Dynamics of Amyloid β Fibrils Revealed by Solid-state NMR*

Background: Alzheimer disease is the most important neurodegenerative disorder; treatment approaches require atomistic knowledge of fibrillar structure and dynamics. Results: We have site-specifically studied the molecular dynamics of amyloid β (Aβ) fibrils by solid-state NMR. Conclusion: The β-sheet motifs of Aβ are essentially rigid, and the termini exhibit more flexibility. Significance: Dynamics studies of Aβ fibrils suggest a structural role of the N terminus of the peptide. We have investigated the site-specific backbone dynamics of mature amyloid β (Aβ) fibrils using solid-state NMR spectroscopy. Overall, the known β-sheet segments and the turn linking these two β-strands exhibit high order parameters between 0.8 and 0.95, suggesting low conformational flexibility. The first approximately eight N-terminal and the last C-terminal residues exhibit lower order parameters between ∼0.4 and 0.8. Interestingly, the order parameters increase again for the first two residues, Asp1 and Ala2, suggesting that the N terminus could carry some structural importance.

Alzheimer disease represents the most widespread neurodegenerative disease affecting particularly the countries with high life expectancy. The existence of extracellular deposits of amyloid fibrils formed from amyloid ␤ (A␤) 2 peptides represents a hallmark of the disease. These deposits constitute the end product of a complicated aggregation pathway that initiates with A␤ monomers, which are the result of the action of several enzymes on the amyloid precursor protein. The folding of A␤ monomers into mature A␤ fibrils occurs via transient oligomeric and protofibrillar states, which are considered to be the toxic intermediates in the disease.
Detailed structural data on mature A␤ fibrils have been provided by solid-state (1)(2)(3)(4)(5) and solution (6, 7) NMR spectroscopy, as well as by cryo-electron microscopy (8,9). It is well established that A␤ fibrils exhibit a significant amount of structural polymorphism (10 -13). Part of this polymorphism is due to preparation conditions, but surprisingly, most structural studies seem to agree on a general U-shaped ␤-strand-turn-␤strand motif of the A␤ peptides in fibrils, with some variation in the register of the opposing ␤-strand zippers (5,14).
In addition, NMR structural data are now available for various A␤ aggregation intermediates, including protofibrils (15) and oligomers (16 -18). Interestingly, the molecular dynamics of A␤ fibrils, which is an essential part of the structural biology of amyloid structures, has not been comprehensively studied. In fact, only a very few reports on fibrillar dynamics in general exist so far (15,19,20). These studies showed that the investigation of the fibril dynamics can significantly support the structural analysis and further characterize the respective structural elements. The combination of structural and dynamical analysis leads to a much more complete picture of the structural biology of the fibrils and therefore improves the understanding of the fibrillation process and the mode of action of the investigated structures. Here, we have studied the molecular dynamics of mature A␤(1-40) fibrils in a site-specific manner using solid-state NMR spectroscopy.

EXPERIMENTAL PROCEDURES
Sample Preparation-For this investigation, eight A␤(1-40) peptides with uniformly 13 C/ 15 N-labeled amino acids in different positions were synthesized using standard N-(9-fluorenyl)methoxycarbonyl protocols. The individual A␤(1-40) peptides with sequence DAEFRHDSGY EVHHQKLVFF AED-VGSNKGA IIGLMVGGVV were isotopically labeled at the following amino acids: peptide I, Val 12 , Phe 20 , Ala 30  A␤(1-40)-labeled peptide was solubilized in 50 mM sodium borate buffer (pH 9) at a concentration of 6 mg/ml. The sample was seeded and incubated at 37°C for 1 week. Seeds consisted of mature A␤(1-40) fibrils previously grown and seeded under the same conditions (second generation) and were sonicated for 10 min before addition to the sample. The presence of mature fibrils was confirmed by transmission electron microscopy (TEM). Mature fibrils were recovered by ultracentrifugation at 100,000 rpm for 2 h at 4°C using a TLA 120. 2  liquid nitrogen and thawing it at 37°C. TEM samples were prepared by applying 5-ml droplets from the sample after 1:10 dilution with pure water onto a carbon film (floating carbon method), counterstained with 2% (w/v) uranyl acetate, and analyzed with a Zeiss 900 electron microscope (80 kV). 13 C Magic Angle Spinning (MAS) NMR Measurements-The 13 C cross-polarized MAS NMR spectra were measured on a Bruker AVANCE 750 spectrometer (Bruker BioSpin GmbH, Rheinstetten, Germany) operating at a resonance frequency of 749.7 MHz for 1 H and 188.5 MHz for 13 C. A 4-mm doubleresonance MAS probe was used. The cross-polarized contact time was 700 s; typical lengths of the 90°pulses were 5 s for 13 C and 4 s for 1 H. For heteronuclear two-pulse phase modulation decoupling, a 1 H radiofrequency field of 65 kHz was applied. 13 C chemical shifts were referenced externally relative to TMS. Two-dimensional 13 C-13 C proton-driven spin exchange spectra with a mixing time of 50 ms were acquired for peak assignment (21).
Constant time dipolar coupling and chemical shift (DIP-SHIFT) experiments using frequency-switched Lee-Goldberg homonuclear decoupling (80-kHz decoupling field) were carried out to measure the 13 C-1 H dipolar couplings (22). DIP-SHIFT experiments were conducted as two-dimensional measurements, where each increment in the indirect dimension refers to one-time increment to sample the dipolar dephasing curve of the respective backbone signal. Spectra were only Fourier-transformed in the direct dimension, allowing us to directly read off the intensities of the dipolar dephasing curve in the indirect dimension for each resolved backbone signal. Thus, a single two-dimensional DIPSHIFT experiment provides the complete dipolar dephasing curve for each resolved carbon (23). After simulating the dipolar dephasing curve over one rotor period, the order parameter was derived by dividing the determined coupling by the known rigid limit (24,25). All experiments were carried out at a temperature of 303 K and a MAS frequency of 7 kHz.

RESULTS
A␤(1-40) fibrils were grown in 50 mM sodium borate buffer at pH 9. To assess the morphological homogeneity of the samples used in this study, Fig. 1 shows typical TEM pictures of four different preparations. TEM pictures of all eight peptide preparations are shown in supplemental Fig. S1.
This choice of buffer system provided the most homogeneous A␤ preparations as can be assessed from the TEM pictures. We also compared the NMR spectra of A␤(1-40) fibrils grown in sodium borate buffer at pH 9 with fibrils grown in sodium phosphate buffer at pH 7.4 (data not shown). The NMR spectra from both fibril preparations showed signals with identical chemical shifts; however, the fibrils prepared in borate buffer provided slightly narrower lines in the NMR spectra. Therefore, we continued our investigations with this system.
To study the molecular dynamics of A␤ fibrils, we carried out solid-state NMR experiments. The DIPSHIFT pulse sequence (22) allowed us to measure the molecular order parameters of mature A␤(1-40) fibrils prepared from the eight differently 13 C/ 15 N-labeled peptides in a site-specific manner. From this measurement of the backbone 13 C-1 H dipolar couplings, we determined the backbone order parameters to characterize the amplitude of motion for the C␣-H␣ bond vectors. A fully rigid C-H bond would exhibit the maximal dipolar coupling strength of 22.8 kHz, corresponding to an order parameter of 1, whereas a value of 0 for the order parameter corresponds to fully isotropic motion expressed by a vanishing dipolar coupling. Molecular motions with a given amplitude lead to partial averaging of the dipolar coupling and can be characterized by a specific order parameter. The 13 C-1 H order parameters sample all motions with correlation times shorter than ϳ10 s. A plot of the backbone order parameter in mature A␤ fibrils at a hydration level of 50 weight % for the C␣ atoms calculated from the measured couplings is shown in Fig. 2. Overall, the order parameters for the mature A␤ fibrils are rather high and never drop below 0.4. Typical dipolar dephasing curves of the MAS signals used for the determination of these order parameters of the A␤ preparations are given in supplemental Fig. S2. For Ile 32 and Val 40 , two slightly different order parameters were determined from the NMR signals that exhibited two sets of isotropic chemical shifts. This corresponds to some structural heterogeneity within the fibrils that has already been observed in previous studies (2,3). We also carried out DIPSHIFT measurements on fibrils prepared at a hydration level of 75 weight % (data not shown). Within experimental error, no differences in the order parameters at these two hydration levels were found.
To correlate these order parameters with secondary structure motifs in A␤ fibrils, we carried out 13 C-13 C correlation experiments and analyzed the isotropic chemical shifts of the 13 C MAS NMR spectra of the individual A␤ preparations. Typical 13 C MAS NMR spectra of our preparations are shown in supplemental Fig. S3. A plot of the secondary chemical shifts (i.e. the chemical shift deviation of a given residue from the random coil values) presented in Fig. 3 shows random coil-like chemical shifts at the peptide N terminus (residues 1-4 and 6 -9). In addition, we found two ␤-sheet regions, comprising residues 11-22 and 30 -38, which are connected by a segment with random coil-like chemical shifts, involving residues 23-29. Our isotropic chemical shift data (supplemental Table S1) agree well with existing literature values obtained with other samples of mature A␤(1-40) fibrils (1)(2)(3)(4)(5). This suggests a very similar secondary structure of the fibrils grown in borate buffer compared with the known models. We further note that the region connecting the two ␤-strands exhibits secondary chemical shifts that are highly variable and show multiple sets of chemical shifts at several positions (supplemental Table S1), which was also observed in previously reported A␤(1-40) fibril preparations (2, 3). Collectively, these data indicate that the intervening loop connecting the two ␤-sheets is affected by significant structural polymorphism. As previous chemical shift measurements were carried out with samples that were obtained under significantly different fibrillation conditions and procedures, the presently observed similarities confirm that A␤ peptides adopt a rather well conserved peptide fold even when fibrillated under somewhat different physicochemical conditions.
Comparison of the order parameters determined in this study with the identified secondary structural elements shows that the two ␤-sheet regions, along with the intervening loop, exhibit relatively high order parameters of 0.8 -0.95. Such values are indicative of a very rigid peptide backbone structure. Such high order parameters would correspond to segmental motions with amplitudes of ϳ15-30°. This is also consistent with previous evidence that a salt bridge between Asp 23 and Lys 26 (14,26) stabilizes this short connection and prevents larger backbone motions. The high order parameter values observed for the two central ␤-sheet regions also correspond well with literature data on the order of PABPN1 fibrils (19) and fibrils of the human prion protein (20).
Slightly lower order parameters (Յ0.8) were obtained at the peptide N and C termini (residues 1-4, 7, and 40). This finding indicates that the terminal ends of the mature A␤ fibrils are less well structured than the more central regions.
A particularly striking result of our data set is the fact that the first two residues in the A␤ sequence (Asp 1 and Ala 2 ) exhibit significantly higher order parameters than the subsequent amino acid (Glu 3 ). This appears to be a very unique feature of the A␤ fibrils that is very unusual and generally not known for soluble proteins.
We have also determined the line width of the NMR signals in the respective spectra of A␤(1-40) fibrils (Fig. 4). Line widths represent a simple measure of the molecular dynamics and have also been analyzed for A␤ fibrils previously (1). The line widths for the labeled residues typically vary between ϳ1 and 3 ppm, with only Ser 8 showing a value of ϳ4.5 ppm. There was no clear trend for in increase in the line width toward the N terminus.

DISCUSSION
We have studied the molecular dynamics of mature A␤(1-40) fibrils by solid-state NMR spectroscopy. 13 C MAS NMR chemical shifts detected for the A␤ fibrils studied here agreed well with what has been observed in the literature (1)(2)(3)(4)(5). In addition, we have presented a site-specific study of the molecular dynamics of the C␣ backbone carbons. The dynamics data are presented as order parameters that provide information about the motional amplitude of the C␣-H bond vector in the protein backbone. A schematic projection of the order parameter values onto the known topology of A␤ fibrils is given in Fig. 5. Collectively, these data indicate the existence of signifi-   cant structural stability at the peptide N terminus, which specifically affects the first two residues.
At first sight, the motional analysis of the mature A␤ fibrils reveals no significant surprises. The two central ␤-strand elements exhibit high order parameters; molecular order is somewhat decreased in the connecting loop region and the N and C termini, respectively.
However, two results of the motional analysis are particularly noteworthy; both concern the N terminus of the mature A␤ fibrils. First, the measured order parameters for the N terminus are still significantly higher than those associated with a purely thermally fluctuating random coil (which is typically ϳ0.1-0.3 even under solid-state conditions) (19). Moreover, such low order parameters would be expected to abolish or to strongly attenuate the cross-polarization signal of these protein parts (27), which was not observed here. Perhaps the tendency of A␤ peptides to form an additional ␤-strand for amino acids 4 -7, which was also observed in some structural studies by solidstate (5) and solution (7) NMR or molecular dynamics simulation (28), can explain the absence of high molecular dynamics in this region. Second, the fact that the order parameters of the first two residues (Asp 1 and Ala 2 ) are actually higher than for Glu 3 suggests that the N terminus represents a somewhat structurally confined segment that may also be stabilized in the course of peptide aggregation.
The molecular dynamics of mature A␤ fibrils has not been systematically studied so far. Therefore, a comparison of our results with the literature can only be done on the basis of a 13 C MAS NMR line width measurement carried out previously (1). This analysis reported large line widths of 3-5 ppm for the first ϳ10 N-terminal residues, whereas the remainder of the sequence showed line widths of ϳ2 ppm. This was interpreted as a structurally disordered N terminus. Although the line widths for our preparations of A␤ fibrils were rather homogeneous between 1 and 3 ppm over the entire peptide sequence (Fig. 4), we measured consistently lower order parameters for the N-terminal residues compared with the ␤-sheet and turn regions. Although the NMR line widths reflect both static structural heterogeneity and the influence of molecular dynamics, order parameters depend only on the fast protein dynamics.
This supports the view that the N terminus of the A␤ fibrils is indeed more mobile than the rest of the protein; however, it does not represent a freely fluctuating polypeptide chain, which would have expressed lower order parameters.
A recent molecular dynamics simulation showed that the conformational flexibility in the loop region of the mature A␤ fibrils is not significantly increased in comparison with the two opposing ␤-sheets (29), which is also reflected in our experimental work. In the simulation, the conformational dynamics of the residues in the A␤ sequence increased toward the N terminus, starting at about Val 12 . Unfortunately, the first eight residues in the A␤ sequence were omitted in the simulation, so no comparison with the interesting dynamical behavior of these amino acids in the sequence observed in our experiments is possible.
Finally, we compared our data with a recently published study on the structure and dynamics of A␤ protofibrils (15). The order parameter profiles for the two A␤ species show a high degree of agreement for most of the amino acids (supplemental Fig. S4), and deviations are mostly within the experimental error. Only Val 12 and Lys 16 show slightly higher order parameters in mature fibrils compared with protofibrils. This is probably caused by the reduced length of the first ␤-sheet in protofibrils (15), which begins at residue 16 in protofibrils but already at residue 10 in mature A␤ fibrils (1,5). At the N terminus, the order parameters of the first two residues in protofibrils are again larger than for residue 3, underlining the putative importance of the N terminus in peptide aggregation and fibril formation. In fact, Asp 1 exhibits yet a higher order parameter of 0.75 in A␤ protofibrils compared with 0.65 in mature A␤ fibrils. Also Phe 4 has a higher value of 0.79 in protofibrils compared with our data on mature A␤ fibrils (0.61). Altogether, these results may suggest that the N terminus of the A␤ peptides could play a role in the maturation of A␤ fibrils. This is particularly interesting as, structurally, protofibrils show a stronger resemblance to oligomers than to mature A␤ fibrils (15). Dynamically, however, protofibrils are actually very similar to the mature A␤ fibrils, with the most pronounced differences in the N-terminal region. A next important step would certainly be a comprehensive dynamical characterization of A␤ oligomers.
In summary, our data provide a comprehensive description of the fast molecular dynamics of mature A␤ fibrils, showing that the two ␤-sheets and the turn region linking these two structural elements are essentially rigid and well ordered. The termini undergo somewhat more dynamic reorientation; however, the first two N-terminal amino acids are again slightly more ordered, suggesting that this protein part could actually play a role in the structural biology of A␤ fibrils. Although it is known that the N terminus of A␤ peptides does not play an important role in fibril growth (1,30), the current data suggest that this protein part is less mobile than typically assumed and might be worth some additional attention.