Insights into the mechanism of interaction between ...

9 downloads 0 Views 120KB Size Report
F. M. Masman, U. I. L. M. Eisel, I. G. Csizmadia, B. Penke, R. D. Enriz, S. J. .... X. Dura, K. Gademann, J. Bernhard, D. Seebach, F. Wilfred van Gunsteren and E.
Insights into the mechanism of interaction between trehalose-conjugated Beta-sheet breaker peptides and Aβ(1-42) fibrils by molecular dynamics simulations. Ida Autiero,a Emma Langellaa* and Michele Savianob* *Correspondence to: Michele Saviano, National Research Council, Institute of Crystallography, 70126 Bari, Italy. E-mail: [email protected]. Emma Langella, National Research Council, Institute of Biostructures and Bioimaging, 80138 Naples, Italy. E-mail: [email protected] a

National Research Council, Institute of Biostructures and Bioimaging, 80138 Naples, Italy

b

National Research Council, Institute of Crystallography, 70126 Bari, Italy

ABSTRACT An attractive strategy to contrast the Alzheimer disease (AD) is represented by the development of β-sheet breaker peptides (BSB). β-sheet breakers constitute a class of compounds which have shown a good efficacy in preventing the Aβ fibrillogenesis, however their mechanism of action has not been precisely understood. In this context, we have studied the structural basis underlying the inhibitory effect of Aβ(1-42) fibrillogenesis explicated by two promising trehalose-conjugated BSB peptides using an all-atom molecular dynamics (MD) approach. Our simulations suggest that the binding on the two protofibril ends occurs through different binding modes. In particular, binding on the odd edge (chain A) is guided by a well defined hydrophobic cleft, which is common to both ligands. Moreover, targeting chain A entails a significant structure destabilization leading to a partial loss of β structure and is an energetically favoured process. A significant contribution of the trehalose moiety to complexes stabilities emerged from our results. The energetically favoured hydrophobic cleft detected on chain A could represent a good starting point for the design of new molecules with improved anti-aggregating features.

INTRODUCTION Insoluble brain plaques characteristic of Alzheimer’s disease (AD) are prominently formed by the aggregation of amyloid peptides (Aβ1-40, Aβ1-42) into ordered fibrils (1-5). AD is a devastating form of dementia, affecting the global cognitive abilities, and the development of anti amyloidplaques remains the main rational strategy to prevent or treat this disorder. Therefore, the structural properties of the amyloid peptides aggregation have been widely studied to investigate at atomic

1

level the deposition of these Aβ-fibrils, and also to support the design of novel drugs against the amyloid plaques formation (6-17). β-sheet breakers (BSB) represent a key for new therapeutic strategies, aimed at contrasting the neuronal dysfunction associated with the deposition of insoluble protein fibrils and Alzheimer's disease (18). Although various experimental data have supported the hypothesis that such ligands act by inhibiting the accumulation of amyloid Aβ peptides or preventing the formation of soluble Aβ oligomers, their mechanism of action has not been yet precisely understood. The Hydrophobic Core Region (HCR) of Aβ1-42, residues 17-20, is the region mainly targeted by BSB ligands, because this contains critical elements for Aβ self-assembly, as experimentally demonstrated by the blocking of aggregation induced by mutating V18, F19 and F20 (19,20). Much efforts have been spent on the development of BSB based on a peptide scaffold homologous to the HCR domain, thus driving the binding through a self recognition mechanism as well as mimicking the growth process of the fibrils, which occurs through the addition of new layers at the end of the oligomer (21-25). Previous studies, on the interaction of KLVFF peptide and related analogues with Aβ protofilaments, have demonstrated that the backbone does’nt significantly contribute to the binding affinity, whereas the side chains are believed to importantly strengthen the interaction with Aβ peptides. The BSB affinity for Aβ oligomers is significantly improved by appending polar groups to the peptide scaffold, and it has been postulated that these groups could enhance the binding, involving also adjacent HCR residues (26-27). In this context, threalose-conjugated peptides (Ac-LPFFD-Th (ThCT) and Th-Succinyl-LPFFDNH2 (ThNT) (see Figure 1) were proposed as effective Aβ inhibitors (28). They were designed to combine a polar group with the peptide portion LPFFD, the well known BSB Soto’s peptide (29), thought to be able to recognize, in a self-complementary manner, the 17-21 Aβ region, by binding to the ends of the growing fibril. As already discussed in our previous study (14), ThCT is able to produce an interesting effect of destabilization targeting the odd end of the Aβ17-42 protofibril. There is a general consensus (6-7) in assessing that the Aβ protofibril is formed by a longitudinal stack of peptide chains, creating two parallel in-register β-sheets. Each peptide chain is arranged in a β-strand-turn-β-strand motif, thereby adopting an U-shaped conformation (Figure 1). The two protofibril ends, which were postulated to be not equivalent by previous investigations (7, 17), are expected to have a different structural and dynamic behavior. Thus, in order to investigate those differences, here we studied the effect of either ThCT or ThNT, interacting with both the ends of the Aβ17-42 protofibril. Moreover,

2

these ligands, having a trehalose moiety anchored to the LPFFD fragment in opposite orientations, allowed us also to deepen the role played by the sugar moiety and the influence of its position relative to the peptide scaffold in increasing the anti-aggregating potential of these new ligands. Trehalose conjugates are among the most suitable ligands to investigate the improper effect of polar groups added to BSB peptide scaffolds, therefore we have performed a detailed analysis by molecular dynamics of their effects on the structural and conformational properties of Aβ(1-42) oligomers, on their influence on the overall complex stability and on the ability to inhibit fibril growth. Moreover, our findings could be useful for the rational design of more effective BSB peptides with improved features.

RESULTS The binding modes of two trehalose-conjugated Aβ inhibitors, ThCT and ThNT, (28) (Figure 1) have been investigated through MD simulations. To this aim, the pentameric NMR structure of Aβ(17-42) (7) (Figure 1) has been used as protofibril model, and the ligands binding to both fibril ends (chain A and chain E) has been evaluated. As a consequence, four different systems have been simulated CTa, CTe, NTa and NTe where the subscripts a and e indicate ligand binding to chain A and chain E, respectively. Indeed, in the case of CTa, data already included in our previous study have been used for further analysis (14). Ligands binding modes In Figure 2, a graphical map, which represents the protofibril regions involved into ligand binding during the simulation time, is reported. Ligands are divided into peptide fragment and sugar moiety. The analysis of the results underlines that, upon binding to chain A, the peptide fragment interacts with a wide region of B1 strand (17-21) and a few residues of B2 strand (34, 36-37) both in CTa and NTa. Therefore, there is a strong similarity in the binding of the peptidic region of the two ligands. More in detail, the inspection of Figures 3-4 reveals that both ligands are accommodated in a very similar hydrophobic cleft, comprising residues V18, F19, F20, A21, L34 and V36 (V36 chain B). So, the peptidic portion is mainly stabilized by vdW contacts with the side chains of the residues belonging to the cleft. For what concerns the sugar moiety, instead, in the two ligands it interacts with different regions of the Aβ protofibril. In detail, in CTa it binds residues 21-23 and 28, close to the U-turn region, whereas in NTa it is involved into stabilizing interactions with N-terminal residues 17-18. The sugar moiety contributes significantly to the energetic stabilization of the

3

complexes, as demonstrated by the huge number of hydrogen bonds that it makes (Table 1) in CTa as well as in NTa. It should be noted that the ThNT sugar portion interacts with L17 residues belonging to chains B and C other than to chain A since it is pending out from the interacting chain, as shown in Figure 3. On the other hand, on chain E, the binding of the peptide portions of the two ligands differs, in contrast to what happens on chain A (Figure 3), and is mainly stabilized by hydrogen bonds, as reported in Table 2. More in detail, in CTe the peptide portion interacts with a wide region of both B1 (residues 17-22) and B2 (residues 34-35, 37-41) strands, whereas in NTe it binds just B1 strand (residues 17-22) (Figure 2). Regarding sugar portion, in CTe it is involved into stabilizing interactions with the U-turn region, as in CTa, while in NTe it makes additional interactions with C-terminal residues other then with Nterminal residues, as in NTa (Figure 2). Similarly to binding on chain A, the sugar moiety forms many hydrogen bonds (Table 2), which stabilize the overall binding to the protofibril. Our analysis suggests that each ligand interacts in a different way with the two protofibril extreme chains. Thus, it emerges that the two protofibril ends (chain A and chain E) are not equivalent. This finding is supported by the different binding free energies computed for all the complexes. Indeed, MMPBSA calculations were performed in order to obtain a theoretical estimation of the binding free energy. These calculations reveal that, in all the cases, the binding is a favorite process (negative binding free energy values). In particular in the case of binding to chain A, energy values of -16.6 kcal mol-1(14) and -18.1 kcal mol-1 were calculated for ThCT and ThNT respectively, whereas -4.7 kcal mol-1 and -5.7 kcal mol-1 were estimated for CTe and NTe systems, respectively. Ligands effects on Aβ protofibril structure and stability In order to evaluate the effect of the trehalose-derivatives on the overall structure and stability of the Aβ protofibril, a comparative analysis with the Aβ system alone, has been carried out including the results of our previous MD study (14). Analysis of RMSD indicates that ligand binding to chain A leads to a significant destabilization of protofibril structure, shown by the higher RMSD values for both ThCT and ThNT (Figure 5). Indeed,

per-chain analysis reveals that the RMSD increase regards all the chains, i. e.

destabilization affects the whole system (Supplementry Figure S1). This is not the case for ligand binding to chain E, where high RMSD values are shown just by the interacting chain and/or by the extreme chains, which are intrinsically less stable, as already shown by Aβ system alone.

4

Protofibril destabilization on ligands binding to chain A rises also from the analysis of the secondary structure elements. In detail, in CTa and NTa there is a decrease of β-sheet content and a concomitant increase of coil structure with respect to Aβ alone, CTe and NTe. (Figure 6). This effect is evident not only on the chain directly interacting with the ligand, but is propagated through the entire system thus affecting all the chains in a comparable way (see Supplementary Figure S2). Targeting chain A significantly destabilizes protofibril packing, since upon ligand binding, the two beta strands (B1 and B2) move slightly apart from each other, causing the opening of the β-hairpinlike structure (Figure 3 C, D) of the first and, to a lesser extent, of the second chain of the protofibril finally leading to the formation of a hydrophobic cleft (described before), which accommodates ligands. The latter conformational rearrangement involves mainly the two strands B1 and B2, whereas the U-turn region does not undergo significant modifications thanks to the presence of the D23-K28 salt bridge, which is essentially retained in CTa and NTa as well as in CTe and NTe, during the whole simulation time (Supplementary Figure S3).

Table . Hydrogen bonds which stabilize CTa and NTa complexes. Percentage of existence, computed over the whole simulation time, has been indicated. ligand system portion atom peptide

CTa

sugar

peptide

sugar

NTa

*

Chain B

5

Aβ residue atom H-bond %

OAC NBW

F19 A21

N O

59 57

OBR N

D23

N

48

A42

O2

18

O4

E22

OE1

35

O2

D23

N

35

O4

A21

O

31

OCP

D23

OD1

18

O6

D23

N

16

OCQ

K28

NZ

11

OCG O

A21 L17

N N

84 55

OBC

L17

N

8

O5

L17

N

20 12

O1

L17

N

OAS

L17

N

11

OAU

A42**

Oterm

35

OAS

A42**

O2

8

O6

V18

N

7

OAT

A42*

O2

6

**

Chain C

Table Hydrogen bonds which stabilize CTe and NTe complexes. Percentage of existence, computed over the whole simulation time, has been indicated. ligand system portion atom

peptide

sugar

CTe

peptide

sugar

NTe

Aβ residue atom H-bond %

OBV

E22

N

58

NBO

F20

O

31

O

V39

N

30

OBV

M35

N

27

NAS

V39

O

13

OBR

E22

N

13

OCR

E22

N

83

OCR

D23

OD2

72

O6

V24

N

55

O3

E22

OE1

33

O6

E22

O

33

O2

D23

OD2

31

O6

D23

OD1

28

O4

E22

OE1

16

O3

V24

O

14

O2

E22

O

12

OCR

F20

N

53

NCW

D23

OD2

34

OCU

E22

N

27

NCW

E22

OE2

13

O6

L17

O

71

OAS

L17

N

63

O2

G37

O

55

O1

L17

N

37

O4

M35

O

30

O3

G37

O

15

OAS

I41

N

14

O4

V39

N

14

OAT

I41

N

14

DISCUSSION AND CONCLUSIONS The structural basis underlying the inhibitory effect of Aβ(17-42) fibrillogenesis explicated by ThCT and ThNT (28), have been studied using a molecular dynamics (MD) simulation approach. In particular, we have investigated the role played by the hydrophilic threalose moiety, the influence of its position relative to the LPFFD peptide scaffold, the possibility that both ligands could efficiently

6

target two binding sites on Aβ oligomers (chain A and chain E, respectively), and the effects of their interaction on the overall oligomer asset. Our simulations suggest that all the complexes are stable. In fact despite the ligands move from their starting positions, they never dissociate from the Aβ(17-42) protofibril and the binding on the two sites on chain A and chain E occurs through different binding modes. In particular, binding on chain A is guided by a well defined hydrophobic cleft, which is common to both ligands and is made of residues V18, F19, F20, A21, L34 and V36. This binding site seems to be favourite from an energetic point of view with respect to binding on chain E for ThCT as well as for ThNT. Moreover, targeting chain A entails a significant structural destabilization leading to a partial loss of βstructure, as already suggested in our previous work (28). The non equivalence of the fibril edges, emerging from our results, is consistent with data reported in early structural studies on Aβ fibrils (7) and with more recent computational studies, indicating that one of the two fibril ends is mainly involved in the fibril growth process, i.e. the odd end or concave face, corresponding to chain A (17,30). Hence, it is of particular interest to find out ligands which preferentially bind to chain A, since they can affect in a significant way fibril growth as in the case of our ligands. It is worth of note that a preferential binding to the odd end of A β42 protofibril has been recently observed also for other anti-aggregation inhibitors, through both explicit and implicit solvent MD simulations, thus further supporting our findings (31-33). As a matter of fact, trehalose solutions have been experimentally shown to exert an inhibitory effect on Aβ fibrillation (34-35). Moreover, computational studies have indicated that a high sugar concentration is needed for Aβ fibril destabilization, hypothesizing that trehalose molecules act though an indirect water-mediated mechanism (36). Interestingly, our simulations indicate a different behavior of the trehalose moiety linked to the LPFFD peptide portion. More in detail, the trehalose moiety is able to directly interacts with Aβ residues adjacent to both sides of the HCR. In particular, residues from the amino and carboxyl terminal regions are involved in ThNT-Aβ complexes, whereas residues from the U-turn of the interacting layer establish several interaction with ThCT. So, the sugar moiety significantly contributes to stabilize the complexes on the HCR through the formation of a huge number of hydrogen bonds. Therefore, we can infer that Thconjugates can represent good candidates for the rational design of new molecules with improved anti-aggregating features. Indeed, targeting Aβ HCR still represents a valuable strategy to interfere with fibril elongation, in agreement with recent works postulating that N-terminal region comprising HCR plays a key role into Aβ fibril elongation/dissociation (30, 37).

7

In summary, our analysis suggests that Th-conjugates are able to i) destabilize protofibril structure upon binding to the odd end (chain A), ii) efficiently interact with both Aβ protofibril ends, hindering the addition of successive peptides, and thus preventing fibril growth. Moreover, we have found an energetically favoured hydrophobic cleft on chain A, which could represent a good starting point for the design of more effective inhibitors, having a strong destabilizing effect on the protofibril asset.

8

METHODS System preparation The pentameric Aβ fibril structure obtained by NMR spectroscopic data by Lührs et al (pdb entry 2BEG) (7) was chosen as the target structure for the ligands under investigation. ThCT and ThNT ligands (28) were built using the Insight Builder module (Insight2000, Accelrys, San Diego) with covalent valence force field (CVFF), then minimized using a conjugate gradient algorithm (38). The conformational space of all structures was sampled through the Simulated Annealing protocol of the Insight Discover module and considering a distance-dependent dielectrics of 80, in order to mimic the aqueous environment. Ligands were firstly minimized for 100 iterations using steep descents algorithm and then with additional 5000 steps using the VA09A algorithm until the maximum derivative was less than 0.001 kcal/A. Structures were thus heated in 1000 fs up to 1000 K for 1000 fs and 200 structures were chosen on the basis of the energy and geometrical criteria. Subsequently, ThCT and ThNT best conformations were docked to both terminal chains of the Aβ pentamer through a manual positioning to allow the self-complementary recognition between the peptide portions of both ligands and the homologous residues 17-21 on chain A and chain E, using the Insight2000 Building module. In order to parametrize ligands, the PRODRG server (http://davapc1.bioch.dundee.ac.uk/prodrg/) was queried to generate automated topologies for use with the GROMOS force field, whereas partial charges were adjusted taking into account, for the sugar portion, those reported in reference (39) and, for the peptide portion, those available in the GROMOS43a1 library. Molecular Dynamics simulation Molecular dynamics (MD) simulations were performed and analyzed using the GROMACS package program and GROMOS43a1 force field (40). The Aβ oligomer and each Aβ-ligand complex were solvated in a cubic box of SPC water with at least 11 Å distance to the border adding counterions to neutralize the system. Periodic boundary conditions were employed and the LINCS algorithm (41) was used to constrain all bond lengths. Simulations were ran under NPT conditions, using Berendsen's coupling algorithm (42) to retain the temperature and pressure constant (300 K with a P = 1 bar). Electrostatic interactions were handled by means of the particle mesh Ewald method (PME) (43-44) whereas for the Lennard-Jones potential a nonbonded cutoff of 1,4 nm was used. For all systems, the solvent was relaxed by energy minimization followed by 10 ps of MD at 300 K, while restraining protein atomic positions with a harmonic potential. The system was then gradually heated up to 300 K in a five step process, from 50 to 300 K, and then simulated under standard NPT conditions for 100 ns without restraints.

9

Finally, the MD ensembles have been clustered, using the algorithm described by Dura et al. (45) with a cut-off of 0.2 nm. Thus, from the most populated clusters we extracted the representative models, which were have been all subjected to the MMPBSA (molecular mechanics (MM) Poisson Boltzmann/Generalized Born surface area MM-GBSA) analysis to obtain the relative binding free energy, using MMPBSA.py program of the AmberTool suite. (46-47).

ACKNOWLEDGEMENTS The authors would like to give their special thanks to Dr. Giuseppe Pappalardo for his help, support, interest and valuable hints. The technical assistance of Mr. Luca de Luca, Dr. Caterina Chiarella and Mr. Giovanni Filograsso are gratefully acknowledged. This work was supported by the M.I.U.R, Ministero dell'Istruzione, dell'Università e della Ricerca (FIRB-MERIT RBNE08HWLZ_002).

10

FIGURE CAPTIONS

Figure 1: Upper panel: 2D structure of the ThCT and ThNT ligands, respectively. Lower panel: Top-view of the Aβ(17-42) pentamer (16) (left), secondary structure elements (B1, B2) and connecting regions (U-turn) are indicated with the relative labels. Different colors in stick are used to represent residues belonging to the different secondary structure regions. Side-view of the pentameric Aβ oligomer colored by chain (right). Figure 2: Graphical map of the contacts between Aβ oligomers and each ligand, representing the region of the oligomer involved into ligand binding during the simulation time, colored by the average of the distance. The secondary structure regions of the oligomer are indicated using labels, and the ligands are divided into peptide- and sugar-fragment. Figure 3: Upper panel: side view of the superimposition between the (A): CTa- (blue) and NTa(red) systems and (B): CTe- (blue) and NTe- (red) systems. The secondary structure of Aβ oligomer are displayed as cartoon and each ligand as sticks. Lower panel: top view of the superimposition between the (C): CTa- (blue) and NTa-(red) systems and (D): CTe- (blue) and NTe- (red) systems. Aβ oligomer is represented in trace and each ligand as sticks. Figure 4: Hydrophobic cleft which accommodates ligands targeting chain A (top view of the superimposition between CTa (blue) and NTa (red) systems). Residues belonging to the cleft are labeled and represented as dashed surfaces. Aβ oligomer is represented in trace and each ligand as sticks. Figure 5: Root mean square deviation (RMSD) of the Cα carbon atoms positions as a function of time. Upper panel: Aβ oligomer alone (black line), ThCT ligand targeting chain A (red line), and chain E (green line). Lower panel: Aβ oligomer alone (black line), ThNT ligand targeting chain A (red line) and chain E (green line). Figure 6: Number of residue in β-sheet and coil secondary structure as function of time. Upper panel: Aβ oligomer alone (black line) ThCT ligand targeting chain A (red line) and chain E (green line). Lower panel: Aβ oligomer alone (black line) ThNT ligand targeting chain A (red line) and chain E (green line). The labels indicate the secondary structures.

11

REFERENCES

1. P.E. Fraser, L. Lévesque and D.R. McLachlan, Clin Biochem, 1993, 26, 339-49 2. W. Annaert, B. De Strooper, Annu. Rev. Cell Dev. Biol., 2002, 18, 25-51 3. M. Hutton, J. Perez-Tur and J. Hardy, Essays Biochem., 1998, 33, 117-131 4. D. J. Selkoe, Nat. Cell Biol., 2004, 6, 1054-1061 5. J. P. Taylor, J. Hardy nad K. H. Fischbeck, Science, 2002, 296, 1991-1995 6. A. T. Petkova, W. M. Yau and R. Tycko, Biochemistry, 2006, 45, 498-512 7. T. Luhrs, C. Ritter, M. Adrian, D. Riek-Loher, B. Bohrmann, H. Doeli, D.Schubert and R. Riek, Proc. Natl. Acad. Sci. U.S.A, 2005, 102, 17342-17347 8. A. H. C. Horn and H. Sticht, J. Phys. Chem. B, 2010, 1141, 2219-222 9. A. J. Lemkul and D. R. Bevan, J. Phys. Chem.B, 2010, 114, 1652-1660 10. B. Ma and R. Nussinov, Proc. Natl. Acad. Sci. U. S. A., 2002, 99, 14126-14131 11. K. Hochdo¨rffer, J. Ma¨rz-Berberich, L. Nagel-Steger, M. Epple, W. Meyer-Zaika, A. H. Horn, H. Sticht, S. Sinha, G. Bitan and T. Schrader, J. Am. Chem. Soc., 2011, 133, 4348-4358 12. P. Rzepecki and T. Schrader, J. Am. Chem. Soc., 2005, 127, 3016-3025 13. F. M. Masman, U. I. L. M. Eisel, I. G. Csizmadia, B. Penke, R. D. Enriz, S. J. Marrink and P. G. M. Luiten, J. Phys. Chem., 2009, 113, 11710-11719. 14. I. Autiero, M. Saviano and E. Langella, Mol.. BioSyst., 2013, DOI: 10.1039/C3MB70086K 15. J. Zheng, H. Jang, B. Ma, C. J. Tsai and R. Nussinov, Biophys. J., 2007, 93, 3046-57 16. R. Friedman, R. Pellarin and A. Caflisch, J. Mol. Biol., 2009, 407-415 17. T. Takeda and D.K. Klimov, Biophys J. 2009, 96, 442-52 18. S. Bieler and C. Soto, Curr Drug Targets, 2004, 5, 553-8 19. C. Hilbich, B. Kisters-Woike, J. Reed, C.L. Masters, K. Beyreuther. J Mol Biol., 1992, 20, 228, 460-73

12

20. W.P. Elser, E.R. Stimson, J.R. Ghilardi, H.V. Vinters, J.P. Lee, P.W. Mantyh and J.E. Maggio, Biochemistry, 1996, 35, 749-757 21. O. Tjernberg, C. Lilliehççk, D. J. E. Callaway, J. Näslund, S. Hahne, J. Thyberg, L. Terenius and C. Nordstedt, J. Biol. Chem., 1997, 272, 12601-12605 22. K. Watanabe, K. Nakamura, S. Akikusa, T. Okada, M. Kodaka, T. Konaka- hara, H. Okuno, Biochem. Biophys. Res. Commun., 2002, 290, 121-124 23. T. L. Lowe, A. Strzelec, L. L. Kiessling and R. M. Murphy, Biochemistry, 2001, 40, 7882-7889 24. G. Zhang, M. J. Leibowitz, P. J. Sinko and S. Stein, Bioconjugate Chem., 2003, 14, 86-92 25. M. A. Chacón, M. I. Barría, C. Soto and N. C. Inestrosa, Mol. Psychiatry, 2004, 9, 953-96 26. C.W. Cairo, A. Strzelec, R.M. Murphy and Kiessling, Biochemistry, 2002, 9, 41, 8620-9 27. A. Ashley A. Reinke and J.E. Gestwicki, Chemical Biology & Drug Design, 2011, 77, 6, 399411 28. P. De Bona, M.L. Giuffrida, F. Caraci, A. Copani, B. Pignataro, F. Attanasio, S. Cataldo, G. Pappalardo and E. Rizzarelli, J. Pept Scie., 2009 , 15, 220-228 29 C. Soto, E. M. Sigurdsson, L. Morelli, R.A. Kumar, E.M. Castaño and B. Frangione, Nat. Med. 1998, 4, 822-826. 30. T. Takeda and D.K. Klimov, Biophys J. 2009, 6, 4428-37 31. C. Wu, J. Scott and J.E. Shea, Biophys. J., 2012, 103, 550-557 32. E.P. Raman, T. Takeda and D.K. Klimov, Biophys. J., 2009, 97, 2070-2079 33. T. Takeda, W.E. Chang, E.P. Raman, and D.K. Klimov, Proteins, 2010, 78, 2849-2860 34. M. Nitz, D. Fenili, A.A. Darabie, L. Wu, J.E. Cousins and J. McLaurin, FEBS J., 2008, 275, 1663-1674 35. R. Liu, H. Barkhordarian, S. Emadi, C.B. Park, M.R. Sierks. Neurobiol. Dis., 2005, 20, 74-81 36 F.F. Liu, L. Ji, X.Y. Dong and Y. Sun, J. Phys. Chem. B 2009, 113, 11320-11329

13

37. N. V. Buchete and G. Hummer, Biophys. J., 2007, 92, 3032-3039 38 Fletcher, R. Unconstrained Optimization. In Pratical Methods of Optimization; John Wiley & Sons: New York, 1980; Vol. 1 39. E. Fadda and R.J. Woods, Drug Discovery today, 2010, 15, 596-609 40. van Gunsteren, W. F.; Billeter, S. R.; Eising, A. A.; Hunenberger, P. H.; Kruger, P.; Mark, A. E.; Scott, W. R. P.; Tironi, I. G. Biomolecular Simulation: The GROMOS96 Manual and User Guide; Vdf Hochschulverlag AG an der ETH Zurich: Zurich, Switzerland, 1996 41. Hess B, Bekker H, Berendsen HJC, Fraaije JGEM. LINCS: a linear constraint solver for molecular simulations. J Comp Chem 1997;18:1463–1472 42. H. J. C. Berendsen, J. P. M. Postma, W. F. Van Gunsteren, A. DiNola, and J. R. Haak, J.Chem. Phy, 1984, 81,3648-3690 43. Darden T, York D, Pedersen L. Particle mesh Ewald: an Nϫlog(N) method for Ewald sums in large systems. J Chem Phys 1993;98: 10089 –10092. 44. Darden T, Perera L, Li L, Pedersen L. New tricks for modelers from the crystallography toolkit: the particle mesh Ewald algorithm and its use in nucleic acid simulations. Structure 1999;7: R55– R60 45. X. Dura, K. Gademann, J. Bernhard, D. Seebach, F. Wilfred van Gunsteren and E. Mark, Angev.Chem.Int.Ed., 1999, 38, 236-240. 46. B.R. Miller, T.D. McGee, Jr. J.M. Swails, Homeyer,H. Gohlke and A.E. Roitberg, J. Chem. Theory Comput., 2012, 8, 3314-3321; 47. D.A. Case, T.A. Darden, T.E. Cheatham, III, C.L. Simmerling, J. Wang, R.E. Duke, R. Luo, R.C. Walker, W. Zhang, K.M. Merz, B. Roberts, S. Hayik, A. Roitberg, G. Seabra, J. Swails, A.W. Goetz, I. Kolossváry, K.F. Wong, F. Paesani, J. Vanicek, R.M. Wolf, J. Liu, X. Wu, S.R. Brozell, T. Steinbrecher, H. Gohlke, Q. Cai, X. Ye, J. Wang, M.-J. Hsieh, G. Cui, D.R. Roe, D.H. Mathews, M.G. Seetin, R. Salomon-Ferrer, C. Sagui, V. Babin, T. Luchko, S. Gusarov, A. Kovalenko, and P.A. Kollman (2012), AMBER 12, University of California, San Francisco

14