Electrophoresis 2002, 23, 1531–1536
1531
Jian Liu1
Markus Enzelberger2*
Stephen Quake2
A nanoliter rotary device for polymerase
chain reaction
1
Polymerase chain reaction (PCR) has revolutionized a variety of assays in biotechnology. The ability to implement PCR in disposable and reliable microfluidic chips will
facilitate its use in applications such as rapid medical diagnostics, food control testing,
and biological weapons detection. We fabricated a microfluidic chip with integrated
heaters and plumbing in which various forms of PCR have been successfully demonstrated. The device uses only 12 nL of sample, one of the smallest sample volumes
demonstrated to date. Minimizing the sample volume allows low power consumption,
reduced reagent costs, and ultimately more rapid thermal cycling.
Department of Chemistry
Department of Applied
Physics, California Institute
of Technology,
Pasadena, CA, USA
2
Keywords: Microfluidic device / Nucleic acid amplification / Polymerase chain reaction / Soft
lithography
EL 4883
Nucleic acid amplification reactions such as PCR have
emerged as powerful tools in a variety of genetic analyses, medical diagnostics, and forensic applications.
These techniques are able to dramatically increase the
concentration of target nucleic acids of interest, which
are initially present at low levels in most cases. For example, a single target DNA molecule can be amplified up to
106 by PCR [1]. PCR involves the repetition of heating
(denaturation) and cooling (annealing) cycles in the presence of a target DNA, two primers, deoxynucleotides, a
polymerase and various cofactors. With the advent of
the notion of “lab-on-a-chip” technology, the idea of making integrated microfluidic PCR chips has attracted the
interest of many researchers [2–8]. Since the cost of PCR
is dominated by the amount of reagents used, miniaturized devices are expected to dramatically reduce the
cost for high-throughput applications. They are also
expected to be easily portable, making them useful for
various field testing situations ranging from crime scene
forensics to biological weapons detection. Finally, since
PCR is exquisitely sensitive to cross-contamination,
“lab-on-a-chip” devices provide a way to resolve this
issue by performing all required analyses in a sealed, disposable device.
There are four key figures of merit that are important in the
evaluation of a microfabricated PCR device. They are
(i) time to perform the reaction, (ii) power consumption of
the chip, (iii) volume of sample used, and (iv) ability to be
integrated with other biological manipulations. Conventional benchtop PCR machines require 1–2 h per reaction,
Correspondence: Dr. Stephen Quake, Caltech MS 128-95,
Pasadena, CA 91125, USA
E-mail: [email protected]
Fax: 1626-793-8675
ª WILEY-VCH Verlag GmbH, 69451 Weinheim, 2002
while miniaturized devices have demonstrated reaction
times ranging from 7 to 40 min. Since the process time is
often limited by the rate at which one can thermally cycle
the reaction, much effort has been placed in reducing the
sample volume in order to shorten cycling time. Power
consumption is a key issue for portable devices and can
also be decreased by reducing the sample volumes.
Sample volume is also an important parameter in determining cost per reaction. Finally, in many cases it will not
be possible to reap the desired economies of scale without a scaleable technology that will allow integration of
PCR with other manipulations such as reagent dispensing
and subsequent analysis. Thus, it is clear that many of the
important parameters for practical microfabricated PCR
devices depend either directly or indirectly on one’s ability
to reduce the sample volumes involved.
A number of microfabricated [2–8] and other miniaturized
PCR devices [9, 10] have been demonstrated to date.
Most have large (.1 mL) sample volumes, although samples as small as 240 nL have been amplified [3]. There are
two main classes of methods to conduct on-chip PCR.
The most common approach is a time domain one [4–8],
which is a direct miniaturization of benchtop methods.
The reaction mixture is kept stationary while the temperature of the reaction chamber is cycled. This method
requires fairly careful optimization of thermal mass in
order to optimize reaction times and power consumption.
The alternative is a space domain approach in which the
device itself is not thermocycled, but the reaction mixture
is moved in a flow channel through three different fixed
temperature zones [2, 3, 9, 10]. This method can be conducted at relatively high speed because it is not necessary to heat and cool the device repeatedly, but it has
required the use of relatively large sample volumes [9, 10].
* Current address: Institute of Technical Biochemistry, Stuttgart,
Germany
0173-0835/02/1005–1531 $17.50+.50/0
Miniaturization
1 Introduction
1532
J. Liu et al.
Electrophoresis 2002, 23, 1531–1536
Here, we describe a microfluidic PCR device that is capable of performing both temporally and spatially cycled
PCR with sample volumes as small as 12 nL. The fluidic
device is fabricated out of silicone elastomer, with integrated valves and pumps [11]. These plumbing components are compatible ([12]; Fu et al., submitted) with a
high degree of integration, as shown by their use in microfabricated cell sorters and other integrated devices [13,
14]. In future devices, these plumbing components will
be useful for metering small amounts of reagents. A key
aspect of the present system is that the fluidic channels
are arranged in a rotary geometry with an inlet and outlet,
which allows compact spatial cycling with no sample loss
or dilution during the amplification process. We previously
showed that such a rotary pump can be used to accelerate chemical kinetics and for efficient microfluidic mixing
[13]. Alternatively, in the temporal cycling scheme the
reaction loop can be sealed off with the valves and the
temperature of the entire loop is cycled. The elastomeric
chip is mounted on a glass slide with microfabricated
tungsten heaters. We have demonstrated nucleic acid
amplification via both the Taqman PCR (2-step) and standard PCR (3-step) in these chips with process time as low
as about 30 min. The elastomeric chips are disposable
because they are significantly cheaper to mass produce
than glass or silicon devices.
2 Materials and methods
2.1 Fabrication and control of rotary
microfluidic device
The rotary microfluidic devices were fabricated by multilayer soft lithography techniques [11, 13, 14]. The channel
pattern of the device is shown in Fig 1. The masks were
printed on transparency films by a high-resolution printer
(3556 dpi). Molds were prepared by spincoating photoresist
(Shipley SJR5740) onto silicon wafers (3000 rpm for 60 s)
which had been pretreated with hexamethyldisilazane
(HDMS). The wafers were then heated at 907C for 1 h, after
which the channel patterns were exposed onto them by
ultraviolet light with a mask aligner (Karl Suss). The patterns
were developed by 5:1 v/v developing solution (Shipley
2401). Then molds were put onto a hot plate (557C) for
30 min to round the remaining photoresist. The devices
were made from GE RTV 615, a two-component poly(dimethylsiloxane) elastomer. A 20:1 mixture of RTV was spincoated on the fluid channel mold at 2000 rpm, resulting in
a thin layer of 30–40 mm. A 5:1 mixture of RTV was then spincoated on the control channel mold, after which 10:1 RTV
was poured onto the top of it to improve the mechanical
properties. After baking both molds in an oven of 807C
Figure 1. (a) Schematic of the PDMS rotary microchip
and the heaters (top view). The control channels appear
in grey; the fluid channel in black; heaters in grey with fill
pattern of upward diagonal. Three heaters arranged in the
following order (clockwise from upper right): denaturation,
annealing, and extension. (In the 2-step Taqman PCR,
only the two heaters for denaturation and extension were
used.) (b) Assembly of the rotary microchip and the
heaters. From the top to the bottom: RTV block with
control channels at the bottom surface; thin RTV layer
with fluid channel at the bottom surface; glass coverslip;
heaters with leads (sputtered onto a glass slide). Arrows
indicate the air and fluid through-holes.
for 1.5 h, the RTV block on the control channel mold was
peeled off to punch air supply through-holes. Then it was
aligned and pressed against the thin RTV layer on the fluid
channel mold. After a post-bake at 807C for 2 h to bond
Electrophoresis 2002, 23, 1531–1536
the two layers, the multilayer RTV chip was peeled off the
mold to punch fluid through-holes. Finally, the device was
completed by sealing the multilayer RTV chip onto a glass
coverslip at 807C for 2 h.
The on-chip valves were actuated with a pneumatic controller supplied by Fluidigm. The Fluid Controller system
allows selective actuation of valves to seal off the loop as
well as peristaltic pumping at variable rates within the
loop. Typical actuation pressures were 10 psi; the pumping frequency was 20 Hz. During operation of the rotary
pump, the sample liquid is transported around the loop
at an average rate of 2–3 revolutions per minute. In the
chip’s control layer, we initially used three separate
valves to peristaltically pump the fluid. Then we found
that one S-shaped channel could accomplish the same
peristaltic pumping as three separate channels (Liu et al.,
to be published). In addition, the S-shaped channels provide more secure closing of the inlet and outlet than a
single channel.
2.2 Fabrication of heaters and temperature
characterization
The pattern of the heaters is shown in Fig 1. They were made
of thin layers of tungsten (heating component) and aluminum (electrical leads) sputtered onto a glass microscope
slide. Similar to the fluidic fabrication, a negative pattern of
photoresist was developed onto a piece of a glass microscope slide (40 mm626 mm61.1 mm). Then a thin layer
(,500 Å) of tungsten was deposited on the sample by a
direct current (DC) sputter system. Acetone was used to
remove the remaining photoresist in order to get the
designed tungsten pattern. The procedure was repeated in
order to sputter another thin layer of aluminum as the electrical leads for the heater. The heaters were calibrated via a
two step process. After the fluidic device was attached to
the heaters, thermochromic liquid crystals (BM/C17–10,
Hallcrest Inc.) were loaded into the fluid channel of the
device. Then TLC were imaged at various heater currents
with a color camera (KR222; Panasonic) and a stereomicroscope (ASA012–3449, Japan). The temperature response
of the TLC was separately calibrated with a thermistor and
hot plate. Images of the TLC particles were digitized and
average hues of them were calculated, with outliers discarded [15, 16]. Hue and temperature have a fairly linear
relationship from 75–100 degrees, sufficient to determine
the denaturing temperature. The lower annealing/extension
temperature was calibrated by the thermistor. During Taqman PCR, the denaturing temperature was set at 957C,
and the annealing temperature at 607C. Two independent
resistive heaters were used to heat the microchip to set temperatures. The power consumed by the heaters was around
380 and 75 mW, respectively.
A nL rotary device for PCR
1533
2.3 Two-step Taqman PCR
The Taqman PCR reagent Kit uses the 5’-3’ nuclease activity of AmpliTaq Gold DNA polymerase to allow direct detection of PCR product by the release of a fluorescent reporter
during PCR [17]. The probe is an oligonucleotide containing
a reporter dye, 6-carboxyfluorescein (FAM) at the 5’-end
and a quencher dye, 6-carboxytetramethylrhodamine
(TAMRA), at the 3’-end. The excitation wavelength is 488
nm, while the emission wavelength of the reporter dye and
quencher dye are 518 nm and 582 nm, respectively. The
probe is cleaved between the reporter and the quencher
by the 5’-3’ nucleolytic activity of the AmpliTaq Gold only
when the probe is hybridized to the target DNA template.
Therefore, the increase in fluorescence during PCR quantitatively reflects the amount of product created [17]. The
DNA template, Human Male genomic DNA, was provided
with the kit (Applied Biosystems, Foster City, CA, USA).
The target was a portion of the b-actin gene; forward primer
and reverse primer were the following: 5’-TCA CCC ACA
CTG TGC CCA TCT ACG A-3’ and 5-’CAG CGG AAC CGC
TCA TTG CCA ATG G-3’, respectively. A segment of DNA
about 294 bp was amplified. Typical benchtop PCR reagent
conditions were as follows (50 mL volumes): template, 0.2
ng/mL; primers, 300 nM; Mg21, 3.5 mM; dATP, 200 mM; dCTP,
200 mM; dGTP, 200 mM; dUTP, 400 mM; AmpliTaq Gold, 0025
U/mL; AmpErase UNG, 0.01 U/mL; Probe, 5’-(FAM)ATG
CCC-X(TAMRA)-CCC CCA TGC CA CCT GCG Tp-3’, 200
nM. A benchtop PCR machine (PTC200 Peltier Thermal
Cycler; MJ Research, Watertown, MA, USA) was used to
verify the reaction by the protocol [17]: 2 min at 507C; 10
min at 957C; 40 cycles: 15 s at 957C; 1 min at 607C.
Figures 2 and 3a show the emission spectra and gel electrophoresis results of the PCR product by this procedure.
2.4 Three-step PCR
A segment (199 bp) of l-phage DNA was selected for
amplification; forward primer and reverse primer were 5’GGT TAT CGA AAT CAG CCA CAG CGC C-3’ and 5’-GGA
TAC GTC TGA ACT GGT CAC-3’, respectively [8]. Typical
benchtop PCR reagent conditions were as follows (100 mL
volumes): primers (Operon), 500 nM; template, 2 ng/mL;
Mg21, 1.5 mM; dNTP, 200 mM; Taq polymerase, 0025 U/mL
(Qiagen, Valencia, CA, USA). An intercalating dye, Sybr
Green I (Molecular Probes, Eugene, OR, USA), was added
to the reagent mixture using a 1:10 000 dilution of the
stock reagent. For amplification by benchtop PCR
machine, the temperature steps were set as follows:
2 min at 947C; 30 cycles: 30 s at 947C; 30 s at 557C; 1 min
at 727C; and 10 min at 727C finally. Figure 3b shows the
gel electrophoresis result of the PCR product of l-phage
DNA by this method.
1534
J. Liu et al.
Electrophoresis 2002, 23, 1531–1536
Figure 2. The emission spectra
of the sample (before and after)
and the negative control (no template). – r –: sample; – x –: no
template control. - - -: Sample
before PCR. The Taqman PCR
was performed in a benchtop
PCR machine (MJ Research).
Note how the fluorescent peak
at 518 nm appears in the sample
after PCR, while there is virtually
no peak from the negative control or the sample before PCR.
The cycling program was described in Section 2.3: 2 min at
507C; 10 min at 957C; 40 cycles:
15 s at 957C; 1 min at 607C.
Figure 3. Gel electrophoresis analysis of PCR products.
(a) Taqman PCR (human b-actin). Lane 1, 123 bp DNA
ladder; lane 3, sample; lane 5, negative control (invisible).
(b) 3-step PCR (l-DNA). Lane 1, 123 bp DNA ladder;
lane 2, l-DNA amplicon; lane 3, negative control (invisible). The sample and the control displayed in 2% agarose gel with easy-cast electrophoresis system (Model
B1A,VWR). All samples were stained with Sybr Green I
stain. The pictures were taken using the Kodak electrophoresis documentation and analysis system.
2.5 Real-time measurement of PCR products
After injecting the reagent mixture into the microchip, we
closed the inlet and outlet valves to the loop by applying
pressure to the control channels above them. We then
actuated the rotary pump in order to circulate the solution
around the loop [13, 14]. The power supply was turned on
to provide direct current for the heaters, and the objective
(206, NA 0.5) of the fluorescence microscope (IX50;
Olympus) was focused on a segment of the fluid channel
which was not shaded by the heaters. The filters (Chroma)
in the fluorescence microscope were chosen as follows:
exciter, HQ 470/406dichroic, Q 495LP; emitter, HQ 525/
50 m. A cooled CCD camera (ST-7; Santa Barbara Instruments Group) was used to acquire pictures of fluorescence in the channels every few minutes during the PCR.
The images were analyzed with the CCDOPS (SBIG) software in the following way: in each picture, three squares
(25 pixels on a side) inside the fluid channel were chosen
to get an average value of the fluorescence intensity.
Similarly, three squares outside the channel were
selected to make another average value as the background of this picture. The latter value was subtracted
from the former to obtain the fluorescence of the solution
inside the fluid channel.
3 Results and discussion
3.1 General
As shown schematically in Fig. 1, the fluid channel is a central loop with an inlet and an outlet; fluid can be circulated in
the loop with the integrated pump. We have previously
shown applications of this rotary pump in mixing and surface binding assays [13]. The dimensions of the central
loop are a radius of 2.5 mm and height of 8.5 mm. The width
of the loop channel is asymmetric: 120 mm on the left semicircle; 70 mm on the right semicircle. In such a pattern the
fluid velocities in the two parts are different from each other.
Therefore, the time that the fluid resides in different temperature zones can be adjusted either by the width of flow
channel or the length of the heater. The fluid motion in the
Electrophoresis 2002, 23, 1531–1536
channel was characterized by using a solution of 2.5 mm
beads. In order to mimic the actual conditions of the experiments, we used the same concentration of Taqman PCR
buffer and human genomic DNA in the bead solution.
At room temperature, the speed of beads was about
3.4 mm/s in the narrow channel and 1.5 mm/s in the wide
channel. However, the velocity was slowed down to about
25–30% of the above velocity when the device was run at its
operating temperatures, resulting in about 20–30 s per full
cycle. The reason for the decrease in velocity could be thermal distortion of the RTV or small bubbles in the solution.
Presently the minimum sample volume required to be
loaded is mainly limited by the amount that can be loaded
on the chip using conventional macroscopic tools, i.e., by
pipetting, and is of the order of 1–2 mL. However, the
actual PCR reaction in the loop uses only about 12 nL of
the sample solution, excluding the dead volume of 2 nL
caused by the inlet and outlet. Future devices will integrate pumps and valves to dispense reagents on chip
into parallel reactions, thus making optimal use of the
most expensive reagents.
A nL rotary device for PCR
1535
more than fourfold increase in fluorescence intensity in
the present of the sample template, while the no-template
control remained flat. This is consistent with the features
of macroscopic real-time PCR [18]. Calibration experiments showed that the recorded increase in fluorescence
was comparable to that of 30–35 cycles of PCR product
by a benchtop PCR machine (PTC200 Peltier Thermal
Cycler; MJ Research). We then used our microchips to
perform a three-step PCR reaction in which we directly
monitored the creation of product by observing the product via an intercalating dye. A fragment of l-DNA was
amplified by Taq Polymerase while the PCR product was
detected with Sybr Green I. Figure 5 (top) shows the fluorescence difference before and after PCR and indicates
that amplification occurs relative to the no-template control.
3.2 Spatial cycling
We first used the chips in the spatial cycling mode to
amplify a segment of the human b-actin gene using Taqman PCR. As the reagent mixture was circulated in the
loop channel for about 10 min, a rapid increase in fluorescence was observed. As shown in Fig. 4, there was a
Figure 4. Fluorescence increase in spatially cycled Taqman assay performed on the rotary-chip. A segment of
the human b-actin gene was amplified from purified
human genomic DNA; fluorescence was monitored in
the channel during the amplification process. (d) Sample;
(j) no template control. The curves are a guide for the
eye.
Figure 5. Three-step PCR. A segment of l DNA was
amplified; the creation of product was monitored directly
with the use of the intercalation dye Sybr Green 1. Top:
spatial cycling. The reagents were pumped through three
different temperature zones: denaturing (947C), annealing
(557C), and extension (727C). The graph shows fluorescence measured in the channel before and after 40 min
of pumping. Dark bar: PCR process with DNA template.
White bar: control without template. Bottom: temporal
cycling. The entire loop chamber was heated to the
designed temperature simultaneously. This temperature
was manually ramped between the same extension,
denaturing, and annealing temperatures as in the spatially
cycled case.
1536
J. Liu et al.
3.3 Temporal cycling
We also amplified the same segment of l-phage DNA in
the chip using the more conventional temporally-cycled
mode. After having injected the reagent mixture into the
microchip, we did not pump it around the loop; instead,
we sealed off the inlet and outlet with the control channels, then thermocycled the entire device. The heaters
were adjusted to identical temperatures so that the contents of the device were uniformly heated; the temperature steps were as follows: 15 s at 947C; 30 s at 557C;
30 s at 727C. Each cycle took about 6 min and 23 cycles
were conducted in the experiment. The change of fluorescence during PCR was monitored by the same optical
setup every few cycles. The fluorescence increased dramatically during the experiment, indicating the amplification of DNA template (Fig. 5, bottom).
4 Concluding remarks
A rotary microfluidic device has successfully been developed to run the PCR in both spatially and temporally
cycled formats. The microchip requires only 12 nL of
reagent mixture, and the microfabricated valves and
pumps should allow a high degree of future integration.
In addition, the polydimethylsiloxane (PDMS) microchips
are disposable, reducing the chances of cross-contamination during the reaction procedure. Finally, we wish to
point out a possible future application enabled by the
unique geometry of the rotary device. Since it is possible
to modify the surface chemistry of the channels, and
indeed to specifically attach proteins, one could conceivably immobilize DNA polymerase [19] in the extension
region of the chip. Then in the spatially cycling mode, the
enzyme is kept at a constant temperature and in particular is not exposed to the denaturing temperature. This is
significant because a tremendous amount of effort has
been spent on developing DNA polymerases that are
both thermostable and of high fidelity, at the expense of
other desired qualities, such as turnover rate. By relaxing
the thermostability requirement, one could perform efficient PCR with enzymes from nonthermophilic organisms
that have been optimized for other parameters, such as
fidelity and speed. Thus, this particular format of the
PCR may allow a number of enzymological advances to
play a role in future nucleic acid amplification technology.
Electrophoresis 2002, 23, 1531–1536
We wish to thank Mark Adams and Prof. Axel Scherer for
assistance with the heater fabrication; Tait Pottebaum, Dr.
Dana Dabiri and Prof. Morteza Gharib for their assistance
and discussion on TLC issue; Fluidigm Co. for their controller box, manifold, and sample microchips. This work
was supported by the NSF, ARO, and DARPA (BioFlips
program).
Received October 30, 2001
5 References
[1] Erlich, H. A., PCR Technology, Stockton Press, New York
1989, pp.1–5.
[2] Kopp, M. U., de Mello, A. J., Manz, A., Science 1998, 280,
1046–1048.
[3] Burns, M. A., Johnson, B. N., Brahmansandra, S. N., Handique, K., Webster, J. R., Krishman, M., Sammarco, T. S., Man,
P. M., Jones, D., Heldsinger, D., Mastrangelo, C. H., Burke,
D. T., Science 1998, 282, 484–487.
[4] Woolley, A. T., Hadley, D., Landre, P., deMello, A. J., Mathies,
R. A., Northrup, M. A., Anal. Chem. 1996, 68, 4081–4086.
[5] Hong, J. W., Fujii, T., Seki, M., Electrophoresis 2001, 22,
328–333.
[6] Legally, E. T., Mathies, R. A., Anal. Chem. 2001, 73, 565–570.
[7] Belgrader, P., Benett, W., Hadley, D., Richards, J., Stratton,
P., Mariella, R., Milanovich, F., Science 1999, 284, 449–450.
[8] Khandurina, J., Mcknight, T. E., Jacobson, S. C., Waters,
L. C., Foote, R. S., Ramsey, J. M., Anal. Chem. 2000, 72,
2995–3000.
[9] Chiou, J., Matsudaira, P., Sonn, A., Ehrlich, D., Anal. Chem.
2001, 73, 2018–2021.
[10] Nakano, H., Matsuda, K., Yohda, M., Nagamune, T., Endo, I.,
Yamane, T., Biosci. Biotechnol. Biochem. 1994, 58, 349–352.
[11] Unger, M. A., Chou, H. P., Thorsen, T., Scherer, A., Quake,
S. R., Science 2000, 288, 113–116.
[12] Quake, S. R., Scherer, A., Science 2000, 290, 1536–1540.
[13] Chou, H. P., Unger, M. A., Scherer, A., Quake, S. R., Biomed.
Microdevices 2001, 3, 323–330.
[14] Chou, H. P., Unger, M. A., Scherer, A., Quake, S. R., Proceedings of the Solid State Actuator and Sensor Workshop,
Hilton Head, S.C. 2000.
[15] Dabiri, D., Gharib, M., Experim. Fluids 1991, 11, 77–86.
[16] Chaudhari, A. M., Woudenberg, T. M., Albin, M., Goodson
K., E., J. Microelectromech. Syst. 1998, 7, 345–355.
[17] Protocol of TaqMan PCR Reagent Kit with AmpliTaq Gold
DNA Polymerase, 2001, Applied Biosystems. http://
www.appliedbiosystems.com.
[18] Ibrahim, M. S., Lofts, R. S., Henchal, E. A., Jahrling, P.,
Weedn, V. W., Northrup, M. A., Belgrader, P., Anal. Chem.
1998, 70, 2013–2017.
[19] Gordon, F. B., Immobilization of Enzyme and Cells, Humana
Press, Totowa, NJ 1997.