Comparative Characterization of Short Monomeric Polyglutamine

May 4, 2010 - 2.2 Replica Exchange Molecular Dynamics (REMD)(23). Generally ...... Looking ahead to the biggest issues for chemistry worldwide, includ...
0 downloads 0 Views 284KB Size
7056

J. Phys. Chem. B 2010, 114, 7056–7061

Comparative Characterization of Short Monomeric Polyglutamine Peptides by Replica Exchange Molecular Dynamics Simulation Miki Nakano,*,† Hirofumi Watanabe,‡,§ Stuart M. Rothstein,| and Shigenori Tanaka*,‡,§ Graduate School of Human DeVelopment and EnVironment, Kobe UniVersity, 3-11 Tsurukabuto, Nada, Kobe 657-8501, Japan, Graduate School of Engineering, Department of Computer Science and Systems Engineering, Kobe UniVersity, 3-11 Tsurukabuto, Nada, Kobe 657-8501, Japan, CREST, Japan Science and Technology Agency, 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan, and Department of Chemistry and Centre for Biotechnology, Brock UniVersity, St. Catharines, Ontario, Canada L2S 3A1 ReceiVed: December 28, 2009; ReVised Manuscript ReceiVed: April 19, 2010

Polyglutamine (polyQ) diseases are caused by an abnormal expansion of CAG repeats. While their detailed structure remains unclear, polyQ peptides assume β-sheet structures when they aggregate. To investigate the conformational ensemble of short, monomeric polyQ peptides, which consist of 15 glutamine residues (Q15), we performed replica exchange molecular dynamics (REMD) simulations. We found that Q15 can assume multiple configurations due to all of the residues affecting the formation of side-chain hydrogen bonds. Analysis of the free energy landscape reveals that Q15 has a basin for random-coil structures and another for R-helix or β-turn structures. To investigate properties of aggregated polyQ peptides, we performed multiple molecular dynamics (MMD) simulations for monomeric and oligomeric Q15. MMD revealed that the formation of oligomers stabilizes the β-turn structure by increasing the number of hydrogen bonds between the main chains. 1. Introduction Patients diagnosed with Huntington’s disease characteristically have glutamine (Q) repeat sequences longer than 37 residues in the Huntingtin protein, while the repeat length in unaffected persons is about 20.1-3 If the polyglutamine (polyQ) region is longer than the critical length of 37 residues, the protein will misfold and aggregate to form insoluble, amyloid-like fibrils that contribute to neuronal dysfunction and eventual cell death.4-8 Expanded polyQ peptides are known to form β-sheet structures, as shown by X-ray fiber diffraction studies, circular dichroism (CD), Fourier transform infrared spectroscopy, and other methods.9-12 Such structural information is important, but, currently, details on the structure of polyQ remain unclear. To date, two main structures have been proposed for the aggregated polyQ peptide. Based on the X-ray diffraction pattern of comparatively short synthetic peptide D2Q15K2, Perutz et al.13 suggested that polyQ might adopt a β-helix structure, thereby explaining the critical length of polyQ peptide. This structure was recently challenged by Sikorski and Atkins14 who, upon reinterpretation of the diffraction pattern, proposed polyQ peptide aggregates that consist of β-strands linked with β-turns. In contrast to the β-helix structure, more hydrogen bonds are present in the structure proposed by Sikorski and Atkins, and it can explain the diffraction pattern more reasonably. Our group has reported previous computational analyses15-17 of these two structures. * To whom correspondence should be addressed. (M.N.) Telephone: +81-78-803-6619. Fax: +81-78-803-7761. E-mail: [email protected]. (S.T.) Telephone: +81-78-803-6620. Fax: +81-78-803-7761. E-mail: [email protected]. † Graduate School of Human Development and Environment, Kobe University. ‡ Graduate School of Engineering, Department of Computer Science and Systems Engineering, Kobe University. | Brock University. § CREST.

Recent studies indicate that monomeric polyQ has no specific secondary structure. Chen et al.18 measured the CD spectrum of K2Q42K2 during aggregation, demonstrating that over time polyQ peptide forms β-sheet structures from the random-coil state. By performing multiple molecular dynamics simulations for short peptides, Wang et al.19 found that the ensemble of polyQ was indeed disordered. Using fluorescence correlation spectroscopy measurements, Crick et al.20 suggested that the ensemble of monomeric polyQ comprised a heterogeneous collection of collapsed structures. Lastly, Masino et al.21 characterized the solution structure of polyQ as random-coil using a model system based on glutathione S-transferase fusion proteins. In the computational study reported here, we performed replica exchange molecular dynamics (REMD) simulations to investigate the conformational ensemble of short, monomeric polyQ in solution. To elucidate how the conformational distribution of polyQ differs from those of similar peptides, we compared the results of simulations among several peptides. Furthermore, to investigate the properties of the aggregated polyQ peptides, we performed multiple molecular dynamics (MMD) simulations for a monomer, dimer, and trimer of Q15, each starting from the β-turn structure. 2. Materials and Methods 2.1. Materials. This study focuses on short, monomeric polyQ, which consists of 15 glutamine residues (Q15). To confirm that our REMD calculations can reproduce the expected Protein Data Bank (PDB) structure, we also chose two peptides from the PDB for simulations: 1E0Q and 1KFP. These are registered in the PDB as peptides having β-turn structures. Furthermore, we compared the results of simulations among these peptides and characterized the properties of polyQ. We also prepared a 15-residue polyglycine peptide (G15), which is not expected to have a specific secondary structure. To further investigate Q15, we prepared two mutant models: Q7PGQ6 and

10.1021/jp9122024  2010 American Chemical Society Published on Web 05/04/2010

Characterization of Monomeric PolyQ Peptide by REMD

J. Phys. Chem. B, Vol. 114, No. 20, 2010 7057

TABLE 1: List of Sequences for All Peptides Which Are Used in This Studya

TABLE 2: List of Number of Replicas and Temperatures Which Were Used in the REMD Simulations for All Peptides PDB ID or peptide name

a Red, blue, green, and black letters represent acidic, basic, polar, and hydrophobic residues, respectively. For all peptides, extended structures are constructed in which the N- and C-termini are capped with acetyl and N-methyl groups.

Q3PGQ10. In Q7PGQ6, a proline-glycine pair was inserted in the center of Q15, while in Q3PGQ10, a proline-glycine pair was inserted near the N-terminal end of Q15. For each peptide, extended structures were constructed using the leap module included in the Amber10 package22 where the N- and C-termini are capped with acetyl and N-methyl groups, respectively. In Table 1, the sequences for all six peptides used in this study are listed. 2.2. Replica Exchange Molecular Dynamics (REMD).23 Generally, with conventional molecular dynamics simulation, the system often becomes trapped in a local minimum of potential energy space. To explore a wide range of energy space without becoming trapped in a local minimum, we performed REMD simulations. In REMD, each replica in a canonical ensemble of the fixed temperature is simulated simultaneously and independently for a certain number of MD steps. Pairs of replicas at neighboring temperatures are exchanged with acceptance probability:

w(xi |xj) )

{

1; for ∆ e 0 exp(-∆); for ∆ > 0

}

(1)

where ∆ ) [1/kBTi - 1/kBTj][Ei - Ej]; subscripts i and j are the replica numbers; xi and xj are the set of coordinates of all the atoms at absolute temperatures Ti and Tj, with energies of Ei and Ej, respectively; and kB is the Boltzmann constant. After exchange, the simulations resume at the new temperatures. 2.3. REMD Conditions. For REMD simulations, we used the REMD module of Amber10 in combination with the parm96 force field and generalized Born (GB) implicit solvent model.24 This combination of force field and solvation model is known to give reasonable structures and a fair balance between R-helix and β-sheet.25,26 We used 16 replicas in each simulation, except for G15, where the small number of atoms led us to use 10 replicas instead. The values of temperature that were used in this study are listed in Table 2. Energy minimizations were performed to relax the initial structures. Minimization was carried out by 1000 steps of conjugate gradient method, followed by 500 steps of steepest descent calculation, using the multisander module included in Amber10. For each replica, before the REMD simulations, we performed a 200 ps simulation with time step of 2 fs to ensure that each replica was equilibrated to its individual temperature. All replicas started from the same structure, whose energy was

number of replicas

1E0Q

16

1KFP

16

G15

10

Q15, Q7PGQ6, Q3PGQ10

16

temperatures [K] 281.1, 300.0, 320.1, 341.6, 364.5, 389.0, 415.1, 442.9, 472.7, 504.4, 538.2, 574.3, 612.9, 654.0, 697.9, 744.7 281.4, 300.0, 319.8, 340.9, 363.4, 387.3, 412.9, 440.1, 469.2, 500.1, 533.1, 568.3, 605.8, 645.7, 688.3, 733.8 270.5, 300.0, 332.7, 369.0, 409.2, 453.8, 503.2, 558.1, 618.9, 686.4 280.2, 300.0, 321.2, 343.8, 368.1, 394.0, 421.8, 451.6, 483.5, 517.6, 554.1, 593.2, 635.0, 679.8, 727.8, 779.1

minimized using the method described above, before they were heated to their respective temperatures using a Langevin thermostat27 with a collision frequency γ of 1.0 ps-1. After equilibration, the REMD simulations were started. All simulations were performed with a 2 fs time step, a replica exchange interval of 1 ps, a nonbonded cutoff length of 99 Å, and an external dielectric constant, ε, equal to 78. Langevin dynamics were applied as temperature control with the collision frequency γ of 1.0 ps-1. Throughout the simulations, bond distances involving hydrogen atoms were fixed using the SHAKE algorithm.28 The simulation time was 300 ns for each replica, and the corresponding acceptance ratios for exchange were 30-40%. 2.4. Multiple Molecular Dynamics Simulation (MMD). To investigate properties of aggregated polyQ peptides, we performed MMD simulations for a monomer, dimer, and trimer consisting of Q15 β-turn structures, with the latter two constructed by placing REMD-generated monomers side-by-side (see Figure 6 below). We performed standard molecular dynamics simulations on these structures with the multisander module included in Amber10. These simulations were carried out for 10 ns at 300 K with the parm96 force field and GB implicit solvent model. One hundred simulations were performed with different random seeds for each peptide. 3. Results 3.1. Reproduction of the PDB Structure. Figure 1 shows the distribution of root-mean-square deviation (RMSD) values from the PDB structures for heavy atoms within the main chains (N, CR, C, and O) of 1E0Q and 1KFP. We discarded the initial 50 ns of REMD data because the systems were not sufficiently equilibrated in this interval. Then we analyzed 12 500 snapshots from the trajectories, taken from the 50-300 ns interval at 300 K. Also shown in Figure 1 are PDB and representative REMD structures for each peptide. We observed that our REMD simulations successfully reproduced the PDB-registered structures for 1E0Q and 1KFP, demonstrating that our REMD conditions are adequate for the short peptides used in this study. Hereafter, we will show the results of analyses for 12 500 snapshots from the trajectories, taken from the 50-300 ns interval at 300 K. 3.2. Classification of Structures. We classified our REMD structures as R-helix, β-turn, and random-coil using the definitions listed in Table 3. The dictionary of protein secondary structure program (DSSP)29 was used to determine the secondary structure of each residue. Figure 2 shows the results of our classification. As registered in the PDB, 1E0Q and 1KFP mainly adopt β-turn structures. Expectedly, G15 only adopts random-coil structures. Q15 favors random-coil structures in implicit solvent,

7058

J. Phys. Chem. B, Vol. 114, No. 20, 2010

Nakano et al.

Figure 2. Results of structure classification for snapshots from REMD according to the definitions listed in Table 3. Green bar: random-coil. Yellow bar: R-helix. Red bar: β-turn.

Figure 1. RMSD values from the PDB structures of (a) 1E0Q and (b) 1KFP for the heavy atoms within the main chain (N, CR, C, and O). 12500 snapshots from the trajectories, taken from the 50-300 ns interval at 300 K, were employed. Yellow bars represent the RMSD value distribution, and blue lines represent its accumulated population ratio. Also shown are the PDB and the representative REMD structures for each peptide, in which green ribbons represent the PDB structures and pink ones represent the REMD snapshots whose RMSD values to the PDB structures are smaller than 2 Å. The structure figures were created with VMD.31

TABLE 3: Definitions of Structure Classification Used for the REMD Resultsa class

definition

R-helix

• the number of residues which have R-helix property g 4 • the number of hydrogen bonds between the main chains g 2 β-turn • the number of residues which have antiparallel-β property g 4 • the number of hydrogen bonds between the residues apart from each other by more than 3 residues g 1 random-coil • structures other than R-helix and β-turn

a The dictionary of protein secondary structure program (DSSP)29 was used to determine the secondary structure of each residue.

but it also demonstrates propensity for R-helix and β-turn structures with the ratios of 27% and 17%, respectively. For Q7PGQ6, one of the mutant models of Q15, the fraction of β-turn structures is enhanced to 41%, while the fraction of randomcoil structures is increased to 70% for mutant model Q3PGQ10. For Q15, we performed simulations starting from both extended and R-helix structures. Their conformational distributions became similar by performing 300 ns simulations (data not shown). 3.3. Hydrogen Bond Counting. We counted and averaged the number of hydrogen bonds for 12 500 snapshots of each of six peptides. The definition of a hydrogen bond is such that the N-H-O or O-H-O angle is 120°-180° and the N-O or O-O distance is shorter than 3.3 Å. These results are shown in Figure 3. For Q15 and its mutants, the average total number of hydrogen bonds is smaller than those of either 1E0Q or 1KFP. However, the contribution from the side chains of these peptides is high in comparison with 1E0Q and 1KFP. For 1E0Q and 1KFP, the

Figure 3. Average number of hydrogen bonds per molecule for each of six peptides. The definition of a hydrogen bond is such that the N-H-O or O-H-O angle is 120°-180° and the N-O or O-O distance is shorter than 3.3 Å. Green bar: hydrogen bonds between the main chains. Red bar: those between the main chain and the side chain. Yellow bar: those between the side chains.

fractions of contribution from the side chains are 16% and 29%, respectively. By contrast, for Q15 and its mutants, the fractions of contributions from the side chains are 49-58%. Naturally, G15 has a small number of hydrogen bonds, and these are only between the main chain atoms. For a more detailed analysis, we investigated the average numbers of hydrogen bonds between the side chains for 1E0Q, 1KFP, and Q15 on a per-residue basis (Figure 4). For 1E0Q and 1KFP, only charged or polar residues are involved in the formation of hydrogen bonds, whereas for Q15 all residues are capable of forming hydrogen bonds with each other. 3.4. MM-GBSA Analysis. We calculated the energies of 1E0Q, 1KFP, and Q15 in vacuum and in implicit solvent using the MM-GBSA module30 included in Amber10. Furthermore, we performed a mutation analysis for 1E0Q and 1KFP by replacing all charged residues included in these peptides with glutamines, glycines, or leucines. Because the intramolecular electrostatic interaction energy of Q15 is approximately between 2 and 3 times larger than those of other peptides, the molecular mechanical energy of Q15 is low. However, the contribution from solvation in this case is smaller than those of others (Q15, 25%; 1E0Q, 78%; 1KFP, 66%). This shows that contributions to the total energy from the solvation of glutamine mutants are similar to those of glycine or leucine, despite the latter being a hydrophobic residue (see the Supporting Information). 3.5. Free Energy Landscape. We calculated free energies for each of six peptides in terms of the potential of mean force:

Characterization of Monomeric PolyQ Peptide by REMD

G ) G0 + kBT ln(p/p0)

J. Phys. Chem. B, Vol. 114, No. 20, 2010 7059

(2)

Here, G is the free energy at absolute temperature T and p is the population of the snapshots at the designated reaction coordinates. G0 is the standard free energy calculated at the area where the population of snapshots is p0, an arbitrary choice. We chose the gyration radius and the number of hydrogen bonds as the reaction coordinates for the free energy landscape. The gyration radius represents the compactness of peptide, and the number of hydrogen bonds relates to its thermal stability. The combination of these two quantities appropriately describes structural features of the peptide; the first two principal components used as independent variables give similar free energy features (Supporting Information). Figure 5 shows free energy landscapes at 300 K for each of six peptides. Representative structures are shown for the designated basins in their respective figures. While the free energy landscapes of 1E0Q and 1KFP have only one basin corresponding to β-turn structure as registered in the PDB, the landscape of Q15 has a basin for random-coil structures and another for R-helix or β-turn structures. The activation energy between random-coil structures and R-helix or β-turn structures is comparable to thermal energy at room temperature (0.6-0.7 kcal/mol). For Q7PGQ6, the relative depth of β-turn structure’s basin is deeper than that for random-coil structures (0.6 kcal/ mol). Conversely, for Q3PGQ10, the activation energy between these two basins is smaller (0.2 kcal/mol). For G15, the sole visible basin corresponds to random-coil structures. 3.6. Multiple Molecular Dynamics (MMD). We performed MMD simulations to study the stability of Q15 dimer and trimer, constructed by placing side-by-side β-turn monomers generated by REMD. The RMSD values become large when the oligomers dissociate with increasing simulation time. Therefore, instead of calculating average RMSD values, we counted the number of snapshots for which the CR atoms’ RMSD values are smaller than 5 Å with respect to their initial structures, (Figure 6). Initial structures of dimer and trimer are also shown in Figure 6. By forming oligomers, the β-turn structure becomes more stable compared to that of the monomer. Figure 7 shows the average number of hydrogen bonds per residue for each peptide during 10 ns of simulation time. For the dimer and trimer, the average numbers of hydrogen bonds between the main chains are larger, and these bonds are very stable compared to those of the monomer.

Figure 4. Average number of hydrogen bonds per molecule for each residue between the side chains for 1E0Q (blue circle), 1KFP (red square), and Q15 (green triangle). The dependence on the residue number is illustrated.

4. Discussion Since we have reproduced the PDB-registered structures for 1E0Q and 1KFP, our REMD conditions are adequate for the short peptides used in this study. According to the results of the classification of structures, 1E0Q and 1KFP mainly form the reported PDB structures, while G15 only forms random-coil structures, as we expected. Interestingly, 1E0Q and 1KFP also form R-helix and random-coil structures in quantities up to 20%, which are not registered in the PDB. While monomeric Q15 favors random-coil structures in implicit solvent, it also forms a certain fraction of β-turn and R-helix structures. It is surprising that such a short, monotonous peptide can form definite secondary structure. For Q7PGQ6, the fraction of β-turn structures is enhanced because of the proline-glycine pair, which is inserted into the center of Q15. This result suggests that the polyQ peptide can easily form a β-turn structure in the presence of a guiding influence. For Q3PGQ10, however, the fraction of random-coil structures is increased. This may be ascribed to the fact that the length of the uninterrupted polyQ sequence becomes shorter than that in Q15 due to the inserted proline-glycine pair near the N-terminal end of Q15. This result indicates that it would be difficult for the combination of short polyQ fragments (Q10 and Q3) to form any definite secondary structure. For Q15, the effect of the side chains on the formation of hydrogen bonds is large in comparison with 1E0Q or 1KFP, since all residues can form side-chain hydrogen bonds with each other. Therefore, the polyQ peptide can assume various secondary structures by means of different combinations of hydrogen bonds. Because glutamine has high polarity, the interaction energy is large, not only between the peptide and water but also intramolecularly. Furthermore, such behavior is observed when charged residues in 1E0Q and 1KFP are replaced by glutamines in the MM-GBSA analyses. These results show that glutamine has hydrophobic properties despite being a polar residue. The free energy landscapes of all peptides are consistent with the classification results described above. The PDB-registered peptides, 1E0Q and 1KFP, have one basin which corresponds to the PDB structure, and G15 has one basin which corresponds to random-coil structures. The landscape of Q15 has a basin for random-coil structures in addition to one for R-helix or β-turn structures. The activation energy and the difference in depth between these basins are small. This suggests that polyQ can readily be transferred between these basins. Compared to Q15, the basin corresponding to β-turn structures becomes deeper for Q7PGQ6, whereas for Q3PGQ10 the activation energy between these two basins becomes smaller. These features of free energy landscapes for Q15 mutant models can be interpreted in the same manner as those of the classification results; the polyQ peptide can easily form a β-turn structure in the presence of a guiding influence, and the length of 10 residues is too short to form any definite secondary structure. We chose the gyration radius and the number of hydrogen bonds as reaction coordinates for the free energy landscapes. These quantities represent structural features of the peptide, and they are used to distinguish between the structured and unstructured state. Notably, in these landscapes, the R-helix and β-turn structures are in the same basin. To differentiate the basin of R-helix and β-turn structures, and to confirm that the activation energy between the structured state and the structureless state is small, we performed a cluster analysis and constructed free energy landscapes for Q15 and its mutants using the principal component values as the reaction coordinates.

7060

J. Phys. Chem. B, Vol. 114, No. 20, 2010

Nakano et al.

Figure 5. Free energy landscapes for each of six peptides which are expressed as a function of the gyration radius and the number of hydrogen bonds. Arrows connect representative structures with their associated basins.

Figure 6. Fraction of snapshots, whose RMSD values of CR atoms from the initial structure are smaller than 5 Å for monomer (green line), dimer (blue line), and trimer (red line). Initial structures of dimer and trimer of the β-turn structure are also shown in the inset.

Thereby, the basins for R-helix, β-turn, and random-coil structures are separated for Q15, and other features of these new free energy landscapes for each peptide are consistent with the results shown in Figure 5. This analysis is described in the Supporting Information. To confirm that Q15 dynamically transfers between different conformational states, we performed MMD simulations of Q15 and 1E0Q from different initial structures (data not shown). We found that Q15 can transfer forward and backward among different conformational states, while transitions among different structures of 1E0Q are one-directional toward the PDB registered structure. Results of the MMD simulations for monomeric and oligomeric Q15 that start from the β-turn structures clearly demonstrate that the β-turn structure becomes more stable by forming oligomers. Additionally, the average number of hydrogen bonds per residue between the main chains increases. For monomeric Q15, hydrogen bonding between side chains causes the peptide to take on various conformations. This is in contrast to oligomeric Q15, where hydrogen bonds between the main chains ensure that the peptides maintain their β-turn structures. It is well-known that a peptide such as Q15, which has low sequence complexity and no hydrophobic residues, will favor a disordered state. In contrast, our calculations reveal that Q15 has a basin of well-defined, conformationally structured states. These surprising

Figure 7. Average numbers of hydrogen bonds per residue during 10 ns MMD simulations for (a) monomer, (b) dimer, and (c) trimer. Green line: hydrogen bonds between the main chains. Red line: those between the main chain and the side chain. Blue line: those between the side chains.

Characterization of Monomeric PolyQ Peptide by REMD results may be due to the hydrophobicity of glutamine and the wide variety of the hydrogen bonds network. Although a peptide of 15 residues is too short to elucidate the mechanism of polyglutamine diseases, our study used such a peptide to reveal the role of hydrogen bonding in polyQ peptides and demonstrate the hydrophobicity of glutamine. Our results allow one to speculate on what will happen upon lengthening the chain. For short, monomeric polyQ, the probability of observing a β-turn structure is low. Consequently, it will be a rare event that two β-turn peptides meet to form a dimer. However, once formed, the dimer structure is more stable than that of the monomer, and aggregation may be accelerated. We expect the probability of nucleation to be higher with increased peptide concentration, or with longer chains. 5. Conclusion In the present study, we performed REMD simulations to investigate a conformational ensemble of short, monomeric polyQ in solution, and MMD simulations to investigate the properties of the aggregated polyQ peptides. Furthermore, we carried out the analyses of free energy landscapes and MMGBSA energies. We compared these results among several peptides to elucidate how the polyQ peptide differs from similar peptides. From the results of the simulations, we revealed the role of hydrogen bonds of polyQ peptide. Q15 takes various kinds of secondary structures with a wide variety of side-chain hydrogenbond combinations. On the other hand, Q15 β-turn structure becomes more stable by forming oligomers, with increased formation of main-chain hydrogen bonds. In the free energy landscape, Q15 has a basin for random-coil structures and one for R-helix or β-turn structures, with a small activation energy between them. It is very interesting that, despite having a short and monotonous sequence, Q15 takes structured states. From the MM-GBSA analyses, we found that Q15 favors intramolecular interactions as well as intermolecular ones due to its high polarity. These interactions give the hydrophobic properties to polyQ peptide. The chain length of Q15 may be too short to elucidate the mechanism of polyglutamine diseases. However, the properties of Q15, which we found in this study, allow us to speculate what will happen when the chain length becomes longer. Acknowledgment. This work has been partially supported by CREST, Japan Science and Technology Agency. We thank Mr. Bryan M. B. VanSchouwen, Mr. Robert Giacometti, Dr. Ikuo Kurisaki, Mr. Takatoshi Fujita, and Prof. Kuniyoshi Ebina for useful comments and discussions. The numerical calculations were partially carried out on a PC-cluster system in Cybermedia Center, Osaka University, and on the Shared Hierarchical Academic Research Computing Network and Compute/Calcul Canada (SHARCNET: www.sharcnet.ca). Supporting Information Available: Detailed procedures and results of MM-GBSA analysis, the cluster analysis, and the free energy landscapes using principal component values. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) The Huntington’s Disease Collaborative Research Group. Cell 1993, 72, 971-983.

J. Phys. Chem. B, Vol. 114, No. 20, 2010 7061 (2) Rubinsztein, D. C.; Leggo, J.; Coles, R.; Almqvist, E.; Biancalana, V.; Cassiman, J. J.; Chotai, K.; Connarty, M.; Crauford, D.; Curtis, A.; Curtis, D.; Davidson, M. J.; Differ, A. M.; Dode, C.; Dodge, A.; Frontali, M.; Ranen, N. G.; Stine, O. C.; Sherr, M.; Abbott, M. H.; Franz, M. L.; Graham, C. A.; Harper, P. S.; Hedreen, J. C.; Jackson, A.; Kaplan, J. C.; Losekoot, M.; MacMillan, J. C.; Morrison, P.; Trottier, Y.; Novelletto, A.; Simpson, S. A.; Theilmann, J.; Whittaker, J. L.; Folstein, S. E.; Ross, C. A.; Hayden, M. R. Am. J. Hum. Genet. 1996, 59, 16–22. (3) de Cristofaro, T.; Affaitati, A.; Cariello, L.; Avvedimento, E. V.; Varrone, S. Biochem. Biophys. Res. Commun. 1999, 260, 150–158. (4) Scherzinger, E.; Lurz, R.; Turmaine, M.; Mangiarini, L.; Hollenbach, B.; Hasenbank, R.; Bates, G. P.; Davies, S. W.; Lehrach, H.; Wanker, E. E. Cell 1997, 90, 549–558. (5) Singer, S. J.; Dewji, N. N. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 1546–1550. (6) Bevivino, A. E.; Loll, P. J. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 11955–11960. (7) Georgalis, Y.; Starikov, E. B.; Hollenbach, B.; Lurz, R.; Scherzinger, E.; Saenger, W.; Lehrach, H.; Wanker, E. E. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 6118–6121. (8) Cooper, J. K.; Schilling, G.; Peters, M. F.; Herring, W. J.; Sharp, A. H.; Kaminsky, Z.; Masone, J.; Khan, F. A.; Delanoy, M.; Borchelt, D. R.; Dawson, V. L.; Dawson, T. M.; Ross, C. A. Hum. Mol. Genet. 1998, 7, 783–790. (9) Perutz, M. F.; Johnson, T.; Suzuki, M.; Finch, J. T. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 5355–5358. (10) Sharma, D.; Shinchuk, L. M.; Inouye, H.; Wetzel, R.; Kirschner, D. A. Proteins 2005, 61, 398–411. (11) Tanaka, M.; Morishima, I.; Akagi, T.; Hashikawa, T.; Nukina, N. J. Biol. Chem. 2001, 276, 45470–45475. (12) Sharma, D.; Sharma, S.; Pasha, S.; Brahmachari, S. K. FEBS Lett. 1999, 456, 181–185. (13) Perutz, M. F.; Finch, J. T.; Berriman, J.; Lesk, A. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 5591–5595. (14) Sikorski, P.; Atkins, E. Biomacromolecules 2005, 6, 425–432. (15) Ogawa, H.; Nakano, M.; Watanabe, H.; Starikov, E. B.; Rothstein, S. M.; Tanaka, S. Comput. Biol. Chem. 2008, 32, 102–110. (16) Nakano, M.; Watanabe, H.; Starikov, E. B.; Rothstein, S. M.; Tanaka, S. Interdiscip. Sci.: Comput. Life Sci. 2009, 1, 21–29. (17) Van Schouwen, B. M. B.; Nakano, M.; Watanabe, H.; Tanaka, S.; Gordon, H. L.; Rothstein, S. M. J. Mol. Struct. 2010, 944, 12–20. (18) Chen, S.; Berthelier, V.; Hamilton, J. B.; O’Nuallain, B.; Wetzel, R. Biochemistry 2002, 41, 7391–7399. (19) Wang, X.; Vitalis, A.; Wyczalkowski, M. A.; Pappu, R. V. Proteins 2006, 63, 297–311. (20) Crick, S. L.; Jayaraman, M.; Frieden, C.; Wetzel, R.; Pappu, R. V. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 16764–16769. (21) Masino, L.; Kelly, G.; Leonard, K.; Trottier, Y.; Pastore, A. FEBS Lett. 2002, 513, 267–272. (22) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Crowley, M.; Walker, R. C.; Zhang, W.; Merz, K. M.; Wang, B.; Hayik, S.; Roitberg, A.; Seabra, G.; Kolossvary, I.; Wong, K. F.; Paesani, F.; Vanicek, J.; Wu, X.; Brozell, S. R.; Steinbrecher, T.; Gohlke, H.; Yang, L.; Tan, C.; Mongan, J.; Hornak, V.; Cui, G.; Mathews, D. H.; Seetin, M. G.; Sagui, C.; Babin, V.; Kollman, P. A. AMBER 10; University of California San Francisco: San Francisco; 2008. (23) Sugita, Y.; Okamoto, Y. Chem. Phys. Lett. 1999, 314, 141–151. (24) Onufriev, A.; Bashford, D.; Case, D. A. Proteins 2004, 55, 383– 394. (25) Zhou, R. Proteins 2003, 53, 148–161. (26) Shell, M. S.; Ritterson, R.; Dill, K. A. J. Phys. Chem. B 2008, 112, 6878–6886. (27) Pastor, R. W.; Brooks, B. R.; Szabo, A. Mol. Phys. 1988, 65, 1409– 1419. (28) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J. Comput. Phys. 1977, 23, 327–341. (29) Kabsch, W.; Sander, C. Biopolymers 1983, 22, 2577–2637. (30) Kollman, P. A.; Massova, I.; Reyes, C.; Kuhn, B.; Huo, S.; Chong, L.; Lee, M.; Lee, T.; Duan, Y.; Wang, W.; Donini, O.; Cieplak, P.; Srinivasan, J.; Case, D. A.; Cheatham, T. E., III. Acc. Chem. Res. 2000, 33, 889–897. (31) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996, 14, 33–38.

JP9122024