Historical Introgression of the Downy Mildew Resistance Gene Rpv12 from the Asian Species Vitis amurensis into Grapevine Varieties

The Amur grape (Vitis amurensis Rupr.) thrives naturally in cool climates of Northeast Asia. Resistance against the introduced pathogen Plasmopara viticola is common among wild ecotypes that were propagated from Manchuria into Chinese vineyards or collected by Soviet botanists in Siberia, and used for the introgression of resistance into wine grapes (Vitis vinifera L.). A QTL analysis revealed a dominant gene Rpv12 that explained 79% of the phenotypic variance for downy mildew resistance and was inherited independently of other resistance genes. A Mendelian component of resistance–a hypersensitive response in leaves challenged with P. viticola–was mapped in an interval of 0.2 cM containing an array of coiled-coil NB-LRR genes on chromosome 14. We sequenced 10-kb genic regions in the Rpv12+ haplotype and identified polymorphisms in 12 varieties of V. vinifera using next-generation sequencing. The combination of two SNPs in single-copy genes flanking the NB-LRR cluster distinguished the resistant haplotype from all others found in 200 accessions of V. vinifera, V. amurensis, and V. amurensis x V. vinifera crosses. The Rpv12+ haplotype is shared by 15 varieties, the most ancestral of which are the century-old ‘Zarja severa’ and ‘Michurinets’. Before this knowledge, the chromosome segment around Rpv12+ became introgressed, shortened, and pyramided with another downy mildew resistance gene from North American grapevines (Rpv3) only by phenotypic selection. Rpv12+ has an additive effect with Rpv3+ to protect vines against natural infections, and confers foliar resistance to strains that are virulent on Rpv3+ plants.


Introduction
The Amur grape (Vitis amurensis Rupr.) is native to cool climates of the Far East, including Siberia, China, Korea, and Japan [1].V. amurensis is a dioecious species.A few forms of what is believed to be pure V. amurensis with domestication traits-i.e.relatively large berries and hermaphroditic flowers-are grown in northeastern China [2] or reported in literature [3,4].This species is a reservoir of important viticultural traits that are absent from the germplasm of the European wine grape Vitis vinifera.The Amur grape is winter hardy-an adaptive trait to withstand temperatures as low as 240uC in its natural habitat and a long-sought character for cool climate viticulture.Some accessions are not significantly damaged by Plasmopara viticola under conditions highly conducive to downy mildew [5].That is why the species has attracted grape breeders' attention since the collecting expeditions in the Far East  organised by the Russian botanist Vavilov [6].Hybridisation with introduced accessions of V. amurensis became the method of choice to strengthen cold hardiness and downy mildew resistance in V. vinifera.Michurin, Negrul, and Potapenko are credited with the first use of V. amurensis for grape breeding in continental climates.Intentional crosses with V. amurensis also started in China after World War II [7].Until recently, grape breeding in the former Soviet Union and China was isolated from the influence of the Western world, where grape improvement for disease resistance has employed only North American genetic resources for almost a century.A few North American founders gave rise to large descent groups of resistant breeding lines.Several haplotypes of the most documented North American resistance gene for downy mildew (Rpv3) have been introgressed since the mid 1800s into many varieties [8].Rpv3 is located on chromosome (chr) 18 [9] and confers a race-specific hypersensitive response (HR) against P. viticola strains that carry the cognate avirulence factor [10].French-American hybrids carrying the Rpv3 gene have become popular in all European countries committed to grape breeding.Thanks to the networking in the former Soviet bloc, grape breeders in Eastern Europe took advantage of sharing V. amurensis germplasm with the Soviet centres of Michurinsk, Novocherkassk, and Magarach, and since the 1960s this material was introduced into continental Europe [11].Rpv10 is the first mapped resistance gene against downy mildew that was introgressed from V. amurensis into grapevine varieties of Western Europe [12].Many resistant varieties such as Severnyi, Rondo, and Solaris-selected in Russia, Czechoslovakia, and Germanyshare Rpv10 by descent [12] but many others are derived from V. amurensis and show high levels of resistance in the absence of known resistance genes.
In this paper, we report on another resistance gene donated by V. amurensis to existing grape varieties.The locus has been named Rpv12 and listed ahead of publication in the catalogue of mapped resistance loci, following the guidelines of the International Grape Genome Program.Rpv12 is inherited independently of other resistance genes introgressed in cultivated grapevines and has become fixed in the genetic background of several varieties, alone or in combination with a resistant haplotype of the gene Rpv3.

Plant material and P. viticola strains for phenotyping
Pedigrees of custom-made populations and varieties derived from V. amurensis are reported in Figure S1.Rpv3 and Rpv10 haplotypes were determined using markers reported in [8] and [12].Varieties and wild accessions used for genotyping at the Rpv12 locus are listed in Figure S2.Accessions of unknown origin were also genotyped with unlinked microsatellite markers (Figure S3).
Phenotyping of mapping populations was done using biological replicates raised from rooted cuttings taken from seedlings.In field experiments, vineyards were maintained following standard crop practices and withholding the application of fungicides with activity against P. viticola.Resistance was scored during two consecutive seasons using the OIV452 descriptor.For leaf disc assays, grapevines were grown in a greenhouse prior to inoculation.QTL analysis was performed upon inoculation with the NE-I isolate of P. viticola [9].Phenotyping of recombinant individuals was extended to two isolates of P. viticola characterised by [10].Sixteen leaf discs were excised from the fourth and fifth leaves beneath the shoot apex of two clonal replicates and plated onto wet paper with the abaxial side up.Discs were sprayed with a suspension of P. viticola at 150,000 sporangia ml 21 and incubated at 21uC and day length of 16 h, for 8 days.Presence/absence of HR and extent of pathogen sporulation were scored as described in [9].Briefly, the OIV452 scale is positively correlated with the magnitude of plant response and inversely correlated with the severity of downy mildew symptoms.A cumulative rating (gOIV452) was obtained by summing the scores from 3 to 8 days post inoculation (dpi).Field experiments were conducted at the Experiment Farm 'A.Servadei' owned by the University of Udine, Italy.The land accessed is not a protected area.No protected or endangered plant species were sampled.All necessary permits for the described field studies were issued by the Regional Government of Friuli Venezia Giulia, Italy, Authorisation no.D.P.Reg.n.0198/Pres.2004no.322/2007.

QTL analysis
Microsatellite and RGA markers were mapped as described in [13].Parental genetic maps were constructed with JoinMap 4.0.Linkage groups were built at an LOD threshold of 4.0, and chromosome number was assigned according to [14].QTL analysis was performed by interval mapping using MapQTL 5.0.
The LOD threshold for significant QTLs at a = 0.05 was estimated using a permutation test with 1,000 iterations.The confidence interval was initially defined as a one-LOD support interval and then refined using informative recombinants with extreme phenotypes for gOIV452 supported by scores of the HR.

Microsatellite and SNP development, genotyping, and haplotype reconstruction
Microsatellite and SNP markers were designed in the upper arm of chr14, and used for linkage mapping and haplotyping (Figure S4).SSR amplicons were sized using an AB3730 capillary sequencer and alleles were called with GeneMapper v4.0 (Applied Biosystems).For SNP discovery and mapping, primers were designed in exons of PN40024 single-copy genes.Amplicons were captured by Agencourt AMPureXP (Beckman Coulter) magnetic beads and used for Sanger sequencing of both DNA strands.Sanger reads were trimmed using Phred and consensus sequences were created using Phrap.Nucleotides were referenced to the chromosomal coordinates of the PN40024 12X assembly by BlastN alignments.For SNP haplotyping, a total of ,3 billion Illumina 100-bp paired-end reads were obtained from nuclear DNA of twelve grapevine varieties (Aglianico, Barbera, Cannonau, Carignan, Kishmish vatkana, Nebbiolo, Pinot noir, PN40024, Primitivo, Sangiovese, Schiava grossa, Vermentino), quality trimmed, and filtered for contaminants and duplicates.Paired reads were aligned with Burrows-Wheeler Alignment (BWA) to the Rpv12 + consensus sequence.Raw sequencing data are deposited in the Sequence Read Archive (SRP018751).Variants were called with VarScan.v2.2.8 using uniquely mapped reads.Haplotypes were determined using PHASE, guided by the evidence of neighbouring phased variants present within a read or in paired reads.Cladograms were constructed using the UPGMA method.The most informative SNPs were validated by PCR amplification and Sanger sequencing in an extended set of grapevine varieties (Figure S1-S2).Measures of DNA sequence variation were estimated using DnaSP.
DM resistance segregated in a bimodal fashion in the progeny of the Rpv12 + /Rpv12 2 , Rpv3 2 /Rpv3 2 parent (Figure 1A).The Rpv12 + locus was located on chr14 and explained 78.7% of the phenotypic variance for the OIV452 parameter (LOD = 15.4) in field-grown grapevines exposed to natural infection (Figure S5C).The QTL peak was associated with markers UDV014 and UDV370.In the progeny, the presence of the UDV014 and UDV370 alleles in coupling with Rpv12 + was associated with an average phenotypic value of 8.68 for the OIV452 parameter, their concomitant absence with a value of 2.96.
QTL analysis in the progeny of the double heterozygous resistant parent (Rpv12 + /Rpv12 2 , Rpv3 + /Rpv3 2 ) confirmed the location of the QTL peak in the interval delimited by markers UDV014 and UDV370, and the independent assortment of Rpv12 and Rpv3 alleles (Figure S5D).In the leaf disc assay, Rpv12 explained 74.5% of the phenotypic variance for the gOIV452 parameter (LOD = 41), while 8.7% of the residual variance was accounted for by a QTL peaking between markers UDV305 and UDV737 on chr18 (LOD = 12.4), which corresponded to the Rpv3 locus (Figure 1B-C).Maximum LOD for the Rpv12 locus was associated with the markers rgVvin180 and UDV304 previously designed in NB-LRR genes [15], which were located within the one-LOD interval delimited by markers UDV014 and UDV370.Rpv12 had a stronger effect than Rpv3 on the ability of the seedlings to restrict pathogen sporulation, and the two loci had an additive effect.Compared to sensitive double homozygous recessive seedlings that scored an average gOIV452 value of 25.4,Rpv3 + seedlings scored 31.2 (+23%), which rose to 39.6 (+56%) in Rpv12 + seedlings.The combination of both resistance genes in Rpv12 + Rpv3 + seedlings was associated with the highest gOIV452 value of 42.7 (+68%).

The Rpv12 phenotype
Rpv12 + conferred the ability to establish an HR within 24-48 hours post inoculation and to restrict sporulation of P. viticola (Figure 2C, G).The type and the timing of the plant reaction are similar to those triggered by Rpv3 [9], but the limitation imposed on pathogen sporulation is more significant (Figure 1B, Figure 2B,  C).Rpv12-dependent HR was successfully established in response to at least one isolate of P. viticola that eludes or suppresses the resistance conferred by Rpv3 (Figure 2F, G).The isolate Pv127 that equally sporulated on Rpv3 2 and Rpv3 + hosts (Figure 2E, F and [10]) was halted on Rpv12 + grapevines to the same extent as the other isolate (Figure 2).The combination of the resistance alleles Rpv12 + and Rpv3 + restricted pathogen growth to the highest extent (Figure 1B), regardless of the pathogen strain (Figure 2D, H).

Mendelisation and fine mapping of Rpv12
The genetic interval underlying the QTL peak for DM resistance was delimited between markers UDV014 and UDV370, based on the assignment of each offspring as being either able or unable to establish the HR (Figure 1A).Validation of this interval was performed in a distinct family of 185 seedlings created by crossing a sensitive grapevine with one of the earliest available ancestors of the Rpv12 + lineage-the variety Kunbara `t (Figure S6).Fine mapping of Rpv12 + was initiated from the markers UDV014 and UDV370, which were used for DNA genotyping of 2,532 individuals segregating for Rpv12 + (Figure S1).The occurrence of 16 recombinants in the interval (0.6 cM) allowed us to dissect the locus by the targeted design of microsatellite and SNP markers (Figure S7A).To this end, we amplified 14 fragments in 8 single-copy genes from genomic DNA of an Rpv12 + homozygous line (hereafter called '156,623') and sequenced 10,675 nucleotides of the Rpv12 + haplotype (Figure S4).The Rpv12 + haplotype was compared with 23 haplotypes identified in 12 varieties of V. vinifera.The nucleotide diversity p value was 0.00313 in the sample of V. vinifera, and the number of net nucleotide substitutions per site was 0.00668 between the Rpv12 + haplotype and V. vinifera haplotypes.All variable sites between Rpv12 + and V. vinifera haplotypes are indicated in Figure S7B-C, and those variants uniquely found in Rpv12 + are highlighted.
Informative SNPs were mapped on the recombinant chromosomes available to this study and reduced the genetic interval of Rpv12 + to a 0.24-cM region spanning a cluster of coiled-coil (CC)-NB-LRR genes (Figure S7A).The borders of the locus are defined upstream by marker SNP 14_ 9036110 (chr14:9.036Mbp) in the gene pentatricopeptide repeat-containing protein (GSVIVT00036207001) and downstream by marker SNP 14_ 9882587 (chr14:9.882Mbp) in the gene serine/threonine protein phosphatase (GSVIVT00023546001).Another five genes (peptidyl-prolyl cistrans isomerase CYP19-3, bHLH104-like transcription factor, AP2/ERF domain-containing transcription factor, hypothetical protein coding region, and activating signal cointegrator) are located between the upper edge of the NB-LRR cluster and the upper border of the locus.Rpv12 + is removed from the upstream SNP 14_ 9036110 by 2 events of crossing over in 2,532 meioses and by 4 crossovers from the downstream marker SNP 14_ 9882587.The resistant haplotype in the homozygous line '156,623' has cytosines at both SNP 14_ 9036110 and SNP 14_ 9882587 positions, which define the CC haplotype of the resistant allele Rpv12 + and distinguish it from the TA haplotype present in all accessions of V. vinifera tested (Figure S2).The C variant at either marker is also present in wild accessions of V. amurensis and V. coignetiae, but we did not find the CC haplotype in any of the V. amurensis breeding lines that do not share Rpv12 + by descent, i.e. 'Severnyi', 'Beichun', and 'Michurinski' (Figure S2).The V. vinifera haplotype TA is ancestral to the V. amurensis haplotype CC because the haplotype TA was also found in other Asian and North American species (Figure S7D).For DNA tests in controlled-cross progeny, these universal SNP markers can be replaced without substantial increase in recombination frequency by nearby microsatellite markers, which reveal allelic variants by size and are heterozygous in Rpv12 + varieties (Figure S8L).On the upper side of Rpv12, UDV343 and UDV345 cosegregate with SNP 14_ 9036110 while UDV350 is placed 0.08 cM outwards; on the lower side, UDV360 is 0.04 cM outward from SNP 14_ 9882587.Another six microsatellite markers (UDV340, UDV347, UDV348, UDV349, UDV380, and UDV390) cosegregate with the Rpv12 phenotype as well as marker SNP 14_ 9036110 (chr14:9.257Mbp), which is located in the single-copy gene (Activating signal cointegrator, GSVIVT00036192001) next to the upper edge of the NB-LRR cluster.
The number of net nucleotide substitutions per site between Rpv12 + and V. vinifera haploypes decreased from each flanking region towards the Rpv12 + NB-LRR gene cluster (Figure S7D).We also detected a reduction of haplotype diversity, which was confirmed by an extended analysis in other Vitis species.In the activating signal cointegrator gene, we failed to find any allelic variant that distinguished the Rpv12 + haplotype from all varieties of V. vinifera, while diversity was detected within the population of V. vinifera and among accessions of other Asian and North American species (i.e.V. yenshanensis, V. monticola, V. cordifolia, Figure S7B).

Origin of the Rpv12 + haplotype and linkage drag in introgression lines
The Rpv12 + haplotype donated by the original individual of V. amurensis was inferred from the comparison of the most ancient varieties 'Zarja severa' and 'Michurinets' that are heterozygous for Rpv12 + /Rpv12 2 and were selected from the first filial generation of pure V. amurensis x V. vinifera crosses (Figure S8G).Phasing of marker alleles in the Rpv12 + haplotype was determined using the seedling '156,623', which carries Rpv12 + in a homozygous state from marker sc30_4 (chr14:4.025Mbp) to VMC2H5 (chr14:20.981Mbp), and the variety V. vinifera 'Italia', because it shares with 'Michurinets' the V. vinifera haplotype from marker VVIC20 (chr14:3.125Mbp) to marker VVIP22 (chr14:18.364Mbp).This analysis also included the control DNA samples of 'PN40024' and its ancestor 'Schiava Grossa', which donated the chromosome fragment containing the Rpv12 2 haplotype to 'PN40024' (see [14] for the pedigree of 'PN40024').
The extent of linkage drag around Rpv12 + was estimated in 21 varieties phenotypically selected from different generations of backcross, and compared to F1 hybrids (Figure S1, S8L).In 'Zarja severa', 'Michurinets', and '156,623', the linkage around Rpv12 + persisted until marker VMC1E12 (chr14:7.137Mbp) upstream of Rpv12 + and until marker VMC2H5 (chr14: 20.981 Mbp) downstream of Rpv12 + .In varieties further removed from the wild parent, linkage drag was eliminated in '13-12-15' immediately downstream of Rpv12 + in the serine/threonine protein phosphatase gene next to the NB-LRR cluster (chr14:9.882Mbp Figure S2), which was confirmed by the alleles at the nearby marker UDV360 (chr14:9.910Mbp, Figure S8L).'Bruskam', 'Kunleany', and 'Taurus' also have minimal linkage drag downstream of Rpv12 + .On the other side of the Rpv12 locus, 'Stepnyak' had the least linkage drag of wild alleles in coupling with Rpv12 + , which persisted for 1.4 Mbp away from the upper edge of the NB-LRR cluster but were not present from VMC1E12 to the top of chr14.With the exception of 'Stepnyak', the varieties that share Rpv12 + are all identical to one another and with the hybrids 'Zarja severa' and 'Michurinets' in the chromosomal region comprised between VMC1E12 and UDV340, but they differ upstream in completely or partially sharing a haplotype with either 'Zarja severa' or 'Michurinets'.
In a set of 48 accessions of V. amurensis and early descendents of V. amurensis x V. vinifera crosses conserved in breeding centres, we did not find any DNA profile to be compatible with the parent that donated Rpv12 + to 'Zarja severa' or 'Michurinets' (Figure S4).We were also unable to find an offspring or a full/half-sibling of 'Zarja severa' or 'Michurinets' among the selections that carry Rpv12 + (Figure S4 and S9).All resistant varieties that contain Rpv12 + tested negative for the resistant haplotype Rpv10 + transmitted by 'Severnyi' and vice versa (Figure S1).However, some Rpv12 + varieties did test positive for the resistant haplotype of Rpv3 transmitted by 'Seibel 4614' (Figure S1).Before this molecular characterisation of Rpv12 + donors, two resistant descendents of V. amurensis ('13-12-17' and '13-12-15') were crossed for breeding purposes and the ten most resistant offspring were retained (Figure S1, family G).Since '13-12-15' was the only parent to contain Rpv12 + , we genotyped all phenotypically selected offspring and assessed that all contained Rpv12 + .

NB-LRR genes in the Rpv12 locus and in the upper arm of chromosome 14
There are 13 CC-NB-LRR genes in the Rpv12 locus of the reference grapevine genome with a high degree of sequence identity (Figure S7F).This array of NB-LRRs spans ,600 kb and it is interspersed with another three genes with unrelated functions (asparagine synthetase B, cytochrome P450 89A2, and a hypothetical protein coding region).The Rpv12 NB-LRR array is part of a more complex structure of 46 clustered NB-LRRs in the upper arm of chr14 (Figure S8N).All NB-LRRs are of CCtype and are physically located in the uppermost 14 Mbp of the chromosome.The pattern of repetitive elements and recombination rate consistently predict the location of the centromere at ,15 Mbp between markers SC8_0062_ 024 and VVIP22 (Figure S8A,C).All but one NB-LRR on chr14 arose from intrachromosomal duplications and form a clade that roots deep in the cladogram of the NB-LRR gene family (Figure S8M,N,P).In addition to the 13 CC-NB-LRR genes in the Rpv12 locus, which derived from short-range duplications, other CC-NB-LRR genes are organised into another four clusters, with the exception of an isolated copy at 14.6 Mbp that is of CC-type but unrelated with respect to the others (Figure S8O).Outside of the Rpv12 locus, the closest NB-LRR genes are arrayed ,3 Mbp away towards the upper telomeric end, and ,2 Mbp away towards the centromere.

Estimation of the sequence gap in the Rpv12 locus of the reference grapevine genome
The assembly of the grapevine reference genome has a scaffold junction of unknown size within the Rpv12 locus due to a high degree of duplication that hampered scaffolding of the arrayed NB-LRR genes.The NB-LRR sequences in the Rpv12 locus have all of their reciprocal best matches in the contiguous terminal ends of scaffolds 81 and 36.We did not find highly similar paralogues in unanchored scaffolds, easing the concern that the gene array in PN40024 may contain more locally duplicated copies.
In order to estimate the size of the physical gap, we used the physical contig_1449 of the 'Cabernet Sauvignon' BAC library, which tested positive for the NB-LRR marker UDV304 [15] that was genetically mapped to Rpv12 (Figure S7E) but lacked any match of its primer sequences in the PN40024 reference sequence.We selected all unique end sequences of the BAC clones included in contig_1449 and projected them onto the PN40024 assembly by BlastN search.All sequences that passed the thresholds of .50%sequence aligned and .90%sequence identity matched the contiguous terminal ends of scaffolds 81 and 36.In particular, BAC_33J08 spanned the scaffold junction.One end of BAC_68L04-the mate of the sequence from which marker UDV304 was developed-matched sequences in scaffold 81, while the other end was repetitive within the Rpv12 NB-LRR cluster.

Discussion
The Rpv12 locus coincides with a cluster of CC-NB-LRR genes on chromosome 14 and the inheritance of the resistant haplotype is associated with a localised HR in leaves challenged with P. viticola.Rpv12 adds to the list of functional NB-LRR genes introgressed into the V. vinifera genome for improving downy mildew resistance, now including Rpv1, Rpv3, Rpv10, and Rpv12 [9,12,16]-the latter two being donated by accessions of V. amurensis.Rpv12 + plants are able to mount a type of host resistance similar to that observed in North American grapevines, raising the intriguing question as to how the functionality of Rpv12 against P. viticola has evolved in the Asian V. amurensis without co-evolution and in the absence of the pathogen.
Compared to the loci Rpv1 and Rpv3, which contain TIR-NB-LRR genes, Rpv12 contains exclusively CC-NB-LRR genes, a typically less dynamic subclass.According to [17], TIR-NB-LRR genes undergo sequence exchange at higher rates and with greater functional constraints to keep pace with changes in the population of specialised pathogens, while CC-NB-LRR genes are evolutionary more stable, preserving established functionalities against a wider range of less-adapted biotic threats.The presence in the Asian continent of Plasmopara cissii and Plasmopara amurensis that are adapted to other Vitaceae species [18] suggests that the Rpv12 locus of the genus Vitis contains functional remnants against Plasmopara species with loose recognition specificity.In our experiments, Rpv12 + protects grapevines from the threat of downy mildew more effectively than the North American gene Rpv3, by expanding the recognition spectrum to isolates of the pathogen that are virulent on Rpv3 + plants.
Rpv12 is inherited independently of another resistance gene Rpv10 introgressed from V. amurensis into varieties of V. vinifera [12].DNA tests provided strong support for the hypothesis that 'Zarja severa', which contains Rpv12 + , and 'Severnyi', which contains Rpv10 + , are full-siblings.According to historical records, the cross that gave rise to 'Zarja severa' and 'Severnyi' was made in 1936 by Potapenko and Zakharova at the Michurin Central Genetics Laboratory in Michurinsk, Russia.The reported direction of the cross indicates that V. amurensis was the pollen parent and, if purely wild, it should have been a staminate plant.Under this hypothesis, Rpv12 + and Rpv10 + were donated by the same male accession of V. amurensis, and they segregated in the selected offspring.Another possible direct descendent of the same resistant accession of V. amurensis is 'Michurinets', which has a coefficient of relatedness compatible with sharing one parent with 'Zarja severa' and 'Severnyi', and contains Rpv12 + but not Rpv10 + .None of the other varieties that were reported to originate from the first filial generation of V. amurensis and contain Rpv12 + ('Bruskam', '13-12-15', and '13-12-11/K10') are full-siblings or half-siblings with 'Zarja severa' and 'Michurinets' (Figure S9), falsifying the hypothesis that they have a parent-offspring relationship with the same wild accession that donated Rpv12 + to 'Zarja severa' and 'Michurinets'.In the set of breeding lines available to this study, we identified 28 varieties and breeding lines that contain Rpv12 + , and another 12 that contain Rpv10 + , but we did not find any variety with both Rpv10 + and Rpv12 + .It is curious that seedlings containing only one of the two genes were accidentally selecteddespite the expectation that seedlings retaining both genes are phenotypically superior-and that the genes did not become combined again by crossing lineages that descended from 'Zarja severa/Michurinets' and 'Severnyi', despite their diffusion in many breeding centres.In contrast, empirical evidence has guided breeders to cross grapevines that we now know to carry the V. amurensis Rpv12 + haplotype with French-American lines of the 'Seibel 4614' lineage that we now know to carry an Rpv3 + resistant haplotype.Pyramidisation of resistance genes was realised before the advent of DNA testing by the phenotypic selection of the best parents to cross and the most resistant seedlings to retain (Figure S1).The resistance phenotype in Rpv12 + /Rpv3 + grapevines was fortified by the combination of two loci that contain CC-NB-LRR and TIR-NB-LRR genes.
An additional 10 descendents of V. amurensis x V. vinifera crosses have neither Rpv12 + nor Rpv10 + (Figure S1).There is evidence to suggest that some of them ('Beichun', 'Bashkanskii krasnyi', data not shown) are able to significantly restrict the sporulation of downy mildew, which leads to the assumption that other resistant haplotypes/genes have been introgressed from V. amurensis.
In this study, the resistant phenotype conferred by Rpv12 + was always dependent on an HR and significant limitation-but not complete absence-of pathogen sporulation in different genetic backgrounds.This phenotype differs from the complete resistance claimed in literature reports for another resistance locus identified in wild V. amurensis (Rpv8, [4]).These phenotypic observations weaken the hypothesis that Rpv8 and Rpv12 resistance specificities are conferred by the same haplotype, prompting us to name the haplotype analysed in this paper separately from Rpv8.The low resolution of the QTL plot of Rpv8 described in [4] does not allow the exclusion of the possibility that Rpv12 and Rpv8 are either alleles in the germplasm of V. amurensis or paralogous loci.In the upper arm of chr14 of the reference grapevine genome, there are 48 NB-LRR genes, pseudogenes, and gene fragments, 13 of which are physically located in the genetic interval of Rpv12.Another two major clusters are located ,2 Mbp and ,3 Mbp away.
The V. amurensis Rpv12 + haplotype is highly diverse with respect to V. vinifera haplotypes (Figure S7), but nucleotide diversity was lower in the single-copy genes most proximal to both sides of the Rpv12 + NB-LRR cluster than in more distant genes.In order to assess whether nucleotide diversity in the Rpv12 locus is also reduced in other Vitis accessions, we sequenced the GSVIVT00036207001 gene fragment in all wild accessions available to this study.The haplotypes detected in the single-copy gene next to the upper edge of the NB-LRR cluster conflicted with the phylogeny, as we observed more differentiation within species (i.e.among varieties of V. vinifera) than between Eastern Asian-Eastern North American disjuncts, and more differentiation between haplotypes of a single diploid accession than between haplotypes of different species (Figure S7D).It is possible that this inconsistency is due to the amount of incomplete lineage sortingthe stochastic retention of ancestral polymorphisms in recently diverged species [19].Divergence among Vitis species has in fact occurred in the past 6 million years [20] and lineages in V. vinifera have larger widths than lengths [21], in the absence of significant bottlenecks during the recent history of domestication and cultivation [22].
By analysing varieties and new seedlings that carry Rpv12 + , we showed that recombination is not suppressed at the chromosomal scale between the V. amurensis haplotype Rpv12 + and haplotypes in V. vinifera.During the process of introgression into the V. vinifera background, linkage drag could be minimised on each side of the Rpv12 NB-LRR gene cluster.In the past, without the assistance of DNA typing, the introgressed segment around Rpv12 + has been shortened in backcross generations, and the varieties '13-12-15' and 'Stepnyak' are those with the least linkage drag.For practical breeding, we demonstrated that screening ,2,500 new seedlings was sufficient to identify four recombinants at the very lower edge of the NB-LRR gene cluster and, on the other side, two recombinants that retained only five genes of the resistant haplotype past the NB-LRR gene cluster.We expect that linkage drag around the NB-LRR gene cluster will be eliminated by backcrossing these individuals to V. vinifera and raising a few thousand seedlings.

Conclusions
A downy mildew resistance gene of the Asian species V. amurensis was initially introgressed by Soviet breeders into V. vinifera and then combined by European breeders with other resistance genes from North American grapevines.Our molecular analyses showed that grape breeders over the past century were visionary in the choice of the parents to cross, the seedlings to retain, and the resistant lines to combine.Marker assisted selection is now feasible with the DNA information made available by this paper.We showed that the haplotype analysis of the introgressed segment provides plenty of markers to track the gene and guide the elimination of linkage drag around the introgressed gene.Background selection is key to the quality of the new grape varieties created by conventional breeding.(panel B) variable sites between Rpv12 + and V. vinifera haplotypes in six single-copy genes at symmetrical positions with respect to Rpv12 (the analysis was extended to additional Vitis accessions in the least variable genes (panel C); plots of nucleotide diversity between sites comparing Rpv12 + and V. vinifera haplotypes, and haplotype diversity in the set of haplotypes (panel D); physically assembled BAC clones of 'Cabernet Sauvignon' that cover the scaffold junction in the PN40024 sequence assembly in the middle of the Rpv12 locus (panel E); sequence identity and hierarchical clustering between physically arrayed NB-LRR gene copies in the Rpv12 locus (panel F). (JPG)

Figure 1 .
Figure 1.Phenotypic distribution of downy mildew resistance.Two families segregating for Rpv12 + (panel A) and for the combination of Rpv12 + and Rpv3 + (panel B) were analysed.Resistance scores in panel A are based on the OIV452 parameter (1 = most sensitive, 9 = most resistant) scored on field-grown seedlings under natural infection.Resistance scores in panel B are based on the cumulative OIV452 parameter (gOIV452 = sum of daily OIV452 scores from 3 to 8 dpi) in artificially inoculated leaf discs.The average phenotypic value in the upper left corner of the panels A-B refers to individuals grouped by their allelic status at the Rpv12 and Rpv3 genes, which was estimated based on the flanking markers UDV014/ UDV370 for Rpv12, and on UDV305/UDV737 for Rpv3 [8].Recombinants in those intervals were excluded from this estimate.QTL plots that explain the phenotypic variance shown in panel B are given in panel C. doi:10.1371/journal.pone.0061228.g001

Figure S1
Figure S1 Pedigrees of custom-made populations for genetic mapping, and varieties containing the resistant haplotype of Rpv12.(JPG) Figure S2 List of 200 grapevine accessions and their genotypes at the Rpv12 locus.DNA typing was performed by Sanger sequencing of the gene fragments GSVIVT00036207001 and GSVIVT00023546001 and by scoring nucleotide variants at the chromosomal positions chr14:9036110 and chr14: 9882587, flanking the Rpv12 NB-LRR cluster.(JPG) Figure S3 DNA fingerprinting and kinship of 48 accessions of V. amurensis, descendents of V. amurensis x V. vinifera crosses with unknown pedigree, and introgression lines in the first two generations of descent with known pedigree.(JPG) Figure S4 Microsatellite and SNP markers on chr14.(JPG) Figure S5 Genetic maps and QTL plots for downy mildew foliar resistance (OIV452 parameter) in an Rpv12 + /Rpv12 2 Rpv3 2 / Rpv3 2 resistant parent (panels A and C) and in an Rpv12 + /Rpv12 2 Rpv3 + /Rpv3 2 resistant parent (panels B and D).For the sake of simplicity, cosegregating markers are shown only once on the xaxis of the QTL plots.(JPG) Figure S6 Validation of the Rpv12 genetic interval in progeny derived from the cross 'Kunbara `t' x 'Sarfeher'.(PDF) Figure S7 Description of the Rpv12 region in the reference genome, in the Rpv12 + resistant haplotype, and diversity among grapevines: (panel A) recombination in custom-made populations;(panel B) variable sites between Rpv12 + and V. vinifera haplotypes in six single-copy genes at symmetrical positions with respect to Rpv12 (the analysis was extended to additional Vitis accessions in the least variable genes (panel C); plots of nucleotide diversity between sites comparing Rpv12 + and V. vinifera haplotypes, and haplotype diversity in the set of haplotypes (panel D); physically assembled BAC clones of 'Cabernet Sauvignon' that cover the scaffold junction in the PN40024 sequence assembly in the middle of the Rpv12 locus (panel E); sequence identity and hierarchical clustering between physically arrayed NB-LRR gene copies in the Rpv12 locus (panel F).(JPG)

Figure S8
Figure S8 Introgression of the Rpv12 + region in grapevine varieties and the surrounding chromosomal landscape.Genetic (panel A) and physical (panels D,H) distance of mapped markers; density of exons, class I and II TEs, unclassified high-copy sequences, and low-copy DNA (panel C); density of NB-LRR genes (panels E,I) and their phylogenetic relationships (panels M-P); and genotypes at markers in 23 varieties (panel L) are given at the chromosomal scale (panels D,H).(JPG) Figure S9 DNA fingerprinting and kinship of six accessions claimed to belong to the first filial generation of V. amurensis x V. vinifera crosses.(JPG)