Several Caucasian V. vinifera were recently described to be resistant to E. necator (28; www.INNOVINE.com). A preliminary survey did not identify a genetic relationship of these accessions to known sources of resistance (data not shown) and suggested that unexplored resistance determinants could be present in the Caucasian germplasm. With the aim of investigating such a hypothesis the Caucasian resistant varieties ‘Shavtsitska’ and ‘Tskhvedianis tetra’ were crossed with the susceptible varieties ‘Glera’ and ‘Chardonnay’, respectively. Phenotyping bioassays were performed on the cross parents and offspring. The two Caucasian varieties showed a partial resistance to E. necator that segregates in the progeny and resulted as being controlled by a major resistance QTL located in chromosome 13.
Caucasian grape varieties show a resistance to E. necator
The Caucasian grapevines showed both similar and different resistance phenotypes in response to E. necator in comparison with the varieties carrying known resistance genes.
‘Shavtsitska’ and its offspring contrasted the pathogen hyphal growth early, which at 1–2 dpi resulted strongly delayed compared to the susceptible control ‘Cabernet sauvignon’ and the susceptible cross parent ‘Glera’ (Fig. 7; Additional file 8: Table S4a). Previous studies reported substantial differences at 2 and 3 dpi for E. necator development on Vitis accessions carrying different resistance sources [29–31]. Run1 gene was described to halt the E. necator conidia penetration and hyphae elongation early at 1–2 dpi, reacting at the infection sites through a programmed cell death (PCD) deployment, callose depositions and reactive-oxygen-species (ROS) production [29, 30]. ‘Kishmish vatkana’ and other varieties carrying Ren1 were described to delay powdery mildew infection , activating the plants reactions with lower intensity and later in comparison to Run1-mediated resistance . Zendler et al.  observed that Ren3/Ren9 effects on E. necator development were evident from 5 dpi. Our results, for the control genotypes ‘RV1-22-8-78’ (carrying the Run1 gene), ‘Kishmish vatkana’ (carrying the Ren1 gene) and ‘Johanniter’ (carrying the Ren3 and Ren9 genes) agree with previous studies and allow it to be speculated that ‘Shavtsitska’ has an effective ‘post-penetration reaction’ to E. necator: the variety does not halt the pathogen growth and show the HR response, associable to a plant PCD  deployment, beneath the appressoria of both conidia and hyphae (Fig. 7). Preliminary studies suggest that such a reaction could be the key one for ‘Shavtsitska’ to control E. necator because the callose depositions appeared very limited in the Caucasian variety in comparison to the control varieties carrying the Run1 and Ren1 genes (data not shown).
According to SEM observations, E. necator produced larger multilobed appressoria from conidia and multiple new appressoria from hyphae on the studied resistant varieties (Fig. 8). Expanded and frequent appressoria in powdery mildew infections were already observed in resistant Vitis spp. accessions [32, 33]. Thus, such events confirmed that the pathogen on the studied resistant plants has more difficulties in establishing effective interactions due to the host response. SEM images also revealed that the conidia falling on prostrate airs of the leaves fail to develop mycelium (Fig. 8). Leaf hairs can influence pathogen infections, acting as a physical barrier or influencing the leaf micro-environmental conditions . While a role of trichomes was often proposed in favouring grape resistance to P. viticola [3, 35, 36], no reference was found about their possible effects on the foliar resistance to E. necator. Our conclusions on this topic need to be confirmed because among all studied accessions only ‘Shavtsitska’ showed a high density of prostrate hairs.
‘Shavtsitska’, ‘Tskhvedianis tetra’ and their resistant offspring showed a partial resistance to E. necator. It means that the pathogen was able to complete his lifecycle but its development is contrasted by the host: E. necator mycelium growth was slowed down from 1–2 dpi and restricted, sporulation was delayed to 7 dpi and was limited to 2–3 conidia per conidiophore at 10–11 dpi. Susceptible genotypes did not influence the mycelium growth, sporulation appeared at 4–5 dpi, reached 5–6 conidia per conidiophore and produced, on average, 2 times more conidia at 10–11 dpi (Additional file 8: Table S4b and Table S4c). The resistance observed in the Caucasian accessions is not as effective as the genotypes carrying either Run1, Ren5 or Ren6 genes, which halt the pathogen hyphal growth and sporulation, therefore offering total resistance [29, 37, 38]. The partial resistance of the study has similar effects to those observed in genotypes carrying the Ren1 and Ren7 genes [18, 37, 39] and, finally, combination of the resistance components resulted in a significantly lower severity of the disease and capability of E. necator to establish either serious or new infections.
In our populations the resistance segregation was neither qualitative as suggested for Run1, Ren4 and Ren6 loci [16, 21, 37] nor quantitative as observed, for example, on V. rupestris . The bimodal distribution of phenotypic scores (Fig. 1) suggests the presence of a major genetic factor for the trait under observation; while the occurrence of more continuous resistance degrees would suggest the presence of further complementary determinants.
A major QTL control the resistance to E. necator in Caucasian grapevines
The GBS approach  performed very well in our study despite the challenges represented by high grape heterozygosity, which might generate erroneous SNP calling, high-percentages of missing data and heterozygote under-calling [40, 42]. The two parental maps, each of about 1,250 cM and 2400 markers, divided in 19 LG (Fig. 2; Table 2; Additional file 4: Figure S3a and Figure S3b) had a marker order consistent with the grape reference genome (Additional file 5: Figure S4)  and agree with previous GBS-derived linkage maps on the length of linkage group and marker density [e.g. 40, 44, 45]. However, ‘Glera’ chr 13 displayed distorted segregations for all markers (Fig. 3). Islands of markers with distorted segregations may be common in interspecific crosses (e.g. 45, 46), but they were also observed in crosses between V. vinifera cultivars (e.g. 47). Distorted segregations may be unpredictable and occur because of post-zygotic lethal combinations that influence the viability of zygotes, germination of seeds and seedlings survival . In our case an haplotypic region on the lower part of chr 13 of ‘Glera’ was defective and not inherited in the progeny, a phenomenon that did not occur with ‘Shavtsitska’. These pieces of evidence may suggest the presence of a new locus responsible for the gamete selection in V. vinifera in addition to the ones described by Riaz et al.  on the chr 14. Markers with distorted segregations can determine spurious linkage, erroneous marker order and imprecise QTL analysis . However, maintaining only SNP not affecting the marker order and distances compared to grape reference genome ‘PN40024’ allowed the genetic map of ‘Glera’ chr 13 to be completed.
All approaches adopted in the QTL analysis (interval mapping and multiple QTL research) identified a single major locus for resistance to E. necator on chr 13 of ‘Shavtsitska’ (Fig. 4; Table 3; Additional file 6: Table S2). The phenotyping data were all performant in mapping the QTL: the LOD peaks scores were different depending on variable and dpi considered but they were always significant and located the QTL in the same interval of 2.2 cM at 47 cM on the LG (Fig. 4; Table 3; Additional file 6: Table S2). The rAUDPC indexes, which summarize the infection progress for pathogen mycelium growth and sporulation, resulted as being the most informative data and explained up to 80% of the phenotypic variance (Table 3). In QTL mapping, the methods of phenotypic data collection, which comprise standardized sampling, handling, infection processing and rating, are as important as the genetic design and analysis. Our results were constant and reproducible and evidenced the effectiveness of the phenotyping strategy in describing the phenotypic patterns as well as the genetic of the trait studied. The properties describing the locus identified in ‘Shavtsitska’ (Table 3), which were further confirmed by investigations on the ‘Tskhvedianis tetra’ cross population (Fig. 6), showed that the QTL is a promising source of resistance to E. necator. It will possibly be important to test the Caucasian resistance with various pathogen strains and experimental conditions (e.g. environments) to understand and confirm the solidity and performance of the gene.
The SNP flanking the locus of resistance of ‘Shavtsitska’ were positioned in chr 13 at 16.8 and 18.2 Mb on the ‘PN40024’ reference genome  (Fig. 5). There are currently 14 known QTL associated with resistance to E. necator and just another locus is located in chr 13 (www.vivc.de). Overlap the ‘Shavtsitska’ QTL, in an interval of 7.1 Mb, Hoffman et al.  identified the locus Ren1 that was mapped starting from the ‘Kishmish vatkana’ SSR-based genetic map. Subsequently, the locus was further saturated with SSR markers and Ren1 was delimited to an area of 1.4 Mb on the grape reference genome . In our study, we report the first high-density genetic map of a V. vinifera variety resistant to E. necator and demonstrate the power of the GBS approaches for QTL mapping and quickly narrowing a region of interest .
We showed that the resistance to E. necator of Caucasian grapevines is coded by a major and effective gene. On the contrary, resistance to P. viticola in Caucasian germplasm appeared to be controlled by three different minor loci . Both the introduction/pyramiding of major, in particular, and minor loci are important to define promising and durable resistance traits. Our results therefore strongly increase the interest in Caucasian grape accessions for breeding resistant grape cultivars . Caucasian varieties carrying resistance to both P. viticola and E. necator may be the most valuable germplasm. The cross-checking of the results of our paper and that of Sargolzaei et al.  did not show varieties carrying a resistance to both pathogens, but an analysis at a larger scale need to be carried out.
Origin of the resistance to E. necator in the Caucasian grapevines
The screening of ‘Shavtsitska’ and ‘Tskhvedianis tetra’ populations with the SC8-0071-014 and Sc47_20 markers , revealed that the allele 149 of SC8-0071-014 and allele 208 of Sc47_20 are in coupling with the resistance to E. necator and present in both Caucasian parents and their resistant offspring. We extended the SSR analysis to 103 Caucasian grapes preserved at the CREA-VE germplasm repository, discovering that the haplotype 149–208 was shared by twenty-four varieties (Table 4; Additional file 7: Table S3). These results suggest that resistance to powdery mildew could be very frequent in Caucasian grape germplasm. We could not phenotypically characterize the Caucasian grapevines but the literature reported eleven of those varieties as partially resistant to E. necator . For six of those grape accessions our molecular analysis showed the presence of the Caucasian resistant haplotype in chr 13 (Additional file 7: Table S3). Other five phenotypically resistant accessions did not share the same haplotype and this would suggest a more complex genetic landscape behind the resistant trait. However, only an extended phenotypic survey in the same experimental conditions, together with the collected molecular data, could provide clearer insights into the spread of resistance to E. necator in the Caucasian cultivated germplasm. Currently, our genetic findings are consistent with the Riaz et al.  study, which identified in the same genomic region of chr 13 of a Caucasian V. vinifera subsp. sylvestris a resistance QTL to E. necator.
Our research, in addition to other studies, identified the resistance to E. necator in many V. vinifera grapevines of different geographic areas (Caucasus and Central Asia) and collected evidence that its inheritance is shared by wild and cultivated V. vinifera subspecies [18, 19, 24, 50]. This information and the long history of grapes isolation in the Caucasus region [51, 52] suggests that the resistance trait might have been inherited from a V. vinifera progenitor thousands of years ago and conserved in Caucasian cultivars until today. In the ancestor/s, probably, the region evolved to fight different fungi-caused diseases, conserving an array of R-genes over time . Maintenance of the trait in V. vinifera through domestication and until today was probably not intentional. The literature supports the hypothesis that E. necator co-evolved in North America on native wild Vitis spp.  and does not report powdery mildew disease in Europe and Asia before the 19th century . It is less likely that the resistance developed recently in the Caucasian germplasm because the historic time of co-evolution between local grapevines and the pathogen has been too brief. The resistance haplotype does not appear to result from an interspecific introgression into ‘Shavtsitska’ and ‘Tskhvedianis tetra’, possibly through a chance cross with an American grapevine introduced in the area, because the Caucasian accessions showed purely V. vinifera genomes in resequencing studies (Magris G., Di Gaspero G., Morgante M. pers. comm.). However, we cannot exclude that either natural or intentional selection took place in the region last two centuries [55, 56], when the pressure of E. necator on grapevine cultivation became evident. Such a selection could explain the high frequency of resistance haplotype 149–208 within the Caucasian V. vinifera.
Anyway, Gur et al.  recently identified an E. necator strain in Israel genetically differentiated from those characterized in Europe and North America, proposing a non-American origin for it and possibly an Asian one. This hypothesis could explain the presence of resistance to E. necator in V. vinifera, and also in other Asian Vitis spp. [18, 19, 24, 50], with the co-evolution theory. However, the Gur et al.  suggestion is in contrast to common notions on E. necator origin and centres of differentiation [e.g. 53] and a more in-depth study would be necessary to confirm their new proposals.
Genetic basis of resistance to E. necator in Caucasian and Central Asia grape germplasm
The mapping of co-located QTL for the resistance to E. necator in many and unrelated V. vinifera revealed a high complexity of the investigated region in chromosome 13, that encompasses some megabase from upstream to downstream of the mapped loci . This would suggest a question: are the Caucasian and Central Asian resistant V. vinifera grapevines, which carry different marker haplotypes, sharing the same resistance genes or are we dealing with different resistance sources developed starting from a common ancestor?
Phenotypic information collected in our research often showed distinct responses to E. necator of Caucasian grapevines (in particular ‘Shavtsitska’) and ‘Kishmish vatkana’. However, the phenotypic resistance of the Caucasian and Central Asian grape accessions, due to the trait variations associated to the genes [24, 39, 50], does not allow to clearly confirm whether the genetic basis of resistances is different or not.
We analysed the region of the QTL in the ‘PN40024’ genome  and found a single putative disease resistance gene, namely the RPP13-like protein 1, an NBS-LRR type R protein with a putative amino-terminal leucine zipper (Additional file 9: Table S5). However, approx. one Mb upstream of the QTL, there are six RPP13-like protein 1 (5 + the one of the QTL) and four At3g14460, a gene isolated first in Arabidopsis thaliana that also belongs to the class of NB-LRR microorganism defence response genes. It is interesting to note that the RPP13-like protein of the QTL contains multiple splicing variants. According to several authors, RPP13 is prone to undergoing evolutionary amino acid divergence within the LRR domain, which might create alleles deputed to recognise different strains of a pathogen . These pieces of evidence would suggest investigating in the future the candidate region in ‘Shavtsitska’, a task that was not possible to accomplish in this study due to the low coverage of the genome in the produced reads.
The literature would suggest other cases where regions rich in R-genes encompass multiple resistance loci. For instance, Ren4, from V. romanetii , and Run2, from V. rotundifolia , loci map in the same position of chr 18 of the grape reference genome ‘PN40024’; furthermore, Run2 is associated to two resistant haplotypes (Run2.1 and Run2.2) that originate from close V. rotundifolia accessions . The Ren1 region in chr 13 contains numerous genes encoding NBS-LRR proteins and appears prone to producing genetic variation . The natural selection and evolution mechanisms at the basis of R-genes clusters [13, 19, 60–62] could also have developed in Caucasian and Central Asian V. vinifera accessions with different resistance genes and/or unique combinations of resistance factors. We therefore consider the identified resistance gene in ‘Shavtsitska', member of a cluster of R genes, of which the region is rich, to name such a variant as Ren1-2 because it is linked with, or possibly allelic to, the previously described Ren1.
Information collected until now does not allow it to be concluded whether grapevines from Central Asia and the Caucasus share the same resistance genes or not. Further narrowing of the genetic region of chromosome 13 explored up to now, as well as comparative sequence analysis and deep transcriptomic study would allow to focus on the precise genetic differences. More precise phenotyping and histochemical observations could complement the information on the origin of the resistance variation and on the mechanism behind the trait.