Chemical Physics Letters 388 (2004) 316–321

www.elsevier.com/locate/cplett

Slowdown of water diffusion around protein

in aqueous solution with ectoine
I. Yu, M. Nagaoka *

Graduate School of Human Informatics, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan
Received 5 December 2003; in final form 9 February 2004
Published online:

Abstract

Ectoine is one of the most common compatible solutes found in halophilic bacteria, and has an effect to introduce a tolerance to
high salt concentration or high temperature. By analyzing 1 ns molecular dynamics simulations at 370 K, we have shown that, in the
ectoine aqueous solution, the water diffusion slows down around a protein (chymotrypsin inhibitor 2 (CI2)), keeping the protein
hydration structure essentially unchanged. It is concluded that the slowdown of water diffusion around the backbone amide protons
must be one of the decisive factors in reducing the exchange rate of the backbone amide protons, whose reduction is experimentally
believed closely related to the tolerance effect.

2004 Elsevier B.V. All rights reserved.

1. Introduction Many experimental works have been performed to

Solvents play important roles not only on the stability
but also on the dynamics of protein conformation in
solution. Although there are a number of factors to alter
the solvent property in a cellular environment, it has
been recently well known that compatible solutes make
the protein conformation stabilized and maintain the
biological activity [1]. Compatible solutes are such
compounds with low molecular weights that are pro-
duced by the cells of organisms and are accumulated in
them in response to such high environmental stresses as
the high salt concentration or the high temperature. Due
to the lack of their strong and specific interaction with
protein surfaces [1], they can maintain water at the
surface where it is most needed. Although they do not
interact directly with macromolecules themselves, they
play an indirect role to alter the solvent properties and
to modify the stability of proteins [2–4].
* itor 2 (CI2) as a target protein. The temperature was set
Corresponding author. Fax: +81-52-789-5623.
E-mail addresses: mnagaoka@info.human.nagoya-u.ac.jp, mnagaoka
@is.nagoya-u.ac.jp (M. Nagaoka).
URL: http://frontier.ncube.human.nagoya-u.ac.jp/.
examine how the addition of compatible solutes to a
protein solution influences on its global thermodynamic
parameters [3–6]. The most striking manifestation of the
stabilization effect of compatible solutes is a consider-
able increase of the thermal unfolding transition tem-
perature (or the melting temperature) Tm of proteins in
their presence. However, its mechanism has still not
been fully investigated from a microscopic viewpoint,
such as where and how the compatible solutes influence
the property of water molecules.
For the purpose to investigate the microscopic be-
havior of solvent molecules around macromolecules, the
molecular dynamics (MD) simulation is one of the most
powerful theoretical tools. In this Letter, therefore, to
obtain information about the effect of compatible solutes
on the dynamic property of water molecules, we focus on
a molecule, ectoine, which belongs to the most common
compatible solutes, found in the cytosol of aerobic het-
erotrophic bacteria [6–8]. Then, we report some curious
results by a number of MD simulations performed for a
model ectoine-water mixture using chymotrypsin inhib-
to 370 K for the present simulations, to elucidate the
microscopic mechanism of ectoine concerning the pro-
tection of CI2 structure from the stress of heat.

0009-2614/$ - see front matter
2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.cplett.2004.02.104
 I. Yu, M. Nagaoka / Chemical Physics Letters 388 (2004) 316–321
From the analysis of trajectories of the simulations,
some characteristic differences in coordination of water
molecules and their dynamic behavior around the pro-
tein were found. The results were consistent with the
current explanation that the compatible solutes do not
interact with the protein surface strongly. In addition,
we have found one of the microscopic factors altering
the solvent property. From our analysis, it was indicated
that the ectoine molecules bring about the large slow-
down of water diffusion and should play an important
role on the stabilization of protein conformation.
This Letter is organized as follows: first, in Section 2,
the computational methods and the model systems are
explained. Numerical results and discussion are pro-
vided in Section 3. Finally, the present study is sum-
marized in Concluding remarks (Section 4).
2. Computational methods
Molecular mechanics (MM) and MD calculations
were all performed using AM B E R 7 program [9]. The
force field parameter set, parm99 [10], was used for all
the molecules in the present systems.
In our previous work [11], using GA U S S I A N 98 pro-
gram, a number of stable zwitter ionic electronic struc-
tures for ectoine were obtained by geometry optimization
at the HF/6-31G* and MP2/6-31G* levels of theory.
Then, two zwitter ionic structures (i.e., ECA and ECB
[11]), which have almost the same stability, were rea-
sonably obtained. In the present work, those partial
atomic charges and the geometry of ECA in which the
COO group is in the axial position, was used to param-
eterize an ectoine molecule. On the other hand, the
starting protein conformation of CI2 was the X-ray
derived crystal structure (2CI2) [12].
One MD simulation for a system of CI2 in water–
ectoine mixture was performed in the following way.
First, a CI2 unit was set, under the periodic boundary
condition, at the center of a cubic box (54 · 54 · 54 A3 )
filled with 4002 TIP3P [13] water molecules and 119 ec-
toine molecules (this simulation is named MD1). Second,
to investigate the influence of the ectoine addition on the
solvent water property, another type of MD simulation
for a system of CI2 in pure TIP3P water was separately
performed (i.e., MD2). Both systems were initially
equilibrated for the first 100 ps duration at 370 K under
the NPT condition and, then, a further simulation of
1000 ps was performed at 370 K under the NVE condi-
tion. Trajectories were numerically integrated by the
Verlet method with a time step of 2.0 fs. The electrostatic
interaction was calculated using the particle-mesh Ewald
(PME) method [14]. All the bonds involving hydrogen
atoms were constrained by the SHAKE method [15]. For
the equilibration, both temperature and pressure were
regulated using the Berendsen algorithm [16].
317
In this work, in order to consider how the concen-
tration of ectoine might influence on the dynamic
property of water molecules, other three sets of MD
simulations (i.e., MD3, MD4 and MD5) were also per-
formed without a CI2 system in three different concen-
trations of ectoine. First, MD3 contains one ectoine and
1009 water molecules, second, MD4, 64 ectoine and 984
water molecules and, third, MD5 is the same as MD4
except those partial atomic charges that are set to 0 for
all the atoms in ectoine. After the starting structures
were minimized molecular mechanically for 1000 cycles
to reduce any bad contact, for all the three systems
having been equilibrated at 370 K and 1 atm under the
NPT condition, three successive 100 ps of NVE simu-
lations were executed at the neutral pH condition,
respectively.
3. Result and discussion
3.1. Slowdown of water diffusion
In order to know some characteristic change in dy-
namic properties of solvent water, we investigated the
diffusion constant as one of the most direct dynamic
properties. Then, it was clearly found that ectoine brings
about the slowdown of water diffusion. Table 1 shows
the diffusion constants of water molecules (Dwat ) and
those of ectoine molecules (DECA ) in those three systems
explained in the previous section, i.e., MD3, MD4 and
MD5, respectively. From the comparison of Dwat values,
12.79 · 10
5 cm2 /s in MD3 and 6.96 · 10
5 cm2 /s in
MD4, it is indicated that the water diffusion decreased
with the number increase of ectoine molecules.
However, taking into consideration such a fact that
the Dwat value 9.73 · 10
5 cm2 /s in MD5 with the ectoine
molecules with no charge did not show so much large
reduction in Dwat as that in MD4, it is understood,
therefore, that charges in ectoine have an essential effect
to reduce the mobility of water molecules. Since ectoine
has a zwitterionic form in water, it might indicate also
that the electrostatic interaction between water and ec-
toine molecules must play a very important role for this
slowdown effect of water diffusion. On the other hand, it
was understood, in our previous work, that an ectoine
molecule could be hydrogen bonding with more than
Table 1
Diffusion constants of water molecules (Dwat ) and ectoine molecules
(DECA ) in MD3, MD4 and MD5, respectively
Numbers of molecules Dwat DECA
(10
5 cm2 /s)
MD3 Water 1009:ECA 1 12.79 5.41
MD4 Water 984:ECA 64 6.96 1.95
MD5 Water 984:ECA (no charge) 64 9.73 2.67
318 I. Yu, M. Nagaoka / Chemical Physics Letters 388 (2004) 316–321
four water molecules simultaneously [11]. Thus, a
number of hydrogen bonds between ectoine and water
molecules must contribute directly to reduce the water
mobility especially around ectoine molecules.
In addition, in Table 1, the diffusion constants of
ectoine molecule DECA are found much smaller than
those of water molecule Dwat in all the three systems.
Moreover, the smaller DECA value 1.95 · 10
5 cm2 /s in
MD4 indicates that the electrostatic interaction among
ectoine molecules has an effect to reduce their mobility
themselves. For this reason, those water molecules in-
teracting with such low-mobility ectoine molecules
should also reduce their mobility directly.
Hence, the slowdown effect of ectoine on the water
diffusion is clearly verified by the analysis of diffusion
constants numerically calculated via the present MD
simulations.
3.2. Coordination of water molecules on the surface of
protein
It is well known that water molecules near the protein
surface should play an important role to maintain the
folding structure of protein. To investigate the effect of
ectoine for the solvent structuring near the protein sur-
face, we have analyzed the integrated coordination
number hNCI2 ðrÞi of constituent atoms of water mole-
cules and ectoine molecules surrounding CI2 (Fig. 1).
The integrated coordination number hNCI2 ðrÞi is pres-
ently defined as
Z nation number of ectoine atoms near the surface of
1 T
hNCI2 ðrÞi ¼ NCI2 ðr; tÞ dt; ð1Þ
T 0
1400
Water (MD1)
1200 Water (MD2)
ECA (MD1)
1000
N CI2 (r)
800
600
400
200
0
1 2 3 4
r (Å)
Fig. 1. Integrated coordination numbers of atoms in water molecules
 from the closest atom in CI2 in MD1 (solid
within the distance r (A)
line) and MD2 (dashed line), and that in ectoine molecules in MD1
(dash-dotted line), respectively.
X
N
NCI2 ðr; tÞ ¼ hðr
qCI2 ðri ðtÞÞÞ; ð2Þ
i¼1
qCI2 ðri Þ ¼ min jri
ra j; ð3Þ
a2CI2
where hðxÞ is a step function as follows:


1 ðx P 0Þ;
hðxÞ ¼ ð4Þ
0 ðx < 0Þ:
Then, hðr
qCI2 ðri ðtÞÞÞ takes 1 if the solvent atom i ex-
ists within a distance r A from the closest CI2 atom, i.e.,
if r
qCI2 ðri Þ > 0, where qCI2 ðri Þ is defined as the mini-
mum distance between the solvent atom i and any CI2
atom a. In this analysis, without distinction on the dif-
ferent types of atom in each solvent molecule, they were
counted in hNCI2 ðrÞi.
In Fig. 1, it was found that hNCI2 ðrÞi of atoms in
water molecules starts to increase at about 1.6 A  in both
MD1 and MD2 and keeps almost the same value (ca.
300) up to 2.5 A,  while that of atoms in ectoine mole-
cules start to increase at a larger distance 2.0 A.  This
observation means that the atoms in ectoine molecules
do not interact so much strongly with those in CI2 as
those atoms in water molecules by their being excluded
from the CI2 surface. In addition, in the region within
 which corresponds to the radius of the first sol-
3.0 A,
vation shell of hydrogen atoms in CI2, there exist only
80 ectoine atoms, i.e., the total number of atoms in four
ectoine molecules. Then, it is indicated that the coordi-
protein is very small in comparison to that of other
solvent atoms.
To investigate the orientation of solvent atoms in
detail, we analyzed the radial distribution functions
(RDFs) of oxygen atoms in water molecules (OW ) and
of the oxygen atoms of COO group in ectoine (OE1 ),
around backbone amide protons (HN ) for all residues of
CI2. This analysis provides the qualitative information
about the number density of OW and OE1 around HN
relative to that of the bulk phase in each system. As a
result, in RDFs of OW for most of residues, there is no
significant difference not only in the distance but also in
the intensity of the first and second peaks. From the
RDFs of all the residues, for example, in Fig. 2a–d, we
have shown four RDFs of OW and OE1 around HN in
four main regions composing CI2; (a) GLU8 in the
N-terminal strand, (b) ILE21 in the a-helix, (c) THR40
in the active site loop, and (d) PHE51 in the b-sheet,
respectively.
Although, for many residues, the first peaks in their
RDFs of OE1 appear at the outside of those of OW , it
was noticed that ectoine molecules are more strongly
excluded especially from the b-sheet region (Fig. 2d),
which is buried in the protein structure, and from the
a-helix region (Fig. 2b), respectively. A few COO
 I. Yu, M. Nagaoka / Chemical Physics Letters 388 (2004) 316–321 319

2. 5 2.5

(a) GLU8 OW (MD1) (b) ILE21

2. 0 2.0
OW (MD2)
OE1 (MD1)
1. 5 1.5

g(r)
1. 0 1.0

0. 5 0.5

0. 0 0.0
5 10 15 20 5 10 15 20

2. 5 2. 5

(c) THR40 (d) PHE51

2. 0 2. 0

1. 5 1. 5
g(r)

1. 0 1. 0

0. 5 0. 5

0. 0 0. 0
5 10 15 20 5 10 15 20

r (Å) r (Å)

Fig. 2. Radial distribution functions (RDFs) of the oxygen atoms in water molecules (OW ) in MD1 (solid line) and MD2 (dashed line) and of the
oxygen atoms of the COO group in ectoine (OE1 ) (dash-dotted line) around backbone amide protons (HN s) in: (a) GLU8 in the N-terminal strand,
(b) ILE21 in the a-helix, (c) THR40 in the active site loop, and (d) PHE51 in the b-sheet, respectively. (The structure of the first 18 residues in 2CI2 is
not resolved [12]. Therefore, we renumbered all the residues such that the protein begins at residue 1 instead of 19.)

groups in ectoine molecules sometimes form hydrogen

bonds with the hydrogen atoms in the polar side chain
of CI2. For this reason, there are some RDFs of OE1 ,
 from
which show a relatively strong peak at about 5.0 A
HN (Fig. 2c). However, the average number of such
COO groups is about 6. It is, therefore, concluded that:
(i) ectoine molecules do not essentially change the
structure of the first hydration shell of CI2 and (ii) the
direct interaction between ectoine and protein is not so
strong.

3.3. Dynamic behavior of water molecules around protein

In this section, we examine the effect of ectoine on the
dynamic behavior of water around CI2. In order to
elucidate this effect, we have analyzed the diffusion of
OW around HN in MD1 (the system with ectoine) and
MD2 (the system without ectoine). Fig. 3 shows a stereo
view of water molecules detained in the first solvation
shells of HN in CI2. During the entire time duration of
our simulations, 10–20 OW s are detained simultaneously
in these shells in both systems and coordinate mainly to
HN s in the exterior surface of CI2.
Fig. 4b, d shows the mean square displacements
(MSDs) of such water molecules that their OW s initially
coordinated in the first solvation shells of HN . Each first
solvation shell was defined as a sphere of radius 3.0 A 
around each HN of CI2, respectively. In both systems,
Fig. 3. Stereo view of the water molecules detained in the first solvation
shell of backbone amide protons in CI2 (CI2: ribbon, water: space
filling, ECA: stick).
MD1 and MD2, about 500 of water molecules were
tagged during the entire simulations and their square
displacements from the starting positions were time-
dependently averaged over the ensemble. This analysis
provides information about how fast the water mole-
cules would diffuse from their initially coordinated HN s.
It was found in MD1 that, the MSD of water molecules
after their coordination in the first solvation shells of HN
(Fig. 4b), shows a large reduction from that of all the
water molecules in this system (Fig. 4a). On the con-
trary, in MD2, such a large reduction was not observed
(Fig. 4c, d).
320 I. Yu, M. Nagaoka / Chemical Physics Letters 388 (2004) 316–321
1500
MD1
1100
MSD (Å2) 700
(a)
300
(b)
- 100
0 50 100
t (ps)
Fig. 4. Mean square displacements (MSDs) of all the water molecules in (a) MD1 and (c) MD2 and of water molecules after the coordination in the
first solvation shell of backbone amide protons (HN s) in (b) MD1 and (d) MD2, respectively.
These water molecules in the first solvation shell can
be classified mainly into three groups: (i) such water
molecules that are just crossing the solvation shells of
HN , (ii) those on the exterior surface of protein, which
weakly interact with HN , and (iii) those in the interior of
protein, which strongly interact with HN s. Among these
three types of water molecules, it is understood that
there should be a large deviation in mobility and, then,
the error bars for MSDs, shown every 25 ps in Fig. 4,
might result in rather large values. For each group of
water molecules, (i) the increase of residence time in the
solvation shells, (ii) the slowdown of diffusion after their
escape from the solvation shells, and (iii) the frequency
increase of revisiting into the same or adjacent solvation
shells, must be three predominant factors for the MSD
reduction in MD1 (Fig. 4b) from that in MD2 (Fig. 4d).
However, since the most slowly exchanging water
molecules have a residence time of more than 100 ps, in
order to analyze exactly how much these factors might
influence on the reduction of MSD by each residue in
CI2, further simulations with much longer time duration
might be necessary to acquire an enough statistical ac-
curacy. Although the larger reduction of MSD in
Fig. 4b from that in Fig. 4d should be brought about by
a combination of some of the above factors, our analysis
could indicate that the stress of heat, which should bring
about a faster diffusion of water molecules from the
Table 2
Diffusion constants of all the water molecules (Dall ) in MD1 and MD2,
and those of water molecules around HN (DHN )
Dall DHN ðDall
DHN Þ=Dall
(10
5 cm2 /s) (10
5 cm2 /s) (%)
MD1 9.62 4.80 50.10
MD2 11.93 9.32 21.88
Reduction rates of diffusion constant DHN relative to Dall , i.e.,
ðDall
DHN Þ=Dall .
1500
MD2
1100
(c)
MSD (Å2)
700
(d)
300
- 100
150 0 50 100 150
t (ps)
surface of protein, is weakened in the slowly moving
ectoine-water binary phase.
Finally, by the present analysis, we have found that
ectoine molecules make the diffusion of water slow not
only in bulk phase but also near the surface of protein.
Diffusion constants for all the water molecules (Dall ) in
both MD1 and MD2 systems are obtained by the slopes
of MSD in Fig. 4a, c, while those of only water mole-
cules (DHN ) are from the average slopes of MSD until
t ¼ 50 ps in Fig. 4b, d. Their numerical values are ob-
tained in Table 2 with the reduction rate (%) of the
diffusion constant DHN relative to Dall , i.e., ðDall

DHN Þ=Dall . In the presence of ectoine (MD1), the diffu-
sion constant DHN is reduced to 50% smaller value in
comparison to Dall , while that in the absence of ectoine
(MD2) is found to only 20%.
4. Concluding remarks
In this work, the influence of compatible solutes on
the hydration structure and the dynamic property of
water molecules near the protein surface, were discussed
at the molecular level. We have found that ectoine does
not change the hydration structure essentially but slows
down the diffusion of water molecules near the protein
surface in addition to in the bulk phase. The results are
consistent with the current explanation that compatible
solutes do not interact strongly with the macromolecule
but act to alter the solvent properties [2–4].
Previously, Foord and Letherbarrow [2] measured the
hydrogen exchange rates of backbone amide protons
(HN ) of CI2 at 300 K using NMR in the presence of 2 M
glycine, which is also the most common zwitter ionic
compatible solute. They found that the presence of
glycine causes a large reduction of the hydrogen ex-
change rates without any significant change in the 3D
structure of CI2. The rate of exchange for individual
 I. Yu, M. Nagaoka / Chemical Physics Letters 388 (2004) 316–321
amide protons with those of the solvent, depends on
such various factors as; (i) the local fluctuation of the
residue that exposes HN to the solvent, (ii) the intensity
of interaction between HN and water molecules coor-
dinated to HN , and (iii) the effect of adjacent residues,
and so on. Thus, it is true that the amide proton ex-
change might be a complex phenomenon and might not
be explained by a single and simple mechanism. How-
ever, it is concluded presently that the slowdown of water
diffusion around backbone amide protons must be one of
the decisive and important factors in reducing the ex-
change rate of the backbone amide protons. For further
understanding the correlation between the slowdown
and the rate reduction, more quantitative analysis is
strongly expected not only computationally but also
experimentally.
The unfolding time for CI2, which is obtained by
extrapolating the experimental data [17] is 9 · 10
2 s at
360 K [18]. On the other hand, it is 4 · 10
8 s (i.e.,
40 ns) at 373 K in the previous MD simulation [19]. It
may be reasonable, therefore, that any large difference
between 3D structure of CI2 in MD1 and that in MD2
should not appear in such a short time duration as that
in our simulations, i.e., 1 ns. However, the effect of ec-
toine found in our analysis, i.e., the slowdown of water
diffusion near the protein surface, must be one of the
important factors to reduce the dynamic fluctuations of
each residues in protein. Furthermore, in addition to
this kinetic stabilization effect, such slowdown should
increase the surface tension of the first hydration layer
on the protein surface and, as a result, this should cause
some compact conformation like the native state to be
favored. In other words, the slowdown of water diffu-
sion on the surface of protein should be the microscopic
origin, which is connected to the kinetic and thermo-
dynamic stabilization of the 3D structure of protein
against the stress of heat or high salt concentration. To
investigate the correlation between solvent property
changes and the kinetic and thermodynamic stability of
protein, a more detailed analysis are now in progress.
321
Acknowledgements
This work was supported partly by a Grant-in-Aid
for Science Research from the Ministry of Education,
Culture, Sport, Science and Technology in Japan and
also by the Research and Development Applying Ad-
vanced Computational Science and Technology from
the Japan Science and Technology Agency (Project
ACT-JST-13B-2-a).
References
[1] P.H. Yancey, M.E. Clark, S.C. Hand, R.D. Bowlus, G.N.
Somero, Science 217 (1982) 1214.
[2] R.L. Foord, R.J. Leatherbarrow, Biochemistry 37 (1998)
2969.
[3] T. Arakawa, S.N. Timasheff, Biophys. J. 47 (1985) 411.
[4] I.M. Plaza del Pino, J.M. Sanchez-Ruiz, Biochemistry 34 (1995)
8621.
[5] M.M. Santoro, Y. Liu, S.M.A. Khan, L.-X. Hou, D.W. Bolen,
Biochemistry 31 (1992) 5278.
[6] S. Knapp, R. Ladenstein, E.A. Galinski, Extremophiles 3 (1999)
191.
[7] E.A. Galinski, Experientia 49 (1993) 487.
[8] E.A. Galinski, Adv. Microb. Physiol. 37 (1995) 273.
[9] D.A. Case et al., AM B E R -7, University of California, San
Francisco, 2002.
[10] J. Wang, P. Cieplak, P.A. Kollman, J. Comput. Chem. 21 (2000)
1049.
[11] K. Suenobu, M. Nagaoka, T. Yamabe, S. Nagata, J. Phys. Chem.
A 102 (1998) 7505.
[12] C.A. Mcphalen, M.N.G. James, Biochemistry 26 (1987) 261.
[13] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey,
M.L. Klein, J. Chem. Phys. 79 (1983) 926.
[14] T. Darden, D. York, L. Pedersen, J. Chem. Phys. 98 (1993)
10089.
[15] J.P. Ryckaert, G. Ciccotti, H.J.C. Berendsen, J. Comput. Phys. 23
(1977) 327.
[16] H.J.C. Berendsen, J.P.M. Postma, W.F. van Gusteren, A. Dinola,
J.R. Haak, J. Chem. Phys. 81 (1984) 3684.
[17] S.E. Jackson, A.R. Fersht, Biochemistry 30 (1991) 10436.
[18] A. Li, V. Daggett, J. Mol. Biol. 257 (1996) 412.
[19] R. Day, B.J. Bennion, S. Ham, V. Daggett, J. Mol. Biol. 322
(2002) 189.