You are on page 1of 4

Chemical Amplification: Continuous-Flow PCR on a

Martin U. Kopp, et al.
Science 280, 1046 (1998);
DOI: 10.1126/science.280.5366.1046

The following resources related to this article are available online at (this information is current as of November 14, 2007 ):

Updated information and services, including high-resolution figures, can be found in the online
version of this article at:

Downloaded from on November 14, 2007

This article has been cited by 435 article(s) on the ISI Web of Science.

This article has been cited by 19 articles hosted by HighWire Press; see:

This article appears in the following subject collections:


Information about obtaining reprints of this article or about obtaining permission to reproduce
this article in whole or in part can be found at:

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
1998 by the American Association for the Advancement of Science; all rights reserved. The title Science is a
registered trademark of AAAS.
performing PCR in a continuous flow at
Chemical Amplification: high speed. The results demonstrate the
Continuous-Flow PCR on a Chip concept of a chemical amplifier for DNA.
The speed of thermal cycling is usually
Martin U. Kopp, Andrew J. de Mello, Andreas Manz* instrument limited, except for a commercial
system that uses an air stream to heat and
A micromachined chemical amplifier was successfully used to perform the polymerase cool sealed glass capillaries containing the
chain reaction (PCR) in continuous flow at high speed. The device is analogous to an PCR mixture; this system has demonstrated
electronic amplifier and relies on the movement of sample through thermostated tem- high thermal cycling speeds and efficient am-
perature zones on a glass microchip. Input and output of material (DNA) is continuous, plification (2). More recently, several groups
and amplification is independent of input concentration. A 20-cycle PCR amplification have reported high cycling speeds for PCR
of a 176 – base pair fragment from the DNA gyrase gene of Neisseria gonorrhoeae was and the ligase chain reaction (LCR) with
performed at various flow rates, resulting in total reaction times of 90 seconds to 18.7 various designs of micromachined heating
minutes. chambers (3, 4). Micromachining can be
defined as the patterning of silicon and its
derivatives to create three-dimensional mi-
crostructures. A wide range of microreactors,
Electronic amplifiers allow weak signals to capillary, or slide containing the reagent mix- microcapillary electrophoresis devices, and
be increased by a large constant factor with ture. The product is then analyzed by an microcell manipulation devices have been
the same time dependency and virtually no endpoint measurement or directly used for described in recent years (5).

Downloaded from on November 14, 2007

time delay. Although some biochemical sys- cloning and sequencing. A continuous-flow PCR system can be
tems, such as hormonal signaling, lead to PCR represents a perfect model reaction realized by a time-space conversion in the
large amplifications, they often do so non- for a chemical amplifier. Conventional PCR system—that is, by keeping tempera-
linearly and with time delays. A true chem- thermal cyclers are based on batch processes tures constant over time at different loca-
ical amplifier, which would continuously and therefore do not represent a chemical tions in the system and moving the sample
amplify the concentration of a compound amplifier in the previously described sense. through the individual temperature zones
independent of the input concentration, Here, we present a microfabricated device (Fig. 1A). The time delay for the sample to
could be extremely useful in analysis and
process control. In particular, a capillary-
shaped chemical reactor with its inherent Buffer Sample
low-dispersion characteristics could main- A B
tain the shape of analyte peaks injected
sequentially, yielding the amplified product
in the same volume element. Unlike elec-
tronic amplification, additional molecules
must be synthesized from building blocks or C Product
transformed from precursor species. This
process involves both chemical reaction ki-

netics and diffusion-limited mass transport,

and will lead to more significant time delays Input Output
than those in an electronic amplifier.
This chemical amplifier concept could be

applied to a number of known reactions, such

A 95°C - melting
as self-activating enzymatic reactions, cyclic
B 77°C - extension
electrochemical reactions, and polymeriza-
C 60°C - annealing
tions. We have focused on PCR (1), which

doubles the number of specific DNA mole- Fig. 1. Chip layout. (A) Schematic of a chip for flow-
cules during each cycle of melting of the through PCR. Three well-defined zones are kept at
double-stranded DNA (dsDNA), binding of 95°, 77°, and 60°C by means of thermostated copper
the specific primers to their target sites (an- blocks. The sample is hydrostatically pumped through 10 mm
a single channel etched into the glass chip. The chan-
nealing), and extending of the primers with nel passing through the three temperature zones defines the thermal cycling process. (B) Layout of the
the thermostable DNA polymerase (such as device used in this study. The device has three inlets on the left side of the device and one outlet to the right.
Taq). The individual steps are simply per- Only two inlets are used: one carrying the sample, the other bringing a constant buffer flow. The whole chip
formed by heating or cooling the sample to incorporates 20 identical cycles, except that the first one includes a threefold increase in DNA melting time.
characteristic temperatures: 95°C for dsDNA The chip was fabricated in Corning 0211 glass at the Alberta Microelectronic Centre, Canada. All channels
melting, 50° to 65°C for primer annealing, are 40 mm deep and 90 mm wide; the etched glass chip and the cover plate are each 0.55 mm thick.
and 72° to 77°C for primer extension at the Access to the channels is provided by holes (400 mm) drilled into the cover plate. Standard fused-silica
optimum enzyme temperature. Convention- capillaries (outside diameter 375 mm, inside diameter 100 mm) are glued with epoxy into the holes of the
ally, this process is performed with a program- chip. Virtually no dead volume is introduced by this connection. Two precision syringe pumps (Kloehn
50300, 25 ml) deliver the PCR sample and the buffer solution onto the chip. The pumps are controlled by
mable instrument that heats and cools a tube,
a program written in Labview running on a PC. Product is collected at the outlet capillary and then analyzed
Zeneca/SmithKline Beecham Centre for Analytical Sci- by slab-gel electrophoresis. The copper blocks are heated by 5-W heating cartridges, and the surface
ences, Department of Chemistry, Imperial College of Sci- temperature is monitored by a Pt100 thin-film resistor mounted on the surface of the block near the chip
ence, Technology and Medicine, London SW7 2AY, UK. contact area. Cooling fins passively cool the two blocks at 77° and 60°C. The temperature controllers are
*To whom correspondence should be addressed. E-mail: built with standard PID (proportional, integral, and derivative) digital temperature controllers (CAL 3200), power supplies, and switching electronics for the heating cartridges.

1046 SCIENCE z VOL. 280 z 15 MAY 1998 z

reach a new temperature depends only on The channel walls in the chip were si- prising because the location of high DNA
the times needed to transport the sample lanized with dichlorodimethylsilane to re- concentration (the last few cycles of amplifi-
into the appropriate temperature zone to duce adsorption of enzyme and DNA to the cation) and the location of the first few crit-
heat a fluid element in the capillary. Accord- glass surface; a zwitterionic buffer and a non- ical amplification cycles were spatially sepa-
ing to Fick’s law (6), the time needed for ionic surfactant were used as the PCR buffer rated. We suggest that cross-contamination
heat dissipation is directly proportional to to impart a dynamic coating (7). Primers in a continuous-flow format may be much
the second power of the channel depth for a were used that defined a 176– base pair (bp) less of a problem than in a stationary-tube
flat rectangular channel, assuming that the product from the quinolone resistance– deter- PCR format. Thus, a flow-through PCR chip
thermostated copper blocks and chip repre- mining region of the Neisseria gonorrhoeae can be reused for extended periods without
sent an infinite heat capacity relative to the DNA gyrase gene (gyrA); the template used major cleaning. All runs in this study were
heated fluid element (see Fig. 1B for instru- was a 1-kbp PCR fragment from the same done on the same chip by alternately inject-
ment layout). Simple calculations have dem- gene. The chip was run with flow rates rang- ing volumes of sample and buffer at a ratio of
onstrated that, in our micromachined de- ing from 5.8 to 72.9 nl/s, which represents a 1: 0.8.
vice, heating and cooling times are each less total flow-through time of 18.7 to 1.5 min for Very small samples can be run on the
than 100 ms. 20 cycles. For each amplification, 10 ml of system. The only constraint is that the sam-
The flow-through PCR system is based on PCR mixture was injected and collected at ple plug must be large enough for accurate
a single channel passing repetitively through the outlet of the chip for analysis. As a injection into the channel, and this can be
the three temperature zones (Fig. 1). The reference, the PCR mixture (10 ml) was run accomplished with sample plug volumes on
pattern of the chip layout determines the on a fast commercial thermal cycler (Hybaid, the order of a few nanoliters. Because mul-
relative time a fluid element is exposed to PCR-Express) with an overall cycling time of tiple sample plugs can travel at the same

Downloaded from on November 14, 2007

each temperature zone. Moreover, the pat- 50 min for 20 cycles. The collected samples speed through the same chip, the potential
tern defines the number of cycles n per- were analyzed by an 8% (29 :1) polyacryl- throughput of a single device can be greatly
formed per run through the chip; the theo- amide gel stained with ethidium bromide. increased.
retical DNA amplification factor is thus 2n. Fluorescence was imaged by a charge-cou- To build a complex system analogous to
We used a chip that generates 20 identical pled-device camera and quantitatively ana- an electronic integrated circuit, it is neces-
cycles, each having a time ratio of 4:4:9 lyzed with the ImageQuant software package sary to integrate a variety of other continu-
(melting:annealing:extension), for a theoret- (Fig. 2A). ously working parts. A cursory glance at
ical amplification factor of 220 (Fig. 1B); the The results demonstrate that with a very developments in miniaturized total analysis
first cycle contained a threefold extended simple device and virtually no optimization, systems (mTAS) during the past 5 years re-
melting time to ensure proper accessibility of PCR can be performed in continuous flow, veals that most of the relevant components,
the template DNA. Two of the three inlets of yielding product quality and quantity compa- such as continuous-flow mixers (8), contin-
the chip were used for the continuous buffer rable to standard thermal cycling methods uous-flow microreactors (9), continuous sep-
flow and the sample injection (the third inlet (Fig. 2B). A negative control (Fig. 2A, lane arations (10), and high-speed capillary gel
was not used in these experiments). All fluids 9), containing the whole PCR mixture ex- electrophoresis (as a conventional batch pro-
were pumped by hydrostatic pressure; the cept for the template, was run at a flow rate cess), have already been presented (11). Ac-
channel dimensions (40 mm by 90 mm by that should yield high amplification (for ex- cordingly, the continuous-flow microampli-
2.2 m) were such that a pressure drop of 1 bar ample, Fig. 2A, lane 4) subsequent to a sam- fier should allow the creation of highly elab-
over the whole chip resulted in a flow- ple run at high template concentration. No orate analysis and synthesis systems.
through time of 4 min for the 20 cycles. product could be detected, which is not sur- Although chemical amplification reac-
tors will obviously not be applicable to all
reactions, neither will they be limited to
Fig. 2. (A) PCR products by flow-through thermal 1 2 3 4 5 6 7 8 9
cycling. All PCR reactions were performed with the A PCR. Devices of this kind open up whole
same PCR mixture: 10 mM tricine (pH 8.4), 0.01% new areas of application in medical diag-
(w/v) Tween 20, 50 mM KCl, 0.2 mM deoxynucle- 700 bp
nostics. For example, the online amplifi-
otide triphosphate, 1.5 mM MgCl2, 1.4 mM polyvi- 500 bp cation and monitoring of a specific gene
nylpyrrolidone, 1 mM primers (59-TGCACCGGC- 300 bp could show the patient’s ability to metab-
GCGTACTGTA-39, 59-CATCACATAACGCATAGC- 176 bp olize a given drug so as to determine the
39), Taq (0.25 U/ml), and ;108 copies of template. Ref. Product 100 bp ideal course of therapy. On-site analysis of
Lane 1: Reference PCR performed in a very rapid patient samples could demonstrate the
commercial thermocycler (Hybaid, PCR Express) presence of bacterial DNA and any sus-
with a cycling time of 50 min for 20 cycles. Lanes 2 to
8: Flow-through PCR with increasing flow rates, 5.8
ceptibility to antibiotic treatment.
to 72.9 nl/s, corresponding to cycling times of 18.8
to 1.5 min for 20 cycles. Lane 9: negative control; REFERENCES AND NOTES
PCR mixture without template run at a flow rate of ___________________________
15.6 nl/s (as in lane 4) just after a run with normal 1. K. B. Mullis, F. Ferré, R. A. Gibbs, The Polymerase
Chain Reaction (Birkhäuser, Boston, 1994).
template concentration. The rightmost lane is a 100-
2. H. Swerdlow, K. Dew-Jager, R. F. Gesteland, Bio-
bp DNA ladder (15628-019, Gibco BRL). (B) Fluo- techniques 15, 512 (1993).
rescence was integrated over the indicated areas in 3. J. Cheng, M. A. Shoffner, K. R. Mitchelson, L. J. Kricka,
(A) with ImageQuant software. Values were normal- P. Wilding, J. Chromatogr. A 732, 151 (1996).
ized to the fluorescence of the product from the 4. A. T. Woolley et al., Anal. Chem. 68, 4081 (1996).
reference cycler (100%) and plotted against their to- 5. A. Manz and H. Becker, Eds., Microsystem Technol-
tal cycling time, that is, the time needed for a fluid ogy in Chemistry and Life Sciences, vol. 194 of Top-
ics in Current Chemistry (Springer, Berlin, 1998).
element to pass through the 20 cycles of the chip.
6. E. L. Cussler, Diffusion Mass Transfer in Fluid Sys-
Because on each run 10 ml had been collected for tems (Cambridge Univ. Press, Cambridge, 1984).
slab-gel analysis, the overall run time increased by the time needed to accumulate this large volume. Numbers 7. N. Chiem and D. J. Harrison, Anal. Chem. 69, 373
correspond to the lanes in (A). A logarithmic regression was fitted to the data points. (1997). z SCIENCE z VOL. 280 z 15 MAY 1998 1047

8. U. D. Larsen, J. Branebjerg, G. Blankenstein, in Pro- mer, ibid., p. 2949. (12)—among the best preserved and most
ceedings of the 2nd International Symposium on 12. Supported by SmithKline Beecham, Zeneca, and
Miniaturized Total Analysis Systems mTAS ’96 (Ana- a grant from the Biotechnology and Biological
complete dinosaur skulls known—and most
lytical Methods & Instrumentation, Base 1, 1996), pp. Sciences Research Council (UK). We thank B. Rob- of the tail. The skull is disarticulated and
228 –230. ertson, for primers and templates used in individual bones are virtually undistorted,
9. T. Laurell and J. Drott, Biosens. Bioelectronics 10, this study, and A. Ivens. We also thank Hybaid for the allowing comprehensive and detailed study
289 (1995). loan of their PCR instrument and the Alberta Micro-
10. D. E. Raymond, A. Manz, H. M. Widmer, Anal. electronic Centre for the production of the microchips. of all elements (Fig. 1). The external sur-
Chem. 66, 2858 (1994). face of many elements is covered in rugose
11. C. S. Effenhauser, A. Paulus, A. Manz, H. M. Wid- 15 December 1997; accepted 23 March 1998 sculpturing, and the skull roof is adorned
with three median ornamentations: thick-
ened, fused nasals; a low frontal horn; and a
Predatory Dinosaur Remains from Madagascar: parietal eminence. The total skull length is
57 cm, and comparisons with a closely re-
Implications for the Cretaceous Biogeography lated taxon, Carnotaurus sastrei from Argen-
of Gondwana tina (13), suggest a total adult body length
of about 7 to 9 m. A second specimen (UA
Scott D. Sampson,* Lawrence M. Witmer, Catherine A. Forster, 8678) of the same taxon includes an incom-
plete and disarticulated skull, most of the
David W. Krause, Patrick M. O’Connor, Peter Dodson, precaudal axial column, and the left ilium.
Florent Ravoavy Several of the vertebrae and ribs, particu-
larly in the cervical region, were recovered

Downloaded from on November 14, 2007

Recent discoveries of fossil vertebrates from the Late Cretaceous of Madagascar include in articulation. The small size of the skull
several specimens of a large theropod dinosaur. One specimen includes a nearly com- elements relative to those of FMNH PR
plete and exquisitely preserved skull with thickened pneumatic nasals, a median frontal 2100, combined with the lack of fusion
horn, and a dorsal projection on the parietals. The new materials are assigned to the between several vertebral centra and corre-
enigmatic theropod group Abelisauridae on the basis of a number of unique features. sponding neural arches, indicates that this
Fossil remains attributable to abelisaurids are restricted to three Gondwanan land- animal was immature at the time of death.
masses: South America, Madagascar, and the Indian subcontinent. This distribution is Although large theropod materials from
consistent with a revised paleogeographic reconstruction that posits prolonged links the Maevarano Formation have generally
between these landmasses (via Antarctica), perhaps until late in the Late Cretaceous. been referred to Majungasaurus crenatissimus
(5, 6, 9), the inadequacy of the holotype
and neotype specimens (14) requires that
this taxon be regarded as a nomen dubium.
Dinosaurs underwent their greatest diver- marine fossils. Plate fragmentation can set Comparison of the recently collected mate-
sification during the Late Jurassic and Cre- minimum dates for the origin of particular rials with the fragmentary holotype speci-
taceous. Although plate tectonics during terrestrial and freshwater clades if members men of the putative Malagasy pachycepha-
this interval had a profound impact on the are present on two or more landmasses. losaur, Majungatholus atopus, demonstrates
evolution of dinosaurs and coeval terrestrial Conversely, phylogenetic patterns can pro- that Majungatholus is not a pachycephalo-
faunas, this impact remains poorly under- vide increased paleogeographic resolution saur but rather a “domed” theropod (15).
stood, in part because of a paucity of fossil and serve as independent tests of tectonic This finding has biogeographic significance
remains from southern continents. Recent models (3). in that it removes the only report of a
expeditions have resulted in important di- Here, we describe theropod dinosaur fos- pachycephalosaur from a Gondwanan land-
nosaurian discoveries on all major land- sils from the Upper Cretaceous (?Campani- mass, thereby restricting occurrences of this
masses that once formed the southern su- an) Maevarano Formation, Mahajanga Ba- dome-headed ornithischian clade to Laur-
percontinent of Gondwana, and it is now sin, northwestern Madagascar (4). Fragmen- asia. Thus, the materials described herein
possible to begin assessing the biogeograph- tary dinosaur remains have been reported are referred to Majungatholus atopus and
ic history of dinosaurian clades, at least on a from the Mahajanga Basin for more than a placed within the enigmatic theropod group
gross scale (1, 2). century (5), with three taxa erected during Abelisauridae (16).
Current models of the sequence and tim- that period (5–7): a sauropod, Titanosaurus The skull of Majungatholus atopus is rela-
ing of Gondwanan fragmentation are based madagascariensis; a theropod, Majungasaurus tively short and broad, with large antorbital,
predominantly on geophysical evidence and crenatissimus; and a pachycephalosaur, Ma- laterotemporal, and external mandibular
have yet to be rigorously tested with non- jungatholus atopus. Recent excavations in this fenestrae (Fig. 1). The snout is blunt and
same field area have yielded a rich diversity relatively deep at the level of the nares, with
S. D. Sampson, Department of Anatomy, New York Col- of fossil vertebrates, including abelisaurid elongate, thickened, and rugose nasals. A
lege of Osteopathic Medicine of New York Institute of theropods, titanosaurid sauropods, birds, large, bilateral pneumatic foramen pierces
Technology, Old Westbury, NY 11568, USA. crocodilians, snakes, turtles, fishes, frogs, and the fused nasals, and computerized tomo-
L. M. Witmer, Department of Biomedical Sciences, Col-
lege of Osteopathic Medicine, Ohio University, Athens, mammals (8–10). The fossils occur predom- graphic (CT) imaging demonstrates that this
OH 45701, USA. inantly in coarse-grained sandstone facies, structure is virtually hollow, supported inter-
C. A. Forster, D. W. Krause, P. M. O’Connor, Department and a variety of indicators suggest a semi- nally only by thin bony struts. The orbital
of Anatomical Sciences, State University of New York,
Stony Brook, NY 11794, USA. arid, seasonal depositional environment fenestra is rounded dorsally, with processes of
P. Dodson, Laboratories of Anatomy, Department of An- (11). Many specimens occur as disarticulated both the lacrimal and postorbital projecting
imal Biology, School of Veterinary Medicine, University of yet associated skeletons amassed into con- into it ventrally, outlining the position of
Pennsylvania, Philadelphia, PA 19104, USA.
F. Ravoavy, Université d’Antananarivo, Service de Palé-
centrations, likely representative of time-av- the eye. Just caudal to the nasals is a rough-
ontologie, Antananarivo (101), Madagascar. eraged assemblages (11). ened, conical median projection arising from
* To whom correspondence should be addressed: E-mail: One of the theropod specimens (FMNH the frontals. CT imaging shows this frontal PR 2100) includes a nearly complete skull horn to be hollow as well. The holotype of

1048 SCIENCE z VOL. 280 z 15 MAY 1998 z