Clinical Chemistry 51:10
1955–1961 (2005)
Oak Ridge Conference
Simultaneous Multiple Immunoassays in a Compact
Disc–Shaped Microfluidic Device Based on
Centrifugal Force
Nobuo Honda,1* Ulrika Lindberg,2 Per Andersson,2 Stephan Hoffmann,2 and
Hiroyuki Takei1
Background: We explored the potential of a microfluidic device based on centrifugal force as an immunoassay platform by examining the imprecision of assays
carried out with 200 nL of sample.
Methods: Biotinylated antibodies against ␣-fetoprotein
(AFP), interleukin-6 (IL-6), and carcinoembryonic antigen
[(CEA); 0.1 g/L each in 15 mmol/L phosphate-buffered
saline (PBS) containing 0.1 mL/L Tween 20] were attached
to a microcolumn packed with streptavidin-coated particles. A 200-nL sample was then allowed to pass through
the microcolumn for 240 s, followed by Alexa 647–labeled
detection antibody (7.5 mg/L in 15 mmol/L PBS containing
10 g/L bovine serum albumin). The flow rate was controlled by altering the rotational speed. Up to 104 sandwich type immunoassays were completed within 50 min.
Results: For AFP, IL-6, and CEA the detection limits
were, respectively, 0.15, 1.25, and 1.31 pmol/L. Inter- and
intraassay imprecisions (CVs) were <10% and <20%,
respectively, for analyte concentrations >5 pmol/L. The
CEA antibody had the lowest affinity according to
fluorescence image analysis of the microcolumn region.
The result was confirmed in a comparative study using
BIAcore 3000.
Conclusions: Day-to-day (total) imprecision (CV) of
immunoassays on the compact disc–shaped device are
<20%. Analysis of fluorescence images allows rapid
ranking of antibodies according to their affinities.
© 2005 American Association for Clinical Chemistry
1
Fujirebio, Inc., Tokyo, Japan
Gyros AB, Uppsala, Sweden.
*Address correspondence to this author at: Methodology Research Group,
Research & Development Division, Fujirebio Inc., 51, Komiya-cho, Hachiojishi, 192-0031, Tokyo, Japan. Fax 81-426-46-8325; e-mail no-honda@fujirebio.
co.jp.
Received April 26, 2005; accepted July 18, 2005.
Previously published online at DOI: 10.1373/clinchem.2005.053348
2
There has been much excitement surrounding use of
microfluidic devices for rapid clinical immunoassays.
Microfluidic devices come in various sizes and shapes
(1, 2 ), but common to all is a structure consisting of
channels and reactors of microscopic dimensions. Minute
quantities of reagents and sample are introduced into the
microstructure and transported from one section to another while undergoing mixing, reaction, and monitoring.
Some of the touted promises are reductions in sample and
reagent volumes, rapid results as a result of enhanced
mixing and reaction efficiencies, and true user-friendliness through full automation. These characteristics are of
utmost importance in the face of ever-increasing medical
costs: whereas accurate and timely results can greatly
reduce the costs of the subsequent medical treatment,
incorrect results not only can lead to complications in
treatment but potentially also to exorbitant legal fees
arising from litigation. Microfluidic devices can also be
useful in the search for improved antibodies for new
generations of diagnostic immunoassays if high throughput via a high degree of parallelism can be achieved with
small reagent volumes.
The development of such microfluidic devices has been
challenging. Some of the technical hurdles are associated
with (a) dispensing of sample and reagents in precisely
defined volumes into a microfluidic device, (b) removal of
trapped air, (c) efficient mixing of discrete fluids, (d)
transfer of the fluid between chambers, and (e) acceptable
signal intensity. Numerous approaches have been proposed and implemented. One way to classify these approaches is by the way fluids are transported within the
microfluidic structures. The dynamics of a liquid of macroscopic quantity are dominated by its mass, but as the
dimension decreases, surface tension begins to dominate
the behavior. Although this might complicate the transport issue, surface effects can be also exploited. With this
in mind, the use of centrifugal force in a rotating device is
rather attractive because under centrifugation, it is possi-
1955
1956
Honda et al.: Centrifugal Force–Based CD Device for Immunoassays
ble to control the apparent “weight” of the fluid; this
approach has been rather successful at addressing some of
above technical hurdles (3– 6 ). When a device is subjected
to centrifugal force, the fluid within experiences a force,
described by:
␳␻2 r⌬r
where ␳ is the density of the fluid, ␻ is the angular
velocity of the rotating device, r is the characteristic
distance of the fluid location from the center of rotation,
and ⌬r is the height of the fluid in the radial direction. The
centrifugal force is particularly well adapted for transporting discrete fluid droplets. In diagnostic applications,
it is droplets of precisely well-defined volume that need to
be transported. Moreover, when centrifugal force is applied to the fluid, separation of air is rather straightforward.
We have been actively engaged in efforts to make use
of centrifugal force for diagnostic devices. In this study,
we evaluated the microfluidic device called the Gyrolab
Bioaffy™ (Gyros AB), which is shaped like a compact disc
(CD)3 with a diameter of 12 cm (Fig. 1). It is a polymer
structure covered by a lid. Formed into this area are 104
identical structures, each capable of carrying out a single
test. The structure consists of a microcolumn filled with
streptavidin-coated particles, and any biotinylated biomolecule can be readily attached to the surface. The rest
of the structure serves to define the volumes of the sample
and reagents as well as keeping the flow rate constant
through the microcolumn. All control is through the
rotational speed of the CD, which is mounted on a drive
mechanism of Gyrolab WorkstationTM. In addition to
driving the CD at speeds in accordance to a preprogrammed sequence, the Workstation is equipped with a
detector unit for laser-induced fluorescence detection. It is
also equipped with robotic arms that can dispense liquid
samples into arbitrary inlets on the CD. The Workstation
has a carousel that carries up to six 96- or 384-well plates
used for storing liquid samples for dispensing. Rapid
dispensing is essential for achieving a high degree of
parallelism through simultaneous use of all 104 structures, which is quite useful for screening of high-affinity
antibodies. More than just the hardware is tailored for
high-throughput screening; special software installed in
the Gyrolab Workstation has a specific capability to rank
antibodies in terms of their affinity. This is accomplished
by monitoring the spatial distribution of the bound antigen along the length of the microcolumn; the narrower
the distribution, the greater the likelihood that the antibody is suitable for a diagnostic application.
Materials and Methods
cd structure
The CD structure is an intricate web of microchannels and
microreaction chambers formed in a 1.2-mm thick polymer plate and covered by a lid film. These fluidic components are separated by hydrophobic regions, so-called
hydrophobic barriers, that function to control the flow of
liquids when the CD is subjected to different rotational
speeds. A series of barriers with different characteristics
can be interconnected, and controlling the passage of fluid
across a barrier is simply a matter of changing the
rotational speed of the CD. The hydrophobic barrier plays
a crucial role in volume definition, which is incorporated
into the test structure. Four such structures are shown in
Fig. 2. When fluid is injected into the CD at an individual
inlet through a hole formed in the lid film, a hydrophilic
coating in the inside helps draw the fluid down to the
hydrophobic barrier located at the bottom end of the
volume definition chamber. When the chamber is full, the
CD is rotated at low speed. This leads to overflow of the
excess volume through a channel located at the connec-
3
Nonstandard abbreviations: CD, compact disc; PBS, phosphate-buffered
saline; AFP ␣-fetoprotein; IL-6, interleukin-6; CEA, carcinoembryonic antigen;
and SPR, surface plasmon resonance.
Fig. 2. Expanded view of 4 of the 104 structures incorporated into a
single CD.
Fig. 1. Photograph of the Bioaffy CD, measuring 12 cm in diameter.
A single structure consists of a column that is fed by 2 separate inlets. The
common inlet serves several structures in parallel through a common channel,
whereas the individual inlet serves only a single structure. A fluid dispensed into
either one of the inlets is subjected to metering, guaranteeing reaction of
precisely metered fluids.
1957
Clinical Chemistry 51, No. 10, 2005
Table 1. Raw data for AFP assays from 3 CDs.a
CD 1 (n ⴝ 6)
CD 2 (n ⴝ 6)
CD 3 (n ⴝ 6)
Interassayb
AFP, pmol/L
Mean
SD
CV, %
Mean
SD
CV, %
Mean
SD
CV, %
Mean
SD
CV, %
0
0.5
1
5
50
500
5000
20 000
0.011
0.051
0.124
0.470
4.73
49.5
488
657
0.005
0.009
0.012
0.049
0.244
3.00
12.3
9.31
46
17
9.7
10
5.2
6.1
2.5
1.4
0.009
0.063
0.113
0.506
5.18
50.0
488
649
0.004
0.009
0.019
0.039
0.399
1.76
10.2
26.9
44
15
17
7.8
7.7
3.5
2.1
4.2
0.008
0.057
0.128
0.437
4.58
47.6
474
641
0.004
0.028
0.040
0.034
0.408
2.46
24.3
20.2
48
49
31
7.9
8.9
5.2
5.1
3.2
0.009
0.057
0.121
0.471
4.83
49.0
483
649
0.001
0.006
0.008
0.035
0.312
1.25
8.32
8.4
15
10
6.5
7.3
6.5
2.5
1.7
1.3
a
The number of replicates is 6 for all concentrations. Intraassay data are calculated from 3 interassay data sets from 3 CDs. Mean and SD are fluorescence intensity
values.
b
Data for CDs 1 through 3 combined.
tion between the inlet and the reservoir. The structure
between the hydrophobic barrier and the overflow channel dictates the sample/reagent volume. At increased
rotational speeds, the hydrophobic barrier is designed to
allow passage of the sample/reagent downstream toward
the microcolumn, which is prepackaged with streptavidin-coated particles so that any biotinylated biomolecule
can be attached to the surface, ready for detection of an
analyte. When microcolumns in 8 consecutive structures
are to be exposed to a common fluid, a common inlet can
be used, which is connected through a zigzag-shaped
common channel to 8 structures so that when a fluid is
dispensed into it, all 8 structures can be filled at the same
time. The amount of biomolecule to be dispensed is
dictated by the concentration and the size of the volume
definition area in the common channel. This option is
useful for minimizing the total dispensing time.
(Pierce) was added to a solution containing 1 g/L capture
antibody, and the mixture was incubated for 1 h at room
temperature. The unreacted biotin was then removed by
gel filtration on NAP5TM columns (Amersham Biosciences). The concentration of the capture reagent was
measured after the desalting step.
labeling of antibodies with Alexa 647
Antibodies labeled with Alexa 647 were prepared as
follows: A 1. 5-fold molar excess of Alexa Fluor® 647
carboxylic acid, succinimidyl ester, per mole of antibody
was added to a solution containing 1 g/L detection
antibody, and the mixture was incubated for 1 h at room
temperature. The unreacted Alexa 647 was then removed
by gel filtration on NAP5 columns. The concentration of
the capture reagent was measured after the desalting step.
assay method
software
An interesting feature of the Gyrolab Workstation LIF is a
proprietary software package that can evaluate the signal
spatial distribution within the microcolumn. The Gyrolab
BioaffyTM software package includes two programs: Gyrolab Evaluator and Gyrolab Viewer. Gyrolab Evaluator
performs an automatic evaluation of the raw data produced by the laser-induced fluorescence detector in the
Gyrolab Workstation. Gyrolab Viewer can then be used to
visualize the fluorescence profile on the column and thus
to study binding profiles. At low analyte concentrations,
the signal is observed principally from the column head,
and the affinity of the antibody used can be estimated
from the sharpness of the distribution.
In the Gyrolab Bioaffy CD, each structure has 15 nL of
streptavidin-coated polystyrene beads (DynospheresTM)
as a solid phase to bind the biotinylated capture antibody.
Sandwich immunoassays were carried out according to
the following protocol:
reagents
The antibodies used for evaluation of Gyrolab Bioaffy
were from lots routinely used for the Lumipulse chemiluminescence system (Fujirebio Inc.).
preparation of biotinylated antibodies
Biotinlyted antibodies were prepared as follows: A 12fold molar excess of EZ-LinkTM Sulfo-NHS-LC-Biotin
Fig. 3. Calibration curve for the AFP assays.
Spline approximation was used. Error bars indicate the SD. The arrow indicates
the detection limit.
1958
Honda et al.: Centrifugal Force–Based CD Device for Immunoassays
Table 2. Raw data for IL-6 assays from 3 CDs.
CD 1 (n ⴝ 6)
CD 2 (n ⴝ 6)
CD 3 (n ⴝ 6)
Interassayb
IL-6, pmol/L
Mean
SD
CV, %
Mean
SD
CV, %
Mean
SD
CV, %
Mean
SD
CV, %
0
0.5
1
5
50
500
5000
20 000
0.020
0.024
0.049
0.151
1.40
14.0
157
382
0.009
0.009
0.015
0.024
0.162
0.943
10.7
10.5
44
36
30
16
12
6.7
6.8
2.7
0.012
0.020
0.065
0.160
1.30
13.6
159
372
0.005
0.008
0.035
0.028
0.102
0.777
5.80
8.06
44
42
53
17
7.8
5.7
3.7
2.2
0.023
0.041
0.059
0.166
1.19
13.1
158
362
0.010
0.023
0.021
0.052
0.130
0.656
5.82
5.77
44
56
36
31
11
5.0
3.7
1.6
0.018
0.028
0.058
0.159
1.30
13.6
158
372
0.005
0.011
0.008
0.008
0.104
0.455
0.907
9.80
29
39
14
4.8
8.0
3.4
0.57
2.6
a
The number of replicates is 6 for all concentrations. Intraassay data are calculated from 3 interassay data sets from 3 CDs. Mean and SD are fluorescence intensity
values.
b
Data for CDs 1 through 3 combined.
Each column bed was washed with two 200-nL volumes of washing solution, after which 200 nL of the
biotinylated capture antibody [0.1 g/L in 15 mmol/L
phosphate-buffered saline (PBS) containing 0.1 mL/L
Tween 20] was introduced to each column bed and
bound to the streptavidin-coated beads. Each column
was then washed twice with 200 nL of washing
solution.
Solutions (200 nL) containing various concentrations of
the analyte (0, 0.5, 1, 5, 50, 500, 5000, and 20 000
pmol/L) in 15 mmol/L PBS–10 g/L bovine serum
albumin were introduced to individuals column
beds, followed by washing as above.
The background fluorescence intensity was measured
from the column bed, with the detection sensitivity of
the photomultiplier tube set at 1%, 5%, and 25%.
The Alexa 467–labeled antibody (200 nL of a solution
containing 0.75 mg/L labeled antibody in 15 mmol/L
PBS–10 g/L bovine serum albumin) was introduced
to each column bed, after which the column was
washed twice with 200 nL of washing solution.
The fluorescence intensity was measured at various
points along the column bed with the photomultiplier tube set at 1%, 5%, and 25% sensitivity. The
fluorescence intensity was converted to immunoreaction signals.
as analytes, were then introduced into the individual
sensor chips. The kinetic data were evaluated by local or
global fitting in simultaneous fitting with BIAevaluation
3.0. The kass and kdiss values were obtained with the
evaluator, and KD was calculated from kdiss/kass.
Results
For the anti-AFP and anti-IL-6 antibodies, measurements
were taken with analyte concentrations of 0, 0.5, 1, 5, 50,
500, 5000, and 20 000 pmol/L. For the anti-CEA antibody,
the concentrations were set as 0, 0.5, 1, 5, 50, 500, and 1000
pmol/L. For calculation of the mean concentrations and
intraassay CVs, 6 measurements were taken from a single
CD. For interassay precision, data obtained from 3 different CDs were used for calculation of the mean and CV.
Data for the AFP assays are shown in Table 1. The
intraassay CVs were ⱕ10% for analyte concentrations ⬎5
pmol/L. The interassay CV was ⱕ10% for analyte concentrations up to 0.5 pmol/L. The corresponding calibration curve is shown in Fig. 3; the detection limit, defined
as the signal for the blank ⫹ 3 SD, was 0.15 pmol/L. The
corresponding data for IL-6 are shown in Table 2. The
precision was somewhat poorer; the CV was near or
Reaction times were 60 s for attachment of the capture
antibody, 240 s for interaction of the analyte with the
column, and 250 s for binding of the detection antibody.
Background monitoring and signal detection took 210 s,
making 420 s for the 2 monitoring steps. Automated
dispensing of all 104 structures took ⬃23 min. With time
for washing added, the total assay time was 50 min.
measurement with BIAcore 3000
Each antibody was immobilized to a BIAcore sensor chip
(BIAcore) as a ligand with an optimum solid antibody
density according to the recommended conditions from
BIAcore. Various concentrations of ␣-fetoprotein (AFP),
interleukin-6 (IL-6), and carcinoembryonic antigen (CEA),
Fig. 4. Calibration curve for the IL-6 assays.
Spline approximation was used. Error bars indicate the SD. The arrow indicates
the detection limit.
1959
Clinical Chemistry 51, No. 10, 2005
below 10% only for analyte concentrations ⬎50 pmol/L.
The detection limit, as indicated on the calibration curve
(Fig. 4), was 1.25 pmol/L. The raw data for CEA are
shown in Table 3, and the detection limit and calibration
curve are shown in Fig. 5.
antibody evaluation with Gyrolab viewer
A series of fluorescence images obtained with the Gyrolab
Viewer with AFP as the analyte are shown in Fig. 6. The
x axis is along the flow of the analyte/reagent, and the
AFP concentrations are 0, 1, 5, 500, 5000, and 20 000
pmol/L. For concentrations ⬎5 pmol/L, the profile
broadens along the x axis, whereas there is some variation
across the column width. The profile reflects the locations
at which the capture antibody binds the antigen. The
initial steep slope indicates that as soon as the antigen
encounters the column, it binds immediately, allowing
only a small portion of the antigen to flow downstream.
As the concentration is increased from 5 to 500 pmol/L,
the peak intensity grows from 0.0015 to 0.15 (arbitrary
units), but the overall profile is maintained. When the
concentration is increased up to 5 nmol/L, the head of the
column becomes saturated, and significant binding begins
to be observed in the downstream region as well. It is
expected that the profile reflects the affinity of the capture
antibody. To illustrate this, 3 images from anti-AFP,
anti-IL-6, and anti-CEA antibody assays are shown in the
top half of Fig. 7 for comparison; the analyte concentration
was set relatively low (50 pmol/L) to prevent saturation
of the column. The peak height is ⬃0.01 for all 3 antibodies, but the profile is distinctly different for the anti-CEA
antibody; there is a long, extended downstream tail; the
graphs in the bottom half of Fig. 7 show integrated
fluorescence along the angle direction corresponding to
the images in the top half of Fig. 7. Because the flow rate
was the same for all analytes, one may suspect that either
the kass and/or kdiss for CEA was low. To confirm this, we
show results from the BIAcore experiment in Table 4, in
which the kass and kdiss values for all of the antibodies are
summarized. The original sensorgrams were clean, but
the somewhat high dependence of these kinetic data on
the fitting procedure led us to treat these data semiquan-
Fig. 5. Calibration curve for the CEA assays.
Spline approximation was used. Error bars indicate the SD. The arrow indicates
the detection limit.
titatively. Nonetheless, the difference among the 3 antibodies were more pronounced for their kass values than
for their kdiss values.
Discussion
We evaluated the Gyrolab Bioaffy with antibody sets used
for the Lumipulse System (Fujirebio). Although this is a
preliminary study with a small number of samples, it
demonstrates that the CD format based on centrifugal
force is well suited for handling small sample/reagent
volumes with good precision. The resulting CVs are
respectable; comfortably under 10% for concentrations ⬎5
pmol/L for the antibody set used in this study. The
detection limits, defined as the signals for the blanks ⫹ 3
SD, ranged from 0.15 to 1.31 pmol/L. This was achieved
with a sample volume of only 200 nL. It is clear that if
some increase in volume can be tolerated, further increases in sensitivity are possible. Assays were carried out
in only 50 min. In comparison, the traditional ELISA in a
96-well plate typically takes a few hours, with sample
volumes of a few hundred microliters. The shortening of
the assay time reflects the proximity of the antigen to the
Table 3. Raw data for CEA assays from 3 CDs.a
CD 1 (n ⴝ 6)
CD 2 (n ⴝ 6)
CD 3 (n ⴝ 6)
Interassayb
CEA, pmol/L
Mean
SD
CV, %
Mean
SD
CV, %
Mean
SD
CV, %
Mean
SD
CV, %
0
0.5
1
5
50
500
1000
0.028
0.04
0.062
0.18
4.81
59.4
142
0.22
0.036
0.048
0.040
0.372
3.98
9.21
78
90
78
22
7.7
6.7
6.5
0.050
0.040
0.084
0.245
4.90
61.1
142
0.032
0.028
0.034
0.048
0.691
6.90
11.8
64
70
41
20
14
11
8.3
0.046
0.039
0.140
0.168
4.63
56.55
130
0.038
0.037
0.085
0.069
0.763
7.15
17.3
83
96
60
41
16
13
13
0.041
0.039
0.095
0.198
4.78
59.0
135
0.011
0.0005
0.040
0.042
0.141
2.30
6.99
28
1.3
42
21
2.9
3.9
5.1
a
The number of replicates is 6 for all concentrations. Intraassay data are calculated from 3 interassay data sets from 3 CDs. Mean and SD are fluorescence intensity
values.
b
Data for CDs 1 through 3 combined.
1960
Honda et al.: Centrifugal Force–Based CD Device for Immunoassays
Fig. 6. Fluorescence images from the column region after AFP assays have been carried out.
The AFP concentrations are 0, 1, and 5 pmol/L (top panel, left to right), and 500, 5000, and 20 000 pmol/L (bottom panel, left to right).
surface-bound capture antibody in a micro system as has
been demonstrated by some other microcolumn systems
(7 ).
The power of the fluorescence imaging of the column
structure was also demonstrated by the comparative
study using the BIAcore system as a benchmark. Among
several methodologies allowing measurement of kinetic
data (8 –11 ), the surface plasmon resonance (SPR) system
from BIAcore has become one of the widely accepted
instruments (8, 12 ). It is capable of measuring binding
Fig. 7. Fluorescence images for AFP, IL-6, and CEA assays with a common concentration of 5 pmol/L.
1961
Clinical Chemistry 51, No. 10, 2005
Table 4. Kinetic constants obtained from the BIAcore 3000
instruments for all antibodies.
Capture Aba
Kass, L/mol䡠s
kdiss, sⴚ1
KD, mol/L
AFP
IL-6
CEA
3.81 ⫻ 10
4.82 ⫻ 105
7.73 ⫻ 104
9.07 ⫻ 10⫺5
1.53 ⫻ 10⫺4
1.52 ⫻ 10⫺4
2.38 ⫻ 10⫺11
3.17 ⫻ 10⫺10
1.97 ⫻ 10⫺9
a
6
Ab, antibody.
events in real time without labeling, and from so-called
“sensorgrams” that reflect accumulation of captured biomolecular species on the sensor surface, it is possible to
calculate kass and kdiss values with relative ease. These
numbers are useful for accurate characterization of biomolecules such as antibodies, but it has turned out to be
surprising difficult to implement an SPR instrument capable of highly parallel processing and still giving quantitative data. With this in mind, the ability of the Gyrolab
Workstation to rank antibodies, in a highly parallel fashion if all 104 structures are used simultaneously, is quite
noteworthy. Strict comparison is possible only after the
flow rate and molecular weight of the antigen under
investigation are taken into account, however. A binding
event on a solid surface is influenced by the masstransport limit; the apparent concentration is reduced
when the flow rate is relatively fast compared with the
diffusion rate dictated by the molecular weight. Because
the diffusion constant is inversely proportional to the
third root of the molecular weight, the mass-transport
limit is encountered more readily with larger molecules.
The upstream slope of the fluorescence profile is likely to
be influenced. Further studies aimed at correlating the
profile in the column and the kinetic properties will
hopefully demonstrate the usefulness of the fluorescence
imagining ability of the Gyrolab Workstation.
Even before it becomes possible to fully extract rich
information from fluorescence images, the highly parallel
platform should be useful. In the current world of multianalyte analysis, the number of target molecule species
has increased drastically from early days (13–16 ). The
explosive growth in parallelism has unfortunately not
been accompanied by ease of use. The Gyrolab system,
however, exploits centrifugal force to truly carry out
assays in parallel under good control while taking care of
the sample/reagent dispensing function with its station.
We thank Y. Kawase for fruitful discussion and D. Miwa
for assistance with the BIAcore experiment.
References
1. Manz A, Fettinger JC, Verpoorte E, Ludi H, Widmer HM, Harrison
DJ. Micromachining of monocrystalline silicon and glass for chemical analysis systems. A look into next century’s technology or just
a fashionable craze? Trends Analyt Chem 1991;10:144 –9.
2. Shi Y, Simpson PC, Scherer JR, Wexler D, Skibola C, Smith MT, et
al. Radial capillary array electrophoresis microplate and scanner
for high-performance nucleic acid analysis. Anal Chem 1999;71:
5354 – 61.
3. Inganäs M, Ekstrand G, Engström, J. Eckersten A, Derand H,
Andersson P. Quantitative bio-affinity assays at nanoliter scale,
parallel analysis of crude protein mixtures. In: Ramsey JM, van
den Berg A, eds. Micro total analysis systems. Dordrecht: Kluwer
Academic, 2001:91pp.
4. Scott CD. A miniature fast analyzer system. Clin Chem 1973;45:
327A– 40A.
5. Puckett LG, Dikici E, Lai S, Madou M, Bachas LG, Daunert S.
Investigation into the applicability of the centrifugal microfluidics
platform for the development of protein-ligand binding assays
incorporating enhanced green fluorescent protein as a fluorescent
reporter. Anal Chem 2004;76:7263– 8.
6. Duffy DC, Gillis HL, Lin J, Norman F, Sheppard J, Kellogg GJ.
Microfabricated centrifugal microfluidic systems: Characterization
and multiple enzymatic assays. Anal Chem 1999;71:4669 –78.
7. Sato K, Tokeshi M, Odake T, Kimura H, Ooi T, Nakao M, et al.
Integration of an immunosorbent assay system: Analysis of secretory human immunoglobulin A on polystyrene beads in a microchip. Anal Chem 2000;72:1144 –7.
8. Jönsson U, Fägerstam L, Ivarsson B, Johnsson B, Karlsson R,
Lundh K, et al. Real-time biospecific interaction analysis using
surface plasmon resonance and a sensor chip technology. Biotechniques 1991;11:620 –7.
9. Kanasawa KK, Gordon JG. A liquid phase piezoelectric detector.
Anal Chem 1985;57:1771–5.
10. Shons A, Dorman F, Najarian J. The piezoelectric quartz immunosensor. J Biomed Mater Res 1972;6:565–70.
11. Cush R, Cronin JM, Stewart WJ, Maule CH, Molloy J, Goddard NJ.
The resonant mirror; a novel optical biosensor for direct sensing of
biomolecular interactions. Biosens Bioelectron 1993;8:347–53.
12. Myszka DG. Kinetic analysis of macromolecular interactions using
surface plasmon resonance biosensors. Curr Opin Biotechnol
1997;8:50 –7.
13. Yamamoto K, Higashimoto K, Minagawa H, Okada M, Kasahara Y.
Simultaneous detection of antibodies to HTLV-1 and HIV-1 by
chemiluminescent enzyme immunoassay [Abstract]. Clin Chem
1991;37(Suppl):1031.
14. Brown CR, Higgins KW, Frazer K, Schoelz LK, Dyminski JW,
Marinkovich VA, et al. Simultaneous determination of total IgE and
allergen-specific IgE in serum by the MAST chemiluminescent
assay system. Clin Chem 1985;31:1500 –5.
15. Fodor SP, Read JL, Pirrung MC. Light-directed, spatially addressable parallel chemical synthesis. Science 1991;251:767–73.
16. Walt DR. Fiber optic imaging sensors. Acc Chem Res 1997;5:
267–78.