Computational and Structural Biotechnology Journal 17 (2019) 963–971 Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/csbj Human Albumin Impairs Amyloid β-peptide Fibrillation Through its C-terminus: From docking Modeling to Protection Against Neurotoxicity in Alzheimer's disease Pol Picón-Pagès a,1, Jaume Bonet b,1, Javier García-García b, Joan Garcia-Buendia a, Daniela Gutierrez c, Javier Valle d, Carmen E.S. Gómez-Casuso a, Valeriya Sidelkivska a, Alejandra Alvarez c, Alex Perálvarez-Marín e, Albert Suades e, Xavier Fernàndez-Busquets f,g, David Andreu d, Rubén Vicente a, Baldomero Oliva b,⁎, Francisco J. Muñoz a,⁎ a Laboratory of Molecular Physiology, Faculty of Health and Life Sciences, Universitat Pompeu Fabra, Barcelona, Spain b Laboratory of Structural Bioinformatics (GRIB), Faculty of Health and Life Sciences, Universitat Pompeu Fabra, Barcelona, Spain c Cell Signaling Laboratory, Centro UC de Envejecimiento y Regeneración (CARE), Department of Cellular and Molecular Biology, Biological Sciences Faculty, Pontificia Universidad Católica de Chile, Santiago, Chile d Laboratory of Proteomics and Protein Chemistry, Faculty of Health and Life Sciences, Universitat Pompeu Fabra, Barcelona, Spain e Unitat de Biofísica, Departament de Bioquímica i de Biologia Molecular, Facultat de Medicina, Centre d'Estudis en Biofísica, Universitat Autònoma de Barcelona, Barcelona, Spain f Nanomalaria Group, Institute for Bioengineering of Catalonia (IBEC), The Barcelona Institute of Science and Technology, Baldiri Reixac 10-12, ES-08028 Barcelona, Spain g Barcelona Institute for Global Health (ISGlobal, Hospital Clínic-Universitat de Barcelona), Rosselló 149-153, ES-08036 Barcelona, Spain a r t i c l e i n f o a b s t r a c t Article history: Alzheimer's disease (AD) is a neurodegenerative process characterized by the accumulation of extracellular de- Received 18 January 2019 posits of amyloid β-peptide (Aβ), which induces neuronal death. Monomeric Aβ is not toxic but tends to aggre- Received in revised form 3 June 2019 gate into β-sheets that are neurotoxic. Therefore to prevent or delay AD onset and progression one of the main Accepted 13 June 2019 therapeutic approaches would be to impair Aβ assembly into oligomers and fibrils and to promote disaggregation Available online 26 June 2019 of the preformed aggregate. Albumin is the most abundant protein in the cerebrospinal fluid and it was reported Keywords: to bind Aβ impeding its aggregation. In a previous work we identified a 35-residue sequence of clusterin, a well- Alzheimer's disease known protein that binds Aβ, that is highly similar to the C-terminus (CTerm) of albumin. In this work, the Amyloid docking experiments show that the average binding free energy of the CTerm-Aβ1–42 simulations was signifi- Albumin cantly lower than that of the clusterin-Aβ1–42 binding, highlighting the possibility that the CTerm retains β-Sheet albumin's binding properties. To validate this observation, we performed in vitro structural analysis of soluble Docking and aggregated 1 μM Aβ1–42 incubated with 5 μM CTerm, equimolar to the albumin concentration in the CSF. Re- versed-phase chromatography and electron microscopy analysis demonstrated a reduction of Aβ1–42 aggregates when the CTerm was present. Furthermore, we treated a human neuroblastoma cell line with soluble and aggre- gated Aβ1–42 incubated with CTerm obtaining a significant protection against Aβ-induced neurotoxicity. These in silico and in vitro data suggest that the albumin CTerm is able to impair Aβ aggregation and to promote disassem- ble of Aβ aggregates protecting neurons. © 2019 Published by Elsevier B.V. on behalf of Research Network of Computational and Structural Biotechnology. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Abbreviations: AD, Alzheimer's disease; APP, amyloid precursor protein; Aß, Amyloid-ß peptide; CD, Circular dichroism; CSF, cerebrospinal fluid; CTerm, albumin C-terminus; fAβ1–42, HiLyte Fluor488 labelled human Aβ1–42; HPLC, high performance liquid chromatography; LC-MS, Liquid chromatography-mass spectrometry; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5- diphenyltetrazolium bromide; NMR, nuclear magnetic resonance; PBS, phosphate-buffered saline; PDB, Protein Data Bank; PPI, protein-protein interactions; SDS, sodium dodecyl sulfate; TEM, transmission electron microscopy; TFA, trifluoroacetic acid; UV, ultraviolet. ⁎ Corresponding authors. E-mail addresses: pol.picon_pages@outlook.es (P. Picón-Pagès), jaume.bonet@gmail.com (J. Bonet), javigx2@gmail.com (J. García-García), joan.garcia04@estudiant.upf.edu (J. Garcia-Buendia), dngutierrez@uc.cl (D. Gutierrez), javier.valle@upf.edu (J. Valle), csgomezcasuso@gmail.com (C.E.S. Gómez-Casuso), valeriesidelkivska@gmail.com (V. Sidelkivska), aalvarez@bio.puc.cl (A. Alvarez), Alex.Peralvarez@uab.cat (A. Perálvarez-Marín), albert.suades@dbb.su.se (A. Suades), xfernandez_busquets@ub.edu (X. Fernàndez-Busquets), david.andreu@upf.edu (D. Andreu), ruben.vicente@upf.edu (R. Vicente), baldo.oliva@upf.edu (B. Oliva), paco.munoz@upf.edu (F.J. Muñoz). 1 These authors have contributed equally to the present work. https://doi.org/10.1016/j.csbj.2019.06.017 2001-0370/© 2019 Published by Elsevier B.V. on behalf of Research Network of Computational and Structural Biotechnology. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 964 P. Picón-Pagès et al. / Computational and Structural Biotechnology Journal 17 (2019) 963–971 1. Introduction interaction with Aβ1–42 peptide by in vitro experiments using different techniques and cell cultures. Alzheimer's disease (AD) histopathological hallmarks are the extra- cellular aggregation of amyloid β-peptide (Aβ) in the brain, and the 2. Methodology intraneuronal aggregation of hyperphosphorylated tau protein [1,2]. The amyloid cascade hypothesis proposes that Aβ aggregation into olig- 2.1. Modeling of the Clusterin Peptide omers and fibrils induces synaptotoxicity and ultimately neuronal death [3–5]. We found the four homologous sequences between clusterin and al- The Aβ is a ~4 kDa peptide released from the amyloid precursor pro- bumin in a previous work through sequence alignments [39]. Due to the tein (APP) [6] by the sequential action of the enzymes β- and γ- lack of available structures for clusterin or the peptide region of interest, secretases [7,8]. Aβ is produced thoughout the life of an individual, 5 decoys of the most promising peptide according to iFrag [39] were being physiologically degraded in the brain by different enzymes or built with MODELLER [43–45] following the alignment shown in Fig. cleared through the blood brain barrier into the blood [9]. During 1A. The five decoys were minimized and scored with Rosetta [46] and aging there is an increase in Aβ production and an impairment of Aβ the best-scored decoy was selected as the structural representative of clearance from brain to blood [10], favoring its aggregation into β- the clusterin-peptide (Fig. 1B). sheets. Then, Aβ forms oligomers known to be highly neurotoxic [3,4], and they progress into forming amyloid fibrils that are packed into se- 2.2. Aβ1–42 -Docking Analyses nile plaques. At present, AD prevalence is increasing as the world population With the exception of the clusterin-peptide model, structures for all ages [11] and has become a major health and social problem ex- the different elements required for the docking analyse were obtained pected to reach pandemic status by 2050 [12]. Unfortunately, de- from the Protein Data Bank (PDB) [47]. The clusterin-peptide and the spite the current knowledge on APP processing, there are no CTerm structure (crystal structure PDB ID: 5FUO [48], residues 504- treatments that effectively prevent AD development or its progres- 538) were docked with two different target binders: (1) the Aβ1–42 α sion. On the other hand, the use of antibodies against Aβ to disas- conformation (nuclear magnetic resonance (NMR) ensemble PDB ID: semble cerebral Aβ aggregates [13–15] has not yielded positive 1IYT [49]), (2) the Aβ1–42 β conformation (cryo-EM structure PDB ID: results in clinical trials. 5OQV [50]). Additionally, CTerm was also docked to the most terminal Interestingly, clusterin (also termed apolipoprotein J) and albumin region of albumin (PDB ID: 5FUO, residues 539-582). to which is nor- are cerebral proteins that have been reported to bind Aβ under physio- mally bound (CTerm-lid). This region of the wild-type albumin hides logical conditions [16–18]. In fact, clusterin impairs Aβ aggregation in the putative recognition site between CTerm and Aβ1–42 (Fig. 2A). the brain [16] and some clusterin polymorphisms are major risk factors From the only two available NMR structures in the PDB [47] present- for late onset AD [19]. Albumin has also been demonstrated to inhibit ing the full 42 residues of the Aβ1–42 peptide, we selected 1IYT over Aβ aggregation [20–23]. These findings make the study of clusterin 1Z0Q [doi:https://doi.org/10.1002/cbic.200500223] (see alignment of and albumin of high relevance to understand AD etiology and to eluci- resolved residue densities for all Aβ1–42 peptides in the PDB at https:// date mechanisms of neuroprotection in AD. github.com/structuralbioinformatics/amyloid/blob/master/albumina/ Albumin is the most abundant protein in the cerebrospinal fluid source/alpha_amyloids/alignment.highlight.aln) for two main reasons. (CSF) produced mostly by ultrafiltrate of plasma proteins [24] and min- On the one hand, most of the standard PDB metric qualities (namely imally by the microglia [25]. It is a single peptide chain of 585 amino clashscore, Ramachandran outliers and sidechain outliers) are far better acids with a secondary structure containing 67% α-helix, 23% extended in 1IYT than in 1Z0Q. On the other hand, 1Z0Q is described as an inter- chain and 10% β-sheet [26–28]. Albumin is a flexible molecule, with the mediate state between the α and β conformations of Aβ1–42 while 1IYT ability to change its structure depending on environmental conditions aims to represent the stable alpha state of the peptide. such as temperature, pH or ionic strength [29,30]. It can bind a plethora For each pair, a total of 8000 decoys were generated with Rosetta's of different molecules [30–33], and it can even act as a free radical scav- docking protocol [51] and optimized with Rosetta's FastRelax [52]. De- enger [34]. More specifically, albumin can bind Aβ [20,35,36] and is re- coys were analyzed in terms of their ddG, shape complementarity and sponsible for 95% of its transport in the blood [18,37,38], thereby structural variation of the target binder. regulating the amount of circulating Aβ. Previously, we used the Protein-Protein Interface Prediction Server 2.3. CTerm Synthesis (iFRAG) to predict the interaction for short protein fragments and found that clusterin shares a common sequence with the C-terminus The CTerm peptide, AETFTFHADICTLSEKERQIKKQTALVELVKHKPK- domain of human albumin (CTerm; Fig. S1) [39]. Here, we address the amide, containing the hydrophobic domains reported to be Aβ binding effect of its CTerm in Aβ aggregation using in silico and in vitro sites by García et al. [39] was produced in C-terminal carboxamide form techniques. by Fmoc solid phase synthesis in a peptide synthesizer Prelude (Gyros Computational methods to predict and evaluate protein-protein in- Protein Technologies). It was purified to near homogeneity (97%) by re- teractions (PPIs) represent a feasible alternative to experimental ap- versed-phase high performance liquid chromatography (HPLC). proaches. Although docking experiments are currently used to predict the conformation of PPIs [40], it has been shown that the dis- 2.4. CTerm Analysis tribution of scores of different docking populations could be used to discern between interacting and non-interacting protein pairs. This is Analytical reversed-phase HPLC was performed on C18 columns valid even when the native conformation cannot be identified (4.6 × 50 mm, 3 μm; Phenomenex) in a liquid chomatrograph (LC- among the poses [41] and the population of docking poses can be 2010A; Shimadzu). Solvent A was 0.045% trifluoroacetic acid (TFA) in used to predict the binding energy if the structure of the complex is H2O; solvent B was 0.036% TFA in acetonitrile. Elution was carried out unknown [42]. In this work, we used similar principles to assess the with linear 15–50% gradients of solvent B into A over 15 min at 1 mL/ binding potential of clusterin, albumin and the CTerm to the alpha min flow rate, with UV detection at 220 nm. LC-mass spectrometry and beta conformations of the Aβ1–42 peptide. We focused our efforts (MS) was performed in a LC-MS 2010EV instrument (Shimadzu) fitted on elucidating the physiological relevance of the CTerm and its with an XBridge column (4.6 × 150 mm, 3.5 μm; Waters), eluted with P. Picón-Pagès et al. / Computational and Structural Biotechnology Journal 17 (2019) 963–971 965 Fig. 1. In silico modeling of clusterin and docking comparison with the CTerm. A) Guiding alignment for the modeling of the clusterin peptide. Common amino acids are labelled in blue. B) Structures of the 5 clusterin decoys with the initial models (transparent) and the minimized structure (red). Rosetta scores for the minimized structure are shown under each structure. C) Comparison of scores between the docking population of clusterin (or albumin) and the Aβ1–42 peptide (in α or β conformation). Using a total of 8000 decoys on each docking population, the scores of albumin peptides systematically outperforms those of clusterin when binding the same target, with a statistical significance of p b .001. a 15–50% linear gradient of B into A (A = 0.1% formic acid in H2O; B = 2.7. Amyloid Preparation 0.08% formic acid in acetonitrile) over 15 min at a flow rate of 1 mL/min, with UV detection at 220 nm. Lyophilized Aβ1–42 (Anaspec) was solubilized as previously de- scribed [54]. Briefly, 1 mg Aβ was dissolved in 250 μL of MilliQ water and pH was adjusted to ≥10.5 using 1 M NaOH solution to avoid the iso- 2.5. CTerm Purification electric point of Aβ disassembling and aggregates that may be present. Peptides were diluted in 250 μL of 20 mM phosphate buffer (pH 7.4). Preparative HPLC runs were performed on a Luna C18 column (21.2 Furthermore, the solutions were sonicated for 1 min in a bath-type mm × 250 mm, 10 μm; Phenomenex), using linear 15–50% gradients of sonicator (Bioruptor, Diagenode). The preparations were immediately solvent B (0.1% in acetonitrile) into A (0.1% TFA in H2O), as required, used for aggregation or stored in 25 μL aliquots at −20 °C until used. with a flow rate of 25 mL/min. Fractions of high (N95%) HPLC homoge- neity were further characterized by electrospray mass spectrometry using a XBridge column C18 (Waters) and a gradient at 1 mL/min of A 2.8. Reversed-Phase Chromatography Analysis of CTerm Effect on Aβ (0.1% formic acid in H2O) into B (0.08% formic acid in CH3CN). Detection Aggregation was performed at 220 nm. Fractions of adequate homogeneity and with the expected mass (4067.77 Da) were combined, lyophilized, and used Soluble and pre-aggregated (24 h at 37 °C) 1 μM Aβ1–42 was dis- in subsequent experiments. See Figs. S2 and S3 and Table S1 for addi- solved in Ham's F12 medium and incubated in the presence/absence tional details. of 5 μM CTerm at 37 °C during 24 h for assembly and disassemble as- says. Samples were quenched in 2% trifluoroacetic acid and injected in a Waters 2690 HPLC coupled to a UV detector set to 214 nm. A linear 2.6. Circular Dichroism (CD) gradient of 25%–45% of 0.1% trifluoroacetic acid in acetonitrile was ap- plied for 90 min into a 250 × 4.6 mm (5 μm) C4 column at a flow rate CTerm was dissolved in 5 mM NaOH at a stock concentration of 1 of 0.75 mL/min. mg/mL. The peptide was further dissolved to yield a 50 mM concentra- tion in 10 mM phosphate buffer at pH 7.2 in the presence or absence of 20 mM SDS. The CD was performed as follows. The spectral region was 2.9. Reversed-Phase Chromatography Analysis of Control Peptide Effect on recorded from 190 to 250 nm, with a 0.5 nm step resolution, on a Jasco Aβ Aggregation J-715 CD spectropolarimeter (Jasco Inc.) using a quartz cell of 1.0 mm optical path length at room temperature. The scanning speed was 100 Random peptides with the same size that CTerm were synthetized: nm/min with 2 s time response, and the spectra were collected and av- peptides X, Y and Z. Soluble and pre-aggregated (24 h at 37 °C) 1 μM eraged over 10 scans. Secondary structure estimation derived from CD Aβ1–42 was dissolved in Ham's F12 medium and incubated in the pres- data was assessed using K2D3 [53]. ence/absence of 5 μM peptides at 37 °C during 24 h for assembly and 966 P. Picón-Pagès et al. / Computational and Structural Biotechnology Journal 17 (2019) 963–971 Fig. 2. Structural in silico analysis of Aβ1–42 interaction with the CTerm. A) Structure of human albumin. The sequence of the CTerm is labelled in red. The continuing structural segment (CTerm-lid) is shown in pink (top). The α and β configurations of Aβ1–42 are shown at the bottom. B) Scores distribution of the docking analysis of CTerm against α-Aβ1–42, β-Aβ1–42, and CTerm-lid. C) Individual contribution of the residues of the CTerm to the binding according to the top 200 decoys (2.5% of the total). D) Representation of the first five top-scored decoys of the docking between the amyloid peptide (blue) and the CTerm (red). ddG score and shape complementarity are provided under each structure. disassemble assays. Analytical reversed-phase HPLC was performed on obtained in a JEM-1010 (JEOL) transmission electron microscope oper- C18 columns (4.6 × 50 mm, 3 μm, Phenomenex) in a model LC-2010A ated at 80 kV accelerating voltage and equipped with an Orius CCD system (Shimadzu). Solvent A was 0.045% TFA in H2O; solvent B was camera. 0.036% TFA in acetonitrile. Elution was carried out with linear 5–95% gradients of solvent B into A over 15 min at 1 mL/min flow rate, with 2.11. Cell Viability Assay by MTT Reduction UV detection at 220 nm. A human neuroblastoma cell line (SH-SY5Y cells) was seeded in poly-L-lysine coated 96-well plates at a concentration of 2 × 104 cells/ 2.10. Transmission Electron Microscopy (TEM) well. Soluble and pre-aggregated (24 h at 37 °C) 1 μM Aβ1–42 was dis- solved in Ham's F12 medium and incubated in the presence/absence Soluble and pre-aggregated (24 h at 37 °C) 1 μM Aβ1–42 was dis- of 5 μM CTerm at 37 °C during 24 h for assembly and disassemble as- solved in Ham's F12 medium and incubated in the presence/absence says. After they were ready the samples were added. Cells were treated of 5 μM CTerm at 37 °C during 24 h for assembly and disassemble as- with the different samples during 24 h at 37 °C. Cell viability was tested says. Samples of aggregated peptides obtained as described above by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide were placed onto carbon-coated 200 mesh copper grids and incubated (MTT) reduction. 10% v/v of MTT stock solution (5 mg/mL) was added. for 5 min. The grids were washed with distilled water and negatively After 2 h the media was replaced with 100 μL of dimethylsulfoxide. stained with 2% (w/v) uranyl acetate for 2 min. Micrographs were MTT absorbance was determined in an Infinite 200 multiplate reader P. Picón-Pagès et al. / Computational and Structural Biotechnology Journal 17 (2019) 963–971 967 at A540 nm and corrected by A650 nm. Data are shown compared to un- the sequences with CLUSTAL [56]. The alignment showed 40% similarity treated controls (100%). and N25% identity, which increased to N70% on a short stretch of 7 amino-acids [39]. The analysis showed that both proteins share a com- 2.12. Aβ1–42 Binding to Neuronal Processes mon sequence located at the albumin's CTerm [39]. The CTerm domain has 35 amino acids (AETFTFHADICTLSEKERQIKKQTALVELVKHKPK) and Primary cultures were carried out as previously reported [4]. The shows high sequence similarity to clusterin (QNAVNGVKQIKTLIEKTN procedure was previously approved by the Ethics Committee of the EERKTLLSNLEEAKKKK) [39]. Institut Municipal d'Investigacions Mèdiques-Universitat Pompeu Here, the binding affinity of the CTerm peptide was computationally Fabra (EC-IMIM-UPF). Cortical neurons were isolated from 18-day-old evaluated against its regular partner (CTerm-lid) and Aβ1–42 in its α and OF1 mouse embryos. The brain was removed and the cortex aseptically β conformations (Fig. 2A). The score distributions obtained from the dissected on ice-cold Hanks' balanced salt solution supplemented with generated decoys (Fig. 2B) suggest a slightly higher affinity for the α 4.5 g/L glucose and trypsinized for 17 min at 37 °C. In order to eliminate conformation of Aβ1–42 than for the β, albeit far from the affinity to its rests of the trypsinization medium and to disaggregate the cells, the cell normal molecular partner. Interestingly, the CTerm peptide seems to solution was washed thrice in Hanks' balanced salt solution+glucose trigger strong conformational changes when binding to the Aβ1–42 pep- and mechanically dissociated. Then, cells were seeded on poly-D-lysine tide. While binding to the peptide in the α conformation translates into coated coverslips in 24 well-plates with DMEM medium plus 10% horse changes between the relative position of the two α-helices, binding to serum at 105 cells/well. Once neurons were attached to the the peptide in the β conformation seems to result in a break of the β polylysinated wells (2 h), the seeding medium was removed and configuration that might affect the pairing with other Aβ1–42 peptides. Neurobasal medium was added containing 2% B27 supplement, 1% The evaluation of the individual contribution of each residue of the GlutaMAX and 1% penicillin/streptomycin. On day 3 of in vitro culture, CTerm peptide to the binding with Aβ1–42 (Fig. 2C) shows the involve- cells were treated with 2 μM 1-β-D-arabinofuranosylcytosine for 24 h ment of residues located throughout the entire peptide, suggesting the to eliminate glia. Primary cortical neurons were used at day 10 [15]. need for the full peptide for the recognition of its binding partners. Cells were treated for 10 min at room temperature in mild agitation When we analyzed the top 5 best scored decoys (Fig. 2D), we observed with 10 μM HiLyte Fluor488 labelled human Aβ1–42 (fAβ1–42) a very similar binding configuration. These configurations compete di- preformed in the presence/absence of 20 μM CTerm. After three washes rectly with the natural helix-loop-helix segment (CTerm-lid) to which with phosphate-buffered saline (PBS) cells were incubated on ice with the CTerm binds. All the data concerning the docking analysis can be 75 μg/mL red concanavaline in neurobasal medium for 20 min. Then found at https://github.com/structuralbioinformatics/amyloid/tree/ cells were washed again in cold PBS and fixed in 4% paraformaldehyde master/albumina. A video showing the docking of the CTerm with in PBS before mounting. Images were obtained in a SP5 Leica confocal Aβ1–42 is available at https://github.com/structuralbioinformatics/ microscope using 40× objective. Image fluorescence analysis was per- amyloid/raw/master/albumina/images/albumina_top5_overlap.mp4. formed using Image J software. To further confirm the different behaviour of the α and β conforma- tions, we docked CTerm and clusterin to the described α-β transition 2.13. Statistical Analysis state structure (PDB ID: 1Z0Q) and compare it with the results obtained for the α (PDB ID: 1IYT) and the β (PDB ID: 5OQV) conformation (Fig. Data are expressed as the mean ± SEM of the values from the num- S4). This comparison also allows to compensate for the difference in ber of experiments as indicated in the corresponding figures. Student's crystallization solvents used between 1IYT and 1Z0Q. As expected t-test or one-way analysis of variance (ANOVA) with Bonferroni multi- from its described condition as transition state, 1Z0Q mostly behaves ple comparisons test were used for statistical analyses. as an intermediate between the two conformations, the only exception being the RosettaScore evaluation. This fact is not that surprising con- 3. Results sidering that some of the same metrics that define entry quality in the PDB do affect the scoring provide by Rosetta. The differences in ddG be- 3.1. Albumin Binds Aß With Higher Affinity Than Clusterin tween the binding to albumin or clusterin in 1Z0Q are lower than those we obtained for 1IYT or the β conformation. However, it is still within Clusterin is known to bind and to inhibit Aβ aggregation [16], in fact the range to support that albumin can bind to the Aβ peptide with at some polymorphisms that cause the dysfunction of this protein are a least the same or better affinity than clusterin. major risk factor for late onset AD [19]. We used Aβ1–42 in in silico and in vitro experiments because it is the most aggregation-prone species 3.3. In vitro Analysis of Aβ1–42-CTerm Interaction of Aβ peptide and, therefore the most neurotoxic in humans [55]. The comparison of the binding affinities of the clusterin peptides and the We synthesized the CTerm as explained in the Methodology Section. CTerm to both the alpha and beta conformations of the Aβ1–42 (Fig. The CD mean residue molar ellipticity of the CTerm is shown in Fig. S5. 1C) shows that the four combinations have relatively acceptable ΔG In solution, the peptide shows α-helical distinct features (minima at values. Still, albumin's CTerm peptide systematically obtained signifi- 204 and 223 nm), with α-helix secondary structure content of 56%. cantly better scores (p b .001) for the global distribution of all poses. SDS micelles (20 mM detergent) used as helical inducer, increases the This observation suggests that the CTerm peptide can bind to the Aβ1– propensity of the CTerm secondary structure to 90%, with minima at 42 peptide with an affinity equal or possibly higher than clusterin. 206 and 223 nm and a maximum at 192 nm. Next, we showed the anti-aggregant properties of the CTerm on 3.2. In silico Analysis of Aβ1–42-CTerm Interaction Aβ1–42. CTerm was used at 5 μM, which is the physiological concentra- tion found in human CSF. The hydrophobicity analysis of the species A peptide region of human albumin that had been previously identi- present in several mixtures by reversed-phase chromatography (Fig. fied with iFRAG [39], was predicted to bind to Aβ1–42. The iFRAG method 3A) reveals several species. At early elution times (left of the chromato- predicts putative fragments that participate in the interface of an inter- graph), the specific peaks due to the presence of Aβ1–42 oligomers are action through the search for similar sequences of known interactions. absent when the CTerm is present. In addition, we observed that the We run iFRAG with the albumin and clusterin sequences interacting CTerm shifts the elution profile to the left, towards more hydrophilic with the sequence of the Aβ1–42 peptide. We obtained several fragments species in the presence of Aβ1–42 oligomers when compared with the of clusterin and human albumin that would putatively interact with elution profile of Aβ1–42 alone (Fig. 3A,a). This behavior indicates that Aβ1–42. We selected the high score fragments from iFRAG and aligned the CTerm prevents the assembly of oligomeric species. Moreover, the 968 P. Picón-Pagès et al. / Computational and Structural Biotechnology Journal 17 (2019) 963–971 Fig. 3. Effect of the CTerm on Aβ1–42 assembly and disassemble in vitro. Soluble or aggregated 1 μM Aβ1–42 and 1 μM Aβ1–42 + 5 μM CTerm were incubated for 24 h to allow soluble Aβ assembly or fibrillar Aβ disassemble. A) Fractions eluted by reverse-phase chromatography. In the central part of the graph are the peaks of the high molecular weight fibrils and on the left the peaks that correspond to low molecular weight oligomers. Representative peaks in terms of hydrophobicity: 1, peak corresponding to Aβ1–42 oligomers; 2, 3, and 4, peaks that shift their hydrophobicity in the presence/absence of Aβ1−42 fibrils. A,a) Aggregation assay performed with soluble Aβ1–42. A,b) Disassemble assay performed with aggregated Aβ1–42. Data are from a representative experiment. B) TEM images of aggregation assays. B,a) 1 μM Aβ1–42 aggregated for 24 h. Insets correspond to oligomers and protofibrils. B,b) 1 μM Aβ1–42 + 5 μM CTerm aggregated for 24 h. B,c) Aggregated 1 μM Aβ1–42 + 5 μM CTerm incubated for 24 h. B,d) CTerm alone incubated for 24 h. Representative images are shown. CTerm was also able to disassemble preformed Aβ1–42 aggregates of Aβ1–42 was prepared as in Fig. 3B, and was incubated for 24 h with the high (fibrils) and low (oligomers) molecular weights (Fig. 3A,b). CTerm, small fibrils and amorphous aggregates were observed (Fig. Three peptides of the same size as the CTerm (termed X, Y, Z) were 3B,c). The interference of CTerm in the analysis was discarded because randomly generated to use as controls. We found that the three pep- no aggregates were observed when the CTerm was incubated alone tides favour Aβ aggregation and stabilize the pre-existing oligomers (Fig. 3B,d). These results are in agreement with those obtained in the and fibrils. The results obtained with Peptide Y, as a representative ex- experiment presented in Fig. 3A and Fig. S7. periment, are shown in Fig. S6. This pattern of interaction between Aβ and different peptides is expected since it has been reported that the 3.4. CTerm Protects Neurons against Amyloid Neurotoxicity binding of Aβ to proteins can act as seeds for amyloid aggregation and stabilization of fibrils [57–60]. We also studied the effect of the CTerm on Aβ-induced neurotoxicity Furthermore, the anti-aggregant properties of the CTerm were stud- on a human neuroblastoma cell line. These studies have been performed ied by TEM (Fig. 3B). When Aβ1–42 was incubated alone for 24 h, we ob- with soluble Aβ1–42 incubated in the presence/absence of the CTerm for served the presence of the typical amyloid unbranched fibrils (Fig. 3B,a), 24 h in vitro. Later, the oligomers were added to cells for 24 h (Fig. 4A). A oligomers and protofibrils (inset in Fig. 3B,a). Upon co-incubation of significant protection was observed when Aβ1–42 aggregates were Aβ1–42 with the CTerm, amorphous aggregates were the only structures preformed in the presence of the CTerm (p b .05). Since Aβ aggregates, present (Fig. 3B,b). To study the disassemble ability of the CTerm, when particularly oligomers, bind to the neuronal membrane impairing its P. Picón-Pagès et al. / Computational and Structural Biotechnology Journal 17 (2019) 963–971 969 Fig. 4. Effect of the CTerm on Aβ-induced neurotoxicity in vitro. A) Human neuroblastoma cells were treated with 1 μM Aβ1–42 or 1 μM Aβ1–42 + 5 μM CTerm, which were previously incubated for 24 h to allow the assembly of Aβ. Cell viability was assayed by MTT reduction after 24 h. Data are the mean ± SEM from 5 independent experiments performed by duplicate. * p b .05 and ** p b .01. B) Mouse cortical neurons were incubated with aggregated 10 μM fAβ1–42 and 10 μM fAβ1–42 + 20 μM CTerm for 10 min. Binding of Aβ aggregates to the membrane was evaluated by colocalization with membrane markers (concanavalin A staining). Representative images of fAβ1–42 (green) and concanavalin A (red). The quantification analysis of fluorescence intensity of fAβ1–42 was normalized by concanavalin A (bottom, left). Concanavalin A fluorescence was also independently analyzed (bottom, right). Data are the mean ± SEM of 8 replicates. *** p b .001. C) Human neuroblastoma cells were treated with 1 μM Aβ1–42 and 1 μM Aβ1–42 + 5 μM CTerm, aggregates that were previously incubated for 24 h to allow Aβ disassemble by the CTerm. Cell viability was assayed by MTT reduction after 24 h. Data are the mean ± SEM from 8 independent experiments performed by duplicate. * p b .05 and ** p b .01. normal function, we studied the binding of Aβ1–42 aggregates formed in the oligomers. Later, human neuroblastoma cells were challenged with the absence/presence of the CTerm to primary cultures of cortical neu- these samples for 24 h (Fig. 4C). Aβ1–42 aggregate-induced neurotoxic- rons. We used these cell cultures to study the binding of the Aβ1–42 ag- ity was significantly impaired when the Aβ1–42 aggregates were pre-in- gregates to the surface of neurites mirroring what happens in vivo. cubated with the CTerm (p b .01; Fig. 4C). The CTerm alone had no effect Experiments were run with concanavalin A, a membrane marker, and on cell viability (Fig. 4D). a fluorescent fAβ1–42 peptide to study their colocalization (Fig. 4B). Cor- tical neurons treated with fAβ1–42 aggregates (at 10 μM to allow its visu- 4. Discussion alization) showed higher binding to the membrane throughout the neurite network than oligomers formed in the presence of the CTerm. In order to elucidate the albumin domain where the anti aggregant Fluorescence was quantified and the aggregates of fAβ1–42 formed in Aβ property is located, we determined the regions of human albumin the presence of the CTerm showed a 25% reduction in binding to neuro- that could interact with Aβ using the iFRAG procedure [39]. In our pre- nal membranes (p b .001; Fig. 4B) while the staining for concanavalin vious work we had compared the sequence of clusterin, a well-known was not modified. Aβ binding protein [16] whose polymorphisms are risk factors for late Finally, we prepared Aβ1–42 aggregates for 24 h in vitro. Then, they onset AD [19] with that of albumin. We found that the albumin CTerm were incubated with the CTerm in vitro for 24 h in order to disassemble has N40% similarity (70% identity on a short 7-residue stretch) with 970 P. Picón-Pagès et al. / Computational and Structural Biotechnology Journal 17 (2019) 963–971 clusterin, and therefore we focused our study on the properties of this [2] Flament S, Delacourte A, Mann DM. Phosphorylation of Tau proteins: a major event during the process of neurofibrillary degeneration. A comparative study between fragment. In the present work we show in silico evidence that the Alzheimer's disease and Down's syndrome. Brain Res 1990;516:15–9. CTerm peptide produces conformational changes on Aβ1–42 upon bind- [3] Klyubin I, Betts V, Welzel AT, Blennow K, Zetterberg H, Wallin A, et al. Amyloid pro- ing. This interaction might impair Aβ1–42 conformation into β sheets tein dimer-containing human CSF disrupts synaptic plasticity: prevention by sys- temic passive immunization. J Neurosci 2008;28:4231–7. https://doi.org/10.1523/ preventing oligomer and fibril formation. These predictions were con- JNEUROSCI.5161-07.2008. firmed in vitro by reversed-phase chromatography and electron micros- [4] Guivernau B, Bonet J, Valls-Comamala V, Bosch-Morató M, Godoy JA, Inestrosa NC, copy. Moreover, the in silico results that suggest a change of the β-sheet et al. Amyloid-β peptide nitrotyrosination stabilizes oligomers and enhances NMDAR-mediated toxicity. J Neurosci 2016;36:11693–703. https://doi.org/10. Aβ1–42 structure fit with the ability of the CTerm to induce disassemble 1523/JNEUROSCI.1081-16.2016. of aggregated Aβ1–42, data that we obtained by reverse phase chroma- [5] Hardy J, Allsop D. Amyloid deposition as the central event in the aetiology of tography and electron microscopy. Alzheimer's disease. Trends Pharmacol Sci 1991;12:383–8. [6] Toro D, Coma M, Uribesalgo I, Guix F, Munoz F. The amyloid β-protein precursor and As expected, the CTerm at physiological CSF concentration was able Alzheimers disease. Therapeutic approaches. Curr Med Chem Nerv Syst Agents to protect neurons against Aβ1–42 neurotoxicity. These results are in 2005;5:271–83. https://doi.org/10.2174/156801505774913053. agreement with our findings of decreased binding of Aβ aggregates to [7] De Strooper B, Saftig P, Craessaerts K, Vanderstichele H, Guhde G, Annaert W, et al. neuronal membranes when the CTerm is co-incubated with Aβ. Since Deficiency of presenilin-1 inhibits the normal cleavage of amyloid precursor protein. Nature 1998;391:387–90. https://doi.org/10.1038/34910. AD is proposed to begin with the insidious interference of Aβ oligomers [8] Vassar R, Bennett BD, Babu-Khan S, Kahn S, Mendiaz EA, Denis P, et al. Beta-secretase on synaptic activity [3,4], we consider that this significant reduction in cleavage of Alzheimer's amyloid precursor protein by the transmembrane aspartic neurotoxicity and binding of Aβ aggregates to neuronal membranes is protease BACE. Science 1999;286:735–41. [9] Qiu Z, Strickland DK, Hyman BT, Rebeck GW. α2-macroglobulin enhances the clear- reinforcing the protective role of albumin CTerm in AD. ance of endogenous soluble β-amyloid peptide via low-density lipoprotein receptor- The experimental results in this work highlight the ability of the pro- related protein in cortical neuron. J Neurochem 2002;73:1393–8. https://doi.org/10. gram iFrag to predict new binding sites for medical relevant targets, for 1046/j.1471-4159.1999.0731393.x. [10] Deane R, Wu Z, Zlokovic BV. RAGE (Yin) versus LRP (Yang) balance regulates example providing the new binder CTerm for Aβ peptides. Despite pre- Alzheimer amyloid beta-peptide clearance through transport across the blood- viously described interactions between albumin and Aβ peptides, brain barrier. Stroke 2004;35:2628–31. https://doi.org/10.1161/01.STR. CTerm is a non-exposed region of the albumin and, as such, this site 0000143452.85382.d1. [11] Prince M, Bryce R, Albanese E, Wimo A, Ribeiro W, Ferri CP. The global prevalence of had been dismissed on the prediction of the interaction by other pro- dementia: a systematic review and metaanalysis. Alzheimers Dement 2013;9: grams. This work confirms that iFrag had successfully identified a fully 63–75. https://doi.org/10.1016/j.jalz.2012.11.007. new binder for Aβ peptides capable of avoiding its aggregation. [12] Holtzman DM, Morris JC, Goate AM. Alzheimer's disease: the challenge of the second century. Sci Transl Med 2011;3:77sr1. https://doi.org/10.1126/scitranslmed. Finally, the interest in clusterin and proteins that share common se- 3002369. quences and Aβ binding properties is based on the effect of subtle [13] Solomon B, Koppel R, Hanan E, Katzav T. Monoclonal antibodies inhibit in vitro fibril- changes in their sequence, as it has been reported for clusterin [19], or lar aggregation of the Alzheimer beta-amyloid peptide. Proc Natl Acad Sci U S A their concentrations, as it has been reported for albumin [61], which 1996;93:452–5. [14] Ivanoiu A, Pariente J, Booth K, Lobello K, Luscan G, Hua L, et al. Long-term safety and along the life of an individual can produce a lack of protection yielding tolerability of bapineuzumab in patients with Alzheimer's disease in two phase 3 ex- to AD onset after 65 years old. tension studies. Alzheimers Res Ther 2016;8:24. https://doi.org/10.1186/s13195- 016-0193-y. [15] Valls-Comamala V, Guivernau B, Bonet J, Puig M, Perálvarez-Marín A, Palomer E, 5. Conclusions et al. The antigen-binding fragment of human gamma immunoglobulin prevents amyloid β-peptide folding into β-sheet to form oligomers. Oncotarget 2017. https://doi.org/10.18632/oncotarget.17074. We propose that the CTerm is a key region of albumin that partici- [16] Matsubara E, Soto C, Governale S, Frangione B, Ghiso J. Apolipoprotein J and pates in the inhibition of Aβ assembly also favour disassemble of already Alzheimer's amyloid beta solubility. Biochem J 1996:671–9. aggregated Aβ, due to its specific Aβ binding capacity. These findings [17] Bohrmann B, Tjernberg L, Kuner P, Poli S, Levet-Trafit B, Näslund J, et al. Endogenous proteins controlling amyloid beta-peptide polymerization. Possible implications for could set the basis for the design of inhibitory peptides as therapeutic beta-amyloid formation in the central nervous system and in peripheral tissues. J tools for the treatment of AD. Biol Chem 1999;274:15990–5. Supplementary data to this article can be found online at https://doi. [18] Biere AL, Ostaszewski B, Stimson ER, Hyman BT, Maggio JE, Selkoe DJ. Amyloid beta- peptide is transported on lipoproteins and albumin in human plasma. J Biol Chem org/10.1016/j.csbj.2019.06.017. 1996;271:32916–22. https://doi.org/10.1074/jbc.271.51.32916. [19] Duguid JR, Bohmont CW, Liu NG, Tourtellotte WW. Changes in brain gene expres- Acknowledgments sion shared by scrapie and Alzheimer disease. Proc Natl Acad Sci U S A 1989;86: 7260–4. [20] Ramos-Fernández E, Tajes M, Palomer E, Ill-Raga G, Bosch-Morató M, Guivernau B, Thanks to Jordi Pujols Pujol for his skillful technical assistance with et al. Posttranslational nitro-glycative modifications of albumin in Alzheimer's dis- the RP-HPLC experiments. This work was supported by the Spanish ease: implications in cytotoxicity and amyloid-β peptide aggregation. J Alzheimers Dis 2014;40:643–57. https://doi.org/10.3233/JAD-130914. Ministry of Economy and Business through grant Plan Estatal [21] Milojevic J, Melacini G. Stoichiometry and affinity of the human serum albumin- SAF2017-83372-R & SAF2014-52228-R (FEDER funds/UE) to FJM and Alzheimer's Aβ peptide interactions. Biophys J 2011;100:183–92. https://doi.org/ RV; BIO2017-85329-R to BO; AGL2014-52395-C2 and AGL2017- 10.1016/j.bpj.2010.11.037. [22] Milojevic J, Esposito V, Das R, Melacini G. Understanding the molecular basis 84097-C2-2-R to D.A.; MDM-2014-0370 through the “María de Maeztu” for the inhibition of the Alzheimer's Abeta-peptide oligomerization by Programme for Units of Excellence in R&D to “Departament de Ciències human serum albumin using saturation transfer difference and off-resonance Experimentals i de la Salut”; the Chilean Government through Fondecyt relaxation NMR spectroscopy. J Am Chem Soc 2007;129:4282–90. https://doi. org/10.1021/ja067367+. 11611065 and AFB170005 to AA, and REDES 180084 to AA and FJM; [23] Algamal M, Ahmed R, Jafari N, Ahsan B, Ortega J, Melacini G. Atomic-resolution map ISGlobal and IBEC are members of the CERCA Programme, Generalitat of the interactions between an amyloid inhibitor protein and amyloid β (Aβ) pep- de Catalunya. tides in the monomer and protofibril states. J Biol Chem 2017;292:17158–68. https://doi.org/10.1074/jbc.M117.792853. [24] Christenson RH, Behlmer P, Howard JF, Winfield JB, Silverman LM. Interpretation of Declaration of interest cerebrospinal fluid protein assays in various neurologic diseases. Clin Chem 1983; 29:1028–30. [25] Ahn S-M, Byun K, Cho K, Kim JY, Yoo JS, Kim D, et al. Human microglial cells synthe- None. size albumin in brain. PLoS One 2008;3:e2829. https://doi.org/10.1371/journal.pone. 0002829. [26] He XM, Carter DC. Atomic structure and chemistry of human serum albumin. Nature References 1992;358:209–15. https://doi.org/10.1038/358209a0. [27] Sugio S, Kashima A, Mochizuki S, Noda M, Kobayashi K. Crystal structure of human [1] Masters CL, Simms G, Weinman NA, Multhaup G, McDonald BL, Beyreuther K. Amy- serum albumin at 2.5 a resolution. Protein Eng 1999;12:439–46. loid plaque core protein in Alzheimer disease and down syndrome. Proc Natl Acad [28] Carter DC, Ho JX. Structure of serum albumin. Adv Protein Chem 1994;45:153–203. Sci U S A 1985;82:4245–9. P. Picón-Pagès et al. / Computational and Structural Biotechnology Journal 17 (2019) 963–971 971 [29] Kragh-Hansen U. Structure and ligand binding properties of human serum albumin. 2007;50:2.9.1–2.9.31. https://doi.org/10.1002/0471140864.ps0209s50 John Wiley Dan Med Bull 1990;37:57–84. & Sons, Inc. [30] Ghuman J, Zunszain PA, Petitpas I, Bhattacharya AA, Otagiri M, Curry S. Structural [46] Alford RF, Leaver-Fay A, Jeliazkov JR, O'Meara MJ, DiMaio FP, Park H, et al. The Ro- basis of the drug-binding specificity of human serum albumin. J Mol Biol 2005; setta all-atom energy function for macromolecular modeling and design. J Chem 353:38–52. https://doi.org/10.1016/j.jmb.2005.07.075. Theory Comput 2017;13:3031–48. https://doi.org/10.1021/acs.jctc.7b00125. [31] Fanali G, Di Masi A, Trezza V, Marino M, Fasano M, Ascenzi P. Human serum albu- [47] Berman HM, Kleywegt GJ, Nakamura H, Markley JL. The protein data Bank archive as min: from bench to bedside. Mol Aspects Med 2012;33:209–90. https://doi.org/10. an open data resource. J Comput Aided Mol Des 2014;28:1009–14. https://doi.org/ 1016/j.mam.2011.12.002. 10.1007/s10822-014-9770-y. [32] Simard JR, Zunszain PA, Ha C-E, Yang JS, Bhagavan NV, Petitpas I, et al. Locating high- [48] Adams R, Griffin L, Compson JE, Jairaj M, Baker T, Ceska T, et al. Extending the half- affinity fatty acid-binding sites on albumin by x-ray crystallography and NMR spec- life of a fab fragment through generation of a humanized anti-human serum albu- troscopy. Proc Natl Acad Sci U S A 2005;102:17958–63. https://doi.org/10.1073/ min Fv domain: an investigation into the correlation between affinity and serum pnas.0506440102. half-life. MAbs 2016;8:1336–46. https://doi.org/10.1080/19420862.2016.1185581. [33] Whitlam JB, Crooks MJ, Brown KF, Veng Pedrrsen P. Binding of nonsteroidal anti-in- [49] Crescenzi O, Tomaselli S, Guerrini R, Salvadori S, D'Ursi AM, Temussi PA, et al. Solu- flammatory agents to proteins—I. Ibuprofen-serum albumin interaction. Biochem tion structure of the Alzheimer amyloid β-peptide (1-42) in an apolar microenvi- Pharmacol 1979;28:675–8. https://doi.org/10.1016/0006-2952(79)90154-0. ronment. Eur J Biochem 2002;269:5642–8. https://doi.org/10.1046/j.1432-1033. [34] Bar-Or D, Rael LT, Lau EP, Rao NKR, Thomas GW, Winkler JV, et al. An analog of the 2002.03271.x. human albumin N-terminus (Asp-Ala-His-Lys) prevents formation of copper-in- [50] Gremer L, Schölzel D, Schenk C, Reinartz E, Labahn J, Ravelli RBG, et al. Fibril struc- duced reactive oxygen species. Biochem Biophys Res Commun 2001;284:856–62. ture of amyloid-β(1–42) by cryo–electron microscopy. Science 2017;358:116–9. https://doi.org/10.1006/bbrc.2001.5042. https://doi.org/10.1126/SCIENCE.AAO2825 (80- ). [35] Stanyon HF, Viles JH. Human serum albumin can regulate amyloid-peptide fiber [51] Kilambi KP, Reddy K, Gray JJ. Protein-protein docking with dynamic residue proton- growth in the brain interstitium: implications for Alzheimer disease. J Biol Chem ation states. PLoS Comput Biol 2014;10:e1004018. https://doi.org/10.1371/journal. 2012;287:28163–8. https://doi.org/10.1074/jbc.C112.360800. pcbi.1004018. [36] Algamal M, Milojevic J, Jafari N, Zhang W, Melacini G. Mapping the interactions be- [52] Nivón LG, Moretti R, Baker D. A pareto-optimal refinement method for protein de- tween the Alzheimer's Aβ-peptide and human serum albumin beyond domain res- sign scaffolds. PLoS One 2013;8:e59004. https://doi.org/10.1371/journal.pone. olution. Biophys J 2013;105:1700–9. https://doi.org/10.1016/j.bpj.2013.08.025. 0059004. [37] Kuo Y-M, Emmerling MR, Lampert HC, Hempelman SR, Kokjohn TA, Woods AS, et al. [53] Louis-Jeune C, Andrade-Navarro MA, Perez-Iratxeta C. Prediction of protein second- High levels of circulating Aβ42 are sequestered by plasma proteins in Alzheimer's ary structure from circular dichroism using theoretically derived spectra. Proteins disease. Biochem Biophys Res Commun 1999;257:787–91. https://doi.org/10.1006/ 2012;80:374–81. https://doi.org/10.1002/prot.23188. bbrc.1999.0552. [54] Bitan G, Teplow DB. Preparation of aggregate-free, low molecular weight amyloid- [38] Kuo Y-M, Kokjohn TA, Kalback W, Luehrs D, Galasko DR, Chevallier N, et al. Amyloid- beta for assembly and toxicity assays. Methods Mol Biol 2005;299:3–9. β peptides interact with plasma proteins and erythrocytes: implications for their [55] Chang Y-J, Chen Y-R. The coexistence of an equal amount of Alzheimer's amyloid-β quantitation in plasma. Biochem Biophys Res Commun 2000;268:750–6. https:// 40 and 42 forms structurally stable and toxic oligomers through a distinct pathway. doi.org/10.1006/bbrc.2000.2222. FEBS J 2014;281:2674–87. https://doi.org/10.1111/febs.12813. [39] Garcia-Garcia J, Valls-Comamala V, Guney E, Andreu D, Muñoz FJ, Fernandez- [56] Sievers F, Higgins DG. Clustal omega, accurate alignment of very large numbers of Fuentes N, et al. iFrag: a protein–protein interface prediction server based on se- sequences. Methods Mol Biol 2014;1079:105–16. https://doi.org/10.1007/978-1- quence fragments. J Mol Biol 2017;429:382–9. https://doi.org/10.1016/j.jmb.2016. 62703-646-7_6. 11.034. [57] Muñoz FJ, Inestrosa NC. Neurotoxicity of acetylcholinesterase amyloid beta-peptide [40] Janin J, Bahadur RP, Chakrabarti P. Protein–protein interaction and quaternary struc- aggregates is dependent on the type of Abeta peptide and the AChE concentration ture. Q Rev Biophys 2008;41:133–80. https://doi.org/10.1017/S0033583508004708. present in the complexes. FEBS Lett 1999;450:205–9. [41] Wass MN, Fuentes G, Pons C, Pazos F, Valencia A. Towards the prediction of protein [58] Castillo GM, Lukito W, Wight TN, Snow AD. The sulfate moieties of glycosaminogly- interaction partners using physical docking. Mol Syst Biol 2011;7:469. https://doi. cans are critical for the enhancement of beta-amyloid protein fibril formation. J org/10.1038/msb.2011.3. Neurochem 1999;72:1681–7. [42] Marín-López MA, Planas-Iglesias J, Aguirre-Plans J, Bonet J, Garcia-Garcia J, [59] Motamedi-Shad N, Monsellier E, Chiti F. Amyloid formation by the model protein Fernandez-Fuentes N, et al. On the mechanisms of protein interactions: predicting muscle acylphosphatase is accelerated by heparin and heparan sulphate through a their affinity from unbound tertiary structures. Bioinformatics 2018;34:592–8. scaffolding-based mechanism. J Biochem 2009;146:805–14. https://doi.org/10. https://doi.org/10.1093/bioinformatics/btx616. 1093/jb/mvp128. [43] Šali A, Blundell TL. Comparative protein modelling by satisfaction of spatial re- [60] Shin TM, Isas JM, Hsieh C-L, Kayed R, Glabe CG, Langen R, et al. Formation of soluble straints. J Mol Biol 1993;234:779–815. https://doi.org/10.1006/jmbi.1993.1626. amyloid oligomers and amyloid fibrils by the multifunctional protein vitronectin. [44] Martí-Renom MA, Stuart AC, Fiser A, Sánchez R, Melo F, Šali A. Comparative protein Mol Neurodegener 2008;3(1):16. https://doi.org/10.1186/1750-1326-3-16. structure modeling of genes and genomes. Annu Rev Biophys Biomol Struct 2000; [61] Llewellyn DJ, Langa KM, Friedland RP, Lang IA. Serum albumin concentration and 29:291–325. https://doi.org/10.1146/annurev.biophys.29.1.291. cognitive impairment. Curr Alzheimer Res 2010;7:91–6. https://doi.org/10.2174/ [45] Eswar N, Webb B, Marti-Renom MA, Madhusudhan MS, Eramian D, Shen M, et al. 156720510790274392. Comparative protein structure modeling using MODELLER. Curr Protoc Protein Sci