Geographical and Cultivar Features Differentiate Grape Microbiota in Northern Italy and Spain Vineyards

Abstract

Recent studies have highlighted the role of the grapevine microbiome in addressing a wide panel of features, ranging from the signature of field origin to wine quality. Although the influence of cultivar and vineyard environmental conditions in shaping the grape microbiome have already been ascertained, several aspects related to this topic, deserve to be further investigated. In this study, we selected three international diffused grapevine cultivars (Cabernet Sauvignon, Syrah, and Sauvignon Blanc) at three germplasm collections characterized by different climatic conditions [Northern Italy (NI), Italian Alps (AI), and Northern Spain (NS)]. The soil and grape microbiome was characterized by 16s rRNA High Throughput Sequencing (HTS), and the obtained results showed that all grape samples shared some bacterial taxa, regardless of sampling locality (e.g., Bacillus, Methylobacterium, Sphingomonas, and other genera belonging to Alphaproteobacteria, Gammaproteobacteria, and Actinobacteria). However, some Operational Taxonomic Units (OTUs) could act as geographical signatures and in some cases as cultivar fingerprint. Concerning the origin of the grape microbiome, our study confirms that vineyard soil represents a primary reservoir for grape associated bacteria with almost 60% of genera shared between the soil and grape. At each locality, grapevine cultivars shared a core of bacterial genera belonging to the vineyard soil, as well as from other local biodiversity elements such as arthropods inhabiting or foraging in the vineyard. Finally, a machine learning analysis showed that it was possible to predict the geographical origin and cultivar of grape starting from its microbiome composition with a high accuracy (9 cases out of 12 tested samples). Overall, these findings open new perspectives for the development of more comprehensive and integrated research activities to test which environmental variables have an effective role in shaping the microbiome composition and dynamics of cultivated species over time and space.

Keywords: Vitis vinifera; fruit microbiome; high throughput sequencing; soil microbiome; wine.

FIGURE 1

PCoA Emperor plots based on Bray–Curtis diversity metric. For each sampling locality (A–C) and cultivar (E–G), samples of soil (triangles) and grape (circles) were compared concerning their microbial community. (A) Overall comparison of grape and soil samples; (B) soil samples; (C) grape samples; (D) overall comparison among grape cultivars; (E) cultivars from the Lombardy regional collection (Northern Italy, NI); (F) cultivars from the germplasm collection of E. Mach Foundation (AI, Alpine Italy); (G) cultivars from the La Rioja collection (NS, Northern Spain). CS, Cabernet Sauvignon; SY, Syrah; SB, Sauvignon Blanc.

FIGURE 2

Bar chart analysis depicting the relative abundance and distribution of the OTUs assigned to class taxonomic rank (A). The legend lists the 18 most abundant Classes. Boxplots (B) show the Faith’s Phylogenetic Diversity (a qualitative measure of community richness incorporating phylogenetic relationships) for each cultivar’s grape and soil sample at each sampling site. Localities and cultivars are reported with the same acronyms detailed in the manuscript and in Figure 1 caption.

FIGURE 3

Venn diagrams showing the number of shared bacterial genera among grape cultivars and soil at each locality (A). In (B), the diagrams show the number of shared bacterial genera among soil samples and grape samples (regardless of cultivar) from the three localities. Localities and cultivars are reported with the same acronyms detailed in the manuscript and in Figure 1 caption.

FIGURE 4

Machine learning analysis performed at the sampling locality level (A), cultivar identity (B), and the combination of the two factors (C). Features’ table was collapsed to the genus level. Overall accuracy levels are indicated as a scatter plot showing predicted vs. true values for each tested sample, along with a linear regression line fitted to the data with 95% intervals (gray shading). Localities and cultivars are reported with the same acronyms detailed in the manuscript and in Figure 1 caption.

FIGURE 5

Heat map showing the relative abundance of the components of grape and soil microbiome mostly contributing to the correct prediction of cultivar-provenance cases in the machine learning analysis. If available, the genus level was considered (italic). In the other cases, we reported the most informative and supported taxonomic rank returned by the classify-consensus-vsearch plugin. Color shading in the heat map indicates the abundance (expressed as log10 frequency) of each genus in the sample. The upper cladogram, shows groups of bacteria genera based on their distribution among samples, whereas the left cladogram shows clusters of grape samples based on genera distribution. The provenance and cultivar of each soil (triangles) and grape (squares) sample are reported on the right of the heat map. Localities and cultivars are indicated with the same acronyms and colors detailed in the manuscript and in Figure 1 caption.