Grapevine cell early activation of specific responses to DIMEB, a resveratrol elicitor

Background In response to pathogen attack, grapevine synthesizes phytoalexins belonging to the family of stilbenes. Grapevine cell cultures represent a good model system for studying the basic mechanisms of plant response to biotic and abiotic elicitors. Among these, modified β-cyclodextrins seem to act as true elicitors inducing strong production of the stilbene resveratrol. Results The transcriptome changes of Vitis riparia × Vitis berlandieri grapevine cells in response to the modified β-cyclodextrin, DIMEB, were analyzed 2 and 6 h after treatment using a suppression subtractive hybridization experiment and a microarray analysis respectively. At both time points, we identified a specific set of induced genes belonging to the general phenylpropanoid metabolism, including stilbenes and hydroxycinnamates, and to defence proteins such as PR proteins and chitinases. At 6 h we also observed a down-regulation of the genes involved in cell division and cell-wall loosening. Conclusions We report the first large-scale study of the molecular effects of DIMEB, a resveratrol inducer, on grapevine cell cultures. This molecule seems to mimic a defence elicitor which enhances the physical barriers of the cell, stops cell division and induces phytoalexin synthesis.


Background
Plants respond to pathogens through constitutive and inducible mechanisms [1]. Structural barriers represent preformed constitutive defences, while the accumulation of pathogenesis-related proteins (PR), phytoalexins and reactive oxygen species is part of an active mechanism stimulated by the pathogen [2]. Grapevine also responds to fungal infection via PR-protein synthesis and phyto-alexin accumulation [3]. Plant phytoalexins are lowmolecular-weight secondary metabolites with antimicrobial properties and they show wide chemical diversity among different plant species [4]. In grapevine they mainly belong to the stilbene family and consist of transresveratrol (3,5,4'-trihydroxystilbene) its oligomers, called viniferins [5][6][7] and pterostilbene, a dimethylated derivative of resveratrol [8]. Stilbene synthesis in berries [9] and leaves can be elicited by fungal infection [5,10], but also by treatment with UV-irradiation [11], ozone [12] and heavy metals [13].
Plant cell cultures are a useful tool for studying plant cell defence response to biotic and abiotic elicitors [14]. Stilbene accumulation has been reported in grapevine cells treated with different elicitors: fungal cell wall fragments [15], Na-orthovanadate, jasmonic acid and methyljasmonate [16,17] and laminarin, a β-glucan polysaccharide from brown algae [18]. In addition, special attention has been given to the β-cyclodextrin molecular class. These are cyclic oligosaccharides consisting of seven α-D-glucopyranose residues linked by α 1 → 4 glucosidic bonds forming a structure with a hydrophobic central cavity and a hydrophilic external surface [19]. Among β-cyclodextrins, heptakis(2,6-di-O-methyl)-β-cyclodextrin (DIMEB), was reported to be the most effective resveratrol elicitor in different Vitis vinifera cultivars [19,20]. The ability of the modified β-cyclodextrins to act as elicitors probably resides in their chemical similarity to the alkyl-derivatized pectic oligosaccharides released from the cell walls during fungal infection [20]. Along with stilbene accumulation these experiments highlighted a more general response involving peroxidase activity as well as inhibition of Botrytis cinerea growth [19,20].
Zamboni et al. [21] further investigated DIMEB activity on additional Vitis genotypes and observed that its effect was more pronounced when tested on Vitis riparia × Vitis berlandieri cell cultures. The kinetics of resveratrol synthesis showed that trans-resveratrol, the induced form, started to accumulate from 6 h after treatment and reached its maximum at 24 h. Moreover, this metabolite was much more localized in the medium than within the cell.
With these results [21] as our starting point, we report here the first large-scale transcriptional characterization of the early response of Vitis riparia × Vitis berlandieri cells to DIMEB treatment.
After 2 h, 127 positively modulated genes were identified by suppression subtractive hybridization (SSH), whereas after 6 h, 371 genes turned out to be differentially expressed when control and treated cells on the Vitis vinifera GeneChip ® Genome Array (Affymetrix) were compared. These results showed that DIMEB specifically modulates the expression of a small number of genes involved in resveratrol and lignin biosynthesis, PR synthesis, cell division and cell wall modification.

Results and discussion
The ability of DIMEB to elicit defence responses in grapevine cell culture was suggested by previous results showing stilbene accumulation, changes in peroxidase activity, as well as inhibition of Botrytis cinerea growth [19,20]. Considerable stilbene accumulation in response to DIMEB treatment was also observed by our group using non-vinifera (Vitis riparia × Vitis berlandieri) liquid cell cultures [21]. In this study we analyzed the changes in gene expression of these cells elicited with DIMEB after 2 h and 6 h using SSH and microarray experiments, respectively.
The rationale behind the two approaches was that after 2 h of treatment, a small number of genes are expected to be modulated, and only to a limited extent, whereas after 6 h an increase in the number of genes and in their expression level is envisaged. The SSH technique appeared then the right choice for identifying the low abundance differential transcripts at 2 h, while the Affymetrix GeneChip ® microarray was used to measure the expression of a larger number of genes (~14,500 unigenes) after 6 h of treatment [22].
Starting with 384 clones from the constructed cDNA subtractive library and then performing a hybridization screening to eliminate clones which were not really differentially expressed (false positives), we obtained 168 highquality sequences which clustered in 127 tentative consensuses (Additional File 1). The microarray experiments instead identified 371 (223 upregulated and 148 downregulated) significantly modulated probe sets in the treated cells compared with the control ones (Additional File 2). Sequence annotation and classification according to Gene Ontology categories [23], revealed that at both time points primary (mainly signal transduction related genes) and secondary metabolisms, together with response to the stimulus, were the most affected categories (Additional Files 3 and 4). At 6 h, the analysis also highlighted downregulation of the cellular component organization and the biogenesis category (Additional file 4).
In general, the two experiments showed modulation of specific mechanisms had already occurred at 2 h and continued more extensively at 6 h after DIMEB treatment. The data summarized in Table 1 suggest that the grapevine cell responds to the elicitor by the activation of a signal transduction cascade which leads to the induction of specific classes of transcription factors. The downstream effect of this process is, on the one hand, the induction of some branches of the secondary metabolism and defence response, and, on the other hand, the blockage of cell duplication (Figure 1).
At 2 h the treatment caused positive transcriptional regulation of a grapevine gene (CLU090) encoding a protein with homology to an Arabidopsis kinase-associated protein phosphatase (KAPP) ( Table 1). KAPP protein may function as a signalling component in the pathway involving the serine-threonine receptor-like kinase, RLK5 Basic endochitinase precursor Putative S-adenosyl-L-methionine:salicylic acid carboxyl methyltransferase Q9C9W8 TC57170 x 1620309_at Putative S-adenosyl-L-methionine:salicylic acid carboxyl methyltransferase Pectin methylesterase Q96497 TC58800 x 1619468_at Pectin methylesterase PME1 Q94B16 TC53043 x 1619522_at Putative beta-galactosidase BG1  Q94B17  TC56838  x  1608756_at  Polygalacturonase-like protein  Q84LI7  TC59719  x  1606763_at Putative beta-1, 3-glucanase  Q8L868  TC67051  x  1609506_at  Putative cellulase CEL2  Q94B13  NP596365  x  1610263_at Putative beta-1,3-glucanase Q8L868 TC67051 x  Cell duplication  1612320_a_at  Tubulin alpha chain  P33629  TC57547  x  1616815_at  Tubulin beta-8 chain  Q41785  TC55048  x  1618413_at  Tubulin alpha chain  P33629  TC63601  x  1619167_at  Tubulin beta-8 chain  Q41785  TC62643  x  1621015_at  Alpha-tubulin 1  Q8H6M1  TC65238  x  1622466_at  Tubulin beta-8 chain  Q41785  TC62809  x  1608927_at  Putative histone H2A  Q6L500  TC53574  x  1612573_at  Histone H3  A2Y533  TC56731  x  1613041_at  Histone H4  Q76H85  TC61904  x  1613076_at  Histone H4  Q76H85  TC62637  x  1620332_at  Histone H3  A2Y533  TC59489  x  1622440_at  Histone H3  A2Y533  TC64779  x  1622737_at  Histone H2B  O22582  TC64405  x  1610854_at  Proliferating cell nuclear antigen  P22177  TC54817  x  1610422_at  Patellin-6  Q9SCU1  TC61622  x  1610607_at  Gip1-like protein  Q93WR4  TC66111  x  1613373_at  Formin-like protein 1  Q8S0F0  TC55249  x  1607792_at Putative DNA polymerase alpha catalytic subunit O48653 TC59012 x a Cluster or Affy ID of transcripts modulated at 2 or 6 h. (+) and (-) refer to up-and down-regulation in the treated sample with respect to the control. b UniprotID [73]of the first hit obtained by "Blast" analysis. c TC: corresponding grapevine Tentative Consensus sequence obtained by a search (BlastN) against the Grape Gene Index database [75]. of Arabidopsis [24]. In rice the RLK XA21 confers resistance to bacterial blight disease [25]. Other genes possibly involved in signal transduction showed overexpression at 6 h: a gene (1620080_at) with homology to a putative receptor-like protein kinase ARK1 of Oryza sativa and a gene (1611172_at) homologous to a Glycine max Salt Overly Sensitive gene encoding a SOS2-like protein kinase (Table 1). In Arabidopsis thaliana ARK genes seem to be involved in plant defence response to wounding and to bacterial infections [26], while SOS2 is a signalling kinase involved in salt tolerance response [27]. Phospholipidderived molecules are emerging as novel second messengers in plant defence signalling and phospholipases are key enzymes for their synthesis [14,28]. In the array experiment we observed the overexpression of a putative phospholipase gene (1608981_at), which may generate lipid messengers for the signalling response ( Table 1).
The activation of a signal cascade generally induces the expression of genes encoding for specific transcription factors, which in turn regulate downstream effector genes.
Two genes, upregulated at 6 h, showed homology to a hot pepper WRKY-b (1610775_s_at) and Arabidopsis WRKY11 (1611285_s_at) respectively (Table 1). WRKY proteins are plant-specific transcription factors whose expression is modulated in response to wounding, pathogen infection and abiotic stress [29]. Other classes of transcription factors appeared to take part in regulation of the response of grapevine cells to DIMEB treatment. The grape homologue (1619311_at) of a tomato pathogenesisrelated gene transcriptional activator PTI5 was upregulated at 6 h ( Table 1). This transcription factor binds to the GCC-box cis element present in the promoter region of many plant PR genes [30] and its upregulation could explain the observed induction of many PR proteins in this experiment. Another sequence (CLU059), induced at 2 h, which might modulate the expression of PR genes is the homologue of the tobacco bZIP TGA10 factor (Table  1). It has been reported that this protein can bind to the regulatory activation sequence-1 (as-1) [31] identified in the promoter of Arabidopsis PR-1 gene [32].
Our results indicated that one of the final grapevine cell responses to the DIMEB-elicited signal consists in the modulation of phenolic metabolism, especially stilbene and monolignol biosynthesis ( Figure 2).
Genes encoding enzymes involved in phenylalanine biosynthesis such as 3-deoxy-d-arabino-heptulosonate 7phosphate synthase (CLU083; 1611211_at; 1614440_at; 1619357_at; 1621405_at), 3-dehydroquinate synthase (1609646_at), prephenate dehydratase (1609932_at; 1621307_at) and chorismate mutase (1611895_at) were positively modulated both at 2 and 6 h after DIMEB treatment (Table 1). These enzymes participate in the synthesis of aromatic amino acids, particularly of phenylalanine, which is the link between primary and secondary metabolism, being a precursor of general phenylpropanoid metabolism. A recent report showed that cyclodextrins stimulates the expression of the structural genes of the general phenylpropanoids metabolism which sustains the synthesis of p-cumaroyl CoA, one of the two precursors of stilbenes [17].
The accumulation of stilbenes in the growth medium requires, besides stilbene biosynthesis, the presence of export machinery. In fact, induction of genes encoding putative secondary metabolite transporters, such as those belonging to the ATP-binding cassette (ABC) transporter family, was found. Genes encoding for pleiotropic drug resistance (PDR)-like ABC transporters (CLU106; CLU119), ABC transporter-like proteins (1613763_at; 1618493_s_at) and a CjMDR transporter (1610363_at) were indeed induced ( Table 1). The ABC transporters play an important role in some host-pathogen interactions [34]. In some pathogenic fungi they are involved in resist-Modulation of secondary metabolism at 2 and 6 h after DIMEB treatment ance to plant phytolexins and antifungal compounds, while in plants they seem to take part in plant defence response [34]. The induction of genes encoding glutathione S-transferase (1609330_at; 1611890_at) at 6 h correlates well with the ABC-mediated transport (Table 1). A glutathione moiety seems to function as a "recognition tag" for the transport of phenols [35]. Resveratrol translocation outside the cells has two main objectives: to mediate the defence response against pathogens and to avoid intracellular accumulation of this compound at cytotoxic levels.
The other branches of phenolic metabolism seemed not to be affected by DIMEB. Only two genes of the anthocyanin pathway (a chalcone-flavonone isomerase (CLU122) and a flavonol-3-O-glucosyltransferase (CLU048)) were induced at 2 h but not at 6 h (Table 1, Figure 2). Interestingly, selective induction of the early steps of phenylpropanoid metabolism and of the late steps leading to monolignol biosynthesis was also described in Arabidopsis in the early response to oligogalacturonide treatment [37].
The results strongly suggest that DIMEB acts as an elicitor modifying cell metabolism to promote the accumulation of phytoalexins and cell wall lignification. These two defence responses have been described as typical biochemical responses occurring in vegetal cells after elicitor exposure [14].
The transcriptional profiling results, however, show that the response to DIMEB seems to include other defence mechanisms. Overexpression of sequences for pathogenesis-related proteins such as chitinase (CLU088; 1613871_at; 1617192_at; 1617430_s_at), PR-10 (CLU001; 1610011_s_at; 1618568_s_at) and PR-4 (CLU021), but also for a prolin-rich protein (CLU036) and a protease inhibitor (1609875_at; 1611666_s_at) was observed in both experiments, while upregulation of two genes encoding the S-adenosyl-L-methyonine:salicylic acid carboxyl methyltransferase (1612552_at; 1620309_at) was recorded at 6 h (Table 1). Interestingly, this enzyme mediates the synthesis of gaseous methyl salicylate which was recently demonstrated to be a key mediator in plant systemic acquired resistance [38] in tobacco, as well as an inducer of the expression of PR-1 gene and TMV resistance [39]. This result strengthens the hypothesis that DIMEB acts as a true elicitor. The increase in the expression of a gene encoding for a 1-aminocyclopropane-1-carboxylate oxidase (1622147_at), would suggest the involvement of ethylene as well (Table 1). This hormone is a major regulator of the plant's reaction to pathogen attack [40] and via the action of a group of ethylene responsive factors it modulates the expression of plant defence-related genes such as, for example, phenylalanine ammonia-lyase, hydroxylproline-rich glycoprotein and acid class II chitinase [41,42]. It appears from the finding that a gene (1616358_at) homologous to an MLO-like 11 of Arabidopsis was downregulated at 6 h (Table 1), that the similarities between the cell's responses upon DIMEB treatment and upon pathogen attack are even greater. In barley, downregulation of the Mlo gene is involved in response to powdery mildew caused by the fungus Blumeria graminis f.sp.hordei [43], and in the dicot Arabidopsis thaliana, resistance to powdery mildews also depends on loss-of-function mlo alleles [44].
Our data support another effect of DIMEB on grapevine cells: blockage of the cell-division process. Upon treatment, we measured a lower expression of the genes involved in modification of the cell wall structure, cell division and microtubule organization. At 6 h, downregulation of genes related to cell wall modification [45], such as those encoding expansins (1608074_s_at; 1620840_at), xyloglucan endotransglycosylase (1615995_at; 1620003_at), pectin methylesterases (1608799_at; 1619468_at), a β-galactosidase (1619522_at), a polygalacturonase (1608756_at) and endoglucanases (1606763_at; 1609506_at; 1610263_at), was observed ( Table 1). The sequence 1609506_at corresponds to the VvCEL2 transcript which encodes a grapevine cellulase. Since in Arabidopsis the expression of the cel1 gene was related to growing tissues [46], downregulation of VvCEL2 could be related to repression of the cell growth. Microtubules play an essential role in cell division and cell elongation too. They set the cellular division planes and axes of elongation and influence the deposition and orientation of cellulose microfibrils [47]. The downregulation of genes coding for αand β-tubulin (1612320_a_at; 1616815_at; 1618413_at; 1619167_at; 1621015_at; 1622466_at) is indication of a stop in cell expansion and cell division (Table 1). mRNA degradation of a β-tubulin isoform was observed in soybean cells elic-ited by Phytophthora sojae-derived glucan fragments suggesting re-routing of the cellular resources towards the defence-related metabolism and repression of the cellular growth [48].
Further indication of cell division reduction were the lower transcription of genes coding for histones H2A, H3, H4 and H2B (1608927_at; 1612573_at; 1613041_at; 1613076_at; 1620332_at; 1622440_at; 1622737_at), a cyclin (1610854_at), a pattelin protein (1610422_at), a GA-induced-like protein (GIP-like) (1610607_at), a putative formin homology (FH) protein (1613373_at) and a DNA polymerase alpha catalytic subunit gene (1607792_at) ( Table 1). All these proteins are either related to DNA organization and synthesis or to the cytokinesis process. The down-regulated grapevine GIP gene is homologous to GIP-5 of Petunia hybrida, which is expressed during the cell division phase in stems and corollas [49]. In Arabidopsis patellin1 plays a role in membrane-trafficking when the cell-plate is formed during cytokinesis [50], and formins are plant cytoskeletonorganizing proteins which take part in cytokinesis and in the establishment and maintenance of cell polarity [51]. Very similar effects on cell growth have been reported upon elicitation of parsley cell cultures with an oligopeptide elicitor. Pep 25 provoked the repression of genes regulating the cell cycle, such as cdc2, cyclin and histones [52].
A likely explanation for the repression of cell division would be the need of the cell to use, almost exclusively, the transcription system as well as the available resources to establish a defence-related metabolism.

Conclusion
The transcriptional profiles measured at 2 h and 6 h after DIMEB treatment highlight the fact that this compound is able to induce an early and specific defence response in grapevine liquid cell cultures, supporting the hypothesis of its role as a true elicitor.
The classes of genes modulated by the treatment reveal that DIMEB triggers a signal transduction cascade which activates different families of transcription factors, in turn modulating the effector genes of specific metabolisms. These results thus suggest that in grapevine cells DIMEB induces a stop in cell division, reinforcement of the cell wall and the production of resveratrol and defence proteins ( Figure 3). This response largely resembles that occurring upon pathogen attack.

Plant material
Liquid cell cultures of a cross between Vitis riparia and Vitis berlandieri were used to carry out the treatment experi-ments with DIMEB (50 mM) [21]. Cell cultures were collected 2 h and 6 h after DIMEB treatment from control and treated samples. Cells and medium were separated by centrifugation at 12.000 ×g for 10 min at room temperature.

Total RNA extraction
Total RNA was extracted from control and treated samples using a modified hot-borate method, as described by Moser et al. [53]. DNA traces were removed by DNase I treatment (Sigma-Aldrich, St.Louis, MO, USA) according to the manufacturer's procedure. RNA was isolated from one replicate for the SSH experiment (2 h) and from 3 biological replicates for the microarray experiment (6 h).

cDNA synthesis and SSH library construction
Double-stranded cDNA was synthesized from 0.6 μg of total RNA of the control and treated samples (2 h) using the SMART™ PCR cDNA synthesis kit (Clontech Laboratories, Mountain View, CA) as recommended by the manufacturer.
Suppression subtractive hybridization (SSH) was carried out using the PCR-Select cDNA subtraction Kit (Clontech Laboratories) according to the manufacturer's procedure. The cDNA from the treated sample was used as the "tester" while the cDNA from the control sample was used as the "driver". Following hybridization, the subtracted cDNA molecules were inserted into a pCR ® 2.1-TOPO ® Vector (Invitrogen, Carlsbad, CA) and then used to transform One Shot ® TOP10 Chemically Competent Escherichia coli cells (Invitrogen). Positive transformants, based on blue/ white screening, were picked and arrayed in a 384-well plate containing LB medium (Sigma-Aldrich) supplemented with ampicillin (50 μg mL -1 ) and glycerol (10% v/ v). The SSH cDNA library was stored at -80°C.

Amplification of cDNA inserts and spotting on filters
The SSH library clones were cultured overnight at 37°C in a 384-well plate with LB medium and ampicillin (50 μg mL -1

Target labelling
To assess whether the isolated clones were truly positive, they were hybridized with the same total RNAs used for SSH library construction. The RNAs were DIG-labelled by reverse transcription according to Vernon et al. [54] with

Sequencing of transcripts identified by SSH
Following the screening procedure, the 289 positive clones were amplified, as described above for filter production, but without betain and Cresol Red in the PCR reaction mix. Five microliters of each PCR reaction were purified from primers and nucleotides using 1.  [55]. Electropherograms were analyzed with Phred [56,57] to assign a quality score and with a perl script using the UniVec Database [58] to identify any vector and adaptors sequences. Interspersed repeats and low complexity DNA sequences were identified through analysis with RepeatMasker [59]. The sequences were then organized in transcript consensus sequences (clusters) using the CAP3 DNA sequence program [60].

Affymetrix GeneChip experiments
Total RNA of the control and treated cells after 6 h of DIMEB treatment (3 biological replicates for each type of sample) were used to hybridize 6 different GeneChip ® Vitis vinifera Genome Arrays (Affymetrix, Santa Clara, CA). Ten micrograms of total RNA for each replicate were purified as described above (Total RNA extraction), subjected to further purification using "RNeasy" columns (Qiagen) and sent to an external service (IFOM-IEO Campus for ONCOGENOMICS, Milan, Italy) for labelling and hybridization. RNA samples passed the quality check as determined by electrophoresis run on a Agilent BioAnalyzer (Agilent, Palo Alto, CA, USA). Biotin-labelling, hybridization, washing, staining and scanning procedures were performed according to the Affymetrix technical manual. Analysis of raw data was performed using the open source software of the Bioconductor project [61,62] with the statistical R programming language [63,64]. The quality of the hybridization reactions was checked using the affyPLM package. Intensity distribution of PM for each chip and the quality of the 3 biological replicates of both control and treated conditions were analyzed with the functions and plots (histogram and MA plots) of the affy package [65][66][67]. Background adjustment, normalization and summarization were performed using gcrma and the affy package. Data, before and after application of the gcrma algorithm [68], were compared through the graphical representation of box-plots and MA plots. Probe sets which were not expressed or were non-differentially expressed between the two conditions considered were eliminated in a filtering step based on the inter-quantile range method (IQR = 0.25) using the genefilter package. A two-class paired SAM analysis (Δ = 0.9; FDR = 13.3%) [69] was performed using the probe sets resulting from the filtering procedure in order to identify differentially expressed probe sets between the control and treated conditions. A fold-change of two was then applied.

Functional annotation of the SSH transcripts and Affymetrix probesets
Protein sequences encoded by the SSH transcripts or by the representative sequence of each probeset as provided by the NetAffx Analysis Center [70] were predicted using a consensus generated by three different CDS predictors [71]. Blastp analyses [72] of the polypeptides obtained from the predicted CDSs were performed by searching against the UniProt database [73]. GO terms (molecular function, biological process and cellular component) [23] were linked at every consensus sequence on the basis of the results of the Blastp analysis (Additional files 1 and 2). The sequences were organized in main functional categories according to the GO term biological process (Additional files 3 and 4). In cases of non significant Blastp results (Evalue <1e-8; sequence alignment length <75% of the query polypeptide length), these were classified as "No hits found".
The SSH transcripts were deposited at the NCBI database [74] under the sequence IDs reported in the Additional file 1. Both SSH transcripts and probesets were also referred to corresponding Tentative Consensus sequences obtained by a search (BlastN) against the Grape Gene Index database [75] and to the corresponding genomic locus on Pinot Noir clone ENTAV 115 [76] (Additional files 1 and 2).

Real-time reverse transcription (RT)-PCR
To validate the SSH and microarray data, 12 genes and 5 genes identified by SSH and GeneChip array respectively, were also analyzed by quantitative RT-PCR experiments (Additional file 5). Specific primers were designed to generate 100-200 bp PCR products (Additional file 5). The actin gene (TC45156) was used to normalize the data (actin forward: 5'-TCCTTGCCTTGCGTCATCTAT-3'; actin reverse: 5'-CACCAATCACTCTCCTGCTACAA-3') since in preliminary trials it appeared to be constantly expressed in the RNA samples subjected to gene expression analyses.
For RT-PCR, total RNA from control and treated samples of the SSH experiment and from 3 biological replicates of control and treated samples of the GeneChip experiments were used. DNA traces were removed with DNase I treatment (Sigma-Aldrich) according to the manufacturer's procedure. Reverse transcription reactions and real-time RT-PCR reactions were performed using the SuperScript™ III Platinum ® Two-Step qRT-PCR Kit with SYBR ® Green (Invitrogen) according to the manufacturer's protocols with minor modification (300 nM of each primer in a final volume of 12.5 μl). PCR reactions contained 20 ng of cDNA and were replicated 3 times (technical replicates). Amplification reactions were performed with an ABI PRISM ® 7000 Sequence Detection System (Applied Biosystems). The following thermal profile was used: 50°C for 2 min; 95°C for 10 min; 40 cycle of 95°C for 15 sec and 55°C for 1 min. Data were analysed with the ABI PRISM ® 7000 SDS Software (Applied Biosystems). PCR reaction efficiencies were calculated with the LinRegPCR program [77]. For all the consensus sequences, the differential expression between treated and control samples was expressed as a ratio calculated with the Pfaffl equation [78]. The overall standard error of the mean normalized expression was obtained by applying the error calculation based on Taylor's series as developed for REST© software [79].

Data Availability
All microarray expression data are available at EBI ArrayExpress under the series entry E-MEXP-2114.