|
|
Int J Plant Genomics. 2007; 2007: 61028. Published online 2007 April 10. doi: 10.1155/2007/61028. | PMCID: PMC1939914 |
Evaluation of Global RNA Amplification and Its Use for High-Throughput Transcript Analysis of Laser-Microdissected Endosperm Robert C. Day,1 Les McNoe,2 and Richard C. Macknight1* 1Department of Biochemistry, University of Otago, P.O. Box 56, Dunedin 9054, New Zealand 2Genomics Facility, Department of Biochemistry, University of Otago, P.O. Box 56, Dunedin 9054, New Zealand Received July 20, 2006; Revised January 13, 2007; Accepted January 22, 2007. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Laser microdissection (LM) provides a useful method for isolating specific
cells or tissues from biological samples. Here, we adapted microdissection
protocols to allow high-resolution transcript analysis of different tissues
from developing Arabidopsis seed. Sufficient RNA (~50 ng) was extracted from
endosperm tissue for RT-PCR. However, to obtain enough RNA for microarray
analyses, it was necessary to amplify the RNA. PCR- and IVT-based
amplification methods were investigated and several important technical
aspects of amplification were identified (such as target truncation and
alterations in signal intensity). We found that when starting from only 50 ng of RNA, amplification methods based on PCR and IVT produced sufficient
product for reliable microarray hybridizations, with two-round IVT giving
the best results. Microarray analyses, using endosperm-derived RNA amplified
by two-round IVT, reproducibly identified endosperm enriched marker genes.
Thus, when combined with RNA-amplification protocols, LM is a robust and
reliable technique for high-throughput tissue-specific gene expression
analysis. Traditionally, to obtain enough material for global transcript
analysis, whole organs or tissues have been used. However, by
starting from a mixture of cells or tissues, key changes occurring
in just one of the cell types may be obscured. To avoid this,
approaches have been developed that allow high-resolution tissue
sampling or enrichment for specific cell types [ 1].
In some cases, such as pollen grains or epidermal cells, the target cells
are easily accessible for harvest or manual dissection. However,
most cell types are embedded and require tissue digestion or
microdissection to negate surrounding cell layers. Laser
microdissection (LM) is an important method for obtaining
individual tissues, cells, and even organelles for biochemical
analysis. Originally developed for isolating cancerous cells from
normal tissue [ 2], LM is beginning to be used in plant
biology (reviewed by [ 3]). Compared to manual
microdissection, modern laser-based systems are easy to use,
highly reproducible, and can avoid direct contact with the
biological sample [ 4]. LM samples have been used successfully
to obtain DNA, RNA, proteins, and metabolites from a range of
plant species and
tissue types [ 5– 7]. One disadvantage of isolating specific cell types is the
relatively large amount of effort required to isolate a very small
amount of material. In the cases of protein and metabolite
analysis, microdissection is still a very inefficient approach for
global analysis [ 6, 7]. However, RNA can be amplified; a
process that can be carried out using a variety of methods
[ 1]. The most common method of amplification is based on
in vitro transcription (IVT) [ 8]. This linear
amplification method involves producing double-stranded cDNA with
a T7 priming sequence at the 3′ end. This template is then used
by T7 RNA polymerase to generate copies of the cDNA template.
Compared to IVT-based methods, the exponential nature of PCR-based
approaches enables much greater yields per round of amplification
[ 9]. This means that far less starting material can be used
to obtain enough target for high-throughput analysis, with just
one round of amplification. For both IVT- and PCR-based methods,
multiple rounds of amplification can be used to increase the
yields when starting with very small amounts of RNA [ 10].
Hybrid methods have also been developed that use PCR and IVT
amplifications, in an attempt to exploit the advantages of both
[ 11, 12]. RNA amplification from very small amounts of RNA by IVT or PCR has
several drawbacks. Reducing the starting amount of RNA to nanogram
levels often leads to a reduction in the reproducibility of data.
Extreme amplifications can also accentuate subtle technical biases
giving loss of fidelity. However, since biases tend to be
systematic, accurate expression data can be obtained if
appropriate amplified controls are used [ 13]. For example,
when comparing expression changes between samples, it is essential
that all samples are similarly amplified. Arabidopsis is being used as a model plant to study many
aspects of seed biology; however, a major drawback is the very
small size of its seeds. This causes numerous technical
difficulties, especially when studying processes of early seed
development. The endosperm appears to play several important roles
during seed growth and development [ 14]. In many plant
species, including Arabidopsis, the triploid primary
endosperm nucleus undergoes several rounds of free-nuclear
division, growing rapidly as a syncytium. Towards the end of this
proliferative phase, three mitotic endosperm domains are
established. The micropylar endosperm surrounds the
embryo, the peripheral domain lines the wall of the
developing embryo sac, and the chalazal domain develops
adjacent to the vascular connection with the seed
parent. The early proliferation of the endosperm is associated
with final seed size and the alteration of the rate and duration
of cell division in the endosperm has been proposed as a biotech
strategy for altering seed size
[ 15, 16]. Seed with severely
defective endosperm cannot complete development which suggests a
role for endosperm in supporting the formation and growth of the
embryo [ 17]. LM provides an ideal tool for analyzing gene
expression changes in specific cell types during the early stages
of Arabidopsis seed development. The key to the success
of such studies is the ability to isolate and amplify RNA from a
small amount of laser-dissected tissue. Here, we present protocols
for the isolation of RNA from different seed compartments of
developing Arabidopsis seed and highlight important
technical issues encountered when amplifying from small amounts of
total RNA for array analysis. The protocols developed in this
study were used to identify a list of approximately 2 700 genes
that are expressed in proliferating endosperm 4 DAP. 2.1. Plant growth Arabidopsis thaliana plants were grown under lights in a
growth room with 16-hour photoperiod and at 20°C. To
produce seed of a known developmental stage, flowers were
emasculated with fine forceps and manual pollinations were carried
out two days after. 2.2. Harvest of siliques and RNA extraction
for amplification
Multiple siliques corresponding to early and late stages of seed
development were harvested into Eppendorf tubes and snap frozen in
liquid nitrogen. Siliques from the early sample were added to the
late sample at a ratio of 1 to 10 (based on fresh weight) to
produce a 90% late sample. Early and 90% late tissues were
ground to a fine powder in liquid nitrogen using a pestle and
mortar. Total RNA was isolated from the samples using Concert Plant RNA
Purification Reagent (Invitrogen, Carlsbad, Calif, USA)
following the manufacturers' large-scale extraction protocol.
Aliquots of these samples were further purified using Qiagen
RNeasy columns and the manufacturers' optional on column DNase
step was carried out.
2.3. Harvest of siliques for LCM tissue preparation Siliques were harvested 4 days after pollination into precooled
(4°C) farmers' fixative (3 : 1 mixture of
ethanol : acetic acid). To enhance penetration of
the fixative, the ends of the siliques were removed with a scalpel
before the tissues were submerged. Samples were kept for 3-4 hours
on an orbital platform at 4°C. The fixative was replaced
with 95% ethanol for 30 minutes. The siliques were
transferred into sample cassettes that were then submerged in
absolute ethanol and were microwaved at 60°C for
10 minutes (Electron Microscopy Sciences EMS-820). Microwave
treatment was repeated in two changes of isopropylalcohol at
70°C for 10 minutes and a third change at
74°C for 10 minutes. Cassettes were then transferred
to molten wax and placed in a heated vacuum chamber. A vacuum was
applied for 20 minutes then released. The vacuum infiltration of
wax was repeated twice more using fresh wax. Siliques were
released into molten wax and multiple siliques were aligned in
single wax blocks and stored at 4°C with desiccant for up
to two weeks before being sectioned using a rotary microtome
(Leica Microsystems, Wetzelar, Germany) to produce
7 μm sections. The sections were floated on distilled water and then 30% ethanol
prior to attachment to superfrost plus slides (BDH/Merck,
Damstadt, Germany). Slides were allowed to dry for 2 hours before
storage with desiccant at 4°C. 2.4. LCM
Slides were dehydrated in xylene for 5 minutes and allowed to
air dry in a flow hood for 5–10 minutes prior to LCM using
the Arcturus LM200 Pixcell II system. This system enables specific
regions of the tissue sample to be targeted microscopically
through a plastic cap coated with a special thermoplastic film.
The laser was then used to melt the film onto the targeted areas
using a spot size of 7.5 μm, power setting of 100 mW,
and pulse durations of 1-2 milliseconds, depending on the ease of
capture from each specific seed. Once cemented to the film,
targeted tissues were removed from the rest of the section by
lifting the cap vertically. Captures were collected onto Arcturus
macrocaps. If nonspecific pickup was apparent a post-it note was
used to remove the contaminating cells (3M, St Paul, MN, USA). The remaining
tissue was sometimes scraped from the glass slide using a sterile
flat-ended spatula, and was subsequently processed using
equivalent protocols to the LCM material. 2.5. RNA extraction from tissues prepared for LCM
Total RNA was obtained from embryo, seed coat, syncytial
endosperm, and the tissue scrape samples using the Picopure
RNA isolation kit (Arcturus, Oxnard, CA, USA) with
optional on-column DNase step (Qiagen, Valencia, CA,
USA). The purified total RNA was quantified using the Ribogreen
RNA quantification kit (Invitrogen) according to the
manufacturers' instructions. 2.6. Conventional target labelling
Aliquots of 30 μg total RNA were reverse transcribed and
labelled using the Superscript Indirect labelling kit (Invitrogen). 2.7. Global transcript amplification by PCR
Aliquots of 50 ng total RNA were subjected to reverse
transcription using superscript III reverse transcriptase
(Invitrogen) and amplified dsDNA was prepared following the
methods described by Petalidis et al. [ 18], with the exception
that the entire cDNA sample was used in the amplification
reaction. A test PCR is essential to avoid overcycling of the
amplification reaction; however, this effectively doubles the
amount of total RNA required to carry out RNA amplification. To
determine the optimal number of cycles required for generating
products during the exponential phase of the PCR reaction, PCR
products from parallel runs were visualised after 15, 18, 21, and
24 cycles on a 1.2% agarose gel. When the yield of PCR product
stops increasing with more cycles, the reaction has reached its
plateau. The optimal number of cycles for the amplification is one
cycle fewer than is needed to reach the plateau; in our case 20
cycles. Amino-allyl dNTPs were incorporated by Klenow and the free
Cy dyes (Amersham Biosciences, Little Chalfont, UK) were
coupled to the target dsDNA as described in the Superscript
Indirect labelling kit manual (Invitrogen). 2.8. Global transcript amplification by IVT IVT-based amplifications were carried out using the Message Amp II
amino-allyl aRNA kit (Ambion Inc., Austin, TX, USA)
following the manufacturers' instructions. Where appropriate,
amino-allyl UTPs were incorporated into the IVT reaction and were
coupled to Cy dyes (Amersham Biosciences) following the kits
recommendations. All labelled aRNA targets were fragmented prior
to hybridization using Ambion fragmentation reagent following the
manufacturers' protocol (Ambion). 2.9. RT-PCR and qRT-PCR The RNA or aRNA was primed with random hexamers and
first-strand cDNA was synthesized using Superscript III
(Invitrogen) according to the manufacturers'
instructions. The amplification conditions for conventional PCR were as
follows: 2 minutes at 94°C; 25–35 cycles of 30
seconds at 94°C, 30 seconds at 58, 30 seconds at
72°C; and 68°C for 2 minutes. The primers
for FWA, MEA, and AP2 are described by
Kinoshita et al.
[ 19, 20].
Real-time qRT-PCR was carried out using reagents from the
LightCycler FastStart DNA Masterplus SYBR Green I kit
(Roche, Penzberg, Germany) in 20 μL volumes using a LightCycler 2.0
(Roche). The amplification conditions for qPCR were Denat:
95°C for 10 minutes; Cycling 94°C for 5
seconds, 58°C for 17 seconds, 72°C for 10
seconds (single acquire); Melt: 95°C for 0 second,
55°C for 20 seconds, 95°C for 0 second with ramp
0.2°C/s (continuous acquire); Cool: 40°C for 20
seconds. Quantification was carried out using Actin-2 as a
reference gene (At3g18780). Actin-2 was selected since it was a
control spot on our microarrays. The differential expression
observed across the 17 spots per slide representing Actin-2 in our
normalized data was 1.004 indicating equivalent expression in both
types of biological sample used to query the arrays and used for
qRT-PCR. Reaction products were confirmed by melting curve
analysis and by running out the product on a 1.2% agarose gel.
Primer sets for qRT-PCR validation experiments were designed to
span at least one intron so we could easily observe PCR products
generated from genomic DNA contamination. For each primer set, a
no-template water control was also used.
The primers used for qRT-PCR of FWA and MEA are
described by Kinoshita et al.
[ 19, 20]
and the FIS2
primers were FIS2-RT-R 5′-agatctcctggcgactaac-3′, and
FIS2-RT-F 5′-tattagcgaacgcctgagac-3′. Actin primers were Q-RT
Actin_F 5′-cgctctttctttccaagctcat-3′ and Q-RT Actin_R
5′-tcctgcaaatccagccttc-3′. TT8 primers were TT8-QRT-F
5′-ctgatcttcatattgaatcaaccca-3′ and TT8-QRT-R
5′-gtgtgacatgagaagtgttgttac-3′. FUS3 primers were FUS3-QRT-F
5′-tcatggtctgcagctaggtga-3′ and FUS3-QRT-R
5′-tacttcttcttcttccgatgcttt-3′. At5g16780 primers were
CAS5G-QRT-F 5′-gtgggtttgcgactgttg-3′ and CAS5G-QRT-R
5′-gagtttccgggctctgatt-3′. At1g48630 primers were CAS1G-QRT-F
5′-aagtctgttgttgaggatttgaag-3 and CAS1G-QRT-R
5′-ttccatctgcactccagtt-3′.
2.10. Array hybridization Prewash and BSA prehybridization: slides were prewashed
in 0.2% SDS for 2 minutes and then washed twice in MilliQ
(MQ) water. The slides were then placed in a 50 mL Falcon tube
containing fresh MQ water and were incubated in a water bath for
20 minutes at 50°C. Slides were then prehybridized in
buffer containing 4x SSC, 0.5% SDS and 5% w/v bovine
serum albumin (BSA, Sigma) at 42°C for 45 minutes.
Slides were then washed 5 times with MQ water at room temperature
to remove BSA residue. Prior to hybridization, the slides were
dried by centrifuged at 2,500 xg. Hybridization: hybridizations were carried out under
LifterSlips (Erie scientific, Portsmouth, NH, USA) placed over the region of
the slide containing the array. A mix of SlideHyb buffer 1
(Ambion) and the labelled probes were then injected under the
LifterSlip. The arrays were then placed in a humidified
hybridization chamber (Corning, New York, NY, USA) and were hybridized over
night in a 55°C water bath. Posthybridization: the slides were removed from the
hybridization chamber and were washed at 55°C for 8
minutes with agitation in 50 mL Falcon tubes containing 2x
SSC, 0.1% SDS. The slides were then moved to new 50 mL Falcon
tubes and were washed with constant agitation in 1x SSC, 0.1%
SDS at room temperature for further 6 minutes. This procedure
was repeated with a solution containing 0.5x SSC and then 0.1x SSC
before the slides were dried by centrifugation at 5800 rpm. Scanning: slide scanning was carried out using an Axon
4000 B dual-color scanner running GenePix 4.1 software (Axon
instruments/Molecular Devices Corporation, Sunnydale,CA, USA).
2.11. Data normalization and analysis Genepix files were normalized by implementing elements of the
SNOMAD gene analysis tools
( http://pevsnerlab.kennedykrieger.org/snomadinput.html) at a
local level using an SPLUS script. This involved global mean
normalization, local mean normalization across the array surface,
and local mean normalization across the element signal intensity.
Both high and low power scans were obtained for each
hybridization. Data from high and low scans were normalized
independently. When calculating correlations, the number of
differentially expressed transcripts and the direction of change,
the saturated spot data was replaced with data from the low
scan. Any spots that had intensity values less than 3x
the mean background in both channels or were only represented in a
single replicate were discarded from this analysis. Genes that had
a signal above the trim level in only one channel remained in the
analysis but had a value of 100 intensity units interpolated for
the low signal to reduce the effect of very low intensity features
on expression ratios. Interpolation introduced large amounts of
variation for transcripts that gave very low signal. Data was
normalized without interpolation for use with genes that had very
low signal, for example, FIS2. When interpreting the data
with regard to movement of intensity values within the population
or loss of signal based on probe design, we only considered the
high intensity scan and the data was not trimmed based on
intensity. 3. RESULTS AND DISCUSSION 3.1. Development of tissue preparation protocols
for laser capture microdissection of Arabidopsis seed Laser capture microdissection (LCM) is a form of LM that relies on
the adhesion of only the targeted cells to a thermoplastic film
attached to a plastic cap [ 2]. We developed tissue
preparation protocols for LCM of immature Arabidopsis
seed using Farmer's fixative based on recommendations of Kerk
et al. [ 5]. Other groups have successfully used acetone or
microwave fixation in the absence of chemical fixative to carry
out transcript analysis on plant material
[ 21, 22]. However,
our preliminary studies indicated that Farmer's fixative gives the
most consistent penetration of the fixative into
Arabidopsis silique tissues and enables good histology
and yields of total RNA. Using Farmer's fixative, our tissue
fixation period (3-4 hours) was shorter than protocols using
acetone (overnight). We also used the heat generated by a
microwave to aid the diffusion of subsequent solvent washes and
the infusion of wax under vacuum, further reducing the processing
time. This enabled silique material to be processed from harvest
to wax block within 9 hours. shows examples
of LCM isolations of seed coat, endosperm, and embryo tissue from
developing siliques. LCM removed the majority of the target tissue
and only the desired cells were visible on the LCM cap
(). Fixed samples were aligned in the wax so
that approximately six siliques were within the footprint of an
LCM cap. This maximized the number of seeds that could be targeted
onto each cap. Three serial sections were mounted onto a single
glass slide, enabling the caps to be rapidly repositioned over
each serial section in turn if target tissue was limiting (as was
the case for embryo samples). 3.2. RNA extracted from microdissected seed is suitable for transcript analysis As well as allowing sufficient histological resolution, the
tissue-processing regime must be conducive to subsequent
extraction of RNA of sufficient quality and in amounts suitable
for meaningful analysis. We used the Arcturus Picopure RNA
isolation kit with an on-column DNase treatment to isolate total
RNA. In an alteration to the kit protocol, we routinely peeled the
thermoplastic film from LCM caps using fine forceps and submerged
these in extraction buffer. This allowed multiple caps to be
extracted in the same tube. By combining two peels containing
isolated material from multiple seeds, we were able to obtain
20–60 ng of total RNA per extraction, depending on the seed
compartment targeted, for example, approximately 50
endosperm fragments were obtained per peel enabling 100 fragments
to be used for RNA extraction (). The RNA
yields were quantified using the Ribogreen RNA quantification kit
(Invitrogen).
As with most studies of this kind, we assessed the quality of the
extracted RNA by generating cDNA templates from unamplified
samples and conducting RT-PCR
[ 5, 21, 23, 24]. We also used
small aliquots of aRNA produced from one round of IVT
amplification for similar analysis (10–15% of total product).
The cDNA yield from the aRNA increased our effective yield of
experimental material and would be sufficient to carry out several
hundred tissue-specific PCR reactions. Our lab is interested in identifying genes that are important
during early endosperm development. To identify transcripts
expressed specifically in the endosperm, we needed to ensure that
the other seed compartment samples had minimal endosperm
contamination and our endosperm sample was suitably enriched. To
investigate this, we carried out RT-PCR on LCM-derived samples. As
shown in , AP2 was amplified from all
samples using conventional RT-PCR, consistent with its reported
expression throughout the early seed [ 19], and as expected
MEA expression was relatively strong in the endosperm and
weak in the embryo (globular to mid-heart) [ 25]. We used aRNA
from a single round of IVT amplification to assess the enrichment
of endosperm expressed transcripts by real-time qPCR. We
calculated the levels of expression (relative to actin) of two
endosperm-specific genes, FWA and FIS2, and
MEA from both LCM-derived endosperm cDNA (4 DAP) and a
similarly amplified tissue scrape cDNA. The scrape sample was
derived from the tissues remaining on the slide after
microdissection of the endosperm. Not all seed tissues enabled
unambiguous identification and unimpeded access to endosperm
tissue, thus the tissue scrape retained a proportion of endosperm
material. Results indicated a strong enrichment for endosperm
(), such that the endosperm-specific genes
FIS2 and FWA reported 9.5-fold and 8-fold
enrichment, respectively. MEA reported a 3.5-fold
enrichment that reflects its simultaneous expression in the embryo
[ 25, 26]. Taken together, our PCR analysis indicated that the RNA obtained
using LCM generated expression data that was consistent with the
expression of known marker genes and that the cell types were well
separated. 3.3. Experimental design for the comparison of amplification methods using oligonucleotide
microarray Our aim was to identify which standalone amplification protocol
performed best in our lab when the starting amount of total RNA
was limited to 50 ng. While other studies have compared
amplification protocols, few start with such low amounts of total
RNA [ 10, 13]. In addition, most of these studies use
ultra-high-quality RNA bought commercially or isolated from cell
lines [ 27– 33].
Since the quality of extracted
RNA is very often tissue-dependent and will affect the
amplification efficiency, we decided to use total RNA extracted
from whole siliques; better representing tissues targeted by LCM.
We set out to compare two basic methods of RNA amplification,
based on IVT and PCR. Preliminary amplifications from LCM prepared
samples indicated that two rounds of IVT or a single round of PCR
were required to generate enough aRNA or cDNA, respectively, for
hybridization to long oligonucleotide arrays (data not shown).
RNA was isolated from siliques during early (2–5 DAP) and late
(6–10 DAP) stages of development. To help identifying major
alterations in the array data due to our target preparation
protocols, the late stage RNA sample was spiked with 10% early
stage sample (90% late). In the absence of compression, the
signals from the genes only expressed in the early sample should
not be more than 10-fold higher than those obtained from the late
sample. Conversely, the late sample should display a full range of
differential expression. This design should result in the
unamplified MA plots (plots of differential expression against
intensity) having a distinctive asymmetrical shape.
Our study compared unamplified samples and three RNA amplification
regimes; one-round IVT, two-round IVT, and a PCR-based method. The
unamplified treatments used 30 μg of total RNA followed by
conventional labelling and hybridization. The one-round IVT
amplifications were made from 5 μg of total RNA (a
standard labelling procedure in many microarray experiments). The
two-round IVT and PCR regimes started from 50 ng of total RNA
and would therefore be appropriate for use with LCM samples. The
unamplified treatment involved duplicate hybridizations of the
90%-late sample against our early sample. For each of the
amplification regimes, a single amplification of the 90%-late
sample was hybridized against two independent amplifications of
our early sample. In all hybridizations the early and 90%-late
samples were consistently labelled with Cy5 and Cy3, respectively.
This consistent labelling strategy for all 8 slides reduced
variation between the microarrays due to dye bias and enabled us
to focus on the variation arising from duplicate amplifications of
a single early silique sample. 3.4. Reproducible microarray results can be obtained
using amplified RNA The RNA yields after amplification are given in
Table 1. We aimed to generate 10–20 μg of aRNA
using IVT and 1–5 μg of cDNA using PCR; such that we
could comfortably perform duplicate hybridizations from each
sample. When starting with 50 ng of RNA, the desired amount of
aRNA/cDNA was obtained after either two rounds of IVT
(103 μg aRNA) or a single round of PCR amplification
(10 μg), respectively ( Table 1). IVT-based
amplification on our LM samples generated aRNA with a similar
efficiency to that obtained from a conventional tissue sample
( Table 1). | Table 1Amount of starting RNA and subsequent yields after
amplification. |
The reproducibility and fidelity of the amplification methods were
assessed by microarray analysis using long oligonucleotide arrays
based on the Operon V1.0 Arabidopsis set. The emergence
of extremely high signal for some array elements post
amplification (hotspots) has been described in the literature
[ 34] and was countered in this study by carrying out both
high and low power scans for each hybridization. Data from high
and low scans were normalized independently and the saturated spot
data replaced with data from the low scan. This enabled the
highest possible number of data points to be included in our
analysis and negated any issues due to alteration of signal spread
due to different amplification regimes. In instances where data
from the combined scan set might introduce artefacts into our
analysis, we considered only data from the high power scans.
To assess the reproducibility of our methods, we plotted the
duplicate data obtained for each target preparation regime
((a)). The between-replicate reproducibility of
the unamplified samples was high with a Pearson correlation of
r = .95. The single-round IVT amplification gave an even higher
between-replicate correlation ( r = .98), a phenomenon that has
been reported by others [ 35, 36]. The extreme amplifications
also showed high between-replicate correlations of r = .8 and
r = .9 for PCR and 2IVT, respectively.
Puskas et al. [ 37] also
found a higher correlation with IVT than with PCR, when the two
amplifications were compared directly. Although our PCR and 2IVT
correlations were somewhat lower than the more conventional target
preparation strategies, they are consistent with other studies
using RNA samples diluted down to tens of nanograms ( r = .87–.94;
[ 38, 39]). In summary, amplification of 50 ng of RNA by either two rounds
of IVT or PCR produced sufficient product for reproducible microarray hybridizations. 3.5. Amplification affects the profile of the RNA population To determine if the different amplification methods fundamentally
altered the global representation of the transcript populations,
we first considered variations in the MA plots. These plots graph
intensity on the x-axis versus Log2 of the differential
expression on the y-axis. Our experimental design generated the
expected asymmetrical form when the unamplified
data were plotted in this way (). All
amplification regimes deviated from this pattern and appeared to
compress the differential signals reported in the unamplified
dataset. The magnitude of compression was (from lowest to highest)
IVT, 2IVT, and PCR. The compression comprised a more reduced
differential in the direction of the 90%-late sample than that
observed in the early sample. This suggests that highly
differential values were more affected than moderate fold changes.
This was also evident in the correlation plots drawn between the
amplified data and the unamplified data, where the most
differential signals fell away from the r = 1 line
((b)). The amplification regimes also reported
a more consistent spread of intensity values
(). To assess the affects of amplification on relative
transcript abundance, we calculated the total intensity (sum of
red and green signals) for each gene within each
treatment. These values were placed in rank order and were split
into five groups based on intensity. The transcripts contained
within each intensity bin were then tracked back to see which
intensity groups the transcript originated from in the unamplified
samples. Amplification caused alterations in the relative
intensity of a subset of genes; an effect that was
most pronounced with extreme IVT- or PCR-based
amplifications (). PCR was the worst affected,
with many transcripts moving into widely different intensity
groups.
Our amplified data gave good positive correlations
to the unamplified samples ( r = .62–.83) and independent studies
describe similar values [ 40, 41]. Many studies that assess
fidelity of amplification have been carried out using Affymetrix
Genechip technology [ 10, 13]. These studies tend to give
higher correlations between amplified samples, than those obtained
here using academic printed slides, a situation that is perhaps
partly attributable to issues of manufacture. However, Genechip
technology can exclude up to 50% of the genes detected by long
oligonucleotide-based arrays, based on their low signal intensity.
This effectively excludes transcripts with low correlations from
the analysis [ 42]. Furthermore, in the majority of Genechip
studies, the conventional target preparation method used is a mild
form of IVT based-amplification. If we compare our two-round IVT
amplification to our IVT labelled sample our correlation increases
from a Pearson correlation of r = .78 to .9. Conversely, our PCR
amplified dataset had a lower correlation to the IVT labelled
sample than it did to the unamplified dataset. This indicates that
the different amplification strategies can introduce different
distinct biases. Other aspects of target preparation can also
introduce sources of variation that can affect the correlation
between datasets. Reverse transcriptases can vary in their ability
to yield useful amounts of cDNA from limiting amounts of RNA
[ 43] and in their ability to generate cDNA that largely
maintains the length of the original transcript, an important
issue when using oligonucleotide-based microarrays (see later).
Comparison of DNA hybridizations (unamplified and PCR), with
fragmented aRNA hybridizations (IVT), may also introduce variation
due to different hybridization kinetics [ 44].
Despite obvious alterations in the observed signal from our
hybridizations after amplification, we wanted to assess whether
the direction of differential expression was consistent. The genes
that reported a two-fold differential expression in the
unamplified treatments were tracked in the amplified data. It
became apparent that while many genes reported an altered level of
differential expression, the majority of these genes gave a
consistent direction of change, irrespective of the method used or
extreme nature of the amplifications ().
Considerable variation in fold change has been observed in
microarray data sets, compared to real-time PCR
[ 45– 47].
Often microarray data changes are compressed and this phenomenon
is more apparent in amplified samples [ 48]. Ignoring or
relaxing the fold change cutoff and using statistical methods can
to a large extent eliminate problems due to compression. Jenson
et al. [ 48] found that by dropping the threshold from 2 to
1.5, they got an increased number of differentially expressed
genes common to both cDNA (unamplified) and aRNA (amplified)
generated microarray data sets.
Overall, our results suggest that while the different
amplification methods report variable differential expression for
the same gene, they are reasonably reliable at indicating if a
gene is up- or down-regulated. Thus, in instances where
compression is likely, such as with extreme amplifications, it
might be advisable to base interpretation of data mainly on the
direction of change. 3.6. Different amplification methods identify different differentially expressed genes Our experiments were not designed to give precise
data about the differential expression of specific genes.
Nevertheless, large fluctuations in the numbers of transcripts
reporting two-fold differential expression were apparent, and
worthy of comment. The different amplification regimes reported a
sizable percentage (6%–12%) of unique differentials
( Table 2(a)). Furthermore, the mild IVT amplification
reported more genes with a differential expression
(>2-fold) than all other treatments, including the
unamplified data ( Table 2(b)). When we plotted the fold
change of the unique IVT and unamplified differentials against
their total intensities, the data were similarly distributed to
the right of the 3x background intensity cutoff
(). However, when we plotted the unique IVT
differentials against the intensity values from the unamplified
data, it became apparent that the majority of the unique IVT
differentials represented transcripts of low intensity in the
unamplified dataset (). This indicated that
in some instances, amplification can increase the absolute signal
intensity allowing more differentials to be identified. The
emergence of such data from low intensity background probably
represents a systematic bias in the amplification of specific
transcripts. Nevertheless, several studies have verified that
differentials specifically identified in amplified samples can be
validated by other methods, such as real-time PCR
[ 31, 49, 50]. | Table 2(a) The number (and %) of unique 2-fold differentials
identified with each target preparation regime. (b) The number of
transcripts identified as being differentially expressed (based on
1.5- and 2-fold cutoffs) using the mean Log2 ratio from
duplicate (more ...) |
Taken together, these data suggest that even if RNA yields were
conducive to conventional cDNA hybridizations, the optimum
strategy for compiling a list of differentially expressed genes
would be several hybridizations using a combination of different
target preparation regimes. 3.7. Extreme amplification can truncate representative
aRNA and cause subsequent reduction in signal on long oligonucleotide arrays The second round of IVT uses a random primer for
first-strand cDNA synthesis, a process that may lead to truncation
of the representative target molecules [ 51]. To assess the
effect of the second-round IVT on our data, each treatment dataset
was ranked based on total intensity and the ranks for each probe
were compared to rankings in the unamplified data.The relative ranking was
calculated such that positive or negative changes in rank
indicated an increase or decrease in relative signal within the
target population, respectively. This change in apparent signal
was then plotted versus the distance of the corresponding target
probe sequence from the 3′ end of the annotated target
transcript (). Our microarrays were printed
using the Operon version 1.0 Arabidopsis oligonucleotide
set. Whilst these oligonucleotides were designed to the 3′
region of genes, their position varies with regard to distance
from the polyA tail. The plot shown in
indicated that more genes were being lost from our dataset if
extreme amplifications were used for target preparation, and that
this correlated with the design of some of the microarray oligos
being more 5′ biased than others. Although not as severe as the
target truncation introduced during the random priming of the
second round of IVT, PCR amplification also has detrimental
affects on target length (). The template
switching method utilized by the PCR regime integrates primer
sites into both the 5′ and the 3′ ends of the representative
cDNA templates that should negate issues of 3′ bias. However,
other studies have also found that PCR can generate products
shorter than those obtained from one round of IVT
[ 52]. This
is presumably due to differences in the abilities of different
reverse transcriptases to generate full-length cDNA.
The latest version of the Operon Arabidopsis
oligonucleotide set should be less sensitive to target truncation,
as it is more 3′ optimized. In addition, not all array types
will be as affected by target truncation. Arrays based on printed
full-length cDNAs, tiling arrays, and conventional Affymetrix
Genechips will have better 3′ representation than our arrays.
The long oligonucleotide arrays used in this study generally had a
single probe representing each transcript. This is in contrast to
Affymetrix Genechips, which have several probes per transcript
that are 3′ optimized. In instances where more than one round of
IVT are used, this can be advantageous. Casson et al.
[ 53] generated target populations from LCM harvested embryo cells
prepared for Genechip analysis by three rounds of IVT-based
amplification. They also found a reduction in 3′ representation
for some transcripts, but by considering fewer representative
probes in their analysis they were able to identify a larger
number of embryo-expressed genes [ 53]. 3.8. Tissue-specific microarray analysis of endosperm
We have shown here that the two-round IVT-based amplification
provided good between replicate correlations and that it was
the extreme amplification method that best represented our
unamplified data. However, since this data was obtained using 50 ng of
conventionally purified RNA rather than RNA obtained from a
microdissected sample, we investigated whether microarray data
produced u sing LCM-derived RNA also gives highly reproducible
results. We used a two-round IVT-based amplification to obtain
expression profiles from three replicate endosperm samples
microdissected from sections of 4 DAP siliques and from the
tissues remaining post-microdissection scraped from one of the
sample slides. The tissue scrape aRNA was used as a universal
control for all endosperm hybridizations. This approach was used
to minimize alterations in the measured differential expression
due to differences in RNA quality [ 54]. 3.9. Reproducibility of array data obtained using LCM endosperm The silique sample (tissue scrape) included a
proportion of endosperm such that the MA plots for hybridizations
between silique tissue scrapes and LM endosperms described
an asymmetrical form similar to our previous amplification
experiments (). Pearson correlations between
our replicates were high (0.88–0.93) indicating that the
microarray data produced using our LM-derived RNA generated highly
reproducible results. The lowest correlations were obtained from
comparisons between biological replicates that incorporated
dye-swaps (Figures and ). 18 220 unique
probes gave signal higher than two-fold background and were
represented in at least three of the four hybridizations. P
values were calculated to assess the variation of the signal
across the four endosperm hybridizations. Twelve thousand seven
hundred and ten (12 710) probes gave consistent signal with P
values of <.05. Of this list, approximately 5 000 probes
allocated to individual loci were differentially expressed in the
endosperm sample using a 1.5-fold cutoff or approximately 2700
probes if a 2-fold cutoff was used (listed in the Supplementary
Table 1 available at doi:10.1155/2007/61028). 3.10. The LM endosperm array data does not appear to be
contaminated with seed coat- or embryo-specific genes During LCM of endosperm from seed at 4 DAP, many individual
microdissections were performed. Tissues such as the inner
endothelium of the developing seed coat and the developing embryo
are immediately adjacent to the endosperm and are the most likely
contaminants of our endosperm samples. Few genes have been
characterized as being preferentially expressed in the inner
endothelium of the seed coat at 4 DAP. However, two such genes TT8
and BAN
[ 55, 56] gave consistent signal in our array data and
were used to assess seed coat contamination (TT8 was also
represented on our slides by two independent probes). Both these
seed coat markers were excluded from our endosperm expressed list
even using a 1.5-fold cutoff, indicating that there is little seed
coat contamination of our endosperm LM samples.
To screen our data for embryo contamination, we compiled a list of
sixteen embryo-expressed genes from the literature. In an LCM
study that targeted Arabidopsis embryos at early stages
of development, Casson et al. [ 53] validated their embryo
Genechip experiments using a list of embryo-expressed
genes. This study also generated
several reporter-GUS lines that provided evidence for
the expression of several additional genes in early embryos.
Thirteen of the sixteen embryonic markers were excluded from our
endosperm-expressed list using a 2-fold cutoff. Of the three
embryonic markers present in the endosperm expressed list,
LEC1 has been reported to be expressed in both the embryo
and the endosperm at heart stage by in situ hybridization
[ 53, 57] and there is evidence from an At1g48630 promoter-GUS
reporter lines that this gene is also expressed in both embryo and
endosperm [ 53, 57].
The expression of the third embryonic marker, FUS3, in
the endosperm at 4 DAP is inconsistent with the genes
characterization in the current literature. Both FUS3
promoter-GUS [ 58] and FUS3-GFP translational fusions
[ 59] give strong and highly specific activity in globular
embryos. Expression of FUS3 in the embryo at heart stage
(approximately 4 DAP) then becomes more restricted to the outer
protodermal tissues. Interestingly during the later stages of
development, FUS3 expression is reported to be strong in
the endosperm tissues that form the aleurone layer. It is
possible, therefore, that at 4 DAP, FUS3 is starting to
make the transition from embryonic-specific expression to a more
widespread expression pattern. Furthermore, the strong signal from
the embryo may obscure a lower more diffuse activity in the
endosperm. Images presented for FUS3 promoter-GUS fusions
by Tsuchiya et al. [ 58] appear to be consistent with this
hypothesis since the endosperm visible in the heart stage sample
(limited to the embryo surrounding region) appears to show GUS
activity. Thus, our endosperm-expressed list does not appear to be
contaminated with any genes that are unequivocally expressed
specifically in the embryo. To further validate our data, the expression of the two embryonic
marker genes not described in the literature as endosperm
expressed ( FUS3 and At4g48630) were validated by
real-time PCR. Real-time PCR of the embryonic marker At5g16780 and
seed coat marker TT8 were also used to clarify the expression of
marker genes close to the 1.5-fold cutoff ( Table 3).
For the embryonic markers (FUS3, At1g48630, and At5g16780)
real-time PCR gave largely equivalent values to those obtained
from the array experiments. For TT8, the differential expression
observed by real-time PCR was slightly higher in the endosperm
direction than that observed in our array data and took the
transcript over the 1.5-fold threshold to a value of 1.66. Since
both TT8 and At5g16780 marker genes gave some evidence of being
over 1.5-fold differentially expressed in the endosperm, it was
thought prudent to limit the endosperm-expressed list to genes
above a 2-fold cutoff ( Table 3). | Table 3Distribution of imprinted genes and endosperm, embryonic,
and seed coat markers in the LM endosperm array data. |
3.11. Known endosperm-expressed genes are present in the LM endosperm data To confirm that our approach could detect differential expression
of known endosperm-expressed genes, we identified a number of
genes from the literature that have been described as
endosperm-preferred or endosperm-specific. Tiwari et al.
[ 16] recently
identified a number of candidate genes that gave strong evidence
for endosperm-preferred expression. A selection of these was used
for validation by promoter-GUS fusions [ 16]. Four of the
genes identified as having upstream regions that can drive GUS
expression during the early stages of endosperm development gave
consistent signal in our LM data (At5g46950, At5g07210, At5g39260,
At2g41000). However, only At5g07210 (ARR21), a two-component
response regulator, was identified as having endosperm-preferred
expression from our 4 DAP data. Interestingly, expression of both At5g46950 and At5g39260 in the
AtGenExpress Genechip developmental series [ 60] is only
evident in siliques at 3 DAP, when the seed contains globular
stage embryos, and not in siliques at 4 DAP (heart stage). This
suggests that the LM endosperm data is extremely accurate with
regard to developmental stage. Further evidence for this is
provided in a study that examined the expression of
isopentenyltransferases (IPT) genes by promoter-GUS reporter
constructs. IPT genes catalyze the rate limiting step of
cytokinin biosynthesis in Arabidospsis, and Miyawaki
et al. [ 61] found that IPT4 and IPT8 were
expressed predominantly in the chalazal endosperm of developing
seed. IPT8 had a slightly more extensive expression
pattern in the endosperm and persisted until late heart stage,
whereas IPT4 was very specific to the chalazal cyst and
the activity in IPT4-GUS transformants disappeared prior to heart
stage. The inclusion of IPT8 in our endosperm-preferred
list (and the exclusion of IPT4) is consistent with
precise sampling of the endosperm at 4 DAP when the seeds are at
the early to mid heart stages of development. At2g41000 encodes a DNAJ heatshock protein that describes a range
of different expression patterns in the literature. The At2g41000
promoter-GUS fusion described by Tiwari et al. [ 16] was able
to drive expression in the endosperm and the embryo sac but was
not expressed in pollen. Clarification by semiquantitative RT-PCR
was attempted to compare the GUS expression pattern with the
endogenous transcript but the authors were unable to detect
At2g41000 transcript in any tissue. Expression of this gene in
online datasets is even more ambiguous with no signal apparent in
dissected ovules and 1 DAP seed but expression in flower buds,
roots, and leaves in the Goldberg Lab Arabidopsis
thaliana Genechip Database
( http://estdb.biology.ucla.edu/genechip) and only pollen
expression being evident in the AtGenExpress Genechip
developmental series [ 60]. A study into abscisic acid (ABA) synthesis has also identified an
endosperm-specific 9-cis-epoxycarotenoid ( NCED6) that
cleaves cis-epoxycarotenoids in the first specific step of ABA
biosynthesis. NCED6 was found to be specific to the
endosperm by promoter-GUS, promoter-GFP, and in situ
hybridization [ 63] and is confirmed by the inclusion of NCED6
in our endosperm expressed list ( Table 3). Expression
constructs were also made for a second gene, NCED9
However, despite using over 3 kb of upstream sequence, neither
GUS nor GFP could be observed in the seed. This was inconsistent
with RT-PCR data from RNA obtained from manually dissected seed
and in situ hybridization, both of which measure the
abundance of the endogenous transcript directly [ 63].
Comparisons of endogenous gene expression and reporter activity
suggest that the upstream sequences used to drive reporter
expression do not necessarily confer expression patterns identical
to those of the endogenous gene as assessed by other methods such
as in situ hybridization or RT-PCR
[ 16, 58,
63]. Another endosperm marker that was consistent with our
endosperm-expressed list is SUC5 ( Table 3).
The SUC5 gene encodes a sucrose transporter that is
thought to contribute to the nutrition of the filial tissues
during seed development. Baud et al. [ 64] examined the
expression of the SUC5 promoter using both GUS and GFP
reporter constructs and linked the expression patterns to
expression of the endogenous transcript by in situ
hybridization studies. SUC5 expression was found to be
endosperm-specific, and at 4 DAP (early to mid heart stage) the
expression is changing from localization in the embryo-surrounding
region to a more general endosperm expression [ 64]. The endosperm-expressed list also contains three components of the
fertilization independent seed (FIS) polycomb group complex; MEDEA
(MEA), FIS2, and FIE ( Table 3). Loss of function
mutations of these components of the FIS-complex causes autonomous
onset of cell division in the central cell without fertilization.
Promoter-GUS lines indicate that these three genes are strongly
transcribed in the central cell prior to fertilization. After
fertilization, only the maternal alleles of these genes appear to
be expressed in the endosperm and persist in the free nuclear
endosperm until cellularization [ 65,
72]. As the endosperm
cellularizes at approximately 5 DAP, GUS activity from MEA and
FIS2 reporter constructs decline, whereas FIE promoter-GUS
activity persists and can be observed in both the embryo and the
cellularized endosperm [ 65]. However, MEA expression analysis
by in situ hybridization and RT-PCR from manually
dissected seed compartments indicate that MEA expression also
occurs in the embryo at heart stage [ 26]. Interestingly, the
transcripts of MEA and FIE have relatively low fold
changes in our LM endosperm data ( Table 3), which is
consistent with MEA and FIE expressions in
additional tissues within developing siliques
[ 26, 65]. The FIS complex has been shown to repress the MADS-box gene
PHERES1 ( PHE1). In contrast to the FIS
genes described above, PHE1 is not expressed in the
central cell prior to fertilization and appears to be only
expressed from the paternal genome after fertilization.
PHE1 promoter-GUS reporter lines indicated that
PHE1 expression is expressed in very early embryos and
the endosperm [ 73]. At heart stage however, PHE1
expression becomes restricted to the chalazal region of the
endosperm, an expression pattern that is consistent with the high
differential described by PHE1 in our endosperm-expressed
list. Another endosperm marker that describes a high differential
in our endosperm list is FWA Although the function of
FWA is unknown, its expression has been characterized
using both reporter construct and in situ hybridization
and has a similar expression to FIS2. FWA
expression in the endosperm also appears to be solely from the
maternal allele [ 19]. Imprinting in plants has been
hypothesized to be an endosperm-specific phenomenon and our list
of endosperm-expressed genes includes all confirmed imprinted
genes in Arabidopsis, that is, PHE,
MEA, FIS2, FWA
[ 19, 26,
62, 73,
74],
and the likely imprinted gene FIE
[ 75] ( Table 3). In summary, our microarray analysis using endosperm-derived RNA
identified approximately 2 700 genes with putative endosperm
expression at 4 DAP using a 2-fold cutoff (Supplementary
Table 1). The presence of known endosperm marker genes
in this gene list indicates that our data is reliable. With the extensive molecular resources available for
Arabidopsis and established methods for RNA
amplification, LM represents an attractive tool capable of
providing high-resolution expression analysis and enables novel
insights into the cellular processes that occur during early seed
development. Here, we have established methods for the generation
of expression profiles from specific seed tissues and have shown
that despite some limitations, global RNA amplification from
50 ng of endosperm total RNA can generate high-quality
expression data. This data should provide a useful resource for
laboratories interested in early seed development. The authors would like to thank Peter Stockwell for mapping of the
array probe sequences to transcripts, Robert Schaffer for
supplying the script for array normalization, and Bart Jansen for
printing the arrays. Sample preparation was carried out at the
Histology Service Unit, University of Otago, with advice from Ken
Turner. The first author was supported by a University
of Otago Ph.D. scholarship and the third author was
supported by the Marsden Fund. 1. Brandt SP. Microgenomics: gene expression analysis at the tissue-specific and single-cell levels. Journal of Experimental Botany. 2005;56(412):495–505. [PubMed] 2. Emmert-Buck MR, Bonner RF, Smith PD, et al. Laser capture microdissection. Science. 1996;274(5289):998–1001. [PubMed] 3. Day RC, Grossniklaus U, Macknight RC. Be more specific! laser-assisted microdissection of plant cells. Trends in Plant Science. 2005;10(8):397–406. [PubMed] 4. Day RC, McNoe L, Macknight RC. Transcript analysis of laser microdissected plant cells. Physiologia Plantarum. 2007;129(2):267–282. 5. Kerk NM, Ceserani T, Tausta SL, Sussex IM, Nelson TM. Laser capture microdissection of cells from plant tissues. Plant Physiology. 2003;132(1):27–35. [PubMed] 6. Schad M, Lipton MS, Giavalisco P, Smith RD, Kehr J. Evaluation of two-dimensional electrophoresis and liquid chromatography—tandem mass spectrometry for tissue-specific protein profilling of laser-microdissected plant samples. Electrophoresis. 2005;26(14):2729–2738. [PubMed] 7. Schad M, Mungur R, Fiehn O, Kehr J. Metabolic profiling of laser microdissected vascular bundles of Arabidopsis thaliana
. Plant Methods. 2005;1(1):2. [PubMed] 8. van Gelder RN, von Zastrow ME, Yool A, Dement WC, Barchas JD, Eberwine JH. Amplified RNA synthesized from limited quantities of heterogeneous cDNA. Proceedings of the National Academy of Sciences of the United States of America. 1990;87(5):1663–1667. [PubMed] 9. Iscove NN, Barbara M, Gu M, Gibson M, Modi C, Winegarden N. Representation is faithfully preserved in global cDNA amplified exponentially from sub-picogram quantities of mRNA. Nature Biotechnology. 2002;20(9):940–943. [PubMed] 10. Wang E. RNA amplification for successful gene profiling analysis. Journal of Translational Medicine. 2005;3:28. [PubMed] 11. Aoyagi K, Tatsuta T, Nishigaki M, et al. A faithful method for PCR-mediated global mRNA amplification and its integration into microarray analysis on laser-captured cells. Biochemical and Biophysical Research Communications. 2003;300(4):915–920. [PubMed] 12. Klur S, Toy K, Williams MP, Certa U. Evaluation of procedures for amplification of small-size samples for hybridization on microarrays. Genomics. 2004;83(3):508–517. [PubMed] 13. Nygaard V, Hovig E. Options available for profiling small samples: a review of sample amplification technology when combined with microarray profiling. Nucleic Acids Research. 2006;34(3):996–1014. [PubMed] 14. Berger F, Grini PE, Schnittger A. Endosperm: an integrator of seed growth and development. Current Opinion in Plant Biology. 2006;9(6):664–670. [PubMed] 15. Sundaresan V. Control of seed size in plants. Proceedings of the National Academy of Sciences of the United States of America. 2005;102(50):17887–17888. [PubMed] 16. Tiwari S, Spielman M, Day RC, Scott RJ. Proliferative phase endosperm promoters from Arabidopsis thaliana
. Plant Biotechnology Journal. 2006;4(4):393–407. [PubMed] 17. Scott RJ, Spielman M, Bailey J, Dickinson HG. Parent-of-origin effects on seed development in Arabidopsis thaliana
. Development. 1998;125(17):3329–3341. [PubMed] 18. Petalidis L, Bhattacharyya S, Morris GA, Collins VP, Freeman TC, Lyons PA. Global amplification of mRNA by template-switching PCR: linearity and application to microarray analysis. Nucleic Acids Research. 2003;31(22):e142. [PubMed] 19. Kinoshita T, Miura A, Choi Y, et al. One-way control of FWA imprinting in Arabidopsis endosperm by DNA methylation. Science. 2004;303(5657):521–523. [PubMed] 20. Kinoshita T, Yadegari R, Harada JJ, Goldberg RB, Fischer RL. Imprinting of the MEDEA Polycomb gene in the Arabidopsis endosperm. The Plant Cell. 1999;11(10):1945–1952. [PubMed] 21. Inada N, Wildermuth MC. Novel tissue preparation method and cell-specific marker for laser microdissection of Arabidopsis mature leaf. Planta. 2005;221(1):9–16. [PubMed] 22. Nakazono M, Qiu F, Borsuk LA, Schnable PS. Laser-capture microdissection, a tool for the global analysis of gene expression in specific plant cell types: identification of genes expressed differentially in epidermal cells or vascular tissues of maize. The Plant Cell. 2003;15(3):583–596. [PubMed] 23. Klink VP, Alkharouf N, MacDonald M, Matthews B. Laser capture microdissection (LCM) and expression analyses of Glycine max (soybean) syncytium containing root regions formed by the plant pathogen Heterodera glycines (soybean cyst nematode). Plant Molecular Biology. 2005;59(6):965–979. [PubMed] 24. Ramsay K, Wang Z, Jones MGK. Using laser capture microdissection to study gene expression in early stages of giant cells induced by root-knot nematodes. Molecular Plant Pathology. 2004;5(6):587–592. 25. Gehring M, Huh JH, Hsieh T-F, et al. DEMETER DNA glycosylase establishes MEDEA Polycomb gene self-imprinting by allele-specific demethylation. Cell. 2006;124(3):495–506. [PubMed] 26. Vielle-Calzada J-P, Thomas J, Spillane C, Coluccio A, Hoeppner MA, Grossniklaus U. Maintenance of genomic imprinting at the Arabidopsis medea locus requires zygotic DDM1 activity. Genes and Development. 1999;13(22):2971–2982. [PubMed] 27. Gallardo TD, Hammer RE, Garry DJ. RNA amplification and transcriptional profiling for analysis of stem cell populations. Genesis. 2003;37(2):57–63. [PubMed] 28. Li J, Adams L, Schwartz SM, Bumgarner RE. RNA amplification, fidelity and reproducibility of expression profiling. Comptes Rendus Biologies. 2003;326(10-11):1021–1030. [PubMed] 29. Marko NF, Frank B, Quackenbush J, Lee NH. A robust method for the amplification of RNA in the sense orientation. BMC Genomics. 2005;6(1):27. [PubMed] 30. Naderi A, Ahmed AA, Barbosa-Morais NL, Aparicio S, Brenton JD, Caldas C. Expression microarray reproducibility is improved by optimising purification steps in RNA amplification and labelling. BMC Genomics. 2004;5(1):9. [PubMed] 31. Polacek DC, Passerini AG, Shi C, et al. Fidelity and enhanced sensitivity of differential transcription profiles following linear amplification of nanogram amounts of endothelial mRNA. Physiological Genomics. 2003;13:147–156. [PubMed] 32. Singh R, Maganti RJ, Jabba SV, et al. Microarray-based comparison of three amplification methods for nanogram amounts of total RNA. American Journal of Physiology—Cell Physiology. 2005;288(5):C1179–C1189. [PubMed] 33. Wilson CL, Pepper SD, Hey Y, Miller CJ. Amplification protocols introduce systematic but reproducible errors into gene expression studies. BioTechniques. 2004;36(3):498–506. [PubMed] 34. Barczak A, Rodriguez MW, Hanspers K, et al. Spotted long oligonucleotide arrays for human gene expression analysis. Genome Research. 2003;13(7):1775–1785. [PubMed] 35. Nygaard V, Løland A, Holden M, et al. Effects of mRNA amplification on gene expression ratios in cDNA experiments estimated by analysis of variance. BMC Genomics. 2003;4(1):11. [PubMed] 36. Stoyanova R, Upson JJ, Patriotis C, et al. Use of RNA amplification in the optimal characterization of global gene expression using cDNA microarrays. Journal of Cellular Physiology. 2004;201(3):359–365. [PubMed] 37. Puskas LG, Zvara A, Hackler L, Jr, van Hummelen P. RNA amplification results in reproducible microarray data with slight ratio bias. BioTechniques. 2002;32(6):1330–1340. [PubMed] 38. Baugh LR, Hill AA, Brown EL, Hunter CP. Quantitative analysis of mRNA amplification by in vitro transcription. Nucleic Acids Research. 2001;29(5):E29. [PubMed] 39. Kenzelmann M, Klären R, Hergenhahn M, et al. High-accuracy amplification of nanogram total RNA amounts for gene profiling. Genomics. 2004;83(4):550–558. [PubMed] 40. Feldman AL, Costouros NG, Wang E, et al. Advantages of mRNA amplification for microarray analysis. BioTechniques. 2002;33(4):906–914. [PubMed] 41. Zhao H, Hastie T, Whitfield ML, Børresen-Dale A-L, Jeffrey SS. Optimization and evaluation of T7 based RNA linear amplification protocols for cDNA microarray analysis. BMC Genomics. 2002;3(1):31. [PubMed] 42. Galbraith DW. Global analysis of cell type-specific gene expression. Comparative and Functional Genomics. 2003;4(2):208–215. 43. Zhou W, Abruzzese RV, Polejaeva I, Davis S, Davis S, Ji W. Amplification of nanogram amounts of total RNA by the SMART-based PCR method for high-density oligonucleotide microarrays. Clinical Chemistry. 2005;51(12):2354–2356. [PubMed] 44. Bowtell D, Sambrook J. DNA Microarrays: A Molecular Cloning Manual. New York, NY, USA: Cold Spring Harbor; 2003. 45. Carter MG, Hamatani T, Sharov AA, et al. In situ-synthesized novel microarray optimized for mouse stem cell and early developmental expression profiling. Genome Research. 2003;13(5):1011–1021. [PubMed] 46. Draghici S, Khatri P, Eklund AC, Szallasi Z. Reliability and reproducibility issues in DNA microarray measurements. Trends in Genetics. 2006;22(2):101–109. [PubMed] 47. Yuen T, Wurmbach E, Pfeffer RL, Ebersole BJ, Sealfon SC. Accuracy and calibration of commercial oligonucleotide and custom cDNA microarrays. Nucleic Acids Research. 2002;30(10):e48. [PubMed] 48. Jenson SD, Robetorye RS, Bohling SD, et al. Validation of cDNA microarray gene expression data obtained from linearly amplified RNA. Journal of Clinical Pathology—Molecular Pathology. 2003;56(6):307–312. 49. Hu L, Wang J, Baggerly K, et al. Obtaining reliable information from minute amounts of RNA using cDNA microarrays. BMC Genomics. 2002;3(1):16. [PubMed] 50. Ji W, Zhou W, Gregg K, Lindpaintner K, Davis S, Davis S. A method for gene expression analysis by oligonucleotide arrays from minute biological . Analytical Biochemistry. 2004;331(2):329–339. [PubMed] 51. McClintick JN, Jerome RE, Nicholson CR, Crabb DW, Edenberg HJ. Reproducibility of oligonucleotide arrays using small samples. BMC Genomics. 2003;4(1):4. [PubMed] 52. Wadenbäck J, Clapham DH, Craig D, et al. Comparison of standard exponential and linear techniques to amplify small cDNA samples for microarrays. BMC Genomics. 2005;6(1):61. [PubMed] 53. Casson S, Spencer M, Walker K, Lindsey K. Laser capture microdissection for the analysis of gene expression during embryogenesis of Arabidopsis
. The Plant Journal. 2005;42(1):111–123. [PubMed] 54. Schoor O, Weinschenk T, Hennenlotter J. Moderate degradation does not preclude microarray analysis of small amounts of RNA. BioTechniques. 2003;35(6):1192–1201. [PubMed] 55. Baudry A, Caboche M, Lepiniec L. TT8 controls its own expression in a feedback regulation involving TTG1 and homologous MYB and bHLH factors, allowing a strong and cell-specific accumulation of flavonoids in Arabidopsis thaliana
. The Plant Journal. 2006;46(5):768–779. [PubMed] 56. Devic M, Guilleminot J, Debeaujon I, et al. The BANYULS gene encodes a DFR-like protein and is a marker of early seed coat development. The Plant Journal. 1999;19(4):387–398. [PubMed] 57. Lotan T, Ohto M-A, Yee KM, et al.
Arabidopsis LEAFY COTYLEDON1 is sufficient to induce embryo development in vegetative cells. Cell. 1998;93(7):1196–1205. 58. Tsuchiya Y, Nambara E, Naito S, McCourt P. The FUS3 transcription factor functions through the epidermal regulator TTG1 during embryogenesis in Arabidopsis
. The Plant Journal. 2004;37(1):73–81. [PubMed] 59. Gazzarrini S, Tsuchiya Y, Lumba S, Okamoto M, McCourt P. The transcription factor FUSCA3 controls developmental timing in Arabidopsis through the hormones gibberellin and abscisic acid. Developmental Cell. 2004;7(3):373–385. [PubMed] 60. Schmid M, Davison TS, Henz SR, et al. A gene expression map of Arabidopsis thaliana development. Nature Genetics. 2005;37(5):501–506. [PubMed] 61. Miyawaki K, Matsumoto-Kitano M, Kakimoto T. Expression of cytokinin biosynthetic isopentenyltransferase genes in Arabidopsis: tissue specificity and regulation by auxin, cytokinin, and nitrate. The Plant Journal. 2004;37(1):128–138. [PubMed] 62. Köhler C, Page DR, Gagliardini V, Grossniklaus U. The Arabidopsis thaliana MEDEA Polycomb group protein controls expression of PHERES1 by parental imprinting. Nature Genetics. 2005;37(1):28–30. [PubMed] 63. Lefebvre V, North H, Frey A, et al. Functional analysis of Arabidopsis NCED6 and NCED9 genes indicates that ABA synthesized in the endosperm is involved in the induction of seed dormancy. The Plant Journal. 2006;45(3):309–319. [PubMed] 64. Baud S, Wuilléme S, Lemoine R, et al. The AtSUC5 sucrose transporter specifically expressed in the endosperm is involved in early seed development in Arabidopsis
. The Plant Journal. 2005;43(6):824–836. [PubMed] 65. Luo M, Bilodeau P, Dennis ES, Peacock WJ, Chaudhury A. Expression and parent-of-origin effects for FIS2, MEA, and FIE in the endosperm and embryo of developing Arabidopsis seeds. Proceedings of the National Academy of Sciences of the United States of America. 2000;97(19):10637–10642. [PubMed] 66. Moussian B, Schoof H, Haecker A, Jürgens G, Laux T. Role of the ZWILLE gene in the regulation of central shoot meristem cell fate during Arabidopsis embryogenesis. EMBO Journal. 1998;17(6):1799–1809. [PubMed] 67. Steinmann T, Geldner N, Grebe M, et al. Coordinated polar localization of auxin efflux carrier PIN1 by GNOM ARF GEF. Science. 1999;286(5438):316–318. [PubMed] 68. Hamann T, Benkova E, Bäurle I, Kientz M, Jürgens G. The Arabidopsis BODENLOS gene encodes an auxin response protein inhibiting MONOPTEROS-mediated embryo patterning. Genes and Development. 2002;16(13):1610–1615. [PubMed] 69. Long JA, Barton MK. The development of apical embryonic pattern in Arabidopsis
. Development. 1998;125(16):3027–3035. [PubMed] 70. Christensen SK, Dagenais N, Chory J, Weigel D. Regulation of auxin response by the protein kinase PINOID. Cell. 2000;100(4):469–478. [PubMed] 71. Friml J. Auxin transport—shaping the plant. Current Opinion in Plant Biology. 2003;6(1):7–12. [PubMed] 72. Ingouff M, Haseloff J, Berger F. Polycomb group genes control developmental timing of endosperm. The Plant Journal. 2005;42(5):663–674. [PubMed] 73. Köhler C, Hennig L, Spillane C, Pien S, Gruissem W, Grossniklaus U. The Polycomb-group protein MEDEA regulates seed development by controlling expression of the MADS-box gene PHERES1
. Genes and Development. 2003;17(12):1540–1553. [PubMed] 74. Jullien PE, Kinoshita T, Ohad N, Berger F. Maintenance of DNA methylation during the Arabidopsis life cycle is essential for parental imprinting. The Plant Cell. 2006;18(6):1360–1372. [PubMed] 75. Yadegari R, Kinoshita T, Lotan O, et al. Mutations in the FIE and MEA genes that encode interacting Polycomb proteins cause parent-of-origin effects on seed development by distinct mechanisms. The Plant Cell. 2000;12(12):2367–2381. [PubMed]
|
