Vitis is a genus widely dispersed and with diverse taxonomy, yet practically most of the world’s commercial grape production is focused on a single species, Vitis vinifera L., which is native to Europe and Asia Minor. Vitis vinifera is a species highly susceptible to various economically devastating pests and diseases, such as powdery mildew. This disease has several causal agents depending on the plant host. In grapevine, the causal agent of powdery mildew is the ascomycete E. necator [(syn. Uncinula necator (Schweinf.) Burrill] (Gadoury et al., 2012; Dry and Thomas, 2015).

Originating from northern America, grapevine powdery mildew was recently discovered in extremely diverse climatic conditions, including temperate regions with high rainfall, especially during spring months (Pirrello et al., 2019). The causal pathogen is obligatorily parasitic on the genus Vitis, as well as on Cissus, Parthenocissus, and Ampelopsis within the Vitaceae family (Gadoury et al., 2012). Erysiphe necator can infect all green tissues of the host and cause significant losses in yield and reduction in berry quality (Pimentel et al., 2021). Due to the devastating effects of the disease, breeding studies have been initiated to develop varieties that are tolerant or resistant to this disease all over the world (Atak and Şen, 2021; Atak, 2022).

During the infection process, E. necator produces conidia that germinate and grow epiphytically on the plant tissue forming a germ tube and a lobed appressorium. This breaks the cell wall invading the underlying epidermal cells with haustoria, a feeding structure. Through it, the fungus retrieves nutrients and secretes effectors that suppress the plant’s immunity, PAMP (pathogen-associated molecular pattern)-triggered immunity (PTI), allowing the colonization of plant tissue surfaces by the development of secondary hypha. The newly formed conidiophores sporulate to infect other host tissues and start a new infection cycle, which leads to an effector-triggered susceptibility (ETS) within the host (Gadoury et al., 2012). As an answer, the plants react using resistance (R) genes that are related to their evolutionary history (Feechan et al., 2011). These genes encode mainly for NBS-LRR (nucleotide-binding site – leucine-rich repeat) proteins that regularly express an interaction of the effector-triggered immunity (ETI) type in which the NB-LRR proteins act as receptors interacting with the strain-specific effectors of the pathogen released during infection. This is likewise true for the R genes that are transcribed into the Vitaceae plant family after E. necator infection (Qiu et al., 2015). The interaction generates a signaling cascade that leads to transcriptional re-programming in the host plant. The R genes activate several defense responses, including programmed cell death, the generation of reactive oxygen species, biosynthesis/signaling of plant stress/defense hormones, phytoalexin biosynthesis, and cell wall strengthening (Agurto et al., 2017; Welter et al., 2017).

Powdery mildew threatens many commercially important grapevine species and varieties, and thus, nowadays, the most used and efficient method of control is based on chemical treatments (Dry and Thomas, 2015). The most suitable fungicides against E. necator are benzimidazoles, ergosterol biosynthesis inhibitors, the quinone-outside inhibitor (QoI) compounds (strobilurins, quinolones), and the succinate dehydrogenase inhibitor (SDHI) group. Since the majority of these fungicides are site-specific, their repeated use leads to fungicide-resistant isolates (Gadoury et al., 2012). Thus, the introduction of resistant cultivars represents the most promising strategy to reduce the use of fungicides in viticulture, avoiding the appearance of E. necator resistance isolates. Although all V. vinifera cultivars are highly susceptible to E. necator, several Vitaceae species belonging to various American and Asian genotypes have developed resistance mechanisms against this pathogen (Gadoury et al., 2012; Agurto et al., 2017; Schneider et al., 2019). The resistance quantitative trait loci (QTLs) in Vitaceae are clustered in tandem repeats of genomic areas that have been genetically mapped, revealing many loci that encode R gene sequences conferring resistance on E. necator and have been utilized to obtain resistant plants by pseudo-backcrossing (Agurto et al., 2017). The R genes identified in Vitaceae are named Ren (i.e. resistance to E. necator) and Run (i.e. resistance to Uncinula necator).

To date, 17 grapevine powdery mildew resistance loci have been identified and described (Sosa-Zuniga et al., 2022); a descriptive list of them is available online (www.vivc.de/loci). It is important to note, however, that the presence of only one gene or locus, even if it has a large effect, can favor the selection of fungus isolates capable of overcoming resistance (McDonald and Linde, 2002). In other words, if the resistance is based solely on the presence of a gene, the fungus may mutate and evade immune recognition through the emergence of new virulent isolates.

In this context, better and longer-lasting disease resistance would be beneficial (Merdinoglu et al., 2018) and a pyramiding technique that integrates multiple resistance loci in the same genotype has been proposed (Mundt, 2018) as a potential solution. To guarantee the longevity of this type of resistance, it is required that loci with different mechanisms of action, spectrums of target isolates and contributions (minor and major) to the resistance be combined. This approach should bring in a more improved, durable and secure implementation strategy, given that, if any mutation or virulence factor occurs, the pathogen will be still recognized by at least one R gene (Peressotti et al., 2010; Cadle-Davidson et al., 2011; Feechan et al., 2015; Pap et al., 2016; Agurto et al., 2017).

Understanding disease resistance or tolerance mechanisms against E. necator in grapevine cultivars with different resistant loci at various time points post-inoculation may provide a holistic interpretation of the incompatible interactions between Vitis and E. necator and provide valuable information for breeding programs. In this respect, characterizing the metabolic profiles associated with disease resistance and susceptibility represents a key step for the identification of trait-related biomarkers. As we have seen in our previous study (Ciubotaru et al., 2021), metabolomics provided novel insights into the resistance mechanisms underlying the hybrid-pathogen interaction by identifying 22 putative biomarkers of grapevine resistance to Plasmopara viticola. Thus, the aim of our study is to provide important metabolomics evidence by monitoring changes in the concentration of a large set of metabolites belonging to four chemical classes in grapevine leaves subjected to artificial infection with E. necator. The significance of these findings is important for experiments studying the different behavior of resistant (totally or partially) varieties and susceptible ones in terms of the biochemical mechanisms involved in disease resistance. A better understanding of resistance biochemistry may lead to an improved selection of resistant plants promoting the reduction of fungicide treatments.

In this sense, metabolomics provides a comprehensive and quantitative investigation of metabolites belonging to both primary and secondary classes, including metabolites that play an important role in fighting pathogens. Moreover, metabolomics studies can help in the identification of key metabolites in plant adaptation to biotic stress. Despite the broad interest in more sustainable agriculture, metabolomics studies performed so far have focused on understanding the mechanism of grapevine defense against downy mildew, while only a limited number of investigations focused on E. necator (Pimentel et al., 2021). Recent studies have shown the mechanisms underlying the synergy between metabolomics and various omics approaches (Maia et al., 2020; Pimentel et al., 2021; Sosa-Zuniga et al., 2022; Yin et al., 2022), the metabolic differences in the composition of the berries and leaves in several grapevine cultivars (Atak et al., 2021; Rienth et al., 2021) as well as control of the pathogen (Gur et al., 2022).

In this work, we focused on two mono-locus resistant genotypes (‘VRH 3082-1-42’- commonly named BC4 - and ‘Kishmish vatkana’) and four pyramided resistant genotypes (‘Bianca’, F29P92, F13P71, and NY42) comparing them with the susceptible cultivar (cv) ‘Teroldego’. To date, our current work is the first study that addresses the way E. necator induces metabolic changes in grapevine genotypes harboring one or more R loci.

The results of E. necator’s inoculation were phenotypically observed and the best infections (highest sporulation observed on the leaves) were obtained with the modified dry methods of Deliere et al. (2010) and Valdés-Gómez et al. (2011). The dry inoculation method provided more effective infections than the wet inoculation, most likely due to conidia germination being inhibited or reduced in the presence of water, which was reported to have a detrimental influence on the viability and infectivity of powdery mildew conidia (Miclot et al., 2012). Furthermore, high humidity has been demonstrated to have a severe negative influence on grapevine powdery mildew conidia germination (Carroll and Wilcox, 2003). The reduced efficacy of the wet inoculation is most likely due to residual water remaining in the leaves during or after the drying step. Pictures of the inoculated genotypes taken during the OIV-455 score evaluated at 3, 7 and 11 dpi are available as Supplementary Figures (2- 22 ).

Over a two-year period, we were able to identify and quantify/semi-quantify 177 metabolites from four chemical classes. These include 60 primary compounds, 56 volatile organic compounds, 43 phenolic compounds and 17 lipids. In the class of primary compounds, we quantified (26) acids, (13) amino acids, (3) amines, one gamma-butyrolactone, and (17) sugars. Within the lipids, we quantified: (2) glycerophospholipids, one sphingolipid, one glycerolipid, one prenol, and (12) fatty acids. We semi-quantified: (4) acids, (9) alcohols, (8) aldehydes, (6) benzenoids, one ester, (2) other volatile organic compounds, (3) fatty acids, (3) fatty acids esters, one fatty alcohol, one benzofuran, (8) terpenoids, (2) terpenes, (3) ketones, one secondary alcohol, and (4) unknowns for the organic volatile compounds. For phenols, we quantified: (3) benzoic acid derivatives, one coumarin, one dihydrochalcone, (12) flavan-3-ols, one flavanone, (12) flavonols, (3) phenylpropanoids, (8) stilbenes and stilbenoids and two other compounds.

The obtained concentrations of all investigated metabolites for each genotype in both years are presented in Supplementary Table 2 for primary compounds, in Supplementary Table 3 for lipids, in Supplementary Table 4 for VOC(s), and in Supplementary Table 5 for phenolic compounds.

In nature, plants protect themselves mostly through mechanical means (spines, trichomes, thick cuticles, and hard or sticky surfaces) and the emission of a variety of poisonous, repellent or unattractive compounds. Plants produce a wide range of metabolites through the latter protection strategy, including fundamental metabolites such as primary compounds and lipids, as well as secondary metabolites like phenolic and volatile organic compounds (Mazid et al., 2011). Secondary metabolism is known to play a defensive role against predators, parasites and diseases (Ali et al., 2010), and primary metabolism, in addition to controlling plant growth, development and reproduction, contributes to plant defense as a source of energy and by signaling molecules that directly or indirectly trigger defense responses (Wolfender et al., 2013).

In this study, we examined the contribution of secondary and primary metabolic components in mediating plant defense responses in resistant grapevine genotypes inoculated by E. necator. To the best of our knowledge, this is the first study to look at the responsiveness of multiple classes of metabolites in varieties with one gene of resistance versus varieties with multiple barriers of resistance against powdery mildew.

Our findings indicated that diverse grapevine genotypes react with different time scales to infection. Interestingly, metabolic response (primary and secondary) was more active in the partial and total resistant varieties (i.e., ‘Bianca’, F26P92, NY42, and BC4) and in the susceptible cultivar ‘Teroldego’ compared to the partially resistant mono-locus (‘Kishmish vatkhana’) and the totally resistant pyramided variety (F13P71) ( Table 1 ). ‘Bianca’, as well as ‘Teroldego’, showed metabolic variability caused by pathogen inoculation at both time points, while F26P92 and NY42 showed metabolic variability only at 24 hpi. This could be explained by the studies of Feechan et al. (2015) and Pap et al. (2016), which indicated that the existence of several resistance genes or loci does not result in a stronger resistance response for all genotypes, thereby suggesting that combinations of loci such as Ren3Ren9 do not always have additive effects (Zendler et al., 2020) when compared to the Run1Ren1 combination that produces an additive effect (Agurto et al., 2017). The method of activating a gene is complicated since just having the gene is not enough; instead, transcription factors are required (Agurto et al., 2017).

Such responses have been observed in some genotypes carrying the combination of Ren3 and Ren9, which did not generate an immune response that has an advantage in terms of the intensity or speed of the response compared to Ren3 alone (Zini et al., 2019; Zendler et al., 2020). The presence of these loci (Ren3 and Ren9) in all three of the pyramided genotypes, ‘Bianca’, F26P92, and NY42 ( Table 1 ), could explain our PCA results, which revealed that ‘Bianca’ had a metabolic variability caused by pathogen inoculation at both time points similar to ‘Teroldego’, followed by F26P92 and NY42, which showed metabolic variability only at 24 hpi ( Figure 3 ).

On the other hand, studies showed that combinations of Run1Ren1 and Run1Ren2 have an additive effect as the combination of both genes/loci generated a stronger immune response than the one triggered by each one individually, however, this effect has been proven to be genotype dependent (Agurto et al., 2017). In fact, in our study, the genotypes F13P71 and NY42 showed little to partial metabolic variability despite possessing the loci Run1Ren1 and Run1Ren2, respectively. Moreover, other studies showed that by powdery mildew isolates could overcome, in some cases, Run1 resistance (Cadle-Davidson et al., 2011; Schneider et al., 2019). This could explain the observed metabolic variability in F13P71. Furthermore, the additive effect of Run1Ren2 can be race-specific (Feechan et al., 2015) and in addition, the existence of the other two extra loci in the genotype NY42, might interfere with the metabolomics response to the pathogen. All these factors contribute to the complexity of the effects of resistance genes in the metabolic variability of infected grapevine genotypes, requiring additional research.

Considering all these aspects, it seems that the level of resistance (partial or total) of the loci is more important than their numbers. The level of resistance is referred to as “total” when there are greatly suppressed symptoms or no observable symptoms of infection at all and “partial” when there is a decrease in symptoms but not a complete disappearance (Julius Kühn-Institut, 2022; Sosa-Zuniga et al., 2022). This is corroborated in our study by the assessment of the OIV-455 descriptor at 7 days after the artificial infection ( Supplementary Table 1 ).

We found for ‘Kishmish vatkana’, a genotype with partial resistance, a high level of resistance (OIV-455 = 7) and for F13P71, a genotype with total resistance, a very high level of resistance (OIV-455 = 9). Indeed, an 84% decrease in the number of cells the fungus invaded has been observed among the responses brought on by Ren1 (‘Kishmish vatkana’). Other reactions include the induction of PCD (programmed cell death), the development of callose deposits at 48 hpi, and the promotion of ROS (reactive oxygen species) at 96 hpi (Agurto et al., 2017). A more intense defense response was likewise observed in genotypes carrying Run1Ren1, such as F13P71, in terms of ROS production, callose accumulation and PCD (Agurto et al., 2017).

NY42 and F26P92, genotypes with partial resistance, scored a high level of resistance (OIV-455 = 7) and BC4, a genotype with total resistance, was assessed as having a very high level of resistance (OIV-455 = 9). Possamai et al. (2021) observed in genotypes carrying Run1Ren2 loci such as NY42 a significant decrease in colony formation, and Feechan et al. (2015) showed that Ren2 confers partial resistance on plants by inducing an efficient immune response that prevents fungal sporulation. Rapid programmed cell death, which hinders the growth of secondary hyphae and sporulation, is one of the immunological responses inflicted by Run1 (BC4) on resistant plants. A quick HR is seen at 48 hpi in cells where the fungus developed secondary hyphae as evidenced by the rise in ROS and the appearance of PCD. The buildup of callose deposits at the E. necator infection site is another reaction caused by Run1 (Agurto et al., 2017). Ren3Ren9 (F26P92) elicits similarly high resistance responses (Zendler et al., 2020), with a high level of resistance score (OIV-455 = 8) assigned to ‘Bianca’, a partial resistant genotype carrying the exact same loci (Ren3Ren9). As expected, the susceptible genotype ‘Teroldego’ was assessed as having a very low level of resistance (OIV-455 = 1). As far as primary metabolites are concerned, powdery mildew induced changes mainly in the class of lipids ( Figure 4 ). Lipids are recognized to be important components of plant cell membranes that provide energy for metabolic activities. In recent years, there has been increasing evidence that lipids play a role in combating biotic stress, such as powdery mildew. Lim et al. (2017) showed that lipids regulate the PCD response during pathogen defense, as well as membrane fluidity, stability, and permeability during plant responses to microbial pathogens. The accumulation of C16:0 might be used to produce C18 fatty acids. Also higher DBI may account for an increase in chloroplasts’ membrane fluidity that may be crucial to avoid any damage in the photosynthetic machinery with inevitable effects on the energy transduction pathways and primary productivity (Laureano et al., 2018; Laureano et al., 2021). Moreover, lipids play important signaling roles also in plant defense and ROS regulating levels. Because of the various functions of lipids, Della Corte et al. (2015) observed that their abundance in plants is influenced by genotype and phenotype. Thus, the fluctuating lipid levels observed in the various resistant genotypes tested may be attributed in part to this aspect as a result of E.necator inoculation.

The role of primary metabolic pathways in the regulation of plant defense responses is not very well known. Mainly, all primary compounds function as signaling molecules that trigger defense responses through signal transduction and pathogen recognition processes (Madiha et al., 2019). The accumulation of the primary compounds in our study, which was comparable to the susceptible genotype, made us lend support to the idea that susceptible plants initiate a basal defense similar to the response in resistant plants, but insufficient in timing and/or intensity to limit disease progression, as observed by Marsh et al. (2010). Similarly, there is a clear alteration of primary compounds in the defense against powdery mildew in resistant genotypes, but the amount raises the question of whether this modulation is a result of resistance or a normal plant reaction.

One of the most important functions of phenolic compounds as secondary metabolites is an antibacterial activity in plants, which acts as a barrier against pathogens like E. necator. Their accumulation in plants is associated with host resistance (Atak, 2017). However, it is noteworthy that Keller (2015) found that despite some Vitis species (such as V.cinerea and V.champinii) exhibiting pathogen resistance, the buildup of stilbenes, the most well-known class of phenolic defense chemicals, did not occur in these plants. This finding could support the hypothesis that metabolite accumulation is not totally linked to the number of loci present in the resistance genotypes. Such was the case in our study where the pyramided genotype F13P71 accumulated very low levels of phenolic compounds. The same genotype displayed low levels of volatile organic compounds (VOCs). Thus, similar assumptions could be made also about VOCs, but further research is needed to confirm it.

Although some chemical classes in some of our resistant varieties showed similar reactions to the susceptible genotype, it should be noted that the resistant genotypes nonetheless produce a number of up-accumulated metabolites that were not found in the susceptible ‘Teroldego’, with the exception of a few whose calculated effect sizes were smaller than in the resistant genotypes. The study of Viret et al. (2018) showed that the induction and accumulation of defensive metabolites increase only during the pathogen’s infectious structure development, which takes around 24 hours (Boddy, 2016). This was noticed in the pyramided genotypes in which the metabolite overaccumulation had a significantly larger impact size at 24 hpi than at 48hpi, when their modulation appeared to subside, except for ‘Bianca’. In our earlier research, we provided evidence that P. viticola caused an early modulation in pyramided genotypes, which began earlier, between 0 and 12 hpi, and peaked at 48hpi. Even though the current work studies E. necator and genotypes that carry different loci than the prior study, we can presume that a similar but somewhat different reaction happened for this study as well. On the other hand, the up-accumulation of metabolites in mono-locus genotypes was shown to be established at 24 hpi and to become stronger at 48hpi. The same finding was obtained in the work of Chitarrini et al. (2017a), in which the plant defense systems were activated 48 hours after inoculation.

Investigating the biological relevance of the ten compounds found as discriminative between resistant and susceptible genotype ( Figure 7 ), we found out that pallidol, oleic acid+cis vaccenic acid and astringin were already discussed as potential biomarkers of resistance in our previous study (Ciubotaru et al., 2021) due to their role in activating plant defense response. Moreover, pallidol has been in some cases linked to the grapevine’s response to fungal attack (Pezet et al., 2004; Jean-Denis et al., 2006). We have also found that the remaining seven- up-accumulated metabolites contribute to plant defense. Isorhapontin, like pallidol and astringin, belongs to the class of stilbenes and stilbenoids, and it has been demonstrated that this class accumulates in larger concentrations in disease-resistant cultivars than in susceptible cultivars, due to its role in plants that inhibits fungal growth (Chitarrini et al., 2017b; Vezzulli et al., 2019). Similarly, the increased accumulation of fatty acids in the plant metabolome, specifically behenic acid, palmitoleic acid, arachidic acid and oleic acid+cis vaccenic suggests that these fatty acids are involved in intracellular signaling processes as well as desaturases-mediated membrane fluidity adjustment (He and Ding, 2020; Ciubotaru et al., 2021). The fatty acid desaturase 7 (FAD7) and fatty acid desaturase 8 (FAD8) genes, which play a key role in the synthesis of fatty acids, have also been linked to protective mechanisms (Rojas et al., 2014; Cavaco et al., 2021). Last but not least, oleanolic acid is known to play a role in plants’ defense mechanisms against pathogens and water loss (Gudoityte et al., 2021), whereas quercetin is a powerful antioxidant that effectively protects plants from a variety of biotic and abiotic challenges (Singh et al., 2021). Our findings confirm previous research about the importance of these compounds in disease resistance because of their different roles in plant defense.

Many metabolomics studies have been conducted on understanding the mechanism of grapevine defense, mainly on downy mildew, but few on powdery mildew. Thus, we designed this study as a promising endeavor in order to contribute to a better understanding of plant defense mechanisms. To our knowledge, this is the first time that metabolic investigations of the most important classes of compounds with a role in plant defense were carried out in artificially inoculated genotypes with mono-locus and pyramided resistance in order to characterize the host’s response to the infection of E. necator.

Overall, the results of this study indicate that how cultivars behaved to pathogen attack can be linked to genotype and/or resistant loci differences; however, resistance is not exclusively related to Run/Ren loci. Additionally, although it cannot be strictly classified as a connection, we saw similar metabolomic responses in our experiment between the mono-locus and pyramided genotypes that share the exact Run/Ren loci. Therefore, additional transcriptome studies are required to fully comprehend the unfavorable interaction between these resistant loci and E. necator. Further research is needed also to validate the molecules identified as biologically relevant compounds produced during the pathogen-host interaction and recommended as possible biomarkers for resistance to E. necator. In terms of plant resistance strength against powdery mildew, our findings show no direct relationship between the number of resistance loci present in plants and the production of metabolites recommended as resistance biomarkers.

The findings of this study add to our understanding of plant defense mechanisms and call for more metabolomics research, as well as additional complementary omics research to clarify which genes are responsible for powdery mildew resistance and how they function in the majority of Run and Ren loci, as only one study in this area has been conducted. The integration of transcriptomics and metabolomics data can be exploited to uncover commonalities and differences between diverse R-gene-mediated resistances to E. necator. This approach will enable breeders to choose more reliable genotypes for marker-assisted breeding by using genetic and biochemical markers.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

RMC, GC, LZ, MS, and UV designed the experiment. MS and SV provided the plant material. RMC, SV and LZ performed the experiment. RMC did the extractions and analytical analysis. PF, RMC, and UV conducted the data treatment, statistical analysis and data visualization. RMC wrote the original draft preparation of the manuscript. UV, SV, PF, MO, PR and GC did the review and editing. UV, MO, and PR supervised the project. All authors discussed the results and implications and commented on the manuscript at all stages. All authors contributed to the article and approved the submitted version.