Phosphorylation at Ser8 as an Intrinsic Regulatory Switch to Regulate the Morphologies and Structures of Alzheimer's 40-residue β-Amyloid (Aβ40) Fibrils*

Polymorphism of amyloid-β (Aβ) fibrils, implying different fibril structures, may play important pathological roles in Alzheimer's disease (AD). Morphologies of Aβ fibrils were found to be sensitive to fibrillation conditions. Herein, the Ser8-phosphorylated Aβ (pAβ), which is assumed to specially associate with symptomatic AD, is reported to modify the morphology, biophysical properties, cellular toxicity, and structures of Aβ fibrils. Under the same fibrillation conditions, pAβ favors the formation of fibrils (Fpβ), which are different from the wild-type Aβ fibrils (Fβ). Both Fβ and Fpβ fibrils show single predominant morphologies. Compared with Fβ, Fpβ exhibits higher propagation efficiency and higher neuronal cell toxicity. The residue-specific structural differences between the Fβ- and Fpβ-seeded Aβ fibrils were identified using magic angle spin NMR. Our results suggest a potential regulatory mechanism of phosphorylation on Aβ fibril formation in AD and imply that the post-translationally modified Aβ, especially the phosphorylated Aβ, may be an important target for the diagnosis or treatment of AD at specific stages.

fibrils derived from the same primary sequences of polypeptides or proteins usually show distinct morphologies in vitro (5,6), and the morphologies of fibrils are sensitive to a variety of fibrillation conditions such as temperature (7), agitation (8), salt concentrations (9), surfactant (10), and seeding effects (11). Recent evidence revealed that amyloids can spread through a prion-like mechanism where fibrils seem to play important roles (12)(13)(14). Different fibrils with their specific morphologies, reminiscent of prion-like strains, may cause distinct pathological phenotypes and link different structures to the variations in disease transmission and pathology (15)(16)(17). Furthermore, fragmentation of fibrils, which always produces new ends for self-or cross-seeded fibrillation, is of critical importance for infectious amyloids (18,19).
Senile plaques consisting of fibrillar A␤ are considered one of the important hallmarks in AD (20). The 40-residue and 42-residue A␤ peptides (i.e. A␤40 and A␤42, respectively) are the two main fibrillar species. Recently, A␤ has also been reported to exhibit prion-like propagation properties (21,22). Distinct strains of A␤ were discerned in Alzheimer's patients (23)(24)(25)(26). Different amyloid propagation properties and structural profiles of A␤40 and A␤42 mimic distinct amyloid strains (10,27). The phenotypes induced by exogenous injection of A␤-containing brain extracts from Alzheimer's patients were dependent on both the hosts and the sources of agents, suggesting that polymorphic A␤ strains might result in varying biological activities (25,28,29). Molecular structures of A␤ fibrils derived from Alzheimer's patients with distinct clinical histories were also different (24). This underlines that the structural variations of A␤ fibrils may correlate with the variations of pathological phenotypes (24). In addition, compared with the brain-derived A␤ fibrils, the synthetic fibrils showed lower prion activities and different molecular structures (23,24,30), implying that some crucial factors in vivo might account for different fibrillar polymorphisms and pathological phenotypes.
Recently,post-translationalmodificationsofA␤,suchasphosphorylation and pyroglutamation, occurring in vivo have been found to promote the progression of AD (31,32). Among different types of post-translational modifications, the phosphorylation of proteins plays crucial roles in protein folding (33). Phosphorylation can alter the structures of a protein and modulate its activities (33). Using Trp-cage as a model protein, Kardos et al. (34) reported that phosphorylation could serve as a conformational switch to trigger the transition from native to amyloid state. Significantly, we and other groups have demonstrated that phosphorylation is involved in the formation of low barrier hydrogen bond (35) and turn conformations (36) as well as the destabilization of ␤-hairpin structure (37). Furthermore, we have also reported that phosphorylation may modulate the fibrillation process of amyloid proteins, such as Tau and ␣-synuclein (38 -40). Herein, we describe a novel regulatory function of phosphorylation at Ser 8 on morphology, biophysical properties, cellular toxicity, and structures of the A␤40 fibrils.
It has been shown that phosphorylation at Ser 8 in A␤ has important roles in late onset sporadic AD (31,41,42). Phosphorylation at Ser 8 was found in the brains of Alzheimer's patients in a hierarchical sequence and was specially suggested to be associated with symptomatic AD (43). Phosphorylation at Ser 8 was modulated by protein kinase A (44). Additionally, this sitespecific phosphorylation was known to accelerate the nucleation-dependent fibrillation of A␤ and to enhance the A␤-mediated amyloid toxicity (44). Attenuation of A␤ clearance via insulin-degrading enzyme and angiotensin-converting enzyme induced by this phosphorylation was also reported (45). Furthermore, the phosphorylation at Ser 8 could elevate numbers of strong hydrogen bonds in the N terminus of A␤ and increase the stability of the resulting pathogenic fibrils. The latter represents one of the crucial factors for disease progression in the brain (46). Despite all the functional importance, it is not clear whether this residue-specific phosphorylation can modify the morphologies and structures of A␤ fibrils, which are closely related to transmission and progression of AD.
Our present study demonstrates that A␤40 with phosphorylation at Ser 8 (pA␤) leads to a cross-␤ fibril (F p␤ ) morphologically and structurally different from the wild-type A␤ fibril (F ␤ ). The F p␤ and F ␤ show distinct single predominant morphologies and structures and have different biophysical properties and cellular toxicities. Residue-specific structural variations between the F p␤ -and F ␤ -seeded 40-residue A␤ fibrils were determined using magic angle spinning (MAS) NMR spectroscopy. Our results show the effects of post-translational modifications on the polymorphism of A␤ fibrils and their potential correlations to the propagation properties and cellular toxicities of the fibrils. Phenomena of strains related to distinct amyloid polymorphism are the subject of extensive interest because of their correlation with different pathological phenotypes (10,25,26,47,48). The current study also strongly suggests the potential relationship between post-translational modifications and strain formation in vivo.

Results
The F p␤ Fibrils Have Different Morphologies Compared with the F ␤ Fibrils-The F ␤ and F p␤ fibrils were prepared using the same conditions with the same initial peptide concentration at 40 M. Circular dichroism (CD) and Fourier transform infrared (FT-IR) spectroscopy were first used to study the structures of both monomers and fibrils. Fig. 1, A and B, shows that CD and FT-IR spectra of the unaggregated monomers of the wild-type A␤40 and pA␤ were very similar. The CD spectra (Fig. 1A) suggested that both monomeric A␤40 and pA␤ had mainly random coil structures. Furthermore, the FT-IR spectra (Fig. 1B) for both monomeric A␤40 and pA␤ were symmetric from 1580 to 1740 cm Ϫ1 with maximum peaks at about 1653 cm Ϫ1 . Interestingly, we observed obvious differences in CD spectroscopy between the F p␤ and F ␤ fibrils. The CD spectra of F p␤ and F ␤ had minimum peaks at about 216 and 222 nm, respectively (Fig.  1A), indicating different ␤-sheet structures in their fibril cores (7,49,50). Compared to the F p␤ fibrils, the fibrillation of F ␤ was accompanied by a reduction of CD spectral intensity presumably due to the increased scattering (50). Transmission electron microscopy (TEM) showed that the morphologies of both F p␤ FIGURE 1. Phosphorylation induces pA␤ misfolding into amyloid fibril morphology (F p␤ ) distinct from wild-type A␤ (F ␤ ). A, CD spectra of unaggregated A␤ monomers (blue dashed line), pA␤ monomers (red dashed line), F ␤ fibrils (blue solid line), and F p␤ fibrils (red solid line). Both unaggregated A␤ and pA␤ monomers mainly showed similar random coil structures (198 nm). F ␤ and F p␤ fibrils showed ␤-sheet-enriched structures with a minimum peak at about 222 and 216 nm, respectively. B, FT-IR absorbance spectra of unaggregated A␤ monomers (blue dashed line), pA␤ monomers (red dashed line), F ␤ fibrils (blue solid line), and F p␤ fibrils (red solid line). Both unaggregated A␤ and pA␤ monomers showed a symmetrical FT-IR spectrum with a maximum absorbance peak at 1653 cm Ϫ1 .The F ␤ and F p␤ fibrils showed the same maximum absorbance peak at 1628 cm Ϫ1 but small variations between 1640 and 1680 cm Ϫ1 that imply certain structural differences. C and D, TEM images of F ␤ fibrils (C) and F p␤ fibrils (D). F p␤ exhibited a twisted cylindrical structure, whereas F ␤ displayed a ribbon-like structure. E and F, XRD images of F ␤ fibrils (E) and F p␤ fibrils (F). The reflections at 4.7 and 10 Å, respectively, indicated that both F p␤ and F ␤ fibrils were rich in cross-␤ structures. mdeg, millidegrees. and F ␤ fibrils were homogenous. The F p␤ fibrils were mainly twisted cylindrical structures (Fig. 1D), which were distinct from the flat ribbon-like morphology for F ␤ fibrils (Fig. 1C). Additionally, the FT-IR spectra (Fig. 1B) and X-ray diffraction (XRD) diffraction patterns (Fig. 1, E and F) revealed that both F p␤ and F ␤ fibrils were rich in cross-␤ structures. The FT-IR spectra of both F ␤ and F p␤ fibrils showed main peaks at ϳ 1628 nm Ϫ1 that were also consistent with previous studies on A␤ fibrils (51)(52)(53). However, two spectra were distinct in the loop or turn regions between 1640 and 1680 cm Ϫ1 , indicating certain levels of structural variations between the two fibrils. The XRD pattern showed reflections at 4.7 and 10 Å for the inter-and intramolecular spacing, respectively, which characterized the cross-␤ structures for both F ␤ and F p␤ fibrils (54). Taken together, we found that phosphorylation of A␤40 at Ser 8 could modify the morphologies of amyloid fibrils. Recent evidence has shown that different microenvironments modulate the free energy landscapes for A␤ folding into different morphological states (8,25,50). It is possible that the side-specific phosphorylation may alter the free energy barrier for the folding of A␤40 and lead to a different morphology for the pA␤.
Phosphorylation Plays Crucial Roles in F p␤ Formation-Recently, Rezaei-Ghaleh et al. (46,55) have shown that phosphorylation at Ser 8 could impel the undergoing changes in local conformational dynamics of A␤40, induce the variations of N-terminal exposure of A␤ aggregates, and increase the fibril stability by promoting the formation of strong hydrogen bonds directly. We propose that different conformations of F p␤ fibrils may also be induced directly by phosphorylation. The thioflavin T (ThT) fluorescence kinetics measurement was first applied to determine the fibrillation rate of pA␤. ThT fluorescence kinetics measurements of 40 M monomeric A␤40 and pA␤ were conducted at 37°C with continuous shaking. Both A␤40 and pA␤ exhibited typical nucleation-dependent sigmoidal curves (Fig. 2). Similar to previous results from Kumar et al. (44), phosphorylation at Ser 8 accelerated the fibrillation process (Fig. 2). The lag times of fibrillation for 40 M pA␤ and A␤40 were 22 and 38 min, respectively. Furthermore, pA␤ seemed to decrease the aggregation half-time (39.1 Ϯ 9.0 min) compared with A␤40 (57.6 Ϯ 13.2 min). These results suggested that phos-phorylation might either shorten or alter the fibrillation process of A␤.
Solid-state 31 P NMR spectra for the F p␤ fibrils and the monomeric pA␤ peptides are shown in Fig. 3. Both spectra were recorded with similar quantities of samples (ϳ8 mg) and the same hydration level (ϳ1 l/mg). The direct polarization 31 P spectroscopy allowed quantitative comparison of the peak intensities between the two samples. The fact that the 31 P intensity was greatly enhanced in the F p␤ sample suggested that the phosphate group at Ser 8 was located in an ordered environment upon fibrillation compared with the monomeric peptides. Furthermore, different from the monomeric pA␤ sample, which might form amorphous aggregates because of the rehydration, the fibrillar F p␤ 31 P spectrum showed only one predominant peak, suggesting a single local conformation for the side chain of Ser 8 .
To further uncover the roles of phosphorylation in F p␤ fibrils, the alkaline phosphatase dephosphorylation assay was applied (56). We utilized calf intestinal alkaline phosphatase (CIP) purified from calf intestinal mucosa that can remove phosphate groups from phosphorylated species (57). Previous work has suggested the involvement of phosphorylated Ser 8 in specific hydrogen bonding to the hydrophobic core of fibrils (46,55). In these cases, the phosphate group might be restricted, and efficiency of dephosphorylation might decrease. In our experimental design, the monomeric pA␤ (Fig. 4, A and C) and F p␤ fibrils ( Fig. 4B) were treated with a series of CIP solutions with different enzymatic concentrations for 30 min, and then the dephosphorylated pA␤ bands were traced using Tricine-SDS-PAGE (Fig. 4, A and B) and native PAGE (Fig. 4C). Only Tricine-SDS-PAGE was used to trace the bands after dephosphorylation of fibrils ( Fig. 4B) because fibrils could not be disassociated and enterintothegelinnativePAGE.Theresultsshowedthatdephosphorylation of pA␤ generated A␤40 in a concentration-dependent manner. Furthermore, the phosphate groups on Ser 8 in monomeric pA␤ could be cleaved by CIP much more easily than those in F p␤ fibrils (Fig. 4, compare the corresponding bands in A with those in B). Even at a lower concentration of CIP (0.1 unit of CIP enzymatic activity), monomeric pA␤ was almost completely dephosphorylated. For F p␤ fibrils, however, the phosphate groups were only partially removed even at high CIP concentrations, such as 1 and 10 units of CIP enzymatic activity (i.e. Fig. 4, two bands for both pA␤ and A␤ in B, second and third columns, compared with only a single band for A␤ in the corresponding columns in A). Therefore, we identified that F p␤ fibril formation clearly limited the dephosphorylation by CIP. Recently, evidence has shown that fibrils can sequester different kinds of proteins and change their corresponding physiological properties (58,59). Fibrils might also influence the enzymatic activities (60,61). Therefore, dephosphorylation of pA␤ monomers mixed with equal moles of F p␤ fibrils (native PAGE in Fig. 4D) was tested to exclude the possibility that fibrils might trap the CIP and lower the enzymatic activity. Only native PAGE was used to identify the new bands after dephosphorylation because the bands from disrupted fibrils in SDS-PAGE would influence the results. Comparison of Fig. 4, C and D, showed that the presence of F p␤ fibrils did not affect the enzymatic activity of CIP because the dephosphorylation effects of this enzyme on pA␤ monomers were similar both in the absence (Fig. 4C) and presence ( Fig. 4D) of fibrils. Collectively, these results suggest that the phosphorylation at Ser 8 would directly induce the formation of certain structures in F p␤ fibril. F p␤ and F ␤ Fibrils Show Distinct Propagation Activities and Cellular Toxicities-Distinct fibril morphologies have been considered to affect the symmetries and lateral association propensities of protofilaments (8). Consequently, the morphological differences between F p␤ and F ␤ fibrils might imply their structural diversities, such as the surface exposure of different amino acid side chains, which might further be correlated with prion-like propagation (30,48) and cellular toxicities (8,9), etc.
First, the prion-like propagation of F p␤ and F ␤ fibrils in vitro was monitored by a ThT fluorescence assay on the seeded fibrillation (62), and TEM was used to examine the structural natures of aggregates after seeding. Both F p␤ and F ␤ fibrils were sonicated in an ice bath to produce small fragments (Fig. 5). The wild-type A␤40, A␤42, and pyroglutamated A␤3-40, which is an important post-translationally modified form of A␤ (named A␤(3pE)40 hereinafter) (63), were tested on the cross-seeding efficiencies for the seeds of F p␤ and F ␤ fibrils. Fig. 6A plots the ThT kinetics curves of A␤40 in the presence of 10% F p␤ and F ␤ seeds. Both seeds greatly shortened the lag phase, and F p␤ initiated the fibrillation more rapidly compared with the F ␤ -seeded fibrillation. The significant difference in the lag time between the F p␤ -(almost no lag time) and F ␤ -seeded fibrillation (296 min) is shown in Fig. 6D. We emphasize that the results in Fig. 6A suggested that the wild-type A␤40 formed fibrils more rapidly in the presence of F p␤ seeds in comparison with its own seeds, meaning that the cross-seeding between pA␤ and wildtype A␤40 was more efficient than the self-seeding of wild-type A␤40. Interestingly, we found that the ThT intensities of seeded plateaus in Fig. 6A were lower than that of unseeded. The concentrations of A␤40 in both the parental and seeded fibrils were 40 M. Therefore, the intensity difference was not likely to originate from the quantities of fibrils. We postulated that the ThT fluorescence intensity was reduced in seeded fibrils because certain fibril structures with low ThT fluorescence emission were selectively amplified during the seeding process as has been suggested by a previous work from Tycko and co-workers (64) . Fig. 6, B and C, further show that, for the A␤(3pE)40 and wild-type A␤42, the F p␤ induced more efficient cross-seeding than the wild-type F ␤ . Overall, our results suggested that the pA␤ (i.e. Ser 8 -phosphorylated A␤40) might serve as an intrinsic trigger to promote the fibrillation of a variety of A␤ species. As revealed by TEM, both the A␤40 and A␤(3pE)40 fibrils seeded from F p␤ showed similar twist-like morphologies (Fig. 6, G and H, right panel), whereas the A␤40 and A␤(3pE)40 fibrils seeded from F ␤ showed ribbon-like morphologies (Fig. 6, G and H, left panel). These results indicated that both F p␤ and F ␤ fibrils could propagate their structures to A␤40 and A␤(3pE)40. However, under the same seeding conditions, although the fibrillation times were significantly shortened, a similar morphology was obtained for both F p␤ -and F ␤ -seeded A␤42 fibrils (Fig. 6I) that was different from either the parental F p␤ or F ␤ fibrils. It was proposed previously that the seeded fibrillation process involves mainly chain elongation from the existing structural patterns of the seeds (65). For A␤40 and A␤(3pE)40, it was interesting that the F p␤ seeds served as a better structural template for the A␤ chain elongation compared with F ␤ seeds, which would reflect the differences in the intrinsic architectures of the two fibrils. The conformational transition from partially folded A␤ to ␤-sheet-enriched state is an essential process for A␤ aggregation, and the seeded fibrillation might follow the "docking conformational change" mechanism in which the existence of fibril seeds as templates would lower the "activation energy" for the conformational switch (66 -68). The differences of the fibrillation process in the presence of the F ␤ and F p␤ seeds revealed that phosphorylated fibrils possibly contained the core structures that would decrease the energy barrier and promote the efficiency of chain elongation much more easily. Surprisingly, despite the high similarity of the A␤40 and A␤42 sequences, monomeric A␤42 was incompatible with the structural nature of either F ␤ or F p␤ seeds. Although the shortened lag time was detected when A␤42 was seeded by both F p␤ and F ␤ fibrils, the resulting A␤42 fibrils did not inherit the parental structural information. This implies that A␤42 cannot be directly converted to fibrils at the termini of the seeds of F p␤ and/or F ␤ fibrils by using the structural template from seeds. A similar phenomenon was also observed in a previous study of ␣-synuclein strains (9). One possible explanation was that the F p␤ and F ␤ seeds only served as a "surface" to catalyze the formation of fibrils but not as structural templates.
To explore whether the F p␤ fibrils possessed different levels of cytotoxicity compared with the wild-type F ␤ fibrils, both fibrils were applied extracellularly to mouse neuroblastoma N2a cells and mouse microglia BV-2 cells. The cell viabilities were investigated using 3-(4-,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (Fig. 7). After 24-h treatment, both fibrils showed concentration-dependent cytotoxicities to these two mammalian cells. The twisted F p␤ fibrils showed significantly higher cytotoxicity than the flat ribbon F ␤ fibrils. Ribbon-like and twisted fibrils are two major types of fibril morphologies formed by amyloid proteins, including A␤ (8,69,70), amylin (71), and ␣-synuclein (9), depending on the growth conditions. Different fibril structures have been shown to correlate with distinct morphologies and different levels of concentration-dependent cytotoxicities (6,69). We observed higher toxicity for the twisted F p␤ fibrils compared with the ribbon-like F ␤ fibrils, which was consistent with the potential correlation between morphology and cytotoxicity observed in other A␤ and ␣-synuclein fibrils (8,9). We suspect that different lateral association and symmetry of twisted and ribbon-like morphologies might lead to distinct packing geometries and different exposures of residues on the fibril surface and therefore might induce different levels of toxicity to cells.
Molecular Structural Differences between the F p␤ -and F ␤ -seeded A␤40 Fibrils-Given the differences in the fibril morphologies, biophysical characterizations, and cytotoxicities, we were interested in whether there are molecular level structural variations between the F p␤ and F ␤ fibrils. This was done by MAS NMR studies on the F p␤ -and F ␤ -seeded A␤ fibrils. The seeded fibrils were prepared by adding 10% F p␤ or F ␤ seeds to freshly dissolved 40 M A␤40 monomers followed by quiescent incubation for 48 h under physiological temperature and pH. This protocol minimized the manipulations on fibrillar structures and was likely to produce seeded fibrils that were similar to their parents (8,72). We verified the structural similarity using TEM and CD spectroscopy and examined the cytotoxicities of daughter fibrils using the MTT assay (Fig. 8). The CD spectra showed minimum peaks at 216 and 222 nm for the F p␤ -and F ␤ -seeded fibrils, respectively (Fig. 8A), that were identical to their parent fibrils. In Fig. 8B, the TEM images show that the features for the F p␤ and F ␤ fibrils, which were the twisted cylindrical and flat ribbon-like morphologies, respectively, were propagated to their daughter fibrils. In Fig. 8C, we found that daughter fibrils from both F ␤ and F p␤ seeds showed concentration-dependent cytotoxicities to the two mammalian cells. Daughter fibrils seeded from F p␤ also showed higher cytotoxicity where the trend was similar to that of their parental fibrils. Therefore, we concluded that F p␤ -and F ␤ -seeded fibrils would retain structural characteristics from their parental fibrils.
Extensive MAS NMR measurements were performed to probe the residue-specific secondary structures of both seeded fibrils. We used six scattering uniformly isotope-labeled A␤ peptides to cover the majority of the primary sequence, especially the typical ␤-strands and interstrand loop segments from the known A␤ fibril structures (24,27,69,(73)(74)(75). Representative NMR spectra and the summary of chemical shift deviations are plotted in Fig. 9. The chemical shift assignments for both seeded fibrils are provided in Tables 1 and 2. Overall, both fibrils showed typical ␤-loop-␤ structures like other known 40-residue A␤ fibrils. However, detailed structural variations were observed. First, the local secondary structure of Gly 9 , which is next to the phosphorylated Ser 8 , changed significantly. The wild-type F ␤ fibrils showed multiple Gly 9 CЈ-C␣ crosspeaks, whereas the phosphorylated F p␤ fibrils had a predominant cross-peak (Fig. 10A). This observation was consistent with the 31 P NMR (Fig. 3) and confirmed that the local conformation around Ser 8 became better defined due to the phosphorylation. Second, the site-specific phosphorylation affected not only the local conformation but also the overall secondary structures of the resulting fibrils. As shown in Fig. 9, we identified significant chemical shift deviations between the F p␤ and F ␤ fibrils, and the largest chemical shift differences were observed for residues Gly 9 , the segments between Leu 17 and Ile 32 , and the residue Val 39 . Particularly, the segments Asp 23 -Ser 26 and Gly 29 -Ile 32 showed the most significant chemical shift deviations between the two fibrillar species. The C-termi- FIGURE 5. TEM images of sonicated seeds of F ␤ and F p␤ . Both A␤ (left) and pA␤ (right) seeds were produced by sonicating the F ␤ and F p␤ fibrils, respectively, in the ice bath three times (for 1.5 min each time). FEBRUARY 17, 2017 • VOLUME 292 • NUMBER 7

JOURNAL OF BIOLOGICAL CHEMISTRY 2615
nal ␤-strand region from Gly 33 to Gly 38 , in contrast, seemed to have little structural variation. Additionally, it seemed that the F p␤ fibrils had more ordered C termini compared with the F ␤ fibrils because the Val 39 cross-peaks were much stronger in the F p␤ -seeded fibrils then in the F ␤ -seeded fibrils (Fig. 10, B and C). Importantly, the segment between Ala 21 and Ala 30 is generally considered as the interstrand loop region based on the known 40-residue A␤ fibril structures and might be involved in the early stages of fibrillation according to previous studies on A␤ and other similar types of amyloid peptides (76,77). Because the seeded fibrils were likely to have the same structures as their parents, our results suggested that the phosphorylation on Ser 8 might affect the intermediate A␤ structures formed during the nucleation steps through specific molecular interactions between the phosphorylated Ser 8 side chains and residues located in the loop segment between Ala 21 and Gly 29 . Interestingly, recent work (46) showed that the phosphorylation of Ser 8 stabilized the A␤ fibrils through a potential inter-residue interaction among Asp 7 and the phosphorylated Ser 8 and Ser 26 , which was consistent with our NMR data.

Discussion
In this decade, polymorphic A␤ fibrils with different neuronal cytotoxicities and propagation efficiencies have attracted intense attention (6). Small variations of microenvironments in vitro will change the kinetics and thermodynamics for aggrega- , and A␤42 (F) showed that both F p␤ (red column) and F ␤ (blue column) can shorten the lag phase for amyloid formation. The structural natures of A␤40, A␤(3pE)40, and A␤42 aggregates after seeding were examined using TEM (G, H, and I, respectively); left and right panels are structural images of F ␤ -seeded fibrils and F p␤ -seeded fibrils, respectively. The TEM samples were prepared after 24-h incubation. For A␤40 (G) and A␤(3pE)40 (H), the structural information of parental F ␤ fibrils and F p␤ fibrils can be inherited. The black and white arrowheads point to ribbon-like and twisted cylindrical morphologies, respectively. However, A␤42 cannot sample the structural information from either F ␤ or F p␤ fibrils (I). ThT fluorescence kinetics of different soluble A␤ species without any seeds and F ␤ or F p␤ seeds only were also conducted as controls. Error bars represent S.E. (n ϭ 3 independent measurements).
tion (8,70). This may be induced by the coexistence of multiple nucleation processes, each of which favors its own fibril structures (78). Morphological differences induced by variations in fibrillation conditions imply that experimentally based microenvironments (agitation, pH, ionic strength, etc.) may lead to a favorable nucleation process (8,9,50,69,70,78). For A␤40 fibrils, two major types of fibrillar morphologies have been reported. One is the twisted filament with a 3-fold molecular symmetry, and the other is the ribbon-like filament with a 2-fold molecular symmetry (8,69,70). Herein, we report the effects of phosphorylation at Ser 8 , which is an existing posttranslational modification of A␤ in vivo (42,43), on the fibrillation of A␤. Similar to the effects of microenvironment variation on amyloid formation, phosphorylation at Ser 8 changes the nucleation process of A␤ and finally leads to fibrils with distinct morphologies. Under the same fibrillation conditions, the pA␤ folds into a twisted fibril, and A␤ folds into a flat ribbon-like fibril. The N-terminal post-translational modifications of A␤, such as phosphorylation (31), pyroglutamation (63), and nitration (79), have been reported to control the kinetic process for A␤ fibrillation, accelerate plaque formation, and alter the prion-like propagation efficiency. We first revealed that phosphorylation at Ser 8 , acting as an intrinsic molecular trigger beyond the variations of fibrillation conditions, can induce pA␤ folding into different fibril morphologies.
The prion-like strain phenomenon where a single amyloid protein gives rise to multiple distinct phenotypes has been correlated with the ability of some polypeptides or proteins to fold into distinct fibril structures in neurodegenerative disease (16,17). A␤ seeds have been found to possess a transmission risk like prion protein aggregates in vivo (22). Our study herein indicates that phosphorylation results in A␤ fibril morphology distinct from the wild-type A␤, which proposes a possible intrinsic origin of polymorphous fibrils of A␤. Neurotoxic studies and seeding experiments show that F p␤ and F ␤ fibrils from pA␤ and wild-type A␤, respectively, have different cytotoxicities and seeding efficiencies, which probably arise from strain-specific structural variation. The results suggest that phosphorylation may act as a possible factor to induce the formation of strains of A␤. A recent study by Rijal Upadhaya et al. (43) showed that phosphorylation of A␤ occurred in the Alzheimer's brain in a hierarchical sequence and promoted the formation of final toxic aggregates, and this phosphorylation was assumed to be indicative for biochemical A␤ stage 3 and associated with FIGURE 7. Cytotoxicity of F p␤ and F ␤ fibrils were measured using MTT assay. Cell viability of BV-2 cells (A) and N2a cells (B) treated with F ␤ and F p␤ fibrils, respectively, for 24 h was measured using an MTT assay. The blank control is shown in gray columns. F p␤ fibrils shows higher toxicity compared with F ␤ fibrils in a concentration-dependent manner. n.s., not significant. S.E. is shown as error bars (n ϭ 3 independent measurements; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 as evaluated with independent Student's t test). FIGURE 8. CD spectroscopy, TEM images, and MTT assay indicated that structural features of F p␤ and F ␤ fibrils could be propagated to their daughter fibrils. 40 M A␤40 was used to examine the structural natures after seeding. A, CD spectroscopy showed that the minimum peak of F p␤ -and F ␤ -seeded fibrils was the same as their parent fibrils, respectively. The inset is normalized CD spectroscopy curves of parent and daughter fibrils. B, TEM images of F p␤ -and F ␤ -seeded fibrils were identical to their parent fibrils. The samples were prepared after 48-h incubation. C, MTT assay showed that F p␤seeded fibrils had higher cytotoxicity of which the trend was the same as for their parental fibrils. S.E. is shown as error bars (n ϭ 3 independent measurements; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; n.s., not significant; as evaluated with independent Student's t test). mdeg, millidegrees. FEBRUARY 17, 2017 • VOLUME 292 • NUMBER 7 FIGURE 9. MAS NMR spectra and chemical shift deviations of F ␤ -and F p␤ -seeded 40-residue A␤ fibrils. The F p␤ -seeded fibrils (A, C, and E) and F ␤ -seeded fibrils (B, D, and F) were isotope-labeled at different residues. Intraresidual C␣-C␤ cross-peaks are highlighted in colored lines. The differences between dashed and solid lines in B, D, and F indicate shifts in cross-peaks. G shows the residue-specific chemical shift deviations (␦(F p␤ -seeded) Ϫ ␦(F ␤ -seeded)) for CЈ (black squares), C␣ (red circles), and C␤ (blue triangles). The solid orange lines highlight the Ϯ1.0-ppm threshold for significant chemical shift deviation based on the estimation of 13 C NMR line widths. Residues with at least one significant chemical shift deviation among CЈ, C␣, and C␤ are highlighted in green on the x axis. symptomatic AD. We suggest that the occurrence of phosphorylation was not just an inducer for the further accumulation of total aggregates but might be an initiator for the formation of certain amyloid morphologies with specific biological activities.

Phosphorylation Regulates Fibrillar Morphology and Structure
Recently, the structures and stabilities of A␤40 phosphorylated at Ser 8 have been studied by Rezaei-Ghaleh et al. (46,55) using solution NMR spectroscopy and molecular dynamics simulations. These studies mainly concluded that the phosphorylated Ser 8 was not included in the fibril core, which typically consisted of residues 15-40. In the monomeric state in solution, the phosphorylation at Ser 8 mainly affected the local conformation from residues 4 to 16. However, it also promoted the inter-residual interactions between segments Asp 23 -Gly 25 and Gly 29 -Ala 30 , which are located in the inter-␤-strand loop region in A␤ fibrils (55). Combined with our solid-state NMR measurements on the chemical shift differences between F p␤and F ␤ -seeded fibrils, we propose that the Ser 8 phosphorylation might affect the initial nucleation of A␤ fibrillation. This sitespecific phosphorylation might stabilize certain conformations within the segment Asp 23 -Ala 30 , presumably through a transient hydrogen bonding interaction based on the previous molecular dynamics simulation (46). The fact that we observed the most significant chemical shift differences within the segments Asp 23 -Ser 26 and Gly 29 -Ile 32 between the two types of fibrils further supports the hypothesis. The residues located between two typical ␤-strands in amyloid fibrils have been proposed as general initial nucleation sites in the fibrillation process of A␤ and other types of amyloid peptides (76,77). Therefore, stabilization of local conformations in this region might accelerate the rates of nucleation and may lead to different fibril morphology.
In summary, amyloid fibrils together with small aggregates, such as oligomers and protofibrils, are considered important pathological agents in AD. Our study raises the possibility that  . 2D 13 C-13 C MAS NMR spectra. A, cross-peaks for Gly 9 in F p␤seeded (left) and F ␤ -seeded (right) fibrils. B and C, cross-peaks for Val 39 in F p␤ -seeded (B) and F ␤ -seeded (C) fibrils. FEBRUARY 17, 2017 • VOLUME 292 • NUMBER 7 different A␤ strains would occur in a phosphorylation-dependent manner. Compared with the wild-type A␤ fibrils, phosphorylation at Ser 8 leads to an overall fibril structural change with significant chemical shift deviations in the segments Asp 23 -Ser 26 and Gly 29 -Ile 32 . Along with the recent study from Rezaei-Ghaleh et al. (46,55), we propose that the effects of phosphorylation at Ser 8 on the local conformation of residues 4 -16 and the inter-residual interactions in the interstrand loop region from residues 21 to 30 would cooperatively promote pA␤ to fold with a distinct nucleation process and finally lead to a fibril morphology different from that of wild-type A␤. Combined with the potential link between phosphorylation and symptomatic AD (43), we propose that pA␤ might serve as a specific therapeutic target for AD. The specific link between A␤ strains and phosphorylation raises the possibility of designing structure-based imaging agents and inhibitors to diagnose or prevent AD at different stages.

Experimental Procedures
Sample Preparation-All A␤ peptides with the same primary sequence (DAEFRHDSGYEVHHQKLVFFAEDVGSNKGAI-IGLMVGGVV) were prepared using Fmoc chemistry with Fmoc-Val-Wang resin (0.18 mmol/g; GL Biochem (Shanghai) Ltd.). Fmoc-Ser(PO 3 Bzl)-OH (where Bzl is benzyl) (GL Biochem (shanghai) Ltd.) and uniformly 13 C-and 15 N-labeled amino acids (Cambridge Isotope Laboratories, Inc.) were incorporated into the peptides to achieve phosphorylated and isotope-labeled A␤ peptide sequences. The Glu 3 -pyroglutamated A␤40 (A␤(3pE)40; pEFRHDSGYEVHHQKLVFFAEDVGSNK-GAIIGLMVGGVV) was also synthesized using Fmoc chemistry. A␤42 peptide was purchased from Your Bio-Tech Partner, Shanghai, China, and used directly. All peptides were purified by reverse-phase HPLC (C 18 column, Waters 600) and identified by matrix-assisted laser desorption ionization time-offlight mass spectrometry (MALDI-TOF/TOF MS, ABI Research). All fibril samples (including the F ␤ and F p␤ fibrils) were prepared following the method described previously (80). Briefly, the A␤ or pA␤ peptides were dissolved in HFIP (1 mg/ml, Sigma) for at least 4 h at ambient temperature to remove any preformed aggregates. Then the HFIP was removed by blow drying with N 2 and further vacuum drying for at least 3 h using an oil pump. The thin film of peptides was then resuspended in 20 mM Tris⅐HCl buffer, 150 mM NaCl, pH 7.4, and incubated at 37°C with 200-rpm continuous shaking for 2 days to produce mature fibrils.
ThT Fluorescence Kinetics Assay-The ThT fluorescence kinetics assay was conducted on the Synergy 4 (BioTek) plate reader using a 96-well plate ( ex ϭ 440 nm and em ϭ 480 nm). For monitoring the fibrillation process of A␤ and pA␤ (40 M), ThT signals (20 M) were recorded at 37°C under continuous shaking (medium) with a time interval of ϳ10 min between each recording. For tracking the aggregation process of different A␤ species (40 M A␤40, 10 M A␤(3pE)40, and 10 M A␤42) in the presence of F ␤ and F p␤ seeds (10%), the seeds of F ␤ and F p␤ were prepared by sonicating the resulted fibrils three times (1.5 min each time) in an ice bath. ThT signals were recorded at 37°C under quiescent conditions with a time interval of 10 min between each recording. The fibrillation ThT kinetics curves of A␤ and pA␤ were fitted using a sigmoidal function, F Ϫ F 0 ϭ 1/(1 ϩ exp(␥ max ( 1/2 Ϫ t))), in which 1/2 is the half-completion time of aggregation and ␥ max is the maximum growth rate (50). The t lag (lag time) was determined as the time intercept of the line best fitted to the linear portion of the F to t coordinate. All the experiments were replicated at least three times.
TEM-Fibrils samples (8 l) were directly incubated on a carbon-coated copper grid for 2 min and then stained with 2% sodium phosphotungstic acid, pH 6.5, for 1 min. TEM images were recorded using a Hitachi-7650B electron microscope at 80 kV.
CD-CD spectroscopy in the 200 -260-nm region (far-UV region) was performed with a Chirascan Plus CD spectrophotometer (Applied Photophysics). All samples (40 M; 200 l) were directly loaded in a 1-mm width CD cuvette, and signals were recorded at room temperature with three-scan signal averaging.
XRD-F ␤ and F p␤ fibrils were centrifuged (12,000 rpm) at 4°C for 30 min, and the resulting pellets were washed with ddH 2 O three times to remove salts. The pellets were then resuspended in 10 l of ddH 2 O, and the final suspension was aligned between two fire-polished glass rods at 4°C overnight. XRD images were recorded on a Rigaku Micromax-007 X-ray generator equipped with an R-Axis IVϩϩ area detector.
FT-IR Spectroscopy -FT-IR spectroscopy was performed on a Frontier FT-IR (PerkinElmer Life Sciences) equipped with an attenuated total reflectance accessory at ambient temperature. 1 ml of F ␤ and F p␤ fibrils was centrifuged (12,000 rpm) at 4°C for 30 min. The pellets were then washed with ddH 2 O three times to remove salts. The pellets were then frozen and dried into powders. The monomeric samples were prepared by dissolving A␤ and pA␤ (1 mg/ml) in HFIP overnight, and then the solutions were frozen and lyophilized into powders. The final powders were directly loaded on the attenuated total reflectance accessory for acquisition of data.
Dephosphorylation by CIP-All samples with or without CIP (Sigma) were incubated at 37°C for 30 min. The final mixtures were then directly loaded and analyzed using Tricine-SDS-PAGE and native PAGE, respectively. The PAGE images were recorded using VersaDoc 3000 (Bio-Rad). For all the experiments, 20 l of 40 M A␤ monomers or fibrils was used.
Neuronal Cell Toxicity Assay-To detect the toxicities of the two fibrils, the fibrils were centrifuged at 12,000 rpm at 4°C for 1 h, and the final pellets were resuspended in the cell culture medium, sonicated in an ice bath three times (1.5 min each time). MTT reduction assays were used to determine the effects of F ␤ and F p␤ fibrils on viability of neuronal cells. Mouse neuroblastoma N2a cells and murine microglia BV-2 cells were plated at a density of 5000 cells/well on 96-well plates in 150 l of culture medium (50% ␣-minimum Eagle's medium and 50% DMEM with 10% fetal bovine serum for N2a cells and 100% DMEM with 10% fetal bovine serum for BV-2 Cells). After 24-h incubation at 37°C, the medium was exchanged for fresh medium (100% DMEM without fetal bovine serum) with F ␤ and F p␤ fibrils. After 24-h incubation, 20 l of MTT (5 mg/ml) was added into each well for a further 4-h incubation. The culture medium was then discarded, and 150 l of DMSO was added to dissolving formazan completely. Absorbance at 490 nm was measured using Synergy 4 plate reader to calculate the cell viability.
MAS NMR-The seeded fibrils for MAS NMR measurements were prepared by incubating the isotope-labeled monomeric A␤40 with 10% F ␤ and F p␤ seeds for 2 days quiescently. The fibrils were then centrifuged at 100,000 ϫ g for 1 h at 4°C (Beckman Coulter). The final pellets were freeze-dried and packed into 2.5-mm MAS rotors with rehydration using deionized water (1 l/mg). All MAS NMR measurements were performed on a 600-MHz Bruker solid-state NMR spectrometer equipped with a 2.5-mm TriGamma MAS probe. The MAS frequency was set to 10 kHz for all experiments. The sample temperature was kept at ϳ 5°C using a 270 K N 2 cooling line. The 31 P spectra were recorded using direct excitation with a 50-kHz 31 P 90°radiofrequency (rf) pulse and a 100-kHz continuous wave 1 H decoupling during the acquisition time. The twodimensional (2D) 13 C-13 C spin diffusion spectra were recorded using a pulse sequence composed of a 60-kHz 1 H excitation 90°p ulse, a linearly ramped 1 H-13 C cross-polarization period, a 10-kHz rf-assisted diffusion 1 H field during the 10-ms mixing period, and a 100-kHz 1 H two-pulse phase modulation decoupling. The 31 P and 2D spectra were processed using Topspin and NMRPipe software, respectively.