Comparative Analysis of Grapevine Epiphytic Microbiomes among Different Varieties, Tissues, and Developmental Stages in the Same Terroir

: There is limited knowledge about the relationships of epiphytic microbiomes associated with the phyllosphere of different Vitis vinifera cultivars in the same vineyard and terroir. To address this research gap, we investigated the microbiome compositionof 36 grapevine genotypes grown in the same vineyard in different plant sections during the growing season. Using high-throughput NGS-based metagenomic analysis targeting the ITS2 and the V4 regions of the 16S ribosomal gene of fungal and bacterial communities, respectively, weassessed the impact of grapevine genotypes on microbial assemblages in various parts of the phyllosphere. The results indicated that different phyllosphere tissues display high microbial diversity regardless of the cultivars’ identity and use. The selected three phyllosphere parts representing three distinct phenological stages, namely bark and bud, berry set, and fruit harvest, had almost a similar number of fungal OTUs, while a difference was recorded for the bacterial species. The fruit harvest stage hosted the highest number of bacterial OTUs, whereas the bark and bud stage contained the lower. Bacterial dominant phyla were Proteobacteria, Bacteroidetes, Actinobacteria, and Firmicutes, and the genera were Gluconacetobacter, Erwinia, Gluconobacter, Zymobacter, Buchnera, Pseudomonas, Pantoea, Hymenobacter, Pedobacter, Frigoribacterium, Sphingomonas, and Massilia . For fungi, the dominant phyla were Ascomycota and Basidiomycota, and the genera were Aureobasidium, Cladosporium, Alternaria, Aspergillus, Davidiella, Phoma, Epicoccum, Rhodosporidium, Glomerella, Botryosphaeria, Metschnikowia, Issatchenkia, and Lewia . Both the genotype of the cultivar and the phenological stage appeared to considerably impact the shape of microbial diversity and structure within the same terroir. Taken together, these results indicate that microbiome analysis could be proved to be an important molecular ﬁngerprint of cultivars and provide an efﬁ-cient management tool for the traceability of wine and grape end products. Moreover, the unique identity of cultivars’ microbial signatures highlights the need for further development of precision management to support viticulture sustainability in the face of climate change.

The most widely studied Vitis-associated microbes are known pathogens and fermentative yeasts.Those include bacteria and fungi such as Agrobacterium spp.[26], Xylella fastidiosa [27], Xylophilus ampelinus [21], Botrytis cinerea [28], Erysiphe necator [29], and Plasmopara viticola [30], causing crown gall, Pierce's disease, blight, gray mold, powdery mildew, and downy mildew, respectively.Among fermentative yeasts (either Saccharomyces or non-Saccharomyces species), S. cerevisiae predominates in must fermentation [31][32][33].Based on the knowledge derived from such studies, technologies have been developed for controlling Vitis pathogens with biological antagonists, improving soil fertility with rhizobial inocula, and improving the wine fermentation process and quality traits with biological commercial products.Nevertheless, Vitis-associated microbes reflect direct and indirect relationships with their host and environment [5,6,10,13,17,23,[34][35][36][37][38][39][40].In a fundamental study, Bokulich et al., 2014 [4] reported that Vitis microbiomes present strong biogeographic regionalization, indicating that local environment shapes distinct microbial consortia, comprising Vitis microbial terroir.Further in-depth analysis implied that the host cultivar might also have an additional controlling effect on Vitis microbial communities across different environments.Thus, apart from the environment, the effect of Vitis genotype on the composition of the hosted microbiome needs to be thoroughly investigated.Additionally, the nature of genotype interaction with tissue and developmental stage on microbiome composition is largely unknown.
A recent metagenomic study revealed that the host genotype may influence Vitis associated bacteria in the same farming system and vineyard [17].This exciting new evidence leads to three key questions:(a) Does the host genotype influence the structure of associated fungal communities as well?(b) Does developmental stage and tissue type affect the composition of microbial communities?Lastly,(c) is there a common core microbiome linked to all Vitis cultivars?Considering the recent view of individual plants as holobionts, where hosted microorganisms are involved in major plant functions such as nutrition and resistance to biotic and abiotic stresses [41], the composition and organization of the microbial community could be a major determinant when selecting a grapevine cultivar.However, the detailed microbial colonization process of grapevine is still poorly understood [42].
Thus, the aim of this study was to comprehensively characterize the role of Vitis cultivar genotype on the structure of the associated microbiome, targeting different epiphytic sections and developmental stages.We sampled microbial communities associated with 36 domestic and international grapevine varieties to provide insights into microbial community size and structure in the semi-conventional vineyard throughout the growing season.Using a high-throughput metagenome analysis for bacterial and fungal fingerprints, the microbiome communities were assessed, for the first time to our knowledge, in such a large collection of grapevine cultivars under the same environment and farming system.

Vineyard, Vine Cultivars and Sampling
Plant material was collected from the Vine Cultivar Collection (VCC), School of Agriculture, Aristotle University of Thessaloniki (AUTh).The vineyard is located at the University Farm (N40.53829,E22.99633).The soil is calcareous sandy loam, and the cultivation follows a semi-conventional low-input system.More specifically, synthetic fertilizers are applied after soil fertility evaluation as needed, and plant protective fungicides, sulfurand Bordeaux mix, are applied three times per year.Fungicide (Switch; Syngenta, Greece) is applied against Botrytis once per year.
The VCC collection hosts more than 150 grapevine varieties of Greek origin along with several of the most important international varieties, including Cabernet Sauvignon, Sauvignon Blanc, Merlot, Syrah, and Riesling, among others.Tissue samples were collected from 36 varieties (names of the varieties are indicated in Tables S1 and S2) and consisted of bark and buds in early spring (March 2019), immature berries at the phenological stages of berry set in the spring season (May-June 2019), and mature berries at harvest (August-September 2019).
Bark and bud tissue samples (10 g) were collected aseptically from a single cane per vine and 5 independent vines (trees) from each cultivar, pooled together, equally mixed, and stored at −80 • C for further analyses [17].For berry sampling, undamaged berries, including their pedicels at the phenological stages of berry set (Stage I, 20 g per sample) and harvest (Stage II, 60 g per sample), were taken from 5 plants from each cultivar.The samples from each phenological stage were placed in 0.5 L sterilized bottles, filledup with TENP buffer (30 mL and 100 mL, respectively), and shaken using a horizontal shaker for 1h at 100 RPM.The supernatants were centrifuged, and subsequently, pellets were collected and stored at −80 • C for DNA extraction.

DNA Extraction
DNA extraction was performed using the NucleoSpin Plant II kit (Macherey-Nagel GmbH, Dylan, Germany) following the instructions.DNA samples were quantified with the NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific, Inc., Walsham, MA, USA) and visualized in 1.0% agarose gel electrophoresis.Samples were kept at −20 • C until further analysis.

Diversity Assessment among Phenological Stages
A total of 4,032,594 high-quality sequences (Table 1A,B) were generated from all studied samples (n = 108).About 1,988,518 sequences were bacterial, with an average of 18,412 sequences per sample, whereas the rest (2,044,076) were fungal sequences, with an average of 18,926 sequences per sample.Moreover, α-diversity indices such as Chao1, Shannon, etc. were calculated as shown in Table 1A,B.The Chao1 scores ranged for bacteria from 192.2 (for bark and bud) to 339.3 (for fruit harvest stage), and for fungi from 172.2 (for bark and bud) to 214.3 (for berry set), whereas the Shannon scores ranged for bacteria from 1.76 (for berry set) to 2.35 (for harvest) and for fungi from 1.53 (for bark and bud) to 1.62 (for berry set).Good's coverage of fungi and bacteria reached 99.5% and 98.8% (Table 1A,B), respectively, indicating that most of the microbial diversity was captured through sequencing [18].Sequences were grouped into 789 OTUs for fungi and 1961 OTUs for bacteria (both at the 97% similarity level) (Table 1A,B).After removing singletons, the OTUs' numbers were 603 for fungi and 1296 for bacteria (Figure 1A,B).The different bacterial OTUs were assigned to 22 phyla, 35 classes, 70 orders, 162 families, and 418 genera.In the case of fungi, the OTUs corresponded to 4 phyla, 17 classes, 53 orders, 124 families, and 267 genera.

Diversity Assessment among Phenological Stages
A total of 4,032,594 high-quality sequences (Table 1A,B) were generated from all studied samples (n = 108).About 1,988,518 sequences were bacterial, with an average of 18,412 sequences per sample, whereas the rest (2,044,076) were fungal sequences, with an average of 18,926 sequences per sample.Moreover, α-diversity indices such as Chao1, Shannon, etc. were calculated as shown in Table 1A,B.The Chao1 scores ranged for bacteria from 192.2 (for bark and bud) to 339.3 (for fruit harvest stage), and for fungi from 172.2 (for bark and bud) to 214.3 (for berry set), whereas the Shannon scores ranged for bacteria from 1.76 (for berry set) to 2.35 (for harvest) and for fungi from 1.53 (for bark and bud) to 1.62 (for berry set).Good's coverage of fungi and bacteria reached 99.5% and 98.8% (Table 1A,B), respectively, indicating that most of the microbial diversity was captured through sequencing [18].
Sequences were grouped into 789 OTUs for fungi and 1961 OTUs for bacteria (both at the 97% similarity level) (Table 1A,B).After removing singletons, the OTUs' numbers were 603 for fungi and 1296 for bacteria (Figure 1A,B

Microbiome Varies According to Grapevine Phenological Stage
In the three phenological stages (bark and bud, berry set, and fruit harvest), fungal OTUs were 442, 433, and 448, respectively(Figure 1A).The bacterial OTUs were 371,713, and 975(Figure 1B), for thebark and bud, berry set, and harvesting phenological stage, respectively.Out of all fungal OTUs identified in the present study, 48% were detected in more than one phenological stage, almost half (311, 52%) of which represented a core across all sample types (Figure 1A).However, many fungal OTUs were unique.A percentage of 11.3%, 11.1%, and 10% (Figure 1A) of the detected OTUs were exclusively observed in berry set, harvest, and bark and bud samples, respectively.The three phenological stages have an almost similar number of fungal OTUs.A percentage of 84% bacterial OTUs were uncovered in more than one grapevine phenological stage, a subset of which (212; 16%) OTUs were shared across all sample types (Figure 1A).A greater number of OTUs (975) were found at the fruit harvest stage, whereasbark and bud samples contained the least OTUs (371).About 48% and 27% of the OTUs detected at the harvest stage were shared with berry set andbark and bud stage, respectively.However, several OTUs were unique; for example, 35.3% of the detected OTUs were exclusively observed in harvest samples (Figure 1B).

Core and Unique Microbiome Vary among the Cultivars in each Phenological Stage
Based on Venn diagram analyses [18], the core bacterial genera among the 36 cultivars assessed were as follows: at harvest 4 (Gluconacetobacter, Erwinia, Gluconobacter, and Zymobacter), at berry set 3 (Buchnera, Pseudomonas, and Pantoea), and at bark and bud 5 (Hymenobacter, Pedobacter, Frigoribacterium, Sphingomonas, and Massilia).The number of unique bacterial genera uncovered in each cultivar at each phenological stage is presented in detail in Table S1.

Temporal Dynamics of Grapevine Microbial Communities through Phenological Stages
Grapevine-associated microbiome analysis in the three phenological stages revealed a very clear differentiation of bacteria community richness and diversity (α-diversity, Shannon index; ANOVA, F = 4.576, p = 0.01243) and taxonomic dissimilarity (β-diversity, Bray-Curtis distance; PERMANOVA, F = 24.95,p = 0.0001, and ANOSIM, R = 0.6432, p = 0.0001) (Table 3).Regarding fungi, no significant differences were detected in community richness and diversity (α-diversity, Shannon index; ANOVA, F = 1.250, p = 0.29030), but such differences were significant in taxonomic dissimilarity (β-diversity, Bray-Curtis distance; PERMANOVA, F = 12.1, p = 0.0001, ANOSIM, R = 0.2100, and p = 0.0001) (Table 3).Principal coordinate analysis (PCoA) showed the phenological stage-specific patterns (95% confidence interval) of bacterial and fungal communities, with 33.78% and 40.95% of total variance explained by the first two principal coordinate (PC) axes, respectively (Figure 3A,B).These results indicated that each phenological stage has a unique microbial signature and that there is a significant difference in taxonomic compositions among the phenological stages.Based on Venn diagram analyses (Figure 4A,B), the abundance of phenological stage-specific genera ranged from 12 (for bark and bud) to 139 (for harvest) and from 22 (for bark and bud) to 56 (for berry set) for bacteria and fungi, respectively.
These results indicated that each phenological stage has a unique microbial signature and that there is a significant difference in taxonomic compositions among the phenological stages.Based on Venn diagram analyses (Figure 4A,B), the abundance of phenologica stage-specific genera ranged from 12 (for bark and bud) to 139 (for harvest) and from 22 (for bark and bud) to 56 (for berry set) for bacteria and fungi, respectively.

Grapevine Genotypes ShapeMicrobial Communities under the Phenological Stage Influence
Principal coordinate analyses on microbial abundance revealed that grapevine genotypessignificantly affect the shaping of their microbiome regardless ofphenological stage (Figures S1-S6).It is noteworthy that these genotypes clustered differently from one phenological stage to another.
At the bark and bud stage, for instance, where the first two PCoA axes (1 and 2) explain 48.7% of the total variation (Figure S1), the cultivars were significantly clustered into six groups based on their bacterial community structure, as confirmed by PERMANOVA (p = 0.0001 and F = 12.5) and ANOSIM (p = 0.0001 and R = 0.66) statistical analyses (Table 3).However, at the berry set stage, where the first two PCoA axes (1 and 2) explain a similar 53.1% of the total variation (Figure S2), the cultivars were clustered into five different groups instead of six, as also confirmed by PERMANOVA (p = 0.0001 and F = 13.1) and ANOSIM (p = 0.0001 and R = 0.76) statistical analyses (Table 3).
Permutational analysis of variance (PERMANOVA) and ANOSIM statistics (Table 3), according to grapevine genotypes and phenological stages, confirmed the hypothesis that these factors might be responsible for shaping the microbial community of the analyzed samples.

Assessment of Vine Microbiome of Economic Importance
Several well-known microbial taxa with economic implications for viticulture and winemaking were uncovered in this study.Some taxa were phenological stage-specific and/or cultivar-specific.Agrobacterium tumefaciens, which is responsible for crown and cane gall disease, was mainly detected in very low abundance at the bark and bud stage in 39% of the studied grapevine cultivars (Voidomatis, Mavrodafni, Karabraimis, Savvatiano, Moschofilero, Asyrtiko, Moschatolefko, Kakotrygis, Ralli, Limniona, Robola, Cabernet Sauvignon, Perlette, and Mavrokorakas).Riesling also hosted A. Vitis as an additional species of Agrobacterium compared to others.Moreover, at the harvest stage, the acetic acid bacteria associated with wine production, Gluconobacter (oxydans and frateurii) and Gluconacetobacter (oboediens, saccharivorans, and hansenii), were detected in a very high abundance in almost all cultivars.Furthermore, Issatchenkia terricola, a non-Saccharomyces yeast associated with winemaking, was also uncovered at the harvest stage in all cultivars but with different relative abundances.For example, Limniona, Victoria, and Soultanina harbored the higher percentages of 43%, 39%, and 19%, respectively (Table S2).In addition, at the berry set stage, Cladosporium cladosporioides responsible for Cladosporiumrot disease of grapevines, was 1.5-fold more abundant than at bark and bud and 3.7-fold more abundant than at harvest stage, in all cultivars (Table 2A).

Discussion
To elucidate the role of Vitis genotype in microbiome assemblage, we examined the composition of epiphytic bacterial and fungal microbial communities in 36 distinct grapevine cultivars grown in the same vineyard and terroir, under the same farming system.We also characterized fluctuations of the microbial communities in different grapevine tissues and developmental stages, examining how such microbial populations change among three different phenological stages (bark and bud, berry set, and harvest) throughout the growing season.Unraveling the associated microbiomes, the relationships among them, and the impact of their hosts, could provide a tool for vine-growers and winemakers to efficiently manage microbiome structure.
Here we analyzed microbiomes in a large selection of 36 cultivars grown in the same environment, including cultivars used for edible grape production as well as for winemaking and colored or white ones.Other studies focused on microbiomes inhabiting the phylospheric compartments in one [5,10,19,20,25,34,46], few [22], or several cultivars [23].High-throughput sequencing revealed rich fungal or bacterial OTUs, and both similarities and differences among microbial communities between the three phenological stages were uncovered.Patterns of microbiome similarity and divergence among cultivars became evident based on PCoA visualization (Figures S1-S6).Similar patterns of microbe divergence were also detected in our previous work [17] and in other studies with different cultivars [23,47].In general, a core microbiome shared by all cultivars was discovered for each phenological stage.However, each cultivar hosted several different microbial species, characteristic of the host genotype.Moreover, the results revealed that some cultivars, such as Limniona (bark and bud), Agiorgitiko (berry set), and Vertzami (harvest), were unique regarding their hosted bacterial genera.Others, such as Savvatiano (bark and bud), Serifiotiko (berry set), and Cabernet Sauvignon (harvest), were also unique regarding their hosted fungal genera.Several studies have documented how geography and environment affect the diversity of microbial community structure [4,12,46].In our study, geographic distance was not a factor controlling microbial diversity because all cultivars were in the same vineyard.Thus, any difference observed could at least partly be considered an effect of the host genotype.Consequently, the genotype of the cultivars appeared to have a drastic impact on the microbial community, influencing the fitness of certain V. vinifera-associated microbes.
A high percentage (79-87%) of fungal and a lower percentage (27-35%) of bacterial OTUs detected in berries could originate from barks and buds, including the majority of the dominant OTUs.Previous studies suggested that the vineyard soil may serve as an origin of primary inoculums of the plant associated microbiomes [10,19,21].The detected microbial OTUs in our study include several taxa that have been typically identified in many studies across the world and suggest an important role of bark and buds as a microbial source (microbial terroir [24]) for secondary direct inoculation of other vines' aerial parts.
The richness of fungal and bacterial species turned out to be lower than that discovered by Singh et al. [23], who targeted similar phylospheric compartments.Additionally, the bacterial OTUs (excluding singletons) uncovered in one-season-old canes' barks and buds in this study were about 40% less than those found in our previous work [17].The plausible explanation could be the different number of 16S RNA variable regions analyzed.
All fungal OTUs discovered in this study were classified similarly to other works [22].On the contrary, about 40% of bacterial OTUs were unclassified for harvest, 48% for berry set, and 80% for bark and bud.This large percentage of unclassified bacterial OTUs, especially for bark and buds, has been encountered in our previous work [17].It has been proposed that richer microbiomes inhabiting such tissues may contain many organisms less well described in the different databases [48] and thus unclassified in these studies.There is also a possibility that the grapevine genotype applies a selection pressure on the microbiome [17,23], and vineyards consisting of different grape cultivars appear to harbor more diverse microbiomes with a higher probability of containing unclassified accessions than those consisting of a single variety.
Fungal richness was lower than that of bacteria, reflecting generally lower fungal diversity and occurrence.Such a difference between bacteria and fungi has been recorded in other studies [5,23,25,46,49]" regardless of the sample type or even the number of cultivars analyzed.Considering the phenological stage, the richness and diversity of the detected bacterial and fungal OTUs on the harvest samples were considerably higher (particularly for bacteria) than that of the berry set, followed by that of the bark and bud samples.This might possibly be due to nutrient availability provided by berries at harvest stage.Such nutrients can support the occurrence and diversity of microbes [19].
Several bacterial or fungal genera detected are known to have distinct beneficial or pathogenic roles.For instance, the presence of Aureobasidium pullulans, a fungus known to have antagonistic activity against Botrytis and Bacillus [23], could explain the lower occurrence of the pathogens.On the other hand, Cladosporium cladosporioides, a fungal species responsible for Cladosporium rot of grapevine [62] reducing yield and affecting the quality of wines [63], was also present.Notably, some economically important microbial genera such as Aspergillus, Acetobacter, Gluconobacter, etc. were observed in trivial abundance at the bark and bud stage with no observable effect, but they became dominant at the harvest stage.Our findings point to the conclusion that grapevine tissues host both beneficial and pathogenic taxa that inflict important economic consequences in viticulture.The dynamics of these beneficial and pathogenic taxa on the three different tissues through seasonal grapevine growth were described.
The evident connection between Vitis microbiomes and the host genotype suggests a genetic component to the host-microbial interactions.Similar interactions were indicated earlier in biogeographic investigations, where local environmental and viticultural practices could not be excluded.Many of the core or unique Vitis microbiota may play an essential role in the quality of grapes and the fermentation or sensory aspects of wines.Viticultural practices such as agrochemical application, canopy thinning, and trellising are commonly applied to regulate Vitis growth and microclimate.Building upon the new evidence reported in this study and the superior capacity of high-throughput sequencing to decipher microbiomes, it could be possible to customize viticultural practices specific to Vitis cultivars and their hosted associated microbiomes.

Conclusions
The composition of microbial communities of 36 grapevine cultivars in three different tissues/developmental stages was studied using NGS.Microbiome structure was significantly different not only among the different tissues/developmental stages but also among the grapevine genotypes and cultivars.This supports the conclusion that even within a terroir, in the same vineyard, and under identical viticultural practices, the cultivar's genotype has a key role in shaping the microbial community.Under these circumstances, all cultivars share a common core microbiome.Remarkably, unique microbial fingerprints were unraveled for each specific cultivar, offering a unique microbial signature of terroir and cultivar/genotype interactions.Further understanding of the host-microbe interactions responsible for the unique signature of grapevine cultivars and microbial distribution could provide a tool to promote precision viticulture for high-quality production with a minimal chemical footprint and climate change impacts.

Figure 1 .
Figure 1.Venn diagrams showing the distribution of (A) fungal and (B) bacterial OTUs among the three phenological stages.

Figure 1 .
Figure 1.Venn diagrams showing the distribution of (A) fungal and (B) bacterial OTUs among the three phenological stages.

Figure 2 .
Figure 2. The relative abundance of the dominant (A) fungal and (B) genera among the three phenological stages.

Figure 2 .
Figure 2. The relative abundance of the dominant (A) fungal and (B) genera among the three phenological stages.

Figure 3 .
Figure 3. Bray-Curtis distance PCoA of bacterial (A) and fungal (B) genera for the three phenological stages for 36 grapevines.

Figure 4 .
Figure 4. Venn diagrams that show the fungal (A) and bacterial (B) genera distribution among th three phenological stages.Figure 4. Venn diagrams that show the fungal (A) and bacterial (B) genera distribution among the three phenological stages.

Figure 4 .
Figure 4. Venn diagrams that show the fungal (A) and bacterial (B) genera distribution among th three phenological stages.Figure 4. Venn diagrams that show the fungal (A) and bacterial (B) genera distribution among the three phenological stages.

Table 1 .
(. Bacterial (A) and fungal (B) Sequence Analysis: number of samples analyzed (n), number of OTUs with and without singletons detected, classified and unclassified individuals, alpha diversity indices (Simpson, Shannon, Chao1), in the three phenological stages.A Phenological Stage

Table 2 .
Relative abundance of (A) fungal and (B) bacterial genera with their percentages detected among the three phenological stages.

Table 2 .
Relative abundance of (A) fungal and (B) bacterial genera with their percentages detected among the three phenological stages.

Table 3 .
Factors predicting αand β-diversity on microbial communities in the vineyard.
NA: Not applied.