Plum pox virus (PPV) gen. Potyvirus, agent of Sharka disease, is the most devastating viral pathogen of stone fruits, particularly peach (P. persica L. Batsch). Since the first report of PPV infections in Bulgaria [1], the virus has gradually spread worldwide. The control of virus spread is currently based on a series of preventive measures, such as outbreaks monitoring and eradication of affected plants. Nevertheless, these strategies have barely slowed spread of the virus, which is by now endemic in many European growing areas [2]. Among several PPV strains identified so far (D, M, Rec, EA, C, W, and T), the M isolate is by far the most virulent in peach [3].

The identification and exploitation of PPV-resistant sources represents the main eligible strategy for the long-term protection of peach cultivation. Genetic sources of both PPV-M and -D resistance have been identified and characterized in apricot (P. armeniaca L.) through a series of linkage and association studies, allowing the development of molecular markers and breeding programmes for the transfer of resistance to novel varieties [4]. In contrast, immunity or resistance against PPV-M have not yet been reported in peach, but only in the related species P. davidiana and P. dulcis (almond) [5, 6]. The introgression of resistance through hybrid selections has been unsuccessful so far due to several drawbacks, including the lack of resistance-associated molecular markers. The vast majority of peach accessions are highly susceptible to the PPV-M strain, although genotypes showing lesser symptom severity upon infection have been reported [7, 8]. Most, if not all, the peach varieties are symptomless when infected by PPV-D strain [9]. The Italian PPVCON research project [10] enabled a wide evaluation of several peach germplasm collections, allowing the identification of a few tolerant accessions, characterized by virus replication and spread throughout graft-infected plants but essentially symptomless or developing only sporadic, mild symptoms. This tolerance may represent an important trait for the short-term preservation of peach cultivation in endemic areas, although there are conflicting views among virologists about the possible impact of cultivating tolerant plants [11].

The range of phenotypic response to virus infection observed in peach germplasm suggests the existence of a quantitative disease tolerance, able to confer a reduction in symptoms intensity rather than absence of the disease. The identification of genetic loci controlling such quantitative tolerance may allow the design of molecular markers to assist breeders in pyramiding favourable alleles and the development of tolerant varieties. However, the genetic architecture of PPV-M tolerance is still unknown in peach and only limited information is available about the resistance alleles present in related species. Linkage mapping experiments in ‘Summergrand’ (peach, susceptible) x ‘clone P1908’ (P. davidiana, resistant) progenies suggest a complex pattern of polygenic inheritance [12, 13]. Minor quantitative resistance loci (QRLs) have been identified on LG (linkage group) 1, 2, 4, 6 and 7. However, the same QRLs were only partially confirmed in a cross ‘Rubira’ (peach) x ‘P1908’, suggesting that their position, number and effect vary depending from the genetic background of peach parents [14]. The small number of molecular markers and progeny sizes adopted in such mapping experiments made it difficult to identify the genomic regions and putative candidate gene(s) associated to the trait. In apricot, a major determinant for PPV resistance -most probably corresponding to a MATH gene [15] belonging to the TRAF-like gene family- has been fine-mapped at the PPVRes locus on LG1, although its effect also varies depending on the genetic background [16]. In the model species Arabidopsis thaliana, dominant resistance against PPV is also conferred by a MATH gene [17, 18], suggesting functional conservation across plant species. A major mechanism conferring resistance against potyviruses involving eukaryotic translation initiation factor (eIF) proteins has also been demonstrated in several model and non-model species [19]. Silencing of the eIF(iso)4E gene confers PPV resistance in plum [20] and peach [21]. However, low genetic diversity in eIF genes seems to be present within peach germplasm (Decroocq, unpublished results).

In peach biparental linkage mapping approaches often offer limited genetic resolution, due to small progeny size and lack of polymorphisms in genomic regions identical-by-descent. Particularly for traits influenced by the genetic background, as virus resistance, it is difficult to achieve an overall picture of involved loci using a limited number of parents. The genome-wide association (GWA) approach is becoming an increasingly powerful tool to identify loci controlling both quantitative and qualitative traits, bearing the potential to improve the power of detection in comparison with classical biparental linkage mapping [22]. This approach relies on historical recombination events occurred in natural populations and collections of landraces, breeding materials and varieties, establishing marker-trait associations based on linkage disequilibrium (LD), the non-random association of alleles at two or more loci [23]. The effectiveness of GWA largely depends on LD extent and distribution, which in turn are affected by biological and evolutionary factors. The presence of population structure and familial relatedness, i.e. systematic difference in allele frequencies between subpopulations, is particularly problematic for GWA since it leads to spurious associations and increase in false positives [24]. As a consequence of self-mating system, peach is characterized by high levels of inbreeding [25]. Moreover, peach germplasm has undergone a series of bottlenecks due to domestication-related events and modern breeding activities, started in the middle of the twentieth century from a low number of parents [26]. These phenomena along with artificial selection contributed to reduce genetic diversity in cultivated peach, increasing LD levels compared to wild relatives and generating different sub-populations [27–30]. Recent examples of the use of GWA approaches in peach showed promising results for the mapping of both quantitative and qualitative traits ([31] [32]). In the present work, a GWA approach was applied in a panel of peach accessions, genotyped with 9 K SNP Array to identify genomic regions associated to Sharka disease tolerance and putative markers to be used for assisted breeding.

Currently, Sharka disease is one of the most important phytosanitary issues in peach. The absence of intraspecific source of resistance and the complexity of introgression from related species has prompted the search of alternative approaches, including genetic engineering [11].

In agreement with several other studies, we confirmed the high susceptibility of peach to PPV. Almost all breeding-derived accessions were rapidly infected by the virus, developing moderate to severe symptoms. A restricted number of accessions clearly showed different responses against viral inoculum in screen house conditions, including delayed disease appearance, mild or no visible symptoms and a certain recovery ability. These accessions mostly belong to the ‘Occidental traditional’ cluster or are ‘admixed’ genotypes with prevalent Oriental ancestry. The tolerance of these accessions was also confirmed by field trials under very high inoculum pressure, although symptoms were slightly more severe, as previously reported for many resistant lines derived from P. davidiana hybrids [37].

The plant response to virus infection is a complex trait, affected by several factors including viral strains and environmental conditions [38]. A common concern in evaluating peach response to PPV infection is the lack of an objective method to measure and compare responses among genotypes. Symptoms evaluation through visual inspection and attribution of score classes are affected by a certain degree of subjectivity, which hampers accurate differentiation of the specific response in each accession. Clearly, the quality of the phenotypic data has significant bearing on the accuracy of GWA, particularly for differentiating a quantitative disease response [39, 40]. The classification of plants behaviour as binary outcome (tolerant vs susceptible) has been proven to be useful in apricot, at least for detecting loci with major effects [41–43]. Anyway, the search for more objective phenotyping methods is a current research priority, as well as a more stringent and reliable evaluation of the plants through experiments in natural conditions.

In this work we provide a first insight into quantitative resistance loci affecting Sharka disease tolerance in peach by using an association mapping approach. A recent study in apricot has demonstrated the power of GWA in detecting both known and novel PPV resistance loci, even in a small size population [42]. GWA requires a genomic map in which marker density is higher than the LD extent [44]. The moderate to high LD levels observed in peach and the effective marker density deployed in this study (about 1 marker every 40 Kbp), appear sufficient to tag associated loci. This is maybe particularly true for traits under biotic selection, often underpinned by a small number of large-effect loci [45].

In terms of accuracy, the effectiveness of the GWA approach is largely determined by the level of population stratification. Although a small panel size may introduce bias in the estimation of genetic relationships, in our case the structure of the analysed population and individual membership reflect those observed in another study comprising more than 1500 peach accessions [32]. Accounting for the complexity of the phenotypic dataset and the known stratification of our peach population, different algorithms were tested for modeling marker-trait associations. The Mixed Linear Model (MLM) outperforms the prediction of both naïve and structured GLM, since it better accounts for genetic relationships among individuals. The different MLM algorithms tested provide substantially similar results in terms of identified loci and their significance. FarmCPU seems to improve statistical power and resolution of GWA analysis, as also demonstrated by detection of already validated loci controlling fruit flesh colour and fruit pubescence. By applying stringent thresholds for reporting significant associations and different models to control for population structure and relatedness, clear signals were identified on chromosomes 2 and 3. In support of these main associations, they were recovered in all tested models. This is the first report about the presence of genetic factors regulating Sharka disease susceptibility in peach and, therefore, we can only compare results with those reported from interspecific peach cross with P. davidiana or in other species of the Prunus genus. The presence of quantitative resistance loci on chromosome 2 was previously observed by linkage mapping in both ‘Summergrand’ x ‘P1908’ and ‘Rubira’ x ‘P1908’ hybrid progenies [12, 14]. In contrast, the locus identified on chromosome 3 has not been reported in such experiments, although weak associations in a collinear region seem to be present in apricot [42]. Logistic regression and MDR analyses on binary phenotypes, and pairwise comparisons on classes of symptoms intensity in a panel of accessions, suggested that SNP_IGA_366639 and either SNP_IGA_214703 or SNP_IGA_185608, sufficiently account for the quantitative reduction of disease severity. However, their effectiveness also varied depending on the specific genotype background, suggesting the presence of additional genetic factors with epistatic and/or minor additive effects. Considering the small panel size and number of SNP markers, it is likely that only major effect loci were detected in our GWAS.

The estimation of linkage disequilibrium (LD) patterns is critical for mapping resolution and the definition of the window size around the significant SNPs [46, 47]. An approximate estimation in the analyzed population suggest a slow LD decay, comparable with those observed with the same markers set in a broader accessions panel [32]. The average extent of LD decay and localized chromosomal LD patterns in peach have also been estimated by genomic re-sequencing data, suggesting a faster decay, although only a small number of occidental breeding-derived accessions were included [28]. For such reason a conservative window size of about 200 Kbp was chosen to search for CGs around most significant regions. Inspecting ‘Kamarat’ and ‘Yumyeong’ re-sequencing data for the regions at about 9 Mb on chromosome 2, we identified two putative high-impact variants on the candidate RTM2-like gene (Prupe.2G065600). Association of the 5′-UTR variant with a reduced disease severity was confirmed in a panel of 73 accessions. As deduced by conceptual translation, PpRTM2-like (encoded by Prupe.2G065600) shares the same functional domains as the Arabidopsis RTM2 protein (AtRTM2), such as the conserved α-crystalline-like domain, a C-terminus transmembrane domain and the coiled-coils regions [48] (Fig. 7). AtRTM2 is expressed in phloem and sieve elements and has been associated to the specific restriction of the long-distance movement of TEV and other potyviruses, including PPV [36, 49], although the mechanism of action is still unknown. The expression pattern of the Prupe.2G065600 transcript was not assessed in this study, and, thus, the effect of the CT/GA-repeat deletion within the 5′-UTR remains to be elucidated. As demonstrated in several species, CT/GA motifs proximal to the ATG start codon play an important role in the regulation of gene expression ([50] and references therein). The 5′-UTR and exon II variants were also evaluated in progenies from ‘Orion’ (a susceptible cultivars) x ‘SDs’ crosses (PPV-resistant). However, at least in this genetic background, such mutations did not appear strongly associated with an increasing PPV tolerance or resistance, suggesting that genetic loci conferring quantitative tolerance in peach could be different from those conferring resistance in P. davidiana. As demonstrated in different studies with P. davidiana hybrids, the peach parent may affect the level of resistance in the progenies, and therefore, the identification of peach determinants could be also important for the introgression of a high level of resistance from related species. The variant found within ENODL-like gene (Prupe.3G291466) cannot be tested, since it is not present in P. davidiana or ‘Orion’ peach background.

In our study, the mapping resolution is mainly limited by the panel size, which does not allow the unequivocal identification of candidate gene(s) in the detected intervals. Despite this limitation, the loss-of-function mutation in the ENODL-like gene (Prupe.3G291466) or the several aminoacidic substitutions in MYB33/65-like (Prupe.2G050100) represents interesting candidates for future studies. For example, the miR159-regulated MYB33/65 plays a role in disease symptom induction by Cucumber Mosaic Virus in Arabidopsis [51]. The functions of Nodulin-like genes in non-nodulating species is still largely unknown, although recent studies highlight their importance in many aspects of plant development and plant-microbe interaction [52]. ENODL-like family members are supposed to be carbohydrate transporters, although some of them, such as AtENODL1,13–15, were differentially phosphorylated by the treatment with elicitors of plant immunity [53].

In perspective, an improvement of mapping resolution could be achieved by increasing markers density, for example through whole-genome re-sequencing. However, this approach could be advantageous only after the increase of sample size, identifying other tolerant/resistant genotypes and/or transferring the trait in different genetic backgrounds, primary in breeding-derived accessions.

The present study is the first effort to identify genetic factors involved in Sharka disease in peach through a GWA approach. The understanding of the genetic basis of peach response to PPV infection is crucial to exploit favourable alleles already present in cultivated peach germplasm, representing a short-term solution for endemic areas and a more feasible approach compared to the introgression from related species. We provide evidence of the presence of quantitative resistant loci in a collection of peach accessions. Although with some limitations due to the small panel size and low number of tolerant individuals, we identified three major loci and three highly informative SNP markers, accounting for most of the phenotypic variability in PPV-M susceptibility that could be useful for marker assisted breeding or selection. Clearly, results should be confirmed by further studies. Biparental populations derived from this set of germplasm may represent a first step of validation and for this purpose, progenies derived from ‘Kamarat’ and ‘Ghiaccio1’ are presently being developed and will be directly evaluated by on-fields trials in endemic areas. Alternatively, the combination of association studies with larger populations and bi-parental linkage mapping could assist the confirmation of the identified loci and also the localization of additional loci or rare variants affecting host susceptibility.