Free Gangliosides can Alter Amyloid-β Aggregation

Abstract

A recently proposed lipid-chaperone hypothesis suggests that free lipid molecules – unbound to membranes - affect the aggregation of amyloidogenic peptides such as the Amyloid-β (Aβ) peptides, whose aggregates are the hallmarks of Alzheimer’s Disease. Here, we combine experiments with all-atom molecular dynamics simulations in explicit solvent to explore the effects of the neuronal ganglioside GM1 – abundant in mammalian brains – on the aggregation of two principal isoforms of Aβ, Aβ40 and Aβ42. Our simulations show that free GM1 forms stable, highly water-soluble complexes with both isoforms, and NMR experiments support the formation of well-ordered, structurally compact GM1-Aβ complexes. By simulation, we also show that Aβ40 monomers display a preference to bind with GM1-containing hetero-oligomers over GM1-lacking homo-oligomers, while Aβ42 monomers have the opposite preference. These observations explain why GM1 dose-dependently inhibits Aβ40 aggregation but has no effect on Aβ42 aggregation, as assessed by ThT fluorescence.

Graphical Abstract

Cell membranes host key proteins that participate in the production and speciation of Aβ peptides, whose aggregation into plaques is the hallmark of Alzheimer's disease (AD)¹. They can also expedite Aβ nucleation, fibrillation and deposition through selective lipid-peptide interactions^2-5. Ganglioside clusters in particular have been implicated in potentiating Aβ deposition. A mechanism described as the G-Aβ hypothesis^4,6-10 posits that Aβ peptides bind to gangliosides in the membrane and undergo conformational changes, which make them efficient endogenous seeds for expedited aggregation.

Cell membranes also release free lipids into the extracellular region, where they exist below their critical micelle concentrations (CMC). Free lipids have been shown to influence the progression of AD via different pathways,^5,11-14 and the recently published lipid-chaperone hypothesis argues that they can modulate aggregation of amyloidogenic peptides and conduce formation of cytotoxic, low molecular-weight hetero-oligomers^15,16.

The abundance of gangliosides in the neuronal membranes¹⁷ is reason to expect that they frequently encounter Aβ monomers - also abundant in the brain. In light of the aforementioned lipid-chaperone hypothesis and a recent work that has explored the effects of free GM1 gangliosides on the aggregation and cytotoxicity of amyloidogenic Yeast Sup35 peptides¹⁸, we were motivated to examine the role of free gangliosides on the aggregation of Aβ.

Here, we present our findings on the nature and the consequences of interactions between free gangliosides and monomers of two commonly occurring Aβ isoforms – Aβ40 and Aβ42. Our use of all-atom molecular dynamics (MD) simulations in explicit solvent environment to model these interactions and complementary NMR and fluorescence experiments have led us to discover a differential effect of GM1 on the aggregation of the two isoforms.

Interaction of Aβ peptides with GM1 gangliosides (Fig. 1a) - the predominantly occurring form in humans^19,20 - was simulated in a system where two Aβ monomers (either Aβ40 or Aβ42) and two GM1 molecules were immersed in a bath of explicit solvent (Fig. S1) leading to a very dilute solution with ~0.003 pM of Aβ and GM1 (<<CMC). Two controls were also considered – (i) DMPC replacing GM1, and (ii) lipid-free. For each system, three independent repeats of 1 μs simulation, starting from different initial structures, were carried out under our simulation conditions of constant temperature (298 K) and pressure (1 atm).

Figure 1:

(a) Chemical structure of the GM1 ganglioside used in this study. (b) Polar component of the absolute solvation free energy, $\Delta {\mathit{G}}_{\mathit{solv}}^{\mathit{polar}}$, for different hetero-oligomers and homo-oligomers (no lipid). (c) Peptide-lipid and inter-peptide van der Waals interaction energy in different hetero-oligomers and homo-oligomers. (d) Representative configurations of Aβ40/42+GM1 dimeric complexes showing GM1 acyl chains as spheres, headgroups by licorice representation, and peptides as ribbons. Each molecule is colored distinctly. (e, f) ¹H NMR spectra of Aβ40 and Aβ42, with and without equimolar GM1, recorded 30 minutes after mixing.

The simulations revealed that Aβ and lipids assembled to form complexes, typically within 250-500 ns, as indicated by the increasing spatial proximity of the peptides and lipids with time (Fig. S2a). No discernable order in which molecules interacted was observed. But the fully assembled configurations – those in which all the peptide and lipids molecules were within 5Å of each other - dominated the population distribution (Fig. S2b), therefore suggesting that formation of hetero- oligomers (complexes between GM1 and oligomeric Aβ) is favored and occurs rapidly. We note, however, that the 2:2 stoichiometry used in our simulations may not reflect the actual composition of these bipartite hetero-oligomers in vivo.

Analyses of the stability of these hetero-oligomers, performed using a continuum solvent treatment with Poisson-Boltzmann formalism, revealed that Aβ+GM1 oligomers were highly water-soluble as indicated by their favorable (more negative) polar component of solvation free energies (Fig. 1b). This arose from hydration of the predominantly polar GM1 headgroup that contains an anionic charge center (the sialic acid moiety or NANA) and a water-soluble polysaccharide moiety. The GM1 headgroup formed approximately three times as many H-bonds with water than the smaller, zwitterionic headgroup of DMPC (Fig. S3b). Notably, no significant differences were found in the per-residue hydrogen bonding pattern of the two Aβ isoforms with water. Only a general trend emerged - the more hydrophilic N-termini were solvent-exposed and the primarily hydrophobic C-termini were not (Fig. S3c).

GM1 acyl chains also engaged in favorable van der Waals interactions with Aβ monomers in the complexes, which were ~20 kcal/mol stronger than the DMPC-Aβ interaction (Fig. 1c) and occured at the cost of reduced inter-peptide interactions. The stronger van der Waals interactions also overcame the destabilizing effect of repulsive Coulombic forces (Fig. S4) resulting from the close placement of negatively charged GM1 (−1e) or DMPC (neutral) and peptides (−3e) in these oligomers.

Stronger van der Waals interactions between GM1 and Aβ can be attributed to the larger number of nonpolar interactions between them, primarily because of the longer tail-group of GM1 (16 and 18 methyl groups vs 14 in DMPC). Compared to DMPC, GM1 formed 30% (with Aβ40) and 25% (with Aβ42) more “prominent” contacts, those present in >20% of the snapshots from the five largest snapshot clusters grouped by similarity of atomic contacts. Though sites in contact with GM1 were spread throughout the sequence of Aβ, they were largely concentrated in the central hydrophobic core (₁₇LVFFA₂₁) and C-terminal (₃₁IIGLMVGGV₄₀IA₄₂) regions (Figs. S5 and S6). Representative snapshots of hetero-oligomers (Fig. 1d) illustrate the close proximity of the acyl chains of GM1 and C-termini of Aβ.

Overall, nonpolar contacts in the hetero-oligomers outnumbered H-bonds and salt-bridges, but they were also short-lived and therefore nonspecific (Fig. S8). Interestingly, they also initiated hetero-oligomerization as is suggested by the rapidly dissipating hydrophobic solvent accessible surface area, in the early stages of simulations (Fig. S9) that can be attributed to the contact formation between hydrophobic sites of the lipid and peptide molecules. This also supports the hydrophobic effect as the primary driver of hetero-oligomerization.

Findings from ¹H-NMR reinforce the conclusions from our simulations. For each isoform, the ¹H NMR spectrum of Aβ+GM1 (Fig. 1e, f) shows a peak near 0 ppm with a higher intensity compared to the same peak in the spectrum of pure peptide. Peaks with this chemical shift usually are assigned to Aβ oligomeric species²¹. The heightened intensities in the mixture samples indicate that GM1 altered the oligomeric species formed by Aβ, either by direct interaction with Aβ homo-oligomers or by the additional formation of distinct Aβ+GM1 hetero-oligomers. Given the sharpness of this peak in the spectra of the Aβ+GM1 mixtures, GM1 appears to encourage formation of more compact and structurally homogeneous oligomers compared to those formed by the peptides alone. Changes in other regions of the spectra in the presence of GM1 further suggest interactions between GM1 and each Aβ isoform. Chemical shift changes were observed for the GM1 acyl chain and not for the headgroup (Fig. S12), which are consistent with hetero-oligomerization stabilized by hydrophobic interactions between GM1 and Aβ.

Using umbrella sampling simulations, we constructed the potential of mean force (PMF) between (a) an Aβ monomer and a dimeric complex and (b) two dimeric complexes as a function of the distance (ζ) between them to examine how different complexes might grow. We note that in both cases we have modeled pre-nucleation oligomerization of the two Aβ species because the final products do not exceed their approximate critical nucleus in size (≈ 6 for Aβ40; ≈ 4 of Aβ42)²². The PMF profiles for the first case suggest that an Aβ40 monomer favors binding to its hetero-oligomer with GM1 over binding to an Aβ40 homodimer while an Aβ42 monomer shows the opposite tendency (Fig. 2a). This is more clearly seen in the probability distribution functions derived from the PMF profiles as PDF(ζ) = Ae^{−β×PMF(ζ)}, Σ_ζ Ae^{−β×PMF(ζ)} = 1 (Fig. 2b). Stable binding of an Aβ40 monomer to an Aβ40-GM1 oligomer can occur when their centers are only 30 Å apart; whereas binding to an Aβ40 homodimer would require the monomer to draw closer (~20 Å). This suggests that Aβ40 monomers can more readily bind to Aβ40+GM1 hetero-oligomers than to Aβ40 homo-dimers. For Aβ42, however, stable binding of a monomer with an Aβ42+GM1 oligomer can occur at ≈20 Å separation while binding to its homodimer appears to be initiated at larger distances. Thus, Aβ42 is more likely to bind to its homo-oligomer than its GM1 hetero-oligomer – a behavior exactly opposite that of Aβ40. In addition, PMF profiles for the growth of dimeric complexes into tetramers (Fig. S10) show free energy basins at small separation distances, indicating that Aβ42 homo-/hetero-oligomers favor facile growth by coalescing with other homo-/hetero-oligomers. Conversely, their Aβ40 counterparts resist coalescence with other homo-/hetero-oligomers for growth, which is indicated by a high free energy barrier met by the approaching dimeric units (Fig. S10).

Figure 2:

(a) PMF profiles between an Aβ monomer and a dimeric complex as a function of their center of geometry separation distance (ζ). Uncertainties in PMF values (magnified by x2 for clarity) are shown using ribbons of corresponding colors. (b) The probability distribution functions (PDF) derived from PMF profiles.

The data presented in Fig. 2a and Fig. S10 also highlight the consistent tendency of Aβ42, without GM1, to oligomerize more easily than Aβ40, which can seemingly form trimers more easily (lower free energy barrier than Aβ42 and deeper basin) but experiences a much higher barrier during tetramerization. Though we have investigated a limited subset of pathways of oligomeric growth in our simulations, our findings, nevertheless, are in alignment with the steeper, uphill free energy landscape of the prefibrillar growth of Aβ40 compared to that of Aβ42²². The higher barrier of Aβ40, leading up to the formation of its critical nucleus, has been ascribed to its larger critical nucleus size and the requisite degree of structural “backtracking” - i.e., structural reconfigurations within prefibrillar oligomers (rich in random coil) involving the breakage of contacts to transform into growth-competent oligomers (rich in β-sheet) ^22,23.

The aforementioned difference in the preferential binding of Aβ homo-/hetero-dimers with monomers can be used to rationalize the time course of Aβ aggregation, assuming that (i) Aβ+GM1 oligomers are off-pathway species that coexist with on-pathway pre-nucleation homo-oligomers and (ii) formation of these oligomers occurs on similar timescales, which are significantly faster than the timescales of nucleation and growth. Using ThT fluorescence we observed that the lag phase of Aβ40 aggregation increased with increasing GM1 concentration (Fig. 3a). This may occur due to the preferential binding of Aβ40 monomers to hetero-oligomers over homo-oligomers, as evidenced by the PMF analyses. As a result, the rate of formation of higher-order homo-oligomeric species leading up to the critical nuclei is reduced and the lag phase is extended.²⁴ Increasing the GM1 concentration likely increases the population of hetero-oligomers and further prolongs the lag phase. Intriguingly, ¹H NMR spectra of Aβ40+GM1 samples, recorded over 24 hours, show a sustained presence of the oligomeric peak near 0 ppm²¹ with relatively higher intensity compared to the peak of pure Aβ40 samples, indicating long-term persistence and therefore, stability of hetero-oligomers (Fig. 3c and Fig. S13). This long-term sequestration, and hence unavailability of Aβ40 monomers, impacts stages of aggregation that occur at timescales much longer than hetero-oligomerization. This is seen in the retarded growth rate (relatively smaller slope) of Aβ40 fibers, most notably for the equimolar and super-stoichiometric mixtures. Albeit delayed, as aggregation continues, fibril growth competes with hetero-oligomerization for monomers, and in the absence of additional GM1 molecules to form hetero-oligomers, the equilibrium shifts away from hetero-oligomerization and towards aggregation. This is consistent with the observed decrease in intensity and integration of the oligomeric peak with time (Fig. S13).

Figure 3:

ThT fluorescence measured over time for (a) Aβ40 and (b) Aβ42 mixed with different concentrations of GM1(< CMC of 100 μM, Fig. S11). (c-d) ¹H NMR spectra of (c) Aβ40 and (d) Aβ42, with and without equimolar GM, recorded at regular intervals over 24 hours. The oligomeric peaks near 0 ppm are indicated by red arrows.

The time course of Aβ42 fibrillation, however, was unaffected by GM1 (Fig. 3b) despite our observation that, like Aβ40, Aβ42 formed hetero-oligomers with GM1 (Fig. 3d). But the higher affinity of monomeric Aβ42 to attach to a growing Aβ42 homo-dimer over a Aβ42+GM1 oligomer, as suggested by the PMF profiles, likely leads to a virtually unaltered rate of primary nucleation and subsequent growth, thereby rendering the presence of GM1 less effective for inhibition of aggregation. Consequently, as Aβ42 monomers self-associate, equilibrium shifts away from hetero-oligomerization towards nucleation/aggregation at the outset and results into a near-total absence of hetero-oligomeric species at later times, which is supported by the progressively decreasing intensity of the oligomeric peak of the Aβ42+GM1 and the similarity in the ¹H NMR lineshapes of Aβ42 with and without GM1 after 24 hours (Figs. 3d and S13).

In conclusion, we report that free GM1 gangliosides have different effects on the aggregation of Aβ40 and Aβ42. Both isoforms form stable, highly water-soluble complexes with GM1 via the concerted roles of its headgroup and acyl chains, but the differences in their abilities, compared to lipid-free aggregates, to attract and bind monomers result in concentration-dependent delayed aggregation of Aβ40 and unaltered aggregation of Aβ42. It is known that Aβ42 aggregates more rapidly than Aβ40, but it is intriguing to observe that encounters with free GM1 molecules can enhance this difference. Our work expands the scope of the G-Aβ hypothesis^4,6-10 by highlighting the role of free gangliosides in Aβ aggregation and possibly linking them to the progression of AD. To expand the physiological scope of this work, it might also be interesting to investigate the role of GM1 on the mixed aggregation kinetics of AB40 and AB42 in combination. However, to fully assess their pathological roles in AD, future studies should explore how free GM1 modulates the capacity of Aβ to interact with lipid membranes and cause cellular toxicity.