The present study investigates how anle138b binds to and affects the L1 Aβ₄₀ fibrils. The structural and functional effects of anle138b treatment are analyzed under two conditions: pre-aggregation (pre-treatment) and post-fibril formation (post-treatment). By utilizing solid-state nuclear magnetic resonance (ssNMR), cryo-EM, and molecular dynamics (MD) simulations, we aim to gain a detailed understanding of the molecular interactions between anle138b and L1 Aβ₄₀ fibrils.

To assess the effect of anle138b on L1 Aβ₄₀ fibril growth, Aβ₄₀ monomers were aggregated in the presence of DMPG liposomes containing anle138b at three different Small Molecule-to-Protein molar Ratios (SMPRs; anle138b: Aβ₄₀) of 0, 0.6, or 1.2. Thioflavin T (ThT) fluorescence and circular dichroism (CD) analyses revealed a concentration-dependent inhibition of fibril formation by anle138b (Fig. 1a, Supplementary Fig. 1). CD spectra showed β-sheet signatures under all conditions at the beginning of incubation after mixing Aβ₄₀ with DMPG liposomes (Supplementary Fig. 1b), indicating similar initial secondary structures regardless of the presence of anle138b. After 48 hours of incubation, a strong β-sheet signal was retained in the control sample (SMPR = 0), whereas β-sheet formation was reduced in samples treated with anle138b (SMPRs = 0.6 and 1.2). Fibril formation was most reduced at SMPR 1.2, notably indicating a dose dependence (Supplementary Fig. 1a, c). These results were further supported by two complementary approaches performed on a dedicated fibril sample (pre-treatment, post-treatment, and control fibril): 1D (H)N CP spectrum signal intensity, and quantification of fibrils by negative stain EM and CD spectroscopy (Fig. 1b, c, Supplementary Fig. 3). All three methods consistently indicated a dose-dependent inhibition of fibril formation by anle138b.

To evaluate the effect of anle138b at different stages of fibril formation, we compared three experimental conditions: 1) pre-treatment (SMPR = 1.2, with anle138b added before initiating fibril formation), 2) post-treatment (SMPR = 1.2, with anle138b applied after fibril formation was completed), and 3) control (fibril without anle138b) (Supplementary Fig. 2a). Pre-formed fibrils used in the post-treatment condition were generated with DMPG liposomes under identical conditions to the control fibrils. Fibril formation under the pre-treatment condition was reduced by approximately 75% compared to the control (Fig. 1c). This observation was supported by analysis of the supernatant following ultracentrifugation: CD and ThT fluorescence spectra showed that β-strand-like species remained in the supernatant under pre-treatment conditions, indicating the presence of soluble, non-fibrillar Aβ₄₀ aggregates (Supplementary Fig. 3c, d). In contrast, these species were absent in the control or post-treatment samples, where nearly all Aβ₄₀ sedimented as fibrils (Supplementary Fig. 3c, d), i.e., post-treatment resulted in no significant change in fibril quantity (Fig. 1c, Supplementary Fig. 3b). The observed reduction in ThT fluorescence intensity in the post-treatment sample (the fibril treated with anle138b after their formation) is likely attributable to competitive binding between anle138b and ThT at shared or nearby fibril binding sites. Thus, the fluorescence intensity cannot be used as a measure of fibril quantity, and complementary readouts suggest that the fibril quantity has not decreased significantly (Fig. 1b, c, Supplementary Fig. 3b). This interpretation is consistent with previously reported interactions between small molecules and amyloid fibrils.

In addition, CD spectroscopy and negative stain EM imaging confirmed that fibrils remained structurally unchanged after post-treatment with anle138b, even after 96 hours of incubation at 37 °C (Supplementary Fig. 4). These findings indicate that mature fibrils retain their structural integrity in the presence of anle138b.

To determine whether the inhibition of fibril formation by anle138b is accompanied by structural changes, we analyzed the fibrils using cryo-EM and ssNMR. Consistent with previous findings, the ssNMR data indicate that L1 Aβ₄₀ fibrils represent the predominant species of the control-fibril sample, and the L1 structure was reconstructed from the cryo-EM data for each preparation (control fibril, post-treatment fibril, and pre-treatment fibril conditions). The cryo-EM analysis revealed no discernible structural differences between L1 Aβ₄₀ fibrils in the presence or absence of anle138b, suggesting that the compound does not significantly alter the L1 Aβ₄₀ fibril architecture (Fig. 2).

By contrast, ssNMR spectra revealed clear chemical shift perturbations (CSPs) in the presence of anle138b. In the 2D (H)NCA spectrum, substantial CSPs were detected for residues in the loop region between the second and third β-strand (Ala21-Gly33) (Supplementary Fig. 2b, c) of the L1 filament fold (Supplementary Fig. 5a). These spectral differences, relative to the untreated (control) fibrils, were observed following anle138b post-treatment of L1 Aβ₄₀ fibrils (SMPR 1.2). Notably, the large CSP values of up to 10 ppm in N and ~1.5 ppm in Cα suggest a direct interaction between anle138b and the fibril (Supplementary Fig. 5a).

Similar localized CSPs were observed under the pre-treatment condition (SMPR 1.2). This observation suggests that anle138b interacts with the loop region for both pre- and post-treatment conditions (Supplementary Fig. 5). Since these CSPs are observed across different parts of the L1 Aβ₄₀ fibril structure, the data indicate that anle138b may bind to multiple regions of the fibril, including the central cavity of the two protofilaments and also to exposed fibril surfaces. However, CSPs can also be the result of allosteric effects (vide infra).

In conclusion, the CSP results indicate that anle138b likely binds in the loop region (Ala21-Gly33), which includes the central cavity formed at the interface between the symmetry-related loop regions of the two protofilaments within the L1 Aβ₄₀ fibril. Cryo-EM analysis indicates that the L1 Aβ₄₀ fibril remains the predominant species in the presence of anle138b. The global fibril structure also remains unchanged.

To directly assess binding locations, we employed amino acid-specific isotope labeling strategies and DNP-enhanced NMR. Chemical shift perturbations (CSPs) may be due to direct binding of anle138b to specific amino acids of L1 Aβ₄₀ fibrils but could also be induced by allosteric structural changes. Therefore, DNP-enhanced NHHC experiments are required to localize the binding sites directly. For this purpose, the protein was labeled first uniformly with C and then with C, N-isoleucine or C, N-lysine, while anle138b was uniformly labeled with N at the central pyrazole group. The cross-peaks in the NHHC spectrum indicate close contact with the nuclei of the protein: the pyrazole N nuclei at 195 ppm are within approximately 5 Å of aliphatic C from the protein. These results provide direct evidence that anle138b binds to specific amino acids within the fibril (Fig. 3a, b), as detailed below.

Under the pre-treatment condition, a cross-peak appears between anle138b and the protein in the Cα region of the spectrum centered at 52 ppm, and a second peak is observed at about 40 ppm. The candidate residues that are consistent with these peaks include the Cα of His13, Ala21, and Ala30, and the sidechain resonances of Lys16, Lys28, Ile31, and Met35 (Fig. 3b, g). For a C, N-isoleucine-selectively labeled fibril, a cross-peak corresponding to Ile31 was identified as the interacting residue based on the Cβ chemical shift (Fig. 3b, g). In contrast, no cross-peaks were observed for a C, N-lysine-selectively labeled fibril, thereby excluding Lys16 and Lys28. Given that Ile31 points towards the central cavity and no contacts were detected with the lysine sidechains on the fibril surface, these data suggest that anle138b predominantly interacts with residues in the central cavity under the pre-treatment condition. The 52 ppm peak is consistent with contact to residue Ala30, which is next to Ile31.

In contrast, under the post-treatment condition, two prominent peaks corresponding to anle138b are observed at different positions, namely 49 ppm and 30 ppm (Fig. 3a, f) in the 2D NHHC spectrum of C-labeled protein. The Cα of residues such as Ala21, Ala30, Gly25, and Gly33 is consistent with the former peak, and the 30 ppm peak is consistent with the resonances of His14 (Cβ), Lys16, Lys28 (Cγ, Cδ), Ile31, Ile32 (Cγ1), Leu17, and Leu34 (Cδ). Leu17, Gly33, and Leu34 were excluded from consideration because of their buried location within the fibril. A DNP-enhanced 2D NHHC spectrum of isoleucine-labeled fibrils confirmed an interaction with Ile31 or Ile32. Lysine labeling also revealed a cross-peak at 30 ppm (Lys Cγ, Cδ), which, considering that these sidechains extend from the fibril, places anle138b on the fibril surface. These data confirm that anle138b interacts with surface-exposed residues such as Lys16, Lys28, and potentially Ile31 and/or Ile32. However, the small chemical shift difference (<1.35 ppm) between Ile31 Cγ1 and Ile32 Cγ1 resulted in peak overlap, preventing definitive assignment of the observed cross-peak to either residue. Therefore, binding to the central cavity cannot be definitively confirmed or excluded under the post-treatment condition.

Interestingly, the cryo-EM structures revealed an additional weak non-proteinaceous density within the fibril's central cavity, going from 0.31 without anle138b to 0.51 in the post-treatment fibril condition to 0.73 in the pre-treatment fibril condition (Fig. 2c, d; Supplementary Fig. 6). Non-proteinaceous density (primarily lipids) is observed outside the fibrils, but there is no systematic density change depending on the presence of anle138b, unlike in the central cavity.

To assess the binding behavior under post-treatment fibril conditions, we characterized concentration-dependent interactions using ITC and NMR.

Isothermal titration calorimetry (ITC) revealed that anle138b binds to L1 Aβ₄₀ fibrils with an exothermic and spontaneous profile (ΔH = -1.84 kcal/mol, ΔG = -8.45 kcal/mol). The dissociation constant (K) was determined to be 0.64 μM, with a binding stoichiometry of ~0.72 molecules per Aβ₄₀ monomer (Fig. 4b, Supplementary Table 2). CD spectroscopy and ThT fluorescence confirmed full conversion of Aβ₄₀ monomers into fibrils under both control and post-treatment conditions, consistent with near-complete sedimentation of L1 Aβ₄₀ after ultracentrifugation (Supplementary Fig. 3).

To further assess drug incorporation, we analyzed the supernatant after ultracentrifugation using 1D solution NMR. No free anle138b signal was detected at anle138b: Aβ₄₀ ratios of 0.2:1 and 0.8:1, while ~30% of the drug remained unbound at 1.2:1 (Supplementary Figs. 21, 22). These results align with the stoichiometry measured by ITC (Fig. 4b, Supplementary Table 2) and NMR titration (Fig. 4a). Together, they indicate that the majority of anle138b associates with the fibrils.

DLPG vesicles lacking anle138b produced no measurable heat signal in ITC (Supplementary Fig. 13b, Supplementary Table 2). This absence of detectable interaction was corroborated by ssNMR NOE and cryo-EM, which showed no DLPG-specific contacts or densities on the L1 Aβ₄₀ fibril surface (Figs. 2b, 5).

To resolve residue-specific changes and explore whether the binding stoichiometry observed in ITC correlates with the NMR spectra, we performed an NMR-based titration experiment using SMPRs of 0.2, 0.4, 0.8, and 1.2. The effects of binding were monitored via 3D (H)CANH spectra, analyzing CSPs and signal intensity changes (I_ratio). This revealed large chemical shift changes and slow exchange behavior in several residues (Fig. 4a).

Peaks associated with anle138b-bound fibrils emerged as early as SMPR 0.2 across the β2 strand, loop, and β3 regions, with residue- and region-specific differences in saturation behavior. Several residues exhibited both free and bound peaks between SMPR 0.2 and 0.8, with most sites saturating at 0.4 or 0.8. Notably, Gly29, Ala30, and Asn27 required SMPR 1.2 to reach saturation (Fig. 4a, Supplementary Fig. 11-13).

In more detail, in the β2 region, Leu17 showed early binding at SMPR 0.2 and saturated at 0.4, whereas Lys16 remained unaffected until signal loss at SMPR 1.2. In the loop region, multiple residues responded at low concentrations: Ala21, Gly25, and Gly33 reacted at 0.2, with Ala21 saturating at 0.4. Other loop residues (Gly29, Ala30, Ile31) responded at 0.4, while Asn27 showed delayed binding (slow exchange at 0.8; saturation at 1.2). Notably, Lys28 and Asp23 showed concentration-dependent signal disappearance at 0.4 and 0.8, respectively, and Ser26 was undetectable at 1.2. In the β3 region, Met35 responded early (0.2), and Leu34 entered slow exchange at 0.4 and saturated at 0.8. Together, these observations highlight region-specific differences in binding sensitivity, with the loop region displaying the broadest range of response behaviors.

Significant N chemical shift perturbations were observed for Ala30 (10.72 ppm), Ala21 (6.48 ppm), and Ile31 (4.65 ppm) in CSP-based titration experiments conducted under saturation conditions (SMPR = 1.2) (Fig. 4a). These large shift changes in amide nitrogen signals can result from changes in the hydrogen bonding environment, and do not necessarily reflect a structural change in the protein. Indeed, upon anle138b binding, water-protein NOE contacts in the 3D H(H)NH NOE spectrum increased for Gly29, Ala30, Ile31, and Ile32, while decreasing for Gly25 and Leu34 (Supplementary Fig. 17).

The same spectrum also revealed that lipid interactions between β-strand residues (Leu17-Val8, Leu34-Val36) and DMPG acyl chains were disrupted, while new lipid contacts were formed with loop residues Ala21-Gly25 (Fig. 5). The disappearance of Asp23 and Lys28 signals indicates changes in the structure or dynamics of charged sidechains at the fibril surface near the loop region. Previous studies reported that Asp23 and Lys28 exist in two distinct populations (Asp23/Asp23' and Lys28/Lys28'), of which only the Asp23-Lys28 pair forms a salt bridge. In the 2D CC-DARR spectra acquired under the SMPR 1.2 condition (mixing times = 20 and 200 ms), the peaks corresponding to the salt bridge conformation were absent, while only those corresponding to Asp23' and Lys28' were observed (Fig. 4a, Supplementary Fig. 8a, 15, 16). In addition to these residues, Ser26 also exhibited peak loss, and no cross-peaks were detected among Asp23(Cβ), Ser26(Cβ), Asn27(Cα/Cβ), and Lys28(Cε). Given that CP-based DARR spectra selectively detect rigid regions of the protein, these observations suggest enhanced flexibility or dynamic averaging in the loop region upon anle138b binding. At cryogenic temperature, differences in the chemical shifts of Lys28 Cγ and Cδ signals are evident: the room temperature assignments match pre-treatment fibrils, and these signals are absent in the post-treatment fibrils (Supplementary Fig. 10a). No major differences between pre- and post-treatment fibrils were observed for the isoleucine C resonances in the 2D CC -DARR spectra acquired at cryogenic temperatures (Supplementary Fig. 10a). Note that the linewidths of C resonances at cryogenic temperature limit the detection of more subtle local differences. Taken together with cryo-EM analyses showing that all three fibril preparations -- pre-, post-treatment, and control -- share the same overall structure (Fig. 2), these findings suggest that the observed changes in the loop region primarily reflect enhanced local flexibility and water/lipid exposure rather than stable or static structural rearrangements.

All-atom molecular dynamics (MD) simulations were employed to investigate the binding of anle138b to the L1 Aβ₄₀ fibril structure observed by cryo-EM and NMR. We first considered the binding of anle138b, in the presence or absence of DMPG lipids, corresponding to an L1 Aβ₄₀ fibril structure with and without surface-bound phospholipids (see Methods, Fig. 6, Supplementary Fig. 23). In both simulation sets, anle138b molecules established contacts with regions of the fibril surface, consistent with the NMR results of post-treatment fibrils (Fig. 6a-c). Notably, anle138b did not bind to the central cavity in either lipid-containing or lipid-free simulations. Lipid-free simulations show short distances (below 5 Å) between the pyrazole NH of anle138b and Cδ of residue Ile32, consistent with NMR data, while the lipid-containing simulations do not, suggesting competition of lipid and anle138b binding at the L1 Aβ₄₀ fibril surface near the position of Ile32 (Fig. 6b, c, Supplementary Fig. 18a, b).

In a second set of simulations, we probed the ability of anle138b to spontaneously bind to the central cavity between the two protofilaments of the L1 Aβ₄₀ fibril as observed under both pre-treatment fibril and post-treatment fibril conditions. To do so, the diffusion of anle138b in the simulation box was restricted to a cylindrical volume centered on the loop region by a flat-bottomed position restraint (see Methods, Supplementary Fig. 19a, b). Indeed, anle138b can insert spontaneously in the central cavity of the loop region, showing contacts with Val24, Gly25, Ser26, Asn27, and Ile31 (Fig. 6a, b). During the microseconds-long MD trajectories, anle138b moved into the filament in eight out of ten simulation replicas, traversing up to 5 or 6 monomer layers into the central cavity. Once fully buried inside the fibril, the anle138b molecules are retained, as no unbinding events were observed (Supplementary Fig. 19a, b).

Interestingly, MD simulation models with deprotonated Lys28 and hence weakened side chain interactions between residues Asp23 and Lys28, allowed for a higher anle138b insertion rate (ten out of ten simulation replicas) and increased probability of deeper penetration into the central cavity of the fibril (Supplementary Fig. 19c). The disruption of the Asp23-Lys28 salt bridge led to an increase in RMSF values in the loop region, indicating enhanced structural flexibility (Supplementary Fig. 19d).

Anle138b molecules binding to the fibril surface and inside the central cavity exhibited contact patterns to residues Lys16, Lys28, Ile31, and Ile32, reproducing the two principal anle138b binding modes observed in the pre- and post-treatment fibril conditions by ssNMR experiments (Fig. 6a, c). Examination of the molecular determinants of anle138b binding revealed that the surface binding around the loop region occurred through contacts to the polar side chains of Asp23 and Lys28 (Fig. 6d, e). Specifically, the MD simulations revealed that the bromine (Br) atom of anle138b interacts with the ε-amino nitrogen (N) of Lys28 through a halogen bond-like contact (Fig. 6d, Supplementary Fig. 20a), whereas the central cavity binding was dominated by close polar interactions between anle138b and the exposed protein backbone atoms, particularly of Gly25 (Fig. 6e). The simultaneous interactions via the pyrazole and bromophenyl moieties allow for a favorable binding of anle138b to the fibril.

Compared to the distribution in the bulk lipid phase, the bromophenyl ring of surface-bound anle138b orients preferentially such that the bromine and pyrazole nitrogen atoms align on the same edge of the molecule (Supplementary Fig. 20b). Interestingly, the prominent binding to Asp23 and Lys28 with this ring orientation leads to a perpendicular alignment of anle138b with respect to the fibril axis and allows for close stacking of the ligands (Supplementary Fig. 20c). In contrast, anle138b binding to other regions of the L1 Aβ₄₀ fibril surface, such as Lys16 or Gly37, as well as the central cavity, occurred with the molecules aligned parallel to the fibril axis.

In contrast to surface binding, the cavity binding mode showed no preferred bromophenyl ring orientation or anle138b alignment (bromophenyl moiety pointing down (0°) or up (180°) with respect to the fibril axis, Supplementary Fig. 18c, 20c, d). The observed anle138b binding mode characteristics were independent of the presence of DMPG lipids in the simulations, as well as the anle138b tautomer (Supplementary Fig. 20c, d).