Kinetic analysis of protein aggregation monitored by real-time 2D solid-state NMR spectroscopy
- 727 Downloads
It is shown that real-time 2D solid-state NMR can be used to obtain kinetic and structural information about the process of protein aggregation. In addition to the incorporation of kinetic information involving intermediate states, this approach can offer atom-specific resolution for all detectable species. The analysis was carried out using experimental data obtained during aggregation of the 10.4 kDa Crh protein, which has been shown to involve a partially unfolded intermediate state prior to aggregation. Based on a single real-time 2D 13C–13C transition spectrum, kinetic information about the refolding and aggregation step could be extracted. In addition, structural rearrangements associated with refolding are estimated and several different aggregation scenarios were compared to the experimental data.
KeywordsAggregation Kinetic Solid-state NMR Real-time spectroscopy Crh
Protein aggregation has become recognized as an important aspect of the protein folding landscape. This process not only interferes with protein expression and recovery assays used in biotechnology, but also impacts the every day live of cells and organisms and is associated with a variety of human diseases (Dobson 2003; Wetzel 2006). Methods to characterize aggregation kinetics comprise predominantly biochemical and biophysical approaches such as gel filtration, sedimentation assays, binding of fluorescent markers, AFM imaging, dynamic light scattering or circular dichroism that report on the oligomerisation state or the overall secondary structure content of the protein (Hurshman et al. 2004; Wetzel 2006).
Real-time solution-state NMR has shown to be useful to follow protein and RNA folding at the level of individual residues (Balbach et al. 1996; Van Nuland et al. 1998; Zeeb and Balbach 2004; Corazza et al. 2010; Lee et al. 2010). In principle, solid-state NMR (ssNMR) offers a complementary spectroscopic means to probe structural and kinetic aspects of protein folding and aggregation, especially as molecular aggregates increase in size. Indeed, ssNMR, has made great progress to structurally study trapped intermediate states of Amyloid (Chimon et al. 2007; Ahmed et al. 2010) and globular (Hu and Tycko 2010) proteins or to examine the effect of protein mutations known to interfere with protein aggregation (Heise et al. 2008; Karpinar et al. 2009; Kim et al. 2009). In principle, kinetic information becomes accessible by repeating ssNMR on cryotrapped intermediates at different time points or by recording NMR data directly during refolding. Indeed, time-resolved 1D ssNMR has been used to detect signal intensity buildup or decay during protein aggregation (Kamihira et al. 2000) and ATP hydrolysis of an ABC transporter (Hellmich et al. 2008) or by combining ssNMR pulse schemes with diffusion measurements using PFGs (Ader et al. 2010). Due to limited spectral resolution 1D ssNMR has to be combined with specific labeling techniques to offer site-specific resolution in proteins. Earlier, we have shown (Etzkorn et al. 2007) that for the Crh protein from B. subtilis, molecular aggregation triggered by a small temperature jump can be followed by two-dimensional ssNMR. Starting from a kinetically destabilized protein precipitate, protein aggregation led to significant increase in β-sheet content, whereas smaller α-helical fragments were retained in the aggregated state. Using Crh as an example, we here demonstrate that 2D ssNMR data sets recording these structural rearrangements in real time offer structural and, in particular, kinetic information about the process of protein aggregation. In general our analysis can offer atom-specific resolution for large segments of the protein and can simultaneously detect and kinetically describe a range of possible intermediate states during protein aggregation under conditions where molecular size or density prohibit the application of other biophysical methods.
Theory, materials and methods
For analysis relevant 1D extracts of real-time 2D spectra are usually fitted to theoretically simulated peak shapes (Balbach et al. 1996; Helgstrand et al. 2000) or a direct analytical solution (Balbach et al. 1999; Zeeb and Balbach 2004). Due to increased spectral overlap in ssNMR, we used (2) to develop a Mathematica (Wolfram scientific) script to calculate the difference of a full theoretical 2D cross peak pattern and the experimental data. Different transition scenarios were examined by varying the mathematical description of the kinetic profile. Amplitude factors a n were set to one, implying the same transfer efficiency for all states. Free theoretical parameters such as λ (1;2),n and ω (1;2),n were chosen according to the line width and position of the peak maxima in the experimental spectrum.
Simulating real-time 2D spectra
Mathematica (Wolfram scientific) version 6.0.1 was used to fit the experimental data to simulated peak patterns by numerically integrating (FT) S(t 1, ω 2) (2) for the population profiles considered. ‘SymbolicProcessing’ was switched off to speed up the integration process. The experimental spectrum was processed using exponential line broadening of 50 Hz. The underlying window function was also implemented in the simulations before FT in the indirect dimension. The effect in the direct dimension was neglected. Resonance frequencies of the occurring states were taken from corresponding cross signals in the experimental spectrum. The line width was measured using the Thr Cβ–Cγ2 cross peaks, which in the real-time 13C–13C 2D spin diffusion spectrum are not symmetric to the Thr Cγ2–Cβ peaks. Here the line width in the direct dimension is resolved and should be largely unaffected by the transition. The line width was comparable for all states, hence justifying in part the assumption a n = 1 for all n.
Calculating difference plots
Experimental data were taken from Etzkorn et al. (2007). Signal intensity in the simulated spectra was read out at the corresponding data points obtained by processing the whole experimental spectrum with 512 points in ω 1 and 2,048 in ω 2. The spectral extract of the Thr Cγ2–Cβ (Cβ–Cγ2) cross section consists of 31 × 21 (21 × 31) values which were treated individually. Free parameters of the kinetic profile were varied in nested loops.
To evaluate the subset of the torsion angle space that is in agreement with the experimentally observed Thr Cβ shift we generated a set of heptapeptides (AATAA) by varying the Thr ψ and ϕ torsion angle in steps of 10°. ShiftX (Neal et al. 2003) was used to predict the expected chemical shift for each peptide. Chemical shift intervals [68.4, 70 ppm], [69.7, 70.8 ppm] and [70.4, 72.5 ppm] were chosen to select a match for the states A, B and C, respectively.
Crh in a classical three-state folding transition
Figure 3b shows the difference plot for the Thr Cβ–Cγ2 cross correlations. The minimum is found for k 1 = 0.77 × 10−4 s−1 and k 2 = 0.52 × 10−4 s−1. Combination of data from both sides of the diagonal leads to the difference plot shown in Fig. 3c. The respective population profiles according to the best fit rate constants allow for an estimation of the accuracy of the method. The rate constants obtained from Fig. 3c are k 1 = 0.91 × 10−4 s−1 and k 2 = 0.65 × 10−4 s−1. Notably, the comparison of the theoretical spectra to the experimental ones (Fig. 3g, h) shows that characteristic features such as line broadening for the initial as well as baseline distortions for the final state are significantly less reproduced by the experimental data than expected for the considered single exponential transition.
Crh in a classical aggregation scenario
Notably, a comparison to the best fit according to a single exponential three-state transition (Fig. 3a, b, lower part), reveals that the assumption of a nucleation step here did not improve the fitting. Indeed the line broadening of the initial state is predominantly related to the unfolding mechanism and, as evident from Fig. 4g, h, not reproduced even by the best fit of the profiles represented by (6a–6c).
Crh in a stretched exponential unfolding scenario
Notably, a completely heterogeneous aggregation scenario also involving a stretched exponential aggregation step as suggested for transthyretin aggregation (Hurshman et al. 2004) could also explain the data (see supporting figure SI 1). However, experimental data analyzed here exclusively report on the population profile. Additional studies on the effect of concentration of native Crh and of seeding with preaggregated Crh may help to discriminate between stretched exponential aggregation, i.e. heterogeneous growth of aggregates with the monomer as the critical nucleus size, and the F–W model, i.e. continuous nucleation followed by an autocatalytic surface growth (Ferrone 1999; Hurshman et al. 2004).
Several methods exist that allow studying protein aggregation kinetics. Most of these techniques rely on the detection of global factors or indirect mechanisms (e.g. the β-strand content in Thioflavin T binding or CD spectroscopy) and may lack the formation and kinetic properties of intermediate species. Here we have shown that ssNMR can contribute to the kinetic and structural analysis of insoluble protein conformations, in particular of those including intermediate folding states which might be an important target to interfere with the aggregation process (Cohen and Kelly 2003).
While intermediate states might be rapidly frozen for a more detailed structural study (Chimon et al. 2007; Ahmed et al. 2010; Hu and Tycko 2010), real-time ssNMR offers unique possibilities to characterize aggregation kinetics. The range of observable states can be modified by a combination of different sets of polarization transfer mechanisms (e.g. based on dipolar or scalar couplings) (Andronesi et al. 2005) or by simply recording ssNMR spectra after direct excitation (Kamihira et al. 2000). Our analysis was solely based on ssNMR cross-signal intensities from a single kinetic transition to investigate the potential of the method. Changes in backbone conformation were estimated from an analysis of conformation-dependent chemical shifts. Rate constants could be extracted using a single exponential three-state transition as well as a conventional aggregation mechanism. No significant difference between the two approaches could be detected, suggesting that the formation of a nucleus is not a significant step in the aggregation of Crh protein precipitates. Remaining differences to the experimental data additionally suggest that a single exponential transition for the initial unfolding step does not suffice to properly describe the time course of the folding process. Instead, a stretched exponential function, as found in downhill folding (Sabelko et al. 1999; Nakamura et al. 2004), significantly improves the agreement between experimental and theoretical data. The stretched exponential decay would be consistent with a heterogeneous unfolding step involving one or several fast as well as slow decaying components.
While limitations regarding sensitivity and resolution do not allow for a more detailed analysis of the aggregation mechanism from ssNMR at this stage, ssNMR studies using selectively labeled protein variants and the combination with other biophysical methods will provide additional opportunities to refine kinetic and structural profiles. Such studies may not only be relevant in the context of molecular aggregation but their application may also facilitate an atomic description of inter-molecular interactions in the context of molecular gel formation (Ader et al. 2010) or protein insertion into membranes.
This work was funded by NWO (Grant number 700.26.121), the CNRS (PICS no. 2424), the French research ministry (ACI Biol. Cell. Mol. Et Struct. 2003; ANR JCJC 2005; ANR-PCV08_321323), and the Max Planck Gesellschaft.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
- Ahmed M, Davis J, Aucoin D, Sato T, Ahuja S, Aimoto S, Elliott JI, Van Nostrand WE, Smith SO (2010) Structural conversion of neurotoxic amyloid-beta(1-42) oligomers to fibrils. Nat Struct Mol Bio 17:556–561Google Scholar
- Corazza A, Rennella E, Schanda P, Mimmi MC, Cutuil T, Raimondi S, Giorgetti S, Fogolari F, Viglino P, Frydman L, Gal M, Bellotti V, Brutscher B, Esposito G (2010) Native-unlike long-lived intermediates along the folding pathway of the amyloidogenic protein beta(2)-microglobulin revealed by real-time two-dimensional NMR. J Biol Chem 285:5827–5835CrossRefGoogle Scholar
- Karpinar DP, Balija MBG, Kugler S, Opazo F, Rezaei-Ghaleh N, Wender N, Kim HY, Taschenberger G, Falkenburger BH, Heise H, Kumar A, Riedel D, Fichtner L, Voigt A, Braus GH, Giller K, Becker S, Herzig A, Baldus M, Jackle H, Eimer S, Schulz JB, Griesinger C, Zweckstetter M (2009) Pre-fibrillar alpha-synuclein variants with impaired beta-structure increase neurotoxicity in Parkinson’s disease models. EMBO J 28:3256–3268CrossRefGoogle Scholar