CHEMICAL, CLINICAL, AND IMMUNOLOGICAL STUDIES ON THE

PRODUCTS OF HUMAN PLASMA FRACTIONATION.

VI. THE OSMOTIC PRESSURE OF PLASMA AND
OF SERUM ALBUMIN " 2
By G. SCATCHARD, A. C. BATCHELDER, AND A. BROWN
(From the Research Laboratories of Physical Chemistry, Massachusetts Institute of Technology,
Cambridge, and the Departmen of Physical Chemistry, Harvard Medical School, Boston)

(Received for publication February 17, 1944)

In order to determine the relative dosage of

plasma and of serum albumin in the treatment of
shock and to determine their efficiencies in increasing blood volume and other characteristics,
we have measured the osmotic pressure of
human plasma and serum, of regenerated solutions from dried plasma up to four-fold physiological concentrations, of human serum albumin
solutions up to 300 grams per liter in 0.15 molar
sodium chloride, and of bovine serum albumin
over large ranges of pH, albumin concentration,
and salt concentration.3 We have not been able
to detect a difference in the osmotic pressures of
human and bovine albumin, so we have used the
measurements on bovine albumin to extend the
range of those on human albumin.
The first function of plasma or serum albumin
in shock is to increase the blood volume by
holding water in the blood stream. Parts of
the water may be injected with the protein, may
be shifted from the extravascular fluids, or may
be drawn from the intestine. This water must
be held in spite of the excess pressure in the
capillary bed. Most of this retention of water
is not caused by any attraction between the
protein and the water, but arises from the fact
1 This work has been carried out under contract, recommended by the Committee on Medical Research, between
the Office of Scientific Research and Development and
Harvard University.
2This paper is Number 16 in the series "Studies on
Plasma Proteins" from the Harvard Medical School,
Boston, Massachusetts, on products developed by the
Department of Physical Chemistry from blood collected
by the American Red Cross.
' All measurements were made with collodion membranes, which are impermeable to plasma proteins. For
technical reasons, the measurements were made at 250 C.,
where the pressure is about 4 per cent less than at 37 C.
All concentrations were calculated using the nitrogen factor 6.25 for both albumin and plasma.

that the chemical potential of water is decreased

by the presence of dissolved molecules which are
not water.4
A convenient way of measuring the change in
the potential of water produced by the addition
of a solute is to measure the decrease in pressure
necessary to maintain equilibrium through a
membrane permeable to the water but not to the
solute, and this pressure is called the osmotic
pressure of the solution. If the membrane is
permeable to any of the solutes, the pressure is
sometimes called the colloid osmotic pressure or
"The difference in the potential of any substance at
two different places is the least work necessary to bring
unit quantity of that substance from the first place to the
second if the temperature and total volume of the system
are constant. The substance will shift spontaneously from
the place of higher potential to that of lower potential
if a path is available, and at equilibrium its -potential
must be the same throughout the system. The rapidity
with which equilibrium is attained depends upon the nature of the available paths as well as upon the difference
in potential, but the position of equilibrium depends only
upon the equality of potential.
There may be a difference in gravitational potential due
to different heights in a gravitational field, and this potential is proportional to the difference in height and to the
weight of unit quantity. There may be a difference in
chemical potential due to different pressures, and this
potential is proportional to the pressure difference and to
the volume of unit quantity. There may also be a difference in chemical potential due to different chemical compositions. We cannot generalize on the value of this
potential, but in very dilute solutions, the difference in
chemical potential of each substance is proportional to the
difference in the logarithm of its mole fraction. Then the
difference in the potential of each solute is proportional to
the difference in the logarithm of its concentration, and
the difference in potential of the solvent is proportional
to the difference in the sum of the concentrations of the
solutes, expressed as moles of solute per unit quantity of
solvent. These are the three potentials of substances
which are physiologically important.

458

OSMOTIC PRESSURE OF PLASMA AND OF SERUM ALBUMIN

459

the oncotic pressure. Since this is the only type

of membrane across which the pressure is important, we will use the simpler term.

Van't Hoff pressure

About two-thirds of the osmotic pressure at
physiological concentrations is explained by the
simple theory of Van't Hoff, which says that
the osmotic pressure is equal to the pressure
difference which would be developed if the solvent were removed and the solutes were gases. C
If the non-diffusible solute is a non-electrolyte,
the osmotic pressure should be proportional to
its concentration, and the pressure-concentra- I)0
tion ratio should be the same for all substances
if the concentration is expressed as moles per 0
unit volume. It is obvious that osmotic pres- C.)(1)
sure per unit mass, however, decreases as the 0
loo
200
size of the molecule increases. If the concentraOsmotic
Pressure,
P(mm.
Hg)
tion is expressed as mass per unit volume, the
pressure-concentration ratio is inversely pro- co
portional to the molecular weight, and one
FIG. 2
measurement of the osmotic pressure and of the
corresponding mass concentration may be used
to determine the molecular weight. The Van't plotted against the concentration, and Van't
Hoff case is illustrated by curve E in Figures 1 Hoff's theory leads to a straight line whose slope
and 2. In Figure 1, the osmotic pressure is is inversely proportional to the molecular weight.
In Figure 2, the pressure-concentration ratio is
plotted against the pressure, and this curve E
is a horizontal line whose ordinate is the slope of
E in Figure 1. The curves A in these figures are
the measured values for serum albumin at pH
7.4, and the curves B are for albumin at pH 5.4,
both in 0.15 M sodium chloride. At pH 5.4,
the albumin is iso-ionic; that is, its average net
charge is zero.

Mcss Concentration, C,(gn/AOOc)

FIG. 1

Donnan pressure
The plasma proteins are not neutral non-electrolytes, and their ionic charges have an important effect on the osmotic pressure. The
extension of Van't Hoff's theory to this case
was carried out by Donnan. We will limit
ourselves to the special case of the sodium salt
of a non-diffusible protein anion and sodium
chloride, which is a rough approximation of
plasma. The solution must be electrically neutral. So, if there is no sodium chloride present,
the pressure-concentration ratio corresponds to

460

G. SCATCHARD, A. C. BATCHELDER, AND A. BROWN

the protein ion plus all the sodium ions associated

with it.
If there is sodium chloride present, its potential must also be the same on the two sides.
Since there is more sodium, there will be less
chlorine on the side with the sodium proteinate.
So the pressure-concentration ratio is smaller.
It corresponds to the protein alone when the
ratio of protein to chloride is very small, and
depends only upon the protein-chloride ratio.
The molecular weight cannot be determined
from a single osmotic pressure measurement.
However, if a series of measurements is made
with varying protein concentration but constant
salt concentration, the pressure-concentration
ratio can be extrapolated to zero protein concentration, and the molecular weight can be
determined from this extrapolated value. At
pH 5.4, the Donnan curve is the same as the
Van't Hoff curve E. The Donnan curve for
pH 7.4, where the measured curve is A, lies
close to the curve B. It is obvious that the
Donnan effect increases rapidly with the net
charge of the protein ion. It is, in fact, proportional to the square of the net charge.'
5 The statement in mathematical equations is more precise and in some ways simpler. We will let C, C+ and C_
be the concentrations in moles per unit volume of protein,
sodium ion, and chloride ion within the membrane, and
C'+ and C'_ the concentrations of sodium and chloride
ions outside: c will be the concentration of the protein in
grams per unit volume, M its molecular weight, which is
c/C, z its valence (a negative number for an anion), T will
be the absolute temperature, and R a universal constant.
Van't Hoff's law is

P = R[C + C+ + C_ - C'+ - C'_]

Pressure of real solutions

The Van't Hoff theory assumes that each individual water molecule in a protein solution behaves like a water molecule in pure water and
that the decrease in potential of the water which
Combination with Van't Hoff's law yields

2C_-zC-2CC

P _ RT
= M[+

C_]

(2
M 2[ - +

2-(l-

lC

It is clear from the last equation that the expression in

square brackets depends only on zC/C_ for a given value
of z, but it is not clear how it approaches unity as zC/C_
approaches zero or even that it does so. To answer these
questions, we expand the radical in a Taylor's series to give
P

zC

91 ~C =1

zC

2C

zC2C .

M [
4CRT RT(z/M)'c RT M(z/M)'P
M
M
4C_
4C_
For a neutral molecule, z is equal to zero, and P/c = RT/M
at all concentrations. For an ion z2C/4C_ is zero only
when C is zero, and in dilute solutions P/c increases as a
linear function of c when C_ is constant.
For a real solution we may also write
P RT
- = M~ + bc + dc2 + * c

or

T + BP + DP' + ...
M
in which b and d or B and D are constants. The latter
form is more useful for our purposes because
1 RT
c

RT

+ C++ c--

C/+ -C'_]

The law of electroneutrality requires that

zC-C_ + C+ = 0
and
C'+ -C'_ = O
The law of equilibrium, or equality of the potential of
sodium chloride, requires that
C'+C'_ = C+C_
Combination of the last three equations yields
C'_= C_

-C.

MP-

In many cases, D is so small that the term DP may be

neglected. Then tIX volume of solution per gram of
protein is equal to the ideal term which is iniversely proportional to the pressure plus a term which is independent
of the pressure.
B will be composed in general of the Donnan term which
is M(z/M)2P/4C_ plus a term arising from the difference
between the force between two protein molecules and that
between one of the molecules and the water displaced by
the other.

OSMOTIC PRESSURE OF PLASMA AND OF SERUM ALBUMIN

accounts for the osmotic pressure is solely a

statistical effect, due to the fact that a certain
fraction of the molecules are not water. This
is equivalent to assuming that each individual
protein molecule in a concentrated solution behaves like a protein molecule in an extremely
dilute solution, or that the activity of the protein
is proportional to its concentration, and this
latter assumption is easier to consider. The
Donnan extension considers the macroscopic effect of the charge on the protein ion, but still
assumes that the activity of each species is proportional to its concentration.
Real protein solutions are not this simple.6
For iso-ionic albumin, the osmotic pressure is
much larger than that calculated by the Van't
Hoff theory. At other values of pH, however,
the increase over the iso-ionic value is much less
than predicted by Donnan, so that at pH 7.4, the
measured value does not differ greatly from that
calculated by his theory, and at still higher
values of the pH, the measured pressures are less
than those calculated. Real solutions do resemble the Donnan solutions in that the molecular weight can be determined by extrapolating
the pressure-concentration ratio to zero concentration or pressure.
Figure 1 shows the curves for osmotic pressure
versus mass concentration for serum albumin at
pH 7.4 (A) and pH 5.4 (B), for human plasma at
pH 7.4 (C), and for an ideal Van't Hoff solute
whose molecular weight is 69,000, which is the
same as that of albumin (E). The curves all

461

start at zero concentration and pressure, and

curves A, B, and E, all have the same initial

slope. The value for the molecular weight of

serum albumin calculated from these osmotic

pressure measurements, 69,000, is in good agreement with that obtained from sedimentation
and diffusion measurements reported in the first
paper of this series (1). The plasma curve (C)
has a smaller initial slope which corresponds to a

larger molecular weight, 93,000. All the experimental curves become steeper at higher concentrations. The difference between A and C is
proportional to the concentration.
In Figure 2 is shown the pressure-concentration ratio, P/c, versus the pressure.7 The ideal
Van't Hoff curve, E, is here a horizontal line
whose ordinate is the slope of E in Figure 1.
A and B intercept E at zero pressure, corresponding to the same initial slopes in Figure 1, but
they deviate sharply at higher pressures. The
plasma curve, C, has a lower intercept, corresponding to the larger average molecular weight
of plasma proteins, and the difference between
A and C is here independent of the pressure.
If 60 grams per cent of the plasma protein is
albumin (1), curve C as drawn would give an
average molecular weight of 194,000 for the
remaining 40 per cent. However, our measurements do not preclude drawing the intercept at
2.1 rather than 2.0, which would correspond to an
average molecular weight of 88,000, and a
molecular weight of 150,000 for the 40 grams per
cent which are not albumin. The probable value
for the average molecular weight of plasma
6 The ionic charges on proteins cause deviations from globulins is 170,000.

random distribution on the microscopic scale as well as

those considered by Donnan, and differences in potential
proportional to the square of the net ionic charge. Even
in iso-ionic albumin, where the average of the charge is
zero, the average of the square of the charge is 6, and
at pH 7.4 it is 300. In either solution, the total charge is
almost 200, and it is not evenly distributed through the
molecule. The resultant electrical field leads to attraction
between a protein molecule and a simple ion, or between
2 protein molecules, which is large compared to the effect
of the net charges. Moreover, the rotein molecule which
bears these charges behaves more like an equal volume of
a normal organic liquid than an equal volume of water.
This molecular framework tends to repel the electrical
fields but to attract another framework. The resultant of
all these forces leads to effects upon the potential of the
water and the resultant osmotic pressure which vary greatly
with the ionic charge.

Osmotic effiiency
We are particularly interested in the volume
of solution per gram of protein, 1/c, at a given
7 The.experimental points are shown in Figure 2, whose
scale is larger than that of Figure 1. The measurements
on plasma are indicated by crosses, those on human albumin by open circles, and those on bovine albumin by
filled circles. The measurements at pH 5.4 have a perpendicular line through the circles.
The measurements on albumin at pH 5.4 are carried to
dilute enough solutions so that there is little uncertainty
about the extrapolation to zero concentration and the
resultant molecular weight. The measurements on plasma
have not been carried to such dilution, so the curve has
been drawn parallel to that for albumin.

462

G. SCATCHARD, A. C. BATCHELDER, AND A. BROWN

8 30

25

_-

40

X--U
10'

20

.I
I

0 * cP *. .7
S

'

Ia

0*

58N.)
-

--

1'* *

15

/0

IC

6. e
3i
E.

.I0%
I.

(I1)

-6 tO

10

V)

aIt
(A
u)

:9
20

_~~~~~~
_-

I
0

so
40
Osmotic Pressure, P. (m Hg)

120

5.0

60

7.0

8.0

pH

FIG. 4

FIG. 3

osmotic pressure. This is shown in Figure 3 for

the same measurements.8 As in the previous
figures, curves A and B refer to albumin at pH
7.4 and 5.4, respectively, and curve C to plasma
at pH 7.4. The curve for an ideal Van't Hoff
solute corresponding to E of Figures and 2 is
omitted. It would lie about 5 cc. below curve B,
or about 7 cc. below A at all values of the pressure, as these three curves are nearly parallel.
The difference between A and C is almost inversely proportional to the pressure.
Figure 4 shows the volume of solution per
gram of protein at 25 mm. pressure as a function of the pH. The curve E represents as
before the ideal Van't Hoff solute, with a molecular weight of 69,000, and the curve D represents
an ideal Donnan solute with the same molecular
weight and the net charge of albumin.9 Although the difference between these curves and
the experimental values is not very great at the
physiological pH, it is evident that this is largely
a coincidence, for the experimental and ideal
curves cross at large angles.
Figure 3 is constructed from the same data as Figures 1
and 2 except that the measurements at very high and very
low pressures are both omitted to permit a larger scale.
The individual points are again omitted.
The titration curve for human albumin lies somewhat to
the left of the curve for bovine albumin, used in estimating
net charge for these calculations.

The experimental values of the volume per

of protein, 1/c, between pH 6 and 8 may
be expressed by the straight line
1/c = 11.1 + 0.9 pH
Since the difference between 1/c and the Van't
Hoff value RT/MP is nearly independent of the
pressure, a good approximation for 1/c at any
small pressure is
1/c = 268/P + 0.4 + 0.9 pH
This corresponds to
gram

P/c=268+(0.4+0.9 pH)P=

268

1+(0.4+0.9 pH)c

COMMENT

The chief uncertainty in determining the

efficiency of albumin in increasing blood volume
lies in the uncertainty of the osmotic pressure of
normal plasma, which goes back to the uncertainty as to the normal concentration of plasma
proteins. The composition of the proteins appears to be very constant in physiological plasma
(1),1o but the total concentrations quoted by
"IWies and Peters (2) have studied the effect on the
osmotic pressure of varying composition of the protein in
pathological plasma. Their results cannot be compared
with ours directly because they determine proteins by
Howe's precipitation method and our compositions are determined by electrophoresis. If we assume with them that

OSMOTIC PRESSURE OF PLASMA AND OF SERUM ALBUMIN

various observers show variations. Gutman

and his coworkers (3) quote protein concentrations in serum for normal adults ranging from
6.5 to 7.9 and averaging 7.2 grams per cent.
Perera and Berliner (4) quote concentrations
from 6.2 to 7.3, averaging 6.8 for normal ambulatory adults, and from 5.4 to 7.0, averaging 6.0
for the same individuals recumbent. The results for plasma should be 0.4 per cent higher
than for serum. So we must consider the range
from 6 to 8 grams per cent. Since the recipient
of a blood substitute will probably be resting,
the lower part of the range is probably more important than the higher.
If c is 6 grams per cent, 1/c is 16.7 cc. per
gram. From curve C, Figure 3, we find the
corresponding pressure to be 20 mm., and from
curve A we find that at 20 mm. pressure, each
gram of albumin retains 20 cc. of solution. If c
is 8 grams per cent, l/c is 12.5 cc. per gram.
The corresponding pressure is 32 mm., and at
that pressure, each gram of albumin retains 15
cc. of solution. If the pressure is 25 mm., each
gram of albumin retains 18 cc. of solution, and
each gram of average plasma protein retains
15 cc., or the concentration of plasma protein
is 6.67 grams per cent. This is the value which
we have taken as the norm, with the realization
that it may vary more than 10 per cent in
normal individuals. This corresponds closely
to the average of 17.4 cc. per gram of albumin,
given for the increase in blood volume per gram
of added albumin by Heyl, Gibson, and Janeway (5, 6). This average corresponds to 6.9
per cent protein. Their extreme values of 13.2
and 24 cc. per gram correspond to 5.1 and 8.8
per cent protein. Probably part of this variation corresponds to the inevitable errors of
measurements of total blood volumes.
If the increase of blood volume on the addithe pressure is a linear function of the ratio of globulin to
total protein, g, at all small pressures and pH's between 6
and 8, we may determine the effect of globulin from our
measurements on normal plasma, which gives

P/c

268(1 - 0.42 g)

1+ (0.4+0.9pH)c
For normal plasma this equation gives a limiting value at
zero concentration which is too large, but it should give a
fair representation of the results in undiluted plasma or
serum.

463

tion of protein is to equal the volume of solution

retained by the protein, the osmotic pressure of
the plasma must be the same after the addition
of the protein as before. It is probable that
this condition is nearly fulfilled in hemorrhagic
shock and also in traumatic shock. Any error
in this assumption will be in the direction of
increasing the pressure during the infusion and
thus giving a smaller increase in blood volume.
It is possible, however, that there are cases in
which the assumption is so greatly in error that
it is more accurate to assume that the blood
volume is constant, and to determine the increase in osmotic pressure per gram of protein
at constant plasma volume. Thus, an increase
in osmotic pressure per gram of protein at constant volume becomes of interest. This will,
of course, depend upon the total plasma volume,
of which we have no measure here. Our measurements do show the change in pressure for
unit change in concentration, and most clearly
in Figure 1. This change, dP/dc, for albumin
is 2.7 at zero concentration, 3.9 when c is 2, 5.1
when c is 4, 6.3 when c is 6, and 8.4 when c is 8
grams per cent. The values of dP/dc for average plasma protein at the same concentrations
are 2.0, 2.8, 3.8, 4.0, and 6.4. For albumin or
for plasma, the change in pressure with changing
concentration increases rapidly as the concentration increases.
The comparison of the relative efficiencies of
serum albumin and of plasma is more certain.
Within the accuracy with which we read our
curves, the relative efficiencies are independent
of the pressure. There is a slight increase in
the relative efficiency of albumin as the pressure
decreases, but we may take, for all physiological
pressures, our result that each gram of albumin
retains as much fluid as 1.2 grams of average
plasma protein. The pooling of plasma from
several donors and the relative constancy of
conditions of collecting blood reduce the fluctuations in the concentration of plasma. At present, the Red Cross pooled citrated plasma contains 6 grams per cent of protein (1). So each
gram of albumin corresponds to 20 cc. of citrated
plasma and 25 grams of albumin corresponds to
500 cc. of citrated plasma. This relationship is
the basis for the present containers used by our
armed forces,-500 cc. of citrated plasma for the

464

G. SCATCHARD, A. C. BATCHELDER, AND A. BROWN

BIBLIOGRAPHY
dried plasma container or 100 cc. of 25 per cent
albumin.
1. Cohn, E. J., OncIey, J. L., Strong, L. E., Hughes, W.
SUMMARY

The osmotic pressures of plasma and of serum

albumin at 250 have been measured over ranges
of concentration and pH, much wider than the
physiological ranges. The extrapolation of the
osmotic pressure-concentration ratios to zero
concentration yields a molecular weight of
69,000 for albumin. A similar extrapolation for
plasma yields an average molecular weight of
about 90,000. This corresponds to an average
weight of about 170,000 for the 40 per cent of
the protein which is not albumin. The osmotic
pressure-concentration ratios increase rapidly
with increasing concentration. At pH 7.4, this
increase corresponds to the Donnan effect of the
ionic charges, but measurements over a pH range
show this to be a coincidence.
The volume of fluid held in the blood stream
by each gram of albumin should be about 18 cc.
but should vary with the protein concentration
of the plasma. Each gram of albumin is equivalent to 1.2 grams of plasma protein or 20 cc. of
the current Red Cross citrated, pooled plasma.

2.
3.

4.
5.

6.

L., Jr., and Armstrong, S. H., Jr., Chemical, clinical,

and immunological studies on the products of human
plasma fractionation. I. The characterization of
the protein fractions of human plasma. J. Clin.
Invest., 1944, 23, 417.
Wies, C. H., and Peters, J. P., Osmotic pressure of proteins in whole serum. J. Clin. Invest., 1937, 16, 93.
Gutman, A. B., Moore, D. H., Gutman, E. B., McClellan, V., and Kabat, E. A., Fractionation of
serum proteins in hyperproteinemia, with special
reference to multiple myeloma. J. Clin. Invest.,
1941, 20, 765.
Perera, G. A., and Berliner, R. W., The relation of
postural hemodilution to paroxysmal dyspnea.
J. Clin. Invest., 1943, 22, 25.
Heyl, J. T., Gibson, J. G., 2nd, and Janeway, C. A.,
Studies on the plasma proteins. V. The effect of
concentrated solutions of human and bovine serum
albumin on blood volume after acute blood loss in
man. J. Clin. Invest., 1943, 22, 763.
Janeway, C. A., Gibson, S. T., Woodruff, L. M., Heyl,
J. T., Bailey, 0. T., and Newhouser, L. R., Chemical, clinical, and immunological studies on the
products of human plasma fractionation. VII.
Concentrated human serum albumin. J. Clin.
Invest., 1944, 23, 465.