Extended diversity analysis of cultivated grapevine Vitis vinifera with 10K genome-wide SNPs

Abstract

Grapevine is a very important crop species that is mainly cultivated worldwide for fruits, wine and juice. Identification of the genetic bases of performance traits through association mapping studies requires a precise knowledge of the available diversity and how this diversity is structured and varies across the whole genome. An 18k SNP genotyping array was evaluated on a panel of Vitis vinifera cultivars and we obtained a data set with no missing values for a total of 10207 SNPs and 783 different genotypes. The average inter-SNP spacing was ~47 kbp, the mean minor allele frequency (MAF) was 0.23 and the genetic diversity in the sample was high (He = 0.32). Fourteen SNPs, chosen from those with the highest MAF values, were sufficient to identify each genotype in the sample. Parentage analysis revealed 118 full parentages and 490 parent-offspring duos, thus confirming the close pedigree relationships within the cultivated grapevine. Structure analyses also confirmed the main divisions due to an eastern-western gradient and human usage (table vs. wine). Using a multivariate approach, we refined the structure and identified a total of eight clusters. Both the genetic diversity (He, 0.26-0.32) and linkage disequilibrium (LD, 28.8-58.2 kbp) varied between clusters. Despite the short span LD, we also identified some non-recombining haplotype blocks that may complicate association mapping. Finally, we performed a genome-wide association study that confirmed previous works and also identified new regions for important performance traits such as acidity. Taken together, all the results contribute to a better knowledge of the genetics of the cultivated grapevine.

Fig 1. Mapping of the 9004 SNPs on the whole PN40024 reference genome sequence (assembly version 12X.V0).

The 14 SNPs that allow identifying the 783 V. vinifera cultivars of this study are indicated with a black arrow. Line 1 corresponds to the chromosome number, line 2 to the total number of SNPs and line 3 to the average distance between 2 SNPs on each chromosome (in kb).

Fig 2. Distruct plot of Bayesian population assignments using STRUCTURE and an admixture model with independent alleles frequencies (K = 4).

Each K cluster is defined by a color: K1 in pink, K2 in green, K3 in blue and K4 in yellow and number individuals and origin are specified.

Fig 3. Geographical distribution of the DAPC clusters.

Clusters were positioned according to the main geographic origin represented in each cluster, the contribution of other geographic zones are indicated by lines and dots (see Table 3b); seven geographic regions were considered with IBER (Iberian peninsula) in light yellow, WCEUR (Western and Central Europe) in blue, BALK (Balkans) in light green, ITAP (Italian peninsula) in purple, EMCA (Eastern Mediterranean and Caucasus) and MFEAS (Middle and Far East) in brown, RUUK (Russia and Ukraine) in green and MAGH (Maghreb) in dark yellow, cluster C4 receives a dotted line and is placed in the middle since it is essentially composed of table cultivars used mainly in the New World (NEWO), but also in several European zones (BALK, ITAP, WCEUR and IBER).

Fig 4. Dendrogram based on the Roger’s distance between DAPC clusters.

All nodes received a 100% bootstrap support.

Fig 5. Haplotype non-recombinant blocks along chromosomes.

Only blocks including more than 5 SNPs and larger than 100 kbp in length are represented. Red regions along the main chromosome lines denote non-recombinant blocks calculated on the whole panel of 783 genotypes. Colored bars above it represent non-recombinant blocks calculated on each of the 8 DAPC subgroups. Chromosome length on X axis is in Mb.

Fig 6. Association signals for performance traits detected in the studied panel.

Green dots are the r2 values, while red and blue dots represent the effect of the major and minor allele, respectively. The location of a gene cluster associated with berry skin color (VVMybA1-VVMYbA3, Fournier-Level et al. 2009 [34]) and the boundaries of the sex locus (Fechter et al. 2012 [35], Picq et al. 2014 [36]) are indicated.