C H A P T E R 1 : I N T R O D U C T I O N T O M I C R O A R R AY A N A LY S I S

Chapter 1
I N T R O D U C T I O N T O M I C R O A R R AY A N A LY S I S

1.0 Introduction
Microarray analysis has emerged in the last few years as a flexible
method for analyzing large numbers of nucleic acid fragments in parallel.
Its origins can be traced to several different disciplines and techniques.
Microarrays can be seen as a continued development of molecular
biology hybridization methods, as an extension of the use of fluorescence
microscopy in cell biology, as well as a diagnostic assay using capture to
solid surface as a way to reduce the amount of analytes needed. The
convergence of ideas and principles utilized in these fields, together with
technological advancements in preparing miniaturized collections of
nucleic acids on solid supports, have all contributed to the emergence
of microarray and microchip technologies.

In molecular biology, analysis of nucleic acids by hybridization is a

universally adopted key method for analysis. Filter-based dot blot
analysis has been used for a long time as a convenient method for
analyzing multiple samples by hybridization. Classical gene expression
analysis methods such as Northern blotting, reverse transcriptase
polymerase chain reaction (RT-PCR) and nuclease protection assays,
are best suited for analyzing a limited number of genes and samples at
a) a time. By reversing the Northern blotting principle so that the labelled
moiety is derived from the mRNA sample and the immobilized fractions
are the known sequences traditionally used as probes, filter-based gene
expression analysis has enabled simultaneous determination of
b)
expression levels of thousands of genes in one experiment. Because of
the ease of use of these filter-based methods and their compatibility with
Fig 1. Comparison of a macroarray and general lab equipment, these macroarrays have been widely adopted for
microarray. A close-up of a filter macroarray (a)
gene expression studies (1). One disadvantage to using this method has
hybridized with 32P-labelled cDNA probe and a
microarray (b) hybridized with two different been the relatively large size and the autofluorescence of the membrane,
cDNA probes, one labelled with Cy™3 and the which prevents efficient use of multiplexed fluorescent probes and
other with Cy5. The array images are shown subsequently limits the number of samples that can be analyzed in
approximately to the same scale. each experiment (Fig 1).

● 1
 M I C R O A R R AY

Utilizing microscope slides to immobilize cells and chromosomes

precedes filter-based gene expression analysis, as well as proven methods
such as immunohistochemistry, immunocytochemistry and in situ
hybridization. By combining fluorescence analysis of multiplexed probes
with microscopy, fluorescent in situ hybridization (FISH) has enabled
detection of nucleic acids within cells and chromosomes, and has been
found useful in gene expression and genomic analysis.

These two methods of analysis were brought together by advancements

made in attaching nucleic acid sequences to a glass support. Borrowing a
technique from semiconductor manufacturing, photolithography was
used to synthesize oligonucleotides directly onto a glass support.
Separately, a procedure called contact printing was used to deposit
purified nucleic acid onto a slide surface (2). These methods have since
made it possible to miniaturize the macroarray experiment, so to speak,
by using microscope slides instead of membrane filters. In 1995 and
1996, the first papers in which the term ‘microarray’ was used in its
current meaning were published by the laboratory of Pat Brown at
Stanford University (3). The rapid adoption of this technique is
illustrated by the publication in 2001 of over 900 papers on the use of
microarray technique.

1.1 Principles of microarray analysis

Despite the variety of technical solutions that have been developed for
performing microarray analysis, all are miniaturized hybridization assays
for studying thousands of nucleic acid fragments simultaneously. All
microarray systems (Fig 2) share the following key components:

■ the array, which contains immobilized nucleic acid sequences,

or ‘targets’

■ one or more labelled samples or ‘probes’, that are hybridized with

the microarray

■ a detection system that quantitates the hybridization signal

hb 63-0048-49 ● 2
 C H A P T E R 1 : I N T R O D U C T I O N T O M I C R O A R R AY A N A LY S I S

Fig 2. Principles of microarrays.

Target preparation mRNA Probe preparation

and deposition

Labelled
probe
Array
spotter

Hybridization

Array scanner

Scanning and
data analysis

9 4 7
3 1
6
2 2

● 3
 M I C R O A R R AY

1.1.1 Nomenclature for microarrays

The terms ‘probe’ and ‘target’ are sometimes used interchangeably to
describe either the labelled sample or the immobilized nucleic acids. In
this handbook the immobilized nucleic acid is referred to as the target
and the labelled sample as the probe (Fig 3).

Fig 3. Target and probe. Targets are the

immobilized nucleic acids on the slide
Probe mixture
surface. A probe consisting of two identical
populations of nucleic acids labelled with Immobilized targets
on slide surface
different fluorescent dyes is shown.

1.1.2 Microarrays
Microarrays consist of a collection of nucleic acid sequences immobilized
onto a solid support so that each unique sequence forms a tiny feature,
called a ‘spot’ or ‘target’. These nucleic acids are obtained in numerous
ways, and there are different methods for depositing them onto microarray
slides (refer to chapter 3). The size of these spots varies from one system to
another, but it is usually less than two hundred micrometers in diameter. A
glass slide or glass wafer acts as the solid support onto which up to tens
of thousands of spots can be arrayed in a total area of a few square
centimeters (Fig 4).

Fig 4. A glass microarray. A standard and

mirrored glass microscope slide that contains 1.1.3 Probe labelling
several thousands of immobilized cDNA The microarray sample that is being analyzed, whether it is mRNA for
fragments. Because of the small amounts of
a gene expression study or DNA derived from genomic analysis, is
nucleic acid present and their tiny size, the
target spots are not visible with the naked eye. converted to a labelled population of nucleic acids, the probe. These
probes frequently consist of several thousands of different labelled
nucleic acid fragments. The complexity of microarray hybridization—
over 10 000 different labelled fragments interrogating up to 100 000
different immobilized sequences—is greater than that encountered in
other routine molecular biology experiments. Therefore, this
hybridization should be carried out under conditions that do not
promote annealing of non-complementary fragments.

Fluorescent dyes, and especially the cyanine dyes Cy3 and Cy5, have
been adopted as the predominant label in microarray analysis.
Fluorescence has the advantage of permitting the detection of two or
more different signals in one experiment. This has allowed investigators
to perform comparative analysis of two or more samples on one
microarray. It has also increased the accuracy and throughput of
microarray analysis over filter-based macroarrays, in which only one
radioactively labelled sample can be conveniently analyzed at a time.

hb 63-0048-49 ● 4
 C H A P T E R 1 : I N T R O D U C T I O N T O M I C R O A R R AY A N A LY S I S

1.1.4 Microarray hybridization

In a microarray hybridization, the labelled fragments in the probe are
expected to form duplexes with their immobilized complementary
targets. This requires that the nucleic acids are single-stranded and
accessible to each other. The number of duplexes formed reflects the
relative number of each specific fragment in the probe, as long as the
amount of immobilized target nucleic acid is in excess and not limiting
the kinetics of hybridization. Two or more samples labelled with
different fluorescent dyes can be hybridized simultaneously, resulting in
simultaneous hybridization taking place at each target spot. By measuring
the different fluorescent signals associated with each spot, the relative
abundance of specific sequences in each of the samples can be determined.

1.1.5 Scanning and data analysis

Microarray scanners typically contain two different lasers that emit light
at wavelengths that are suitable for exciting the fluorescent dyes used as
labels. A confocal microscope attached to a detector system records the
emitted light from each of the microarray spots, allowing high-resolution
detection of the hybridization signals.

Despite their small size, microarrays generate large quantities of data

even from a single experiment. As a typical experiment will involve
the use of several analyzed samples on replicate arrays, the use of
computerized data processing is necessary in order to handle the
amount of data generated and to gain maximum information from the
experiment. This can be achieved by specialized software that extracts
primary data from scanned microarray slide images, normalizes this
data to remove the influence of experimental variation, and finally
manipulates the data so that biologically meaningful conclusions can
be made.

● 5
 M I C R O A R R AY

1.2 Applications of microarray analysis

The versatility of microarray analysis is confirmed by its rapid emergence
as a general molecular biology analytical technique. Increasing numbers
of researchers within academic institutions and industrial laboratories
are now exploiting this technology in diverse biomedical disciplines.
Microarrays have not become a replacement to established techniques,
but more a novel, high-power approach to perform analyses that were
previously time consuming.

By using information derived from the several complete or near complete

genome sequences, including the human genome, it is now possible to
perform genome-wide experiments using microarray technology. This has
already been demonstrated for S. cerevisiae where all the expressed genes
are known. As microarrays can contain thousands of targets, both
characterized and uncharacterized, experiments can be conducted without
prior hypotheses. This combined with the millions of data points that are
possible to analyze in one experiment, microarray analysis has enabled
global analysis of biological processes. Gene expression analysis, genome
analysis, and drug discovery have been three of the main areas in which
microarray analysis has been applied so far.

1.3 Gene expression analysis

Gene expression analysis examines the composition of cellular messenger
RNA populations. The identity of transcripts that make up these
populations and their expression levels are informative of cell state and
activity of genes and, as the precursors of translated proteins, changes in
mRNA levels are related to changes in the proteome.

1.3.1 Traditional techniques

Traditional gene expression analysis has used techniques such as
Northern blotting, RT-PCR and nuclease protection assays. More
advanced methods—some of these include differential display, subtractive
hybridization, representational difference analysis, expressed sequence
tags, cDNA fragment fingerprinting, and serial analysis of gene
expression—have enabled the discovery of novel differentially expressed
genes (4). However, the technical challenges of these methods still limit
their use to the analysis of just a few samples at a time. Microarray
analysis, in contrast, allows the analysis of thousands of genes in
multiple samples with relative ease.

hb 63-0048-49 ● 6
 C H A P T E R 1 : I N T R O D U C T I O N T O M I C R O A R R AY A N A LY S I S

1.3.2 Gene expression analysis with microarrays

A typical microarray gene expression analysis experiment compares
the relative expression levels of specific transcripts in two samples.
Signal from
differentially One of these samples is a control and the other is derived from cells
expressed genes
whose response or status is being investigated. Each of these samples is
Genes expressed labelled with a different fluorescent dye, and equal amounts of the
at low levels
labelled samples are combined and hybridized with the microarray.
Equally expressed
The fluorescent signals corresponding to the two dyes are measured
genes independently from each spot after hybridization. After normalization,
the intensity of the two hybridization signals can be compared. Equal
signal from both samples suggests equal expression in both samples (Fig 5).
Fig 5. Dual color differential microarray
analysis. Dual color microarray hybridization Microarray analysis does not give information about absolute gene
signals are typically represented as false color
expression levels in the samples. This is because the intensity of the
images in which signals from one dye are
presented in red (Cy5 in this case) and
fluorescent signals is not only proportional to the number of hybridized
signals from the other dye in green (Cy3). If fragments but also to the length of these fragments and the number of
equal signal is obtained from a spot, it will fluorescent labels each fragment carries, i.e. labelling density. As these
appear yellow. Shades of green and red are determined by the unique nucleotide sequence of each gene and
denote differences in relative abundance in transcript, they will vary from gene to gene. If two samples have been
favor of one or the other sample. As the
labelled under similar conditions, the length and labelling density of
screen appearance of microarray images can
be easily manipulated, information gained specific transcripts will be similar in the two samples, making it possible
from such images can be misleading. to compare the relative abundance of the transcripts in the two samples.
A strong hybridization signal from microarray analysis does not
necessarily correspond to a highly expressed gene; it could be derived, for
example, from a gene that is expressed at a relatively low level but yields
long, highly-labelled probe fragments.

Gene expression analysis with microarrays has been applied to

numerous mammalian tissues, plants, yeast, and bacteria alike (1, 5,
6, 7, 8). These studies have examined the effects of treating cells with
chemicals, the consequences of over-expression of regulatory factors in
transfected cells, and compared mutant strains with parental strains to
delineate functional pathways. In cancer research microarrays have been
used to find gene expression changes in transformed cells and metastases,
to identify diagnostic markers, and to classify tumors based on their gene
expression profiles (9, 10, 11).

● 7
 M I C R O A R R AY

1.4 Genomic analysis

Microarrays are proving to be useful tools for genomic analysis.
Identification of new genes by examining nucleic acid sequences
derived from open reading frames has proved to be an efficient way
of annotating the human genome and facilitating the use of genomic
information for experimental purposes (12). Understanding of gene
regulation is advanced by elucidation of transcription factor gene
interactions. For example, by combining immunoprecipitation of
transcription factor-DNA complexes to microarray identification of
DNA fragments on a genomic microarray, it was possible to identify
functional regulatory elements in the yeast genome (13). Furthermore,
microarrays can be used for predicting splice variants of transcripts
and analyzing genomic fragments derived from genetic analysis methods,
such as genomic mismatch scanning and representational difference
analysis (14, 15). Oligonucleotide microarrays have been applied to
analysis of known single nucleotide polymorphisms (SNPs) and
mutations (16, 17). Samples can be sequenced using microarray
hybridization (18), thus providing convenient means for identifying
new genetic variants.

1.5 Drug discovery

As a typical drug discovery process takes several years and incurs high
costs, and only a few drug candidates result in approved drugs, methods
that increase the efficiency of the process and improve the chances of
developing effective drugs have been welcomed.

Microarrays have been found to provide useful information in the

different stages of the drug discovery process (8, 15, 19). Identification of
potential drug targets can be aided by elucidating metabolic pathways
by looking for co-expressed genes. The protein targets of drug treatments
can be identified by finding a protein that causes the same changes as a
drug when removed from cells. Once drug candidates have been
identified and selected, microarrays can be used to define their toxic
properties by examining expression profiles induced by drug treatments
(20). On the other hand, different function modes of drugs were
identified based on the gene expression changes they elicited (21).

hb 63-0048-49 ● 8
 C H A P T E R 2 : G E N E T I C C O N T E N T O F M I C R O A R R AY S

Chapter 2
G E N E T I C C O N T E N T O F M I C R O A R R AY S

2.0 Introduction
The genetic content of microarrays resides in the immobilized nucleic
acid sequences on the microarray. The identity of these sequences
determines what information can be obtained from array experiments
and how reliable this information is. As microarrays enable simultaneous
interrogation of up to tens or hundreds of thousands of targets with one
or more labelled probes, generation of accurate data demands that only
specific interactions result in detectable signals. Several strategies for
preparing the immobilized target nucleic acids for microarrays exist
(Fig 6). These nucleic acids can be synthesized directly on the microarray
or they can be purified cDNA clones, other DNA fragments or
oligonucleotides, which are deposited onto the array by a printing
process. This flexibility of using either partially characterized sequences
or defined oligonucleotides as targets has improved the application of
microarray analysis to different biological problems in a number of
species.

Cloned cDNA library Genomic DNA Gene database

A G T T C G A G A T T C C A
T C G A C G C A T G T G C A

PCR with PCR with

Oligonucleotide
vector-specific gene-specific
synthesis
primers primers

Fig 6. Sources of microarray target

sequences. Some of the common strategies
for obtaining targets for microarray analysis
are illustrated.

● 9
 M I C R O A R R AY

2.1 Oligonucleotides as genetic content

2.1.1 Printed oligonucleotide arrays

Oligonucletides can be attached onto microarrays by depositing modified
oligos onto a specially treated glass surface. The deposition can be
achieved with a variety of contact and non-contact printing methods.
See chapter 3, section 3.1.2 on common deposition methods for a
detailed overview of this process.

Depending on the source of oligonucleotide content, the length of these

oligonucleotides typically varies between 50–70 nucleotides (22, 23).
Different microarrays can be easily prepared with the deposition method
by choosing different sets of oligonucleotides for array printing. This
method is increasingly advantageous because customized microarrays
can be prepared in a researcher’s own laboratory using microarray
spotter equipment.

Regardless of the array fabrication method, the use of oligonucleotides

requires that the nucleotide sequences of the intended targets are known.
The publication of the human genome sequence as well as partial or
complete sequences of several other organisms has facilitated this task.
However, the accuracy of the information in the databases has a critical
impact on the quality of the arrays. Errors in sequence entries can result
in oligonucleotides that do not function in hybridization, because
nucleotide mismatches can prevent efficient hybridization from taking
place or non-complementary target strands are used by mistake. As more
and more genes are identified and their sequences elucidated, the power
of oligonucleotide arrays will increase.

2.1.2 Benefits of oligonucleotide arrays

Oligonucleotide targets have several benefits over cDNA targets.
■ Different parts of the same gene can be represented on the array. This
enables a more robust design of microarray experiments as the same
gene can be probed independently for the same information in the
same experiment.

■ Oligonucleotides can be designed to distinguish between alternative

splicing variants as well as different alleles. Oligonucleotides offer
precise control over the genetic composition on the arrays. With a
judicious choice of oligos, it is possible to discriminate between
related gene sequences and study different members of gene families
simultaneously.

hb 63-0048-49 ● 10
 C H A P T E R 2 : G E N E T I C C O N T E N T O F M I C R O A R R AY S

■ Oligonucleotide targets are readily available from commercial

manufacturers or synthesized by researchers.

■ The time and effort required to prepare oligonucleotides for array

printing is less than that required for preparing cloned targets by
molecular biology methods.

2.1.3 Design of oligonucleotides

The design of oligonucleotide targets should take into account factors
that influence the specificity and strength of hybridization with labelled
probes. The specificity can be estimated by comparing the oligonucleotide
sequence with known gene sequences. Predicting the strength of
hybridization is more difficult, however. Computer algorithms have been
developed for selecting target oligonucleotides. Some general rules for
oligonucleotide selection have been established:

■ Repeat sequences should be avoided, including polynucleotide

stretches, repetitive genomic elements, and palindromic sequences.

■ The chosen sequences should not be homologous to other genes, but

one short homologous stretch may still produce enough specificity in
hybridization (22, 24, 25).

■ The length of the oligonucleotide, its nucleotide sequence, as well as

the positions of mismatches in the oligo, all influence the behavior of
the oligo in hybridization.

■ It is important to choose a fairly even distribution of all four

nucleotides in the sequence.

■ Testing of oligonucleotide targets before including them on arrays can

help to eliminate sequences that will not perform well.

■ The use of computer algorithms may also facilitate the selection of

target oligonucleotides (23).

Target sequences may not be accessible to probe molecules near the

attachment site on the solid support. mSpacer sequences can be used to
increase hybridization efficiency. These are additional sequences added to
the oligo sequence to move it further away from the solid support (2).
40-atom long spacers were found to result in up to 150-fold increase in
hybridization signal on oligonucleotide arrays (26).

● 11
 M I C R O A R R AY

2.2 DNA fragments as genetic content

Development of mechanical microspotting methods and instruments,
which can be used to deposit nucleic acid solutions onto glass surface,
has enabled the use of cDNA clones and other DNA fragments as
microarray targets (27). These methods allow quick and adaptable
construction of microarrays that can be customized according to
different experimental needs.

2.2.1 Sources of DNA targets

The nucleic acid fragments used for microarray construction can be
derived from a number of sources. For gene expression microarrays the
fragments are typically derived from either cDNA clones or amplified
from exon sequences. Libraries of cDNA clones, expressed sequence tags,
clones isolated from subtraction libraries in which the number of highly
expressed sequences has been minimized, or PCR-amplified fragments
corresponding to open reading frames in genomic DNA have been used
as targets (28, 29). It is not always necessary to fully sequence the cDNA
clones before using them on microarrays, nor have prior information
about their expression in tissues. On the other hand, if the cDNA
sequence is known, it is possible to select areas of cDNAs that hybridize
with higher specificity to sequences derived from one gene only and
which do not hybridize with other related sequences. Many 3' untranslated
sequences can also contain repetitive genomic elements that will
compromise hybridization specificity and should not be present in
microarray targets.

As microarray analysis will usually involve the examination of thousands

of fragments in one experiment, acquiring and maintaining large
collections of nucleic acid fragments is labor-intensive and expensive.
While access to genetic content has previously limited, to some extent,
the adoption of microarray technology, the availability of ready-printed
microarray slides from both commercial companies and academic
consortiums has helped alleviate this problem.

2.2.2 Preparation of DNA targets

Before using for microarray spotting, DNA targets need to be amplified
and purified. Typically, PCR amplification is used, and universal primers
complementary for vector sequences simplify the process (30). It is
possible to amplify the target sequences starting from bacterial cultures,
purified plasmids, or RNA, if reverse transcription is performed before
amplification. With PCR, it is possible to amplify only part of the DNA
target or clone. This allows for the removal of sequences that might
compromise hybridization specificity. The amplified DNA needs to be
purified to remove enzymes, nucleotides, and buffer components, all

hb 63-0048-49 ● 12
 C H A P T E R 2 : G E N E T I C C O N T E N T O F M I C R O A R R AY S

of which can interfere with the microarray analysis if present in target

solution. Column purification methods, such as GFXTM PCR DNA
and Gel Band Purification Kit, can be used for this purpose. Whatever
methods are used for amplification and purification, it is most important
to verify that the amplified fragments are the right size, do not contain
other contaminating sequences and that they are present in known
quantities. Agarose gel electrophoresis is a convenient way of
performing this analysis. The Ready-to-Run Electrophoresis System,
which is capable of separating up to 96 samples in 5 min, is well suited
for this task (Fig 7).

a) Special care is needed when large collections of nucleic acid fragments

are handled simultaneously. It has been estimated that as much as
5–30% of clones in some collections are wrongly labelled or
contaminated with other sequences (31, 32). It is important that the
genes identified with microarray analysis are verified with other
techniques.

2.2.3 Desired properties of DNA targets

a)
An optimal length for DNA targets is between 300–800 nucleotides.
Fragments of this length can be efficiently attached to the microarray
Fig 7. Separation of PCR* fragments with
slide surface, where they form specific and stable hybrids. Figure 8 shows
Ready-to-Run Electrophoresis System.
that the retention of UV-immobilized double-stranded DNA targets
on aminosilane-treated microarray slides increases slightly with
increasing length of the molecules. With increasing length, however,
the concentration of DNA required to guarantee the deposition of a
sufficiently high number of target molecules within a spot increases.
This creates practical problems for using long DNA sequences as targets,
as it can become difficult to ensure that the targets are not limiting the
hybridization reaction. In order to obtain accurate results from
competitive microarray hybridization, the target molecules must be in
excess of the corresponding labelled probe molecules. Otherwise,
hybridization signals will be saturated.

% retention
100
90
80
Fig 8. Retention of microarray targets of 70
different lengths on aminosilane-treated 60
microarray slide. 50
40
30
20
10
0
200 300 400 500 600 700 800
Size of PCR insert

● 13
 M I C R O A R R AY

As was a requirement for oligonucleotide targets, DNA targets should

not contain repetitive sequences, and they should contain sequences that
are unique to one particular gene. Examination of potential cross
hybridization between related sequences, such as those derived from a
gene family, has revealed that more than 80% homology between targets
results in hybridization signals that are not specific for one gene.
However, even a lower degree of similarity was found to result in cross
hybridization, suggesting that interpretation of microarray data must
take the nature of the target sequences into account.

DNA targets do not need to be single-stranded. Spotting from denaturing

solutions is enough to render even double-stranded targets available for
hybridization. However, single-stranded DNA, which can be generated
by asymmetric PCR or by exonuclease digestion of partially protected
fragments (2), can also be used as targets in microarray analysis.

2.3 Control targets

Because microarray analysis is a complex process, there is a need for the
use of effective controls during the whole process. For gene expression
microarrays, the hybridization signal is influenced by a number of
variable factors, including the number of specific transcripts in the
labelled samples, the labelling method, the properties of microarray
printer pens, hybridization conditions, and slide surface chemistry.
Furthermore, variation in microarray signal is observed not only between
different slides but also between different replica targets spotted onto
different locations of the microarray slide. In this context, it is important
to be able to draw conclusions from the validity of microarray results
and to identify experiments that did not proceed optimally. Control
sequences included on microarrays are the key factor to aid in these
functions.

Different strategies have been devised for microarray control purposes.

These include the use of spiked exogenous sequences of known quantities
(25, 33) and housekeeping genes (34), i.e. genes whose expression is not
expected to change under experimental conditions.

A control system that consists of both control targets and RNA spikes
can monitor most aspects of the microarray process. A control strategy
adopted in the Lucidea™ Universal ScoreCard™ combines the use of
different types of control targets for spotting onto microarray slides and
spikes added to samples before labelling. Together these elements cover
aspects of slide printing, sample labelling, slide pretreatment, and
hybridization. In a typical experiment, up to 24 replicas of the ScoreCard

hb 63-0048-49 ● 14
 C H A P T E R 2 : G E N E T I C C O N T E N T O F M I C R O A R R AY S

sequences would be included on a slide printed with 12 pens. This

number of replicas allows calculation of quality indicators that report on
variation between different pens and spot sets as well as the overall
dynamic range and precision of signals.

In order to gain maximum information from the quality of microarrays,

positive, negative, ratio, dynamic range controls, and normalization
controls are typically used. Table 1 lists the properties and main utilities
of these control types. As a microarray hybridization involves thousands
of different nucleic acid fragments, it is important that sequences used as
controls are selected and functionally tested to avoid unspecific or cross
species hybridization. Negative controls, on the other hand, need to
represent different nucleic acid sequences to be able to capture the
occurrence of random hybridization events. Blank spots that contain
no DNA are useful as negative controls too, but are not sufficient on
their own.

The use of oligonucleotides as targets allows the use of mismatched

sequences as controls. By comparing the signal from the correct sequence
to that from the mismatched sequence, the reliability of each signal can
be assessed individually (24).

Table 1. Control target types.

Control type Composition Purpose
Positive control ■ Pooled genomic DNA ■ Control for labelling and
hybridization success
Negative control ■ DNA fragments derived from ■ Specificity of hybridization
unrelated species ■ Detection limit
Ratio control ■ Two different sequences spiked ■ Success of labelling and
into each sample before labelling hybridization
at different amounts ■ Color discrimination
Dynamic range control ■ Different sequences spiked into ■ Success of labelling and
samples before labelling at hybridization
different molar amounts ■ Color balance
■ Dynamic range of detection
■ Detection limit and saturation
of signal

● 15
 M I C R O A R R AY

hb 63-0048-49 ● 16
 C H A P T E R 3 : M A N U F A C T U R I N G O F M I C R O A R R AY S L I D E S

Chapter 3
M A N U F A C T U R I N G O F M I C R O A R R AY S L I D E S

3.0 Introduction
Microarray analysis is invariably performed on a glass slide, which
enables the performance of high-throughput miniaturized hybridization
assays with fluorescently labelled samples—a significant improvement
over the use of membrane support.

Microarray manufacture requires three distinct components:

■ production method

■ microarray slide

■ target genetic content

In this chapter the properties of deposition methods, instruments, and

microarray slides are discussed.

3.1 Production methods

3.1.1 Oligo synthesis

Two parallel approaches have been developed for the production of
microarray slides. Nucleic acid targets can either be synthesized directly
onto the microarray slide, or purified targets can be deposited onto a
solid surface that is capable of binding nucleic acids.

Adopting a photolithographic masking method used in the semiconductor

industry, oligo synthesis is begun by attaching chemically modified linker
groups, which contain photochemically removable protective groups,
onto the glass surface (39). By masking different predefined positions of
the glass at different steps, it is possible to synthesize different
oligonucleotides at different locations. Target synthesis proceeds in a
step-wise fashion using a different light-impermeable mask for each
round. In each step, the unprotected areas are first activated with light
which removes the light sensitive protective groups. Exposure of the
activated areas to a nucleoside solution results in chemical attachment of
the nucleoside to the activated positions. This process is then repeated by
using a different mask and a new nucleotide until all nucleotides have
been added to the oligo (35).

● 17
 M I C R O A R R AY

This method (Fig 9) produces arrays of small features that are anchored
at their 3' ends to the array surface. Each feature is made up of
oligonucleotides that all have the same nucleotide sequence. These
arrays have a high density: an area of 1.6 cm can contain up to 400 000
features. Additionally, multiple arrays can be synthesized simultaneously
onto a large glass wafer, further automating the manufacturing process.
The wafers are then cut into individual arrays in preparation for use.

Light deprotection Light deprotection

Mask Mask

C A T A T
Chemical
coupling A G C T G
O O O O O HO HO O O O T T O O O T T O O O T T C C O T T C C G
T– C– Repeat

Substrate Substrate Substrate Substrate Substrate Substrate

Fig 9. Microarray manufacturing using

photolithography.

3.1.2 Deposition
Using common deposition methods, purified nucleic acids are attached to
a modified glass slide. Typically, small volumes of nucleic acid solution—
nanoliters or picoliters—are transferred onto the glass slide. Deposition
methods are equally suitable for preparing microarrays containing
oligonucleotides, cDNA sequences, as well as genomic DNA. Deposition
methods are commonly used for preparing customized microarray slides.

The deposition chemistry involves a chemical reation between molecular

groups on the glass surface and the oligo, resulting in the formation of
covalent bonds that bind the ologonucleotide onto the array. There are
many different suitable attachment chemistries. One is the coupling of
amine-modified oligonucleotides to aldehyde slides. Another is the
derivatization of 5' phosphate groups with imidazole, followed by
reaction with the amine slide surface. A third is the use of bifunctional
cross-linkers to couple aminated oligos to aminated glass (2).

Many different techniques have been developed for the deposition

process, some of which are reviewed in this chapter. Regardless of the
technique used, however, the manufacturing process should meet several
criteria. Variation in the quantity of targets deposited, the shape of spots,
the regularity of the array pattern, and the carryover of targets could all
detrimentally affect the accuracy of microarray data.

hb 63-0048-49 ● 18
 C H A P T E R 3 : M A N U F A C T U R I N G O F M I C R O A R R AY S L I D E S

3.1.3 Requirements of microarray spotting methods and

instruments
Spot size and density
The microarray spots should be small and discernible from each other.
The spots should be deposited in grid-like fashion, at equal distances
from each other. It is important to immobilize the slides during printing,
as even the slightest movement can distort the microarray pattern.

Spot reproducibility
The spots should be of uniform size and shape, and they should contain
equal amounts of the target nucleic acids. This requires careful
calibration and matching of individual printing pens.

Environmental control
Environmental conditions can have a significant effect on the quality
of the spotted slides. Clean environment is important because dust
particles can become fixed onto slides, causing background signals
in microarray hybridization and spot finding problems during data
analysis. Controlling the humidity helps to avoid changes in sample
concentrations due to evaporation during printing and when spots are
drying. High humidity levels may cause spots to smear whereas low
humidity levels may cause evaporation from the sample plate. Under
high temperature conditions targets will dry rapidly at the outer edges
of the spot, thereby causing poor spot uniformity. This effect can cause a
donut-shaped spot morphology. A humidity between 10–70% has been
found to be most suitable for a microarray application.

Sample carryover
There must not be any carryover of previous target during the printing
process.

Throughput
The printing process should be fast to allow timely printing of slides. The
total time need for sample retrieval, printing, and washing of the printer
pens needs to be considered.

● 19
 M I C R O A R R AY

3.1.4 Non-contact deposition

Non-contact deposition has been adapted for microarray manufacture
from the modern ink jet printing industry. As the name implies, the
printing heads do not touch the surface of the microarray. Piezoelectric
Reservoir printing and syringe-solenoid methods are the two common variations of
this method.

■ In piezoelectric printing (Fig 10) the target solution is drawn into a

Piezoelectric crystal capillary that is in contact with a piezoelectric crystal. Application
Glass capillary of voltage to the crystal results in a slight conformational change,
squeezing the capillary. A small volume of sample is deposited onto
the glass surface. This method allows for very rapid spotting times.
Fig 10. Diagram of a piezoelectric
Very small volumes can be delivered, as the distortion of the crystal
microarray printing system.
shape can be accurately controlled. However, this deposition method
is prone to problems caused by air bubbles, which can cause poor
spot morphology.

■ Syringe-solenoid deposition (Fig 11) uses a syringe pump positive

displacement method to deposit nanoliter volumes onto a slide.
A syringe that provides the pressure source is connected to a
micro-solenoid valve. The sample is drawn up the dispensing tip
via the syringe. The system is pressurized and the opening of the
micro-solenoid valve allows small volumes of sample to be deposited
onto the surface. This system is not as rapid as that of piezoelectric
printing, and it is not able to deposit sub-nanoliter volumes; however,
deposition volumes are very precise and reproducible.

Fig 11. Diagram of a syringe-solenoid

microarray printing system.
Solenoid valve

Syringe

Pressure

hb 63-0048-49 ● 20
 C H A P T E R 3 : M A N U F A C T U R I N G O F M I C R O A R R AY S L I D E S

3.1.5 Contact deposition

In contact deposition, solid, hollow, or split-open pen designs are used to
transfer target nucleic acid onto the slide surface. These pens are dipped
into the target solution, a small volume of which adheres to the pen.
When the pen comes into contact with the slide surface, a fraction of the
nucleic acid solution on the pen is deposited onto the glass surface. One
sample uptake of the pen allows for several spots to be printed. For
achieving high throughput, several pens are used simultaneously, each of
which typically deposits a different nucleic acid solution. Successful spot
deposition is achieved by using pens that are quantitatively tested to
ensure performance, such as those made by Amersham Biosciences.

Contact deposition requires less target nucleic acid solution than the
non-contact methods and also results in smaller spots that can be packed
more densely on the microarray surface.

Table 2. Comparison of characteristics of different

microarray printing methods.

Contact Piezoelectric Syringe-Solenoid

pen printing printing printing
Microtiter plate
well volume
(microliters) 10–30 20 – 50 20 – 50
Uptake volume
(microliters) 0.2 – 1.0 5 – 10 5 – 10
Spot volume
(nanoliters) 0.5 – 2.5 5 – 100 0.1 – 10
Spot size
(nanometers) 75 – 250 250 – 500 125 – 175

● 21
 M I C R O A R R AY

3.2 Lucidea Array Spotter

Lucidea Array Spotter (Fig 12) is a new contact deposition microarray
spotter from Amersham Biosciences. The Lucidea Array Spotter is part of
the Lucidea platform of products offered by Amersham Biosciences for
microarray analysis. These products include Lucidea Array Spotter,
Lucidea SlidePro Hybridizer (see chapter 9), and Lucidea Universal
ScoreCard (see chapter 11). The key features of Lucidea Array Spotter are:

■ Patent pending, stainless steel capillary pens that conserve sample and
uniformly deposit picoliter volumes of target (Fig 13). From a single
sample uptake of less than 200 nl, up to 150 spots can be spotted in
duplicate, across each of 75 slides. The design of the pens (Fig 14)
minimizes clogging with target solution and simplifies washing after
each sample, resulting in no detectable carryover or mixing of samples
Fig 12. Lucidea Array Spotter.
during printing. To achieve good spot uniformity, the pens in each pen
set are quantitatively tested during manufacturing to ensure performance.

■ A newly designed five-step wash system eliminates the possibility of

sample carryover as shown with dye-labelled DNA testing (Fig 15).

■ Lucidea Array Spotter allows for monitoring and control of humidity

and temperature monitoring during spotting (Fig 16).

■ The target plates are kept in an area with minimized airflow to

reduce evaporation while the printing is in progress.

■ The spotting chamber is encased inside the enclosure of the instrument,

thus limiting the access of particulates during slide printing.

■ Several user-defined spotting modes are available to print arrays in

up to four replicates per slide. Lucidea Spotting Pens can handle
Fig 13. Lucidea Spotting Pen Set. spotting fluids with significantly different viscosity.

■ Control software integrates all aspects of spotter operation.

Fig 15. Wash system in Lucidea Array Spotter

Fig 14. Side view of the tip of Lucidea removes previous target solution efficiently,
Spotting Pen. Target solution is drawn resulting in virtually no detectable carryover of
by capillary action to the narrow sample during printing. The above experiment
opening in the tip of the pen. was performed with radiolabelled 32P-cDNA
spotted on a nylon membrane. The image on
the right shows results after the pens have
been washed, dipped into a sample blank,
and then spotted.
32P-CDNA 32P-CDNA
before wash after wash

hb 63-0048-49 ● 22
 C H A P T E R 3 : M A N U F A C T U R I N G O F M I C R O A R R AY S L I D E S

Uniform signals

Fig 16. Controlled temperature and humidity

result in relatively even spot intensities and
morphology.

Spots Scanned

3.3 Microarray slides

At present the most commonly used support for microarrays are
standard glass microscope slides that offer flat and rigid support with
low intrinsic background fluorescence. However, there are quality
differences between different manufacturer’s glass slides. Careful analysis
of slides before they are used for microarray printing is recommended.
Furthermore, it is very important to ensure that microarray slides are
absolutely clean.

3.3.1 Slide surface chemistries

Nucleic acids will not attach efficiently to an untreated glass slide.
Therefore, different surface chemistries have been developed to facilitate
the attachment of targets to the slide. These treatments not only enable
the binding of targets, but also determine the density of molecules that
can be attached per surface unit.

The uniformity and thickness of the surface coating on the slide is critical
for good quality microarray results, as this will influence spot uniformity
and morphology, DNA binding, as well as background signals from
microarray hybridization. Variation in slide coating can contribute to the
variation in microarray signals and decrease the resolution of a
microarray experiment. Uneven slide coating can also lead to poor
attachment of deposited nucleic acid, which may come loose during
microarray hybridization.

Commonly used slide surface modifications include the introduction of

aldehyde, amino, or poly-lysine groups onto the slide surface.
Aminosilane slides give highly consistent and reproducible data with high
signal to noise values, and they are most favorable for use in microarray
experiments.

● 23
 M I C R O A R R AY

3.3.2 Common slide types

Aldehyde slides
Amino-modified DNA can be attached to microarray slides that have
been modified with aldehyde groups (Fig 17). The amino group can
be introduced into DNA in a PCR amplification reaction using amino-
modified oligonucleotides. The aliphatic amine on the amino-modified
DNA acts as a nucleophile, attacking the carbon atom on reactive
aldehydes covalently attached to the surface of the slide. An unstable
intermediate is converted to a Schiff base through a dehydration reaction
(-H2O), and the DNA is bound to the surface. To minimize fluorescent
background, the unreacted aldehyde groups are reduced to non-reactive
primary alcohols by treatment with sodium borohydride (NaBH4).
Aromatic amines on the G, C, and A bases of naturally occurring DNA
can also react with aldehyde groups. The efficiency of this side reaction is
~0.01% for short oligonucleotides and ~10% for double-stranded PCR
products (36).

DNA

NH2 DNA DNA

NH H NH

H-C=O H-C=O H-C=O H-C=OH H-C=OH H-C=OH

sodium
borohydride

Fig 17. Attachment of amine-modified

DNA to aldehyde slide.

hb 63-0048-49 ● 24
 C H A P T E R 3 : M A N U F A C T U R I N G O F M I C R O A R R AY S L I D E S

Amine slides
Amine groups can be introduced onto microarray slides by treating
cleaned glass with aminosilane, such as 3-aminopropyltrimethoxysilane
(Fig 18). Vapor treatment of slides gives generally better results than
deposition by a dipping method (37). Unmodified DNA can be attached
to amine-modified slides, via interactions between negatively charged
phosphate groups on the DNA and the positively charged slide surface.
This interaction helps ensure denaturation of the DNA as well as increase
its binding affinity to the slide surface. UV treatment can be used to
further immobilize the DNA onto the slide surface. Attachment via
electrostatic interactions is suitable for binding DNA fragments that are
longer than 60–70 nucleotides. For attaching oligonucleotides to amine-
modified glass, chemical coupling methods must be used (2).

DNA
Fig 18. Attachment of unmodified DNA to an
amine-modified slide surface. The DNA binds
O O O O O O
to the surface of the slide via an electrostatic
P=O P=O P=O
interaction. The positive amines in the silane
coating will attract the negative phosphate O- O- O-
backbone of the DNA.
+ + +
NH3 NH3 NH3

Solid support

Poly-lysine slides
Treatment of the slide with poly-lysine creates a positively charged
surface to which unmodified DNA can bind via ionic interactions (38).

● 25
 M I C R O A R R AY

3.3.3 Reflective slides

A large proportion of the fluorescent light emitted from the hybridized
probe is scattered in all directions when using regular glass arrays. The
introduction of a reflective surface below the spotting surface enables a
significant amount of this scattered output to be directed towards the
detector, hence increasing the amount of signal detected by the system.
These reflective slides are constructed by adding a layer of aluminium
above the glass surface. Figure 19 shows a diagram of a reflective slide.

Sample layer
Fig 19. Diagram of the structure of a Silane layer
reflective microarray slide. Silicon dioxide layer
Aluminium layer
Glass slide

Signal enhancement is further achieved if an optimal thickness of silicon

dioxide is used as a spacer on top of the reflective layer (Fig 20). It has
been found that a thickness corresponding to 1/4 of the excitation
wavelength results in optimal signal enhancement for a particular dye. In
a typical microarray experiment two different dyes, such as Cy3 and
Cy5, are used. As these dyes have different excitation maxima, it is not
possible to enhance the excitation of both dyes simultaneously. Since the
fluorescence from Cy3 is already enhanced when the dye is bound to
molecules, it is more critical to increase the fluorescent signal from Cy5-
labelled molecules. Hence, Lucidea Reflective Slides have been designed
to specifically enhance Cy5 signals from microarray experiments,
resulting in better balanced signals from both dyes.

Incident light Fluorescence

sample
Fig 20. The enhancement of signal by a 1/4 Wavelength
reflective slide. Silicon dioxide layer
Aluminium layer
Glass slide

hb 63-0048-49 ● 26
 C H A P T E R 3 : M A N U F A C T U R I N G O F M I C R O A R R AY S L I D E S

3.4 Target nucleic acids

The third critical component in microarray manufacturing is the target
nucleic acid. Factors influencing the choice of target sequences are
Purified
described in Chapter 2.

Microarray targets must be available in high enough concentration

Unpurified
to allow a sufficient number of molecules to be deposited onto the slide.
The purity of target solutions is important for both the efficient
Purified + Taq attachment of nucleic acids to the slide surface and the availability of
the immobilized targets for hybridization. PCR-amplified targets must
Purified + dNTPs be purified to remove dNTPs, primers, DNA polymerase, buffer salts,
and detergents. Column chromatography methods are suitable for this
purpose. As shown in Figure 21, the presence of these compounds is
Purified + buffer
detrimental to the success of microarray hybridization.

Purified + primers
The targets, once attached to the microarray surface, are only available
for hybridization when they are present in a denatured, single-stranded
form. This can be achieved by spotting the targets under denaturing
Fig 21. Purification of PCR-amplified
conditions. Typically, targets are dissolved in high salt solutions such
targets. cDNA targets were amplified
as 3 × SSC, or in denaturing solvents such as DMSO.
with PCR, then purified with column
chromatography. The indicated reagents
were added to the purified target DNA
before spotting. Significantly decreased
3.5 Critical success factors for microarray
hybridization signals were observed from preparation
all targets containing impurities as ■ Always handle microarray slides in a clean environment.
compared with purified targets.
■ Never use gloves that contain powder as the powder will invariably
get onto the slides and cause background signals.

■ Never touch the array surface, only handle slides from sides.

■ Only use low fluorescence microscope slides for microarray

manufacturing.

■ Clean microarray slides efficiently before applying slide surface

treatment to them.

■ Verify the purity and concentration of targets before using them

for slide printing.

■ Handle target plates with care to avoid drying out or mixing of

targets.

■ Follow the instructions provided with the microarray spotter carefully.

■ Make sure that spotting is carried out under known and controlled
temperature and humidity.

■ Microarray slides have a limited shelf life, so prepare and use

microarray slides in a timely fashion.

■ Always store slides dessicated and protected from light.

● 27
 M I C R O A R R AY

hb 63-0048-49 ● 28
 C H A P T E R 4 : F L U O R E S C E N T L A B E L S I N M I C R O A R R AY A N A LY S I S

Chapter 4
F L U O R E S C E N T L A B E L S I N M I C R O A R R AY A N A LY S I S

4.0 Introduction
Most researchers performing microarray analysis prefer to use
fluorescent dyes as labels in these experiments as their use offers high
sensitivity of detection and enables detection of different dyes
simultaneously. Furthermore, fluorescent dyes do not carry the hazards
associated with radioactive markers. In this chapter, the general
properties of fluorescence and CyDyeTM fluorophores are discussed.

4.1 Definition of fluorescence

Fluorescence can be defined as the molecular absorption of light energy
(photon) at one wavelength and its re-emission at another wavelength.
Molecules that absorb light are known as chromophores. Molecules that
both absorb and emit light are known as fluorochromes, or fluorophores.

Light is a high frequency electromagnetic wave, and the energy of the

photon is inversely proportional to its wavelength (λ). Thus photons
towards the blue end of the spectrum, i.e. light photons with shorter
wavelengths, have a higher energy than those towards the red end of
light spectrum (Fig 22).
The process of fluorescence is a three-phase one, consisting of excitation,
the excited state, and emission.

Near ultraviolet Visible Near infrared

Fig 22. Light spectrum.

300 400 500 600 700 800
Wavelength (nm)

● 29
 M I C R O A R R AY

4.1.1 Excitation
Extinction coefficient (ε) is a measure for a fluorophore’s ability to
absorb light energy. When a photon of light energy (hvEX) of the
appropriate wavelength is absorbed by a fluorophore, an electron is
boosted to a higher, unstable excited energy state. The difference between
the ground state (S0) and the higher energy state (Sn) is a property of the
fluorophore; it is equivalent to the energy of the photon absorbed.
Because photons with shorter wavelengths have higher energy, the
shorter the wavelength of the absorbed photon, the higher the excited
energy state reached by the fluorophore. The wavelength at which the
fluorophore has maximum excitation is determined by the structural
properties of that fluorophore.

4.1.2 The excited state

The excited state typically lasts a fraction of a second. During this state
some of the energy absorbed may be dissipated in the form of vibrational
and rotational energy, often resulting in localized heating. The
fluorophore thus loses some of the energy it has absorbed from
excitation, prior to any fluorescent emission taking place. It is for this
reason that the quantum yield (φ) of a particular fluorophore (ratio of
the number of photons emitted to the number of photons absorbed) is
between 0 and 1.0. The quantum yield of a fluorophore can be greatly
influenced by the medium in which it resides. For example, unconjugated
Cy3 in phosphate buffered saline solution (PBS) has a quantum yield of
0.04; thus a large proportion of the energy absorbed by each dye
molecule is lost to its surrounding solution. However, in glycerol the
quantum yield increases more than ten-fold to 0.52.

4.1.3 Emission
Once the fluorophore has reached the lowest vibrational energy level
Sn
Vibration decay
within the electronic excited state (S1), the electron falls from the excited
S1 state to the ground state (S0). It is at this point in the decay process that
light is emitted at a specific wavelength (hvEM). Because some energy
Absorption Emission between excitation and emission has already been lost, the emitted
photon has less energy than the original photon absorbed by the
fluorophore (Fig 23). Therefore, the emitted light has a longer
S0
wavelength. The difference between the maximum excitation wavelength
and the maximum emission wavelength is known as the Stokes shift
Fig 23. Diagram illustrating the energy (hvEX – hvEM).
levels of the fluorescence process.

hb 63-0048-49 ● 30
 C H A P T E R 4 : F L U O R E S C E N T L A B E L S I N M I C R O A R R AY A N A LY S I S

4.1.4 Photobleaching
The fluorescent process is rapid (10-8 seconds) and cyclical, enabling the
fluorophore molecule to be excited repeatedly. It must be considered,
however, that the excited state of a fluorophore is generally more
chemically reactive than the ground state. In conditions of intense light,
the fluor may gradually lose its fluorescent properties, a phenomenon
known as "photobleaching". This results in lower fluorescent output
from the fluorophore after prolonged exposure to light or repeated
excitation. For more photosensitive dyes, such as fluorescein,
photobleaching may be a significant factor when using instrumentation
with high laser power. In contrast, as seen in Figure 24, CyDye fluors are
more resistant to photobleaching, which makes them more suitable for
multiple applications.

An excellent approach to reduce photobleaching is to maximize detection

sensitivity so that the excitation intensity can be reduced.

The brightness (intensity of output) of a fluorophore is proportional to

both the extinction coefficient (ε, the molecule’s ability to absorb light
energy) and the quantum yield (φ, the molecule’s efficiency to re-emit
light). Both of these are constants under specific static environmental
conditions. Consequently, fluorophores with very different characteristics
may give a comparable signal brightness. Fluorescein, for example, has a
molar extinction coefficient of ~70 000 and a quantum yield of ~0.9,
whereas Cy5 has values of ~200 000 and 0.3 respectively. However both
are of similar brightness.

Normalized
fluorescence
100
90
80
70
Fig 24. Photostability of fluorophores 60
Cy3
under "natural" light conditions. 50
Cy3.5
40 Cy3B
30 Cy5
20 Fluorescein
10
0
0 10 20 30 40 50 60

● 31
 M I C R O A R R AY

4.2 Applications of fluorescence

The fluorescent process, combined with appropriate imaging
instrumentation, enables a sensitive and quantitative detection method
for microarray expression analysis.

4.2.1 Benefits of fluorescent labelling

The particular Stokes shift properties of individual fluorophores make it
possible to separate excitation light from emission light with the use of
optical filters. Good spectral separation enables high sensitivity of
detection and yields low background.

By choosing fluorophores that have different pairs of excitation and

detection wavelengths, it is possible to excite and detect multiple dyes in
the same sample. This method enables multiplexing with dyes—labelling
two or more samples with different dyes that have different absorption
and emission spectra—and makes it possible to analyze several samples
simultaneously. Figure 25 shows the absorbance and emission spectra
of Cy3 and Cy5, the most widely used pair of fluorescent dyes in
microarray analysis. The minimal overlap between the Cy3 and Cy5
spectra demonstrates how it is possible to detect both dyes with minimal
cross talk (overlap) between their respective signals.

Another advantage of using fluorescent dyes as labels is that they are less
hazardous than radioactive compounds and offer significantly increased
stability, or longer shelf life. The availability of fluorophores conjugated
to many different chemical groups enables the labelling of nucleic acids,
proteins, lipids, and carbohydrate molecules. Furthermore, fluorescent
dyes can also be used as reporters to detect changes in pH, ion
concentrations, or dye environment.

Fluorescence
100
Cy3 excitation
90
Cy3 emission
80
Cy5 excitation
70 Cy5 emission
60
Fig 25. Excitation and emission spectra of 50
Cy3 and Cy5 dyes. 40
30
20
10
0
400 450 500 550 600 650 700 750 800
Wavelength (nm)

hb 63-0048-49 ● 32
 C H A P T E R 4 : F L U O R E S C E N T L A B E L S I N M I C R O A R R AY A N A LY S I S

4.2.2 Fluorescent quenching

Fluorescent quenching (Fig 26) causes decreased fluorescent signals. It
occurs when two or more fluorophores are in close proximity to each
other and the excitation energy is dissipated in interactions between the
adjacent dye molecules, rather than emitted as fluorescent light. Chemical
structures as well as photochemical properties of dyes determine the
distance at which two fluors will quench.

Quenching can occur when samples have been labelled too densely or
when too much labelled sample is used in hybridization. Over-labelling
not only results in the loss of linearity between fluorescent signal emitted
and the number of fluors present, but will, in extreme cases, reduce the
signal to levels that cannot be observed. In microarray analysis, quenching
may also occur when two probe strands come into close proximity of
each other. This is likely to be most apparent in the presence of highly
expressed transcripts, where a very large number of labelled molecules
are bound densely at a precise location. See chapter 6, section 6.2.4 for
further overview of this process.

High density labelling

CyDye

Fluorescence
A C C C T A G G A T G C G T G C G C T A A G A A C A T G G G G A T T Interaction & absorption

Lower density labelling

A C C C T A G G A T G C G T G C G C T A A G A A C A T G G G G A T T

Fig 26. A schematic representation of the effect of quenching

under varying labelling conditions. Fluorescent signal from the
same nucleic acid fragment labelled with high and low densities
is depicted. Green shows fluorescent emission, whereas purple
denotes dissipation of energy between closely spaced fluors
(pink circles).

● 33
 M I C R O A R R AY

4.3 CyDye fluorophores

4.3.1 Chemical structure of cyanine dyes

The cyanine, or CyDye, family of fluorescent dyes were first used in the
photographic industry as film sensitizers. They were subsequently
discovered for use in molecular biology applications when a CyDye was
coupled to a N-succinimide ester, to form the first dUTP (40, 41).

Fluorescence
100 Cy2
90 FluorX
80 Cy3
Cy3.5
70
Cy5
60
Cy5.5
Fig 27. The emission spectra of some 50 Cy7
CyDye fluors. 40
30
20
10
0
450 500 550 600 650 700 750 800 8500
Wavelength (nm)

This family of fluors consists of a chemically related group of dyes whose

emission spectra span the spectrum of visible light (Fig 27). CyDye fluors
X X share a core structure consisting of two heterocyclic indocyanine ring
structures joined by a polymethine bridge (Fig 28). Each dye differs in
N+ [ ]n N
the structure of this bridge. Adding pairs of conjugated C atoms to the
R R′ polymethine chain results in a wavelength shift of ~100 nm, for example
Cy3 (550 nm) and Cy5 (650 nm). An important modification of Cyanine
X = O, NR, C(CH3)2, S, Se
R and R′ = alkyl or aryl dyes is sulfonation of the indocyanine rings, as shown in Figure 29. The
n = 0, 1, 2, 3, 4 or 5 sulfonate acid groups increase the solubility of the dyes. In addition they
reduce aggregation of dye molecules, as the introduction of negative
Fig 28. Core structure of the cyanine dyes. charge makes the dye molecules repel each other. This results in a
decrease of fluorescence quenching.

The multiplexing properties of CyDye fluors were further increased by

synthesizing Cy3.5 and Cy5.5 (Fig 30). The addition of benzene rings
shifts their absorbance and emission spectra up by approximately 30 nm
to the red. Two additional sulfonate groups are needed to increase
solubility in order to overcome aggregation due to benzene rings.

hb 63-0048-49 ● 34
 C H A P T E R 4 : F L U O R E S C E N T L A B E L S I N M I C R O A R R AY A N A LY S I S

HO3S SO3H

N+ N
Fig 29. Structure of Cy3 NHS ester
(3 carbon bridge) and Cy5 NHS ester
(5 carbon bridge ).
Cy3 NHS ester O
Excitation max = 548 nm
Emission max = 562 nm O O N
O

HO3S SO3H

N+ N

Cy5 NHS ester O

Excitation max = 646 nm
Emission max = 664 nm O O N
O

Fig 30. Structures of Cy3 and HO3S SO3H

Cy3.5 NHS esters.
N+ N

Cy3 NHS ester O

Excitation max = 548 nm
Emission max = 562 nm O O N
O

SO3H SO3H

HO3S SO3H
+
N N

Cy3.5 NHS ester O

Excitation max = 581 nm
Emission max = 596 nm O O N
O

● 35
 M I C R O A R R AY

4.3.2 Fluorescent and chemical properties of CyDyes

Cy3 and Cy5 have become the most commonly used pair of CyDyes.
This can be attributed to the following factors:

■ The photostability of Cy3 and Cy5 is higher than that of other widely
used dyes (Fig 24), making the use of CyDye fluors more practical.

■ Cy3 and Cy5 are bright dyes that give strong fluorescent signals.

■ The good spectral separation of Cy3 and Cy5 means that each can be
excited at a different wavelength and their emissions can be detected
separately. This requires the use of two different lasers, typically a
532 nm and 633 nm laser for Cy3 and Cy5, respectively. By choosing
optical filters that only collect emitted light from part of the spectra,
Cy3 and Cy5 signals can be measured with minimal overlap.

■ The fluorescence of Cy3 and Cy5 is minimally affected by factors

such as pH and the presence of DMSO. Additionally, these dyes can
withstand temperatures and conditions normally encountered in
molecular biology applications.

4.3.3 Handling of CyDye fluors

Correct handling of CyDye and CyDye compounds helps to conserve
their fluorescent properties. When handling these substances, observe the
following precautions:

■ Minimize the exposure of CyDye compounds to all light sources.

■ Store CyDye in amber tubes, in light-safe containers, or wrapped in

aluminium foil to protect them from light during storage.

■ Take CyDye out of their protective container only when ready to

use and return to dark immediately after use.

■ If using CyDye nucleotides, prepare single use aliquots to avoid

freeze thaw.

■ Protect CyDye NHS esters and other CyDye labelling conjugates

such as CyDye Direct™ from moisture by storing dessicated.

■ Do not store CyDye in solutions containing concentrated

amines. Phosphate buffer or water is preferred over Tris-buffers.

hb 63-0048-49 ● 36
 C H A P T E R 4 : F L U O R E S C E N T L A B E L S I N M I C R O A R R AY A N A LY S I S

Table 3. Photochemical properties of selected CyDye fluors.

CyDye Abs. Max. Em. Max. Molar extinction coefficient
(nm) (nm) (M-1cm-1)
Cy2 489 506 ~150 000

Cy3 550 570 150 000

Cy3.5 581 596 150 000

Cy5 649 670 250 000

Cy5.5 675 694 250 000

Cy7 743 767 ~250 000

● 37
 M I C R O A R R AY

4.4 Use of fluorescent dyes in microarray analysis

Multiple labelling strategies have been developed for incorporating
fluorescent labels into microarray probes. These techniques are discussed
in chapter 6. The photochemical properties of fluorescent dyes, especially
the positions of their excitation and emission peaks, determine the
specification of scanning instruments, the laser type, and the choice of
emission filters that are required to separate and detect fluorescent
signals from particular dyes. The popularity of pairing Cy3 and Cy5 as
labels has led to the development of microarray scanning instruments
that are optimally suited for detection of these dyes.

General requirements for detecting fluorescent signals in microarray

analysis are:

■ High enough resolution to image a large area (2 × 6 cm) in a short

period of time.

■ At least two fluorescent spectra must be distinguished to

accommodate differential gene expression experiments using two
fluorescent dyes.

■ The wide range of message abundance levels requires an instrument

with a low fluorescent detection threshold to allow detection of rare
messages and wide linear dynamic range to measure the more
abundant messages.

■ The entire area of the microarray must be scanned uniformly to

ensure reproducibility.

4.4.1 Characteristics of fluorescent detection

The energy of the emitted fluorescent light is a statistical function of the
available energy levels in the fluorochrome, but it is independent of the
intensity of the light used to excite the fluorophore. In contrast, the
intensity of the emitted fluorescent light varies with the intensity and
wavelength of incident light and the brightness and concentration of the
fluorochrome. When more intense light is used to illuminate a sample,
more fluorochrome molecules are excited, and the number of photons
emitted, i.e. the number of electrons falling to the ground state, increases.
If the illumination is very intense, all the fluorochrome molecules are in
the excited state most of the time—this is called saturation.

hb 63-0048-49 ● 38
 C H A P T E R 4 : F L U O R E S C E N T L A B E L S I N M I C R O A R R AY A N A LY S I S

When the illumination wavelength and intensity are held constant, as

with the use of a controlled laser light source, the number of photons
emitted is a linear function of the number of fluorochrome molecules
present. At very high fluorochrome concentrations, the signal becomes
non-linear because the fluorochrome molecules are so dense that
excitation occurs only at or near the surface of the sample. Additionally,
some of the emitted light is reabsorbed by other fluorochrome molecules
(self-absorption).

The amount of light emitted by a given number of fluorochrome

molecules can be increased by repeated cycles of excitation. In practice,
however, if the excitation light intensity and fluorochrome concentration
are held constant, the total emitted light becomes a function of how long
the excitation beam continues to illuminate those fluorochrome
molecules (dwell time). If the dwell time is long relative to the lifetime of
the excited state, each fluorochrome molecule can undergo many
excitation and emission cycles.

Measuring fluorescent light intensity (emitted photons) can be

accomplished with any photosensitive device. For example, for detection
of low-intensity light, a photomultiplier tube (PMT) can be used. This is
simply a photoelectric cell with a built-in amplifier. When light of
sufficient energy hits the photocathode in the PMT, electrons are emitted,
and the resulting current is amplified. The strength of the current is
proportional to the intensity of the light detected. The light intensity is
usually reported in arbitrary units, such as relative fluorescence units (RFU).

If fluorescent samples are detected with a system that uses an intense

light source to excite the dyes, photobleaching can occur. This irreversible
destruction of an excited fluorophore will result in a loss of brightness,
or emission light intensity. As photobleaching is a consequence of
excitation, altering the characteristics of detection, such as increasing the
voltage of PMT to allow more sensitive detection, will not cause it.
Microarray slides hybridized with Cy3- or Cy5-labelled probes can be
scanned several times with commercial microarray scanners without
considerable loss of fluorescent emission.

● 39
 M I C R O A R R AY

hb 63-0048-49 ● 40
C H A P T E R 5 : P R E P A R AT I O N O F R N A S A M P L E S F O R M I C R O A R R AY A N A LY S I S

Chapter 5
P R E P A R AT I O N O F R N A S A M P L E S F O R M I C R O A R R AY
A N A LY S I S

5.0 Introduction
There are multiple steps involved in isolating and preparing RNA
samples to be used for microarray analysis. Discussed in this chapter
are factors affecting the quality of analyzed RNA, protecting RNA
samples from contamination and degradation, purifying RNA, and
finally, characterizing the purified RNA.

5.1 Factors affecting RNA quality

The quality of information obtained from microarray experiments
is primarily dependent upon the quality of RNA analyzed. Ideally, the
RNA should be devoid of DNA, protein, carbohydrates, lipids, and
other compounds. The presence of these substances will not only make it
difficult to correctly estimate the amount of RNA present in the sample,
but can contribute to fluorescent background signals in the array
hybridization. Degradation of RNA, whether by enzymatic or chemical
means, results in the loss of gene expression information from the
labelled samples. Furthermore, if the quantity or quality of the two
samples being compared differ, misleading conclusions can be made.
Compared with DNA, RNA is relatively unstable and can be degraded
either enzymatically, chemically or physically. Ribonuclease enzymes
(RNAses) degrade RNA into short oligonucleotides in a rapid reaction.
They are present in all cells and can be derived from a variety of
environmental sources, such as the hands, skin, and hair; bacterial or
viral contamination of solutions; or remnants of previous reagents in lab
glassware. Inactivation of these enzymes is difficult; therefore it is
essential that precautions are taken to ensure that RNA degradation is
minimized during isolation, purification, and storage.

● 41
 M I C R O A R R AY

5.2 Protecting RNA from degradation

RNA can be degraded if it comes into contact with any source of RNAse.
Discussed in this section are the many ways in which RNA samples can
become contaminated and how to protect them from subsequent
degradation if contamination occurs.

5.2.1 Protecting RNA from contamination by environment

An easy way of accidentally contaminating RNA preparations is by
transferring nucleases from the investigator’s hands, skin, or hair to the
sample. The transfer of RNAses can also take place via equipment, bench
surfaces, and door handles. Common molecular biology protocols, such
as plasmid preparation, involve the use of high amounts of RNAses and
can lead to RNAse contamination if handled in the same location. In
order to protect RNA samples from contamination, the following
precautions are recommended:

■ Wear disposable gloves while handling RNA samples and while

preparing other solutions for RNA work. Use a clean pair of gloves if
potential contamination of any kind occurs.

■ Perform RNA work in a separate area of the laboratory where no

RNAses are allowed. Before RNA handling, clean the bench surface
with a detergent, such as RNAse ZAPTM (Ambion).

■ Lab equipment, such as tissue homogenizers, non-disposable

centrifuge tubes, gel tanks, and trays that can come into contact with
RNA, should be reserved for RNA work. If this is not possible, large
equipment or those items made from materials that do not withstand
autoclaving temperatures should be cleaned with RNAse inactivating
detergents.

5.2.2 Protecting RNA during preparation of reagents

In order to obtain good quality RNA and to maintain its integrity during
subsequent analysis, all reagents should be prepared so that they do not
contain any traces of RNAses.

■ Use only disposable plastic tubes and pipette tips for RNA work.
Clean plastic-ware should be baked at 120 °C for at least 20 min to
reduce RNAse (and other nuclease) contamination. Before baking,
pack centrifuge tubes into glass beakers and cover them with
aluminium foil. Additionally, wrap tip racks in foil to keep the
baked items free of contamination after treatment.

■ Set aside reagents for use in RNA work only. Always use baked or
disposable spatulas and weighing trays for measuring out reagents.

hb 63-0048-49 ● 42
C H A P T E R 5 : P R E P A R AT I O N O F R N A S A M P L E S F O R M I C R O A R R AY A N A LY S I S

■ Treatment with diethyl pyrocarbonate (DEPC) renders RNAses

inactive and can be used to clean solutions and labware before use
in protocols involving RNA. As DEPC is toxic it should be handled
with appropriate care. Prepare 0.2% (v/v) solution of DEPC in water
and soak clean equipment and glassware in the solution for at least
1 h. Rinse the treated equipment with sterile water and let dry in a
clean place where no further contamination is encountered. Finally,
autoclave the equipment/labware in sealed autoclave bags. This is
necessary as autoclaves become contaminated by RNAses from
other autoclaved materials or from the water used in the autoclaving
process.

■ Reagent solutions can be treated by adding 0.2% (v/v) DEPC into

the solution. After the solution has been left to stand for a couple of
hours, it can be autoclaved to remove DEPC by heat degradation.
Solutions containing Tris-buffers cannot be prepared in this way.
However, water that has been treated with DEPC and then autoclaved,
can be used to make up any Tris-buffers.

■ Use DEPC-treated water in all RNA protocols.

■ Store sterilized equipment and solutions unopened in a clean

environment, away from any potential sources of contamination.

5.2.3 Protecting RNA during isolation process

Isolation of RNA requires disruption of cellular structures, which leads
to the release of RNAses. Rapid degradation of RNA follows if these
RNAses come in contact with RNA during cell homogenization. The
following protocols are recommended:

■ Use strong denaturing agents such as guanidium isothiocyanate

(GITC) or guanidium hydrochloride in RNA extraction protocols as
they will denature RNAses (and other enzymes) efficiently and
quickly.

■ RNAse inhibitor, which is a placental protein that inactivates

RNAses by binding to them, can also be used during RNA isolation.
However, the effectiveness of the RNAse inhibitor can be affected by
the composition of extraction solutions, and denaturation of the
inhibitor can release active RNAses.

■ The longer the time elapsed between collecting cells or tissue and
preparing a denatured homogenate, the more chance there is for RNA
degradation to take place. Furthermore, during this time, changes in
gene expression can also take place as cells react to the change in their
environment. Therefore it is good practice to work as quickly as
possible when preparing biological samples for RNA extraction.

● 43
 M I C R O A R R AY

■ Harvest cultured cells directly into a solution that contains GITC to

ensure minimal degradation. Pipette cell lysis solution directly onto a
washed cell monolayer. This will lead to immediate cell lysis. Scrape
the cells off and transfer into a centrifuge tube. The lysate can be
homogenized by drawing it several times through a needle with a
syringe or by vigorous mixing until the lysate becomes clear and
homogeneous. Cell lysates prepared in this way can be stored frozen,
preferably at -70 °C, until RNA isolation protocol is completed.

■ Freeze tissue samples rapidly in liquid nitrogen, as this helps to

minimize RNA degradation. It is advisable to cut large samples into
small pieces before freezing them as this will greatly facilitate their
subsequent use for RNA extraction. These samples must be stored
frozen, preferably in liquid nitrogen. Thawing of the samples, at any
stage, will result in RNA degradation as the freezing process causes
some cellular damage. For best results, mechanically pulverize frozen
tissue samples while they are kept cold with liquid nitrogen. This can
be done with a pestle and mortar. The cold powder should then be
dissolved in lysis buffer containing GITC to prepare a clear
homogenate.

5.2.4 Protecting RNA during storage and handling

RNA should be protected during storage and handling. The following
protocols are recommended:

■ RNA is not stable at alkaline pH. This property can be exploited

to degrade RNA selectively from DNA. However, if the aim is to
maintain RNA intact, all solutions that come into contact with RNA
should be neutral or mildly acidic. As the pH of laboratory water can
vary, using dilute buffers, such as TE pH 7.6, is recommended.

■ RNA degradation is more rapid at high temperatures. For long term

storage, storing RNA solutions at -70 °C is recommended. All RNA
solutions should be stored on ice while working with them and kept
thawed for the minimum time needed. As with other nucleic acids,
avoiding freeze-thaw cycles is important. If large amounts of RNA are
prepared at one time, it is recommended that the purified nucleic acid
is aliquoted for storage and only the required number of aliquots are
thawed at any time.

hb 63-0048-49 ● 44
C H A P T E R 5 : P R E P A R AT I O N O F R N A S A M P L E S F O R M I C R O A R R AY A N A LY S I S

5.3 Choosing an RNA isolation method

5.3.1 Methods for purifying total RNA

Numerous RNA isolation methods have been published and a variety of
RNA isolation kits are available (42, 43). The key criteria in choosing a
method should be to achieve a high yield of intact and pure RNA.
Obtaining long RNA molecules can be problematic, and the choice of
purification method will influence results (44).

5.3.2 Critical factors in RNA isolation

The main factors in isolating good RNA are the composition of the cell
lysis buffer, the method of cell disruption, and the method used for
separating RNA from protein, DNA, and other compounds. The nature
of the biological sample is also relevant as some RNA isolation methods
may be more suitable for certain tissues. Whereas soft tissues or cultured
cells disrupt quickly and efficiently, and methods using mild cell lysis
buffers can give good results, harder tissues containing large amounts of
connective tissue, such as muscle, will require the use of strong chaotropic
agents such as GITC. Amersham Biosciences RNA extraction kits, such
as QuickPrepTM Total RNA Extraction Kit, RNA Extraction Kit, and
QuickPrep Micro mRNA Purification Kit, all contain either guanidium
hydrochloride or guanidium isothiocyanate in the lysis buffer, and are
suitable for use with a wide variety of cells and tissues. In general, RNA
preparations that use chaotropic agents in the lysis buffer tend to give the
best results. However, the protocols are more laborious, involve the use
of toxic chemicals, and take longer.

Choosing an efficient cell/tissue disruption method for RNA extraction

is important. If cell lysis is not complete, the yield of RNA will be
compromised. Mechanical disruption using tissue homogenizers,
vigorous vortexing, needle and syringe, sonication, pestle and mortar,
and bead milling are among commonly used methods. The main
considerations in choosing a disruption method are the amount of each
sample and the time that is needed for preparation of a cell lysate.

A variety of techniques can be used to differentially separate RNA from

other cellular compounds. Precipitation with lithium chloride, acid
extraction phenol/chloroform, binding to an absorbent matrix or cesium
chloride gradient centrifugation can be used successfully to purify RNA.
However, the quality and quantity of RNA obtained with these methods
can vary.

● 45
 M I C R O A R R AY

The presence of contaminating DNA in total RNA samples can cause

problems in microarray analysis. Most labelling methods will label both
RNA and DNA with equal efficiency. Labelled DNA can hybridize with
microarray targets and can lead to high level hybridization signals that
are not derived from transcripts. RNA cannot be quantitated separately
from DNA, so an accurate estimation of the amount of RNA in
contaminated RNA preparations is impossible. Therefore, it is advisable
to treat total RNA preparations to remove DNA contaminants before
using the RNA for labelling. This can be achieved with DNase I
treatment or by using CsCl gradient centrifugation to separate RNA
from DNA.

5.3.3 Isolating RNA from difficult samples

The nature of some biological samples may necessitate the use of
modified RNA extraction strategies to avoid contamination of RNA
samples with other compounds. For example, plant tissues can contain
polyphenols and polysaccharides. Precipitation with polyvinyl
pyrrolidone can be used to remove these substances from RNA
preparations. The hardness of cell walls and outer protective structures
can also pose a problem. Freezing the samples followed by mechanical
grinding may be necessary to efficiently disrupt cell walls and to release
cellular RNA. In some cases, as with yeast that has cell walls that can
form capsules, disruption of cellular structures increases access to RNA.
Digestion with enzymes, such as zymolase, can be used to weaken the
cell walls before mechanical disruption with bead milling to lyse the cells.
Similarly, isolation of bacterial RNA benefits from the use of enzymes
that digest and weaken outer supportive structures. Lysozyme treatment
followed by mechanical bead milling is a suitable approach for disrupting
bacterial cells for RNA extraction.

5.3.4 Purification of eukaryotic mRNA

Most eukaryotic transcripts contain a poly-A tail, and this property
can be exploited to separate transcripts from other RNA molecules.
Incubation of total RNA with oligonucleotides containing a poly-T
sequence, otherwise called oligo(dT), will result in the hybridization
between the poly-A tail of transcripts and the oligonucleotides. By
attaching the oligo(dT) to a solid support, it is possible to specifically
separate transcripts away from other RNA molecules. QuickPrep Micro
mRNA Purification Kit uses oligo(dT) cellulose for extraction of mRNA.

hb 63-0048-49 ● 46
C H A P T E R 5 : P R E P A R AT I O N O F R N A S A M P L E S F O R M I C R O A R R AY A N A LY S I S

Although purification of mRNA lengthens the sample preparation

protocols, it provides several benefits for microarray analysis:

■ Probes prepared from mRNA usually give higher signal to noise values
on arrays than probes prepared from similar amounts of
total RNA.

■ Total RNA preparations are more likely to contain compounds other

than RNA, which can interfere with the labelling or hybridization
steps.

■ The yield of labelled cDNA is higher from mRNA than from total
RNA, because alternative priming strategies that use oligo(dT) can be
used.

■ It is easier to prepare labelled probes corresponding to the 5' ends

of transcripts from mRNA populations than from total RNA.

5.3.5 Purification of prokaryotic mRNA

Purification of mRNA from prokaryotes is difficult as most
transcripts lack poly-A tails. However, strategies have been developed to
polyadenylate 3' ends of bacterial transcripts in crude extracts.

Enrichment for bacterial mRNA can also be achieved by selective

degradation of ribosomal RNA. By synthesizing first-strand cDNA
selectively from ribosomal RNA with the use of specific primers,
RNAse H can be used to degrade the RNA strand in the resulting
double-stranded hybrid. DNAse I digestion can then remove the DNA
strand, resulting in the enrichment of transcripts. Up to 80% enrichment
can be achieved with this method (45).

● 47
 M I C R O A R R AY

5.4 Characterization of purified RNA

Microarray gene expression data is derived from a comparison of
Kb hybridization signals obtained from two samples. Accurate results are
only obtained when the two samples are of the same quality. Comparison
of a partially degraded RNA sample to an intact sample can artificially
4.0
show some genes as being more highly expressed in the better quality
2.0
sample. Likewise, differences in the amount of RNA in different samples
0.5 can also give biased data. Therefore, it is highly advisable to verify both
the quantity and quality of the RNA or mRNA sample before their use in
1 2 3 4 5
microarray analysis.

Fig 31. Schematic illustration of typical 5.4.1 Measuring the amount of RNA
results obtained from analyzing the The amount of RNA can be quantified by measuring the absorbance of
quality of RNA with a denaturing RNA solution at 260 nm. Pure RNA solution that contains 40 µg of
agarose gel. Denatured RNA samples
RNA per mL will give absorbance of 1 AU. This method works well with
were electrophoretically separated in
an agarose gel and visualized with UV clean RNA samples that are devoid of other contaminating substances.
light after ethidium bromide staining. Unfortunately, this is rarely the case with total RNA samples purified
Lane 1. Intact total RNA with both with simple methods. Proteins and DNA or other compounds, such as
ribosomal RNA bands present as sharp those released from some affinity chromatography columns, will absorb
and bright bands. High abundance
at 260 nm. This absorbance is indistinguishable from that of RNA and
transcripts can be discerned as distinct
will give an artificially high estimate for the amount of RNA in the
bands. Lane 2. Partially degraded total
RNA. Although ribosomal bands are sample. By measuring absorption spectra from 200 to 350 nm, some
still visible, the average size of RNA is conclusions on the purity of RNA can be made. However, the presence of
smaller, and no distinct transcript DNA in the sample can still go undetected. Therefore, it is important to
bands are visible. Lane 3. Badly use an RNA isolation method that specifically removes DNA from the
degraded total RNA. Most of the RNA
purified sample.
is shorter than 1 kb, and ribosomal
RNA bands appear as smears. Lane 4.
Intact mRNA. Fragments longer than 4
kb should be visible and abundant
5.4.2 Verifying the quality of RNA
transcripts should appear as distinct Agarose gel electrophoresis performed under denaturing conditions can
bands. Lane 5. Degraded mRNA. The be used to analyze the quality of RNA. Gels containing formaldehyde
transcripts appear as a fast migrating
have been traditionally used for this purpose, but denaturation of RNA
smear.
hairpins by glyoxal has gained popularity, as this method does not
involve the use of large quantities of harmful chemicals (42, 43, 46).
Both methods, followed by staining of the gels with nucleic acid
binding stains, such as ethidium bromide or Vistra Green™, are useful
for observing overall differences in RNA quality. It is relatively easy
to see if a sample is badly degraded as the ribosomal bands appear as
smears. However, it may be difficult to detect more subtle degradation of
transcripts unless large amounts of sample are used. Figure 31 shows a
schematic illustration of typical results from RNA gel electrophoresis.
Transferring the gel onto a membrane for Northern blotting analysis
can give more precise information, as the status of specific transcripts
in different samples can be analyzed.

hb 63-0048-49 ● 48
C H A P T E R 5 : P R E P A R AT I O N O F R N A S A M P L E S F O R M I C R O A R R AY A N A LY S I S

Reverse transcriptase PCR (RT-PCR) offers a convenient, fast, and

versatile method for obtaining information regarding the quality of RNA
preparations from small sample amounts. First-strand cDNA synthesized
with oligo(dT) priming can be used as PCR template.

By choosing specific primer pairs, it is possible to determine whether

different transcripts are intact in the samples. Performing PCR
amplification for a limited number of cycles, or with real time detection,
makes it possible to estimate the relative amount of specific transcripts in
different samples. By amplifying transcripts derived from genes whose
expression is not expected to vary under experimental conditions, it is
also possible to compare the amount of mRNA in the different samples.
By choosing several pairs of primers from one preferably long transcript,
and targeting its 5' central and 3' regions, it may be possible to observe
partial degradation of samples. Performing PCR with primers that are
derived from gene introns can reveal the presence of genomic DNA.

5.5 General recommendations for preparing

RNA samples for microarray analysis
In conclusion, the following general recommendations for preparing
RNA for microarray analysis are given:

■ Minimize the degradation of RNA at all handling stages.

■ Choose an RNA purification method that gives good yields of

pure and intact RNA from your samples, even if this means using
a complicated protocol.

■ Measure the amount of RNA before using it for microarray

labelling.

■ Verify the quality of the RNA before using it for microarray

labelling.

■ If possible, purify mRNA for use in microarray analysis.

■ Prepare all the samples for microarray analysis with the same
protocol.

● 49
 M I C R O A R R AY

hb 63-0048-49 ● 50
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

Chapter 6
S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.0 Introduction
In differential gene expression analysis two or more RNA samples are
compared to identify differences in the abundance and identity of the
transcripts they contain. In order to convert the information contained in
the transcript populations into a form that can be hybridized with
microarrays and subsequently detected, the transcript populations need
to be labelled. This can be achieved using different methods; an ideal
method retains both the information carried by the identity of the
transcripts as well as their relative abundance in the sample.

6.1 The diversity of transcript populations

Messenger RNA molecules, otherwise called transcripts, carry the
genetic information encoded in genes. In most cells these transcripts
constitute only a small proportion of the total RNA, whereas
ribosomal and transfer RNA account for more than 98%. In any
cell type, the transcript population typically consists of thousands of
distinct transcripts, most of which are transcribed from different genes
(although splice variants of genes exist too). These transcripts can be
present in widely varying amounts ranging from just a few copies per
cell to thousands of copies. Furthermore, the relative levels of transcripts
are constantly changing as the cell responds to different environmental
signals. The amount of transcripts is estimated to follow a normal
distribution in which a small number of genes are expressed at high or
very low levels. The majority of the genes are expressed at intermediate
levels.
Genes come in different sizes, with different numbers and sizes of exons.
The size of transcripts reflects this by varying from a few hundred
nucleotides to up to about 20 000 nucleotides. Average length of
transcripts is estimated to be between 1.5–1.7 kb.

● 51
 M I C R O A R R AY

6.2 Requirements of labelling methods

6.2.1 Retaining gene expression information

The labelling methods used in microarray analysis must cope with the
inherent diversity of transcript sequences and create representations that
contain all the information present in the original transcript population.
Thus an ideal labelling system is neither biased towards any nucleotide
sequences, nor does it label differently transcripts of different sizes or
sequences that are expressed at different levels.

In reality, existing labelling methods do not convert all information into

labelled form. Enzymatic methods are limited to copying certain nucleic
acid sequences, whereas the instability of some transcripts is a general
problem for all methods.

6.2.2 Length of labelled fragments

Accurate information about gene expression can only be deduced from
microarray experiments if the labelled nucleic acids can hybridize
efficiently and with specificity to their complementary targets. The
length of the labelled fragment is an important factor in determining
these parameters. Fragments longer than 100 nucleotides can hybridize
strongly enough with their target sequences to withstand stringent
hybridization and wash conditions. The hybridization kinetics of
shorter fragments is faster. For optimal hybridization, probes consisting
of fragments of 200–500 nucleotides long are recommended. Longer
fragments may not find their targets as efficiently as shorter fragments,
but will produce a higher signal when hybridized as they carry more
labelled molecules.

6.2.3 Yield of labelled probe

The amount of labelled probe prepared by the labelling method is
important to the sensitivity of microarray experiments. This is because
the efficiency of the labelling process is critical in determining the lowest
amount of mRNA that can be used to generate detectable signals from
microarrays. With higher amounts of mRNA available, differences in
probe yield from different labelling methods can make the difference
between being able to hybridize one or several slides with one probe.
Ideally, the labelling method should transform each transcript into a
labelled fragment, without any bias towards more highly expressed
sequences. If the labelling method results in net amplification of nucleic
acid in the labelling process, the amplification process should be linear,
i.e. the original ratios of expression levels within the sample should not
be changed in the amplification process.

hb 63-0048-49 ● 52
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.2.4 Optimum labelling density

The fluorescent labels used in microarray analysis bring their own
restrictions to labelling protocols. If two or more fluorescent molecules
are in close proximity of each other, a significant portion of the absorbed
light energy can be spent on interactions between different molecules and
dissipated as heat. This will result in less than the expected amount of
fluorescence being emitted from the sample, and moreover, the amount
of fluorescence is no longer directly proportional to the number of fluors
in the sample. This phenomenon is called “quenching”, and it is an
inherent property of fluorescent molecules. Each fluorophore has slightly
different quenching properties that are determined by its chemical
structure. In practical terms this means that for each fluor there is an
optimal labelling density, or distance between attached labels, which will
produce maximum fluorescence from a labelled nucleic acid fragment.
Exceeding this optimum labelling density results in decreased fluorescent
signal (Fig 32). Therefore, in order to achieve highest sensitivity of
detection, the labelling method used in microarray experiments should be
optimized to yield fragments that are labelled at the maximum density as
determined by the labelling fluors. See chapter 4, section 4.2.2 for
additional information.

High density labelling

CyDye

Fluorescence
A C C C T A G G A T G C G T G C G C T A A G A A C A T G G G G A T T Interaction & absorption

Lower density labelling

A C C C T A G G A T G C G T G C G C T A A G A A C A T G G G G A T T

Fig 32. Quenching and labelling density. Fluorescent output

from two identical nucleic acid strands labelled to different
labelling densities is depicted. Green denotes fluorescent
emission whereas purple shows energy being lost in
intermolecular interactions between adjacent fluorophores.
Lower labelling density results in higher fluorescent signal.

● 53
 M I C R O A R R AY

6.2.5 Equal labelling with different fluors

The purpose of differential gene expression analysis is to detect relative
differences in the number of specific transcripts between two or more
samples. This requires that the two samples hybridize competitively
with the immobilized targets and differences in relative signals primarily
reflect changes in the number of the transcripts. From a technical point
of view, this is best achieved when equal numbers of equally labelled
nucleic acid fragments are compared. As two different fluors are used
in two-color analysis, no imbalance due to the properties of the fluors
should be present in the labelled populations. In extreme cases such
imbalances can lead to false positive signals from gene expression
microarrays. The labelling method should produce the same labelling
density and size distribution of labelled fragments, regardless of the
fluorescent dye label used.

6.2.6 Nucleotide sequence preferences

As all labelling methods attach the fluorescent label in a specific manner,
usually via certain nucleotides, the nucleotide sequence of the molecules
being labelled can have a significant effect on labelling density and also
on the length of labelled fragments (Fig 33). For example, incorporation
of CyDye dUTP instead of CyDye dCTP into cDNA that is C-rich is less
likely to result in quenching because the likelihood of incorporating
two CyDye dCTP into close proximity will be lower. Conversely, highly
A-rich sequences are best labelled using CyDye dCTP as label.

Labelling with Cy-dCTP

CyDye

Fluorescence
T G C C A C C C T A C G C C C C C G C T C C C T C T A C C C C T A A
Cy-dCTP Interaction & absorption

Labelling with Cy-dUTP

U G C C A C C C U A C G C C C C C G C U C C C T C U A C C C C U A
Cy-dUTP

Fig 33. The effect of sequence on labelling density and

fluorescent signal. The same nucleotide fragment has been
labelled with Cy-dCTP and Cy-dUTP, using a similar ratio of
label to cold nucleotide. As the fragment is GC-rich, the use of
Cy-dCTP results in much higher labelling density on this
fragment than the use of Cy-dUTP, which can only be
incorporated in a few positions. However, the fluorescent signal
from the Cy-dUTP labelled fragment is greater than from the
more densely labelled fragment, because quenching becomes
a significant factor in the situation of high labelling density.

hb 63-0048-49 ● 54
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

As probes used in microarray analysis are complex mixtures of nucleic

acid sequences, careful optimization of labelling methods is needed to
ensure that quenching is not a practical problem for the majority of
transcripts. However, it may not be possible to create labelling conditions
that would be perfect for all sequences. Performing a ‘yellow experiment’,
in which the same sample is labelled with both fluors and then used in
microarray hybridization, can help to identify the targets that do not give
balanced signals, as well as to optimize the microarray system in general.
By choosing target sequences that do not show high nucleotide sequence
bias, it is also possible to control this aspect of labelling.

Increasing labelling density is not the best solution for increasing signal
from microarrays. Using a lower labelling density is more likely to result
in higher signal. As fluorescent molecules tend to be fairly large, high
labelling densities can lead to changes in the hybridization kinetics of
labelled molecules (47). The melting temperature (Tm) of highly
substituted nucleic acids are lower than those of unlabelled nucleic acids
and can result in lower hybridization signal.

6.3 Labelling strategies

6.3.1 mRNA vs total RNA

Only a small proportion, about 1.5–2.5%, of cellular RNA is mRNA.
Prior to hybridization, mRNA must be purified and labelled. Since most
of the cellular RNA is ribosomal RNA, specific protocols are used to
separate mRNA from ribosomal RNA.

Eukaryotic transcripts usually have a poly-dA tail at their 3' ends. This
property can be exploited by using a complementary poly-dT sequence
to capture polyadenylated transcripts away from other RNA species,
as well as from other molecules and impurities. Because of the additional
purification steps involved in the preparation of mRNA, mRNA samples
tend to give higher signal to noise values on microarrays than total
RNA samples.

● 55
 M I C R O A R R AY

6.3.2 Priming with oligo(dT)

In the labelling reaction, mRNA can be selected from total RNA for use
as the labelling template by using oligo(dT) primers that will hybridize
with the poly-A tail in transcripts. Addition of an anchoring base to the
3' end of these primers directs cDNA synthesis to the beginning of the
poly-A stretch. This has the advantage of producing labelled fragments
that are devoid of most of the repetitive sequence. This priming method
will result in only one copy of cDNA that contains primarily 3' sequences
synthesized from each transcript. If the targets on the microarray are
derived from 5' ends of long cDNAs, probes labelled directly in cDNA
synthesis using oligo(dT) priming may not produce complementary
fragments to these targets, resulting in absence of signal in hybridization.

6.3.3 Random priming

Random priming, in which a mixture of oligonucleotides comprised of
all sequence variants of a short sequence of defined length are used as
primers, can be used to produce probes that contain sequences derived
from all parts of transcripts. Typically, each transcript is copied into
several non-overlapping probe fragments. Because of their longer length
and ability to form more stable duplexes, nonamers are preferred over
hexamers and give higher yields of cDNA. Random priming is only
compatible with mRNA templates, as random primers can anneal to all
RNA molecules. cDNA synthesis from total RNA with random priming
will produce a large quantity of short fragments that lack specificity in
hybridization and usually give rise to high background signals. As the
proportion of label incorporated into cDNA derived from mRNA is
going to be very small under these conditions, the specific signals from
microarray spots will be low. Conversely, as most of the fluorescent label
is incorporated into sequences derived from ribosomal RNA, unspecific
hybridization can become a problem.

6.3.4 Other priming strategies

Highest yield of cDNA from mRNA (without probe amplification) is
achieved with the use of both oligo(dT) and random primers together
(Fig 34). This strategy has the highest likelihood of copying all parts of
transcripts into probe, and therefore, is suitable for use with target
sequences that are derived from varying parts of genes.

It is also possible to use specific primers to copy transcripts into probe.

As each sequence requires the synthesis of a specific primer, this
approach can be costly and require a new set of primers to be prepared
for each different microarray. The advantage of this approach is that
only those sequences that are analyzed on the microarray are labelled. It
is also possible to use total RNA as a sample.

hb 63-0048-49 ● 56
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.3.5 Amount of primer

Regardless of the type of primer used, its concentration should be in
excess of the number of possible binding sites on transcripts so that its
availability is not limiting cDNA synthesis. cDNA synthesis with general
priming strategies (oligo[dT] and random nonamers) may be biased
towards highly expressed transcripts if too much mRNA is used. The
specific priming strategies may be biased against high-expressing
transcripts if the amount of each primer is not sufficient to cover the
whole expression range.

6.3.6 Labelling bacterial RNA

Bacterial mRNA lacks poly-A tails and selecting transcripts for labelling
from total RNA is not as easy as with eukaryotic RNA. It is possible to
remove ribosomal RNA sequences by converting them into cDNA:
RNA hybrids, followed by digestion with RNAse H and DNAse I to
selectively remove the double-stranded sequences. Random priming
strategies can be used successfully with bacterial RNA, if the stringency
of hybridization is controlled carefully to counteract the high proportion
of label associated with ribosomal RNA sequences. Alternatively, priming
strategies that utilize gene specific primers or short primers that are able
to prime from several genes have been used (48, 49). As gene specific
primers prime cDNA synthesis only from those genes that are being
studied on the array, they can help to increase specificity of hybridization
and signal to noise values.

C T T T T T T T T T T T
cDNA strand primed
with anchored oligo(dT) G A A A A A A A A A A A A A A

cDNA strand primed

G A A A A A A A A A A A A A A
with random nonamers

cDNA strand primed

C T T T T T T T T T T T
with anchored oligo(dT) G A A A A A A A A A A A A A A
and random nonamers

Fig 34. The use of different primers in cDNA synthesis.

cDNA strand synthesized with anchored oligo(dT) priming is
shown in yellow and those primed with nonamers in green.
The combined use of both primers results in the probe being
prepared from all parts of the transcript.

● 57
 M I C R O A R R AY

6.4 Enzymatic labelling methods

6.4.1 Labelling strategies

Several strategies based on molecular biology or chemical reactions have
been developed for labelling samples for gene expression microarray
analysis. The availability of fluorescent labels in different reactive forms
has contributed to the diversity of labelling methods. All these strategies
have in common that they start with an RNA population (Fig 35).
Molecular biology strategies rely on the use of enzymes to convert
mRNA into new populations of nucleic acids, either DNA or RNA.
Combining two or more enzymatic reactions into one protocol widens
the choice further. Using more than one enzyme for labelling, however,
has the disadvantage that the information carried by the original
population is likely to change more than by using a single enzyme. This
is because some information is lost in each enzymatic conversion step,
and as the lost information is dependent on the sequence of the
transcripts and the properties of the enzyme, the representations
synthesized by each enzyme will be different. Chemical methods have the
advantage that no copying of nucleic acid to another form takes place;
instead the labelling moiety reacts with the nucleic acids to form
covalently modified, labelled probe population.

mRNA or RNA

Incorporation Incorporation Incorporation cDNA cDNA Labelling

of radioactive of fluorescent of AA-dUTP with chemical
label nucleotides reagents
with RT

Post labelling Labelling Incorporation

with CyDye with chemical of CyDye
NHS esters reagents ribonucleotides

cDNA probe RNA probe

Fig 35. Labelling strategies for gene expression microarrays.

hb 63-0048-49 ● 58
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.4.2 Fluorescent labels

Fluorescent dyes, especially the cyanine dyes Cy3 and Cy5, are the most
popular choice for dual color microarray analysis. The main benefit of
using CyDye fluors in particular is that two dyes can be excited and
detected from the same slide. CyDye fluors also produce bright signals
and have a wide dynamic range of detection, so both weak and strong
signals can be detected in the same experiment. Fluorescent dyes can be
directly incorporated into nucleic acid by either enzymatic or chemical
methods.

● 59
 M I C R O A R R AY

6.5 Labelling in first-strand synthesis

6.5.1 Principle
One of the simplest and most popular labelling strategies is to convert
mRNA population into a labelled first-strand cDNA population. This is
achieved by copying the transcripts into cDNA molecules with a reverse
transcriptase while incorporating a modified CyDye nucleotide. The
cDNA synthesis can be primed with a choice of primers including
random primers, anchored oligo(dT) as well as gene specific primers.
This allows the use of both mRNA and total RNA as sample.

Fig 36. Principle of labelling in first-strand mRNA or total RNA

cDNA synthesis. Fluorescent nucleotide
(pink circles) is incorporated into first-strand
Primer added
cDNA by a reverse transcriptase. After
degradation of the mRNA template strand,
labelled single-stranded cDNA probe
can be purified.
Annealing

Nucleotide,
CyDye nucleotide
and enzyme added

NaOH/Hepes added

cDNA synthesis

NaOH/Hepes added

RNA degradation

Purification columns

cDNA purification

hb 63-0048-49 ● 60
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.5.2 Optimization
The incorporation of fluorescently labelled nucleotides is the rate-
limiting step of this labelling method. This is because all polymerase
enzymes, both DNA and RNA dependent polymerases, incorporate
unlabelled nucleotides more efficiently than larger fluorescent dye
nucleotides. The incorporation kinetics are very much dependent on
the identity of the enzyme, and even homologous enzymes can have
very different properties. The ratio between the fluorescently labelled
nucleotide and the corresponding unlabelled nucleotide in the labelling
reaction determines the incorporation of fluors into cDNA. As different
fluorescent dyes and nucleotides have different structures, it is necessary
to optimize this ratio separately for each combination of dye and
nucleotide. Furthermore, optimized ratios determined for one enzyme
will not necessarily give optimal results with other enzymes.

Labelling in first-strand synthesis does not produce long cDNA molecules.

This is because elongation of the nucleotide chain is dependent on the
previous nucleotides having been incorporated. The incorporation of
more than one fluorescent label consecutively into cDNA is not favored
by polymerases. If the ratio of fluorescent nucleotide to unlabelled
nucleotide is high, a highly labelled cDNA can be transcribed, but cDNA
synthesis will stall if the mRNA sequence requires several labelled
nucleotides to be incorporated in a row. Only short fragments will be
made, resulting in low yield of cDNA. Lowering the nucleotide ratio can
increase the yield of cDNA, but this will compromise labelling density
and the brightness of the probe. A balance between these two factors and
the consequences of quenching at high labelling densities must be attained
for optimal results. In practice this requires evaluation of fluorescent
signals on microarrays from probes labelled at different nucleotide ratios,
and this demands considerable effort.

● 61
 M I C R O A R R AY

6.6 cDNA Post Labelling

6.6.1 Principle
The shortcomings of the first-strand cDNA labelling method and the
availability of CyDye as reactive N-hydroxyl succinimidal dyes (NHS-dyes)
have led to the development of cDNA post-labelling method.

mRNA or total RNA Aminoallyl labelled cDNA

Primer added CyDye NHS ester added

Annealing

Nucleotide,
CyDye nucleotide
and enzyme added

Coupling of CyDye
NaOH/Hepes added
to modified cDNA

cDNA synthesis
Hydroxylamine
quencher added
NaOH/Hepes added

RNA degradation

Purification columns

cDNA purification

Elimination of
free reactive dye

Purification columns

Purification of
CyDye labeled cDNA

Fig 37. The principle of post-labelling is illustrated. mRNA is

converted into first-strand cDNA that contains aminoallyl-dUTP. After
elimination of mRNA template, the amine groups on cDNA are
reacted with CyDye-NHS ester, resulting in the generation of
fluorescently labelled cDNA. Excess of NHS-ester can be neutralized
with hydroxylamine, and labelled cDNA is purified for use.

hb 63-0048-49 ● 62
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

Amersham Biosciences has developed CyScribe™ Post-Labelling Kit

(see section 6.14) for this method. In this method, amine-modified cDNA
is first synthesized by incorporating aminoallyl-modified nucleotide in
first-strand cDNA by a reverse transcriptase. After removal of RNA
template and purification of the amine-modified cDNA, chemical
labelling with N-hydroxyl succinimidyl-ester derivative of CyDye is
performed. NHS-esters react with amine groups to form a covalent
bond between CyDye and the amine group while the NHS-ester group
is released (Fig 38). A high excess of CyDye NHS-ester is required for
efficient reaction, and any non-reacted label is neutralized with an
excess of small amine such as hydroxylamine. The cDNA needs to be
re-purified after labelling to remove CyDye that is not incorporated
into labelled cDNA.

R′ = CyDye Conjugate

O O
O O
R NH2 + R + N OH
N R′
R′ N
O O
O H
Amide group on cDNA NHS ester dye Amide bond coupling NHS leaving group
CyDye to cDNA

Fig 38. The use of CyDye NHS-ester in

labelling amine-modified cDNA

● 63
 M I C R O A R R AY

6.6.2 Benefits
The main benefits of this method over the first-strand cDNA labelling
method are derived from the more efficient incorporation of the smaller
aminoallyl nucleotide compared with the bulkier CyDye nucleotides. As
a consequence, the yield of cDNA is considerably higher than with the
first-strand labelling method. The cDNA fragments synthesized in the
post-labelling method are also longer. Further benefits of the post-labelling
method stem from the chemical labelling step itself. The number of
amine groups on cDNA is the main factor influencing labelling density.
The sequences of the cDNAs being labelled do not have a major impact
on labelling outcome. Because of this, more random attachment of labels
is achieved than with the first-strand cDNA labelling method. Furthermore,
as the labelling process is not dependent on the structure of different
fluorescent dyes, it is easier to achieve equal labelling with both Cy3 and
Cy5, and the labelling method introduces less variation into microarray
analysis. This is illustrated in tighter scatter plots derived from microarray
hybridizations in which the mRNA samples were labelled with the post-
labelling method (Fig 39). As the extent of experimental variation is
reduced, the detection of smaller changes in gene expression between two
samples is improved.

CyScribe Post-Labelling Kit CyScribe First-Strand cDNA

scatter plot Labelling Kit scatter plot
Cy3 volume Cy3 volume
1000000 1000000

Fig 39. Scatterplots of gene expression

microarray analyses comparing skeletal 100000 100000
muscle and placental mRNA samples.
Identical mRNA samples were labelled
10000 10000
with the post-labelling method or with
first-strand cDNA labelling. Identical slides
were hybridized with equal amounts of 1000 1000

all probes.

100 100
100 1000 10000 100000 1000000 100 1000 10000 100000 1000000

Cy5 volume Cy5 volume

hb 63-0048-49 ● 64
6.6.3 Chemical considerations
Coupling of CyDye NHS-ester to amine-modified cDNA requires mildly
alkaline pH, which is provided by the coupling buffer. However, if they
are not carefully removed, buffer components or acidic residues from
preceding steps can alter the pH of this buffer. Furthermore, any
amine groups present on other compounds will compete with the
amine-modified cDNA for CyDye incorporation. Hence the free
aminoallyl-dUTP, Tris-buffer, and reverse transcriptase enzyme must
be removed from the cDNA preparation before labelling. This can be
best achieved with affinity column chromatography methods, such as
GFX columns, in which amine-modified cDNA is bound to the matrix,
other compounds are washed away, and cDNA is finally eluted with
water. Alternatively, standard ethanol precipitation also works well
and provides the added benefit of concentrating the cDNA ready for
CyDye coupling.

NHS-esters are readily hydrolyzed with water, even with the small
amount of moisture present in laboratory air. Because of this, aliquots
of CyDye NHS-esters must always be stored desiccated and protected
from light. Storing CyDye NHS-esters in solution can lead to rapid
loss of reactivity. As these reactive dyes were originally developed for
protein labelling, the quantities provided commercially were adjusted
for this application. This necessitated aliquoting of reactive dyes before
their use for microarray sample labelling and frequently resulted in
decreased activity as a consequence of handling and storage. The
availability of individually foil-packed, pre-dispensed, and freeze-dried
aliquots of Cy3 and Cy5 NHS-esters for microarray analysis has
removed this problem. Furthermore, the quality of the CyDye NHS-ester
in CyDye Post-Labelling Reactive Dye Packs is higher than available
otherwise.

● 65
 M I C R O A R R AY

6.7 RNA amplification and labelling in RNA synthesis

6.7.1 RNA amplification

The amount of RNA sample can be a limiting factor for microarray
analysis, and it may be necessary to amplify RNA before analysis. In the
most commonly used protocol, the mRNA population is first converted
into a double-stranded cDNA that contains a promoter sequence for viral
RNA polymerase, such as T7, T3, or SP6 polymerase (Fig 40).
This can be achieved by using a modified oligo(dT) primer containing
a 5' extension coding for the viral promoter. Each resulting cDNA
molecule will contain one RNA polymerase promoter sequence. By
including the corresponding RNA polymerase and ribonucleotides in
the reaction, several RNA copies can be synthesized from each template.

mRNA or total RNA A A A A A A A A A

A A A A A A A A A

A A A A A A A A A

Reverse transcriptase dNTPs T T T T T T T T T T T T - T 7

NTPs
Double-stranded
T7 IVT templates
(T7 RNA polymerase + NTPs)

Single-stranded
amplified RNA
(T7 RNA polymerase + NTPs)

Fig 40. Principle of RNA amplification with

RNA polymerases.

hb 63-0048-49 ● 66
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

This results in a linear amplification of the cDNA pool into RNA

without altering the relative abundance of sequences in the mixture
significantly. However, the length of synthesized fragments will be
shorter than the original templates. 2000-fold amplification of starting
RNA can be achieved with this method, in one round of amplification
(50, 51). This RNA amplification strategy can be used on its own, and
the amplified RNA population can be labelled separately using other
methods. Alternatively, labelling and amplification can be performed
together by including a CyDye ribonucleotide in the reaction (Fig 41).
In some cases when only a few cells are available for the preparation of
RNA sample, such as when laser capture micro (LCM) dissected samples
are analyzed, several rounds of RNA amplification can be performed to
acquire enough RNA for labelling (52).

mRNA or total RNA A A A A A A A A A

A A A A A A A A A

A A A A A A A A A

Reverse transcriptase dNTPs T T T T T T T T T T T T - T 7

NTPs
Double-stranded
T7 IVT templates CyFye-NTPs
(T7 RNA polymerase + NTPs)

Single-stranded
amplified RNA
(T7 RNA polymerase + Cy-dUTP
+ unlabelled NTPs)

Fig 41. Principle of preparation of

fluorescently labelled RNA probe with RNA
polymerase amplification.

● 67
 M I C R O A R R AY

6.7.2 Other amplification methods

Amplification of RNA can also be achieved by other means, including
using limited numbers of PCR cycles to amplify double-stranded cDNA
(53). However, PCR-based methods have not gained wide popularity in
microarray analysis, because of the problems associated in logarithmic
amplification of complex nucleic acid mixtures. Sequence and size
differences in nucleic acid fragments can influence their amplification
rate and result in selection of sub-populations of sequences (50).
Regardless of the RNA amplification strategy, the method should neither
alter the relative abundance of different transcripts in the sample nor
result in the creation of sequence chimeras in which sequences from two
or more transcripts are joined together. Linear amplification strategies
avoid these pitfalls and are favored in microarray analysis.

6.7.3 Labelling with RNA amplification

Synthetic RNA can be labelled in synthesis reactions by incorporating
a CyDye ribonucleotide. The main factor determining both labelling
density and yield of labelled RNA in this method is the ratio of CyDye
ribonucleotide to the corresponding unlabelled ribonucleotide. As is the
case for DNA polymerases, RNA polymerases incorporate the unlabelled
nucleotide more efficiently, and a compromise between labelling density
and yield of RNA is necessary. This requires careful optimization of the
labelling conditions, combined with the analysis of fluorescent signal
derived from the probes.

RNA labelling can also be performed by using hapten-labelled

ribonucleotide, such as biotinylated ribonucleotides. This strategy has the
advantage of giving higher yields of cDNA, as the incorporation of large
dye nucleotide is not a rate-limiting factor. However, the need to perform
additional detection steps adds to the length of the protocol and makes
the use of more than one color at a time more difficult.

6.7.4 Fragmentation of RNA probes

The secondary structure of RNA can interfere with hybridization to
targets. Fragmentation of RNA probe into smaller fragments of 50–200
nucleotides can be performed to overcome this. This can be achieved in
controlled fashion by exposure of RNA to potassium and magnesium
ions (24). Careful handling of RNA is necessary at all stages to minimize
uncontrolled degradation of RNA to nucleotides and short fragments. If
RNA probes are used in hybridization, precautions must be taken
throughout the whole microarray procedure.

hb 63-0048-49 ● 68
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.8 Expression patterns

The intensity of hybridization signals from microarrays is determined not
only by the number of hybridized probe fragments but also by their
CyScribe First-Strand
cDNA Labelling Kit
length and the number of fluorescent labels that are associated with
them, or the labelling density. Similar labelling densities can be achieved
with different labelling methods. However, with different methods, the
average length of the labelled fragments varies. In practice, the intensity
CyScribe
Post-Labelling Kit of signals derived from two identical samples labelled with different
methods will be different. Hence, gene expression microarrays do not
give quantitative information about gene expression. However, if two
Fig 42. Expression patterns obtained from samples are labelled by the same method using two different fluorescent
gene expression microarrays comparing labels, the relative intensity of the signals will reflect the relative
skeletal muscle and placental mRNA abundance of the specific transcripts in those two samples. Information
samples. Identical mRNA samples were
about differential gene expression can therefore be gained (Fig 42).
labelled with the post-labelling method
and first-strand cDNA labelling methods Labelling the two samples to be compared under identical conditions is
using CyScribe Labelling Kits. Equal therefore extremely important for guaranteeing that experimental
amounts of all probes were hybridized with artifacts do no lead to wrong conclusions from microarray results.
identical microarrays. Although the overall
hybridization patterns are very similar, some
significant differences in signal intensities
of individual spots can be seen. These
6.9 Random prime labelling of DNA
reflect the size differences of labelled A modified random prime labelling method can be used to label
fragments obtained with the two labelling DNA with CyDye. In this method, Klenow polymerase incorporates
methods. Despite the different appearance fluorescently labelled nucleotides in a DNA synthesis reaction, which
of the two microarrays, both gave similar
is primed with random nonamer or hexamer primers. This method is
information on differential gene expression.
a practical solution for genomic microarrays, although direct chemical
labelling methods can also be used. Random prime labelling methods
are not recommended because two different enzymes are needed to
convert mRNA into labelled form.

6.10 Direct chemical labelling of mRNA

When an enzyme is used to convert a mRNA population into another
nucleic acid population, some information is lost because the ability of
different enzymes to copy through different nucleic acid sequences varies.
By using a chemical labelling method, it is possible to label mRNA
directly by using a chemical reaction, coupling modified CyDye reagent
to RNA molecules. These methods are simple to perform, as no
modification of RNA is required before labelling. Furthermore, these
methods are less prone to discrimination against certain nucleotide
sequences that are difficult templates for polymerase-based labelling
systems. It is important to note that any chemical that avidly reacts with
nucleic acid molecules is potentially toxic and will require careful
handling and adequate safety measures to be taken.

● 69
 M I C R O A R R AY

6.11 Purification of labelled probes

Regardless of the labelling strategy, it is necessary to purify the labelled
nucleic acid after labelling, as the amount of incorporated fluorescent
dye is typically only a small fraction of all the dye present in the sample.
The recovery of labelled nucleic acid from purification is a major limiting
factor for most labelling methods. Products such as AutoSeq™ G-50,
GFX, and QIAquick™ spin columns have been developed for the
purification of double-stranded cDNA. They can also be used to remove
unincorporated CyDye nucleotides away from fluorescently labelled
single-stranded cDNA, but their use does not result in optimal recovery
of labelled material. Typically, less than 40% of the labelled probe is
recovered, and the recovery can vary considerably between different
samples. This variation not only reduces the amount of data that can be
generated with microarray analysis, but also significantly contributes to
poor quality of results if the amount of probes are not adjusted before
hybridization. When small amounts of template are labelled, the loss of
labelled cDNA tends to be higher, and as little as 5% of probe may be
recovered.

All of these methods, however, are relatively successful in removing the

free dye nucleotide. Ethanol precipitation is not an option for use with
most labelling methods, as it can result in the formation of dye
aggregates that will produce intense speckled background in array
hybridization.

Appreciation of these problems has lead to the development of a novel

GFX purification system, CyScribe GFX Purification Kit, which has been
specifically tailored for purification of single-stranded CyDye labelled
cDNA. As this purification system is based on binding of the labelled
nucleic acid to a customized GFX matrix, it is also suitable for use with
the post-labelling method. These CyScribe GFX columns give excellent
yield of purified cDNA from different synthesis scales, including small-
scale samples. Therefore, the use of this purification system downstream
of optimized labelling kits can improve the sensitivity of microarray
analysis, enable the use of smaller amounts of RNA samples, or increase
the number of replica slides that can be hybridized with one sample.

6.12 The CyScribe family of labelling kits

The CyScribe family of labelling kits from Amersham Biosciences has
been developed to offer a range of optimized labelling products for
producing CyDye labelled microarray probes. These kits enable flexible
choice of different labelling methods to suit the different needs of
researchers.

hb 63-0048-49 ● 70
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.13 CyScribe First-Strand cDNA Labelling Kit

6.13.1 Features of the kit

CyScribe First-Strand cDNA Labelling Kit has been developed for preparing
highly fluorescent CyDye labelled probes for microarray analysis using
first-strand cDNA labelling. The kit has been optimized for the use of either
CyDye dCTP or CyDye dUTP nucleotides as labels. Two nucleotide
mixes are provided in the kit to provide optimal nucleotide ratios for
each type of nucleotide when 1 nmol of CyDye nucleotide is used per
reaction. The compositions of these solutions have been optimized so
that labelled cDNA will contain an attached CyDye fluor to every 12–25
nucleotides synthesized, regardless of the nucleotide used.

CyScribe First-Strand cDNA Labelling Kit contains both anchored

oligo(dT) primers and random nonamer primers, allowing flexible choice
of templates. As little as 50 ng of mRNA and 2.5 µg of total RNA can be
labelled per reaction with the kit and used successfully in microarray
hybridization on one slide (Fig 43). Recommended highest amounts of

Signal from microarrays

120

100

Cy3
80
250 ng Cy5
Total signal from
spot set 60

100 ng 40

20

50 ng
0
a) b) 500 ng mRNA 250 ng mRNA 100 ng mRNA 50 ng mRNA
Probe reaction template

150
Cy3
10 µg of total RNA Cy5
100
Relative signal
50

5 µg of total RNA
0
10 µg 5 µg 2.5 µg
c) d)
Amount of total RNA

Fig 43. The use of CyScribe First-Strand cDNA Labelling Kit with small
template amounts. The indicated amounts of mRNA and total RNA
templates were labelled with Cy3 and Cy5 in duplicate reactions, purified,
and all of the recovered probe was used in a microarray hybridization with
duplicate slides. Panels A and C show part of microarray images obtained
with mRNA probes and total RNA probes, respectively. Panels B and D
show quantified Cy3 and Cy5 signals from these arrays. As the amounts of
probes were not adjusted before hybridization, a high amount of variation
in signal intensities was observed in this experiment.

● 71
 M I C R O A R R AY

template are 500 ng of mRNA and 25 µg of total RNA per reaction.

These amounts of template will generate enough probe for several
microarray hybridizations. Highest yield of labelled probe is obtained
with dual priming, when both oligo(dT) and random primers are used
together. This priming strategy is only compatible with the use of mRNA
templates. The choice or amount of RNA template does not affect the
labelling density achieved with the kit.

6.13.2 Degradation of RNA template

Degradation of the RNA template after cDNA synthesis is necessary to
prevent the labelled probe from hybridizing with the original template in
solution instead of the microarray targets during microarray hybridization.
The removal of RNA can be performed enzymatically, by using RNAse H
enzyme to digest the RNA component of RNA DNA heterohybrid
molecules. A simpler option is to degrade the RNA strands by raising
the pH of the probe solution. The CyScribe Kits contain an efficient
RNA degradation protocol which has been developed to minimize pH
fluctuation observed in earlier protocols in which small volumes of
concentrated alkali and acids were used. In this improved protocol, RNA
is degraded by the addition of 2 µl of 2.5 M sodium hydroxide, and after
incubation the solution is neutralized with 10 µl of 2 M Hepes free acid.

The CyScribe First-Strand cDNA Labelling Kit also contains a mixture of

synthetic mRNA molecules, as control RNA template, that can be used
to gain familiarity of the labelling technique, or to troubleshoot problems.
Because of its synthetic nature this control RNA is not suitable for
microarray hybridization.

6.13.3 Critical success factors for CyScribe First-Strand cDNA

Labelling Kit
■ Only label RNA that is intact, clean, and in known quantity.
■ If possible, purify mRNA for best labelling results and highest
yield of labelled probe.
■ Do not exceed the recommendations for template amount.
■ Do not alter the amount of CyDye in reaction.
■ Pipette all volumes exactly.
■ Protect CyDye from light during all handling and storage.
■ Do not alter the RNA degradation protocol.
■ Monitor the success of purification.
■ Measure the amount of purified probe before performing
microarray hybridization.

hb 63-0048-49 ● 72
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.14 CyScribe Post-Labelling Kit

6.14.1 Features of the kit

CyScribe Post-Labelling Kit has been developed to offer an optimal and
convenient solution for using post-labelling methods in microarray
analysis. Each kit provides reagents for performing 12 cDNA synthesis
and labelling reactions with both Cy3 and Cy5 fluors. CyScript reverse
transcriptase, which is a highly efficient enzyme and gives high yields of
cDNA, is used to synthesize amine-modified cDNA by incorporation of
aminoallyl-dUTP into first-strand cDNA. The amount of this modified
nucleotide in the cDNA synthesis reaction has been adjusted to give an
optimal labelling density with CyDye fluors.

The kit includes a protocol for an improved RNA degradation method

and two alternative methods for removing amine-containing impurities
from amine-modified cDNA. Purified cDNA is reacted with an amount
of reactive CyDye NHS-ester that has been chosen to give high and
reproducible labelling density, similar to that achieved with the CyScribe
First-Strand cDNA Labelling Kit. Protocols for reacting excess
NHS-esters with hydroxylamine and subsequent purification of labelled
cDNA with column chromatography are also provided.

A practical difficulty in performing CyDye post-labelling of microarray

samples has been the necessity for aliquoting and storing CyDye
NHS-esters when traditional reagents developed originally for protein
labelling have been used. Because of the instability of NHS-esters in
moist conditions, their storage in standard laboratory conditions can
result in significant loss of reactivity in just a few weeks. CyScribe Post-
Labelling Kit solves this problem by providing ready-to-use CyDye NHS-
esters in individually dispensed aliquots. These dyes have been sealed in
foil to protect them from light and contain desiccant for extra protection.
The reactive dyes in these aliquots are also guaranteed to contain over
75% reactive dye content, thus providing the highest quality reagents for
microarray labelling.

CyScribe Post-Labelling Kit includes both oligo(dT) and random

nonamer primers, offering flexible use of both total and messenger
RNA as template. Because this kit yields high amounts of cDNA, it is
recommended that 500 ng or less of mRNA and 25 µg or less of total
RNA are used per reaction. Highest yield of cDNA is obtained from
mRNA using both types of primers together.

● 73
 M I C R O A R R AY

6.14.2 Comparison of performance with CyScribe First-Strand

cDNA Labelling Kit
Properties of labelled probe:
■ The CyScribe Post-Labelling Kit synthesizes about three times
as much cDNA than CyScribe First-Strand cDNA Labelling Kit
from an equal amount of template. However, because this kit involves
two purification steps in which some of the cDNA is lost,
approximately two-fold increase in final probe yield is achieved.
■ Both kits have been optimized to give similar labelling densities: on
average a CyDye fluor is attached at every 12–25 nucleotides.
■ As the average length of cDNAs synthesized with the post-labelling kit
is longer, higher signals can be obtained from some targets (ones with
long transcripts) with this kit.
■ The post-labelling method is not influenced by individual nucleotide
sequences to the same extent as the first-strand cDNA labelling
method is. Therefore, this method can cope better with sequences that
contain repeated nucleotide stretches, which can lead to chain
termination in first-strand labelling.
■ Because of the chemical nature of the labelling process, more random
distribution of CyDye fluors over labelled cDNAs is obtained than
with the first-strand labelling method in which labelling is modified by
sequence-specific events.

6.14.3 Identification of differential gene expression with

CyScribe kits
The performance of the two CyScribe labelling kits in identifying
differential gene expression was investigated in a model experiment
in which human skeletal muscle and placental mRNA populations
were compared. Replicate labelling reactions were performed with
both systems, purified probes were pooled, and replica slides were
hybridized with 25 pmol of each probe. Examples of the expression
patterns produced by the two CyScribe Labelling Kits from these

Normalized
log ratio
Fig 44. Identification of muscle-specific 2.75

gene expression with CyScribe Labelling 2.25 P1

Kits. Data is shown from two replica slides 1.75
P2
hybridized with skeletal muscle and 1.25
C1
placental cDNA probes labelled with C2
0.75
CyScribe First-Strand cDNA Labelling Kit
0.25
and CyScribe-Post Labelling Kit. 1 3 5 7 9 11 13 15 17 19 21 23 25

Gene number

hb 63-0048-49 ● 74
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

experiments, as shown in Figure 44, highlight the different intensity

of several gene-specific signals. However, when muscle-specific gene
expression was examined, essentially the same genes were identified
with both methods as being more highly expressed in muscle tissue.

In order to compare the performance of the CyScribe Labelling Kits in

analyzing gene expression over a wide range of expression values, 31
random gene sequences were selected for analysis. In skeletal muscle
mRNA, the expression levels of these genes cover the whole dynamic
range of values (data not shown).

Figure 45 depicts normalized gene expression log ratios for the 31

selected gene sequences. Despite the differences in the intensity of the
observed Cy3 and Cy5 signals generated by the two labelling systems
(data not shown), the extent of differential gene expression revealed by
the two methods is concordant for most of the genes. On the whole,
variation between replica slides is greater than variation between the two
labelling methods. Some genes show slightly different (within 0.5 log
units) values for differential gene expression and may indicate the
presence of gene sequence-specific labelling events. However, this data
does not support the conclusion that either of the methods is labelling
with a systematic bias towards one of the fluors. Rather, a particular
labelling method may be more suitable for extracting information from
certain gene sequences, and combining data from experiments using
different labelling principles could enhance chances of identifying
significant difference in gene expression between two samples.

Average
normalized
log ratio
Fig 45. Differential gene expression determined 2.5
2.0
with CyScribe First-Strand cDNA Labelling Kit 1.5 CyScribe Post-Labelling Kit
and CyScribe Post-Labelling Kit for 31 randomly 1.0 CyScribe First-Strand
0.5 cDNA Labelling Kit
chosen genes in skeletal muscle and placental 0
-0.5
mRNA samples. Data from two replica slides, -1.0
each of which contained two identical spot sets -1.5
-2.0
for each of the labelling methods, is shown. 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31

Gene number

● 75
 M I C R O A R R AY

6.14.4 Critical success factors for CyScribe Post-Labelling Kit

■ Only label RNA that is intact, clean, and in known quantity.
■ If possible, purify mRNA for best labelling results and highest yield
of labelled probe.
■ Do not exceed the recommendations for template amount.
■ Do not use random nonamers with total RNA.
■ Purification of amine-modified cDNA is critical for labelling
success: other amines and acidic ions should be removed.
■ Do not alter the RNA degradation protocol.
■ Only dissolve CyDye-NHS esters immediately before use.
■ Do not reuse CyDye-NHS ester solutions.
■ Pipette all volumes exactly.
■ Protect CyDye from light during all handling and storage.
■ Monitor the success of purification.

CyDye

Reagent

G A U T C G C C A A U C A G G A G C U C A G A U G C A A A A A A

G A U T C G C C A A U C A G G A G C U C A G A U G C A A A A A A

Fig 46. Principle of mRNA labelling with CyScribe Direct

mRNA Labelling Kit.

hb 63-0048-49 ● 76
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.15 CyScribe Direct mRNA Labelling Kit

Reconstitute active
labelling reagent 6.15.1 Principle
CyScribe Direct™ mRNA Labelling Kit utilizes Cy3 Direct and Cy5 Direct
labelling reagents to generate highly labelled mRNA probes for use in
microarray applications. These labelling reagents contain a CyDye fluor
attached to a chemical group which can react efficiently with guanine
Add labelling reagent to
mRNA and mix reaction residues, resulting in covalent attachment of CyDye to mRNA (Fig 46).
Because of this high reactivity with nucleic acids, Cy Direct™ reagent is
hazardous and requires careful handling. When in contact with water, the
reagent hydrolyzes and loses its reactivity. As seen in Figure 47, the whole
labelling process, including removal of excess dyes, can be performed in
Incubate at 37 °C
less than 2 h.
for 1 o 1.5 h
Because of the chemical nature of the labelling, attachment of both Cy3
and Cy5 is equally efficient and even. Furthermore, the reaction does
not require any enzymatic modification of the mRNA prior to labelling.
The attachment of CyDye to mRNA does not interfere with subsequent
hybridization with DNA targets; thus mRNA labelled with CyScribe Direct
Spin down mRNA Labelling Kit is suitable for use as microarray probe. Because
reaction
mRNA can be used directly as the probe, no loss of information because of
incomplete or biased transcription process occurs.

Purify labelled mRNA

from active reagent
(Column purification or
ethanol precipitation)

Quantify pmoles of
CyDye per labelling

Hybridize
Hybridize

Fig 47. The method for labelling mRNA

with CyScribe Direct mRNA Labelling Kit.

● 77
 M I C R O A R R AY

6.15.2 Application
Labelling reactions performed with CyScribe Direct mRNA Labelling
Kit can be scaled up or down to accommodate different amounts of
template: as little as 250 ng or as much as 5 µg of mRNA can be labelled
in a single reaction. The ratio of the CyDye Direct Labelling Reagent to
mRNA determines the labelling density achieved with the kit. Extending
the labelling time beyond 1 h can result in higher labelling efficiency
than shorter times. The length of transcripts is an important factor in
determining the intensity of hybridization signals from mRNA probes
labelled with the CyScribe Direct mRNA Labelling Kit. As seen in
Figure 48, the expression patterns obtained with the direct labelling
method differ considerably from those generated with the first-strand
labelling method. This evidence suggests that the direct labelling method
may be advantageous for analyzing gene expression from transcripts
that are difficult templates for reverse transcription.

The CyScribe Direct mRNA Labelling Kit has been developed for use
with purified mRNA that is free from contaminating DNA, proteins,
or nucleotides. The kit requires the use of purified mRNA as template
because the chemical labelling reaction cannot discriminate between
transcripts and other species of RNA. Labelling of total RNA will
result in low signal to noise values from microarray experiment.

CyScribe
15 pmol dye/slide
Fig 48. Hybridization of microarrays with
CyScribe Direct
varying amounts of mRNA labelled with Cy3 1 µg mRNA/slide
using CyScribe Direct mRNA Labelling Kit.
Results from a DNA probe generated with CyScribe Direct
10.5 µg mRNA/slide
CyScribe First-Strand cDNA Labelling Kit
are shown for comparison. CyScribe Direct
0.25 µg mRNA/slide

hb 63-0048-49 ● 78
 C H A P T E R 6 : S A M P L E L A B E L L I N G F O R G E N E E X P R E S S I O N A N A LY S I S

6.16 Critical success factors for sample labelling

■ Observe safety precautions while performing the labelling
reaction.
■ It is imperative that the mRNA is free from contaminating
ribosomal RNA, DNA, and proteins.
■ Handle RNA so that degradation is avoided. See Chapter 5
for advice. Protect the CyDye Direct labelling reagent from
water or any moisture, as it can be inactivated on contact.
■ Store the reagent vial and reaction tubes sealed from the
environment.
■ Protect all reactions and labelling reagents from light when
storing and handling.
■ Minimize RNAse contamination of mRNA probes during
microarray hybridization.

Table 4. Choosing the right CyScribe Kit.

CyScribe Direct CyScribe First Strand CyScribe Post
Feature
mRNA Labelling Kit cDNA Labelling Kit Labelling Kit
Brightness of signals
Even incorporation of Cy3 and Cy5
Starting material mRNA mRNA or total RNA mRNA or total RNA
Quantity of starting material using total RNA – 2.5 – 25µg 2.5 – 25µg
Quantity of starting material using mRNA 250 ng – 1µg 50 ng – 1µg 100 – 500 ng
Possible to prepare a batch of unlabelled cDNA and store no no yes
Simplicity of protocol
Time from RNA to probe 2h 3h 5.5 h
Suitable for less experienced users
Labelling density 20 – 35 nuc 12 – 25 nuc 12 – 25 nuc

= Highly recommended = Recommended = Suitable – = Not suitable

● 79
 M I C R O A R R AY

hb 63-0048-49 ● 80
 C H A P T E R 7 : C H A R A C T E R I Z AT I O N O F L A B E L L E D M I C R O A R R AY P R O B E S

Chapter 7
C H A R A C T E R I Z AT I O N O F L A B E L L I N G M I C R O A R R AY P R O B E S

7.0 Introduction
When performing gene expression microarrays, it is important to
characterize the labelled probes in order to avoid experimental error
derived from variation between the probes. Even under carefully
controlled conditions, some differences in the amounts of nucleic acid
samples, labelling success, and recovery of material from the purification
system will occur. In order to account for these artifacts and to ensure
that these variations are not carried through to hybridization, it is highly
recommended that the properties of the labelled probes are determined
before microarray hybridization.

Several methods can be used for characterizing the properties of the

labelled probes. As a minimum we recommend routinely measuring the
amount of CyDye in the labelled and purified sample. If problems do
occur in microarray hybridizations, the other methods described below
can be used for troubleshooting purposes.

■ The quantity of CyDye and nucleic acid in the labelled probe can be
determined using spectrophotometry.

■ Radioactive spiking can be used to derive information about the

amount of nucleic acid in the sample and the success of purification.

■ Gel electrophoresis and fluorescence scanning can be used to analyze the

molecular distribution of the probe, its purity, and relative fluorescence.

7.1 Determination of the amount of CyDye in a

labelled sample with spectrophotometry
Spectrophotometry can be used to determine the amount of CyDye
incorporated into labelled nucleic acid. This can be achieved by
measuring the absorbance of the solution containing the nucleic acid
at the absorption maximum for Cy3 and Cy5. These wavelengths are
550 nm for Cy3 and 650 nm for Cy5. From the known extinction
coefficients corresponding to these wavelengths, the concentration and
amount of CyDye in the sample can be calculated. The amount of CyDye
in the purified sample can be used as a guide to optimize the amount of
probe in the hybridization. Best results are achieved when the amounts
of Cy3 and Cy5 in dual color hybridization are equal.

It is necessary to purify the labelled nucleic acid before performing the

spectrophotometry analysis, as any residual, unincorporated CyDye
labelled nucleotides will interfere with the detection of CyDye labelled cDNA.

● 81
 M I C R O A R R AY

7.1.1 Measuring the amount of CyDye in the probe with

spectrophotometry
Follow the steps below to measure the amount of incorporated CyDye:

❶
Sample preparation
■ Dilute an aliquot of purified probe with water. The required volume
depends on the size of available measuring cells. It is recommended
to use the smallest cells possible. Small volume disposable measuring
cells are available, and they offer the added advantage that a separate
cell can be used for each sample, thus minimizing cross contamination
from one sample to another. However, it is also possible to use regular
100 µl glass cells, but these need to be cleaned thoroughly by rinsing
with sterile water between samples.

❷
Measurement
■ Measure the absorbance spectrum of the sample from 200 to
700 nm against a blank. Although only the absorbance values at
the absorbance peaks are needed to calculate the amount of CyDye
in the sample, the shape of the absorption spectra can give additional
information about the quality of the labelled sample. The absorption
spectrum of CyDye contains two peaks, of which the second should be
of higher intensity (Fig 49). The first peak indicates the presence
of intermolecular interactions between different dye molecules. If both
peaks are of nearly similar intensity, this is a sign of quenching, i.e.
loss of fluorescent signal because absorbed light energy is spent
on intermolecular interactions. This is usually associated with over-
labelling of the sample.

Normalized
Fig 49. Quantification of the amount Absorbance
Aggregation
of CyDye in a labelled probe with 1.0 quenching
spectrophotometry: examples of 0.9

absorption spectra from Cy3-labelled 0.8

samples. The presence of two 0.7

0.6
absorption peaks of near identical
0.5
intensity (gray line) is a sign of
0.4
intermolecular interactions and
0.3
aggregation, which can reduce
0.2
fluorescent signal from the probe.
0.1
The red line shows a typical 0
profile of Cy3 absorption spectrum. 550
Wavelength (nm)

hb 63-0048-49 ● 82
 C H A P T E R 7 : C H A R A C T E R I Z AT I O N O F L A B E L L E D M I C R O A R R AY P R O B E S

■ Measure absorbance at 550 nm for Cy3 and at 650 nm for Cy5

using cuvettes with 1 cm path length to determine the amount of
CyDye in the sample. The observed absorbance values depend on how
much the sample was diluted before measurement, how much RNA
was used in the labelling, the labelling method used, and the efficiency
of labelling. Typically, values around 0.050 would be expected from
first-strand cDNA labelling reactions in which 1 µg of mRNA is used
as a template, when the sample is diluted to 100 µl before measurement.
Choose a dilution that will give a measurement that is clearly discernible
from background values. Sometimes it is necessary to use all of the
labelled sample for analysis.

■ Recover the measured sample for future use. It may be necessary

to concentrate the sample before using for microarray hybridization.
This can be performed by drying down the sample in a vacuum
concentrator or by letting sample that has been heated to 60 °C
evaporate to dryness. It is important to protect the sample from
light during all handling.

❸
Calculation
■ The amounts of Cy3 and Cy5 incorporated into probes can be
calculated from their respective extinction coefficients and using
the following equations:

Extinction coefficients
150 000 M–1 cm –1 at 550 nm for Cy3 and
250 000 M–1 cm –1 at 650 nm for Cy5

Calculation equations
pmol Cy3 in purified sample =
(A550 / 150 000) × dilution factor × (z µl) × (w cm) × 1012
where
A550 = absorbance at 550 nm
z µl = the volume of sample after purification
w cm = optical path in cuvette

pmol Cy5 in measured sample =

(A650 / 250 000) × dilution factor × (z µl ) × (w cm) × 1012
where
A650 = absorbance at 650 nm
z µl = the volume of sample after purification
w cm = optical path in cuvette

● 83
 M I C R O A R R AY

7.2 Determination of the amount of labelled nucleic

acid in the sample

7.2.1 UV spectrophotometry
Two methods can be used to determine the amount of labelled nucleic
acid in the sample: spectrophotometry and radioactive spiking. Nucleic
acids absorb at 260 nm, and absorption measurement at this wavelength
can be used for quantitative purposes. However, several purification
systems release other materials absorbing near or at this wavelength and,
depending on the amounts of these compounds, most of the absorbance
at 260 nm will be derived from something other than nucleic acid.
Therefore absorption spectra from 200 to 400 nm should be measured,
and only if the peak at 260 nm is clearly distinguishable from absorption
at near wavelengths, should estimation of nucleic acid amount be made
(Fig 50).

The following approximations can be used to calculate the amount of

nucleic acid in the probe:

1 A260 unit of double-stranded DNA corresponds to 50 µg/ml

1 A260 unit of single-stranded DNA corresponds to 37 µg/ml

1 A260 unit of single-stranded RNA corresponds to 40 µg/ml

Fig 50. Quantification of the amount of nucleic Unreliable

information
acid in probe with UV spectrophotometry. The about the
sample, from which the black spectrum was amount of
cDNA
generated, contains impurities absorbing near
Absorbance
260 nm and does not give reliable information
about the amount of nucleic acid in the sample. Reliable
The sample from which the red spectrum was information
about the
measured is cleaner and the A260 can be used amount of
for quantification. In both cases, the absorbance CyDye

spectra for Cy3 at 550 nm gives usable 260 550

information. Wavelength (nm)

7.2.2 Spiking with radioactive nucleotide

If more accurate quantification of the labelled nucleic acid is needed, the
labelling reactions can be spiked with a small amount of a radioactive
nucleotide. This enables the determination of the amount of nucleic acid
synthesized as well as the amount of sample recovered from purification.
This approach can be used with all labelling methods in which new
nucleic acid is synthesized. For labelling in cDNA synthesis, a

hb 63-0048-49 ● 84
 C H A P T E R 7 : C H A R A C T E R I Z AT I O N O F L A B E L L E D M I C R O A R R AY P R O B E S

deoxynucleotide must be used, and for labelling of synthetic RNA, a

ribonucleotide is needed. In order to minimize interference with the
labelling process, the nucleotides used as spike and label should be
different. For example, dATP is suitable for use as a spike with both
CyDye-dCTPs and CyDye-dUTPs. Only small amounts of the radioactive
nucleotide are needed—2 µCi per sample or less is adequate—and
isotopes of lower energy, such as 33P, can be used. Note that the absolute
amount of the radioactive spike is not critical, as long as the radioactivity
can be measured accurately. The low concentration of radioactive
nucleotide solutions means that the spiked nucleotide will not contribute
significantly to the total amount of that nucleotide in the reaction. The
amount of the radioactive isotope incorporated into labelled sample is
relatively small and does not interfere with the detection of fluorescence.

The information derived from the radioactive spike can be used to

estimate the amount of nucleic acid synthesized. It is reasonable to
assume that the radioactive nucleotide and unlabelled nucleotide are
incorporated by enzymes at a similar rate. Therefore, if the percentage
of the radioactive nucleotide incorporated into the synthesized product is
known, it can be concluded that the same percentage of the unlabelled
nucleotide is incorporated as well. If the total amount of this nucleotide
in the labelling mix is known, and it is assumed that all four nucleotides
are incorporated with equal efficiency (this is probably not absolutely
true), it is possible to calculate the amount of nucleic acid synthesized.
This method only estimates the amount of nucleic acid synthesized, but
in most cases gives more accurate data than UV spectrophotometry.

In order to quantify the amount of nucleic acid synthesized by spiking, a

small amount of radioactive nucleotide is added to the labelling reaction.
In order to keep the total reaction volume unchanged, the amount of water
Direction of separation pipetted into the reaction must be adjusted. The radioactive nucleotide
should not contain any colored marker or stabilization agents, as these
Bottom Top compounds are fluorescent in the same regions of visible spectrum as
of plate of plate
CyDye and can interfere with determination of CyDye amount.

Note: Because of the radioactivity present in the spike, necessary

cDNA Unincorporated precautions for working with radioactive materials must be followed.
spiked nucleotide
No other special modifications are needed for the labelling reactions.
Fig 51. Thin layer chromatography analysis After the labelling reaction has been completed and before the sample
of the incorporation of radioactive spike into is purified, the incorporation of the radioactive nucleotide into the
labelled cDNA sample. The labelled nucleic
synthesized nucleic acid needs to be determined. The incorporated
acid does not migrate far from the sample
application line, whereas the unincorporated
radioactive nucleotide can be easily separated from free nucleotides
radioactive nucleotide moves up toward the by thin layer chromatography on PEI cellulose chromatography plates
top of the plate. (Fig 51).

● 85
 M I C R O A R R AY

7.2.3 Thin layer chromatography analysis of spike

incorporation
To perform thin layer chromatography, follow the steps outlined below:

❶
Preparation
■ Prepare a 20 × 20 cm glass-backed PEI cellulose chromatography
plate (available from Merck) for use by cutting a thin linear groove
into the PEI cellulose layer at 1 cm distance from the top edge of
the plate. Mark sample positions along a line that is 3 cm from the
bottom edge of the plate.

❷
Titration
■ Pipet 0.5 µl samples of labelling reactions in duplicate along the
marked line. The sample spots should be about 1 cm apart. Do
not damage the PEI cellulose layer with the tip.

❸
Separation
■ Place the plate in a rectangular chromatography tank so that the
bottom of the plate is immersed 2 cm deep in 1 M K2HPO4. Make
sure that the level of the buffer is below the level of the marked
sample line. Cover the tank and let sample separation take place.
Newly synthesized nucleic acid will not move far from the sample
line whereas free nucleotides will move progressively towards the
top of the plate.

■ When the buffer has reached the top groove, i.e. when the
samples have migrated the full available length of the 20 cm
plate, remove the plate from the tank and let it air dry.

❹
Imaging
■ Wrap the plate in cling-film and expose it to a phosphor screen for
1 – 6 h. Take care not to overexpose the phosphor screen as it will
saturate the signal. Some trial and error may be needed to determine
correct exposure time.

■ Scan the phosphor screen on a Typhoon ™ Variable Mode Imager,

using the recommended settings. As an alternative to using phosphor
screens, any instrument that can accommodate chromatography plates
and measure radioactivity quantitatively can be used.

hb 63-0048-49 ● 86
 C H A P T E R 7 : C H A R A C T E R I Z AT I O N O F L A B E L L E D M I C R O A R R AY P R O B E S

➎
Analysis
■ Calculate the proportion of the radioactive nucleotide that is
associated with the synthesized nucleic acid. This is the incorporation
percentage. For example, the ImageQuantTM software can be used for
this purpose.

■ Calculate the yield of nucleic acid as follows:

Yield of nucleic acid =

(mol of cold nucleotide in labelling mix) ×

(incorporation percentage) × 4 × 330 g/mol

This formula assumes that all four nucleotides are incorporated

in equal proportions (hence, times 4) and that the molecular weight
of average nucleotide is 330 g/mol.

For example first-strand cDNA labelling was performed with 2 nmol

of dATP in the reaction. 20% of [α-33P]dATP was incorporated into
cDNA.

Yield of cDNA =
2 × 10-9 mol × 20% × 4 × 330 g/mol = 528 ng

● 87
 M I C R O A R R AY

7.3 Calculation of labelling density

Labelling density can be defined as the amount of CyDye incorporated
into a known amount of nucleic acid. It can be expressed as pmol of
CyDye incorporated into µg of nucleic acid, or can be converted to the
number of CyDye molecules per 100 nucleotides. Labelling density is a
measure of the average distance between CyDye fluors on the labelled
nucleic acid. Samples labelled successfully, with an optimized protocol,
will usually have similar labelling densities, but problems with the
labelling reagents or in performing the protocol can result in variation
in the incorporation of CyDye into cDNA. For example, in the cDNA
post-labelling method, exposure of the CyDye-NHS esters to moisture
during storage would result in low labelling density without affecting
the amount of cDNA synthesized.

Once the amount of CyDye incorporated into nucleic acid and the
amount of the nucleic acid are known, labelling density can be
calculated.

Labelling density = pmol CyDye in labelled sample

µg nucleic acid in labelled sample

1 µg of cDNA contains approximately 1 × 10-6 /330 = 3030 pmol of

nucleotides.

Hence labelling density of 100 pmol/µg equals 100 pmol/3030 pmol of

nucleotides = 3.3 CyDye nucleotides per 100 nucleotides = 3030 pmol of
nucleotides.

7.4 Recovering labelled nucleic acid after purification

The recovery of CyDye labelled nucleic acids from purification systems
can be variable. In order to draw conclusions about the performance
of the purification process, two aspects need to be considered. First,
the recovery of the labelled material needs to be determined. Poor
recovery can limit the amount of slides that can be hybridized with
the sample, and the true benefits of labelling methods may not be
realized. Second, the presence of free CyDye needs to be assessed.
Free CyDye in the purified probe will affect the determination of
CyDye amount in the labelled nucleic acid and can result in too little
probe being used in hybridization. Free CyDye, especially free CyDye-

recovery % = (cpm after) × (volume recovered) × 100%

(cpm before) × (incorporation %/100) × (volume purified)

hb 63-0048-49 ● 88
 C H A P T E R 7 : C H A R A C T E R I Z AT I O N O F L A B E L L E D M I C R O A R R AY P R O B E S

nucleotide, can also contribute to hybridization background and give rise

to speckled images that are difficult to quantify. Gel electrophoresis is a
convenient method for determining whether the purified probes are free
of unincorporated dye.

Radioactive spiking, as explained above, can be used to obtain information

about the purification process. Scintillation counting of a small aliquot
(0.5 – 1 µl) of labelling reaction sampled before and after purification
can be used to calculate the proportion of cDNA (or other nucleic acid)
recovered. For this calculation it is necessary to know what proportion
of the radioactive nucleotide was incorporated into the nucleic acid.
This can be determined from the unpurified labelling reaction with thin
layer chromatography analysis as explained above.

7.5 Analysis of the composition and fluorescence of

labelled sample
Gel electrophoresis can be used to investigate the quality of the labelled
sample. Separation of unincorporated CyDye from labelled nucleic acids
can be achieved with denaturing polyacrylamide gel electrophoresis (PAGE)
or agarose gel electrophoresis. After electrophoretic separation, the gel
can be scanned to detect Cy3 and/or Cy5 fluorescence using a multipurpose
scanner, such as Typhoon 9410 Variable Mode Imager. From the scanned
image it is possible to detect how much free CyDye there is present in the
sample. If PAGE gels are used, the size range of the labelled nucleic acid
population can be examined. This can give an indication of the quality of
the starting population of RNA and success of the labelling reactions. If
unpurified sample is separated alongside purified samples, it is possible
to estimate the recovery of probe from purification. By comparing the
relative fluorescence of different labelled samples, the relative amount of
CyDye in each sample can be estimated (Fig 52).

Fig 52. PAGE analysis of Cy3-labelled cDNA

probes. 1 µl samples of Cy3-labelled and
purified cDNA probes prepared with cDNA 20 x 20 cm slab gel
post-labelling method were separated in 6% RapidGel™ mix

6% sequencing gel. The gel was scanned 1 µl of Cy3-labelled cDNA

for Cy3 and Cy5 fluorescence with Typhoon ALFexpress
scanner to detect ALFexpress™ Sizer and the Cy5 sizing ladder
50-500 nt
labelled probes The cDNA probe consists of Free CyDye
fragments longer than 150 nucleotides and Bromphenol blue
some unincorporated Cy3-reactive dye is
present, as indicated. All four samples show
similar relative fluorescence and quality.

● 89
 M I C R O A R R AY

7.5.1 Analyzing CyDye labelled probes with PAGE

In order to analyze CyDye labelled probes using PAGE, follow the
procedures detailed below:

❶
Preparation
■ PAGE gels suitable for analysis of CyDye labelled samples can be
prepared from standard 6% or 8% (w/v) sequencing gel mix such
as RapidGel-XL-6%. The gel can be cast using equipment designed
for sequencing, but since single base pair resolution is not required,
slab gel instruments can be used. These provide the added benefit of
thicker and deeper wells that simplify sample loading.

■ 0.1 – 1 µl of purified labelling reaction is enough for detection of

CyDye labelled nucleic acid by fluorescence scanning. Add 2 µl of
formamide and 8 µl of water to each sample. Do not use normal
loading/denaturation buffers which contain dyes such as bromophenol
blue or xylene cyanol, as these will interfere with the detection of
CyDye fluorescence.

■ Dilute 4 µl of fluorescent markers such as ALFexpress Sizer 50-500

with 4 µl of water and 2 µl of formamide. This sizer contains a
Cy5-labelled DNA marker ladder.

❷
Denature
■ Denature all samples by boiling for 2 min at 95 °C. Snap cool
on ice before loading on to the gel.

❸
Electrophoresis
■ Load samples to a gel that has been pre-electrophoresed for
15 – 30 min. Perform electrophoresis according to the instructions
provided with your equipment. Use 1× TBE as buffer. Protect the
samples from light during the electrophoresis.

■ In order to help monitor the progress of electrophoresis, you

can load a small aliquot of loading buffer (1 µl) containing
bromphenol blue into a side well of the gel that is well separated
from the wells containing labelled cDNAs. Stop the electrophoresis
when the bromphenol dye is still well within the gel. Unincorporated
CyDye will migrate faster than bromphenol blue.

■ Remove one of the gel plates before scanning and make sure that
the back of the remaining plate is clean. Do not let the gel dry
before scanning.

hb 63-0048-49 ● 90
 C H A P T E R 7 : C H A R A C T E R I Z AT I O N O F L A B E L L E D M I C R O A R R AY P R O B E S

❹
Imaging
■ Scan the gel on a Typhoon Variable Mode Imager. Detect Cy3 by
excitation with 532 nm laser and using emission filter 555 BP 20.
Detect Cy5 by excitation with 633 nm laser and using emission filter
670 BP 30. Set PMT to 800 V, focal plane to +3 mm, and use normal
sensitivity. The PMT values may need to be adjusted to account for
different amounts of sample in the gel.

Agarose gel electrophoresis can be used instead of PAGE. Single-stranded

CyDye labelled nucleic acid fragments will not migrate true to their size
in standard agarose gels, but valid information about the purity and
fluorescence of labelled sample can be obtained. 0.1 – 1 µl of labelling
reaction can be diluted by adding 2 µl of 50% (v/v) glycerol and water to
10 µl. No loading dyes should be used. Denaturation of samples is not
necessary before gel loading. Run the gel in 1× TBE or 1× TAE.

● 91
 M I C R O A R R AY

hb 63-0048-49 ● 92
 C H A P T E R 8 : M I C R O A R R AY H Y B R I D I Z AT I O N

Chapter 8
M I C R O A R R AY H Y B R I D I Z AT I O N

8.0 Introduction
The process of hybridization is typically performed in order to identify
and quantitate nucleic acids within a larger sample. Generally, it involves
annealing a single-stranded nucleic acid to a target complementary
strand. Southern blotting is one well-established hybridization method.
In this technique, samples of genomic DNA—the target—are attached
to a membrane and incubated within a solution of fluorescently labelled
DNA—the probe (54). Binding of probe molecules to the sample on the
membrane highlights complementary sequences, and the intensity of
signal is proportional to the amount of immobilized sample.

The microarray hybridization technique works in a very similar way to

that of Southern blotting, except that it is carried out in reverse, and the
target is first attached to a slide instead of a membrane.

There are several critical factors to performing a successful microarray

hybridization. The following are discussed in detail in this chapter:

■ Pre-hybridization

■ Hybridization conditions

■ Hybridization buffer

■ Stringency washes

● 93
 M I C R O A R R AY

8.1 Overview of the microarray hybridization

process
❶
Attachment
Genes of interest are spotted onto a solid surface by the array spotter.
These are known as the targets. Attachment chemistry will often be
required to ensure that the DNA remains attached to the slide surface
throughout the hybridization process.

❷
Hybridization
Hybridization buffer containing a known amount of labelled sample
DNA—often referred to as probe— is then placed on the slide surface.
A coverslip can then be carefully placed on top of the slide.

The slide is then incubated in a humid environment for up to 16 h.

During this time the labelled probe is in contact with the targets on
the slide. If the sequence homology is good then the probe will adhere
to the target.

❸
Washing
Once the hybridization is complete, the slides are washed, and buffer and
probe of little or no homology to the target will be washed away, leaving
the labelled probe of high homology attached to the target and available
for detection.

hb 63-0048-49 ● 94
 C H A P T E R 8 : M I C R O A R R AY H Y B R I D I Z AT I O N

8.2 Pre-hybridization
Pre-hybridization consists of incubating the spotted slide in a buffer
in the absence of probe. Different slide chemistries require slightly
different pre-hybridization protocols that vary in the type of buffer
used. Consult manufacturers’ recommendations to find out what is the
best procedure for each slide. Pre-hybridization prepares the microarray
for hybridization in the following two ways:

■ Badly adhered target is washed away during pre-treatment. If this step

is omitted, this target will often wash off the slide surface during the
hybridization and will hybridize to the probe in solution, thus competing
with the immobilized targets. This can decrease hybridization signal.
■ Pre-treatment ensures that the target is available for hybridization.
The target is normally double-stranded DNA and, although targets
are frequently spotted in a denaturing solution such as DMSO, most
microarray protocols do not contain a specific denaturization step.
Pre-hybridization may also act to block any sites on the slide surface
that are capable of binding the probe nonspecifically, thus improving
the backgrounds.

8.3 Hybridization
There are several widely used methods for carrying out the hybridization,
either using automated instruments or performing the procedure manually.
In this chapter we describe the manual hybridization method, which is
the most widely used. General properties of manual hybridization will be
discussed, followed by advice for choosing a suitable hybridization method
for different microarray slide types.

8.3.1 Coverslip method

In the coverslip method, hybridization buffer containing the probe is
incubated on the microarray under a coverslip. This way only a small
volume, typically about 30 µl, of buffer is needed. The coverslip method
is a very convenient one in that it requires no special equipment. However,
a microarray slide under coverslip is prone to drying out, especially
around the edges, causing most of the array to become unusable.
Originally, coverslips were sealed on the slide. This prevented drying
but made the coverslip difficult to remove prior to detection. A more
practical approach is to carry out the hybridization in a humid
environment, thus preventing evaporation of the hybridization buffer
from beneath the coverslip. This can be achieved with anything from a

● 95
 M I C R O A R R AY

humid chamber with slide trays and a reservoir, to a plastic box lined
with wet tissues at the base (Fig 53). In either case, the atmosphere
within the chamber must remain at >95% humidity throughout the
16-h hybridization. Equally important is not to allow the slide to come in
contact with the water, which may dilute the probe or cause water marks
on the slide surface.
a)

8.3.2 Hybridization buffer

The hybridization buffer and conditions used are vital for successful
results. Hybridization buffers vary considerably but will normally
b) contain the following components:
Fig 53. Practical solutions for ensuring ■ a buffering component that acts to stabilize variations in pH
that a humid environment is maintained
during manual microarray hybridization. A ■ a detergent that acts to lower the surface tension and allow
plastic box (a) containing a platform raised the buffer to flow easily under a coverslip
above moistened tissues is sufficient for
hybridizing a few slides. Commercially ■ compounds that act as rate enhancers, volume excluders, or
available humid chamber (b) holds up to to speed up the hybridization and lower the Tm
40 slides on removable trays and fits into
most lab ovens. Melting temperature (Tm) is the temperature at which 50% of the probe
is denatured. This temperature will be affected by both the size and
G-C content of the probe fragments, but the effect is minimized by
optimizing the salt content and formulation of commercial hybridization
buffers, thus making them suitable for use with most probes without
optimization. Formamide is a denaturing reagent that is often used to
lower the Tm of the probe and hence the temperature of hybridization.
The optimum hybridization temperature for microarrays, in aqueous
buffers, will be high (65 – 75 °C). At these high temperatures drying out
of the slide becomes more of a problem; the probe is also more likely
to degrade. The addition of formamide to a buffer decreases the Tm
by 0.65 °C for every 1% concentration; therefore, the addition of 50%
formamide to the hybridization lowers the optimum temperature to a
more reasonable 42 °C (55). However, hybridizations carried out in
formamide should be left for 16 h, unless the probe concentration
is increased.

hb 63-0048-49 ● 96
 C H A P T E R 8 : M I C R O A R R AY H Y B R I D I Z AT I O N

8.3.3 Probe blocking

Most manufacturers recommend some type of probe blocking either
75 pmol probe
prior to or during the hybridization to prevent nonspecific hybridization
of probe to common genetic elements. One common blocking agent
25 pmol probe
is poly-dA oligo, which hybridizes to poly-dT tails (formed during the
cDNA probe synthesis by the poly-dA oligo, and prevents probe from
hybridizing with poly-A sequences often present in targets. Other types
8 pmol probe
provide more general forms of blocking, such as salmon sperm and yeast
tRNA to block non-specific binding and the inclusion of Cot1 DNA™
3 pmol probe
mop up repetitive sequences. The blocking agents are normally added
to the labelled probe/hybridization buffer solution prior to applying to
Fig 54. Cy3- and Cy5-labelled probes are
the slide surface. The solution can then be heated to denature any
hybridized on Amersham Biosciences
reflective slides at 75, 25, 8 and 3 pmol of double-stranded DNA and to allow the blocking to take place, before
each dye per slide. Although hybridization setting up the hybridization reaction.
signals can be detected using as little as
3 pmol of each probe, the intensity of signals
is greatly increased by using 25 pmol of each 8.3.4 Probe concentration
probe, thus allowing the rarer messages to
The amount of probe to add to a hybridization will vary, depending
be visualized.
on the samples used, the slide type, and what information is expected
to be gained. Slide manufacturers will recommend optimum probe
concentrations to use in hybridizations with their slides. If two or more
colors are being used, it is important that exactly the same amount of
probe labelled with each dye is added so as not to skew the results in
favor of one of the probes. For glass slides 30 pmol of each labelled dye
is sufficient for most systems, but this should be reduced by as much as
half when using mirrored slides, which contain a reflective layer capable
of enhancing signal intensities. Increasing the amount of probe used will
increase the result obtained but only up to a certain point (Fig 54),
beyond which the increase in background levels will actually decrease
the amount of background-corrected signal (Fig 55).

Effect of probe concentration on signal intensity

RFU
700000

600000
Fig 55. When increased amounts of
Cy3
labelled probe are used, there is an 500000
Cy5
increase in background levels. This 400000
decreases the amount of background-
300000
corrected signal detected from the
microarray slide. 200000

100000

0
3 pM 8 pM 25 pM 75 pM

Probe concentration

● 97
 M I C R O A R R AY

8.3.5 Probe depletion and target saturation

As in most manual hybridizations the reaction is carried out under a
coverslip, which means that there is sufficient solution under the coverslip
for complete coverage of the array but no room for any movement or
flow of the buffer under the coverslip during the hybridization. Whether
a labelled probe fragment finds its complementary target on the microarray
is therefore simply relying on diffusion of the probe. Research has shown
that a 20 bp oligo diffuses a distance of 3.6 mm in an 18 h period,
therefore a labelled cDNA sample is hardly going to move during an
overnight hybridization. This means that if there were several replicate
target spots concentrated in a small area on an array, these spots would
be competing for a limited amount of complementary sequences within
diffusion distance. This is called probe depletion, and it can limit the
signal obtained from microarray. This will be most relevant for those
transcripts that are present in low numbers in the labelled samples, as
the signal from these spots may fall below detection sensitivity of the
microarray system.
The sensitivity of the microarray system is determined by several factors
including the amount of label attached to the probe molecules, the level
of background signal, and the sensitivity of the scanner. Furthermore,
the rate at which the probe molecules find their targets is a more critical
determinant of sensitivity than the amount of spotted target. For most
low- to medium-abundance genes, the amount of spotted target is in a
huge excess over the probe molecules. For a high abundance gene, the
amount of probe in solution starts to approach the amount of target
present, which can lead to target saturation. Target saturation will be
determined by factors such as the amount of target initially spotted on
the slide and the amount retained on that slide after pre-treatment, as
well as the percentage of that target available for hybridization and the
efficiency of hybridization. Together, the sensitivity of detection and
target saturation determine the dynamic range of the microarray
experiment.

hb 63-0048-49 ● 98
 C H A P T E R 8 : M I C R O A R R AY H Y B R I D I Z AT I O N

8.3.6 Use of hybridization chambers

Hybridization chambers are used in order to overcome any problem
arising from the amount of buffer used under coverslips and possible
probe depletion. These are plastic chambers that hold an individual
slide and a larger volume of hybridization buffer. The amount of
labelled probe added remains the same. These are then incubated
overnight in a water bath or oven. In order to introduce a significant
amount of mixing, using an automated hybridizer is recommended.

8.3.7 Practical tips for setting up coverslip hybridization

This protocol gives instructions how to set up a microarray hybridization
using coverslips. It should be performed with clean slides and coverslips,
preferably in a clean room or under a hood. It is recommended to use
plastic coverslips that have been packed in plastic film. Do not use gloves
that contain powder, as this can easily get onto microarray slides and
cause background problems.

❶
Pre-hybridization
■ Store spotted slides in a desiccator until use.
■ Read through the pre-treatment and hybridization protocols thoroughly
before use, as buffers often need preparing and preheating before use.
■ Pretreat the number of slides required for the experiment. Some
manufacturers say that pre-treated slides can be stored before use,
but not all, so it is worth checking before doing a large batch.
■ During the pre-treatment stage, prepare the probes. This will often
involve drying down equivalent amounts of the two probes of interest
together and reconstituting them in the manufacturer’s recommended
buffer. Some protocols require heating the probe before use; the probe
prepared by reverse transcription will be single-stranded and therefore
should not require denaturing before use.

● 99
 M I C R O A R R AY

❷
Hybridization
■ Once the probe is prepared, lay the spotted slide, DNA side up, on a
clean surface. Absorbent tissue that does not release any fibers is a
good choice of surface. This is important, as most slides are glass, and
dirt on the rear of the slide will affect the result on the front of the
slide (this is obviously not an issue with opaque or mirrored slides).
■ Using a pipette, transfer the required amount of hybridization
buffer/probe mixture onto the slide. Avoid touching the slide surface
with the pipette tip. Try to deposit the mixture along the short side of
the slide, away from the spotted area.
■ Take a clean coverslip and place it on the slide near the probe
mixture, and allow surface tension to speed the buffer along the
coverslip. Then gently lower the coverslip, avoiding trapping air
bubbles underneath. There are several ways of lowering the coverslip,
two of which are illustrated in Figure 56.
Fig 56. Lowering the coverslip.
■ If air bubbles become trapped beneath a coverslip (Fig 57), do not
move the coverslip to try and remove them. Movement of the
coverslip will result in damage to the targets themselves. Most small
air bubbles will disperse once the slide is transferred to hybridization
temperature. Larger bubbles can be ‘encouraged’ to move by gently
pressing on the surface of the coverslip with a pipette tip.
■ The probe mixture is light sensitive, so once the coverslip is on, place
the slide in the humid chamber and incubate overnight in the dark.

Note: If using RNA probes, it is important to take appropriate

precautions to protect all reagents from nucleases (see Chapter 5 on
RNA handling).

8.4 Stringency washes

The purpose of the post-hybridization washes is to remove all unattached
and loosely bound probe molecules. This prevents false positive signals
a) b) and removes all components of the hybridization buffer, preventing
background noise in the form of smearing and speckles. Again, as the
Fig 57. Typical problems encountered in
slides are light sensitive at this stage, the washing steps should be carried
microarray hybridization. Trapping of air
bubbles (a) beneath the coverslip will lead out in the dark so as to minimize signal loss due to bleaching of the
to areas on the array that fail to hybridize fluorescent dyes. Once the slides have been washed, they should
at all. Allowing the slide to dry out during immediately be dried by centrifugation or nitrogen steam to prevent
the hybridization will lead to high patchy smearing while drying. The slides should then be stored in the dark in a
backgrounds (b) that may cause difficulty
desiccator and scanned as soon as possible. If, once scanned, it is found
at the analysis stage.
that the slides have high background or low stringency, it is worth re-
washing the slide and re-scanning.

hb 63-0048-49 ● 100
 C H A P T E R 8 : M I C R O A R R AY H Y B R I D I Z AT I O N

The stringency washes will affect the amount of labelled probe retained
on the slide for the final analysis. While it is obviously important to
remove all the loosely bound probe, it is important to not strip the
bound probe. Generally, stringency washes are carried out in a SSC/SDS
solutions of different concentrations, with the primary washes often
being carried out at the same temperature as the hybridization. Primary
wash solutions have a high salt content (typically 1–2× SSC/0.1% SDS
buffer), and they remove most of the hybridization buffer components.
The secondary washes are performed with low salt buffer (typically
0.1× SSC/0.1% SDS), and they will remove the loosely bound probe from
the blot. It also serves to remove any remaining salt from the primary
washes. Failing to warm solutions thoroughly before use will lower their
effectiveness and may lead to increases in background noise. Conversely,
warming solutions too much (as often happens if a microwave oven is
used) or using too low a salt concentration in the buffers, will strip
precious signal from the blot. Check the manufacturer’s protocols for
exact wash dilution volumes. Manufacturers often suggest a water dip
prior to drying the slides to prevent smearing. Check the protocols
provided with the buffer components for instructions on performing the
water dip.

8.5 Microarray slides

8.5.1 Choosing the right hybridization protocol for different

slide types
Most manufacturers of microarray slides will provide a hybridization
protocol that they have optimized for their system. The following table
lists some of the more commonly used slides and a brief summary of
tested hybridization protocols for them. This in not an exhaustive list
and the protocols are only for reference. Please refer to the
manufacturer’s own protocols prior to use.

● 101
 M I C R O A R R AY

Table 5. Microarray slide types and their characteristics.

Slide type Manufacturer Spotting Hybridization Pretreatment Hybridization
chemistry buffer protocol
Lucidea Reflective Amersham 50% DMSO Version 2 (4×)- None required 42 °C overnight
Slides Biosciences formamide based

CMT-GAPS Corning 50% DMSO 25% formamide, 25% formamide, 42 °C overnight

or 3× SSC 5× SSC, 0.1% SDS 5× SSC, 0.1%
SDS for 45
min at 42 °C

SigmaScreen™ Sigma-Aldrich 3× SSC ArrayHyb 1% SDS for 2 min, 50 °C 6 h

water rinse. Boiling overnight
water for 2 min,
ethanol dip

Type I Clontech 150 mM Na GlassHyb Optional: 50 °C overnight

phosphate 70 mM succinic
anhydride in 315 ml
1-methyl-2-pyrrolidinone
and 35 ml Na borate pH
8 – 15 min at RT. Boiling
water for 2 min. Ethanol dip.

Type II Clontech 150 mM Na GlassHyb 70 mM succinic 50 °C overnight

phosphate anhydride in 315 ml
1-methyl-2-pyrrolidinone
and 35 ml Na borate
pH 8 – 15 min at RT.
Boiling water for for 2 min.
Ethanol dip.

Super Amine Telechem Int. 5× SSC UniHyb 0.1% SDS twice 42 – 65 °C

(1.25×) for 2 min at overnight
RT followed by
a water rinse.
Incubate in
bolling water for
3 min before
drying.

hb 63-0048-49 ● 102
 CHAPTER 9 : LUCIDEA SLIDEPRO HYBRIDIZER

Chapter 9
LUCIDEA SLIDEPRO HYBRIDIZER

9.0 Introduction
A complex system, microarray analysis is affected by a number of
experimental factors, including target preparation, physical deposition of
the targets, slide chemistry, probe chemistry, hybridization, and detection
of the fluorescent signal (37). In order to improve the efficiency of
microarray analysis, each source of variation must be eliminated or
minimized. Control of environmental conditions during hybridization in
particular, is critical in producing and maintaining a consistent fluorescent
signal. Lucidea SlidePro was designed to overcome the problems associated
with hybridization variability.

Fig 58. Lucidea SlidePro Hybridizer. 9.1 Features of Lucidea SlidePro Hybridizer
Lucidea SlidePro Hybridizer (Fig 58) has a modular format that consists
of a base control unit, up to four additional modules, and control
software on a laptop computer. Each unit contains six individually
temperature-controlled chambers, each of which holds a standard
microscope slide. With all five modules a total of 30 slides can be
processed. The multi-module protocol software allows each module to
be started at different times and different experiments can be conducted
on each module, thereby increasing user flexibility. Also, each module
has its own pump, which allows a faster processing time.
Heating block
Slide Lucidea SlidePro is capable of automating a variety of chemical
O-ring seal O-ring seal
PEEK chamber Port and biochemical techniques in which incubation and wash steps are
(100 µm recess)
performed at varying temperatures. It is primarily used in microarray
Outlet port analysis to automate the pre-hybridization, hybridization, and washing
of microarray slides. It has been designed to be used in conjunction with
the Lucidea range of microarray instruments and reagents.
Hinged clamp
In Lucidea SlidePro, each slide is held in a chamber sealed with a patented
O-ring. Pre-treatment and wash solutions are drawn into the chamber,
from up to five reservoirs, and deposited into a waste bottle. Each
Inlet port module can run off one set of wash botttles, or multiple modules can run
off the same set of wash bottles. Hybridization samples are injected
Septum
injection port through a septum port at the lower end of the chamber (Fig 59).
Side view Bottom view
of chamber

Fig 59. Schematic of individual slide chamber.

● 103
 M I C R O A R R AY

9.2 Benefits of Lucidea SlidePro Hybridizer

The key benefits of Lucidea SlidePro include:

■ Improved uniformity of signals and Cy3/Cy5 ratios within and

between slides.

■ Temperature and mixing controls for each chamber. Small volumes

are drawn in and out of the chamber to provide continuous mixing.
The volume, speed, and length of mixing can also be individually
controlled, making optimization easier.

■ Temperature and wash solution controls for each chamber. This

improves the reproducibility of signals both within and between
slides and from user to user. Greater reproducibility can increase
the accuracy of results with fewer numbers of replicates.

■ Enhanced detection of rare messages. The mixing of probe during

hybridization results in enhanced signals, without the need to use
increased amounts of probe as compared with manual hybridization.
Increased signal strength compared to background enables detection
of low signals from rare messages.

■ Rapid start-up time and ease of use. Experiments to determine optimal

hybridization parameters, such as temperature and washing, are
performed with ease. Standard protocols are provided to decrease time
to optimize experimental procedures. Different conditions can be tested
within a single run, using up to 30 slides with five modules. The
software has a help feature for fast start-up and troubleshooting.

■ Facilitates reuse of probe. Probe can be removed from the hybridization

chamber via the injection port, allowing samples to be reused multiple
times. The instrument is paused after hybridization and the probe
removed with a syringe. The instrument continues with washing and
drying of slides. While overall signals decrease with probe reuse, the
Cy3/Cy5 ratios are not significantly altered.

■ Reduced demands on user time. The user needs only to load slides and
inject the probe. Following automated hybridization, washing, and
drying, the slides are removed from the instrument ready to scan.

hb 63-0048-49 ● 104
 CHAPTER 9 : LUCIDEA SLIDEPRO HYBRIDIZER

9.3 Validation in microarray hybridization

We describe here how a number of hybridization factors—including time,
probe reuse and probe concentration—affect the signal-to-noise values
detected from microarray experiments. Despite the effects of these factors
on overall signal strength, the Cy3/Cy5 ratio remains constant, suggesting
that differential expression may be determined under conditions which
do not provide an optimal signal. Three experimental applications were
used to demonstrate the utility of Lucidea SlidePro.

9.3.1 Comparison of automated and manual hybridization

Coefficient of variation Twenty-four standard silanized glass slides were spotted with p53
Type 5 Lucidea Manual cDNA. Twelve slides were hybridized manually (approximately 30 µl of
SlidePro hybridization sample was placed on a microarray slide, under a coverslip
and incubated in a humidified container in a hybridization oven) and
Within slide 8.6% 35.9% twelve were hybridized in Lucidea SlidePro. All were hybridized using
Between slides 4.1% 10.7% Cy3-labelled human skeletal muscle and Cy5-labelled human skeletal
Total 12.7% 46.6% muscle.

Hybridization in Lucidea SlidePro produced increased signal intensity

Fig 60. Lucidea SlidePro vs manual ANOVA and more consistent Cy3/Cy5 ratios with very low variability compared
coefficient of variation values for glass slides to the manual method. Analysis of variance (ANOVA) comparison of the
hybridized in Lucidea SlidePro or manually.
Cy3/Cy5 ratios showed significantly less variation in the Lucidea SlidePro
processed slides compared to manually processed slides (Fig 60).

● 105
 M I C R O A R R AY

Results suggest that hybridization efficiency and data reproducibility

could be improved using Lucidea SlidePro as compared to the manual
methodology (Fig 61-62).

9.3.2 The effect of probe mixing

Since diffusion rates on solid surfaces are much lower than those in
solution, compensation for localized depletion of probe may not occur
within the time frame of a hybridization (37, 56). Lucidea SlidePro
provides mixing during hybridization, which ensures a constant probe
concentration and thereby eliminates depletion effects.

Initial experiments were designed to determine whether mixing could

enhance hybridization efficiency. The relative signals of a serial dilution
of target hybridized under static or mixing conditions were assessed
(Fig 63). Despite greater than 300-fold dilution of probe, mixing during
hybridization enhanced the signal detected around 5-fold at the highest
concentration of target. This result suggests that localized probe
depletion effects could be reduced by mixing the sample.

RFU
800000

Cy3 manual
Cy3 Lucidea SlidePro
600000
Cy5 manual
Cy5 Lucidea SlidePro

400000

200000

0
1 2 3 4 5 6 7 8 9 10 11 12
Slide

Fig 61. Lucidea SlidePro vs manual mean signal intensities.

Lucidea SlidePro vs manual mean ratios

Ratio
1
Manual ratio average
Lucidea SlidePro ratio average
0.8

0.6

0.4

0.2

0
1 2 3 4 5 6 7 8 9 10 11 12
Slide

Fig 62. Lucidea SlidePro vs manual mean Cy3/Cy5 ratios.

hb 63-0048-49 ● 106
 CHAPTER 9 : LUCIDEA SLIDEPRO HYBRIDIZER

RFU Constant probe concentration

6.E+06

5.E+06

4.E+06 with mixing

no mixing
Fig 63. The effect of mixing on 3.E+06

hybridization signal. 2.E+06

1.E+06

0.E+00
2 50 100 150 200

Pg/spot PCR product deposited

9.3.3 Reuse of probe

Since probe is often generated from mRNA samples that are in short
supply, it may be desirable to use a labelled probe for multiple rounds
of hybridization. This is possible using Lucidea SlidePro, since probe can
be removed directly through the injection port following hybridization.
The effect of probe reuse on overall signal strength and Cy3/Cy5 ratios
was assessed. Duplicate reflective slides were hybridized in Lucidea
SlidePro with 200 µl of Cy3-labelled skeletal muscle and Cy5-labelled
placenta cDNA (40 pmol total) per chamber. Following hybridization
and immediately prior to washing the slides, probe was recovered from
all chambers using a syringe/needle through the injection port.
Approximately 60% of the injection volume (120 µl) was recovered
from each chamber. Each probe was then reconstituted to the original
volume of 200 µl in 1× version 2 hybridization buffer and reinjected
into chambers containing fresh slides. This constituted the first reuse of
probe. The procedure was repeated for the second reuse.

Although the signal strength decreased over multiple hybridizations

(Fig 64), the Cy3/Cy5 ratios were relatively unaffected (Fig 65). This
suggests that the individual Cy3/Cy5 ratios for each gene should remain
constant, provided the signal is strong enough to be detected above
background. Furthermore, overall background was reduced with
multiple rounds of hybridization, providing an additional benefit.

Fig 64. Reuse of probe. Scanned

images showing arrays following
overnight hybridization in Lucidea
SlidePro with fresh probe, first r
euse, and second reuse.

Fresh probe (day 1) Reuse 1 (day 2) Diluted reuse (day 3)

● 107
 M I C R O A R R AY

9.3.4 Length of hybridization

Standard microarray protocols call for overnight hybridization
Scatter plot Cy3/Cy5 log ratio (12–18 h). Time course studies suggest that effective hybridization can
Normalized log ratio
take much longer than this (data not shown). In order to assess whether
1 Lucidea SlidePro enhances the kinetics of hybridization, the relative
signal strength following different hybridization times was investigated.
Arrayed reflective slides were hybridized with 20 pmol each of Cy3
0
skeletal muscle and Cy5 placenta probe per slide. Total mean signal
intensities (Fig 66) increased with time up to 16 h hybridization, but
-1 Cy3/Cy5 ratios (Fig 67) remained consistent between the 3-h, 6-h, and
15-h timepoints.
-2 Another set of microarray slides were spotted with Lucidea Universal
Fresh Reuse 1 Reuse 2
ScoreCard dynamic range controls (control elements that are used to
Probe
evaluate the dynamic range and sensitivity of the system), and the
Fig 65. Reuse of probe. Cy3/Cy5 ratios of
amount of probe was increased to 60 pmol labelled probe per slide.
selected genes following hybridization Slides were hybridized for 3, 6, 9, 12, and 15 h in Lucidea SlidePro.
with fresh probe, reuse 1, and reuse 2. Mean signal intensities were similar (Fig 68) for all hybridization times
with dynamic ranges representing 2000 copies (DR 2), 200 copies
(DR 3), and 20 copies (DR 5). These results suggest that shorter
hybridization times may be used for higher throughput, provided that
sufficient signal is obtained to detect low expressing genes of interest.
Furthermore, increased probe concentrations may be used to decrease
the hybridization time required for signal detection provided that the
mean signal above background is not compromised.

Total Cy3 signal intensity with Total Cy5 signal intensity with
hybridization time hybridization time

500000 500000

400000 500000

Fig 66. Effect of hybridization time. 300000 500000

Total mean Cy3 and Cy5 signals
following 3-h, 6-h, and 15-h 200000 500000
hybridization in Lucidea SlidePro.
100000 500000

0 0
3h 6h 15 h 3h 6h 15 h

hb 63-0048-49 ● 108
 CHAPTER 9 : LUCIDEA SLIDEPRO HYBRIDIZER

Scatter plot Cy3/Cy5 log ratio

Normalized log ratio

Fig 67. Effect of hybridization time. Cy3/Cy5

1
ratios of selected genes following 3-h, 6-h,
and 15-h hybridization in Lucidea SlidePro.
0

-1

-2
3h 6h 15 h

Hybridization

Menu RFU

10000000

1000000
Cy3 mean
Cy5 mean
100000

10000

1000
3 6 9 12 15 3 6 9 12 15 3 6 9 12 15 15
Control
DR2 DR3 DR5
(20 pmol)
Hybridization time (hr)

Fig 68. Effect of hybridization time. Total mean Cy3 and Cy5
signals of dynamic range controls.

● 109
 M I C R O A R R AY

9.4 Using different types of microarray slides

The dimensions of the slide chamber are specific and only standard
microscope slides—25–25.5 mm × 75.5–76.0 mm, and 0.95–1.15 mm
thick—may be used with Lucidea SlidePro. The hybridization chamber
will fit arrays of up to 20 × 59 mm, allowing for a barcode on one end of
the slide. The array should be spotted at least 2.5 mm from the edge. It is
important that the bar code does not exceed 10 mm in width, or it will
lie under the O-ring seal and cause leakage.

Lucidea SlidePro has been optimized for use with Amersham Biosciences
version 2 hybridization buffer and microarray slides. Lucidea SlidePro
can also be used to hybridize other manufacturer’s slides (Fig 69).
These protocols can be found on the Amersham Biosciences web site
(www.amershambiosciences.com).

a) d)
Fig 69. Lucidea SlidePro hybridization
with various commercially available
slides.
b) e)

c)
T28 direct labeled
Cy3 heart / Cy5 muscle

hb 63-0048-49 ● 110
 C H A P T E R 1 0 : F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY A N A LY S I S

Chapter 10
F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY
A N A LY S I S

10.0 Introduction
All fluorescence imaging systems require the following key elements
(Fig 70):

■ Excitation source

■ Light delivery optics

■ Light collection optics

■ Filtration of the emitted light

■ Detection, amplification and digitization of the emitted light

In this chapter, various types of scanner systems are discussed. Their

light delivery and light collection mechanisms, signal detection and
amplification, and overall performance are detailed, as well as criteria for
selecting appropriate fluorochromes and filters for use with the scanner.

Sample

Fig 70. Components of a general

fluorescence imaging system. Light delivery optics Light collection optics

Emission filter
Excitation source

Detection and
amplification

● 111
 M I C R O A R R AY

10.1 Requirements of a fluorescence imaging system

532
10.1.1 Excitation sources and light delivery optics
Relative output (a.u.)

Light energy is essential to fluorescence. Light sources fall into two broad
categories—wide-area, broad-wavelength sources, such as UV and xenon
arc lamps, and line sources with discrete wavelengths, such as lasers
(Fig 71). Broad-wavelength excitation sources are used in fluorescence
spectrometers and camera imaging systems. Although the spectral output
300 500 700 900 1100 of a lamp is broad, it can be tuned to a narrow band of excitation light
Wavelength (nm) with the use of gratings or filters. In contrast, lasers deliver a narrow
beam of collimated light that is predominantly monochromatic. In most
camera systems, excitation light is delivered to the sample by direct
Fig 71. Spectral output of light from a xenon
illumination of the imaging field, with the excitation source positioned
lamp and Nd:YAG laser. The “relative output”
axis is scaled arbitrarily for the two light sources. either above, below, or to the side of the sample.
The 532-nm line of the Nd:YAG laser is shown
Laser-based imaging systems, on the other hand, use more sophisticated
in green.
optical paths, comprising mirrors and lenses, to direct the excitation
beam to the sample. Some filtering of the laser light may also be required
before the excitation beam is directed to the sample. For microarray
applications, laser-based instruments are substantially favored, therefore
CCD scanners will not be discussed.

10.1.2 Light collection optics

High-quality optical elements, such as lenses, mirrors, and filters, are
integral components of any efficient imaging system. Optical filters are
typically made from laminates of multiple glass elements. Filters can be
coated to selectively absorb or reflect different wavelengths of light,
thus creating the best combination of wavelength selection, linearity,
and transmission properties.

10.1.3 Filtration of the emitted light

Although emitted fluorescent light radiates from a fluorochrome in all
directions, it is typically collected from only a relatively small cone angle
on one side of the sample. For this reason, light collection optics must be
as efficient as possible. Any laser light that is reflected or scattered by
the sample must be rejected from the collection pathway by a series of
optical filters. Emitted light can also be filtered to select only the range
or band of wavelengths that is of interest to the user. Systems that
employ more than one detector require additional beam splitter filters
to separate and direct the emitted light along separate paths to the
individual detectors.

hb 63-0048-49 ● 112
C H A P T E R 1 0 : F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY A N A LY S I S

10.1.4 Detection, amplification, and digitization

For detection and quantification of emitted light, either a photomultiplier
tube (PMT) or a charge-coupled device (CCD) can be used. In both cases,
photon energy from emitted fluorescent light is converted into electrical
energy, thereby producing a measurable signal that is proportional to
the number of photons detected. After the emitted light is detected and
amplified, the analog signal from a PMT or CCD detector is converted to
a digital signal. The process of digitization turns a measured continuous
analog signal into discrete numbers representing intensity levels. The
number of intensity levels available is based on the digital resolution of
the instrument, which is usually given as a number of bits, which increases
exponentially by two. 8-bit, 12-bit, and 16-bit digital files correspond to
the number of intensity levels allocated within that image file (256, 4096
and 65 536, respectively).

Digital resolution defines the ability to resolve two signals with similar
intensities. Since only a limited number of intensity levels are available, it
is unavoidable that this conversion process introduces a certain amount
of error. To allow ample discrimination between similar signals and to
keep the error as low as possible, the distribution of the available
intensity levels should correspond well to the linear dynamic range of a
detector. There are two methods of distributing intensity levels. A linear
(even) distribution has the same spacing for all the intensity levels,
allowing measurement across the dynamic range with the same absolute
accuracy. However, relative digitization error increases as signals become
smaller. A non-linear distribution (e.g. logarithmic or square root
functions) divides the lower end of the signal range into more levels
while combining the high end signals into fewer intensity levels. Thus,
the absolute accuracy decreases with higher signals, but the relative
digitization error remains more constant across the dynamic range.

● 113
 M I C R O A R R AY

10.2 Scanner systems

10.2.1 Excitation sources

Most fluorescence scanner devices used in life science research employ
laser light for excitation. A laser source produces a narrow beam of
highly monochromatic, coherent, and collimated light. The combination
of focused energy and narrow beam-width contributes to the excellent
sensitivity and resolution possible with a laser scanner. The active
medium of a laser—the material that is made to emit light—is commonly
a solid state (glass, crystal), liquid, or gas (57). Gas lasers and solid-state
lasers both provide a wide range of specific wavelength choices for
different imaging needs. Other light sources used in imaging systems
include light emitting diodes (LEDs), which are more compact and less
expensive than lasers, but produce a wide-band, low-power output.

Lasers
There are several commonly used types of lasers.

■ Argon ion lasers produce a variety of wavelengths including

457 nm, 488 nm and 514 nm that are useful for excitation of
many common fluorochromes, such as fluorescein and Cy2.

■ Helium neon or HeNe lasers, which generate a single wavelength

of light (633 nm), are popular in many laser scanners, and
can be used to excite Cy5.

■ Neodymium:Yttrium Aluminum Garnet (Nd:YAG) solid-state

lasers, when frequency-doubled, generate a strong line at
532 nm which can be used to excite Cy3.

■ Diode lasers (or semiconductor diode lasers) are compact lasers.

Because of their small size and light weight, these light sources
can be integrated directly into the scanning mechanism of a
fluorescence imager. Diode lasers are inexpensive and are generally
limited to wavelengths above 635 nm.

Light Emitting Diodes (LEDs)

As an alternative to lasers, the LED produces an output with a much
wider bandwidth (over 60 nm) and a wide range of power from low to
moderate output. Because LED light emissions are doughnut-shaped, and
not collimated, the source must be mounted very close to the sample
using lenses to tightly focus the light. LEDs are considerably smaller,
lighter, and less expensive than lasers. They are available in the visible
wavelength range above 430 nm.

hb 63-0048-49 ● 114
 C H A P T E R 1 0 : F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY A N A LY S I S

10.3 Excitation light delivery

Because light from a laser is well-collimated and of sufficient power,
delivery of excitation light to the sample is relatively straightforward,
with only negligible losses incurred during the process. For lasers that
produce multiple wavelengths of light, the desired line(s) can be selected
by using filters that exclude unwanted wavelengths, while allowing the
selected line to pass at a very high transmission percentage. Excitation
filters are also necessary with single-line lasers, as their output is not
100% pure. Optical lenses are used to align the laser beam, and mirrors
can be used to redirect the beam within the instrument. One of the main
considerations in delivering light using a laser scanning system is that the
light source is a point, while the sample typically occupies a relatively
large two-dimensional space. Effective sample coverage can be achieved
by rapidly moving the excitation beam across the sample in two
dimensions. There are two ways to move and spread the point source
across the sample, which are discussed below.

10.3.1 Galvanometer-based systems

Galvanometer-based systems use a small, rapidly oscillating mirror to
deflect the laser beam, effectively creating a line source (Fig 72). By using
relatively simple optics, the beam can be deflected very quickly, resulting
in a short scan time. Compared to confocal systems, galvanometer-based
scanners are useful for imaging thick samples due to the ability to collect
more fluorescent signal in the vertical dimension. However, since the

Sample

Fig 72. Galvanometer-controlled scanning

mechanism. Light is emitted from the laser
in a single, straight line. The galvanometer Sample tray

mirror moves rapidly back and forth

redirecting the laser beam and illuminating
the sample across its entire width (X-axis).
The f-theta lens reduces the angle of the
excitation beam delivered to the sample.
The entire sample is illuminated either by the f-theta lens

galvanometer mechanism moving along the

length of the sample (Y-axis) or the sample Galvo mirror

moving relative to the scanning mechanism.

Laser

● 115
 M I C R O A R R AY

excitation beam does not illuminate the sample from the same angle in
every position, a parallax effect can result. The term parallax here refers
to the shift in apparent position of targets, predominately at the outer
boundaries of the scan area. Additionally, the arc of excitation light
created by the galvanometer mirror produces some variations in the
effective excitation energy reaching the sample at different points across
the arc. These effects can be minimized with an f-theta lens, but when the
angle of incident excitation light varies over the imaging field, some
spatial distortion can still occur in the resulting image.

10.3.2 Moving-head scanners

Moving-head scanners use an optical mechanism that is equidistant
from the sample. This means that the angle and path length of the
excitation beam is identical at any point on the sample (Fig 73). This
eliminates variations in power density and spatial distortion common
with galvanometer-based systems. Although scan times are longer with
a moving-head design, the benefits of uniformity in both light delivery
and collection of fluorescence are indispensable for accurate signal
quantification. For microarray scanners an alternative method is to move
the stage that contains the microarray slide. In some scanners the stage is
moved in one direction while the galvanometer moves the laser beam
across in the second dimension.

10.4 Light collection

The light collection optics in a scanner system must be designed to
efficiently collect as much of the emitted fluorescent light as possible.
Laser light that is reflected or scattered by the sample is generally

Sample
Glass platen

Fig 73. Moving-head scanning mechanism.

The light beam from the laser is folded by a
Scan head rail Lens
series of mirrors and ultimately reflected onto
the sample. The sample is illuminated across
its width as the scan head moves along the
scan head rail (X-axis). The entire sample is Mirror
illuminated by the scan head, laser, and
mirrors tracking along the length of the
Scan head
sample (Y-axis).

Mirror

Laser

hb 63-0048-49 ● 116
 C H A P T E R 1 0 : F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY A N A LY S I S

rejected from the collection pathway by a laser-blocking filter, the design

of which is to exclude the light produced by the laser source, while
passing all other emitted light. Light collection schemes vary depending
on the nature of the excitation system. With galvanometer systems, the
emitted fluorescence must be gathered in a wide line across the sample.
This is usually achieved with a linear lens (fiber bundle or light bar),
positioned beneath the sample, that tracks with the excitation line,
collecting fluorescence independently at each pixel. Although this system
is effective, it can produce image artifacts. At the edges of the scan area
where the angle of the excitation beam, relative to the sample, is farthest
from perpendicular, some spatial distortion may occur. Where very high
signal levels are present, stimulation of fluorescence from sample areas
that are adjacent to the pixel under investigation can result in an
inaccurate signal measurement from that pixel, an artifact known as
flaring or blooming.

With moving-head systems, emitted light is collected directly below the

point of sample excitation. Again, it is important to collect as much of
the emitted light as possible to maintain high sensitivity. This can be
achieved by using large collection lenses, or lenses with large numerical
apertures (NA). Since the NA is directly related to the full angle of the
cone of light rays that a lens can collect, the higher the NA, the greater
the signal resolution and brightness (58). Moving-head designs can also
include confocal optical elements that detect light from only a narrow
vertical plane in the sample. This improves sensitivity by focusing and
collecting emission light from the point of interest while reducing the
background signal and noise from out-of-focus regions in the sample
(Fig 74). Additionally, the parallel motion of moving head designs
removes other artifacts associated with galvanometer-based systems,
such as spatial distortion and the flaring or blooming associated with
high activity samples.

Sample
Glass platen

Fig 74. Illustration of confocal optics.

Fluorescence from the sample is collected
by an objective lens and directed toward a Objective lens

pinhole aperture. The pinhole allows the

emitted light from a narrow focal plane
(red solid lines) to pass to the detector,
while blocking most of the out-of-focus Pinhole
light (black dashed lines).

Detector

● 117
 M I C R O A R R AY

10.5 Signal detection and amplification

The first stage in fluorescent signal detection is selection of only the
desired emission wavelengths from the label or dye. In single-channel
or single-label experiments, emission filters are designed to allow only
a well-defined spectrum of emitted light to reach the detector. Any
remaining stray excitation or scattered light is rejected. Because the
intensity of the laser light is many orders of magnitude greater than the
emitted light, even a small fraction of laser light reaching the detector
will significantly increase background. Filtration is also used to reduce
background fluorescence or inherent autofluorescence originating
from either the sample itself or the sample matrix gel, membrane, or
microplate. In multichannel or multi-label experiments using
instrumentation with dual detectors, additional filtering is required
upstream of the previously described emission filter. During the initial
stage of collection in these experiments, fluorescence from two different
labels within the same sample is collected simultaneously as a mixed
signal. A dichroic beam splitter must be included to spectrally resolve
the contribution from each label and then direct the light to appropriate
emission filters (Fig 75). At a specified wavelength, the beam splitter
partitions the incident fluorescent light beam into two beams, passing
one and reflecting the other. The reflected light creates a second channel
that is filtered independently and detected by a separate detector. In this
way, the fluorescent signal from each label is determined accurately in
both spatial and quantitative terms.

Emission filter

Fig 75. Use of a beam splitter or dichroic Mirror

PMT
filter with two separate PMTs. Light from a Short wavelength
dual color sample enters the emission optics
as a combination of wavelengths. A dichroic
beam splitter distinguishes light on the basis
of wavelength. Wavelengths above the beam
splitter range pass through, those below are
reflected. In this way two channels are Emitted light
PMT
created. These two channels can then be Long wavelength
Beamsplitter
filtered and detected independently.
Emission filter

hb 63-0048-49 ● 118
 C H A P T E R 1 0 : F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY A N A LY S I S

After the fluorescent emission has been filtered and only the desired
wavelengths remain, the light is detected and quantified. Because the
100 intensity of light at this stage is very small, a PMT must be used to detect
it. In the PMT, photons of light hit a photocathode and are converted
into electrons which are then accelerated in a voltage gradient and
multiplied from 106 to 107 times. This produces a measurable electrical
Cathode radiant sensitivity (mA/W)

10
signal that is proportional to the number of photons detected. The
response of a PMT is typically useful over a wavelength range of
300–800 nm (Fig 76). High-performance PMTs extend this range to
1
200–900 nm.

0.1
10.6 System performance
The performance of a laser scanner system is described in terms of system
resolution, linearity, uniformity, and sensitivity.
0.01 Resolution can be defined in terms of both spatial and amplitude
100 200 300 400 500 600 700 800 900 1000

Wavelength (nm)
resolution. Spatial resolution of an instrument refers to its ability to
distinguish between two very closely positioned objects. It is a function
Fig 76. An example of the response of a of the diameter of the light beam when it reaches the sample and the
PMT versus wavelength. distance between adjacent measurements. Spatial resolution is dependent
on, but not equivalent to, the pixel size of the image. Spatial resolution
improves as pixel size is reduced. Systems with higher spatial resolution
can not only detect smaller objects, but can also discriminate more
accurately between closely spaced targets. However, an image with a
100-µm pixel size will not have a spatial resolution of 100 µm. The
pixel size refers to the collection sampling interval of the image.
According to a fundamental sampling principle, the Nyquist Criterion,
the smallest resolvable object in an image is no better than twice the
sampling interval (59). Thus, to resolve a 100-µm sample, the sampling
interval must be at most 50 µm. Amplitude resolution, or gray-level
quantification, describes the minimum difference that is distinguishable
between levels of light intensity (or fluorescence) detected from the
sample (60). For example, an imaging system with 16-bit digitization can
resolve and accurately quantify 65 536 different values of light intensity
from a fluorescent sample.

● 119
 M I C R O A R R AY

Linearity of a laser scanner is the signal range over which the instrument
yields a linear response to fluorochrome concentration and is therefore
useful for accurate quantification. The linear dynamic range can be
defined in at least 3 ways:

1) the electronic dynamic range of the scanner

2) the chemical dynamic range of the fluorescent dyes used

3) the biological dynamic range of the system under study

A scanning system with a wide dynamic range can detect and accurately
quantify signals from both very low- and very high-intensity targets in
the same scan. The linear dynamic range of most laser scanner
instruments is between 104 and 105.

Uniformity across the entire scan area is critical for reliable quantitation.
A given fluorescent signal should yield the same measurement at any
position within the imaging field. Moving-head scanners, in particular,
deliver flat-field illumination and uniform collection of fluorescent
emissions across the entire scan area.

Detection limit is the minimum amount of sample that can be detected

by an instrument at a known confidence level. From an economical
standpoint, instruments with better detection limits are more cost-
effective because they require less fluorescent sample for analysis.

10.7 Fluorochrome and Filter Selection

To generate fluorescence, excitation light delivered to the sample must
be within the absorption spectrum of the fluorochrome. Generally, the
closer the excitation wavelength is to the peak absorption wavelength
of the fluorochrome, the greater the excitation efficiency. Appropriate
filters are usually built into scanner instruments for laser line selection
and elimination of unwanted background light. Fixed or interchangeable
optical filters that are suitable for the emission profile of the
fluorochromes are then used to refine the emitted fluorescence, such
that only the desired wavelengths are passed to the detector. Matching
a fluorochrome label with a suitable excitation source and emission
filter is the key to optimal detection efficiency.

10.7.1 Types of emission filters

The composition of emission filters used in fluorescence scanners and
cameras ranges from simple colored glass to glass laminates coated with
thin interference films. Coated interference filters generally deliver
excellent performance through their selective reflection and transmission
effects. Three types of optical emission filters are commonly used.

hb 63-0048-49 ● 120
 C H A P T E R 1 0 : F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY A N A LY S I S

Long-pass (LP) filters pass light that is longer than a specified wavelength
and reject all shorter wavelengths. A good quality long-pass filter is
100 characterized by a steep transition between rejected and transmitted
Transmission (%)

80 wavelengths (Fig 77a). Long-pass filters are named for the wavelength
60 at the midpoint of the transition between the rejected and transmitted
cutoff point
40 light (cutoff point). For example, the cutoff point in the transmission
20 spectrum of a 560 LP filter is 560 nm, where 50% of the maximum
0 transmittance is rejected. The name of a long-pass filter may also
550 560 570 580 590 600

a) Wavelength (nm)
include other designations, such as OG (orange glass), RG (red glass),
E (emission), LP (long-pass), or EFLP (edge filter long-pass). OG and RG
are colored glass absorption filters, whereas E, LP, and EFLP filters are
100 coated interference filters. Colored glass filters are less expensive and
Transmission (%)

80 have more gradual transition slopes than coated interference filters.

60

40
cutoff point Short-pass (SP) filters reject wavelengths that are longer than a specified
20
value and pass shorter wavelengths. Like long-pass filters, short-pass
0
filters are named according to their cutoff point. For example, a 526 SP
500 510 520 530 540 550
filter rejects 50% of the maximum transmittance at 526 nm (Fig 77b).
b) Wavelength (nm)
Band-pass (BP) filters allow a band of selected wavelengths to pass
Fig 77. Transmission profiles for a (a) 560-nm
through, while rejecting all shorter and longer wavelengths. Band-pass
long-pass and a (b) 526-nm short-pass filter. filters provides very sharp cutoffs with very little transmission of the
The cutoff points are noted. rejected wavelengths. High-performance band-pass filters are also
referred to as Discriminating Filters (DF). The name of a band-pass filter
is typically made up of two parts:

■ the wavelength of the band center (the 670 BP 30 filter passes a

band of light centered at 670 nm [Fig 78]);
100
■ the full-width at half-maximum transmission (FWHM) (the
80
670 BP 30 filter passes light over a wavelength range of 30 nm
Transmission (%)

60
FWHM
[655–685 nm] with an efficiency equal to or greater than half
40 the maximum transmittance of the filter).
20
Band-pass filters with an FWHM of 20–30 nm are optimal for most
0
650 660 670 680 690 700
fluorescence applications, including multi-label experiments. Filters
Wavelength (nm) with FWHMs greater than 30 nm allow collection of light at more
wavelengths and give a higher total signal; however, they are less able
Fig 78. Transmission profile for a band-pass to discriminate between closely spaced, overlapping emission spectra
(670 BP 30) filter. The full-width at half- in multichannel experiments. Filters with FWHMs narrower than
maximum (FWHM) transmission of 30 nm 20 nm transmit less signal and are most useful with fluorochromes
is indicated by the arrows.
with very narrow emission spectra.

● 121
 M I C R O A R R AY

10.8 Using emission filters to improve sensitivity and

linearity range
When selectable emission filters are available in an imaging system,
filter choice will influence the sensitivity and dynamic range of an assay.
In general, if image background signal is high, adding an interchangeable
filter may improve the sensitivity and dynamic range of the assay. The
background signal from some matrices (gels and membranes) has a broad,
relatively flat spectrum. In such cases, a band-pass filter can remove the
portion of the background signal comprising wavelengths that are longer
or shorter than the fluorochrome emissions. By selecting a filter that
transmits a band at or near the emission peak of the fluorochrome of
interest, the background signal is typically reduced with only slight
attenuation of the signal from the fluorochrome. Therefore, the use of an
appropriate band-pass filter should improve the overall signal-to-noise
ratio (S/N).

To determine if a filter is needed, scans should be performed with and

without the filter while other conditions remain constant. The resulting
S/N values should then be compared to determine the more efficient
configuration. Interchangeable filters can also be used in fluorescence
scanners to attenuate the sample signal itself so that it falls within the
linear range of the system. Although scanning the sample at a reduced
PMT voltage can attenuate the signal, the response of the PMT may not
be linear if the voltage is set below the instrument manufacturer’s
recommendation. If further attenuation is necessary to prevent saturation
of the PMT, the addition of an appropriate emission filter can decrease
the signal reaching the detector.

hb 63-0048-49 ● 122
 C H A P T E R 1 0 : F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY A N A LY S I S

10.9 General guidelines for selecting fluorochromes

and filters
532
490 552

10.9.1 Single-color imaging

Excitation efficiency is usually highest when the fluorochrome’s absorption
Excitation

•
maximum correlates closely with the excitation wavelength of the imaging
system. However, the absorption profiles of most fluorochromes are
• rather broad, and some fluorochromes have a second (or additional)
300 350 400 450 500 550 600 absorption peak or a long “tail” in their spectra. It is not mandatory that
Wavelength (nm) the fluorochrome’s major absorption peak matches exactly the available
excitation wavelength for efficient excitation. For example, the absorption
Fig 79. Excitation of fluorescein (green) and maxima of the fluorescein and Cy3 fluorochromes are 490 nm and 552
Cy3 (orange) using 532-nm laser light. The
nm respectively (Fig 79). Excitation of either dye using the 532-nm
absorption spectra of Cy3 and fluorescein are
wavelength line of the Nd:YAG laser may seem to be inefficient, since
overlaid with the 532-nm wavelength line of
the Nd:YAG laser. the laser produces light that is 40 nm above the absorption peak of
fluorescein and 20 nm below that of Cy3. In practice, however, delivery
of a high level of excitation energy at 532 nm does efficiently excite both
fluorochromes. For emission, selecting a filter that transmits a band at or
near the emission peak of the fluorochrome generally improves the
sensitivity and linear range of the measurement. Figure 80 shows
560 LP
collection of Cy3 fluorescence using either a 580 BP 30 or a 560 LP
emission filter.
Emission

580 BP 30
10.9.2 Multicolor imaging
Multicolor imaging allows detection and resolution of multiple targets
500 550 600 650 700
using fluorescent labels with different spectral properties. The ability to
Wavelength (nm)
multiplex or detect multiple labels in the same experiment is both time
and cost-effective and improves accuracy for some assays. Analysis using
Fig 80. Filtering of Cy3 fluorescence using
a single label can require a set of experiments or many repetitions of the
either a 580 BP 30 (dark gray area) or a
560 LP filter (light and dark gray areas). same experiment to generate one set of data. For example, single-label
analysis of gene expression from two different tissues requires two
separate hybridization to different gene arrays, or consecutive
hybridization to the same array with stripping and re-probing. With a
dual-label approach, however, the DNA probes from the two tissue types
are labelled with different fluorochromes and used simultaneously with
the same gene array. In this way, experimental error is reduced because
only one array is used, and hybridization conditions for the two probes
are identical. Additionally, by using a two-channel scan, expression data
is rapidly collected from both tissues, thus streamlining analysis.

● 123
 M I C R O A R R AY

The process for multicolor image acquisition varies depending on the

imaging system. An imager with a single detector acquires consecutive
images using different emission filters and, in some cases, different
excitation light. When two detectors are available, the combined or
mixed fluorescence from two different labels is collected at the same
time and then resolved by filtering before the signal reaches the detectors.
Implementation of dual detection requires a beam splitter filter to
spectrally split the mixed fluorescent signal, directing the resulting two
emission beams to separate emission filters (optimal for each
fluorochrome), and finally to the detectors. A beam splitter, or dichroic
reflector, is specified to function as either a short-pass or long-pass filter
relative to the desired transition wavelength. For example, a beam splitter
that reflects light shorter than the transition wavelength and passes
longer wavelengths is effectively acting as a long-pass filter (Fig 75).

10.9.3 Fluorochrome selection in multicolor experiments

When designing multicolor experiments, two key elements must be
considered: the fluorochromes used and the emission filters available.
As with any fluorescence experiment, the excitation wavelength of the
scanner must fall within the absorption spectrum of the fluorochromes
used. Additionally, the emission spectra of different fluorochromes
selected for an experiment should be relatively well resolved from each
other. However, some spectral overlap between emission profiles is
almost unavoidable. To minimize cross-contamination, fluorochromes
with well-separated emission peaks should be chosen along with emission
filters that allow reasonable spectral discrimination between the
fluorochrome emission profiles. Figure 86 shows the emission overlap
between two common fluorochromes and the use of band-pass filters to
discriminate the spectra. For best results, fluorochromes with emission
peaks at least 30 nm apart should be chosen. A fluorescence scanner is
most useful for multicolor experiments when it provides selectable
emission filters suitable for a variety of labels. A range of narrow band-
pass filters that match the peak emission wavelengths of commonly used
fluorochrome labels will address most multicolor imaging needs.

hb 63-0048-49 ● 124
 C H A P T E R 1 0 : F L U O R E S C E N C E I M A G I N G S Y S T E M S I N M I C R O A R R AY A N A LY S I S

10.10 Amersham Biosciences imaging systems

Amersham Biosciences offers a variety of imaging instruments that are
well suited for use in microarray analysis. For more information, please
consult Amersham Biosciences web site at www.amershambiosciences.com.

10.10.1 Typhoon 9210: High performance laser scanning

system
Excitation sources: 532-nm Nd:YAG and 633-nm HeNe lasers

Filters: 6 emission filters and 3 beamsplitters (up to 13 emission

Fig 81. Typhoon Variable Mode Imager. filter positions)

Detection: 2 high sensitivity PMTs

Imaging modes: 4 modes. 2 modes for fluorescence detection,
chemiluminescence, storage phosphor

Scanning area: 35 × 43 cm

Maximum resolution: 10 µm

Sample types: microarrays, gel sandwiches, agarose and polyacrylamide

gels, blots, microplates, TLC plates, and macroarrays

10.10.2 Typhoon 9410: High performance laser scanning

system
Excitation sources: 532-nm Nd:YAG, 633-nm HeNe,
and 457-nm and 488-nm Argon lasers

Filters: 7 emission filters and 3 beamsplitters (up to 13

emission filter positions)

Detection: 2 high sensitivity PMTs

Imaging modes: 5 modes. 3 modes for fluorescence detection,

chemiluminescence, storage phosphor, chemifluorescence

Scanning area: 35 × 43 cm

Maximum resolution: 10 µm

Sample types: microarrays, gel sandwiches, agarose and

polyacrylamide gels, blots, microplates, TLC plates, and
macroarrays

Notes: Versatile fluorescence and radioactive imager that can scan

microarrays but also contains an extra blue laser

● 125
 M I C R O A R R AY

hb 63-0048-49 ● 126
C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

Chapter 11
D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F
M I C R O A R R AY E X P E R I M E N T S

11.0 Introduction
Methods for the analysis of microarray data are still evolving, and
there is no standard experimental design or method of data analysis for
microarray experiments at the present time. However, some efforts are
being made to set a common annotation and standards for microarray
data in order to create public databases for microarray results (61, 62,
63). Meanwhile, in this chapter some important considerations for
analyzing microarray data are discussed.

11.1 Experimental design

Data analysis begins with experimental design. When planning a
microarray experiment it is important to consider sources of variation
within the experiment. These can arise from the samples reflecting
differences in gene expression between individual animals or different
tissue culture plates. Furthermore, time-dependent variation in gene
expression levels resulting from circadian rhythms can also be a factor.
Experimental variations may also occur due to variation within the
experiment itself. In order to ensure that these experimental errors
can be identified, slides should contain replicate spots of each target
and replicate slides should be analyzed with pooled or multiple mRNA
samples. This replication enables the use of statistical tools such as
averages and standard deviations to monitor the extent of experimental
variation (64).

Lucidea Universal ScoreCard has been developed by Amersham

Biosciences to address the need for controls. It is a set of controls used
to validate and normalize microarray experimental data. It is further
described in section 6 of this chapter.

Another prudent measure is to perform reverse color or ‘flip-flop’

experiments. In these experiments the two mRNA samples being
compared in a microarray experiment are labelled separately with both
Cy3 and Cy5. Replica slides are hybridized with both combinations of
probes. By comparing the signal ratios from the reversed slides, it is
possible to identify data that is affected more by the labelling process
or quality of mRNA than by changes in gene expression levels.

● 127
 M I C R O A R R AY

Once the mRNA extraction, labelling, hybridization, and scanning are

complete, the final stage in the microarray experiment is data analysis.
This is a complex multi-step process and is illustrated in Figure 82. The
steps of data analysis are described in further detail in this chapter.

Glass slide with sample

Replicate slides

Image analysis

Discard spots with

poor quality values

Calculate ratios and

examine controls

Discard slides or areas

of slides where controls
indicate a problem

Average data and

examine variation of
ratios between replicates

Visualize data

Cluster data

9 4 7
3 Store data in database
6 1
2 2

Fig 82. Stages of microarray data analysis.

hb 63-0048-49 ● 128
C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

11.2 Overview of microarray data analysis

Microarray data analysis consists of four main steps:

■ Image analysis

■ Examination of controls

■ Data normalization

■ Visualization and clustering

Image analysis uses a dedicated software to quantitate the fluorescent

intensity at each spot. Normally, this involves a process called
spotfinding. The second step is to examine the controls on the arrays.
Normalization is performed next, followed by calculation of mean
ratios. The data can then be visualized in a graphical form, and
clustered, such that meaningful trends can be found among data
from multiple slides and experiments.

The amount of data obtained from microarray experiments is vast and

can be generated very rapidly. However, it is important to know the
quality of the data. There are three types of quality values that can be
used. It is recommended that all three are used within a microarray
experiment. The types of values are:

■ A series of metrics reported by the image analysis software to ensure

that the spots that have been quantitated appear to be good spots,
for example, regularly shaped.

■ A series of controls on your microarray to ensure the hybridization

has occurred correctly. These will indicate how specific and efficient
the hybridization has been.

■ Analysis of the data from replicate targets. Replicates are critical

for indicating how good the overall data is and whether the results
obtained are statistically meaningful (64).

● 129
 M I C R O A R R AY

11.3 Image analysis

A scanned microarray image records the fluorescent intensities of all
pixels in the image area, including pixels from within (signal) and outside
(background) the DNA spots. The first step in the microarray workflow
is to locate the spots. Consider the following objectives:

■ accurately define the positions of every DNA spot in the image

■ provide appropriate measurement of fluorescence intensity for each

spot by quantifying the intensities of pixels within and outside the
DNA spots

■ provide quality metrics that give estimation of the accuracy of the

intensity measurement
The first step, alternately called “gridding”, can usually be performed
using dedicated software, such as ArrayVision or Lucidea Automated
Spotfinder.
The process of spotfinding begins by defining a grid, or an array of
circles, that indicates the expected size of each spot, how far away they
are spaced, and how they are arranged in an array, all regardless of the
intensities of individual spots. This information can be measured in pixel
or micron units. Once a grid is defined, it is overlaid onto the scanned
image such that the circles are nearly exactly aligned with the spots on
the microarray image. This spotfinding process can be automated using
spotfinding software, which serves to eliminate the tedious task of
manual alignment. In addition to the following descriptions, see the
spotfinding software help guide for instructions how to most effectively
use this tool.

■ Manual: This method involves first dividing the grid into several
subgrids, and then visually aligning each smaller subgrid with the
corresponding area of the image by adjusting the position of circles.

■ Semi-automated: This is where the software algorithm finds the spots,

but some user intervention is required.

■ Automated: This is where minimal user intervention is required. These

software packages can automatically analyze multiple images while
eliminating the need for supervision.

hb 63-0048-49 ● 130
 C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

The next step in image analysis is to determine the signal present inside
and outside the spot. The background signal is then subtracted from the
spot signal to give background-corrected signal. Whereas spot signal is
calculated from within the positioned grid, the background signal is
determined by calculating the average pixel intensity in a user-defined
region. Although mean or median background signal can be used,
median values are more resistant to variation in background caused, for
example, by fluorescent speckles. Some of the various regions in which
the background can be calculated are illustrated in Figure 83. It is
important to use a background correction method such that any pixels
from the spot do not get included in the background. This can easily
occur if the spots are close together.

Around spot groups Corner between spots User defined areas Around individual spots User defined spots

Fig 83. Some of the background region options available to the user
in the ArrayVision™ Image Analysis Software. The green represents the
spots enclosed by the grid while the blue encloses the background
region. Image analysis quality metrics calculated by analysis software
are increasingly used to highlight data that may be unreliable and
should be omitted from further analysis. Typical causes of suspect
data include dust speckles over spots, poor spot morphology, very low
or very high signal.

● 131
 M I C R O A R R AY

11.4 Spotfinding software offered by Amersham

Biosciences

11.4.1 Lucidea Automated Spotfinder

Lucidea Automated Spotfinder processes microarray images by performing
spotfinding and data extraction in an automated fashion with virtually
no manual intervention (Fig 84). The output from Lucidea Automated
Spotfinder includes the signal intensity for each spot, plus quality metrics
to assess individual spots as well as the overall image. The software is
compatible with images produced by commercially available scanners.
Several images can be analyzed at once in a batch mode, without manual
Fig 84. Lucidea Automated Spotfinder has inspection or image manipulation. Lucidea Automated Spotfinder
a simple, intuitive user interface to initiate features include:
the automatic processing of microarray
■ fully automated spot finding and data extraction
images. In this example, four images are
loaded for analysis. ■ multiple reporting options and data export

■ user-defined templates for analyzing single or multiple images

■ metrics for assessing data quality

■ background subtraction

11.4.2 ArrayVision
ArrayVision software is a semi-automated software used for performing
image analysis (Fig 85). Some of its features include:

■ automated alignment of quantification grid over array

■ choice of methods for background signal removal

■ quality metrics
■ reporting tools and data export

■ visualization tools for viewing array images

Fig 85. ArrayVision software provides
automated analysis of radioisotopic or
fluorescent macro- and microarrays.

hb 63-0048-49 ● 132
C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

11.5 Use of controls in microarray experiments

As with all experiments, microarrays should contain a series of controls
to ensure that the data obtained from the arrays is accurate. Therefore,
included on microarrays should be some cDNA or oligos which are
expected to give a negative result, and some which should give a positive
result. The types of controls that should be included are discussed below.

11.5.1 Negative controls

Negative controls are spotted DNA sequences that should not hybridize
with any labelled probe. The negative controls used should ideally come
from organisms that are only distantly related to the organism being
studied in the experiment. For example with human microarrays, sequences
from bacterial genes, intergenic regions, plant genes, or double-stranded
poly-dA are often used for this purpose (25, 65, 66). Under optimal
analysis conditions negative control spots should not give any signal at
all. However, if the stringency of the hybridization is not high enough,
non-specific hybridization between labelled cDNA molecules in the probe
and unrelated target sequences on the array may take place, resulting in
detectable signals from negative control spots. The higher the signal from
negative controls, the less reliable is the data from the whole slide. These
negative controls are particularly important to include if the spotted
array consists of oligonucleotides because in these types of arrays, a
lower hybridization stringency may be used. Negative controls can also
be used to detect contamination between targets during spotting. Placing
negative control targets after positive control targets that are always
expected to give signal can do this.

11.5.2 Poly-adenylated DNA and CotI DNA

When using oligo(dT) to prime first-strand cDNA synthesis, it is possible
that the oligo(dT) will prime within the poly-A tail of the mRNA. If this
occurs there will be a string of dT bases within the probe. It is possible
that the targets spotted may also contain a similar string of poly-A
sequences, particularly if the targets were derived originally from an EST
library which had been made by the use of oligo(dT) primer. In order to
prevent cross reactivity of the poly-dT sequences within the probe with
potential poly-dA sequences in the targets, a poly-dA oligo of 80 bases
can be included in the hybridization to block the poly-dT (65). In order
to ensure that this process has occurred correctly, it is good to include as
a negative control, a poly-dA sequence spotted on the microarray. If
these spots are negative this suggests that the blocking has occurred
effectively.

● 133
 M I C R O A R R AY

Another sequence that may cause problems within the probe is

derived from repetitive sequences such as Alu-repeat sequences. These
sequences can be blocked by the inclusion of Cot-1 DNA in the
hybridization. To ensure that this blocking has occurred correctly, the
spotting of Cot-1 DNA as a negative control is recommended.

11.5.3 Positive controls

Labelled DNA
DNA can be labelled with CyDye fluors using polymerase chain reaction
(PCR), or in the case of oligos, during the synthesis of the oligos (many
oligo manufacturers offer this service). When the labelled DNA is spotted
onto the array, the DNA will be fluorescent and serve as a useful positive
control for verifying that the target DNA is binding effectively to the
slide surface during the hybridization and washes. Total genomic DNA
can also be used as a positive target. Positive controls placed on different
locations of the slide can help in the spotfinding process by providing
clearly detectable signals in known positions, regardless of the type of
probe used.

1 2 3 4 1

Fig 86. To validate, normalize, and filter microarray data, four

different types of controls are supplied:
1. Calibration controls: the signal intensities of ten individual controls
span 4.5 orders of magnitude for both Cy3 and Cy5 channels. These
controls can be used to generate a calibration curve.
2. Ratio controls: eight ratio controls are provided at both low and high
expression levels and are used to evaluate precision of ratios.
3. Negative controls: two controls are used to estimate non-specific
hybridization and potential carryover with the microarray system.
4. Utility controls: three individual controls can be used to troubleshoot
and examine sample preparations, or they can be used as additional
ratio or calibration controls.

hb 63-0048-49 ● 134
C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

Spikes for determining sensitivity and dynamic range

Using some of the negative control genes discussed above can be used
to make a different set of positive controls. RNA can be synthesized by
in vitro transcription from plasmids carrying these negative control
sequences. This synthetic mRNA can then be included in the Cy3 and
Cy5 labelling reactions at known concentrations. This process is known
as spiking, and the synthetic mRNA are the spikes. Several different
mRNA spikes can be used and spiked in at different concentrations,
resulting in a set of controls that can give the researcher a value for the
linear dynamic range and sensitivity of the assay. These controls are known
as dynamic range controls. In addition, different spikes can be spiked
into the Cy3 and Cy5 labelling reactions at different concentrations.
These types of controls are known as ratio controls. After hybridization
of the probe to the slide, washing, and scanning, the Cy3 and Cy5 signals
obtained from the ratio controls can be compared with the known
amount of mRNA spiked into the labelling reactions and the theoretical
known ratios. Therefore a series of spikes can determine how sensitive
the hybridization has been and how accurate the data obtained is.

11.5.4 Housekeeping gene controls

Some genes are expressed relatively consistently within many different
cell types. These are called housekeeping genes. Examples of such genes
are actin, glyceraldehyde 3-phosphate dehydrogenase (GAPDH), and
tubulin. These housekeeping genes can be included in microarray
experiments as controls to ensure that the hybridization has occurred,
and they can also act as a normalization factor. However, the expression
of these genes can vary under experimental conditions, and relying on the
use of one or few housekeeping genes can result in skewed data.

11.5.5 Controls for measuring pen-to-pen variation

Housekeeping gene, spikes, or positive controls can be spotted in replicate
across the slide. If a different pen on the microarray spotter spots each of
these replicates, this may provide some information on the pen-to-pen
variation of the spotter. It should be considered that the variation of
each of these controls will be dependent not only on pen-to-pen variation
but also on the variation in the slide surface and any variation in the
hybridization. Therefore, if a high pen-to-pen variation is seen, a pen
test on the spotter should be performed.

● 135
 M I C R O A R R AY

11.6 Control products offered by Amersham

Biosciences

11.6.1 Lucidea Universal Scorecard

Lucidea Universal ScoreCard contains a set of 23 artificial genes that
serve as analytical controls to validate and normalize microarray data.
The controls are composed of DNA sequences from yeast intergenic
regions, and their performance has been shown to be independent of a
wide variety of species. This system can be used as a universal reference
for validating and normalizing microarray data as well as for creating a
calibration curve for determining limit of detection, linear range, and
saturation of microarray experiment (Fig 86).

Log Number of spots

(Cy3/Cy5) per bin
Normalized
1.5 800 log ratio

1 700 2

1.5
0.5 600

1
0 500
0.5
-0.5 400
0
-1 300
-0.5
-1.5 200
-1

-2 100 -1.5

-2.5 0 -2
a) 2 3 4 5 6 7 8 b) 2 3 4 5 6 7 8

Log (Cy5 sDxA) Log (Cy5 sDxA)

Fig 87. Applying non-linear normalization to microarray data.

Panel A shows distribution of log(Cy3/Cy5) values plotted against
Cy5 signal values (blue crosses) from a typical microarray experiment.
The orange line denotes the average relationship between these log
ratios as a function of Cy5 signals. As can be seen from the shape of
the curve, this relationship is non-linear over the distribution of Cy5
signals. The blue line shows distribution of Cy5 signal intensities.
Panel B shows the same data after it has been normalized using
the non-linear normalization algorithm.

hb 63-0048-49 ● 136
C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

11.7 Normalization
In order to compare ratio data from one microarray slide to another
microarray slide, the ratio data needs to be normalized to correct for
experimental variation. The reason for this is that from one slide to
another there will be differences between the relative Cy3 and Cy5
signals due to one or more of the following:

■ the amounts of mRNA used in the Cy3 and Cy5 labelling reaction

■ efficiency of detection of the Cy3 and Cy5 by the detection system

within the scanner

■ relative incorporation differences of the Cy3 and Cy5 reverse

transcriptases

11.7.1 Linear normalization

Linear normalization assumes that there is a single normalization factor
required over the whole signal range. There are three methods to obtain
this factor:

■ use signal ratios from housekeeping genes

■ use signal ratios from spikes which have been added to the labelling
reaction in equal amounts

■ use total signal, which is the summation of all the Cy3 signals and all
the Cy5 signals

In the housekeeping gene or spike gene methods, it is assumed that the

spots corresponding to these targets give a ratio of 1. The total signal
method assumes that addition of all the signals in the Cy3 and the Cy5
channel should give a Cy3/Cy5 ratio of 1. The normalization factor can
be calculated from the observed ratios for the housekeeping genes,
spikes, or total signals to give a conversion factor that results in the
expected ratio for the control spots.
For example, a housekeeping gene in experiment A gives a ratio of 2,
while the gene of interest has a Cy3/Cy5 ratio of 5. Therefore the
normalized ratio for the gene of interest is 5/2 = 2.5. In experiment C,
the housekeeping gene has a ratio of 3, while the gene of interest has a
Cy3/Cy5 ratio of 7.5. The normalized ratio is 7.5/3 = 2.5.

The three methods discussed above assume that the normalization factor
is constant over the whole signal range, which in most cases is not.

● 137
 M I C R O A R R AY

11.7.2 Non-linear normalization

It has been found that linear normalization is not necessarily accurate
(67). Examining the data shown in Figure 87a can show this. In this
experiment, Cy3-labelled muscle and Cy5-labelled muscle probes
(identical mRNA labelled with two different fluors) were hybridized to
a single slide. The results of this experiment are plotted below as a plot
of log Cy3/Cy5 ratio against the log Cy5 signal (Fig 87b). As the same
mRNA was used, it would be expected that the log Cy3/Cy5 ratio should
be constant over all Cy5 signals. However, as can be seen from the graph,
the ratio is higher at low Cy5 signal levels compared to the figure at high
Cy5. Therefore, the normalization factor used for spots with a low Cy5
signal should be different from the normalization factor that is used at
high Cy5 signals.

There are two principal non-linear methods of calculating the normalization

factor. One method is to rank all the data points according to their
Cy3+Cy5 signal. Then for each 50 genes calculate the normalization
factor for those 50, in the same way as total normalization is carried out.
A more precise way is to fit a curve to the data so that the normalization
factor for each point can be calculated. Software packages are commercially
available that can perform this kind of normalization. A non-linear
normalization generally results in a more accurate normalization than
linear normalization.

11.7.3 Post normalization

Once the normalization procedure has been carried out for all the data
points, the behavior of controls is examined next. The negative controls
should have a signal-to-noise ratio of less than 3 [(SNR = [average signal
– average background]/standard deviation of background). Any higher
SNR than 3 suggests that the data obtained from this experiment may
not be accurate.
If a series of dynamic range controls have been included, this is one
way to estimate sensitivity. A control which has a SNR above 3 would be
regarded as having been detected. If spikes have been included, such that
the spikes have been placed in the Cy3 reaction at a different amount
compared to the Cy5 reaction, then the theoretical ratio can
be compared to the actual ratio. Finally, controls such as pen-to-pen
variation controls may suggest other potential problems within the
experiment.

If the data from the slide meets the criteria set by the researcher then the
next step is to look at the variation between replicates. It is recommended
that each experiment be repeated several times, and there are statistical

hb 63-0048-49 ● 138
C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

criteria as to how many replicates should be carried out to give a

certain degree of confidence in your results (64). Replicates could be
within multiple spots on the same slide, but can also be carried out using
several different slides. Typically log ratios are calculated so that ratios
less than 1 appear as a negative value. The coefficient of variation
(CV = [Standard deviation of the ratio] / [mean ratio]) can also be
calculated and provides a simple measure of the value of a particular
data point. Data points with high CVs can be highlighted or discarded.
Often ratios with high CVs are due to low signal in one or both of the
channels. A standard practice by some researchers is therefore to discard
spots which have a signal in both channels with a SNR below 3. If there
is a low signal (below a SNR of 3) in one channel, then the ratio could be
considered to be an arbitrary fixed value to avoid very large ratios
(to prevent, for example if Cy5 signal is zero, the Cy3/Cy5 would be
infinite). In addition, it should be remembered that there may be
significant biological variation that must be taken into account when
designing experiments.

11.8 Visualization and clustering

After microarray data is normalized to account for differences in Cy3
and Cy5 signals, it can subsequently be exported to any number of data
visualization software for further analysis. These software products can
be used to mine the data for significant changes in gene expression. The
process of visualization can significantly enhance data analysis. It can
provide helpful features, such as data integration, customized query
devices, and pattern recognition. Clustering data points, or genes, that
show similar responses on microarray analysis can be used to identify
genes that have similar gene expression patterns and which possibly
belong to the same pathway.

● 139
 M I C R O A R R AY

11.9 Visualization software products offered by

Amersham Biosciences

11.9.1 Spotfire DecisionSite for Functional Genomics

This software combines the core capabilities of Spotfire™ DecisionSite™
with specialized tools for interrogating and extracting information from
microarray data. In addition to simplified access to data and information,
Spotfire DecisionSite for Functional Genomics provides researchers with
leading analytical methods used in gene expression analysis. Dynamic
visualizations and interaction with computational results help researchers
Fig 88. Exploring gene expression in validating and prioritizing target genes (Fig 88).
data using hierarchical clustering, and
Some of the features of this software include:
principal component analysis with
Spotfire DecisionSite for Functional ■ easy access to information, in any format, wherever it resides
Genomics.
■ visually interactive representations of enriched data sets for
enhanced analysis

■ publication and sharing of results for collaborative decision-making

■ idenification of key patterns with distinction calculation, hierarchical,

bi-directional hierarchical, and K-means cluster analysis

■ preparation of data for analysis with standard array and gene

based normalization

■ identification of entities exhibiting characteristic or signature

profiles with ad hoc profile search and analysis

hb 63-0048-49 ● 140
C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

11.10 Scierra Microarray Laboratory Workflow

Systems: Information Management Systems for
Microarray Laboratories
Microarray technology is rapidly becoming the mainstream platform for
high-throughput gene expression analysis. As microarray experiments
generate vast amounts of experimental and biological data, an urgent
need is created for informatics tools that can manage the microarray
workflow process more effectively and efficiently. Scierra™ Microarray
Laboratory Workflow System, as a part of a larger integrated laboratory
information managment system, is designed to address this need.

11.10.1 Scierra Laboratory Workflow Systems

Scierra Laboratory Workflow Systems (LWS) are a series of bioinformatics
solutions to aid in the collection, annotation, collation, curation, and
analysis of biological data. The platform includes four products:

■ Scierra Sequencing LWS System

■ Scierra Genotyping LWS System

■ Scierra Microarray LWS System

■ Scierra Proteomics LWS System

These Scierra LWS products are built on a common software framework

that manages and tracks all aspects of an experiment. Each system
mirrors the natural workflow found within the laboratory, and succeeds
in linking manual processes, instrumentation, software, and reagent use
into one system. This integration enables the collection and comparison
of each type of biological information. The open framework design
allows the introduction of new instruments and reagents, thereby
providing a scalable system that will grow as needs increase.

11.10.2 Scierra LWS architecture

Scierra LWS is a three-tiered application comprised of an Oracle™
database, a middleware application server, and multiple clients. Scierra
LWS can accept data from most available network client sources,
including computers running Windows™ 2000 or Windows NT™. Work
is easily requested and results readily accessed through a standard
browser-based user interface that communicates to the middleware
application server.

● 141
 M I C R O A R R AY

11.10.3 Scierra Microarray Laboratory Workflow System

(MA-LWS)
A typical microarray workflow involves array content preparation,
array production, sample preparation and labelling, array hybridization,
scanning, and image analysis.

Scierra MA-LWS has been designed to mirror the workflow of the typical
microarray lab (Fig 89). It allows users to perform many tasks. First,
users can organize large numbers of samples and experiments based on
projects. Users can also manage and track a large variety of samples and
reagents, including:

■ crude biological samples, including blood, tissue, and cells

■ total RNA, mRNA, and labelled cDNA

■ spotting plates and spotting sample types, including cDNA and oligo

■ spotting substrates, chemistry, and protocol

■ arrayed slides

Scierra MA-LWS allows users to manage and track every activity in the
microarray workflow, including:

■ array preparation—spotting custom arrays and pre-arrayed slides

■ sample preparation and labelling

■ hybridization, scanning, and image analysis

Scierra MA-LWS makes it possible to integrate components of different

microarray systems, including spotters and scanners, and image analysis
software. Users can store and effectively retrieve large amounts of
information. Flexible reporting tools provide for standard and user-defined
queries across different activities. With Scierra MA-LWS, the precious
microarray data is completely captured and securely stored in the
database for further analysis.

hb 63-0048-49 ● 142
 C H A P T E R 1 1 : D E S I G N , C O N T R O L S A N D D ATA A N A LY S I S O F M I C R O A R R AY E X P E R I M E N T S

Sample
Definition

Experimental
Design
Biological Samples Array
Sample
Submission
Sample Spotting plate
Receiving Receiving

Total RNA Compression/

Extraction Decompression

mRNA
Preparation Quantification
Fig 89. Scierra MA-LWS mirrors
Sample Normalize and
the microarray workflow. Labeling Aliquot

Sample In House Array

Clean Up Design

Sample
Quantification Spotting

Sample Treatment
Fragmentation Post-spotting

Hybridization Pre-arrayed
slides

Array
Scanning

Spotfinding

Normalization

Result Reports Result

Retrieval Query

● 143
 M I C R O A R R AY

hb 63-0048-49 ● 144
 C H A P T E R 1 2 : T R O U B L E S H O O T I N G M I C R O A R R AY E X P E R I M E N T S

Chapter 12
T R O U B L E S H O O T I N G M I C R O A R R AY E X P E R I M E N T S

12.0 Introduction
Microarray analysis is a complicated multistep process consisting of
discrete steps as shown in Figure 90. Each of these steps is critical in
determining whether the experiment is successful, and problems
encountered at any stage will be detrimental to the quality of data
obtained. Troubleshooting microarray experiments needs to consider
all these steps.

The microarray process

Glass slide

Array spotter Arrayed slide

with DNA

Slide processor Hybridized slide with

CyDye-labelled cDNA or RNA probe

Array scanner
Scanned slide

Analysis software Analysis of

scanned data

9 4 7
Biological data
3 1
generated by microarrays 6
2 2

Fig 90. Flowchart of microarray experiment.

● 145
 M I C R O A R R AY

The output from a microarray experiment is the intensity of

hybridization signals, which reflect the expression levels of the
corresponding genes in the analyzed samples. However, other
experimental factors also have significant influence on the
magnitude of these signals. Some of these factors are listed
in Figure 91.

Amount of target in Length of labelled

hybridization reaction target molecules

Labelling
density

Number of target
molecules in sample

Hybridization Hybridization Detection

conditions: efficiency: sample set up:
• buffer • diffusion • PMT
• time • kinetics Slide • Cy3 vs Cy5
• stringency • Tm • lasers

Fig 91. Factors influencing the intensity of

observed microarray signals.
12.1 Optimization of the microarray system
As microarray analysis is not trivial to perform, careful optimization of the
microarray system is recommended. Due to the wide choice of reagents,
consumables and instruments from various manufacturers, it is important
to optimize the selected combination of materials and protocols before
starting real experiments. Some reagents may not work well with other
reagents; for example, spotting buffer may not be compatible with all
different slide surfaces, and different hybridization buffers may give very
different results on identical slides.

System optimization should test all microarray components together

using target sequences and probes that are as close to real samples as
possible. Such experiments should incorporate controls, as discussed in
Chapter 11. Combining the use of realistic sets of targets and control
reagents, as illustrated in Figure 92, gives the best results. The aim of
Fig 92. Panel A shows a scanned microarray
image from a yellow experiment in which
skeletal muscle mRNA was labelled with
Cy3 and Cy5. Only a proportion of the
microarray slide is shown. Panel B shows a
scatterplot derived from a yellow experiment.
a) Scatter plot
The microarray used in this experiment Cy3 volume
contained ratio control targets in addition to 1000000

cDNA clones and mRNA corresponding to ScoreCard

these control sequences was spiked into the ratio controls
100000
mRNA before labelling. These ratio control
spots appear as green and red dots on the
array image and are also shown on the Non-specific
10000
Background
scatterplot, where they fall above and below random signals corrected and
normalized RFUs
the scatter line. 1000

1000 10000 100000 1000000

b) Cy5 volume

hb 63-0048-49 ● 146
 C H A P T E R 1 2 : T R O U B L E S H O O T I N G M I C R O A R R AY E X P E R I M E N T S

system optimization is to find the best overall protocol for the system
and to determine what is the standard performance of the system.

12.1.1 The yellow experiment

The yellow experiment is an efficient tool for microarray system
optimization. In this experiment, the same RNA sample is labelled with
both fluorescent dyes, typically with Cy3 and Cy5 fluors. Hybridization
of equal amounts of both probes onto a microarray should produce equal
hybridization signals from both colors. In a false color array image, all
target spots should appear as yellow dots. As computer screen images of
microarray data can be misleading, analyzing numerical data from a
yellow experiment as a scatter plot is more informative. As no differential
gene expression is expected, normalized Cy3 signals plotted against
normalized Cy5 signals should appear as a straight line. Figure 92 shows
an example of microarray data generated from a typical yellow experiment.
Because RNA isolated from different cell cultures or individuals is likely
to contain slightly different levels of some transcripts, it is recommended
that pooled RNA obtained from several RNA isolations is used for system
optimization. Alternatively, purified RNA is also available commercially.

Yellow experiments only require one type of RNA sample, but provide
information on all aspects of the microarray process. In contrast,
experiments in which two different RNA species are used have the added
complication that differential gene expression will be present in unknown
quantities. In reverse-color experiments, in which both samples are
labelled separately with two colors and hybridizations are performed
with both possible combinations, any inherent variation between the
quality of the two arrays can complicate the optimization process and
result in wrong conclusions. For example, high and uneven Cy3
background on one of the slides in a reverse-color experiment can give
the appearance of unbalanced labelling with one fluorescent dye. For
‘real’ experiments in which information about gene expression is being
sought, reverse color experiments are useful.

● 147
 M I C R O A R R AY

12.2 Experimental design and execution

12.2.1 Experimental variation

Experimental variation must be taken into account when designing and
carrying out microarray experiments (Fig 93). No two microarray
experiments, even if replicas of each other, will give exactly the same
results. Each step of the process contributes to this variation, which may
mask the presence of differential gene expression, leading to false negative
results. If the amount of variation is not known, false positive results can
also be obtained if randomly varying results are taken at face value.

Fig 93. Experimental variation. Four

identical microarray slides were hybridized 200

simultaneously with equal aliquots of the

150
same probe. Scatterplots of normalized
gene expression data are shown. Variation 100
in background fluorescence arising from
uneven slide surfaces was the major 0
Slide A Slide B Slide C Slide D
contributing factor to experimental
variation in this case.

12.2.2 Replication in microarray analysis

Replication is the key to identifying and quantifying variation in
microarray experiments. It has been found that performing three replica
microarray hybridizations with different slides reduced the misclassification
of gene expression compared with performing single hybridizations (64).
Data calculated and pooled from all these replicas enables statistical
determination of experimental variation in terms of standard deviation
and coefficient of variance (CV).

Microarray experiments should contain the following:

■ Each target should be present in at least two, preferably more,

copies on the microarray.

■ Multiple slides should be hybridized with each probe pair.

■ Multiple RNA samples should be obtained for each experimental

condition.

hb 63-0048-49 ● 148
 C H A P T E R 1 2 : T R O U B L E S H O O T I N G M I C R O A R R AY E X P E R I M E N T S

12.2.3 Step controls

Separately monitoring the various steps in microarray analysis, while
lengthening the protocols, provides quality control information. Step
controls also offer the following benefits:

■ Researchers’ resources are used most efficiently, with the

least amount of waste.

■ Problems can be identified on a step-by-step basis, and

most likely causes obtained.

■ Information from intermediate steps allows conclusions

about the overall validity of microarray results to be drawn.

Ideal step controls provide numerical or visual information that

unequivocally characterizes the success of that step. Recommended
control procedures are listed in Table 6. The control measures indicated
in bold should be included in every microarray experiment. Additional
controls are recommended to be performed when new protocols or
reagents are being tested. It is advisable to prepare some extra reagents if
control procedures are performed. Greatest benefits from step controls
are derived when the results are evaluated before continuing with the
experiment.

12.3 Performing microarray analysis

There are several critical points to performing a successful microarray
hybridization experiment. Please take note of the following:

■ Follow all protocols precisely.

■ Be consistent across experiments when several separate

hybridizations are involved.

■ Maintain precise technique when performing the microarray

experiment and analyzing the results.

■ Always use appropriate reagents for each protocol; alterations

can be a source of error in final data analysis.

■ Keep record of individual reagents used in each experiment

as it can be useful in identifying causes of problems.

Table 7 lists some typical problems in microarray analysis and their

likely causes. Most of these problems can be identified and avoided by
following the recommended quality control measures listed in Table 7,
which can also aid in identifying the cause of the problem. Often there
can be several contributing causes to any problem.

● 149
 M I C R O A R R AY

Table 6. Quality control measures for microarray analysis.

Step Control mearsures
Preparation of cDNA targets by PCR ■ Verify presence of only one band in agarose electrophoresis.
■ Determine quantity of target DNA.
■ Sequence target to verify identity.
Microarray slide coating ■ Scan to detect presence of background fluorescence or
dirt particles.
■ Perform additional surface tests.
Microarray printing ■ Use fluorescent DNA as control for printing quality.
■ Use DNA-binding dyes to detect printing DNA on slide.
■ Perform a test yellow experiment with well-characterized
RNA preparation for each printing batch and especially
when new targets are introduced.
■ Incude plenty of control targets on slide.
■ Note down temperature and humidity of printing chamber.
■ Keep track of how many times targets have been used.
■ Keep track of when slides were printed.
RNA isolation ■ Check purity and integrity of RNA by gel electrophoresis
or RT-PCR.
■ Determine quantity of RNA.
Sample labelling and purification ■ Spike in synthetic control mRNA.
■ Perform control labelling reaction with control RNA.
■ Determine incorporation of CyDye into labelled
sample by spectrophotometry.
■ Determine whether purified sample contains free
CyDye by gel electrophoresis.
■ Determine size of labelled nucleic acid fragments
with gel electrophoresis.
■ Determine the amount of labelled nucleic acid.
■ Determine the amount of CyDye per microgram of
labelled nucleic acid (labelling density).
■ Determine the recovery of labelled cDNA from purification
step by radioactive spiking or gel analysis.
Hybridization and stringency washes ■ Use equal and optimal amounts of Cy3- and Cy5-labelled
probes in the hybridization.
■ Include positive and negative hybridization controls.
■ Perform hybridization without probe.
Scanning ■ Test scanner performance in order to verify
correct functioning.
■ Perform scans at different settings to ensure
optimal data collection.
■ Visually inspect scanned images to detect any
obvious blemishes or areas of poor data.
Data analysis ■ Background correct and normalize data before
drawing conclusions.
■ Examine how well normalization worked.
■ Determine the amount of variation in experiment.
■ Never trust data from one slide.
Verification of microarray results ■ Use independent analytical techniques to verify whether
the results obtained from microarray analysis are
reproducible and biologically significant.

hb 63-0048-49 ● 150
 C H A P T E R 1 2 : T R O U B L E S H O O T I N G M I C R O A R R AY E X P E R I M E N T S

Table 7. Troubleshooting microarray experiments,

Symptom Possible cause Remedy
No hybridization signal. ■ Target concentration too low. ■ Determine target concentration before
slide spotting.
■ Targets not clean enough. ■ Remove PCR components from targets
before slide spotting.
■ Poor retention of targets on slide. ■ Prepare new microarray slides. Check
that spotting buffer and protocol are
compatible with slide type.
■ No transcripts in RNA sample. ■ Obtain new RNA/mRNA sample and
test it before labelling.
■ Failed labelling reaction. ■ Always check the success of labelling
reaction before using it in hybridization.
■ Faulty component in labelling reaction. ■ Test components of labelling reaction
against new reagents. Use control RNA.
■ Loss of probe in purification step. ■ Check success of probe purification
before use.
■ Poor hybridization. ■ Check that hybridization buffer and
protocol are compatible with slide type.
■ Failure of scanning instrument. ■ Test performance of scanner with known
amounts of fluors.
■ Detection sensitivity too low. ■ Adjust detection sensitivity.
■ CyDye have been exposed to light ■ Protect CyDye from light.
during handling.
■ Target genes not expressed in ■ Use housekeeping genes and positive
examined tissue. controls to ascertain proper functioning
of the system.
■ Human error at some stage. ■ Repeat experiment and use step controls
to monitor progress.
Low or undetectable Cy3 and/or ■ Poor retention of targets on slide. ■ Check purity and concentration of
Cy5 signal. Identical probes were hybridized targets. Use slides of different batch.
with two different types of slide that Use different slide type.
contained the same targets
(Fig 94n, 94o).
■ Targets are old. ■ Prepare new targets for spotting.
■ Poor labelling reaction with one dye. ■ Check success of labelling reaction. Check
One or more components faulty in performance of all components of labelling
labelling reaction. reaction such as nucleotides, enzyme and
fluorescent nucleotides or reactive dyes.
■ Loss of probe in purification. ■ Check performance of purification. Do not
purify Cy3 and Cy5 probes together.
■ Unequal amount of Cy3 and Cy5. ■ Use equal amounts of
probes in hybridization.
■ Too little probe in hybridization. ■ Measure the amount of probe before
hybridization. Use more RNA to prepare
probe.
■ Free CyDye in probe. ■ Optimize probe purification.
■ Poor quality RNA sample or samples. ■ Test RNA before labelling.
■ Too much or too little quantity of RNA. ■ Measure amount of RNA before labelling.
■ RNA contaminated by DNA. ■ Use DNAse I to remove DNA.

● 151
 M I C R O A R R AY

Table 7 cont’d. Troubleshooting microarray experiments.

Symptom Possible cause Remedy
■ Hybridization conditions not optimal. ■ Check compatibility of hybridization
buffer and slide surface. Use lower
hybridization stringency.
■ High background in hybridization. ■ Use slides of a different batch.
Optimize hybridization and wash
protocol.
■ Overexposure of Cy3 and Cy5 to light ■ Protect CyDye from light always.
during storage and handling.
■ Laser source not working optimally. ■ Check laser performance.
■ Detection sensitivity not optimal. ■ Optimize laser power and detection
sensitivity settings.
■ Detection sensitivity not optimal. ■ Optimize laser power and detection
sensitivity settings.
■ Human error. ■ Repeat experiment with step controls.
Unbalanced Cy3 and ■ Too high labelling density leading to ■ Label to a lower density.
Cy5 signals. quenching of one fluorescent dye.
■ Too much CyDye nucleotide in ■ Use less CyDye nucleotide.
labelling reaction.
■ Too much of one probe in hybridization ■ Optimize the amount of probe in hybridization.
leading to quenching. Measure the amount of probe in hybridization.
■ Some nucleotide sequences label poorly. ■ Use a different labelled nucleotide.
Use a different labelling method.
■ Poor or variable quality of RNA ■ Use good quality RNA for labelling.
sample/samples. Biological variation Use pooled RNA samples.
in RNA samples.
■ High fluorescent background in . ■ See separate entry.
one color.
■ Normalization method is not adequate. ■ Use a different normalization method.
■ High amount of variation in experiment. ■ Optimize and standardize experimental
conditions to reduce amount of variation.
High background, weak ■ Poor labelling reaction. ■ Check success of labelling reaction
specific signals. before hybridization.
■ Random nonamers used for labelling ■ Prepare probe with oligo(dT) primer.
total RNA.
■ Probe fragments very short. ■ Use good quality RNA. Check purity
of RNA. Re-purify RNA.
Uneven fluorescent background ■ Poor slide quality with an uneven ■ Use slides of a different batch. Optimize slide
on slide. coating. Often background is higher surface treatment protocol. Use different
on one side of slide (Fig 94a). source of microscope slides.
■ Fluorescent background from ■ Optimize wash protocol. Include water dip at
pre-hybridization or the end. Dry slides quickly.
hybridization solution (Fig 94f).

hb 63-0048-49 ● 152
 C H A P T E R 1 2 : T R O U B L E S H O O T I N G M I C R O A R R AY E X P E R I M E N T S

Table 7 cont’d. Troubleshooting microarray experiments.

Symptom Possible cause Remedy
■ Salts in wash buffer dried onto ■ Dip slide into water before drying.
dried onto slide (Fig 94b).
■ Powder from lab gloves adheres ■ Handle slides using powder-free gloves.
to slide.
■ Edge of coverslip has dried during ■ Perform hybridization under humid
manual hybridization (Fig 94m). conditions.
Most spots give high ■ High amount of unspecific signals. ■ Increase stringency of hybridization
uniform signal. and washes.
Even fluorescent background. ■ Poor slide quality (Fig 94g, 94i). ■ Use slides from different batch.
Use different source of microscope slides.
■ Too much probe used in hybridization. ■ Quantify probe before use.
■ Over labelling of sample. ■ Optimize amount of probe to use.
Optimize labelling density.
Speckled background on slide. ■ CyDye nucleotides remain in probe ■ Optimize purification of probe.
(Fig 94d). Check probe for presence of free CyDye.
Particles seen on slide. ■ Dust particles have been fixed ■ Always handle slides in clean environment.
onto slide. Use air stream to remove any dust particles
from dry slides before spotting and use.
■ Slide surface is scratched (Fig 94p). ■ Handle slides with care using forceps.
■ Finger prints seen on slide. ■ Never touch slides with bare hands.
Bubble effect on slide. ■ Air has been trapped under ■ Remove air bubbles from hybridization.
coverslip (Fig 94c).
Spots appear as comets ■ Hybridized probe is coming loose ■ Optimize wash conditions. Dry slides
with tails. during low stringency wash/water quickly after washes and water dip.
dip (Fig 94j).
Deformed spots. ■ Doughnut spots (Fig 94l). ■ Control humidity of spotting process.
■ Tiny spots.(Fig 94e). ■ Test printing pen performance.
■ Variably sized spots (Fig 94k). ■ Wrong spotting buffer for slide
chemistry.
■ Negative spots caused by slide ■ Use slides of a different batch.
background which is higher than
hybridization signals (Fig 94h).

● 153
 M I C R O A R R AY

Fig 94. Troubleshooting microarray experiments.

a) b) c)

d) e) f)

g) h) i)

hb 63-0048-49 ● 154
 C H A P T E R 1 2 : T R O U B L E S H O O T I N G M I C R O A R R AY E X P E R I M E N T S

j) k) l)

m) n) o)

p)

● 155
 M I C R O A R R AY

hb 63-0048-49 ● 156
 REFERENCES

References
References cited in text
1. Baldwin, D. et al. A comparison of gel-based, nylon filter and
microarray techniques to detect differential RNA expression in
plants. Curr Opinions in Biol. 2, 96–103 (1999).

2. Watson, A. et al. Technology for microarray analysis of

gene expression. Current Opinions in Biotech. 9 609–614 (1998).

3. Schena, M. et al. Microarrays: Biotechnology’s discovery platform

for functional genomics. Trends in Biotech. 16, 301–306 (1998).

4. Kozian, D. H. and Kirschbaum, B. J. Comparative gene expression

analysis. Trends in Biotech. 17, 73–78 (1999).

5. Braxton, S. and Bedilion, T. The integration of microarray

information in the drug development process. Current Opinions
in Biotech. 9, 643–649 (1998).

6. Mirnics, K. et al. Analysis of complex brain disorders with gene

expression microarrays: Schizophrenia as a disease of the synapse.
Trends in Neuroscience 24, 479–486 (2001).

7. Schulze, A. and Downward, J. Navigating gene expression using

microarrays — a technology review. Nature Cell Biology 3,
e190–e195 (2001).

8. Van Berkum, N. L. and Holstege, F. C. P. DNA microarrays: raising

the profile. Current Opinions in Biotech. 12, 48–52 (2001).

9. Alizadeh, A. A. et al. Towards a novel classification of human

malignancies based on gene expression patterns. J Pathol. 195(1),
41–52 (2001).

10. DeRisi, J. et al. Use of cDNA microarray to analyze gene expression

patterns in human cancer. Nature Genetics 14, 457–460 (1996).

11. Rew, D. A. DNA microarray technology in cancer research.

European Journal of Surgical Oncology 27, 504–508 (2001).

12. Shoemaker, D. D. et al. Experimental annotation of the human

genome using microarray technology. Nature 409, 922–927 (2001).

13. Lieb, J. D. et al. Promoter-specific binding of Rap1 revealed by

genome-wide maps of protein-DNA association. Nature Genetics 28,
327–334 (2001).

14. Hu, G. K. et al. Predicting splice variant from DNA chip expression
data. Genome Research 11, 1237–1245 (2001).

● 157
 M I C R O A R R AY

15. Meltzer, P. S. Spotting the target: microarrays for disease gene

discovery. Current Opin in Genetics and Dev. 11, 258–263 (2001).

16. Sapolsky, R. J. et al. High-throughput polymorphism screening and

genotyping with high-density oligonucleotide arrays. Genet Anal.
14(5-6), 187–92 (1999).

17. Larsen, L. A. et al. Recent developments in high-throughput

mutation screening. Pharmacogenomics 2(4), 38799 (2001).

18. Drobyshev, A. et al. Sequence analysis by hybridization with

oligonucleotide microchip: identification of beta-thalassemia
mutations. Gene 188(1), 45–52 (1997).

19. Lockhardt, D. J. and Winzeler, E. A. Genomics, gene expression

and DNA arrays. Nature 405, 827–836 (2000).

20. Jain, K. K. Applications of biochip and microarray systems in

pharmacogenomics. Pharmacogenomics 1, 289–307 (2000).

21. Gray, N. S. et al. Exploiting chemical libraries, structure and

genomics in the search for kinase inhibitors. Science 281,
533–538 (1998).

22. Kane, M. D. et al. Assessment of the sensitivity and specificity

of oligonucleotide (50mer) microarrays. Nucleic Acids Res. 28,
4552–4557 (2000).

23. Li, F. and Stormo, G. D. Selection of optimal DNA oligos for gene
expression arrays. Bioinformatics 17, 1067–1076 (2001).

24. Lockhart, D. J. et al. Expression monitoring by hybridization to

high-density oligonucleotide arrays. Nat Biotech. 14(13), 1675–1680
(1996).

25. Schena, M. et al. Parallel human genome analysis: Microarray-based

expression monitoring of 1000 genes. Proc. Natl. Acad. Sci. 93,
10614–10619 (1996).

26. Schepinov, M. S. et al. Steric factors influencing hybridization of

nucleic acid to oligonucleotide arrays. Nucl Acids Res. 25,
1155–1161 (1997).

27. Schena, M. et al. Microarrays: biotechnology's discovery platform

for functional genomics. Trends Biotechnol. 17, 217–218 (1997).

28. Welford, S. M. et al. Detection of differentially expressed genes in

primary tumor tissues using representational differences analysis
coupled to microarray hybridization. Nucleic Acids Res. 26,
3059–65 (1998).

hb 63-0048-49 ● 158
 REFERENCES

29. Penn, S. G. et al. Mining the human genome using microarrays

of open reading frames. Nat Genet 2000 26, 315–318 (2000).

30. Hedge, P. et al. A concise guide to cDNA microarray analysis.

BioTechniques 29, 548–562 (2000).

31. Knight, J. When the chips are down. Nature 410, 860–861 (2001).

32. Taylor, E. et al. Sequence verification as quality control.

Biotechniques 31, 62–65 (2001).

33. Schuchhardt, J. et al. Normalization strategies for cDNA

microarrays. Nucl. Acids Res. 28, e47 (2000).

34. Yamanaka, Y. et al. Gene expression profiles of human small airway

epithelial cells treated with low doses of 14- and 16-membered
macrolides. Biochem. Biophys. Res. Comm. 287, 198–203 (2001).

35. Lipschutz, R. J. et al. High density synthetic oligonucleotide arrays.

Nature genetics 21(supplement), 20–24 (1999).

36. Zammatteo, N. et al. Comparison between different strategies of

covalent attachment of DNA to glass surfaces to build DNA
microarrays. Anal Biochem. 280(1), 143–50 (2000).

37. Worley, J. et al. Microarray Biochip Technology (Schena, M., ed.),

Eaton Publishing/BioTechniques Books, Natick, MA, pp. 65–85
(2000).

38. DeRisi, J. L. et al. Exploring the metabolic and genetic control of

gene expression on genomic scale. Science 278, 680–686 (1997).

39. Fodor, S. P. et al. Light-directed, spatially addressable parallel

chemical synthesis. Science 251, 767–73 (1991).

40. Mujumdar, R. B. et al. Cyanine dye labeling reagents:

sulfoindocyanine succinimidyl esters. Bioconjug Chem. 4(2),
105–11 (1993).

41. Yu, H. et al. Cyanine dye dUTP analogs for enzymatic labeling of
DNA probes. Nucleic Acids Res. 22, 3226–32 (1994).

42. Sambrook, J. et al. Molecular Cloning: A Laboratory Manual, Cold

Spring Harbor Laboratory Press, Cold Spring Harbor, New York
(1989).

● 159
 M I C R O A R R AY

43. Ausubel, F. M., Current Protocols in Molecular Biology, Greene

Publishing Associates and Wiley-Interscience, New York (2000).

44. Gruffat, D. et al. Comparison of four methods for isolating large

mRNA: Apolipoprotein B mRNA in bovine and rat livers. Anal
Biochem. 249, 77–83 (1996).

45. Rosenow, C. et al. Prokaryotic RNA preparation methods useful

for high density array analysis: comparison of two approaches.
Nucl. Acid Res. 29, e112 (2001).

46. (Farrell, R. E., ed.), RNA Methodologies, A laboratory guide for

Isolation and Characterization, Academic Press, Inc., New York,
pp. 125–157 (1997).

47. Randolp, J. B. and Waggoner, A. S. Stability, specificity and

fluorescence brightness of multiply-labeled fluorescent DNA
probes. Nucl Acid Res. 25, 2923–2929 (1997).

48. Smoot L. M. et al. Global differential gene expression in response to

growth temperature alteration in group A Streptococcus. Proc. Natl.
Acad. Sci. 98, 10416–421 (2001).

49. Talaat A. M. et al. Genome-directed primers for selective labeling of

bacterial transcripts for DNA microarray analysis. Nature Biotech.
18, 679–682 (2000).

50. Van Gelder, R. N. et al. Amplified RNA synthesized from limited

quantities of heterogeneous cDNA. Proc Natl Acad Sci. 87,
1663–1667 (1990).

51. Eberwine, J. H. et al. Analysis of gene expression in single live

neurons. Proc Natl Acad Sci. 89, 3010–3014 (1992).

52. Luo, L. et al. Gene expression profiles of laser-captured adjacent

neuronal subtypes. Nature Medicine 5, 117–122 (1999).

53. Herrler, M. Use of SMART-generated cDNA for differential gene

expression. J Molecular Medicine 78, B23 (2000).

54. Southern, E. M. Detection of specific sequences among DNA

fragments separated by gel electrophoresis. J. Mol. Biol. 98, 503
(1975).

55. Casey, J. and Davidson, N. Rates of formation and thermal

stabilities of RNA:DNA and DNA:DNA duplexes at high
concentrations of formamide. Nucleic Acids Res. 4, 1539–1551
(1977).

hb 63-0048-49 ● 160
 REFERENCES

56. Southern, E. et al. Molecular interactions on microarrays. Nat

Genet. 21(1 Suppl), 5–9 (1999).

57. O’Shea, D., in Introduction to Lasers and their applications,

Addison-Wesley, Reading, MA, pp. 51–78 (1978).

58. Smith, W. J., in Modern Optical Engineering, McGraw Hill, Boston,

MA, pp. 142–145 (1990).

59. Skoog, D. A. et al., in Principles of Instrumental analysis, Harcourt

Brace, Philadelphia, p. 108 (1998).

60. Gonzalez, R. C. and Woods, R. E., in Digital Image Processing,

Addison-Wesley, Reading, MA, pp. 31–37 (1978).

61. Brazma A. et al. One-stop shop for microarray data. Nature. 403,
699–700 (2000).

62. Brazma, A. and Vilo, J. Gene expression data analysis. FEBS Letters
480, 17–24 (2000).

63. Brazma, A. et al. Minimum information about a microarray

experiment (MIAME) — toward standards for microarray data.
Nature genetics. 29, 365–371 (2001).

64. Lee, M. et al. Importance of replication in microarray gene

expression studies: Statistical methods and evidence from repetitive
cDNA hybridizations. Proc Natl Acad Sci. 97, 9834–9839 (2000).

65. Bernard, K. et al. Multiplex messenger assays: simultaneous,

quantitative measurement of expression of many genes in the context
of T cell activation. Nucl Acids Res. 24, 1435–1442 (1996).

66. Yue, H. et al. An evaluation of the performance of cDNA

microarrays for detecting changes in global mRNA expression.
Nucleic Acids Res. 29, e41 (2001).

67. Tseng, G. C. et al. Issues in cDNA microarray analysis: quality

filtering, channel normalization, models of variations and
assessment of gene effects. Nucl. Acids Res. 29, 2549–2557 (2001).

68. Hemmilä, I.A., Applications of Fluorescence in Immunoassays,

John Wiely and Sons, Inc. New York (1991).

● 161
 M I C R O A R R AY

Product Quantity Code Number

Microarray probe preparation

CyScribe First-Strand cDNA Labelling Kit 25 reactions RPN6200
CyScribe First-Strand cDNA Labelling Kit 25 reactions RPN6200X
with CyScribe GFX Purification Kit
CyScribe First-Strand cDNA Labelling System-dUTP 50 reactions RPN6201
CyScribe First-Strand cDNA Labelling System-dUTP 50 reactions RPN6201X
with CyScribe GFX Purification Kit
CyScribe First-Strand cDNA Labelling System-dCTP 50 reactions RPN6202
CyScribe First-Strand cDNA Labelling System-dCTP 50 reactions RPN6202X
with CyScribe GFX Purification Kit
CyScribe Post-Labelling Kit 24 reactions RPN5660
CyScribe Post-Labelling Kit with CyScribe 24 reactions RPN5660X
GFX Purification Kit
CyScribe Direct mRNA Labelling Kit 24 reactions RPN5665
Cy3-dCTP 25 nmol PA53021
Cy5-dCTP 25 nmol PA55021
Cy3-dUTP 25 nmol PA53022
Cy5-dUTP 25 nmol PA55022
Cy3-UTP 100 nmol PA53026
Cy5-UTP 100 nmol PA55026
CyDye Post-Labelling Reactive Dye Pack 24 reactions RPN5661

Lucidea brand products

Lucidea Array Spotter 1 63-0040-09
Lucidea SlidePro Module 1 1 18-1162-01
Lucidea SlidePro Module 2 1 18-1162-02
Lucidea SlidePro Module 3 1 18-1162-03
Lucidea SlidePro Module 4 1 18-1162-04
Lucidea SlidePro Module 5 1 18-1162-05
Lucidea Automated Spotfinder 1 63-0038-18
Lucidea Universal ScoreCard 200 hybridizations 63-0042-85
Lucidea Reflective Slides Available soon

Microarray scanning systems

Typhoon 9610 Variable Mode Imager 1 63-0038-55

hb 63-0048-49 ● 162
 O R D E R I N G I N F O R M AT I O N

Product Quantity Code Number

Image analysis software

ArrayVision for Scanners 1 ARV-100
Spotfire DecisionSite for Functional Genomics 1 63-0036-56
ImageQuant Solutions for Windows 2000 for Scanners 1 63-0035-16

Bioinformatics products
Scierra Microarray Laboratory Workflow System

Electrophoresis systems
Ready-to-Run Separations Unit 1 80-6460-95

RNA purification products

RNase-free water 500 ml US70783
QuickPrep Total RNA Extraction Kit 1 27-9271-01
RNA Extraction Kit 1 27-9270-01
QuickPrep Micro mRNA Purification Kit 1 27-9255-01

Spectrophotometry products
Ultraspec 3300 pro UV/Visible Spectrophotometer 1 80-2112-33

Hybridization equipment and reagents

Microarray Hybridization Solutions version 2 1 RPK0325
Humid Hybridization Cabinet for microarrays 1 RPK0176
SSC 20
100 ml US19629
SDS 20% 500 ml US75832

Other
Hybond-N+ Membrane 50 RPN82B
Vistra Green Nucleic Acid Stain 500 ml RPN5786
RapidGel-XL - 6% 100 ml US75861
RapidGel-XL - 8% 100 ml US75862
ALFexpress Sizer 50-500 50 27-4539-01
TBE Buffer Pre-mixed Powder 10
6 bottles US70454

● 163
 M I C R O A R R AY

Index

A CyScribe Post-Labelling Kit 73-76

aldehyde 23-24 CyScribe Direct mRNA Labelling

Kit 77
ALFexpress Sizer 89
CyScribe GFX Purification Kits 70
algorithms 11, 130

aminosilane 13, 23, 25

D
ArrayVision 130, 132
data analysis 129
AutoSeq 70
deposition 10, 13, 18, 20-22, 24

differential gene expression 6-7, 38,

B
51, 54, 69, 74-75, 105, 147-148
bacterial RNA 46-47, 57 DNA fragments
beamsplitters, fluorescence 125 as genetic content 10-13

bioinformatics 141
E

C electrophoresis 13, 48, 81, 89, 90, 91

characterizing labelled probe eurkaryotic 46, 55, 57

spectrophotometry 81-84 extinction coefficient 30-31, 37, 81, 83
thin layer chromatography 84-87
PAGE 90 experimental design 148-149

chromatography 27, 48, 85-86, 89

F
clustering 129, 139-140
fluorescence
confocal scanning 115, 117
detection 38-39, 89-91, 114,
contact printing 2, 10, 21 124-125
controls 4-15, 104, 108-109, excitation 30, 112-114
127-129, 133-138, 149 emission 30
collection of 116-117
CyDye Fluorophores 34-37 filtration 118-123
CyDirect 77, 79 intensity 130
variation 148
CyScribe First-Strand cDNA
Labelling Kits 70-72

hb 63-0048-49 ● 164
 INDEX

fluorescent dyes 4-5, 7, 29, 32, 34,

38, 54, 59, 61, 64, 70, 100, 120, M
147
melting temperature 96
fluorochrome 29, 38-39, 111-112,
114, 120-124 microarray
applications 6-8
fluorophore 29, 30-34, 38-39, 53 bioinformatics 141-142
hybridization 5, 94-104
H slides 23-26, 101-102

housekeeping genes 14, 135, 137

N
hybridization
manual 95-101 normalization 15, 129, 135-139
automated 103-109
O
I
oligo(dT) 46-47, 49, 56-57, 60, 66,
image analysis 130-131 71-73, 133

ImageQuant Image and Analysis oligonucleotides 2, 8-11, 14-15

Software 87 deposition 18, 24-25
synthesis 17-18

L open reading frames 8, 12

labelling density 7, 33, 53-55, 61, 64,

P
68-69, 72-73, 78-79, 88

labelling methods 55-57 photobleaching 31, 39, 100

enzymatic 58-59 photolithography 2,18
first-strand synthesis 60-61
cDNA post-labelling 62-65 piezoelectric printing 20

Lucidea Array Spotter 22 poly-lysine 23, 25

Lucidea Automated Spotfinder 130, post-labelling 62-65, 73-76, 79

132 probe labelling 4
Lucidea Reflective Slides 26, 102 prokaryotic 47
Lucidea SlidePro Hybridizer 22, 103

Lucidea Universal ScoreCard 22, 14,

136

● 165
 M I C R O A R R AY

Q signal-to-noise ratios 105, 138

quality controls 150 specificity 11-12, 15, 52, 57

quantum yield 30-31 spectrophotometry 81-84

quenching 33-34, 53-55, 61, 82 spot morphology 19-20, 23, 131

spotfinding software 132

R Spotfire DecisionSite for Functional
Genomics 140
radioactive spiking 84-85, 89
stokes shift 30, 32
random priming 56-57, 69, 71-73
syringe-solenoid deposition 20
reflective slides 26, 97, 102

relative fluorescence units (RFU) 39

T
reverse trancriptase 1, 49, 60, 63,
target nucleic acids 27
65-67, 73, 137
target sequences 9, 11-12, 14, 52,
ribonuclease enzymes 41-43, 47, 79
55-56, 133, 146
RNA
troubleshooting 151-155
contamination 41-43, 46, 49, 79
degradation 41-44, 48-49 Typhoon Imager 86, 89, 91, 125
isolation 45-49
measurement 48
V
RNA ampification 66-68
visualization software 140

S
Y
saturation 38, 98, 122, 136
yellow experiment 55, 147
scanners 39, 81, 89, 111-116,
119-124

Scierra Laboratory Workflow System

141-142

hb 63-0048-49 ● 166