Results and Discussion

Mutation-Based Analysis of Higher-Order Structural Changes. To provide new tools for the in vitro analysis of NM assembly we engineered unique sites for the covalent attachment of a variety of biophysical probes. Because wild type NM (NM^wt) does not contain cysteine residues, we introduced NM^cys distributed throughout, with each mutation carrying just one substitution to provide a unique location for attachments (NM^S2C, NM^Y35C, NM^Q38C, NM^Q40C, NM^G43C, NM^G68C, NM^M124C NM^L144C, NM^T158C, NM^E167C, NM^K184C, NM^E203C, NM^S234C, and NM^L238C; ref. 26). Three of the mutants, NM^Y35C, NM^Q40C, and NM^M124C, were not expressed recombinantly in Escherichia coli at a level sufficient for in vitro analysis and were not studied further.

All other mutants were tested in vivo for their ability to function like wild-type Sup35p, by replacing the NM region of the genomic wild-type copy of SUP35 with each of the Sup35^cys variants by site-directed gene conversion in the yeast strain 74-D694. Replacements were performed both in [psi ^–] and [PSI ⁺] cells (26). All of Sup35^cys proteins were stably expressed at levels indistinguishable from that of Sup35p^wt (data not shown). All stably maintained the [psi ^–] phenotype in [psi ^–] cells and the [PSI ⁺] phenotype in [PSI ⁺] cells. Further, when [psi ^–] and [PSI ⁺] cells were mated inter se, as expected with cytoplasmic inheritance, the [PSI ⁺] state was dominant and segregated 4:0 in progeny. In the [PSI ⁺] state, each of the mutants efficiently converted [psi ^–] cells to [PSI ⁺] by cytoduction, the exchange of cytoplasm without nuclear transfer (data not shown). Finally, each was cured of the [PSI ⁺] state by growth on plates containing 5 mM GdmCl. Thus, all of the mutants behaved like wild-type protein with respect to the full spectrum of epigenetic characteristics of the [psi ^–] and [PSI ⁺] states.

Next, each of the purified mutant proteins was analyzed in vitro in NM assembly reactions. All assembled into β-sheet-rich fibers in seeded and unseeded reactions, and seven mutants had slightly different assembly kinetics, produced shorter fibers, or had molar ellipticity values at 222 nm, which were somewhat different from wild-type (Fig. 1A and data not shown). Only the mutants NM^S2C, NM^Q38C, NM^T158C, and NM^E167C were completely indistinguishable from NM^wt by every criteria examined.

Fluorescent Labeling of NM^cys. We labeled these four mutant proteins with acrylodan and, separately, with N,N′-dimethyl-N-(iodoacetyl)-N′-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)ethylenediamine (IANBD amide). Both fluorescent labels are commonly used to detect protein conformational changes and assembly because their fluorescent properties vary depending on solvent accessibility (27–30). The labeling efficiency ranged from 0.40 to 0.78 M label/M protein, as determined by UV/Vis absorption spectrophotometry, depending on the mutant and label used. Nonspecific labeling, assessed with NM^wt, was <0.05 M/M (data not shown).

To determine whether the labels themselves affected fiber assembly, mixed-assembly reactions were performed with equal quantities of labeled and unlabeled protein of each mutant. The ratio of labeled protein to unlabeled protein that remained in the soluble phase was constant throughout the assembly time course (data not shown), and the final level of assembly was the same (Fig. 1B and data not shown). The fibers formed with each of the labeled NM^cys mutants were indistinguishable from unlabeled NM^cys fibers in terms of their diameter (11.5 ± 1.5 nm) and concentration (Fig. 1C and data not shown). Thus, covalent attachment of acrylodan/IANBD amide did not influence the assembly of these mutants.

Fluorescence Assay for Conformational Conversion. Next, we asked which residues were located in positions that would provide a change in fluorescent signal on assembly in conformational conversion reactions (during seeded fiber elongation). For all four NM^cys mutants tested, acrylodan showed a blue shift in fluorescence emission maximum (λ_max), indicating that the environment of each cysteine substitution changed. To determine whether these changes were based on the conformational transitions that are associated with the transition from soluble protein into fibers, fluorescent changes were analyzed for 12 h in undisturbed, nonseeded reactions. Such reactions depend on spontaneous nucleation, and no NM fibers are detected in this time frame (4). This experiment revealed that acrylodan fluorescence emission showed a gradual change of λ_max during the preassembly stage for NM^S2C and NM^Q38C (data not shown). The N region of NM has by many criteria been established as the region responsible for nucleation. Thus, these changes most likely reflect early conformational transitions involved in the first stage of NCC, and warrants further study.

Acrylodan fluorescence emission of NM^T158C and NM^E167C, both located in the M region, revealed no significant change after 12 h in nonseeded samples (Fig. 2A and data not shown). However, coincident with seeded fiber assembly, solutions of NM^T158C- and NM^E167C-acrylodan showed increased fluorescence intensities, accompanied by a blue shift of λ_max (NM^T158C: 521 to 486 nm, Fig. 2 A ; NM^E167C: 528 to 502 nm). Thus, acrylodan labels at cysteine 158 and 167 are sensitive to the conformational differences between soluble and fibrous NM.

Seeded Fiber Elongation Occurs in Two Steps. In the second stage of NCC, soluble protein is converted into amyloid fibers by the nuclei. Nucleation of fiber assembly can be bypassed by the addition of preformed fibers to soluble protein (3, 4). Both NM^T158C- and NM^E167C-acrylodan (2 μM each) showed a rate of fiber assembly of v ^fluor = 8 ± 0.4 × 10^–4 μmol·s^–1 at 25°C (Fig. 2B and data not shown) in the presence of seed (4% wt/wt). At this seed concentration soluble NM is present in excess over the seeding fiber ends by ≈50,000-fold (3). This fiber assembly rate was similar to that measured for NM^wt by far-UV CD (3 × 10^–4 μmol·s^–1; refs. 3 and 4) and light scattering (5 ± 0.3 × 10^–4 μmol·s^–1) at identical experimental conditions.

To determine the kinetic parameters of fiber assembly, it was essential to ensure that both the substrate and the seed concentrations were not limiting in the reactions. Therefore, we determined fiber assembly rates with constant seed concentrations (4% wt/wt calculated for a 5 μM protein concentration) and varying soluble protein concentrations. Decreasing the soluble NM concentration 100-fold only decreased fiber assembly rates by a factor of two (Fig. 2B ). Hence, we are confident that soluble protein is in excess with 4% wt/wt of seed and 5 μM soluble NM.

The kinetics of seeded fiber elongation reproducibly showed a lag-phase of 80 ± 10 s at 25°C, then exhibited linear kinetics (Fig. 2C ). The delay in fiber assembly suggested that an assembly intermediate is formed. To confirm the existence of an intermediate, we took advantage of the fact that nonfibrous NM is soluble in SDS, whereas fibrous NM shows SDS resistance (3). Based on this determination, we developed an assay to detect intermediate complexes; i.e., soluble NM that is associated with seed but not converted into the fiber state. Seeds were prepared from NM^K184C, a cysteine-substitution mutant with surface-accessible sulfhydryl groups that allows labeling after fiber formation and that has a seeding efficiency indistinguishable to that of NM^wt (26). These NM^K184C seeds were biotinylated as described (26). Further, NM^T158C was labeled with iodo[1-¹⁴C]acetamide. Reactions were started by addition of biotinylated NM^K184C seed (50% wt/wt) to soluble NM^T158C-iodo[1-¹⁴C]acetamide. At distinct time points aliquots of the reaction were incubated with Streptavidin-coated Dynabeads. A high ratio of seed to soluble protein was used to ensure that the fiber ends (i.e., the seeds) were saturated with soluble NM, to allow the best opportunity of observing short-lived intermediate complexes. The beads were removed at different time points by using a magnet and washed with SDS to detect nonconverted intermediates. Both the SDS-soluble protein and the SDS-resistant fibers, which were attached to the beads, were analyzed by scintillation counting (Fig. 2D ). At early time points, a substantial fraction (≈50%) of the NM assembled with bead-bound seeds was soluble in SDS, at later time points the fraction of SDS-soluble material diminished. In a control experiment, in which the NM^K184C seeds were not biotinylated, no radioactivity could be detected attached to the beads. The ability to capture material bound to the seed that had not completely converted, established the formation of a detergent susceptible complex. However, this method did not have sufficient resolving power to analyze kinetic parameters of the assembly process.

To establish kinetic parameters, it was necessary to precisely discriminate between soluble and seed-bound NM. Therefore, we developed a sedimentation assay to detect the disappearance of soluble NM^T158C-acrylodan during fiber assembly. The total acrylodan concentration was plotted against its concentration in the supernatant, and each measurement was repeated six times to estimate the variation (Fig. 2E ). In combination with the wavelength shift assay (above), this assay provided data to kinetically analyze fiber assembly and develop a model for nucleated fiber elongation. These reactions have two reactants: the seed and the soluble NM, with the soluble NM as the substrate being in excess of the seed, and a catalyst that is not used up as the reaction progresses (the catalyst is the fiber ends, to which the soluble NM bind, but the same number of ends are present as the fiber elongates). These components and that these reactions reach steady state kinetics suggest that they can be analyzed with the Michaelis–Menten equation used to describe enzyme kinetics: where S is soluble NM, A is assembled protein (seed), SA is bound but not converted intermediate (akin to an enzyme:substrate complex), and AA is converted fiber, which again can act as seed. Importantly, we were unable to discriminate whether seed associates with monomers or oligomers or both. The observed rate of conformational conversion is determined experimentally by k ₁ and k _{–1 ,} the rate constants for binding and dissociation, and k _conf, the first-order conformational conversion rate. Because the dissociation rate of converted protein from the amyloid fibers is too slow to be detected in our experimental set-up, the back reaction is quasi-irreversible and is ignored in our model. We analyzed our experimental data by using a Lineweaver–Burk plot (Fig. 2F ), which yielded a straight line and a protein concentration of K _m = 0.12 ± 0.01 μM, at which the rate of reaction is half of the maximum rate. We also calculated a maximal rate of conformational conversion V _max = 10 ± 0.3 × 10^–4 μmol·s^–1, the rate constant of conformational conversion of k _conf = 5 ± 0.1 × 10^–3 s^–1, and a conformational conversion efficiency of k _conf/K _m = 42,000 M^–1·s^–1, which is equivalent to an enzyme's specificity constant.

Influences of Temperature on Seeded Fiber Elongation. We investigated the effect of increased temperature on seeded fiber elongation with NM^T158C-acrylodan in the presence of 4% wt/wt of seed, because previous studies detected an inhibitory effect of elevated temperature (4). In accordance to the previous study, we found a low-temperature optimum for the rate of fiber assembly as seen in the logarithm of NCC velocities plotted against the reciprocal temperature (Arrhenius plot; Fig. 3A , 4% seed). The sticking probability of soluble protein, which is reflected by k _conf/k _–1, characterizes the rate at which soluble NM associates with seed relative to dissociation; i.e., the sticking probability is high if k _–1 < k _conf. In our experiments, the abnormal temperature dependence with decreasing ratios of k _conf/k _–1 at elevated temperature indicates a significant rate enhancement for the dissociation of the seed-NM (SA) complex in comparison to its conversion into an assembled fiber (AA). At low temperature k _–1 « k _conf and k _conf/K _m becomes equal to k ₁. Because the dissociation of nonconverted, but seed-bound NM, has a high activation energy, k _–1 becomes predominant at high temperature.

To test this hypothesis experimentally, we measured the velocities of fiber elongation at 25°C and 40°C with a constant soluble NM^T158C-acrylodan concentration (2 μM) and increasing seed concentrations. We confirmed that increasing seed concentrations led to increasing fiber elongation velocities at both temperatures yielding maximal elongation rates at >10% wt/wt of seed (Fig. 3B ). Therefore, we plotted fiber elongation velocities at 12% wt/wt of seed, which should be a non-rate-limiting seed concentration for fiber elongation, against the reciprocal temperature. The plot revealed a temperature dependence of fiber elongation that is consistent with the collision theory of Arrhenius (Fig. 3A , 12% seed). The Arrhenius plot gives a straight line and its slope is equivalent to the activation energy E_a divided by the gas constant R = 8.3145 J K^–1·mol^–1. By using this equation, the activation energy for fiber elongation was calculated to be E_a = 11.7 ± 0.2 kJ mol^–1.

Acquisition of Secondary and Tertiary Structure of Soluble NM. To elucidate the influence of the conformation of soluble NM on the association with seed, we investigated the rate at which secondary, tertiary, and quaternary structures were acquired in soluble material. When NM is first diluted out of denaturants such as urea or guanidinium chloride (GdmCl), it adopts the characteristics of a molecule that is rich in random coil (typical for intrinsically unstructured proteins; ref. 23), indistinguishable from NM purified under nondenaturing conditions (4). To analyze whether the rate of this process influences seeded fiber assembly, we used 6 M GdmCl to form a homogenous and monomeric population of denatured NM (Fig. 4D ). After dilution into 5 mM sodium phosphate, pH 7.4, 150 mM NaCl, we monitored the time course of far-UV CD changes at 222 nm (refs. 4 and 26 and Fig. 4A ). The acquisition of secondary structure reached half-maximal amplitude after 24 ± 2 s with a rate constant of k _gain ^farUV = 2.1 ± 0.2 × 10^–2 s^–1. Thus, the formation of secondary structure is not rate determining for seeded fiber elongation.

The kinetics of acquisition of NM tertiary structure was investigated by using the four fluorescently labeled NM^cys mutants. Changes in tertiary structure of NM on dilution into buffer from 6 M GdmCl were investigated with two different techniques: IANDB amide-labeled protein was detected by fluorescence emission, and acrylodan-labeled protein was detected with near-UV CD. The fluorescence emission of IANBD amide revealed solvent exposure in all four mutants in 6 M GdmCl, as expected. A stable IANDB amide emission signal was reached after dilution into buffer indicative of a higher ordered environment. The time course had a half-maximal amplitude at 31 ± 4 s and a rate constant of k _gain ^fluor = 1.6 ± 0.2 × 10^–2 s^–1 (Fig. 4B and data not shown). Similarly, near-UV CD time courses with acrylodan-labeled NM (all four mutants led to the same results) showed a half-maximal amplitude after 33 ± 2 s and a rate constant of k _gain ^nearUV = 1.5 ± 0.1 × 10^–2 s^–1 (Fig. 4C and data not shown). Both independent measurements revealed that formation of some tertiary structure is also not rate limiting for seeded fiber assembly under the experimental conditions chosen.

Quaternary Structure Analysis. Dilution of NM^wt out of denaturant led to the formation of a mixed population of monomers (87 ± 5%) and heterogeneous oligomers, with varying molecular masses from tetramers to 30-mers (Fig. 4D ). Oligomerization was preceded by a lag phase of ≈60 s after dilution out of denaturant, which suggests that some acquisition of secondary and tertiary structure is required before oligomerization. Populations of monomers and oligomers were established after a half-time of 75 ± 5 s and remained constant for 3 h (Fig. 4D Upper). Because this steady state was achieved far before spontaneous nucleation (and well before seed was added), NM oligomerization is not likely to be rate determining for seeded fiber assembly in our experiments.

Our data suggested the following mechanism for initial structural changes of soluble NM, starting from the denatured state: where M _u is the unfolded monomer, M is the random-coil monomer with some structure, and O_x are the oligomers. The rate constant for structural gain of monomeric NM from the denatured state was k _gain = 1.5 ± 0.2 × 10^–2 s^–1. Remarkably, the rate of oligomerization and establishment of a steady-state distribution of monomers and oligomers showed little dependence on the concentration of NM between 0.7 and 46 μM (Fig. 4E ). This observation agrees with a previous study (3) where NM fiber assembly proceeds by means of the conversion of oligomers to nuclei with little concentration dependence. Nuclei form by conformational rearrangements of NM within the context of oligomeric intermediates and not by assembly of structurally converted monomers (3).

Implications. Self-perpetuating conformational conversion of NM is responsible for epigenetic inheritance of an altered conformational state in yeast. Previous work proposed that conformational conversion follows a mechanism of NCC (ref. 3 and Fig. 5), which is similar to that for conformational conversion of Alzheimer's A-β and islet amyloid polypeptide, suggesting that it may be widely applicable for describing amyloid fibrillogenesis, especially for intrinsically unstructured proteins (17, 31, 32). Because high-resolution techniques to analyze mechanistic details of NCC are unavailable, we used a generally applicable, sensitive method that measures changes in the intrinsic fluorescence of residues whose environments change during conformational transitions.

Oligomerization of NM is necessary to form nuclei. Here, we determined that oligomerization of soluble NM is not strongly dependent on the concentration on NM. Further, we found that the rate of oligomerization cannot be rate limiting in the first stage of NCC. It is, rather, a structural rearrangement of oligomers: most likely some oligomers convert to a conformational state competent for nucleation during prolonged incubation (Fig. 5). However, to investigate this first stage and to resolve the mechanistic details of NCC, it is crucial to first define the steps of the second stage.

The second stage of NCC can be imitated when a sufficient amount of preformed fibers (seed) is added to soluble NM. We identified two rate-determining steps of the second stage of NCC by using seeded reactions: (i) binding of soluble protein to the seed and (ii) conformational conversion initiated by the seed (Fig. 5). We have not determined the soluble protein fraction (monomeric, oligomeric, or both) that interacts with seed. At 25°C, soluble NM (S) dissociates more slowly from the complex with seed (SA) than it is converted. However, dissociation of nonconverted, but seed-bound NM, becomes predominant at high temperature. The final step of conformational conversion is quasi-irreversible under steady growth conditions, which are obtained by continuously adding new protein at the seeding ends, leading to the formation of amyloid fibers.

It has been shown recently for another amyloid, transthyretin, that amyloidoses inhibitors prevent amyloid formation by stabilizing the native state (33). Our results highlight a possibility that inhibition of amyloid assembly with chemicals that influence the association/dissociation properties of soluble protein with preexisting amyloid fibers might also be successful. Furthermore, our results contribute to understanding how to control amyloid formation kinetically, which will aid in the development of materials, and in the elucidation of protein-based mechanisms of inheritance and processes leading to protein-folding diseases.