Preferential solvation of lysozyme in water/ethanol mixtures

Abstract

We provide a quantitative description of the solvation properties of lysozyme in water/ethanol mixtures, which has been obtained by a simultaneous analysis of small-angle neutron scattering and differential scanning calorimetry experiments. All data sets were analyzed by an original method, which integrates the exchange equilibrium model between water and ethanol molecules at the protein surface and activity coefficients data of water/ethanol binary mixtures. As a result, the preferential binding of ethanol molecules at the protein surface was obtained for both native and thermal unfolded protein states. Excess solvation numbers reveal a critical point at ethanol molar fraction ≈0.06, corresponding to the triggering of the hydrophobic clustering of alcohol molecules detected in water/ethanol binary mixtures.

Full text

Preferential solvation of lysozyme in water/ethanol mixtures
Maria Grazia Ortore,1Paolo Mariani,1Flavio Carsughi,2, 3 Stefania Cinelli,4Giuseppe
Onori,4Jos´e Teixeira,5and Francesco Spinozzia)1 , b)
1)Department of Life and Environment Sciences, Marche Polytechnic University and
CNISM, I-60131 Ancona, Italy
2)Department of Agricultural Food and Environmental Sciences,
Marche Polytechnic University and CNISM, I-60131 Ancona,
Italy
3)Forschungszentrum J¨ulich, J¨ulich Centre for Neutron Science, D-85747 Garching,
Germany
4)Department of Physics, University of Perugia and CEMIN, I-06123 Perugia,
Italy
5)Laboratoire L´eon-Brillouin, F-91191 Gif-sur-Yvette, France
(Dated: 24 November 2011)
We provide a quantitative description of the solvation properties of lysozyme in wa-
ter/ethanol mixtures, which has been obtained by a simultaneous analysis of Small-
Angle Neutron Scattering and Differential Scanning Calorimetry experiments. All
data sets were analyzed by an original method, which integrates the exchange equi-
librium model between water and ethanol molecules at the protein surface and activ-
ity coefficients data of water/ethanol binary mixtures. As a result, the preferential
binding of ethanol molecules at the protein surface was obtained for both native and
thermal unfolded protein states. Excess solvation numbers reveal a critical point at
ethanol molar fraction ≈0.06, corresponding to the triggering of the hydrophobic
clustering of alcohol molecules detected in water/ethanol binary mixtures.
PACS numbers: 87.15.kr
Keywords: Small Angle Neutron Scattering; Differential Scanning Calorimetry;
Lysozyme; Preferential hydration; Global fit analysis
a) Corresponding author
b)Electronic mail: f.spinozzi@univpm.it
1

I. INTRODUCTION
It is widely acknowledged that the structure and dynamics of proteins are strongly in-
fluenced by interactions with water1,2. Measurements on the properties of water molecules
associated with a protein, and particularly on their changes after conformational transitions
and binding reactions, are then considered fundamental for furthering the understanding of
protein hydration.
When a protein is dissolved in a mixed aqueous solution its structural and dynamic
properties change as a function of the solvent composition3, according to a preferential
solvation process, which accounts for the accumulation or reduction of the different solvent
molecules at the protein surface4,5. It has been demonstrated, for example, that a co-solvent
can act as plasticizer or stabilizer, allowing or blocking the protein to jump between the so-
called conformational substates6, or to play as modulator for biochemical reactions7. Also
the perspective of being able to regulate the characteristics of a protein by changing the
physical and chemical environment in which it is dissolved has a relevant practical interest
and noticeable consequences.
The molecular structural characterization of the protein solvation shell in a binary mix-
ture appears very problematic due to sensitivity requirements for detecting possible small
modifications of solvent composition at the protein surface. Moreover, until a few years ago,
molecular dynamics simulations to study this issue were scarcely applied8.
On the other hand, due to a fine tuning of the scattering contrast between protein, solu-
tion, and solvation shell, we have recently showed that the Small Angle Neutron Scattering
(SANS) technique can be an optimum tool to experimentally obtain the composition of the
solvation shell of a protein dissolved in binary solvents9–11. In particular, we demonstrated
that lysozyme is preferentially hydrated when dissolved in a water/glycerol mixture, in full
agreement with previous literature indications obtained at infinite protein dilution12, but
depleted of water molecules when dissolved in a water/urea mixture10, again in agreement
with previous data13. In water/glycerol mixtures one result is specially noticeable: the ex-
cess solvation number for water (e.g., the number of water molecules around a lysozyme in
solution, in excess - augmentation or depletion - to the number of water molecules contained
in a volume of bulk solvent equivalent to the solvation shell volume) reported as a function
of the water molar fraction in the solvent, xw, shows a maximum at around 0.69. Using Dif-
2

ferential Scanning Calorimetry (DSC) studies, a clear relationship between the preferential
hydration and the lysozyme stability was then established14, as the stability of lysozyme in
water/glycerol mixtures at 298 K was observed to be maximal exactly at xw= 0.6.
Among other binary mixtures, monohydric alcohols/water solutions have largely been
investigated. In particular, it has been clearly established that monohydric alcohols, such
as methanol, ethanol and 1-propanol, destabilize the native structure of proteins, although
the promotion of the α-helix conformation in unfolded proteins and peptides15 and the sta-
bilization of the native state for diluted solutions16 have been also reported. In particular,
ethanol has been demonstrated to modify the amyloidogenic self-assembly of insulin 17. A
large number of experimental18–21, theoretical22 and molecular dynamics studies23–25 have
been focused on the structure of the molecular clusters formed by alcohol in water. As a
result, micro-heterogeneous structures of alcohol and water in binary mixtures have come
to be generally accepted, and their relationship with protein stability has been highlighted.
In particular, the hydrophobic clustering of ethanol molecules in a bulk aqueous solution
has been experimentally investigated through detailed measurements of compressibility and
frequency of C−H stretching18,26. Such investigations have shown that ethanol is essen-
tially monomeric for an ethanol molar fraction in the solvent, xe, lower than 0.06, while
self-associates in the range 0.06 < xe<0.29. Correlations between the microstructure of
the binary solvent and the conformation of a protein therein dissolved have been demon-
strated by studying the thermal denaturation of lysozyme in ethanol-water mixtures27. In
particular, the transition enthalpy, ∆H, has been observed to depend in a complex manner
on the amount of alcohol present in the solution: as ethanol concentration is increased, ∆H
increases until a maximum value is reached at a mole fraction x∗
eof about 0.06. A maximum
at the same mole fraction was also detected in the denaturation entropy change, ∆S.
Actually, proteins appear as sensible probes for monitoring solution properties and their
complicated evolution with composition. Since existing results collectively indicate that
the conformational stability of a protein is closely related to the features of the solvation
shell, a complete description of the phenomenon requires considering the properties of the
plain alcohol-water mixtures and the thermodynamic characteristics of the protein-solvent
interactions.
The aim of the present study is to discuss the properties of the water/ethanol binary
solution comparing SANS and DSC results on lysozyme dissolved in the mixture. Experi-
3

ments were performed in parallel on similar samples, and the whole experimental data were
analyzed using a global fit procedure, an approach that we have demonstrated to be very
efficient for extracting tiny structural details from a wide experimental context9,10,14. By
adopting Schellman’s thermodynamic model28,29, based on an equilibrium exchange between
water and ethanol in the bulk and in the solvation layer, we are able to estimate, for both
the native and the denatured protein states, the thermodynamic equilibrium constant and
related thermodynamic parameters, such as the excess solvation numbers. Since data were
evaluated as a function of solvent composition, correlations with the water/ethanol binary
mixture properties were also derived.
II. MATERIALS AND METHODS
A. Sample preparation
Hen egg-white lysozyme (Fluka Chemie AG) was dissolved in 0.1 M HCl/glycine wa-
ter/ethanol solutions at pH 3.0. The lysozyme concentration was checked spectrophoto-
metrically by using an extinction coefficient of ε1%
1 cm = 26.9 at 280 nm. 10 samples were
investigated by DSC, with lysozyme concentration fixed to 20 mg/mL and ethanol molar
fraction in the solvent, xe, ranging from 0 to 0.09. 16 samples were studied by SANS: for
each lysozyme concentration equal to 30, 60, 90, and 120 mg/mL, the values of xewere 0,
0.03, 0.06 and 0.09. All the samples were prepared at a deuteration grade xDequal to 0.88,
according to the procedure described in the supplementary material at30.
B. SANS Experiments
SANS measurements were carried out at the Laboratoire Leon Brillouin, France, on the
PAXE diffractometer. Two sample-detector distances (1 and 3 m) and 7 ˚
A wavelength
neutrons were used. The investigated exchanged wave vector modulus qranged from 0.015
to 0.3 ˚
A−1. Samples were measured at room temperature in a 1 mm thick quartz cell.
Experimental detector counts were radially averaged and corrected for sample transmis-
sion, detector inhomogeneities and scattering from buffer, and converted into macroscopic
differential scattering cross section in absolute units (cm−1) by direct beam measurements31.
4

C. DSC Data
The DSC experiments were performed on a micro-DSC II (Setaram, France) at a scan
rate of 18 Kh−1with sample masses of 0.85 g. More details can be found in Ref.27, where
these data were already published.
III. THERMODYNAMIC MODEL AND DATA ANALYSIS
In brief, SANS and DSC results obtained on lysozyme dissolved in water/ethanol binary
solutions were analyzed using a global fit procedure based on the mixed solvent exchange
thermodynamic model. In the following, the thermodynamic model as well as the SANS
and DSC data analysis will be detailed. Here, it should be noticed that few basic assump-
tions have been made, i.e. the lysozyme structure was considered unchanging in the range
where water/ethanol ratio is varied and the unfolding process has been assumed to be co-
operative and two-state at all the investigated conditions. Both assumptions are consistent
with the low alcohol concentration used (less than 25% v/v) and are supported by previ-
ous results. First, in ref.32 the observed slight variations of lysozyme hydrodynamic radius
with ethanol mole fraction between 0 and 0.08 were attributed to the effect of alcohol in
modulating solvent-mediated interactions and not on protein structural changes. Second, in
ref.33 data analysis demonstrated that the increase of the osmotic second virial coefficient
in lysozyme water-ethanol solutions (with xeranging from 0.002 to 0.072) could be fully
understood considering a Mean Force Potential for “globular proteins of constant radius”.
Third, the unfolding of lysozyme dissolved both in D2O and CH3CH2OD/D2O was studied
by Fourier Transform Infrared (FTIR) absorption spectroscopy at different protein concen-
trations (xe=0.15, 30 and 120 mg/mL lysozyme concentrations)34. The detailed description
of the local and global rearrangements was compatible with a cooperative and two-state
process at both the considered sample compositions.
A. Solvent exchange thermodynamic model
As shown in recent papers9–11,14, the number of water molecules adsorbed or released
from a protein surface, as a result of modifications on the composition of the binary mix-
ture where the protein is dissolved, can be experimentally determined both from SANS and
5

DSC data. In particular, the mixed solvent exchange thermodynamic model formulated
by Schellman28, which describes the protein solvation process as a thermodynamic equilib-
rium between solvent molecules in contact with the protein surface and in the bulk, has
been demonstrated to be appropriate. By using a global fit strategy, Schellman’s thermody-
namic equilibrium constant appears as the main data fitting parameter, responsible for the
protein shell structural, compositional and energetic features in all the set of investigated
experimental conditions, which differ for protein and co-solvent concentrations11.
The model can be summarized in this way. The protein surface is considered to provide
msites, which can be occupied either by water or ethanol molecules, and none of them can
be unoccupied. The water and ethanol molecules in close vicinity to the protein define the
“local domain” region l, which includes the solvent molecules filling the gaps between the
protein bound ethanol molecules (see Fig. 4 in Ref.9). The composition of the local domain,
represented by the ethanol molar fraction xe,l, can be different from the one of the bulk
solvent, xe,b. The ethanol-water exchange equilibrium taking place at msites can be written
as:
es+wb⇋eb+ws(1)
where es,eband ws,wbrepresent ethanol (e) and water (w) molecules in the bulk phase (b)
and in direct contact with the protein surface (s), respectively. Note that domain scontains
only water and ethanol molecules in contact with the protein surface and then is a non-
uniform shell around the protein. Considering the probability φthat msites are occupied
by water molecules, the thermodynamic exchange equilibrium constant can be translated
into the equation
Kex =φ
1−φ
ae
aw
(2)
where ae=γexe,b and aw=γw(1 −xe,b) are the activities of ethanol and water in the
bulk domain. The corresponding activity coefficients, γeand γw, have been expressed as
analytical functions of xe,b, according to excess free energy data35.
Two important aspects of Eq. 2 merit to be considered. First: since it is well known that
water-ethanol mixtures differ in a relevant way from ideality, the thermodynamic availability
of both water and ethanol molecules in the bulk domain is properly described by activities
instead than by molar fractions28.Second: at our knowledge, it is not possible to design
experiments by which to measure activity coefficients of molecules bound on a protein sur-
6

face. Hence we can only quantitatively express the thermodynamic availability of water and
ethanol molecules bound to the lysozyme surface by the simple water occupation probability
φof the msites.
Eq. 2 is then central in the global fit strategy adopted here. However, as DSC data
contain information on lysozyme unfolding, two different thermodynamic exchange equilibria
were considered in the calorimetric data analysis14, the first related to the solvent exchange
between the bulk and the native protein surface (described by the thermodynamic exchange
equilibrium constant Kex
N) and the second related to the solvent exchange between the bulk
and the unfolded protein surface (described by the thermodynamic exchange equilibrium
constant Kex
U). In this framework, the SANS experiments performed on native lysozyme
enhanced the information concerning the first thermodynamic equilibria already provided
by DSC data.
B. SANS data analysis
For monodisperse and randomly oriented protein particles dissolved in a homogeneous
solvent, the SANS macroscopic differential scattering cross section can be expressed as
dΣ
dΩ(q) = npP(q)SM(q) + B, (3)
where npis the protein number density, SM(q) the effective (measured) structure factor
defined by
SM(q) = 1 + [F(q)]2
P(q)[S(q)−1] (4)
in which S(q) is the protein-protein structure factor, P(q) the form factor, and F(q) the
angular average of the Fourier transformation of the density distribution of a particle36.
Finally, Bis a flat background, which accounts for incoherent scattering effects.
In the present case, the protein-protein structure factor S(q) was modeled on the basis
of a three-terms pair interaction potential, u(r) = u0(r) + u1(r) + u2(r), which includes a
hard sphere potential,
u0(r) = 


∞r≤2R
0r > 2R(5)
and two Yukawian terms, namely a repulsive coulombic screened potential, u1(r), and an
attractive potential, u2(r), both of the form uj(r) = Ajexp[−κj(r−2R)]/r 9,11. In these
7

expressions, Rrepresents the effective hard-sphere protein radius. For the coulombic term,
the constant A1depends on the net protein charge Z, on the bulk dielectric constant ε
(which is a known function of the composition of the bulk, xb,e ) and on the inverse Debye
length κ1(which is usually written as a function of Zand the ionic strength ISdue to
all the microionic species in solution, as in eq. 11 in ref. 37 ). Concerning the attractive
Yukawian term, the pre-exponential constant is written as A2=−2R J, where Jis the
so-called potential at protein-protein contact, whereas the constant κ2is simply written as
the inverse of the decay length d.
The calculaton of S(q) from the potential u(r) requires the solution of the Ornstein-
Zernike (OZ) equation under a proper choice of a closure. A very efficient closure method for
a two-Yukawa potential, based on a numerical solution of the mean-spherical approximation
(MSA), originally developed by Liu, Chen, and Chen 38 , was successfully applied in the
analysis of SANS data of proteins in a wide range of concentration38–41. These studies
have shown that the subtle balance between attractive and repulsive potentials leads to a
transition from a monomer fluid to a cluster fluid at increasing protein concentration. In
particular, the appearence of a peak in the structure factor whose position does not scale with
the n1/3
p, as expected for a fluid of not interacting monomers, suggests an intermediate range
order, as it has been recently discussed in many experimental and theoretical works41–43.
In this work, the strong repulsion regime due to the acidic conditions of our samples led
us to choice the simple and analytical Random-Phase Approximation (RPA), where S(q) is
expressed as a perturbation of the well-known hard-sphere structure factor S0(q) according
to PY closure 44,45 (see supplementary material at30 for details).
F(q) and P(q) were evaluated from the lysozyme atomic coordinates reported in the
Protein Data Bank (PDB entry 6lyz46) and considering the composition of the protein
solvation layer. The SASMOL approach47,48, which locates the sites of the water molecules
belonging to the different protein hydration shells and describes them as Gaussian spheres,
was used. According to the SASMOL method, the number of sites belonging to the lysozyme
first hydration shell (corresponding to the msites of the s-domain) was 385, whereas in the
second shell m′= 470 sites were found. Since the molecular volume of ethanol is larger than
that of water, the local domain lhas been considered to include both the first hydration
shell, of which a fraction φis occupied by water molecules, and the second hydration shell of
the protein, whose ethanol composition corresponds to xe,b. The calculation of the values of
8

φand xe,b, which for a fixed thermodynamic exchange constant are functions of the nominal
sample composition (expressed by npand xe), has been performed by numerically solving the
system constituted by the Eq. 2 (using the dependency of γeand γwon xe,b) and the mass
balance of the system. The local domain composition, xe,l , was then calculated by averaging
the compositions of the sites in the first and in the second hydration shells. Therefore, the
form factor depends on solvent composition (ethanol concentration and deuteration grade)
and on the thermodynamic parameters fixing the compositional relationships among the
different domains9(see the supplementary material at30 for the resulting equations).
C. DSC data analysis
DSC analysis was based on the two-state transition model described by Schwarz and
Kirchhoff49 , updated under the quantitative advisement of the 4 components of the sys-
tem: folded and unfolded lysozyme, water and ethanol. On the basis of previous reports,
the unfolding process has been assumed to be cooperative and two-state at all the investi-
gated conditions34. Thermodynamic exchange constants, Kex
Nand Kex
U, and the number of
exchange sites on the unfolded protein (mU) had to be determined.
DSC thermograms, reporting the temperature dependence of the heat capacity Cpof the
protein dissolved in the water/ethanol mixtures, were calculated following the Schellman’s
exchange model as reported by Spinozzi et al. 14 ,
Cp=CpN+np,U
np∆Cp+(∆H)2
(exp (−∆G/kBT) + 1)kBT2,(6)
where np,U is the number density of unfolded lysozyme, ∆Cp=CpU−CpN, ∆Hand ∆Gare
the differencies of heat capacity, enthalpy and free energy for the transition between native
and unfolded protein, each conformation having a different composition of the solvation
shell, as represented by the following expression:
NwmNφNemN(1−φN)+ (mUφU−mNφN)w+ [mU(1 −φU)−mN(1 −φN)]e⇋
UwmUφUemU(1−φU).(7)
All terms in Eq. 6 are complex functions of composition (xeand np) and temperature14. It
has to be stressed that the whole unfolding equilibrium described by Eq. 7 can be decom-
posed in three elementary parts: i) an unfolding equilibrium of lysozyme in water, ii) mN
9

water/ethanol exchange equilibria at the surface of the protein in the native conformation,
iii) mUwater/ethanol exchange equilibria at the surface of the unfolded protein. Because
any water/ethanol exchange over folded and unfolded protein modifies the composition of
the bulk solvent (xe,b), the model takes into account the contribution of excess enthalpy,
excess heat capacity and excess volume of water/ethanol mixtures, which are calculated as
analytical functions of xe,b by means of published phenomenological expressions50–52. Hence,
all thermodynamic functions were written as proper combinations of the thermodynamic
functions of each of the three elementary transitions.
D. Global fit analysis
A global fitting procedure was used to provide a unique interpretation of the whole set
of experimental data. All the 26 experimental curves (10 DSC thermograms and 16 SANS
curves) were analyzed together, and three different classes of fitting parameters were con-
sidered. The first class includes common parameters, which are independent of the type
of experiment and of the experimental conditions, such us Kex
N; the second class includes
common parameters which are pertinent only to SANS or DSC, but are still independent
of the experimental conditions; the third class includes parameters which depend on sample
composition, deuteration grade, and type of experiment and should be fitted independently
for each curve. Further details concerning the fitting parameters can be found in the sup-
plementary material at30.
IV. RESULTS AND DISCUSSION
SANS and DSC fitting results are superimposed to experimental data in Fig.s 1 and 2:
the good quality of the fitting procedure, also indicated by the rather low value of 1.23 for
the global reduced χ2, can be easily appreciated. The common parameters resulting from
the global fittings are reported in Table I.
Before to analyze the thermodynamic exchange constants, which dictate the composition
of the local domain, few comments on the fitted structural parameters are in order. The
protein effective charge Zresulted nearly independent on protein and ethanol concentration:
its average value resulted <Z >= 16.0±0.3 e, which roughly corresponds to the number of
10

xe= 0.09xe= 0.06
xe= 0.03
120
90
60
30
xe= 0.00
120
90
60
30
120
90
60
30
120
90
60
30
q(˚
A−1)
dΣ/dΩ(q) (cm−1)
0.30.20.100.30.20.10
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.6
0.5
0.4
0.3
0.2
0.1
0
FIG. 1. SANS curves of lysozyme in water/ethanol mixtures at different compositions, as shown
by the xevalues reported in the bottom of each column and by the protein concentration reported
on the right of each curve in mg/mL unit. All the samples are at a deuteration grade xD= 0.88.
Circles correspond to experimental data, solid lines are the best fits obtained by the global fit
analysis. Curve are scaled by a factor 0.1 cm−1for clarity.
charges carried by lysozyme at pH 3 according to its aminoacid composition53. Considering
the attractive term, the range dwas found to be almost constant with the ethanol mole frac-
tion, with an average value <d >= 3.0±0.4˚
A, whereas the depth Jwas observed to slightly
decrease (∼3%, from 12.2±0.4 kJ mol−1at xe= 0.00 to 10.9±0.4 kJ mol−1at xe= 0.09).
It can be noticed that the depth of the attractive term can be overestimated because of RPA
closure adopted for the structure factor. Indeed the fitted short-range attraction strength
results to be larger in respect to the ones resulting in similar experimental conditions, but
11

xe= 0.09
xe= 0.08
xe= 0.07
xe= 0.06
xe= 0.05
xe= 0.04
xe= 0.03
xe= 0.02
xe= 0.01
xe= 0.00
20 kJ mol−1K−1
T(K)
Cp−CpN0
360350340330320310300
FIG. 2. Experimental DSC thermal scans of lysozyme in water/ethanol mixtures at different com-
positions, as shown by the indicated xevalues. Circles correspond to the experimental data. Solid
lines represent the profiles obtained by the global fit analysis. Curves are scaled by 40 kJ mol−1K−1
for clarity.
obtained by MSA closure40,41 and by HCN closure54. However, this approximation does not
influence the observed trend of the depth of the attractive potential versus ethanol molar
content in solution.
From one hand, this result is rather interesting, as it confirms previous observations sug-
gesting that the increase of lysozyme repulsive forces in the presence of ethanol (detected as
an increase of the osmotic second virial coefficient induced by addition of ethanol on aqueous
lysozyme solution) is mainly determined by reduced attractive hydrophobic interactions and
not by changes in the coulombic repulsion33.
On the other hand, the observation of a nearly constant Zdisagrees with recent SANS
12

n1/3
p(˚
A−1)
qmax (˚
A−1)
0.0160.012
0.14
0.1
120
90
60
30
q(˚
A−1)
SM(q)
0.30.250.20.150.10.050
3
2.5
2
1.5
1
0.5
0
xe= 0.09
xe= 0.06
xe= 0.03
xe= 0.00
q(˚
A−1)
P(q) (105barn)
0.30.250.20.150.10.050
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
FIG. 3. Left panel. Effective structure factors calculated by [ dΣ
dΩ(q)−B]/P (q). Solid lines are the
effective structure factors obtained by the global-fit analysis. The four groups of curves, scaled
by a factor 0.5, refer to the different protein concentration, as reported on the right in mg/mL
unit. Symbols represent ethanol molar fractions: ,xe= 0.00; ,xe= 0.03 ; △,xe= 0.06; ▽,
xe= 0.09. The insert reports the first maximum positions qmax of SM(q) as a function of the cubic
root of lysozyme number density, n1/3
p. The solid straight line shows the behavior qmax ∼n1/3
p
expected for purely repulsive charged monomers. Right panel. Form factors obtained by the fit of
SANS curves, reported in absolute unit (barn). Variations of P(q) with protein concentration, at
fixed xe, are within the width of the curves.
data analysis on lysozyme aqueous solution, which showed a net charge increase from ca. 8
to 10 e when protein concentration changes from 5 to 22.5%41. However, in that case the
pD of the protein solutions was barely controlled (the pD value was 6.8 at 5% and 4.5 at
22.5%41), and the different used closure of the OZ equation to obtain the structure fac-
tor can maybe explain the discrepancy (indeed, some discordances between the attraction
strengths calculated by using MSA and Hypernetted Chain closures were even underlined
by the authors41), too. As the structure factors are not expected to change with the choice
13

of the closure (and then can be directly compared), effective structure factors at different
lysozyme and ethanol concentrations have been extracted from SANS curves by dividing the
experimental macroscopic differential scattering cross section by the fitted form factor of the
solvated protein. Results are reported in Fig. 3. At one side, it can be observed that the
form factors strongly depend on xe(e.g., on the composition of the solvation layer), while
do not change with protein concentrations, as expected. On the other side, the structure
factors appear almost independent on ethanol addition but dependent on protein concentra-
tion, confirming that the inter-protein potential is dominated by a strong repulsive regime
(probably due to the large number of charges on lysozyme at this pH) and suggesting that
the eventual formation of small equilibrium clusters43,55 is prevented. Moreover, SM(q) show
only one peak, at about 0.1 ˚
A−1, while the second peak, observed by Liu et al. 41 at about
0.23 ˚
A−1and prominent at high lysozyme concentration, is absent (or only barely visible).
The pronounced shift of the first maximum toward higher qwith increasing concentration,
although does not follow a n1/3
pbehavior (see the insert in the left panel of Fig. 3), appears in
contrast with the attribution of this peak to cluster-cluster correlations55 and is clearly more
compatible with a system dominated by largely repulsive individual lysozyme molecules in
solution, as also observed by Shukla et al. 42 .
The main quantitative results of the present analysis are however the thermodynamic
exchange constants Kex
Nand Kex
U. Fig. 4 reports the ethanol molar fraction in the local
domain, xe,l, calculated on the basis of the fitted thermodynamic exchange constants, as a
function of ethanol nominal content of the solution, xe. Considering that the thin dashed
line represents the effective constant Kexγw/γe= 1, it appears evident that in water/ethanol
mixtures there is a release of water molecules from the lysozyme surface, which in opposition
enriches in alcohol. This is a very interesting result, as according to our knowledge no
quantitative description of the protein solvation shell in a water/ethanol mixture has been
previously reported, even though in a SANS study of ribonuclease in water/glycerol and
water/ethanol mixtures, Lehmann and Zaccai already suggested that “For ethanol it is even
possible that there might be some preferential binding”56. Here, however, the quantitative
description of the lysozyme solvation shell also concerns the unfolded protein: this is a very
interesting issue, which can be related to the contribution of co-solvents with respect to
protein unfolding “in vivo”, as recently evidenced with force spectroscopy measurements5.
Concerning the results, it appears that Kex
Nis slightly larger than Kex
U(2.67 ±0.04 and
14

U
N
xe
xe,l
0.10.080.060.040.020
0.1
0.08
0.06
0.04
0.02
0
FIG. 4. Ethanol molar fraction in the local domain as a function of ethanol nominal content. The
thin dashed line represents the effective constant Kexγw/γe= 1. Continuous lines with points
result from the global fit of native (open circles) and unfolded (close circles) lysozyme dissolved in
water/ethanol mixtures. The dotted line shows the local domain vs. the nominal glycerol molar
fraction for native lysozyme in water/glycerol mixtures, as obtained by Sinibaldi et al. 9.
2.54 ±0.02, respectively, while the number of sites appears to increase from mN= 385,
observed in the native state, to mU= 418 ±2, observed for the unfolded lysozyme. The
increase in the number of sites is in full agreement with the presence of a higher protein
surface exposed to the solvent upon thermal unfolding (even if the small difference is con-
sistent with the occurrence of a molten globule conformation34), while the meaning of the
thermodynamic exchange constant values merits to be further discussed.
The first parameter that can be derived is the so-called excess solvation number, Npj,
which represents the difference between the number of j-molecules (j=w, e) occupying the
volume of the protein shell and those occurring in a similar volume in the bulk solvent. Npj
values, calculated by using Eq. 16 of Sinibaldi et al. 9, are reported as a function of solvent
composition in Fig. 5 (note that the excess solvation number for water is different from zero
even in pure water because of the higher density of water molecules in the local domain).
Two points could be underlined: at one side, the composition of ethanol in the local domain
is higher than that in the bulk, in close resemblance of the behavior observed in the case of
15

U,e
U,w
N,e
N,w
xe
Npj
0.10.080.060.040.020
10
5
0
-5
-10
FIG. 5. Number of excess molecules of the species jin the lysozyme solvation layer as a function of
ethanol nominal content. Circles refers to ethanol and squares to water. Empty and filled symbols
refer to the native and to the unfolded protein, respectively.
BSA13,57 and of lysozyme10 dissolved in urea aqueous solutions. Noticeable, in both solvents
the protein structure stability is strongly reduced. At the other side, the excess solvation
numbers for native lysozyme show a critical point exactly at the same concentration in
which a transition in water/ethanol mixtures has been detected. In fact, both the apparent
ethanol molar compressibility and the frequency of the C−H vibrations vary as a function
of solvent composition starting from about x∗
e≃0.0527. This finding confirms the idea that
a protein can be a probe to evidence binary mixtures properties. Also interesting is the
fact that a criticality in the melting temperature of Yeast Frataxin cold denaturation in
water/ethanol mixtures was observed at similar ethanol concentration58, while some of us
recently determined that the same critical ethanol concentration triggers DNA condensation
in water/ethanol mixture59. The criticality in the excess solvation numbers can hence be
associated with the appearance of a sort of hydrophobic clustering of alcohol molecules26.
While in the water rich region, a simple thermodynamic equilibrium between water and
ethanol molecules in the bulk and in the solvation shell exists, it is possible that ethanol
clustering affects this equilibrium beyond x∗
e. Consequently, ethanol can be found in solution
as free molecules in the bulk, molecules bound on the protein surface or molecules forming
clusters in the bulk. This picture is however implicitly taken into consideration in our model,
16

First class parameters
Kex
N
2.67 ±0.04
Second class parameters and average third class parameters, SANS
<Z > <d > νw,s
(e) (˚
A) (˚
A3)
16.0±0.3 3.0±0.4 29.3±0.1
Third class parameters, SANS
<J >xe=0.00 < J >xe=0.03 <J >xe=0.06 < J >xe=0.09
(kJ mol−1) (kJ mol−1) (kJ mol−1) (kJ mol−1)
12.2±0.4 11.8±0.3 11.4±0.3 10.9±0.4
Second class parameters, DSC
Kex
UmU
2.54 ±0.02 418 ±2
∆Hw0∆Gw0
(kJ mol−1) (kJ mol−1)
270 ±10 48 ±1
CpwU 0CpwN 1CpwU 1
(kJ mol−1K−1) (J mol−1K−2) (J mol−1K−2)
7.6±0.6 130 ±20 32 ±7
∆Hex
N0∆Hex
U0∆Cpex
N0∆Cpex
U0
(kJ mol−1) (kJ mol−1) (J mol−1K−1) (J mol−1K−1)
6.8±0.5−0.2±0.1 12 ±4 220 ±20
TABLE I. First and second class global fit parameters of SANS and DSC data (symbols as in the
text and in the supplementary material, at30). Errors in fitting parameters were established by
iteratively moving all the SANS and DSC curve points within their experimental error, by then
repeating the minimization process and by calculating the average and the standard deviation of
each fitting parameter after a number of iterations equal to 20.
17

Uat 360 K
Nat 360 K
Uat 298 K
Nat 298 K
xe
mi(1 −φi)
0.10.080.060.040.020
45
40
35
30
25
20
15
10
5
0
FIG. 6. Number of ethanol molecules bound to the surface of lysozyme in the native, mN(1 −φN),
and in the unfolded, mU(1 −φU), state, for the two reported temperatures, as a function of the
ethanol nominal content.
which implements activity coefficients: for this region, a unique thermodynamic exchange
constant succeeds in analyzing all the investigated experimental data.
Fig. 5 also shows that the protein in the native state releases more water than the protein
in the unfolded state. This finding could be considered counterintuitive, as it is commonly
assumed that unfolded states expose more hydrophobic residues to the solvent than the
native state, so that higher affinity to ethanol (that is more hydrophobic than water) for the
unfolded protein could be expected. However, such a result needs to be read together with
the number of protein sites upon unfolding, which increases from 385 to 418. Indeed, Fig. 6
shows the total number of ethanol molecules bound to the surface of native and unfolded
lysozyme calculated as a function of xeat two different temperatures (298 K and 360 K). As
expected, more ethanol interacts with the lysozyme in the unfolded state than in the native
state.
Even if the small decreasing in the ethanol preferentially affinity over the binding site from
the folded to the thermally unfolded state, followed by a small increase of the total number of
sites, has to be regarded within the model approximations, it should be observed that protein
folded and unfolded forms can experience different solvation mechanisms, due for example
to a competition engaged by alcohol molecules in the interaction with the apolar groups
18

of the protein58 or according to the remarkable observation that hydrophobicity manifests
itself differently on large and small length scales 60. Hence the difference between Kex
Uand
Kex
Ncannot easily be explained without a deeper knowledge of the thermally exposed sites.
It is worth noting that the data analysis strategy exploited in our study is able to deter-
mine not only whether there is a preferential hydration or not, but it additionally provides
values for the excesses or deficits of water and co-solvent molecules in the vicinity of lysozyme.
Note that the excess or deficit of solvent molecules close to the protein surface represents
the behavior of the whole protein molecule, but not for particular functional groups of the
protein. In fact it is surely possible that ethanol is in excess around the whole lysozyme
but in deficit in the vicinity of certain functional groups61. However the molecular approach
of this study cannot be denied by the absence of details on specific functional binding of
ethanol on protein surface.
Protein solvation issue is intimately tied to protein structural integrity and flexibility,
hence to its dynamics and function. For the first time, this study provides a simultaneous
analysis of SANS and DSC experimental data in order to obtain a more quantitative knowl-
edge of the thermodynamic processes at the basis of protein solvation equilibria in binary
mixtures.
REFERENCES
1M. C. Bellissent-Funel, Hydration processes in biology (IOS Press, Amsterdam, 1999).
2T. M. Raschke, Curr. Opin. Struct. Biol. 16, 152 (2006).
3C. M. Othon, O. Kwon, M. M. Lin, and A. H. Zewail, Proc. Natl. Acad. Sci. 106, 12593
(2009).
4J. A. Schellman, Q. Rev. Biophys. 38, 351 (2005).
5L. Dougan, G. Feng, H. Lu, and J. M. Fernandez, Proc. Natl. Acad. Sci. 105, 3185 (2008).
6H. Fraunfelder, S. G. Sligar, and P. G. Wolynes, Science 254, 1598 (1991).
7S. N. Timasheff, Proc. Natl. Acad. Sci. 99, 9721 (2002).
8Q. Zou, B. J. Bennion, V. Daggett, and K. P. Murphy, J. Am. Chem. Soc. 124, 1192
(2002).
9R. Sinibaldi, M. G. Ortore, F. Spinozzi, F. Carsughi, H. Frielinghaus, S. Cinelli, G. Onori,
and P. Mariani, J. Chem. Phys. 126, 235101 (2007).
19

10M. G. Ortore, R. Sinibaldi, F. Spinozzi, F. Carsughi, D. Clemens, A. Bonincontro, and
P. Mariani, J. Phys. Chem. B 112, 12881 (2008).
11M. G. Ortore, R. Sinibaldi, F. Spinozzi, A. Carbini, F. Carsughi, and P. Mariani, JPCS
177, 012007 (2009).
12K. Gekko and S. N. Timasheff, Biochem. 20, 4667 (1981).
13S. N. Timasheff and G. Xie, Biophys. Chem. 105, 421 (2003).
14F. Spinozzi, M. G. Ortore, R. Sinibaldi, P. Mariani, A. Esposito, S. Cinelli, and G. Onori,
J. Chem. Phys. 129, 35101 (2008).
15R. M. Parodi, E. Bianchi, and A. Ciferri, J. Biol. Chem. 248, 4047 (1973).
16T. S. Banipal and G. Singh, Thermochim. Acta 412, 63 (2004).
17W. Dzwolak, S. Grudzielanek, V. Smirnovas, R. Ravindra, C. Nicolini, R. Jansen, A. Lok-
sztejn, S. Porowski, and R. Winter, Biochemistry 44, 8948 (2005).
18G. Onori and A. Santucci, J. Mol. Liq. 69, 161 (1996).
19A. Wakisaka and T. Ohki, Faraday Discuss. 129, 231 (2005).
20A. K. Soper, L. Dougan, J. Crain, and J. L. Finney, J. Phys. Chem. 110, 3472 (2006).
21J. H. Guo, Y. Luo, A. Augustsson, S. Kashtanov, J. E. Rubensson, D. K. Shuh, H. ˚
Agren,
and J. Nordgren, Phys. Rev. Lett. 91, 1574011 (2003).
22S. J. Suresh and V. M. Naik, J. Chem. Phys. 116, 4212 (2002).
23S. K. Allison, J. P. Fox, J. P. Hargreaves, and S. P. Bates, Phys. Rev. B 71, 024201 (2005).
24C. Zhang and X. Yang, Fluid Phase Equilib. 231, 1 (2005).
25A. B. Roney, B. Space, E. W. Castner, R. L. Napoleon, and P. B. Moore, J. Phys. Chem.
B108, 7389 (2004).
26M. D’Angelo, G. Onori, and A. Santucci, J. Chem. Phys. 100, 3107 (1994).
27S. Cinelli, G. Onori, and A. Santucci, J. Phys. Chem. B 101, 8029 (1997).
28J. A. Schellman, Biopolymers 34, 1015 (1994).
29J. A. Schellmann, Biophys. J. 85, 108 (2003).
30“Supplemental material,” http://... (2011).
31J. Teixeira, in Structure and Dynamics of Strongly Interacting Colloids and Supramolecular
Aggregates in Solution, edited by J. S. H. S. H. Chen and P. Tartaglia (Springer, 1992) pp.
635–658.
32V. Calandrini, G. Onori, and A. Santucci, J. of Molecular Structure 565, 183 (2001).
33W. Liu, D. Bratko, J. M. Prausnitz, and H. W. Blanch, Biophys. Chemistry 107, 289
20

(2004).
34P. Sassi, G. Onori, A. Giugliarelli, M. Paolantoni, S. Cinelli, and A. Morresi, J. Mol.
Liquids 159, 112 (2011).
35R. S. Hansen and F. A. Miller, J. Phys. Chem. 58, 193 (1954).
36S. H. Chen, Annu. Rev. Phys. Chem. 37, 351 (1986).
37F. Spinozzi, D. Gazzillo, A. Giacometti, P. Mariani, and F. Carsughi, Biophys. J. 82,
2165 (2002).
38Y. Liu, W. R. Chen, and S. H. Chen, J. Chem. Phys. 122, 044507 (2005).
39M. Broccio, D. Costa, Y. Liu, and S. H. Chen, J. Chem. Phys. 124, 084501 (2006).
40S. H. Chen, M. Broccio, Y. Liu, E. Fratini, and P. Baglioni, J. Appl. Cryst. 40, s321
(2007).
41Y. Liu, L. Porcar, J. Chen, W. R. Chen, P. Falus, A. Faraone, E. Fratini, and K. H. P.
Baglioni, J. Phys. Chem. B 115, 7238 (2011).
42A. Shukla, E. Mylonas, E. D. Cola, S. Finet, P. Timmins, T. Narayanan, and D. I.
Svergun, PNAS 105, 5075 (2008).
43F. Cardinaux, E. Zaccarelli, A. Stradner, S. Bucciarelli, B. Farago, S. U. Egelhaaf, . F.
Sciortino, and P. Schurtenberger, J. Phys. Chem. B 115, 7227 (2011).
44M. S. Wertheim, Phys. Rev. Lett. 10, 321 (1963).
45J. P. Hansen and I. R. McDonald, Theory of Simple Liquids (Academic Press, London,
1976).
46R. Diamond, J. Mol. Biol. 82, 371 (1974).
47M. G. Ortore, F. Spinozzi, P. Mariani, A. Paciaroni, L. Barbosa, H. Amenitsch, M. Stein-
hart, J. Ollivier, and D. Russo, J. R. Soc. Interface 6, S619 (2009).
48F. Spinozzi, submitted (2011).
49F. P. Schwarz and W. H. Kirchhoff, Thermochim. Acta 128, 267 (1988).
50J. A. Boyne and A. G. Williamson, J. Chem. Eng. Data 12, 318 (1967).
51J. P. E. Golier and E. Wilhelm, Fluid Phase Equilibria 6, 283 (1981).
52B. Gonzalez, N. Calvar, E. Gomez, and A. Dominguez, J. Chem. Thermodynamics 39,
1578 (2007).
53L. I. G. Toldo, “Jambw chapter 3.1.6,” (1997).
54A. Tardieu, A. L. Verge, M. Malfois, F. Bonnete, S. Finet, M. Reis-Kautt, and L. Belloni,
J. Crystal Growth 196, 193 (1999).
21

55A. Stradner, H. Sedgwick, F. Cardinaux, W. C. K. Poon, S. U. Egelhaaf, and P. Schurten-
berger, Nature 432, 492 (2004).
56M. S. Lehmann and G. Zaccai, Biochem. 23, 1939 (1984).
57R. Sinibaldi, M. G. Ortore, F. Spinozzi, S. de Souza Funari, J. Teixeira, and P. Mariani,
Eur. Biophys. J. 37, 673 (2008).
58S. R. Martin, V. Esposito, P. D. L. Rios, A. Pastore, and P. A. Temussi, J. Am. Chem.
Soc. 130, 9963 (2008).
59S. Marchetti, S. Cinelli, and G. Onori, Chem. Phys. Lett. 493, 158 (2010).
60B. J. Berne, J. D.Weeks, and R. Zhou, Annu. Rev. Phys. Chem. 60, 85 (2009).
61I. L. Shulgin and E. Ruckenstein, J. Chem. Phys. 123, 054909 (2005).
22