Molecular Breeding 4: 277289, 1998.

1998 Kluwer Academic Publishers. Printed in The Netherlands.

277
Overview of DNA chip technology
Bertrand Lemieux
1
, Asaph Aharoni
2
and Mark Schena
3,
1
Department of Plant and Soil Sciences, University of Delaware, Newark, Delaware 19717, USA;
2
Department
of Cell Biology, DLO Center for Plant Breeding and Reproduction Research, P.O. Box 16, 6700 AA Wagenin-
gen, The Netherlands;
3
Department of Biochemistry, Beckman Center, Stanford University School of Medicine,
Stanford CA 94305, USA (

author for correspondence; phone: (650) 7235484, fax: (650) 7236783, e-mail:
schena@cmgm.stanford.edu)
Received 10 March 1998; accepted in revised form 20 April 1998
Key words: genomics, genotyping, microarrays, mutation detection, transcript monitoring
Abstract
DNA chip technology utilizes microscopic arrays (microarrays) of molecules immobilized on solid surfaces for
biochemical analysis. Microarrays can be used for expression analysis, polymorphism detection, DNA resequenc-
ing, and genotyping on a genomic scale. Advanced arraying technologies such as photolithograpy, micro-spotting
and ink jetting, coupled with sophisticated uorescence detection systems and bioinformatics, permit molecular
data gathering at an unprecedented rate. Microarray-based characterization of plant genomes has the potential
to revolutionize plant breeding and agricultural biotechnology. This review provides an overview of DNA chip
technology, focusing on manufacturing approaches and biological applications.
Introduction
Microarray assays [711, 3945] arose out of bio-
chemical experiments on solid surfaces [17, 31, 35].
Chip-based approaches utilize high-density molecu-
lar arrays to examine biochemical samples. Mixtures
of deoxyribonucleic acid (DNA) or ribonucleic acid
(RNA) isolated from biological sources, are labeled
enzymatically by incorporating nucleotides bearing
reporter tags and hybridized to microarrays. Hy-
bridization reactions yield heteroduplexes between in-
dividual components of the uorescent sample (probe)
and a complementary sequence (target) on the chip
surface. Because each target element or feature is
chemically homogeneous and occupies a known loca-
tion, the identity and quantity of each component in
the uorescent mixture can be ascertained by measur-
ing the uorescence intensity at each position on the
microarray.
Solid surface micro-scale assays permit the use
of sophisticated uorescence detection technology in-
cluding confocal laser scanning and charge-coupled
device (CCD) imaging which allow high-speed, quan-
titative data gathering. Though the basic principals
behind DNA chips (e.g., the hybridization of sam-
ples to immobilized DNA molecules) are conceptually
similar to those used in earlier lter-based assays
[30] the precision, speed and scale afforded by DNA
chip assays is unprecedented and represents a major
technological advance in molecular biology.
Microarrays are prepared by various synthesis or
deposition strategies. Synthesis strategies typically
produce microarrays consisting of groups of oligonu-
cleotides ranging in size from 1025 bases, while
chips prepared by micro-deposition technologies are
usually composed of 0.52.0 kb cDNAs amplied
by the polymerase chain reaction (PCR). In this re-
view, we will refer to the two types of microarrays
as oligonucleotide microarrays and cDNA microar-
rays, though this distinction is somewhat arbitrary
in that it is possible to prepare oligonucleotide mi-
croarrays by micro-deposition; conversely, improve-
ments in DNA synthesis strategies may soon allow the
synthesis of 50100 base cDNA sequences.
We will restrict our review to DNA microarrays
produced on glass surfaces at densities of 400250000
278
features/cm
2
. Low-density arrays on porous surfaces
such as nylon and nitrocellulose are not considered
in this report. Each different type of DNA chip tech-
nology will be examined in the context of specic
research applications. Future trends in microarray re-
search and the potential impact of this technology on
agriculture will also be addressed.
Physical chemistry of microarray hybridization
Several parameters inuence the rate of heteroduplex
formation during the hybridization reaction including
the concentration of the target, concentration of the
probe, sequence composition of the heteroduplexes,
salt and temperature. When the target concentration
is 10-fold greater than its cognate species in the
probe mixture, pseudo-rst order reaction kinetics
ensue such that the hybridization rate is determined
largely by the probe concentration (Schena and Davis,
unpublished results); thus, a two-fold increase in
probe concentration produces a two-fold increase in
signal. Pseudo-rst order conditions greatly improve
quantitation by minimizing the effects of minor differ-
ences in target concentration caused by imprecision in
microarray manufacturing methods.
Hybridization reactions involving target concentra-
tions that are equal to or less than the probe concen-
tration, display second order hybridization kinetics.
Under these conditions, small differences in the con-
centration of immobilized DNA (target) can have a
large effect on the rate of hybridization and the ab-
solute signal. Efcient synthesis and micro-deposition
strategies, used in conjunction with robust coupling
chemistries, permit the preparation of microarrays that
have a high target concentration and therefore display
pseudo-rst order kinetics over a wide range of sam-
ple concentrations. Most of the published assays with
oligonucleotide and cDNA microarrays utilize chips
that contain an excess of immobilized target DNA
relative to probe [79, 19, 29, 3945].
The presence of monovalent cations, such as
sodium, increase the rate of heteroduplex forma-
tion by shielding the negatively charged phosphate
backbones that would otherwise hinder base pairing
interactions between target and probe molecules. Typ-
ically, sodium ion concentrations of 1 molar are
used for both oligonucleotide and cDNA microar-
rays experiments. Temperature exerts a positive ef-
fect on hybridization rate constants, providing that
the hybridization temperature is sufciently below the
melting temperature (Tm) of the heteroduplexes. Hy-
bridization temperatures of 2542

C and 5570

C
are typically used for oligonucleotide and cDNA mi-
croarray experiments, respectively. Optimal salt con-
centration and hybridization temperature must be em-
pirically determined for a given set of experimental
conditions.
Sequence composition is the parameter over which
the experimenter has least control, and is a much
greater concern with oligonucleotides than cDNAs.
Pairs of G:C form three hydrogen bonds whereas A:T
pairs form only two; consequently, increased duplex
stability is observed in GC-rich regions. When these
stable regions are 50 bases long, they are thought
to serve as nucleation sites from which heterodu-
plex formation spreads. Considering that thousands of
different heteroduplexes are being formed in a single
microarray hybridization, it is critical to select a set
of compromise reaction conditions that gives opti-
mal signal to noise discrimination for as many of the
heteroduplexes probes as possible. This is empirically
done by introducing appropriate positive and negative
control sequences on the array as well as spiking-in a
number of positive controls.
The physical chemistry of hybridization to
oligonucleotide microarrays is clearly different from
that of cDNA microarrays. As previously mentioned,
hybridizations involving oligonucleotides are more
sensitive to the GC content of individual heterodu-
plexes than experiments involving longer DNA se-
quences. To minimize this problem, agents that equal-
ize the binding energies of G:C and A:T base pairs
such as tetramethylammonium chloride (TMAC) can
be added to the hybridization reaction. Because base
composition issues are much less important with 0.5
2.0 kb sequences, the use of TMAC is an unnecessary
precaution for cDNA microarray experiments. Sin-
gle base mismatches can have a pronounced impact
on the hybridization reassociation of short sequences
(i.e., oligonucleotides) and relatively little effect on
cDNA hybridization. This fact can be exploited in
resequencing and diagnostic applications of oligonu-
cletide microarrays. In general, issues of hybridization
specicity and cross-hybridization are important and
complex issues and must be carefully considered and
controlled for in each of the applications involving
microarrays of oligonucleotides and cDNAs.
279
Oligonucleotide microarrays
Manufacturing of oligonucleotide microarrays
The most efcient strategy for oligonucleotide mi-
croarray manufacturing involves DNA synthesis on
solid surfaces using combinatorial chemistry methods.
Glass is currently used as the preferred synthesis sub-
strate because of the inert chemical properties of this
material, the ability to chemically derivatize the sur-
face, and its low level of intrinsic uorescence. As
it is impossible to purify full-length reaction prod-
ucts from the shorter precursors, in-situ synthesis
strategies have limitations as far as oligonucleotide
length is concerned. On the other hand, considering
that DNA is synthesized at a large number of sites
in parallel and that sequence data can be taken di-
rectly from databases without the physical handling
of clones, synthesis methods have a signicant advan-
tage over deposition strategies when very high density
microarrays are required.
Of the three approaches currently used to manu-
facture oligonucleotide microarrays, the light-directed
deprotection method of Fodor et al. [17] is the most
effective strategy for generating very high density
chips. This method uses modied nucleotide phos-
phoramidites bearing a photo-labile protecting group,
allowing light to be used as the activating agent in the
synthesis reaction (Figure 1). A single round of syn-
thesis involves light-directed deprotection, followed
by nucleotide coupling (Figure 1). Photolithographic
masks are used to control the regions of the chip desig-
nated for illumination (Figure 1). Flooding the surface
of the chip with one of the four bases results in the se-
lective coupling of that base to each deprotected region
on the synthesis surface (Figure 1).
Considering that a separate deprotection and cou-
pling cycle is required for each of the four possible
bases, the maximum number of cycles needed to pro-
duce all of the possible 20-mer primers would be 80
(i.e., 4 n, where n is the length of the oligonu-
cleotide). This method currently allows for very high
density (250000 features/cm
2
) microarrays to be man-
ufactured; however, current coupling efciencies im-
pose a limit of 25 bases to these chips. Beyond
100 cycles, the accumulation of incomplete synthe-
sis products becomes a problem. Recent advances in
related photolithographic technologies, such as those
involving acid resists, may increase the efciency of
in situ synthesis such that oligonucleotides containing
>25 bases are feasible.
Figure 1. Light-directed synthesis strategy used in the manufacture
of oligonucleotide microarrays (Fodor et al., 1991). A solid surface
derivatized with chemical linkers containing photolabile protect-
ing groups (X), is activated selectively by shining light through a
photomask (M1). The wafer is ooded with a modied nucleotide
(AX), resulting in the coupling of Aresidues to the activated region
of the chip. A second region of the chip is activated selectively with
light and a second photomask (M2), allowing the coupling of g
residues to a different activated region. A repeated series of steps
allows parallel oligonucleotide synthesis at many addresses.
Generally speaking, photolithograpy is a relatively
poor prototyping technology and a rather good pro-
duction technology. A principal disadvantage of this
method is that a signicant amount of design work
and cost is associated with mask design. Once a set of
photolithographic masks have been made, however, a
large number of chips can be produced at a reasonable
cost. The availability of Affymetrix chips is currently
limited by how fast this company can grow and the
ease by which the photolithographic process can be
scaled.
Other in situ manufacturing technologies can be
used to make oligonucleotide microarrays [46]. For
example, ink-jet printing technology is being used to
direct the delivery of phosphoramidites to specied
addresses [6]. In this method, a glass surface is coated
with a light-sensitive hydrophobic material. Synthesis
areas are then prepared by photolithographic etch-
ing. Phosphoramidites are delivered to the resulting
hydrophilic wells to allow chemical coupling. The
wafer is then washed with oxidizer and deblocking
agent to complete a round of coupling. Although only
modest densities (10 000 features/cm
2
) are attainable
by this technology, the simplicity of ink-jetting may
bring oligonucleotide microarrays to a large number
of researchers in the near future. Efforts at Protogene,
Incyte, Hewlett Packard, Rosetta and several other
companies suggest that the ink-jetting approach to mi-
croarray preparation has serious commercial viability.
280
Mutation detection with oligonucleotide microarrays
Microarrays of all the possible 8-mers were once
thought to hold the promise of providing an alterna-
tive the Sanger-based DNA sequencing. The method
of sequencing by hybridization (SBH), proposed by
Drmanac and others [4, 13, 23], seemed to be an
appealing application of oligonucleotide microarray
technology. To date however, reaching this objec-
tive has proven elusive. The principle shortcoming
in microarray-based sequencing by hybridization de-
rives from the fact that it is difcult to obtain uniform
hybridization signals for a large number of oligonu-
cleotides in parallel. The lack of signal uniformity
caused by sequence-dependent variability in heterodu-
plex formation leads to the acquisition of both false
positives and false negatives. The resulting data set
poses a classic signal-to-noise dilemma that pre-
cludes the unambiguous assembly of de novo se-
quences. This limitation, coupled with the presence
of repeated sequences in complex genomes leading to
assembly branchpoints, suggests that the sequencing
by hybridization application with microarrays may not
be realized unless signicant improvements are made
in chip design and experimental approach.
As an alternative to sequencing by hybridization,
attention has turned to more conventional diagnos-
tic applications of oligonucleotide microarrays [7, 8,
15, 19, 25, 50] such as the identication of single
base mismatches in DNA sequences whose primary
composition has been determined by conventional se-
quencing. The so-called resequencing approaches rely
on the capacity to distinguish a perfect mismatch from
a single base mismatch, by hybridization to a re-
lated set of four groups of oligonucleotides that are
identical except for the central base which contains
either A, G, C or T (Figure 2). Mismatches in the
center of the oligonucleotide sequence have a greater
destabilizing effect than mispairing at distal positions
(Figure 2). The Affymetrix tiling strategy utilizes a
set of four oligonucleotides for each base to be rese-
quenced (Figure 2). By this approach resequencing a
10 kb gene requires a microarray of 40 000 oligonu-
cleotides, a density that is easily accomplished using
the Affymetrix technology.
The 4L tiling pattern is used to detect point muta-
tions in a known DNA sequence (Figure 2). In one
study, the authors designed a chip based on the re-
ported sequence of the human mitochondrial DNA [7].
The complete sequence of the human mitochondrial
genome was synthesized on a chip as a set of 16000
Figure 2. Tiling strategy used for DNA resequencing with oligonu-
cleotide microarrays. Schematic of a chip containing 6 columns
of 15-mers, with each column containing four groups of oligonu-
cleotides that differ only at the central position (underlined).
Incubation of the chip with a uorescent sample results in
sequence-specic hybridization events between the probe and
oligonucleotide targets on the surface (solid blocks). Comparative
uorescence intensity analysis of each column allows base-by-base
resequencing of a wild type (TACGCA) and mutant (TACTCA)
gene. The single base mutation in the mutant sequence leads a
general reduction of signal (stippled blocks) in the proximity of the
point mutation.
20-mers. The resulting 64 000 feature chip allowed
mutation detection of the entire mitochondrial genome
in a single hybridization. Out of a total of 180 poly-
morphisms, the chip succeeded in indentifying 179
base changes [7].
Exploiting a two-color hybridization scheme anal-
ogous to that originally proposed by Kallioniemi et
al. [22], mitochondrial resequencing is accomplished
by preparing uorescent samples from experimental
and control sources. Two mitochondrial DNA samples
are labeled by PCR amplication in the presence of
nucleotide analogues derivatized with biotin and uo-
rescein, respectively, followed by fragmentation of the
amplicon mixtures with DNaseI. Following hybridiza-
tion of the two samples, the microarray is incubated
with a streptavidin-phycoerythrin conjugate after the
unhybridized material has been washed away. Detec-
tion is then accomplished by sequential excitation and
detection of phycoerythrin and uorescein emission,
respectively.
Point mutations are detected by plotting the ratios
of the two signals at each feature on the chip, with
the presence of a mismatch in the patient sample seen
as a reduced phycoerythrin signal compared to the
uorescein positive control. In some cases, the substi-
281
tuted nucleotide can be ascertained by identifying the
feature with highest phycoerythrin signal in the tiling
footprint. It should be noted that this tiling strategy
does not allow detection of all possible mutations. For
example, some GC- rich regions hybridize to several
adjacent features on the chip probes such that muta-
tions in these regions do not leave a foot print in the
tiling block.
Genotyping by hybridization to oligonucleotide
microarrays
The tiling strategy outlined above is typically used to
detect new mutant alleles or polymorphisms in a given
sample. However, in diagnostics and genetic mapping,
researchers often know which alleles are likely to be
informative and the search can be focused on these
polymorphisms directly. The focus on a targeted set of
loci can simplify the design of robust tiling strategies.
Diagnostic chips have been designed to detect
previously reported mutant alleles in the CFTR [8],
BRCA1 [19], and beta-globin [50]. For the CFTR ex-
periments, two microarrays were manufactured. One
microarray, containing 428 features, was designed to
identify mutations in exon 11 of CFTR, while the
second type of microarray contained 1480 elements
chosen to detect known deletions, insertions, or base
substitution mutations. The validity of the oligonu-
cleotide microarrays was established by hybridization
of labeled control oligonucleotides targets. Hybridiza-
tion of CFTR genomic DNA samples from patients
bearing previously characterized mutations were then
used to further test hybridization specicity. Finally,
ten unknown patient samples were genotyped using
the CFTR oligonucleotide microarray assay and the
genotype assignments were identical to those obtained
by restriction fragment analysis of PCR products.
In the BRCAI experiments, a microarray contain-
ing 96 600 20-mers was used to identify mutations
over the entire 3.45 kilobases (kb) of exon 11. Ref-
erence and test samples were co-hybridized to these
microarrays and differences in hybridization patterns
were quantitated by two-color analysis. Fourteen of
fteen patient samples with known mutations were
accurately diagnosed, and no false positive mutations
were identied in 20 control samples. Eight single
nucleotide polymorphisms were also identied.
cDNA microarrays
Production of cDNA microarrays
Microarrays containing large DNA segments such as
cDNAs are generated by physically depositing small
amounts of each DNA of interest onto known loca-
tions on glass surfaces. Two advanced liquid deliv-
ery technologies have been used extensively thus far.
The rst consists of the mechanical microspotting ap-
proaches that utilize pins, tweezers, capillaries and
other pronged devices to deposit small quantities of
DNA onto known addresses using motion control sys-
tems [2, 14, 24, 39, 44, 46]. The rate and accuracy by
which the pre-synthesized material can be delivered
to the surface depends on a large number of para-
meters including the precision of the robotic control
system and the micro-machining technology used to
manufacture the printing devices.
As originally proposed by Southern [46], the most
convenient printing substrates are standard 1 3 inch
microscope slides bearing chemically modied sur-
faces that allow DNA attachment. A variety of surface
coupling chemistries, derived in part from oligonu-
cleotide microarray experiments [18, 26], have re-
cently been employed including polylysine coatings
[9, 39, 44] and reactive aldehydes [21, 41].
Modern gridding robots can easily print up to 100
chips in a single session (see http://cmgm.stanford.edu
/pbrown/ for details on one such instrument). Recent
advances in microspotting technology, such as the
availability of high-precisionprint heads containing 32
pins (Figure 3A, B), allow for the preparation of 100
microarrays containing >10 000 features in 12 h.
Further improvements in micro-spotting technologies
will eventually permit the automated fabrication of
cDNA microarrays containing 100000 features. High-
density DNA printing, together with suitable uo-
rescence detection equipment, enable efcient and
inexpensive genome analysis of any eukaryotic organ-
ism.
The second approach to cDNA microarray fabri-
cation utilizes ink-jet nozzles to deliver concentrated
DNA solutions to glass surfaces. This technology re-
lies on the so-called piezoelectric effect (review by
Blanchard, [6]), whereby certain materials such as ce-
ramics expand when exposed to an electric potential.
Piezoelectric ttings, tightly attached to glass capillar-
ies, allow the selective contraction of the capillaries in
an electrically controlled manner. Piezojets, attached
to suitable motion control systems, are capable of de-
282
Figure 3. Mechanical microspotting technology. (A) ChipMaker
TM
microspotting device. A high-precision print head and 16 microspotting
pins. The print head holds 32 pins capable of delivering 32 samples simultaneously. A feature size of 100150 m permits the manufacture of
100 microarrays containing 10 000 spots in 12 h. The region of the microspotting pin shown in (B) is denoted (arrow). Space bar = 1.0 cm. (B)
Enlarged view of a microspotting pin tip. A 0.2 l sample is loaded into the microspotting pin by capillary action. Printing is accomplished by
direct surface contact between the pin and the printing substrate, which results in the deposition of 1.0 nl samples at dened addresses. Space
bar = 200 m. (For more information, see http://www.hooked.net/users/telechem/).
283
livery of 100 picoliter droplets at a rate of 10000
droplets per second. Several companies including In-
cyte Pharmaceuticals and Protogene have proprietary
ink-jetting technologies. The recent preparation of
high-density cDNA microarrays for human gene ex-
pression analysis [42] demonstrate the viability of ink
jet technology.
One major advantage of the deposition technolo-
gies, relative to the synthesis approaches, is the ca-
pacity to prepare microarrays of virtually any mole-
cule of interest including genomic DNAs, antibodies,
lipids, carbohydrates, small molecules and so forth.
At the current printing densities, it would be possi-
ble to prepare microarrays of bacterial articial chro-
mosomes (BACs) sufcient to cover the entire 3
10
9
bp of the human genome. Microarrays of human
BACs might nd use in a variety of genomics exper-
iments including gene mapping, DNA ngerprinting
and promoter identication. Though microarrays of
long DNA molecules are unsuitable for examining se-
quences at single nucleotide resolution, microarrays of
genomic DNA can be used for massive, parallel analy-
sis of genomes at single gene resolution. One of these
applications is outlined in the next section.
Genomic mismatch scanning
Genomic mismatch scanning (GMS) is a hybridization-
based method of linkage analysis that allows rapid
identication of regions of identity-by-descent be-
tween two related individuals [34]. Genetically com-
mon chromosomal segments are identied based on
the ability of these DNA sequences to form exten-
sive regions of mismatch-free heteroduplexes. A series
of enzymatic steps, coupled with lter binding, is
used to selectively remove heteroduplexes that con-
tain mismatches (i.e. chromosomal regions that do
not share identity-by-descent). Fragments of chromo-
somal DNA representing inherited regions are hy-
bridized to a microarray of ordered genomic clones,
and positive hybridization signals pinpoint regions of
identity-by-descent at high resolution.
Because the entire genomic sequence of yeast is
known, GMS has been most extensively used for
genetic studies in S. cerevisiae. This approach is par-
ticularly promising as a tool to map multigenic traits.
For example, the ability of yeast strains to grow at
42

C is believed to be an required for the patho-
genicity of certain yeast strains [32]. By analyzing the
inheritance of large numbers of tetrads derived from
crosses of pathogenic and wild type strains, it should
be possible to identify all the genes that confer growth
at elevated temperatures.
Considering that inter-alu PCR products gener-
ated with human DNA templates are currently being
physically mapped with radiation hybrids, it is con-
ceivable that GMS will eventually become a tool for
genetic linkage studies in humans. These mapped
products could be used to prepare a microarray of
physically-ordered fragments, for use in detecting
meiotic recombination breakpoints.
Plant genomes also contain a number of repeat el-
ement families that could serve as the equivalent of
AluI elements for the generation of PCR fragments.
For example, the maize genome contains moder-
ately repetitive DNA sequences (ZLRS) representing
2500 copies per haploid genome; moreover, the
ZLRS sequences are present in the genus Zea and ab-
sent in other graminaceous species [36]. Analysis of
the ZLRS sequence from Zea diploperennis revealed
the presence of a long terminal repeat-like region, two
clusters of different tandem repeats and several ORFs,
suggesting that ZLRS may be derived from transpos-
able elements [33]. Limited genomic analysis of the
Adh region of the maize genome suggests that these
dispersed repeats are very numerous in this genome
[37]. Ananiev et al. [1] have reported unusual plants
with individual maize chromosomes added to a com-
plete oat genome generated by embryo rescue fromoat
(Avena sativa) maize (Zea mays) crosses. By using
highly repetitive maize-specic sequences as probes,
these authors were able to selectively isolate cosmid
clones containing maize genomic DNA. In a sense,
these lines have the potential of being used to cre-
ate radiation hybrid panels of maize, with the maize
repetitive DNA sequences playing the role of the AluI
repeat in identifying the maize genomic DNA in an oat
genomic background. Therefore, the plant equivalent
of radiation hybrids could eventually be available.
GMS is only one example of the use of gene
microarrays to characterize the composition of nu-
cleic acid mixtures subjected to in vitro selection.
Restriction endonuclease protection selection and am-
plication (REPSA) is another example of a selection
method that could be adapted to a DNA microar-
ray based-detection [20]. REPSA utilizes a combina-
tion of restriction enyzme cutting, PCR amplication
and lter binding to selectively identify consensus
sequences for DNA binding proteins.
284
DNA chips and functional genomics
Transcript monitoring with DNA chips
Measuring transcript levels for thousands of genes in
parallel is one of the more widespread applications
of DNA chip technology. Microarrays for gene ex-
pression analysis were the rst biological application
of DNA chip technology. Both oligonucleotide and
cDNA microarrays work well for transcript monitor-
ing [911, 21, 27, 29, 3943].
One advantage of oligonucleotide microarrays for
expression studies is that chips can be prepared di-
rectly from sequence databases, obviating the need
for cumbersome clone handling and sample tracking.
Another advantage of oligonucleotide microarrays is
that transcripts from individual members of multi-
gene families that share extensive sequence homology
can be easily distinguished by synthesizing oligonu-
cleotides to regions of non-identity. Microarrays of
cDNAs, however, also possess some distinct advan-
tages over oligonucleotide microarrays, including the
ease of prototyping and data analysis, immediate ac-
cessibility to the research community, and the capacity
to examine large numbers of novel cDNAs in gene
discovery applications. If the present research trends
continue, it would appear that both oligonucleotide
and cDNA microarrays will be widely used for gene
expression analysis.
Three strategies can be adopted when developing a
DNA chip for gene expression studies. The rst con-
sists of strategically selecting a limited genes which
are know (or suspected) to play an important role
in a particular biological pathway. One example of
this is type of study was made by Heller [21] in the
analysis of inammation. The second strategy, which
is restricted to cDNA microarrays, is to use clones
from a library prior to sequence analysis [41]. The
third strategy is to generate a chip with the com-
plete expressed sequence content of an organism [10,
49]. Genome chips provide the best chances for
discovering new interactions between metabolic and
genetic pathways and for gaining functional insights
into novel expressed sequences.
DNA microarrays and proteomics
The elucidation of protein/protein interactions within
cells, as well as the identication of proteins that bind
small ligands, is another area in which DNA chips
could signicantly increase the rate of discovery.
Bartel [5] have demonstrated that a protein-
linkage map can be created using genomic sequence
information. The authors correctly suggest that the
yeast two-hybrid system is probably the best tool
available for the systematic determination of protein-
protein interactions in complex organisms. The two-
hybrid system uses two fusion proteins to activate the
transcription of reporter genes in yeast [16]. The rst
fusion protein contains a DNA binding domain fused
to a protein of interest, while the second is an acidic
transcriptional activation domain fused to a second
protein of interest. Specic interactions between the
two chimeric proteins leads to transcriptional activa-
tion the reporter genes, which is easily scored with
either color-based assays or by auxotrophic comple-
mentation.
In the conventional two-hybrid approach, the iden-
tity of interacting proteins is conrmed by sequence
analysis of each clone identied in the yeast as-
say. Systematic two-hybrid analysis requires a large
amount of DNA sequencing, which represents a major
source of time and expense. As an alternative to con-
ventional DNA sequencing, it is possible to use chip
hybridization to identify the genes involved in protein-
protein interactions. In the case where entire genome
sequences are available (e.g., yeast), DNA chips can
be used for massive, parallel gene resequencing. With
the chips, it would be possible to rapidly identify all
of the clones whose encoded sequences interact in the
two-hybrid assay. In this experimental design, PCR
would be used to amplify and label each cDNA insert
that encodes an interacting protein. Hybridization to
genome chips would allows identication of the all of
the genes involved in protein-protein interaction in a
single hybridization.
Phage presentation libraries are also amenable to
DNA chip-based detection systems. Phage presenta-
tion utilizes fusion proteins encoded by chimeric se-
quences of bacteriophage viral coat proteins and genes
of interest (reviewed by Winter and Milstein [48]). As
in the case of the two-hybrid system, cDNA libraries
encoding fusion proteins are created. Phage display li-
braries can be made for plant species by fusing coding
sequences to plant virus coat proteins [12]. Individ-
ual clones in the library can be selected by panning
with any ligand of interest (e.g., a herbicide whose
mode of action is not fully understood). If this ligand
is immobilized to an inert support, the bacteriophage
clones that bind can be puried by elution. The bac-
teriophage particles can then be used as templates for
PCR, with a set of uor-labeled primers that ank the
285
cDNA inserts. The cDNA fragments in this enriched
population could then be characterized by microarray
hybridization.
Molecular bar coding and reverse genetics
Another post-genome era activity in which DNA chips
will play a central role is the characterization of pop-
ulations of mutant organisms exposed to various se-
lective pressures. The completion of the yeast genome
sequencing project has catapulted the yeast genetics
community into the post-genome era. One of the next
logical steps for the yeast community is the systematic
preparation of single gene deletion mutants (knock-
outs) corresponding to all 6,000 open reading frames
(ORFs). Indeed, the ease with which yeast genes can
be targeted by homologous recombination is a central
advantage of this model system.
Shoemaker [45] have proposed a new strategy for
screening large populations of knock-out mutants in
parallel. Their strategy consists of introducing unique
molecular sequences or bar codes into each of the
6000 ORFs in the yeast genome. These unique 20-
mers, which were designed to have similar melting
temperatures and minimal sequence homology, can
then be used for parallel hybridization-analysis with
oligonucleotide microarrays. In this strategy, a pool
of yeast strains containing individual bar codes for
all 6000 genes is subjected to a selective pressure.
Samples of cells growing under selective conditions
are taken at incremental times during the course of
the experiment and the bar-code sequences are labeled
by multiplex PCR with uorescent primers. Each
pool of uorescent amplicons is then hybridized to
an oligonucleotide microarray containing sequences
complementary to each of the amplied bar codes.
Comparative analysis of uorescent intensities at each
bar code position over time provides a quantitative
measure of the tness of each strain under a given
selective pressure. Correlations between strain dis-
appearance and selective pressure allow global func-
tional analysis of yeast gene function.
Because genetic bar-coding is not yet feasible in
higher organisms, systematic searches for mutant al-
leles will be required for genetic analysis in these
systems. In the case of plants, Arabidopsis is likely
the organismof choice for these efforts. Feldmann and
collaborators have proposed the use of expressed se-
quence tags (ESTs) as a global means of identifying
"insertion elements" in genes of interest (Azpiroz-
Leehan and Feldmann, 1997). In this approach, PCR
is used to screen pools of Arabidopsis lines bearing
insertion elements at random locations. Lines bear-
ing a mutational insert in a gene of interest yield a
specic amplicon in the PCR amplication. Although
this method can be performed for all of sequenced
genes of Arabidopsis, it is both labor intensive and
costly in terms of primer synthesis. Liu et al. [28] have
shown that interlaced asymmetric PCR can be used to
generate products of plant DNA / T-DNA insert junc-
tions. Hybridization of PCR amplicons to microarrays
of expressed sequences could be used to speed the
identication of mutant lines of Arabidopsis.
Potential impact on agribusiness
Functional analysis, through parallel expression mon-
itoring, should help researchers better understand the
fundamental mechanisms that underlie plant growth
and development. By accumulating databases of ex-
pression information as a function of tissue type,
developmental stage (Figure 4A, B), hormone and
herbicide treatment, genetic background and environ-
mental condition, it should be possible to identify
the genes involved many aspects of plant biology.
Microarray analysis provides a way to link genomic
sequence information and functional analysis. Recent
experiments involving the use of cDNA microarrays
for expression monitoring in strawberry and petunia
(Figure 4A, B), indicate the immediate applicability
of DNA chips in agricultural biotechnology.
Several specic research areas will be of sig-
nicant commercial interest. Because of the central
role of plant hormones in plant growth and develop-
ment, microarray-based gene expression analysis of
plant hormone action will be an important commercial
project. The interplay of genes and the environment is
also of particular importance in plants, and will con-
stitute another area of research interest. Microarrays
will assist plant biotechnology companies by allow-
ing rapid analysis of transgenic plants. These data will
permit genome-wide correlations between expression
patterns and a host of desirable traits such as fertility,
seed set, yield and resistance to environmental stress
and insects. It may ultimately be possible to reduce
the need for costly eld trials by chip-based analysis
of transgenic lines. The use of microarray technol-
ogy to understand the effect of small molecules on
gene expression might serve to speed the discovery of
herbicides and elucidate their mechanism of action.
286
287
Figure 4. (Previous page) Transcript monitoring with microarrays. (A) Pseudocolor representation of a uorescent scan of a microarray
containing 5184 groups of cDNAs, representing 1728 expressed sequence tags (ESTs) from strawberry (Fragaria annanasa cv. Elsanta) and
480 from petunia (Petunia hybreda, line W115). The cDNA inserts were amplied by PCR, puried, arrayed in duplicate with a microspotting
device, and attached covalently to the surface via Schiffs base linkages [41]. Columns 114 and 1518 of each 18 18 subarray correspond
to strawberry and petunia cDNAs, respectively. The microarray was hybridized with a uorescent probe prepared from mRNA isolated from
strawberry fruits at a late stage of ripening. The microarray was scanned for Cy3 emission with a ScanArray 3000 (General Scanning). The fth
subarray (box) is shown enlarged in (B). Color bar (right) denotes relative gene expression levels. (B) Subarray number ve from (A). Space
bar = 200 m.
The use microarrays in gene resequencing and di-
agnostics will greatly speed the identication of DNA
polymorphisms, which in turn could be used to ex-
pedite plant and animal breeding. Polymorphic mark-
ers should also nd use in a large number of other
applications, including the diagnosis and prognosis
of infectious diseases in plants and animals. Recent
results indicate that chips will allow simultaneous
analysis of thousands of polymorphisms in a sin-
gle experiment (Sapolsky et al., unpublished results).
The near-term application of DNA chip technology
in agriculture will, to a very large extent, depend on
whether cost-effective microarray technologies can be
provided to the agricultural research community. The
current urry of activity in the private sector [42] sug-
gests that complete microarray systems will soon be
available at an affordable price.
Future developments
Though microarray technology is still in its infancy,
there is a growing sense that the biochemistry of the
future will be performed on chips. This trend suggests
that many gel- and lter-based methods will eventually
give way to microarray analysis. The speed of this
transition will largely depend upon the commercial
availability of chips, detection systems, software and
reagents. The enthusiasmin the private sector suggests
that affordable systems will be widely available in the
near future [42]. The availability of high-density mi-
crospotting technologies (Figure 3A, B), for example,
will accelerate the proliferation of microarray-based
assays.
Microarray technology development can be expe-
dited by following a simple set of methodological
procedures. The so-called twelve rules of parallel gene
analysis were drafted to assist in the development of
tools and technologies in the microarray industry [43].
Methodological guidelines describe what needs to be
done and why, as opposed to dening how a partic-
ular task is to be accomplished. The methodological
principles contained in the twelve rules are intended
to be timeless and to survive the rapid proliferation
and genesis of enabling technologies. Though the
twelve rules were originally drafted for gene expres-
sion analysis, the architectural framework dened by
these guidelines is applicable to other types of assays.
Several trends are anticipated in microarray tech-
nology, including automation, miniaturization, diver-
sity and reduced cost. Improvements in automated mi-
croarray fabrication will provide increased throughput
and decreasing the cost of biological chips. Microarray
feature size will continue to decrease, allowing in-
creased target density. Though the current microarrays
consist of nucleic acids, chips containing other types
of biomolecules will eventually become available.
References
1. Ananiev, EV, Riera-Lizarazu O, Rines HW, Phillips RL: Oat-
maize chromosome addition lines: a new system for mapping
the maize genome. Proc Natl Acad Sci USA 94: 35243529
(1997).
2. Augenlicht L: United States Patent 4, 981, 783 (1991).
3. Azpiroz-Leehan R, Feldman KA: T-DNA insertion mutagen-
esis in Arabidopsis: going back and forth. Trends Genet 13:
152156 (1997).
4. Bains W, Smith GC: Anovel method for nucleic acid sequence
determination. J Theor Biol 135: 303307 (1988).
5. Bartel PL, Roecklein JA, SenGupta D, Fields S: A protein
linkage map of Escherichia coli bacteriophage T7. Nat Genet
12: 7277 (1996).
6. Blanchard A: Synthetic DNA arrays. in Genetic Engineering,
Principles and Methods, Plenum Press, in press (1998).
7. Chee M, Yang R, Hubbell E, Berno A, Huang XC, Stern D,
Winkler J, Lockhart DJ, Morris MS, Fodor SPA: Accessing
genetic information with high-density DNA arrays. Science
274: 610614 (1996).
8. Cronin MT, Fucini RV, Kim SM, Masino RS, Wespi RM,
Miyada CG: Cystic Fibrosis Mutation Detection by Hybridiza-
tion to Light-Generated DNA Probe Arrays. Human Mutation
7: 244255 (1996).
9. DeRisi J, Penland L, Brown PO, Bittner ML, Meltzer PS, Ray
M, Chen Y, Su YA, Trent JM: Use of a cDNA microarray to
analyze gene expression patterns in human cancer. Nat Genet
14: 457460 (1996).
10. DeRisi JL, Iyer VR, Brown PO (1997) Exploring the
metabolic and genetic control of gene expression on a genomic
scale. Science 278: 680686.
288
11. de Saizieu A, Certa U, Warrington J, Gray C, Keck W, Mous
J: Bacterial transcript imaging by hybridization of total RNA
to oligonucleotide arrays. Nature Biotech 16: 4548 (1988).
12. Donson J, Kearney CM, Hilf ME, Dawson WO: Systemic ex-
pression of a bacterial gene by a tobacco mosaic virus-based
vector. Proc Natl Acad Sci USA 88: 72047208 (1991).
13. Drmanac R, Crkvenjakov R: Method of sequencing of
genomes by hybridization with oligonucleotide probes. Yu-
goslav patent application 570/87 (1987).
14. Drmanac RT, Crkvenjakov RB (1993) Method of sequencing
of genomes by hybridization of oligonucleotide probes. United
States Patent 5,202,231.
15. Drmanac S, Kita D, Labat I, Hauser B, Schmidt C, Burczak
JD, Drmanac R: Accurate sequencing by hybridization for
DNA diagnostics and individual genomics. Nature Biotech.
16: 5458 (1998).
16. Fields S, Song O: A novel genetic system to detect protein-
protein interactions. Nature 340: 245246 (1989).
17. Fodor SPA, Read JL, Pirrung MC, Stryer L, Tsai Lu A, So-
las D: Light-directed, spatially addressable parallel chemical
synthesis. Science 251: 767773 (1991).
18. Guo Z, Guilfoyle RA, Thiel AJ, Wang R, Smith LM: Direct
uorescence analysis of genetic polymorphisms by hybridiza-
tion with oligonucleotide arrays on glass supports. Nucl Acids
Res 22: 54565465 (1994).
19. Hacia JG, Brody LC, Chee MS, Fodor SPA, Collins FS: Detec-
tion of heterozygous mutations in BRCA1 using high density
oligonucleotide arrays and two-colour uorescence analysis.
Nature Genet 14: 441447 (1996).
20. Hardenbol, P, Wang JC, Van Dyke MW: Identication of pre-
ferred hTBP DNA binding sites by the combinatorial method
REPSA. Nucl Acids Res 25: 33393344 (1997).
21. Heller RA, Schena M, Chai A, Shalon D, Bedilion T, Gilmore
J, Woolley DE, Davis RW: Discovery and analysis of inam-
matory disease-related genes using cDNA microarrays. Proc
Natl Acad Sci USA. 94: 21502155 (1997).
22. Kallioniemi A. Kallioniemi OP, Sudar D, Rutovitz D, Gray
JW, Waldman F, Pinkel D: Comparative genomic hybridiza-
tion for molecular cytogenetic analysis of solid tumors. Sci-
ence 258: 818821 (1992).
23. Khrapko KR, Lysov Yu P, Khorlyn AA, Shick VV, Floren-
tiev VL, Mirzabekov AD: An oligonucleotide hybridization
approach to DNA sequencing. Febs Letters: 256: 118122
(1989).
24. Khrapko KR, Khorlin AA, Ivanov IB, Chernov BK, Lysov
Yu P, Vasilenko SK, Florentev VL, Mirzabekov AD: Hy-
bridization of DNA with oligonucleotides immobilized in gel:
a convenient method for detecting single base substitutions.
Mol Biol 25: 581591 (1991).
25. Kozal MJ, Shah N, Shen N, Yang R, Fucini R, Merigan TC,
Richman DD, Morris D, Hubbell E, Chee M, Gingeras TR:
Extensive polymorphisms observed in HIV-1 clade B protease
gene using high-density oligonucleotide arrays. Nature Med.
2: 793799 (1996).
26. Lamture JB, Beattie KL, Burke BE, Eggers MD, Ehrlich, DJ,
Fowler R, Hollis MA, Kosicki BB, Reich RK, Smith SR,
Varma RS, Hogan ME: Direct detection of nucleic acid hybrdi-
cation on the surface of a charge coupled device. Nucl Acids
Res 22: 21212125 (1994).
27. Lashkari DA, DeRisi JL, McCusker, JH, Namath AF, Gentile
C, Hwang SY, Brown PO, Davis RW: Yeast microarrays for
genome wide parallel genetic and gene expression analysis.
Proc Natl Acad Sci USA 94: 1305713062 (1997).
28. Liu Y G, Mitsukawa N, Oosumi T, Whittier RF: Efcient
isolation and mapping of Arabidopsis thaliana T-DNA insert
junctions by thermal asymmetric interlaced PCR. Plant J 8:
457463 (1995).
29. Lockhart DJ, Dong H, Byrne MC, Follettie MT, Gallo MV,
Chee MS, Mittmann M, Wang C, Kobayashi M, Horton H,
Brown EL: Expression Monitoring by Hybridization to High-
Density Oligonucleotide Arrays. Nature Biotechnol 14: 1675
1680 (1996).
30. Maier E, Meier-Ewert S, Ahmadi AR, Curtis J, Lehrach H:
Application of robotic technology to automated sequence n-
gerprint analysis by oligonucleotide hybridisation. J Biotech-
nol 35: 191203 (1994).
31. Maskos U, Southern EM: Oligonucleotide hybridizations on
glass supports: a novel linker for oligonucleotide synthesis
and hybridization properties of oligonucleotides synthesised
in situ. Nucl Acids Res 20: 16791684 (1992).
32. McCusker JH, Clemons KV, Stevens DA, Davis RW: Ge-
netic characterization of pathogenic Saccharomyces cerevisiae
isolates. Genetics 136: 12611269 (1994).
33. Monfort A, Vicient CM, Raz R, Puigdomenech P, Martinez-
Izquierdo JA: Molecular analysis of a putative transposable
retroelement from the Zea genus with internal clusters of
tandem repeats. DNA Res 2: 255261 (1995).
34. Nelson SF, McCusker JH, Sander MA, Kee Y, Modrich P,
Brown PO: Genomic mismatch scanning: a new approach to
genetic linkage mapping. Nature Genet 4: 1118 (1993).
35. Pease AC, Solas D, Sullivan EJ, Cronin MT, Holmes CP, Fodor
SPA: Light-generated oligonucleotide arrays for rapid DNA
sequence analysis. Proc Natl Acad Sci USA 91: 50225026
(1994).
36. Raz R, Puigdomenech P, Martinez-Izquierdo JA: A new fam-
ily of repetitive nucleotide sequences is restricted to the genus
Zea. Gene 105: 151158 (1991).
37. SanMiguel P, Tikhonov A, Jin YK, Motchoulskaia N, Za-
kharov D, Melake-Berhan A, Springer PS, Edwards KJ, Lee
M, Avramova Z, Bennetzen JL: Nested retrotransposons in the
intergenic regions of the maize genome. Science 274: 765768
(1996).
38. Sapolsky RJ, Lipshutz RJ: Mapping Genomic Library Clones
Using Oligonucleotide Arrays. Genomics 33: 445456 (1996).
39. Schena M, Shalon D, Davis RW, Brown PO: Quantitative
monitoring of gene expression patterns with a complementary
DNA microarray. Science 270: 467470 (1995).
40. Schena M: Genome Analysis with Gene Expression Microar-
rays. BioEssays 18: 427431 (1996).
41. Schena M, Shalon D, Heller R, Chai A, Brown PO, Davis
RW: Parallel human genome analysis: microarray-based ex-
pression monitoring of 1000 genes. Proc Natl Acad Sci USA
93: 1061410619 (1996).
42. Schena M, Heller RA, Theriault TP, Konrad K, Lachenmeier
E, Davis RW: Microarrays: Biotechs discovery platform for
functional genomics. Trends Biotech 16: 301306 (1998).
43. Schena M, Davis RW: Parallel Analysis with Biological Chips.
in PCRMethods Manual, Academic Press, San Diego, in press
(1998).
44. Shalon D, Smith SJ, Brown PO: A DNA micro-array sys-
tem for analyzing complex DNA samples using two-color
uorescent probe hybridization. Genome Res 6: 639645
(1996).
45. Shoemaker DD, Lashkari DA, Morris D, Mittmann M, Davis
RW: Quantitative phenotypic analysis of yeast deletion mu-
tants using a highly parallel molecular bar-coding strategy.
Nature Genet 14: 450456 (1996).
289
46. Southern E: Method and apparatus for analysing
polynucleotide sequences. European Patent Specication
PCT/GB89/00460 (1989).
47. Southern EM: DNA chips: analysing sequence by hybridiza-
tion to oligonucleotides on a large scale. Trends Genet 12:
110115 (1996).
48. Winter G, Milstein C: Man-made antibodies. Nature 349: 293
299 (1991).
49. Wodicka L, Dong H, Mittmann M, Ho M-H, Lockhart DJ:
Genome-wide expression monitoring in Saccharomyces cere-
visiae. Nature Biotech. 15: 13591367 (1997).
50. Yershov G, Barsky V, Belgovsky A, Kirillov E, Kreindlin E,
Ivanov I, Parinov S, Guschin D, Drobishev A, Dubiley S,
Mirzabekov A: DNA analysis and diagnostics on oligonu-
cleotide microchips. Proc Natl Acad Sci USA 93: 49134918
(1996).