High-throughput sequencing of microbial community diversity in soil, grapes, leaves, grape juice and wine of grapevine from China

In this study Illumina MiSeq was performed to investigate microbial diversity in soil, leaves, grape, grape juice and wine. A total of 1,043,102 fungal Internal Transcribed Spacer (ITS) reads and 2,422,188 high quality bacterial 16S rDNA sequences were used for taxonomic classification, revealed five fungal and eight bacterial phyla. At the genus level, the dominant fungi were Ascomycota, Sordariales, Tetracladium and Geomyces in soil, Aureobasidium and Pleosporaceae in grapes leaves, Aureobasidium in grape and grape juice. The dominant bacteria were Kaistobacter, Arthrobacter, Skermanella and Sphingomonas in soil, Pseudomonas, Acinetobacter and Kaistobacter in grape and grapes leaves, and Oenococcus in grape juice and wine. Principal coordinate analysis showed structural separation between the composition of fungi and bacteria in all samples. This is the first study to understand microbiome population in soil, grape, grapes leaves, grape juice and wine in Xinjiang through High-throughput Sequencing and identify microorganisms like Saccharomyces cerevisiae and Oenococcus spp. that may contribute to the quality and flavor of wine.


Introduction
Microbial biodiversity is highly essential for harboring a healthy environment with sustainable economy, especially at agricultural field. In recent years, several studies on biodiversity tried to characterize the microbiome in different agricultural ecosystems, as to understand the dynamics of plant and microbe interaction. It is discovered that plant-associated bacteria and fungi colonize on both exterior (epiphytes) and interior surfaces (endophytes) of plants, while the surrounding soil around the plant acts as the major resource for these microbes [1]. As plant species and soil type play a predominant role in soil microbial community, the interaction between plant and soil microbes are highly complex. Vitis vinifera (wine grape) is an economically important agricultural crop. V. vinifera phyllosphere is easy to be colonized by both PLOS  microbiome in the vineyard from in Xinjiang region, and help to establish correlation between wine quality and regional microbiome.

Sampling
Samples were collected from three different winery regions in Xinjiang province, China, named Fukang (A), Manasi (B) and Changji (C) area, both are cabernet sauvignon samples. A total of 20 samples were taken up in July and October, 2015, categorized as vineyard soils (T), grape (P), grape leaves (Y), grape juice (Z) and wine (J) ( Table 1). Soil samples were collected in randomlychosen plots around each vine root, collected at 5-10cm depth using a shovel and sieved to remove plant residues, macrofaunal, and stones. Undamaged grape samples were collected from several bunches with their pedicels attached. Grape leaves were collected from the vines to prevent cross contamination. Grape juices were sampled during grape crushing period. Wine samples from winery were collected from the final wine product. After collection in triplicates, each sample was stored immediately at -20˚C for further study. The samples were collected from the private properties, permissions from the owners were received before sample collection.
Sample preparation for grape and grape leaves 4g of grape skin or grape leaves (no flesh and fruit stalk), was taken into 50mL sterilized centrifuge tube (L1), and 6mL of TENP buffer was added and mixed vigorously in vortex for 10min. The mixture was then centrifuged at 3000×g for 5min, supernatant was transferred to another 50mL centrifuge tube (L2). This step was repeated thrice, and a total of around 18ml supernatant was transferred to L2. L2 was then centrifuged at 9000×g for 10min. The supernatant was discarded and pellet was stored at -20˚C for DNA extraction [21].

Sample preparation for grape juice and wine
About 4mL of grape juice or wine were taken into 50mL sterilized centrifuge tube (L1), vigorously shake to mix, then vortex for 1min. The mixture was then centrifuged at 3000×g for 5min, and supernatant was discarded. For the pellet, 4mL of sterile water was added and vortexed for 1min. Mixture was centrifuged at 3000×g for 5min and supernatant was discarded. This wash step was repeater thrice, and the pellet was finally stored at -20˚C for DNA extraction.

DNA extraction and amplification
DNA from soil samples were extracted using Mag-Bind Soil DNA Kit (OMEGA) following manufacturer's instructions. DNA from grape, grape leaves, grape juice and wine were The quantity and quality of extracted DNA were assessed by spectrophotometry (Eppendorf, Germany) and agarose gel (1%) electrophoresis, respectively. For fungal ITS regions, PCR amplification was performed using ITS 1F (5 0 -CTTGGTCATTTAGAGGAGTAA-3 0 ) and ITS 1R (5 0 -GCTGCGTTCTTCATCGATGC-3 0 ) as primers and genome DNA as template. PCR was performed at a final volume of 50μL mixture containing 4μL dNTPs mixture, 5μL of 10×PCR buffer (Mg 2+ plus), 5μL of template DNA, 1μL of each primer, 0.25μL Ex Taq, and added to final volume of 50μL using ddH 2 O. PCR conditions were 3min at 98˚C for initial, followed by 35 cycles at 98˚C for 45s, annealing at 53˚C for 30s, and extension at 72˚C for 45s, and final extension at 72˚C for 8 min. PCR products were stored at -20˚C.
For bacterial 16s rRNA gene region, PCR amplification was performed using 16S 515F(5 0 -GTGCCAGCMGCCGCGGTAA-3 0 ) and 16S 806R(5 0 -GGACTACHVGGGTWTCTAAT-3 0 ) as primers and genome DNA as template. PCR was performed at a final volume of 50μL mixture containing 4μL dNTPs mixture, 5μL 10×PCR buffer (Mg 2+ plus), 5μL template DNA, 1μL of each primer, 0.25μL Ex Taq, and added to final volume using ddH 2 O. PCR conditions were 3min at 98˚C for initial, followed by at 95˚C for 45s, annealing at 50˚C for 30s, and extension at 72˚C for 45s, and final extension at 72˚C for 5 min. PCR products were stored at -20˚C. Both fungal and bacterial PCR products were analyzed with 1% gel using DL2000 marker for quality examination.

High-throughput sequencing and statistical analysis
High-throughput sequencing was performed using an Illumina Miseq platform at the International Joint Research Center of National Liquor Quality and Safety in Chinese Food Fermentation Industry Research Institute, Beijing, China. After obtaining the sequencing result and calculation of operational taxonomic units (OTUs) matrix, statistical analysis was applied using alpha indices (Shannon, Simpson, Chao 1 and ACE), heatmap of genera, principal coordinate analysis (PCoA) and UPGMA. The alpha diversity index (Shannon Index), and the species richness estimator (Chao1) were calculated by using the R package phyloseq from the OTU matrix. Principal coordinate analysis (PCoA) was performed using statistical software based on the Bray-Curtis dissimilarities calculated for the composition of the bacterial or fungal communities at the genus level.

Abundance and diversity of members of the bacterial and fungal microbiota
Illumina Miseq sequencing generated 1,043,102 high quality fungal sequences with an average of 52,155 sequences per sample, and a total of 2,422,188 high quality bacterial sequences with an average of 121,109 sequences per sample. However, the fungal population in T1, T5, and J1 was not detected. The sequencing results were shown in Supporting Information(S1 and S2 Figs, S1 and S2 Tables). Though the rarefaction curve was not parallel with the x-axis, the Good's coverage of fungal and bacterial reached 99.9% and 97.7% respectively, with majority of microbial diversity being captured. As we applied the Chao1 index and ACE index, it reflected the species richness of sample communities, while the Shannon and Simpson index reflected the species diversity. The Chao1 and ACE scores from 139.565 to 2941.088 and from 150.4 to 3354.01, respectively. The Shannon and Simpson scores ranged from 1.538 to 8.939, and 0.241 to 0.994, respectively. Sequences were grouped into 4597 OTUs for fungi and 23,458 OTUs for bacteria (both at the 97% similarity level). As shown in Fig 1a and 1b, after removing singletons, OTUs number was 91 for fungi and 227 for bacteria, which were similar in the T, Y, P and Z samples. Fungal OTUs were 787, 568, 299 and 158. Bacterial OTUs were 3932, 1766, 850 and 370 in T, Y, P and Z respectively.
Earlier works suggested that soil provides essential nutrition for microbial growth, such as carbon sources (including amino acids, organic acids, and carbohydrates), nitrogen sources, and growth factors [22]. The physical and chemical properties of the soil, such as soil texture, mineral composition and organic matter, affect the growth and distribution of microbes [23][24]. Suitable growth environment favors the growth of most microbes, which in-turn alters the microbial community composition to larger extent in the soil and the plant [25]. And in contrast, external factors like pollution by heavy metals, organic pollutants, pesticide fertilizers, domestic sewage, and factory waste will directly alter the soil quality, causing major changes in the microbial communities [26-28]. It is described that microbes in vineyard soil are the source of primary inoculum to affect the structure of the microbial community on vine's aerial parts [29], and this pattern is observed in our study. Diversity indices indicated that the diversity of fungal and bacterial community in T were significantly higher than Y, P, Z, J in October. The inoculation effect can be extended to wine brewing, as microbial community in P, Y and J were very similar with some minor differences according to Fig 1. Besides brewing microbes added in wine production process, these minor differences are possibly come from picking, transportation, crushing and other factors in grape crushing process, or fermenters and oak barrels in fermentation process [30][31][32][33].
The comparison of microbial communities in soil and vine plants suggests an inoculation effect between soil microbes and vine plants. But with the growth of vine plants and climate change, interaction between soil microbe and plants should be more complex. Comparison of soil samples collected from July and October indicate that fungal microbes in soil have higher diversity in July, while bacterial populations changed little in our study. As the weather turns gradually colder, temperature and humidity can inhibit fungal species that are not tolerant to these environmental challenges, while bacterial species showed higher adaptation in this environment. Besides, vine plants were found to limit the growth of bacteria by limiting nutrients in the early stage of growth [2,5,[34][35][36]. And in response to biotic stress, stilbenes are produced in vine plants to effectively inhibit microbial activity [37][38]. To better understand such interaction and impact of soil microbes on vine plants, more samples should be taken gradually to provide systematic and detail results on microbial communities.

Comparison of fungal communities in soil, grape, grapes leaves, grape juice and wine
Five fungal phyla identified were Ascomycota, Basidiomycot, Chytridiomycota, Un-s-fungal sp CC 06_28 and Zygomycota. Heatmap revealed that Ascomycota and Basidiomycot were found as predominant phyla in T, P, Y, Z and J (Fig 2). This finding was consistent with previous reports [39][40]. Majority of OTUs in T and Y were Ascomycota, and similar pattern was discovered across the samples in T and Y. Basidiomycot remained less abundant in T and Y, but its abundance was found to be higher in P and Z samples. Chytridiomycota, Un-s-fungal sp CC 06_28, and Zygomycota has less abundance in T samples, but rarely detected in Y, P and Z samples. The results suggested that Ascomycota adapted to the T and Y environment, and Basidiomycot adapted to the P and Z environment.
At the genus level, 271 fungal genera were detected in T, P, Y, Z and J, in which 14 fungal genera which had their relative abundance greater than 1% were selected for further analysis (Fig 3). Ascomycota, Sordariales, Tetracladium and Geomyces were the predominant genera in T samples, Aureobasidium, Pleosporaceae, Cryptococcus and Dothideales were the predominant genera in Y, P and Z samples. Other major genera were Aspergillus, Pleosporales, Penicillium, Erysisphe, Alternaria and Scleroderma. Our finding is consistent with previous reports [41]. The Ascomycota was sharply increased from July to October in A and B, however this was not obvious in C. In contrast, the Sordariales, Tetracladium and Geomyces of A, B and C were decreased in October, and Sordariales in C was found increased. Such difference might be due to interspecies competition. For Y samples, Aureobasidium, Cryptococcus and Dothideales in A, B and C have higher abundances in Octoberexcept Dothideales in A. In contrast, Pleosporaceae was found to have lower abundance in October. It is worth noticing that Erysisphe declined sharply in B, from 93% (Y11) in July to 1% (Y13) in October. This finding was consistent with an earlier report, which indicated that grapevine powdery mildew is one of the most damaging fungal diseases and it often occurs during July [42], suggesting that grapevine powdery mildew in B is quite serious and needs proper preventive measures [43][44]. Meanwhile, Erysisphe was found in Z. Interestingly, the Penicillium increased from 2% (Y11) to 44% (Y13), had become a dominant genus in B of Y samples in October. Consistent with other reports, Aspergillus, Penicillium, and Alternaria are other discovered major genera and might play an important role in T, P and J [45][46]. Studies have pointed out that ochratoxin A (OTA) produced by Aspergillus is a predominant global wine contaminant causing health hazards, the safe limit of OTA in wine is established at 2ug/L [47][48][49].
Our results indicated that fungal community changes in the grape leaves and grapes are more complex than changes in soil. This may be due to the competition between species, or natural conditions such as light intensity, light time, wind, rain, etc. or insects, human activities that causing microbial migration [29,36]. In this metagenomics analysis, Brettanomyces bruxellensis has not been detected. It is able to convert hydroxycinnamic acids into volatile phenols, create 'spicy', 'barnyard', 'animal', 'horse sweat' and 'medicinal' odors in final wine product. Though a large number of culture-dependent techniques are available to assess the presence of this undesired yeast during the vinification processes, in several cases Brettanomyces is undetectable. It is been reveal that while keep alive and maintain the metabolic activities, Brettanomyces cells were tend to enter in a Viable But Not Culturable (VBNC) state. This could be the reason for not been detected in our wine samples. PCoA and cluster analyses were performed to evaluate similarities in fungal communities of T, Y, P, Z and J. One weighted (PC1 variance = 63.28%, PC2 variance = 20.92%) and another weighted (PC1 variance = 31.71%, PC2 variance = 11.01%) PCoA were performed (Fig 4). UPGMA clustering obtained a phylogenetic tree by using unweighted group averaging   High-throughput sequencing of microbial community diversity method (Fig 5). The branch of the sample intuitively reflected the similarity between the samples. Result indicates that same type of samples showed high similarity of fungal communities, as the soil samples and leave samples formed two big clusters. Soil samples from the same region each formed a small cluster, while similarity of leave samples from the same vineyard is relatively weaker. Compare to soil, grape leaves are highly exposed in environment. Microbes on leave are easier to be affected by natural conditions like insects, pathogens and human activities. For soil from vineyards, it is reported that fertilizing the soil, applying pesticides, notillage, continuous cropping and rotation, etc., affects the structure and physicochemical properties of the soil caused by human measures which in-turn change the composition and distribution of microbial communities [50][51][52][53]. The management of the vineyard may also be a factor in the composition of the microbial community.

Comparison of bacterial communities in soil, grape, grapes leaves, grape juice and wine
Compared to fungal population variation, the overall diversity of the bacterial microbiota in T, P, Y, Z and J samples were higher, especially in P samples. Grapes tend to mature quickly by sunlight, rain, and other conditions, as these factors can easily cause them to rot. And in juice samples, the high sugar and nutrients can increase bacterial growth [29,36]. Detected major bacteria phyla include Actinobacteria, Bacteroidetes, Crenarchaeota, Firmicutes, Nitrospirae, Planctomycetes, Proteobacteria and Verrucomicrobia (Fig 6). Heatmap revealed that Proteobacteria, Firmicutes, Bacteroidetes and Actinobacteria were the predominant phyla in T, P, Y, Z and J (Fig 6). Proteobacteria and Firmicutes were found in all the samples especially in T1, T5 and J1, and Bacteroidetes and Actinobacteria were mainly found in T samples. Crenarchaeota, Nitrospirae, Planctomycetes and Verrucomicrobia were rarely detected except in T samples. The results suggested that Proteobacteria and Firmicutes adapt better to all environment, and Bacteroidetes and Actinobacteria adapted well in soil environment.
At the genus level, 317 bacterial genera were detected in T, P, Y, Z and J samples. The 14 bacterial genera with their relative abundance greater than 1% were selected (Fig 7). Kaistobacter, Arthrobacter, Skermanella and Sphingomonas were the predominant genera in T, Pseudomonas, Acinetobacter and Kaistobacter were the predominant genera in Y and P, and Oenococcus was the predominant genera in Z and J. Other major genera in samples include Steroidobacter, Rubrobacter, Flavisolibacter, Pontibacter, Nitrospira, Rhodoplanes and Adhaeribacter. These findings were consistent with previous reports [54][55]. The Kaistobacter abundance in October T samples was found increased compare with July sasmples, and Arthrobacter abundance was found declined. In contrast, no significant changes of Skermanella and Sphingomonas abundances in T samples were found. For Y and P samples collected in July and October, Kaistobacter abundance was not consistent, Pseudomonas and Acinetobacter in Y of A and B decreased in October. C sample was an exception that Pseudomonas and Acinetobacter have higher abundance in October samples, which might be due to interspecific completion that Oenococcus decreased from 35%(Y15) to 0%(Y17). Oenococcus was the dominant genus in Z and J samples, increased sharply to 95% in Z1 and 98% in J1. Oenococcus sp is a slow growing lactic acid bacterium. It is a necessary bacterium in winemaking process with the function of malolactic fermentation. It's accumulation in wine is a natural process, and similar phenomenon was also observed in other reported studies [56]. Lactobacillus plantarum is another well studied bacterium with some strains been commercially used as malolactic fermentation (MLF) starter cultures. It is able to conduct MLF under high pH condition and in co-inoculation with yeasts. It has not been detected in this metagenomics analysis, possibly because of the special climate conditions in Xinjiang region is more suitable for Oenococcus sp.
It is discussed that the geographical location, natural climatic conditions [57], host plant phenology [58], physical and chemical properties of the soil [59][60], economic characteristics of the phyllosphere or soil [61], artificial vineyard management model, external pollution and other factors both can affected the T, Y and P microbial community composition to a certain extent, which in-turn determines the community species composition, quantity and distribution. The metagenomics analysis in this study indicates a strong correlation between grape and leave samples, differences of microbial communities in various regions were also discovered.
PCoA of bacterial communities in T, Y, P, Z and J samples was performed using one weighted (PC1 variance = 70.61%, PC2 variance = 15.50%) and another weighted (PC1 variance = 41.87%, PC2 variance = 10.35%) (Fig 8). UPGMA cluster map (Fig 9) reflected the bacterial microbial community structure, indicate that the similarity of bacteria in all soils were higher except T4. P, Y, Z and J had some similarity in the bacterial microbial community, though differences were existed between vineyard management in the A, B, C region.

Conclusions
The specific microbes in Xinjiang region served as autochthonous starter cultures, which play a major role in local winery production. In this study, microbial diversity of soil, grape, grape High-throughput sequencing of microbial community diversity leaves, grape juice and wine were studied by high-throughput sequencing and identified the dominant genera and phyla. It is the first study to apply this technology on the winery research in Xinjiang region of China, aimed to understand the microbial diversity in Xinjiang region, and identify microorganisms that could improve the quality and flavor of wine product. The extensive revealing of microbial diversity in Xinjiang region is helpful in building wine microbial germplasm repository. In addition, this study identified various microbes which are beneficial for the wine production. Further identification and research on these microbes can help to figure out their indigenous beneficial nature, which will in-turn help the winery to improve quality and local characteristics of their wine product.