Characterizing Intermolecular Interactions That Initiate Native-Like Protein Aggregation

Abstract

Folded proteins can access aggregation-prone states without the need for transitions that cross the energy barriers for unfolding. In this study we characterized the initial steps of aggregation from a native-like state of the acylphosphatase from Sulfolobus solfataricus (Sso AcP). Using computer simulations restrained by experimental hydrogen/deuterium (H/D) exchange data, we provide direct evidence that under aggregation-promoting conditions Sso AcP populates a conformational ensemble in which native-like structure is retained throughout the sequence in the absence of local unfolding (N^∗), although the protein exhibits an increase in hydrodynamic radius and dynamics. This transition leads an edge strand to experience an increased affinity for a specific unfolded segment of the protein. Direct measurements by means of H/D exchange rates, isothermal titration calorimetry, and intermolecular relaxation enhancements show that after formation of N^∗, an intermolecular interaction with an antiparallel arrangement is established between the edge strand and the unfolded segment of the protein. However, under conditions that favor the fully native state of Sso AcP, such an interaction is not established. Thus, these results reveal a novel (to our knowledge) self-assembly mechanism for a folded protein that is based on the increased flexibility of highly aggregation-prone segments in the absence of local unfolding.

Introduction

A wide range of human diseases is associated with the conversion of peptides and proteins from their soluble functional states into insoluble aggregates characterized by extensive β-sheet structure and a fibrillar and unbranched morphology, generally referred to as amyloid fibrils (1). These pathologies range from neurodegenerative disorders, such as Parkinson's disease and Alzheimer's disease, to systemic amyloidoses involving the formation of proteinaceous deposits in a wide range of vital organs (1).

The process that leads to the formation of amyloid aggregates is a heterogeneous multistep reaction in which many parallel events can occur (1). For soluble proteins, the first stage of the process is typically a conversion from their native states into aggregation-prone states in which they are able to self-assemble through a variety of subsequent nucleation and growth steps. One intriguing aspect of amyloid formation is that although the resulting fibrils share many common structural features (2,3), the ensemble that is populated at the initiation of the aggregation process by different proteins can be significantly diverse (2). Thus, although many proteins have been found to aggregate from ensembles of largely unfolded conformers, in other cases normally folded proteins can access amyloidogenic states without the involvement of a global unfolding reaction, for example, as a result of thermal fluctuations of the native state (4,5) or disruption of the quaternary structure (6,7). In these cases, the amyloidogenic state can be described as an ensemble of native-like conformations characterized by the presence of locally unfolded elements (referred to as N^∗) (4,6,8,9). The characterization of the properties of these amyloidogenic N^∗species is crucial for our understanding of aggregation under native-like conditions, and is likely to be particularly important in the context of aggregation events that occur in vivo.

In this study we focused our attention on the acylphosphatase from the archaebacterium Sulfolobus solfataricus (Sso AcP). This 101-residue enzyme is an α/β globular protein that has a ferredoxin-like topology and is characterized by the motif βαββαβ (10). Unlike the other acylphosphatases characterized thus far, Sso AcP has in addition to the globular domain an 11-residue-long unfolded N-terminal segment (see Fig. S1 in the Supporting Material) (10). Aggregation of Sso AcP is not linked to any known disease, but its properties, as discussed below, are such that the protein has been used extensively as a model system in which to study the mechanisms of protein aggregation involving N^∗ states (11).

Sso AcP aggregates at 25°C in 50 mM acetate buffer at pH 5.5 in the presence of 15–25% (v/v) 2,2,2-trifluoroethanol (TFE) (12). Several pieces of evidence suggest that under these conditions, aggregation of Sso AcP is initiated after the conversion of the native state (N) into an ensemble of N^∗ conformations. Not only is the protein still enzymatically active, but the secondary structure content and the environments in the vicinity of aromatic residues are still native-like, as revealed by circular dichroism (CD) analysis (9). Moreover, [¹H, ¹⁵N] heteronuclear single quantum coherence (HSQC) spectra show that monomeric Sso AcP remains highly native-like upon addition of up to 25% (v/v) TFE, with only a small number of minor changes in chemical shifts being evident (5). In addition, folding is faster than unfolding under these conditions, indicating that the protein populates predominantly a folded state before aggregation occurs (12). Furthermore, aggregation of the protein is faster than unfolding, confirming that an unfolding event is not required, at least for the initial steps of aggregation (12). Taken together, these results indicate that Sso AcP aggregation is initiated from an ensemble of N^∗ conformations (11).

The first phase of the Sso AcP aggregation process (∼1 min long under the conditions used here) involves the self-assembly of N^∗ monomers, resulting in the formation of a variety of aggregated species ranging from tetramers up to large oligomers containing ∼100 molecules (5). These assemblies involve monomeric units that retain their N^∗ conformation, as monitored by far-UV CD, Fourier transform infrared (FTIR), and NMR spectroscopies, do not bind Thioflavin T (ThT) or Congo red (CR) and have enzymatic activity (5,13). In a second phase of aggregation, the early aggregates convert into amyloid-like protofibrils that appear in electron microscopy images as spheres or thin filaments with diameter of 3–5 nm (12,13), possess a high content of β-sheet structure, are able to bind ThT and CR, and lack enzymatic activity (13).

A series of experiments involving kinetic analysis of the aggregation behavior of a set of Sso AcP variants suggested that two regions of the structure play a fundamental role in the formation of the early aggregates that takes place during the first phase of the self-assembly process: β-strand 4, spanning residues 83–90, and the 11-residue unfolded N-terminal segment (Fig. S1) (9,14,15). Moreover, the analysis of the kinetic data pointed to an intermolecular interaction between the N-terminal segment of one Sso AcP molecule and the globular unit of another molecule (14). However, no direct structural evidence has been obtained for this model, and key questions remain unanswered, such as the specific region of the protein targeted by the N-terminal segment, the stoichiometry and the orientation of the interaction, and the possible existence of poorly populated, locally unfolded states, as observed for another member of the acylphosphatase family (16).

To address these fundamental issues, we studied the Sso AcP protein variant lacking the N-terminal segment (ΔN11 Sso AcP) in the absence or presence of a peptide having the same sequence as the N-terminal segment of the protein (N11). This model offers unique opportunities to add information to that previously obtained from the analysis of kinetic tests on Sso AcP (14), for the following reasons: 1) The ΔN11 variant does not self-assemble under the conditions we used to explore aggregation of the wild-type protein. Thus, it can be studied under these conditions over the long periods of time required to obtain quantitative and direct information about the interactions that trigger aggregation. 2) The excision of the initial 11 residues does not destabilize Sso AcP and does not affect its enzymatic activity or the secondary and tertiary structure content of the protein (9,14). Thus, conclusions from the study of ΔN11 Sso AcP can be extended to obtain insights into the behavior of the wild-type protein. 3) The absence of the first 11 residues allows results to be obtained in the absence of possible intramolecular interactions between the globular unit and the N-terminal segment. (Such interactions are not productive for aggregation (14), and their characterization is therefore beyond the scope of this work.) 4) One can compare results obtained for ΔN11 Sso AcP in the presence/absence of the N11 peptide under conditions that promote aggregation of the wild-type protein to distinguish between effects associated with the conversion of N into N^∗ and those associated with the interaction between monomeric species in the initial steps of aggregation.

Materials and Methods

Sample preparation

We obtained (1-oxyl-2,2,5,5-tetramethyl-Δ3-pyrroline-3-methyl) methanethiosulfonate (MTSL) from Toronto Research Chemicals (North York, Ontario, Canada), TFE from Sigma, and synthetic peptides with amidated C-termini from Genscript (Piscataway, NJ). In all cases, the purity of the peptides was >95%. The gene encoding ΔN11 Sso AcP was subcloned into the pGEX-2T plasmid (9,10,14), and the protein was expressed and purified as previously described (17). The resulting protein contains the GS dipeptide at the N-terminus (not considered for residue numbering). ¹⁵N-labeled ΔN11 Sso AcP was expressed in minimal medium enriched with ¹⁵NH₄Cl and purified according to the protocol used for the unlabeled protein. The purity of samples was checked by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and electrospray ionization mass spectrometry (ESI-MS). In all cases, the degree of isotopic labeling was >97%.

Labeling of the M1C-N11 and E11C-N11 peptides with the spin-label MTSL was carried out according to protocols reported previously (18). Briefly, a 2 mg quantity of peptide was dissolved in 300 μl of a 10 mM HEPES buffer at pH 7.4 containing 100 mM NaCl and a 10-fold molar excess of MTSL. After 1.5 h of incubation at room temperature, unbound MTSL was removed by gel filtration using a PD MiniTrap G-10 gravity column (GE Healthcare). The peptide was then lyophilized and examined by ESI-MS. In no case was any unlabeled peptide detected in the mass spectra, indicating that the degree of labeling was >99%.

Chemical shift analysis

[¹H, ¹⁵N] HSQC spectra of ΔN11 Sso AcP were acquired with a Bruker Avance 500 TCI-Cryo spectrometer (Bruker) at 25°C in H₂O/D₂O (90/10 v/v) solutions with a variety of buffer and salt concentrations. All solutions contained 150 μM ΔN11 Sso AcP in 50 mM acetate buffer at pH 5.5 with or without 20% (v/v) TFE. Furthermore, in some cases the N11 peptide at a concentration of 150 μM or NaH₂PO₄ at a concentration of 50 mM was added to the protein samples. Chemical shifts under all the experimental conditions described above were measured by titrations starting from solutions for which spectra had previously been analyzed and assigned (10). Data were acquired and processed with the use of Topspin 2.1 (Bruker Biospin) and analyzed wth Sparky 3.114 for Windows (SPARKY 3; T. D. Goddard and D. G. Kneller, University of California, San Francisco, CA). Chemical shift changes were calculated according to the method described by Mulder et al. (19). Assignment of the N11 peptide was carried out via a set of nuclear Overhauser enhancement spectroscopy and total correlation spectroscopy experiments in 50 mM acetate buffer at pH 5.5 with and without 20% (v/v) TFE.

Hydrogen/deuterium exchange experiments

Hydrogen/deuterium exchange (HX) measurements were carried out on ΔN11 Sso AcP under different experimental conditions. Briefly, we lyophilized purified Sso AcP after buffer exchange with 20 mM ammonium carbonate at pH 7.4 using a Centricon YM-3 (3000 Da; Millipore, Billerica, MA). We then dissolved 1 mg of the protein in 550 μl of D₂O solution containing one of the buffers described in the previous section. A series of [¹H, ¹⁵N] HSQC spectra were then acquired at 25°C with a sweep width of 13 ppm and numbers of scans ranging from 2 to 16. The time between ΔN11 Sso AcP dissolution and the first spectrum to be recorded ranged from 4 to 7 min, and spectra were acquired up to 8 weeks after the beginning of the reaction. Peak heights were normalized to protein concentrations determined from 1D ¹H spectra. We processed the data using Topspin 2.1, analyzed them with Sparky 3.114, and fitted them to a single exponential function to obtain exchange rate constants k_exp for the amide groups. Protection factors (PFs) were calculated as k_int/k_exp, where k_int is the intrinsic decay rate observed for unfolded peptides (20).

Restrained simulations

We carried out restrained simulations starting from the Sso AcP crystal structure (PDB entry 2BJD) in a 6 Å shell of TIP3 water molecules using PFs as experimental restraints and the CHARMM22 force field in the CHARMM program to obtain structural ensembles describing the conformation of Sso AcP in the presence and absence of 20% (v/v) TFE (16). Details of these simulations are reported in the Supporting Material.

Intermolecular paramagnetic relaxation enhancements

We dissolved 1 mg of ΔN11 Sso AcP in H₂O/D₂O (90/10) solution containing 50 mM acetate buffer at pH 5.5 and 20% (v/v) TFE, to obtain a final protein concentration of 150 μM. [¹H, ¹⁵N] HSQC spectra were acquired under these conditions with a sweep width of 30 ppm. In other experiments, the protein was dissolved in the same solutions but with the MTSL-labeled peptide at relative concentrations ranging from 0.25 to 1.50 peptide/protein. Peak heights from [¹H, ¹⁵N] HSQC spectra were measured and the ratios between values in the presence and absence of the peptide were calculated with the use of Sparky 3.114.

Isothermal titration calorimetry

We conducted isothermal titration calorimetry (ITC) experiments to study the binding between the N11 peptide and ΔN11 Sso AcP using an ITC-200 instrument from Microcal (Wien, Austria). In brief, a solution of 100 μl of N11 at a concentration of 700 μM was titrated in 4 μl aliquots into the calorimetric cell containing 250 μl of 50 μM ΔN11 Sso AcP. Before measurements were obtained, both the protein and the peptide were dialyzed against a buffer containing 50 mM acetate at pH 5.5 with or without 20% (v/v) TFE. Injections were performed every 300 s at 25°C, and titration of N11 in the sample cell containing buffer alone was subtracted from the signal obtained before it was analyzed with the use of Microcal software (Origin 7.0).

Results

Chemical shift changes under aggregation promoting conditions

To identify the residues of the Sso AcP structure that was most affected by the addition of a cosolvent promoting aggregation of the wild-type protein, and to distinguish between the effects of the addition of the cosolvent on the monomer and subsequent aggregation, we measured ¹H^N and ¹⁵N NMR chemical shifts when ΔN11 Sso AcP was titrated from conditions under which aggregation of wild-type protein does not occur at a detectable rate to conditions under which aggregation to yield amyloid species occurs within 1 h (9,12–15). In particular, ΔN11 Sso AcP was titrated in 50 mM acetate buffer at pH 5.5 with TFE concentrations ranging from 0% to 20% (v/v). Of importance, the fact that ΔN11 Sso AcP does not aggregate under these conditions enables the protein to be studied in detail without problems arising from precipitation in the sample tube. The [¹H, ¹⁵N] HSQC spectrum measured at the end of the titration (i.e., the spectrum of the protein in 20% (v/v) TFE) is still that of a globally folded protein, although the addition of TFE induces changes of chemical shifts (Δδ) throughout the Sso AcP sequence (Fig. 1 A and B, and Fig. S2 A). We introduced a threshold corresponding to the average Δδ (Δδ) + 2 standard deviations (SD, σ). Residues whose Δδ values deviate from Δδ by >2σ are located in two regions: helix 1 (F32) and the segment encompassing β-strand 4 (E82 and Y86) and the long loop that follows the strand (K92 and G93). This is consistent with previous reports pointing to the importance of β-strand 4 (9,14) and the C-terminal region (residues 78–101) in the initial steps of aggregation (5). We also assigned the chemical shifts of the N11 peptide in the presence and absence of 20% (v/v) TFE and did not detect any significant changes between the two conditions. This finding shows that TFE has no effect on the structure of the N-terminal segment of the protein, and suggests that aggregation is induced by its effect on the globular segment of the molecule.

Figure 1.

(A) Perturbations of the amide chemical shifts of ΔN11 Sso AcP in 50 mM acetate buffer at pH 5.5, 25°C, resulting from the addition of 20% (v/v) TFE. The dashed black line represents the mean chemical shift change Δδ. Continuous black lines represent the mean chemical shift change Δδ + one standard deviation σ (Δδ + σ) and two standard deviations (Δδ + 2σ). Chemical shift changes between Δδ + σ and Δδ + 2σ are colored in light gray (yellow in the online version), and those larger than Δδ + 2σ are in dark gray (red in the online version). The topology of ΔN11 Sso AcP is also shown. Chemical shift changes were calculated according to the method of Bai et al. (20). (B) The structure of ΔN11 Sso AcP is colored to show the regions affected most strongly by the addition of TFE. Color code as in panel A.

It was previously shown that ΔN11 Sso AcP does not aggregate detectably under conditions used to induce aggregation of the wild-type protein, even in the presence of a 25 molar excess of the 11-residue peptide (N11) that has the same sequence as the N-terminal segment of Sso AcP (14). This observation is consistent with the notion of an intermolecular interaction between the N-terminus and the globular segment. Indeed, the N11 peptide inhibits aggregation of wild-type Sso AcP, indicating the presence of an intermolecular interaction involving N11 that competes with that of the N-terminal segment (14). Therefore, we titrated ΔN11 Sso AcP with this peptide in 50 mM acetate buffer at pH 5.5 and 20% (v/v) TFE, under conditions ranging from a zero- to fourfold molar excess of the peptide relative to the protein concentration. The results of this experiment show that there are only minor perturbations to the chemical shifts of the protein (Fig. S2 B). The average chemical shift change is <0.01 ppm and no regions with a significant number of large Δδ values could be identified. The experiment suggests that the intermolecular interactions between the N-terminal segment and the globular unit do not affect the conformation of the protein, and that the interactions leading to aggregation of Sso AcP may be highly labile, with peptide molecules continuously binding and releasing from the globular segment.

HX experiments

To obtain information about the dynamics of the Sso AcP globular unit before and during aggregation, we performed HX measurements on ΔN11 Sso AcP under a range of experimental conditions. As a preliminary experiment, we verified that exchange of the amide protons of all residues are in the EX2 regime under the conditions used in this work. We measured exchange rates as a function of pH (from pH 4 to pH 7). The observed linear pH dependence of the log (k_exp) indicates that exchange is occurring in the EX2 regime in this range of pH values. In the EX2 regime, HX experiments provide information about the degree of protection of the amide groups that interact with solvent (21).

In a first experiment, PFs were measured for ΔN11 Sso AcP in 50 mM acetate buffer at pH 5.5, 25°C. The results are shown in Fig. 2 A and B, and Table S1. We calculated the average of the observed PF values (symbolized as log (PF)) and the SD σ, and used these two values to classify each amide group in one of the following four categories: 1), log (PF) values lower than log (PF) − σ (highly dynamic); 2), log (PF) higher than log (PF) − σ and lower than log (PF) (dynamic); 3), log (PF) higher than log (PF) and lower than log (PF) + σ (protected); and 4), log (PF) value higher than log (PF) + σ (highly protected). The results show that secondary structure elements have many protected and some highly protected amide groups; indeed, β-strands 4 and 5 are the only regions of secondary structure that show more than one highly dynamic amide group. The loops in the structure are much more flexible than the regions of secondary structure and possess many dynamic residues. Thus, although the Sso AcP globular unit is native under these conditions, the region encompassing β-strands 4 and 5 and the loop between them is highly flexible and more solvent-exposed relative to other regions of the sequence, even in solutions that do not promote aggregation.

Figure 2.

HX measurements carried out on ΔN11 Sso AcP under four different sets of experimental conditions. (A) log (PF) measured in 50 mM acetate buffer at pH 5.5, 25°C; log (PF) values have been classified into four categories (see text): highly dynamic (red), dynamic (yellow), protected (blue), and highly protected (black). Dashed and continuous black lines represent the boundaries among these four classes. (B) Structure of ΔN11 Sso AcP illustrating the results shown in Fig. 2A. The color code is shown in panel A. (C) log (PF) measured in 50 mM acetate buffer at pH 5.5 with 20% (v/v) TFE, 25°C. The residue positions that show a decrease in log (PF) relative to the values measured in panel A that is higher than the average value of Δlog (PF) + 1 SD (1σ) are colored in red. (D) The structure of Sso AcP colored according to the code used in panel C. (E) Log (PF) values measured in 50 mM acetate buffer at pH 5.5 with 20% (v/v) TFE, 25°C, and in the presence of the N11 peptide at the same concentration as ΔN11 Sso AcP. The positions that show an increase in log (PF) relative to the values measured in panel C that is greater than Δlog (PF) + 2σ are colored in blue. (F) Structure of ΔN11 Sso AcP colored according to the code used in panel E. (G) Log (PF) measured in 50 mM acetate buffer at pH 5.5, 25°C, with 20% (v/v) TFE, in the presence of the N11 peptide at the same concentration as ΔN11 Sso AcP and of 50 mM phosphate buffer. The positions that show a decrease or an increase in log (PF) relative to the values measured in panel A that is higher than Δlog (PF) + (or −) 1σ deviation are colored in red and blue, respectively. (H) Structure of Sso AcP colored according to the code used in panel G.

Although previous observations obtained by bulk techniques showed that the N^∗ state has a globally native-like fold (9), the existence of local unfolding events involving even a few residues of the Sso AcP sequence has not yet been tested. Thus, we measured PFs in 50 mM acetate buffer at pH 5.5, 25°C, containing 20% (v/v) TFE. The results are shown in Fig. 2, C and D, and Table S1, and reveal that TFE induces a widespread reduction in protection. For each residue, we calculated the difference between the log (PF) values measured in the presence and absence of TFE (Fig. S3 A). The average of these values is equal to −0.52, representing a decrease in PF of 70.3% on average. To identify the regions of the protein most affected by the presence of the cosolvent, we depicted in red the amide groups whose Δlog (PF) values are lower than Δlog (PF) − σ (Fig. 2, C and D). Of interest, one amide group located in β-strand 4 (S87) and two amide groups located in β-strand 5 (E99 and T100) show decreases in protection larger than Δlog (PF) − σ, suggesting that although the C-terminal region is the most dynamic region even in the absence of TFE, the presence of this cosolvent induces a further increase in dynamics. Overall, for the protein in 20% (v/v) TFE, the region encompassing β-strands 4 and 5 and the interconnecting loop appears to be the most flexible and solvent-exposed.

Next, we measured PFs in 50 mM acetate buffer at pH 5.5 at 25°C containing 20% (v/v) TFE, in the presence of the N11 peptide at the same concentration as ΔN11 Sso AcP. The results are shown in Table S1 and Fig. 2, E and F, where we have identified those amide groups whose Δlog (PF) values relative to the values obtained in the presence of TFE but in the absence of N11 are higher than Δlog (PF) + 2σ (depicted in blue in Fig. 2, E and F; see also Fig. S3 B). Intriguingly, we find that the three amide groups belonging to β-strand 4 and pointing toward the solvent in the native structure (V84, Y86, and F88) show increases in PFs in the presence of the N11 peptide that are greater than the average value by >2σ. This observation is consistent with the existence of an intermolecular interaction involving the N-terminal segment of the full-length sequence (represented here by the N11 peptide) and the outer edge of β-strand 4, which causes a decrease in solvent accessibility. Under conditions that promote aggregation of wild-type Sso AcP, the presence of the peptide induces a small but significant overall increase in log (PF) values throughout the sequence. The Δlog (PF) upon addition of N11, excluding those residues that differ by >2 SDs, is 0.12, suggesting that the interaction between the N-terminal segment of the protein and the globular unit induces a general decrease in dynamics throughout the protein.

As a final control experiment, we measured amide PFs in 50 mM acetate buffer at pH 5.5, 25°C, containing 20% (v/v) TFE, in the presence of the N11 peptide at the same concentration as ΔN11 Sso AcP, and of 50 mM sodium phosphate. Aggregation of Sso AcP is inhibited in the presence of phosphate due to the stabilizing interaction that results from the binding of a phosphate ion in the catalytic site (5,14). The results of these measurements are reported in Table S1 and Fig. 2, G and H, where it is evident that the addition of the phosphate ion induces an increase of the PFs at almost all positions in the protein. Indeed, the average value of Δlog (PF) (Δlog (PF)) is 0.95 relative to the measurements in the presence of TFE and N11 (Fig. S3, C), 1.07 relative to the values obtained in the presence of TFE but without N11, and 0.54 relative to the measurements performed in the absence of TFE and N11. These data suggest that when a phosphate ion is bound in the catalytic site of Sso AcP, the dynamics of the protein are markedly reduced, especially under conditions where aggregation is disfavored. We identified the amide groups showing Δlog (PF) values higher than Δlog (PF) + σ (depicted in blue in Fig. 2, G and H) or lower than Δlog (PF) − σ (depicted in red in Fig. 2, G and H) compared with the condition in which N11, but not phosphate, is present. Of interest, we found that two amide groups located in β-strand 4 (Y86 and F88) showed a decrease (rather than the increase generally observed) in their PF upon addition of phosphate. This result suggests that the addition of a phosphate ion reduces the interaction between the N-terminal peptide and β-strand 4, resulting in an increase in dynamical fluctuations in this region.

Restrained simulations

The direct interpretation of HX data to characterize poorly populated partially unfolded conformational states that exist in equilibrium with the N^∗ state is a complex challenge. The combination of NMR data and molecular-dynamics simulations has proved to be a powerful method for sampling the complex conformational space of proteins and peptides (22). Moreover, the use of experimental restraints is a powerful tool for driving the conformational exploration and the free energy toward an accurate solution (16). Thus, we carried out simulations on ΔN11 Sso AcP using experimental PFs as restraints to generate an ensemble of structures (16). In the absence of TFE (the N state), this ensemble indicates that the globular segment of Sso AcP maintains a native-like topology in which the secondary structure elements are rigid (Fig. 3, A and B, and Fig. S4 A). Upon addition of 20% (v/v) TFE, the calculated structure ensemble (the N^∗ state) reveals that the protein maintains its native-like fold, and there is no evidence that subpopulations exhibit local unfolding (Fig. 3, C and D). However, the ensemble describing N^∗ differs from that describing N in a number of ways. First, in the presence of TFE, the N^∗ ensemble of ΔN11 Sso AcP structures suggests larger fluctuations throughout its sequence than those observed in the ensemble describing the N state (Fig. S4 A). This effect is particularly evident in the first helix and in the two loops spanning residues 22–29 and residues 76–82 (Fig. 3, C and D). Second, the presence of the cosolvent induces an increase in the radius of gyration, with a median value of 1.23 ± 0.01 nm and 1.26 ± 0.01 nm in the absence and presence of TFE, respectively (Fig. S4 B). Third, after addition of TFE, there is a small but significant decrease in the number of residues involved in native contacts (58 ± 1 and 56 ± 1 residues in the absence and presence of TFE, respectively; Fig. S3 C) and a slight increase in the number of residues assuming coil conformation (21 ± 1 and 24 ± 1 residues in the absence and presence of TFE, respectively; Fig. S4 D). Thus, our simulations reveal that although N^∗ maintains its native-like character, the globular unit in the presence of TFE has a higher tendency to populate a wider range of conformations than in its absence, and represents a slightly more dynamic and expanded version of the native-state conformation.

Figure 3.

Structure ensembles obtained by restrained simulations in the absence (A and B) and presence (C and D) of 20% (v/v) TFE. Front (A and C) and side (B and D) views are shown. β-strand 4 is colored in dark gray (red in the online version).

Detection of the interaction by intermolecular PRE measurements

Up to now, the existence of intermolecular interactions between the N^∗ state of Sso AcP and N11 has been supported only by indirect evidence (14). Furthermore, there is little direct information about the specific region on ΔN11 Sso AcP that is targeted by N11, or about the arrangement of N11 in this interaction. Thus, we attached a paramagnetic spin label, MTSL, to two variants of the N11 peptide: one with a cysteine residue at the N-terminus (M1C-N11) and one with such a residue at the C-terminus (E11C-N11). The two MTSL-labeled peptides were then mixed individually with ΔN11 Sso AcP at ratios ranging from 0:1 to 1.25:1 peptide/protein in 50 mM acetate buffer at pH 5.5, 25°C containing 20% (v/v) TFE. The ratios (I_diam/I_param) of the peak intensities of all residues of Sso AcP in the [¹H, ¹⁵N] HSQC spectra in the absence and presence of an equimolar amount of M1C-N11 and of E11C-N11 are shown in Fig. 4, A and B, respectively. Of importance, this experimental procedure enables a direct investigation of the interactions between N11 and ΔN11 Sso AcP in the absence of signals arising from possible intramolecular or intermolecular interactions involving only N11, which do not stimulate aggregation (14). We calculated the average value m of the observed I_diam/I_param ratio in each experiment and introduced a threshold value corresponding to m – 2σ. In each case, the region containing the highest density of peaks below this threshold value (i.e., the region where the highest paramagnetic relaxation due to interaction with the labeled peptide occurs) is that extending from the loop preceding β-strand 4 to the loop between β-strands 4 and 5 (six and seven peaks in Fig. 4, A and B, respectively; see also Fig. 4 C). Furthermore, M1C-N11, which has the spin label positioned at the N-terminus of the peptide, has the greatest relaxation effect at the C-terminus of β-strand 4 (Fig. 4 A), whereas E11C-N11, which carries the spin label at the C-terminus of the peptide, induces the greatest effects on the residues located at the beginning of β-strand 4 (Fig. 4 B). Therefore, these results show that N11 binds to β-strand 4 in an antiparallel manner.

Figure 4.

Ratio between amide peak heights measured from [¹H, ¹⁵N] HSQC spectra of ΔN11 Sso AcP in the presence of an equimolar amount of the MTSL-labeled M1C-N11 (A) and E11C-N11 (B), and the corresponding amide peak heights measured in the absence of either peptide (I_param/I_diam). Amide groups displaying a low ratio (lower than the average – 2σ) are highlighted in orange (A) and cyan (B); the threshold is depicted as a continuous line. Residues that could not be analyzed are depicted in gray; their values have been arbitrarily set to 1.0. (C) Structure of Sso AcP colored to show the residues whose I_param/I_diam values are below the threshold shown in panels A (orange) and B (cyan).

Detection of the intermolecular interactions by ITC

Another crucial question concerns the formation of the N^∗ state and whether it follows or precedes the establishment of the interaction between ΔN11 Sso AcP and N11. To answer this question and to obtain quantitative parameters describing the interaction, we performed ITC measurements in 50 mM acetate buffer with and without 20% (v/v) TFE. The results obtained in the presence of TFE (Fig. 5 A) could be fitted well to a binding model involving a 1:1 binding stoichiometry. The enthalpy change associated with the interaction is small (−0.44 Kcal mol⁻¹), whereas the entropy change is 0.024 Kcal mol⁻¹ K⁻¹, resulting in a free-energy change associated with the interaction (ΔG) of 7.6 ± 0.8 Kcal mol⁻¹ and a dissociation constant of the complex ΔN11 Sso AcP-N11 (K_D) of 2.3 ± 0.8 μM. By contrast, in the absence of TFE (Fig. 5 B) no binding could be detected, and the only heat change that was observed during addition of the peptide to the protein resulted from mixing of the solutions. These results confirm the existence of an interaction between the N-terminal segment and the globular unit of Sso AcP, and reveal the thermodynamic parameters of such an interaction. In addition, they show that the interaction of the N-terminal residues with the globular unit of the protein is a consequence of the addition of TFE, because the peptide is not able to bind to Sso AcP molecules in their native conformation due to the change in conformational properties resulting from the addition of TFE.

Figure 5.

ITC data for the interaction between ΔN11 Sso AcP and the N11 peptide in 50 mM acetate buffer at pH 5.5, 25°C, with (A) and without (B) 20% (v/v) TFE. In each panel, the upper graph shows the heat change over the time of the experiment and the lower graph shows the heat change per mole of injectant. The data in panel A have been fitted to a 1:1 bimolecular association model (continuous line).

Discussion

Previous studies (9,11,14) proposed that Sso AcP aggregates through the establishment of an intermolecular interaction between the unfolded N-terminal segment of one molecule and the globular unit of another molecule, possibly involving β-strand 4. This conclusion was drawn from a variety of evidence. First, it was found that ΔN11 Sso AcP, which lacks the 11-residue N-terminal segment, is not able to aggregate under conditions promoting aggregation of the wild-type protein, indicating that a major role in aggregation is played by this disordered region of the protein sequence (9). In addition, aggregation of wild-type Sso AcP was shown to be slower in the presence of N11, suggesting that this peptide competes with the intermolecular aggregation interaction (14). Furthermore, an inverse correlation was observed for a set of variants of Sso AcP between their conformational stability and the rate at which they form early aggregates. Only variants carrying single point mutations in β-strand 4 or in the N-terminal region of the protein were found to deviate significantly from this general correlation, suggesting a direct role in aggregation for residues located in these two segments (9). Also, although edge strands of proteins belonging to the acylphosphatase superfamily were observed to be protected against aggregation, in the case of Sso AcP, β-strand 4 has a relatively low level of protection, making this region more susceptible to intermolecular interactions than it is in other homologous proteins (15). Accordingly, mutations designed to increase the level of protection in the edge β-strand 4 inhibit native-like aggregation, indicating that this segment needs to unfold before forming aggregates (15). Finally, the region spanning β-strands 4 and 5 and the loop binding them is the region that shows the largest number of chemical shift changes upon formation of the initial aggregated species (5).

Despite this evidence, many details about the early steps of Sso AcP aggregation are still lacking. First, no direct structural information describing these interactions had been reported. Second, our understanding of the system did not include a characterization of the dynamics experienced by Sso AcP at the level of individual residues before aggregation. Third, it is unclear whether the interaction between the protein and the peptide is established through a conformational-selection mechanism or an induced-fit mechanism. In this work, however, we have provided quantitative information at the residue-specific level concerning the dynamics of the Sso AcP globular unit before aggregation, and we reveal direct evidence about the intermolecular interactions that lead to the formation of the initial aggregates formed by the protein. We obtained this information by studying the behavior of ΔN11 Sso AcP, a variant that lacks the 11-residue N-terminal segment but has the same globular unit as the wild-type protein and is found to resist aggregation under conditions that otherwise promote native-like aggregation of the intact protein.

The results show that when TFE is added to a concentration that stimulates rapid aggregation of Sso AcP, it results in the conversion of the protein into a N^∗ conformational state that, although native-like, is characterized by structural perturbations, particularly in the C-terminal region of the molecule, spanning residues 78–101. In the latter region, a substantial number of NMR chemical shift changes are observed under these conditions. Of importance, the NMR data do not indicate any significant degree of local unfolding. However, the simulations we carried out using HX PFs as experimental restraints provide an explanation for the different properties exhibited by N and N^∗. In particular, they show that in the presence of TFE, Sso AcP experiences a slight decrease in the number of native contacts throughout the sequence, a larger radius of gyration, and a slight increase in the magnitude of dynamical fluctuations (Fig. S5). These findings suggest that the aggregation-prone N^∗ state is best described as a folded but slightly expanded variant of the native state, with the highest degree of flexibility and solvent exposure being located at the C-terminus.

In this conformational ensemble, the increase in dynamics enables intermolecular interactions to be established with the N-terminal segment (here the N11 peptide) and the outer part of β-strand 4, resulting in an antiparallel arrangement between N11 and the edge strand. Rather than suggesting an induced-fit model, in which the N11 peptide induces a conformational change in the globular unit of the protein, our data point toward a conformational selection model for binding, in which the peptide can bind to Sso AcP molecules only if they populate the N^∗ state (Fig. S5). Indeed, population of N^∗ occurs even in the absence of the N11 peptide, and no chemical shift changes are observed upon binding to the peptide, suggesting that the latter does not significantly affect the structure of the globular unit. Moreover, ITC data indicate that binding does not occur in the absence of TFE, showing that Sso AcP molecules cannot bind the N-terminal segment when they populate the fully native state.

Our results regarding Sso AcP are particularly relevant when compared with models proposed for other proteins that aggregate from N^∗ ensembles. Aggregation from the native-like state has been found to occur for a number of proteins through a local unfolding event (23). Thus, for example, in human superoxide dismutase 1 (SOD1), the S134N mutation induces the dislocation of a specific electrostatic loop (24), and NMR studies showed that S134N SOD1 can form soluble oligomers through intermolecular interactions involving this loop (24). In human transthyretin, a variety of familial mutations lead to the unfolding of two peripheral β-strands (25), an event that exposes other regions of the protein structure that are normally buried, enabling intermolecular interactions to occur (23). In the case of human β₂-microglobulin, the protein becomes prone to form fibrils after a structural conversion that involves conformational changes in two peripheral β-strands and minor structural perturbations throughout the sequence (26). The same regions were found to be perturbed in the crystallographic structure (27).

In human lysozyme, a set of mutations induces a locally cooperative loss of native tertiary structure (28), with subsequent generation of an amyloidogenic state on the native side of the energy barrier for denaturation, that is characterized by unfolding of the β-sheet-rich domain while the rest of the molecule appears to be largely native-like (23). Moreover, this unfolded segment appears to form the core of fibrils (29). Recently, we showed that an acylphosphatase from Drosophila melanogaster aggregates after the displacement of a single peripheral β-strand, corresponding to β5 of Sso AcP, and this dislocation leads to the exposure of a highly amyloidogenic segment (16).

Our data on Sso AcP suggest a different type of behavior: A widespread increase in dynamical fluctuations occurs upon formation of the N^∗ state in the absence of a significant global or local unfolding and exposure of hydrophobic regions. In this conformational ensemble, β-strand 4 undergoes fluctuations of sufficient magnitude to allow formation of new interactions with the highly flexible and solvent exposed N-terminal segment. Thus, our findings could have important consequences for understanding aggregation in vivo and for biotechnological purposes. Indeed, although globular proteins spend the majority of their lifespan in a folded conformation, an increase in dynamics in the absence of local unfolding may trigger their self-assembly.

Conclusion

In conclusion, we have shown that the native-like, aggregation-prone state of Sso AcP is a slightly expanded version of the fully native state, lacking local unfolding but exhibiting larger dynamical fluctuations that enable the protein to establish amyloidogenic interactions. Thus, the aggregation-prone N^∗ state of a protein can be (as in the various examples described above) an ensemble of locally unfolded conformations or (as in the case of Sso AcP) a fully folded state that experiences increased dynamical properties relative to the fully folded state. Thus, our study provides important information about the dynamical properties of a native globular protein in solution, and the manner in which these properties can affect aggregation that is initiated from such a state.