Amplification biases: possible differences among deviating gene expressions
© Degrelle et al; licensee BioMed Central Ltd. 2008
Received: 09 August 2007
Accepted: 28 January 2008
Published: 28 January 2008
Gene expression profiling has become a tool of choice to study pathological or developmental questions but in most cases the material is scarce and requires sample amplification. Two main procedures have been used: in vitro transcription (IVT) and polymerase chain reaction (PCR), the former known as linear and the latter as exponential. Previous reports identified enzymatic pitfalls in PCR and IVT protocols; however the possible differences between the sequences affected by these amplification defaults were only rarely explored.
Screening a bovine cDNA array dedicated to embryonic stages with embryonic (n = 3) and somatic tissues (n = 2), we proceeded to moderate amplifications starting from 1 μg of total RNA (global PCR or IVT one round). Whatever the tissue, 16% of the probes were involved in deviating gene expressions due to amplification defaults. These distortions were likely due to the molecular features of the affected sequences (position within a gene, GC content, hairpin number) but also to the relative abundance of these transcripts within the tissues. These deviating genes mainly encoded housekeeping genes from physiological or cellular processes (70%) and constituted 2 subsets which did not overlap (molecular features, signal intensities, gene ID). However, the differential expressions identified between embryonic stages were both reliable (minor intersect with biased expressions) and relevant (biologically validated). In addition, the relative expression levels of those genes were biologically similar between amplified and unamplified samples.
Conversely to the most recent reports which challenged the use of intense amplification procedures on minute amounts of RNA, we chose moderate PCR and IVT amplifications for our gene profiling study. Conclusively, it appeared that systematic biases arose even with moderate amplification procedures, independently of (i) the sample used: brain, ovary or embryos, (ii) the enzymatic properties initially inferred (exponential or linear) and (iii) the preliminary optimization of the protocols. Moreover the use of an in-house developed array, small-sized but well suited to the tissues we worked with, was of real interest for the search of differential expressions.
Several years ago gene expression profiling has emerged as a tool of choice to study developmental kinetics  and is now widely used to study mammalian oocytes or embryos (mouse: ; bovine: [3, 4], human , porcine ) including questions on cell lineage differences . However intermingled cells within complex tissues, biopsies, early embryos or single cells give rise to ng or pg amounts of RNA so that amplification has become a prerequisite, coupled sometimes to laser capture micro-dissection (for example ).
Two amplification methods have been reported in the 90ties [9, 10], a linear procedure based on the use of In Vitro Transcription (IVT) and an exponential procedure based on the use of the Polymerase Chain Reaction (PCR). These exponential and linear definitions have been based on the dynamics of the corresponding enzymatic reactions with no implicit reference to their intrinsic pitfalls. However, it has quickly become "obvious" that the linear process was a high-fidelity process which guaranteed the conservation of the initial transcript abundances . It thus became the tool of choice for gene profiling studies on cDNA and oligo-nucleotide arrays . In the meantime, global PCR amplification procedures have been optimised and claimed better than IVT for array screening when starting from the sub-pg quantities of RNA isolated from single cells [7, 13]. As a consequence, numerous reports have compared the performance of PCR and IVT on decreasing amounts of starting material. Some of these extended the comparison to mRNA or total RNA as un-amplified standards although these standards appeared debatable .
Since their first use, both linear and exponential amplification processes became commercially available (reviewed in [14, 15]) and evolved into 8 to 10 different protocols. The original IVT protocol  has been worked out by Baugh et al.  to make it more specific, by Moll et al.  to make it more efficient and by Schlingemann et al.  to adapt it to oligo-arrays composed of sense oriented oligonucleotides. Similarly, the original PCR protocol has evolved into a new amplification process known as the SMART protocol . Several reports compared SMART amplified targets to IVT  and/or global PCR . Improvements of the original Brady's protocol have also been worked out and compared to total RNA . At last, alternatives to IVT or PCR have also been proposed such as a PCR step followed by an IVT step, an IVT step followed by a PCR step, the use of single-stranded cDNA instead of double-stranded cDNA (ribo-SPIA protocol) or the use of subtraction prior to RNA amplification (STAR protocol). These recent procedures have also been compared to IVT or total RNA and sound promising (reviewed in [14, 15]). They fall beyond the scope of this paper.
In all these studies fidelity, reproducibility and linearity of RNA amplification has been a major concern and increasingly refined statistics have been used accordingly (correlation, fold change, T-test, ANOVA; ) to identify amplification biases through deviating expression patterns between amplified and un-amplified material. However, the possible differences between the sequences affected by PCR and IVT amplification defaults were only rarely explored. We thus aimed at studying the biases of moderate amplification protocols as well as their major characteristics, using an in-house developed protocol for the global PCR amplification [24, 25] and taking into account in-house criteria linked to our biological purposes. We thus decided to use a small set of in vivo elongating embryos, recovered after uterus flushing, to screen different arrays with the same embryonic material (this study, [26, 27]). Second, we chose a bovine array dedicated to these bovine stages rather than a larger Affymetrix array where sequences from this embryonic repertoire were not present. Third, we preferred moderate amplifications to intense ones, as routinely practiced on oocytes and earlier embryonic stages, since intense amplifications cost more and have drawbacks too.
On this basis, we showed that deviating expressions affected 16% of the array after PCR or IVT amplifications, formed 2 gene subsets which did not overlap (molecular features, signal intensities, gene ID) and corresponded to housekeeping genes from physiological or cellular processes. Nevertheless, differential expressions were relevant and displayed relative expression levels which were biologically similar, though not identical, between amplified and unamplified samples.
Our purpose was to analyse moderate amplifications on tissues of similar but also different molecular complexities to analyse the relevance and biases in gene profiling studies following RNA amplification. To this aim, we selected two protocols (IVT one round and global PCR) and five tissues: three embryonic tissues of equivalent molecular complexities as revealed by SAGE data on ovoid, tubular and filamentous stages in pig  and two adult tissues of different complexities (brain and ovary) as described by human SAGE and EST data .
Amplified material from each tissue (brain, ovary, ovoid embryos, tubular embryos and filamentous embryos) was indirectly labelled using "random" hexamers. As advised  replicates were emphasized and deliberately focused on technical points. Three or two independent targets for each tissue and each protocol (target replicates) were thus generated and hybridised to 4 replicates of the same array (array replicates), so that 48 measurements per probe were generated for somatic (3 targets × 2 tissues × 2 protocols × 4 arrays) and embryonic samples (2 targets × 3 tissues × 2 protocols × 4 arrays). To find out gene expression differences between protocols or embryonic stages, appropriate statistical analyses have been applied on each set of data.
RNA amplifications: optimisation and quality
Since drawbacks were reported for both IVT and PCR based protocols which could originate from a too long IVT (degradation effect reported by Spiess ) or too many cycles of PCR (saturation effect reported by Cha or Nagy [33, 34]), we first challenged our protocols on our tissues to define optimal amplification conditions. The protocols were tested using increasing in vitro transcription time or PCR cycle number, with a special look at 5 transcripts: 3 endogenous and 1 exogenous transcripts in brain and ovary, 1 endogenous transcript in developing embryos. Transcripts encoding EF1α, L23a and Cytochrome oxidase III were selected as somatic controls because of a differential expression between brain and ovary, an easy detection on slot blots with poly A+ RNA and a different length: 1.7, 0.9 and 0.7 kb, respectively (preliminary data, not shown). As a result, an in vitro transcription of 10 h and 2 rounds of 12 PCR cycles on 1/10 of the reverse transcription looked optimal since (i) the size of the "spiking" transcripts was conserved (ii) no shortage or degradation of the amplified material was observed and (iii) the amount of spiking transcripts had increased linearly with the time of the IVT or the number of cycles in the PCR [see Additional file 1].
Comparing the 3 amplified targets generated on the somatic tissues, it appeared that the anti-sense RNA (or aRNA) obtained after IVT corresponded to molecules of 0.1 to 4 kb with a mean size of 600 bp while the cDNA fragments generated by global PCR were reduced in size: 0.1 to 1 kb, with a mean size of 150 bp [see Additional file 2A–B]. Interestingly cDNA populations were similar for brain and ovary (panel B) whereas aRNA populations displayed slightly different patterns (panel A). However, the 3 aRNA targets generated from brain or ovary were similar [see Additional file 2C]. These results underlined a good reproducibility in the production of target replicates, a slightly different distribution of RNA species between tissues with IVT and a more homogeneous pattern between tissues with PCR.
Global features of somatic and embryonic hybridisations
As previously advised by N'Guyen , we first determined the amount of labelled target to be used for each hybridisation so that no additional signal appeared but the intensity of the positive signals increased when the amount of target did (50, 125, 250 and 500 ng for aRNA or cDNA labelled targets; data not shown). On this basis, 125 ng of each target has been hybridised to each array with no particular focus on the relative amplification rates and the subsequent equivalence between these targets.
We thus confirmed a high correlation within methods, an intermediate correlation between methods and evidenced a divergence between methods for at least a subset of the array. We thus aimed at its characterisation.
Amplification distortions in somatic and embryonic hybridisations
Additional features on the gene subsets affected by amplification defaults. Hairpins, A stretches and promoter like sequences have been investigated. The parameters were the following: hairpins (minimal length: 10 nucleotides, maximal length: 100, maximal gap: 50), A stretches (size: 18A, maximal gap: 3), promoter of the T7 RNA polymerase (forward sequence: CCCTATAGTGAGTCGTATTA and reverse sequence, maximal gap: 6). Results on Panel 1 and 2 are summarised here. Statistical significance between subsets or features has not been evaluated.
Conclusively, it appeared that systematic biases arose during both amplification procedures independently of (i) the sample used: brain, ovary or embryos, (ii) the enzymatic properties initially inferred (exponential or linear) and (iii) the preliminary optimisation of the protocols. These distortions affected 16% of the core array (154/987) and involved different subsets of genes (Panels 1 and 2) which harboured different molecular properties.
Gene expression differences between embryonic stages
Knowing from above that systematic biases arose during amplification (global PCR and IVT one round) and affected 16% of the core array (987 EST), we wondered whether gene expression differences identified between embryonic stages with amplified samples could be both reliable and relevant.
49 gene expression differences were identified between stages (ovoid, tubular and filamentous) with PCR amplified samples and 28 with IVT amplified ones. Among these, 14 were IVT specific, 35 PCR specific and 14 were commonly identified (Fig. 8A). The common ones (n = 14 EST) encoded 4 genes referenced in the Unigene Bos taurus index and corresponded to transcripts identified in another study using IVT amplified samples only . We showed therein that c12, c93, c88 and TKDP1 transcripts were differentially expressed among these stages (c12, c93: Northern blots; TKDP1 ). The IVT specific differences (n = 14) encoded 8 genes referenced in the Bos taurus index, 4 of which were known as reliable differential expressions: IFN-tau (our endogenous control for embryos, Fig. 2B), Cox2 , c12 and PAG11 . Similarly, the PCR specific differences (n = 35) encoded 15 genes referenced in the Bos taurus index, 5 of which were also known as reliable differences: c12, c93, TKDP1, PAG11 and IFN-tau. Surprisingly, they were not identified as common differences between PCR and IVT amplified samples. Looking in more details at the corresponding EST it appeared clearly that, although located in the same Bt., they did not overlap. Extending this analysis to the list of specific differences (n = 49; 35 +14) we found that the EST from the PCR group were frequently located at the 3'end of the referenced cDNAs (or Bt.), as compared to those from the IVT group (Fig. 8B), and displayed reduced sizes (Fig. 8C). Last but not least, a few differences identified between embryonic stages with PCR amplified samples (2 Bt./15) matched with those identified in Panel 2 (2 Bt./32) whereas no intercept was detected with Panel 1.
Comparison of the differential expression ratios observed between embryonic stages using amplified and unamplified material. The expression ratios observed with amplified targets (IVT and PCR) come from the array datasets presented in this study. Those originating from Real-Time PCR come from the gene specific validations (IFN-tau, Cox2) we performed on unamplified RNA from each stage. Those originating from Northern blots (c12, c93, PAG11) come from previous results on unamplified RNA from each stage .
IVT amplified material
PCR amplified material
As a final view, gene expression differences identified between embryonic stages with amplified samples were both reliable (tiny intersect with deviating expressions) and relevant (biologically valid). In addition, the molecular features observed on the differential EST identified by IVT or PCR amplifications suggest that global PCR favoured the representation of short cDNA harbouring rather low GC contents.
This work illustrated the questions frequently asked since 2002 about RNA amplification and showed that even with optimised and reproducible protocols deviating gene expressions affected 16% of our array and appeared whatever the tissue. These biases, linked to the abundance or the molecular features of the sequences affected by amplification defaults, corresponded mainly to housekeeping genes from physiological and cellular processes. Differential expressions, however, were found reliable and relevant with biologically similar expression ratios between amplified and unamplified material.
Similar biases were reported in previous studies using also moderate IVT and PCR amplifications. They evidenced either contradictory expression ratios or missing spots [39, 40] but also a vast majority of expression patterns which differed only in the magnitude of the differential expression . In our study, only one gene out of the five tested showed an inversed ratio at one stage after PCR amplification, whereas most of them showed ratios which differed only in their magnitudes. All of them however were relevant as confirmed by Northern blots or Real-Time PCR. Interestingly, the deviating genes from our study corresponded mainly to housekeeping genes whereas those identified by van Haaften (genes lost during IVT amplification) rather included transcription factors. As an alternative to minimise distortions, Real-Time detection of amplified products has been proposed to prevent over-amplification in PCR-based protocols  and a similar approach has been used before and after IVT amplifications to discriminate between well and badly amplified samples . This has also been used to follow IVT amplifications on bovine oocytes and early embryos (Robert & Sirard, personal communication).
The possible differences between the sequences affected by amplification defaults were however rarely explored. Van Haaften observed that the reporters that disappeared after IVT amplification (20% of them) had a GC content of about 54% and displayed more hairpins of longer sizes than the other reporters (80%). A higher GC content has also been observed in deviating genes after PCR amplification with the SMART protocol . The authors correlated this feature to the temperature of the enzymatic reaction (68 to 72°C for the Taq Polymerase) and to the GC content of their plant genome. This was surprising to us since GC rich fragments are often difficult PCR templates, requiring sometimes DMSO or betaine addition. In our study, we could not assign the distortions from Panel 1 and Panel 2 to IVT or PCR defaults since, without a standard, it was impossible to distinguish IVT over-expression from PCR under-expression and vice versa. It was clear however that these 2 gene subsets did not overlap: different molecular features, different signal intensities and different gene ID. EST from Panel 2 displayed reduced sizes, were more frequently located in the 3'end of the cDNAs and displayed a lower GC content than those from Panel 1. They also contained more hairpins (60% versus 37%) and A stretches (10% versus 5%) than those from Panel 1 but displayed similar contents of promoter-like sequences. Since EST corresponding to true differential expressions identified by PCR targets were frequently located at the 3'end of the referenced cDNAs and displayed reduced sizes (as compared to IVT specific ones), one would suggest that deviating genes from Panel 2 could display a PCR signature.
From this work, it was not really possible to favour PCR over IVT amplification or vice versa. Both generated distortions and revealed true differential expressions between embryonic stages (minor intersect between differential patterns and biases), so that one would rather advise (i) using only one protocol to keep amplification factors and biases equal (ii) monitoring the amplification process as offered now through Real-Time PCR and (iii) searching for protocol specific expression differences or gene-protocol interactions before any differential analysis on a new dataset or a new array. Obviously, the choice between those protocols is also a question of total RNA input, time, cost and available arrays since amplified targets enriched in 3'end fragments will not hybridise to SSH fragments or 5'positioned oligos. Last but not least, knowing that Taq Polymerases make more mistakes than RNA polymerases do, IVT may be favoured over PCR to hybridise highly discriminating oligo-arrays or arrays from other species.
Estrus synchronized heifers of the Charolais breed were inseminated (day 0) and day 12 to day 17 blastocysts were collected by non surgical flushing in warm PBS. Ovoid blastocysts (1–12 mm) came from collects at 12 dpi (day post insemination) whereas tubular and early filamentous stages (50–60 mm and 140–160 mm) were obtained at 14 to 15 and 16 to 17 dpi, respectively. Brain and ovaries were collected on Day-50 pregnant cows. To take adult somatic tissues, animals were humanly put down in the accredited experimental slaughterhouse of INRA under the supervision of veterinary services.
Total RNA from ovoid (n = 4), tubular (n = 4) and filamentous (n = 4) embryos was extracted with RNA-Plus™ (QBioGene). RNA quality was first verified by intact ribosomal bands on a 1% agarose gel (28S and 18S) and A260/280 absorbance ratios. Total RNA from brain and ovary was isolated in the same way. RNA quality was also verified by intact ribosomal bands on a 1% agarose gel (28S and 18S) and A260/280 absorbance ratios. A spiking mRNA was then added to brain and ovary as 1% of the estimated polyA+ amount to test whether highly expressed genes can be biased through amplification. This CG03 mRNA from A. thaliana was in vitro synthesized (with a T7 Megascript kit, Ambion) from the c554 containing plasmid, given to us by H. Hofte (LBC, INRA Versailles, France). Brain and ovary polyA+ RNA were further extracted using a Dynabeads mRNA purification kit (Dynal).
Amplified RNA from each sample was synthesized with the MessageAmp™ aRNA Kit (Ambion) according to the manufacturer instructions. Briefly, 1 μg of total RNA was incubated with 500 ng of an anchored T7-(dT) primer in 12 μl (water) at 70°C for 10 min. The 1rst cDNA strand was synthesized by the addition of 2 μl first-strand buffer, 1 μl RNAse inhibitor, 4 μl dNTP mix and 1 μl reverse transcriptase mix and incubation at 42°C for 2 h. Second-strand synthesis was performed by the addition of 63 μl DEPC-treated water, 10 μl second-strand buffer, 4 μl dNTP mix, 2 μl DNA polymerase, 1 μl RNAse H- and incubation at 16°C for 2 h. DNA was extracted with phenol:chloroform:isoamyl alcohol and precipitated in ethanol with 20 μg glycogen (Ambion). In vitro transcription was carried out at 37°C for 10 h in a 20 μl reaction volume. 1 μl DNAse was added and incubated at 37°C for 30 min. RNA was purified on Mini Quick Spin RNA columns (Roche Diagnostic) and its quality verified on RNA 6000 lab-chips (BioAnalyser 2100; Agilent Technologies).
RNA target labelling
aRNA was retro-transcribed and directly labelled with [α-33P]dATP as described for polyA+RNA . 500 ng of aRNA was mixed with 500 ng of random hexamers in a volume of 25 μl. The mixture was incubated at 70°C for 10 min and chilled on ice. cDNA was synthesised by the addition of 5 μl 10× PCR buffer, 5 μl 25 mM MgCl2, 5 μl 0,1 mM DTT, 2,5 μl 10 mM mix dGTP, dCTP and dTTP, 2,5 μl water, 50 μCi [α-33P]dATP and 200 U Superscript II (Invitrogen) at 42°C for 50 min. The RNA template was removed by the addition of 1 μl RNAse H- and incubation at 37°C for 20 min.
Global RT-PCR amplification
Amplified cDNA was prepared as described  with few modifications. Briefly, 1 μg total RNA was incubated with 1 μl 10 μM oligo(dT), 1 μl 10 mM dNTPs, 1 μl 10% NP40, 1 μl 20 mM DTT, 2 μl first-strand buffer 5×, 1 μl RNAse inhibitor (Ambion) at 65°C for 2 min, at room temperature for 3 min and cooled on ice. cDNA was synthesised by the addition of 200 U Superscript II (Invitrogen) and 2 U AMV (Gibco BRL) and incubation at 42°C for 30 min. First-strand cDNA were poly(dG)-tailed by incubation with 1 μl 20 mM dGTP, 4 μl TdT buffer 5×, 2,5 μl water, 2,5 μl TdT enzyme (Promega) at 37°C for 1 h. The first PCR was performed in a volume of 50 μl using 1/10 of the RT and the second PCR was performed on 1/4 of the first PCR. Samples were incubated at 94°C for 10 min before the two rounds of PCR cycles (12 cycles each; 94°C for 2 min, 63°C for 50 sec and 72°C for 6 min). PCR products were then purified using Qiaquick PCR purification (Qiagen) and their quality verified on DNA 7500 lab-chips (BioAnalyser 2100; Agilent Technologies).
cDNA target labelling
PCR-amplified cDNA was labelled with [α-33P]dATP using random hexamers and Klenow included in Atlas SMART Probe Amplification kit (Clontech). 500 ng of amplified cDNA was mixed with 500 ng random hexamers in a volume of 34 μl. The mixture was incubated at 98°C for 8 min and at 50°C for 3 min. After addition of 5 μl 10× buffer, 5 μl dNTPs for ATP label, 5 μl [α-33P]dATP and 1 μl Klenow, the reaction mixture was incubated at 50°C for 30 min and stopped with 2 μl 0,5 M EDTA. Labelled targets were then purified on Sephadex columns (G-50).
Quantitative Real time PCR
Real-time PCR was carried out in a final volume of 30 μl with 1 μl of diluted reverse transcriptions (1/100; 1/1000) in a 1× SYBR green Master Mix (Applied Biosystems) with 0.3 μM of gene-specific primers. Reactions were run on ABI Prism 7000 HT (Applied Biosystems). The presence of a specific and unique PCR product was checked by ABI Prism melting curves. The relative quantification of the initial amount of target was extrapolated from the appropriate standard curve, which was generated simultaneously while using serial dilutions of the corresponding PCR product. IFN-tau and Gapdh primers were as published [43, 44] but Cox2 primers were a kind gift from G. Charpigny. Their sequences (unpublished so far) will be available upon request email@example.com.
125 ng of polyA+RNA, aRNA or cDNA were spotted and cross-linked to HybondN+ membranes (Amersham) at 80°C for 2 h. DNA probes encoding IFN-tau, CG03, EF1α, RPL23a or Cytochrome oxidase III were [α-32P]dCTP-labelled using the Ready-Prime kit (Amersham). Apart from CG03, those DNA probes originate from the array. Hybridisations were conducted at 65°C for 16 h and washes performed once in 2 × SSC, 0,1% SDS at 65°C for 30 min and twice in 0,1 × SSC, 0,1% SDS at 65°C for 10 min. Slot blots were then exposed to phosphor-imaging for 24 hours and signal intensities quantified with the ImageQuant 3.3 software (Molecular Dynamics).
The bovine embryonic array used here originates from a bovine cDNA library established at the ovoid stage, starting from 1.6 μg of RNA and using the Cap Finder cDNA kit from Clontech as described in Degrelle et al. . Briefly, cDNA inserts from the arrayed library were amplified by PCR using the flanking primers from the Cap Finder kit and selected for spotting after a short run on a 2.5% agarose gel. 1855 probes were then spotted and fixed (UV light, 1 min, 1200 J, twice) onto nylon N+ membranes (8 cm × 12 cm, Amersham Biosciences) with a 5 × 5 pattern (BioRobotics). This was achieved with the kind help of C. Matingou and G. Piétu at the Genexpress Laboratory headed by C. Auffray (CNRS FRE 2571, Villejuif, France). The library has been called "bcai" and indexed in TGI and NCBI database as "#FJB" and "15979", respectively [45, 46] and the array published as "INRA-BDR Bovine D14 Embryo 1K" (GPL6284) in NCBI Gene Expression Omnibus database . Bacterial clones are available upon request at the CRB GADIE (INRA Jouy en Josas, France )
Array hybridization, image acquisition and quantification
Each target was hybridized to 4 array replicates using ExpressHyb™ Hybridization Solution (Clontech) at 68°C overnight. Arrays were washed four times in 2 × SSC, 1% SDS and once in 0.1 × SSC, 0.5% SDS at 68°C for 30 min each. They were then exposed to phosphor-screens for 7 days. The hybridization signals were quantified with the Imagene 3.1 software from BioDiscovery (Proteigene) on the PICT plateform (INRA Jouy en Josas, France). These raw datasets are accessible in NCBI Gene Expression Omnibus database (experimental series "GSE9929" ). Internal controls within the array corresponded to 65 probes and either positive or negative controls were as expected in all the hybridizations. A signal was considered "valid" when the Imagene software did not flag it (flag = 0) and when the same signal was observed on 2 thirds of the arrays, namely: 8 out of 12 for the somatic targets and 5 out of 8 for the embryonic ones.
Gene expression analyses
All the plots (scatter plots, histograms) were performed on R environment .
Gene expression differences between protocols
These analyses were performed either on the whole array (1855 inserts plus 65 controls = 1920 probes) or on the biological core of the array also called 1 K array (1097 informative sequences submitted to the EBI – 110 mitochondrial sequences = 987 probes). With 2 protocols, 5 tissues, 2 to 3 target replicates per protocol and 4 array replicates per target (as indicated in the experimental design, Fig. 1), these analyses involved 184 320 (1920*2*2*3*4+1920*2*3*2*4) and 94 752 (987*2*2*3*4+987*2*3*2*4) pieces of data, respectively. Statistical and clustering analyses were performed using TIGR MeV 3.0 (MultiExperiment Viewer software ). Before calculations, the data were log2 transformed and standardised within each protocol. Differences between PCR and IVT methods were assessed by a Student's t-test assuming an unequal variance (Welch approximation). The adjusted Bonferroni correction was considered at P < 0.05. An unsupervised hierarchical clustering, based on Euclidean distance and complete linkage, was performed on the significant gene expression differences between the 2 methods.
Gene expression differences between embryonic stages
These analyses were performed on the biological core of the array (987 probes). With 2 protocols, 3 embryonic stages, 2 target replicates per protocol and 4 array replicates per target (as indicated in the design), these analyses involved 47 376 (987*3*2*2*4) pieces of data. To identify gene expression differences between stages, we used a set of SAS macros called AnovArray and performed an analysis of variance considering a homogeneous variance for all the genes (HOM option) and a multiple testing (False Discovery Rate) at the threshold 5% (details in [26, 51]). AnovArray has been originally conceived to analyse these datasets.
Biological processes were analysed through Gene Ontology annotations  considering the Indentation 1. EST size, GC content, EST position according to the referenced mRNA of the Bos taurus gene index  were performed using Perl scripts and box plot function from the R environment . Presence of hairpins, dA stretches and sequences similar to RNA polymerase promoters was evaluated using the palindrome and fuzznuc programs of the Emboss package .
The authors wish to thank Véronique Duranthon for introducing us to her protocol of global PCR amplification, Olivier Dubois for his precious help in Real-Time PCR experiments and Philippe Bardou from SIGENAE team for the submission of the "INRA-BDR Bovine D14 Embryo 1K" to the GEO database. This work was supported by the EEC (contract BOI4-CT95-0190) and INRA (AIP P00183). Degrelle S. A. was a MNERT fellow.
- Ko MS, Kitchen JR, Wang X, Threat TA, Wang X, Hasegawa A, Sun T, Grahovac MJ, Kargul GJ, Lim MK: Large-scale cDNA analysis reveals phased gene expression patterns during preimplantation mouse development. Development. 2000, 127 (8): 1737-1749.PubMedGoogle Scholar
- Zeng F, Baldwin DA, Schultz RM: Transcript profiling during preimplantation mouse development. Dev Biol. 2004, 272 (2): 483-496. 10.1016/j.ydbio.2004.05.018.PubMedView ArticleGoogle Scholar
- Misirlioglu M, Page GP, Sagirkaya H, Kaya A, Parrish JJ, First NL, Memili E: Dynamics of global transcriptome in bovine matured oocytes and preimplantation embryos. Proc Natl Acad Sci USA. 2006, 103 (50): 18905-18910. 10.1073/pnas.0608247103.PubMedPubMed CentralView ArticleGoogle Scholar
- Mamo S, Sargent CA, Affara NA, Tesfaye D, El-Halawany N, Wimmers K, Gilles M, Schellander K, Ponsuksili S: Transcript profiles of some developmentally important genes detected in bovine oocytes and in vitro-produced blastocysts using RNA amplification and cDNA microarrays. Reprod Domest Anim. 2006, 41 (6): 527-534. 10.1111/j.1439-0531.2006.00708.x.PubMedView ArticleGoogle Scholar
- Dobson AT, Raja R, Abeyta MJ, Taylor T, Shen S, Haqq C, Pera RA: The unique transcriptome through day 3 of human preimplantation development. Hum Mol Genet. 2004, 13 (14): 1461-1470. 10.1093/hmg/ddh157.PubMedView ArticleGoogle Scholar
- Whitworth KM, Agca C, Kim JG, Patel RV, Springer GK, Bivens NJ, Forrester LJ, Mathialagan N, Green JA, Prather RS: Transcriptional profiling of pig embryogenesis by using a 15-K member unigene set specific for pig reproductive tissues and embryos. Biol Reprod. 2005, 72 (6): 1437-1451. 10.1095/biolreprod.104.037952.PubMedView ArticleGoogle Scholar
- Kurimoto K, Yabuta Y, Ohinata Y, Ono Y, Uno KD, Yamada RG, Ueda HR, Saitou M: An improved single-cell cDNA amplification method for efficient high-density oligonucleotide microarray analysis. Nucleic Acids Res. 2006, 34 (5): e42-10.1093/nar/gkl050.PubMedPubMed CentralView ArticleGoogle Scholar
- Luo L, Salunga RC, Guo H, Bittner A, Joy KC, Galindo JE, Xiao H, Rogers KE, Wan JS, Jackson MR: Gene expression profiles of laser-captured adjacent neuronal subtypes. Nat Med. 1999, 5 (1): 117-122. 10.1038/4806.PubMedView ArticleGoogle Scholar
- Van Gelder RN, von Zastrow ME, Yool A, Dement WC, Barchas JD, Eberwine JH: Amplified RNA synthesized from limited quantities of heterogeneous cDNA. Proc Natl Acad Sci USA. 1990, 87 (5): 1663-1667. 10.1073/pnas.87.5.1663.PubMedPubMed CentralView ArticleGoogle Scholar
- Brady G, Billia F, Knox J, Hoang T, Kirsch IR, Voura EB, Hawley RG, Cumming R, Buchwald M, Siminovitch K: Analysis of gene expression in a complex differentiation hierarchy by global amplification of cDNA from single cells. Curr Biol. 1995, 5 (8): 909-922. 10.1016/S0960-9822(95)00181-3.PubMedView ArticleGoogle Scholar
- Wang E, Miller LD, Ohnmacht GA, Liu ET, Marincola FM: High-fidelity mRNA amplification for gene profiling. Nat Biotechnol. 2000, 18 (4): 457-459. 10.1038/74546.PubMedView ArticleGoogle Scholar
- Affymetrix. [http://affymetrix.com]
- 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. Nat Biotechnol. 2002, 20 (9): 940-943. 10.1038/nbt729.PubMedView ArticleGoogle Scholar
- Nygaard V, Hovig E: Options available for profiling small samples: a review of sample amplification technology when combined with microarray profiling. Nucleic Acids Res. 2006, 34 (3): 996-1014. 10.1093/nar/gkj499.PubMedPubMed CentralView ArticleGoogle Scholar
- Peano C, Severgnini M, Cifola I, De Bellis G, Battaglia C: Transcriptome amplification methods in gene expression profiling. Expert Rev Mol Diagn. 2006, 6 (3): 465-480. 10.1586/14737220.127.116.115.PubMedView ArticleGoogle Scholar
- Eberwine J, Yeh H, Miyashiro K, Cao Y, Nair S, Finnell R, Zettel M, Coleman P: Analysis of gene expression in single live neurons. Proc Natl Acad Sci USA. 1992, 89 (7): 3010-3014. 10.1073/pnas.89.7.3010.PubMedPubMed CentralView ArticleGoogle Scholar
- Baugh LR, Hill AA, Brown EL, Hunter CP: Quantitative analysis of mRNA amplification by in vitro transcription. Nucleic Acids Res. 2001, 29 (5): E29-10.1093/nar/29.5.e29.PubMedPubMed CentralView ArticleGoogle Scholar
- Moll PR, Duschl J, Richter K: Optimized RNA amplification using T7-RNA-polymerase based in vitro transcription. Anal Biochem. 2004, 334 (1): 164-174. 10.1016/j.ab.2004.07.013.PubMedView ArticleGoogle Scholar
- Schlingemann J, Thuerigen O, Ittrich C, Toedt G, Kramer H, Hahn M, Lichter P: Effective transcriptome amplification for expression profiling on sense-oriented oligonucleotide microarrays. Nucleic Acids Res. 2005, 33 (3): e29-10.1093/nar/gni029.PubMedPubMed CentralView ArticleGoogle Scholar
- Clontech. [http://www.clontech.com]
- Wadenback J, Clapham DH, Craig D, Sederoff R, Peter GF, von Arnold S, Egertsdotter U: Comparison of standard exponential and linear techniques to amplify small cDNA samples for microarrays. BMC Genomics. 2005, 6 (1): 61-10.1186/1471-2164-6-61.PubMedPubMed CentralView ArticleGoogle Scholar
- Subkhankulova T, Livesey FJ: Comparative evaluation of linear and exponential amplification techniques for expression profiling at the single-cell level. Genome Biol. 2006, 7 (3): R18-10.1186/gb-2006-7-3-r18.PubMedPubMed CentralView ArticleGoogle Scholar
- Cui X, Churchill GA: Statistical tests for differential expression in cDNA microarray experiments. Genome Biol. 2003, 4 (4): 210-10.1186/gb-2003-4-4-210.PubMedPubMed CentralView ArticleGoogle Scholar
- Pacheco-Trigon S, Hennequet-Antier C, Oudin JF, Piumi F, Renard JP, Duranthon V: Molecular characterization of genomic activities at the onset of zygotic transcription in mammals. Biol Reprod. 2002, 67 (6): 1907-1918. 10.1095/biolreprod67.6.1907.PubMedView ArticleGoogle Scholar
- Revel F, Renard JP, Duranthon V: PCR-generated cDNA libraries from reduced numbers of mouse oocytes. Zygote. 1995, 3 (3): 241-250.PubMedView ArticleGoogle Scholar
- Degrelle SA, Campion E, Cabau C, Piumi F, Reinaud P, Richard C, Renard JP, Hue I: Molecular evidence for a critical period in mural trophoblast development in bovine blastocysts. Dev Biol. 2005, 288 (2): 448-460. 10.1016/j.ydbio.2005.09.043.PubMedView ArticleGoogle Scholar
- Hue I, Degrelle SA, Campion E, Renard JP: Gene expression in elongating and gastrulating embryos from ruminants. Soc Reprod Fertil Suppl. 2007, 64: 365-377.PubMedGoogle Scholar
- Blomberg LA, Long EL, Sonstegard TS, Van Tassell CP, Dobrinsky JR, Zuelke KA: Serial analysis of gene expression during elongation of the peri-implantation porcine trophectoderm (conceptus). Physiol Genomics. 2005, 20 (2): 188-194. 10.1152/physiolgenomics.00157.2004.PubMedView ArticleGoogle Scholar
- Huminiecki L, Lloyd AT, Wolfe KH: Congruence of tissue expression profiles from Gene Expression Atlas, SAGEmap and TissueInfo databases. BMC Genomics. 2003, 4 (1): 31-10.1186/1471-2164-4-31.PubMedPubMed CentralView ArticleGoogle Scholar
- Kendziorski C, Irizarry RA, Chen KS, Haag JD, Gould MN: On the utility of pooling biological samples in microarray experiments. Proc Natl Acad Sci USA. 2005, 102 (12): 4252-4257. 10.1073/pnas.0500607102.PubMedPubMed CentralView ArticleGoogle Scholar
- Herwig R, Aanstad P, Clark M, Lehrach H: Statistical evaluation of differential expression on cDNA nylon arrays with replicated experiments. Nucleic Acids Res. 2001, 29 (23): E117-10.1093/nar/29.23.e117.PubMedPubMed CentralView ArticleGoogle Scholar
- Spiess AN, Mueller N, Ivell R: Amplified RNA degradation in T7-amplification methods results in biased microarray hybridizations. BMC Genomics. 2003, 4 (1): 44-10.1186/1471-2164-4-44.PubMedPubMed CentralView ArticleGoogle Scholar
- Cha RS, Thilly WG: PCR Methods Appl. 1993, 3: S18-29.PubMedView ArticleGoogle Scholar
- Nagy ZB, Kelemen JZ, Feher LZ, Zvara A, Juhasz K, Puskas LG: Real-time polymerase chain reaction-based exponential sample amplification for microarray gene expression profiling. Anal Biochem. 2005, 337 (1): 76-83. 10.1016/j.ab.2004.09.044.PubMedView ArticleGoogle Scholar
- Roberts RM, Ezashi T, Rosenfeld CS, Ealy AD, Kubisch HM: Evolution of the interferon tau genes and their promoters, and maternal-trophoblast interactions in control of their expression. Reprod Suppl. 2003, 61: 239-251.PubMedGoogle Scholar
- Nguyen C, Rocha D, Granjeaud S, Baldit M, Bernard K, Naquet P, Jordan BR: Differential gene expression in the murine thymus assayed by quantitative hybridization of arrayed cDNA clones. Genomics. 1995, 29 (1): 207-216. 10.1006/geno.1995.1233.PubMedView ArticleGoogle Scholar
- MacLean JA, Chakrabarty A, Xie S, Bixby JA, Roberts RM, Green JA: Family of Kunitz proteins from trophoblast: expression of the trophoblast Kunitz domain proteins (TKDP) in cattle and sheep. Mol Reprod Dev. 2003, 65 (1): 30-40. 10.1002/mrd.10262.PubMedView ArticleGoogle Scholar
- Charpigny G, Reinaud P, Tamby JP, Creminon C, Martal J, Maclouf J, Guillomot M: Expression of cyclooxygenase-1 and -2 in ovine endometrium during the estrous cycle and early pregnancy. Endocrinology. 1997, 138 (5): 2163-2171. 10.1210/en.138.5.2163.PubMedGoogle Scholar
- Puskas LG, Zvara A, Hackler L, Van Hummelen P: RNA amplification results in reproducible microarray data with slight ratio bias. Biotechniques. 2002, 32 (6): 1330-1334. 1336, 1338, 1340.PubMedGoogle Scholar
- van Haaften RI, Schroen B, Janssen BJ, van Erk A, Debets JJ, Smeets HJ, Smits JF, van den Wijngaard A, Pinto YM, Evelo CT: Biologically relevant effects of mRNA amplification on gene expression profiles. BMC Bioinformatics. 2006, 7: 200-10.1186/1471-2105-7-200.PubMedPubMed CentralView ArticleGoogle Scholar
- Laurell C, Wirta V, Nilsson P, Lundeberg J: Comparative analysis of a 3' end tag PCR and a linear RNA amplification approach for microarray analysis. J Biotechnol. 2007, 127 (4): 638-646. 10.1016/j.jbiotec.2006.08.016.PubMedView ArticleGoogle Scholar
- Decraene C, Reguigne-Arnould I, Auffray C, Pietu G: Reverse transcription in the presence of dideoxynucleotides to increase the sensitivity of expression monitoring with cDNA arrays. Biotechniques. 1999, 27 (5): 962-966.PubMedGoogle Scholar
- Bertolini M, Beam SW, Shim H, Bertolini LR, Moyer AL, Famula TR, Anderson GB: Growth, development, and gene expression by in vivo- and in vitro-produced day 7 and 16 bovine embryos. Mol Reprod Dev. 2002, 63 (3): 318-328. 10.1002/mrd.90015.PubMedView ArticleGoogle Scholar
- Smith JL, Sheffield LG: Production and regulation of leptin in bovine mammary epithelial cells. Domest Anim Endocrinol. 2002, 22 (3): 145-154. 10.1016/S0739-7240(02)00121-2.PubMedView ArticleGoogle Scholar
- The Gene Index database (TGI). [http://compbio.dfci.harvard.edu/tgi/]
- National Center for Biotechnology Information (NCBI).
- Gene Expression Omnibus (GEO). [http://www.ncbi.nlm.nih.gov/geo/]
- Centre de Ressources Biologiques GADIE. [http://www-crb.jouy.inra.fr/]
- The R Project for Statistical Computing. [http://www.r-project.org/]
- MultiExperiment Viewer Software. [http://www.tm4.org/mev.html]
- Hennequet-Antier C, Chiapello H, Piot K, Degrelle S, Hue I, Renard JP, Rodolphe F, Robin S: AnovArray: a set of SAS macros for the analysis of variance of gene expression data. BMC Bioinformatics. 2005, 6: 150-10.1186/1471-2105-6-150.PubMedPubMed CentralView ArticleGoogle Scholar
- Gene Ontology Database. [http://www.geneontology.org/]
- Emboss package. [http://emboss.sourceforge.net/]
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.