Membrane-Induced Dichotomous Conformation of Amyloid β with the Disordered N-Terminal Segment Followed by the Stable C-Terminal β Structure

Various neurodegenerative disorders are ascribed to pathogenic molecular processes involving conformational transitions of amyloidogenic proteins into toxic aggregates characterized by their β structures. Accumulating evidence indicates that neuronal cell membranes provide platforms for such conformational transitions of pathogenic proteins as best exemplified by amyloid β (Aβ). Therefore, membrane-bound Aβ species can be promising targets for the development of novel drugs for Alzheimer’s disease. In the present study, solid-state nuclear magnetic resonance spectroscopy has elucidated the membrane-induced conformation of Aβ, in which the disordered N-terminal segment is followed by the stable C-terminal β strand. The data provides an insight into the molecular processes of the conformational transition of Aβ coupled with its assembly into parallel β structures.


Introduction
Various neurodegenerative disorders are ascribed to pathogenic molecular processes involving conformational transitions of amyloidogenic proteins into toxic aggregates characterized by their β structures [1,2]. Amyloid β (Aβ) is a major player in the onset and development of Alzheimer's disease (AD) [3]. The major isoform of this protein is composed of 40 or 42 amino acid residues, which is proteolytically cleaved from its precursor membrane protein. Aβ has a high propensity to form cross-β-fibrils [4][5][6][7][8][9][10][11], which are supposed to be a major component of senile plaque as a hallmark of AD. Accumulating evidence indicates that neuronal cell membranes provide platforms for conformational transition and subsequent aggregation of pathogenic proteins, including Aβ [12]. In particular, the formation of toxic Aβ aggregates is facilitated in membrane environments containing GM1 ganglioside, a glycosphingolipid abundant in neuronal cell membranes [13][14][15]. Therefore, membrane-bound Aβ species can be promising targets for the development of novel anti-AD drugs.
To characterize the membrane-bound conformation of Aβ, spectroscopic approaches, including solution nuclear magnetic resonance (NMR), circular dichroism (CD), and Fourier transform infrared spectroscopy (FT-IR), have been attempted by employing membrane mimics typified by aqueous micelles and even organic solvents [16][17][18][19][20]. The data thus obtained have underscored the formation of α-helical structures of Aβ in membrane-like environments. However, it has been suggested that the Aβ conformation is influenced by the size of micelles and the Aβ density thereon [21]. For example, Aβ has been reported to exhibit a thioflavin Treactive β structure under conditions where Aβ density on micelles is high [21], although detailed structural information remains unavailable. These data emphasize the necessity of the conformational characterization of Aβ bound to large vesicles as more realistic model membranes. However, the membrane-bound Aβ molecules are not accessible with solution NMR techniques because of its slower molecular tumbling. In contrast, solid-state NMR studies have elucidated various structures of Aβ fibrils formed in solution [5,6,8,9,11] and membrane-associated Aβ fibrils [10,[22][23][24], and also membrane perturbations caused by interactions of Aβ with lipid bilayers [25][26][27][28]. However, few detailed structural data have been available for the membrane-bound Aβ molecules prior to formation of amyloid fibrils.
Hence, in this study, we applied solid-state NMR spectroscopy to characterize the conformation of Aβ  bound to multilamella vesicles (MLVs) composed of 1,2-dimyristoyl-snglycero-3-phosphocholine (DMPC).  [29] homonuclear correlation experiment at 20°C. We observed 22 cross peaks, indicating that the Aβ(1-40) molecule bound to the DMPC membrane has a heterogeneous nature in terms of conformational variability. To identify the well-ordered region (s) of the membrane-induced Aβ(1-40) molecule, we attempted to assign the observed cross peaks based on through-bond connectivities. Sequential signal assignments were achieved by the 13 C-CTUC-COSY homonuclear correlation experiment in conjunction with 13 C observed NCO and NCA heteronuclear correlation experiments using 13 C-15 N double cross polarization (DCP) [30] as shown in Fig 1B and 1C. The peak assignments of Aβ  bound to DMPC lipid bilayers are summarized in Table 1. The results indicate that the observed peaks originated exclusively from the C-terminal segment of Aβ(1-40), i.e., Val24-Val39. The conformational-dependent isotropic chemical shift values of the peaks thus assigned were inspected using TALOS-N [31], revealing that the C-terminal segment exhibited a β structure (S1 Fig). In general, the β strands of Aβ peptides are stabilized through either intra-or intermolecular hydrogen-bonding interactions [4][5][6][7][8][9][10][11]14,[32][33][34][35][36]. For example, it has been suggested that SDS can induce Aβ(1-42) oligomerization forming β-strands, Val18-Asp23 and Lys28-Gly33, which are arranged into strand-turn-strand structure [35]. Furthermore, an Aβ(1-40) oligomer induced by epigallocatechin-3-gallate adopts an intramolecular anti-parallel β-sheet formed between Glu22-Asp23 and Gly29-Val39 [36]. However, β-strand conformation has been continuously observed for the Gly25-Val39 segment as mentioned in present study (S1 Fig). Thus, the possibility of an anti-parallel β-sheet structure through intramolecular hydrogen bonds at this C-terminal regions was excluded. Moreover, previous 2 H and 31 P NMR data have demonstrated that Aβ(1-40) was not inserted in neutral lipid bilayers [26,27]. Based on these data, we concluded that the Aβ(1-40) molecules bound to DMPC lipid bilayers exhibit dichotomous conformation, in which the disordered N-terminal segment is followed by the stable C-terminal β structure. The remaining possibilities of the β-strand structures due to intermolecular hydrogen bonding, i.e., parallel and/or anti-parallel β-sheet arrangements, were inspected using dipolar-assisted rotational resonance [37]/RF-assisted diffusion [38] (DARR/RAD) experiments. Huang et. al. indicated that, in 150-kDa oligomers of Aβ(1-42), its C-terminal β-strand is arranged into an intermolecular antiparallel β-sheet based on detection of inter-residue cross peaks in DARR spectra [32]. To clarify intermolecular proximity among Aβ(1-40) peptides in the DMPCinduced structure, we employed DARR/RAD experiments at various mixing times as shown in Fig 2 and S2 Fig. The 13 C-homonuclear through-space DARR/RAD correlation spectra with mixing time up to 400 ms gave only intra-residue cross peaks and inter-residue cross peaks between adjacent residues as a hallmark of the anti-parallel β structure, suggesting that the Cterminal segment of Aβ(1-40) forms a parallel β structure in the membrane environment.

Results and Discussion
The previously reported NMR studies have demonstrated the conformational versatility of the Aβ(1-40) segments. The free form of Aβ(1-40) is unstructured in an aqueous solution but exhibits an α-helical conformation at Gln15-Asp23 and Ile31-Met35 in organic solvents such as trifluoroethanol [18]. Upon binding to aqueous micelles composed of sodium dodecyl sulfate (SDS) or lyso-GM1, Aβ(1-40) forms discontinuous α-helices at His14-Val24 and Ile31--Val36 under micelle excess conditions [20]. In contrast, the C-terminal hydrophobic dipeptide segment shows two distinct conformational states that are reactive with thioflavin T in GM1 micelles under conditions where the Aβ density on the micelles is high [21]. In the typical mature cross-fibrils, Aβ(1-40) molecules adopt a strand-turn-strand motif at Tyr10-Val24 and Ala30-Val39 and assemble into parallel β sheets [8], although Aβ fibril exhibits polymorph because of the variety in the biological environments [10,39], not only because of the experimental conditions. Niu et. al. has recently reported solid-state NMR study for structure of Aβ(1-40) fibrils formed in the presence of lipid vesicles indicating significant structural differences from the fibrils formed in solution [24]. Furthermore, Qiang et. al. has recently reported solid-state NMR study characterizing the formation of a complex between membrane-associated Aβ(1-40) fibrils and lipids, which contains a typical β-loop-β motif similar to the mature fibril, suggesting that formation of such complex could be associated with lipid uptake and possibly membrane disruption [22]. On the other hand, CD and 2 H and 31 P NMR data have indicated that Aβ(1-40) could form either an α-helical structure or a β-structure depending on molar ratios of neutral/negatively charged lipids as well as the peptide/lipid molar ratios [26,27]. However, despite these cumulative observations, α-to-β conformational transition processes of Aβ  remain largely unknown at the atomic level. The present solid-state NMR study identified a partially ordered conformation of membrane-induced Aβ(1-40) molecules, in which only the C-terminal segments are involved in a parallel β structure, while leaving the N-terminal segment disordered (Fig 3). The stabilized C-terminal segment encompasses the C-terminal half of the strand-turn-strand structure in the amyloid fibrils. Based on all these data, we conclude that DMPC vesicles provide large, flat surfaces compared with those of micelles and thereby promote intermolecular interactions of Aβ(1-40) through their C-terminal segments, which form a stable core of the Aβ(1-40) assembly. We suggest that DMPC vesicles capture an intermediate form of Aβ(1-40) during its conformational transition coupled with assembly into the toxic β structure, which could be delineated by our solid-state NMR approach. Our findings will not only provide unique insights into the mechanisms underlying Aβ assembly but also offer structural clues for designing drugs targeting the assembly intermediates of Aβ for the development of anti-AD therapeutics.

Sample preparation
Expression and purification of uniformly 13 C-and 15 N-labeled Aβ(1-40) were performed as previously described [21]. DMPC was purchased from Avanti Polar Lipids, Inc

Solid-state NMR experiments
All solid-state NMR experiments were performed using a Bruker Avance 600 spectrometer at 1 H resonant frequency of 600 MHz, equipped with a 2.5 mm outer diameter 1 H-13 C-15 N triple resonance magic angle spinning (MAS) probe at a spinning frequency of 13.5 kHz. The sample was packed into a 4 mm space at the center of a sample tube using original Diflon spacers of 1 mm thickness to maintain RF homogeneity. Temperatures were controlled to 20°C using a VT controller. Typical RF fields of 1 H and 13 C pulses were 100 and 88 kHz, respectively. The spinning frequency of the sample was actively controlled by Bruker MAS controller to 13.5 kHz ± 5 Hz. 1 H heteronuclear dipolar decoupling was achieved by SPINAL64 [40] at a RF field of 100 kHz. 1 H-X RAMP-Cross polarizations [41] were used to enhance the initial magnetization of rare nuclei with 1 H RF fields of 100 and 80 kHz for X = 13 C, 15 N, respectively with repetition time of 2 s. The 13 C CTUC-COSY experiment was acquired at a constant evolution time of 6.96 ms with t 1 of 348 and t 2 of 2500 points. The number of scans was 1024 for each t 1 point. Z-filtering of 1 ms was inserted before detection.
The NCO and NCA-DCP experiments were acquired with t 1 of 256 and t 2 of 2048 points, respectively. The number of scans was 1024 for each t 1 point. Contact times of 1.5 and 5.0 ms were used for 1 H-15 N and 15 N-13 C CPs, respectively. 15 N-13 C CP was achieved by a 13 C spin locking field of 20 kHz at -1 SSB HH matching condition with 5% of RF ramp of 15 N RF spin locking field. LG decoupling at RF field of 100 kHz was applied during the second CP period.
The 13 C DARR/RAD experiments were performed at n = 1 R 3 condition for 1 H RF field during the mixing period with t 1 of 500 and t 2 of 2500 points, respectively. The number of scans was 256 for each t 1 point. The experiments were performed at mixing times of 4, 10, 100, 200, and 400 ms.
Quadrature detection in the indirect dimensions was achieved by States procedure for all two-dimensional (2D) experiments. Other parameters for 2D experiments are described in individual figure legends. The t 1 time domain of 2D FID for 13 C CTUC-COSY experiment was extended by forward linear prediction. The t 1 time domains of 2D FIDs were apodized by a sin window function. 2D FIDs were zero filled to 4096 for both t 1 and t 2 time domains prior to Fourier transformation. All data processing was performed using TOPSPIN2.1 (Bruker Biospin, Japan). Spectral analyses were performed using Sparky. Secondary structural elements were identified by TALOS-N software [31] according to 13 C and 15 N chemical shifts.
Supporting Information S1 Fig. Dihedral angles (φ, ψ) obtained by TALOS-N analysis according to 13 C and 15 N chemical shifts of Aβ  bound to DMPC MLVs. The secondary structure analysis by the TALOS software indicated a β-strand region (yellow shadow and yellow arrows) at the residues of Gly25-Val39. The chemical shift references were corrected before analyses of TALOS-N.