Theoretical Screening of Peptide Scaffold

The primary driving force behind the aggregation of intrinsically disordered Aβ42 conformers to form fibrils is the hydrophobic clustering of nonpolar residues.^{[44, 45]} Therefore, disruption of hydrophobic clustering is critical for inhibiting the early oligomerization stages of Aβ42. Secondary structural predictions of Aβ42, obtained through the s2D method, reveal that the central region (K16–A21) and the C-terminal region (G38–A42) inherently favor β-sheet formation (Figure 1a).^[46] Given the crucial role of β-sheet structures in aberrant protein assembly, region-specific targeting of these segments could be a promising strategy to disrupt Aβ42 aggregate formation. To test this hypothesis, we employed peptide inhibitor candidates—P_a and P_b, listed in Table S1—that were previously designed based on the two cross-β regions in the fibrillar structure of Aβ42.^[35] We performed replica exchange molecular dynamics (REMD) simulations (detailed in the Methods section) to characterize the molecular behavior of these peptide candidates in the presence of Aβ42 and to assess their impact on the structural dynamics of Aβ42 (Figures S1–S3). We systematically varied their lengths from five to thirteen amino acids to explore the influence of one-dimensional extension of the peptide length on molecular interactions (Figure 1b).

The interactions between P_a-derived peptides and Aβ42 are predominantly concentrated in the central region (K16–A21) and the C-terminal region (A30–V36). The contact probability (detailed in the Methods section) between Aβ42 and the inhibitors increases with peptide length. Notably, the contribution of the C-terminal region of Aβ42 in the contact probability gradually increases from P_a5 to P_a13 (Figures 1b and S4). Our simulations revealed that the positively charged N-terminus and lysine residues of P_a-derived peptides interact with the negatively charged Asp7, Glu11, Glu22, or Asp23 residues of Aβ42 (Figures S4 and S5). Conversely, the C-terminus and negatively charged residues (Asp, Glu) of P_a-derived peptides form interactions with the N-terminus, Lys16, or Lys28 of Aβ42. Additionally, the nonpolar side chains of Aβ42, particularly residues ¹⁷LVFF²⁰, exhibit limited exposure to water molecules, favoring intermolecular interactions with P_a peptides. The longest P_a-derived peptide, P_a13, forms a salt bridge between its N-terminus and the C-terminus of Aβ42, while its C-terminus interacts with Lys28 of Aβ42, stabilizing the complex differently from the P_a11 (Figure S4). Overall, the arrangement of charged and nonpolar residues in P_a-derived peptides promotes frequent contacts with the central region of Aβ42, a trend that becomes more prominent with increasing peptide length. Similarly, the interaction map of P_b-derived peptides shows a gradual increase in contact probability with the central region (K16–A21) and the C-terminal region of Aβ42 as peptide length increases (Figure 1b).

To investigate the structural rearrangement of Aβ42 induced by the designed peptides, we analyzed the number of hydrogen bonds, the theoretical interaction energy, and the free energy landscape concerning for the dihedral angles of Aβ42 using the REMD simulation trajectories (Figure 1c–h). The number of hydrogen bonds between water molecules and Aβ42–peptide complexes gradually increases but plateaus with P_a11 and P_a13.

This trend suggests that the displacement of water molecules from the hydrophobic regions of Aβ42 into the bulk solvent, indicative of enhanced hydrophobic interactions, is maximized at P_a11 and P_a13 (Figure 1c).^[47] Similarly, the interaction energy between Aβ42 and each peptide levels off with P_a11 and P_a13, indicating that increasing peptide length beyond eleven residues does not contribute to stronger interactions with Aβ42 (Figure 1d). The representative structure of Aβ42 in complex with P_a11 reveals that P_a11 binds to the central region of Aβ42, effectively shielding the critical oligomerization domain responsible for aggregation (Figure 1e).^{[35, 48]} In the case of P_b-derived peptides, a higher degree of interactions is observed in the longer peptide (P_b9, P_b, and P_b13) (Figure 1f,g). However, the conformational landscape of Aβ42 is perturbed differently by P_b-derived peptides (Figures 1h and S6). When complexed with P_b13, Aβ42 continues to adopt partially folded conformations where the central region interacts with the C-terminal region (Figure 1h).

Although both P_a- and P_b-derived peptides interact with Aβ42, distinct interaction patterns were observed in the structural dynamics of Aβ42. Also, the molecular interaction with C-terminal region of Aβ42 is enhanced as the peptide length increases. These insights from computational screening were examined experimentally using the peptide inhibitor candidates in Table S1.

Experimental Validation of Optimal Inhibitor Scaffold

To evaluate the effects of the inhibitor candidates, we monitored the formation of Aβ42 fibrillar aggregates using a thioflavin T (ThT) fluorescence assay (Figure 2a–e). The normalized fluorescence intensity at the endpoint (24 h) of P_a5 treated Aβ42 showed a fluorescence intensity (105.6 ± 1.8%) comparable to the control group (Aβ42 w/o inhibitor). However, the relative fluorescence intensity at the endpoint decreased to 86.1 ± 5.9% and 85.1 ± 9.6% when treated with P_a and P_a9, respectively. A more significant reduction in fluorescence intensity was observed with P_a11 (71.4 ± 4.7%), whereas the inhibitory effect of P_a13 was not significantly improved (78.3 ± 1.6%). Following the ThT assay, transmission electron microscope (TEM) images after negative staining showed a gradual shift in the length distribution of Aβ42 fibrils (Figures 2f–g and S7). The control group formed short fibrillar aggregates after the 24 h of incubation (279.4 ± 19.0 nm). In contrast, all P_a-derived peptides induced broadly dispersed, longer fibrils, especially for P_a11 and P_a13 treated Aβ42 samples (816.9 ± 44.8 nm and 784.1 ± 48.9 nm, respectively). These results suggest that P_a-derived peptides impede nuclei formation during the early stages of Aβ42 fibrillar aggregation, thereby reducing the number of nuclei available for elongation. Consequently, this mechanism results in the formation of longer fibrils with a lower number density at the endpoint.^[49]

To assess the cytotoxicity of Aβ42 fibrils altered by the treatment of P_a-derived peptide, we assessed viability changes of the SH-SY5Y neuroblastoma cells (Figure 2h). Aβ42 fibrils reduced cell viability to 76.9 ± 1.0%, as measured by the modified thiazolyl blue tetrazolium bromide (MTT) assay. However, the Aβ42 fibrils co-incubated with P_a-derived peptides alleviated the cytotoxicity, and the cell viability increased up to 88.4 ± 1.4% in Aβ42 sample co-incubated with P_a11 peptide. This reduction in cytotoxicity is attributed to the inhibitory effects of P_a-derived peptides that decrease the number density of fibrillar aggregates and limit the formation of cytotoxic shorter aggregates. Consistent with MD simulations, P_a11 demonstrated the highest performance in both ThT and MTT assays. We hypothesize that the interaction energies observed in MD simulations correlate with the actual binding affinity between Aβ42 and P_a peptides, thereby influencing the suppression of Aβ42 aggregation.

To further validate this hypothesis, we examined the relative intrinsic binding affinities between Aβ42 and P_a peptides using electrospray ionization tandem mass spectrometry (ESI-MS/MS) (Figures S8, and S9). The results revealed a clear upward trend in binding affinity from P_a5 to P_a11, with a slight decline observed at P_a13. Notably, these findings are in good agreement with the trend observed in the interaction energies of MD simulations (Figure 1d). Considering that P_a11 exhibited the highest inhibitory effect among the P_a-derived peptides, we conclude that the strong association with Aβ42 is a critical factor in effectively suppressing Aβ42 aggregation and mitigating the cytotoxicity of Aβ42 aggregates. Following the analysis of the P_a-derived peptides, we investigated the effects of P_b-derived peptides on fibrillar aggregation of Aβ42. The tested P_b-derived peptides exhibited negligible effects on Aβ42 aggregation (Figure S10). The normalized ThT fluorescence after 24 h of incubation showed only a slight reduction of up to 6.4% in the P_b13-treated Aβ42 sample (Figure S10a). Also, P_b-derived peptides did not induce significant changes in the length distribution of Aβ42 fibrils (Figures S10b and S11). Consistent with these findings, the cytotoxicity of Aβ42 fibrils remained unchanged in the presence of P_b-derived peptides. This lack of reduction in cytotoxicity is attributed to the unaltered quantity and morphology of fibrillar aggregates (Figure S10c). Although MD simulations resulted in the length-dependent increase of the Aβ42-peptide contacts near the C-terminal region of Aβ42, this was not reflected in significant improvements for ThT or MTT assay results, regardless of the P_b-derived peptide length. These results are in line with our experimental data, which show the length dependent increase in the CID binding energy of P_b-derived peptide is not directly related to the suppression of Aβ42 fibrillar aggregation (Figures S10d, S12, and S13). Additionally, we confirmed that the N-terminal fragment (¹DAEFRHDSGYEVH¹³, P_N) of Aβ42 does not show any suppression effect on the Aβ42 fibrillar aggregation (Figure S14).

As discussed above, the fundamental difference between P_a and P_b is their modes of molecular interaction. The interaction mechanism of P_b-derived peptides is predominantly driven by hydrophobic interactions. By contrast, the contact mechanism of P_a peptide involves in the orchestration of electrostatic interactions by salt bridges and hydrophobic interaction facilitated by water release. Given that the early oligomerization stage of Aβ42 is primarily driven by hydrophobic interactions, a peptide inhibitor that relies solely on hydrophobicity for complex formation should be less competitive in disrupting aggregation. In this context, the eleven-residue peptide P_a11 emerges as the optimal candidate for suppressing Aβ42 aggregation. Its ability to shield the critical oligomerization domain of Aβ42 by specifically targeting the central region enhances its inhibitory efficacy.

Antiparallel β-sheet Motif of Dimerized Scaffold via Disulfide Bond Formation

By theoretical and experimental screening of peptide length and sequence, P_a11 was identified as the best candidate with the highest inhibitory performance among P_a, P_b, and P_N candidates. While linear extension of peptide length improves inhibitory effects, single-chain forms (AA) can still suffer from structural fluctuations and thus face higher entropic barriers, making it difficult to achieve the desired rigid structure for effective Aβ42 interaction (Figure 3a). Therefore, we sought to further enhance the inhibitory potential of P_a11 through topological modification. Modifying structural connectivity through disulfide bond formation (e.g., Terlipressin,^[50] DOTATOC,^[51] Pramlintide^[6]), or cyclization (e.g., Pasireotide^[43] and Bremelanotide^[52]) is well-known strategies to enhance molecular recognition in biologically active peptides. Covalent constraints imposed by these modifications restrict the conformational flexibility of peptides, favoring structural arrangements optimal for binding to target proteins.

Considering the predicted secondary structure populations of Aβ42 and its fibrillar form, the regions interacting with P_a11 predominantly prefers β-sheet formation (Figure 1a).^{[39, 40]} However, our MD simulations display that P_a11 facilitates transient helical conformations of residues 16–22, whereas random coil conformations remain dominant in the remaining residues (Figure S6). Due to these observations, we hypothesized that although P_a11 exhibits molecular recognition functionality, its preferential recruitment of the transient helical conformation may be suboptimal; therefore, stabilizing P_a11 in a β-sheet-enriched conformation could be more efficient to enhance its interaction with Aβ42. A disulfide bond is the common stabilizer of antiparallel β-sheet orientations by constraining bond lengths and supporting the hydrogen bond network of antiparallel β-sheet structures.^[53-55] Given the β-sheet propensity of the P_a11 sequence, we anticipated that disulfide-bond design may promote the formation of a stabilized antiparallel orientation of the P_a11 dimer that enables multivalent interaction with a single Aβ42 molecule (Figure 3a). Consequently, we implemented a disulfide bond-mediated dimerization strategy (A*-A*) to stabilize P_a11 in the orientation necessary for enhanced inhibitory potency. To examine this idea, we theoretically generated the dimeric peptide P_a*-a* by substituting the asparagine residue in the center of P_a11 to cysteine.

To evaluate the interaction between the dimeric P_a*-a* peptide and Aβ42, we conducted REMD simulations. The theoretical interaction energy of the Aβ42–P_a*-a* complex nearly doubled compared with that of the Aβ42–P_a11 complex (−613.51 and -322.25 kJ mol⁻¹ for P_a*-a* and P_a11, respectively) (Figures 1d and 3b). This substantial increase in binding energy suggests that the dimeric P_a*-a* forms a more stable complex with Aβ42 through multivalent interactions. The degree of molecular contact between P_a*-a* and Aβ42 is also significantly higher than that observed for P_a11 (Figure 3c). Contact probability analysis revealed that P_a*-a* preferentially interacts with the central region of Aβ42, further supporting its enhanced binding capability (Figure 3c). Furthermore, when interacting with Aβ42, the two units of P_a*-a* adopt an antiparallel orientation, with covalent constraints and interstrand salt bridges between N-termini and C-termini that further stabilize the dimerization. Each unit of P_a*-a* contains two positively charged residues (the N-terminus and a lysine) and three negatively charged residues (the C-terminus, a glutamic acid, and an aspartic acid), enabling simultaneous interstrand salt-bridge formation (within P_a*-a*) and additional salt-bridge formation with Aβ42. For example, of the two Antiparallel strands, the N-terminus of one strand forms a salt bridge with Glu22 or Asp23 of Aβ42, while the other unit binds to the C-terminal region (or end) of Aβ42 (Figures 3d, and S15). This relatively higher probability of salt-bridge formation, compared with P_a11, likely underlies the enhanced selectivity of the dimeric peptide for the central region of Aβ42 (Figures 3c, S5, and S15). Potential energy surface analysis of the Aβ42–P_a*-a* complex revealed two predominant binding modes. A deep free energy well on the lower right of the energy surface corresponds to simultaneous binding of P_a*-a* to both the central and C-terminal hydrophobic regions of Aβ42 (Figure 3d). A secondary, less populated binding mode on the lower left of the energy surface involves P_a*-a* binding primarily to the C-terminus with contacts to the central region of Aβ42 (Figures 3d and S16). Collectively, these molecular dynamics simulation results suggest that P_a*-a* not only enhances recognition of the central region of Aβ42 but also establishes stronger, multivalent interactions, effectively shielding the key oligomerization domain to potentially inhibit Aβ42 aggregation.^{[35, 48]}

Suppression of Aβ42 Aggregation by P_a*-a* Inhibitor

We investigated the suppression of Aβ42 aggregation by the disulfide bond-linked P_a*-a* dimer. To quantify the β-sheet contents at the endpoint (24 h) of Aβ42 aggregation, we analyzed the secondary structural features of the Aβ42 in the absence or presence of P_a*-a* using circular dichroism (CD) spectroscopy (Figure 4a). The control group, consisting of intact Aβ42, showed the formation of β-sheet rich fibrillar structure. By contrast, a progressive decrease in the β-sheet content was observed with increasing P_a*-a* concentration. Notably, at a 10-fold excess of P_a*-a*, the feature of a random coil structure was dominant in the CD spectrum, indicating significant disruption of β-sheet formation (Figure 4a). This trend was observed consistently in the ThT fluorescence assay and TEM analysis after 24 h of incubation.

The aggregation of Aβ42 was markedly suppressed in the presence of P_a*-a* (Figure 4b,c). At an equimolar concentration of P_a*-a*, the ThT fluorescence intensity decreased to 48.7% relative to the control Aβ42 group. It is noteworthy that P_a*-a* added Aβ42 at an equimolar ratio showed lower ThT fluorescence intensity compared to the 5-fold excess of P_a11 added Aβ42 (Figures 4b, and S19). This implies that the inhibitory effect of the peptide is significantly enhanced by disulfide linked dimer. Further, reductions in ThT fluorescence intensity were observed with 5- and 10-fold excesses of P_a*-a*, decreasing to 15.7% and 8.8%, respectively (Figure 4b). TEM image analysis corroborated these findings, showing a significant reduction in fibril abundance in P_a*-a*-treated Aβ42 samples, without evidence of alternative aggregate formation (Figure 4c). Importantly, the reduction of Aβ42 aggregates led to the significant alleviation of cytotoxicity. The MTT assay demonstrated that cell viability increased from 70.7% in the control group to 95.8% in the presence of a 10-fold excess of P_a*-a* (Figure 4d), and P_a*-a* did not exhibit cytotoxicity itself (Figure S22). Moreover, P_a*-a* addition not only reduced toxicity of Aβ42 aggregates obtained at the endpoint (24 h incubation) but also mitigated the cytotoxicity of various toxic oligomeric species (Figure S23). These results collectively highlight the potent inhibitory effect of the disulfide bond-linked P_a*-a* dimer on Aβ42 aggregation and its ability to mitigate aggregate-induced cytotoxicity.

We anticipated that the binding mode of P_a*-a* with Aβ42 would lead to compelling results in ThT and MTT assays. Thus, the molecular interaction of P_a*-a* with Aβ42 was investigated using ion mobility-mass spectrometry (IM-MS) (Figure 5a,b).^[56-59] The conformation of gaseous ion of the Aβ42 and P_a*-a* complex showed 1124 Å² and 1216 Å² of collision cross sections (CCSs) in the +5 and +6 charge states, respectively. Experimental CCS values of the complex were compared with theoretical values of in vacuo MD simulated conformations (see details in Methods section) (Figure 5c,d). The representative MD-simulated structures in solution from the free energy landscape were employed as a structural pool for modeling the gas-phase conformations (Figures 3d and S26). One hundred MD-simulated structures originated from the more stable, abundant population in solution MD simulations show a broad distribution of theoretical CCS values ranging from 1070 to 1351 Å² (Figures S26a and S26b). It is noteworthy that this distribution includes a subpopulation centered around 1203 Å², closely matching the experimental CCS of 1216 Å² in the +6 charge state. If conformations of Aβ42 and P_a*-a* complex are stabilized with five or six extra charge states in the gas phase, the conformations merge to the experimental CCS values observed in +5 or +6 charge states. By contrast, MD-simulated structures of the less stable, abundant conformers show a narrow distribution of theoretical CCS values (Figure S26c) ranging from 1078 to 1243 Å² that does not sample the ∼1216 Å² region, indicating these conformers contribute less to the observed gas-phase structure. Therefore, we anticipate that the minor conformers in Figure 3d may not play a significant role in the suppression of Aβ42 aggregation. Furthermore, Fourier transform infrared (FT-IR) spectroscopy analysis of Aβ42 and P_a*-a* mixture exhibited a characteristic shoulder corresponding to the conventional antiparallel β-sheet conformation (Figure S28). Consequently, the IM-MS and FT-IR results strongly support the conclusion that P_a*-a* stabilizes Aβ42 through multivalent interactions as intended in our disulfide bond design, thereby suppressing the aggregation pathway (Figures 3d and 5c).

To understand the role of P_a*-a* after the initial stage, we measured relative abundances of Aβ42 oligomers and P_a*-a* complexes with monomeric or small oligomeric Aβ42 using ESI-MS (Figure S29). We found that, as the concentrations of P_a*-a* increased, the relative abundance of Aβ42 oligomers formed within 1 h decreased in a concentration-dependent manner. Moreover, MS spectra revealed clear single or multiple binding of P_a*-a* with monomeric or small oligomeric Aβ42, indicating that P_a*-a* contributes to the sequestration of Aβ42 species in the early stage. Next, we treated P_a*-a* to preincubated Aβ42 for 1, 3, or 6 h (Figure S30). We observed a decline in the inhibitory effect as the preincubation time increased, confirming that P_a*-a* acts predominantly at the early stage of Aβ42 aggregation. Finally, we tested seed-dependent aggregation in the presence of 0.01% (w/w) preformed fibrils (Figure S31). While monomeric P_a-derived peptides were not effective, P_a*-a* suppressed Aβ42 aggregation in a concentration-dependent manner. However, its inhibitory effect disappeared at higher seed ratio (0.1% w/w).

To further characterize chemical properties of P_a*-a*, We measured its solubility in buffer and stability in human plasma to assess chemical properties of P_a*-a*. Turbidity (OD₆₀₀) assay in 20 mM Tris-HCl (pH 7.4) showed that P_a*-a* remained fully soluble up to 0.5 mM (1.29 mg mL⁻¹), whereas P_a11 began to show a noticeable increase in OD₆₀₀ (∼0.02) at 1 mM (1.30 mg mL⁻¹; Figure S32). In human plasma, P_a*-a* exhibited an apparent half-life of 7.7 min, while it was 57.9 min for P_a11 (Figure S33). MS binding studies indicate that both peptides associate with human serum albumin (Figure S34), which is likely protected from proteolytic degradation. Additionally, in a confluent calf pulmonary artery endothelial cell (CPAE) monolayer transwell assay, P_a*-a* achieved 35% translocation of the donor load at 48 h, whereas no flux was observed in a parallel artificial membrane permeability assay (PAMPA) blood-brain barrier (BBB) kit (Figure S35). These data demonstrate P_a*-a* has a potential cell-mediated transportation mechanism, rather than a simple diffusion across the lipid membrane.