In vitro reconstitution demonstrates the amyloid-beta mediated myelin membrane deformation

Amyloid-beta (Aβ) aggregation mediated neuronal membrane deformation, although poorly understood, is implicated in Alzheimer’s Disease (AD). Particularly, whether Aβ aggregation can induce neuronal demyelination remains unknown. Here we show that Aβ-40 binds and induces extensive tubulation in the myelin membrane in vitro. The binding of Aβ-40 depends predominantly on the lipid packing defect densities and electrostatic interactions and results in rigidification of the myelin membrane in the early time scales. Furthermore, elongation of Aβ-40 into higher oligomeric and fibrillar species leads to eventual fluidization of the myelin membrane followed by extensive membrane tubulation observed in the late phase. Taken together, our results capture mechanistic insights into snapshots of temporal dynamics of Aβ-40 - myelin membrane interaction and demonstrate how short timescale, local phenomena of binding, and fibril mediated load generation manifests into long timescale, global phenomena of myelin tubulation and demonstrates the ability of Aβ-40 to demyelinate.

Aggregation and accumulation of Amyloid-β (Aβ)peptides in the extracellular space of 62 neurons, forming senile plaques, is considered a hallmark of Alzheimer's Disease 63 (AD) (Selkoe & Hardy, 2016;Winblad et al, 2016). The cleavage of a transmembrane amyloid 64 precursor protein results in the generation of two pathologically important amyloid-ß 65 (Aβ)peptides -Aβ-40 and Aβ-42 (Haass & Selkoe, 1998). While the non-aggregated forms of 66 Aβ are known to be physiological constituents of cerebrospinal fluid, the neuritic plaques 67 are rich in fibrillar forms of Aβ (Selkoe, 2000;Seubert et al, 1992). The understanding of the 68 pathology of AD has undergone a paradigm shift over time. It is now believed that the 69 soluble oligomeric Aβ forms are critical factors in driving the onset and progression of 70 neuronal injury and have more potent neurotoxic effects than mature aggregates 71 (Bucciantini et al, 2002;Kirkitadze et al, 2002). While the synaptic membrane disruption has 72 attracted significant focus in AD research, however, the myelin membrane's potential role 73 remains unknown. Emerging clinical evidence strongly suggests that demyelination is an 74 essential pathological feature that may be one of the earliest characteristics during AD 75 progression (Dean et al, 2017). Despite the evidence that the soluble forms of Aβ are 76 elevated in the white matter independent of the amyloid burden in the cortical plaque, 77 whether Aβ aggregation can alter the myelin membrane remains elusive(Collins-Praino et al, 78 2014).

80
The etiology of AD remains poorly understood due to the interplay of highly complex 81 mechanisms that govern Aβ aggregation and dynamic cellular interface. The complexity of 82 the problem arises primarily due to -the existence of soluble aggregates that are highly 83 heterogeneous in size, shape, and structure (Fusco et al, 2017); aggregation induced by 84 seeding (Jucker & Walker, 2013); and the interplay between highly diverse physicochemical 85 properties of lipid membranes and Aβ aggregation kinetics(Di Paolo & Kim, 2011). The 86 fibrillogenic properties of Aβ-40 and membrane damage have been observed to be 87 significantly correlated (Yip & McLaurin, 2001). The formation of heterogeneous ion 88 channels observed before the plaque formation has been considered the early steps during 89 neuronal damage formed by soluble, small-sized aggregates of Aβ-40 (Kourie, 2001). 90 However, the inertness of Aβ-40 mature fibrils is not fully validated as both fibrils and 91 oligomeric Aβ-40 assemblies induced a decrease in mitochondrial membrane potential in 92 neurons (Eckert et al, 2008). More importantly, the same was not observed for neurons 93 incubated with Aβ-40 monomeric forms (Eckert et al., 2008). Gangliosides and cholesterol 94 have also been observed to enhance the Aβ-40 membrane interaction and influence the 95 peptide's penetration into the membrane ( Sciacca et al, 2012b). Therefore, despite the body of reported work supporting that the 97 amyloid cytotoxicity is majorly due to small oligomeric Aβ-40, toxic peptide's real nature is 98 still a matter of intense debate (Sciacca et al, 2018). Although the two-step mechanism 99 involving both the pore formation and fibril mediated membrane deformation is a relatively 100 newly proposed general mechanism behind neuronal membrane damage, however, a full 101 understanding of the interplay of the membrane parameters, early stages of membrane-102 associated aggregation, and membrane deformation remains elusive (Lauwers et al, 2016;103 Sciacca et al., 2018; Shrivastava et al, 2017). More recently, studies indicate that the 104 templated protein misfolding (seeding) could be a crucial mechanism in the initiation and 105 propagation of Aβ as the slow primary nucleation cannot by itself account for the steep 106 aggregation kinetics that follows (Jucker & Walker, 2013; Kane et al, 2000). Therefore, secondary nucleation mechanisms such as seeding result in the acceleration of amyloid fibril 108 formation by reducing or inhibiting the lag phase of fibril formation and are now considered 109 the major driving force in the progression of Aβ aggregation (Cohen et al, 2013). The current 110 understanding is obtained mostly using small liposomes or supported bilayers composed of 111 single/binary lipid membrane mixtures and at peptide concentrations towards the higher 112 side in micromolar concentrations (i.e., 10-40 M). Further, the studies reported so far did 113 not take into account the complex lipid compositions such as myelin, the existence of 114 phase-separation, and the secondary nucleation mechanism into account that are crucial for 115 the initiation and progression of neuronal membrane damage. Also, there is a lack of 116 temporal mapping of the membrane deformation induced by Aβ during its aggregation. 117 118 Here we reconstitute in vitro the binding of seeded Aβ on the myelin model membrane and 119 visualize snapshots of the changes at membrane interface over a period of early, mid and 120 late phases of A aggregation spanning 24 hours, using a combination of photonic and 121 electron microscopy, fluorescence spectroscopy, and membrane monolayer experiments. 122 Enhanced fluidization of the myelin membrane during early binding and elongation of A 123 was found to precede the extensive membrane tubulation observed in the late phase 124 involving fibril mediated weak phase separation. Dissection of the early binding of A to 125 myelin lipid components varying in their shape and charge revealed particularly high lipid 126 specificity for DOPG, PI, BSM, and PIP 2 membranes. Coarse grain MD simulations showed 127 the highest density of lipid packing defects in the myelin membrane compared to other 128 membranes containing diverse lipid shapes. Indeed, the binding and diffusion of Aβ and the 129 reduction in lipid diffusion were strongest in the case of myelin membrane during the early 130 to mid and phase, suggesting a correlation between Aβ binding and lipid packing defects at 131 the membrane interface. Furthermore, the seeded Aβ was found to significantly deform the 132 myelin monolayer membrane equilibrated at a bilayer equivalent pressure within the early 133 time regime. Our findings suggest that under the seeded environment, the early binding of 134 Aβ on the myelin membrane strongly depends on the density of the surface lipid packing 135 defects that help drive the fibril load generation and deformation of the myelin membrane. 136

137
Binding and deformation of myelin membrane by A-40 138 To examine whether the A-40 interaction could trigger the deformation of the myelin 139 membrane, we sought to use the reconstitution methodology to allow us to mimic myelin 140 membrane exposed to a bulk concentration of soluble A-40 and map any morphological 141 changes in the membrane over different phases of A-40 aggregation. This is particularly 142 important as more and more emerging clinical evidence suggests that demyelination takes 143 place before neuronal damage and the fact that elevated levels of soluble chosen composition mimics the complexity of a myelin membrane, both in terms of the 160 topological aspects of the lipids (i.e., a mixture of cylindrical, conical, inverted conical lipids 161 with varying hydrophobic volumes) and surface charge (Fig. 1, Fig. S2). A-40 showed 162 significant binding to the myelin membrane starting early phase (i.e., visualized at 1 hour) as 163 evident from the binding intensity, followed by a decrease in the binding at 4-hour time 164 point. The early phase likely involves the binding of the population of A-40 that is 165 predominantly monomeric to lower oligomeric. However, the binding of short fibrils cannot 166 be ruled out as it is technically challenging to quantify the same in the bound state given the 167 highly dynamic state of aggregation. Interestingly, at the 12-hour time point again, an 168 increase in the binding intensity was observed (Fig. 1). The variation in the binding intensity 169 of A-40 at the observed time points is likely due to the destabilization of the membrane by 170 A-40 as a result of its early phase of aggregation and subsequent opposing equilibrium 171 forces operating within the membrane. Interestingly, at 24 hours, A-40 was found to 172 induce striking membrane tubulation at the interface, as evident from the tubulation profile 173 of the large distorted membrane regions (Fig. 1) and 24 hours confirmed the fibrillar network as well as the progression of A-40 induced 182 disruption of the myelin membrane (Fig. 1). The fluorescence microscopy observation 183 suggests that while the process of destabilization of the myelin membrane is triggered by 184 the early binding of A-40 aggregation, however, significant tubular disruption takes place 185 only as the A-40 progresses through the late fibrillar phase of aggregation. Moreover, A-186 40 aggregation is known to be enhanced by curvature, and therefore, it might be reasonable 187 to believe that these are indeed membrane tubules coated by A-40 (Terakawa et al, 2018).

207
The interplay of lipid specificity and fibrillation of A-40 208 Considering that several lipid components constitute the myelin membrane, we next asked 209 whether there is specificity for diverse lipid geometry and the head group that could play an 210 important role in dictating A-40 binding and aggregation. The shape of lipids depends on 211 the aspect ratio of their headgroups and acyl chains that determine their packing, the 212 degree of local defects, and spontaneous curvature within the membrane (McMahon &  213 Boucrot, 2015). To evaluate this, we set out to investigate microscopic visualization of the 214 temporal changes in membrane upon binding by A-40. To this end, we first focused on the 215 early phase of its interaction and aggregation corresponding to a time scale of the first hour.

216
Interestingly, while no significantly visible early binding of A-40 was observed in DOPC 217 (conical/zwitterionic lipid) membrane in the first hour(Strandberg et al, 2012) (Fig. 1a), 218 homogenous binding was observed in the case of DOPG (cylindrical/negatively charged 219 lipid) and PI membrane (inverted conical/negatively charged lipids (Fig. 2a, Fig. S3). cylindrical DOPC and negatively charged cylindrical DOPG were found to have no significant 252 effect on the fibrillation in the early phase kinetics as evident from extracted growth rates 253 (Fig. 2d) Interestingly, despite the early binding of A-40 on the myelin membrane, the observed 259 growth rate was lowest amongst all, hinting at slow fibrillation (Fig. 2d). The myelin 260 membrane was also found to facilitate A-40 fibrillation most strongly and sustain it over a 261 longer duration, as evident from the ThT fluorescence intensity observed at the plateau ( Fig.  262 2e). Amongst all the membranes, only PIP2 was found to facilitate fibrillation as evident 263 from higher plateau (Fig. 2e). PIP2 membrane with a net negative charge, the apparent 264 facilitation of aggregation of A-40 which itself is a negatively charged peptide seems 265 counter intuitive. This affinity of the peptide for the negatively charged bilayer could be 266 attributed to the coulombic interaction of positive residues present in A-40 ( Myelin membrane contains higher lipid packing density defects 277 We hypothesized that the interplay of lipid packing defects and electrostatics drive the early 278 binding, fibrillation of A-40 and subsequent deformation of the membrane. To quantify the 279 lipid packing defect density, we used coarse-grained molecular dynamics simulation. The 280 following four membrane surfaces containing lipid compositions with different shapes, 281 hydrophobic volumes, and charge were adopted to mimic different degrees of lipid packing 282 defects -i) zwitterionic conical lipids (DOPC), ii) zwitterionic conical and cylindrical lipids 283 (DOPC/BSM/Chol), iii) negatively charged inverted conical, zwitterionic conical and 284 cylindrical lipids (DOPC/BSM/PIP2/Chol), iv) a more complex surface containing negatively 285 charged inverted conical, zwitterionic conical and cylindrical lipids (myelin). We used 286 PackMem (Gautier et al, 2018) to quantify the lipid packing defects that follow the Cartesian 287 grid system for mapping the membrane surface where the grid dimension is set to 1 Å x 1 Å. 288 This tool computes the defects by characterizing them into deep and shallow defects. The 289 deep defects represent the voids created due to the presence of aliphatic atoms deeper The higher the π constant, the more abundant and larger the packing defects (Gautier et al.,317 2018). Overall, of the four bilayer systems studied, the myelin membrane has more 318 numbers of packing defects, and out of deep and shallow defects, shallow defects are more 319 abundant. The scale of the observed variation in the lipid packing defect densities of chosen 320 membrane surfaces is in the range of 3-5 Å holds significant relevance in the biological 321 regime.

323
The interplay of lipid packing defects and electrostatics drives A-40 fibrillation and membrane 324 deformation 325 To validate our hypothesis experimentally, we next questioned if the observed differences 326 in the lipid packing defect densities might be the driving forces for the degree of early 327 binding and subsequent deformation by A-40. Since A-40 aggregation is a slow process, 328 we monitored the fate of A-40 binding and changes induced in membrane morphology on 329 a time scale of 24 hours and captured the snapshots at early (1-4 hour), mid (4-12 hour), 330 and late aggregation phases (12-24 hour) (i.e., at 1 hour, 4 hours, 12 hours, and 24 hours 331 respectively) (Fig. 4, Fig S3-5). We observed that in the case of DOPC, A-40 was weakly 332 membrane-bound at 4 hours with no microscopically visible deformation of the membrane, 333 as evident from the contour on the equatorial plane of the GUV (Fig. 4). Striking 334 deformation of the membrane into large tubular structures was observed at 24 hours, 335 followed by visualization of a pool of GUVs that either has intense membrane tubulation or 336 completely ruptured membrane in the late phase (around 24 hours) ( Fig. 4a-b). presence of PIP2 during early to mid-phase, even though relatively excess cholesterol was 381 present in the membrane. Likewise, strong tubulation of the membrane was observed 382 during the late phase ( Fig. 4e-f). No binding of A-40 was observed from the early to the 383 late phase when PIP2 in the DOPC/BSM/Chol/PIP2 membrane was replaced with DOPG (a 384 negatively charged cylindrical lipid) that results in a reduction in lipid packing defects (Fig.  385 4g). This has also been found in related membrane composition where DOPG likely interacts 386 with neighboring lipid molecules by hydrogen bonding. This hydrogen bonding is between 387 glycerol moiety of DOPG and the phosphate oxygen of the neighboring phospholipid which 388 leads to ordering of the membrane (Greiner et al, 2009). The early membrane binding of 389 A-40 could be predominantly driven by the lipid packing defects and the limiting bulk 390 peptide concentration; however, the polar interactions should be essential for the 391 stabilization of the interactions. This is evident from the lack of early binding of A-40 to 392 DOPC that has significant lipid packing defects but not sufficiently strong electrostatic forces 393 (Fig. 3, 4a). Further, considering the membrane deformation and comparing the membrane 394 binding induced by A-40 over time across the membrane conditions investigated, the 395 degree of binding of A-40 on membranes decreases in the following order -myelin 396 membrane (cylindrical lipids with height mismatch, inverted conical and conical lipids) > 397 DOPC/BSM/Chol/PIP2 (conical, cylindrical, inverted conical lipids) > DOPC/BSM/Chol 398 (conical and cylindrical lipids) > DOPC (conical lipids)(Box plot, Fig. S6-8). Together, the 399 above observations suggest that although the early binding and fibrillation of A-40 400 depends on the interplay of both lipid geometry (that defines local lipid packing defects) 401 and electrostatics. However, lipid packing defects could be the predominant factor amongst 402 the two that dictates both the kinetics of A-40 binding and membrane deformation.

404
Contribution of lipid shape in the absence of charge for A-40 binding 405 We then looked into the role of coupling between proportions of different lipid shapes (i.e., 406 a ratio of cones to cylinder) and cholesterol, in the absence of negative charge, that might 407 affect the binding of A-40. To test this, we modulated the ratios of DOPC (conical lipid), 408 BSM (cylindrical lipid), and cholesterol (Fig. S9). We observed significant early binding of A-409 40 on DOPC/BSM membranes in the absence of cholesterol both at a ratio 5:5 as well as 3:2 410 ( Fig. S9 and Fig. 2). Interestingly the binding efficiency of A-40 was found to vary as the 411 proportions of conical/cylindrical lipid changed, suggesting that not only the lipid shape is binding of A-40 is observed to the liquid disordered regions of the phase-separated 418 membrane (Fig. S9). This further reinforces our hypothesis that lipid geometry that dictates 419 packing defects contribute predominantly in the initial binding of A-40. 420 421 A-40 drives myelin deformation through its fluidization 422 We reasoned that the observed differences in early and late binding of A-40 should also 423 reflect in the fluidity changes in the membrane at a comparable time point. Irrespective of 424 the density of lipid packing defects, A-40 seems to bind to most membrane conditions by 425 the 4 hour time points (early to mid-phase) (Fig. 4). Thus, we next probed the fluidity   hours and therefore allow bleaching of a region of interest (ROI) on the GUV equatorial 483 plain (Fig. 6, Table S2). We observed complete recovery of fluorescent signal from A-40 for 484 the myelin membrane (containing highest lipid packing defects) at 12 hours, suggesting a 485 dynamic interaction of A-40 at the membrane interface. Photobleaching at 24hour could 486 only bleach 35-40% of the fluorescence at the ROI, likely, due to dense coating of A-40 487 fibrils on the membrane, which was recovered fully (Fig. 6 a-b, d-e). Likewise,  90% of the 488 A-40 fluorescence was found to recover upon photobleaching of the membrane containing 489 moderate defects (DOPC/BSM/Chol/PIP2) at 12 hours, which eventually decreased three-490 folds at 24 hours (Fig. 6 a-b, d-e). No significant fluorescence recovery of A-40 was 491 observed in the case of the membrane with the least defects (PC/SM/Chol) both at 12 and 492 24 hours (Fig. 6 a-b, d-e). A-40 seems to be in a highly dynamic and deformative 493 interaction with the myelin membrane resulting in significant extraction of lipid tubules. 494 This, in turn, leads to a generation of free space on the GUVs, allowing continuous binding 495 as evident from recovery of A-40 signal at 24 hours. This also hints at a likely enhancement 496 of fluidization of the myelin membrane. Indeed, monitoring the FRAP curves of the myelin 497 membrane lipid suggests a  20 % increased recovery in fluorescence at 24 hours compared 498 to that at 12 hours, after photobleaching (Fig. 6c, f). A three-fold drop in the fluorescence 499 recovery in the lipid channel at 24 hours compared to 12 hours, in the case of 500 DOPC/BSM/Chol/PIP2 membrane, suggests a restricted movement of the lipids and more 501 stable interfacial interaction (Fig. 6c, f). Finally, in the case of the DOPC/BSM/Chol 502 (membrane with least amounts of defects), no significant change in the fluorescence 503 recovery signal in the lipid channel is observed at 12 and 24 hours (Fig. 6c, f). 504 505 Aβ-40 mediated changes in phase behavior and compressibility modulus of membrane 530 monolayer at short timescale 531 Bilayer experiments allowed us to probe the long time-scale phenomena (i.e, from 1 hour to 532 24 hours), that cannot capture the molecular aspects of interaction during earliest time 533 scales. We therefore next aimed to investigate the short time-scale local phenomena 534 capturing the molecular events of the earliest binding as well as changes in the mechanical 535 properties of the membrane within 1 hour. To address this, we used two-dimensional 536 models of a biological membrane, i.e., Langmuir monolayers that are highly sensitive tools 537 to study mixing behavior, binding/insertion ( with the lipid monolayer. This is further corroborated by the difference in the collapse 558 pressure indicative of material loss during the compression. 559 560 We then looked at the mixing behavior of the DOPC/BSM/Chol (4:4:2) membrane and Aβ-561 40. A left shift quite early in the isotherm was observed seen just at the start of the L e phase 562 indicating Aβ-40 induced condensation upon compression. The collapse of the membrane 563 condition with Aβ-40 in the subphase happens to be higher (~ 50 mN/m) than the control 564 without Aβ-40. This condensing effect caused by Aβ-40 stabilized the solid phase to a 565 greater extent indicated by an increase in the surface pressure corresponding to the 566 collapse of the membrane models. The higher surface pressure collapse of these 567 membranes could also be attributed to Aβ-40 having a lesser degree of affinity to these 568 membranes, albeit having enough affinity to non-disruptively condense and stabilize the 569 membrane. A similar observation was found for a membrane containing DOPC/BSM/Chol in 570 3:3:4. 571 572 The difference in the observed initial binding and phase behaviour of the membrane should 573 manifest into changes in its mechanical properties. We therefore quantified the changes in 574 the elastic compressibility modulus (C s   induced by interacting Aβ-40 that showed the same trend as observed for the peak values 624 of C s -1 (Fig. 7c-d) Aβ-40 fibril load generation on monolayers compressed at bilayer lateral pressure 641 After establishing the mixing behavior of Aβ-40 with free-standing monolayer and its effect 642 on elasticity, we next wondered how would the Aβ-40 binding/fibrillation affect lipid 643 monolayer membranes compressed to a surface pressure of 30-35 mN/m to mimic bilayer 644 lateral pressure over the duration of fibrillation. The membrane monolayer was allowed to 645 equilibrate for 15-20 minutes, after which the Aβ-40 was injected into the monolayer sub-646 phase (Fig. 7e). It was observed that immediately after injection of Aβ-40 within 1 hour, 647

670
Together, we show that the myelin model membrane can be deformed by the Aβ-40 671 mixture mimicking a high monomeric to oligomeric ratio. The deformation is driven by the 672 interplay of the lipid packing defect densities mediated by early binding and the lipid-673 specific enhancement or retardation of the fibrillation of Aβ-40. The binding and fibrillation 674 of Aβ-40 induce rigidification on short timescales followed by fluidization of the membrane 675 over long time-scales, resulting in significant lipid extraction and tubulation. This increase in 676 lipid extraction and tubulation with the Aß-40 fibrillation might hint at the possibility of 677 biological demyelination during neurodegeneration (Fig. 8a-b). 678 679

680
Emerging clinical evidence strongly suggests that myelin alteration is an important 681 pathological feature in the pathogenesis of AD that may be one of the earliest 682 characteristics of the disease progression (Dean et al., 2017). Despite the evidence that the 683 soluble forms of Aβ are elevated in the white matter independent of the amyloid burden in 684 the cortical plaque, whether Aβ aggregation can alter the myelin membrane remains 685 elusive (Collins-Praino et al., 2014). In this study, we first reconstitute the binding of seeded 686 Aβ-40 on the myelin lipid membrane, mimic and visualize snapshots of the changes at 687 membrane interface over a period of 24 hours. The presence of a monomer/ oligomer 688 mixture of Aβ-40 is known to accelerate the fibrillation through a physiologically more 689 favored secondary nucleation mechanism, which is considered a major driving force during 690 the progression of protein aggregation (Cohen et al., 2013). We see that the binding of Aβ-691 40 to the myelin membrane, although homogenous, shows an interesting oscillating pattern 692 of increasing and decreasing binding intensity at early (1-4hour), mid (4-12 hour), and late 693 phase (12-24 hour) (Fig. 1). The differences in binding might arise due to structural changes 694 in the soluble forms of the Aβ aggregates within the heterogeneous population that exist in 695 a highly dynamic metastable state. Such an ensemble of Aβ-40 forms is not only 696 physiologically relevant but also results in different modes of interactions with the 697 membranes (De et al, 2019). Extensive tubulation of the myelin membrane is observed in 698 the late phase involving fibril-mediated phase separation within the membrane as a result 699 of lipid sequestration by the growing fibril (Fig. 1). Aβ-40, particularly at the mid-late phase 700 (4-12 hour), triggered the clustering of vesicles, likely entangled by the growing fibrils as also 701 evident from the dynamic light scattering measurements showing the increased 702 hydrodynamic radii due to the fusion/clustering of vesicles (Fig. 1, Fig. S12). Similar 703 clustering of vesicles was reported earlier, induced by β 2 M fibrils, wherein ends of the fibrils 704 were found to distort the membrane and extraction of lipids (Milanesi et al, 2012).

705
Aggregation kinetics of Aβ-40 revealed that while myelin membrane enhances sustained 706 aggregation of Aβ-40, however, except PIP2 other myelin lipid components such as DOPC, 707 DOPG, PI, DOPS decrease the overall aggregation rate (Fig. 2d). There are contrasting 708 observations previously reported on the effect of DOPC on A fibrillation. While one study 709 observed DOPC mediated retardation of A-40 and A-42 fibrillation (Hellstrand et al.,710 2010), another study reported enhancement of A-42 fibrillation (Lindberg et al, 2017). The 711 later study observed enhancement of DOPC driven A-42 fibrillation was driven by 712 augmentation of the secondary nucleation that facilitates fibril fragmentation. It was also 713 observed that A-42 monomers do not directly interact with DOPC membrane till fibrils are 714 formed, something in line with our observation of delayed binding of A-40 to DOPC 715 membranes (Fig. 4a). While we could see significant binding of A-40 to DOPG membrane 716 however, it did not seem to have much effect on the aggregation kinetics (Fig 2a, 4d-e). This 717 is in line with previous observations wherein POPC/POPG membrane had no effect on Aβ-40 718 fibrillation although DPPG induced templated crystalline ordering that triggers 719 fibrillation (Chi et al., 2008). Interestingly, although we did not observe any significant 720 binding of Aβ-40 to DOPS membrane in our study, however, Aβ-42 is shown to significantly 721 interact with DOPS membranes in the presence of Ca 2+ that help bridge Aβ-42 nucleate on 722 the membrane (Yi et al, 2015). Likewise, the aggregation kinetics of another peptide 723 amyloidogenic peptide IAPP was also found to enhance in the presence of higher mole % of

733
A key finding from our results is that the degree of binding by Aβ-40 depends on the 734 interplay of lipid packing defects and electrostatics that eventually drive membrane 735 deformation. Aβ-40 binds early when the membrane contains either a mixture of lipids with 736 shapes that differ geometrically or possess a negatively charged surface (Fig. 2). The general 737 trend for the average binding intensity of Aβ-40 during the early, mid, and late phase is 738 Myelin membrane > PC/BSM/Chol/PIP2 > PC/BSM/Chol that reflect the decreasing density 739 of the surface voids or the lipid packing defects at the membrane interface (Fig. 1, 3). In effect on lipid dynamics in mid to late phases (12h and 24h) showed that Aβ-40 is highly 754 mobile on myelin membrane reflected in fluorescence recovery upon fluorescence 755 photobleaching (Fig. 6). The mobility of Aβ-40 was also found to follow the same trend as 756 the average binding, i.e., myelin membrane > PC/BSM/Chol/PIP2 > PC/BSM/Chol, suggesting 757 that Aβ-40 has a highly dynamic interaction on membranes with higher densities of lipid 758 packing defects. However, the strongest increase in the lipid mobility was observed in the 759 case of the myelin membrane, followed by PC/BSM/Chol/PIP2 and least changed in the case 760 of PC/BSM/Chol membrane (Fig. 6). This suggests that the growth of Aβ-40 on myelin 761 membrane interface, although highly dynamic, however, the interaction with the underlying 762 lipids is strong enough to reduce the average diffusion that increases over time due to 763 membrane deformation (Fig. 6c,  The monolayer experiments helped us dissect the earliest timescales of molecular 770 interaction of Aβ-40 with the membrane (i.e., within a time scale of 30mins to 1hour). The 771 surface pressure-area isotherms of the myelin membrane show a significant increase in the 772 area per lipid molecule around 30mN/m followed by a decrease in the collapse surface 773 pressure hinting at Aβ-40 induced reduction in the compressibility (i.e., C s ) of the membrane 774 (Fig. 7). A similar observation was seen in the case of PC/SM/Chol/PIP2 membrane. On the 775 contrary, Aβ-40 was found to induce a rise in collapse pressure in the case of PC/SM/Chol 776 membranes (4:4:2 and 3:3:4 ratio), suggesting an overall increase in the compressibility (Fig.  777  7). 778 779 Furthermore, the Aβ-40 was found to significantly deform the monolayers of myelin and 780 PC/SM/Chol/PIP2 membrane equilibrated at 30mN/m, as evident from the steep drop in the 781 surface pressure within the early time regime (Fig. 7). The deformative effect seemed to be with a recent report suggesting insertion of oligomer and protofibrils on liposomes observed 791 by 3D Cryo-electron tomography (Tian et al, 2021). We think that although the monomeric 792 or low oligomeric Aβ-40 may insert into the membrane depending on the depth of the lipid 793 packing defect, however, the fibrillation takes place at the membrane interface, as evident 794 from the lack of increase in surface pressure corresponding to surface insertion (Fig. 7) 795 which is in-line with the purview that the deformation of a membrane in response to a force 796 can be described on the basis of membrane compression, area expansion, and bending 797 moduli (Tyler, 2012).

799
Taken together, the work captures both the molecular insights of both early and late events 800 of the Aβ-40 mediated myelin membrane deformation. The findings demonstrate how lipid 801 packing defects can be exploited by Aβ-40, to manipulate the compressibility modulus, 802 anisotropy and diffusion of the myelin membrane to drive the aggregation mediated 803 deformation (Fig. 8a). This may also be critical for the elusive mutual interference of the 804 early pore formation and late fibril mediated lipid extraction that contribute to the two-step Peptide reconstitution 830 The commercially available amyloid Beta 40 (A 40) was purchased from AnaSpec, inc. 831 which was stored at -20C. At the time of preparation, the stored peptide was made to 832 equilibrate at room temperature. The peptide powder was then dissolved in 40 L of 1% 833 NH 4 OH diluting it with milliQ water up to 1mL, bringing the concentration of the peptide at 834 1mg/mL. Further, 10L aliquots of this preparation were flash freezed and lyophilized, 835 which was then stored at -20C. The peptide was then dissolved in the desired buffer for 836 further experiments. The reconstituted peptide was incubated for two hours to allow 837 aggregation prior to quantifying different populations of soluble forms of A-40 by 838 fluorescence correlation spectroscopy. 839

840
Thioflavin T assay for the measurement of fibrillation Kinetics of Amyloid Beta 40 841 The freeze-dried peptide was then reconstituted in phosphate-buffered saline (PBS). The 842 Final concentration of the peptide for ThT experiments was kept at 1M. The Thioflavin T 843 concentration used for the experiments was 20M. The lipid specificity of A 40 was 844 screened by incubating the peptide with giant unilamellar vesicles (GUVs) which acted like 845 lipid templates for amyloid aggregation. The fluorescence intensity was followed against 846 time to monitor the A 40 fibrillation kinetics using BioTek Synergy H1 fluorescence plate 847 reader at an excitation wavelength of 440 nm and an emission wavelength of 490 nm. 848 Readings in triplicate were recorded every 30 min for 6 hrs. To minimize evaporation an 849 Opti-seal was applied over the microplate. The data was then normalised by the lipid 850 controls for each condition and plotted using Origin pro. The initial growth rate was 851 calculated by fitting the initial log phase of the aggregation kinetics to the equation y = A + 852 B*exp(-kx). 853

854
Preparation of fluorescently labelled large unilamellar vesicles.

855
Each LUV sample contains 120 nmol of the respective lipid which is doped with 1 mol% of 856 di-8-ANEPPS. This lipid solution was then dried under a gentle nitrogen gas stream, 857 subsequently it was vacuum dried for an hour to remove the residual solvent from the lipid 858 film. These lipid films were then rehydrated in 250 μL of PBS of pH 7.4 and then were 859 incubated for about 15 min in a water bath, making sure the temperature remained above 860 the transition temperature of the lipid. The heated samples were then vortexed for 4-5mins. 861 For the preparation of LUVs, the MLV suspension was then sonicated for 5 min at 0.9 pulse 862 rate and 100% amplitude. Hylite-488 A 40. The cover slip was wipe cleaned with 70% ethanol and was then air-dried. 879 Then, 90 μL of the GUVs from the microcentrifuge tube was added to the chamber along 880 with 10 μL of A 40 in equi-osmolar PBS buffer with the effective concentration of the 881 peptide being 2.5M. This chamber was then sealed with an opti-seal so as to minimize 882 evaporation during prolonged incubation of the sample. Imaging was performed on a Leica 883 TCS-SP8 confocal instrument using appropriate lasers for rhodamine-PE (DPSS-561) and 884 Hylite-488 A 40 (argon-488). An identical laser power and gain settings were used during 885 the course of all the experiments. The image processing was done using ImageJ. 886

Quantification of Membrane Packing Defects 887
The biological membranes are composed of lipids of different shapes such as cylindrical, 888 conical etc. and when such lipids get together to frame a bilayer structure with their polar 889 heads exposed to the aqueous environment, several voids get created which leads to the 890 packing defects in membrane. To study these defects, membranes of four different 891 compositions were set up and simulated using coarse-grained (CG) molecular dynamics. where P(A) denotes the probability of finding a defect with area A, b is the pre-exponential 925 factor and π is the packing defect constant (Gautier et al., 2018). Finally, an R script provided 926 with the package computes the mean packing defect constants. The barplot for the packing 927 defect constants for each type of packing defects is prepared using GNUPLOT version 928 5.2 (Williams et al, 1986).

930
Fluorescence recovery after photobleaching (FRAP) 931 For the FRAP measurements on the GUVs, the GUVs were doped with 1 mol% of rhodamine 932 PE and were incubated with 2.5 μM of A 40 doped with 10% Hylite-488 A 40. First, pre-933 bleach images at an attenuated laser intensity were acquired. Photobleaching was 934 performed using DPSS-561 (to photo-bleach the lipid rhodamine-PE) and argon-488 (to 935 photo-bleach bound the bound Hylite-488 labelled A 40) at 100% laser power for 30 s, 936 achieving a partial bleach through a circular region of interest (ROI) of a nominal radius r = 937 2.2-2.4 μm. The laser was then switched back to the attenuated intensity, and the recovery 938 curve along with the images was recorded for several seconds. The photobleaching was 939 executed at the equatorial plane of the GUV being visualised. The FRAP curves for each 940 condition was repeated five times and then normalized. The diffusion coefficient was 941 calculated using the Soumpasis equation for 2D-diffusion 942 = 0.224 2 1/2 (2) 943 Where, 0.244 is the numerically determined value, r (2.2 μm) stands for the radius of the 944 laser beam focused on the region of interest, and τ 1/2 is the time required for half the 945 recovery. The time for half the recovery was determined by plotting the normalized 946 recovery curve. 947 Fluorescence spectrophotometric assay 949 The Fluorescence spectroscopy experiments were performed on LUVs (preparation 950 described earlier). The LUVs prepared were then incubated with A 40 at an effective 951 concentration of 1M and making up the total volume of the LUV including the A 40 to 300 952 μL. This mix was then incubated in dark. Readings were then taken at the 4 th hour of 953 incubation of aggregation kinetics of A 40. The fluorescence anisotropy was measured for 954 24s, with an integration time of 1s. 460nm and 560nm were the excitation and emission 955 wavelength, respectively. The anisotropy (r) was automatically calculated by the instrument 956 using the equation; FCS experiments were carried out using a dual channel ISS Alba V system equipped with a 967 60X water-immersion objective (NA 1.2). Samples were excited with an argon laser at 488 968 nm. All protein data were normalized using the τ D value obtained with the free dye 969 (Alexa488) which was measured under identical conditions. For a single-component system, where, S is the structure parameter, which is the depth-to-diameter ratio.