Multivirulent Plasmopara viticola strains from Cilaos on Réunion Island break down Rpv1, Rpv3.1 and Rpv10 mediated resistance of grapevine

Abstract

Grapevine downy mildew, caused by Plasmopara viticola, is one of the most destructive diseases in viticulture. Resistance-based management strategies rely on grapevine varieties carrying major resistance loci (Rpv). Although breakdown of several loci has been reported, Rpv1 had remained effective until now. Here, we provide the first evidence of Rpv1 breakdown by P. viticola in Cilaos on Réunion Island (France) and the first case of strains simultaneously overcoming three resistances of grapevine. We combined pathogenicity assays with whole-genome sequencing to characterize strains of P. viticola collected in Cilaos in 2023, alongside a panel reference strains. Hypersensitive response and sporulation were assessed for each strain on Chardonnay (susceptible variety) and four resistant varieties carrying Rpv1, Rpv3.1, Rpv10, or Rpv12. The strains collected in Cilaos were able to overcome not only Rpv1 but also Rpv3.1 and Rpv10. On Rpv1, we observed a complete loss of host recognition with high sporulation. On Rpv3.1, the phenotype was consistent with previous breakdowns, and the strains carried the vir1 allele previously described in France. By contrast, the breakdown of Rpv10 differed from that reported in European populations: whereas European strains displayed only partial breakdown of resistance, the strains of Cilaos showed complete loss of host recognition with high sporulation. Moreover, genomic analyses revealed a novel mutation, a large homozygous deletion in the corresponding avr locus. Population genomic analyses further suggests that the P. viticola strains from Cilaos share a genetic background with populations from western Europe, raising serious concerns about the potential emergence and spread of multivirulent lineages in that area. These findings highlight the need for large-scale virulence monitoring of P. viticola and improved strategies for the sustainable management of grapevine resistance in Europe.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09004-9.

Introduction

Grapevine downy mildew is caused by the oomycete Plasmopara viticola (Berk. & M.A. Curt.) Berl. & De Toni, an obligate biotrophic pathogen that infects grapevine species. Native to North America, this pathogen has spread to all major grapevine-growing regions worldwide, where it causes significant yield and economic losses [1]. In Europe, P. viticola was introduced in the 1870s from its center of origin [2–4]. At that time, the grapevine varieties (Vitis vinifera L.) that were cultivated there were highly susceptible to the pathogen, leading to a severe outbreak of the disease and widespread crop devastation [5].

The discovery of resistance sources within the Vitis genus led to breeding efforts that aimed at introgressing resistance to P. viticola (Rpv) loci — genomic regions responsible for grapevine downy-mildew resistance— into susceptible V. vinifera varieties. To date, more than 30 Rpv loci have been identified [6] but only a few have been effectively incorporated into breeding programs to develop resistant grapevine varieties currently used in European vineyards. Among those, four loci —Rpv1, Rpv3.1, Rpv10, and Rpv12— are the most widely used and prevalent in resistant grapevine crops [7].

The genetic basis of these Rpv loci involves major quantitative trait loci (QTL) that exhibit monogenic inheritance pattern and mediate partial resistance phenotypes [8], meaning that individually they are able to limit the growth and reproduction of the pathogen, but not entirely, and when combined, their inhibitory effect increases. Moreover, these loci follow a gene-for-gene interaction model, which is characterized by strain-specific responses on the same host and by a hypersensitive response when the host’s immune system recognizes the pathogen [9].

Despite the limited deployment of resistant varieties, P. viticola has already overcome Rpv loci-mediated resistance. The first documented breakdown resistance was reported on Rpv3.1 in 2010 [10]. The erosion of Rpv10-mediated resistance, was also reported by several studies, having initially been noted in the adaptation of specific strains to Rpv10-carrying varieties in 2016 [11], and later confirmed by studies in 2021 [12] and 2022 [13]. For the Rpv12 locus, breakdown of resistance was reported in 2021 [14] and 2022 [13]. Virulence against Rpv loci can be recombined into multivirulent pathotypes, and P. viticola strains breaking down the resistance conferred by up to two loci at a time have been described in Europe [13, 15]. So far, Rpv1 was the last of the main Rpv loci for which no breakdown had been reported, however, in 2023 an outbreak was spotted in Cilaos, Réunion Island (I. Hoarau, Personal communication, 2023), and in 2024 in France in the Rhône valley [16], in both cases on Rpv1-carrying varieties.

In recent years, gene-for-gene interactions between grapevine resistance loci and P. viticola effectors have been progressively deciphered. On the plant side, it has been observed that the Rpv loci are generally located in genomic regions that are rich in nucleotide binding domain and leucine-rich repeat genes (NBS-LRR) which are involved in the recognition of specialized pathogen effectors [7, 17, 18]. On the pathogen side, the RxLR family is the largest and most studied family of cytoplasmic effectors in oomycetes. These effectors are characterized by the presence of a signal peptide (SP), a RxLR-EER motif at their N-terminal sequence with one or more WY-domains, and a common fold found only in this family of proteins [19].

More specifically, loci associated with the virulence of P. viticola on Rpv3.1, Rpv10 and Rpv12 have been identified through Genome Wide Association Studies (GWAS) [20] and Quantitative Trait Loci (QTL) mapping [21]. Virulence towards Rpv3.1 and Rpv12-carrying varieties was found to be recessive and associated with deletions or disruption of genes encoding RxLR effector-like proteins [20, 21]. Outbreaks associated with Rpv3.1 and Rpv12 breakdown appear to result from recurrent, independent mutation events across different geographical regions [20, 22]. The locus involved in the Rpv10 breakdown involved a different underlying mechanism, the virulence was dominant and linked to the presence of additional putative secreted proteins, suggesting a possible virulence-suppressor mechanism [21]. The haplotype carrying the allele conferring virulence towards Rpv10 varieties seemed to have originated from P. viticola north-American populations and spread into some European areas via secondary introduction of the pathogen [21].

In this study, we investigated a grapevine downy mildew outbreak detected in 2023 in Cilaos, where a marked increase in disease incidence was observed in vineyards planted with the ResDuR varieties Floréal and Artaban, which carry the Rpv1 and Rpv3.1 loci. We isolated and characterized strains, referred hereafter as the “Cilaos strains”, assessing their virulence profile and genetic background together with a panel of strains representing P. viticola genetic and virulence diversity. Whole-genome sequencing of these strains was used to assess the genetic origin of the Cilaos strains and characterize the alleles at the known avirulence loci [20, 21]. Our results provide the first evidence for the breakdown of Rpv1 locus. Moreover, this represents the first documented case of P. viticola strains able to simultaneously overcome three out of the four mainly deployed Rpv loci (Rpv1, Rpv3.1 and Rpv10). Remarkably, the Cilaos strains exhibited an unprecedent complete loss of host recognition on Rpv10-carrying varieties and harbored a novel mutation at the avirulence locus, distinct from the previous breakdown of Rpv10 in Europe.

Materials and methods

Incidence and severity of the Plasmopara viticola outbreak in Cilaos

Grapevine downy mildew was assessed in an experimental vineyard plot at Bras Sec in Cilaos, Réunion Island, in 2023, where symptoms were observed in both foliage and grape bunches (see Fig. 1A-H). The plot contained two ResDur grapevine varieties, Floréal (Rpv1–Rpv3.1) and Artaban (Rpv1–Rpv3.1) that were grafted on Couderc-13 plants in 2019, covering a total area of 400 m². Each variety occupied approximately half of the plot, as shown in Fig. 1I, and comprised 121 plants.

Fig. 1.

Downy mildew symptoms on Artaban and Floréal (Rpv1-Rpv3.1) varieties in Cilaos, Réunion Island. A-C Symptoms on grape bunches of the variety Artaban. Many berries have dried out and collapsed, other remain with brownish lesions, and rachises display brownish necrosis. D-E Leaves of the variety Artaban covered by Plasmopara viticola sporangia in the lower surface. F Shoot of leaves of the Artaban variety with yellowing and necrosis. G-H Leaves of the variety Floréal displaying oil spot (upper surface) and P. viticola sporulation (lower surface). I graphical representation of the plot where 18 rows were present, half of them planted with the Artaban variety and the rest with Floréal

Disease incidence and severity were measured separately for each variety. A total of 34 plants per variety were randomly selected, and all grape bunches, along with 20 leaves from each plant, were assessed for visible symptoms of downy mildew. Incidence was expressed as the percentage of symptomatic plants, while severity was expressed as the percentage of symptomatic leaves and grape bunches per plant. To assess the impact of the disease in the plot, a scale ranging from 0 to 6 was used, where: 0 corresponded to less than 1% incidence with very scarce symptoms; 1 to 1–10% incidence with sparse symptoms; 2 to 10–50% incidence; 3 to incidence greater than 50% with severity below 1%; 4 to incidence greater than 50% and severity between 1% and 5%; 5 to high incidence (>50%) and severity between 5% and 20%; and 6 to high incidence (>50%) with severity exceeding 20%.

Isolation of the strains from Cilaos

From the affected plot we sampled symptomatic leaves, and through monosporic isolation, following the protocol described by Paineau et al. [13], we successfully isolated three strains of P. viticola. The strains were designated Pv8586_1, Pv8588_1 and Pv8583_1. The first two strains were isolated from Floréal plants, while the other one was obtained from an Artaban plant. All three strains have been deposited in the Plasmopara viticola collection maintained by the SAVE research unit at INRAE Bordeaux where infected leaf tissue with sporulation is kept frozen at −20°C after being soft dried in silica desiccator overnight.

Genetic relatedness of Plasmopara viticola strains from Cilaos

To assess the genetic relatedness of the stains isolated from symptomatic leaves in the affected vineyard, we used the KING-robust kinship estimator [23]. For this analysis, we generated a set of high-quality SNPs from whole-genome sequencing data of the strains collected in Cilaos (Pv8586_1, Pv8588_1 and Pv8583_1) and included eight additional, unrelated control strains for which genomic data was available in public databases. The origin and characteristics of all strains used in the analysis are summarized in Table 1.

Table 1.

Features of Plasmopara viticola strains used in this study

Strain ID	Reported virulence	Year of collection		Country	Locality	Reference	Genomic data Accession number
Pv8586_1 Pv8588_1 Pv8583_1	-	2023	France		Cilaos-Réunion Island	Present study	PRJNA1255592
Pv2664_1	Rpv3.1	2013	France		Villenave d’Ornon	Paineau et al. [13]	PRJNA1255592
Pv412_11	Rpv3.1	2010	Switzerland		Cugnasco	Paineau et al. [13]	PRJNA1095879
Pv8522_1a	Rpv3.1/Rpv12	NA	Switzerland		Soyhières	Wingerter et al. [14]	PRJNA875296
Pv1419_1	Rpv3.1/Rpv10	2012	Germany		Ihringen	Paineau et al. [13]	PRJNA1095879
Pv2543_1	Rpv12	2013	Hungary		Pécs	Paineau et al. [13]	PRJNA1095879
Pv2578_1	Avirulent	2013	France		Le Landreau	Paineau et al. [13]	PRJNA1255592
Pv3174_11	Avirulent	2016	Italy		Piateda	Paineau et al. [13]	PRJNA644583
Pv2909_11	Avirulent	2014	Germany		Geisenheim	Paineau et al. [13]	PRJNA1255592

NA no information

^aName of the strain in Wingerter et al. (2021) publication was “VB-isolate”

To obtain the DNA from the Cilaos strains, sporangia, and sporangiophores were produced and collected following the protocol described by Dussert et al. [22, 24]. DNA extraction was performed using a CTAB-based protocol adapted from Möller et al. [25] and further modified by Dvorak et al. [26]. The purity and concentration of the extracted DNA were assessed using a DeNovix DS-11 spectrophotometer and a Qubit 3 fluorometer (Invitrogen), respectively. Libraries were prepared using an Illumina DNA TruSeqkit and sequencing was performed at the GeT-PlaGe facility (Toulouse, France https://get.genotoul.fr/) with a NovaSeq6000 to produce 2x150 bp paired-end read.

The obtained reads from the Cilaos strains together with the publicly available reads from the other strains were mapped against the reference genome Pv221_FU (genome available: https://doi.org/10.57745/MXJWZS) using BWA mem2 v2.2.1 [27] and bam files were handled using SAMtools v1.10 [28]. The PCR duplicates were removed with the tool markduplicates in GATK v 4.2.6.1 [29]. Variant calling was performed in the same software with the tool HaplotypeCaller setting the option output-mode to EMIT_ALL_ACTIVE_SITES. This procedure was performed parallelizing mapping and variant calling using the RattleSNP pipeline (https://rattlesnp.readthedocs.io/).

After the variant calling step, high-quality sites were filtered using BCFtools v1.21 [30] and VCFtools v0.1.16 [31] to obtain a dataset with the following characteristics: indels were removed, only biallelic SNPs (Single Nucleotide Polymorphisms) were kept, SNPs in repetitive regions were removed, and SNPs were kept when MQ ≥ 59, QUAL ≥ 25, mean depth ≥ 5, then genotypes at each site with depth values outside the aforementioned limit were set to missing, and finally 5% of missing data per site was allowed.

The resulting high-quality SNPs were used as input to calculate the KING-robust kinship estimator implemented on PLINK2 [32].

Virulence profile characterization

Plasmopara viticola strains

In addition to the strains isolated from Cilaos, the eight other strains from which the genomic data was used in the previous section of this paper, were obtained from previous studies [13, 14]. These strains originating from various European wine-producing countries were included in this part of the study due to their diverse virulence profile against Rpv loci (see Table 1). From these eight strains, five (Pv1419_1, Pv2543_1, Pv2664_1, Pv412_11, and Pv8522_1) were previously shown to overcome resistance conferred by some of the most commonly used Rpv loci, Rpv3.1, Rpv10, and Rpv12. In contrast, the remaining three strains (Pv2578_1, Pv2909_11, and Pv3174_11) were isolated from grapevine varieties lacking any known Rpv loci and appear to be avirulent on Rpv-carrying varieties (Table 1).

Plant material

To assess the virulence of the Cilaos strains together with the other chosen strains, a cross-inoculation experiment was performed in which four grapevine varieties carrying different resistance loci were used, i.e. 3179-90-7 (Rpv1), Regent (Rpv3.1), Muscaris (Rpv10) and Fleurtai (Rpv12). Additionally, the susceptible variety Chardonnay that carries no resistance loci was included as a control, for a total of five varieties.

A total of 20 cuttings per each variety were obtained from nurseries (Reben Sibbus: https://sibbus.com/fr/index.html, VCR France: www.vcrfrance.fr/ and IFV: https://www.vignevin.com/innovation-varietale/) and grafted onto the Selektion Oppeinheim 4 (SO4) rootstock. All the plants were grown simultaneously in a glasshouse under natural photoperiod conditions, without chemical treatment. After two months of growth, the 5^th leaf below the apex was collected and used for the inoculation.

Cross-inoculation assay

Inoculation was carried out according to the methodology of Paineau et al. [13], with modifications. To obtain the inoculum, strains were propagated on leaves of Cabernet Sauvignon and kept on wet chambers at 23 °C during a week. After that, sporangia were gently rinsed off to ensure that freshly produced, uniformly aged sporangia could be collected the following day for inoculum preparation. The day of the inoculation, sporangia suspension was prepared and adjusted to a concentration of 1×10⁵sporangia/mL using a portable particle counter (Scepter 2.0 Automated Cell Counter; Millipore).

The eleven strains were inoculated on the five grapevine varieties, resulting in a dataset of 55 interactions. For each interaction, six repetitions were performed as described below. Inoculation was carried out on 15 mm diameter leaf discs excised from the fifth fully expanded leaf below the apex of each plant. To prepare the discs, leaves were first washed with reverse-osmosis water, dried on absorbent paper, and then leaf discs were excised using a cork borer. For each interaction, six discs (representing repetitions) from different leaves were placed abaxial side up on wet filter paper within a square Petri dish (25x25 cm). All the leaf discs that were inoculated with one strain were randomly arranged within the dish. This represented a total of 30 discs per Petri dish (five varieties with six leaf discs each), sprayed with 4 mL of the sporangia suspension of a strain. Mock inoculations were also performed using sterile water instead of the sporangia suspension.

In summary, six repetitions were conducted for each combination of variety and strain, resulting in a total of 330 inoculated leaf discs for the experiment. Additionally, 30 mock-inoculated discs were inoculated.

One day after inoculation the leaf discs were wipe-dried with filter paper. The Petri dishes, used here as moist chambers, were sealed with plastic film to maintain high relative humidity and incubated for 6 days at 18 °C under a 12/12 h photoperiod.

After this assay, a second experiment following a similar procedure was performed to confirm the results regarding the breakdown of Rpv1. In this case, only the varieties Chardonnay (susceptible), 3179-90-7 (Rpv1) and Regent (Rpv3.1) were inoculated with all the strains except Pv412_11, which failed to grow during the strain propagation step. Three types of leaves were included: the fifth, eight and tenth fully expanded leaves below the apex of the plant. Each leaf type was treated independently, thus, there were three Petri dishes per strain, each for a type of leaf.

The main differences from the first experiment were the plant varieties used and the inclusion of multiple leaf types; all other procedures, including leaf disc preparation, inoculation, and incubation, remained unchanged.

Symptoms and virulence assessment

Sporulation and necrosis were measured at the 6^th day post-inoculation on all the leaf discs used. Sporulation area was quantified by taking high resolution pictures and then performing an image analysis to estimate the proportion of leaf area covered by sporangia. Leaf discs were photographed four at a time. Pictures were taken with a Canon EOS 650D camera equipped with a macro lens (Canon EF 100 mm f/2.8 USM). Photographs were taken in manual mode (f/5.6; ISO-100). Each 17.9-megapixel image was analyzed with the Python-based tool CATIA (available at: https://gitlab.com/grapevinedownymildew/notebook_image_analysis) trained to isolate the discs and then detect the sporulation area.

Necrosis was assessed as an indicator of the hypersensitive response (HR) using a modified version of the visual scale proposed by Paineau et al. ([13], Fig. S1) and Dvorak et al. [21], with the addition of a value of 0 to indicate the absence of necrosis. We developed a visual scale ranging from 0 to 5, where 0 and 1 correspond to the absence of an HR, and values from 2 to 5 indicate the presence of HR in the host variety in different categories.

The different types of HR were classified based on the shape and distribution of necrotic lesions across the leaf tissue, as illustrated in Fig. 2, to reflect the type of necrosis involved in the plant’s immune response.

Fig. 2.

Visual scale used to assess Hypersensitive Response (HR) to Plasmopara viticola infection on grapevine leaf discs. The top of the figure shows the classification of the different values of the scale according to the absence (no HR) or presence of HR (HR). Pictures of infected leaf discs illustrating the different categories of the scale are displayed with a brief description of the different values of HR according to the observed necrosis pattern. At the bottom of the figure, the corresponding graphical representation of the different categories is shown, created using Inkscape v1.3.2 (https://inkscape.org). For the “no HR” reaction there are two categories named “0” and “1”, while for the “HR” reaction four categories were identified: “2”, “3”, “4” and “5”. 0: no necrosis nor visible tissue damage, 1: Dense brownish lesions that extend without clear delimitation, typical of tissue damage due to cell starvation but not caused by immune response of the plant 2: Light diffused necrosis spots localized inside the limits of the polygons formed by the small veins of the leaf, 3: Smaller necrosis spots, still inside the limits created by the small veins of the leaf, with a central point that diffuses inside those limits, 4: Small circular or oblong spots that do not diffuse and localize in the corners formed by the intersection of the small veins of the leaf, 5: Very small and well defined dots that tend to be circular and localize in the intersection of the small veins of the leaf.

Additionally, we used a parameter that we called Efficiency of the Resistance Factor (ERF) to assess the capacity of a given strain to sporulate on a variety carrying a given Rpv locus in comparison with the capacity to sporulate on a variety with no resistance locus, i.e. susceptible, following:

The value is expressed as the complementary ratio between the sporulation of the strain on the variety carrying the resistance locus and the sporulation of the strain on a susceptible variety, as defined in Formula 1. The lower the value of the ERF, the higher the sporulation on the locus-carrying variety having as reference the sporulation on the susceptible variety. Values close to 0 correspond to similar sporulation level than in the susceptible variety, and negative values can be obtained when sporulation on the Rpv-carrying variety is higher than on the susceptible variety.

We considered that a strain was able to breakdown the resistance conferred by a given locus similarly to the method described by Paineau et al. [13], based on the principle that a virulent strain should be able to evade the host’s immune system, i.e. not producing HR, and being able to propagate, i.e. have efficient spore production on both susceptible and Rpv-carrying variety. However, in some cases, partial breakdown has been described when incomplete scape of host’s recognition is observed, resulting in an unstable phenotype showing HR at low frequency (this is the case of the Rpv10 breakdown [13, 21]).

The values of sporulation area were analyzed statistically. Differences in the percentage of sporulating area produced by each strain across the various Rpv-carrying varieties and the susceptible variety were assessed using ANOVA followed by a paired Tukey test using R v4.4.1 [33]. For these analyses, the percentage values were square root transformed to meet homogeneity and normal distribution of residues.

To compare sporulation among strains on each Rpv-carrying variety, we first calculated, for each strain, the mean sporulation observed on the susceptible variety Chardonnay. We then normalized the sporulation of each leaf disc by subtracting this strain-specific mean value from its corresponding measurement on the Rpv-carrying variety (sporulation on Rpv-carrying variety minus the mean sporulation on Chardonnay). This normalization step accounted for the strain-dependent developmental differences in P. viticola. Subsequently, we performed a pairwise Wilcoxon tests to evaluate statistical differences and to generate the corresponding plots in R.

The visual scale for HR, as a categorical variant was not analyzed statistically, it was only used to characterize the plant’s response to the strain’s infection. Additionally, the HR scoring was also used to quantify the number of leaf discs (repetitions) exhibiting any type of hypersensitive response (HR) in each variety × strain interaction. This allowed us to evaluate whether our predefined criteria for breakdown were met. Specifically, breakdown was considered when, besides efficient sporulation, HR was observed in less than 100% of the inoculated leaf discs for a given plant–strain interaction.

Characterization of the Plasmopara viticola loci AvrRpv3.1, S-AvrRpv10 and AvrRpv12 in the Cilaos strains

To further characterize the Cilaos strains, we analyzed the previously identified P. viticola loci that interact with the grapevine resistance loci Rpv3.1, Rpv10, and Rpv12 [20, 21]. This involved mapping of the whole genome sequencing reads from the Cilaos strains, together with publicly available reads from the other phenotyped strains, against different reference genomes, according to their virulence profiles. The accession numbers for the used sequencing reads are listed in Table 1.

Reads were mapped to different reference genomes using BWA mem2 v2.2.1 [27]. For the P. viticola locus associated with the Rpv3.1 breakdown, we used the reference genome Pv221_FU [20] (genome available: https://doi.org/10.57745/MXJWZS). The interval between 640 and 740 kbp of the Primary_000014F contig was then analyzed using the corresponding genome annotation as guide. For the locus associated with virulence on Rpv10, we used the diploid genome assembly Pv1419_1 and its gene annotation [26], examining both haplotypes, as this region appears to have undergone substantial rearrangements and the gene content differ between them. In Haplotype 1, we examined the region of chromosome 16 (Chr16) spanning 3,900–4,400 kbp. In Haplotype 2, which is known to harbor the virulent allele, the analyzed region was on chromosome 16, but between 4,050–4,750 kbp. Finally, for the locus associated with Rpv12, we used the same reference genome as for Rpv3.1, but analyzing the region on the Primary_000017 contig from 250 to 400 kbp.

Afterwards, we computed both the average genomic depth and the depth across sliding windows within the target loci to obtain normalized read depth. This allowed us to estimate gene copy number as a proxy for ploidy and to identify allele variation at each locus. The sliding window sizes used for the genomic regions associated with Rpv3.1, Rpv10, and Rpv12 were 1 kbp, 4 kbp, and 1 kbp respectively. To identify nonsense mutations and any open reading frame disruptions, mapping depths and alignments within the targeted loci were visualized using igv v2.19 [34].

Mating type

To continue with the characterization of the Cilaos strains and the other strains in the panel, mating type was determined through crosses with reference strains of known mating types P1 and P2, following the method described by Dussert et al. [35]. In summary, each strain was first grown on grapevine leaves, a sporangial suspension was then prepared for each strain and mixed with a sporangial suspension of either the P1 or P2 reference strain. The mixtures were inoculated onto grapevine leaf discs placed on water agar and incubated at 22 °C for 5 days (12 h/12 h photoperiod), after which the temperature was lowered to 12 °C and maintained for three weeks. After the incubation period, the presence of oospores was assessed using a stereo microscope (SMZ1270, Nikon, Tokyo, Japan). A strain that produced oospores only with one of the reference strains was classified as the opposite mating type of that reference.

Population structure analysis

To investigate the genetic relationships of the Cilaos strains within the global P. viticola population, we integrated whole genome Illumina sequencing data from additional isolates and performed population structure analysis through Principal Components Analysis (PCA) and Sparse Non-negative Matrix Factorization (SNMF) clustering.

In our dataset, we included a set of strains previously used to assess the population structure of this pathogen [21], which already encompassed most of the reference strains employed in the present study except Pv2909_11; this strain was therefore added.

For the Cilaos strains, genetic relatedness analysis showed that the three strains were clonal (see Results section). To satisfy the assumption of independence of observations required for both PCA and SNMF clustering, and to avoid inflated variance toward clonal lineages, only one representative strain (Pv8586_1) was retained in the dataset.

The final dataset comprised 59 strains from several wine producing countries, France (n=16 including the one from Cilaos), Spain (n=2), Italy (n=7), Hungary (n=15), Switzerland (n=5), Germany (n=8) Canda (n=2), United States of America (n=1), China (n=1), Georgia (n=1) and Lebanon (n=1). Details on the strains used for this section including their genomic data availability are shown in Supplementary Table 1.

To genotype the strains, a set of high-quality SNPs was produced following a similar methodology as described before on the section about the relatedness of the strains from Cilaos. Briefly, the reads were mapped against the reference genome Pv221_FU, PCR duplicates were removed, and variant calling was performed. After, high-quality sites were filtered to obtain a dataset with the following characteristics: SNPs in repetitive regions were removed and SNPs were kept when biallelic, QUAL ≥ 20, mean depth ≤ 5 and ≥ to twice the average individual coverage. Then genotypes at each site with depth values outside the threshold were set to missing, and only 10% of missing data per site was allowed. Finally, SNPs were pruned for linkage disequilibrium using PLINK 2, with a window size of 50 kb, a step size of 10 variants, and an r² threshold of 0.1

The resulting high-quality SNPs were used to analyze the placement of the Cilaos strain (Pv8586_1) into the group of 58 P. viticola strains. For this, we performed a Principal Components Analysis (PCA) using the adegenet package [36] in R, which does not scale the variables (SNPs) but rather centers them to account for the nature of nucleotide genomic variants. And we also used the clustering algorithm SNMF implemented in the package LEA [37] in the same software.

Results

High incidence and severity of Plasmopara viticola in the studied vineyard in Cilaos

After assessing grapevine downy mildew epidemics in both parts of the vineyard, one planted with Floréal and the other with Artaban, at Bras Sec in Cilaos, we found that disease incidence and severity on the leaves were 100% for both varieties; all assessed leaves showed symptoms corresponding to the maximum value on the rating scale used (score = 6). Regarding grape bunches, no fruit was present on the Floréal plants at the time of evaluation, therefore, no assessment was possible. Conversely, the Artaban plants that were evaluated bore six to seven bunches each, all exhibiting symptoms of P. viticola infection (100% incidence and severity; score = 6). Overall, the disease exerted a strong impact on the foliage and, where present, on the grape bunches, causing the severe symptoms shown in Fig. 1.

The Cilaos strains are clones

After the monosporic isolation from symptomatic tissue, three strains were obtained, Pv8586_1 and Pv8588_1 from the Floréal variety and Pv8583_1 from Artaban variety. Later, analysis of their genomic information, and the data from the other eight strains, revealed that the three strains from Cilaos were clones. Using the KING-robust kinship estimator, the pairwise tests performed between the three strains consistently yielded high values (above the clone/identical twin threshold of 0.354) for these strains [23]. Conversely, all the other strains were confirmed to be distantly related, scoring negative values (unrelated strains should score <0.088), which was consistent with our expectations (see Supplementary table 2).

The Cilaos strains are able to break down the resistance conferred by Rpv1 and other Rpv loci and their mating type is P2

In the first inoculation experiment five varieties were inoculated with the Cilaos strains along with the other eight strains of known virulence profile. First, we analyzed the results of sporulating area and the presence/absence of HR in the inoculated discs to check if a breakdown had occurred in the interactions variety x strain tested (Fig. 3). We show the results in Fig. 3A where a grid was plotted inside which each interaction is represented by a cell. The percentage of discs where necrosis was seen is shown as a color gradient, while the sporulation values for each disc are represented by the circles. Additionally, statistical differences between the sporulation values on Chardonnay and each of the other varieties per strain are shown (For more details on the statistical comparisons see Supplementary Figure 1). All strains successfully infected the susceptible variety Chardonnay, evading host’s immune response (no HR was observed).

Fig. 3.

Mating type and virulence characterization of Plasmopara viticola strains from Cilaos. Eight extra strains with different virulence profiles and geographical origins were included as reference. For virulence characterization results of a cross-inoculation experiment are displayed, in which all strains were inoculated on different six leaf discs of grapevine varieties carrying the loci Rpv1, Rpv3.1, Rpv10 and Rpv12 and on a susceptible variety with no Rpv loci (Chardonnay). A From left to right the plot displays: strain names along the y-axis; a color-coded bar indicating each strain’s mating type; and a grid representing interactions between grapevine varieties (x-axis) and strains (y-axis). In the grid, each cell corresponds to a single variety–strain combination. In each cell, the background color indicates the percentage of leaf discs (out of six) that exhibited Hypersensitive Response (HR), based on a continuous color scale. Superimposed on the cell are up to six circles, each representing one individual disc: the size of each circle is proportional to the sporulation area (%) observed on that disc. Thus, multiple circles per cell reflect the replicate measurements for that combination strain-variety. Statistical comparisons of sporulation between each variety and the reference variety ‘Chardonnay’, within each strain, are shown to the top right of the cells, where “=” denotes no significant difference (p > 0.05), while arrows indicate significant differences (p < 0.05). Upward-pointing arrows indicate higher sporulation on the tested variety than on Chardonnay; downward-pointing arrows indicate lower sporulation. Statistical comparisons were made using Tukey pairwise test after square root transformation of the percentage of the sporulating area. B Heatmap that shows the values of Efficiency of the Resistance Factor (ERF index) for each interaction variety (x-axis) - strain (y-axis) tested, except for the susceptible variety. The whiter the color, the lower the sporulation efficiency. Values are represented in a color scale and explicitly showed in each cell. C Normalized sporulation of the strains on the three resistant grapevine varieties: 3179-90-7 (Rpv1), Regent (Rpv3.1), and Muscaris (Rpv10). A boxplot-type of graphic is displayed for each variety showing the distribution of the data, and all the values are shown as dots. Normalization of sporulation values was performed by subtracting the mean sporulation value observed on the susceptible reference variety Chardonnay from each individual measurement. Each dot represents the normalized value of a single leaf disc, and each dot is colored according to the absence or presence of HR. Statistic groups are shown as letters, after pairwise Wilcoxon test was applied. D Characterization of the type of HR observed on the Rpv-carrying varieties 3179-90-7 (Rpv1), Regent (Rpv3.1), Muscaris (Rpv10) and Fleurtai (Rpv12) when infected by the strains. A bar-plot-type of graph per variety is shown, where the proportion of discs with each type of HR (x axis) per strain (y axis) is represented by the proportion of the bar colored with a given color. Each category (HR type 2, 3, 4 and 5) has its own color, and No color indicates that no HR was observed. Cilaos strains are shown in bold letters. HR= Hypersensitive Response

As expected, the strains Pv3174_11, Pv2909_11, and Pv2578_1 exhibited an avirulent phenotype, being unable to evade the immune response of grapevine varieties that carry Rpv loci (HR was always observed). Sporulation was either absent or substantially lower on varieties with the Rpv1 and Rpv3.1 loci compared to the susceptible variety. As for the rest of the panel of strains, Pv2664_1, Pv412_11, Pv1419_1, Pv2543_1 and Pv8522_1, they were shown to be virulent and met our definition of total resistance breakdown, displaying no HR and sporulation values similar or higher on the Rpv loci on which they were reported to be virulent (Fig. 3A, Supplementary table 3) (see reported virulence in Table 1). The only exception was strain Pv1419_1 which breaks down the resistance conferred by Rpv3.1 but only partially breaks down Rpv10. It was able to sporulate efficiently on the Rpv10-carrying variety with similar values of sporulation than in the susceptible variety Chardonnay, but did not evade HR in all inoculated discs of this variety (Fig. 3). Overall, these results aligned with the known virulence profiles of the tested strains and confirmed the reliability of the assay.

For the Cilaos strains (Pv8586_1, Pv8588_1 and Pv8583_1), this assay showed that all of them were able to breakdown Rpv1-mediated resistance consistent with the field observations on Réunion Island. In fact, as these strains are clones, they can be treated as biological replicates of the same individual/genotype within our experiments. This assay further revealed that the Cilaos strains were also able to overcome Rpv3.1 and, most notably, Rpv10. When inoculated on varieties carrying Rpv1, Rpv3.1, or Rpv10, the Cilaos strains did not trigger a HR in any of the leaf discs and sporulated at the same level or even more abundantly than on the susceptible variety (Fig. 3A).

We also used the ERF values to assess the sporulation efficiency (Fig. 3B). The identified breakdowns consistently corresponded to low or near to 0 index values (−2.02 to 0.19), and this observed pattern aligns with the results shown in Fig. 3A. Moreover, for the Rpv loci that are broken down by the Cilaos strains, i.e. Rpv1, Rpv3.1 and Rpv10, we show the comparison of the normalized values of sporulation among all strains per variety in Fig. 3C. This figure shows the different behaviors of the Cilaos strains relative to their growth on Chardonnay: when inoculated on the Rpv1-carrying variety 3197-90-7, they sporulate better than the rest of the strains, on Regent (Rpv3.1) they have a similar behavior than the other Rpv3.1-breaking strains, and on Muscaris (Rpv10), the Cilaos strains showed normalized sporulation values comparable to the partial breaking down strain Pv1419_1, but without inducing any HR.

Finally, we assessed which type of HR from the proposed scale was characteristic of each strain -Rpv-carrying variety interaction. Results are shown in Fig. 3D which presents the proportion of infected discs with each HR category per variety. As the Cilaos strains breakdown most of the used loci, we only could see the type of HR that they cause on the Rpv12-carrying variety Fleurtai, on which they almost always display a HR type “5”. As for the rest of the interactions, we could see that for the varieties that carry Rpv1 and Rpv3.1 the pattern is that they always display the HR type “2” and “3” when the strains are not able to breakdown these loci, whereas for the varieties that carry Rpv10 and Rp12 the pattern is that they tend to display the categories “3”, “4” and “5” when infected with strains that are not able to break them down. As for the strain Pv1419_1 that partially breaks down Rpv10, the few discs that displayed HR, the type was “2”.

The virulence profiles of the Cilaos strains on Rpv1 and Rpv3.1 were also confirmed in other three independent experiments (Supplementary Figure 2, Supplementary table 4).

It’s worth mentioning that control mock inoculated disks remained healthy until the end of the experiment.

To further characterize the Cilaos strains, along with the other strains used in the inoculation experiments, we assessed the mating type through crossing tests. The Cilaos strains were classified as P2; four additional strains were also classified as P2, while the remaining strains belonged to the P1 type (Fig. 3A).

Characterization of the Plasmopara viticola loci AvrRpv3.1, S-AvrRpv10 and AvrRpv12 in the Cilaos strains

We examined genomic regions and their gene content at P. viticola loci previously associated with the breakdown of Rpv3.1, Rpv10, and Rpv12 resistance (Figs. 4, 5 and 6), analyzing normalized mapping depths (Supplementary Figures 3–6), and compared these results with those from the other eight strains used as references.

Fig. 4.

Plasmopara viticola locus associated with Rpv3.1 breakdown in Cilaos strains (Pv8586_1, Pv8588_1 and Pv8583_1) and eight additional strains of variated virulence profiles. Strains that are virulent on Rpv3.1-carrying grapevine varieties are highlighted in light peach. For each strain the locus is represented by two lines that correspond to the haplotypes (P. viticola is diploid), with genes shown as boxes along the sequence, gene names are shown above each graphic. Black boxes indicate candidate-effector genes. Gray boxes denote other genes, while dashed white boxes indicate gene absence. Nonsense mutations in genes are shown in bright orange.

Fig. 5.

Comparison of the Plasmopara viticola genomic region associated with virulence on Rpv10 among a reference virulent strain (Pv1419_1), virulent strains from Cilaos - Réunion Island (Pv8586_1, Pv8588_1 and Pv8583_1), and seven avirulent strains (Pv3174_11, Pv2909_11, Pv2578_1, Pv2664_1, Pv412_11, Pv2543_1 and Pv8522_1). The figure shows a non-scaled diploidized representation of the associated genomic region. For each strain or strain group, each band corresponds to a haplotype sequence (P. viticola is diploid); dashed lines indicate deletions or absence of sequence. Genes encoding secreted proteins are shown as ellipsoidal gray boxes, with black boxes representing candidate virulence suppressors (S-AvrRpv10). At the top of the graphic the haplotypes of the virulent European strain Pv1419_1 described by Dvorak et al. 2025 https://doi.org/10.1101/2025.05.18.654733) are displayed. Hap 1 represents the European haplotype found in avirulent strains across Europe, whereas Hap 2 corresponds to the rearranged haplotype of American origin that carries putative virulent suppressors and that is thought to confer virulence on Rpv10 by dominant inheritance. The American-derived allele is shown in pale green. Below, the Cilaos strains are shown, characterized by a homozygous deletion in the region and absence of the candidate suppressors and the American allele. The rest of the strains (other strains) show no modification in this region and do not carry the American allele and the candidate suppressors. Strains that are virulent on Rpv10-carrying grapevine varieties are highlighted in light peach.

Fig. 6.

Plasmopara viticola locus associated with Rpv12 resistance breakdown in Cilaos strains (Pv8586_1, Pv8588_1 and Pv8583_1) and eight additional strains of variated virulence profiles. Strains that are virulent on Rpv12-carrying grapevine varieties are highlighted in light peach. For each strain the locus is represented by two lines that correspond to the haplotypes, with genes shown as boxes along the sequence, gene names are shown above each graphic. Black boxes indicate candidate-effector genes. Gray boxes denote other genes, while dashed white boxes indicate gene absence. Premature stop codon in genes is shown in bright orange

For the locus associated with Rpv3.1 interaction, the Cilaos strains showed a complete absence of the genes encoding effectors as expected for a virulent genotype (Fig. 4, Supplementary Figure 3) and was shown to carry a virulence allele previously described as avr1 by Paineau et al. [20]. All strains that were virulent on Rpv3.1-carrying varieties, including the Cilaos strain, displayed either a complete absence, partial deletions, or nonsense mutations in these effectors (Fig. 4).

To study the characteristics of the genomic region associated with Rpv10 resistance breakdown, and to take into account the complexity of the genomic architecture of the region, we used as reference both haplotypes of the partially virulent strain Pv1419_1, in which the Rpv10 virulence mechanism was originally identified. These were compared to the corresponding region in the fully virulent Cilaos strains and in the remaining strains of the panel, all of which are avirulent on Rpv10.

In the strain used as reference we cand find the haplotypes previously described by Dvorak et al. (2025) [20]: one that is present in avirulent European populations (Haplotype 1), and the other (Haplotype 2), which is thought to be inherited from American populations and confers virulence on Rpv10 by a dominant suppressor mechanism. The presence or absence of these haplotypes, as well as their genes encoding secreted proteins, including the candidate suppressors (S-AvrRpv10), was assessed (Supplementary Figure 4 and 5). As expected, we found that all avirulent strains lacked the American allele and the putative suppressors (Fig. 5). However, the Cilaos strains also lacked this allele despite displaying a virulent phenotype on Rpv10-carrying varieties. Instead, these strains harbor a homozygous large deletion (~120–215 kb) spanning most of the examined region which includes 13 genes encoding secreted proteins (Fig. 5, Supplementary Figure 4–5).

Finally, for the AvrRpv12 locus, the two strains Pv8522_1 and Pv2543_1 that were virulent on Rpv12 lacked the putative effectors linked to avirulence. In contrast, the Cilaos strains, along with the other strains that were avirulent on the Rpv12-carrying variety, had the putative effectors (Fig. 6, Supplementary figure 6).

The Cilaos strains are related to strains from western Europe

After variant calling and quality filtering, a total of 92,069 high-quality biallelic SNPs were retained. This dataset was then used for a population structure analysis.

A principal component analysis (PCA) was performed, obtaining that the first principal component (PC1) accounted for 14.8% of the total variance and the PC2 accounted for 4.8%. When plotting these principal components (Fig. 7A) it was shown that the first axis separates European and Asian strains on the left from North American strains, representing the pathogen’s center of origin. Between these two groups there are strains from France and Germany (like Pv1419_1) that we know from previous studies that are admixed between American and European populations [21]. The second axis reflects the variability resulting from the genetic gradient observed in Europe between western and eastern populations (Alpine chain barrier), as previously described by Fontaine et al. [2, 3]. PCA analysis indicates that the Cilaos strain Pv8586_1 is genetically related to the western European population of P. viticola.

Fig. 7.

Population structure analysis showing that the Cilaos strain Pv8586_1 from Cilaos (Réunion Island) clusters closely with strains from continental Europe, particularly those from western Europe (France, Spain, and Germany). For the analyses, 92,069 high-quality single-nucleotide polymorphisms (SNPs), obtained from whole-genome data analysis, were used. Analyses included a total of 59 Plasmopara viticola strains: the strain from Cilaos, strains from Canda (n=2), United States of America (n=1), China (n=1), Georgia (n=1), Lebanon (n=1), France(n=15), Hungary (n=15), Switzerland (n=5), Italy (n=7) and Germany (n=8). A Principal Component Analysis (PCA) of the high-quality SNPs, where each dot represents a strain and the color represents the location from where it was isolated. Labeled dots correspond to reference panel of strains of variated origin. B Genetic clustering results inferred using the SNMF (sparse non-negative matrix factorization) algorithm on the SNPs dataset. Strains are indicated of the x axis. Each vertical bar represents an individual strain, labeled along the x-axis. The colors within each bar represent the estimated proportion of ancestry derived from each of K inferred genetic clusters, where each color corresponds to a distinct cluster. Bars with a single dominant color indicate strains with ancestry predominantly from one cluster, while bars with multiple colors indicate strains with ancestry from two or more clusters. K values are shown on the right margin of the plot. Strains were arranged according to their geographical origin, except for French and German strains that showed admixture with American populations; those were plotted close to the American strains. Cilaos strain is indicated by and bold letters. Switz. = Switzerland

The sparse non-negative matrix factorization (SNMF) algorithm was also applied. Results are shown in Fig. 7B and indicated a similar result than the PCA, with the most likely division of the P. viticola population into three groups, with the K=3 having the lowest value of cross-entropy (Supplementary Figure 7). The three populations corresponded to the two groups of mostly European strains and the North-American one. This analysis also confirms the admixed status of some French and German strains with North-American genotypes, and most importantly, this analysis also shows that the Cilaos strain shares a genetic background with the population from western Europe that includes French, Spanish and German strains.

Discussion

Our results demonstrate that the Plasmopara viticola strains (Pv8586_1, Pv8588_1, Pv8583_1) isolated from Cilaos on Réunion Island, are overcoming the resistance conferred by the Rpv1 locus. The virulent phenotype of the Cilaos strains on Rpv1-carrying grapevine varieties is clear and reproducible: the strains produce abundant sporulation and completely escapes host recognition, as evidenced by the absence of HR in all the replicates. Population genomics analysis revealed that the Cilaos strains shares a genetic background closely related to P. viticola strains sampled in continental France, Spain and Germany, suggesting a likely European origin.

In addition, phenotypic assays also showed that the Cilaos strains exhibits a multi-virulent profile, being able to overcome resistance mediated by both the Rpv3.1 and Rpv10 loci. The breakdown of Rpv3.1 in these strains follows the previously described mechanism [20, 21], with a recessive mode of inheritance and involving the same virulence determinants. For Rpv10, however, our results indicate a novel breakdown mechanism. In this case, the loss of resistance is associated with a complete absence of pathogen recognition by the plant, distinct from the incomplete HR previously reported in Europe. Furthermore, the strains exhibited a large homozygous deletion in the corresponding avr locus, representing a previously unreported mutational event that might be associated with Rpv10 breakdown. These findings revealed that P. viticola can evade recognition by more than two resistance genes, exceeding the previously reported maximum of two [13], and is capable of accumulating multiple virulence factors (three in this case) within a single genotype, highlighting its unprecedented evolutionary potential to overcome a broad spectrum of major grapevine resistance loci.

The emergence of these multivirulent strains may have been facilitated by local environmental conditions in Cilaos, which are highly favorable to downy mildew development, in combination with the deployment of grapevine varieties carrying distinct Rpv loci. Local climate could indeed play a role in the emergence of strains capable of breaking down Rpv loci. The Cilaos cirque, situated at an elevation of approximately 600–1,300 m a.s.l. and surrounded by dense forests, has a humid climate with rainfall throughout the year, more abundant in summer, and cool but not cold winters [38, 40]. This is characteristic of a temperate oceanic climate according to the Köppen–Geiger classification, which can more broadly be described as subtropical [39, 41]. These climatic conditions are highly conducive to grapevine downy mildew epidemics [42].

Regarding grapevine cultivation and resistance sources in Cilaos, vineyard areas on Réunion Island have always remained limited, largely due to the challenges posed by the subtropical climate, which hampers dormancy breaking and flower induction [43]. Since the introduction of grapevine downy mildew to the island in the late 19^th century, growers have relied on French–American hybrids that naturally exhibit partial resistance to the pathogen, some of which carry the Rpv3.1 locus. More recently, in 2019–2020, INRAE-ResDur1 varieties [44] combining Rpv1 and Rpv3.1 resistance loci were planted, followed shortly afterward, in 2020, by German-bred varieties such as Solaris and Muscaris, carrying Rpv10 [45]. As a result, all three loci — Rpv1, Rpv3.1, and Rpv10 — are now present in Cilaos, exerting selective pressure on P. viticola. Furthermore, between 2019 and 2023, several vineyards in Cilaos were reported to have received no phytosanitary treatments, which likely contributed to the severity of disease outbreaks (O. Yobrégat, personal communication, 2025).

Consistent with their virulent phenotype, the Cilaos strains exhibit a homozygous deletion of effector genes at the AvrRpv3.1 locus. In addition, they harbor the avr1 allele, one of the most prevalent virulence alleles in France [20]. This might suggest that this allele may have been introduced into Réunion Island through a founder population from mainland France, a hypothesis further supported by with our population genetic structure analyses. Nonetheless, we do not exclude the possibility of an independent adaptation, with mutations converging into the avr1 allele, given the long-lasting cultivation of Rpv3.1-carrying hybrids in Cilaos.

The breakdown of Rpv10 in the Cilaos strains presents a different narrative. Several lines of evidence suggest that the Rpv10 breakdown observed in this study is fundamentally distinct from that described in Europe by Paineau et al. [13] and Dvorak et al. [21]. First, the HR triggered by the plant is not completely suppressed in European strains, leading to only partial breakdown of resistance. In contrast, infection with the Cilaos strains produced no necrosis in any repetition. Secondly, the Cilaos strains lacked the known suppressor allele (S-AvrRpv10) previously associated with Rpv10 breakdown in continental Europe populations, in fact the Cilaos strains harbor a novel mutation at the S-AvrRpv10 locus described by Dvorak et al. [21], featuring a large homozygous deletion encompassing 13 genes coding for secreted proteins. Together, these findings support the hypothesis of a local and independent breakdown of Rpv10 in Cilaos, likely involving a novel virulence mechanism remaining to be confirmed. The genes encompassed by this deletion are excellent candidate effectors for functional studies aimed at demonstrating interaction with the Rpv10 resistance gene.

The breakdown of Rpv10 by the Cilaos strains was unexpected, as these strains were not collected from a plot containing Rpv10 varieties, raising questions about its origin. The vineyard of origin is located 4 km in a straight line from the nearest vineyard planted with Rpv10-carrying varieties (O. Yobrégat, personal communication, 2025). Although P. viticola is reported to disperse over several tens of meters [46, 47], it may be capable of long-distance aerial dispersal, as observed in other oomycete pathogens at continental scales [48]. An alternative explanation is that the breakdown of Rpv10 and Rpv1 in the Cilaos strains represents a case of cross-infectivity. According to the definition of Moury et al. [49], the mutation conferring virulence on Rpv1 varieties could also confer virulence on Rpv10, with selective pressure from Rpv1 being responsible for the Rpv10 breakdown. This could explain why the Cilaos strain, collected from a variety carrying Rpv1 and Rpv3.1, is also virulent on Rpv10. Confirming this hypothesis will require identifying the virulence determinants and further studying the P. viticola population in Cilaos. We discard the possibility of a cross contamination and co-infections as the possible cause for this phenotype since monosporic isolation was performed, and several strains yielded the same results several times. The origin of Rpv10 breakdown in the Cilaos strain, therefore, remains unresolved.

The adaptation of P. viticola to host resistance (Rpv1 and Rpv10) occurred on Réunion Island despite the very limited deployment of resistant varieties and within a remarkably short time frame. Assuming independent breakdown events and disregarding the cross-infectivity hypothesis, virulence against Rpv1 and Rpv10 emerged only four and three years, respectively, after the introduction of varieties carrying these loci in 2019 and 2020. The fast local adaptation capacity of P. viticola has been documented before, including resistance to fungicides [50, 51] as well as for the breakdown of Rpv3.1 which occurred in multiple locations within just five years [22].

In addition to genomic plasticity [24, 26, 52], the evolutionary potential of P. viticola is further enhanced by large population sizes and a mixed reproductive strategy involving both asexual and sexual reproduction [53]. During the grapevine growing season, Rpv loci confer partial resistance to P. viticola by limiting its growth and spread, but at fall, the pathogen can undergo sexual reproduction on resistant varieties allowing recombination between existing genotypes [54]. This genetic exchange increases standing variation in the pathogen population for the following season and can recombine different virulences into one genotype. These dynamics underscore the importance of continuous surveillance of pathogen populations across seasons and, more importantly, highlight the need for new disease management strategies that limit sexual reproduction.

Our findings raise important concerns for the sustainable management of resistant grapevine varieties. They highlight a potential risk of a similar Rpv1 breakdown in continental Europe and the emergence of more strains able to breakdown multiple Rpv loci. This is particularly critical in France where grapevine breeding programs rely on pyramiding Rpv1 with Rpv3.1 and Rpv10 (program ResDur3)[55]. While the less conducive climatic conditions in Europe and the recommended targeted phytosanitary treatments on resistant varieties may reduce the risk of multivirulent strain emergence, evidence from 2024 prompts vigilance. An increase in disease incidence was observed in vineyards planted with the Artaban variety (Rpv1–Rpv3.1) in southeastern France, and subsequent phenotyping assays confirmed that several strains were capable of overcoming Rpv1-mediated resistance [16]. These events, however, remain localized, with outbreaks detected only in that region [56]. Overall, vineyards planted with Rpv1-carrying varieties, such as Bouquet and ResDur lines, continue to benefit from effective resistance conferred by this locus. Nevertheless, these observations underscore the need for ongoing monitoring and further investigation of Rpv1 breakdown.

The emergence of multi-virulent P. viticola strains capable of overcoming major resistance loci including Rpv1, Rpv3.1, and Rpv10, raises serious concerns about the long-term durability of a strategy relying solely on resistance loci pyramiding. This highlights that sustainable management of resistant grapevine varieties must combine intensive field monitoring of grapevine downy mildew epidemics with laboratory-based phenotyping to detect emerging adaptations. Achieving this requires coordinated monitoring networks, standardized methods, reference strains for phenotyping, and stakeholders alert protocols. When integrated with population genomics, such effort can identify avr loci and the genetic mechanisms underlying plant–pathogen interactions, improving our understanding of evolutionary dynamics and enabling early detection of virulence. This underscores the need for future research to pinpoint the genetic determinants of Rpv1 and Rpv10 breakdown on Réunion Island, using approaches such as GWAS or QTL mapping, which have proven effective in this pathosystem [20, 21]. Without such coordinated efforts, decades of breeding could risk being rendered ineffective.

Conclusion

We report the breakdown of Rpv1 by P. viticola in Cilaos by characterizing the first strains capable of overcoming three Rpv loci. Genomic analyses reveal that these strains share a common genetic background with western European populations. Furthermore, we provide evidence for a novel mechanism underlying the breakdown of Rpv10 demonstrating that the phenotype, the virulence determinants and the genetic background are distinct from previously known Rpv10-breaking strains. Our findings raise important concerns for viticulture in Europe. The emergence of these multi-virulent strains that share genetic traits with European populations, suggests that similar breakdowns could soon occur on Europe. While this represents a significant threat to grapevine disease management, our study also provides a timely opportunity to anticipate and respond to such host-adaptation of the pathogen. Beyond its practical implications for resistance monitoring and breeding strategies, this work makes a significant contribution to our understanding of host-pathogen interactions and the evolutionary dynamics that shape the durability of plant disease resistance. Furthermore, it lays the groundwork for further investigation on the identification of the virulence determinants of these new resistance breakdowns.

Abbreviations

P. viticola

Plasmopara viticola

GWAS

Genome Wide Association Studies

QTL

quantitative trait loci

ERF

Efficiency of the Resistance Factor

SNPs

Single Nucleotide Polymorphisms

SNMF

sparse non-negative matrix factorization

PCA

Principal Components Analysis

PC

Principal Component

INRAE

Institut national de recherche pour l’agriculture, l’alimentation et l’environnement

SAVE 

Santé et Agroécologie du vignoble

Authors’ contributions

J.R.M.: Drafting of the manuscript; analysis of phenotypic and genomic data. A.-S.M.: Lead of phenotypic experiments; analysis of phenotypic results; critical review of the manuscript. E.D.: Part of the population structure analysis; critical review of the manuscript; photos and guidance for the description of visual scale of necrosis. I.D.M.: Carried out experiments and provided technical expertise. C.C.: Management of genomic data. L.D.: Lead of phenotypic experiments. F.F.: Critical review of the manuscript. I.H.: Field monitoring and isolation of the strains from Cilaos, Réunion Island. O.Y.: Field monitoring and isolation of the strains from Cilaos, Réunion Island; contribution to drafting a part of the discussion. M.F.-O.: Project leadership; drafting and critical review of the manuscript. F.D.: Project leadership; drafting and critical review of the manuscript.

Funding

This study was funded by the Plant Health and Environment Division of the French National Research Institute for Agriculture, Food and Environment (INRAE), the IdEX Bordeaux University “Investments for the Future” program GPR Bordeaux Plant Sciences, the French National Research Agency (PPR VITAE, grant 20-PCPA-0010), the European Union project GrapeBreed4IPM (Horizon Europe funding program, grant number 101132223), the Plant2Pro® Carnot Institute (ANR agreement 23-CARN-0024-01), and the French government as part of France 2030 (Project BPI France, Innovitiplant).

Data availability

The sequencing data of the Cilaos strains produced in this study has been deposited under BioProject PRJNA1255592.

Declarations

Ethics approval and consent to participate

Not applicable. The present experimental research, including the collection of symptomatic plant tissue and the subsequent isolation of a Plasmopara viticola strains from this material, was conducted in full compliance with applicable institutional, national, and international guidelines and legislation. The resulting strains and associated data are used exclusively for research purposes.

Consent for publication

Not applicable

Competing interests

The authors declare no competing interests.