The phyllosphere consists of the aerial parts of the plant and it is one of the most prevalent microbial habitats on earth [1
]. Its heterogeneous environment harbors a myriad of microorganisms, like yeast, bacteria, and filamentous fungi and many uncultured organisms [1
]. The phyllosphere or carposphere microbial communities (PMCs) live at the plant-climate interface and its ability to establish, thrive and reproduce on the leaf or fruit surface depends on several microbial functional traits, such as the ability to attach to the cuticle and to use the foliar nutrients, as well as to the prevailing climatic conditions, like temperature, air humidity, and rain [3
]. Leaf or fruit chemistry, physiology, and morphological structure differ among plant genotype or species, and as all of these traits have a genetic basis, these variations may lead to a different combination of PMCs among plant genotypes [6
The plant genotype may exert selection pressure on PMCs, as often reported in A. thaliana [7
]. In the literature, impacts of climatic stressors have received much more attention, especially on soil communities than on the PMCs. Nevertheless, phyllosphere faces constant direct exposure to the outside conditions and available pieces of evidence suggest that PMCs significantly alters in response to the climatic stressors like heat, rain or drought [9
]. Air pollutants (e.g., oxides of nitrogen and sulfur and particulate matters) that are produced by human activities can alter foliar traits, including cuticle properties [13
], leaf chemistry, and phenology [14
] may also affect the structure of PMCs. Moreover, some of the pollutants can be used as a carbon source by PMCs [16
The PMCs that are associated with Vitis vinifera
L., the major crop for fruit and wine production in the world, is less extensively studied when compared to the other habitats (e.g., soil, rhizosphere, and endosphere), especially in relation with the genotypes and the variable climatic conditions or geographic locations. One study suggested that the leaf PMCs are minimally affected by the chemical and biological treatments tested on the plant, but mainly differed according to the grapevine location [17
]. Berry surfaces also exhibit a huge bacterial and fungal diversity and that can have an impact on grapevine health and wine qualities [19
In this study, we assessed both the effect of grapevine genotype and environmental factors on the diversity and structure of phyllosphere and carposphere microbiome. When considering that the PMCs on leaf and berry surface plays a crucial role in plant health and fitness as it can modulate leaf or fruit susceptibility to infection [19
], this study could bring new insights to develop innovative and natural biocontrol methods or phytostimulators against grapevine pathogens or rethink breeding schemes for the creation of innovative resistant varieties.
2. Materials and Methods
2.1. Sample Preparation for PMCs and DNA Extraction
Samples were collected in two sets. In Set1, A total of 279 grapevine cultivars was grown in two vineyards, Chapitre (Supagro field station, Villeneuve-les-Maguelone, Hérault, France) and Vassal (INRA Experimental Unit, Marseillan-Plage, France) near Montpellier (French Mediterranean region). A panel of cultivars representing three genetic pools (western Europe, WW; from eastern Europe, WE; and table grape, TE) was constructed for genome-wide association studies while minimizing relatedness and retaining the main founders of modern cultivated grapevine to optimize the genetic diversity [22
]. Five cultivars from each genetic pool, which are far apart based on their distances on PCoA map shown by Nicolas et al. 2016 [22
], were selected (Table 1
) to maximize the distance between genetic pools. Leaf or berry samples were taken from four to five plants of each cultivar at Spring season (mid of May 2017, before spraying of the fungicides) and harvesting season of (September 2017). Berries were also collected from eleven of these varieties during the harvest season.
In Set2, leaf samples from five commercially important varieties (Cabernet Sauvignon, Chardonnay, Syrah, Grenache, Sauvignon Blanc) were taken from three different geographic locations, (INRA field stations from Bordeaux, Montpellier, and Colmar) within France, representing the three agro-climate zones (Oceanic, Mediterranean, and Continental) of France or different terroirs at the mid of spring season (before spraying of fungicides).
All the samples from both of the sets were washed with an isotonic solution of sodium chloride (0.15 M) containing 0.01% Tween 20 in 50 mL propylene tubes (2–3 leaves and 50–80 g of berries were washed per tube) while using a horizontal shaker for 1h at 100 RPM. Afterward, samples were given an ultrasonic bath for 7–10 min while using Ultrasonic Cleaner (Branson 5510, Marshall Scientific, Hampton, NH, USA) for maximum recovery of microbes from the sample surface. The remaining solution was centrifuged at 4000×g and microbial pellets containing PMCs were transferred in a 2 mL Eppendorf tube and were collected and stored at −20 °C. PMCs from two of these tubes were mixed to make one biological replicate of a single variety and a total of three biological replicates were made for each variety per vineyard. DNA was extracted from each sample by using the ZymoBiomics DNA MicroPrep Kit (Zymo Research, Irvine, CA, USA) following the manufacturer’s instructions.
2.2. DNA Amplification and Amplicon Sequence Library Preparation
To access bacterial communities, the V4 region of the 16S ribosomal gene was amplified using primers 515F and 806R and fungal community diversity and abundance were accessed using modified ITS9 and ITS4 primers targeting the ITS2 region [23
]. Two-step PCR was performed to prepare sequencing libraries. PCR1 was designed to perform amplification of the target regions and to add Illumina Nextera transposase sequence to the amplicons. Both forward and reverse primers for PCR1 were amended with frameshift (FS) sequences in their 5′ overhang to improve sequence diversity and overall read quality [25
]. PCR1 was performed in 25 µL reactions with 30 ng of sample DNA while using the KAPA HiFi HotStart (KAPA Biosystems, Wilmington, MA, USA) PCR mix (Initial denaturing at 95 °C followed by 30 cycles of denaturing at 95 °C for 30 s, primer annealing at 57 °C for 60 s, and primer extension at 68 °C for 60 s). Amplicons were purified while using Agencourt AMPure XP beads (Beckman Coulter, Brea, CA, USA) at a bead-to-DNA ratio of 0.7:1, resuspended in 30 μL MilliQ water, and evaluated in agarose gels. In PCR2, Primers from Illumina kit for dual indexing of the amplicons was used. Each cleaned PCR1 product within the same sample received a unique combination of forward and reverse primers (respectively, N7 and S5 Illumina dual index oligos). Afterward, samples were again cleaned while using AmPure XP magnetic beads, pooled in equimolar concentrations, and sequenced using 2×250 bp MiSeq v2 sequencing (Illumina Inc., San Diego, CA, USA).
2.3. Data Processing and Analysis
Demultiplexed RAW data files covering all of the samples were imported into the R-environment, (R Core Team, Vienna, Austria). The entire amplicon sequences data was uploaded to the institutional server (http://agap-ng6.supagro.inra.fr/inra
). Paired forward and reverse reads from raw data files were trimmed (primer removal) and filtered (base quality) while using the fastqPairedFilter function of the dada2
] and bases with low-quality scores (<11) were discarded. These filtered files were then processed using Divisive Amplicon Denoising Algorithm (DADA) pipeline which included the steps of dereplication, core denoising algorithm (that models and corrects Illumina-sequenced amplicon errors) and the merging of the base pairs. Merging function provided global ends-free alignment between paired forward and reverse reads and merged them together if they overlapped exactly and a table for amplicon sequence variants (ASVs, a higher analog of operational taxonomic units—OTUs) was constructed. It records the number of times each amplicon sequence variant is observed in each sample. DADA infers sample sequences exactly and resolves differences of as little as one nucleotide [26
]. Chimeras were removed using the removeBimeraDenovo function of the same dada2
package (Table 2
represents the total number of reads available during these steps). ASVs or OTU sequences were assigned a taxonomy using the RDP classifier [27
] with k-mer size 8 and 100 bootstrap replicates. Afterward, a phyloseq data object was created using phyloseq function of the phyloseq
package in R [29
]. Unassigned taxa and singletons were removed and this data object was then used to calculate microbial abundances, α, β diversity analysis and for other statistical tests using various functions in the phyloseq
Estimates of observed α-diversity [31
] were measured within sample categories using estimate_richness function of the phyloseq
package. Relative abundances of microbial genera and phylum were plotted using the ggplot2
] after transforming abundance data into relative abundances. Multidimensional scaling (MDS, also known as principal coordinate analysis; PCoA) was performed while using the Bray-Curtis dissimilarity matrix between samples and visualized by using their base functions in the phyloseq
2.4. Statistical Analysis
We analyzed all of the amplicon sequences in R version 3.3.4 using above mentioned Bioconductor packages. CRAN packages plyr
] were also used to draw the figures. We assessed the statistical significance (p
< 0.05) throughout and whenever necessary, we adjusted p
-values for multiple comparisons according to the Benjamini and Hochberg method to control False Discovery Rate [34
], while performing multiple testing on taxa abundance according to sample categories. We performed an analysis of variance or ANOVA [35
] among sample categories while measuring the Observed
estimates of α-diversity (richness of unique OTUs). Stratified permutational multivariate analysis of variance (PERMANOVA) with 999 permutations was conducted on all principal coordinates that were obtained during PCoA with the adonis function of the vegan
package, to observe the statistical significance of clusters according to the sample categories.
Phyllosphere of the grapevines is quite a neglected milieu and many questions related to this microbial habitat are still unanswered, especially the relative impacts of potential factors that could play key roles in shaping the microbial community structure in the phyllosphere. A better understanding of the principal factors affecting community structure and multitrophic interactions in the phyllosphere will be the key to develop new strategies for grapevine protection. The better we understand the role of these stressors and PMCs that they affect, the better we would be able to predict and protect grapevine against pathogen infection.
In this study, we first explored the microbial communities present in the Mediterranean, Continental and Oceanic vineyards. Major bacterial and fungal taxa (at genus level) were Pseudomonas
, Pantoea, Skermanella
, and Stemphylium
, respectively. Differences in relative abundances of major taxa were quite visible according to agro-climate zones (or growing region) as compared to cultivars (grouped in three genetic pool), growing in the Mediterranean (Supplementary Figures S2 and S3
). We mainly investigated the impacts of grapevine genotypes (or cultivars) and of terroir on the assemblage of PMCs using a culture-independent method. In the Mediterranean vineyards, grapevine cultivars, and their genetic pools had a significant impact on leaf and berry microbiome and the impact is stronger on the berry surface. Assuming that the PMCs on berries would also be present on wine must this result is in line with reports, suggesting that the microbiota exhibits varietal level differences in wine musts of Chardonnay and Cabernet Sauvignon [36
While comparing the impacts of climatic stressors and cultivars at three different locations, we observed a very strong impact of French agro-climate zones or terroirs. Although the impact of genetic factors was significant but much lower in comparison with terroir, which suggests that genotype-by-environment interactions contributed to the complexity of microbiome assembly. Such interactions also represent the cumulative influence of a potentially large number of environmental factors can be involved: soil type, for example, was different in the locations tested. Since the epiphytes (PMCs) that are associated with grapevine could originate from soil [38
], leaf communities could be influenced by soil chemistry or other abiotic factors of the regions where plants are grown, leading to these region-specific unique microbial signatures.
Few strains of Sphingomonas, which was found quite abundant in all three regions (Figure 4
E) were recently reported in plant protection against a bacterial pathogen (P. syringae
DC3000) in A. thaliana
model system [39
]. Although, the molecular basis of pathogen reduction is unknown, but available evidence suggests that several traits contribute to the outcome of plant protection [40
]. Differential abundance of Sphingomonas in grapevine grown in different regions should thus be studied in future in relation to plant traits to assess its impacts on grapevine health. Similarly, a fungal genera Aureobasidium
was also quite abundant in all three regions (Figure 4
F) and this prevalence of Aureobasidium
was due to the presence of A. pullulans
(relative abundance >12%, Supplementary Figure S1
). A. pullulans
have an antagonistic activity for Botrytis molds and for certain bacteria like Bacillus [28
], which probably explains the lower prevalence of Bacillus and Botrytis in our data (Figure 4
Seasonal shifts in phyllosphere microbiome structure and the impacts of plant organs were also observed. At a particular location, the microbial composition of few bacterial phyla might change while bacterial diversity does not change during season shift. For example, cyanobacteria (photosynthetic bacteria) change its abundance from high to lower due to season change from spring to harvest. Lower daylight presence in harvest season probably explains these changes. These results are coherent with another grapevine (Tempranillo) related study [38
]. On contrary, fungal community diversity and their relative abundances, both were significantly impacted by season. Apart from genotype and terroir, the vineyard management practices could also be the possible reason for these differences [9
]. Although a significant fraction of the members of PMCs were shared between plant organs (leaves and berries), we observed distinct assemblage patterns between both organs, which is also in accordance with recently published reports [25
]. These differences among organs do not only reflect the compositional differences (or difference in the relative abundance of shared OTUs), but also the diversity in taxa present.