Annotation of grapevine Gene Index

Annotation of grapevine Gene Index (VvGI version 5) using the MapMan ontology was performed by including information on grapevine genome and plant protein domains, the latter found in SwissProt/Uniprot plant proteins PPAP [32], the Conserved Domain Database CDD [33], Clusters of orthologous groups KOG [34] and InterPro [35]. Manual annotation was performed by different contributors for the BINs they have most expertise for. A manual correction usually consisted of blasting the appropriate tentative contig (TC) sequence, followed by classification using expert knowledge and a literature search. Altogether, 1728 manual corrections of automated annotation were made using this approach (Table 1). Because many genes from VvGI are expected to be grape-specific, special emphasis was put on genes present on either array and classified into BIN 35.2, where typically genes with no or only weak similarity to Arabidopsis and other plants included in MapMan are found. We were able to successfully annotate 13 TCs from this BIN. Some BINs, usually the smaller ones and the ones where a weaker emphasis was assigned, had a lower number/percentage of clones manually checked and corrected when necessary. Other BINs (e.g. BIN 16), which served as the basis of constructing new pictorial representations for phenylpropanoid, carotenoid and terpenoid metabolism were more thoroughly checked. A special case is BIN 20, where half of it was manually corrected due to recent changes in its annotation [7]. 34 TCs from the grape Gene Index were found to have high similarity to grapevine pathogen sequences, for example Ralstonia solanacearum or do not belong to Vitis vinifera and thus they were assigned to BIN 35.2 (not assigned. unknown). 24 out of these 34 TCs can be found in the more recent grape Gene Index (VVGI version 6).

To enable visualisation of results of transcriptomics experiments two final mapping files were created, one to be used with Operon microarrays and the other to be used with Affymetrix microarrays. Both final mapping files thus consist of the following data: BINcode, BINname, probe identifier from Operon/Affymetrix microarrays, respectively and gene description from grape gene index (with grape gene index and the protein domains information included). Due to specificities of grape physiology, we have created new pictorial representations of the grape mapping file focusing on three selected pathways: carotenoid pathway, terpenoid pathway and phenylpropanoid pathway.

Validation of mapping file for Affymetrix platform

The mapping file built for the analysis of data obtained with the Affymetrix GeneChip was validated on a dataset of genes differentially expressed during berry ripening, reported in [21]. This study analysed two pre-véraison (A, B) and one post-véraison (C) stages during Pinot Noir grape berry development along three years, identifying a set of 1477 genes conservatively modulated. They were manually annotated using the Gene Ontology vocabulary and grouped into biological process categories. Here, we have annotated the 1477 genes with the MapMan mapping file. Two processes important for berry development have been chosen for visualization in Figure 1 and 2: an overview of ripening regulation and the phenylpropanoid pathway. The representation of berry development regulation provided by MapMan immediately highlights the large number of genes involved in the berry developmental control (226), including 95 genes involved in transcription regulation, 49 in the hormonal metabolism and 88 in signalling and protein modification (Figure 1). A clear trend of prevalent gene induction from stage A to B and prevalent repression from stage B to C is evident. These observations are in agreement with the results reported in [21], in which these three classes (transcription factors, hormone metabolism and signal transduction) included respectively 125, 65 and 92 genes. The higher numbers are probably due to a wider functional classification and to annotation redundancy of the published analysis. The general trend of these classes is the same as reported in Figure 5 of [21], in which it is evident that all of them are prevalently induced from stage A to B and then mostly repressed. The main advantage of using MapMan is no doubt the ease and speed of the analysis.

The visualization of the genes distribution among the phytohormones confirms the large involvement of genes linked to auxins and ethylene and fewer genes involved in the abscisic acid, brassinosteroid and gibberellic acid pathways as previously reported [36–38], and as showed in Table 1 of [21].

Finally, the two categories of receptor kinases and calcium regulation (Figure 1), which were not investigated in detail in the published analysis [21], appear to be quite highly represented according to the MapMan annotation and, in agreement with [25]. Nonetheless, the two categories of light signalling and redox control, on which the published analysis had focused (see Table 1 and Figure 7 in [25]), include fewer transcripts and gene families, respectively.

This comparison shows that MapMan is a reliable annotation and data display tool which is easy to use and particularly efficient for representing genome-wide gene expression experiments. Obviously, when the interest is focused on a particular pathway or metabolism, further investigation is needed.

The pictorial representation of the general phenylpropanoid pathway is very effective, as it displays the information about the members of the gene families, the relative gene expression and the name of the enzyme in a single image. It is interesting to observe that before véraison the pathway is not modulated, with the exception of the flavonol synthase gene which is responsible for the flavonols biosynthesis. On the contrary during ripening, specific enzymes are strongly induced: phenylalanine ammonialyase (PAL), at the beginning of the pathway; stilbene synthase (SS), responsible for resveratrol accumulation; chalcone isomerase (CHI) and flavanone-3 hydroxylase (F3H), at the beginning of flavonoid synthesis; UDPglucose:flavonoid 3-O-glucosyltransferase (F3OGT), catalysing the final steps of colour accumulation. The picture clearly shows that, at the same time, the genes encoding the enzymes leucoanthocyanidin reductase (LAR) and anthocyanidin reductase (ANR) are repressed. Altogether these results suggest that in the ripening process the branches of the phenylpropanoid pathway which lead to the accumulation of stilbenes and anthocyanins are favoured with respect to those which lead to tannins production.

Overview of transcriptional changes in grapevine response to infection with Eutypa lata

20 Cabernet Sauvignon plantlets were experimentally infected with NE85-1 E. lata strain in three independent series. Seven weeks after infection, the plantlets with confirmed eutypiosis and healthy plantlets were collected for the microarray hybridisations each in two technical replicates. The symptoms consisted of a clear necrosis appearing few mm above the infected zone (Figure 3C). Symptoms sometimes extended beyond this zone, inducing leaf yellowing and necrosis (Figure 3A). Similar symptoms were obtained after infection of various grapevine varieties with either E. lata mycelium (symptoms observed after 6 weeks) or culture filtrate (symptoms observed after 4 weeks) [39].

Stress related responses in E. lata infected plants

In Figure 4A we can see that changes in expression of genes involved in cellular responses clearly indicate that the only genes induced belong to the category of biotic stress. The exceptions are the two genes classified to development corresponding to legumins, a gene family that has been frequently mentioned as up-regulated by biotic stress [40]. This is more specifically shown in Figure 4B showing biotic stress related genes which 107 out of the 312 strictly defined DE genes are annotated to. The strongly regulated biotic stress related genes include several PR-proteins (endochitinase (TC60929), chitinase (Q7XB39), osmotin-like proteins (P93621), thaumatin (Q7XAU7), disease response protein (Q45W75), tumor related proteins (P93378)), three 1,3-β-glucanases and several proteinase inhibitors. Early induction of genes encoding chitinases and 1,3-β-glucanases is a typical response of plants towards fungal pathogens. In the interaction between Cladosporium fulvum and tomato, resistance against the fungus correlates with early induction of transcription of genes encoding apoplastic chitinase and 1,3-β-glucanase and the accumulation of these proteins in inoculated tomato leaves [41]. It is additionally considered that genes encoding chitinases or β-1,3-glucanases are the most attractive candidates for the genetic manipulation approach to increase anti-fungal tolerance in grapevine [42]. Also strong induction of polygalacturonase inhibiting proteins (PGIPs, TC69081 CF604851 TC69081) was observed. In plant tissues, the activity of pathogen's PGs is counteracted by PGIPs, leucine-rich repeat proteins (LRR) located in the cell wall. A role for PGIPs in plant defence has been demonstrated by showing that transgenic Arabidopsis plants over-expressing PGIPs exhibit enhanced resistance to Botrytis cinerea [43].

Induction of phenylpropanoid pathway in eutypiosis

Secondary metabolism was also strongly induced after infection of grapevine by E. lata (Figure 5). Phenylalanine ammonia-lyase, PAL (P45735 and P45726), the first enzyme of the pathway, was up-regulated 1.5 to 1.7 fold. Likewise, chalcone synthase (Q8W3P6, O80407), chalcone isomerase (P51117), dihydroflavonol 4-reductase, DFR (P93799), and anthocyanidin reductase, ANR (Q7PCC4) which control the pathways leading to flavonols, tannins, and anthocyanins were up-regulated. Also the induction of secondary metabolites was shown to be associated with defence and pathogenesis related processes [44].

However, these results are not in complete agreement with those obtained by treating grapevine cell suspension cultures with the Eutypa pathogenic toxin, eutypine. In that experiment the authors observed a strong repression of the UDP glucose-flavonoid glucosyltransferase gene (UFGT) and no significant modulation of the chalcone synthase and DFR encoding genes [45]. A possible explanation of the incongruity can be different experimental systems used: complex response to the pathogen is most probably going to be different in a population of relatively uniform cells of a cell culture than in whole plants. In addition, the treatment with toxic compounds alone might not be directly related to effects of the real infection with pathogen.

Similarly, no changes in activity of respiratory metabolism related genes were observed in infected plantlets, while a model bioassay using the yeast Saccharomyces cerevisiae showed that the E. lata derived toxic metabolites inhibited the respiratory metabolism, causing the reduced growth of the cells. One possible explanation is that such a response of grapevine cells is masked due to the dilution of infected area with surrounding healthy tissue. The other explanation is that grapevine reacts differently from yeast and that primary target of toxic compounds is not yet determined [46].

Plant material

Cabernet Sauvignon in vitro plantlets were experimentally infected with the E. lata strain NE85-1 previously characterized as an aggressive strain [50]. Cabernet Sauvignon is a major variety used in various countries around the world and is particularly sensitive to E. lata. Infection was achieved by applying 10–15 days growing mycelium directly onto the cut surface of de-topped plantlets. Uninfected de-topped plantlets were used as healthy controls. Samples were further on characterized according to symptoms, re-isolation of the fungus, and formal identification of re-isolated E. lata mycelium by PCR. Foliar symptoms of eutypiosis were evaluated for each plantlet seven weeks after the experimental infection. 46 plantlets out of 60 infected plantlets were showing eutypiosis symptoms. The control plantlets [40] did not show any symptoms. After inspection of symptoms, each plantlet was frozen individually in liquid nitrogen and stored at -80 ‰.

Confirmation of infection

E. lata was re-isolated from the first internode of each plantlet. Internodes were surface-sterilized for 2 min with 1% sodium hypochlorite, rinsed in sterile water three times for 5 min, cut into two pieces and put on sterile PDA medium containing streptomycin (0.1 mg mL^-1). After incubation in darkness for 10 days at 22 ‰ the plates were visually assessed for the presence of typical E. lata cottony white mycelium. DNA was extracted from mycelium and E. lata was detected with PCR as described in [51]. The re-isolation and PCR detection of E. lata was used to check whether the non inoculated control was indeed axenic and to distinguish the experimentally inoculated plantlets that became infected from those that did not. Fungus was successfully re-isolated from all plantlets that were characterised as symptomatic and its identity confirmed by PCR. From the non-symptomatic plantlets no fungus could be isolated. The plantlets where infection was not successful were eliminated from further analysis.

Microarray hybridizations

RNA was isolated as described in [52]. To prepare fluorescently (cy3/Cy5) labelled antisense RNA (aRNA) targets, total RNA was amplified and purified using the Amino Allyl MessageAmp II aRNA Amplification Kit (Ambion, USA) following the manufacturer's instructions.

Array-Ready Oligo Set™ for the Grape (Vitis vinifera) Genome Version 1.0 designed and synthesised by Operon (Qiagen) was used for microarrays fabrication. Probes were spotted on epoxy mirror slides (Amersham, GE HealthCare) at the Montpellier Languedoc Roussillon Genopole, Institut de Génomique fonctionnelle. Just before hybridization, oligonucleotides were fixed on the slide by UV (254 nm) radiation of 120 mJ in a UV Stratalinker 2400-crosslinker (Stratagene). The slides were then washed twice in 0.2% SDS and positioned into the hybridization chambers. Approximately 4 μg of Cy3 and 4 μg of Cy5-labelled aRNA was mixed, fragmented, volume brought to 100 μL with hybridization buffer and denatured.

Hybridization was then conduced 16 h at 37 ‰ with medium agitation in the automated microarray station HS4800 Mastersystem (Tecan). Slides were washed sequentially with 1× SSC/0.2% SDS for 20 min, twice with 0.1× SSC/0.2% SDS for 10 min and finally with 0.1× SSC for 10 min. Washed arrays were quickly dried with nitrogen gas (2.7 × 10⁵ Pa) and immediately scanned with Genepix 4000B fluorescence reader (Axon Instruments, Canada) using GenePix 4.0 image acquisition software.

Statistical analysis of microarray results

Signals were extracted from microarray images using the Maia tool version 2.75 [53]. Median intensity gene expression data without background subtraction were normalized by a global lowess method followed by a print-tip median method with a modified version of the Goulphar script version 1.1.2 [54]. Differentially expressed genes were identified with the R/Bioconductor package Limma[55] using linear models and by taking into account technical and biological replicates.

Microarray data accession number

Microarray data analyzed in this study have been deposited in the ArrayExpress database. The accession number is E-MEXP-2102 [56].

Classification of TIGR Gene Index using the MapMan ontology

Vitis vinifera Gene Index sequences were downloaded from TIGR (VvGI version 5; 34,134 unique sequences). These were blasted (BLASTx, version 2.2.14) against Arabidopsis proteins release TAIR7 [57] under default settings. A RPS-blast against SwissProt/Uniprot plant proteins PPAP [32], the Conserved Domain Database CDD [33], Clusters of orthologous groups KOG [34] and InterProScan [35] was also performed.

The results of all searches were compiled into one table and using the Arabidopsis as well as RPS-Blast hits an initial classification into the MapMan BINs was achieved. This was then checked manually based on the annotations provided by TIGR, as well as based on the results obtained by the different similarity searches [7]. Furthermore several genes were annotated based on the expertise of the different contributors. The final mapping file has 35,246 entries, of those 34,132 are unique and the others are represented in two or more subBINs.

Grapevine specific BINs and subBINs organization

In order to get pictorial representations for the selected pathways (carotenoid, terpenoid and phenylpropanoid) BIN structures for the secondary metabolism had to be further refined. The carotenoid biosynthesis pathway was adopted from potato carotenoid pathway [58] with the help of already published MapMan representation for potato [7]. The flavonoid biosynthesis pathway was constructed based on [24, 59] where grape gene names, playing a role in flavonoid biosynthesis were directly used for creating our improved, grape-based flavonoid pathway. The existing pathway mapping file for terpenoid biosynthesis from the MapMan website [60] which included only the mevalonate pathway was modified in order to include the non-mevalonate pathway, too. Thus the present terpenoid mapping file includes both the mevalonate pathway and non-mevalonate pathways, which are essential for plant development [16]. Pictorial representations for all three pathways can be uploaded from the MapMan website.

Linking the mapping file to Operon and Affymetrix microarrays

In order to make the mapping file useful for studies involving grape microarray experiments, information from two microarray platforms was used, the Array Ready Oligo Set Vitis vinifera (grape) AROS V1.0 Oligo Set (Operon, Qiagen) covering 14,562 transcripts and the GeneChip^® Vitis vinifera Genome Array with 14,496 probesets (Affymetrix). Sequences from both microarrays were blasted against the grape gene index and grapevine genome sequences using the default parameters (BLASTN [61]). Since the first hit from this blast was typically by orders of magnitude (as shown with the E value) better than the other hits, only the information from the first hit was taken into account. Hits displaying less than 70% sequence identity were ignored. Subsequently all found blast hits were aligned to the corresponding oligonucleotide using the smith-waterman algorithm (as implemented in the "water" program of the EMBOSS package [62]), again discarding hits with more than 10% gaps and/or less than 70% sequence identity over the full length of the oligonucleotide sequence. The remaining hits were classified into MapMan BINs [63] using the manually curated mapping file for grapevine as a reference. If all hits for one specific oligonucleotide probe belonged to the same BIN (or multiple BINs) these BINs were assigned to the oligonucleotide probe. In case of disagreeing hits the BIN code "35.3 not assigned. disagreeing hits" was assigned unless manual inspection suggested a different BIN. For each oligonucleotide that showed multiple hits, the corresponding gene identifiers of all found hits were given in the description line in order not to lose potentially important information.

Berry development study

Details on experimental setup, lab work and statistical analysis of microarray results are described in [21]. Briefly, 50 berries of Vitis vinifera cv. Pinot Noir were randomly chosen from ten grape clusters harvested at the peak of acidity, two weeks before and three weeks after in the 2003, 2005 and 2006 seasons. Data representing individual years (2003, 2005 and 2006) were analyzed independently in order to obtain one set of differentially expressed genes for each season. Following stringent data analysis, a set of 1477 genes modulated during berry ripening was obtained. Genes were grouped into 17 functional categories based on GO 'biological process' terms annotation.