Liquid–liquid phase separation in supersaturated lysozyme solutions and

associated precipitate formation/crystallization

Martin Muschol and Franz Rosenberger

Citation: J. Chem. Phys. 107, 1953 (1997); doi: 10.1063/1.474547

View online: http://dx.doi.org/10.1063/1.474547
View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v107/i6
Published by the American Institute of Physics.

Additional information on J. Chem. Phys.

Journal Homepage: http://jcp.aip.org/
Journal Information: http://jcp.aip.org/about/about_the_journal
Top downloads: http://jcp.aip.org/features/most_downloaded
Information for Authors: http://jcp.aip.org/authors

Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 Liquid–liquid phase separation in supersaturated lysozyme solutions
and associated precipitate formation/crystallization
Martin Muschola) and Franz Rosenberger
Center for Microgravity and Materials Research, University of Alabama in Huntsville, Huntsville,
Alabama 35899
~Received 16 October 1996; accepted 2 May 1997!
Using cloud point determinations, the phase boundaries ~binodals! for metastable liquid–liquid
~L–L! separation in supersaturated hen egg white lysozyme solutions with 3%, 5%, and 7%
(w/ v ) NaCl at pH54.5 and protein concentrations c between 40 and 400 mg/ml were determined.
The critical temperature for the binodal increased approximately linearly with salt concentration.
The coexisting liquid phases both remained supersaturated but differed widely in protein
concentration. No salt repartitioning was observed between the initial and the two separated liquid
phases. After the L–L separation, due to the presence of the high protein concentration phase,
crystallization occurred much more rapidly than in the initial solution. At high initial protein
concentrations, a metastable gel phase formed at temperatures above the liquid binodal. Both crystal
nucleation and gel formation were accelerated in samples that had been cycled through the binodal.
Solutions in the gel and L–L regions yielded various types of precipitates. Based on theoretical
considerations, previous observations with other proteins, and our experimental results with
lysozyme, a generic phase diagram for globular proteins is put forth. A limited region in the
(T,c) plane favorable for the growth of protein single crystals is delineated. © 1997 American
Institute of Physics. @S0021-9606~97!51130-5#

I. INTRODUCTION existing crystal solubilities. These data, together with obser-

There is growing evidence that supersaturated protein
solutions can undergo a thermally induced separation into
two metastable liquid phases of widely different concentra-
tion. This was first indicated by light scattering from unbuf-
fered lysozyme–salt solutions,1 and later extended to solu-
tions buffered to pH'7.2,3 The most detailed investigations
were concerned with g-crystallin and its natural variants.4–9
We are interested in the impact of the demixing transi-
tion on protein crystallization. In the high protein-
concentration phase resulting from the demixing, crystalliza-
tion is typically too rapid for the growth of single crystals
with low defect densities. In addition, the question arises to
what extent various forms of precipitate10 and the fractal
aggregates11 that have been observed in protein solutions are
related to liquid–liquid phase separations? Answers to this
question are complicated by the noticeable lack of character-
ization of precipitates and their morphologies, which are
typically considered a nuisance rather than a source of valu-
able information.
Hence, we investigated the existence and salt-
concentration dependence of the demixing transition in
lysozyme solutions under crystallization conditions, and the
resulting precipitate morphologies. Using light scattering
measurements of the cloud point temperature, we have de-
termined coexistence curves for binary liquid lysozyme so-
lutions at pH54.5 and three different NaCl concentrations
that have been used in crystal growth studies of lysozyme.
This allowed for comparison of the cloud point curves with
a!
Current address: Dept. of Neuroscience, University of Pennsylvania School
of Medicine, Philadelphia, PA 19104.
vations of gel formation, delineate the phase regions useful
for crystal growth.
Recent phase diagram calculations for hard spheres with
attractive interactions indicate that liquid phase separation
can occur underneath the solidus line as the range of the
potential is sufficiently reduced.12,13 In crystallizing protein
solutions, the range of attractive interactions appears to fall
in the relevant range.14 The high salt ~precipitant! concentra-
tions typically used, reduce the range of the electrostatic re-
pulsion between the macroions such that the attractive inter-
actions dominate.3,15,16 These considerations suggest that
metastable L–L phase separation in crystallizing protein so-
lutions may be widespread.
II. MATERIALS AND METHODS
Our earlier work on lysozyme’s crystallization kinetics
and its relation to impurity concentrations has shown signifi-
cant differences with respect to various source materials.17–19
Six times recrystallized and lyophilized lysozyme, obtained
from Seikagaku America, showed kinetics comparable to
that obtained with highly purified material. Hence, we used
this material ~lots E94203 and E94Z05! without further pu-
rification. Given the high salt concentrations of our solutions
~see below!, contamination of the solutions by the small
amounts of salt introduced by the added protein was ne-
glected. All other chemicals used were reagent grade or
purer.
The solutions were buffered with 0.1 M sodium acetate/
acetic acid ~NaAc! at pH54.5 as determined with an Orion
Sa 520 pH meter. After dissolution of either 3%, 5%, or 7%
NaCl (w/ v ) in the buffered solution, the pH was readjusted.

J. Chem. Phys. 107 (6), 8 August 1997 0021-9606/97/107(6)/1953/10/$10.00 © 1997 American Institute of Physics 1953

Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 1954 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions

Up to protein concentrations of c<150 mg/ml, the lyo-

philized lysozyme powder was directly dissolved in the
buffer/salt mixtures. To remove undissolved protein, air
bubbles, and dust, samples were centrifuged in a Savant
HSC10K centrifuge at 9000 g ~7500 rpm! for 5–10 min and
filtered through a 0.22 mm Millipore Millex-GV filter into a
clean sample vial. For the 5% and 7% NaCl samples, buffer/
salt solutions were preheated to about 45–50 °C to expedite
dissolution of the protein and to keep sample temperatures
well above the cloud point. To obtain higher protein concen-
trations ~up to 400 mg/ml!, about 2 ml of a lysozyme solu-
tion with c'150 mg/ml was prepared as described above.
Then, after insertion of the vial into a 15 ml plastic centri-
fuge tube containing an ice/water mixture for cooling, it was
centrifuged at 3500 g for 5–10 min in a Hermle Labnet Z232
centrifuge. The concurrent cooling and centrifugation re-
sulted in two visibly distinct layers of different protein con-
centration. Lack of adequate temperature control and partial
remixing after centrifugation prevented the use of these sepa-
rated phases for direct determination of the binodal. After
decanting of the top layer ~low c!, this process was repeated
until the desired c ~as measured by uv absorption of dilute
samples; a 28052.64 ml/mg cm!20 was obtained. Prior to the
cloud point measurements, these samples were homogenized
by centrifugation at 9000 g for 3 min at some elevated tem-
perature.
To check for salt repartitioning possibly associated with
the L–L separation, we phase separated a solution containing
110 mg/ml lysozyme and 3% NaCl, and determined the
Na1 and Cl2 concentrations in the initial and resulting two
phases. Na1 determinations were carried out by atomic ab-
sorption spectroscopy. The Cl2 concentrations were obtained
from a colorimetric procedure employing the Sigma Chemi-
cals diagnostics kit #461-3.
The coexistence curves for the binary liquid phases were
determined through light scattering intensity measurements.
A light scattering cell of approx. 200 ml volume and 1 mm
path length ~Starna Cells, Inc.; type 37! was filled with the
solution, glass stoppered, and placed into a miniaturized and
computer-controlled scintillation cell setup similar to one de-
scribed elsewhere.21 This setup provides for programmed
control (60.1 °C) and monitoring of the sample tempera-
ture while simultaneously recording the 90° light scattering
signal. Determination of the cloud temperature T cloud as a
function of protein concentration c at a given NaCl concen-
tration maps out the coexistence curve in the (T,c) plane.
III. RESULTS AND DISCUSSION
A. Cloud points and phase diagram
Figure 1 shows time traces of the sample temperature
and corresponding scattering signal obtained from a solution
with c5360 mg/ml in 3% NaCl. One sees that on lowering
of the temperature, at first the scattering intensity increases
slowly. This reflects an increase in concentration fluctuations
~due to attractive protein interactions! with decreasing ther-
mal energy. Then, at about 7.0 °C, a jump in scattering in-
tensity occurs. This sudden increase in sample turbidity
FIG. 1. Time traces of sample temperature and light scattering signal for
lysozyme at c5360 mg/ml in 3% NaCl/0.1 M NaAc buffer at pH54.5;
T sat553 °C estimated from Eq. ~2!. Note the two reversible increases in
scattering intensity associated with brief decreases in sample temperature by
0.5 °C. T cloud57.0 °C.
originates from the appearance of liquid droplets ~see below!
or the onset of spinodal decomposition. Hence, we designate
this temperature as T cloud for this sample composition. On
raising the temperature shortly after the L–L separation by as
little as 0.5 °C above T cloud, the cloudiness responsible for
the strong signal peak readily disappears. The second peak in
Fig. 1 shows that the cloud point determination is reproduc-
ible within a few tenths of a degree.
The L–L phase separation enhances the crystal nucle-
ation rate in the solution. This is illustrated by Fig. 2, that
was obtained from a sample of initial c5140 mg/ml with 5%
NaCl. Signal peak A, resulting from the demixing transition
associated with the small dip in temperature from 22 to
21.5 °C, suggests complete re-establishment of the initial,
homogeneous solution state. However, about 10 min later
~point B! the scattering intensity steeply increased, even after
the temperature had been increased to 25 °C. This strong rise
in signal is due to scattering from small crystals that nucleate
upon phase separation in the high-c phase and continue to
grow at any T below the saturation temperature for the
solid–liquid transition T sat. For this combination of protein
and salt concentrations, T sat was estimated from a van’t Hoff
extrapolation of the experimental solubility data ~see below!
to be 67 °C. Thus it is not surprising that, as Fig. 2 shows,
raising the temperature to 38 °C could not remove the tur-
bidity. ~The final plateau in the scattering intensity is due to
the competition between enhanced scattering and enhanced
absorption of light in these turbid samples, rather than a
steady state in crystal population.!
At higher salt and protein concentrations, enhanced
nucleation was often noticeable right after cycling of the
sample through the L–L region: The scattering intensity

J. Chem. Phys., Vol. 107, No. 6, 8 August 1997

Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions
FIG. 2. Time traces for a sample with c5140 mg/ml; T sat567 °C, as esti-
mated from Eq. ~2!. Point A: small dip in temperature reversibly induces
liquid–liquid phase separation. Point B: spontaneous crystal nucleation.
above T cloud
remained elevated above the level prior to the
phase separation and continued to grow. Such measurements
were excluded from our cloud point data. It should be em-
phasized that without cycling through the binodal, nucleation
in the above solution at 25 °C would have required a much
longer time. In general, we have observed, that after cycling
through the binodal, nucleation induction times ~as judged by
the appearance of crystallites! were reduced from hours to
minutes. Two factors contribute to this pronounced lowering
of the kinetic barrier to crystallization: ~a! as we will see
below, the supersaturation in the high concentration phase is
about an order of magnitude higher than in the starting solu-
tion; and ~b! the difference in surface energy between the
high-c liquid and the solid phases is reduced due to the
closer match in protein densities.
Direct evidence for the L–L separation was obtained
from microscopic observations as well as centrifugation of
the samples below T cloud. Microscopy revealed that the high
concentration protein phase emerged either in the form of
spherical droplets or as bulging tubes intertwined with the
low-c phase. Both morphologies are characteristic of sepa-
rating liquid phases.22 Centrifugation of samples at T
,T cloud, followed by reheating above T cloud, yielded two
transparent liquid layers ~see also Ref. 2!. Without centrifu-
gation the high concentration liquid droplets tended to trans-
form into microcrystals before settling out.
In order to detect whether the protein demixing is ac-
companied by salt repartitioning, we determined the Na1 and
Cl2 concentrations in the two liquid phases forming from a
solution with c5110 mg/ml and 3% NaCl. At this lowest of
the salt concentrations used, salt repartitioning associated
with protein–salt interactions should result in the largest
relative changes in salt levels. However, within the experi-
mental errors of our measurement techniques ~see Sec. II! of
J. Chem. Phys., Vol. 107, No. 6, 8 August 1997
Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
1955
FIG. 3. Cloud point data for lysozyme at pH54.5 in 0.1 M NaAc buffer and
three different NaCl concentration. Dashed lines: fit to Eq. ~1!. Open down-
ward pointing triangles: Cloud point data for lysozyme at pH56.0 in 0.6 M
sodium phosphate from Ref. 2.
610% and 62.5%, respectively, no changes from the initial
ion concentrations were found. Hence, this metastable phase
separation in lysozyme is essentially binary and the T cloud
data can be justifiably presented in a pseudobinary ~protein-
solution! phase diagram. This is fundamentally different
from coacervation, where the phase separation of a solution
component other than the protein leads to a repartitioning of
the protein across that phase boundary. Coacervation has
been reported for phosphoglucomutase/polyethylene-glycol
mixtures23 and is frequently observed with membrane pro-
teins when using non-ionic detergents.24
Our cloud point data obtained at the three different NaCl
concentrations are presented as full symbols in T – c plots in
Fig. 3. At the highest c’s and thus high supersaturations,
T cloud was exceedingly difficult to determine due to the onset
of crystallization shortly after sample preparation. Hence, the
scarcity of data above c'300 mg/ml. With the broad flat
maxima of the binodal curves, the descending branch is
barely present. Yet, the above phase separation on centrifu-
gation unambiguously established the existence of a de-
scending high-c branch. For comparison, we have included
in Fig. 3 the L–L transition data previously obtained at
pH56 and 0.6 M sodium phosphate.2
The dashed lines in Fig. 3 are fits of our three cloud
point data sets to the scaling relation for binary demixing
from renormalization-group theory with a critical exponent
of b 50.32525,26
 1956 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions

FIG. 5. Phase diagram for lysozyme with 5% NaCl at pH54.5 in 0.1 M

FIG. 4. Phase diagram for lysozyme with 3% NaCl at pH54.5 in 0.1 M
NaAc buffer. Full circles: cloud point data from this work. Dashed line: fit
NaAc buffer. Full triangles: cloud point and solubility data from this work.
to Eq. ~1!. Open circles and diamonds: solubility of tetragonal and ortho-
Dashed line: fit to Eq. ~1!. Open, upward pointing triangles solubility of
rhombic lysozyme, respectively, from Refs. 29 and 30. Solid lines: extrapo-
tetragonal lysozyme from Ref. 21. Open, downward pointing triangles: solu-
lated fits to van’t Hoff equation @Eq. ~2! with c 0 53.4 mg/ml, DH
bility of tetragonal lysozyme from Refs. 29 and 30. Solid line: fit to van’t
526.7 k B T 0 at T 0 5298.15 K#. Points A, B, C mark composition and tem-
Hoff equation @Eq. ~2! with c 0 510.3 mg/ml, DH530.5 k B T 0 at T 0
perature of samples used in crystal morphology study; for details see the
5298.15 K# and extrapolation.
text.

H U
T cloud5T crit 12A
c crit2c p
c crit
U J
1/b
~1!
resent fits to the solubility data and extrapolations to higher
and T crit, c crit, and A as adjustable parameters. Within the
protein concentrations based on an expansion of van’t Hoff’s
large fitting uncertainties, all salt concentrations yielded a
relation around the reference temperature T 0 5298 K
critical concentration c crit'~255630! mg/ml. T crit increased
approximately linearly with salt concentration from well be-
low to well above ambient temperature. Interestingly, the
amplitude A for the 3% NaCl solutions is about twice that for
the higher salt concentrations. Despite the experimental un-
c sat5c 0 exp S DH T sat2T 0
kT 0 T0 D ~2!

certainties associated with the measurements of metastable

states, we feel that this difference lies outside the experimen-
tal errors. Note that lysozyme in growth from solution un-
dergoes a transition from the tetragonal to the orthorhombic
crystal structure around T'25 °C.27 Hence, it is reasonable
to expect some change in the solution structure as well. Re-
cent Monte Carlo simulations of protein interactions with
attractive square well potentials reproduced the experimental
values for c crit in g-crystallin but did not yield the large
width of the two-phase liquid coexistence region.28 The au-
thors suggested spatial anisotropy of the protein interactions
as possible cause for the broadness of the experimental data.
It will be interesting to see if modifications in such interac-
tion anisotropies are, in turn, responsible for the change in
amplitude of the phase separation curve and the tetragonal–
orthorhombic structure transition in lysozyme crystals.
To complete the phase diagram for lysozyme, we have
added in Figs. 4–6 the crystal solubility data for the same pH
and the three salt concentrations.21,29,30 The solid lines rep-
with c 0 5c sat(298 K). Values of the fitting parameters c 0 and
DH are given in the respective figure captions. To check the
validity of this extrapolation, we determined the c sat of two
high concentration lysozyme samples in 3% NaCl solutions;
see the upper full triangles in Fig. 4. Predicted and measured
values agree reasonably well. The resulting T sat’s lie safely
below the unfolding temperature for lysozyme of 83 °C ~Eq.
4 in Ref. 31!. Again, having the transition from the tetrago-
nal to the orthorhombic crystal structure at higher tempera-
tures in mind we have plotted solubility curves for both
forms in Figs. 5 and 6. For our ensuing discussion, however,
only the generic shape of the solubility curves and their lo-
cation with respect to the demixing boundary is of signifi-
cance. Most recently, a similar phase diagram was reported
for lysozyme at pH57.8 and 0.5 M NaCl.3 The authors com-
mented on the strong correlations of shifts in T cloud and T sat
with salt concentration. Such correlations are equally appar-
ent in our experimental data.

J. Chem. Phys., Vol. 107, No. 6, 8 August 1997

Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions
FIG. 6. Phase diagram for lysozyme with 7% NaCl at pH54.5 in 0.1 M
NaAc buffer. Full squares: cloud point data from this work. Dashed line: fit
to Eq. ~1!. Open squares and diamonds: solubility of tetragonal and ortho-
rhombic lysozyme, respectively, from Refs. 29 and 30. Solid lines: extrapo-
lated fits to van’t Hoff equation @Eq. ~2! with c 0 51.9 mg/ml, DH
524.3 k B T 0 at T 0 5298.15 K#.
B. Relevance for protein crystal growth
As described above, the solution demixing severely im-
pacts crystallization. When a typical crystal growth solution
(c!c crit) is cycled through the two-phase region, crystallites
develop within minutes compared to hours required for
nucleation slightly above T cloud. Most solutions that were left
in the two-phase liquid region formed white, microcrystal-
line precipitates. Even brief cycling through the binodal lead
to the undesirable formation of a large number of irregular
crystallites. Both morphologies are unsuitable for x-ray
structure determinations.
At low protein concentrations with 5% and 7% salt, cy-
cling resulted in spherulitic crystal morphologies that con-
sisted of thin needles growing radially outward from a nucle-
ation center; see Fig. 7. This ‘‘sea urchin’’ morphology,
which has also been obtained in high-pressure crystal growth
studies of lysozyme,32 can readily be understood in terms of
the L–L demixing. In low-c solutions, only few submicron-
sized droplets of the high-c phase appear upon phase sepa-
ration. Through multiple nucleation within a highly super-
saturated droplet the protein is rapidly transformed into a
polycrystalline solid. In the surrounding low-c phase, nucle-
ation is much less likely, but since this phase is also super-
saturated, further growth of the polycrystalline spheres that
emerge from the high-c droplets is still supported. Competi-
tion for nutrient supply among the crystals with various ori-
entations will select only those favorably oriented to grow
radially outwards from the polycrystalline center.
We have also investigated the crystal morphologies that
J. Chem. Phys., Vol. 107, No. 6, 8 August 1997
Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
1957
FIG. 7. ‘‘Sea urchin’’ crystal morphology obtained in lysozyme solution
with c520 mg/ml and 7% NaCl at room temperature after prior cooling
below T cloud.
evolve from high-c solutions after short cycling through the
binodal at temperatures slightly above T crit. Three 5% NaCl
containing solutions, designated A, B, and C, with respec-
tive c’s5147, 250, and 330 mg/ml, were, after the cloud
point measurements, held at 26 °C. As can be seen from Fig.
5, these concentrations were well below, close to and well
above c crit, respectively. Figure 8 shows low-magnification
images of these samples taken within a few hours after
preparation and cycling through T cloud. Sample A, depicted
in Fig. 8~a!, yielded predominately tetragonal lysozyme crys-
tals, with some granularly structured background ~precipi-
tate! which disappeared within a day. In sample B, that is
close to the critical concentration, collections of beads of
about 200 mm diameter formed instead, presumably from the
high-c phase; see Fig. 8~b!. Probing with a needle revealed a
mechanically highly resistant gel nature of these beads. This
morphology persisted for many hours before crystals ap-
peared. As control, a fraction of the same sample was left
above T cloud. It showed no gel formation. In sample C, the
whole solution gelled with small crystallites becoming ap-
parent under a microscope with crossed polarizers; Fig. 8~c!.
To test the impact on crystallization of passing through
the two-phase boundary, a control sample with same c as in
A was held above T cloud at all times. This sample, labeled
A2, yielded predominately crystals. While much less pro-
nounced than in sample A, A2 also contained ramified clus-
 1958 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions
FIG. 8. Photomicrographs of morphologies obtained in samples A, B, and C
in Fig. 5 ~c5147, 250 and 330 mg/ml, respectively, 5% NaCl! incubated at
26 °C after the cloud point measurements. ~a! Sample A: tetragonal
lysozyme crystals coexist with a granular background. ~b! Sample B: spheri-
cal, isotropic gel beads, several hundred mm in diameter. ~c! Sample C
between crossed polarizers: crystals ~bright spots! that continued to grow
from within this gelled sample.
J. Chem. Phys., Vol. 107, No. 6, 8 August 1997
Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
ters; see Fig. 9~a!. Figure 9~b! shows one of the crystals that
grew in A2. Scanning of the focal plane through the crystal
revealed that it had incorporated some ramified clusters. The
cluster formation at c’s below those causing the whole solu-
tion to gel @Figs. 9~a! and 9~b!# is strongly suggestive of a
sol–gel transition in lysozyme.33
In solutions with c>c crit and 7% NaCl, extensive gela-
tion occurred even without cycling through T cloud. Thus gel
formation is not subject to a prior L–L phase separation,
although it appears to be enhanced by passage through the
binodal. These gels can be clearly distinguished from both
the high-c liquid and solid phases: They withstand high shear
stress without showing any macroscopic order. Also, gelled
samples turned translucent or slightly milky compared to the
transparent fluid state and were much more difficult to redis-
solve upon dilution. Furthermore, gels were considerably
more stable towards transformation into the crystalline phase
than the high-concentration liquid droplets that form during a
L–L separation. Highly twinned, irregular protein crystals
emerged from gels within a day, but significant portions of
the sample remained gelled for several days.
The 3% NaCl solutions showed no gelation, neither be-
low nor above room temperature. This suggests that the en-
hanced short-range protein–protein attraction at higher salt
concentration triggers or substantially expands the gelation
region for lysozyme solutions. Alternatively, one might in-
terpret the gelation as a consequence of thermal or salt-
induced denaturation. However, this is not supported by the
finding that lysozyme, obtained from dissolved orthorhombic
crystals grown at pH54.7 and at temperatures as high as
60 °C, was found to posses full biological activity.34
Besides the microscopic inspection, the L–L separation
and gel formation can also be detected from changes in the
speckle patterns obtained with a HeNe laser beam passing
through a sample. For instance, a solution with c
550 mg/ml and 5% NaCl showed no significant forward
scattering before the sample temperature reached T cloud.
Once cooled below this binodal, however, a fluctuating
speckle pattern in the forward direction appeared, and per-
sisted long after rewarming the sample above T cloud. The
time scale t for changes in the speckle intensity is given by
t 51/(Dq 2 ). 35 Here, D is the diffusivity, q
5(4 p n/l)sin(u/2) the magnitude of the scattering vector,
n51.33 the solution’s refractive index, l5633 nm the laser
wave length, and u the scattering angle which in our arrange-
ment was 1°–3°. With an estimated t '1 s we obtain an
upper bound for D<231028 cm2/s. The corresponding
lower bound on the cluster radius can be estimated from the
Stokes–Einstein relation D5kT/6p h r h as r h >100 nm.
From this crude observation, however, we can not distin-
guish between crystal nuclei or gel clusters as scatterers.
For practical purposes, it is interesting to note, that the
forward scattering intensity in the above observations de-
pended on the route of mixing the protein and buffer/salt
solutions. As mentioned in Sec. II, for our measurements
appropriate amounts of protein were directly dissolved into
the salt/buffer solution. However, when samples of the same
final composition were prepared by mixing buffered stock
 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions
FIG. 9. Photomicrographs obtained from sample A2 (c5147 mg/ml in 5%
NaCl, incubation at 35 °C overnight, T. above T cloud at all times!. Gel
clusters ~a! coexist with crystals ~b! which appear to have incorporated such
clusters.
J. Chem. Phys., Vol. 107, No. 6, 8 August 1997
Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
1959
FIG. 10. Typical ~c, NaCl!-phase diagram for lysozyme. Heavy solid line:
L–L coexistence curve ~based on two values, see solid triangles, for c cloud at
15 °C in 5% and 7% NaCl!. Dashed line: Solubility curve based on data
~open, downward pointing triangles! for tetragonal lysozyme at 15 °C from
Ref. 29. Arrows indicate two different pathways for obtaining a lysozyme
solution with c540 mg/ml and 4% NaCl. Vertical upward arrow: direct
dissolution of protein into solution. Inclined arrows: mixing of protein and
salt stock solutions of twice the final concentrations. Large lens-shaped re-
gion: possible local sample compositions during second mixing process,
with parts of the sample passing through the cross-hatched demixing region
even though the final solution lies outside.
solutions of protein and salt at twice their final concentration,
we observed much stronger forward scattering and much
shorter delay times for crystal nucleation. We tentatively as-
sign this enhanced forward scattering to the transient forma-
tion of gel clusters. As schematically depicted in Fig. 10, the
transient concentration heterogeneities unavoidably associ-
ated with the second mixing path are bound to pass part of
the sample through the L–L phase separation region, thus
enhancing gel cluster formation.
Based on these considerations and the observations pre-
sented in Fig. 8, we suggest that gelation underlies some of
the ‘‘amorphous precipitate’’ formation, which plagues
many protein crystal growth experiments. It should be em-
phasized that lysozyme is atypical for globular proteins in
this respect, since, as our experiments show, amorphous pre-
cipitation hardly occurs at the c’s <70 mg/ml that are com-
monly used for crystal growth. However, dynamic light scat-
tering data on lysozyme aggregation in supersaturated
solutions at 20 °C with 5% NaCl revealed two types of ag-
gregates: micron-size fractal clusters and compact clusters
intermediate in size between the lysozyme monomer and the
fractal clusters.36 Since in these experiments the samples
were mixed from concentrated stock solutions, based on Fig.
10, we tentatively identify these two populations as gel clus-
ters and crystal nuclei, respectively.
 1960 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions
IV. PHASE BEHAVIOR OF OTHER PROTEINS AND
GENERIC PHASE DIAGRAMS
The protein most thoroughly studied for its phase-
separation behavior is g-crystallin. Cloud point data7 and
solubility curves6 were determined for several of its native
variants. Similar to lysozyme, the authors report a two-phase
liquid region which is metastable with respect to crystalliza-
tion. Pronounced shifts in solubility and cloud point curves
were observed among the four variants of g-crystallin stud-
ied. The overall shape and relative positions of both phase
boundaries, however, remained essentially unchanged. In ad-
dition, the c crit for all variants were similar, and quite close to
the value found for lysozyme. No gel formation was re-
ported.
Recently we have performed screening experiments for
crystallization conditions in apoferritin/CdSO4 solutions.
Again, a gel phase appeared which was preceded by sample
turbidity. This phase transition occurred within the meta-
stable zone for crystallization.37 Qualitatively similar behav-
ior was observed in our laboratory with concanavalin A,
where crystals grew out of murky, turbid solutions. Further-
more, ‘‘metastable phase diagrams’’ in the ~pH, c!-plane
have been reported for bovine serum albumin, with the shape
of the demixing and gelation boundary ~see Fig. 60 in Ref.
38! closely resembling our (c,T) diagram for lysozyme. The
authors noted that these states were metastable, without ex-
panding on the origin of the metastability. Another example
for this complex phase behavior is the gelation of sickle cell
hemoglobin HbS. Under the same conditions that promote
the gelation of deoxy HbS, eventually crystals are formed.39
In addition, both a spinodal L–L phase separation line and a
gelation line have been reported for the HbS system.40
Note that the above examples span proteins of widely
different molecular weight, isoelectric point, quaternary
structure, and biological function. These observations sug-
gest that the features in the metastable phase diagram of
lysozyme are quite common to globular proteins. This can be
understood by considering the prevailing attractive protein
interactions under high salt conditions, which underlie the
observed phase behavior. Figure 11 shows schematic
temperature-density phase diagrams based on Monte Carlo
simulations for hard spheres with attractive Yukawa
potentials.12 In Fig. 11~a!, gas is the stable low-density phase
at high temperatures. On lowering the temperature, a phase
separation into coexisting gas and liquid phases occurs. At
high densities the gas–liquid and liquid–solid coexistence
curves intersect at the triple point. Further increase in density
leads to complete solidification. This standard phase diagram
undergoes pronounced changes @Fig. 11~b!# as the range d of
the interactions is shortened to about one-seventh of the
hard-core diameter s. The triple-point vanishes and the gas–
liquid coexistence region moves underneath the gas–solid
coexistence curve, forming a metastable state. An analytical
theory based on hard spheres with attractive square-well po-
tentials yielded the same result for d ,0.25s . 13 To relate the
results of Refs. 12 and 13 to protein solutions, we refer to
Mayer and McMillan’s treatment of solutions as gaseous
J. Chem. Phys., Vol. 107, No. 6, 8 August 1997
Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
FIG. 11. Schematic temperature-density phase diagrams for hard spheres
~diameter s! with Yukawa attraction ~range d! after Ref. 12. ~a! Standard
phase diagram with a stable gas–liquid coexistence region below the critical
point ~CP!. The sublimation curve intersects the gas–liquid curve at the
triple point ~TP!. ~b! Phase diagram for d < s /7. The sublimation curve
moves above the then metastable gas–liquid coexistence region.
suspensions.41 The gas phase can then be identified with the
fluid state of uniform protein dispersion, the gas–liquid co-
existence curve as the liquid–liquid binodal and the sublima-
tion line as the crystal solubility curve. Similarly, the pres-
ence of a sol–gel transition has been anticipated in models
employing adhesive hard spheres.42,43 The percolation
threshold obtained in these calculations originated along the
L–L phase boundary, as suggested by our experimental ob-
servations.
Monte Carlo simulations for g-crystallin and lysozyme
provide considerable evidence that the interaction between
the macroions in protein solutions is limited to the short
range underlying the diagram of Fig. 11~b!.28 Most recently,
attractive Yukawa potentials were used44 to model experi-
mental x-ray structure factors of lysozyme and g-crystallin
and to reproduce the spinodal lines from previous phase
separation experiments in g-crystallin.9 The range of the at-
tractive interactions thus extracted from the experiments was
only of the order of one-tenth of the protein diameter. Hence,
metastability of the L–L coexistence region can be expected
in general for systems of ~colloidal! particles with dominat-
ing short-ranged centrosymmetric attraction. Colloidal mod-
els of protein interactions have also proven successful for
descriptions of the solution behavior of proteins, including
lysozyme’s solubility,14 phase separation behavior,3 and in-
teraction effects on light scattering.16
Based on the above experimental and theoretical consid-
erations we put forth a generic phase diagram for globular
proteins. As depicted in Fig. 12~a!, three zones are identified
within the solid–liquid coexistence region. Zone II represent
the metastable L–L region and Zone III the gelation region.
Note that the latter is either very narrow or absent for 3%
NaCl–lysozyme solutions. Zone I is the region most suitable
for protein crystal growth. Previous static and dynamic light
scattering16,45 and low angle x-ray scattering46 studies in this
zone showed that aggregation during the induction period for
nucleation was insignificant. This is not surprising, since su-
 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions
FIG. 12. Postulated generic phase diagram for globular proteins. ~a! Normal
solubility, ~b! retrograde solubility. Solution conditions below and above the
solubility curve in ~a! and ~b!, respectively, are metastable with respect to
crystallization. Zone II: separation into two metastable immiscible liquid
phases ~formation of cloudy precipitate!. Gelation in Zone III. Cross-
hatched area: region of incomplete gelation ~amorphous precipitation!. Op-
timal region for protein crystallization: Zone I outside cross-hatched area.
persaturations are relatively modest and interfering second-
ary phase transitions are absent. The shaded area in Zone I,
however, indicates the possibility that pregelation clustering
can affect crystal growth even in this zone. Sample A2 in our
experiments ~see Fig. 9! displayed such behavior with the
simultaneous presence of crystals and ramified gel clusters.
The extent of the shaded area is likely to depend on the
specific protein and sample preparation procedure.
The phase diagram in Fig. 12~a! applies to proteins with
normal temperature dependence of the solubility, i.e., with
an increase in c sat with temperature. However, both HbS and
apoferritin have retrograde solubility, i.e., c sat decreases with
increasing temperature. Interestingly, the reported tempera-
ture dependence of the metastable states in HbS follows re-
versed temperature dependence as well.40 Therefore, as sche-
matically depicted in Fig. 12~b!, the phase diagram of
proteins with retrograde solubility appears to be a mirror
image of Fig. 12~a!.
As our work and the above references suggest, the loca-
tion of the various phase boundaries can vary considerably
with changes in solution conditions ~lysozyme, bovine serum
albumin! or differences in the surface groups of the protein
~hemoglobin and g-crystallin!. Therefore, some proteins
could develop metastable states, which are undesirable for
controlled crystal growth, at rather low c’s. The overall
shape of the phase diagram, however, should be retained.
Hence, any search for advantageous protein crystallization
conditions can take advantage of information on precipitate
formation and associated solution conditions. To date, this
information has frequently been considered useless. From
our investigation it appears, however, that such data can pro-
vide valuable guidance for adjustments of pH, temperature or
precipitating salt concentration to move the solution condi-
tions towards the ‘‘crystallization slot’’ of zone I in the sche-
matic phase diagram.
VI. CONCLUSIONS
We have determined the cloud point curves for the meta-
stable liquid–liquid phase transition in lysozyme solutions
J. Chem. Phys., Vol. 107, No. 6, 8 August 1997
Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
1961
with different NaCl concentration. Combining these results
with crystal solubility data for the same solution conditions,
phase diagrams for lysozyme were obtained. Relations be-
tween crystal growth and metastable demixing were ex-
plored. Within the two-phase liquid region, nucleation was
enhanced and microcrystalline precipitates formed. Outside
of this region, a sol–gel transition was found at higher salt
concentrations. Both, the L–L phase separation as well as
gelation appear closely related to two common forms of un-
desirable morphologies frequently encountered in protein
crystallization: cloudy and amorphous precipitate, respec-
tively. We associate cloudy precipitates with L–L phase
separations. Amorphous precipitate formation, on the other
hand, is often ascribed to the rapid aggregation prevailing at
excessive supersaturations.10 We rather suggest as a possible
origin of amorphous precipitate formation the presence of an
additional sol–gel transition within the solid–liquid coexist-
ence region. The intimate coupling between demixing, gela-
tion and crystallization results in complex kinetics of the
crystallization process. As we illustrated with an example on
mixing procedures, details of the thermal and preparative
history of a sample can result in differences in the outcome
of crystal growth experiments under seemingly equivalent
experimental conditions.
Most significantly, the presence of a metastable two-
phase liquid region and a sol–gel transition within supersatu-
rated lysozyme solutions appear to be intrinsic features for a
large variety of globular proteins. This finding led us to a
generic shape for phase diagrams of globular proteins, with a
limited region suitable for high-quality crystal growth.
Clearly, further experimental and theoretical work is re-
quired to test our hypothesis.
ACKNOWLEDGMENTS
It is a pleasure to acknowledge valuable experimental
support and suggestions by B. R. Thomas, and M. Banish
who also carried out the atomic absorption spectroscopy
measurements. We are equally grateful to C. Carter, Jr., A.
Chernov, F. Ferrone, K. Murphy, R. Nagel, and J. Wienzek
for pointing out some of the references cited in this work. L.
Carter has expertly prepared the figures. This research has
been supported by NASA ~Grants No. NAG8-1161 and No.
NAG8-1168! and the State of Alabama through the Center
for Microgravity and Materials Research at the University of
Alabama in Huntsville.
1
C. Ishimoto and T. Tanaka, Phys. Rev. Lett. 39, 474 ~1977!.
2
V. G. Taratuta, A. Holschbach, G. M. Thurston, D. Blankschtein, and G.
B. Benedek, J. Phys. Chem. 94, 2140 ~1990!.
3
M. L. Broide, T. M. Tomine, and M. D. Sasowsky, Phys. Rev. E 53, 6325
~1996!.
4
B. M. Fine, J. Pande, A. Lomakin, O. Ogun, and G. B. Benedek, Phys.
Rev. Lett. 74, 198 ~1995!.
5
C. Liu, A. Lomakin, G. M. Thurston, D. Hayden, A. Pande, J. Pande, O.
Ogun, N. Asherie, and G. B. Benedek, J. Phys. Chem. 99, 454 ~1995!.
6
C. R. Berland, G. M. Thurston, M. Kondo, M. L. Broide, J. Pande, O.
Ogun, and G. B. Benedek, Proc. Natl. Acad. Sci. USA 89, 1214 ~1992!.
7
M. L. Broide, C. R. Berland, J. Pande, O. Ogun, and G. B. Benedek, Proc.
Natl. Acad. Sci. USA 88, 5660 ~1991!.
 1962 M. Muschol and F. Rosenberger: Phase separation in lysozyme solutions

8 25
P. Schurtenberger, R. A. Chamberlin, G. M. Thurston, J. A. Thomson, and J. V. Senger, in Phase Transitions, edited by M. Lévy, J. C. Le Guillou,
G. B. Benedek, Phys. Rev. Lett. 63, 2064 ~1989!. and J. Zinn-Justin ~Plenum, New York, 1980!, p. 95.
9 26
J. A. Thomson, P. Schurtenberger, G. M. Thurston, and G. B. Benedek, E. Stanley, Introduction to Phase Transitions and Critical Phenomena
Proc. Natl. Acad. Sci. USA 84, 7079 ~1987!. ~Oxford, New York, 1971!.
10
A. Ducruix and R. Giegé, in Crystallization of Nucleic Acids and Proteins. 27
J. Jollès and J. Berthou, FEBS Lett. 23, 21 ~1972!.
A Practical Approach, edited by A. Ducruix and R. Giegé ~Oxford, New 28
A. Lomakin, N. Asherie, and G. B. Benedek, J. Chem. Phys. 104, 1646
York, 1992!, Chap. 4. ~1996!.
11
Y. Georgalis, A. Zouni, W. Eberstein, and W. Saenger, J. Cryst. Growth 29
E. Cacioppo and M. L. Pusey, J. Cryst. Growth 114, 286 ~1991!.
126, 245 ~1993!. 30
F. Ewing, E. Forsythe, and M. L. Pusey, Acta Cryst. D 50, 424 ~1994!.
12
M. J. Hagen and D. Frenkel, J. Chem. Phys. 101, 4093 ~1994!. 31
F. P. Schwarz, Thermochim. Acta 147, 71 ~1989!.
13
N. Asherie, A. Lomakin, and G. B. Benedek, Phys. Rev. Lett. 77, 4832 32
B. Lorber, G. Jenner, and R. Giegé, J. Cryst. Growth 158, 103 ~1996!.
~1996!.
14
33
D. Stauffer, A. Coniglio, and M. Adam, Adv. Polym. Sci. 44, 103 ~1982!.
D. Rosenbaum, P. C. Zamora, and C. F. Zukoski, Phys. Rev. Lett. 76, 150 34
J. Berthou and P. Jollès, Biochim. Biophys. Acta 336, 222 ~1974!.
~1996!. 35
B. J. Berne and R. Pecora, Dynamic Light Scattering: With Applications to
15
A. George and W. W. Wilson, Acta Cryst. D 50, 361 ~1994!.
Chemistry, Biology, and Physics ~Wiley, New York, 1976!.
16
M. Muschol and F. Rosenberger, J. Chem. Phys. 103, 10424 ~1995!. 36
17 Y. Georgalis, J. Schüler, J. Frank, M. D. Soumpasis, and W. Saenger,
B. R. Thomas, P. G. Vekilov, and F. Rosenberger, Acta Cryst. D 52, 776
~1996!. Adv. Colloid Interf. Sci. 58, 57 ~1995!.
18
P. G. Vekilov and F. Rosenberger, J. Cryst. Growth 158, 540 ~1996!.
37
M. Muschol ~unpublished!.
19
P. G. Vekilov, L. A. Monaco, B. R. Thomas, V. Stojanoff, and F. Rosen-
38
A. H. Clark and S. B. Ross-Murphy, Adv. Polym Sci. 83, 57 ~1987!.
berger, Acta Cryst. D 52, 785 ~1996!.
39
W. A. Eaton and J. Hofrichter, Adv. Prot. Chem. 40, 63 ~1990!.
20
A. J. Sophianopoulos, C. K. Rhodes, D. N. Holcomb, and K. E. Van
40
P. L. San Biagio and M. U. Palma, Biophys. J. 60, 508 ~1991!.
Holde, J. Biol. Chem. 237, 1107 ~1962!.
41
D. A. McQuarrie, Statistical Mechanics ~Harper and Row, New York,
21
F. Rosenberger, S. B. Howard, J. W. Sowers, and T. A. Nyce, J. Cryst. 1976!, Chap. 15.
Growth 129, 1 ~1993!.
42
Y. C. Chiew and E. D. Glandt, J. Phys. A. 16, 2599 ~1983!.
22
P. Guenoun, P. R. Gastaud, F. Perrot, and D. Beysens, Phys. Rev. A 36,
43
W. G. Kranendonk and D. Frenkel, Mol. Phys. 64, 403 ~1988!.
44
4876 ~1987!. M. Malfois, F. Bonneté, L. Belloni, and A. Tardieu, J. Chem. Phys. 105,
23
W. J. Ray, Jr. and C. E. Bracker, J. Cryst. Growth 76, 562 ~1986!. 3290 ~1996!.
24
F. Reiss-Husson, in Crystallization of Nucleic Acids and Proteins. A Prac- 45
M. Muschol and F. Rosenberger J. Cryst. Growth 167, 738 ~1996!.
tical Approach, edited by A. Ducruix and R. Giége ~Oxford, New York, 46
A. Ducruix, J. P. Guilloteau, M. Ries-Kautt, and A. Tardieu, J. Cryst.
1992!, Chap. 8. Growth 168, 28 ~1996!.

J. Chem. Phys., Vol. 107, No. 6, 8 August 1997

Downloaded 30 Mar 2013 to 128.233.210.97. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions