Introduction

Non-specific membrane disruption induced by extracellular β-amyloid (Aβ) aggregates is considered a tenable mechanism of Aβ cytotoxicity and pathology in Alzheimer’s disease (AD)1,2,3. Upon enzymatic cleavage of the amyloid precursor protein, the released Aβ is likely to initiate aggregation at the membrane surface and within the extracellular matrix. Pathological amyloid plaques are commonly associated with lipids, sterols, and extracellular proteins4,5,6. Recent updates to the amyloid cascade hypothesis (ACH) highlight cellular plasma membranes as a platform for the co-aggregation of multiple amyloidogenic proteins, such as the coaggregation between Aβ and tau, which may be more directly relevant to AD pathology4,7,8,9. Despite its biological importance, understanding the molecular structural features and membrane interactions of Aβ intermediate states during the primary nucleation stage remains a major challenge10,11,12,13.

Current experimental studies on the nucleation intermediate states of Aβ fibrillation primarily use kinetic assays and microscopic techniques. The kinetic assays provide insights into the molecular mechanisms of nucleation, including the involvement of multistep nucleation processes, the rates of individual steps, and potential critical oligomer sizes. For instance, a recent kinetic study suggested that Aβ1-42 may adopt two distinct nucleation mechanisms at varying initial monomeric concentration, with a predicted critical nucleus size of seven to eight monomers14. Kinetic studies have also revealed distinct nucleation mechanisms in Aβ1-40 and Aβ1-42, particularly regarding the dominance of secondary nucleation15. Microscopic studies enable direct observation of the presence, morphology, and interactions of Aβ intermediates during nucleation processes. For instance, a previous atomic force microscopy (AFM) study showed that the N-terminal segments of Aβ preferentially bind to ganglioside-enriched membrane domains, while the C-terminal segments serve as anchors to the bilayer interior through hydrophobic interactions16. Single-molecule microscopy also directly observes the attachment of Aβ monomers and low-order oligomers to fibrillar surfaces through the secondary nucleation process17. Complementary to the experimental findings, computational simulations provide insights into residue-specific interactions between Aβ and membrane compositions during nucleation. A recent all-atom molecular dynamics (MD) simulation highlighted the presence and membrane interactions of an Aβ1-42 dimer during the nucleation phase in phosphatidylcholine bilayers and proposed the formation of an F19-F20 π-π ring stacking and a D23-K28 salt bridge in this early-stage dimeric intermediate18. Currently, there is limited residue-specific experimental information available regarding the membrane-associated Aβ1-42 nucleation intermediate states.

Solid-state nuclear magnetic resonance (ssNMR) spectroscopy has been widely used to characterize the molecular structure and dynamics of Aβ fibrils and stable oligomers, providing a highly effective approach for exploring the residue-specific features of Aβ-membrane nucleation intermediates. The current study examined site-specific interstrand distances in membrane-associated Aβ1-42 intermediate states and their proximities to lipid phosphate groups, using 13C-PITHIRDs-CT19 and 13C-31P REDOR20 ssNMR spectroscopy. Both quantitative ssNMR approaches detect the decay of 13C signals from isotope-labeled sites due to the reintroduction of 13C-13C or 13C-31P dipolar couplings, which are inversely proportional to the internuclear distances. A faster time-dependent evolution of 13C signal intensities indicates stronger dipolar coupling and therefore corresponds to shorter internuclear distances. Combined with selectively single-13C isotope-labeled samples, these approaches mapped quantitative, time-dependent changes in Aβ peptide chain assembly and membrane localization during the early stages of membrane-associated Aβ aggregation. Additionally, our previous dynamic nuclear polarization (DNP)-enhanced ssNMR studies demonstrated significant signal enhancements in isotope-labeled Aβ samples in the presence of excess natural abundance lipids, enabling the detection of molecular interactions through multidimensional NMR spectroscopy21,22.

Results

Aβ1-42 achieves enriched β-strand conformation and inter-strand proximities throughout the primary sequence before the onset of fibrillation

We first applied CD spectroscopy (spectra shown in Supplementary Fig. S2) and ThT fluorescence spectroscopy to determine the timeframe of the nucleation phase in the membrane-associated Aβ fibrillation process. As shown in Fig. 1A, both Aβ1-40 and Aβ1-42 exhibited typical nucleation-driven aggregation processes with sigmoidal ThT kinetics. Fitting these data yielded \({t}_{{lag}}\) values of 20.0 ± 1.5 h for 10 μM Aβ1-40 and 10.5 ± 2.2 h for 5 μM Aβ1-42. Notably, Aβ1-42 exhibited approximately two-fold faster fibrillation kinetics at half the concentration of Aβ1-40, consistent with our previous finding that Aβ1-42 nucleates first in mixtures containing three times the amount of Aβ1-40. The time-dependent changes in CD intensity at 218 nm, typically assigned to the β-strand conformation, revealed an earlier onset of the β-strand secondary structure compared to fibril elongation. The best-fit half-time of CD signal for 10 μM Aβ1-40 and 5 μM Aβ1-42 was 15.7 h and 2.1 h, respectively.

Fig. 1: Characterization of the membrane-associated Aβ1-42 assembling.
figure 1

A ThT fluorescence (left y-axis) and CD intensity at 218 nm (right y-axis) along the incubation of membrane-associated Aβ1-40 (10 μM, top) and Aβ1-42 (5 μM, bottom). Red curves show sigmoidal fitting of ThT kinetics to the equation \({I}_{t}={I}_{0}+({I}_{\infty }-{I}_{0})[1-\exp \left(-{\left({kt}\right)}^{n}\right)]\), from which the lag periods were derived. The blue curves and dashed lines show the same sigmoidal fitting for the time-dependent CD intensities and the derived \({{{\rm{t}}}}_{1/2}\) values, where \({{{\rm{I}}}}_{{{\rm{t}}}}-{I}_{0}\equiv \frac{1}{2}\left({I}_{\infty }-{I}_{0}\right)\). B Representative 13C-PITHIRDs-CT spectra for residues A2 and A30 with 3- and 15-h incubation in POPC liposomes. 13C-13C dipolar evolution times were given on top of the spectra. C Plots of corrected residue-specific 13C-PITHIRDs-CT dephasing curves (see S.I. for the natural abundance correction) at different incubation times. Color-coding: black, 3 h; red, 6 h; blue, 9 h; orange, 12 h; purple, 15 h. Error bars represent the signal-to-noise levels of individual sets of 13C-PITHIRDs-CT spectra. D Contour plots for the fitting of 13C-PITHIRDs-CT dephasing results using SIMPSON (details provided in S.I.). Contour levels show 1-7 times of \(\sigma \equiv \sqrt{2(N-n)}\) around \({{{\rm{\chi }}}}_{\min }^{2}\), where N is the number of data points \((N=7)\) and n is number of fitting parameters \((n=2)\).

Full size image

To understand the molecular-level structural changes of Aβ1-42 during nucleation and its differences from Aβ1-40, we applied 13C-PITHIDRs-CT spectroscopy19 to singly 13C-labeled samples (see Supplementary Table S1 for a list of samples and ssNMR experiments), with incubation times ranging from 3 to 15 h. Representative 13C-PITHIRDs-CT spectra with 3- and 15-h incubation were shown in Fig. 1B (with additional spectra provided in Supplementary Fig. S3). Quantitative analysis took into account the natural abundance of lipid carbons that overlapped with the 13C chemical shifts with C’, Cα, or Cβ. The actual fractions of membrane-bound Aβ1-42 at various incubation times were determined by the bicinchoninic acid (BCA) assay (Supplementary Fig. S4). The corrected 13C-PITHRIDs-CT decay curves (Fig. 1C) were fit to derive the average 13C-13C distances (Table 1 and Fig. 1D) using the SIMPSON simulation package23.

Table 1 Best-fit residue-specific 13C-13C distances (in Å) for intermediate membrane-associated Aβ1-42 assemblya
Full size table

In general, the 13C-PITHIRDs results clearly demonstrated the time-dependent formation of parallel-β-sheet structures, as indicated by more rapid decay curves with longer incubation times. Before the onset of fibrillar elongation (i.e., at 3, 6, and 9 h of incubation), all tested residues in Aβ1-42 showed significant inter-strand proximity. The most significant parallel-β-sheet structural convergence occurred between 6 and 9 h of incubation, where the inter-strand distances at N-terminal residues such as A2, F4, and V12 decreased from >8.5 Å to 5.5–6.0 Å. This timescale was consistent with the CD and ThT results, which showed the maturation of the secondary β-strand conformation and the onset of fibrillation between 6 and 9 h. In comparison, we previously showed that only residues G25, G29, L34, and V36 in membrane-associated Aβ1-40 with 15-h incubation had inter-strand distances ranging from 6.0 to 6.5 Å20. Residues closer to the N-terminus, such as F19 and A21, remained farther apart with interstrand distances longer than 8.0 Å. Thus, structural convergence during the nucleation phase is significantly accelerated throughout the entire primary sequence in Aβ1-42 compared to Aβ1-40.

The nucleation of Aβ1-42 modulates membrane fluidity and phospholipid dynamics, driven by membrane-surface-embedded oligomers

Compared to Aβ1-40, Aβ1-42 induced more significant local membrane disruptions, as evidenced by shortened lag periods and higher levels of leakage (Supplementary Fig. S5). Since the Aβ peptides were added exogenously to liposomes, we anticipated that the physicochemical properties of phospholipid headgroups would be modulated during the membrane-associated nucleation process. Figure 2A shows the time-dependent changes of Laurdan fluorescence general polarization (GP) upon incubation with Aβ, a measurement of membrane fluidity24. An increase in GP was induced by the nucleation and fibrillation of Aβ1-42, but not Aβ1-40, correlating with the stronger membrane-modulation effects of Aβ1-42. Similar increase of Laurdan fluorescence GP was observed in cholesterol-inserted model membranes, indicating the increase of hydrophobic interactions20. Modulation of microscopic phospholipid dynamics by Aβ1-42 was further assessed by 31P relaxation spectroscopy (Supplementary Figs. S6 and S7 for representative 31P spectra and decay curves). Quantitative analysis of 31P T1 and T2 relaxation rates provided correlation times for the microsecond timescale lateral diffusive motion (Fig. 2B) and the nanosecond timescale uniaxial rotational/wobbling motion (Fig. 2C) of lipids12,25,26. A general decrease trend in these correlation times (Supplementary Table S2) as a function of incubation time implied an increase of local phospholipid dynamics due to membrane-associated nucleation of Aβ1-42. Temperature-dependent changes in correlation times revealed decreases in activation energies for both motions (Supplementary Fig. S8). The modulations of phospholipid dynamics agreed with the enhancement of membrane content leakage observed for Aβ1-42, indicating the induction of local defects by the aggregation process27.

Fig. 2: Characterization of the phospholipid dynamics of bilayers.
figure 2

A Plot of the change of Laurdan fluorescence general polarization of POPC liposomes upon incubation of 10 μM Aβ1-40 and 5 μM Aβ1-42. Error bars represent standard deviation from three repetitions (n = 3). Plots of microsecond timescale (B) and nanosecond timescale (C) correlation times derived from the 31P T1 and T2 relaxation spectroscopy.

Full size image

To understand how the Aβ1-42-lipid interaction proceeds during the nucleation process of membrane-associated Aβ1-42 fibrillation, we performed 13C-31P REDOR spectroscopy28 (Fig. 3A and Supplementary Fig. S9) to probe residue-specific proximities to lipid headgroups. Experiments were conducted under the same conditions as those for the 13C-PITHIRDs-CT, allowing us to track the time-dependent changes in Aβ1-42 assembly and membrane location simultaneously. Plots of REDOR dephasing (i.e., \(\Delta S/{S}_{0}\equiv ({S}_{0}-{S}_{1})/{S}_{0}\), Fig. 3B; see Supplementary Methods for details of quantitative analysis) were fitted to derive the average 13C-31P distances and lipid-phosphate-proximity 13C populations (Fig. 3C and Supplementary Table S3). Most of the fittings yielded similarly correlated \({\chi }^{2}\) value patterns as functions of the two variables (i.e., 13C-31P distances and lipid-phosphate-proximity 13C populations), suggesting that for individual residues, it is the time-dependent trend, rather than the best-fit distances or populations values, that was validated. A few samples showed well-defined minimum \({\chi }^{2}\) ranges, especially for the samples with 12-h incubation time (Table S3). Several residues, such as G9, A21, A30, and G33, are located in proximity with lipid phosphates with ~5 Å 13C-31P distances. Interestingly, a global best-fit lipid-phosphate-proximity population of ~0.3 was obtained, suggesting that about one-third of the membrane-bound Aβ1-42 located close to lipids at this incubation time.

Fig. 3: Characterization of the membrane lipid proximity to Aβ1-42.
figure 3

A Representative 13C-31P REDOR spectra for residues A2, A21, and G38 with 3–15-h’ incubation in POPC liposomes and 24.1 ms 13C-31P dipolar evolution time. Horizontal dashed lines highlight the differences between S0 and S1 spectra. B Plots of corrected residue-specific 13C-31P REDOR dephasing (\({{\rm{dephasing}}}\equiv ({S}_{0}-{S}_{1})/{S}_{0}\), see S.I. for the natural abundance correction) at different incubation times. Color-coding: black, 3 h; red, 6 h; blue, 9 h; orange, 12 h; purple, 15 h. Error bars represent the signal-to-noise levels of individual sets of 13C-31P REDOR spectra. C Contour plots for the fitting of 13C-31P REDOR dephasing results using SIMPSON (details provided in S.I.). Contour levels show 1–7 times of \(\sigma \equiv \sqrt{2(N-n)}\) around \({\chi }_{\min }^{2}\), where N is the number of data points \((N=6)\) and n is number of fitting parameters \((n=2)\).

Full size image

Combined with the 13C-PITHIRDs-CT results, the interpretation of the intermediate states is summarized schematically in Fig. 4. The time-dependent behavior of individual residues is presented in three color-coded columns: (1) The greyscale (left, Fig. 4) shows the convergence to parallel-β-sheet assemblies. All residues exhibited shorter inter-strand distances at longer incubation times, but with residue-specific rates of structural convergence. The segment F19-A30, as well as discrete regions around M35 and the C-terminus, reached shorter average inter-strand distances earlier in the nucleation process compared to other parts of the Aβ1-42 sequence. Interestingly, residues F19 and A21 showed a unique trend, with longer interim inter-strand distances at 6 h. These observations suggest that shorter segments from the middle to the C-terminus of Aβ1-42 primary sequence may form discrete nucleation sites, and transient registry shift could occur within segments like F19-A21. Out-of-registry β-sheets and antiparallel β-sheets structures have been reported in Aβ oligomers and protofibrils, and these structures are typically associated with higher level of cytotoxicity compared to mature parallel-in-register β-sheet fibrils29,30,31. (2) The red and blue scales (middle and right, Fig. 4) represent residue-specific lipid-phosphate proximity. Several residues, including G9, F19, A21, G25, A30, G33 and G38, exhibited a consistent decrease in lipid-phosphate-proximity populations. Some of these residues, such as A30 (~5.0–7.0 Å), A21 (~5.5–7.5 Å), G9 (~5.0–7.0 Å), and G38 (5.0–7.5 Å), maintained relatively fixed 13C-31P distance range, indicating their proximity to lipid headgroups throughout the nucleation phase. As the β-sheet structures grew, their lipid-phosphate-proximity population decreased because of the shielding of effective 13C from lipids. (3) A few residues, such as F4, V12 and M35, remained distant from lipid headgroups throughout nucleation. Other residues, especially the N-terminal A2 and C-terminal A42, consistently stayed close to lipid 31P. Given the broad distribution of 31P-proximity residues across the entire Aβ1-42 sequence, it is reasonable to conclude that the peptide adopts a surface-embedded position during nucleation. A recent all-atom MD simulation investigated the early-stage membrane interactions of Aβ1-42 dimers32, which revealed that residues with strong lipid headgroup interactions were primarily located at the N-terminus and from F20 to A30. This latter segment aligns well with the current 13C-31P REDOR results. While the MD simulation suggested that N-terminal interactions may involve more specific interactions with PE and PS lipids, which are absent in the present membrane bilayer model, the current REDOR data still show proximity at A2, consistent with the strong interactions observed between D1 and PC lipids in the simulation.

Fig. 4: Color-scale presentation of the time-dependent and residue-specific Aβ1-42 chain assembly and lipid headgroup proximity.
figure 4

Color scales are derived from quantitative fitting of 13C-PITHIRDs-CT and 13C-31P REDOR results described in previous sections.

Full size image

DNP-ssNMR results demonstrate molecular-level structural convergence during the membrane-associated nucleation of Aβ1-42

DNP-ssNMR spectroscopy has proven effective in providing detailed information about molecular structures and interactions during the Aβ fibrillation process21,33,34,35. In this study, membrane-bound Aβ1-42 was collected at various incubation times, and the membrane-permeative bi-radical AsymPol-POK36,37 was added to the hydrated membrane-Aβ1-42 pellets. The samples typically contained 0.4–0.7 mg of isotope-labeled Aβ1-42 aggregates, diluted by a 100–150 molar excess of phospholipids. DNP application at 14.1 T/600 MHz yielded between 24 and 52-fold signal enhancements, characterized as the intensity ratio between NMR spectra acquired with and without microwave irradiation (\({\varepsilon }_{{{\rm{on}}}}/{\varepsilon }_{{{\rm{off}}}}\)), depending on the sample incubation time and spectral frequency range (Supplementary Fig. S10). The nuclear hyperpolarization buildup times range from 1.2 and 3.0 s.

Two-dimensional (2D) 13C-13C DARR spectra with 25 ms (Supplementary Fig. S11) and 250 ms mixing times (Fig. 5A) were collected to investigate the time-dependent structural convergence of membrane-bound Aβ1-42. At short incubation times (e.g., one and 2 h), most intra-residue cross peaks, including the backbone C’/Cα and Cα/Cβ cross peaks, were poorly resolved. Only G38, a residue near the C-terminus of Aβ1-42, exhibited a relatively strong C’/Cα cross peak (representative 1D slices are shown in Fig. 5B). After 6 h of incubation, intra-residue cross peak intensities were significantly increased, and certain inter-residue cross peaks, such as F19-I32 and/or F19-L34, began to emerge. These inter-residue contacts, which have been reported in several previous Aβ1-42 fibril structures as part of the fibrillar cores38,39,40, suggest that specific tertiary structural domains may already be forming during the early stages of nucleation. Structural convergence continued with further incubation at 9 and 15 h, where a dominant set of cross peaks appeared for most residues. Notably, the N-terminal E3 also showed a predominant set of intra-residue cross peaks, indicating that the entire Aβ1-42 peptide had adopted a rigid and structured conformation at this stage. This observation is consistent with the 13C-PITHIRDs results, where residues A2 and F4 exhibited parallel-β-sheet structural features at 15 h. The structural convergence of membrane-associated Aβ1-42 was further confirmed by 2D DQ-SQ correlation spectra (Fig. 5C), where well-defined intra-residue cross peaks were observed only after 9 h of incubation.

Fig. 5: DNP-ssNMR spectra of the Aβ1-42-membrane samples at intermediate incubation times in POPC liposomes.
figure 5

All ssNMR samples contain selective-13C isotope labeling at residues E3, V12, F19, A21, I32, L34 and G38. A 2D 13C-13C DARR spectra with 250 ms mixing time. Colored rectangles highlight the spectral regions from which the 1D slides are extracted. B 1D slides that represent the 13C chemical shifts in the t1 dimension of the rectangle regions. Color-coding for panels (A, B): red, intra-residue cross peaks; blue, inter-residue cross peaks; purple, cross peaks between 13C-sites in Aβ1-42 and natural abundance 13Cs in lipids. C 2D DQ-SQ correlation spectra acquired on the same samples as for the DARR spectra. Colored-lines in the 15-h spectrum highlight the intra-residue aliphatic cross peaks for individual labeled residues. All spectra were processed with 50 Hz Gaussian line broadening in each dimension.

Full size image

Lastly, we observed cross peaks between the aromatic 13C sites in F19 and the natural abundance lipid aliphatic 13C sites (i.e., ~31–33.5 ppm, corresponding to the most abundant lipid carbons C4-C13). Interestingly, these cross peaks exhibited higher intensities at shorter incubation times (e.g., 1–6 h) but diminished afterward, showing an opposite trend compared to the inter-residue cross peaks between F19 and I32/L34. This observation suggests that the aromatic side chain of F19 was initially exposed to the hydrophobic interior of the bilayer and later incorporated into the formation of early-stage tertiary structural domains in fibrillar core. Additionally, F19 likely inserted into the outer leaflet of the bilayer, which is consistent with the 13C-31P REDOR results, where F19 was found to be far from the phospholipid headgroups. Similar membrane positioning and hydrophobic interactions involving F19 were observed in a previous all-atom MD simulation18.

Discussion

Molecular-level structural convergence along the membrane-associated Aβ1-42 nucleation process

Multiple sets of biophysical and structural characterizations were conducted in the current study to systematically map the intermediate states of membrane-associated Aβ1-42 throughout the fibrillation process. Fitting the sigmoidal ThT kinetics curves revealed three characteristic phases at the early stages of nucleation and fibrillation: (1) A “silence phase” where the ThT fluorescence intensity remains unchanged up to ~8 h, suggesting the absence of parallel-β-sheet-like stacking constructs in Aβ1-42; (2) A “transition phase” between 9 and 12 h, during which the increase of ThT fluorescence indicates a transition towards fibrillar elongation; (3) The ThT kinetics then exhibit an exponential increase, signifying the typical “elongation phase”. By combining 13C-PITHIRDs-CT, 13C-31P REDOR, and 2D DNP-ssNMR spectroscopy, we examined the molecular-level and residue-specific details of the structures and lipid interactions of Aβ1-42 intermediates across these three phases.

As illustrated in Fig. 6, significant molecular-level structural transitions have already emerged during the “silence phase”. First, membrane-associated Aβ1-42 achieves secondary structural convergence, supported by the development of negative ellipticity at 218 nm in CD spectroscopy and the amplification of predominant intra-residue cross peaks in DNP-ssNMR from 1 to 6 h. Second, segments within F19-A30 and M35-A42 show close inter-strand distances, though parallel-β-sheets have not yet formed. These segments represent the typical hydrophobic cores in mature Aβ1-42 fibrillar structures41,42,43,44. In the meanwhile, certain tertiary contacts, such as F19-I32/L34, start to form, which also imply the assembly of early fibrillar core. A recent ssNMR study captured the early-stage Aβ1-40 intermediate states using rapid mixing and freezing techniques, reporting a similar time-dependent increase in the cross-peak volumes, which indicates stronger intra- and inter-residue interactions34. We observed that individual residues within these segments showed distinct time-dependent assembling trends. For instance, a transient increase of interstrand distances was observed for residues F19 and A21. In addition, most glycine residues (e.g., G33 and G38) exhibited a slower assembling rate compared to other sites. These results suggest that the formation of parallel-β-sheet occurs at discrete short pieces within the F19 to C-terminal segment.

Fig. 6: Schematic presentation of the membrane-associated Aβ1-42 intermediate states.
figure 6

The blue curve indicates a typical ThT fluorescence build-up trace. Red arrows indicate the segments that adopt the β-strand conformation. Cyan, purple, and orange circles highlight the assemblies of β-sheets, tertiary contacts, and N-terminal domains, respectively.

Full size image

During the “transition phase”, a predominant set of intra-residue cross peaks (i.e., secondary structure) matures, and the inter-strand distances decrease further to 5–7 Å across the Aβ1-42 sequence. At this stage, the most significant structural convergence occurs at the N-terminal segment of Aβ1-42, including residues A2, F4, and V12. This indicates the formation of a rigid, fibrillation-prone nucleus. Highly ordered N-terminal segments were shown to stabilize the tertiary and/or quaternary interfaces of Aβ1-42 fibrillar cores. Additionally, the lipid-phosphate-proximity population for residues F19-A30 shows a decrease trend between 6 and 9 h, suggesting the initiation of parallel-β-sheet-like assemblies, during which most 13C sites are incorporated into larger β-sheet-like plaques and shielded from lipids. By 15 h, when the ThT kinetics enter the “elongation phase”, no further changes are observed in the secondary structure of Aβ1-42. However, the inter-strand distances and lipid-phosphate-proximity populations further decrease, highlighting the maturation of parallel-β-sheet structures.

The molecular basis of different membrane-disrupting properties between Aβ1-42 and Aβ1-40

Compared to the abundant Aβ1-40 in the human brain, the more pathological Aβ1-42 is known to be more aggregation-prone and to induce greater cellular toxicity at lower concentrations45,46. Co-incubation of Aβ1-40 and Aβ1-42 at the physiological molar ratio (i.e., ~3:1) exhibited multi-stage fibrillation features, both in the presence and absence of membrane-mimicked liposomes, with the initial aggregation originating from the lower-abundance Aβ1-4222,47. Recent cryogenic electron tomography (Cryo-ET) studies indicated that both Aβ1-40 and Aβ1-42 induced significant membrane curvature changes and local defects within membrane leaflets in their oligomeric states, but not in their monomeric forms27,48. However, few pronounced differences between Aβ1-40 and Aβ1-42 were observed. Therefore, from a membrane disruption perspective, it remains unclear whether and how the differences in their aggregation properties correlate with their varying cellular toxicity levels.

Our results provide insights into the molecular basis for the differing membrane-disruptive effects of Aβ1-42 and Aβ1-40, with Aβ1-42 inducing more significant content leakage and membrane fluidity modulation. A recent single molecular fluorescence spectroscopy study also showed different membrane insertion abilities of Aβ1-42 and Aβ1-40 might correlate with their distinct cytotoxicity levels49. We previously characterized the residue-specific inter-strand distances and lipid-phosphate-proximity populations in membrane-associated Aβ1-40 at 5 and 15-h incubation times20. Given the ThT-based lag period of ~20 h, these time points correspond to the first phase of fibrillation in Aβ1-40, analogous to the 1–6-h time points for Aβ1-42. Three major differences between Aβ1-42 and Aβ1-40 were observed: First, Aβ1-42 adopts closer inter-strand distances, indicating tighter β-sheet-like assemblies compared to Aβ1-40. At the residue-specific level, F19, A21, G25, A30, and M35 in Aβ1-42 all showed inter-strand distances of 5–6 Å, and G33 showed 6–7 Å. In contrast, in Aβ1-40, F19, A21, and A30 remained at 7.5–8 Å. This trend of tighter packing at early stages was extended to the C-terminal residues G38 and A42, suggesting that the entire C-terminal half of Aβ1-42 (e.g., F19-A42) tends to assemble more rapidly. For Aβ1-40, however, the initial assembly is likely limited to the shorter segments within the C-terminal and/or loop domains (e.g., G25-G29 and around L34).

Second, the lipid-phosphate-proximity populations at the early stage for Aβ1-42 were significantly higher compared to Aβ1-40. For example, residues A2, G9, A21, G25, and A30 in Aβ1-42 all exhibited > 50% 31P-proximity populations at the 3–6-h incubation stage, while the maximum lipid-phosphate-proximity population for any probed residue in Aβ1-40 was less than 10%. This result suggests that low-order Aβ1-42 oligomers, which form rapidly and possess well-defined parallel-β-sheet-like structures in the F19-A42 domain, dominate the interactions with membranes at the nucleation stage. It has been shown previously that Aβ1-42 is more likely to form stable small oligomers (e.g., trimer, pentamer, hexamer, etc.) in solution, which act as effective nucleation species50,51,52. The size of these proposed oligomers is consistent with the lipid-phosphate-proximity populations, where about one-half to one-third of the isotope-labeled 13C sites locate close to 31P between 9 and 12 h. Similar oligomers may form rapidly in the membrane environment and directly contribute to membrane disruption.

Third, the same region in the Aβ1-42 primary sequence, likely within the F19-A30 segment, initiates both the interstrand assembly and membrane-disruptive interactions. This is supported by the combination of 13C-PITHIRDs-CT and 13C-31P REDOR results (Fig. 4), where residues F19, A21, G25, and A30 exhibited both close inter-strand distances and high lipid-phosphate-proximity populations at the 3–6-h incubation stage. This contrasts our previous findings for Aβ1-40, where residues with high lipid-phosphate-proximity populations (e.g., F19, A21, and G25) and those with close inter-strand distances (e.g., G25, G29, L34, and V36) were separated20. We hypothesize that in Aβ1-42, this overlapping between inter-strand packing and membrane interaction leads to competition between peptide chain assembly and membrane insertion, which may stabilize the rapidly forming low-order oligomers that can more efficiently insert into the bilayer. For Aβ1-40, on the contrast, the binding of peptides and the initial interstrand assembling may occur in different primary sequence regions, which allows the formation of larger exterior Aβ assemblies with less membrane-disruptive effects. These differences may explain the higher levels of content leakage and membrane fluidity changes observed with Aβ1-42 compared to Aβ1-40.

Conclusions

The present ssNMR studies provide a molecular-level understanding of the membrane-associated nucleation process of Aβ1-42. Through quantitative analysis of intermediate-state samples with selective isotope labeling, we demonstrate that surface-embedded, low-order Aβ1-42 oligomers strongly interact with the phospholipid headgroups of membranes during nucleation towards fibril formation. Compared to the less pathological Aβ1-40, the membrane-interactive Aβ1-42 shows closer proximity to lipid headgroups, broader lipid-contacting segments, and lower oligomer states, which may allow for more effective insertion into bilayers. These differences correlate positively with the higher levels of membrane-disruptive effects and cytotoxicity exhibited by Aβ1-42. Future studies may also involve mixture of Aβ1-40 and Aβ1-42 in more biologically relevant membrane-mimicked model systems. Additionally, residues that undergo structural transitions and converge at various stages of membrane-associated nucleation are identified. This information may assist in the future design of agents targeting these early-stage intermediates.

Methods

Peptide synthesis and purification

All Aβ peptides (Aβ1-40 m/z = 4329.8 Da; Aβ1-42 m/z = 4514.1 Da) were synthesized using a microwave-assisted peptide synthesizer (Biotage Initiator+ AlstraTM) following standard Fmoc-based solid-phase peptide synthesis protocols. The crude peptides were purified by HPLC (Agilent 1260 Infinity) employing a reversed-phase C18 column, and the products were verified by LC-MS to >95% purity (Shimadzu LCMS 2020, see Supplementary Fig. S1). Detailed synthesis and purification conditions are provided in the Supporting Information.

Liposome preparation and exogeneous addition of Aβ

Liposomes were prepared following a standard protocol established in our previous studies10,20,53,54. Briefly, the desired phospholipids were dissolved in chloroform, which was then removed by gentle N2 flow, followed by the formation of lipid films under vacuum for at least 8 h. The dried lipid film was resuspended in 10 mM phosphate buffer (pH 7.4, containing 0.01% NaN3) and subjected to 10 freeze-thaw cycles, alternating between liquid N2 and 50–60 °C water bath. The resuspended liposomes were extruded through 100 nm (for Thioflavin-T fluorescence and circular dichroism spectroscopy), 200 nm (for Laurdan fluorescence assay), or 400 nm (for ssNMR measurements) pore-size membrane filters (using a mini-extruder set from Avanti Polar Lipids, Inc.). Liposome solutions were freshly prepared before use. For all experiments, except for the calcein leakage assay, a single-component liposome containing only 1-palmitoyl-2-oleoyl-glycerol-3-phosphocholine (POPC, Avanti Polar Lipids Inc.) was utilized. For the calcein leakage assay, liposomes were prepared with the following three compositions: POPC only, POPC and 1-palmitoyl-2-oleoyl-glycerol-3-phosphoethyleneamine (POPE) in a 3:1 molar ratio, and POPC, POPE, and 1-palmitoyl-2-oleoyl-glycerol-3-phosphoserine (POPS) in a 3:1:1 ratio.

Lyophilized Aβ peptides were dissolved in hexafluoro-isopropanol (HFIP, Sigma Aldrich Inc.) at a concentration of 2 mg/mL. The solution was bath-sonicated for 1 min, followed by overnight incubation at room temperature to ensure complete dissolution. HFIP was then removed by gentle N2, and the sample was dried under vacuum for at least 8 h. The peptide film was dissolved in dimethyl sulfoxide (DMSO, Sigma Aldrich Inc.), and the concentration was determined using a nanodrop UV-VIS spectrometer (Thermo Scientific Inc.) at 280 nm, based on a molar extinction coefficient of 1420 M−1 cm−1 for both Aβ1-40 and Aβ1-42. For all samples, the Aβ-DMSO stock was diluted into liposome solution to reach the final peptide concentration of 10 μM for Aβ1-40 and 5 μM for Aβ1-42, with a 1:30 Aβ to total lipids molar ratio for Aβ1-40 and 1:90 molar ratio for Aβ1-42.

Biophysical Characterizations of Aβ Aggregation and Aβ-Membrane Interactions

The present work employed a combination of thioflavin-T (ThT) fluorescence assay, circular dichroism (CD) spectroscopy, calcein leakage assay, and Laurdan fluorescence assay to characterize Aβ aggregation and membrane interactions. All assays followed standard procedures for liposome preparation, exogenous addition of Aβ peptides, and incubation. Most assays used a simplified liposome model consisting of POPC. For the calcein leakage assay, three liposome models containing mixtures of POPC, POPE, and POPS were employed. Experimental details and associated data analysis for each assay are provided in the Supporting Information.

The ssNMR Spectroscopy

For 13C-PITHIRDs-CT and 13C-31P REDOR experiments, samples that contain Aβ1-42 in POPC liposomes were incubated for the desired times and collected by ultracentrifugation on a Beckmann Coulter ultracentrifuge (Optima MAX-TL, 50,000 rpm for 30 min at 4 °C). The pellets were lyophilized, packed into a 2.5 mm magic angle spinning (MAS) rotor, and rehydrated using 1.0 μL/mg deionized water. For 31P relaxation spectroscopy, the pellets were packed directly into the 2.5 mm MAS rotors without lyophilization. For dynamic nuclear polarization (DNP) assisted ssNMR spectroscopy, samples were prepared by mixing the Aβ1-42-liposome pellets with a stock solution that contains 10 mM Asympol-POK36 in 13C-depleted d8-glycerol/D2O/H2O in a ratio of 60/30/10 vol%. Around 35 μL of the stock solution was directly added into the wet sample pellets; the mixture was vortexed vigorously for about 1 min55 and transferred to a 3.2 mm sapphire rotor and closed with Vespel caps. The DNP-ssNMR rotors were frozen immediately and kept frozen before measurements.

Non-DNP ssNMR spectroscopy, including 13C-PITHIRDs-CT, 13C-31P REDOR, and 31P relaxation experiments, was performed on a 600 MHz Bruker Avance III spectrometer equipped with a 2.5 mm TriGamma HXY magic-angle spinning (MAS) probe tuned to 1H/31P/13C. PITHIRDs and REDOR experiments were conducted on rehydrated Aβ-membrane pellets containing ~2 mg of selectively 13C-labeled Aβ aggregates and about 10 mg of lipids. The pulsed-spin locking (PSL) acquisition method56 was applied to individual labeled sites to enhance spectral sensitivity. The magic-angle spinning (MAS) frequencies were set to 20,000 ± 2 Hz and 10,000 ± 2 Hz, respectively, for PITHIRDs-CT and REDOR experiments. The For PITHIRDs-CT and REDOR, a complete set of 13C spectra from a single labeled site (comprising eight and six dephasing time points, respectively) was collected over ~16 and ~24 h. The 31P relaxation experiments were performed on fully hydrated Aβ-membrane pellets, with sample temperatures monitored via 1H chemical shifts of H2O. A full set of T1/T2 spectra at four different temperatures was acquired within 6 h.

MAS-DNP experiments were conducted at the National High Magnetic Field Laboratory (Tallahassee, USA) using a 600 MHz/395 GHz system, equipped with a Bruker NEO console and a 3.2 mm wide-bore MAS-DNP probe57. For each sample, a 1H-13C cross-polarization (CP) experiment was performed to assess sensitivity enhancement. Additionally, two 2D SQ-SQ DARR spectra with 25 ms and 250 ms mixing periods, along with a 2D DQ-SQ spectrum, were acquired within 8–15 h. Further details on the experimental setup, sample preparation, and quantitative data analysis procedures for ssNMR measurements are provided in the Supporting Information.