| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Biophys J, October 2002, p. 1774-1783, Vol. 83, No. 4
*LURE, Bât 209D, Centre Universitaire Paris-Sud,
F-91898 Orsay Cedex, France; Bioinformatique
Génomique et Moléculaire, INSERM U436, Université
Paris 7, 75251 Paris Cedex 05, France
 |
ABSTRACT |
|---|
 TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
CONCLUSIONS
REFERENCES
|
|---|
We describe a combined use of experimental and simulation
techniques to configure side chains in a coiled coil structure. As
already demonstrated in a previous work, x-ray diffraction patterns
from hard 
-keratin fibers in the 5.15 Å meridian zone reflect the
global configuration of the 
1 dihedral angle of the coiled coil side chains. Molecular simulations, such as energy minimization and molecular dynamics, and rotameric representation in
the PDB, are used here on a heterodimeric coiled coil to investigate the dihedral angle distribution along the sequence. Different procedures have been used to build the structure, the quality assessment was based on the agreement between the simulated diffraction patterns and the experimental ones in the fingerprint region of coiled
coils (5.15 Å). The best one for building a realistic coiled coil
structure consists of placing the side chains using molecular dynamics
(MD) simulations, followed by side chain positioning using SMD or SCWRL
procedures. The side chains and the backbone are equilibrated during
the MD until they reach an equilibrium state for the t/g+
ratio. Positioning the side chains on the resulting backbone, using the
above procedures, gives rise to a well-defined 5.15 Å meridian reflection.
| |
INTRODUCTION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
CONCLUSIONS
REFERENCES
|
|---|
Coiled coil structures are widespread motifs
involved in a large number of oligomerization domains in various
proteins. These structures are also the basic units at a molecular
level in the fibrous proteins, such as intermediate filaments (Parry
and Steinert, 1995
). Regular coiled coil structures are obtained by two
or more -helices winding around each other (Crick, 1953b) to form a
left-handed supercoil. Their stabilization is mainly achieved by
interfacial hydrophobic interactions between residue side chains. The
amino acid sequences composing these structures have a characteristic repetition of a seven-amino acid motif, usually noted
(abcdefg)n, with the a and
d positions occupied by mostly hydrophobic residues (Parry
et al., 1985; Parry and Fraser, 1985) facing inward to form a
hydrophobic core into a knob-into-hole packing manner (Crick, 1953b).
The side chains at the interface pack against each other in a specific
configuration (O'Shea et al., 1991; Burkhard et al., 2000).
Until recently, the atomic structure of coiled coils was only known for
short pieces involved in protein domains. The longest coiled coil
structure determined at atomic resolution was that of the cFos-cJun
bZip leucine zipper, which is 39 residues long (Glover and Harrison,
1995). In fibrous proteins, the coiled coil structures can be many
heptads long. Our structural knowledge of this class of fibrous
proteins is very limited because of the lack of adapted conditions for
their crystallization to be studied by x-ray diffraction techniques;
moreover, they are not soluble, so they cannot be studied by
spectroscopic techniques such as NMR. High-resolution structural
information on a long coiled coil piece has only been obtained recently
by solving the crystalline structure of the 18 heptad repeat coiled
coil domain of the actin bundling protein cortexillin (Burkhard et al.,
2000).
Despite this lack in the fiber structures field, some information about
these structures can be deduced from x-ray fiber diffraction. In the
wide angle region, along the meridian axis (along the fiber axis),
x-ray fiber diffraction patterns from hard -keratins exhibit an
intense reflection at 5.15 Å distance. This has been shown to arise
essentially from the molecular structure. In a previous study we have
demonstrated the side chain configuration dependence of the 5.15 Å reflection intensity (Busson et al., 1999). The origin of this
reflection, which is the fingerprint on the diffraction pattern of the
coiled coil structure in keratinous tissues, was unclear until this
recent study. It was shown that an all-g+ side
chain configuration could switch off the intensity of the 5.15 Å reflection, whereas an all-trans configuration gives rise to
a very intense reflection.
A fine analysis is, however, needed to complete this information,
aiming at quantifying the proportion of the two main
1 (g+ and
trans) configurations in a coiled coil. Progress in homology modeling and protein design has generated considerable interest in
methods for predicting side chains packing in hydrophobic cores of
proteins (Bower et al., 1997; Dunbrack and Cohen, 1997; Tufféry et al., 1997; Koehl and Delarue, 1994). However, the present techniques are not practically validated because of the lack of an experimental tool able to provide information on this aspect. Moreover, most of the
side chain configuration prediction studies are based on a statistical
representation of the side chain conformations in the Protein Data Bank
(PDB) (Bower et al., 1997; Tufféry et al., 1997). Actually, in
their paper, Bower et al. developed a method for configuring side
chains on the basis of a rotameric library per residue (fewer than 10 per residue) extracted from the PDB and depending on the backbone
parameters. Tufféry et al. used an approach in which a
conformational search was performed in the rotameric space again built
from the PDB structures. The effect of the backbone flexibility has
also been analyzed using an approach consisting of small deviations of
the structure from its experimental form; this study shows that only
deviations larger than 2 Å RMSD give rise to significant changes on
the predicted side chain configurations. However, all of these
prediction methods are based on the rotameric and structural
representation in the PDB. This criterion was seriously limiting for
fibrous and membrane proteins, which are nearly absent in this database.
The objective of our study deals with the determination of a method for
obtaining a realistic initial structure for molecular simulations on
coiled coils. This is much needed for different studies in the field of
fibrous proteins, like intermediate filaments, whose structures cannot
be obtained from usual experimental techniques. Modeling the coiled
coil structures is all the more important because these proteins are
naturally submitted to different stresses whose effect is to induce
structural changes that are hard to analyze experimentally. In
addition, various mutations in the sequences affect the structure and
the activity of these proteins (Kreis and Vale, 1999). We have already
studied the effect of some stresses on the molecular structure of
keratin (Kreplak et al., 2001), but an atomic structure is still
lacking for understanding the behavior at this scale.
In the present work, we perform different simulations based on various
techniques and force fields. Starting from an ideal backbone
conformation, based on Crick equations (Crick, 1953a), the first
simulation consisted of optimally placing the side chains with
different side-chain positioning algorithms. In the second simulation,
the side chains were placed in an all-g+
1 initial configuration, then the system was
relaxed and explored through a short MD simulation monitored by the
trans 1 configuration content. The
third simulation consisted of using different positioning side chain
algorithms on a backbone conformation selected along the dynamic
trajectory. In all the cases the main purpose was to get the best
agreement between the simulated diffraction patterns and the
experimental one in the fingerprint region of the coiled coils (5.15 Å). In addition, the results were compared to the known
high-resolution coiled coil crystallographic structures.
As a result, the side chain configurations can be inferred in the
specific case of the coiled coil conformation, taking into account the
dynamical aspect due to the backbone flexibility that had been shown to
play an important role in side chain conformations (Harbury et al.,
1995).
Two x-ray diffraction references have been used: 1) an experimental
fiber diffraction pattern from hard -keratin in hair; and 2) a
calculated diffraction pattern from the crystallographic cortexillin
structure, to avoid many extra signals originating from different parts
of the hair fiber.
This study shows that the backbone flexibility, which plays a key role in the quality of the side chain positioning, can be achieved by a short molecular dynamics. For the side chain configuration we have used three positioning methods; two of them satisfy the x-ray criterion (SMD and SCWRL, see below). We show that the approach presented below can be followed for all coiled coils.
| |
MATERIALS AND METHODS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
CONCLUSIONS
REFERENCES
|
|---|
Initial model building
A coiled coil dimer 101 residues long was built from the amino
acid residues of part 1B of a type I 8C-1 (Dowling et al., 1986) and a
type II 7C (Sparrow et al., 1989) keratin molecule. These have been
shown experimentally to associate preferentially onto a heterodimer of
the type I and the type II keratins (Herrling and Sparrow, 1991). The
1B piece of the molecule has been chosen because of its long
uninterrupted heptad repeat sequence (18 heptads).
The coiled coil structure can be modeled by curves consisting of
circles of radius r2 processing with a
frequency 
1 around a helix path of radius
r0 and a pitch
p0. The model parameterization was
performed according to Busson and Doucet, 1999, using upgraded Crick's
equations. These are based on a rotation matrix transformation relating
a regular simple -helix in the laboratory frame to a structure in a
frame defined with its x axis on the radial vector and
z axis tangent to a major helix. Therefore, a coiled helix was obtained by rotating a circle in the new frame when moving up the
major helix path.
A value of p0 = 150 Å has been
attributed to the coiled coil pitch on the basis of experimental values
for coiled coil pieces in the literature (O'Shea et al., 1991; Seo and
Cohen, 1993). In this way, the trace of the C
atoms was built and the other backbone atoms are placed according to a
systematic construction with the O package (Jones, 1985). Side chains
were then added with a chosen configuration. The dimeric molecule was
built with two chains in exact register, as it has already been
determined experimentally (Parry et al., 1985; Coulombe and Fuchs,
1990).
X-ray diffraction pattern calculation
X-ray diffraction patterns were simulated using a home-made
program that calculates the scattered intensity on a plane detector by
the model coiled coil. It determines successively for each pixel
(x, z) of the simulated zone (x along
the radial axis and z along the meridian) the corresponding
scattering vector S. The intensity
I(x, z) was then calculated as the
square modulus of the structure factor F(S)
multiplied by the polarization and Lorentz factors. The polarization
factor expression for a synchrotron radiation beam selected by a
cylindrically curved crystal monochromator had been calculated (Kahn et
al., 1982). The Lorentz factor, expressed as (cos
2
)3, where 2 was the diffraction angle,
accounts for both the beam divergence and the still geometry with a
plane detector perpendicular to the x-ray beam.
The structure factor, which is a complex function, sums up the atomic
contributions (phase and amplitude):
|
= 1.450 Å and a sample-to-detector distance of 125 mm (experimental conditions
for Fig. 1). The pixel size along the
x and z axes was 0.1 mm, which corresponds in the reciprocal units to 5 × 10
4
Å1. Simulated patterns were finally
cylindrically averaged by summing up the intensities produced by eight
coiled coils related by successive 45° rotations around their axes to
take into account the different orientations of coiled coils with
respect to the x-ray beam. The graphic representation was simply
achieved using gray scale images with a darkness proportional to the
intensity.
|
Molecular simulations
Once the backbone model structure is built, the side chain
dihedrals have to be chosen. The 1 dihedral
angles are subjected to certain restrictions arising from steric
hindrance between 
side chain atoms and the main chain atoms.
Different procedures have been followed for the positioning and
configuration of the side chains: use of the extended rotameric
database from the PDB, energy minimization, and MD simulations.
Side chain positioning using a rotameric-based approach
Three approaches have been used to position the side chains on the coiled-coil backbone. All of them are based on a rotamer description for the side chain configurations, but they differ both by the rotamer database and the positioning algorithm. They are briefly summarized in the following paragraphs. The first approach, called SMD (sparse matrix driven), developed by Tufféry et al., 1997, consists of a conformational search in the
rotameric space using an energy criterion. The energy is computed using
the complete FLEX force field (Lavery et al., 1986, 1995; Zakrzewska
and Pullman, 1985). The rotameric space is constituted of 214 rotamers
for the 20 amino acids. To overcome the combinatorial problem of side
chain conformation prediction, the 3D structure is clustered into
groups with a small number of strongly interacting residues. In each
cluster, the set of energetically optimal rotamers is determined. The
residue conformations are obtained through an appropriate combination
of the different local solutions and the process is repeated two or
three times, slightly modifying the definition of the clusters at each
step. The final solution is the one corresponding to the minimal energy
value. This procedure is systematically followed by a simplified energy
minimization process in the dihedral conformational space of side
chains. The residues are first ranked according to their interaction
energy with the remaining residues, the first residue being that with the most unfavorable energy terms. Side chains are tested one at a
time and the process is iterated until no interaction energy exceeds a given threshold (1 kcal/mol). All the side chains with energy
conflicts are thus allowed to leave their rotamer state. At the end of
the process the side chains are generally still close to their
rotameric state, but do not present any energy clash.
The second approach (Koehl and Delarue, 1994), hereafter referred to as
Confmat, uses the same backbone-independent rotamer library as
Tufféry et al. in the SMD method. The prediction of the side
chain conformations is obtained through a mean field approximation. The
mean field potential is defined as a sum of real potentials (only van
der Waals interactions are considered) weighted by normalization
factors. These factors are varied to yield a minimum of the effective
potential of the effective system. The weighting factors are obtained
from the elements of the conformational matrix. The procedure is first
initialized with a conformational matrix CM0,
whose terms are the probability for each side chain to adopt a given
rotamer. The mean potential is computed and the energies are then
converted to probabilities defining a new conformational matrix
CM1 that is compared to
CM0. Then the process is iterated, modifying the
CM0 conformational matrix slightly at each step, taking into account the CM1 values. When no
significant difference exists between the CM0 and
CM1 matrices, the convergence is considered to be
reached and the process is stopped. The final solution corresponds, for
each residue, to the rotamer with the highest probability. The side
chains are then energy-minimized with the CHARMM force field in the
Cartesian space with 1000 Powell steps.
The last approach (Bower et al., 1997) is called SCWRL. It uses a
backbone-dependent rotamer library. The search strategy is made
hierarchically. First, it chooses a structure with residues in their
most probable backbone-dependent rotamers and systematically resolves
the conflicts that arise from that structure, changing progressively to
less probable rotamers until one is found without any conflict with the
main chain. At this step, the side chain-side chains conflicts are
examined. If residues clash, they are put into a cluster. The residues
are allowed to explore all of their respective rotamers. When residues
outside the cluster clash with the cluster residues for a given rotamer
conformation, they are added to the cluster, leading to a progressive
growth of the cluster size. At the end of the SCWRL procedure, the
combination with the minimum steric clash score is assumed to be the
best possible solution.
Configuring side chains using molecular dynamics
Because no simple criteria exist for attributing side chains in fibrous proteins, we have used the three most abundant1 configurations (g+, t,
g) corresponding respectively to 
60°,
180°, and 60° dihedral angles. A default choice was made to assign
to the initial structure a g+
1 content of 82%, which is one of the default
libraries in the O package. This rate results from the proportion of
Val residues, which are all constructed in trans
configuration, and Ser and Thr residues in g
form. The other side chain dihedrals (2,
... ) have been taken from the default library of the package. We
note that the initial chosen structures are unfavorable, their
optimization has to be achieved by energy minimization.
The structure was energy-minimized with the GROMOS (van Gunsteren and
Berendsefn, 1987) package using a combination of 500 steepest descent
steps followed by a 500-step conjugate gradient. Simulations were
carried out in vacuo. The constant temperature MD calculations were
carried out at T = 300 K, with an integration time step
of 0.001 ps.
Two different protocols have been followed to achieve the
dynamics. On the one hand, constraints were applied on the backbone (C, O, N, H) atoms during all the 0.4-ns
molecular dynamics. This calculation (hereafter called
"constrained-molecular dynamics") was achieved to save the initial
global geometric parameters for the coiled coil (pitch, radius) and it
also allowed the effect of these parameters on the conformational
transitions of the side chains to be tested, as well as the influence
of the geometry on these transitions. On the other hand, a molecular
dynamics (hereafter called "free-molecular dynamics") has been
carried out under different conditions: during the equilibrating
period, the structure was harmonically constrained to its initial
configuration with a force constant of 9000 kJ/mol·nm2. The constraints were progressively
reduced (1000 kJ/mol·nm2/0.005 ns) until the
temperature equilibrium was reached. When after 0.100 ns the
equilibrium was reached, a free MD was then carried out for 0.4 ns. For
the two protocols (free and restrained MD), two different MD
calculations using different sets of initial velocities were accomplished.
Molecular dynamics and rotameric procedure combination
To take into account the backbone flexibility in the case of the use of a rotameric database, we used an approach combining an MD simulation for equilibrating the structure, followed by side chains positioning from a rotameric database). Structures from the free MD were picked up every 0.01 ns and the side chains placed using the SMD procedure, followed or not by energy minimization (see above).| |
RESULTS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
CONCLUSIONS
REFERENCES
|
|---|
In all the following simulations the structures have been shown to remain in a global coiled coil conformation whatever their statistical composition in side chain configuration. The global average geometric coiled coil parameters have only undergone small variations during the minimization steps. During the MD process the coiled coil pitch p0 shifted progressively from 150 Å to an equilibrium average value of 140 Å.
Experimental information
Fig. 1 presents the experimental x-ray diffraction pattern
obtained on the whole keratin fiber. The pattern shows a well-defined meridian reflection in the 5 Å zone, characteristic of both a coiled-coil conformation (Crick, 1953b) and a peculiar amino acid side
chain configuration (Busson et al., 1999). We will focus all the
analyses on the position, shape, and relative intensity of this
reflection. It should be mentioned that the intense reinforcement of
the 5.15 Å reflection is also associated with the supramolecular organization. This organization was not taken into account in the
computed patterns on single molecules.
For ease of comparison, on the calculated diffraction patterns the intensity representation was selected to reveal the meridian 5.15 Å signal relative to the neighboring reflections, which are not seen on the experimental pattern because of the existence of a diffracted signal from other parts of the sample than the coiled coils. To help in reading and interpreting the calculated patterns, there are two relevant parameters: first, the meridian signal position near 5 Å (indicated by the central dashed arc), and second, the intensity of this reflection relative to the neighboring ones.
These facts are illustrated in Fig. 2
A, which was calculated on an all-g+
structure that shows much more intense satellite reflections around the
5 Å position at the 5.15 Å fingerprint meridian. Clearly, this
extinction is due to destructive interferences between the backbone
atoms and the g+ side chain atoms. Moreover, Fig.
2 E, which was calculated on an all-trans
structure, shows constructive interferences between the backbone atoms
and the trans side chains, which gives rise to the very
intense signal at 5.15 Å. These facts have already been demonstrated
in our previous paper (Busson et al., 1999).
|
Side chain positioning using rotameric databases and different algorithms on an ideal keratin backbone
Fig. 2 shows the diffraction patterns computed on the ideal backbone with the side chains positioned using the SMD procedure, the mean field procedure (Confmat), and the SCWRL procedure. For comparison, the pattern obtained from the structure with an all-g+ side chain configuration is presented in Fig. 2 A. Clearly, no signal in the characteristic 5 Å zone is visible in this pattern. The structure obtained with the SMD procedure leads to a signal in the correct zone, but the ratio between the main reflection intensity compared to the secondary reflections is not satisfactory (Fig. 2 B); the intensity value is of the same order, instead of being significant at 5 Å. The structure with the mean-field approach leads to a different pattern from the previous one with a reflection split into two reflections in the 5 Å zone; this shape is rather different from the experimental one, which presents a unique well-centered maximum on the meridian axis. Moreover, the intensity is too low, not more significant than the neighboring satellite reflections (Fig. 2 C). The structure with the SCWRL procedure (Fig. 2 D) presents a shape similar to the one produced by the SMD method. The statistical distribution of the1 dihedral
angle of the side chains is given for each pattern (Fig.
3). In the case of SMD, the proportion of
trans configurations reaches ~50%. Analyzing the
sequence, we notice that almost all the Glu, Lys, Arg, and Asp residues
are in trans configurations regardless of their position in
the heptad motif. In the case of the mean field approach, the proportion of trans configurations (~12%) was very small.
For the SCWRL, the distribution was quite similar to that observed with
SMD. The distribution of the side chains seems to be better for
reproducing some important features of the diffraction pattern in the
case of SMD and SCWRL compared to Confmat. Nevertheless, none of the
structures produces a pattern sufficiently close to the experimental
one. Energy-minimizing the global structure (backbone + side chains)
after side chain positioning did not improve the pattern quality (data
not shown).
|
Side chain positioning with different procedures on the experimental backbone of the cortexillin coiled coil protein
To check the validity of the different approaches for a coiled coil structure, we selected the unique long high-resolution coiled coil structure available in the PDB, namely the cortexillin protein (code: 1D7M). Its simulated diffraction pattern is shown in Fig. 4 A. It displays well-defined features in the 5 Å zone, as for the keratin fiber. Starting from the experimental backbone, the side chains were positioned using the above procedures (Fig. 4, B-D). As in the previous case, the Confmat approach fails to give an appropriate diffraction signal, its shape was again different from the calculated pattern on the experimental structure, and its intensity was very low. The produced trans configuration rate was again close to 16%, while it amounts to ~45% in the experimental structure. A slightly better signal was obtained with the SCWRL and SMD procedures: both the shape and the relative intensity of the 5 Å reflection were similar to the ones produced on the experimental crystallographic structure. Moreover, each of the three procedures produced better signal quality than they did on the ideal backbone keratin structure; the signal-to-noise ratio is improved as compared to Fig. 2 (see above).
|
1 deviations <30° from the experimental
1 value. According to this criterion, for all
the residues in the structure the SMD procedure shows overall 54%
correctly positioned side chains, with >77% side chains correctly
positioned at the interface (a, d positions in
the heptad repeat), and the SCWRL method predicts 57% residues in the
right configuration with >81% at the interface. These results show
that SMD and SCWRL produce very similar results, but SCWRL is slightly better.
Clearly, these approaches were able to place the side chains correctly
at the interface of the backbone coiled coil conformation. For the
other residues, it is likely that transitions between different
rotamers occur because of the lack in the representation of the environment.
The quality of the theoretical solution is assessed by comparison with
the unique combination of rotamers provided by the x-ray experiment.
Recent studies have mentioned the possible existence of a mixture of
rotamers for a given structure. For instance, MacKenzie et al. (1996)
have reported a fast exchange among 2 rotamers
for all leucines in the glycophorin dimer studied with NMR experiments.
Examination of the recently solved structure (MacKenzie et al., 1997)
of the transmembrane part of the glycophorin dimer (code: 1AFO.pdb)
indicates that first, the coiled coil extent is very short compared to
the two fiber molecules examined here, and second, that most of the
rotational flexibility of the leucine residues may be due to their
occurrence in poorly constrained regions. In contrast, examination of
the B factor values for the x-ray cortexillin dimer shows that most of
the residues at the interface exhibit B values lower than the mean B
value of the protein. This observation could indicate that in the
cortexillin dimer, the interface residues are less flexible than the
exposed ones. Thus, the probability of obtaining different combinations of rotameric states for the interface residues could be small.
Backbone structure influence
Aiming to test the backbone effect on the quality of the diffraction pattern, an MD simulation was carried out for a short period to allow backbone flexibility. All the side chains are initially in g+ configuration. During the dynamics, the transitions were monitored using the increasing trans configuration rate. As we can see in Fig. 5 A, the free MD reaches an equilibrium state after ~0.2 ns, resulting in a rate of ~25% of the residues in trans configuration. In Fig. 5 the variations of the backbone RMSD compared to the initial ideal backbone B are indicated. When the equilibrium t/g+ ratio was reached, one structure was selected at 0.21 ns. The dihedral distribution of the equilibrated system is shown in Fig. 6. Its backbone parameters are shown in Table 1. We observe for (
, 
) values an average shift of 5° from the ideal initial
structure. The coiled coil helix pitch decreased from 150 Å to
p = 140 Å. The decrease in the pitch value corresponds
to an overwinding of the helices, but was still compatible with the
observed pitches in coiled coil (Seo and Cohen, 1993).
|
|
|
|
|
| |
DISCUSSION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
CONCLUSIONS
REFERENCES
|
|---|
Different methods for positioning the side chains on a given backbone have been proposed, but the main interest of our study was to use experimental information as a validating criterion. The main features that emerge from the present study concern two essential points: the side chain configuration and its coupling with an appropriate backbone representation. These two points are discussed below.
Accuracy of the positioning side chain procedure: rotamer database and algorithm effects
The three different procedures explored differ in two main criteria: the rotamer databases (SCWRL and SMD/Confmat) and the algorithm. The results obtained for the cortexillin have allowed us to assess their influence on the structure.
It seems that a fine-backbone-dependent database was not necessary in
the present case due to the extreme regularity of the structure. In
other words, the rotamers in the SCWRL database in the (, )
helical region are close to that defined in the average SMD database.
On the contrary, the results obtained for the cortexillin show that the
side chain positioning algorithm plays a key role. The SMD and SCWRL
exhibit very similar 1 distribution (see Fig.
9) although based on a different
positioning method. In the same way, Confmat and SMD, while using the
same database, produce very different results. It may be noted that,
although the different procedures use various simplified force fields, their roles do not seem crucial for determining the correct
distribution of the side-chain angles.
|
Backbone influence
The different results obtained from the different starting structures show clearly that the backbone conformation is as important as the side chain conformations to provide a well-defined diffraction pattern. For instance, when the backbone conformation was restrained to its ideal conformation, letting only the side chains relax, the quality of the diffraction pattern was not improved.
At this point, it may be asked how a correct conformation for the
backbone may be obtained. The simple procedure proposed, consisting of
relaxing an initial "ideal" coiled coil structure with a short MD
simulation, seems to be appropriate. The distribution of the (, )
dihedral angles at the end of the MD simulation is quite interesting.
While the criterion used to select the final conformation concerns the
g+/t ratio of the side chain
1 angles, the distribution obtained for the
backbone angles was quite similar to that observed in the
crystallographic cortexillin dimer structure. In Table 1 the mean
values, the standard deviations, the minimum and maximum along the
sequence for the (, ) angles in keratin in the initial and final
conformations, and the corresponding values for the cortexillin and the
GCN4 coiled coils are shown. Clearly, the mean and the deviation for
the keratin final structure are close to that of cortexillin. The fine
analysis of the possible sequence dependence has not been performed
because of the few data available. Nevertheless, it might be possible
to avoid the MD step by selecting (, ) angles in the given
distribution that appear characteristic of a long coiled coil
structure. The GCN4 leucine zipper, which was considered as a template
for a coiled coil structure, exhibits a slightly different
distribution, closer to the ideal values chosen for constructing the
initial backbone keratin structure.
This shows that the GCN4 backbone structure may not be appropriate for long fibrous proteins. Further studies need to be carried out to assess the influence of various parameters such as sequence specificity, sequence length, or environment on this coiled coil backbone structure.
| |
CONCLUSIONS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
CONCLUSIONS
REFERENCES
|
|---|
In the present study we propose a new strategy combining a molecular simulation technique (molecular dynamics and side chain positioning procedure) with information from x-ray diffraction fiber patterns to obtain a detailed atomic coiled coil structure. The main result shows that a right side chain positioning method alone is not sufficient to provide correct diffraction patterns, an accurate backbone conformation is also necessary. The way to obtain both backbone and side chain conformations consists of selecting a short MD final structure and positioning the side chains with the SMD (or SCWRL) method using a rotamer database.
The present approach could be particularly interesting for most of the fiber proteins that are difficult to crystallize, but in contrast are easy to study with fiber diffraction techniques. Our strategy consists of selecting between putative coiled coil structures, built from various modeling methods, on the basis of an experimental criterion from x-ray fiber diffraction. Indeed, the present analysis shows that the keratin molecule is a good representative example of the ensemble of coiled coil structures. Its well-defined diffraction pattern could be used as a reference for other coiled coil proteins.
| |
FOOTNOTES |
|---|
Address reprint requests to Dr. Fatma Briki, LURE, Bât 209D, Centre Universitaire Paris-Sud, B.P. 34, F-91898 Orsay Cedex, France. Tel.: 33-1-64-46-88-20; Fax: 33-1-64-46-41-48; E-mail: briki{at}lure.u-psud.fr.
Submitted August 14, 2001, and accepted for publication May 17, 2002.
| |
REFERENCES |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
CONCLUSIONS
REFERENCES
|
|---|
Biophys J, October 2002, p. 1774-1783, Vol. 83, No. 4
© 2002 by the Biophysical Society 0006-3495/02/10/1774/10 $2.00
This article has been cited by other articles:
 |
|
 |
|
  A. Pineiro, A. Villa, T. Vagt, B. Koksch, and A. E. Mark A Molecular Dynamics Study of the Formation, Stability, and Oligomerization State of Two Designed Coiled Coils: Possibilities and Limitations Biophys. J., December 1, 2005; 89(6): 3701 - 3713. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. M. Bernot, C.-H. Lee, and P. A. Coulombe A small surface hydrophobic stripe in the coiled-coil domain of type I keratins mediates tetramer stability J. Cell Biol., March 14, 2005; 168(6): 965 - 974. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
