Use of Transcriptor Reverse Transcriptase
in Microarray Analysis
Researchers studying tissue and cellular differences in gene expression
have applied microarray analysis to questions ranging from cancer research
to plant development studies. Although the method is well established, many
of the steps in the process continue to be optimised to increase overall
sensitivity and reliability. Here, we monitor gene expression in the obligate
intracellular pathogen Toxoplasma gondii in order to compare reverse
transcriptase target labelling performance on a custom T. gondii 12,000 cDNA
microarray. A histogram of signal intensities shows that the Transcriptor
Reverse Transcriptase produced a major shift in the signal distribution to higher signal intensity
classes compared with the most commonly used RNase H- M-MuLV reverse transcriptase. This
indicates the generation of a labelled target with an overall higher specific activity. The correlation
of Cy5/Cy3 ratio data from four replicate microarrays demonstrate that target labelling with
Transcriptor Reverse Transcriptase is highly reproducible and produces fewer “outlier” signals
compared with the updated version of the RNase H- M-MuLV reverse transcriptase. The known
enzymatic properties of Transcriptor Reverse Transcriptase suggest that it will produce a labelled
target that closely reflects the experimental input RNA.
Introduction
Microarray analysis of gene expression is rapidly
becoming a common tool for molecular biological investigations.
The ability to simultaneously monitor tens of
thousands of genes affords an unprecedented view of
the transcriptional changes that underlie cellular
processes. Microarrays are ideal for genetic investigations
of intracellular pathogens such as Toxoplasma
gondii because it is possible to simultaneously monitor
gene expression in both host and pathogen [1].
However, the system must be optimised for pathogen
gene expression since the T. gondii RNA can be more
than an order of magnitude less concentrated compared
with the host cell RNA. Over the past few years, technological
advances have made the method much more
reliable and robust.
Curiously, the standard protocols for generating the
labelled cDNA target from the experimental RNA,
arguably the most critical step in the procedure, remain
virtually unchanged. The vast majority of investigators
use either a direct incorporation of deoxynucleotides
with covalently attached fluorophores or the incorporation
of amino-allyl deoxynucleotides with subsequent
fluorophore conjugation. In both methods, the efficiency
of the target labelling reaction and quality of the labelled cDNA product depend almost entirely on the reverse
transcription enzyme used.
We employ an amino-allyl labelling reaction to generate
and compare microarray data from different commercially
available reverse transcriptases.
Materials and Methods
Parasite and RNA preparation
Parasites (RH strain) were cultured according to a standard
protocol. NIH 3T3 cells were used as host cells, and
the parasites were harvested by centrifugation just prior
to monolayer lysis. The cells were suspended in TRI
reagent (Sigma). RNA was extracted according to the
manufacturer’s protocol, except that the RNA was not
ethanol precipitated following incubation with
2-propanol, but was added directly to an RNeasy Midi
column (Qiagen). Binding of the sample was followed
directly with the recommended wash and two extra
washes. The RNA was eluted with a total of 450 µl
RNase-free water. This was followed by a standard
ethanol precipitation. The RNA was vacuum dried and
resuspended in RNase-free water to a final concentration
of 1 µg/µl.
Target labelling reaction
Labelling was performed using a modified version of the
published protocol from the DeRisi laboratory
(www.microarrays.org). The modifications included an
increase in the ratio of amino-allyl-dUTP:dTTP to 6:4. The
neutralisation was performed with Tris-HCl. All reverse
transcription reactions were performed at 42°C using
15 µg of total RNA previously extracted as described
above. The total unit amount of Transcriptor Reverse
Transcriptase was 20 units per reaction (10 units added,
and 1 hour later another 10 units were added). The total unit amount of product Y was 400 units per reaction
(400 units added and left for 2 hours according to the
manufacturer’s recommended protocol). The labelled
product was vacuum dried and finally resuspended in
34.5 µl HGMP hybridisation buffer (40% formamide,
5x Denhardt’s solution, 5x SSC, 0.05 M Tris-HCl pH 7.4,
1.0 mM sodium pyrophosphate, 0.1% SDS), 0.4 µg yeast
tRNA, 0.96 µg poly-dA and 2.7% SDS
(www.hgmp.mrc.ac.uk/Research/Microarray/ index.jsp).
Hybridisation
The microarrays used in this study are from a single
printing run, custom made with approximately 12,000
sequenced expressed-sequence tags (ESTs) from a
nonnormalised library. The ESTs were PCR amplified
and the products used for printing were analysed for
content and quality on agarose gels. Every microarray
was visually inspected and samples were prescanned for
printing defects. The microarrays were preblocked (10%
BSA, 3x SSC) for 1 hour at 55°C. Slides were washed with
Milli-Q water (Millipore), and dried by centrifugation.
Fixation of cDNA products was achieved after boiling the
product and submersing it in 95% ethanol. It was prehybridised
(0.001% SDS, 10% BSA, 3x SSC) at 55°C for
25 minutes, washed twice in Milli-Q water and washed in
100% ethanol. Hybridisation was carried out in a humid
Arraylt hybridisation cassette using a lifter slip coverslip
(Erie Scientific) overnight in a 42°C water bath. Post
hybridisation, the slides were washed for 4 minutes in
progressively reduced SDS concentrations (solution 1:
0.5x SSC, 0.1% SDS; solution 2: 0.5 x SSC, 0.01% SDS;
solution 3: 0.06 x SSC). Finally, the slides were dried by
centrifugation.
Spot finding and scanning
The slides were scanned using an Axon 4000A Scanner
according to the manufacturer’s instructions. The images were spot-found using GenePix 4.1 software with a fixed
spot diameter of 150 mm. The median fluorescence value
for the spot was used for final analysis without background
correction or within chip normalisation.
Data analysis
Microsoft Excel was used to create the correlation
curves and histogram data. The Cy5/Cy3 data were also
exported from Microsoft Excel into GeneSpring 6.0 for
final analysis of gene-specific effects.
Results and Discussion
Since a single microarray experiment consists of multiple
steps, the evaluation of any particular step is
assessed from the resulting data. Ideally, the labelled
target should accurately reflect the input mRNA in both
content and representation, have a high specific activity
to maximise sensitivity and be very consistent between
reactions using the same input mRNA. The first of these
properties is inferred from prior enzymology. The second
and third properties are experimentally evaluated using
the scheme outlined in Figure 1. Briefly, a single RNA
source is used throughout and each target labelling represents
an independent reverse-transcriptase and dyecoupling
reaction. Data from each microarray are
derived from two independently labelled targets simultaneously
hybridised to a single microarray.
Although the relative specific activities of a hybridised
labelled target can be estimated by total mean fluorescence
over an entire microarray signal set, the distribution
of signal intensities on a per-spot (probe) basis is
more informative. Figure 2 illustrates a histogram of three
representative sets of 9,737 raw (nonnormalised) fluorescent
signals from the Cy5 channel. Compared with
Transcriptor Reverse Transcriptase there is a much
greater abundance of signals in the 0 – 5,000 class and
relative paucity of signals in the higher fluorescence
classes for product X. This signal distribution indicates
that Transcriptor Reverse Transcriptase labels cDNA
more efficiently than product X. In contrast, Transcriptor
Reverse Transcriptase and product Y show similar signal
distributions where more than half of the signals were in
classes greater than 5,000. Members of the highest signal
class are signals at saturation, some of which are
T. gondii gene transcripts known to be in very high abundance.
The high number of signals in this class for
Transcriptor Reverse Transcriptase indicates a relatively
strong labelling reaction compared with the low signal
numbers for product X. Since the optimal laser power
and PMT levels were similar for all three microarray
scans, image acquisition can not account for the difference
in results.
Consistency of target labelling is critical for microarray
analysis of gene expression to detect real transcriptional
differences between experimental treatments. Since
the same input RNA was used in all experiments, it was
possible to test an enzyme’s ability to generate reproducible
results; microarrays using Transcriptor Reverse
Transcriptase and product Y were replicated four times
each. Ratio (Cy5/Cy3) data from each of the four replicates
plotted against the mean show similar results,
where the Transcriptor Reverse Transcriptase replicates
show generally better correlation to the mean than product
Y (Figure 3). This appears to be due to fewer “outlier”
spots that depart substantially from the trend lines in the
four Transcriptor Reverse Transcriptase replicates.
The replicate data were processed in GeneSpring to
detect particular spots that are consistently biased
towards greater signal in either Transcriptor Reverse
Transcriptase or product Y. Inspection of a few cDNA
probes showed that the greatest signal bias towards
Transcriptor Reverse Transcriptase appear to be in
those probes that are substantially truncated, and
those with a greater signal bias towards product Y are truncated much closer to the 5′ end. Although this
anecdotal observation could be interpreted in several
ways, it is at least consistent with the notion that the
high processivity of Transcriptor Reverse Transcriptase
may produce a better labelled target and more robust
data in cases where the spot/probe cDNA is substantially
truncated.
Conclusion
The histogram of signal intensity distributions show that
Transcriptor Reverse Transcriptase can be used for high
efficient target labelling. Since many investigators eliminate
data with signal intensities below a particular
threshold (e.g.,1,000–2,000), an increased specific activity
of the labelled target may be critical for analysing
genes with low expression.
In general, the data above suggest that Transcriptor
Reverse Transcriptase produces a labelled target with
results similar to those of the updated RNase HM-
MuLV reverse transcriptase. But Transcriptor Reverse
Transcriptase appears to produce a more consistent
labelled target, as the four replicates generally correlate
better, with noticeably fewer “outlier” spots. The known enzymatic properties of Transcriptor Reverse Transcriptase
(high processivity and RNase H activity) should
produce a labelled target that closely reflects the experimental
input mRNA. This will be critical for using the
latest generation of oligonucleotide microarrays, which
include probes designed to the 3′ end of a transcript to
facilitate the analysis of mRNA splice variants, multiple
poly-A+ addition sites, etc.