Protein Purification by Bulk
Crystallization: The Recovery
of Ovalbumin
Russell A. Judge, Michael R. Johns,* and Edward T. White
Department of Chemical Engineering, The University of Queensland,
Brisbane, Australia, 4072
Received November 22, 1994lAccepted May 30, 1995
Crystallization is used industrially for the recovery and
purification of many inorganic and organic materials.
However, very little is reported on the application of bulk
crystallization for proteins. In this work, ovalbumin was
selected as a model protein t o investigate the feasibility
of using bulk crystallization for the recovery and purifi-
cation of proteins. A stirred I - L seeded batch crystallizer
was used t o obtain the crystal growth kinetics of ovalbu-
min in ammonium sulfate solutions at 30C. The width of
the metastable region, i n which crystal growth can occur
without any nucleation, is equivalent t o a relative super-
saturation of about 20. The bulk crystallizations were un-
dertaken within this range (using initial relative supersat-
urations less than 10) and nucleation was not observed.
The ovalbumin concentration in solution was measured
by UV absorbance and checked by crystal content mea-
surement. Crystal size distributions were measured both
by using a Malvern Mastersizer and by counting crystals
through a microscope. The crystal growth rate was found
to have a second-order dependence upon the ovalbumin
supersaturation. While there is no discernible effect of
ammonium sulfate concentration at pH 4.90, there is a
slight effect at higher pH values. Overall the effect of
ammonium sulfate concentration is small compared to
the effect of pH, for which there is a 10-fold increase in
the growth rate constant, kG<,,over the range pH 4.6-5.4.
To demonstrate the degree of purification which can be
achieved by bulk crystallization, ovalbumin was crystal-
lized f r om a solution containing conalbumin (80,000 Da)
and lysozyme (14,600 Da). After one crystallization and a
crystal wash, ovalbumin crystals were produced with a
protein purity greater than 99%. No contamination by the
other proteins was observed when using overloaded so-
dium dodecyl sulfate-polyacrylamide gel electrophore-
sis (SDS-PAGE) stained with Coomassie blue stain and
only trace amounts of lysozyme were observed using a
silver stain. The presence of these other proteins i n so-
lution did not effect the crystal growth rate constant,
k,,,. The study demonstrates the feasibility of using bulk
crystallization for the recovery and purification of oval-
bumin. It should be readily applicable to other protein
systems. 0 1995 John Wiley & Sons, Inc.
Key words: ovalbumin bulk crystallization crystal
growth rate nucleation purification
INTRODUCTION
For many materials, including organic compounds such as
antibiotics, organic acids, and amino acids, bulk crystalli-
* To whom all correspondence should be addressed
Biotechnology and Bioengineering, Vol. 48, Pp. 316-323 (1995)
0 1995 John Wiley & Sons, Inc.
zation in agitated vessels is an important, cheap, and easily
scaled-up separation and purification process. l 3 It provides
a pure solid product in a form convenient for further han-
dling, storage, or packaging.
Protein crystallization literature is orientated toward the
use of microcrystallization techniques, such as sitting and
hanging drops, for the spontaneous nucleation and growth
of large, single, highly ordered crystals from pure protein
solutions for X-ray diffraction studies. As this requires the
screening of numerous crystallization conditions, with
much trial and error, protein crystallization has gained a
reputation for being a difficult and exacting task. Although
successful in identifying the crystal growth conditions for
many proteins, these microtechniques are unsuitable for
large-scale protein crystallization.
In this study, the feasibility of applying bulk crystalliza-
tion in an agitated vessel for the recovery and purification of
proteins is examined using ovalbumin as a model protein.
To do this, ovalbumin crystal growth kinetics were studied
in bulk solution under a range of operating conditions (in-
cluding pH, ammonium sulfate concentration, and the pres-
ence of other proteins) using a stirred, seeded, batch crys-
tallizer. Studies of this sort are rare for proteins but are
essential for the design of bulk crystallization processes.
Crystallization kinetic data have been reported for
canavalin,2 l y ~ o z y r n e , glucose
~,~ isomerase,22 and insu-
lin. l6
For canavalin and lysozyme, the crystal face growth rates
of a small number of individual crystals in pure protein
solutions were measured in a small growth cell using a
microscope to record the change in crystal size. For both
materials the growth is controlled by a surface integration
mechanism, and for lysozyme, the effect of temperature and
salt concentration is such that any change that caused a
decrease in solubility caused a decrease in the growth rate
c o n ~ t a n t The
. ~ crystals in these studies, however, were not
freely moving in an agitated bulk solution so crystal growth
behavior may have been influenced by contact with the
growth chamber. As only a small number of crystals were
studied, growth dispersion may also be influencing the re-
sults.
A crystallization study with pure glucose isomerase so-
lution22 was performed in bulk in an unstirred vessel using
pressure to generate supersaturation. Crystal generation,
CCC 0006-3592/95/040316-08
however, was not controlled (crystals were generated by
primary nucleation), and it is unclear if the reported in-
crease in the amount of crystal with increase in pressure was
due to higher growth rates or higher nucleation rates. The
smaller final crystal size indicates that high pressure in-
creased nucleation.
Schlichtkrull'6 was the only investigator to use a stirred
vessel to obtain crystallization kinetic data. He crystallized
insulin from pure solution. Size-independent growth and a
second-order dependence of growth rate upon supersatura-
tion were observed. No results were given for the effect of
crystallization variables upon the growth rate kinetics.
Typically the growth rates of the crystal faces reported
for canavalin, lysozyme, and insulin are about 1 p d m i n .
For inorganic and simple organic compounds, bulk crys-
tallization in stirred vessels achieves a high-purity product
free of solvent and solution contaminants. However, up to
SO% of a protein crystal volume is comprised of voids large
enough to contain solvent molecules,' so a protein crystal
will always contain an amount of solvent. A high degree of
purification of a protein from other proteins by bulk crys-
tallization in a stirred vessel should be possible; however,
there are no published studies of such purification. In this
work, the purification of ovalbumin from a mixture of
smaller (lysozyme) and larger (conalbumin) molecular
weight proteins was investigated and the resulting growth
rate kinetics determined.
Ovalbumin, the principal protein component of chicken
egg white, was selected for study since reasonably large
quantities of pure ovalbumin could be obtained easily at low
cost and it crystallizes relatively easily. Also, extensive
solubility data are available. l 9
Further, the information from this study should be useful
in indicating the value of the bulk crystallization of pro-
teins.
MATERIALS AND METHODS
Ovalbumin
Ovalbumin (45,000 Da) is a phosphoglycoprotein consist-
ing of a single chain of 385 amino acid residues with a
single disulfide bond.," It has an isoelectric point at pH
4.S8.23Three naturally occurring, phosphorylated forms of
ovalbumin exist: A,, A,, and A,, which have two, one, and
zero phosphorylated sites per molecule, respectively. The
relative proportions of A , :A,:A, can be determined using
high-performance liquid chromatography (HPLC). These
forms are not separated by repeated crystallization.8
Ovalbumin is irreversibly converted to S-ovalbumin on
storage in alkaline solutions. l 8 S-ovalbumin is distin-
guished from native ovalbumin by its greater resistance to
heat denaturation. In every other respect the properties of
both forms of the protein are identical. The heat denatur-
ation method of Smith and Back18 was used to determine
the proportion of ovalbumin present as S-ovalbumin in our
samples.
JUDGE, JOHNS, AND WHITE: PROTEIN PURIFICATION BY BULK CRYSTALLIZATION
X-ray diffraction data for ovalbumin have only recently
been published for both the monoclinic" and triclinic crys-
tal forms.*' As this study was conducted at 30"C, at solution
conditions similar to those at which Miller et al." (working
at 37C) observed monoclinic crystals, the crystal form in
this article was expected to be monoclinic.
The void volume fraction of the ovalbumin crystal lattice
available for internal solvent was calculated, using the
method of Mikol and Giege," as 33%, taking the specific
volume v,, = 0.74 cm3/g for the ovalbumin molecule and
the crystal density p,. = 1.268 g/cm3.'
Source of Ovalbumin
Purified crystalline ovalbumin was prepared from frozen
chicken egg white using a method similar to Sorensen and
Hoyrup" as modified by Warner.23
The frozen egg white (10 kg) was thawed and gently
mixed to degrade membrane material. An equal volume of
saturated ammonium sulfate solution was added slowly
while stirring. After standing for 2 h, the precipitate was
removed by centrifugation (14,00Og, 25"C, 30 min) and
discarded. The filtrate was adjusted to pH 4.6 with 0.1 M
H,SO,, briefly stirred, and left to crystallize overnight at
room temperature. Crystalline ovalbumin was separated
from the liquor by centrifugation (14,00Og, 25"C, 10 min)
and dissolved in water (700 mL). Saturated ammonium sul-
fate solution was then slowly added until a slight permanent
turbidity was observed (23-25 g ammonium sulfate1100 g
water). The solution was briefly stirred to promote nucle-
ation and left to crystallize overnight. After centrifugation,
the crystals were washed and stored in a 10:7 by volume
saturated ammonium sulfate-water mixture at 4C. The
ovalbumin so prepared showed no contaminating proteins
[overloaded sodium dodecyl sulfate-polyacrylamide gel
electrophoresis (SDS-PAGE) using Coomassie blue stain
and HPLC].
Analysis
The ultraviolet (UV) absorbance of ovalbumin solutions
was measured at 280 nm using a Hitachi U110 spectropho-
tometer (Tokyo, Japan) and the ovalbumin solution concen-
tration calculated using an absorbance coefficient of A ;Trn
= 7.0.,'
In this work all concentrations are given as ratios to water
and expressed in the units of g1100 g of water. The driving
force for crystallization is taken as the relative supersatura-
tion, (T = [(OvlW)/(Ov/W),,J - I , where (OvlW)is the
ovalbumin bulk solution concentration and (OvlW),,, is the
ovalbumin solubility concentration.
Ovalbumin purity was measured by HPLC and SDS-
PAGE analysis. HPLC was performed on a Waters (Waters
Associates, MA) system operating at ambient temperature,
equipped with a Waters 486 tunable UV absorbance detec-
tor set at 280 nm. Samples (20 pL) containing approxi-
mately 1.5 mglmL ovalbumin were injected onto a Bio-gel
317
TSK DEAE-5-pw (75 x 7.5 mm) anion exchange column
(Bio-Rad, Richmond, CA). A gradient system comprising
solvents A (Tris 20 mM,pH 8.0) and B (solvent A plus 0.5
M NaC1) was run from 0 to 100% B in 20 min.
The SDS-PAGE analysis used precast Tris-HCl Bio-Rad
gels (15%) run on a Mini-Protean 11 electrophoresis cell
(Bio-Rad) powered by a Bio-Rad model 25012.5 power sup-
ply. Broad range molecular weight markers covering the
range 6500-200,000 Da were used. The gels were stained
with 0.1% Coomassie blue R-250 in water-methanol-acetic
acid (5:4:1 by volume) and destained in the same solution
without the dye.
Ammonia concentration was measured using an ammonia
specific ion probe (Orion Research, Boston, MA).
Nucleation Thresholds
If practical, crystallization is best carried out in the absence
of nucleation, i.e., in the metastable zone below the nucle-
ation threshold using controlled seeding. There are two nu-
cleation thresholds, the primary, below which crystal-free
solutions do not nucleate, and the secondary (in the pres-
ence of solute crystals), below which crystals grow but
nucleation does not occur. The secondary threshold is of the
greater interest in bulk crystallizations.
For the primary nucleation threshold (nucleation in crys-
tal-free solutions), 100 mL of solution of known ovalbumin
composition was contained in a 250-mL Schott bottle main-
tained at 30C and stirred by a magnetic stirrer. As agitation
has a complex influence and usually enhances nucleation,
the bottles were all stirred by identical stirrers at the same
speed of 180 rpm.
The initial solution pH was set by adjustment with either
0.5 M NH,OH or 0.1 M H,SO, solution. Saturated ammo-
nium sulfate solution then was added slowly at about 2
mllmin. Nucleation was deemed to have occurred when the
solution became a permanent grey or white color. A micro-
scope was used to check that the precipitate was crystalline.
The ovalbumin and ammonium sulfate concentration in so-
lution at the moment of nucleation was determined by both
a mass balance on the quantities added and by direct anal-
ysis. Ovalbumin purity was also checked to ensure that
protein degradation had not occurred.
The secondary nucleation threshold (nucleation in the
presence of ovalbumin crystals) was determined in the same
way, except that the initial solution contained 0.8 g/L of
30-pm crystals (size as volume equivalent size) of ovalbu-
min. The initial crystals were large enough that the onset of
fine crystals as nuclei could still be observed.
Crystal Growth Kinetics
It is assumed that the crystal growth rate, G , varies with
supersaturation, u, as
G = kGuua (1)
The growth rate, G, is the change in the characteristic di-
mension, L (here taken as the volume equivalent size), with
318 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 48, NO. 4, NOVEMBER 20, 1995
time, k,, is the growth rate constant, and a is the kinetic
order constant. The order, a, depends on the controlling
crystallization mechanism. For growth of a crystal, the sol-
ute must diffuse through the solution to the crystal surface
and then become integrated into the growing crystal lattice.
If the diffusion is the rate-controlling mechanism, a usually
is 1. If the surface integration is the controlling mechanism,
a is often higher.14
Ovalbumin seed crystals were produced by breakage us-
ing high-speed stirring. A crystal slurry (350 mL) in a 600-
mL glass beaker was stirred at 500 rpm for 4 h. A small
amount of water was then added to dissolve any fines. The
remaining seed crystals were stored in saturated solution
(23.5 g ammonium sulfate/100 g water) at pH 4.75 at a
crystal content of 68 mg ovalbumidg solution. The mean
crystal size (volume equivalent) of the seed was 2.5 pm.
A stirred, seeded batch crystallizer was used to determine
the crystal growth rates. The flat-flanged glass crystallizer
(1 L) was clamped to the glass lid with a wire clasp. The
unbaffled vessel was stirred with a single stainless steel,
axial turbine impeller at 35 rpm using an overhead mixer
(Janke and Kunkel, Staufen, Germany). Too high a speed
causes crystal breakage. A constant temperature of 30C
was maintained by immersion in a water bath.
To prepare the feed solution, ovalbumin was dissolved in
water and any denatured protein was removed by centrifu-
gation (9600g, 25C, 10 min). Ammonium sulfate solution
then was gently mixed into the clarified protein solution to
give the required concentrations. If necessary, the pH was
adjusted. Seed crystals (about 0.8 g) were added to the
supersaturated solution (600 g) to start the run. During crys-
tallization, crystal and solution samples were taken for the
measurement of crystal size distribution, crystal content,
and ovalbumin and ammonium sulfate concentrations. The
pH was also monitored and provision was made for adjust-
ment, but generally this was unnecessary as ammonium
sulfate concentration and solution pH remained substan-
tially constant during all crystallization runs.
Ovalbumin solution concentration was measured by UV
absorbance. Crystal content, which gives a mass balance
check on the ovalbumin solution concentration, was deter-
mined by filtering and washing the crystals, then dissolving
in a known amount of water and measuring the absorbance.
During crystallization, the crystals formed loose agglom-
erates of small numbers of crystals. This behavior was also
reported by Sorensen and Hoyrup and could not be pre-
vented, either using additives or by changing operating con-
ditions. Crystal growth rates then were measured using two
techniques.
In the first, the seed size distribution was measured using
a Malvern MasterSizedE (Malvern Instruments, Malvern,
United Kingdom), which measures volume equivalent size.
As the agglomerates created difficulties in obtaining size
distributions of the product crystals, the growth rates were
calculated using the seed crystal distribution and the drop in
ovalbumin concentration. This method assumes that crystal
growth is size independent with no growth dispersion, that
there is no nucleation, and that the crystals grow as indi-
vidual crystals without agglomeration or breakage.
The second method involved the measurement of crystal
size distributions from enlargements of photographs of sam-
pled crystals taken through a microscope. The measured
crystal dimensions were converted to volume equivalent
size, taking a given crystal shape. Growth rates calculated
using both techniques were in good agreement, so no dis-
tinction is made between them when presenting results.
When the ovalbumin solution had fallen to near the equi-
librium solubility and the growth rate measurements were
deemed complete, the crystallizer contents were emptied
into a bottle and stirred at 30C (for about 11 days) to give
the ovalbumin solubility concentration, c*, which is neces-
sary for the accurate calculation of the supersaturation.
Purification Experiments
A solution (600 g) containing 12 g ovalbumin, 1 g conal-
bumin (Sigma Aldrich), 1 g lysozyme (Mauri Laboratories,
Moorebank, Australia), and ammonium sulfate (27 g/100 g
water) at pH 4.9 and 30C was prepared and centrifuged
(lO,OOOg, 25"C, 5 min) to remove undissolved residue and
denatured protein. Duplicate seeded batch crystallizations
(300 g each) were then performed to determine the effect of
these protein contaminants on the rate of crystallization and
upon the crystal purity. The data were treated using the
methods described previously.
After 48 h, the crystals produced were recovered by cen-
trifugation (9600g, 25"C, 5 min), washed with ovalbumin
saturated solution (27 g ammonium sulfate/100 g water),
and filtered using 0.45-pm cellulose acetate membrane fil-
ters (Sartorius, Gottingen, Germany). The crystals were
washed with 10:7 (v/v saturated ammonium sulfate-water)
to remove soluble protein in the adhering liquor. To assess
their purity, the final crystals were dissolved in water, fil-
tered (0.45 pm membrane filters), and frozen for HPLC and
SDS-PAGE analysis.
The SDS-PAGE using Coomassie blue R-250 stain was
done with lysozyme and conalbumin at different concentra-
tions to obtain an estimate of the smallest detectable con-
centration. Overloaded ovalbumin crystal samples (6 and 9
k g per well) were run on SDS-PAGE stained with
Coomassie blue R-250 to assess contaminant protein levels.
For higher resolution silver staining was also used.
RESULTS AND DISCUSSION
Ovalbumin
The proportions of the phosphorylated forms A,:A,:A, in
the feed were 79:18:3, which is in substantial agreement
with the reported values for ovalbumin of 85:12:3.* Oval-
bumin preparations contained 40 ? 10% S-ovalbumin,
which is typical for eggs obtained from retail stores."
JUDGE, JOHNS, AND WHITE: PROTEIN PURIFICATION BY BULK CRYSTALLIZATION
Nucleation Thresholds
Figure 1 shows the concentrations at which primary and
secondary nucleation were observed at pH 5.2. Ovalbumin
solubility data are from Sorensen and Hoyrup'' and Judge.'
Figure 2 shows the results for all pH, with the thresholds
now expressed in terms of the amount of extra ammonium
sulfate that can be added to an ovalbumin solution at equi-
librium before nucleation is observed. For both primary and
secondary nucleation, the threshold is a constant with an
average value of 8.4 ? 0.5 g AS/100 g water and 6.3 -t 0.5
g AS1100 g water, respectively. The values appear to be
independent of solution pH and ovalbumin concentration.
In terms of ovalbumin concentration, the primary nucle-
ation threshold can be expressed as a critical supersatura-
tion, ucp= (Ov/Wj,,,l(OvlW),,, - 1 --- 50, while the cor-
responding value for secondary nucleation, ucs, is approx-
imately 20. Here (OvlW),,, is the ovalbumin solution
concentration at the nucleation threshold and (OvlW),,, is
the ovalbumin solubility concentration at the same ammo-
nium sulfate concentration, pH, and temperature.
The secondary nucleation threshold was investigated at
only one ovalbumin concentration, but at several values of
pH. In Figure 1, the curves for the primary and secondary
nucleation thresholds have been drawn on the assumption
that the threshold critical supersaturations are constant, in-
dependent of ovalbumin concentration. The secondary nu-
cleation threshold, as expected, is lower than that for pri-
mary nucleation.
For ovalbumin a wide metastable zone exists, which
would allow the use of crystallizer feeds with an initial
supersaturation of at least 10. In this region, crystals can be
grown without nucleation and the crystal size distribution is
controlled solely by seeding. Good final crystal sizes and
crystallization times should be able to be obtained while still
achieving yields of 90%.
6 t \ HE'CION
-1
e 4
?
0 3
1
0
20 24 28 32 36 40 44
Ammonlum Sulphate g, 100 g Water
Figure 1. Nucleation thresholds and metastable region for ovalbumin at
pH 5.2 and 30C.
319
 'j
L
- Figure 3. Ovalbumin crystals grown from seed crystals at pH 4.90 and
0 2 4 6 8 LO 12 I4 16 18
Ovalburnin g 100 g Watpr
Figure 2. Influence of ovalbumin concentration and bulk solution pH on
ovalbumin pnmary and secondary nucleation thresholds
For hen egg lysozyme, it was reported that crystals ap-
pear in larger amounts with increasing super~aturation'~and
that precipitation (which was difficult to distinguish from
showers of microcrystals) occurred at 30-100 times the sol-
ubility concentration. This is comparable with our results.
Nucleation thresholds in the presence of crystals were not
reported however.
Knowledge of the nucleation thresholds may have as-
sisted studies producing ovalbumin crystals for X-ray crys-
tallography. Miller et a1.I' experienced showers of micro-
crystals working close to the nucleation thresholds with
both seeded and unseeded solutions. To avoid this, Stein et
a1." worked at conditions further away from the nucleation
thresholds, which gave large crystals but required 1-3
months before any crystals were observed since no seed
crystals were added.
Crystal Growth Kinetics
Seeded batch crystallizations of ovalbumin were performed,
without nucleation, over a wide range of conditions. Figure
3 illustrates typical product crystals grown from seed crys-
tals in this study. The different operating conditions had no
effect upon the final crystal shape, typically having length-
width-thickness ratios of 8: 1:0.2. Similar crystals were
grown by Miller et al." using a hanging drop with vapor
equilibrium and by Sorensen and Hoyrup," who reported
that the larger crystals grown using unstirred flasks had
dimensions 13:1:0.07.
Figure 4 illustrates the successive crystal size distribu-
tions obtained using the microscope counting technique.
The crystal size displayed is the size of the longest face. The
parallel nature of the curves illustrates size-independent
crystal growth and negligible growth dispersion. The results
320 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 48, NO. 4, NOVEMBER 20, 1995
30C in 22 g ammonium sulfatei100 g water solution.
20
also confirm that growth does not cease upon reaching any
particular crystal size. The calculated total number of crys-
tals remained constant, indicating that little nucleation, ag-
glomeration, or breakage occurred.
Good agreement (within _t 1.5%) was found between the
ovalbumin solution concentrations measured by UV absor-
bance and by crystal content.
TheoryI4 indicates that the crystal growth rate can be
obtained from the magnitude of the shift of the parallel
curves in Figure 4 divided by the time interval. The crystal
growth rates of ovalbumin so determined are plotted in Fig-
ure 5 on a log-log scale as a function of the relative super-
saturation. Note that the growth rate now is the change in
the volume equivalent size of a crystal, not the change in
any prescribed face dimension. Face growth rates are easily
converted to a volume equivalent size basis using the indi-
cated crystal shape. There is a second-order dependence of
growth rate upon the supersaturation. This accords with
previous work on other proteins which reported a second-
I I
"I
0
c rl,tcli irr~giii (/mo
Figure 4. Ovalbumin crystal size distnbutrons dunng a crystallization at
pH 4 9 and 24 g ammonium sulfateilO0 g water
 , 1 , , , , I 1 1 ,
001 I
0 1 1 10 01 1 in
Kelatl\r iupei s,ituration ( u ) R c l < i t i \ e yiipel-ccltiir d t i o n ( 0 )

Figure 5. Crystal growth rate as a function of supersaturation at pH 4.9 Figure 6. Effect of ammonium sulfate concentration upon the crystal
and 24 g ammonium sulfateilO0 g water growth rate kinetics at pH 4.9

order dependence of growth rate on supersaturation for in- easily. For production purposes, the crystallization time
sulin16 and surface kinetic controlled growth mechanisms would be dramatically reduced by using higher crystal con-
for l y ~ o z y m eand
~ , ~canavalin.2 tents. This is illustrated in Figure 8, where the behavior of
Growth rates for ovalbumin range up to several microme- batches using the seed crystals has been predicted from the
ters per hour (i.e., 0.1 p d m i n ) on a volume equivalent size growth kinetics obtained in this study. The time to recover
basis. Taking the ratio of dimensions for an ovalbumin crys- 95% of the ovalbumin (above solubility values) would be
tal as 8: 1:0.2 gives the ratio of the growth rate of the length reduced from 18 h, when seeded at 0.2 g/100 g water, to
of the crystal to be about 10 times that of the volume equiv- less than 3 h at 10 g/100 g water.
alent size. The length growth rate (1 prdmin) is comparable
to quoted growth rates for other proteins that range from 0.4
Purification Experiments
F d m i n for canavalin to 1 F d m i n for lysozyme and insu-
lin. In Figure 9 the values of the growth rate constant derived
The effects of ammonium sulfate concentration and pH from the crystallizations of ovalbumin in the presence of
upon the growth rate kinetics are shown in Figures 6 and 7.
Ammonium sulfate concentration (AS/W) had no discern-
ible effect upon the crystal growth kinetics at pH 4.90 (Fig.
6), however a slight effect can be seen at higher pH values
(Fig. 7). In Figure 7, the average growth rate constant k,,
= G/u2 has been plotted against pH. There is about a
10-fold increase in K,, over the pH range 4.6-5.4. As the
effect of ammonium sulfate concentration is small com-
pared to the effect of pH, the growth rate constant has only
been correlated against the pH by

log,,(k,,) = -0.92 + 1.21(pH - PI) (2)

Here pZ = 4.58 was taken as the limit in the correlation as

a solubility minimum exists for ovalbumin at this point.
Forsythe and Pusey' report that, for lysozyme, changes in
temperature and salt concentration that cause a decrease in
solubility also cause a decrease in the growth rate constant.
Ovalbumin fits this generalization.
Long crystallization times resulted from the use of the
low initial crystal content (0.2 g/100 g W) chosen to give a
large extent of crystal growth which could be measured Figure 7. Effect of pH upon the crystal growth rate constant

JUDGE, JOHNS, AND WHITE: PROTEIN PURIFICATION BY BULK CRYSTALLIZATION 321

 3 , , , , , , , , , , , , I ' l""I liquor. On overloaded SDS-PAGE using Coomassie blue

1 staining, neither conalbumin nor lysozyme were detected in

the washed ovalbumin crystals obtained (Fig. 10, lanes 7
and 8). The detection limit for both lysozyme and conalbu-
min using this stain was 0.05 &well. Consequently, the
crystals contained more than 99 wt % ovalbumin on a pro-
tein basis. Conalbumin and lysozyme are clearly present in
the final mother liquor (lanes 5 and 6) and trace amounts of
both were detected in the wash liquors, presumably re-
moved in the liquor adhering to crystals after filtration.
The results for silver-stained SDS-PAGE gels, which
have greater sensitivity than gels stained with Coomassie
blue, are given in Figure 11. Whereas the washed crystals
contain no conalbumin, trace amounts of lysozyme are vis-
ible. Nevertheless, the crystals contain more than 99.4 wt %
ovalbumin on a protein basis. These results, although
n ' 1 1 1 1 1 1 1 1 1 1 1 / 1 : 1 1 1 1 1 1 , 1 ,
sparse, represent the first published quantified demonstra-
0 n 10 15 20 25 30

T i ~ i i r( h )
tion of the powerful selectivity of the crystallization process
for obtaining highly pure crystals of one protein from a
Figure 8. Predicted drop in ovalbumin concentration with time for three mixture of several, in contradiction to frequent reports in
crystallizations with different initial seed crystal contents (g/100 g water) the literature that crystallization of a protein is only possible
at 27 g ammonium sulfateilO0 g W and pH 4.9 at 30C.
from a highly purified solutions.6 This highlights the dif-
ferent purity requirements of those seeking single, large,
both conalbumin and lysozyme are compared with those and highly homogeneous protein crystals for protein struc-
from crystallizations of pure ovalbumin. There was no ef- ture determination by X-ray crystallography and those seek-
fect of the presence of conalbumin and lysozyme on either ing a lesser, but still high, degree of purification for bulk
the rate of ovalbumin crystallization or the dependence of protein crystallization, such as enzyme manufacturers. The
the ovalbumin crystal growth rate upon relative supersatu- high selectivity obtained is not surprising, since the crystal
ration. As some solute impurities can have dramatic effects lattice presumably acts as a selective template in essentially
upon the crystal growth rate," this result is extremely ben- the same manner as the "molecular imprinting" phenom-
eficial, simplifying the design of a bulk crystallizer to re- ena being used in the development of more selective chro-
cover ovalbumin from a heterogeneous mix of proteins. matography matrices.
The degree of purification of ovalbumin achieved in a The crystals, nevertheless, contain high salt concentra-
single crystallization was considerable. The ovalbumin
comprised 86 wt % of total protein in the initial mother

.
5
.- 0 1
i
Y
+- Pure ovalbumin
m Ovalbumin a n d
other proteins

0 01
~I.I 1
I Figure 10. Purity of ovalbumin crystals obtained from a solution mixture
0.5 I 5
of ovalbumin, conalbumin, and lysozyme by overloaded SDS-PAGE anal-
Relative S u p r r s a t u r a t i o n (a) ysis with Coomassie blue R-250 staining. Lanes: (1) molecular weight
markers (sizes in kDa on left), (2) original ovalbumin prepared in this
Figure 9. Effect of conalbumin and lysozyme upon the crystal growth study, (3) conalbumin, (4)lysozyme, ( 5 , 6 ) final mother liquor, and (7,8)
rate at 27 g ammonium sulfate/100 g water and pH 4.90. washed ovalbumin crystals.

322 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 48, NO.4, NOVEMBER 20, 1995

Figure 11. Purity of ovalbumin crystals obtained from a solution mixture
of ovalbumin, conalbumin, and lysozyme by overloaded SDS-PAGE anal-
ysis with silver staining. Lanes: (1) molecular weight markers (sizes in kDa
on left), (2) conalbumin, (3) lysozyme, (4, 5) final mother liquor, and (6,
7) washed ovalbumin crystals.
tions in the internal solvent filled voids, and apparently
variants of the same protein (i.e., different phosphate
forms) are typically incorporated into the crystal lattice in
the proportion of their occurrence in the mother liquor,
although this has not been systematically studied.
In conclusion, the results demonstrate that bulk crystal-
lization in ammonium sulfate solutions can be successfully
applied for the recovery and purification of ovalbumin. The
process is scalable to any production requirement, and pro-
vided solution and kinetic data are obtained, it should be
applicable to any protein. Studies to extend our understand-
ing of protein crystallization are continuing in our labora-
tory.
We thank Susan Hamilton for guidance in SDS-PAGE silver
staining. Thanks also to Peter Abeydeera for his assistance with
HPLC analysis. Financial support was provided for one of us (R.
A. J.) by the Arthur Joseph Deakin Scholarship and by the De-
partment of Chemical Engineering, The University of Queens-
land.
References
1. Cohn, E. J . , Edsall, J . T. 1943. Proteins, amino acids and peptides as
ions and dipolar ions. Reinhold, New York.
JUDGE, JOHNS, AND WHITE: PROTEIN PURIFICATION BY BULK CRYSTALLIZATION
2. DeMattei, R. C . , Feigelson, R. S . 1989. Growth rate study of
canavalin single crystals. J. Crystal Growth 97: 333-336.
3. Dhal, P. K., Arnold, F. H. 1991. Template-mediated synthesis of
metal-complexing polymers for molecular recognition. J. Am. Chem.
SOC.113: 7417-7418.
4. Durbin, S . D., Feher, G . 1986. Crystal growth studies of lysozyme as
a model for protein crystallization. J. Crystal Growth 76: 583-592.
5 . Forsythe, E., Pusey, M. 1994. The effects of temperature and NaCl
concentration on tetragonal lysozyme face growth rates. J. Crystal
Growth 139: 89-94.
6. Giege, R., Dock, A. C., Kern, D., Lorber, B., Thieny, J. C., Mo-
ras, D. 1986. The role of purification in the crystallization of proteins
and nucleic acids. J. Crystal Growth 76: 554-561.
7. Judge, R. A. 1995. Investigating the bulk crystallization of proteins.
Ph.D. Dissertation, University of Queensland, Brisbane, Australia.
8. Kitabatake, N., Ishida, A., Doi, E. 1988. Physicochemical and func-
tional properties of hen ovalbumin dephosphorylated by acid phos-
phatase. Agric. Biol. Chem. 52: 967-973.
9. McPherson, A. 1990. Current approaches to macromolecular crystal-
lization. Eur. J. Biochem. 189: 1-23.
10. Mikol, V., Giege, R. 1992. The physical chemistry of protein crys-
tallization. pp. 219-238. In: A. Ducruix and R. Giege (eds.), Crys-
tallization of nucleic acids and proteins: A practical approach. Oxford
University Press, New York.
11. Miller, M . , Weinstein, J., Wlodawer, A. 1983. Preliminary X-ray
analysis of single crystals of ovalbumin and plakalbumin. J. Biol.
Chem. 258: 58665866.
12. Mullin, J. W . 1993. Crystallization, 3rd edition. Butterworth-
Heinemann, Oxford.
13. Myerson, A. S., Toyokura, K. (eds.). 1990. Crystallization as a sep-
arations process. ACS Symposium Series 438. American Chemical
Society, Washington, DC.
14. Randolph, A. D., Larson, M. A. 1988. Theory of particulate pro-
cesses: Analysis and techniques of continuous crystallization, 2nd
edition. Academic, San Diego.
15. Ries-Kautt, M . , Ducruix, A. 1992. Phase diagrams. pp. 195-281. In:
A. Ducruix and R. Giege (eds.), Crystallization of nucleic acids and
proteins: A practical approach. Oxford University Press, New York.
16. Schlichtkrull, J. 1957. Insulin crystals. V. The nucleation and growth
of insulin crystals. Acta Chem. Scand. 11: 439460.
17. Smith, M. B. 1964. Studies on ovalbumin I. Denaturation by heat,
and the heterogeneity of ovalbumin. Aust. J. Biol. Sci. 17: 261-270.
18. Smith, M. B., Back, 1. F. 1965. Studies on ovalbumin II. The for-
mation and properties of S-ovalbumin, a more stable form of ovalbu-
min. Aust. J. Biol. Sci. 18: 365-377.
19. Sorensen, S. P. L., Hoyrup, M. 1915-1917. Studies on proteins.
Compt. Rend. Lab. Carlsberg. 12: 12-67, 164-212, 213-261.
20. Stein, P. E., Leslie, A. G. W . , Finch, J. T., Carrell, R. W. 1991.
Crystal structure of uncleaved ovalbumin at 1.95 A resolution. J. Mol.
Biol. 221: 941-959.
21. Stoscheck, C. M. 1990. Quantitation of protein. General methods for
handling proteins and enzymes. Methods Enzymol. 182: 5CL68.
22. Visuri, K., Kaipainen, E., Kivimaki, J., Niemi, H., Leisola, M . ,
Palosaari, S . 1990. A new method for protein crystallization using
high pressure. Bioltechnology 8: 547-549.
23. Warner, R. C. 1954. Egg Proteins. pp. 435485. In: H. Neurath and
K. Bailey (eds.), The proteins: Chemistry, biological activity, and
methods, vol. 11, part A. Academic New York.
323