1999 microfluidic devices for dna analysis

NANOTECHNOLOGY

applied to Pascal’s-triangle coefficients, as emphasized 6 Milton, R. C., Milton, S. C. and Kent, S. B. H. (1992) Science 256,
by the topologist Pontryagin20. 1445–1448
7 Figliozzi, G. M., Siani, M. A., Canne, L. E., Robson, B. and
The way forward Simon, R. J. (1996) Protein Sci. 5 (Suppl.), 72
8 Robson, B. (1996) Nat. Biotechnol. 14, 892–893
The traditional protein and materials chemists might 9 Levinthal, C. (1966) Sci. Am. 214, 42–44
well feel, faced with such emerging complexity, that 10 Biegler, T. L., Coleman, T. F., Conn, A. R. and Santosa, F. N., eds
this is some Cabalistic vision of hell! However, this is (1997) Large Scale Optimization with Apllications, Part III: Molecular
perhaps not unexpected: these difficulties relate ulti- Strcuture and Optimization (IMA Volumes in Mathematics and its
mately to notoriously tough, quite general problems Applications) (Vol. 94), Springer-Verlag
about the development of physical systems with time 11 Ward, D. J., Brass, A. M., Li, J., Platt, E., Chen, Y. and Robson, B.
and to what Dirac meant when he said that ‘the exact (1991) Peptide Pharmaceuticals (Ward, D. J., ed.), pp. 83–129, Open
application of these laws leads to equations much too University Press
complicated to handle’. Certainly, the chemist should 12 Becker, O. M. and Karplus, M. (1997) J. Chem. Phys. 106, 1495–1517
not venture forth unaided but should turn increasingly 13 Li, J., Platt, E., Waszkowycz, B., Cotterill, R. and Robson, B. (1992)
Biophys. Chem. 43, 221–238
to the mathematicians and computation theorists to 14 Robson, B., Brass, A., Chen, Y. and Pendleton, B. J. (1993)
iron out the Cabalistic details and provide better vehi- Biopolymers 33, 1307–1315
cles and tools for the navigation of conformational 15 Turner, J., Weiner, P. K., Robson, B., Venugopal, R., Schubele, H., II
space. and Singh, R. (1997) in Computer Simulation of Biomolecular Systems:
The artillery to provide the way forward glimpsed Theoretical and Experimental Applications (Vol. III) (van Gunsteren,
first by Dirac21 and computational visionaries like Boys W. F., Weiner, P. K. and Wilkinson, A. J., eds), pp. 122–149,
and Clementi is on its way. The new generations of Kluwer–ESCOM
multi-teraflop machines, special chips for molecular 16 Robson, B., Platt, E. and Li, J. (1992) in Theoretical Biochemistry and
work, sophisticated configurations of software with Molecular Biophysics: Proteins (Vol. 2) (Beveridge, D. L. and Lavery, R.,
hierarchically organized programs and new generations eds), pp. 207–222, Adenine Press
17 Henle, M. (1994) A Combinatorial Introduction to Topology, Dover
of search tools should crack the problem. If the protein Publications, New York, NY, USA
can do it, so can we. 18 Alexandroff, P. (1961) Elementary Concepts of Topology, Dover
Publications, New York, NY, USA
19 Graham, R. L., Knuth, D. E. and Patashnik, O. (1989) Concrete
References Mathematics: A Foundation for Computer Science, Addison–Wesley
1 Robson, B. (1976) Trends Biochem. Sci. 1, 49–51 20 Pontryagin, L. S. (1952) Foundations of Combinatorial Topology,
2 Drexler, E. K. (1986) Engines of Creation, Anchor/Doubleday Graylock Press, New York, NY, USA
3 Robson, B. and Garnier, J. (1986) Introduction to Proteins and Protein 21 Dirac, P. A. M. (1930) The Principles of Quantum Mechanics, Oxford
Engineering, Elsevier University Press
4 Levitt, M. and Warschel, A. (1975) Nature 253, 694–698 22 Tulp, A. (1972) Permeabiliteit en Reguleirung van het Metabolisme van
5 Anfinsen, C. (1962) Brookhaven Symp. Biol. 184, 15–17 Mitochondriën uit de Vleugspier van Musca domestica L, Elsevier

Microfluidic devices for DNA analysis
Daniel J. Ehrlich and Paul Matsudaira

Microfabricated electrophoresis devices allow us to perform short-tandem-repeat genotyping assays in under 2 min and
sequence single-stranded DNA in under 15 min. This is 10–100 times faster than standard slab-gel and capillary systems.
The microdevice format is the natural extension of 100 years of gradual improvements to electrophoresis but operates in an
almost-perfect way, limited only by the sieving medium.

arge-scale genome-sequencing initiatives and No single technology can satisfy the various demands

L recent decisions to generate large genotype data-
bases for human forensics have generated a nearly
insatiable need for improved technology at a lower cost,
for better assays. For example, hybridization arrays1 are
currently an intriguing way to obtain semiquantitative
information for gene-expression studies and other
for high-throughput and fast, nominally real-time, massive sampling applications. By contrast, the leading
DNA assays. methods for more-quantitative assays and de novo
sequencing are still based on electrophoresis. However,
D. J. Ehrlich (ehrlich@wi.mit.edu) and P. Matsudaira are at the the format for electrophoresis will be optimized differ-
Whitehead Institute for Biomedical Research and the Massachussetts ently for various categories of DNA assay. A primary
Institute of Technology, 9 Cambridge Center, Cambridge, MA 02142, tool for this optimization will be photolithography and
USA. microfabrication.

TIBTECH AUGUST 1999 (VOL 17) 0167-7799/99/$ – see front matter © 1999 Elsevier Science. All rights reserved. PII: S0167-7799(99)01310-4 315

NANOTECHNOLOGY

The procedures for making these devices are simple
Cathode reservoir using current technology for semiconductor devices9,12;
Sample waste typically, they require one or two lithography steps at
low (~10 m) resolution. The channels are semicircular
in cross section with an etched depth of 40–100 m.
After fabrication, DNA separations require a rigorous,
stable neutralization of the intrinsic negative charge that
resides on ambient SiO2 surfaces. This method was
originally described by Hjerten13 to produce a co-
valently bonded layer of polyacylimide on internal sur-
Sample faces. Following coating, the devices are injected with
Separation load the polyacrylamide (PAA) sieving solutions.
channel In a DNA assay, samples are pipetted into reservoirs
connected to the input ends of the injector channels.
Eight separation channels A voltage is applied to electrophorese the DNA past
(sample moves to bottom of page) the channel-intersection points. An orthogonal bias

trends in Biotechnology
(along the separation channel) is then applied to run
Anode reservoir the sample from the channel intersection (typically only
several hundred picoliters) down the separation channel,
where it undergoes electrophoretic separation followed
by detection via laser-induced fluorescence.
1 cm Most of this protocol is taken directly from well-
established techniques in capillary electrophoresis (CE),
Figure 1 with the exception of sample injection. Microma-
A typical eight-lane microelectrophoresis genotyping device. The channels are micro- chined devices offer unmatched control of the injection
machined between two 15-cm-diameter fused-silica wafers, which are then diced into volume and uniform delivery of the sample compo-
six to eight individual devices. The network at the broad end of the device is matched nents. In CE devices, injection volumes are difficult to
to a four-tip pipette, used for sample and buffer loading. DNA is detected by laser control and the separations show undesired bias towards
fluorescence at the narrow end of the device. During operation, the device is housed low molecular weights. A second new aspect is the
in a ceramic cassette with integrated electrodes and microfluidic reservoirs. practicality of short, 2–15 cm, devices, which are dif-
ficult to implement other than in microdevice formats.
Finally, there are also numerous system-level advantages
In the early 1990s, the concept of micromachined stemming from the fact that microfabricated devices are
devices for electrophoretic separation was proved2,3. vast planar channel structures that allow complicated
This set the stage for DNA microelectrophoresis assays intersections and other features to be fabricated easily.
to separate oligonucleotides4, restriction fragments5–7, In fact, these aspects are likely to be the most important
sequencing mixtures8,9, PCR products10, genotyping asset of the microfabricated format.
samples11 and short tandem repeats12 (STRs). We are
now at a stage where microdevices will be used at the Genotyping applications
forefront of genetic applications, some of which require Genotyping may be the first application in which the
a very high electrophoretic performance. new format outperforms the established technologies
For example, high cost, long run times, large sample in almost all practical measures. One example is the
volumes and manual operation of gel-based electro- analysis of the STR ‘CTTv’ system, which consists of
phoresis devices are among the most important the four loci CSF1PO, TPOX, THO1 and vWA. Each
factors that presently limit the pace of the Human of these loci contains STR alleles that differ in length
Genome Project. Hence, slab-gel and capillary methods by four base pairs. This assay is directly compared in
have been optimized to near their theoretical limits. As slab, capillary and microdevice formats in Fig. 2, which
a result, to be useful for such applications, any micro- shows the difference in time scales for comparable-
device must improve on the already very high perfor- quality results. In addition, the microdevice trace shows
mance standards of current methods. In the past year, a series of periodic signals, which arises from a CTTv
it has become clear that this will be achieved. This article ladder (ranging from 140 to 330 bases) that is used as
describes the current state of the art in DNA micro- an internal sizing standard for the allelic profiling. The
electrophoresis devices and some of the engineering device performance is optimized according to the
issues that limit further development of the new format. required electrophoretic resolution, R (Eqn 1),

The anatomy of a microelectrophoresis device R [(2 ln2)1/2 (t1 t2)] [(w1 w2) b] (1)
Figure 1 illustrates a typical eight-lane microelectro-
phoresis device. Micromachined into fused silica, the where t is the migration time of the fragment, w is the
heart of the structure is a series of eight pairs of inter- full width at half maximum of the peak and b is the
secting enclosed channels, each set including an injec- base-number difference between the two DNA frag-
tor channel and a separation channel. This insert is ments. The alleles of all the four loci are resolved in
made by the photolithography, etching and bonding of 2 min with resolutions (R) ranging from 1.7 for the
two fused silica or glass plates, and is housed in a vWA locus to 1.1 for the CSF1PO locus. Forensic
ceramic cassette that contains microfluidic reservoirs, applications typically require a resolution greater than
electrodes and an integrated heater. one.

316 TIBTECH AUGUST 1999 (VOL 17)

NANOTECHNOLOGY

Microdevices are filled with a polyacrylamide sepa-
ration matrix using a syringe inserted into the separation- a
channel exit reservoir, and the detector is placed
~26 mm from the injector. The device is then pre-
electrophoresed for 3 min at 200 V cm 1 and 50 C
across the separation channel. To load the sample,
400 V cm 1 is applied across the load and sample chan-
nels (Fig. 1). Fields of 40 V cm 1 are applied to the
cathode reservoir and separation channel to prevent the 1.3 1.6 1.9 2.2
sample from entering them. This results in a stable Time (h)
injection-plug length of 100 m and an injection b
volume of approximately 0.36 nl. The sample plug is
injected into the separation channel and the voltages
are switched to create a field strength of 200 V cm 1
in the separation channel and approximately 20 V cm 1
in the load and sample channels. This bias generates a
well-defined plug entering the separation channel, with
no excess sample leakage from the side channels.
25 30 35 40
The field strength determines the migration speed Time (min)
within the device and influences the performance of
the sieving matrix. To optimize this parameter, the field c
strength is increased stepwise from 200 V cm 1 (typical
for capillary devices) to as high as 800 V cm 1. At high

trends in Biotechnology
field strengths, the resolution suffers because of factors
including field-induced orientation of the DNA and
matrix distortions. The strongest permissible field (i.e.
the field that maintains a resolution of R 1) depends
on the specific locus (i.e. its molecular length). 70 80 90 100 110 120
For field strengths below 600 V cm 1, the device and Time (sec)
sieving matrix exhibit excellent long-term stability.
Migration times increase by approximately 10% during Figure 2
ten consecutive runs, but the original migration time A comparison of three different DNA-analysis techniques. (a) A slab-
can be restored by reinjecting the gel–buffer system. In gel allelic assay of a single individual for the four-locus CTTv short-
addition, the accuracy of the allele assignment is not tandem-repeat system (4-bp repeats), with a run time of 2.2 h.
affected by small changes in migration time because an (b) An identical assay performed on a capillary system (run time
internal standard can be used for allele identification. 40 min). (c) The same sample run on a microdevice system9. The
No other changes in separation results are observed small peaks are an internal standard achieved by spiking the sam-
even after 20 consecutive runs without replacing the ple with a CTTv ladder composed of all common alleles [not added
gel–buffer system in STR experiments. Therefore, a to (a) or (b)]; run time 2 min. The microdevice run time represents
single microdevice can be used for allelic profiling for a speed increase of 20 over the capillary system and 70 over
prolonged periods and multiple applications with the the slab gel.
need for only periodic reinjection of the gel.
laries tend to load excessive sample and distort sample
DNA sequencing concentrations, leading to a more-complicated, non-
DNA sequencing applications require longer devices ideal performance. As a result, only a few matrix-related
than those used for genotyping8,9. The trade-off in assay parameters need to be measured for a microdevice
speed against required resolution and read length has before a general model can be developed. At the limit,
still to be defined over the full range of PAA sieving when injection and diffusion are the sole contributors
materials. Figure 3 shows data from a four-color detec- to peak width, the theoretical resolution Rt achievable
tor with a total read length of ~525 bases. The signal for two adjacent DNA sequencing fragments during an
remains strong throughout the separation, with an electrophoretic separation can be described by Eqn 29,
almost uniform signal to noise ratio of 50:1. Analy-
sis of such traces indicates that single-base resolution Rt [ L (sinj2 2Dt) 1/2] 4 (2)
can be achieved for 200 bases in 8 min, for 300 bases
in 11 min, for 400 bases in only 13 min and for where is the difference in the electrophoretic
525 bases in only 20 min. The same sequencing analy- mobilities of the two DNA fragments, is their aver-
sis, when performed at 400 V cm 1 under otherwise- age electrophoretic mobility, L is the effective sepa-
identical conditions, requires only 7 min and generates ration distance, sinj2 is the variance of the injected sample
a maximum read length of approximately 350 bases. plug, D is the longitudinal diffusion coefficient of the
fragments and t is the separation time. / is a meas-
Scaling considerations and the limits of ure of the selectivity of the separation process and
microelectrophoresis devices depends on the matrix type, fragment size and field
Microelectrophoresis devices can be operated in a strength. For DNA analyses, the essential point is that
nearly ideal regime and free of injection-related broad- D and / are dependent on the field strength,
ening factors. More-conventional slab gels and capil- owing to orientation or conformation changes of the

TIBTECH AUGUST 1999 (VOL 17) 317

NANOTECHNOLOGY

data up to a read length of approximately 500 bases
using a 11.5-cm-long device in 20 min or less (Fig. 3).
Somewhat better results and a longer read length will
be possible with very-well-adjusted base-calling software
(B. L. Karger et al., unpublished).
The application of these techniques to genotyping
requires the optimization of different parameters,
particularly for extremely fast separations; for example,
DNA fingerprinting using various 4-base-repeat STR
systems might be achieved in near-real time9. The prin-
cipal alleles from the first three loci of the CTTv STR
system should be resolved in less than 4 sec (4% PAA,
500 V cm 1, 10 m injector, 50 C) but requires an
optimized injector geometry.

Full automation
One of the most important needs for commercial
applications is the full automation of DNA sequenc-
ing. Current commercial technology requires a num-
ber of manual steps (e.g. pouring slab gels, loading
DNA samples), but important inroads are being made
with multicolumn capillaries14,15. Nevertheless, it is
anticipated that microfabricated devices will rapidly
become the format of choice for automated sequenc-
ing. Compared with capillary systems, the microfabri-
cated devices offer identical or improved performance,
bulk manufacture with automated equipment (i.e. no
manual assembly of columns) and the potential to
develop built-in interfaces to a precision robotic front
end (sample and gel loader) and detector (i.e. litho-
graphically defined tolerances at the fluid and data
interfaces of the multichannel device).

Massive parallelism
Given the compelling need for high sample through-
put in DNA sequencing, one obvious question is, how
many samples can be analysed simultaneously on one
system? Microfabrication per se will not be limiting for
many generations of devices, particularly because only
the simplest methods have been used to date. Robots
trends in Biotechnology can easily achieve the spatial resolution required and
should not be cost prohibitive. If a conventional CCD
Figure 3
large-field imaging detector was used, the pixels avail-
DNA sequencing results for a sample run on an electrophoresis
able in the array could limit the device to approximately
microdevice 11.5 cm long and containing a 3% linear polyacrylamide
200 lanes. However, the pixel count requirement ( 50
sieving matrix (molecular weight 6 000 000 Da) separated at
pixels per channel) is not demanding, opening the way
150 V cm 1 and 45 C. The analysis time required is only 20 min,
for the less-conventional use of imaging devices. More-
with 100% accuracy between 30 and 525 base pairs; between 525
over, a scanning detector is certainly capable of greater
and 570 bases, the accuracy decreases to 90%.
than 2000 channels. Consequently, microfabricated
devices of several thousand lanes or more are potentially
possible.
DNA fragments. These dependencies are difficult to
measure but are crucial to generating a predictive model. Conclusions
The greatest improvement in channel performance
Device length and assay speed using microdevices applies to assays such as genotyping,
Based on the measurement of the field-dependent which can take advantage of the unique practicality of
selectivity and lateral-diffusion coefficient8, the maxi- short devices in the new format, but longer sequenc-
mum read length can be predicted12 for a given PAA ing reads will require long devices. However, micro-
concentration, separation length, temperature and bias. fabricated devices have fewer single-channel performance
These calculations, however, require careful measure- advantages over capillary systems as the channel length
ment of field-dependent broadening, mainly anisotropic increases. Systems advantages such as interface compat-
diffusion of the DNA in the presence of a field. ibility, geometrical flexibility, microfluidics integration
Careful measurements of dynamic PAA-sieving per- and ease of device manufacture are all intrinsic to the
formance have only been made in a few cases12. A res- microfabricated format and are relevant to all highly
olution of R 0.5 results in high-quality separation multiplexed applications, even for long-channel systems.

318 TIBTECH AUGUST 1999 (VOL 17)

FOCUS

One of the strengths of microelectrophoresis devices 3 Pace, S. J. (1990) US Patent 4 908 112
is clearly the increase in assay speed. DNA-finger- 4 Effenhauser, C. S., Paulus, A., Manz, A. and Widmer, H. M. (1994)
printing applications, such as STR analysis, are now Anal. Chem. 66, 2949–2953
5 Woolley, A. T. and Mathies, R. A. (1994) Proc. Natl. Acad. Sci.
possible in minutes or even seconds. This speed offers
U. S. A. 91, 11348–11352
many new applications for genotyping. However, the 6 Jacobson, S. C. and Ramsey, J. M. (1996) Anal. Chem. 68,
advantages of highly multiplexed or automated systems 720–723
remain to be demonstrated but should not really be a 7 McCormick, R. M., Nelson, R. J., Alonso-Amigo, M. G.,
problem once resources are applied to the problem of Benvegnu, D. J. and Hooper, H. H. (1997) Anal. Chem. 69,
scale up. Automating the microdevice format may well 2626–2630
push the sequencing-throughput bottleneck back into 8 Woolley, A. T. and Mathies, R. A. (1995) Anal. Chem. 67,
sample preparation. 3676–3680
9 Schmalzing, D., Koutny, L., Adourian, A., Belgrader, P.,
Acknowledgments Matsudaira, P. and Ehrlich, D. (1997) Proc. Natl. Acad. Sci. U. S. A.
94, 10273–10278
The authors’ work was supported by the US
10 Woolley, A. T., Hadley, D., Landre, P., deMello, A. J., Mathies, R. A.
National Institutes of Health and the US Air Force and Northrup, M. A. (1996) Anal. Chem. 68, 4081–4086
Office of Scientific Research. We thank A. Adourian, 11 Woolley, A. T., Sensabaugh, G. F. and Mathies, R. A. (1997) Anal.
L. Koutny and D. Schmalzing for extensive contribu- Chem. 69, 2181–2186
tions to the Whitehead Institute results discussed here. 12 Schmalzing, D., Adourian, A., Koutny, L., Ziaugra, L., Matsudaira, P.
and Ehrlich, D. J. (1998) Anal. Chem. 70, 2303–2310
References 13 Hjerten, S. J. (1985) J. Chromatogr. 347, 191–198
1 Fodor, S. et al. (1996) Science 274, 610–614 14 Mathies, R. A. and Huang, X. C. (1992) Nature 359, 167–169
2 Manz, A. et al. (1992) J. Chromatogr. 593, 253–258 15 Ueno, T. and Yeung, E. S. (1994) Anal. Chem. 66, 1424–1431

Cytokines: from technology to therapeutics
Anthony R. Mire-Sluis

Cytokines are playing an ever-increasing role in the treatment of human disease. The characterization of these proteins plays
a vital role in their development as useful therapeutic agents. Physicochemical techniques can produce information about the
structure and composition of cytokine therapeutics but cannot yet predict their biological activity, for which biological assays
are required. Because of the large number of techniques available and the variety of products requiring analysis, the tests
used to characterize cytokine products must be both appropriate for the product and adequately controlled if the information
they provide is to be of value.

ytokines and growth factors mediate a wide DNA screening that appear to have cytokine-like

C range of physiological processes, including
haematopoiesis, immune responses, wound heal-
ing and general tissue maintenance1; for the purposes
properties but whose true functions remain unidentified3.
Widespread application of recombinant DNA technol-
ogy within the biotechnology industry has dramatically
of this review, growth factors will be included in the increased the number of cytokines available for clinical
term ‘cytokine’. As cytokines are involved in many evaluation (Table 1). New cytokines are being dis-
physiological processes, it is not surprising that they are covered, cloned and entered into clinical trials at such
also involved in the pathogenesis of many diseases. a rate that their structural properties and biological activ-
Associated with this is their vast potential in replace- ities are often poorly understood during their develop-
ment or modulatory therapy2. Recombinant DNA ment as therapeutic agents.
technology has resulted in the discovery of increasing Safety, efficacy and quality are major concerns for the
numbers of proteins that have been categorized as success of any biological product. Safety, involving tox-
either cytokines or growth factors. There are over 150 icity and possible infectious agents, is dealt with through
well-established cytokines and growth factors3, and many well-documented procedures (regulatory guidelines
other proteins have been discovered through random available from the European Medicines Evaluation
Agency website http://www.eudra.org/emea.html) but
A. R. Mire-Sluis (amire-sluis@nibsc.ac.uk) is at the Division of efficacy can only be evaluated through clinical trials4.
Immunobiology, National Institute for Biological Standards and Control, Quality assessment needs to address a variety of issues
Blanche Lane, South Mimms, Potters Bar, UK EN6 3QG. including heterogeneity, consistency, potency, stability

TIBTECH AUGUST 1999 (VOL 17) 0167-7799/99/$ – see front matter © 1999 Elsevier Science. All rights reserved. PII: S0167-7799(99)01330-X 319

1999 microfluidic devices for dna analysis

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to 1999 microfluidic devices for dna analysis (20)

Recently uploaded (20)

1999 microfluidic devices for dna analysis