Vox Sang.

28: 90-100 (1975)

Alternatives to Freeze-Drying for the Removal

of Ethanol from Plasma Proteins
II. Gel Filtration of Albumin

A. J. DICKSON
and J. K. SMITH
Protein Fractionation Centre, Edinburgh

Abstract. Removal of ethanol from highly concentrated solutions of human albumin

(Cohn fraction V) by gel filtration on Sephadex G-25 is hindered by the contraction of
the gel in ethanolic solution, by incomplete retardation of ethanol compared with other
low MW solutes, and by restricted diffusion of ethanol from the albumin zone. Despite
these obstacles, the hourly capacity of such gel filtration columns, for approximately
100-fold reduction of ethanol concentration, may exceed 0.06 kg albumin per litre of
column volume. The gel can be used safely at 5°C for several years. The ethanol content
of the final product is higher than that achieved by vacuum distillation, and it may be
desirable to operate the two techniques sequentially.

Albumin prepared for clinical use by ethanol fractionation of human

plasma is subject to polymerisation at several stages [4], including the
processes used for the removal of ethanol. These processes may also influence
the amount and nature of the polymer formed during later pasteurisation.
Desalting of proteins, or ‘group separation’, on cross-linked dextran or
polyacrylamide gels has long been a familiar laboratory process, and equip-
ment is available for its extension to the industrial scale [I]. Gel filtration is
unlikely to initiate polymerisation or denaturation of albumin, if carried out
at moderate temperature and pH. The predominant disadvantage compared
with freeze-drying or vacuum distillation [9] is that albumin emerging from
the column must be more dilute than the feed solution.
FRIEDLI and KISTLER[5] have recently investigated the feasibility of
applying large-scale gel filtration on Sephadex G-25 to the production of
5% albumin solution. They concluded that the process was safe and eco-

Received: April 15, 1974; accepted: July 15, 1974.

 DICKSON/SMITH 91

nomically attractive, but suggested certain limitations, e.g., small charge

volume and operation at room temperature.
Our original interest in gel filtration was to replace freeze-drying for the
removal of ethanol from solutions of IgG and to develop general methods
of removing precipitants and other solutes of low molecular weight from
coagulation factors and immunoglobulins. Group separation of concen-
trated solutions of albumin and ethanol has proved to be a convenient
model system, although it also presents some peculiar problems. Our results
support FRIEDLI and KISTLER’S [5] general conclusions and extend their
observations on the potential value of gel filtration in this application.

Materials and Methods

Protein concentration was measured by the method of GORNALL et al. [a]. Ethanol
concentration was measured by an alcohol dehydrogenase method (Boehringer Corp.,
Mannheim; Blood Alcohol Test Kit). The deproteinisation stage was omitted.
Viscosity was measured in a temperaturecontrolled Haake falling ball viscometer at
4 and at 13°C.
Sephadex G-25 (coarse) and LH-20 were manufactured by Pharmacia, Uppsala.
Biogel P2. P6 and P10, all 100 mesh, were manufactured by Biorad Laboratories Ltd.,
Richmond, USA. Controlled-pore glass CPG-10, 75 Angstroms, was manufactured by
Corning Glass Co.
Columns 2.5 cm in diameter (K25/100) and 21.5 cm in diameter (K215/100), adjustable
between 60 and 90 cm in length by piston end-pieces, were manufactured by Pharmacia.
Columns were packed under at least 60 cm head or under downward flow at more than
300 ml/cma/h. Samples were applied and eluted in the upward direction to minimise
differential compression and particle size classification along the column. Flow was
maintained either by peristaltic pump or, in the case. of the large column, by a stainless
steei gear pump. The latter gave particularly smooth flow, but the inlet assembly of the
large column, which includes a manometer chamber, provided effective damping of the
small pulses generated by large peristaltic pumps. Flow-rate in these laboratory columns
was limited by their mechanical strength; a limit of 0.6 kg/cma was never exceeded in
these experiments. All operations were carried out at an ambient temperature of 5°C.
Fine particles were removed from the swollen gels by at least two decantations before their
first use. Columns were eluted with distilled water, and stored in 0.01% merthiolate while
not in use. Before each experiment, the K215/100 column was washed with 25 litres of
1 % NaOH solution and 100 litres of pyrogen-free water. After very long storage, the
upper piston was replaced by a large reservoir and the column backwashed with a large
volume of water to remove fines and accumulated debris.
The volume parameters describing the elution of solutes in gel chromatography have
been d e h e d by FLODIN [3]. Vt is the total volume enclosed between the pistons of the
column. Ve is the elution volume of a solute applied to the column. For broad peaks,
resulting from the application of large sample volumes, V, has been calculated as the
92 DICKSON/SMITH

elution volume at half-height of the solute front. Vo (void volume) is the elution volume
of a completely excluded solute, e.g., albumin. One expression of the distribution coef-
ficient of a solute is:

Ve solute - Vo
Ksv solute =
Vt - v o

Results

Preliminary experiments on the separation of human albumin (a fully

excluded solute) and sodium chloride (a fully retarded solute) on Sephadex
G-25 (coarse) confirmed that columns of diameters 2.5 and 21.5 cm were
equally efficient in separating these solutes when loaded with sample volumes
of the order of 0.3 Vt.
These experiments were extended by the addition of ethanol to the sample
to simulate an aqueous solution of Cohn fraction V; typically, the sample
contained 10% w/v protein, 10% w/v ethanol and 0.1 M NaCl included as a
reference solute. Sample viscosity was 3.9 cP. Two serious difficulties were
encountered immediately; contraction of the gel occurred in the lower part
of the column during sample application, and ethanol was not eluted in the
same position as the reference solute.

Gel Contraction
During application of an ethanolic sample, a gap formed between the
bottom of the gel bed and the lower piston. This gap filled first with the
sample, then with eluant which, being less dense, mixed with it and diluted
the tail of the sample.
When individual components of the sample were omitted or greatly
diluted in further experiments, it was confirmed that gel contraction was
caused largely by high ethanol concentration in the sample and to only
a minor extent by the osmotic effect of high concentrations of protein. The
consequent mixing of the tail of the sample with the incoming eluate was,
however, aggravated by the high density of 10% protein solution. Contrac-
tion was not caused by high concentrations of salt, or by high flow rates
during sample application or elution.
Attempts were made to minimise the effects of gel shrinkage by varying
the rate of application and elution, and by introducing a delay between sample
application and elution. Since the gel remained contracted while in contact
with ethanol in the sample, it was preferable to apply the sample at maximum
flow-rate and to elute without delay. A high rate of elution enabled the
 Gel Filtration of Albumin 93

Table I. Elution parameters for Sephadex G-25 and LH-20

Gel type K*V Cycle volume/Vt

Albumin Ethanol NaCl

G-25 coarse 0 0.68 0.86 1 .o

LH-20 0 0.75 0.64 1 .o

Elution parameters were calculated from the elution volume of each solute after
separation of samples containing 8% w/v albumin, 10% w/v ethanol and 0.1 M NaCI. The
G-25 column measured approximately 21.5 x 80 cm, and the LH-20 column 2.5 x 80 cm.

viscous sample to travel through the gap between gel and piston with the
minimum of mixing.
Operation of the column by downward flow would be expected to mini-
mise mixing of the dense sample with eluant following gel contraction. This
alternative was rejected on the ground that gradients, or even discontinuities,
of gel compression would develop progressively in the bed, and could be
relieved only by repacking or back-washing the column frequently. The
21.5 cm column was packed under varying rates of downward flow, in order
to vary gel compression, after which the sample was applied and eluted in
the upward direction. It was observed that a bed under high uniform com-
pression recovered more rapidly from contraction and that separation
efficiency was not impaired by compression, up to at least 0.6 kg/cma.

Incomplete Retardation of Ethanol

Table I includes the K,, values for ethanol and sodium chloride on
Sephadex G-25. The difference in K,, for the two solutes was not markedly
affected by the concentration of albumin or ethanol in the sample. It was
concluded that the non-ideal elution of ethanol (MW46) reflected an un-
avoidable interaction between the solute and the gel matrix. A similar inter-
action was found with gels based on polyacrylamide, and an opposite
‘retarding’ interaction of ethanol with lipophilic Sephadex LH-20.

Gel Filtration of Large Charges of Concentrated Albumin Solution on

Sephadex G-25
Two large-scale separations using approximately 27 litres Sephadex G-25
coarse in a column of diameter 21.5 cm clarify the compromises which may
be made between capacity and separation efficiency.
Q
TJ.5
'' Albumin concentration. gll00 mt Albumin concentration, gll00 m i
Ethand concentration, gll00 ml Ethanol concentration, gl100 m l
d
0 N P m m s 0 N P m a 0

0
w
A
N a, s N m
0 0 s om 0 0 0 o 0
P o 0
a, 0

Cumulative albumin recovery. *h Cumulatlve albumin recovery, %

Cumulative ethanollalbumin, rng l g Cumulative ethanol/albumin. mg/g
 Gel Filtration of Albumin 95

In the experiment illustrated in figure 1 the charge was 8 litres (0.3 Vt)
of 10% albumin containing 10% ethanol, and in the experiment illustrated
in figure 3, the charge was 4 litres (0.15 Vt) of a similar solution. In both
experiments, elution was with distilled water at 5 "C, and fractions collected
at 1- or 0.5-litre intervals were sampled for determination of conductivity
and protein and ethanol concentrations.
The total amount of albumin eluted between the front of the albumin
peak and successive points across the albumin peak was plotted (cumulative
percentage albumin recovery) as a percentage of the total albumin recovered
in all fractions; estimates of total recovery varied between 95 and 105%
of the albumin initially applied to the column. Similarly, the relative recovery
of ethanol and albumin (cumulative ethanol/albumin) was plotted at
successive points across the albumin peak.
It is evident that, using the lower charge volume, ethanol was better
separated from the albumin peak, although KBvfor ethanol was unaffected;
on the other hand, the albumin emerged at a higher dilution. Figure 1
typifies the upper limit of charge volume imposed by resolution efficiency,
while figure 2 typifies the lower limit of charge volume imposed by dilution
and economic considerations.

Alternative Molecular Sieve Media

G-25 was considered the best dextran-based (Sephadex) medium because
the swollen gel particles are sufficiently rigid to permit higher flow-rates
than, say, G-50; albumin would emerge less dilute than from G-50; less gel
would be required, and less gel-solute interaction would be expected than
with G-10.
The polyacrylamide-based (Biogel) media were examined as alternatives
to Sephadex. P6 resolved albumin and ethanol less well than the equivalent
bed volume of G-25, and was even more subject to contraction on contact
with ethanol. If the sample contained no ethanol, beds of Biogel P6 and P10
gave higher flow-rates than Sephadex G-25. However, beds of Biogel com-
pressed sufficiently to counteract the effect of gel contraction in ethanolic
samples gave much greater resistance to flow than comparable beds of
Sephadex G-25.
Controlled-pore glass might seem an ideal molecular sieving medium
for therapeutic products because of its inertness, permanence, resistance
to heat and agressive cleaning agents, and complete rigidity, permitting a
linear increase of flow-rate with pressure. However, it was found that, even
after treating CPG-10 porous glass with PEG or albumin to diminish inter-
96 DICKSON/SMITH

action of the medium with the solutes, the charge volume for effective
separation was much less than 0.01 Vt. It was concluded that very large
bed volumes of expensive controlled-pore glass, operated at high pressures,
would be required to exploit the real advantages of the medium.
Sephadex LH-20, a dextran-based, lipophilic gel with an exclusion limit
of 20,000 dalton, suffered almost no contraction in the presence of samples
containing 10% albumin and 10% ethanol. The difference in Kav for albumin
and ethanol was greater on LH-20 than on G-25 (table I) and the cycle
volume was the same. Although LH-20 is initially more expensive than G-25,
and has inferior mechanical properties, its potentially higher separation
efficiency merits further study of gel stability and optimal column configura-
tion.

Discussion

Sample Concentration
Redissolved Cohn fraction V contains similar w/v concentrations of
ethanol and albumin. To make the most economical use of the column, it is
necessary to apply a highly concentrated solution. A lower limit of sample
concentration is set by the need to recover almost all of the applied albumin
at a mean concentration greater than 5% in this application. The upper limit
of sample concentration might be set by viscosity, density or ethanol con-
centration. Sample viscosity, discussed in more detail below, is not con-
sidered to be the primary limitation in this particular application. Sample
density and ethanol concentration, which increase proportionately with
viscosity, appear to have more serious effects on separation. Dense solutions
of albumin tend to form a streaky ‘tail’ in the column because of excessive
mixing with less dense eluant following the sample; this tailing is aggravated
if contraction of the gel in high concentrations of ethanol creates even more
mixing at the bottom of the column.
The upper limit on sample concentration is therefore fixed primarily by
the design of the column end-pieces, which should minimise dead space, and
the mechanical strength of the column, which limits the gel compression
which can be used to counteract gel contraction. It is not easy to reduce the
dead space of the end-pieces while maintaining efficient distribution and
collection of liquid at high flow-rates. A column capable of operating under
high pressures would offer increased daily capacity and would eliminate
the problem of gel contraction. The column could be packed under hydro-
static pressure or while contracted in ethanolic solution.
 Gel Filtration of Albumin 97

Charge Volume and Selection of Fractions

The volume of each charge is h e d by the difference in the distribution
coefficients of albumin and ethanol, by the lowest mean concentration of
albumin, or by the highest mean concentration of ethanol that can be
accepted in the pooled eluate. Increased charge volume does not lead to a
linear increase of hourly capacity since slightly longer wash periods are
required, but a larger charge volume permits recovery of albumin at a higher
mean concentration, i.e. the albumin ‘plateau’ constitutes a higher pro-
portion of the eluted albumin.
Although this would not necessarily be true in other applications, it is
anticipated that industrial scale gel filtration of ethanolic albumin would be
carried out by sub-dividing a large batch into many column charges. This
system is the most favourable for the maximum recovery of albumin at the
highest concentration, since the dilute tail of the albumin peak could be
recycled into the concentrated feed. In this mode, the proportion of tailings
collected would not be expected to exceed 10% of the charge volume. A
higher proportion of tailings, up to 80% charge volume, might be acceptable
if pooled tailings were used to dissolve the next batch of fraction V paste,
rather than recycled into the current feed.
It has been shown that, in the present application, it is economical to
apply a large charge of ethanolic 10% w/v albumin. Under these conditions,
virtually all albumin can be recovered at a mean concentration of about
7%, and the ethanol content of the product becomes the more important
criterion for selection of fractions.
Figures 1 and 2 illustrate two rational criteria for selecting fractions,
namely the cumulative ‘percentage albumin recovery’ plot, or the cumulative
‘milligramme ethanol per gramme protein’ plot. Even in the experiment
where the column was deliberately overloaded (charge volume 0.3 Vt) to
demonstrate the limits of the system, it would have been possible to recover
75% of applied albumin at less than 10 mg ethanol/g albumin, or 88% of
applied albumin at less than 15 mg/g. 93% of applied albumin could have
been recovered by recycling tailings up to 10% of the charge volume, or
98% by collecting tailings up to 25% of charge volume. A further 30%
charge volume would have to have been collected to achieve 99.9% recovery,
and this would not normally be considered worthwhile.
Further improvements in albumin recovery at any fixed maximum ethanol
content would have to be achieved by reducing charge volume. For example,
with a charge of 0.15 Vt, 90% of applied albumin would be recovered at less
than 10 mg ethanol/g albumin, or 99% at less than 15 mgfg, without resort-
98 DICKSON/SMITH

ing to recycling. At this sample loading, however, the dilution factor would
be approximately 1.7, compared with approximately 1.4 at a charge of 0.3 Vt.
FRIEDLI and KISTLER [5] obtained a dilution factor of 1.7 with a charge of
0.18 Vt, at room temperature.

Effects of Temperature and Sample Viscosity on the Ethanol Concentration

of the Product
The same batch of Sephadex G-25 has been used for more than 3 years.
The gel has been left in the column, sometimes for several months, in water
or 0.01 % merthiolate solution. Subsequent washing with 1% sodium
hydroxide solution and water has always sufficed to produce a clear and
pyrogen-free effluent.
FRIEDLI and KISTLER[5] elected to carry out separations at room tem-
perature, thereby incurring a greater risk of bacterial contamination, in
order to reduce the effects of sample viscosity. The viscosity of 10% albumin
is only 3.9 CP at 5"C, compared with 3.0 CP at I4"C, and the compatibility
of our separations with those of FRIEDLI and KISTLER [5] confirms that such
a small difference in viscosity is not a major determinant of separation
efficiency. FLODIN [3] claimed that separation of salt from haemoglobin was
impaired only at a sample viscosity of 10 cP.
In addition to ethanol included with the tail of the albumin fraction by
overlapping with the ethanol peak (a function of sample volume and the
difference in distribution coefficients of the two solutes), a relatively low
'background' concentration of ethanol was eluted throughout the albumin
peak. This appeared to be related to sample concentration rather than
column loading, and was attributed to restricted diffusion of ethanol from
the albumin zone. This effect was even more dramatically illustrated when
attempts were made to separate ethanol from a very viscous solution of IgG.
About 30% of the initial ethanol remained in the protein peak. Very viscous
protein solutions may be desalted in a basket centrifuge [2] but resolution
is much poorer than that achieved by chromatographic separation.
FRIEDLI and KISTLER[5] set a limit of 15 mg ethanol/g albumin in the
final product, presumably by reference to albumin prepared by conventional
(condenser) freeze-drying. They showed that albumin containing 10.4 mg
ethanol/g albumin could be pasteurised without obvious damage to the
protein, and we have made similar observations.However, some fractionators
are obliged to use ethanol mixed with up to 4% methanol as denaturant,
and there are possible dangers of intoxication following the infusion of
arge volumes of albumin solution containing high levels of both alcohols.
 Gel Filtration of Albumin 99

Starting from Cohn fraction V, gel filtration effectively reduces ethanol

from about 1,OOO mg/g albumin to 10-15 mg/g albumin, and offers some
discretion in choosing a limiting ethanol concentration. Further reduction
of ethanol to approximately 2 mg/g albumin would require approximately
twice the gel filtration capacity, or the acceptance of a more dilute eluate.
Ethanol concentrations of about 2 mg/g can be achieved by alternative
means such as steam-ejection freeze-drying or vacuum distillation. The
latter method uses several sequential passages through similar distillation
cycles to reduce ethanol concentration progressively. A hybrid system,
using a single distillation stage followed by gel filtration, might prove to be
the most efficient. Gel filtration has economic advantages over a second
vacuum distillation stage of comparable capacity. A single distillation step
would reduce ethanol concentration below that which causes contraction
of the gel, and the final ethanol concentration in the column effluent should
be significantly reduced by feeding an albumin solution containing less than
1% ethanol. Concentrated salt-poor albumin might be made by reversing
the order of the vacuum distillation and gel filtration stages.

Plant Capacity
Our experiments using Sephadex G-25 (coarse) did not indicate any loss
of resolution with increase of flow-rate up to 300 ml/cm*/h, the limit set by
our apparatus. FRIEDLIand KISTLER[5] used similar rates and EK [l]
described certain applications in which the flow-rate was as high as 420 ml/
cm2/h. All such limits are affected by variations in temperature, viscosity,
gel compression, etc. The experiment illustrated in figure 2 showed that,
using a charge volume of approximately 0.15 Vt, it is possible to achieve
99% recovery of albumin at an ethanol concentration of less than 15 mg/g
albumin, without resorting to recycling. The cycle volume, i.e. the interval
between fresh charges, was equal to Vt, and it was possible to achieve
flow-rates greater than 4 Vt/h without loss of resolution. It follows that the
minimum capacity of the column is approximately 0.6 Vt/h. Since at least
10% w/v albumin can be used as the feed, the capacity of an automatically
operated 75-litre column is 4.5 kg albumin/h.
The capital cost of high-capacity gel filtration plant is much lower than
that of freeze-dryers of comparable capacity. Some typical costs have been
calculated by HORTON [7land by KISTLER and FRIEDLI [8] who have developed
their earlier study to a practical industrial-scale application. The total
operating cost depends largely on the local cost of pyrogen-free water, and
to a lesser extent on the complexity of the fraction-selecting equipment
100 DICKSON/SMITH

chosen to make the most efficient use of the column. Different strategies, e.g.
the inclusion of recycling, or the choice of limits on final ethanol concen-
tration, make very large differences in the theoretical daily capacity.
To achieve final concentrations of ethanol comparable to those found
after freeze-drying, with nearly complete recovery of albumin in a single pass,
a charge of 0.15-0.20 Vt should be applied, and continuous monitoring of
absorbance or refractive index should suffice to detect the end of the protein
peak. However, if maximum economy of operation is required, and if a
proportion of recycling is accepted, the charge should be increased to 0.25-
0.30 Vt, and fractions selected by precise metering of effluent volumes. If the
ethanol concentration of the effluent must be reduced below lOmg/galbumin,
it may be necessary to operate at higher temperature, at a reduced flow-rate,
or to accept less complete recovery of a more dilute effluent.

References

1 EK, L.:Gel filtration - a unit operation. Process Biochem. 3: 25-28 (1968).

2 EMNEWS. N. I. A.: A procedure for gel filtration of viscous solutions. J. Chromat. 32:
243-257 (1968).
3 FLQDIN,P. : Dextran gels and their applications in gel filtration (Pharmacia, Uppsala
1962).
4 FRIEDLI, H. and KISTLER,P. : Polymers in preparations of human serum albumin. VOX
Sang. 18: 542-546 (1970).
5 FRIEDLI, H. and KISTLER,P. : Removal of ethanol from albumin by gel filtration in the
manufacturing of human serum albumin solutions for clinical use. Chimia 26: 25-27
(1972).
6 GORNALL,A. G.; BARDAWILL, C. J., and DAVID,M. M.: Determination of serum
proteins by means of the biuret reaction. J. biol. Chem. 177: 751-766 (1949).
7 HORTON, T.: Large-scale gel filtration for purification of natural products. Int. Labora-
tory, pp. 43-49 (July/August 1972).
8 KISTLER,P. and FRIEDLI, H.: Working Party EFRAC, Bruges 1974.
9 S m ,J. K.; W A ~J. ,G.; WATSON, C. M., and MASTENBROEK, G. G. A.: Alternatives
to freezedrying for the removal of ethanol from plasma proteins. I. Vacuum distillation
of human albumin. Vox Sang. 22: 120-130 (1972).

Request reprints from: Dr. J. K. S m , Scottish National Blood Transfusion Service,

Protein Fractionation Centre, Ellen's Glen, Edinburgh EH17 7QT (UK)