SDS capillary gel electrophoresis of proteins in microfabricated channels

Abstract

Analysis of variations in the concentrations or structures of biomolecules (e.g., mRNAs, proteins, peptides, natural products) that occur either naturally or in response to environmental or genetic perturbations can provide important insight into complex biological processes. Many biological samples are mixtures that require a separation step before quantitation of variations in the individual components. Two-dimensional denaturing gel electrophoresis has been used very effectively to separate complex mixtures of proteins, but it is time consuming and requires considerable amounts of sample. Microchannel-based separations have proven very effective in rapidly separating small amounts of nucleic acids; more recently, isoelectric focusing of proteins also has been adapted to the microchannel format. Here, we describe microchannel-based SDS capillary gel electrophoresis of proteins and demonstrate the speed and high resolution it provides. This development is an important step toward the miniaturization and integration of multidimensional and array separation methods for complex protein mixtures.

Full text

Proc. Natl. Acad. Sci. USA
Vol. 96, pp. 5372–5377, May 1999
Chemistry
SDS capillary gel electrophoresis of proteins in
microfabricated channels
SHAO YAO*, DEON S. ANEX
†
,W.BRETT CALDWELL*, DON W. ARNOLD
†
,KATHERINE B. SMITH
†
,
AND PETER G. SCHULTZ*
‡§
*Department of Chemistry, Scripps Research Institute, La Jolla, CA 92037; and
†
Sandia National Laboratories, P.O. Box 969, M.S. 9671, Livermore, CA 94551
Contributed by Peter G. Schultz, March 12, 1999
ABSTRACT Analysis of variations in the concentrations
or structures of biomolecules (e.g., mRNAs, proteins, peptides,
natural products) that occur either naturally or in response to
environmental or genetic perturbations can provide impor-
tant insight into complex biological processes. Many biolog-
ical samples are mixtures that require a separation step before
quantitation of variations in the individual components. Two-
dimensional denaturing gel electrophoresis has been used
very effectively to separate complex mixtures of proteins, but
it is time consuming and requires considerable amounts of
sample. Microchannel-based separations have proven very
effective in rapidly separating small amounts of nucleic acids;
more recently, isoelectric focusing of proteins also has been
adapted to the microchannel format. Here, we describe mi-
crochannel-based SDS capillary gel electrophoresis of pro-
teins and demonstrate the speed and high resolution it
provides. This development is an important step toward the
miniaturization and integration of multidimensional and ar-
ray separation methods for complex protein mixtures.
Methods for the direct measurement of changes in the con-
centrations and posttranslational states of proteins in complex
biological systems are useful in analyzing protein function and
identifying proteins of potential diagnostic or therapeutic
value (1, 2). For a typical biological sample, the large number
of different proteins present (up to thousands) and the small
concentrations at which they can exist make such experiments
difficult. Two-dimensional SDSyPAGE has proven to be a
powerful tool for the profiling of protein expression (3). By
combining isoelectric focusing for charge-based separation in
one dimension and SDSyPAGE for size-based separation in a
second dimension, hundreds to thousands of proteins have
been resolved on a single two-dimensional SDSyPAGE slab
gel (4, 5). Subsequently, chemical degradation andyor mass
spectrometry can be used to identify the separated compo-
nents. However, increased sensitivity and speed of detection
would improve the ability to profile proteins on a routine basis
significantly.
Improvements can be achieved by using capillary electro-
phoresis (CE), which offers many advantages for the separa-
tion of a wide variety of molecules. The method offers high
efficiency, versatility, speed, and economy of analysis (i.e., very
low consumption of reagents and analytes; ref. 6–22). Con-
siderable effort also has been directed toward miniaturization
in the field of CE—primarily in the adaptation of CE from
capillaries to a planar microchannel format (9–19). Although
chip-based CE allows better control of sample introduction
and leads to better performance in terms of speed and
efficiency, its greatest advantage lies in the opportunity to
perform parallel-array (17) or multidimensional types of anal-
yses. Conventional photolithography and microfabrication
technology enable these approaches by providing a means of
fabricating channel structures of many different sizes and
integrated devices on chips made of silicon, glass, quartz,
plastics, or elastomers (12–16).
Among the known CE methods for protein analysis, capil-
lary-zone electrophoresis has been performed in a microchan-
nel-based format with some degree of success (18, 19). More
recently, capillary isoelectric focusing has also been adapted to
the chip format (20). SDSyPAGE (23) was demonstrated in a
capillary over a decade ago (21). However, in spite of its
importance as the most commonly used protein-separation
method, it has not yet been adapted to the planar microchannel
format. We demonstrate the on-chip application of a closely
related size-dependent separation technique, SDS capillary gel
electrophoresis (SDSyCGE). By transferring this method to
planar microchannels, protein separations are accelerated by
a factor of 20, and separation efficiencies typical of conven-
tional capillary-based SDSyCGE experiments are retained.
EXPERIMENTAL PROCEDURES
Chemical Derivatization of Proteins. All protein molecular
mass (MM) markers (ranging from 9 kDa to 116 kDa) were
labeled with 5- and 6-carboxyfluorescein succinimidyl ester
(fluorescein-NHS) or fluorescein-5-maleimide (fluorescein-
MAL) (Molecular Probes) following the manufacturer’s pro-
tocols. Calmodulin,
a
-lactalbumin, pepsinogen, egg albumin,
BSA,
b
-galactosidase, and hen egg white lysozyme were pur-
chased from Sigma. Calmodulin was obtained from Sigma with
an uncharacterized MM. Its apparent MM (9 kDa) was
determined experimentally by using an SDSyPAGE mini-gel.
Staphylococcal nuclease 16-cys mutant was prepared as de-
scribed (24). All other reagents were purchased from Aldrich,
unless otherwise indicated.
For the fluorescein-NHS reactions, the protein was dis-
solved in a 0.2-M NaHCO
3
buffer (pH 8.3) to a final concen-
tration of 1 mgyml. Freshly prepared fluorescein-NHS solution
(100
m
lof10mgyml in DMSO) was slowly added to the protein
solution, and the reaction was stirred in the dark for 1 h. The
mixture (0.5 ml) was passed through a NAP-5 gel filtration
column (Amersham Pharmacia) preequilibrated with H
2
O,
and the fluorescein-labeled protein was eluted with 1 ml of
H
2
O. Proteins were purified further by either (i) two more
rounds of NAP-5 purification or (ii) overnight dialysis (MM
cutoff 5 3 kDa) into H
2
O at 4°C, followed by separation on a
FPLC Mono Q column (Amersham Pharmacia; eluent of 0.01
M NaH
2
PO
4
, pH 7.0, with a gradient of 0–1 M NaCl over 30
The publication costs of this article were defrayed in part by page charge
payment. This article must therefore be hereby marked ‘‘advertisement’’ in
accordance with 18 U.S.C. §1734 solely to indicate this fact.
PNAS is available online at www.pnas.org.
Abbreviation: CE, capillary electrophoresis; CGE, capillary gel elec-
trophoresis; MM, molecular mass; fluorescein-NHS, 5- and 6-carboxy-
fluorescein succinimidyl ester; fluorescein-MAL, fluorescein-5-
maleimide.
‡
To whom reprint requests should be addressed. e-mail:
pschultz@lbl.gov.
§
Present address: Department of Chemistry, 824 Latimer Hall, Uni-
versity of California, Berkeley, CA 94720-1460.
5372

min) and desalting of the desired fractions with Centricon
columns (Amicon).
For the fluorescein-MAL reactions, the protein was dis-
solved in 100 mM sodium phosphate buffer (pH 7.0) to a final
concentration of 100
m
M and then degassed under vacuum for
10 min. A freshly prepared fluorescein-MAL solution (20
equiv of 10 mM dye in DMSO) was slowly added to the protein,
the reaction was stirred in the dark under N
2
for 2 h, and then
it was desalted with a NAP-5 column. The fluorescein-labeled
product was purified further by using the Mono Q protocol
described above. Identities of all fluorescein-labeled proteins
were confirmed on the basis of their MM by using an SDSy
PAGE mini-gel (Bio-Rad).
Calmodulin, hen egg white lysozyme, and the staphylococcal
nuclease 16-cys mutant (100
m
M each) were treated with the
two different dyes (fluorescein-NHS and fluorescein-MAL) as
described above. Equimolar amounts of the three maleimide-
labeled proteins were mixed to yield a sample of specifically
labeled proteins. For comparison purposes, a similar sample of
nonspecifically labeled proteins was prepared by mixing
equimolar amounts of the three NHS-labeled proteins.
Capillary-Based SDSyCGE. All capillary-based SDSyCGE
separations were performed by using a PyACE 5000 capillary
electrophoresis system (Beckman Coulter) with absorbance
detection (214 nm) or laser-induced fluorescence detection
(excitation at 488 nm; emission at 520 nm). Separations were
performed by using an eCAP SDS 14-200 kit from Beckman
Coulter with either an eCAP neutral capillary (27 cm 3
100-
m
m i.d.; Beckman Coulter) or an uncoated fused-silica
capillary (27 cm 3 100-
m
m i.d.; Polymicro Technologies,
Phoenix). The distance from inlet to the detection window was
20 cm.
Before each analysis, the capillary was conditioned with a
series of pressure washes (20 psi; 1 psi 5 6.89 kPa): (i)1M
NaOH for 2.5 min; (ii)H
2
O for 1 min; (iii) 1 M HCl for 1 min;
(iv)H
2
O for 1 min; and (v) SDS 14-200 gel buffer (Beckman
Coulter) for 3 min. The performance of the instrument and the
capillary was ensured by periodically testing with the Beckman
Coulter protein test mix. The sample solution was prepared
freshly by mixing 50
m
l of the protein mixture, 50
m
l of the SDS
sample buffer (provided with the SDS 14-200 kit), and 3
m
lof
b
-mercaptoethanol and then by heating the mixture at 95°C for
3 min. The sample was injected into the capillary either
electrokinetically (4 kV) or by pressure (20 psi). The separa-
tion voltage was 8.1 kV (300 Vycm) as recommended by the
manufacturer. All voltage operations were performed in a
reverse polarity mode (inlet at negative potential with respect
to outlet).
Chip Fabrication. Chips were fabricated from borofloat
glass wafers (4-inch diameter; 1-mm thickness; Schott Labo-
ratories, Yonkers, NY) by using standard methods (17). The
wafers were cleaned before the deposition of an amorphous
silicon sacrificial layer (1,500 Å) in a plasma-enhanced chem-
ical vapor deposition (PECVD) system (PEII-A, Technics
West, San Jose, CA). After Piranha cleaning, the wafers were
dried at 120°C for 5 min before being primed with hexameth-
yldisilazane, spin-coated with photoresist (Shipley 1818, Marl-
borough, MA) at 6,000 rpm for 30 s, and then soft baked at
90°C for 30 min. The mask pattern was transferred to the
photoresist on the wafers by exposing the photoresist to UV
radiation in a Quintel contact mask aligner. The photoresist
was developed in a 1:1 mixture of Microposit developer
concentrate (Shipley) and H
2
O, followed by hard baking at
120°C for 25 min. The mask pattern on the photoresist then was
transferred to the amorphous silicon by a CF
4
plasma etch
performed in the PECVD reactor. The wafers were etched in
49% (volyvol) HF for different lengths of time at a vertical-
etch rate of 7
m
mymin. The photoresist was stripped in a spin
dryer, and the remaining amorphous silicon was removed with
aCF
4
plasma etch. Access holes (0.75 or 1.1 mm) were drilled
through the etched wafers by using diamond-tipped drill bits
(Olympic Mountain Gems, Port Orchard, WA). An unetched
Borofloat wafer was then thermally bonded, as a cover, to the
etched channel wafer in an N
2
-purged programmable furnace
(Thermolyne, Dubuque, IA).
Microchannel Layout for Chip-Based SDSyCGE. Because
of the high viscosity of the SDS 14-200 gel used in the
separation, it was anticipated that a microchannel size com-
parable to that of a 100-
m
m i.d. capillary would be needed for
easy handling. Channels etched to a depth of 20
m
m proved to
be too shallow for gel to pass through without frequent
clogging. A 40-
m
m-deep channel was found to be better and
was used for all subsequent chip-based separations. The mi-
crochannel layout for chip-based SDSyCGE used in our
experiments is shown in Fig. 1. The channel is 40
m
m deep, with
a 100-
m
m width at the top and a 20-
m
m width at the bottom.
The distances between reservoirs 1 and 4 and reservoirs 2 and
3 are 5 cm and 0.5 cm, respectively. Laser-induced fluores-
cence detection was typically performed 0.25 cm from reser-
voir 4, affording a separation channel 4.5 cm in length. In
addition, the distances from 1, 2, and 3 to the cross-channel
intersection are 0.25 cm each.
Laser-Induced Fluorescence Detection for Chip-Based
SDSyCGE. The detection system for the chip-based SDSyCGE
is shown schematically in Fig. 1. Similar epifluorescencey
microscope-based detectors for microfabricated separation
systems have been reported in the literature (25). Light from
an argon-ion laser (Model 532, Omnichrom, Chino, CA)
operating at 488 nm is passed through a bandpass filter and
directed by a dichroic mirror through an aspheric lens (Model
5722-H-A, New Focus, Santa Clara, CA). The aspheric lens
serves as a microscope objective (403; 0.55 numerical aper-
FIG. 1. Schematic of the instrumentation set up used for chip-
based SDSyCGE separations. The microchannel layout used in the
chip-based SDSyCGE separations also is shown. F1, excitation band-
pass filter; F2, emission bandpass filter; PMT, photomultiplier tube;
TS
x,y,z
, x-y-z translational stage.
Chemistry: Yao et al. Proc. Natl. Acad. Sci. USA 96 (1999) 5373

ture; 2.9-mm working distance), focusing the laser light into
the separation channel and collecting the fluorescence. Col-
lected light passes through the dichroic mirror and a bandpass
filter centered on 535 nm with a 55-nm bandwidth. The filters
and the dichroic mirror were purchased as a set (XF100
fluorescein filter set, Omega Optical, Brattleboro, VT). A
50-mm focal length lens (Spindler and Hoyer, Milford, MA)
then focuses the emitted light through an adjustable iris onto
a photomultiplier tube (Model HC-120-05, Hamamatsu,
Bridgewater, NJ). A home-built power supply controls the
photomultiplier tube bias and provides the power for an
internal preamplifier. Output from the photomultiplier tube is
sent to a personal computer via an interface (PCI-1200 DAQ
card with a BNC-2081 board, National Instruments, Austin,
TX), and the data are collected with a custom
LABVIEW
(National Instruments) program. A low-pass resistance–
capacitance filter was used to reduce the high-frequency noise.
The microfabricated separation chip is held in a x-y-z translator
for positioning the detection region of the chip at the focus of
the objective. High voltage to drive the separations is provided
by 5-kV power supplies (SRS model PS350, Stanford Re-
search, Sunnyvale, CA).
Chip-Based SDSyCGE Separations. Pipette tips were in-
serted into the drilled holes in the chip to serve as fluid
reservoirs. All fluids used in the chip-based separation were
prefiltered with 0.22-
m
m syringe filters. Before sample loading,
the channels first were cleaned and rinsed as follows: (i)1M
NaOH for 1 min; (ii)H
2
O for 1 min (iii) 1 M HCl for 1 min;
and (iv)H
2
O for 1 min. The SDS 14-200 gel was loaded into
the channels by first filling reservoirs 1–3 with the gel and then
applying a vacuum at reservoir 4 to pull the gel into the
channels. Care was taken to ensure no air bubbles were
introduced in the channels during gel loading. After the
channels were filled completely, vacuum at reservoir 4 was
removed and replaced with a pipette tip prefilled with gel. The
gel in reservoirs 2 and 3 was removed subsequently and
replaced with a 1:1 mixture of fresh SDS sample buffer and
H
2
O. The channels were equilibrated by applying 200 V
between reservoirs 2 and 3 and then 2 kV between 1 and 4,
each for more than 5 min. Immediately before injection, the
content of reservoir 2 was replaced with the protein sample
(prepared as described earlier for the capillary-based separa-
tion). A simple floating injection (26) was used; an injection
voltage (200 V) was applied between reservoirs 2 and 3 for
5–30 s to draw the sample across the cross-channel region, and
then a separation voltage (1–5 kV) was applied between
reservoirs 1 and 4. Because a stable current is necessary for
good separations, current in the separation channel was mon-
itored closely during the entire separation process. Data
collection was performed at 400 Hz, with intervals of 10 points
averaged to yield a rate of 40 points per s.
RESULTS AND DISCUSSION
The main focus of this work is chip-based protein separations
with SDSyCGE. However, for performance comparisons, both
capillary-based and chip-based SDSyCGE separations were
performed by using identical protein samples and separation
media. Two types of labeled proteins were used in these
experiments: proteins with multiple labels (nonspecifically
labeled with amine-reactive fluorescein-NHS) and proteins
with single labels (labeled specifically with the thiol-reactive
dye fluorescein-MAL on a single cysteine in the protein
sequence). We will refer to these proteins as ‘‘nonspecifically’’
and ‘‘specifically’’ labeled, respectively.
The first sample consisted of six fluorescently labeled pro-
teins (ranging from 9 kDa to 116 kDa, see the legend in Fig.
2 for details). Calmodulin, which contains only one cysteine in
its protein sequence, was labeled specifically and purified to
yield the singly fluorescein-modified product. The other five
proteins were labeled nonspecifically.
All six proteins were well resolved within 15 min with
capillary-based SDSyCGE by using Beckman Coulter’s eCAP-
coated capillary as indicated in Fig. 2a. The peak correspond-
ing to singly labeled calmodulin (peak 1 in Fig. 2a)issym-
metrical and sharp, indicating the high-resolving nature of
capillary-based SDSyCGE. The other protein peaks are sep-
arated but are much broader, indicating the presence of a
heterogeneously labeled protein population caused by non-
specific labeling (see below), the effects of longer migration
times for the larger proteins, or a combination of these.
For comparison with the chip-based separation conditions,
where uncoated channels are used, the capillary SDSyCGE
was performed with an uncoated fused-silica capillary. The
proteins were still separated on the uncoated capillary (Fig. 2b)
but with reduced resolution. Peak broadening was likely
caused by nonspecific interactions between proteins and the
charged inner surface of the uncoated capillary.
The same protein mixture was then used for chip-based
separations. As shown in Fig. 3, a separation voltage of 1 kV
resolves all six proteins within 3.5 min. In fact, the pattern of
the peaks is quite similar to the capillary separation (Fig. 2b),
but the time required for the on-chip separation is more than
a factor of five shorter. Increasing the voltage from 1 kV to 5
kV increases the speed concomitantly, with the separation
FIG. 2. Electropherograms from capillary-based SDSyCGE on a
six-protein mixture with an eCAP-coated capillary (a) and an un-
coated fused-silica capillary (b). Capillary dimensions: 100-
m
m i.d. 3
20 cm (27 cm in total length). Separation voltage: 8.1 kV. Injection: 4
kV for 30 sec. Peaks: (1) calmodulin (MM 5 9 kDa); (2)
a
-lactalbumin
(MM 5 14.4 kDa); (3) pepsinogen (MM 5 39 kDa); (4) egg albumin
(MM 5 45 kDa); (5) BSA (Mm 5 66 kDa); and (6)
b
-galactosidase
(MM 5 116 kDa). Each protein concentration in the mixture was '1 3
10
29
Mto103 10
29
M. Injection: 4 kV for 30 s. (These conditions
were optimized.)
5374 Chemistry: Yao et al. Proc. Natl. Acad. Sci. USA 96 (1999)

being complete in less than 35 s at a separation voltage of 5 kV.
The increase in separation speed compromised the separation
efficiencies, as evidenced by decreases in theoretical plate
numbers and increases in plate heights at higher separation
voltages (see Table 1). However, a maximum in time-based
separation efficiencies (Nys) occurs at intermediate operating
voltages.
In the floating-injection approach used here, the sample is
migrated across the intersection by applying a voltage between
reservoirs 2 and 3 (see Fig. 1), while the other two reservoirs
are allowed to float. The effect of migrating the sample across
the intersection for different lengths time was investigated.
Variations of the migration time from5sto30swhile
maintaining a constant injection voltage (200 V) did not
appreciably alter either the quality of the separation or the
peak intensity (data not shown). These results indicate mini-
mal leakage of the sample into the separation channel during
the injection (presumably because of the viscous nature of the
gel used in the chip and the slow diffusion of the proteins) and
that sufficient time was allowed for all components to migrate
to the intersection.
For the coated and uncoated capillaries and the chip, the
separation performance for the nonspecifically labeled pro-
teins was quantified for several peaks (see Table 1). Strictly
speaking, theoretical plate numbers should be calculated by
using a peak from a single component. By their nature, protein
samples are heterogeneous. In addition, the nonspecifically
labeled proteins have heterogeneity because of variation in the
number of attached dyes. As a result, the calculated theoretical
plate numbers are a reflection of both the column efficiency
and the heterogeneity of a particular sample component. The
calculations are intended for comparison of performance
between capillaries and for comparison of the column-based
separations with the chip-based separations.
For the coated capillary, peak 2 yields a theoretical plate
number of 6.6 3 10
3
, which corresponds to a plate height of 30
m
m. The separation efficiency decreases in the uncoated
capillary, where peak 2 has a theoretical plate number of 3.5 3
10
3
(58-
m
m plate height). A similar trend is observed for peak
5, where the efficiency decreases from 3.1 3 10
3
theoretical
plates (65-
m
m plate height) in the coated capillary to 9.7 3 10
2
theoretical plates (207-
m
m plate height) in the uncoated one.
For peaks 3 and 4, which correspond to proteins differing in
MM by '15% (39 kDa and 45 kDa, respectively), a resolution
of 1.6 was achieved by using the coated capillary. With the
FIG. 3. Electropherograms of protein mixture (Fig. 2) with micro-
channel-based SDSyCGE with different separation voltages: 1 kV (a),
2kV(b),3kV(c),4kV(d),and5kV(e).
Table 1. Comparison of capillary-based vs. chip-based SDSyCGE for nonspecifically labeled proteins
a
a
Calculated from Figs. 2 and 3. All calculations were based on standard equations (27).
b
For peaks 3 and 4, the resolution was calculated by using the expression R 5 2 3 (t
3
2 t
4
)y(Dt
1y2,3
1Dt
1y2,4
), where t
3
and t
4
are the migration times of peaks 3 and 4 and Dt
1y2,3
and Dt
1y2,4
are the respective full widths at half maximum.
c
Theoretical plate numbers (N) were calculated for peaks 2 and 5. These calculations were based on the measured full width at
half maximum (Dt
1y2
) of a peak by using the expression N 5 5.54 3 (tyDt
1y2
)
2
, where t is the migration time.
d
Plate heights (H) are calculated by dividing the column length by the theoretical plate numbers. For CE-based separations, a
column length of 20 cm was used. For chip-based separations, a column length of 4.5 cm was used.
e
Theoretical plates per second (Nys) were calculated by dividing theoretical plate numbers (N) by migration time.
Chemistry: Yao et al. Proc. Natl. Acad. Sci. USA 96 (1999) 5375

uncoated fused-silica capillary, however, a decrease in reso-
lution to 1.3 was observed.
For the chip-based separations, the number of theoretical
plates was generally better than for the uncoated capillary
(Table 1). The efficiency, in terms of plate height, was roughly
an order of magnitude better for the chip-based separation
than for the capillary. For the time-based efficiency, the
performance generally improved with increasing separation
voltage. For example, peak 2 has 5.2 3 10
2
theoretical plates
per s at 3 kV, corresponding to an electric field of 600 Vycm.
For peaks 3 and 4, good resolution was achieved with sepa-
ration voltages of 1 kV and 2 kV, exceeding the resolution of
the coated capillary for these peaks. Even with higher sepa-
ration voltages (3–5 kV), moderate resolution comparable to
that of capillary-based SDSyCGE with the uncoated fused-
silica capillary was still maintained.
As indicated in Figs. 2 and 3, specifically labeled calmodulin
eluted as a very sharp and symmetric peak in both the
capillary-based and chip-based SDSyCGE. In contrast, larger
nonspecifically labeled proteins eluted as much broader peaks.
Because the reaction conditions for fluorescein-NHS labeling
lead to nonspecific labeling of all primary amines in a protein,
each labeling reaction likely generates a mixture containing the
same protein modified with a variable number of dyes. This
likely is a major cause of peak broadening and loss of resolu-
tion in our SDSyCGE analyses.
To test the separation performance of the chip-based system
with minimized sample-dependant contributions to broaden-
ing, mixtures of three single-cysteine-containing proteins were
analyzed. Calmodulin, lysozyme, and staphylococcal nuclease
16-cys mutant, labeled either nonspecifically or specifically,
were first analyzed by SDSyCGE by using an uncoated fused-
silica capillary (Fig. 4a). These proteins, which differ by '20%
in MM, were resolved completely in the specifically labeled
mixture (Fig. 4a, Bottom). In contrast, the nonspecifically
labeled protein mixture produced an electropherogram in
which the three protein peaks were resolved poorly and much
more broadly (Fig. 4a, Bottom), showing the effect of the
presence of differently modified proteins on the separation
efficiency. Finally, the specifically labeled protein mixture was
analyzed with chip-based SDSyCGE at a range of separation
voltages (Fig. 4b). All three proteins were well resolved under
all separation voltages used, with separation being complete in
less than 25 s at a separation voltage of 5 kV (Fig. 4b, Inset).
The calculated separation performance parameters for the
specifically labeled proteins are included in Table 2. The
capillary-based separation yield theoretical plate numbers
ranging from 9.4 3 10
3
to 1.8 3 10
4
, which correspond to plate
heights from 21
m
mto11
m
m (Table 2). For the chip-based
separations, the plate numbers are again generally higher than
for the capillary, and the plate heights are about an order of
magnitude smaller. As was true for the nonspecifically labeled
proteins, the optimum voltage for the theoretical plate number
is between 2 kV and 3 kV (electric field of 400–600 Vycm). In
this voltage range, the theoretical plate number varies from
3.2 3 10
4
to 4.6 3 10
4
(1.4-
m
m to 0.98-
m
m plate heights). The
time-based efficiencies range from 1.9 3 10
2
plates per s to
4.1 3 10
2
plates per s at 1 kV. They increase with driving
voltage, reaching a plateau near 1.2 3 10
3
plates per s for peaks
1 and 2. For peak 3, the time-based efficiencies reach a
maximum of 8.6 3 10
2
plates per s at 3 kV and then fall to 4.2 3
10
2
plates per s at 5 kV.
We have been able to transfer successfully the capillary-
based SDSyCGE for protein separation onto a microchannel-
based format. Compared with capillary-based separations, the
chip-based SDSyCGE separation of proteins is clearly superior
in terms of separation efficiency and speed. We demonstrate
that six proteins, ranging from 9 kDa to 116 kDa in MM, can
be separated in less than 35 s, while comparable separation
efficiencies are maintained. Specifically labeled proteins dif-
fering in MM by '20% can be resolved easily with high
resolution in less than 25 s. With chip-based separations, plate
heights near 1
m
m can be obtained under optimized conditions,
which is 10 times better than the capillary-based SDSyCGE. As
for time-based efficiency, more than 1,000 plates per s were
obtained in chip-based separations, which is more than a
20-fold improvement from the capillary-based separation.
Further improvements in chip-based separations can be ex-
pected through the use of more sophisticated electronic equip-
ment for fluidic control, increasing the length of the separation
FIG. 4. Electropherograms of three cysteine-containing, fluores-
cently labeled proteins (peak 1, calmodulin with an apparent MM of
9 kDa; peak 2, lysozyme (hen egg white) with a MM of 14.5 kDa; peak
3, staphylococcal nuclease 16-cys mutant with a MM of 18 kDa). (a)
Capillary-based SDSyCGE with the proteins labeled nonspecifically
with fluorescein-NHS (Top) or labeled specifically with fluorescein-
MAL (Bottom). Capillary: 100-
m
m i.d. 3 20 cm (27 cm total length)
uncoated fused-silica. Separation voltage: 8.1 kV. Injection: 4 kV for
30 s. Note the two different free dyes migrate differently under
identical separation conditions. (b) Chip-based SDSyCGE of the
specifically labeled proteins with different separation voltages. The
Inset highlights the three well resolved protein peaks at a separation
voltage of 5 kV. Each protein concentration in the mixture was
estimated to be '1 3 10
29
Mto103 10
29
M.
5376 Chemistry: Yao et al. Proc. Natl. Acad. Sci. USA 96 (1999)

channel, coating the separation channel, optimizing the chan-
nel depth, and improving the sample preparation.
Extending the current work, chip-based SDSyCGE could be
incorporated into more complex separation schemes. The
simplest incorporation would be chip-based SDSyCGE with
postcolumn fluorescent labeling. A second area could include
chip-based SDSyCGE in an array format where hundreds of
channels containing the same or different separation media
can be run in parallel on a single glass wafer. A third would be
a chip-based two-dimensional SDSyPAGE analog in which
tandem separations (for example, isoelectric focusing followed
by SDSyCGE) can be integrated into one device. Such meth-
ods could potentially provide extremely fast and powerful tools
for analyzing complex protein mixtures.
We thank Dave Neyer for helpful discussions and Gary Hux for
technical assistance (both from Sandia National Laboratories). Mi-
crofabrication was performed at the University of California, Berke-
ley, Microfabrication Laboratory. S.Y. would like to thank the Cancer
Research Institute for a postdoctoral fellowship. W.B.C. would like to
acknowledge an Alexander Hollaender Postdoctoral Fellowship spon-
sored by the Department of Energy and administered by the Oak
Ridge Institute for Science and Education. This work was supported
in part by the Laboratory Directed Research and Development
program of Sandia National Laboratories and by the National Insti-
tutes of Health. Sandia is a multiprogram laboratory operated by
Sandia Corporation, a Lockheed Martin Company, for the United
States Department of Energy under contract DE-AC04-94AL85000.
1. James, P. (1997) Q. Rev. Biophys. 30, 279–331.
2. Celis, J. E. (1998) Drug Discovery Today 3, 193–195.
3. Celis, J. E. & Gromov, P. (1999) Curr. Opin. Biotechnol. 10,
16–21.
4. Lopez, M. F. (1999) J. Chromatogr. B 722, 191–202.
5. Haynes, P. A., Gygi, S. P., Figeys, D. & Aebersold, R. (1998)
Electrophoresis 19, 1862–1871.
6. Watzig, H., Degenhardt, M. & Kunkel, A. (1998) Electrophoresis
19, 2695–2752.
7. Landers, J. P. (1997) Handbook of Capillary Electrophoresis
(CRC, Boca Raton, FL), 2nd Ed.
8. Weinberger, R. (1993) Practical Capillary Electrophoresis (Aca-
demic, San Diego).
9. Effenhauser, C. S. (1998) in Microsystem Technology in Chemistry
and Life Sciences, eds. Manz, A. & Becker, H. (Springer, Hei-
delberg), pp. 51–82.
10. Effenhauser, C. S., Bruin, G. J. M. & Paulus, A. (1997) Electro-
phoresis 18, 2203–2213.
11. Van den Berg, A. & Bergveld, P., eds. (1995) Micro Total Analysis
Systems (Kluwer, Boston).
12. Harrison, J. D., Fluri, K., Seiler, K., Fan, Z., Effenhauser, C. S.
& Manz, A. (1993) Science 261, 895–897.
13. Burns, M. A., Johnson, B. N., Brahmasandra, S. N., Handique, K.,
Webster, J. R., Krishnan, M., Sammarco, T. S., Man, P. M., Jones,
D., Heldsinger, D., Mastrangelo, C. H. & Burke, D. T. (1998)
Science 282, 484–487.
14. Boone, T. D. & Hooper, H. H., eds. (1998) in Micro Total Analysis
Systems ‘98 (Kluwer, Boston), pp. 257–260.
15. Duffy, D. C., McDonald, J. C., Schueller, O. J. & Whitesides,
G. M. (1998) Anal. Chem. 70, 4974–4984.
16. Manz, A., Harrison, J. D., Verpoorte, E. & Widmer, M. H. (1993)
Adv. Chromatogr. 33, 1–65.
17. Simpson, P. C., Roach, D., Woolley, A. T., Thorsen, T., Johnson,
R., Sensabaugh, G. F. & Mathies, R. A. (1998) Proc. Natl. Acad.
Sci. USA 95, 2256–2261.
18. Chiem, N. & Harrison, J. D. (1997) Anal. Chem. 69, 373–378.
19. Colyer, C. L., Mangru, S. D. & Harrison, D. J. (1997) J. Chro-
matogr. A 781, 271–276.
20. Hofmann, O., Che, D., Cruickshank, K. A. & Muller, U. R. (1999)
Anal. Chem. 71, 678–686.
21. Cohen, A. S. & Karger, B. L. (1987) J. Chromatogr. 397, 409–417.
22. Guttman, A., Nolan, J. A. & Cooke, N. (1993) J. Chromatogr. 632,
171–175.
23. Laemmli, U. K. (1970) Nature (London) 277, 680.
24. Shin, I. J., Ting, A. Y. & Schultz, P. G. (1997) J. Am. Chem. Soc.
119, 12667–12668.
25. Ocvirk, G., Tang, T. & Harrison, D. J. (1998) Analyst 123,
1429–1434.
26. Jacobson, S. C., Hergenroder, R., Koutny, L. B., Warmack, R. J.
& Ramsey, J. M. (1994) Anal. Chem. 66, 1107–1113.
27. Giddings, J. C. (1991) Unified Separation Sci. (Wiley, New York).
Table 2. Comparison of capillary-based vs. chip-based SDSyCGE for specifically labeled proteins
a
a
Calculated from Fig. 4a (Bottom) and b. See legend in Table 1 for details.
Chemistry: Yao et al. Proc. Natl. Acad. Sci. USA 96 (1999) 5377