Anal. Chem.

1998, 70, 1581-1584

Measurement of Protein Charge and Ion Binding

Using Capillary Electrophoresis
Manoj K. Menon and Andrew L. Zydney*

Department of Chemical Engineering, University of Delaware, Newark, Delaware 19716

A new technique is described for the rapid and accurate

measurement of electrophoretic mobilities of proteins in
different solution environments using capillary electrophoresis. Data were obtained at different pH using
surface-modified capillaries to reduce nonspecific protein
adsorption and using hydrodynamic mobilization to improve reproducibility and overall accuracy. The net
protein charge and extent of anion binding were evaluated
from the mobility data obtained in different pH and ionic
environments for bovine serum albumin. The results
were in good agreement with titration data obtained using
ion-selective electrodes and mobility data obtained using
free solution electrophoresis. The method requires extremely small amounts of protein (picogram quantities
and nanoliter volumes) and is easily automated, making
it very suitable for protein characterization and for initial
screening of possible separation techniques.
Electrostatic interactions can significantly affect the stability,
activity, conformation, and purification/isolation of many proteins.
The charge characteristics of a given protein are determined by
the extent of ionization of the various acidic and basic amino acid
residues in combination with the binding of specific anions,
cations, and charged ligands. The magnitude of the protein
charge is thus a complex function of the pH, ionic strength, and
detailed composition of the electrolyte solution.
Previous studies have generally evaluated protein charge and
ion binding by direct titration, equilibrium dialysis, free solution
electrophoresis, or NMR. Titration experiments evaluate the
extent of H+ and ion binding from the change in free concentration
of a given ion upon addition of a known amount of that species.
Glass electrodes are typically used for pH measurement. Ionselective electrodes are required to study the binding of specific
anions or cations. Equilibrium dialysis experiments are based
on the Donnan re-distribution of ions across a semipermeable
membrane caused by the addition of a charged protein to the
solution on one side of the membrane. Both the titration and
dialysis methods require relatively large amounts of protein and
very accurate measurement of ion concentrations since the extent
of ion binding is determined from (small) differences in concentration (either before and after addition of a known volume of
solution or between solutions on the two sides of the dialysis
membrane). Nuclear magnetic resonance techniques can provide
detailed information on the properties of specific ion binding sites;
however, these techniques are awkward and expensive to use to
evaluate the overall protein charge.
S0003-2700(97)00902-5 CCC: $15.00
Published on Web 03/12/1998

1998 American Chemical Society

Electrophoretic techniques provide a direct measure of the

protein mobility, with the protein charge evaluated using theoretical models that relate the mobility to the surface potential (typically
by assuming a relatively simple protein shape and uniform charge
distribution). The idea of using electrophoresis to study ion
binding is well-established. Many investigators have used the
Tiselius moving boundary technique to measure protein electrophoretic mobility in different buffers.1-3 More recently, Douglas
et al.4 developed a free-flow electrophoresis cell to measure protein
mobilities. However, both of these techniques require relatively
large (typically microgram to milligram) quantities of protein and
use complex analytical systems to evaluate the mobilities. In
addition, the boundary patterns in the Tiselius technique can be
difficult to quantify. Whitesides and co-workers5,6 used a technique that they termed affinity capillary electrophoresis to
evaluate binding constants of specific charged ligands from
changes in protein mobility. However, this technique was focused
on binding of a high-affinity ligand to a single specific binding
site on the protein. The technique described in this paper can
be used to evaluate the extent of ion binding from the background
electrolyte over a wide range of solution conditions, including
conditions where the protein may be positively or negatively
charged.
The importance of intermolecular electrostatic interactions in
protein solutions provided the motivation for the development of
a technique for the rapid and accurate measurement of electrophoretic mobilities, protein charge, and ion binding using available
capillary electrophoresis equipment. The usefulness of this
approach is demonstrated for the protein bovine serum albumin
(BSA), with data obtained in different buffer solutions used to
evaluate the protein charge as a function of pH as well as the
extent of anion (chloride and perchlorate) binding.
EXPERIMENTAL METHODS
Bovine serum albumin fraction V Powder (A7906), fatty acidfree BSA (A7030), BisTris, Tris, and HCl were all obtained from
Sigma Chemical (St. Louis, MO). Sodium acetate and sodium
phosphate were obtained from Fisher Scientific (Pittsburgh, PA).
(1) Longsworth, L. G.; Jacobsen, C. F. J. Phys. Colloid Chem. 1949, 53, 126135.
(2) Schlessinger, B. S. J. Phys. Chem. 1958, 62, 916-920.
(3) Norde, W.; Lyklema J. J. Colloid Interface Sci. 1978, 66, 277-284.
(4) Douglas, N. G.; Humffray, A. A.; Pratt, H. R. C.; Stevens, G. W. Chem. Eng.
Sci. 1995, 50, 743-754.
(5) Gomez, F. A.; Avila L. Z.; Chu, Y.-H.; Whitesides, G. M. Anal. Chem. 1994,
66, 1785-1791.
(6) Chu, Y.-H.; Avila L. Z.; Gao, J.; Whitesides, G. M. Acc. Chem. Res. 1995,
28, 461-468.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998 1581

Sodium perchlorate and perchloric acid were obtained from

Aldrich (Milwaukee, WI), and sodium chloride was obtained from
EM Sciences (Gibbstown, NJ). All chemicals were analytical
reagent grade. Mesityl oxide, obtained from Sigma Chemical, was
used as a neutral marker.
Salt solutions were prepared using deionized water obtained
from a Nanopure water purification system (Barnstead, Dubuque,
IA) with resistivity greater than 18 M cm. To minimize pH
variations during electrophoresis, all solutions were buffered using
1 mM sodium phosphate (pH 3-3.5), 1 mM sodium acetate (pH
3.5-5.6), 1 mM BisTris (pH 5.4-8), 1 mM Tris (pH 9), or 0.3
mM sodium tetraborate (pH 8-11). Solution pH was adjusted
using the corresponding acid and base for a given salt (i.e., HCl
and NaOH were used for NaCl solutions). pH was measured
using an Acumet 915 pH meter (Fisher Scientific). Solution
conductivity was determined using a Horiba ES-12 conductivity
meter (Kyoto, Japan). Protein samples were prepared by dissolving the protein (BSA) in the buffer solution and adjusting the pH
to the same value as the buffer. A protein concentration of 2 g/L
was used in all experiments reported here. The neutral marker
was prepared as a 1 mM mesityl oxide solution in the desired
salt and buffer.
Capillary electrophoresis was performed using an Isco model
3850 capillary electropherograph equipped with a dual-polarity
variable high-voltage dc supply (0-30 kV) and a variablewavelength UV-visible absorbance detector (Isco, Lincoln, NE)
with detection at 214 nm. Fused-silica capillaries (i.d. ) 75 m,
o.d. ) 363 m, total length 50 cm, distance to detector 30 cm)
with different surface modifications were used to minimize protein
adsorption to the capillary surface. A CElect Amine capillary with
a positively charged polymeric coating (Supelco Inc., Bellefonte,
PA) was used at low pH (pH <4.5) where BSA has a net positive
charge. A polyacrylamide-coated7 capillary with low (negative)
surface charge and high hydrophilicity was used for runs near
and above the isoelectric point of BSA (pH 4.5-9.0). The
polyacrylamide capillary was not used above pH 9.0 to avoid
destabilization of the polymer coating. An unmodified (negatively
charged) fused-silica capillary (Isco) was used at high pH (pH
6.0-11.0) where BSA has a strong net negative charge.
The hydrodynamic (or pressure) mobilization technique developed by Williams and Vigh8 was used to minimize the time
required for the mobility measurements and improve the overall
accuracy of the technique. The capillary and solution reservoirs
were filled with the buffered electrolyte solution of interest. The
injection end of the capillary was then immersed in a vial
containing the protein sample. A narrow slug of sample (8 nL)
was drawn into the capillary by applying a small vacuum (0.5 psi)
at the opposite end of the capillary for 1 s. The capillary was
then returned to the reservoir containing the background electrolyte solution, the vacuum reapplied, and the data acquisition
system initiated. The hydrodynamic flow was allowed to move
the protein sample down the capillary to a position some distance
past the detector, with the actual distance chosen based on the
expected magnitude of the protein migration velocity during the
subsequent electrophoretic step. Smaller distances were used
when the protein migration velocity was expected to be small in
order to maintain the electrophoresis time to below 10 min. The
vacuum was then disconnected, and an electric field of 8-18 kV
was applied so that the protein (or marker) would migrate back
toward the detector. The field strength was chosen such that
the current produced during electrophoresis was always less than
(7) Xueying, H.; Wirth, M. Anal. Chem., submitted.
(8) Williams B. A.; Vigh, G. Anal. Chem. 1996, 68, 1174-1180.

1582 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

60 A to minimize Joule heating. pH changes during electrophoresis were less than 0.05 unit due to the added buffer and the
short times required for data acquisition. Runs performed without
buffering and/or hydrodynamic mobilization had unacceptable pH
variations.
The migration of the protein and neutral marker through the
negatively charged silica capillary was dominated by the electroosmotic flow; thus the field polarity was chosen so that the
direction of the electroosmotic flow was back toward the detector;
i.e., the sample injection end of the capillary was maintained as
the cathode. The field polarity was reversed for the CElect Amine
capillary, which has a strong positive charge. The situation was
more complex for the weakly charged polyacrylamide-coated
capillary. Since the capillary has a small negative charge, the
sample injection end of the capillary was maintained as the cathode
whenever the electroosmotic flow dominated the migration or
when the protein migrated in the same direction as the electroosmotic flow (i.e., when the protein is positively charged). However,
above pH 5.4, the protein had a sufficient negative charge that its
electrophoretic velocity was opposite to, and greater than, the
electroosmotic flow. The field polarity under these conditions was
thus reversed, with the injection end of the capillary maintained
as the anode. Note that this made it necessary to run the protein
and neutral marker separately, using opposite field polarity for
the two runs. The protein and neutral marker could be run
simultaneously under conditions where the electroosmotic velocity
was much higher than the protein migration velocity, yielding the
protein migration velocity and the electroosmotic flow velocity in
a single run. However, under conditions of low protein migration
velocity, much better peak resolution and more accurate mobility
data were obtained by running the protein and neutral marker
separately. This also allowed the time for the electrophoretic
migration of the protein and marker to be optimized individually.
These experiments were performed with the neutral marker
injected into the capillary immediately after completing the protein
run, without any flushing of the capillary. This essentially
eliminated any changes in the magnitude of the electroosmotic
flow between runs.
Data Analysis. The mean velocity during the hydrodynamic
mobilization was evaluated directly from the time, t1, at which the
protein peak first passed the detector:

Vh )

Ld

(1)

t1 + (tinj/2)

where Ld is the distance from the injection end of the capillary to

the midpoint of the detection window. The second term in the
denominator accounts for the movement of the protein centroid
during its injection into the capillary. The protein then moves a
distance h past the detector:

h ) Vh(t2 - t1)

(2)

where t2 is the time at which the vacuum was disconnected. The

observed migration velocity of the protein in response to the
applied electric field was then evaluated as

Vprotein )

t2 - t 1
h
) Vh
t
t

(3)

where t is the time required for the peak centroid to migrate

surface charge distribution. The mobility is then a linear function

of the surface potential, :9

U)

2
f (a)
3 1

(4)

where is the solution viscosity,  is the permittivity of the solvent,

a is the protein radius, and is the inverse Debye length:

Figure 1. BSA mobility in 10 mM NaCl. Results are shown for the

different buffer solutions and capillaries: (b) phosphate, CElect
Amine; (4) acetate, CElect Amine; (0) acetate, polyacrylamide; (2)
BisTris, polyacrylamide; (+) Tris, polyacrylamide; (O) borate, fused
silica. The dashed curve represents the free-flow electrophoresis
results obtained by Douglas et al. in 20 mM NaOAc and NaMes.

back past the detector in response to the applied electric field.

The migration velocity of the neutral marker (equal to the
electroosmotic flow velocity) is also given by eq 3 but with t the
time required for the marker to migrate back past the detector.
The actual electrophoretic velocity of the protein was then
determined from the difference between Vmarker and Vprotein (or from
the sum of Vmarker and Vprotein for runs with the polyacrylamidecoated capillary in which opposite field polarity was used for the
protein and the marker). The electrophoretic mobility, U, was
calculated as the ratio of the electrophoretic velocity (cm/s) to
the applied field strength (V/cm).
RESULTS
Experimental data for BSA mobility in 10 mM sodium chloride
solutions are shown in Figure 1. The results were obtained using
the three different columns (CElect Amine, polyacrylamide-coated,
and fused silica) and five different buffers depending on the pH
range. All data points represent the average of at least two repeat
measurements. The data were highly reproducible, with repeat
measurements differing by less than 4%. (The error bars are not
shown since they were smaller than the size of the symbols used.)
The mobility values for these different systems collapse onto a
single curve, demonstrating that the calculated mobilities are
unaffected by the millimolar levels of added buffer or the specific
charge characteristics of the capillary. For example, the mobility
at pH 8.0 determined using the polyacrylamide capillary and
BisTris buffer was -22.6 ( 0.9 10-5 cm2 V-1 s-1 compared to
-22.3 ( 1.0 10-5 cm2 V-1 s-1 for the mobility in the fused-silica
capillary and borate buffer. The dashed line in Figure 1 represents
the mobility data obtained by Douglas et al.4 in 20 mM sodium
acetate or sodium 2-(N-morpholino)ethanesulfonate (Mes) using
free-flow electrophoresis. The results are in good agreement with
the data obtained in this study with the small discrepancies likely
due to the difference in total ionic strength and/or differences in
anion binding (chloride versus acetate or Mes).
The net protein charge can be evaluated directly from the
mobility data using an appropriate model for electrophoretic
motion. In this case, we have used the simplest model in which
the protein is treated as a nonconducting sphere with uniform

F2

RT

2 1/2

izi

(5)

where F is Faradays constant, R is the universal gas constant, T

is absolute temperature, and ci and zi are the bulk concentration
and valence of ion i, respectively, with the summation performed
over all ions in solution. The function f1(a) accounts for the
distortion of the electric field caused by the presence of the sphere
and was evaluated by Henry10 as

1
5
1
1
(a)2 (a)3 (a)4 +
(a)5 +
16
48
96
96
a exp(-t)
1
1
(a)4 (a)6 exp(a)
dt (6)
8
96
t

f1(a) ) 1 +

Overbeek and Wiersema11 obtained a more complete solution

to the governing equations, with the results in good agreement
with eq 4 for < 50 mV for 1-1 electrolytes. The net charge of
the particle is then evaluated from the surface potential as11

Z ) 4a(1 + a)/e

(7)

where e is the electronic charge and a is the particle radius (a )

34.5 for BSA12,13).
The mobility data in Figure 1 were used to evaluate the net
BSA charge as a function of pH with the results shown in Figure
2. The solid curve represents the net charge calculated using
the model for H+ ion binding developed by Tanford et al.14 from
hydrogen ion titration data in combination with the chloride ion
binding evaluated using the model developed by Scatchard et al.15
The results are in good agreement for pH 4-9. The deviations
at the very high and low pH are likely due to the ionic relaxation
effect,11 which has been neglected in the model (eq 6) as well as
any errors associated with the use of the Debye-Huckel (low
surface potential) approximation in eqs 4 and 7. The acid
expansion of BSA13 at pH <4.0 could also contribute to the error
at low pH.
Figure 3 shows the calculated values of the net BSA charge at
pH 5.3 as a function of the anion concentration for data obtained
using NaCl and NaClO4 solutions. In both cases, the protein
(9) Hunter, R. J. Introduction to Modern Colloid Science; Oxford University
Press: New York, 1993; Chapter 8.
(10) Henry, D. C. Proc. R. Soc. 1931, A133, 106-129.
(11) Overbeek, J. Th. G.; Wiersema, P. H. In Electrophoresis: Theory, Methods
and Applications; Bier, M., Ed.; Academic Press: New York, 1967; Vol. 2,
Chapter 1.
(12) Kronman, M. J.; Foster, J. F. Arch. Biochem. Biophys. 1957, 72, 205-218.
(13) Tanford, C. Physical Chemistry of Macromolecules; John Wiley & Sons: New
York, 1961; Chapter 7.
(14) Tanford, C.; Swanson, S. A.; Shore, W. S. J. Am. Chem. Soc. 1955, 77, 64146421.
(15) Scatchard, G.; Coleman, J. S.; Shen, A. L. J. Am. Chem. Soc. 1957, 79, 1220.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1583

Figure 2. Calculated values of BSA charge determined from mobility

data as a function of pH. The solid curve represents model calculations as described in text.

Figure 4. BSA charge at pH 4.0 in mixtures of 10 mM NaCl and 10

mM NaClO4 as a function of the perchlorate fraction.

that small differences in the specific BSA formulations used in

these studies would also have an effect on the BSA charge. For
example, limited CE data obtained with fatty acid-free BSA as a
part of this study yielded a charge that was one to two electronic
charges smaller (less negative) than that for the fraction V powder.
One of the real advantages of capillary electrophoresis is that
data can be obtained in complex electrolyte solutions without the
need for different ion-selective electrodes. Figure 4 shows the
calculated BSA charge for mixtures of 10 mM NaCl and 10 mM
NaClO4 at pH 4.0 from mobility data obtained using the CElect
Amine capillary. BSA has a positive charge at this pH, with the
magnitude of the charge decreasing with increasing perchlorate
concentration due to the greater affinity of BSA for perchlorate.
This effect is quite pronounced, with the BSA charge decreasing
by nearly five electronic charge units as the solution is changed
from 10 mM NaCl to 10 mM NaClO4.

Figure 3. BSA charge at pH 5.3 as a function of the anion

concentration determined with b NaCl and 9 NaClO4.

charge becomes more negative as the anion concentration

increases due to anion binding. The BSA charge in NaCl varies
from about +1 at very low NaCl concentration to about -8 in the
70 mM NaCl. These results are in good agreement with data
obtained by Scatchard et al.15 (using ion-selective electrodes)
which yielded a BSA charge of +1.5 at very low Cl- and pH 5.1
and a charge of -7.5 in 80 mM NaCl and pH 5.4. The BSA charge
in NaClO4 was more negative than that in NaCl due to the higher
BSA affinity for perchlorate. The charge evaluated in the 43 mM
NaClO4 solution (Z ) -10.5) is in good agreement with the 10
perchlorate anions bound to BSA at pH 5.2 in a 45 mM NaClO4
solution evaluated by Carr16 using ion-selective electrodes. Note
(16) Carr, C. W. Arch. Biochem. Biophys. 1952, 40, 286-294.

1584 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

DISCUSSION
The capillary electrophoresis technique described in this study
provides a rapid, accurate, and reliable means to measure the
mobility and in turn the net protein charge in different solution
environments. The method requires extremely small amounts
of protein (picogram quantities and nanoliter volumes) and is
easily automated, making it very suitable for protein characterization, ion binding studies, and initial screening of possible separation techniques that exploit differences in electrostatic interactions.
ACKNOWLEDGMENT
This work was supported in part by a grant from Genentech
Inc., Millipore Corp., and the Delaware Research Partnership
Program.
Received for review August 18, 1997. Accepted January
29, 1998.
AC970902R