Analysis of P. viticola developmental stages
Infected tissues were examined under a microscope at 12, 24, 48 and 96 hpi, to determine the most suitable time-points for microarray analysis and to observe sporulation. The localization of zoospores over stomata at 12 hpi in both V. riparia and V. vinifera confirmed previous reports that zoospores can locate stomata with equal efficiency in susceptible and resistant species [27, 30]. Restriction of pathogen growth in V. riparia is a post-infection phenomenon that begins when the first haustoria enter mesophyll cells, resulting in the thickening of cell walls, necrosis of guard cells, the accumulation of phenolics and peroxidases, and in some cases a hypersensitive reaction depending on environmental conditions [9, 30, 47]. This correlates well with the specific induction of genes related to hypersensitivity and phenylpropanoid synthesis. Pathogen spread was severely impaired between 24 and 48 hpi in comparison to V. vinifera, suggesting that the resistance mechanism is already in effect before this time point, consistent with the strong transcriptional reprogramming observed at 12 hpi, when the first haustoria form.
Reliability of hybridization data
Because we used a V. vinifera microarray to assess differential gene expression in V. vinifera and V. riparia we performed experiments to confirm the reliability of cross-species hybridization. The successful outcome was not unexpected because V. vinifera arrays have previously been hybridized with RNA from other Vitis species [22, 23, 48]. Indeed, cross-species microarray hybridization is widely used in animals and plants [49–52], and although the data must be interpreted with caution, it remains a valid approach when dealing with groups of closely related species where sequence information is only available for one member . The average signal intensity and the number of absent calls in the hybridization data were similar in V. riparia and V. vinifera, and comparison of replicates within each species suggested a similar level of variation. This probably indicates that polymorphisms within each species provide nearly as much sequence variation as the differences between species, as previously shown by singlenucleotide polymorphism analysis . Moreover, the only direct comparison between V. vinifera and V. riparia was performed to assess differences in basal gene expression, while most of the comparisons were made between sampling time points in the same species, preventing such misinterpretation of hybridization results.
Interspecies differences in basal gene expression
The comparison of basal gene expression in healthy V. vinifera and V. riparia plants 12 and 24 h after a mock infection procedure revealed substantial variation in the expression of thousands of genes, but no overall bias towards either species.
V. riparia is a major source of resistance against P. viticola [4, 6, 13, 55, 56] and although major resistance genes have been identified  it has been suggested that some resistance may be conferred by constitutive differences in defense-related gene expression. We therefore focused on defense-related transcripts (resistance, stress, cell wall and secondary metabolism categories) to see if there were any broad trends. Although the levels of individual transcripts varied widely, overall levels were similar in the two species (Additional file 3).
The 'cell wall' category contained more transcripts expressed preferentially in V. riparia and the average signal intensity was also higher, but the differential expression of various cell wall enzymes did not explain how the modified cell wall might help to prevent pathogen spread. The 'resistance' and 'stress' categories, in contrast, included more transcripts preferentially expressed in V. vinifera. Many grapevine species accumulate stilbene derivatives, such as resveratrols and viniferines, in response to pathogens [57, 58] and we found that one stilbene synthase was preferentially expressed in V. riparia at 12 hpi, two were more abundant in V. riparia at 24 hpi, whereas five were more abundant in V. vinifera. Several PR protein genes were also more strongly expressed in V. vinifera, which is perhaps surprising because the genes are strongly induced by infection in V. riparia but not in V. vinifera. These data confirm that the response to P. viticola infection in V. riparia is not mediated by higher constitutive expression of defense genes and is essentially a post-infection process [26, 28, 30].
The absence of any significant differential expression of 'secondary metabolism' transcripts in pre-infection samples supports this conclusion, given that secondary metabolism, especially the phenypropanoid pathway, is often considered an important component of plant resistance . In a previous microarray-based comparison of a susceptible and a resistant V. vinifera cultivars, Figueiredo and co-workers  identified only 12 genes preferentially expressed in the uninfected resistant cultivar, one of which encoded phenylalanine ammonia lyase, whereas 17 genes were preferentially expressed in the susceptible cultivar. Other authors have reported that stilbene synthase and phenylalanine ammonia lyase mRNA are not detected in healthy leaves but are induced by infection or abiotic stresses, proportionally to the resistance phenotype observed and are therefore considered elicitor-induced responses [24, 25].
In the subset of transcripts showing differential basal expression at both time points, about 75% were more strongly expressed in V. vinifera and about 25% were more strongly expressed in V. riparia. When the analysis was restricted to defense-related transcripts the same broad trend was observed. Taken together, these findings suggest there is a stronger diurnal fluctuation in basal gene expression in V. riparia compared to V. vinifera, but provide no evidence that the resistance phenotype in V. riparia is caused by the constitutive expression of resistance genes maintaining a constant state of readiness.
Broad transcriptional changes associated with P. viticola infection
The infection of both species with P. viticola results in the rapid induction of many genes, although their number and the magnitude of induction are much greater in V. riparia (Figure 3). Transcript profiling in other grapevine diseases [23, 33, 36, 37] has focused on compatible interactions, for which large transcriptional changes are observed. The only incompatible interaction studied in this manner is that between V. aestivalis and the powdery mildew agent Erysiphe necator . This is another biotrophic, haustoria-forming grapevine pathogen, which might be expected to adopt strategies similar to P. viticola with similar consequences. In V. aestivalis only three genes were shown to be modulated by infection by E. necator. The same authors also investigated the compatible interaction with V. vinifera, which responded with a broad remodeling of the transcriptome. Our data show that both V. vinifera and V. riparia respond to downy mildew infection with a massive transcriptional change, which is much more pronounced in the resistant species as suggested by several large scale analyses of incompatible interactions in other species [60–63]. Many similarities can be identified between the responses against powdery and downy mildew in V. vinifera based on the annotation of probes on the chips, although a complete and detailed comparison cannot be carried out because different array platforms were used in each case.
Overlapping transcriptional responses to infection in V. vinifera and V. riparia
As expected, there were overlaps in the transcriptional changes in each species in response to infection, with most of the genes induced in V. vinifera constituting a weak subset of those induced in V. riparia at the same time-points (Figures 4 and 5). The limited response in V. vinifera appears to reflect an abortive attempt to achieve resistance, since most of the common modulated transcripts fall into the 'resistance' category (Figure 5). The activation of genes encoding PR proteins and enzymes in the phenylpropanoid pathway was anticipated based on data from model species [19, 59]. Interestingly, many of the common modulated transcripts are not only expressed at higher levels in V. riparia than V. vinifera, but also at higher levels than the genes in the same family that are uniquely expressed in V. riparia, e.g. PR-10, stilbene synthases and WRKY transcription factors. For example, the six WRKY genes whose induction is common to both species (TC59548, TC66456, TC71038, TC57604, TC53734, TC68615) are induced 6-22-fold in V. riparia, whereas those solely expressed in V. riparia are induced 2-5-fold (TC60897, TC51831, TC51732, TC53072, TC55553, TC64282). It therefore appears that V. vinifera can only weakly execute those responses that are strongly induced in V. riparia.
It is interesting to highlight the induction of an apoplastic invertase (TC56057), a sink-specific enzyme that catalyzes the irreversible cleavage of sucrose into hexoses, both in V. vinifera and V. riparia (2-3-fold and 7-9-fold, respectively). The rapid induction of invertase activity has also been observed in tomato roots resistant to the necrotrophic fungal pathogen Fusarium oxysporum . Likewise, in barley challenged with powdery mildew, an apoplastic invertase was induced more strongly and rapidly in a resistant cultivar . Hexoses produced by the invertase could be seen as a nutrient source for pathogens, but also as a supply of extra energy required for the activation of defense responses [66, 67] whose accumulation might suppress photosynthesis in line with our data on photosynthetic genes. Most importantly, sugar can also be used to trigger defense gene expression [68, 69] hence the suggestion to consider apoplastic invertase as a true PR protein .
All the common genes were modulated in the same direction by both species, indicating they probably fulfill the same functions in defense. Inverse regulation of the same gene in genotypes with different infection outcomes could be interpreted as part of a pathogen defense suppression strategy . Indeed, susceptibility to P. viticola is associated with broad downregulation of gene expression at later time-points  but our data show that such downregulation does not occur early in the infection.
Quantitative and kinetic differences between compatible and incompatible interactions have been elegantly described in Arabidopsis . The incompatible interactions produced a more robust and intense transcriptional response and the proposed quantitative model suggested that a high level input signal is generated in resistant plants in response to infection, determining the robustness of the system.
The specific transcriptional response in V. riparia
Although both species responded to infection with broad changes in gene expression, the response was strongest and fastest in V. riparia, with a peak of gene induction at 12 hpi. This response had transient and permanent components, since the expression of about half the genes fell back by 24 hpi (Figure 2). The strong transcriptional response of V. riparia together with its histological reactions to the pathogen is reminiscent of R-gene dependent resistance in other species , although the molecular determinants are unknown in this case.
When transcripts with unknown functions are excluded, the genes induced specifically in V. riparia fall into a number of functional categories whose expression appears to be coordinated. At 12 hpi, many genes encoding signal transduction components are induced, and this is followed by a wave of metabolic genes that are induced 24 hpi. This may indicate that an initial burst of signaling activity reprograms metabolism to provide a 'defense mode'. Among the different signaling pathways affected, calcium is known to be an important second messenger in resistance  as shown by the induction specifically in infected V. riparia, of calmodulins and calmodulin-binding proteins, calcium transporting ATPases, and proteins with similarity to calreticulin and calcineurin B-like proteins, all known to contribute to calcium homeostasis in the cell and to the definition of specific calcium signatures . Several different ethylene response factors are also strongly induced solely in V. riparia at 12 hpi, and this hormone has also been implicated in resistance . The possible involvement of ethylene in P. viticola resistance is further supported by the very strong induction of the ACC oxidase gene TC64623 (20-fold in V. riparia compared to only 3-fold in V. vinifera) and the 5-fold induction of an ACC synthase gene (TC60326) specifically in V. riparia.
Several genes with homology to known receptor-like protein kinases and leucine-rich repeat receptor-like proteins are specifically induced in V. riparia, especially at 12 hpi. These genes are known to mediate pathogen recognition and trigger defense responses in many species . Although the ligands for these receptors are unknown, hundreds of genes encoding receptor-like proteins have been identified in V. vinifera [12, 13], some of which map in linkage groups associated with resistance. Two MAP kinase kinase genes (TC62930, TC53469) were induced specifically in V. riparia at 12 hpi, consistent with the upregulation of three MAP kinases, two specifically in V. riparia at 12 hpi (TC66292, TC56256) and one also induced in V. vinifera at 24 hpi (TC61436). Interestingly, the TC66292 and TC56256 genes are related to Arabidopsis MAP kinase 3 (MPK3), the ortholog of tobacco wound-induced protein kinase (WIPK), which acts together with salicylic acid-induced protein kinase (SIPK) in resistance responses . The absence of a SIPK homolog among our induced genes is consistent with its predominantly post-translational mode of regulation .
Several families of transcription factors are also specifically upregulated in V. riparia, especially WRKY factors and other zinc-finger proteins. WRKY factors are regulated by interaction with MAP kinase in other species [78, 79] which provides a link in the signaling network we have outlined above. WRKY factors bind to DNA motifs known as W-boxes which are often found in defense genes, so they are regarded as important regulators of resistance .
It is well established that primary metabolic reprogramming underlies defense in biotrophic interactions and many genes in this category are specifically induced in V. riparia. Further analysis of our data suggests that specific pathways are involved: gycolysis (GADPH, enolase), the pentose phosphate pathway (glucose 6-phosphate dehydrogenase) and the Krebs cycle (pyruvate dehydrogenase, citrate synthases, succinyl-CoA ligase) are all induced, and could supply both energy and precursors for the biosynthesis of aromatic amino acids. Indeed, we observed the strong and specific induction of a group of genes controlling all the key steps in phenylalanine biosynthesis, including genes with homology to 3-deoxy-D-arabino-heptulosonate 7-phosphate synthases (6-30 fold at 12 hpi), chorismate synthase and mutase, and prephenate dehydratase, correlating with the induction of PAL (GSVIVT00013936001) and other genes involved in the hydroxycinnamic acid biosynthesis. Enzymes involved in lipid metabolism are also induced specifically in V. riparia. These include enzymes involved in lipid synthesis (e.g. acetyl-CoA carboxylase, β-ketoacyl-CoA synthase) and degradation (e.g. 13-lipoxygenase, acyl-CoA oxidase, acetoacetyl-CoA thiolase), and enzymes involved in the synthesis of jasmonates (omega-3 fatty acid desaturase, allene oxide cyclase, allene oxide synthase).
Genes encoding anti-oxidant enzymes and genes involved in protein degradation are also strongly and specifically induced in V. riparia, e.g. many RING-H2 domain proteins involved in ubiquitinylation are induced at 12 hpi. Interestingly, a rice RING-H2 protein associated with incompatible (but not compatible) interactions with Magnaporthe grisea is induced following treatment with different resistance-inducing chemicals, and transgenic plants constitutively expressing this gene are resistant to several pathogens, as well as drought and oxidative stress . This demonstrates how modulated transcripts identified in our experiments provide promising candidates for biotechnology-based disease resistance programs.
Surprisingly, 'resistance' as a functional category, is relatively poorly represented among genes expressed specifically in V. riparia, many of them instead being common to both species. However, as already stated, many of the common resistance genes are more strongly modulated in V. riparia, and the V. riparia -specific group does include a number of genes strictly related to hypersensitivity, such as those encoding rapidly elicited Avr9/Cf-9 proteins (e.g. TC63609, TC61603) , two hypersensitive-induced response proteins (TC63023, TC63883) and two homologs of known HR markers in other species - tobacco Hin1  and tomato hsr203J [45, 46] - both of which are specifically or preferentially induced in V. riparia at 12 hpi. The HR has previously been implicated in resistance response to downy mildew in V. riparia . Several additional defense genes are strongly induced in V. riparia, including those encoding PR proteins (such PR-4 and PR-10) and enzymes involved in the synthesis of antimicrobial compounds, as already reported in grapevine infected with powdery and downy mildew [23, 28].
The specific transcriptional response in V. vinifera
Although most modulated transcripts in V. vinifera are also modulated in V. riparia, there is a small collection of genes induced specifically in V. vinifera. The genes involved in this specific response do not suggest any coordinated and explicit mechanism related to the establishment of compatibility in V. vinifera. It is possible that the analysis of early transcriptional changes provides more information on resistance than susceptibility (the former involving a pro-active transcriptional response by the plant) and transcriptional changes associated with compatibility are established later .
Jasmonate levels in healthy and infected plants
Resistance to biotrophic pathogens is often dependent on salicylic acid-mediated defense responses . Jasmonates were originally associated with defense against herbivores and necrotrophic pathogens  but have more recently been implicated in resistance against biotrophes, such as powdery and downy mildews in Arabidopsis and in grapevine [84–87] and in resistance induced by BABA and by β-1,3-glucan sulfate against P. viticola [88, 89]. Jasmonates interact with other danger signals such as salicylic acid and ethylene to determine the ultimate outcome of an infection, in a manner dependent on the specific plant-microbe interaction. Our data support a role for jasmonates in establishing or maintaining V. riparia resistance against P. viticola, given the significant increase in the levels of both jasmonic acid and MeJA at 48 hpi only in this species, concomitant with the effective arrest of pathogen growth, although much later in comparison to the transcriptional reprogramming described above. More experiments are needed to determine the precise timing of this accumulation in relation to pathogen arrest and to reveal how much of the response to P. viticola can be considered jasmonate-dependent in grapevine.