European Journal of Medicinal Chemistry 95 (2015) 136e152

Contents lists available at ScienceDirect

European Journal of Medicinal Chemistry journal homepage: http://www.elsevier.com/locate/ejmech

Original article

New mimetic peptides inhibitors of Аb aggregation. Molecular guidance for rational drug design
n A. Andujar a, b, Ellen Hubin c, d, e, Exequiel E. Barrera Guisasola a, b, Sebastia
ndez f, Carina M.L. Delpiccolo f, Kerensa Broersen c, Ivonne M. Kraan c, Luciana Me g a, b , Ricardo D. Enriz a, b, * Marcelo F. Masman , Ana M. Rodríguez a

Departamento de Química, Universidad Nacional de San Luis, Chacabuco 915, 5700 San Luis, Argentina IMIBIO-SL (CONICET), Chacabuco 915, 5700 San Luis, Argentina Nanobiophysics Group, MIRA Institute for Biomedical Technology and Technical Medicine, Faculty of Science and Technology, University of Twente, 7500 AE Enschede, The Netherlands d Structural Biology Brussels, Department of Biotechnology (DBIT), Vrije Universiteit Brussel (VUB), Pleinlaan 2, B-1050 Brussels, Belgium e VIB Department of Structural Biology, Pleinlaan 2, B-1050 Brussels, Belgium f Instituto de Química Rosario (CONICET-UNR), Facultad de Ciencias Bioquímicas y Farmac
euticas, Universidad Nacional de Rosario, Suipacha 531, 2000 Rosario, Argentina g Department of Molecular Dynamics, University of Groningen, Nijenborgh 7, 9747 AG Groningen, The Netherlands b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 10 June 2014 Received in revised form 17 March 2015 Accepted 18 March 2015 Available online 19 March 2015

A new series of mimetic peptides possessing a significant Ab aggregation modulating effect was reported here. These compounds were obtained based on a molecular modelling study which allowed us to perform a structural-based virtual selection. Monitoring Ab aggregation by thioflavin T fluorescence and transmission electron microscopy revealed that fibril formation was significantly decreased upon prolonged incubation in presence of the active compounds. Dot blot analysis suggested a decrease of soluble oligomers strongly associated with cognitive decline in Alzheimer's disease. For the molecular dynamics simulations, we used an Ab42 pentameric model where the compounds were docked using a blind docking technique. To analyze the dynamic behaviour of the complexes, extensive molecular dynamics simulations were carried out in explicit water. We also measured parameters or descriptors that allowed us to quantify the effect of these compounds as potential inhibitors of Ab aggregation. Thus, significant alterations in the structure of our Ab42 protofibril model were identified. Among others we observed the destruction of the regular helical twist, the loss of a stabilizing salt bridge and the loss of a stabilizing hydrophobic interaction in the b1 region. Our results may be helpful in the structural identification and understanding of the minimum structural requirements for these molecules and might provide a guide in the design of new aggregation modulating ligands. © 2015 Published by Elsevier Masson SAS.

Keywords: Molecular modelling Amyloid b-protein Mimetic peptides Aggregation modulating effect Molecular dynamics simulations

1. Introduction Alzheimer's disease (AD) is a devastating neurodegenerative disease that affects approximately 35 million people worldwide [1]. With the increase in human life span, AD will continue to afflict ever-increasing numbers of the elderly, augmenting an already serious health problem. Current AD drugs are targeting just the symptomatic mechanisms with limited benefit. In the absence of

* Corresponding author. Departamento de Química, Universidad Nacional de San Luis, Chacabuco 915, 5700 San Luis, Argentina. E-mail address: [email protected] (R.D. Enriz). http://dx.doi.org/10.1016/j.ejmech.2015.03.042 0223-5234/© 2015 Published by Elsevier Masson SAS.

effective drugs, the incidence of AD is expected to rise rapidly over the coming years. Thus, the discovery of disease-modifying therapeutics that can slow or ultimately halt disease progression is paramount. Currently, several therapeutic approaches targeting amyloidogenic proteins are under development, which include reducing the expression level of the amyloidogenic protein, increasing the clearance rates of misfolded amyloidogenic proteins, increasing the stability of properly folded amyloidogenic proteins, and inhibiting the self-assembly of amyloidogenic proteins into oligomers and fibrils [2]. The most widely accepted theory regarding the etiology of AD is known as the “amyloid cascade hypothesis”, which features the

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

amyloid b-protein (Ab) as the central pathological agent. Amyloid is a generic term used to refer to protein aggregates adopting a crossb-sheet structure [3]. Several studies have shown that the aggregation and fibril formation of Ab involve a change in conformation, from an intrinsically disordered peptide to insoluble fibrils [4]. Experimental data have revealed an underlying superstructure of anti-parallel, intra-molecular b-strands stabilized predominantly by backbone hydrogen bonds and salt bridges and parallel, intermolecular b-sheets stabilized predominantly by backbone hydrogen bonds and hydrophobic interactions [5e8]. A network of inter and intra-molecular hydrogen bonds maintains the stability of the fibril once it has formed [9e12]. Ab is a short peptide, ranging in length from 37 to 43 residues, with its most common alloforms being 40 and 42 residues in length. The aggregation and deposition of Ab in neural tissue is believed to be linked to neuronal cell death and loss of cognitive function seen in patients with AD [13]. Although the characteristic lesion of AD are large plaques, consisting of fibrilar Ab, the most toxic forms of Ab are generally believed to be soluble oligomers [14,15]. Extensive studies exploring the mechanism of Ab misfolding and aggregation have been ongoing for several decades [16e18] leading to the identification of a variety of inhibitors targeting self-assemble of Ab. These compounds can be divided into three general categories: small molecules, short peptides and antibodies [2]. Two types of small molecule inhibitors for amyloidogenic proteins have been identified so far: polyphenols and nonpolyphenols. Polyphenols comprise a large group of aromatic compounds containing one or more phenolic hydroxyl groups including epigallocatechin gallate, curcumin, wine-related polyphenols, coffee derived polyphenols and oleuropein. These groups may competitively interact with aromatic residues in amyloidogenic proteins, be sandwiched between them, prevent the pep interaction, and block the amyloid formation [19]. This process is explained in the “p-stacking” theory, in which the aromatic residues of amyloidogenic proteins interact with each other via pstacking [20]. Non-polyphenol inhibitors include small molecules like alkaloids, flavonoids, glycosides and phenazines. Most of these inhibitors contain one or several aromatic rings, which are key factors to the inhibitory effect according to the “p-stacking” theory mentioned above. Aside from the compounds described above, other small molecule inhibitors of Ab aggregation have been modeled based on the histological dyes used to characterize amyloid both in vitro and in vivo. This class of compounds includes sulfonated dyes like congo red, chrysamine G, and thioflavin S. Congo red is capable of binding to various Ab species ranging from monomers to mature fibrils and it has been shown to interrupt the normal aggregation of Ab peptides, thereby reducing toxicity. MD simulations revealed that congo red may play a role in interrupting sheet-to-sheet stacking and blocking strand-to-sheet extension [21]. With respect to peptide compounds, a pioneering effort by Ghanta et al., in 1996 demonstrated the possibility to use small peptides related to protein self-recognition regions as a viable strategy for the development of new Ab antiaggregants [22]. Tjernberg and coworkers synthesized a series of short peptides derived from Ab and found that KLVFF (Ab 16e20) is one of the nucleation sites in Ab [23]. They further observed that KLVFF can bind to full-length Ab and prevent its self-assembly into b-sheetrich amyloid fibrils. These results indicated that peptide based selfrecognition regions may inhibit the self-assembly of amyloidogenic proteins by acting as b-sheet breakers. Subsequent work expanded upon these original findings [24,25]. Soto and coworkers designed and synthesized b-sheet breaker peptides (BSB) based on the 17e20 region of Ab. Τhese BSB inhibited and disassembled amyloid

137

fibrils in vitro and, prevented Ab induced neurotoxicity in cells [26,27]. Moreover, treatment with these peptides inhibited neuronal death, brain inflammation and memory impairment in vivo [28]. These peptides showed low toxicity, low immunogenicity, high solubility and reasonably high brain uptake. Unfortunately, their biodisponibility was very poor as they were rapidly degradable and had low permeability to cross biological barriers imposing serious limitations for their therapeutic use [29]. The BSB affinity for Ab oligomers is significantly improved by appending polar groups to the peptide scaffold. Autiero and co-workers [30] combined polar groups with the peptide portion LPFFD, the wellknown BSB Soto's peptide. These compounds were able to efficiently interact with the protofibril ends preventing fibril growth. Chen et al. reported small mimetic peptides with similar ability to inhibit and reverse Ab misfolding and aggregation, but with potentially better drug-like properties [31]. Among these molecules, BSBM6 (1) (Fig. 1) possessed a significant antiaggregant effect and appeared as one of the best candidates for further studies. Chen et al. suggested that this compound interferes with electrostatic interactions during aggregation and destabilizes the Ab fibril internal hydrogen bond network that is necessary to maintain the bsheet structure, by forming strong hydrogen bonds with the Ab subunits [31]. Although this hypothesis seems reasonable, it should be noted that such assumptions have been made on the basis of simple docking analysis. Therefore, several questions regarding the molecular mechanism of these compounds at a sub-molecular level remain unanswered. The main limitation in using docking to identify binding modes of antiaggregating compounds against Ab is the rigidity of the target structure. If an antiaggregating small molecule alters the structure of Ab, such changes cannot be reflected during docking, and thus a structure candidate will be identified largely based on rigid interactions. Another challenge associated with docking is the scoring and refinement of docked poses. This particular hurdle is not unique to the study of Ab, but it affects all docking studies even those involving well-characterized receptors [32e34]. Molecular Dynamics (MD) simulations provide a molecularlevel view of a system as it evolves over time [35], making them particularly useful in studying Ab and its interactions with small molecules. Since Ab adopts a variety of structures along the aggregation pathway [36], MD simulations can be used to examine how these different structures may interact with small molecules that may serve as therapeutic antiaggregating compounds. Over the past few years, a growing number of theoretical studies have been conducted to analyze the interactions of Ab with antiaggregating molecules. Liu and co-workers conducted simulations of polyphenol (e)-epigallocatechin-3-gallate binding to Ab42 [37]. Raman and co-workers [38] and Takeda and co-workers [39,40] have reported the interactions of naproxen and ibuprofen with Ab fibrils using simulations employing implicit solvents. Lemkul and co-workers conducted simulations of binding of the flavonoid morin to a model of Ab42 protofibrils [41] and more recently examined the binding of this molecule to monomers and dimers of full-length Ab40 and Ab42 [42]. Attanasio and co-workers used MD simulations and NMR experiments to study the effect of carnosine on the inhibition of Ab42 aggregation by perturbing the H-bond network near the central hydrophobic core [43]. Bruce and coworkers [44] used MD simulations to compare the mode of interaction of an active (LPFFD) and inactive (LHFFD) b-sheet breaker peptide with an Ab fibril structure from solid-state NMR studies. All these efforts have shown several strengths and limitations of current techniques. Lemkul and Bevan reported a comprehensive review about the role of molecular simulations in the development of inhibitors of Ab peptide aggregation, which is a very complete and concise contribution on this topic [45]. Among their conclusions,

138

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

Fig. 1. Structural features of the mimetic peptides reported here.

the authors indicated that “combining docking and MD provides the most efficient and informative means of assessment of candidate molecules”. Despite some progress in this field, mechanistic details have only slowly emerged using experimental techniques, while the specific chemical features of small molecules responsible for antiaggregant activity remain largely uncharacterized. In aim of developing novel compounds capable of preventing Ab aggregation, it will be necessary to obtain more detailed information about their mechanism of action. In this paper, we report a possible mechanism of action for the mimetic peptide 1, previously reported by Chen and co-workers [31] to be a potent inhibitor of Ab aggregation. Moreover, we designed new compounds with significant Ab aggregation modulating activity on the basis of our molecular modelling study. 2. Results and discussion 2.1. Synthesis of 6-Benzyloxycarbonylamino-2-(4-methylaminobenzoylamino)-hexanoic acid (2) In a first attempt to obtain a new compound with antiaggregant activity, we synthesized the lower homolog of 1 (compound 2 in Fig. 1). The procedure of this synthesis is described in the Experimental section (4.1.2). Surprisingly, this compound only showed antiaggregant activity at very high concentrations as is discussed below (Section 2.3). This result was surprising because we did not expect a relatively small structural change (note that the difference between 1 and 2 is only a methylene group) to have such a strong effect on activity. It must be considered that equimolar concentration of compound 1 and Ab42 led to >70% inhibition of fibril formation [31]. Considering the low antiaggregant effect showed by compound 2 compared with 1, we decided to change our strategy to obtain new antiaggregating compounds. First, we tried to elucidate the mechanism of action of these compounds at a molecular level and then, with

this information in hand, we started the search of new compounds with the desired properties. 2.2. Molecular modelling study Despite of experimental observations [31], many questions remained about Ab-1 and Ab-2 interactions on a molecular level: i) what are the binding sites of 1 and 2 in Ab oligomers ii) which physicochemical factors control 1 and 2 binding, and iii) does 1 and 2 binding induce changes in the Ab peptide structure? Answering these questions will be important to further our understanding of the mechanism of Ab protofibril dissociation induced by these compounds and may aid the design of new antiaggregating agents. Computer simulations, such as MD, are well suited to provide molecular-level details of Ab-1 interactions. For MD simulations, we used the Ab42 pentameric model previously reported and designed by our own group [46]. Amino acid sequence of human Ab42 peptide is shown in Fig. S1 (Supplementary material). A graphical snapshot of the resulting Ab42 pentameric structure is shown in Fig. S2 in the Supplementary material. Compound 1 was docked into the Ab42 pentamer using the program Autodock Vina. These calculations were performed using the blind docking strategy, where the possible binding sites were obtained by scanning the entire surface of the pentameric peptide target [47]. Four distinct binding sites were identified for 1 (indicated as IeIV in Fig. 2). One of them was located in the polar, disordered region (site I in Fig. 2), site II corresponded to the top site of the pentamer, where an incoming monomer may take position during oligomerization, whereas sites III and IV were located at the b1 site. Owing to the disorder of site I region it is likely to represent just a momentary mode of 1 complexation. To determine the most likely site of attachment for 1, we performed relatively short MD simulations (10 ns each) to calculate the relative binding energy of each complex, using the Molecular Mechanics/Generalized Born Surface Area (MM/GBSA) method. Structural snapshots collected throughout the short MD

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

139

Fig. 2. Binding modes between 1 and the Ab42 pentamer. This figure shows the binding pockets IeIV, identified from docking procedures. Compound 1 is indicated with different colors depending on its location at the different sites: yellow, magenta, green and turquoise for sites I, II, III and IV, respectively. Binding free energies (in units of kcal mol
1) as estimated from MM/GBSA calculations for 1 to the Ab42 pentamer at the four binding sites are also indicated. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

simulations were used to estimate binding free energies (DG) of compound 1 complexation following MM/GBSA directives. Fig. 2 shows that the type III complex revealed the lowest energy of DG, indicating that this is the strongest binding mode for 1 to the Ab42 pentamer when compared to the other three binding sites (I, II and IV). Taken together these initial results, a central role for the type III arrangement was identified together with the need for a more detailed analysis of the structural alterations affecting the Ab42 pentamer. All further evaluations were therefore restricted to the third type of complexes (indicated as site III in Fig. 2), where the ligand is bound at the central hydrophobic core. The docking study performed for 2 gave similar results to those obtained for 1 (data not shown). To explore the dynamic behaviour of the type III complex, more extensive MD simulations were carried out in explicit water (200 ns sampling time, GROMACS software for molecular modelling, version 4.5.3). Initial and final structures obtained from the MD simulations were compared for the wild type (WT) Ab42 pentamer (Fig. 3A), as well as for complexes with 1 (Fig. 3B) and 2 (Fig. 3C). Results indicated that only minor changes occurred in the WT Ab42 pentameric structure within the time-frame of 200 ns (Fig. 3A). The pentameric structure was slightly more compact or “tight” in comparison to its initial structure. In contrast, significant changes occurred for the complex with 1, mainly in the plane-parallel organization of individual b-sheets. A trend towards the formation of a more twisted arrangement became apparent, which might prevent the fibril structure formation [46] (Fig. 3B). It is interesting to note that the 2-Ab42 complex shows an intermediate situation between WT Ab42 pentamer and 1-Ab42 complex (Fig. 3C), in agreement with the initially obtained experimental results suggesting activity of 2 only at high concentrations. Another major difference between the two simulated complexes is that 1 is inserted deeper into the Ab42 pentamer with respect to 2 at the end of the simulation. This difference can be appreciated when comparing the final structures (Fig. 3B and C). Moreover, the secondary structure of Ab42 pentamer in the complex with compounds

Fig. 3. A spatial view comparing the initial (red) and final (blue) structural snapshots from a 200 ns MD simulation of the Ab42 pentamer in the absence of compound (A). The equivalent snapshots are shown for the complex 1-Ab42 pentamer (B) and for the complex 2-Ab42 pentamer (C). The carboxylic group is remarked in a yellow circle. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

1 and 2 was characterized by DSSP method. Figure S3, in the Supplementary material, illustrates the temporal development of the secondary structure content for these complexes along the 200 ns trajectory. In addition to the significant trend predicted by the MD simulations, we also needed to obtain a quantitative difference for the antiaggregant effect of 1 and 2. Therefore we measured parameters or descriptors that allowed us to quantify the effect of these compounds as potential antiaggregating agents. 2.2.1. Compounds 1 and 2 disrupt the fibril structure in the Ab42 pentamer We previously described a quantitative method using parameters to characterize Ab protofibril structure likely to be relevant for AD outbreak and clinical diagnosis [46]. In another study, we

140

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

employed some of these parameters to quantify the antiaggregant effect of fullerene [48]. Here the dihedral angle d, defined between two adjacent layers of Ab sheets is used to quantify the helix twist that is indicative for the development of Ab fibrils. The b-helix is a plausible structural motif for amyloid fibrils, since it is primarily a b-sheet structure with the proper cross-b orientation and is known to occur in bona fide proteins [49]. This type of helical structure enables the hydrogen bonding between the b-strands to be extended over the total length of the amyloid fibrils, thereby accounting for their characteristic rigidity and stability. A schematic of the definition of the dihedral angle d is given in Fig. 4A. It involves residues Ala42X, Val36X and Val36X
1, Ala42X
1, referring to adjacent Ab chains. Positions of Ca atoms (depicted as spheres in Fig. 4A) of the above residues will determine the value of d, which was estimated to be in the order of
10 in the original report [46]. Dihedral angles d were computed in the current study and the corresponding results are summarized for the WT Ab42 pentamer (Fig. 4B) and its complexes with 1 and 2 (Fig. 4C and D, respectively). WT-quantification of d appears to be slightly above the
10 base line with a rather homogeneous trend for all various pairs of chains (Fig. 4B). However, Ab42 pentamer in complex with 1 showed significant deviation from WT with respect to d defined between chains D and E (Fig. 4C). The sudden rise to a value of about
100/
110 occurring after approximately 50 ns of MD

simulation is indicative of b
helix structure disruption caused by 1. It should be noted that a significant increment of about 25 also occurred for the d defined between chains A and B after 150 ns of simulation. Regarding the complex between 2 and Ab42 pentamer we can also observe a deviation from WT with respect to d defined between chains D and E. However in this case the value is markedly lower in comparison with the complex of 1 (about
60 after 50 ns of simulations). 2.2.2. Compound 1 alters key-interactions in the structure of the WT Ab42 pentamer Lemkul and Bevan highlighted the importance of the interstrand salt bridge Asp23XeLys28Xþ1 for fibril structure formation [50]. Salt bridge formation between residues Asp23XeLys28Xþ1 was also analyzed here (Fig. 5). Results are summarized in Fig. 5B for the WT Ab42 pentamer, Fig. 5C for the complex with 1 and Fig. 5D for the complex with 2. After approximately 80 ns, the distance between the charged moieties of chains B and C was greater than 1.5 nm in the presence of 1 (Fig. 5C) suggesting that this compound disrupts inter-strand salt bridge formation. This increased distance persisted for much of the remaining simulation time, far longer than the results obtained for WT Ab42 that indicated the salt bridge distance persisted at a distance of approximately 1 nm throughout the MD

Fig. 4. Ab fibril disruption by 1 and 2. (A) Schematic of the characteristic dihedral angle, d, introduced by Masman et al. [44] to describe the anticipated b-helix structure of Ab fibrils. d is a measure of b-helix twist. It is defined from the position of four Ca atoms of residues Ala42x, Val36x, Val36x
1 and Ala42x
1, on two adjacent chains of the Ab42 pentamer. (B) Dihedral angle d for the Ab42 pentamer in the absence of compound as a function of MD simulation time. (C) Dihedral angle d obtained for the complex 1-Ab42 pentamer as a function of MD simulation time. (D) Dihedral angle d obtained for the complex 2-Ab42 pentamer as a function of MD simulation time.

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

141

Fig. 5. Snapshot showing the interchain salt bridges on the Ab42 pentameric model (A). Maintenance and recovery of the inter-chain salt bridge between residues Asp23x and Lys28xþ1 as observed during 200 ns of MD simulation of the Ab42 pentamer in the absence of compound (B), for complex 1-Ab42 pentamer (C) and for complex 2-Ab42 pentamer (D).

simulations. It is interesting to remark that 2 did not alter keyinteractions in the structure of the Ab42 pentamer (Fig. 5D). 2.2.3. Compound 1 mainly affects the structure of the Ab42 pentamer near its binding site To evaluate how the structure of the Ab42 pentamer in close proximity of the binding site of 1 is affected we evaluated the distances between the different chains in the area called b1 (amino acids Val18-Ser26). Interestingly, only in the presence of this compound the distance between chains B and C was significantly increased from the 150 ns of simulation (Fig. 6A and B). In contrast, these distances were not altered in presence of 2 (Fig. 6C). Region b1 is formed by nine amino acids. To evaluate in more detail the behaviour of 1 in this region, we performed a punctual analysis per residue for chains B and C which were the most affected by 1. During the MD simulations chains B and C were opened gradually by compound 1 (Fig. S4, Supplementary material). Apart from changes in close proximity of 1, regions distant from the binding site type III were also significantly modified by the presence of this compound. Therefore, to further assess the structural effects caused by the binding of 1, we evaluated distances between the different chains in areas remote from the binding site. The presence of 1 produced a significant increase in the distance between the two outer chains (D and E) in the b2 region (amino acids Ile31-Ala42), while the remainder of the distances between the chains were similar to those observed for Ab42 pentamer in absence of any compound (Fig. 7A and B). On the contrary, 2 did not produce alterations far away from the binding site (Fig. S5,

Supplementary material). Visual inspection of the behaviour of 1 and 2 suggested that the carboxylic group plays an important role in the process by which these compounds interact with the Ab42 pentamer. As this group is highly exposed to the solvent, the internalization of the compounds into the hydrophobic core is significantly affected. For 2 the carboxylic group was not able to be inserted in the hydrophobic core during the 200 ns of MD simulation (Fig. 3C). Conversely, the interaction between Lys100 and the carboxylic group of 1 allowed the deep insertion of this compound (Fig. 3B and Fig. 8B). For compound 2 this interaction was not observed (Fig. 8C). It should be noted that this structural aspect was taken into account in the design of new potential inhibitors replacing in some compounds the carboxylic by a succinimide group. The insertions of these compounds are also shown as side views (Fig. S6, Supplementary material). On the other hand, it is important to remark that our simulations suggested that aromatic rings placed at R2 are preventing the p-stacking interaction between residues Phe61 and Phe103. These interactions are well appreciated in Fig. 8, which displays binding pockets of compounds 1 and 2. As well, the aromatic ring at R3 is involved in the insertion of compound 1 into the hydrophobic core. This interaction zone is also shown for the WT Ab42 pentamer to visualize the typical p-stacking interaction between Phe residues (Fig. 8A). It is not an easy task to explore all possibilities for these mimetic peptides acting as antiaggregating agents considering the many possible molecular mechanisms between these compounds and Ab. However, we have carried out sufficient simulations to feel confident that this might be one of the possible mechanisms for these

142

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

compounds interacting with low size oligomers. Once a possible mechanism of action of these compounds was proposed, our next step was to obtain new compounds structurally related to 1 with stronger antiaggregant properties. Therefore we performed a structural based virtual selection in order to select the best candidate structures. 2.2.4. Structural-based virtual selection On the basis of their structural similarity with 1, 28 compounds were selected. These compounds maintained the lysine backbone but showed variations in three portions of the molecule: i) the aamine group, ii) the a-carboxylic group and iii) the ε-amino group (Fig. 1). Simulations suggested that the aromatic character of both rings was important for the molecular interactions, while replacement of the carboxylic group by a polar group could also be beneficial for a better insertion of the ligand. Insertion of the compound into the Ab42 pentameric model appeared one of the key components to induce the anti-aggregating effect of compound 1. We analyzed the binding mode of these compounds using a docking analysis with the same procedure described previously. We took into account only those compounds that showed the b1 portion as the preferential zone to form the complex with the Ab peptide. Compounds 3e12 (Fig. 1) were selected to perform the MD simulations. The simulated compounds showed different behaviours ranging from similar to 1 (noticeably affecting the pentameric structure of Ab42) to no significant effect on structure, whereas other compounds showed an intermediate behaviour similar to that of 2. We discuss here the results obtained for four compounds which are representative of the entire series: two compounds that produced the most noticeable change (4 and 7) and two compounds that least affected the pentameric structure (5 and 6). The MD simulations results obtained for 4 and 7 were comparable to those observed for 1. Fig. 9A and B shows how these compounds are deeply inserted in to the core of the Ab42 pentamer. In contrast, 5 and 6 were detached from the core of the pentamer at the end of the 200 ns simulation (Fig. 9C and D, respectively). Once again, the carboxylic group seems to have a significant role in the insertion process into the Ab42 pentamer. In contrast with 5 and 6, compounds 4 and 7 have a substituted carboxylic group. We suggest that this substitution might facilitate the insertion process mentioned above. Additional information regarding compounds 6 and 7 is given in Fig. S7 in the Supplementary material. From these data we postulate that compounds 5 and 6 are marginally effective in modulating aggregation behaviour of Ab42 while compounds 4 and 7 are expected to inhibit Ab42 aggregation more strongly. To corroborate our findings we performed biophysical assays for the most representative compounds. 2.3. Experimental corroboration for the molecular modelling study and MD simulations To experimentally confirm our previous results we measured the activities of four representative compounds of the entire series. Molecular modelling study predicts that compounds 4 and 7 have significant antiaggregant effect at least comparable to that previously reported for 1. Hence the antiaggregant activities of both compounds 4 and 7 were studied. As the simulations indicated that compounds 5 and 6 did not significantly affect the structure of the

Fig. 6. (A) Average inter-chain distances of the mass center of the Ca's versus time of the b1 region of the Ab42 pentamer in the absence of compound. (B) and (C) show the average inter-chain distances of the mass center of the Ca’s versus time of the b1 region of the complexes 1-Ab42 pentamer and 2-Ab42 pentamer, respectively.

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

143

Fig. 7. Average inter-chain distances of the mass center of the Ca’s versus time of the b2 region of Ab42 pentamer in the absence of compound (A) and complex 1-Ab42 pentamer (B).

Ab42 pentamer, the activities of these compounds were included as a negative control. The experimental results obtained for 2 are also shown. To assess the antiaggregant activities of the representative compounds, Ab42 fibril formation after 7 days of incubation at 37  C was measured in the presence and absence of these compounds, using thioflavin T (ThT) fluorescence (Fig. 10). Thioflavin T is a dye commonly used to detect fibrils as its fluorescence emission is largely enhanced upon binding to amyloid fibrils [51]. Ab42 fibril formation at this time point was significantly inhibited in a dosedependent manner by compounds 2, 4 and 7 as seen by decreased ThT fluorescence intensities (Fig. 10A, B and C). Compounds 4 and 7 were maximally active at a concentration of 100 mM, whereas a higher dose of 2 was necessary to induce a similar effect (3200 mM). On the contrary, negative control compound 6 did not lower the Ab42 response to ThT (Fig. S8, Supplementary material). Accordingly, transmission electron microscopy (TEM) revealed networks of intertwined and negatively stained fibrils for samples containing Ab42 and one of the nonactive compounds 5 and 6 (Fig. 11). For Ab42 incubated in presence of compounds 2 or 4, fewer fibrils were detected compared to Ab42 alone (Fig. 11). No fibrils were seen for Ab42 incubated in presence of 100 mM of 7 indicating the high anti-aggregating effect of this compound (Fig. 11). However, it should be taken into consideration that TEM cannot be exclusively used as a quantitative method for fibril formation. As a final assessment of the effect of compounds on the formation of oligomeric Ab42 reported previously to be responsible for the neurotoxic response and cognitive defects observed in AD patients, dot blot analysis was performed with the anti-amyloid precursor protein antibody DE2B4, for detection of the total amount of Ab present, and with the oligomerspecific A11 antibody (Fig. 12). Samples were blotted after 1.5 h of incubation at 37  C. In the absence of antiaggregating compounds, a high A11 response is expected at this time point as it has been shown previously that this corresponds to an Ab oligomer-enriched fraction [52]. Consistently, high dot blot intensities (and thus high reactivity with DE2B4 and A11 antibodies) were recorded for Ab42 in presence of the inactive compounds 5 and 6. Compounds 4 and 7 however inhibited aggregation from 100 mM and higher, as demonstrated by the lack of reactivity with DE2B4 and A11 antibodies at these concentrations. Interestingly, antibody DE2B4 shows a lack of reactivity with Ab in the presence of compounds 4 and 7. We hypothesize, based on results obtained from our molecular modelling studies, that this lack of reactivity can be explained by a partial overlap of the Ab epitope that is recognized by DE2B4 and the binding site of compounds 4 and 7. Based on the results of these doseeresponse assays, a fixed concentration of

100 mM compound was used to obtain further insight into the effect of selected compounds on the kinetics of Ab aggregation. Ab42 was incubated in the presence of active compounds 4 and 7, moderately active compound 2 and inactive compounds 5 and 6. Aggregate formation was monitored by means of ThT fluorescence intensity (Fig. 13) and oligomer formation by A11 antibody detection (Fig. 14). As reference, the effect of a known inhibitor, scyllo-inositol (SI), on Ab aggregation kinetics was compared with the compounds under study [53]. Ab alone showed significant A11-positive oligomer staining upon incubation up to 30 h after which presumably these oligomers had been converted into A11-negative fibrils. Inactive compounds 2, 5, and 6 or SI did not reduce A11-positive oligomer recognition compared to their respective controls. Contrary, active compounds 4 and 7 completely inhibited A11-positive oligomer formation during the time frame of the experiment. The effect of compounds on the kinetics of Ab aggregation was also monitored at various time points. The aggregation kinetics did not change in the presence of compounds 2, 5, 6 or SI compared to their respective controls. Interestingly, at early time points, up to 1.5 h of incubation, rapid formation of ThT-positive but A11-negative species was observed in the presence of active compounds 4 and 7. This initial rapid development of ThT-positive species was abolished after 5 h of incubation, beyond which ThT levels decreased. This observation implies that ThT positive aggregates are present at early time points of incubation which largely disappear at later time points. ThT is known as a reporter for both amyloid fibril morphology as well as amyloid fibril mass [54]. A decrease of ThT fluorescence intensity therefore can be accommodated by either a conversion from soluble to insoluble material or a structural conversion from ThT positive to ThT negative aggregates. The first scenario can be excluded as TEM showed no indication of such amorphous aggregates which are often correlated to low ThT binding affinity. We therefore stipulate that ThT positive aggregates are rapidly formed, also in the presence of compounds 4 and 7, but that these aggregates then destabilize upon prolonged incubation with the compounds and disaggregate/disintegrate, resulting in a decreased ThT signal. Contrary, ThT levels of Ab in presence of inactive compounds increased upon prolonged incubation, indicating the formation of fibrils. From these results it is evident that the experimental results are in agreement with those obtained from the molecular modelling, indicating that compounds 4 and 7 possess a significant aggregation modulating effect and deserve further studies. It should be noted that compounds 4 and 7 are N-hydroxysuccinimide active ester aminoacid derivatives commonly used as reactants in peptide synthesis [55]. Therefore, possible alternative pathways, including chemical reaction with Ab, cannot be ruled out to explain the experimentally observed aggregation modulating activity of these

144

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

to the Ab42 monomer. Thus, MD simulations were carried out in explicit water for these complexes (50 ns sampling time, GROMACS software for molecular modelling, version 4.5.3). Previous experimental and theoretical studies have proven that conformational conversion from the initial a-helix to b-sheet is a crucial early step in Ab amyloidogenesis [56]. To characterize the effect of 6 and 7 on the conformational transition of Ab42 monomer, the secondary structure was determined by the DSSP method. Fig. 16 illustrates the temporal development of the secondary structure content for the Ab42 monomeric model along the 50 ns trajectory. During the simulation a significant decrease of helical forms was observed for WT Ab42 monomer (Fig. 16A). In contrast, for complex 7-Ab42 monomer the a-helix content was highly conserved during the 50 ns of simulation (Fig. 16B). Regarding 6Ab42 complex a loss of helical content with short b-sheet structures in the region comprised at positions 31e33 and 39e42 was observed (Fig. S9, Supplementary material). Masuda et al. identified the toxic conformer of Ab42 with a turn at positions 22 and 23, “toxic turn”, by solid-state NMR [57]. More recently, experimental results suggested that Ab42-mediated toxicity is caused by this turn that generates toxic oligomers [58,59]. On the basis of these findings, we also investigated the effect of compounds 6 and 7 on the formation of the so called “toxic turn” at positions 22 and 23 in the Ab42 monomer. Fig. 17 displays the structural conformations of complexes 6 and 7-Ab42 monomer (A and B, respectively). This figure shows that in the presence of the active compound 7 the ahelix content is highly preserved including residues Glu22 and Asp23. We hypothesized that the interaction of compound 7 with Ab42 hinders the formation of the “toxic turn” in this Ab42 monomeric model. In contrast, in presence of the inactive compound 6, the toxic conformer of Ab42 with a turn at positions Glu22 and Asp23 was observed. Considering that Ab42 monomers are the seeds in the aggregation process, these theoretical results are in complete agreement with our experimental results and contribute to understand the action mechanism of these aggregation modulating compounds. Taking into account the high level of intrinsic degrees of freedom of Ab42 monomer, the binding of this putative antiamyloid compound is extremely difficult to achieve in a suitable manner. However, these results obtained for compound 7 are encouraging and represent the first step of a more complete exploration of the conformational landscape of this complex interacting system. 3. Conclusions

Fig. 8. Structural snapshot (the final instance of a 200 ns MD simulation) of the WT Ab42 pentamer (A), complex 1-Ab42 (B) and complex 2-Ab42 (C). The interacting residues are labeled and shown in sticks. The Ab42 pentamer is shown in gray ribbon, while the compounds 1 and 2 are shown in green and orange, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

compounds. At this stage of our study a question which arises is: May these compounds affect the molecular behaviour of the Ab42 monomer? To answer this question we performed first a blind docking study of compounds 6 and 7 into an Ab42 monomeric model using the program Autodock Vina. Three distinct binding sites were identified for each compound (indicated as IeIII in Fig. 15). The type II complex revealed the strongest binding mode for both compounds

We report a new series of mimetic peptides possessing a significant Ab aggregation modulating effect. Among the compounds tested, 4 and 7 displayed the strongest activities, both in MD/ docking studies as well as in biophysical assays. These compounds were obtained based on a molecular modelling study which allowed us to perform a structural-based virtual selection. By combining docking analysis with MD simulations, we reported a simple and generally applicable procedure to evaluate the binding process of ligands interacting with an Ab42 pentameric model while providing to some extent a detailed picture for the binding mechanism of these ligands at the molecular level and their potential action mechanism. Monitoring Ab aggregation by ThT fluorescence and TEM revealed that fibril formation is significantly decreased in presence of 4 and 7. Moreover, dot blot analysis showed a decrease of soluble oligomers strongly associated with cognitive decline in AD [60]. On the other hand, significant alterations in the structure of an Ab42 protofibril model have been identified by computer

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

145

Fig. 9. Lateral spatial view of complexes obtained for compounds 4 (A), 7 (B), 5 (C) and 6 (D) with the Ab42 pentamer. This view allows us to appreciate the depth at which the different complexes are inserted after 200 ns of simulation.

Fig. 10. Ab42 amyloid fibril formation is affected in a dose-dependent manner by compounds 2, 4 and 7. Dose-response amyloid fibril formation of 25 mM Ab42 incubated in the presence of compounds at 37  C for 7 days was monitored using ThT fluorescence intensity at 485 nm (lex ¼ 450 nm): (A) 4, (B) 7 and (C) 2. Values represent results of three independent replicates. Statistical significance of the results was established by P-values using paired two-tailed t-tests. Statistical significance levels were * P < 0.05, **P < 0.005 and ***P < 0.0005.

146

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

Fig. 11. Characterization of Ab42 amyloid fibril formation by transmission electron microscopy. Images were obtained from 25 mM Ab42 incubated in the presence of compounds 2, 4e7 for 7 days at 37  C. The scale bar represents 500 nm.

simulations. These include: (i) the destruction of the regular helical twist, (ii) the loss of a stabilizing salt bridge and (iii) the loss of a stabilizing hydrophobic interaction in the b1 region. All these structural changes are caused by the complexation with those compounds possessing the strongest activities. In addition, we performed MD simulations by using an Ab42 monomeric model. In these simulations, the active compound 7 affected the monomer structure preserving its helical content and preventing the formation of the “toxic turn” involved in the propagation of toxic oligomers. Our theoretical and experimental results give interesting information which may be helpful in obtaining a better understanding of the molecular mechanism between these ligands and their potential to disrupt Ab aggregation at molecular level and can provide a guide in the design of new antiaggregating ligands. 4. Experimental section 4.1. Synthesis 4.1.1. General procedures Chemical reagents were purchased from commercial sources and were used without further purification. Solvents were

analytical grade or were purified by standard procedures prior to use. 1H NMR spectra were recorded on a Bruker Avance 300 spectrophotometer operating at 300 MHz, using deuterated chloroform (CDCl3) as solvent and tetramethylsilane as internal standard. 13C NMR spectra were recorded on the same apparatus at 75 MHz with CDCl3 as solvent and reference (76.9 ppm). Mass spectra were performed on an LC-MS Bruker micrOTOF-Q II spectrophotometer. Silica gel aluminum plates (Merck 60 F254) were used for analytical TLC. Flash column chromatography was performed using Merck silica gel 60 (230e400 mesh). Merck Silica gel 60 PF254 containing gypsum was used for preparative layer chromatography. 4.1.2. Synthesis of 6-Benzyloxycarbonylamino-2-(4-methylaminobenzoylamino)-hexanoic acid (2) H-Lys(Z)-OBzl hydrochloride (568.3 mg, 1.4 mmol) was dissolved in dimethylformamide (DMF) (16 mL) and 4-methylamino benzoic acid (211 mg, 1.4 mmol), 1-hydroxibenzotriazole (98 mg, 0.73 mmol), N-(3-Dimethylaminopropyl)-N0 -ethylcarbodiimide hydrochloride (268.4 mg, 1.4 mmol) and N,N-diisopropylethylamine (0.24 mL, 1.4 mmol) were successively added. The solution was stirred at room temperature for 16 h. After this time, the

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

147

Fig. 12. Dot blot analysis of the effect of compounds 4e7 on Ab42 fibril formation. Images were obtained from 25 mM Ab42 incubated in the presence of compounds 4e7 at 37  C for 1.5 h. Dot blot analysis was performed with the anti-amyloid precursor protein antibody DE2B4, for detection of the total amount of Ab42 present, and with the oligomer-specific A11 antibody.

reaction mixture was poured over 73 mL of H2O and NaCl (4.5 g) was added. The organic material was then extracted with ethyl acetate (4 x 45 mL). The combined organic phases were washed with H2O (10 mL) and Brine (10 mL), dried over Na2SO4 and evaporated under reduced pressure. This crude material was purified by silica gel column chromatography (hexane/EtOAc, 45:55) providing 537.4 mg of pure compound 6-benzyloxycarbonylamino2-(4-methylamino-benzoylamino)-hexanoic acid benzyl ester (74% yield). 1HNMR (300 MHz, CDCl3) d (ppm) 7.66 (d, J ¼ 8.7 Hz, 2H, Ar), 7.34 (m, 9H, Ar), 6.63e6.50 (m, 3H, Ar), 5.20 (d, J ¼ 12.3 Hz, 1H, eCH2Ph), 5.14 (d, J ¼ 12.3 Hz, 1H, eCH2Ph), 5.04 (m, 2H, eCH2Ph), 4.85 (m, 1H, eNeCHeCOe), 3.12 (m, 2H, eCH2eNHCO), 2.84 (s, 3H, eNeCH3), 1.99e1.75 (m, 2H; eCH2CH2e), 1.53e1.30 (m, 4H, eCH2CH2e); 13CNMR (75 MHz, CDCl3) d (ppm) 172.8, 167.1, 156.5, 152.1, 136.6, 135.3, 128.8, 128.6, 128.4, 128.3, 128.0, 121.6, 111.3, 67.1, 66.5, 52.1, 40.5, 32.3, 30.2, 29.2, 22.3. HRMS calcd for C29H33N3O5 (MNaþ, m/z): 526.2312; found: 526.2305. The 6-benzyloxycarbonylamino-2-(4-methylamino-benzoylamino)-hexanoic acid benzyl ester obtained at the previous step (462.7 mg, 0.89 mmol) was dissolved in MeOH (6 mL) and treated with 1.2 mL of 1 N NaOH at room temperature for 3.5 h, optimum reaction time was determined by HPLC analysis of the reaction mixture. After that, the solvent was evaporated under reduced pressure and dried under high vacuum. An aliquot of this crude material (99 mg) was purified by preparative layer chromatography (90 mL DCM/10 mL MeOH/0.5 mL formic acid) obtaining 36 mg of

pure 6-benzyloxycarbonylamino-2-(4-methylamino-benzoylamino)-hexanoic acid (2) (45% yield). 1HNMR (300 MHz, CDCl3) d (ppm) 7.65 (d, J ¼ 8.4 Hz, 2H, Ar), 7.29 (s, 5H, Ar), 6.46 (d, J ¼ 8.4 Hz, 2H, Ar), 5.02 (m, 2H, eCH2Ph), 4.61 (m, 1H, eNeCHeCOe), 3.14 (m, 2H, eCH2eNHCO), 2.79 (s, 3H, eNeCH3), 1.92e1.74 (m, 2H, eCH2CH2e), 1.52e1.35 (m, 4H, eCH2CH2e); 13 CNMR (75 MHz, CDCl3) d (ppm) 175.0, 168.3, 156.9, 152.3, 136.5, 129.1, 128.5, 128.1, 128.0, 127.0, 120.9, 111.3, 66.7, 52.9, 40.4, 30.2, 29.7, 29.4, 22.4; HRMS calcd for C22H27N3O5 (MNaþ, m/z): 436.1843; found: 436.1823. 4.2. Chemical compounds Compounds 4 (Na-Boc-Nε-Cbz-L-lysine N-hydroxysuccinimide ester, CAS: 34404-36-9), 5 (Na-Fmoc-Nε-Z-L-lysine, CAS: 8606082-4), 6 (Na,Nε-Di-Z-L-lysine, CAS: 405-39-0) and 7 (Na,Nε-Di-Z-Llysine hydroxysuccinimide ester, CAS: 21160-83-8) and scylloinositol (CAS: 488-59-5) were purchased at SigmaeAldrich possessing a purity of 98.0% (HPLC or TLC). 4.3. Biophysical studies 4.3.1. Preparation of Ab e compound solutions Ab1-42 (rPeptide) was dissolved according to the standard procedure developed and validated in our laboratory [51]. In short, Ab1-42 was dissolved in 1,1,1,3,3,3-hexafluoro-2-propanol

148

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

(HFIP). HFIP was evaporated using nitrogen gas and the peptide film was redissolved using dimethyl sulfoxide (DMSO). The peptide was separated from DMSO by elution from a HiTrapTM desalting column (GE Healthcare, cat. # 17-1408-01) into a PBS buffer. The resulting samples were kept on ice until experiments started with a maximum lag time of 30 min. Peptide concentration was determined using absorbance at 280 nm and an extinction coefficient of 1490 M
1 cm
1 and corrected to 25 mM in PBS. Ab peptide was incubated under quiescent conditions at 37  C in the presence of chemically synthesized compounds at a range of concentrations from 0 to 3200 mM in DMSO (final DMSO concentration 5%). 4.3.2. Dot blot After 1.5 h of incubation a volume of 5 mL Ab incubated with a concentration range of compounds was spotted onto a nitrocellulose membrane. The membranes were blocked in phosphate buffered saline containing 0.2% Tween-20 (1 h, 25  C), and incubated (overnight, 4  C) with primary oligomer-recognizing A11 antibody (Invitrogen, cat. # AHB0052), diluted 1:4000 in 100 mM Hepes, pH 7.0 [61]. After incubation (0.5 h, 25  C) with a secondary antirabbit-HRP-tagged antibody (Promega, cat. #W4011), diluted 1:8000 in phosphate buffered saline containing 0.05% Tween-20, the membranes were visualized using the ImmobilonTM Western chemiluminescent HRP substrate system. The procedure was repeated to detect total Ab present using anti-amyloid precursor protein mouse monoclonal antibody DE2B4 (Abcam, cat. # ab12266) by first blocking the blot for 1 h at 25  C in phosphate buffered saline containing 0.05% Tween-20 and 5% milk, followed by overnight incubation at 4  C in 1:20,000 diluted DE2B4 in phosphate buffered saline containing 0.05% Tween-20 and 5% milk. Goat anti-mouse-HRP-tagged antibody was used as secondary antibody and membranes were visualized using ImmobilonTM Western chemiluminescent HRP substrate. The A11 staining procedure was repeated by incubation of Ab in the absence and presence of 100 mM compounds 2, 4, 5, 6, 7 and scyllo-inositol by collecting aliquots upon incubation times of 0.5 h, 1.5 h, 5 h, 20 h, 30 h, 55 h and 70 h. 4.3.3. Thioflavin T fluorescence After 7 days of incubation, a volume of 100 mL incubated Ab/ compound samples was incubated in presence of 25 mM ThT for 10 min at room temperature. The extent of aggregation was monitored using a Varian Cary Eclipse fluorimeter upon excitation at a wavelength of 450 nm while recording the emission spectrum at a wavelength range from 470 to 560 nm. Measurements were performed as independent triplicates. Recorded values were averaged and background measurements (buffer containing 25 mM ThT and compound) were subtracted. Further, in situ ThT measurements were performed by incubating a volume of 100 mL Ab at a concentration of 25 mM, prepared as described above, but in the presence of 15 mM ThT and 100 mM compounds 2, 4, 5, 6, 7 and scyllo-inositol. Samples were deposited into a Greiner Bio-one 96well Cellstar plate (cat. 655185) and sealed with transparent Greiner VIEWseal platesealer to prevent evaporation. Samples were incubated at 37  C and at various time intervals, the ThT fluorescence intensity was determined using a Tecan Infinite M200Pro platereader upon excitation at a wavelength of 450 nm while recording the emission spectra at a wavelength range from 470 to

Fig. 13. The effect of compounds on Ab42 aggregation kinetics. Controls include the incubation of Ab in 5% DMSO (compounds were dissolved in 100% DMSO after which they were diluted 20-fold in the presence of Ab) and Ab with 5% Millipore H2O (SI was dissolved in H2O followed by 20-fold dilution in the presence of Ab).

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

149

with 2% (wt/vol) phosphatidic acid. Samples were studied with a JEOL JEM-1400 microscope (JEOL Ltd., Tokyo, Japan) at 80 kV. Images were collected from three independently prepared Ab solutions.

4.4. Molecular dynamic simulations

Fig. 14. The effect of compounds on Ab42 oligomer formation. Dot blot images were obtained upon incubation of 25 mM Ab42 in the presence of 100 mM of compounds 2, 4, 5, 6, 7 or scyllo-inositol (SI) as reference compound at 37  C for various amounts of time. Dot blot analysis was performed using the oligomer-specific A11 antibody. Controls include the incubation of Ab in 5% DMSO (compounds were dissolved in 100% DMSO after which they were diluted 20-fold in the presence of Ab) and Ab with 5% Millipore H2O (SI was dissolved in H2O followed by 20-fold dilution in the presence of Ab).

600 nm, at a gain of 90, an integration time of 100 ms, 50 flashes, and a stepsize of 1 nm. Data were background corrected for 15 mM ThT and accordingly for either 5% DMSO (Ab alone), 5% DMSO plus compounds, 5% H2O, or 5% H2O in the presence of 100 mM SI. 4.3.4. Transmission electron microscopy After 7 days of incubation, Ab aliquots (5 mL) were adsorbed to carbon-coated Formvar 400-mesh copper grids (Agar Scientific, cat. #S162-4) for 1 min. The grids were blotted, washed, and stained

Fig. 15. Binding modes between 7 and the Ab42 monomer. This figure shows the binding pockets IeIII, identified from docking procedures. Compound 7 is indicated with different colors depending on its location at the different sites: yellow, red and blue for sites I, II and III, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

We used the full-length pentameric aggregate Ab42 model developed by Masman et al. as a target system [46]. For all of our compounds Gasteiger charges were assigned and non-polar hydrogen atoms were merged. All torsions of the ligand were allowed to rotate during docking. The grid dimensions were 50  50  50 points along the x-, y- and z-axes, with points separated by 1 Å. The grid was chosen to be sufficiently large to cover the whole system (blind docking method). The value of exhaustiveness of search was 400 and the number of poses collected were 10. All graphic manipulations and visualizations were performed by means of the AutoDock Tools 1.5.4 [62] and ligand docking with Autodock Vina 1.1.1 [63]. A total of ten different binding modes were obtained in every docking calculation, these were ranked according their binding affinity expressed in kcal mol
1. The best binding modes obtained for each compound in the Docking calculations were used as initial complexes of a 10 ns MD simulation. The purpose of this procedure was to generate side chain relaxation and improve the ligandeprotein interactions. The coordinates were saved every 2 ps. MD simulations were performed using the GROMACS 4.5.3 [64] package of programs, with the GROMOS 96 53a6 force field [65]. The Nterminus of each peptide was protonated and the C-terminus, deprotonated, giving zwitterionic conditions to the protofibril. All other tritatable amino acids were assigned their canonical state at physiological pH. The system was placed in a dodecahedric box of simple point charge (SPC) water [66], to which 100 mM was added, including neutralizing counterions. The simulations were run under NPT conditions, using V-rescale coupling algorithm for keeping the temperature constant (T ¼ 310 K, tT ¼ 0.1 ps) and the ParrinelloeRahman barostate [67] to isotropically regulate pressure (P ¼ 1 bar, tP ¼ 2.0 ps). The LINCS algorithm was used to constrain the lengths of hydrogen containing bonds [68,69]; the waters were restrained using the SETTLE algorithm [70]. The time step for the simulations was 0.002 ps and the compressibility 4.5x10-5 bar-1. Van der Waals forces were treated using a 1.4 nm cutoff. Long-range electrostatic forces were treated using the particle mesh Ewald method (PME) [71,72]. For MM-GBSA methodology, snapshots were taken at 10 ps time intervals from the corresponding last 1000 ps MD trajectories, and the explicit water molecules were removed from the snapshots. Once determined the binding site we proceed with longer MD simulations reaching 200 ns overall simulation time. To warrantee a good reproducibility of our simulations and discard the effect of any simulation artifact, three MD simulations of 200 ns were conducted for each system under different starting velocity distribution functions. For the Ab42 monomeric model we used as a starting conformation the PDB-entry 1IYT, which is composed of two helices [helix-1 (residues 8e25) and helix-2 (residues 28e38)] and a turn region (residues 26e27) linking the helices. Three MD simulations of 50 ns were conducted for each system under different starting velocity distribution functions. These simulations were run with the same conditions used for the calculations on the pentameric model. Secondary structure analyses were carried out employing the DSSP method [73]. Post MD analysis was carried out also using the GROMACS 4.5.3 package of programs.

150

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

Fig. 16. Change in the secondary structure during molecular dynamics simulations for the Ab42 monomer in the absence of compound (A), and the complex 7-Ab42 monomer (B).

Fig. 17. Structural snapshots (the final instance of a 50 ns MD simulation) of complex 6-Ab42 monomer (A) and complex 7-Ab42 monomer (B). The Ab42 monomer is shown in cyan ribbon, and compounds 6 and 7 are shown in yellow and purple, respectively. Residues Glu22 and Asp23 are shown in red sticks. Complex 6-Ab42 monomer (A) shows the “toxic turn” at positions 22 and 23 while for complex 7-Ab42 monomer (B) the toxic conformer is not present. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152

Acknowledgments Grants from Universidad Nacional de San Luis (UNSL), partially supported this work. E.E.B.G. thanks a doctoral fellowship of CONICET-Argentina. R.D.E. and S.A.A. are members of the Consejo
cnicas (CONICETNacional de Investigaciones Científicas y Te Argentina) staff. L.M. and C.M.L.D thank CONICET and UNR for support. E.H. is supported by an FWO doctoral fellowship. K.B. thanks FWO Odysseus and the MIRA UTWIST program for financial support. M.F.M. was supported by a Hersenstichting Nederland post-doctoral fellowship.

[26]

[27]

[28]

[29]

Appendix A. Supplementary data [30]

Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.ejmech.2015.03.042.

[31]

References

[32]

[1] World Alzheimer Report 2013, Alzheimer's Disease International, London, 2013. [2] B. Cheng, H. Gong, H. Xiao, R.B. Petersen, L. Zheng, K. Huang, Inhibiting toxic aggregation of amyloidogenic proteins: a therapeutic strategy for protein misfolding diseases, Biochim. Biophys. Acta 1830 (2013) 4860e4871. [3] A.S. Cohen, L.A. Jones, Amyloidosis. Curr. Opin. Rheumatol. 3 (1991) 125e138. [4] C. Soto, M.C. Branes, J. Alvarez, N.C. Inestrosa, Structural determinants of the Alzheimer's amyloid b-peptide, J. Neurochem. 63 (1994) 1191e1198. [5] A.T. Petkova, G. Buntkowsky, F. Dyda, R.D. Leapman, W.M. Yau, R. Tycko, Solid state NMR reveals a pH-dependent antiparallel b-sheet registry in fibrils formed by a b-amyloid peptide, J. Mol. Biol. 335 (2004) 247e260. [6] R. Tycko, Solid-state NMR as a probe of amyloid fibril structure, Curr. Opin. Chem. Biol. 4 (2000) 500e506. [7] R. Tycko, Characterization of amyloid structures at the molecular level by solid state nuclear magnetic resonance spectroscopy, Methods Enzymol. 413 (2006) 103e122. [8] R. Tycko, Molecular structure of amyloid fibrils: Insights from solid-state NMR, Q. Rev. Biophys. 39 (2006) 1e55. [9] I. Kheterpal, M. Chen, K.D. Cook, R. Wetzel, Structural differences in Ab amyloid protofibrils and fibrils mapped by hydrogen exchange-mass spectrometry with on-line proteolytic fragmentation, J. Mol. Biol. 361 (2006) 785e795. [10] I. Kheterpal, R. Wetzel, Hydrogen/deuterium exchange mass spectrometry e a window into amyloid structure, Acc. Chem. Res. 39 (2006) 584e593. [11] T. Luhrs, C. Ritter, M. Adrian, D. Riek-Loher, B. Bohrmann, H. Dobeli, D. Schubert, R. Riek, 3D structure of Alzheimer's amyloid-b(1-42) fibrils, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 17342e17347. [12] A.T. Petkova, W.M. Yau, R. Tycko, Experimental constraints on quaternary structure in Alzheimer's b-amyloid fibrils, Biochemistry 45 (2006) 498e512. [13] J.A. Hardy, G.A. Higgins, Alzheimer's disease: the amyloid cascade hypothesis, Science 256 (1992) 184e185. [14] R. Kayed, E. Head, J.L. Thompson, T.M. McIntire, S.C. Milton, C.W. Cotman, C.G. Glabe, Common structure of soluble amyloid oligomers implies common mechanism of pathogenesis, Science 300 (2003) 486e489. [15] R. Kayed, Y. Sokolov, B. Edmonds, T.M. McIntire, S.C. Milton, J.E. Hall, C.G. Glabe, Permeabilization of lipid bilayers is a common conformationdependent activity of soluble amyloid oligomers in protein misfolding diseases, J. Biol. Chem. 279 (2004) 46363e46366. [16] S.J. Wood, R. Wetzel, J.D. Martin, M.R. Hurle, Prolines and amyloidogenicity in fragments of the alzheimer's peptide b/A4, Biochemistry 34 (1995) 724e730. [17] M.A. Etienne, N.J. Edwin, J.P. Aucoin, P.S. Russo, R.L. McCarley, R.P. Hammer, Beta-amyloid protein aggregation, Methods Mol. Biol. 386 (2007) 203e225. [18] D.B. Teplow, Structural and kinetic features of amyloid b-protein fibrillogenesis, Biochemistry 37 (1998) 121e142. [19] E. Ahmad, A. Ahmad, S. Singh, M. Arshad, A.H. Khan, R.H. Khan, A mechanistic approach for islet amyloid polypeptide aggregation to develop antiamyloidogenic agents for type-2 diabetes, Biochimie 93 (2011) 793e805. [20] E. Gazit, A possible role for pi-stacking in the self-assembly of amyloid fibrils, FASEB J. 16 (2002) 77e83. [21] Ch Wu, J. Scott, J.E. Shea, Binding of congo red to amyloid protofibrils of the Alzheimer Ab9e40 peptide probed by molecular dynamics simulations, Biophys. J. 103 (2012) 550e557. [22] J. Ghanta, C.L. Shen, L.L. Kiessling, R.M.A. Murphy, Strategy for designing inhibitors of b-amyloid toxicity, J. Biol. Chem. 271 (1996) 29525e29528. [23] L.O. Tjernberg, J. Naslund, F. Lindqvist, J. Johansson, A.R. Karlstrom, J. Thyberg, L. Terenius, C. Nordstedt, Arrest of beta-amyloid fibril formation by a pentapeptide ligand, J. Biol. Chem. 271 (1996) 8545e8548. [24] T.L. Lowe, A. Strzelec, L.L. Kiessling, R.M. Murphy, Structure-function relationships for inhibitors of b-amyloid toxicity containing the recognition sequence KLVFF, Biochemistry 40 (2001) 7882e7889. [25] T.J. Gibson, R.M. Murphy, Design of peptidyl compounds that affect b-amyloid

[33]

[34]

[35] [36] [37]

[38] [39]

[40] [41]

[42]

[43]

[44]

[45]

[46]

[47] [48]

[49]

[50] [51] [52]

[53]

151

aggregation: importance of surface tension and context, Biochemistry 44 (2005) 8898e8907. C. Soto, M.S. Kindy, M. Baumann, B. Frangione, Inhibition of Alzheimer's amyloidosis by peptides that prevent beta-sheet conformation, Biochem. Biophys. Res. Commun. 226 (1996) 672e680. C. Soto, E.M. Sigurdsson, L. Morelli, R.A. Kumar, E.M. Castano, B. Frangione, Beta-sheet breaker peptides inhibit fibrillogenesis in a rat brain model of amyloidosis: implications for Alzheimer's therapy, Nat. Med. 4 (1998) 822e826. B. Permanne, C. Adessi, G.P. Saborio, S. Fraga, M.J. Frossard, J. Van Dorpe, I. Dewachter, W.A. Banks, F. Van Leuven, C. Soto, Reduction of amyloid load and cerebral damage in a transgenic mouse model of Alzheimer's disease by treatment with a beta-sheet breaker peptide, FASEB J. 16 (2002) 860e862. C. Adessi, M.J. Frossard, C. Boissard, S. Fraga, S. Bieler, T. Ruckle, F. Vilbois, S.M. Robinson, M. Mutter, W.A. Banks, C. Soto, Pharmacological profiles of peptide drug candidates for the treatment of Alzheimer's disease, J. Biol. Chem. 278 (2003) 13905e13911. I. Autiero, E. Langella, A. Saviano, Insights into the mechanism of interaction between trehalose-conjugated beta-sheet breaker peptides and Ab(1e42) fibrils by molecular dynamics simulations, Mol. Biosyst. 9 (2013) 2835e2841. D. Chen, Z.S. Martin, C. Soto, C.H. Schein, Computational selection of inhibitors of Ab aggregation and neuronal toxicity, Bioorg. Med. Chem. 17 (2009) 5189e5197. D.B. Kitchen, H. Decornez, J.R. Furr, J. Bajorath, Docking and scoring in virtual screening for drug discovery: methods and applications, Nat. Rev. Drug Discov. 3 (2004) 935e949. A.R. Leach, B.K. Schoichet, C.E. Peishoff, Prediction of protein-ligand interactions. Docking and scoring: successes and gaps, J. Med. Chem. 49 (2006) 5851e5855. N. Moitessier, P. Englebienne, D. Lee, J. Lawandi, C.R. Corbeil, Towards the development of universal, fast and highly accurate docking/scoring methods: a long way to go, Br. J. Pharmacol. 153 (2008) S7eS26. H. Zhao, A. Caflisch, Molecular dynamics in drug design, Eur. J. Med. Chem. 91 (2015) 4e14. R. Roychaudhuri, M. Yang, M.M. Hoshi, D.B. Teplow, Amyloid b-protein assembly and Alzheimer disease, J. Biol. Chem. 284 (2009) 4749e4753. F.-F. Liu, X.-Y. Dong, L. He, A.P.J. Middelberg, Y. Sun, Molecular insight into conformational transition of amyloid b-peptide 42 inhibited by (e)-epigallocatechin-3-gallate probed by molecular simulations, J. Phys. Chem. B 115 (2011) 11879e11887. E.P. Raman, T. Takeda, D.K. Klimov, Molecular dynamics simulations of ibuprofen binding to Ab peptides, Biophys. J. 97 (2009) 2070e2079. T. Takeda, R. Kumar, E.P. Raman, D.K. Klimov, Nonsteroidal anti-inflammatory drug naproxen destabilizes Ab amyloid fibrils: a molecular dynamics investigation, J. Phys. Chem. B 114 (2010) 15394e15402. T. Takeda, W.E. Chang, E.P. Raman, D.K. Klimov, Binding of nonsteroidal antiinflammatory drugs to Ab fibril, Proteins 78 (2010) 2849e2860. J.A. Lemkul, D.R. Bevan, Destabilizing Alzheimer's Ab42 protofibrils with morin: mechanistic insights from molecular dynamics simulations, Biochemistry 49 (2010) 3935e3946. J.A. Lemkul, D.R. Bevan, Morin inhibits the early stages of amyloid b-peptide aggregation by altering tertiary and quaternary interactions to produce “offpathway” structures, Biochemistry 51 (2012) 5990e6009. F. Attanasio, M. Convertino, A. Magno, A. Caflisch, A. Corazza, H. Haridas, G. Esposito, S. Cataldo, B. Pignataro, D. Milardi, E. Rizzarelli, ChemBioChem 14 (2013) 583e592. N.J. Bruce, D. Chen, S.G. Dastidar, G.E. Marks, C.H. Schein, R.A. Bryce, Molecular dynamics simulations of Ab fibril interactions with b-sheet breaker peptides, Peptides 31 (2010) 2100e2108. J.A. Lemkul, D.R. Bevan, The role of molecular simulations in the development of inhibitors of amyloid b-peptide aggregation for the treatment of Alzheimer's disease, ACS Chem. Neurosci. 3 (2012) 845e856. M.F. Masman, U.L.M. Eisel, I.G. Csizmadia, B. Penke, R.D. Enriz, S.J. Marrink, P.G.M. Luiten, In silico study of full-length amyloid b 1-42 tri- and pentaoligomers in solution, J. Phys. Chem. B 113 (2009) 11710e11719. C. Hetenyi, D. van der Spoel, Blind docking of drug-sized compounds to proteins with up to a thousand residues, FEBS Lett. 580 (2006) 1447e1450. S.A. Andujar, F. Lugli, S. Hofinger, R.D. Enriz, F. Zerbetto, Amyloid-b fibril disruption by C60-molecular guidance for rational drug design, Phys. Chem. Chem. Phys. 14 (2012) 8599e8607. N. Ferguson, J. Becker, H. Tidow, S. Tremmel, T.D. Sharpe, G. Krause, J. Flinders, M. Petrovich, J. Berriman, H. Oschkinat, A.R. Fersht, General structural motifs of amyloid protofilaments, Proc. Natl. Acad. Sci. U. S. A. 103 (2006) 16248e16253. J.A. Lemkul, D.R. Bevan, Assessing the stability of Alzheimer's amyloid protofibrils using molecular dynamics, J. Phys. Chem. B 114 (2010) 1652e1660. H. Le Vine, Quantification of beta-sheet amyloid fibril structures with thioflavin T, Methods Enzymol. 309 (1999) 274e284. K. Broersern, W. Jonckheere, J. Rozenski, A. Vandersteen, K. Pauwels, A. Pastore, F. Rousseau, J. Schymkowitz, A standardized and biocompatible preparation of aggregate-free amyloid beta peptide for biophysical and biological studies of Alzheimer's disease, Protein Eng. Des. Sel. 24 (2011) 743e750. €rner, T. Schrader, C. Wang, G. Bitan, ComS. Sinha, Z. Du, P. Maiti, F.G. Kla parison of three amyloid assembly inhibitors: the sugar scyllo-inositol, the

152

[54] [55]

[56]

[57]

[58]

[59]

[60]

[61]

[62]

E.E. Barrera Guisasola et al. / European Journal of Medicinal Chemistry 95 (2015) 136e152 polyphenol epigallocatechin gallate, and the molecular tweezer CLR01, ACS Chem. Neurosci. 3 (2012) 451e458. M. Biancalana, S. Koide, Molecular mechanism of Thioflavin-T binding to amyloid fibrils, Biochim. Biophys. Acta 1804 (2010) 1405e1412. G.W. Anderson, J.E. Zimmerman, F.M. Callahan, The use of esters of Nhydroxysuccinimide in peptide synthesis, J. Am. Chem. Soc. 86 (1964) 1839e1842. N. Cerda-Costa, A. Esteras-Chopo, F.X. Aviles, L. Serrano, V. Villegas, Early kinetics of amyloid fibril formation reveals conformational reorganisation of initial aggregates, J. Mol. Biol. 366 (2007) 1351e1363. Y. Masuda, S. Uemura, R. Ohashi, A. Nakanishi, K. Takegoshi, T. Shimizu, T. Shirasawa, K. Irie, Identification of physiological and toxic conformations in Ab42 aggregates, ChemBioChem 10 (2009) 287e295. N. Izuo, T. Kume, M. Sato, K. Murakami, K. Irie, Y. Izumi, A. Akaike, Toxicity in rat primary neurons through the cellular oxidative stress induced by the turn formation at positions 22 and 23 of Ab42, ACS Chem. Neurosci. 3 (2012) 674e681. N. Izuo, K. Murakami, M. Sato, M. Iwasaki, Y. Izumi, T. Shimizu, A. Akaike, K. Irie, T. Kume, Non-toxic conformer of amyloid b may suppress amyloid binduced toxicity in rat primary neurons: Implications for a novel therapeutic strategy for Alzheimer's disease, Biochem. Biophys. Res. Commun. 438 (2013) 1e5. M. Necula, R. Kayed, S. Milton, C.G. Glabe, Small molecule inhibitors of aggregation indicate that amyloid-b oligomerization and fibrillization pathways are independent and distinct, J. Biol. Chem. 282 (2007) 10311e10324. R. Kayed, E. Head, J.L. Thompson, T.M. McIntire, S.C. Milton, C.W. Cotman, C.G. Glabe, Common structure of soluble amyloid oligomers implies common mechanism of pathogenesis, Science 300 (2003) 486e489. M.F. Sanner, Python: a programming language for software integration and

development, J. Mol. Graph. Model. 17 (1999) 57e61. [63] O. Trot, A.J. Olson, AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading, J. Comput. Chem. 31 (2010) 455e461. [64] B. Hess, C. Kutzner, D. van der Spoel, E. Lindahl, GROMACS 4: Algorithms for highly efficient, load-balanced, and scalable molecular simulation, J. Chem. Theory Comput. 4 (2008) 435e447. [65] C. Oostenbrink, A. Villa, A.E. Mark, W.F. Van Gunsteren, A biomolecular force field based on the free enthalpy of hydration and solvation: the GROMOS force-field parameter sets 53A5 and 53A6, J. Comput. Chem. 25 (2004) 1656e1676. [66] H.J.C. Berendsen, J.P.M. Postma, Interaction models for water in relation to protein hydration, Jerusalem Symp. Quantum Chem. Biochem. 14 (1981) 331e338. [67] M. Parrinello, A. Rahman, Polymorphic transitions in single crystals: a new molecular dynamics method, J. Appl. Phys. 52 (1981) 7182e7190. [68] B. Hess, LINCS: a linear constraint solver for molecular simulations, J. Comput. Chem. 18 (1997) 1463e1472. [69] B. Hess, P-LINCS: a parallel linear constraint solver for molecular simulation, J. Chem. Theory Comput. 4 (2008) 116e122. [70] S. Miyamoto, P.A. Kollman, Settle: an analytical version of the SHAKE and RATTLE algorithm for rigid water models, J. Comp. Chem. 13 (1992) 952e962. [71] T. Darden, Particle mesh Ewald: an N$log(N) method for Ewald sums in large systems, J. Chem. Phys. 98 (1993) 10089e10092. [72] U. Essmann, A smooth particle mesh Ewald method, J. Chem. Phys. 103 (1995) 8577e8593. [73] W. Kabsch, C. Sander, Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features, Biopolymers 22 (1983) 2577e2637.