Localized surface plasmon resonance and atomic force microscopy study of model lipid membranes and their interactions with amyloid and melatonin

Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by the accumulation of amyloid plaques in the brain. The toxicity of amyloid to neuronal cell surfaces arises from interactions between small intermediate aggregates, namely amyloid oligomers, and the cell membrane. The nature of these interactions changes with age and disease progression. In our previous work, we demonstrated that both membrane composition and nanoscale structure play crucial roles in amyloid toxicity, and that membrane models mimicking healthy neuron were less affected by amyloid than model membranes mimicking AD neuronal membranes. This understanding introduces the possibility of modifying membrane properties with membrane-active molecules, such as melatonin, to protect them from amyloid-induced damage. In this study, we employed atomic force microscopy and localized surface plasmon resonance to investigate the protective effects of melatonin. We utilized synthetic lipid membranes that mimic the neuronal cellular membrane at various stages of AD and explored their interactions with amyloid-β (1–42) in the presence of melatonin. Our findings reveal that the early diseased membrane model is particularly vulnerable to amyloid binding and subsequent damage. However, melatonin exerts its most potent protective effect on this early-stage membrane. These results suggest that melatonin could act at the membrane level to alleviate amyloid toxicity, offering the most protection during the initial stages of AD.


Introduction
Alzheimer's disease (AD) is a progressive neurodegenerative disorder that primarily affects the elderly population [1][2][3][4][5].Despite the extensive research, the molecular mechanism of AD is not well understood which delays the development of Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.cure and preventive strategies.Amyloid plaques, a hallmark of AD [3] are found on neuronal cell surfaces and are composed of misfolded and entangled amyloid-β (Aβ) fibrils [6].
Aβ  and Aβ  are most common fragments which originate from the transmembrane amyloid precursor protein (APP) cleavage [7][8][9][10].Through process of misfolding and aggregation these monomer fragments result to formation of small soluble amyloid oligomers and large fibrils, all of which interact with cellular membranes inducing cellular damage (amyloid toxicity).Previous clinical, animal and cellular studies have demonstrated that small soluble amyloid oligomers are more toxic to neurons than large fibrils [11,12].
While specific amyloid receptors are present in neuronal cells, it is known that Aβ can also interacts with cellular membranes non-specifically, inserting into the membrane and leading to membrane damage in the way of defects, pore and ion-channel formation [7,8,[13][14][15], which has been linked to the initiation of pathogenesis in the early stages of AD [16].
In spite of more than hundred years of research, there is currently no cure available for AD [17], and therefore the developments of preventive measures are in much demand and are highly researched.Multiple cellular and animal studies have demonstrated that melatonin can be protective against Aβ neurotoxicity but the molecular mechanisms of this protection are not well understood [18].Melatonin is a hormone secreted by the pineal gland in the brain and regulates circadian rhythms, seasonality, and has antioxidative and immune response effects [5,19,20].It has been shown that the levels of melatonin decrease with aging and neurodegeneration [21].Previous studies on melatonin's effect on model membranes have revealed that it is able to fluidize the membrane while enhancing the lipid dynamics and increasing the area per lipid [20,22,23].It counteracts the ordering effect of cholesterol [24,25] and acts similarly to its biosynthetic precursors tryptophan, serotonin and N-Acetylserotonin (NAS) [26].With nuclear magnetic resonance (NMR), melatonin was also shown to influence the phase separation in the DPPC/POPC/cholesterol membrane model [27].In order to understand the mechanisms of melatonin protection against amyloid toxicity, it is important to study the role of melatonin in Aβ-membrane interactions.Due to its ability to alter the surface properties of the membrane, melatonin may exhibit an effect on non-specific interactions between amyloid and cellular membranes, which may protect the membranes from amyloid toxicity.
Previous studies have demonstrated that changes in lipid composition, consequently involving different lipid charges and phases, affect membrane interactions with amyloid peptides [13,14,[28][29][30][31][32][33][34][35][36][37].As reported earlier, the changes in the lipid composition of the plasma membrane due to aging have been observed in early-state AD, and such changes are closely related to the early development of AD and affect Aβ toxicity [28,38].Due to the high complexity of cellular membranes, model membranes composed of synthetic lipids are widely used to investigate the interaction between Aβ and the membrane [39,40].A previous study has demonstrated that changes in the composition of model lipid membranes, mimicking the changes in neuronal membranes in health and in AD, cause changes in the membrane nanoscale structure (nanodomains) and dramatically alter the damaging effect of Aβ in such model membranes [28].Therefore, we hypothesized that protecting the membrane with membrane-active molecules, such as melatonin, can reduce Aβ toxicity.In this study, complex model lipid systems developed by Drolle et al [28] were employed to model neuronal cell membranes in various states of AD (table 1).The interactions between amyloid and the membranes, in the presence or absence of melatonin, were studied using localized surface plasmon resonance (LSPR).Complementary atomic force microscopy (AFM) images were also utilized to visualize the membrane surfaces and Aβ-membrane interactions.

Materials and methods
In this study, supported lipid membranes were prepared by vesicle fusion method [41] onto mica and sensor surface for AFM and LSPR, respectively.Once attached to the substrate, individual vesicles or liposomes rupture into a single membrane and then connect to form a continuous membrane that covers the entire substrate surface.Details of preparing the membranes and experimental techniques are discussed in the following sections.

Lipid membrane
1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC), 1-palmitoy l-2-oleoyl-glycero-3-phosphocholine (POPC), cholesterol, sphingomyelin (SM), ganglioside (GM1) and melatonin were purchased from Sigma-Aldrich and all chemicals were used without further purification.In this study, three lipid models that mimic neuronal cell membranes at different disease states of AD were utilized.Composition details are given in table 1.Stock solutions of DPPC, POPC and cholesterol at 10 mg ml −1 , and SM and GM1 at 1 mg ml −1 were prepared with chloroform and stored at −20 • C. The stock solutions were mixed and dried with a gentle flow of nitrogen and then resuspended into a buffer to produce a homogeneous lipid vesicle solution [42].
Melatonin is known to have high solubility in hydrophobic solvents [43], and thus readily partitions into the lipid membranes from water-based solutions, as melatonin has limited solubility in polar solvents.Melatonin stock solution was prepared with ethanol, and it was dried with nitrogen flow and then resuspended in buffer.It was dissolved in buffer at near saturation (400 μM) and the buffer was added to the supported membranes and incubated at least 30 min to allow melatonin to partition into the membrane.Freshly prepared buffer solutions, composed of 20 mM HEPES and 150 mM NaCl (pH = 7.4), were used.The supported lipid membrane (SLB) on the LSPR sensors was created within the LSPR system, with further details to be discussed in the LSPR section.After depositing the membrane on a sensor surface, the LSPR sensor was removed from the LSPR liquid cell and affixed to a microscope slide for AFM imaging.A Teflon ring was positioned around the sensor to retain water during AFM imaging.For AFM imaging of amyloid binding to the membrane, the SLB was formed on mica by vesicle fusion.Excess liquid was then carefully removed using a pipette, followed by rinsing the surface five times with 600 μl of deionized (DI) water.The liquid cell was subsequently left with approximately 800 μl of water inside for 90 min before collecting AFM images.

Aβ solution preparation
The Aβ (1-42) peptide was acquired from rPeptide and aliquoted using the hexafluoroisopropanol (HFIP) protocol [44].A concentration of 1 mg ml −1 was achieved by adding HFIP to the amyloid vial, which was then sealed and sonicated for 30 s to ensure complete dissolution.The vial was subsequently left in a fume hood for incubation for 15 min, after which 45 μl aliquots of the Aβ-HFIP solution (0.01 μmol Aβ peptides) were prepared and dried in the fume hood overnight.The aliquots were stored at −20 • C for future use.
An appropriate buffer solution, with or without melatonin, was used for all solution preparation.Aβ solution (50 μM) was prepared immediately prior to each injection by adding 200 μl of buffer solution to one amyloid aliquot and then sonicating for 10 min.This preparation of Aβ solution ensures the monomeric and oligomeric forms of the peptide when it interacts with the membrane surface.

Atomic force microscopy
AFM is a scanning probe microscopy technique involving the scanning of a sharp probe over a sample surface to measure topography with nanoscale resolution.AFM can operate in air or liquid environments [28,45,46], critical to maintaining physiological conditions for biological studies [15,42,47], and to our study of lipid membranes here.For this study, the JPK NanoWizard II AFM system was used with qp-BioAC AFM probes from Bruker to image samples in the intermittent contact mode.The CB1 tip (resonance frequency = 90 kHz) was utilized for unused LSPR sensor imaging and the CB3 tip (resonance frequency = 30 kHz) was used for all other imaging purposes.
All images were processed using Gwyddion [48] software with minimal filtering and correction steps to level the image background and eliminate noise.Surface roughness, along with difference in height of the surface (Δh), were determined by the software.and small-molecule detection [55], etc.In surface plasmon resonance, an oscillation of charge density at the interface of a metal and a dielectric, such as a thin layer of gold or silver in air or water, is excited by incident light at a certain angle, resulting in a propagating wave along the surface [56].Localized surface plasmon resonance further confines this resonance to a nanoparticle smaller than the wavelength of the incident light, resulting in the collective oscillation of free electrons of the metal nanoparticle, leading to a highly localized and enhanced plasmon wave [57], as shown in figure 1.The resonance wavelength of LSPR is highly dependent on the local environment of the metal nanoparticle, such that a change in the local refractive index will correspond to a change in the resonance wavelength.This parameter is exploited in LSPR biosensing, where the resonance wavelength is measured, and its shift can be correlated to changes in the local dielectric environment of the metal nanoparticle sensor, including binding events.Figure 2(A) illustrates the experimental set-up of the gold nanoparticle (GNP) sensor, while figure 2(B) depicts the strong absorption peak associated with the LSPR and its shift (Δλ) corresponding to a change in the local dielectric environment of the gold nanoparticle sensor.
The measured wavelength shifts Δλ can be translated into changes in the thickness, d, of the adsorbate layers building up on the LSPR sensing surface.The following expression explains the general relationship between these two parameters [58]: where m represents the sensitivity factor of the sensor, Δn denotes the change in refractive index in the vicinity of the sensing surface due to molecular binding and l d is the evanescent field decay length that defines the sensing distance of the GNP sensor.When no molecules bind to the sensor, the first layer is the bulk buffer solution that has approximately infinite thickness (d = ∞) compared to the sensing distance of the GNPs.Therefore, equation (1) can be simplified into In this case, m can be determined as the slope of a calibration curve, where various Δλ values are measured as a function of refractive indices for sequentially diluted standard solutions.Equation (1) is valid for LSPR responses to a single layer with a uniform thickness (or the effective thickness of a non-uniform layer).In our experiments, this can be applied directly to the membrane formed on the GNPs.Ultimately, a second layer of Aβ peptides will cover the pre-formed membrane.This adsorption on the sensor is illustrated in figure 2 with the membrane being the first layer and the Aβ layer being the second.The LSPR response to the formation of the Aβ layer will be distinct from the formation of the membrane layer, as the membrane layer dampens the response of the nanoparticle sensor to the second binding event.To account for this, equation (1) can be modified by multiplying an additional scaling factor, - e d l d 2 [58].Therefore, Δλ Aβ for Aβ binding can be expressed as ( ) where d Aβ is the effective thickness of the Aβ layer and d mem is the effective thickness of the membrane.As our goal is to measure the thickness of the membrane and Aβ layers, we can rearrange equations (1) and (3) as expressions of d mem and d Aβ , respectively: Δλ mem and Δλ Aβ are the measurable experiment responses; Δn mem and Δn Aβ are the differences of the refractive indices of the membrane and Aβ to the bulk buffer solution, respectively.The refractive index of the buffer solution was taken to be 1.334 [59], the refractive index of the membrane was estimated to be 1.47 [60][61][62][63][64], and the refractive index of Aβ was calculated as 1.6 [65].As discussed with equation (1), m is the sensor's sensitivity factor, which can be determined as the slope of the calibration curve.We used standard glucose solutions with a series of concentration gradients and known refractive indices [66] to generate the calibration curve and obtain m.The parameter l d , in the commercial LSPR sensor, has a defined range of 20-40 nm.Due to the >50% uncertainty of this parameter, it is kept as an unknown constant in the final calculations, expressed as d/l d , as shown in equation (4a) for each layer (membrane, d mem or amyloid, d Aβ ).Therefore, we define a new parameter T as the thickness profile to represent the ratio: Overview of the LSPR experiments.In this study, the OpenSPR and gold nanoparticle sensors from Nicoya Lifesciences were used.Figure 2(A) illustrates the experimental set-up inside the fluidic chamber for the LSPR experiment.The LSPR chip with the GNP size ≈100 nm was used and was optimized previously in our collaboration with Nicoya Lifesciences with a specific goal to provide maximum enhancement for the LSPR signal meanwhile preparing supported lipid bilayer on GNP sensors.A GNP sensor is pressed and sealed onto the fluidic chamber during the experiment.Before conducting the experiment, the fluid chamber was thoroughly cleaned with 80% isopropanol to eliminate any air bubbles in the loop, while maintaining the pump speed at 150 μl min −1 .Then at a pump speed of 20 μl min −1 , two consecutive injections of lipid solution were carried out, with a 5 min interval in between to produce a stable membrane covering the sensing surface through lipid fusion.Following this step, the pump speed was maintained at 20 μl min −1 to allow for stabilization of the membrane.The amyloid solution was prepared as detailed in section 2.2 and injected immediately afterward.After the injection of amyloid, the system is allowed to run for another 10 min to finish the experiment.A representative LSPR sensorgram is shown in figure 3.

AFM imaging
In this study, we employed AFM imaging for two main purposes: first, to verify membrane formation on the LSPR sensor, thereby confirming that the amyloid interactions in LSPR were with the membrane and not the GNPs on the sensor; second, to capture images on the lipid membrane before and after amyloid-membrane interaction for illustration.To demonstrate membrane formation on the LSPR sensor, we compared the image of a new sensor with the image of an membrane-covered sensor.We also imaged pure mica and lipid membrane supported on mica for reference.Figure 4 presents AFM images of mica surface, membrane surface, GNP sensor surface and membrane on the GNP sensor that was used in the LSPR experiments.Table 2 shows surface roughness and average height difference (Δh) between the highest and the lowest points in each image.Figure 4(A) shows a fresh-cleaved mica surface for reference to figure 4(B), which shows the AFM topography image of the HM membrane supported on the mica slide.The crosssection profile of mica surface, together with its' surface roughness (122.6 pm), indicate that mica surface is relatively smoother than the measured HM membrane surface (roughness = 225.95pm).When the HM membrane is formed on mica (figure 4(B)), it shows well-defined defect structures that have distinct height difference between domains, which originate due to phase separation of various lipids in multicomponent systems.These domains form due to the nature of lipid molecules, which have different chain lengths and head group sizes.Drolle et al reported AFM images on lipid monolayers of the lipid models used in this study and showed various domain structures [28].also drops significantly to 2.52 ± 0.28 nm.This indicates that a continuous stable sheet of lipid membrane has been formed over the rough GNP surface, demonstrating that the injection of liposome solution into the LSPR fluidic cell creates complete lipid membrane on the sensor.
To illustrate and confirm the amyloid binding to the membranes, figure 5 shows representative AFM topography images demonstrating amyloid binding to the membrane in the same concentration as in LSPR experiments.The AFM topography images of the early diseased model (EDM) membrane supported on mica were collected before and after adding amyloid to the AFM liquid cell for a bulk concentration of 50 μM, achieving the same amyloid level as in the LSPR experiments.The images with amyloid interacting with the lipid membrane were collected at 60, 90, 120, and 180 min after adding amyloid.Figure 5 displays the sequential changes caused by amyloid interactions with the EDM membrane over time.Before adding amyloid, as shown in figure 5(A), the imaged surface is relatively flat, but with small height variations displayed as brighter or darker regions in the image, indicating the presence of small domains distributed on the surface.The largest height difference of the features is within the 1-2 nanometer range.Figure 5(B) shows the image taken at the same scale after incubating the membrane with amyloid for 60 min.The image reveals amyloid-induced surface indentations with a distinct height difference (2-3 nanometers) relative to the surrounding area.
Further images focusing on a smaller area covering the apparent damage reveal evidence of amyloid accumulation over time.Figure 5(C) (60 min) shows lipid loss, which produces defects (holes) in the membrane near the bottom of the image.The darker (lower) area is not a smooth plane; it has many clusters that stand up to the same height as the surrounding taller region.These clusters start to grow larger and taller over time, suggesting amyloid-lipid complex clustering when amyloid damages the membrane.Simultaneously, amyloid aggregates attached to the edge of the induced domains continue to grow. Figure 5(D) displays an image taken at 180 min when the sporadic spots of aggregates finally link together.This set of AFM images clearly demonstrates that amyloid binds to the lipid membrane and induces visible damage to the membrane integrity at a concentration of 50 μM.

LSPR results
Calibration of the GNP sensor was accomplished with standard glucose solutions at concentrations of 5%, 10%, 15%, 20%, 30%, 40%, 50% to ensure successful recording of the SPR shift (Δλ).The measured response Δλ was plotted against the refractive index change Δn between the glucose solution and the deionized water (n = 1.33).The slope of the trend line for each plot was then extracted and averaged among the five repeats based on equation (2), giving the value Table 2. Roughness and height difference (Δh) for the AFM images shown in figure 4. Roughness was calculated by built-in statistical functions in Gwyddion [48], which is computed from the second central moment of data values.The average Δh is calculated from the difference between the minimum and maximum values of the image, also using the functions from Gwyddion [48].As shown in figure 6 and Part 1 of table 3, the thickness of the membrane varies between models, with the thinnest membrane (T mem = 0.0678 ± 0.0104) found in the healthy model (HM) membranes and the thickest in the late diseased model (LDM) membranes (T mem = 0.0881 ± 0.0075).The average membrane thickness for each membrane model is reduced when melatonin is present in the buffer, with the most significant decrease observed in the LDM (T mem = 0.0705 ± 0.0043).The reduction in membrane thickness caused by the incorporation of melatonin into the HM and the EDM is insignificant, indicating limited fluidity changes in the membranes.The substantial reduction in membrane thickness due to melatonin in the LDM suggests that fluidity enhancement is the greatest among all three lipid models.
These varying levels of thickness profile changes might be attributed to the differing effects of melatonin on spacing lipid molecules apart, depending on the lipid species.The average area for lipids with PC head groups is 64 Å 2 [67], while the average area of the GM1 head group is approximately 192 Å 2 [68], triple the size of the PC head groups.Consequently, the effect of melatonin in expanding the distance between the head groups of the PC group lipid molecules is relatively larger compared to GM1.As a result, the lipid tails of DPPC and POPC exhibit more flexibility due to melatonin incorporation, causing the entire membrane to reduce its thickness.The LDM, which contains the highest proportion of DPPC and POPC among all three lipid models, experiences the most significant reduction in membrane thickness.
Figure 7 displays the distribution of thickness profile T of the amyloid layer on the top of three lipid models, with the corresponding values provided in Part 2 of table 3.For the control groups of all membrane models, the thickest amyloid layer build-up is observed in the EDM membrane (T Aβ = 0.0137 ± 0.0013), followed by the HM membrane (T Aβ = 0.0078 ± 0.0005), while the LDM membrane exhibits the thinnest amyloid layer formation (T Aβ = 0.0040 ± 0.0005).The presence of melatonin in the buffer decreases the thickness of the amyloid layer on both the HM and EDM membranes, with the most significant effect (p = 0.003) on the EDM resulting in approximately a 40% reduction in amyloid layer thickness (T Aβ = 0.0083 ± 0.0009).For the HM membrane, a decrease of amyloid layer thickness was observed, yet not significant.In contrast, the addition of melatonin substantially increases the amyloid layer thickness on the LDM membrane (p = 0.008) resulting in approximately a 40% thickness change (T Aβ = 0.0058 ± 0.0004).These observations demonstrate that the presence of melatonin reduces the binding of amyloid to the HM and EDM membranes.However, for the LDM membrane, the opposite effect is observed, with melatonin increasing the adsorbed layer thickness.
Before delving into the details of the results, it is crucial to note that amyloid interacts with membranes in multiple ways, including attaching or inserting into the membranes and behaving like a surface surfactant that disrupts the lipid layer and uptakes some lipid molecules [8,31].The size of Aβ oligomers, ranging from 2 to 6 nm, is comparable to the lipid membrane thickness [69,70].However, because of various types of Aβ-membrane interactions, the amyloid layer on the membrane is not continuous or uniform, thus the thickness profile T that we are discussing is indeed dependent on the effective thickness of this layer.
The results indicate that amyloid interacts more strongly with the EDM membrane than with HM or LDM membranes.This suggest that neuronal membrane at early stages of AD may be more suseptible to amyloid damage.As membranes progress to the late diseased phase, the degree of amyloid insertion or membrane adsorption is reduced, resulting in the thinnest amyloid layer on the LDM membrane.From this perspective, the amyloid-membrane interaction depends on the diseased state of the cellular membrane, with the EDM being most susceptible to damage, followed by the HM and LDM.One possible explanation for such lipid-dependent behavior of amyloid is that the composition of membranes plays a crucial role in amyloid-membrane interaction.A review from Mrdenovic et al (2022) [13] research from Ewald et al (2019) [31] suggested that lipid components of the membrane model systems would cause different Aβ-membrane affinity and toxic damage, with the GM1 and cholesterol-containing membrane being the most susceptible to destructions.This could help explain that the LDM model interacts with Aβ the least, given that the amount of cholesterol and GM1 is the lowest in this model comparing to the other two models used.This finding also aligns with an AFM study previously reported by Drolle et al in 2017 [28], where the first hour of Aβ incubation on the three lipid models exhibited the most binding and accumulation on the EDM monolayer, which has the largest difference in the domain height.
Furthermore, our results indicate that melatonin protect the EDM membrane more efficiently than the HM and LDM membranes.This indicates that the presence of melatonin in the EDM membranes significantly alters the surface structure of the membrane, thereby reduces the affinity of Aβ to the surface, protecting this membrane most effectively.Melatonin potentially acts as a mediator between the lipid head groups, diminishing the roughness of the membrane.Figure 7 and table 2 Part 3 demonstrate that the presence of melatonin leads to a decrease in the amyloid layer thickness on EDM membranes, an increase in thickness on LDM membranes, and relatively constant thickness on HM membranes.These more convergent amyloid layer thicknesses suggest that melatonin tends to induce similar surface properties in neuronal membranes with different diseased states.This effect of melatonin is likely due to the interaction of melatonin with the membrane lipids and changing the membrane properties.The fluidizing properties of melatonin, which eliminate the structural differences among all models, as discussed earlier [20,22,23], could be one of the reasons.Notably, the fluidizing effect is most prominent in the EDM membrane.
Figure 8 and table 3 Part 3 display the results of the amyloid thickness divided by the membrane thickness, which is equivalent to dividing the amyloid response by the membrane response.In LSPR experiments, one GNP sensor was only used once for data collection; however, the sensitivity factor of each sensor would differ slightly.This quotient eliminates the small uncertainty of the results introduced by the varying sensor sensitivities.Figure 8 confirms that the relative distribution of amyloid layer thicknesses remains consistent across all lipid models, as depicted in figure 7.Among the three models, the EDM membrane is most likely to be damaged by amyloid, while melatonin contributes to the most significant reduction in amyloid layer build-up.Conversely, the presence of melatonin leads to increased amyloid layer formation on LDM membranes, although the amount of amyloid binding is the lowest in the absence of melatonin.In the case of HM membranes, the difference between the control group and the melatonin group is not statistically significant (p = 0.1299).Combining these results with the previously discussed findings on T mem and T Aβ , it becomes evident that the difference in amyloid layer build-up due to melatonin's effect is the most significant for both diseased Table 3.The thicknesses profiles of lipid membrane models (T mem ) and adsorbed amyloid layers (T Aβ ) with and without melatonin.This table shows the thicknesses of each layer calculated from LSPR response using equation (4a) for each membrane model.The thickness values are divided by the evanescent field decay length l d , denoted as a dimensionless parameter T as the thickness profile (see equation (5)), and the table is divided into three parts.Part 1 and Part 2 show T of membrane (T mem ) and adsorbed amyloid (T Aβ ) layer, respectively.Part 3 shows the ratio of amyloid layer thickness to the membrane thickness T Aβ /T mem .The table includes the thickness or ratio results of the control group (without melatonin) and the melatonin group (with 400 μM melatonin).models, while its effect on HM is negligible.Melatonin protects the EDM membrane from amyloid binding but promotes the amyloid accumulation on the LDM membrane.

Conclusion
We studied interaction of Aβ (1-42) with model membranes mimicking neuronal membranes in healthy (HM), early diseased (EDM) and late diseased (LDM) states and determined the effect of melatonin on these non-specific interactions utilizing AFM and LSPR.The study demonstrated that amyloid interactions with membrane models depend on the disease states of the models, with the EDM membrane being affected the most significantly.It also revealed that melatonin inhibits amyloid-membrane interactions of the three lipid models, but this protective effect was the most pronounced in the EDM membranes.Interestingly, melatonin has an opposite effect on the LDM, slightly promoting amyloid-membrane interaction in the LDM.The results suggest that melatonin may inhibit amyloid toxicity in the early stages of AD, but this effect diminishes in the late stages as the disease advances.

Figure 1 .
Figure 1.Schematic of the LSPR induced by electric field.The electric field associated with the incident light induces the electron cloud of the gold nanoparticles to oscillate collectively to generate a localized surface plasmon.

Figure 2 .
Figure 2. The LSPR experiment design and measurement principle.(A) Schematic view of an LSPR sensing surface used in the experiment.A lipid membrane assembles onto the GNP sensor.The solution containing Aβ flows over this membrane, leading to the accumulation of oligomers on the surface.Incident light travels through the sensing area, starting from the GNP side, and the transmitted light is then measured on the opposite side.Simplified model demonstrating the adsorbed layers producing LSPR signals for analysis is shown on the right.(B) The initial absorption spectrum of the transmitted light is shown as a black curve.As layers of molecules accumulate on the sensor, the absorption peak of the spectrum shifts, shown as the orange curve.This shift, represented as Δλ, is recorded for analysis.

Figure 4 (
C) displays the surface structure of a new LSPR sensor composed by a layer of GNPs.The GNPs are closely packed and entirely covering the glass chip, forming the sensing surface with a large surface roughness of 16.53 nm, as shown in table 2. The surface is a continuous sheet of GNPs with height difference Δh = 73.06± 12.23 nm.In

Figure 3 .
Figure 3. LSPR Sensorgram.LSPR sensogram shows the change in wavelength of absorbance peak versus time after each injection: first liposome solution and formation of lipid bilayer and next after amyloid injection and building of amyloid layer on the top of the membrane.Each pair of blue dashed line and circle indicates the time point of each solution injection.The two orange arrows indicate the respective sensor responses due to each layer formation.The responses are labeled as Δλ mem and Δλ Aβ , respectively.The orange dashed lines indicate where the baseline of each injection response is taken.

Figure 4 .
Figure 4. AFM topography images of [A] a fresh mica surface, [B] an LSPR sensor surface covered with GNPs, [C] a supported HM membrane formed on mica, and [D] the HM membrane supported on the GNP covered LSPR sensor.Each image size is 1 μm×1 μm.The cross-section plot below each image shows the height profile taken along the white line on the AFM image.

Figure 5 .
Figure 5. AFM topography images of supported early diseased model (EDM) MEMBRANES ON MICA WITH AND WITHOUT AMyloid.(A) untreated EDM membrane and (B) treated with 50 μM amyloid for 60 min.Images (C) to (F) are enlarged images of the black square area shown in (B) after amyloid treatment of 60 min, 90 min, 120 min and 180 min.

[
A] Mica [B] Membrane on Mica [C] LSPR Sensor [D] ± 0.12 nm 73.06 ± 12.23 nm 2.52 ± 0.28 nm of the sensitivity factor.In the end, the sensitivity factor of the GNP sensor was determined to be m = 107.99± 1.95.Details on calibration curves for the sensors are included in the supplementary material.Then m can be substituted into equation (4a) to calculate for d mem /l d and d Aβ /l d , which are represented as T mem and T Aβ according to equation((5), respectively.

Figure 6 .
Figure 6.Lipid membrane thickness profiles (T mem ) of HM, EDM, and LDM membranes from left to right (T = adsorbedlayerthickness/l d ).The vertical axis is unitless because the membrane thickness and l d are both in the nanometer units.The red bars show the control group while the blue bars show the experimental group treated with 400 μM melatonin.

1 . 2 .Figure 7 .
Figure 7. Adsorbed amyloid layer thickness profiles (T Aβ ) of HM, EDM, and LDM membranes from left to right (T = adsorbed layer thickness/l d ).The vertical axis is unitless because the membrane thickness and l d are both in the nanometer units.The red bars show the control group while the blue bars show the experimental group treated with 400 μM melatonin.

Figure 8 .
Figure 8. T mem /T Aβ of HM, EDM, and LDM membranes (from left to right).Red bars show the control group, and blue bars show the experimental group treated with 400 μM melatonin.

Table 1 .
Lipid compositions of the three model membranes mimicking different disease states of neuronal cell membranes in AD: healthy model (HM), early diseased model (EDM), and late diseased model (LDM) membranes.The ratios are shown by weight.Adapted from Drolle et al 2017 [28].