Vous êtes sur la page 1sur 23

Non-Gaussian Statistics of the Vibrational Fluctuations of

Myoglobin
J. A. Tuszynski
a
, T. Luchko
a
, E. J. Carpenter
a
, J. M. Dixon
b
, M. Peyrard
c
and Y.
Engelborghs
d
a
Department of Physics, University of Alberta, Edmonton, AB, T6G 2J, Canada;
b
Department of Physics, University of Warwick, Coventry, CV4 7AL, U.K.;
c
Ecole Normale Suprieure de Lyon, 69364 Lyon Cedex 07, France;
d
Laboratory of Biomolecular Dynamics, University of Leuven, Leuven,Celestijnenlaan 200D,
B-3001, Belgium
ABSTRACT
Experiments on the dynamics of vibrational uctuations in myoglobin revealed an interesting behavioral cross-
over occurring in the range of 180-200 K. In this temperature range the mean square displacement of atomic
positions versus temperature sharply increases its slope indicating the dissociation of CO from the haeme group.
In this paper we develop a theoretical model that provides a framework for the quantitative description of this
phenomenon. The basis of our calculations is an assumption of an eective potential with multiple local minima.
In particular we consider a quartic potential in place of the simple quadratic. We then use non-Gaussian statistics
to obtain a relationship between the mean square displacement and model parameters. We compare our model to
published experimental data and show that it can describe the data set using physically meaningful parameters
which are tted to the experimental data. In the process we verify the Gaussian approximations applicability
only to the low-temperature regime. In the high-temperature limit, however, deviations from the Gaussian
approximation are due to the double-well nature of our eective potential. We nd that the published datasets
showing the thermal transition display the qualitative trends predicted by appropriate algebraic approximations
to our predicted myoglobin behaviors.
Keywords: myoglobin, conformational changes, non-Gaussian statistics, thermal uctuations, protein dynamics
Further author information: Send correspondence to J.A.T.:
E-mail: jtus@phys.ualberta.ca
1
1. INTRODUCTION
Protein structure and function determination is one of the central problems of modern biophysics and biochem-
istry.
1
Among the million or so proteins identied by cell biologists, a special place is held by functional
proteins such as motor proteins or enzymes.
2
Despite the many advances in our understanding of how proteins
fold and function, molecular level insights into these mechanisms often remain elusive. Proteins exhibit dynamic
complexity with dierent regions, referred to as functional domains, of the same protein often showing dierent
types of activity. Of special importance to the biological functioning of proteins are structure uctuations that
are induced by binding to ligands or substrates. Such is the case of myoglobin where the oxygen ligand has to
diuse through the protein matrix before it is able to bind to the iron in the haeme group.
3
Due to a wide range of characteristic time scales, from picoseconds to tens of seconds, experimental studies
are fraught with diculties. One of the best characterized functional proteins is myoglobin, the dynamics of
which have been experimentally studied using inelastic scattering of synchrotron radiation,
4
phonon assisted
Mossbauer scattering,
5
time resolved spectroscopy,
6
and M ossbauer absorption spectroscopy
79
in addition to
standard X-ray crystallography.
10
In this paper we analyze some of the experimental results for myoglobin using
non-Gaussian statistical methods
11, 12
applied to the temperature dependence of the positional uctuations. The
objective of the present study is to provide a powerful theoretical tool in the analysis of the conformational
changes that are essential for protein molecules in order that they may carry out their functions.
2. NON-GAUSSIAN ANALYSIS OF CONFORMATIONAL DYNAMICS
Myoglobin is a protein (see Figure 1 for an illustration of its structure) consisting of 153 amino acids and one
haeme group of proto-porphyrin IX with an iron atom in the center.
4
The molecular weight of myoglobin
is approximately 17.8 kDa. The three-dimensional folding of myoglobin produces a globular structure with a
diameter of about 3 nm. Myoglobins function is to store oxygen provided by the blood until it is released for
some form of metabolic activity. The molecular oxygen binds at the active site of the protein and one of the six
co-ordinations of the iron atom. The remaining ve co-ordinates are taken up by four nitrogen atoms and one
bond to the protein matrix. The process of binding and releasing of the ligand is reversible.
4, 6
2
Figure 1. Three dimensional structure of myoglobin.
Recent experiments using Mossbauer absorption
5
as well as several other experimental methods
4, 6
provided
insights into the kinetics and dynamics of myoglobins conformational changes. In particular, ash photolysis
was used to obtain an eective potential in terms of a reaction co-ordinate.
6
It shows a global minimum at a
bound state A and two local minima states at photoproduct states B
2
and B
3
whose activation energies have
been estimated as 6.1 kJ/mol between B
2
and A, 11.1 kJ/mol between B
2
and B
3
and 22.4 kJ/mol between
B
3
and B
2
. This qualitatively indicates a three-well eective potential structure that underlies the protein
conformational dynamics.
Parak and Achterhold
4
used inelastic Rayleigh scattering of Mossbauer radiation to determine the temper-
ature dependence of the mean square displacement, x
2
, in myoglobin. It is interesting to note that in the
low-temperature range from 0K to approximately 180K the dependence on temperature is linear, and given by
x
2
/T = 2.1 10
4

A
2
/K for the average position of atoms in the protein from linear regression while the slope
is x
2
/T = 0.6 10
4

A
2
/K was obtained for the iron atom in deoxymyoglobin (Mbdeoxy) crystals. In the
high-temperature regime, between 180K and 300K, the slope of these curves increases dramatically, by a factor
of more than 10 in one case. The authors attribute the low-temperature behavior to harmonic motions of the
protein with the high-temperature behavior indicative of diusive motion limited by the bonding forces within
the protein.
In parallel to the experimental studies a formidable computational eort was undertaken to evaluate myo-
3
globins dynamics using molecular dynamics simulations.
13, 14
Interestingly, a very similar qualitative picture of
the mean square uctuations as a function of temperature emerged. The eect depends on the level of hydration
with the extreme of dry samples
15
or suitably coated ones,
16
remaining harmonic at or above room temperatures.
In the remainder of this paper we will attempt to present a relatively simple explanation that is consistent with
the model of thermal uctuations in a system with multiple potential minima corresponding to the proteins
stable conformations. This is in agreement with the results of conformational kinetics obtained by Ostermann
et al..
6
To make a direct comparison we will additionally use the experimental data points for the iron atoms
positional uctuations measured by Mossbauer absorption as well as results for independent molecular dynamics
simulations performed for myoglobin.
3. THE MODEL AND ITS RESULTS
Parak and Achterhold report two sets of data involving the temperature dependence of the mean square displace-
ment, x
2
, in myoglobin.
4
The rst set refers to the protein as a whole and the second only to the iron atom in
Mbdeoxy crystals. Both plots (Figure 6 and Figure 7) exhibit a cross-over from a linear x
2
vs. T dependence
at low temperatures to a more rapidly increasing function at high temperatures. A double well potential has
been used by Doster et al. to explain elastic incoherent neutron scattering data on myoglobin.
17
It is evident
from the ts produced by Parak and Achterhold, on the basis of normal-mode analysis, that this approach fails
to capture the cross-over and deviates from the high-temperature regime producing a rapidly increasing error
as the temperature increases. In this paper we attempt to develop a self consistent model for both experiments
and both thermal regimes.
In order to provide a simple model of uctuational dynamics in the myoglobin matrix we postulate an eective
potential of the form (see Figure 2):
V (x) = ax
4
+bx
3
+cx
2
+V
0
(1)
which accounts for the bistability of the potential under certain assumptions about the coecients. We selected
this form using Fig.4c
6
as a guide. We are ignoring the intermediate region at B
2
due to the small potential
barrier between B
2
and A and the fact that if we were to include it we would be forced to introduce additional
4
Figure 2. An eective potential illustrating Equation (1).
adjustable parameters with no experimental basis. The variable x in Equation (1) is a spatial co-ordinate that
we map to the reaction co-ordinate shown in the original gure.
We intend to develop a statistical approach for the calculation of average uctuations over the entire range
of temperatures. The conceptual reasoning involved is the Boltzmann probability distribution P(x) given by
P(x) = Z
1
e
V (x)
(2)
where the normalizing factor is the corresponding partition function given by
Z =
_

e
V (x)
dx (3)
This Boltzmann averaging technique leads to
x
2
=
_
x
2
e
V (x)
dx
_
e
V (x)
dx
(4)
for the positional variance, x
2
, measured about the origin, where the integrations are over all space. Unfor-
tunately, our experimentally guided model requires the use of an anharmonic potential so these integrals are
non-Gaussian and the solutions are not simple. We note that the experimental data we use is of uctuations
about a mean position which is not exactly the same as our choice of about the potential minimum; this dierence
will change x
2
by the square of the mean position which is zero for harmonic wells. We estimate the size of
this dierence later. Also note that the constant term, V
0
, in the potential becomes the multiplicative constant
e
V0
in Equation (2) to Equation (4) and therefore does not aect the distribution.
5
Figure 3. The probability distribution P(x) corresponding to Figure 2.
It is convenient to introduce parameters which can be directly linked to the experimental data. The rst is
the depths of the potential energy wells which we denote by d
A
and d
B
, where the A and B refer to the deep
and shallow wells respectively.
A third parameter which controls the width of the lowest energy well we call w. At low temperatures, the
uctuational motion is limited to within the connes of the dominant probability peak of P(x) (see Figure 3).
Above some transition temperature, T
t
, the molecule begins to explore the vicinity of the second equilibrium
conformation as manifested by the smaller probability peak. In this regime the Taylor expansion of V (x) about
its lowest minimum, x
A
, is
V (x) =

i=0
_
1
i!
_
d
i
V
dx
i
_
x
A
(x x
A
)
i
_
(5)
and is dominated by the parabolic term in the expansion near the minimum (the linear coecient is zero at
the minimum). For a pure harmonic (parabolic) potential, V (x) = V
0
+ w
0
x
2
, the mean square displacement is
linear with temperature and is just x
2
= (k
B
T)/(2w
0
). Using (x
2
/T)
T<Tt
for the low-temperature slope, we
therefore require that
w =
k
B
2(x
2
/T)
T<Tt
(6)
and
w =
1
2
_
d
2
V
dx
2
_
x
A
.
Appendix A provides a description of how any given choice of the three physical constants d
A
, d
B
, and
(x
2
/T)
T<Tt
gives a unique potential by xing the three coecients a, b, and c.
6
We can now proceed to estimate the temperature dependence of the actual mean square displacement over
the entire temperature regime.
4. METHOD
To calculate the mean square displacement at a temperature, we use Boltzmann statistics so that
x
2
= I
2
/I
0
(7)
where
I
k
=
_

x
k
exp((ax
4
+bx
3
+cx
2
+V
0
)) dx (8)
In order to obtain a rapidly convergent series approximation for I
k
we expand the cubic term in the exponential
bearing in mind that the quartic term will be dominant. Thus
I
k
= e
V0
_

x
k
exp((ax
4
+cx
2
))
_
1 bx
3
+

2
b
2
x
6
2!
+

3
b
3
x
9
3!
+. . .
_
dx (9)
Each of the integrals in Equation (9) can be evaluated analytically in terms of parabolic cylinder functions (see
Appendix B). Using the procedure described in Appendix B we nd that in the high-temperature limit for which
x is small we obtain
x
2

= 0.478
1

2a
e
c/

a
_
1 +

2
b
2
2!
1
(2a)
3/2
2
3/2
_
2.028e

3c/

2a
+ 7.397e
c/

2a
_
. . .
_
(10)
In the opposite regime, for low temperatures, with x >> 1, we nd that
x
2

=
1
2c
_
1 +
45
8

b
2
c
3
. . .
_
(11)
where in both equations Equation (10) and Equation (11) we have retained two terms in I
2
and I
0
from Eq.(B2)
and Eq.(B3) and used the limits in Eq.(B4) and Eq.(B5).
We can therefore conclude that spatial uctuations in an anharmonic potential should manifest themselves
by a predominantly linear temperature dependence with small power law corrections provided the temperature
is low enough. In the high-temperature regime however, positional uctuations are dominated by the product of
an exponential decay with an inverse square root dependence on temperature and a square root of temperature
pre-factor. Corrections to this behavior are approximately similar to the leading term but without the pre-factor.
7
(a) (b)
Figure 4. Temperature dependence of the mean square displacement, x
2
, over all atoms in myoglobin measured by
inelasic Rayleigh scattering. Radiation from
5
7CoCr Mossbauer source (lled) and sychrotron (open). Plotted as (a)
log-log with best-t line logx
2
= 1.28 log(T)4.35 and (b) log-inverse square root with best-t line ln(x
2
=

T) =
176/

T + 5.82. Data source: Fig. 6.


4
i dont see a best t line
In order to support the results of our calculations we have replotted the data sets from reference
4
for the
average atomic co-ordinates and for the iron atom using two dierent co-ordinate representations. First, plotting
the data with a log-log scale we show a linear dependence of the rst half of the curve x
2
vs. T for T < T
c
.
The high-temperature range, T > T
c
, in both cases follows a modied exponential dependence and thus we
have also plotted ln(x
2
/

T vs. 1/

T. It is clear from these two data sets (see Figs. 4 and 5)


4
that the low-
temperature data points follow a linear trend with temperature while the high-temperature data points are best
described by a function proportional to

T exp(/

T). Furthermore, we have similarly re-analysed data from


the molecular dynamics simulations of hydrated and dehydrated carboxymyoglobin (MbCO) published originally
by Steinbach and Brooks.
13, 14
Again, the same qualitative trend prevails showing a cross-over from a linear to a
modied exponential dependence on temperature in myoglobins vibrational uctuations. Figures 6, 7, 8, 9 and
10 summarize the data of references
13, 14
appropriately replotted. The rst set of gures describes the data sets
of Parak and Achterhold for x
2
vs. T as an average of all atoms in metmyoglobin crystals and the iron atoms
in Mbdeoxy crystals. This set is followed by the data for dehydrated and hydrated MbCO
13
and nally the
molecular dynamics simulations
14
for hydrated MbCO. All these data sets fully support our model predictions
in terms of the trend in the x
2
vs. T dependence.
It is worth emphasizing that the above calculation is quite general and applies to any anharmonic potential
8
(a) (b)
Figure 5. Temperature dependence of x
2
for the iron atom in Mbdeoxy crystals. Plotted as (a) log-log and (b)
log-inverse square root. Data source: Fig. 7.
4
(a) (b)
Figure 6. Temperature dependence of x
2
for MbCO hydrated with 0 (boxes) and 350 (circles) water molecules. Plotted
as (a) log-log and (b) log-inverse square root. Data source: Fig. 3A.
14
(a) (b)
Figure 7. Temperature dependence of x
2
for dehydrated (0 water molecules) MbCO averaged over all atoms with
dihedral transitions (boxes) and without (crosses). Plotted as (a) log-log and (b) log-inverse square root. Data source:
Fig. 2A.
13
9
(a) (b)
Figure 8. Temperature dependence of x
2
for hydrated (350 water molecules) MbCO averaged over all atoms with
dihedral transitions (circles) and without (pluses). Plotted as (a) log-log and (b) log-inverse square root. Data source:
Fig. 2B.
13
(a) (b)
Figure 9. As in Figure 7, but plotting the mass weighted variance in position r
2
over the eight -helix centers of
mass. Data source: Fig. 3A.
13
(a) (b)
Figure 10. As in Figure 8, but plotting the mass weighted variance in position r
2
over the eight -helix centers of
mass. Data source: Fig. 3B.
13
10
Figure 11. Temperature dependence of x
2
for a) Equation (1) using the parameters in Equation (12) (solid line) and
b) as in Fig. 7
4
(dots). is this our gure 7 or Paraks? also, there is no a and b plots.
including situations with two local minima separated by a potential barrier. The key message of these calculations
is that a cross-over occurs in the temperature dependence of x
2
from a linear function of T to one which is
dominated by an exponential dependence. We have supported this prediction with a number of data sets for
myoglobin originating from the work of other authors. What this means physically is that at temperatures below
the cross-over temperature of 200 K spatial uctuations are dominated by oscillations in the harmonic potential
around the ground state. Once the cross-over temperature is reached the molecule is allowed to explore higher
energy states including the vicinity of the second local minimum (see Figure 2).
The actual shape of this eective potential may be more complicated than the postulated form of Equation (1).
However, the qualitative features should remain the same. Furthermore, the eective potential depends on the
type of myoglobin, the degree of hydration and solvation of the molecule, as well as whether the average is taken
over all atoms or just the iron atom. As an illustration we have selected the data set from Fig. 7 of
4
and made a
fairly accurate t to the thermal uctuation data using a quartic polynomial function for the eective potential
as described above. We set the low-temperature slope to the value 0.6 10
4

A
2
/K, that the authors reported
in the text and choose dB = 22.4 kJ/mol as reported in Fig.4c from.
6
We can then plot x
2
vs. T using
Equation (4). We nd the choice of d
A
= 4150 K = 34.5 kJ/mol results in the curve shown in Figure 11.
Note that the heights of the potential barriers are of the same order of magnitude as typical hydrogen bond
energies. This is consistent with the role of thermal uctuations in bond-breaking leading to a conformational
11
change. The resulting eective potential coecients are
a = +1.52 10
20
J/

A
4
b = +4.33 10
21
J/

A
3
c = 5.29 10
20
J/

A
2
(12)
It can be immediately seen from the diagram that the quantitative agreement with the data is very good; the
dierences in our t value of x
2
resulting from the choice of a xed origin rather than the thermally variable
mean are less than 4% in the curve of Figure 11. Notably the deviations below 200 K are less than 1%.
Finally we wish to comment about the calculations presented in a recent paper
5
using the vibrational mode
energy and the density of states D() through the formula
x
2
(T, ) =
h
2
m
Fe
_

0
D( )

_
1
e
k
B
T
1
+
1
2
_
d (13)
This leads to good agreement at low temperatures but is very poor at high temperatures. Above 200 K the
above formula predicts a continuation of the linear trend, but with small corrections. Clearly, since this is based
on a harmonic approximation, there is no possibility of obtaining behavior consistent with the experimental
data unless the anharmonicity of the system is accounted for. We believe that myoglobin is no exception here
and similar conclusions can be expected for all proteins which undergo signicant conformational changes. A
recent paper
18
analyzed molecular dynamics simulations of human lysozome and concluded that the large volume
uctuations observed in the simulations require the inclusion of anharmonic potential terms and consequently
non-Gaussian uctuations. We could not agree more. Moreover, we foresee the usefulness of our non-Gaussian
approach in the description of virtually all functional proteins. A case in point is the motor-domain of kinesin
that changes conformation depending on whether ATP or ADP molecules are bound to it. Since activation of
the processivity of kinesins motion is temperature dependent,
19
this eect is also expected to strongly involve
non-Gaussian uctuations. Clearly this connection between a proteins structural bistability, and the statistics
of thermal uctuations oers a key insight into the activation of biological functions.
12
5.
::::::::::::::::::
MOLECULAR
:::::::::::::::::
DYNAMICS
::::::::::::::::::::
SIMULATIONS
:::::
OF
:::::::::::::::::::
MYOGLOBIN
::::::::::::::::::
HYDRATION
5.1.
::::::::::
Method
Myoglobin was simulated at a variety of hydration levels using a methodology similar to that of Steinbach and
Brooks.
14
Preparation began with removing a lone SO
4
molecule and 137 water molecules from the crystal
structure (PDB ID: 1MBC).
20
All of the histidines were set to be protonated on the ND1 site only (neutral
histidine) giving the molecule a zero net charge. (Later, using the web-based interface for WHAT IF,
21
it was
determined that while all of the histidines were neutral some were protonated on the NE2 site instead.) Since
the net charge was zero, no ions were added and the system was solvated by a 66

A diameter sphere of exible
TIP3P water molecules. After molecules overlapping with the protein had been removed 4103 water molecules
remained. This was followed by a steepest decent minimization of 100 steps, 1 ps of heating from 0 K to 300 K
and 50 ps of equilibration.
After the initial 50ps of equilibration was performed waters were removed based on their distance from the
protein surface. This resulted in 14 systems of 0, 35, 50, 60, 80, 100, 125, 150, 225, 350, 600, 1000, 1900 and
3830 water molecules, closely following Steinbach and Brooks.
14
Each of these systems was then simulated at
an equilibrium temperature of 100K and 300K, making 24 total simulations. Each simulation was heated for 1ps
to reach its equilibrium temperature and then equilibrated for 150 ps. Productions simulations of 300 ps were
run for each system and the atomic positions recorded every 0.1 ps.
All simulations were performed with CHARMM
22
28b using the CHARMM27 forceeld.
23
Constant temper-
ature dynamics simulations used the Berendsen thermostat
24
with a coupling constant of 1.0 ps for equilibration
and 5.0 ps for production runs. During heating simple temperature scaling was used. A distance based cuto of
28

A was used for all long range interactions(Lennard-Jones and electrostatic) with a switching function starting
at 24

A
14
Im not sure that this reference is necessary. Distance based cutos of this length have been shown
to capture almost all of the electrostatic energy of the system
25
and should account for polarization eects.
13
Figure 12. Electrostatic potential of myoglobin. Red is negative and blue is positive. I have included both a colour and
a black and white image.Image created with VMD.
26
5.2.
::::::::
Dipole
:::::::::::
Moment
For system of zero net charge the dominant term in the multipole expansion of the electrostatic potential is the
dipole moment,
V (x) =
1
4
0

_
_
q
r
+
p x
r
3
+
1
2

i,j
Q
ij
x
i
x
j
r
5
+. . .
_
_
, (14)
where p for a system of point charges is
p =
N

i=1
q
i
r
i
. (15)
As such it is a measure of the shape and strength of an electric eld but unlike the net charge it may uctuate.
Thus, for a system eletrostatically neutral system like hydrated myoglobin it can be used to characterize the
protein, water and the system as a whole. Figure 12 show that the potential of the myoglobin is overwhelmingly
dipolar.
Figure 13 show the relationship between the hydration level of the protein and the dipole moment of the
protein and surrounding water. Clearly, the addition water reduces the total dipole moment of the system.
Temperature plays a major role as water at 100 K is unable to conform to the electric eld of the protein. For
the 300 K case the total dipole of the system appears to plateau after hydration of 100 waters. This, however,
maybe due to due in the proteins dipole moment at this level.
14
Water 300K
Water 100K
Protein 300K
Protein 100K
log
10
(# of waters)
p
(
D
e
b
y
e
)
4 3.5 3 2.5 2 1.5 1 0.5 0
250
200
150
100
50
0
(a)
Protein & water 300K
Protein & water 100K
log
10
(# of waters)
p
(
D
e
b
y
e
)
4 3.5 3 2.5 2 1.5 1 0.5 0
160
140
120
100
80
60
40
20
(b)
Figure 13. Average dipole moment as a function of hydration. Values for protein and water are shown at 100 K (solid
lines) and 200 K (dashed lines).
Figure 14. Dipole moments of surface water on myoglobin.Image created with VMD.
26
.
5.3.
::::::::
Dipole
::::::::::::::::
Correlations
Water has a natural hexagonal geometry when in the pure bulk form. In close proximity to protein the network
of hydrogen bonds is disrupted as can be seen in Figure 14. In this close contact individual water molecules
conform to the electric eld of the protein rather than that of their neighbours. The extent to which water
networks are disrupted can be seen by calculating the distance dependence of the correlations between water
dipoles and their nearest-neighbours as well as with the electric eld produced by the protein.
Nearest-neighbour dipole-dipole correlations are calculated
cos(
ij
)
R
surf
i
= p
i
p
j
(16)
where
ij
is the angle between the dipoles, R
surf
i
is the distance of the oxygen atom of the i
th
water molecule from
the protein. This is only calculated if the distance between the i and j oxygens is less than 3.5364

A, slightly
larger than the diameter of a water molecule. Figure 15(a) shows the temperature and proximity dependence
15
300 K
100 K
Distance from Protein (

A)

c
o
s
(

i
j
)

20 15 10 5 0
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
(a)
300 K
100 K
Distance from Protein (

A)

c
o
s
(

i
)

20 15 10 5 0
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
(b)
Figure 15. Nearest neighbour dipole-dipole (a) and dipole-electric eld correlations (b) for myoglobin hydrated by 1900
water molecules.
of this. At the higher temperature the water has greater mobility and is able to more easily conform to the
potential of the protein near the surface. This is reected in the more gradual transition between no correlation
and a correlation value of about 0.45. This mean correlation of bulk water corresponds to an angle of 64

which
one would expect for a hexagonal lattice. The slow decay of this value demonstrates the range of the proteins
inuence. Large deviations from this, starting at about 15

A at 300 K, are distortions due to the water-vacuum
interface.
Dipole-electric eld correlations directly measure the impact of the electric eld of the protein on the orienta-
tion of the water molecule. Rather than calculate the angle between the dipoles we calculate the angle between
the dipole and the in vacuo electric eld of protein at the center of the waters oxygen atom
cos(
i
)
R
surf
i
= p
i
E
protein
. (17)
Figure 15(b) shows the results of this calculation for myoglobin hydrated by 1900 water molecules. From this
we can see that the higher temperature water is able to get closer to the protein where it is strongly correlated
to the proteins electric eld. Within about 2.5

A, about one water layer, of the protein the water is strongly
correlated to proteins electric eld. Near the edge of the water sphere we see what appear to be surface eects
that distort the dynamics of these waters.
16
5.4.
::::::::::::::::
Fluctuations,
::::::::::::::
Deviations
:::::
and
:::::::::
Radius
:::
of
::::::::::::
Gyration
Thermal uctuations are necessary for protein function. How susceptible various parts of a protein are to these
uctuations determines in large part the what the function of the protein is. We can quantify these uctua-
tions by looking at the root-mean squared uctuations (RMSF) and deviations (RMSD). A further indicator of
conformational change within the protein is the radius of gyration.
RMSF are dened as
r
2
i

1/2
=
_
1
M
M

k=1
(r
i
(t
k
) r
i
)
2
_
1/2
(18)
where r
i
is average position of atom i over M congurations. This relates to the temperature B-factor, directly
measurable from experiment,
B
i
=
8
3

2
r
2
i
. (19)
This gives us a measure of the exibility of dierent parts of the protein that can be easily visualized as in
Figure 16.
Figure 16 demonstrates the temperature dependence of the RMSF. Thermal motion is generally conned
to the side chains of the amino acids while internal structure, such as -helices remain rigid even at room
temperature. Some parts of the backbone are relatively exible, such as the N and C-termini.
By averaging the RMSF for dierent groups of atoms we can see the eect of both temperature and hydration.
Hydration does not increase the exibility of the myoglobin back bone as shown in Figure 17(a) and even seems
to stabilize it at low temperatures. However, the protein side chains are signicantly aected by the addition
of water. Figure 17(b) shows that after being hydrated by at least 350 waters the exibility of the side changes
greatly increased to almost threes times that of the backbone. At low temperatures this is not the case.
The dierence between two structures is the RMSD and is given by
RMSD =
1
M
M

i=1
_
1
N
N

i=1
(r
A
i
r
B
i
)
2
_
1/2
(20)
where N is the number of atoms being compared between structures A and B. Figure 18 shows the time averaged
RMSD of myoglobin heavy atoms at 100 and 300 K compared to the crystal structure. The low temperature
17
(a) (b)
Figure 16. RMSF of myoglobin. scale 0.330000 3.150000 and 0.130000 1.04000Images created with VMD.
26
.
300K
100K
log
10
(# of waters)
r
m
s
f
(
A
)
4 3.5 3 2.5 2 1.5 1 0.5 0
0.65
0.6
0.55
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
(a)
300K
100K
log
10
(# of waters)
r
m
s
f
(
A
)
4 3.5 3 2.5 2 1.5 1 0.5 0
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
(b)
Figure 17. Average RMSF of myoglobin backbone (a) and side chains (b).
18
300K
100K
log
10
(# of waters)
R
M
S
D
(
A
)
4 3.5 3 2.5 2 1.5 1 0.5 0
2.6
2.4
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
Figure 18. RMSD of myoglobin heavy atoms.
structure is actually closer to the crystal structure. This should be expected since the crystal structure is was
taken at low temperatures and, since the high temperature structure has larger thermal uctuations, it should
naturally deviate more signicantly. Of interest is that in the 300 K case the deviations plateau after the addition
of 350 waters.
Another indicator of conformational change is the radius of gyration. This is dened as
R
gyr
=
_
I
m
=
_

m
i
r
2
i
m
. (21)
where I is the moment of inertia. This is an indicator of changes in size and shape as the moment of inertia. In
the case of myoglobin, being a globular protein, R
gyr
increases as the protein expands.
Figure 19 shows R
gyr
for the heavy atoms of myoglobin. This value oscillates considerably until hydrated by
at least 350 waters. At this points R
gyr
grows predictably, at strong indication of the minimum hydration being
reached at 350 waters/myoglobin. Once again a temperature dierence is obvious though volume contractions
at low temperature are expected.
A few other calculations that we could add but I havent yet done are surface area and volume changes with
hydration and the distance from the protein of water vs. the RMSF.
6. SUMMARY
This paper has been concerned with the behavior of the positional thermal uctuations of myoglobin over a
large temperature range, from 0 K to 300 K. The data that we analyzed came from both experiments on and
19
300K
100K
log
10
(# of waters)
R
g
y
r
(
A
)
4 3.5 3 2.5 2 1.5
15.4
15.3
15.2
15.1
15
14.9
14.8
Figure 19. Radius of gyration of myoglobin heavy atoms.
molecular dynamics simulations of myoglobin. The experiments on the vibrational dynamics of myoglobin used
a variety of techniques including the phonon-assisted M ossbauer eect, M ossbauer absorption spectroscopy and
inelastic scattering of synchrotron radiation. Several variants of myoglobin were investigated and the molecular
dynamics simulations of MbCO involved several levels of hydration. All these data sets, however, had one thing
in common, namely a pronounced signature of anharmonic uctuations above a threshold temperature of 200
K. Until now modeling eorts have stopped short of accounting for anharmonicity using non-Gaussian statistics
and this therefore was the objective of our paper. We have provided a relatively simple eective potential that
accounts for the two local minima corresponding to the two conformational states of the protein separated by
a potential barrier. The height of the potential dierence determines the value of the cross-over temperature.
The curvature of the potential around the ground state minimum denes the slope is the x
2
vs. T line at
low temperatures. Finally, anharmonicity in the form of a quartic term produces good qualitative agreement
with all the data sets analyzed. We have also found a very reasonable quantitative t to one particular case
which was selected (it had the most data points) to provide an example. We believe the method presented here
is suciently general to be applicable to the analysis of conformational eects in proteins. While molecular
dynamics simulations can provide an accurate tool in the quantication of the conformational energy landscape,
the non-Gaussian statistical analysis described in this paper can be used in the prediction of functional transitions
that are triggered by thermal uctuations.
20
Acknowledgments
This research has been supported by MITACS, NSERC and the Theoretical Physics Institute of the University
of Alberta. Part of this research was carried out with the support of a senior fellowship for J. A. T. at the
University of Leuven. We wish to thank Mr. W. Malinski for his assistance in data analysis. J. M. D. would
like to thank the sta and members of the Physics Department, University of Alberta, for all their kindness and
thoughtfulness during his stay.
REFERENCES
1. C. Branden and J. Tooze, Introduction to Protein Structure, Taylor and Francis, London, 2
nd
ed., 1999.
2. J. M. Berg, J. Tymoczko, and L. Stryer, Biochemistry, W. H. Freeman and Co., New York, 5
th
ed., 2002.
3. H. Frauenfelder, B. McMahon, and B. P. Stojkovi, Is myoglobin like a swiss watch?, in Proceedings of the
Adriatico Research Conference on Nonlinear Co-operative Phenomena in Biological Systems, L. Matsson,
ed., p. 145, World Scientic, 1998.
4. F. G. Parak and K. Achterhold, Protein dynamics studied on myoglobin, Hyperne Interactions 123,
pp. 825840, 1999.
5. K. Achterhold, C. Keppler, A. Ostermann, U. van B urck, W. Sturhahn, E. E. Alp, and F. G. Park,
Vibrational dynamics of myoglobin determined by the phonon-assisted Mossbauer eect, Phys. Rev. E
65(051916).
6. A. Ostermann, R. Waschipky, F. G. Parak, and G. U. Nienhaus, Ligand binding and conformational
motions in myoglobin, Nature 404, pp. 205208, 2000.
7. F. Parak, E. Knapp, and D. Kucheida, Protein dynamics. Mossbauer spectroscopy on deoxymyoglobin
crystals, J. of Mol. Biol. 161, pp. 177194, 1982.
8. F. G. Parak, Fluctuations and relaxations in proteins, in Proceedings of the First Biological Physics
Conference, Bangkok, Thailand, p. 7, World Scientic, Singapore, 2001.
9. G. U. Nienhaus, R. Waschiply, K. Nienhaus, O. Minkow, A. Ostermann, and F. G. Parak, Ligand migration
and binding in myoglobin mutant l29w, in Proceedings of the First Biological Physics Conference, Bangkok,
Thailand, p. 56, World Scientic, Singapore, 2001.
21
10. G. U. Nienhaus, J. Heinzl, E. Huenges, and F. Parak, Protein crystal dynamics studied by time-resolved
analysis of x-ray diuse scattering, Nature 338, p. 665, 1989.
11. J. A. Tuszynski, M. J. Clouter, and H. Kiefte, Beyond the Gaussian approximation: Some new results in
the Landau theory of phase transitions, Physics Letters A 108, pp. 272276, 1985.
12. J. A. Tuszynski and A. Wierzbicki, Non-Gaussian approach to critical uctuations in the Landau-Ginzburg
model and nite-size scaling, Phys. Rev. B 43, pp. 84728481, 1991.
13. P. J. Steinbach and B. R. Brooks, Hydrated myoglobins anharmonic uctuations are not primarily due to
dihedral transitions, 93, pp. 5559, 1996.
14. P. J. Steinbach and B. R. Brooks, Protein hydration elucidated by molecular dynamics simulation, 90,
pp. 91359139, 1993.
15. M. Ferrand, A. Dianoux, W. Petry, and G. Zaccai, Thermal motions and function of bacteriorhodopsin
in purple membranes: Eects of temperature and hydration studied by neutron scattering, PNAS 90,
pp. 96689672, 1993.
16. L. Cordone, M. Ferrand, E. Vitrano, and G. Zaccai, Harmonic behavior of trehalose-coated carbon-monoxy-
myoglobin at high temperature, Biophys. J. 76, pp. 10431047, 1999.
17. W. Doster, S. Cusack, and W. Petry, Dynamical transition of myoglobin revealed by inelastic neutron
scattering, Nature 337, pp. 754756, 1989.
18. F. Tama, O. Miyashita, A. Kitao, and N. Go, Molecular dynamics simulation shows large volume uctua-
tions of proteins, Eur. Biophys. J. 29, pp. 472480, 2000.
19. J. Howard, Mechanics of motor proteins and the cycloskeleton, Sinauer Associated Inc., 2001.
20. J. Kuriyan, S. Wilz, M. Karplus, and G. A. Petsko, X-ray structure and renement of carbon-monoxy (Fe
II)-myoglobin at 1.5

A resolution, J. Mol. Biol. 192, p. 133.
21. G. Vriend, What if: A molecular modeling and drug design program, J. Mol. Graph. 8, pp. 5256, 1990.
22. B. R. Brooks, R. E. Bruccoleri, B. D. Olafson, D. J. States, S. Swaminathan, and M. Karplus, Charmm: A
program for macromolecular energy, minimization, and dynamics calculations, J. Comp. Chem 4, pp. 187
217, 1983.
22
23. A. D. MacKerell, Jr., D. Bashford, M. Bellott, R. Dunbrack Jr., J. Evanseck, M. Field, S. Fischer, J. Gao,
H. Guo, S. Ha, D. Joseph-McCarthy, L. Kuchnir, K. Kuczera, F. Lau, C. Mattos, S. Michnick, T. Ngo,
D. Nguyen, B. Prodhom, W. Reiher, III, B. Roux, M. Schlenkrich, J. Smith, R. Stote, J. Straub, M. Watan-
abe, J. Wiorkiewicz-Kuczera, D. Yin, and M. Karplus, All-atom empirical potential for molecular modeling
and dynamics studies of proteins, J. Phys. Chem. B 102, pp. 35863616, 1998.
24. H. Berendsen, J. Postma, W. van Gunsteren, A. DiNola, and J. Haak, Molecular dynamics with coupling
to an external bath, J. Chem. Phys 81, pp. 36843690, 1984.
25. R. H. Stote, D. J. States, and M. Karplus, On the treatment of electrostatic interactions in biomolecular
simulation, J. Chim. Phys 88, pp. 21492433, 1991.
26. W. Humphrey, A. Dalke, and K. Schulten, Vmd - visual molecular dynamics, J. Molecular Graphics 14,
pp. 3338, 1996. http://www.ks.uiuc.edu/Research/vmd/.
23

Vous aimerez peut-être aussi