Content uploaded by Sylvia Cohen-Kaminsky
Author content
All content in this area was uploaded by Sylvia Cohen-Kaminsky
Content may be subject to copyright.
Available via license: CC BY 4.0
Content may be subject to copyright.
Vol. 44 ı No. 1 ı 2008 www.biotechniques.com ı BioTechniques ı 109
Research Reports
INTRODUCTION
The ability to quantify the expression
levels of all genes in a given tissue or
cell with a single assay is an exciting
promise. Two technologies are widely
used: quantitative reverse transcription
PCR (RT-PCR) and DNA microarray.
While quantitative RT-PCR is easily
applicable to a few hundred genes
(1,2), DNA microarray technology
has proven very useful for studying
gene expression variations between
many biological samples at the whole
transcriptome scale. Although the
successful application of microarray-
based expression analysis was demon-
strated in a number of applications,
the main problem with this approach
is the fact that hybridization signals
do not necessarily correlate with target
concentrations (3–5).
To get reliable microarray hybrid-
ization data, each probe should be
specific for a unique transcript, and all
probes on the same microarray should
have similar thermodynamic stability.
In addition, hybridization signals should
be maximal in order to reach the highest
sensitivity. During probe design, these
requirements can be considered using
a theoretical sequence-based approach
for prediction of duplex stability
and hairpin formation. However, all
predictions are based on hybridization
models deduced from experiments
in solution and not from experiments
in which probes are immobilized on
DNA microarrays. Even fully comple-
mentary oligonucleotide targets
hybridized to immobilized probes
of the same size might give different
hybridization signals at equilibrium.
Evidently, hybridization with complex
cDNA targets will worsen the situation
because of stable secondary structures
of the targets themselves or interac-
tions between targets. A possibility of
competition between the targets can
also complicate the hybridization data
(6–8). Since expression levels can vary
considerably from one gene to another,
there will be high variations in the
concentration of the different cDNAs
hybridized to the microarray. According
to standard protocols, hybridization is
usually performed overnight (15–18
h) (6). For low-copy genes, reaching
equilibrium might require hybrid-
ization times considerably longer.
Consequently for some probes, the
measured fluorescent signals will not
correspond to a plateau. There are
several important consequences to this.
First, as it is theoretically expected and
experimentally proven (9,10), in non-
equilibrium conditions, the proportion
of cross-hybridized targets is higher as
compared with equilibrium conditions.
Moreover, other factors including
thermodynamic stability of on-chip
formed duplexes and the presence
of secondary structures can change
the kinetic properties of probe-target
combinations.
If different oligonucleotide probes
corresponding to the same cDNA, even
of the same length and similar base
composition, show different kinetic
properties, this will have an impact on
hybridization signals and sensitivity.
In addition, for low-copy genes at
typically non-equilibrium conditions,
different expression results will be
obtained with different probes corre-
sponding to the same gene if the kinetic
properties of the probes are different.
Thus, depending on gene expression
levels and kinetic properties of the
On-chip hybridization kinetics for
optimization of gene expression experiments
Elena Khomyakova
1
, Mikheil A. Livshits
2
, Marie-Caroline Steinhauser
3
, Luce Dauphinot
3
,
Sylvia Cohen-Kaminsky
4
, Jean Rossier
3
, Françoise Soussaline
1
, and Marie-Claude Potier
3
BioTechniques 44:109-117 (January 2008)
doi 10.2144/000112622
DNA microarray technology is a powerful tool for getting an overview of gene expression in biological samples. Although the success-
ful use of microarray-based expression analysis was demonstrated in a number of applications, the main problem with this approach is
the fact that expression levels deduced from hybridization experiments do not necessarily correlate with RNA concentrations. Moreover,
oligonucleotide probes corresponding to the same gene can give different hybridization signals. Apart from cross-hybridizations and
differential splicing, this could be due to secondary structures of probes or targets. In addition, for low-copy genes, hybridization
equilibrium may be reached after hybridization times much longer than the one commonly used (overnight, i.e., 15 h). Thus, hybridiza-
tion signals could depend on kinetic properties of the probe, which may vary between different oligonucleotide probes immobilized on
the same microarray. To validate this hypothesis, on-chip hybridization kinetics and duplex thermostability analysis were performed
using oligonucleotide microarrays containing 50-mer probes corresponding to 10 mouse genes. We demonstrate that differences in
hybridization kinetics between the probes exist and can influence the interpretation of expression data. In addition, we show that using
on-chip hybridization kinetics, quantification of targets is feasible using calibration curves.
1
IMSTAR, Paris, France,
2
EIMB, Moscow, Russia,
3
CNRS UMR 7637, Paris, France, and
4
CNRS UMR 8078, Le Plessis-Robinson, France
110 ı BioTechniques ı www.biotechniques.com Vol. 44 ı No. 1 ı 2008
Research Reports
probes, contradictory results might be
obtained.
Indeed this is what we have
observed when studying mouse gene
expression in various tissues using
DNA microarrays probing neuro-
transmission genes, represented by
two or three 50-mer probes each. Out
of the 280 genes represented on the
microarray, 29 corresponding to 14
genes were selected for detailed experi-
ments. We performed complex analysis
of hybridization including on-chip
analysis of duplex thermostability and
measurement of on-chip hybridization
kinetics at different concentrations of
targets. Antisense Texas Red-labeled
oligonucleotide targets fully comple-
mentary to immobilized probes were
used. Based on these experiments,
we have analyzed some of the causes
for the observed variations in cDNA
hybridization to different probes and
propose the recommendations for
the improvement of the reliability of
expression data, particularly for low-
copy genes.
MATERIALS AND METHODS
Oligonucleotides
Twenty-nine oligonucleotide probes
corresponding to 14 mouse genes
involved in neurotransmission were
purchased from Eurogentec (Liège,
Belgium) as 5′ end amino-modified 50-
mer deoxyribonucleotides (Table 1).
Oligonucleotide targets complementary
to the 29 oligonucleotide probes were
purchased from Eurogentec as well.
Hybridization of cDNAs to the
NeuroTrans Oligoarrays
Total RNA was extracted from
thymus or cerebral cortex from B10D2
mice using the RNeasy kit (Qiagen,
Valencia, CA, USA). The quality of
RNA was checked on the Agilent 2100
Bioanalyzer (Agilent Technologies,
Santa Clara, CA, USA). Fifteen micro-
grams RNA were labeled with Cy5 by
reverse transcription into cDNAs using
7.5 μM random hexanucleotides, 10 μM
dithiothreitol (DTT), 125 μM dNTP, 25
μM dUTP-cyanine (Amersham plc,
Little Chalfont, Buckinghamshire, UK),
and 400 U of Superscript II (Invitrogen,
Carlsbad, CA, USA). Reference RNA
(equimolar mixture of thymus, thymo-
cytes, and cerebral cortex RNAs) was
labeled with Cy3. Labeled targets
were purified on QIAquick columns
(Qiagen). Cy3 and Cy5 cDNAs were
hybridized to NeuroTrans NT280M
containing oligonucleotides for 280
genes involved in neurotransmission
(two to three oligonucleotides per
gene; GPL4746 on GEO; www.ncbi.
nlm.nih.gov/geo). Hybridization was
performed at 42°C in 50% formamide,
4× standard saline citrate (SSC), 0.1%
sodium dodecyl sulfate (SDS), and
5× Denhart’s overnight. Slides were
washed for 2 min in 2× SSC, 0.1% SDS,
for 1 min in 0.2× SSC, for 1 min in 0.1×
SSC, and scanned on the ScanArray
Lite (Packard Biochip, Billerica, MA,
USA) using the same procedure for
each slide. Fluorescence on each spot
was extracted using the optical scan
array (OSA) image analysis software
package (www.imstarsa.com). Data are
Table 1. Probe Sets Corresponding to 14 Genes Selected for the Kinetic Experiments
Figure 1. The differences between hybridiza-
tion signals for probe pairs corresponding to
the same gene. (A) Expression levels (signal
from a spot divided by the total signal from the
microarray) after hybridizing cDNA targets from
mouse cerebral cortex or thymus overnight. (B)
Signal ratios for pairs of probes to mutual genes
in cerebral cortex and in thymus. (C) Signal ra-
tios between cerebral cortex and thymus for each
probe, deduced from panel A. Values are from
three independent microarray experiments.
C
A
B
Vol. 44 ı No. 1 ı 2008 www.biotechniques.com ı BioTechniques ı 111
Research Reports
accessible on GEO database under the
number GSE7574.
On-Chip Hybridization Kinetic
Experiments
Microarrays specially designed for
kinetic experiments were used. Amino-
modified oligonucleotide probes (Table 1)
were dissolved in 150 mM sodium
phosphate buffer, pH 8.3, at 12.5 μM
and spotted on activated glass slides
according to patent no. WO03068712
using the BioRobotics MicroGrid II
spotter (Genomic Solutions, Ann Arbor,
MI, USA). Each oligonucleotide probe
was spotted in quadruplet. Distance
between spots was 400 μm.
Oligonucleotide targets were labeled
at the 3′ end using terminal transferase.
One hundred picomoles of an oligonu-
cleotide were mixed with buffer, 5 mM
CoCl
2
, 0.05 mM Texas Red-5′-ddATP
(PerkinElmer, Waltham, MA, USA),
400 U terminal transferase (Roche,
Basel, Switzerland), and incubated for
2 h at 37°C. Reaction was stopped by
adding 2 μL 0.2 M EDTA, pH 8.0, on
ice. Labeled oligonucleotides were
purified on DyeEx columns (Qiagen)
and precipitated in 10 volumes of
acetone and 2% LiClO
4
overnight at
-20°C. After a 10-min centrifugation,
the pellet was rinsed once with acetone,
centrifuged again for 5 min, and
dissolved in water at 12.5 μM final
concentration. The oligonucleotide
concentration was adjusted after quanti-
fication using a ND-1000 spectropho-
tometer (NanoDrop Technologies,
Wilmington, DE, USA) at 260 nm.
Oligonucleotide labeling efficiency was
0.65 as estimated from the absorbance
ratio measured at 260 nm (DNA) and at
612 nm (Texas Red).
Labeled oligonucleotide targets
(equimolar mixture) were deposited
on the microarray and covered with
Gene Frame (Abgene, Epsom, UK).
The volume of the hybridization
chamber was 25 μL (1 cm × 1 cm).
Hybridization was carried out at 60°C
in 2.5× SSC, 0.2% SDS at different
target concentrations from 0.05 to 1
nM for kinetic experiments and 10 nM
for the thermostability measurement.
For registration of on-chip kinetics
and melting curves, OSA instrument
(IMSTAR, Paris, France) equipped
with a thermo-controlled microarray
holder was used (Reference 11 and
patent no. WO2004059302). Curves
were recorded by measuring the
fluorescence intensity on all the spots
of the microarray after scanning,
either at a fixed temperature (60°C
for kinetic curves) or at 2°C intervals
starting from 20° to 95°C (2°C/25 min
at lower temperatures to 2°C/200 s at
higher temperatures for equilibrium
melting curves). Each experiment was
started by the denaturation of probes
and targets within the hybridization
chamber at 95°C for 5 min. For kinetic
experiments, after on-chip denatur-
ation, the temperature was decreased to
60°C followed by the immediate start
of hybridization signals registration at
60°C.
Images were captured and analyzed
using OSA image analysis software
package. The positions of the micro-
array spots were determined using
Figure 2. On-chip melting curves for the 21 probes. Each probe pair is represented on a separate
graph. a.u., arbitrary units.
112 ı BioTechniques ı www.biotechniques.com Vol. 44 ı No. 1 ı 2008
Research Reports
an integrated algorithm of automated
spot finding on the hybridized micro-
array images. For kinetics and melting
experiments, the hybridization signal
S and local background fluorescence
intensity B corresponding to each
spot was calculated as described in
Reference 11. To compensate for the
temperature dependence of Texas
Red and its bleaching after several
measurements, the intensity I(T) of
each spot was calculated as described
in Reference 12:
The number of binding sites (N
binding
site
) involved in duplex formation on the
area of a microarray spot A
spot
[cm
2
]
was estimated from the maximum value
of the fluorescence intensity I
max
(upper plateau of the melting curve),
the concentration of oligonucleotide
targets in solution (C
target
), the thickness
of hybridization chamber (l = 500 μm),
and spot area A
spot
[cm
2
] as follows:.
N
binding sites
= 10
3
I
max
A
spot
l
û
c
target
The maximum value I
max
= 0.5 of
hybridization signal registered in the
experiment with 10 nM oligonucleotide
target at 20°C corresponds to 5 amol (5
× 10
-18
) of binding sites. Thus, the target
quantity exceeds at least 10
2
- to 10
3
-fold
the quantities of immobilized oligo-
nucleotide probes, thus guaranteeing
the constancy of target concentration
during hybridization and melting. The
invariability of background intensity at
a fixed temperature during the kinetics
measurements is an additional proof of
the constancy of the target concentration
during hybridization (data not shown).
RESULTS
Selection of Probes Giving Different
Hybridization Signals for the Same
Gene
Out of 22 hybridization experiments
performed on the NeuroTrans oligoarray
containing 50-mer probes with cDNAs
from mouse forebrain or cerebellum,
14 genes showed signal differences of
at least 10-fold between the two probes
corresponding to the same gene in a
minimum of four experiments (data not
shown). Table 1 shows the list of the
genes. We selected for detailed analysis
10 of these genes for which we obtained
sufficient fluorescent signals in hybrid-
ization kinetics (Table 1, not in gray).
For these 10 genes, signal intensities
were determined from six additional
hybridization experiments performed
with either cerebral cortex or thymus
cDNAs on NeuroTrans oligoarrays.
Figure 1A shows the hybridization
signals for each probe normalized to
the total signal on the oligoarray. Figure
1B gives the signal ratios for the pairs
of probes corresponding to the same
gene. There are 11 ratios for 10 genes,
because one gene was represented by
three probes instead of two. As expected,
the relative expression levels in cerebral
cortex and thymus were contradictory
depending on the oligonucleotide probe
(Figure 1C). To analyze the impact
of hybridization kinetics on gene
expression data, we performed complex
analysis, including on-chip study of
duplex thermostability and on-chip
hybridization kinetics at different target
concentrations. Antisense Texas Red-
labeled oligonucleotide targets fully
complementary to immobilized probes
were used for on-chip experiments in the
same experimental conditions (buffer
composition and temperature) as with
hybridization of cDNA.
Theoretical Model of On-Chip
Hybridization
The on-chip hybridization is a
bimolecular reaction between 50-mers
single-stranded oligonucelotide probes
(P) immobilized to the microarray
surface and the antisense target (T)
molecules, either oligonucleotide or
cDNA, present in the hybridization
solution.
P + T J
The kinetic equation of the corre-
sponding reaction describes the
increase of the quantity of target
molecules hybridized to probes on a
spot and the corresponding increase
of the fluorescent hybridization signal
(13):
[Eq. 1]
Figure 3. An example of on-chip hybridization
kinetics (probe pair 22/23). (A) Normalized
fluorescent signals obtained after 9 h hybridiza-
tion of targets complementary to probes 22 and
23. (B) Fluorescence ratios between probes 22
and 23 during hybridization kinetics deduced
from panel A. (C) Dependence of the hybridiza-
tion time (τ) on target concentration: example for
probes 22 and 23. 1/τ, hybridization rate.
A
B
C
Vol. 44 ı No. 1 ı 2008 www.biotechniques.com ı BioTechniques ı 113
Research Reports
Here, p is the surface density of the
immobilized probe, τ is the hybrid-
ization characteristic time, determined
by the effective rate constants of
association (k
ass
) and dissociation (k
diss
)
and the target concentration (c):
1/τ = k
diss
+k
ass
c = k
diss
(1+Kc)
[Eq. 2]
K = k
ass
/k
diss
is the equilibrium
binding constant for the considered
probe-target. Equation 2 is true also
in case of diffusion-limited kinetics
in spite of a different interpretation of
effective rate constants (14).
It is well known that the forward rate
constants k
ass
are very similar for perfect
match (perf) and for mismatch (mism)
duplexes, while reverse rate constants
k
diss
may differ considerably (9,15,16).
Typically, k
diss
(perf) << k
diss
(mism),
indicating that perfect duplexes are much
more stable than mismatch duplexes
formed upon cross-hybridization (16,17).
Thus, the equilibration time τ is longer
for perfect duplexes than for mismatch
duplexes (see Equation 2 and Reference
9). It was shown both theoretically
(10,18) and experimentally (19) that at
early stages of hybridization the highest
concentration species dominate (either
specific or non-specific), while at later
stages the highest affinity species displace
non-specific binding. Considerable
amount of cross-hybridization (non-
specific binding) is present under non-
equilibrium conditions (9,20).
The possible formation of secondary
structures in the probes and in the targets
can influence both the rate and the extent
of hybridization. The kinetic effects of
secondary structures are documented in
a number of studies (21,22).
To model the on-chip hybridization
kinetics, we used oligonucleotide
targets that were strictly comple-
mentary to the oligonucleotide probes
immobilized on the NeuroTrans micro-
array. The linear dependence (Eq. 2)
of the hybridization rate (1/τ) on target
concentration c allows to measure k
diss
and K and then to deduce τ at the cDNA
target concentrations used during gene
expression experiment.
Thermodynamic Analysis of On-
chip Formed Duplexes
Thermodynamic stability of the
hybridized duplexes was estimated
in a separate experimental study.
Oligonucleotides listed in Table 1 were
spotted onto custom microarrays. These
microarrays were hybridized with Texas
Red-labeled 50-mer oligonucelotide
targets complementary to immobilized
probes until equilibrium was reached.
Rather high oligonucleotide target
concentration (10 nM) was chosen
to ensure fast attainment of hybrid-
ization equilibrium. Hybridization was
performed in the same buffer as the
one used for cDNA hybridization. On-
chip melting of the duplexes was then
recorded. Normalized melting curves
are shown in Figure 2.
Equilibrium melting temperatures
(T
m
) obtained by fitting the theoretical
equation for the occupation level (12):
Figure 4. Simulation of hybridization kinetics for 10 probe sets corresponding to 10 different
genes. The simulation is based on extrapolated oligonucleotide kinetics data. Hybridization signal is
represented by the fraction of occupied probe molecules in a spot.
114 ı BioTechniques ı www.biotechniques.com Vol. 44 ı No. 1 ı 2008
Research Reports
to experimentally obtained melting
curves are represented in Table 2.
However, the similarity of melting
curves observed in the thermodynamic
experiments does not guarantee the
identity of hybridization kinetics. An
example using the experimental data for
the probe pair 22-23 is represented in
Figure 3. Figure 3A shows the difference
in on-chip normalized kinetic curves
for probes 22 and 23 recorded at target
concentrations 0.05, 0.1, 0.25, and 1
nM. The dependence of hybridization
signal ratios between probes 22 and 23
on time is represented in Figure 3B. At
high target concentrations (1 nM), the
equilibration time was short, and the
final value of the ratio between probe 22
and probe 23 was close to 1 (Figure 3B).
For lower target concentrations (0.1 and
0.25 nM), because of the difference in
hybridization kinetics between probes 22
and 23, at the beginning of hybridization,
the signal ratio was high, especially at
concentration 0.1 nM. However, when
approaching equilibrium, the ratio
decreased to smaller values close to 1.
At the lowest target concentration (0.05
nM) even after 50,000 s of hybridization,
the ratio 22/23 was only approaching 1
(Figure 3B).
On-chip Hybridization Kinetics
For a detailed study of the kinetic
effects, we have measured the on-chip
hybridization kinetics of Texas Red-
labeled 50-mer oligonucleotide targets
at four concentrations: 0.05, 0.1, 0.25,
and 1 nM. Results of fitting (K and
k
diss
values) are listed in Table 2. These
experiments allowed the calculation
of τ as well as the modeling of hybrid-
ization kinetic curves at any target
concentration.
Simulation of Kinetics Curves from
the Theoretical Model
As an example we simulated the
kinetic curves for 0.01 and 0.005 nM
target concentrations, using Equation
1, and K and k
diss
values listed in
Table 2. Evidently, such estimation
of kinetic parameters K and k
diss
has a
limitation coming from experimental
errors that are more essential at lower
target concentrations. Figure 4 shows
the curves obtained for all the probes
tested. Figure 5 indicates the simulated
hybridization signal ratios after 15
h of hybridization (panel A) and at
saturation (panel B) for the analyzed
probe pairs at four concentrations of
oligonucleotide targets: 0.05, 0.01,
0.005, and 0.001 nM. For slow kinetics,
duplex hybridization ratios are different
depending on the hybridization time.
DISCUSSION
The discrepancy between hybrid-
ization signals obtained from different
probes corresponding to the same
gene is a problem for gene expression
analysis. Except for the well-known
alternative splicing and stable secondary
conformation of cDNA targets, some
Figure 5. Comparison of pair ratios for
simulated hybridization signals after 15 h
of hybridization (A) and at equilibrium (B).
Melting temperature (T
m
) deduced from on-chip
experiments.
Probes T
m
(°C) k
ass
K
3 60.8 1.64 × 10
6
3.23 × 10
10
4 62.1 1.45 × 10
6
2.67 × 10
10
5 59.3 9.74 × 10
5
3.97 × 10
10
6 56.4 4.96 × 10
5
1.46 × 10
11
7 58.3 7.93 × 10
5
1.01 × 10
10
8 55.7 1.55 × 10
6
1.23 × 10
10
9 57.1 1.19 × 10
6
1.03 × 10
10
12 57.8 9.37 × 10
5
3.25 × 10
10
13 56.7 1.71 × 10
6
2.63 × 10
10
14 62.3 1.65 × 10
6
4.95 × 10
10
15 64.4 1.05 × 10
6
2.17 × 10
10
18 59.3 1.10 × 10
6
5.24 × 10
10
19 59.6 1.72 × 10
6
4.85 × 10
10
20 58.4 1.53 × 10
6
2.62 × 10
10
21 64.6 9.44 × 10
5
1.17 × 10
10
22 68.7 1.30 × 10
6
4.46 × 10
10
23 67.0 8.23 × 10
5
9.90 × 10
10
26 62.3 7.51 × 10
5
1.39 × 10
10
27 56.6 1.35 × 10
6
1.63 × 10
10
28 65.6 1.14 × 10
6
2.58 × 10
10
29 63.7 7.79 × 10
5
8.70 × 10
9
T
m
, melting temperature; k
ass
, rate constant of association; K, equilibrium binding constant.
Table 2. Kinetics Parameters and T
m
Deduced from On-Chip Experiments
A
B
Vol. 44 ı No. 1 ı 2008 www.biotechniques.com ı BioTechniques ı 115
Research Reports
other factors could influence micro-
array expression results. To analyze
these factors we performed thermo-
stability study and analysis of on-chip
hybridization kinetics at four concen-
trations of complementary oligonucel-
otide targets: 1, 0.25, 0.1, and 0.05 nM.
Theoretical analysis applied to our on-
chip kinetics data allowed estimation of
kinetic behavior at target concentrations
corresponding to cDNA concentrations
in gene expression experiments.
The probes selected for the study
showed high hybridization signals as
well as a significant difference between
probes corresponding to the same gene
(Figure 1). Figure 1B indicates at least
a 2-fold difference in the hybridization
signals for each pair of probes in
cerebral cortex and/or thymus. When
the difference is the same for both
tissues, it is most probably caused by
different secondary structures of the
probes or/and of the targets. Such a case
makes no confusion in the interpretation
of the expression data. However, several
probe pair ratios were found to differ
in cerebral cortex and thymus (probes
3-4, 7-8, 7-9, 14-15, 18-19, 20-21, 22-
23, 28-29) (Figure 1B). This entailed
a difference in expression level ratios
between cerebral cortex and thymus
depending on the probe (Figure 1C).
This difference could be due to the
presence or absence of alternative
transcripts in cerebral cortex or thymus
that would correspond to one probe or to
the other. Indeed probes from pairs 19-
18, 20-21, 22-23, and 29-28 map either
to different exons or to alternative 5′ or
3′ untranslated regions (UTRs) in two
databases (www.ensembl.org; genome.
ewha.ac.kr). However, for pairs 3-4, 14-
15, and 27-28, the two probes mapped
to the same transcript. Alternatively, the
observed tissue-dependent variations
in signal ratios between two probes
corresponding to the same gene could
result from differences in hybridization
kinetics at different cDNA target concen-
trations. In addition, kinetic properties
of the probes could be different within
the same microarray. Thus, if a gene is
highly expressed, hybridization signals
at the end of the experiment may reach
saturation levels. In this condition, signal
ratio between oligonucleotide pairs will
only depend on the equilibrium binding
constant K of on-chip duplex formation
(see above). However, for low-copy
genes, hybridization signals might be
rather far from saturation, and thus
hybridization signal ratios will depend
on the hybridization time and target
concentration.
To illustrate this, we performed
on-chip kinetics analysis at different
target concentrations and modeled
the behavior of the duplexes. We first
analyzed the temperature stability
of on-chip formed duplexes. The
experimental data (Figure 2) demon-
strate that thermodynamic stability
of the duplexes, including also the
enthalpy ΔH of duplex formation,
reflected by the shape of the melting
curves, is similar for all immobilized
probes tested. The minor differences
observed cannot explain the difference
in hybridization signals of oligonucle-
otide probes belonging to the same
gene. In addition, the experimental
data confirmed the correct design of
oligonucleotide probes and the absence
of degradation of either probes or
targets. However, the similarity of
melting curves does not guarantee
the identity of hybridization kinetics.
Indeed, probes 22 and 23 have similar
melting curves but different hybridiza-
tions curves (Figure 3, A and B). This
suggests that signal ratios between the
two probes depend on target concen-
tration and hybridization time.
Based on the kinetic data obtained at
higher concentrations of targets using
Equations 1 and 2, it was possible to
model the kinetics behavior of the
duplexes at target concentrations corre-
sponding to real cDNA hybridization
experiments. DNA target concentration
in biological samples was estimated
using the data of Reference 23.
Approximately 10 μg RNA are typically
used for microarray experiments. This
amount corresponds to RNA quantities
contained in 10
6
cells. The copy number
of single genes in a cell varies from
1 to 10
4
, and the efficiency of reverse
transcription is about 20% (24). Thus,
the expected concentration of cDNA
in 20 μL hybridization solution is 10
-14
to 10
-10
M. Unfortunately, the on-chip
kinetic measurements at concentrations
lower than 5 × 10
-11
M (0.05 nM) are
hardly feasible because the fluorescent
hybridization signals are too low as
compared with the fluorescence of the
hybridization solution. Thus, according
to Equation 2, using the linear depen-
dence of 1/τ value measured at concen-
trations 0.05, 0.1, and 0.25 nM, we can
estimate the K and k
diss
values for each
probe-target pair (Figure 4 and Table
2). Based on these data, we were able to
extrapolate the kinetics behavior at low
target concentrations after 15 h hybrid-
ization and at saturation (Figure 5, A
and B). Figure 4 shows that 15 h (54,000
s) hybridization time was not always
sufficient for reaching at least 70% of
maximal hybridization signal, especially
at 0.005 nM target concentration (probes
6, 12 14, 18, 23, and 28). Obviously, for
target concentration below 0.005 nM,
the hybridization kinetics would be even
slower. Thus, one can expect that after
hybridization times used in classical
experiments (roughly 15 h), the majority
of hybridization signals is rather far
from saturation level.
While at equilibrium (saturated
signals) a decrease in target concen-
tration leads to an increase of the
absolute value of the hybridization signal
ratio (Figure 5B), at non-equilibrium
conditions (15 h of hybridization), and
at low concentrations, some probe pairs
(5-6, 12-13, 18-19, and 22-23) had
ratios inverted as compared with the one
obtained at equilibrium. Moreover, two
pairs (5-6 and 28-29) had lower ratios
after 15 h as compared with equilibrium
conditions. Since probes 6 and 28 are
characterized by slow hybridization
kinetics (Figure 4), 15 h is definitively
not enough for reaching equilibrium.
In pairs of probes addressed to
mutual genes, the highest hybridization
signals were obtained with probes 4, 6,
8, 13, 14, 19, 20, 23, 27, and 29 (Figure
1A). Only some of these probes (14,
19, 20, and 27) showed hybridization
signals that differed more than 4-fold
for cDNA targets from cerebral cortex
and thymus (Figure 1C). Simulation of
kinetics indicated that for these probes,
a decrease in target concentration is
expected to induce an increase of the
hybridization signal ratio between
probes within the same pair (Figure 5A).
Indeed for probe 14, which corresponds
to a gene that is expressed more in
cerebral cortex than in thymus (Figure
1A), a decrease in target concentration
induced a decrease in hybridization
signal (Figure 5A). However, for probes
116 ı BioTechniques ı www.biotechniques.com Vol. 44 ı No. 1 ı 2008
Research Reports
19 and 20, a decrease in target concen-
tration in cerebral cortex as compared
with thymus induced an increase in
hybridization signal, which is opposite
to what was predicted. Thus for these
two probes, the kinetics model did not
comply with the expression data. The
most likely explanation may be the
presence of splicing variants in cerebral
cortex but not in thymus.
The tissue-dependent alternative
splicing could be proposed as an alter-
native explanation for the low hybrid-
ization signal for probe 22 as well.
However, since a very low hybridization
signal was observed for probe 22 both
in cerebral cortex and thymus, the most
probable explanation seems to be the
hairpin formation in cDNA fragment
corresponding to probe 22. Stable
secondary structure in cDNA could also
explain the hybridization results obtained
for probes 5, 6, 9, 12, 26, and 28.
Finally, for probe pair 3-4, there was
no difference in expression levels of the
corresponding gene between cerebral
cortex and thymus (Figure 1A). Thus,
the comparatively lower intensity of
hybridization signal corresponding to
probe 3 for both thymus and cerebral
cortex could also be the consequence of
cDNA secondary structure formation.
Based on our results and to improve
microarray gene expression results, we
can now make several recommenda-
tions. First, one should apply the best
experimental conditions for reaching
equilibrium for the majority of cDNAs
hybridized on the microarray. One way
would be to increase the hybridization
time at least up to 24 h. Another solution,
which seems to be more appropriate,
would be to decrease the volume of the
hybridization solution. Using micro-
fluidic stations will greatly accelerate
the hybridization process and essentially
improve the expression data analysis
(6). Secondly, fragmentation of cDNA
targets will result in destabilization of
the secondary conformations. Also, more
attention should be brought to oligonu-
cleotide probe design, including minimi-
zation of hairpin formation. It should be
noted, too, that the kinetic properties of
oligonucleotides are difficult to predict a
priori. Performing experimental studies
of on-chip kinetics for all the immobilized
probes is rather complicated. Thus, the
optimal solution could be to study on the
same microarray several oligonucelotide
probes (at least 3, ideally 5–10) comple-
mentary to different regions of the gene
and to select for expression analysis only
the probes showing the highest hybrid-
ization signals. This should be helpful in
avoiding the misinterpretation coming
from slow-kinetics probes as well as in
excluding the influence of alternative
splicing and of secondary structures of
DNA targets on the hybridization results.
Finally, we demonstrate that
following the linear dependence
between the characteristic hybridization
kinetics and the target concentration, it
is possible, using calibration curves, to
deduce the target copy numbers from
on-chip hybridization experiments.
ACKNOWLEDGMENTS
The authors wish to thank E.
Tatarinova for very helpful discussion, C.
Mikonio, D. Le Clerre, and F. Richard for
spotting the oligoarrays, and S. Ursulet
and E. Dreval for technical help.
COMPETING INTERESTS
STATEMENT
The authors declare no competing
interests
REFERENCES
1. Dolganov, G.M., P.G. Woodruff, A.A.
Novikov, Y. Zhang, R.E. Ferrando, R.
Szubin, and J.V. Fahy. 2001. A novel method
of gene transcript profiling in airway biopsy
homogenates reveals increased expression of
a Na+-K+-Cl- cotransporter (NKCC1) in asth-
matic subjects. Genome Res. 11:1473-1483.
2. Czechowski, T., R.P. Bari, M. Stitt, W.-R.
Scheible, and M.K. Udvardi. 2004. Real-time
RT-PCR profiling of over 1400 Arabidopsis
transcription factors: unprecedented sensitivity
reveals novel root- and shoot-specific genes.
Plant J. 38:366-379.
3. Hekstra, D., A.R. Taussig, M. Magnasco,
and F. Naef. 2003. Absolute mRNA concen-
trations from sequence-specific calibration of
oligonucleotide arrays. Nucleic Acids Res.
31:1962-1968.
4. Held, G.A., G. Grinstein, and Y. Tu. 2006.
Relationship between gene expression and
observed intensities in DNA microarrays—a
modeling study. Nucleic Acids Res. 24:E70.
5. Kuo, W.P., F. Liu, J. Trimarchi, C. Punzo,
M. Lombardi, J. Sarang, M.E. Whipple, M.
Maysuria, et al. 2006. A sequence-oriented
comparison of gene expression measurements
across different hybridization-based technolo-
gies. Nat. Biotechnol. 24:832-840.
6. Bishop, J., S. Blair, and A. Chagovetz. 2006.
A competitive kinetic model of nucleic acid
surface hybridization in the presence of point
mutants. Biophys. J. 90:831-840.
7. Horne, M.T., D.J. Fish, and A.S. Benight.
2006. Statistical thermodynamics and kinet-
ics of DNA multiplex hybridization reactions.
Biophys. J. 91:4133-4153.
8. Fish, D.J., M.T. Horne, R.P. Searles, G.P.
Brewood, and A.S. Benight. 2007. Multiplex
SNP discrimination. Biophys. J. 92:L89-L91.
9. Dai, H., M. Meyer, S. Stepaniants, M.
Ziman, and R. Stoughton. 2002. Use of hy-
bridization kinetics for differentiating specific
from non-specific binding to oligonucleotide
microarrays. Nucleic Acids Res. 30:E86.
10. Bhanot, G., Y. Louzoun, J. Zhu, and C.
DeLisi. 2003. The importance of thermody-
namic equilibrium for high throughput gene
expression arrays. Biophys. J. 84:124-135.
11. Khomyakova, E.B., E.V. Dreval, M. Tran-
Dang, M.-C. Potier, and F.P. Soussaline.
2004. Innovative instrumentation for micro-
array scanning and analysis: application for
characterization of oligonucleotide duplexes
behavior. Cell. Mol. Biol. 50:217-248.
12. Fotin, A.V., A.L. Drobyshev, D.Y.
Proudnikov, A.N. Perov, and A.D.
Mirzabekov. 1998. Parallel thermodynamic
analysis of duplexes on oligodeoxyribonucleo-
tide microchips. Nucleic Acids Res. 15:1515-
1521.
13. Lauffenderberger, D.A. and J.J. Linderman.
1996. Receptors. Oxford University Press,
New York, NY.
14. Sorokin, N.V., V.R. Chechetkin, S.V.
Pan’kov, O.G. Somova, M.A. Livshits,
M.Y. Donnikov, A.Y. Turygin, V.E. Barsky,
and A.S. Zasedatelev. 2006. Kinetics of hy-
bridization on surface oligonucleotide micro-
chips: theory, experiment, and comparison
with hybridization on gel-based microchips. J.
Biomol. Struct. Dyn. 24:57-66.
15. Livshits, M.A. and A.D. Mirzabekov. 1996.
Theoretical analysis of the kinetics of DNA
hybridization with gel-immobilized oligos.
Biophys. J. 71:2795-2801.
16. Wang, S., A.E. Friedman, and E.T. Kool.
1995. Origins of high sequence selectivity:
a stopped-flow kinetics study of DNA/RNA
hybridization by duplex- and triplex-forming
oligos. Biochemistry 34:9774-9784.
17. Ikuta, S., K. Takagi, R.B. Wallace, and K.
Itakura. 1987. Dissociation kinetics of 19
base paired oligo-DNA duplexes contain-
ing different single mismatched base pairs.
Nucleic Acids Res. 15:797-811.
18. Zhang, Y., D.A. Hammer, and D.J. Graves.
2005. Competitive hybridization kinetics re-
veals unexpected behavior patterns. Biophys.
J. 89:2950-2959.
19. Bishop, J., C. Wilson, A.M. Chagovetz, and
S. Blair. 2007. Competitive displacement of
DNA during surface hybridization. Biophys. J.
92:L10-L12.
20. Sartor, M., J. Schwanekamp, D. Halbleib, I.
Mohamed, S. Karyala, M. Medvedovic, and
C.R. Tomlinson. 2004. Microarray results im-
prove significantly as hybridization approaches
equilibrium. BioTechniques 36:790-796.
21. Sekar, M.M.A., W. Bloch, and P.M. St. John.
2005. Comparative study of sequence-depen-
Vol. 44 ı No. 1 ı 2008 www.biotechniques.com ı BioTechniques ı 117
Research Reports
dent hybridization kinetics in solution and on
microspheres. Nucleic Acids Res. 33:366-375.
22. Gao, Y., L.K. Wolf, and R.M. Georgiadis.
2006. Secondary structure effects on DNA
hybridization kinetics: a solution versus sur-
face comparison. Nucleic Acids Res. 34:3370-
3377.
23. Carter, M.G., A.A. Sharov, V. VanBuren,
D.B. Dudekula, C.E. Carmack, C. Nelson,
and M.S. Ko. 2005. Transcript copy number
estimation using a mouse whole-genome oli-
gonucleotide microarray. Genome Biol. 6:
R61.
24. Tsuzuki, K., B. Lambolez, J. Rossier, and
S. Ozawa. 2001. Absolute quantification of
AMPA receptor subunit mRNAs in single hip-
pocampal neurons. J. Neurochem. 77:1650-
1659.
Received 29 June 2007; accepted
18 September 2007.
M.-C.S.’s present address is Max Planck
Institute of Molecular Plant Physiology, Am
Muehlenberg 1, 14476 Golm, Germany.
Address correspondence to Elena
Khomyakova, IMSTAR, 60 rue Notre Dame
des Champs, 75006 Paris, France. e-mail:
elena.khomiakova@imstar.fr; or Marie-
Claude Potier, CNRS UMR 7637, ESPCI 10
rue Vauquelin, 75005 Paris, France. e-mail:
marie-claude.potier@espci.fr
To purchase reprints of this article, contact:
Reprints@BioTechniques.com