, 600mesh copper grids for 30 s, stained with freshly filtered 2% ammonium molybdate or 2% uranyl acetate, and were viewed in a JEOL JEM 100CX II electron microscope at 80 kV at standard magnifications of 40,000.

RESULTS AND DISCUSSION
To estimate the thermodynamic stability of ␣-MoPrP, its urea-induced unfolding and refolding was measured using the far-UV CD as a probe of its secondary structure. In a low salt buffer, pH 3.6, the urea-induced unfolding profile of the molar ellipticity at 222 nm shows a single cooperative transition between the ␣-isoform and the unfolded state (Fig. 1a). When ␣-MoPrP is unfolded in 10 M urea and then refolded by diluting the urea concentration, its refolding curve expresses hysteresis, a phenomenon indicative of a non-two-state process (Fig.  1a). Both the unfolding and refolding limbs of the curve remain stable for at least 5 weeks when MoPrP is kept in a low salt buffer (20 mM sodium acetate). However, when refolding of MoPrP at 10 M concentration is performed in a high salt buffer (0.2 M NaCl, 20 mM sodium acetate), the refolding curve undergoes a gradual time-dependent transformation from a single cooperative transition to a transition with local intermediates (Fig. 1b). If a similar experiment is performed at 30 M MoPrP, the migration of refolding curve occurs more rapidly (Fig. 1c).
Unfolded MoPrP folds first to the ␣-helical form upon dilution from 10 M urea (Fig. 1d). During incubation for 5 weeks in the high salt buffer, it undergoes a slow conformational transition to the ␤-rich form as illustrated by the change in the overall CD spectra, as well as by reduction of the CD signal at 222 nm ( Fig.  1, b and d). The conformational transition from the ␣-helical to a ␤-sheet-rich isoform is accompanied by oligomerization as judged by size-exclusion chromatography (SEC) (Fig. 1e). Immediately after dilution from 10 M urea, a new peak corresponding to an oligomer appears, in addition to the peak that represents a monomer. During the conformational transition, the population of monomer decreases whereas the fraction of oligomer grows. Although the square variance analysis of the oligomer peak indicates that there is certain heterogeneity of the oligomer species, electrospray mass spectrometry suggests that an octamer is the dominant multimeric assembly (data not shown).
The unfolding and refolding behavior of MoPrP demonstrates hysteresis, a time-dependent transformation of the single transition curve into a double transition curve, and a concentration-dependence for this process. These observations challenge the application of either of the two possible classical three-state models used previously to estimate the thermodynamic parameters for PrP unfolding (16,17). In contrast, a model with two independent transitions, one between the ␣-isoform and unfolded and the other between the ␤-isoform and unfolded, can be used to fit the data. We have observed that the refolding to the ␣-isoform is much faster than the refolding to a ␤-isoform, whereas the time-dependent accumulation of a ␤-isoform indicates that it is thermodynamically more stable than the ␣-isoform. Thus, MoPrP diluted out of urea folds predominantly to the ␣-isoform with little ␤-isoform present. The presence of a ␤-isoform would account for the hysteresis between the unfolding and the refolding curves (Fig. 1b). With time, the refolding curve transforms from an apparent single transition to the double transition, demonstrating equilibration of the ␣and the ␤-isoforms.
Direct comparison of the thermodynamic stability of the ␣-and the ␤-isoforms illustrate that the ␣-isoform is not the lowest energy state. First, we estimated the thermodynamic parameters for the ␣-isoform using the urea-induced unfolding curve and applying a classical two-state model (see Fig. 1a and Table I) (18). To evaluate the thermodynamic stability of the ␤-isoform, two parameters, the molar ellipticity at 222 nm and the fraction of the oligomer, were monitored in parallel as a function of urea concentration after re-equilibration of MoPrP for 5 weeks. Despite the fact that a small fraction of MoPrP remains trapped in the ␣-helical conformation even after 5 weeks, we have exploited the fraction of the ␤-oligomer as directly measured by SEC to analyze the "unfolded 7 ␤-isoform" equilibrium using the two-state model (Fig. 2a). The unfolding curve measured by CD requires deconvolution, because it is composed of signals from the ␤-isoform, the unfolded state, and the ␣-isoform. Using the population of the monomer measured by SEC as a function of urea concentration and the thermodynamic parameters estimated previously for the "␣isoform 7 unfolded" equilibrium, we calculated the contribution of the ␣-isoform to the CD curve (Fig. 2b). When this contribution is subtracted from the original curve, a curve reflecting the unfolded 7 ␤-isoform transition results. As shown in Fig. 2c, the transition curves for the ␤-isoform are superimposible, with ⌬G, m, and C 1/2 determined from the two techniques equal within the uncertainty of the experiment (Table I). Both ⌬G and C 1/2 demonstrate that the ␤-isoform is thermodynamically more stable than the ␣-isoform (see Fig. 2c and Table I). Because both isoforms can be refolded directly from the unfolded state, we have used the unfolded state as a reference in the free energy diagram (Fig. 2d).
Although the ␤-isoform is thermodynamically more stable than the ␣-isoform, it might be not a true global energy minimum state, because the ␤-isoform can undergo an additional time-dependent transition to a polymeric amyloid form. Incubation of ␤-MoPrP at 37°C and constant shaking lead to the formation of higher molecular weight aggregates that possess amyloid properties. The process of amyloid formation monitored by thioflavin T binding displays an apparent latent period and then an exponential accumulation of the aggregate (Fig. 3a). In addition to thioflavin T, the amyloid of MoPrP binds Congo red in a specific manner as judged by birefringence of polarized light and typical red shift of absorbance spectra (Fig. 3b). Aggregated MoPrP forms numerous twisted fibrilar filaments as seen by electron microscopy (Fig. 3c).
Why is the thermodynamically more stable ␤-isoform not accessible during folding under native conditions? Previously, it has been shown that the folding of PrP to the ␣-isoform is an extremely fast, first-order process (14). Folding to the ␤-isoform is slower by several orders of magnitude and is concentration-dependent. To prevent the conformational conversion, the ␣-isoform has to be separated by a large energetic barrier from the ␤-isoform. Although the free energy diagram does not provide a view of the actual kinetic pathway for the conformational transition, several important observations can be made concerning the origin of the energetic barrier. First, the ␣-isoform has to unfold substantially on route to the ␤-isoform. As we have seen before, the ␣-isoform converts very slowly to the ␤-isoform at pH 3.6 in the absence of urea (Fig. 1b). This process can be accelerated by shifting the ␣-isoform 7 unfolded equilibrium toward the unfolded state. After jumping the urea concentration from 0 to 5 M, we observed a very fast loss of secondary structure by the ␣-monomer within the dead time of manual mixing, followed by an accumulation of a ␤-sheet-rich conformation (Fig. 4a). This result illustrates that a substantial portion of the energetic bar-rier requires partial unfolding of the ␣-isoform. The connection between the structural complexity of the pretransition state and the energetic barrier is demonstrated by previous observations  that conversion of PrP-derived peptides with low structural complexity into ␤-rich isoforms occurs spontaneously and does not require partially denaturing conditions (19 -21). Whether the transition state on the way from the ␣to the ␤-isoform is predominantly unfolded under native conditions or whether it has residual ␤-sheet or ␣-helical structure remains to be established.
A significant contribution to the energetic barrier seems to be associated with the process of oligomerization. As shown on Fig. 4b, the accumulation of a ␤-rich conformation is accompanied by oligomerization. The fact that both kinetic curves are superimposible illustrates that the two processes are coupled (Fig. 4a). MoPrP can be refolded directly to the ␤-isoform if the unfolded protein is diluted first to 5 M urea (Fig. 4b). When dialyzed out of urea and salt, ␤-MoPrP is stable for months at room temperature with no detectable conversion to the ␣-isoform. Analysis of the kinetic traces indicates that the process of folding to the ␤-isoform represents a single transition with apparent reaction order of 5, regardless of whether the refolding is initiated by dilution of urea from 10 to 5 M, a jump of the urea concentration from 0 to 5 M, or if the conformational transition occurs in the absence of urea. Such a high order of reaction suggests that the conformational transition will depend upon the concentration of the transition state.
To estimate the energy of activation (⌬E ‡ ) of the conformational transition, the Arrhenius relation, can be used, in which k obs is the constant rate of the conformational transition measured experimentally, and k 0 is the rate of the process under diffusion control. Under experimental conditions employed (pH 3.6 and 10 M MoPrP), we found that the ␣-isoform is separated from the ␤-isoform by an energy barrier of 20 kcal/mol (Fig. 4c). The energetic barrier is predicted to be much higher under physiological conditions because of the lower concentration of PrP and the higher thermodynamic stability of the ␣-isoform at pH 5-7 (Fig. 4d). For wild-type MoPrP, the calculated energy barrier of 35-45 kcal/mol is sufficient to prevent the process of conformational transition over the functional lifetime of the protein. Hence, a large energetic barrier prevents the conversion of the ␣-isoform to the thermodynamically more stable ␤-isoform. From the kinetic perspective, the process of conformational transition can be facilitated by the reduction of the energetic barrier (22). Thus, single point mutations associated with inherited forms of prion diseases might reduce the energetic barrier by stabilizing the transition state. Additionally, if PrP Sc provides a template for the conversion of PrP C to PrP Sc by binding and stabilizing the transition state, this would also speed up the conformational conversion.
Our results clearly indicate that the folding of native PrP C is under kinetic control. The observations that many proteins are able to adopt alternative amyloid-like folds require us to revisit the role of kinetic traps in protein folding (8 -12). If a ␤-rich amyloid competent structure is an intrinsic preference especially at a high protein concentration, then compartmentalization of partially folded intermediates and proteins that mediate unfolding and clearance of misfolded proteins play critical roles in cellular health. In addition, side-chain patterns that favor the formation of amyloid, such as alternating polar and nonpolar amino acid residues, will be avoided (23). Despite these strategies, some proteins, including PrP, A␤, ␣-synuclein, parkin and tau, find a route to a ␤-rich, multimeric structure with unfortunate consequences. FIG. 4. a, the kinetic trace of transition from the ␣-isoform to the ␤-isoform (10 M MoPrP) induced by jumping the urea concentration from 0 to 5 M at pH 3.6, as monitored simultaneously by SEC (open squares) and CD (filled circles). b, the kinetic trace of folding the ␤-isoform induced by jumping the urea concentration from 10 to 5 M at pH 3.6 monitored by CD. In the inset, the linearity of the fifth-order (1/ normalized n Ϫ 1 versus time) plot suggests that the process of folding may follow an apparent fifth-order kinetics. c, free energy diagram of the conformational transition, representing the activation energy (kcal/ mol) estimated at pH 3.6 and 10 M of MoPrP. TS represents the transitional state. The ␤-MoPrP undergoes an additional transition to the amyloid form, represented by the dotted line. d, the activation energy versus the concentration of MoPrP shown at different pH values, as calculated from the Arrhenius relation applying the diffusion-controlled reaction rates. The physiological concentration range of MoPrP is shown in the shaded bar.