Molecular Dynamics Conformational Search of Six Cyclic Peptides Used in the Template Assembled Synthetic Protein Approach for Protein De Novo Design RAINER FLOECEL and MANFRED MUTTER* Universite de Lausanne, Section de chirnie, Rue de la Barre 2, CH-1005 Lausanne, Switzerland
SYNOPSIS
Six cyclic peptides, designed to act as topological templates in the TASP (template assembled synthetic protein) approach in protein de novo design, were investigated employing a 100ps, 900-K molecular dynamics conformational search. The peptides are composed of two Lys-X-Lys ( X = Gly, Ala) tripeptides connected at its N- and C-terminal end by a ProGly motif and a cystine bridge ( I ) , two Pro-Gly units (11) , naphthalene derivatives (III), and tetrahydronaphthalene derivatives of different stereochemistry ( IV-VI) . The molecular dynamics conformational search established that template I had @-sheetlike geometry. Templates 11-VI showed different preferential geometries, among them, e.g., distinct preferences for type V turns in Pro-Gly containing peptides and close spatial arrangement of hydrophobic naphthalene moieties. The orientation of the lysine side chains within preferential geometries of the individual templates is analyzed and a tentative evaluation for their potential to stabilize TASP molecules of 4-helix bundle topology is given. 01992 John Wiley & Sons, Inc.
INTRODUCTION Design of protein-like conformations has been the goal of a number of approaches in the last decade.’-5 As a general strategy for the design of artificial tertiary structures employing synthetic chemistry, the construction of template assembled synthetic proteins, commonly referred to as TASP, has been proIn this approach secondary structure forming peptide blocks of amphiphilic nature are covalently attached to a topological carrier molecule termed template. The template molecule then is thought to direct the individual peptide blocks to a characteristic folding unit. Evidently the spatial arrangement of the peptide blocks is to some extent predetermined by the functional groups on the template serving as “anchoring points” and thus by their mutual spatial arrangement. The purpose of this study is to obtain information about conformational
* To whom correspondence should be addressed. Biopolymers, Val. 32, 1283-1310 (1992) 0 1992 John Wiley & Sons, Inc.
CCC 0006-3525/92/101283-28$04.00
features of a number of templates currently in our laboratory by means of using a MD conformational search. The six templates IVI investigated in this study are all cyclic peptide molecules with two tripeptide motifs of the form Lys-Gly-Lys (11-VI), in one case Lys-Ala-Lys ( I ) being interconnected at each side through dipeptides or dipeptide mimetics (Figure 1A-F). Those are a Pro-Gly motif and a cystine bridge ( I ) , two Pro-Gly motifs ( I1) , two 2-carboxy-8-aminomethylenenaphthalene residues (1111, and the three combinations resulting from the stereoisomers of 2-carboxy-8-R( S ) -aminomethylene- (5,6,7,8) -tetrahydronaphthalene with two Lys-Gly-Lys tripeptides (IV-VI) . The €-aminogroups of the lysine residues were treated in the acetylated form, thus mimicking partially the situation in the TASP molecule. The conformational preferences found among the templates by this study, then, were used to perform a tentative grouping of their suitability in the design of 4-helical bundle TASP molecules. Independently, the conformational preferences defined by this MD study can be used in comparison to computational investigations l5 of other cyclic peptides. 1283
1284
FLOEGEL AND MUTTER
Ac-C y s-Ly s (A c)- A Ia- Ly s (A c,) I S
/
4
-
Pro 5
58 G ~ Y
Gly
4 8 Pro
I
6
-
/
1
8.
3B Figure 1A-F. VI.
28
1B
U
1B
2B 2
3
38
28
n
1B
a) * = S b) * = R c) * = s
a)*=R b)*=R
[(lD-F)I
c)
*=s
Schematic drawing of the six templates investigated. A-F: Templates I-
METHODS Model building of initial starting conformations of templates I-VI was performed with the aid of interactive molecular graphics. It proceeded with the following underlying criteria. The tripeptide motifs were aligned in antiparallel @-sheetlike manner, so that the side chains of Lys ( Ac) were assembled on the same face of the template being in extended conformations (X, 180" ) . The 0-sheet like alignment provided for possible hydrogen bonding between peptide bond N-H hydrogens and the C =0 oxygen of Lys ( Ac) residues. Torsion angles within the tripeptide motifs were selected close to values ( @ = - 1 5 0 " / 1 = 160") found in the P-sheet region. Restrictions to the alignment of the tripeptides in a P-sheet conformation were imposed by the need to maintain proper geometry for the building blocks intended to connect the tripeptides. Templates I and I1 (Figure 1A,B) were most favorable in terms of introducing, in the tripeptides, mainchain torsion angles close to the values found in 0sheets. The maximum deviation allowed for the above listed 0-sheet values was 25". This was facilitated by the interconnecting Pro-Gly residues, which are commonly found in 0-turns, and by the
-
4
\~ys(~c)-~iy-Lys(Ac) 3B
CHP-NH-Lys(Ac)-Gly-L.ys (A C)-Cd
\ Pro I
I
NH2- CyI s-Ly s (Ac) A I a- Ly s(Ac)
1 2 3 C O - L ~ S ( A C ) - G ~ ~ - L ~ SmNH-CHP (AC)-
-
Ly s (A C) G Iy Ly s (A C)
high flexibility provided by the cystine bridge. In all cases type I1 turn dihedrals ( @ i + l = -60", Pi+l = 120" = 80",li+z = 0" ) were selected for the Pro-Gly dipeptides.16 The maximum deviation of those dihedral values was allowed to be 15". In the resulting model sets, Lys ( Ac ) hydrogen-bond donor and acceptor distances in the antiparallel 0-sheets were between 1.9 and 2.4 A, and the C a distances between the turn i and i 3 residue were then 5.0 A in template I (Figure 1 A ) , and 5.4 A in template I1 (Figure 1 B ) .This is well within the conventional definition of a 0-turn demanding that the i and i 3 distances are less than 7.0 The initial intention upon synthesis of 2-carboxy8-aminomethylene-naphthaleneand its tetrahydronaphthalene derivative was to obtain 0-turn mimetics resembling geometrical features of a @-turn motif, yet providing more rigidity in order to maintain or induce a more stable 0-sheet. Preliminary model building studies indicated that at least in the case of the 2-carboxy-8-aminomethylene-naphthalene the geometry was not optimal for adjacent residues. Thus in the case of the template I11 with 2carboxy-8-aminomethylene-naphthalene the C a distances of Lys ( Ac) in the turn i and i 3 positions were 7.7 A and the distances for main-chain hydro-
+
+
+
1285
MOLECULAR DYNAMICS CONFORMATIONAL SEARCH
gen-bond donors and acceptors were 4.1 and 4.3 A, which clearly cannot account for possible hydrogen b ~ n d i n g . Upon '~ employing tetrahydronaphthalene derivatives in templates IV-VI the Ca distances were in all cases 7.1 A, and possible hydrogen-bond distances were between 2.9 and 3.5 A. The mainchain torsional angles among the tripeptides of templates 111-VI were still in the P-sheet region, but compared to templates I and I1 one had to allow for larger deviations (up to 45") from above values. In order to obtain partial charges for the naphthalene derivatives and the acetylated lysines that are not parameterized within the consistent valence f orce f ield ( CVFF) ,l8 partial charges were derived from residues having groups similar to the ones used in calculations. The partial charges were chosen such that microcompensated groups resulted. All other parameters were used as implemented in the generic parameter set of the CVFF. The model conformations of the six templates were subjected to energy minimization with the aid of the CVFF until the maximum derivatives were less than 0.001 kcal/mol. Starting with the minimized conformations, MD at 900 K was performed for 100 ps after an initial 10ps equilibration time. The integration time step was set to 1.0 fs. Along the trajectory at 900 K, for every 1 ps a conformation was taken and subsequently energy minimized until the maximum derivatives were less that 0.001 kcal/mole. For the calculation of nonbonded interaction, no cutoff distances were employed. A distance-dependent macroscopic dielectric constant of 2.5 *r was used. In order to avoid isomerization about peptide bonds during the 900 K conformational search and equilibration period, peptide bond w dihedrals were kept at the trans conformation by imposing an energetic penalty of 3.3 kcal /rad. In order to compare for each template the resulting 100 minimized conformations, rms cluster graphs were prepared. Figure 2A-F indicates conformational relationships by accumulation of dark small squares representing low mutual rms deviations. Areas corresponding to conformational families are framed in Figure 2A-F. Selection of those areas was also supported by other figure sets throughout and visual examination. All calculations were performed on a Silicon Graphics Personal Iris 4D-25-TG workstation. For model building and the MD calculations, the programs Insight and Discover from Biosym, Inc., were used. Graphical representations were prepared with the programs Schakall' and GraphPAD" on an IBM-PS2.
RESULTS AND DISCUSSION Template I
One of the distinct clusters, which could be defined in Figure 2A, is one extending from 9 to 21 ps. Comparing Figure 2A to Figure 3A, it appears that this area coincides with a time comprising many lowenergy conformations and among them the one with the lowest energy of all at 19 ps time. Table I demonstrates that this conformational class is characterized by main-chain torsional angles close to the P-sheet region. The only value that deviates more from the ones commonly found in the P-sheet region is the ?T! angle of Lys(Ac) 9. By consequence, the four Lys ( Ac) side chains are assembled on the same face of the template, which was considered desirable for the TASP design of 4-helical bundles. For possible main-chain hydrogen bonds, time-averaged distances of between 9 and 2 1 ps bonds were found to be 3.7 A between Lys ( Ac ) 4 and Lys ( Ac ) 7; for Lys(Ac) 2 and Lys(Ac) 9, the distances were 2.8 A. Thus one cannot assign i -P i 3 hydrogen bonds in residues that are part of the turn motif.7 The main-chain torsion angles of residues Pro 5 and Gly 6 (Table I ) conform very well to dihedrals of a type V turnbeing characterized by @ / \k values close to -80"/80" for i 1 and SO"/-SO" for i 2. Average Ca distance between turn i and i 3 Lys ( Ac ) 4 and Lys(Ac) 7 residues were found to be 5.8 A. The general definition for a turn demands a distance of less than 7.0 A between the Cai and Ccqt3, and specifically for the type V turn, that there is no intraturn i and i 3 hydrogen-bond capability.16 Therefore the Pro-Gly motif constitutes an ideal type V turn in this case. Moving on in time there is a steep rise in energy after 2 1 ps by more than 18 kcal/mole. It marks the beginning of the next cluster extending from 22 to 57 ps. Among this cluster two subsets were defined, one from 27 to 43 ps and one from 46 to 57 ps. The major feature of this cluster is that either one or two Lys ( Ac ) side chains are not assembled on the same face of the template anymore relative to their orientations in the first cluster. In the subset from 27 to 43 ps it is the side chain of Lys(Ac) 2 that points to the opposite face. During the second subset from 46 to 57 ps this reorientation happens equally with both, Lys(Ac) 2 and Lys(Ac) 9. This reorientation is additionally accompanied by a flip of the disulfide bridge to the opposite face of the template, too. Independent of the described changes of the disulfide bridge, Lys ( Ac) 2 and Lys ( Ac) 9, the ele-
-
-
+
+
+
+
+
1286
FLOEGEL AND MUTTER
10
20
30
40
50 60 Time [ps]
0.48 < x
SYNOPSIS
Six cyclic peptides, designed to act as topological templates in the TASP (template assembled synthetic protein) approach in protein de novo design, were investigated employing a 100ps, 900-K molecular dynamics conformational search. The peptides are composed of two Lys-X-Lys ( X = Gly, Ala) tripeptides connected at its N- and C-terminal end by a ProGly motif and a cystine bridge ( I ) , two Pro-Gly units (11) , naphthalene derivatives (III), and tetrahydronaphthalene derivatives of different stereochemistry ( IV-VI) . The molecular dynamics conformational search established that template I had @-sheetlike geometry. Templates 11-VI showed different preferential geometries, among them, e.g., distinct preferences for type V turns in Pro-Gly containing peptides and close spatial arrangement of hydrophobic naphthalene moieties. The orientation of the lysine side chains within preferential geometries of the individual templates is analyzed and a tentative evaluation for their potential to stabilize TASP molecules of 4-helix bundle topology is given. 01992 John Wiley & Sons, Inc.
INTRODUCTION Design of protein-like conformations has been the goal of a number of approaches in the last decade.’-5 As a general strategy for the design of artificial tertiary structures employing synthetic chemistry, the construction of template assembled synthetic proteins, commonly referred to as TASP, has been proIn this approach secondary structure forming peptide blocks of amphiphilic nature are covalently attached to a topological carrier molecule termed template. The template molecule then is thought to direct the individual peptide blocks to a characteristic folding unit. Evidently the spatial arrangement of the peptide blocks is to some extent predetermined by the functional groups on the template serving as “anchoring points” and thus by their mutual spatial arrangement. The purpose of this study is to obtain information about conformational
* To whom correspondence should be addressed. Biopolymers, Val. 32, 1283-1310 (1992) 0 1992 John Wiley & Sons, Inc.
CCC 0006-3525/92/101283-28$04.00
features of a number of templates currently in our laboratory by means of using a MD conformational search. The six templates IVI investigated in this study are all cyclic peptide molecules with two tripeptide motifs of the form Lys-Gly-Lys (11-VI), in one case Lys-Ala-Lys ( I ) being interconnected at each side through dipeptides or dipeptide mimetics (Figure 1A-F). Those are a Pro-Gly motif and a cystine bridge ( I ) , two Pro-Gly motifs ( I1) , two 2-carboxy-8-aminomethylenenaphthalene residues (1111, and the three combinations resulting from the stereoisomers of 2-carboxy-8-R( S ) -aminomethylene- (5,6,7,8) -tetrahydronaphthalene with two Lys-Gly-Lys tripeptides (IV-VI) . The €-aminogroups of the lysine residues were treated in the acetylated form, thus mimicking partially the situation in the TASP molecule. The conformational preferences found among the templates by this study, then, were used to perform a tentative grouping of their suitability in the design of 4-helical bundle TASP molecules. Independently, the conformational preferences defined by this MD study can be used in comparison to computational investigations l5 of other cyclic peptides. 1283
1284
FLOEGEL AND MUTTER
Ac-C y s-Ly s (A c)- A Ia- Ly s (A c,) I S
/
4
-
Pro 5
58 G ~ Y
Gly
4 8 Pro
I
6
-
/
1
8.
3B Figure 1A-F. VI.
28
1B
U
1B
2B 2
3
38
28
n
1B
a) * = S b) * = R c) * = s
a)*=R b)*=R
[(lD-F)I
c)
*=s
Schematic drawing of the six templates investigated. A-F: Templates I-
METHODS Model building of initial starting conformations of templates I-VI was performed with the aid of interactive molecular graphics. It proceeded with the following underlying criteria. The tripeptide motifs were aligned in antiparallel @-sheetlike manner, so that the side chains of Lys ( Ac) were assembled on the same face of the template being in extended conformations (X, 180" ) . The 0-sheet like alignment provided for possible hydrogen bonding between peptide bond N-H hydrogens and the C =0 oxygen of Lys ( Ac) residues. Torsion angles within the tripeptide motifs were selected close to values ( @ = - 1 5 0 " / 1 = 160") found in the P-sheet region. Restrictions to the alignment of the tripeptides in a P-sheet conformation were imposed by the need to maintain proper geometry for the building blocks intended to connect the tripeptides. Templates I and I1 (Figure 1A,B) were most favorable in terms of introducing, in the tripeptides, mainchain torsion angles close to the values found in 0sheets. The maximum deviation allowed for the above listed 0-sheet values was 25". This was facilitated by the interconnecting Pro-Gly residues, which are commonly found in 0-turns, and by the
-
4
\~ys(~c)-~iy-Lys(Ac) 3B
CHP-NH-Lys(Ac)-Gly-L.ys (A C)-Cd
\ Pro I
I
NH2- CyI s-Ly s (Ac) A I a- Ly s(Ac)
1 2 3 C O - L ~ S ( A C ) - G ~ ~ - L ~ SmNH-CHP (AC)-
-
Ly s (A C) G Iy Ly s (A C)
high flexibility provided by the cystine bridge. In all cases type I1 turn dihedrals ( @ i + l = -60", Pi+l = 120" = 80",li+z = 0" ) were selected for the Pro-Gly dipeptides.16 The maximum deviation of those dihedral values was allowed to be 15". In the resulting model sets, Lys ( Ac ) hydrogen-bond donor and acceptor distances in the antiparallel 0-sheets were between 1.9 and 2.4 A, and the C a distances between the turn i and i 3 residue were then 5.0 A in template I (Figure 1 A ) , and 5.4 A in template I1 (Figure 1 B ) .This is well within the conventional definition of a 0-turn demanding that the i and i 3 distances are less than 7.0 The initial intention upon synthesis of 2-carboxy8-aminomethylene-naphthaleneand its tetrahydronaphthalene derivative was to obtain 0-turn mimetics resembling geometrical features of a @-turn motif, yet providing more rigidity in order to maintain or induce a more stable 0-sheet. Preliminary model building studies indicated that at least in the case of the 2-carboxy-8-aminomethylene-naphthalene the geometry was not optimal for adjacent residues. Thus in the case of the template I11 with 2carboxy-8-aminomethylene-naphthalene the C a distances of Lys ( Ac) in the turn i and i 3 positions were 7.7 A and the distances for main-chain hydro-
+
+
+
1285
MOLECULAR DYNAMICS CONFORMATIONAL SEARCH
gen-bond donors and acceptors were 4.1 and 4.3 A, which clearly cannot account for possible hydrogen b ~ n d i n g . Upon '~ employing tetrahydronaphthalene derivatives in templates IV-VI the Ca distances were in all cases 7.1 A, and possible hydrogen-bond distances were between 2.9 and 3.5 A. The mainchain torsional angles among the tripeptides of templates 111-VI were still in the P-sheet region, but compared to templates I and I1 one had to allow for larger deviations (up to 45") from above values. In order to obtain partial charges for the naphthalene derivatives and the acetylated lysines that are not parameterized within the consistent valence f orce f ield ( CVFF) ,l8 partial charges were derived from residues having groups similar to the ones used in calculations. The partial charges were chosen such that microcompensated groups resulted. All other parameters were used as implemented in the generic parameter set of the CVFF. The model conformations of the six templates were subjected to energy minimization with the aid of the CVFF until the maximum derivatives were less than 0.001 kcal/mol. Starting with the minimized conformations, MD at 900 K was performed for 100 ps after an initial 10ps equilibration time. The integration time step was set to 1.0 fs. Along the trajectory at 900 K, for every 1 ps a conformation was taken and subsequently energy minimized until the maximum derivatives were less that 0.001 kcal/mole. For the calculation of nonbonded interaction, no cutoff distances were employed. A distance-dependent macroscopic dielectric constant of 2.5 *r was used. In order to avoid isomerization about peptide bonds during the 900 K conformational search and equilibration period, peptide bond w dihedrals were kept at the trans conformation by imposing an energetic penalty of 3.3 kcal /rad. In order to compare for each template the resulting 100 minimized conformations, rms cluster graphs were prepared. Figure 2A-F indicates conformational relationships by accumulation of dark small squares representing low mutual rms deviations. Areas corresponding to conformational families are framed in Figure 2A-F. Selection of those areas was also supported by other figure sets throughout and visual examination. All calculations were performed on a Silicon Graphics Personal Iris 4D-25-TG workstation. For model building and the MD calculations, the programs Insight and Discover from Biosym, Inc., were used. Graphical representations were prepared with the programs Schakall' and GraphPAD" on an IBM-PS2.
RESULTS AND DISCUSSION Template I
One of the distinct clusters, which could be defined in Figure 2A, is one extending from 9 to 21 ps. Comparing Figure 2A to Figure 3A, it appears that this area coincides with a time comprising many lowenergy conformations and among them the one with the lowest energy of all at 19 ps time. Table I demonstrates that this conformational class is characterized by main-chain torsional angles close to the P-sheet region. The only value that deviates more from the ones commonly found in the P-sheet region is the ?T! angle of Lys(Ac) 9. By consequence, the four Lys ( Ac) side chains are assembled on the same face of the template, which was considered desirable for the TASP design of 4-helical bundles. For possible main-chain hydrogen bonds, time-averaged distances of between 9 and 2 1 ps bonds were found to be 3.7 A between Lys ( Ac ) 4 and Lys ( Ac ) 7; for Lys(Ac) 2 and Lys(Ac) 9, the distances were 2.8 A. Thus one cannot assign i -P i 3 hydrogen bonds in residues that are part of the turn motif.7 The main-chain torsion angles of residues Pro 5 and Gly 6 (Table I ) conform very well to dihedrals of a type V turnbeing characterized by @ / \k values close to -80"/80" for i 1 and SO"/-SO" for i 2. Average Ca distance between turn i and i 3 Lys ( Ac ) 4 and Lys(Ac) 7 residues were found to be 5.8 A. The general definition for a turn demands a distance of less than 7.0 A between the Cai and Ccqt3, and specifically for the type V turn, that there is no intraturn i and i 3 hydrogen-bond capability.16 Therefore the Pro-Gly motif constitutes an ideal type V turn in this case. Moving on in time there is a steep rise in energy after 2 1 ps by more than 18 kcal/mole. It marks the beginning of the next cluster extending from 22 to 57 ps. Among this cluster two subsets were defined, one from 27 to 43 ps and one from 46 to 57 ps. The major feature of this cluster is that either one or two Lys ( Ac ) side chains are not assembled on the same face of the template anymore relative to their orientations in the first cluster. In the subset from 27 to 43 ps it is the side chain of Lys(Ac) 2 that points to the opposite face. During the second subset from 46 to 57 ps this reorientation happens equally with both, Lys(Ac) 2 and Lys(Ac) 9. This reorientation is additionally accompanied by a flip of the disulfide bridge to the opposite face of the template, too. Independent of the described changes of the disulfide bridge, Lys ( Ac) 2 and Lys ( Ac) 9, the ele-
-
-
+
+
+
+
+
1286
FLOEGEL AND MUTTER
10
20
30
40
50 60 Time [ps]
0.48 < x
