Plant-associated microbial communities are complex and diverse. As with most microbial communities, there is a limited understanding of the factors and mechanisms that establish and stabilize plant-associated microbiotas. It is unclear how specific populations of microorganisms are established and maintained and what promotes the appropriate balance of different microbes (Ramey et al., 2004). It is widely accepted that greater than 99% of the microbes present in many environments are not readily culturable or not-yet-cultured and therefore not easily accessible for basic and applied research (Bidle et al., 2007). The species diversity in many unique environments has never been described. To more fully understand novel environmental niches, several DNA-based methods have been developed including 16S rRNA gene analyses and metagenomics (Whitman et al., 1998; Williamson et al., 2005). The former provides information about taxa present in an environmental sample while the latter offers insight into the functional roles of different microbes within a community (Handelsman, 2004; Riesenfeld et al., 2004).

The grapevine-associated microbiota has been a subject of several studies due to the importance of grape cultivation for the production of wine, fresh grapes, raisins, jelly, juice, jam and grape seed extracts, and oil. The tissue saps of grapevines are rich in nutrients that include organic acids, amino acids, sugars, and several inorganic compounds with a pH 5.7–6.9 (Roubelakis-Angelakis and Kliewer, 1979; Andersen et al., 1989; Glad et al., 1992; Prima-Putra and Botton, 1998). This nutrient-rich tissue environment supports the growth of several bacteria resulting in an epiphytic and endophytic population with several species including important pathogens such as Agrobacterium spp. and Allorhizobium vitis (A. vitis) (Bell et al., 1995; Szegedi and Bottka, 2002; Bulgari et al., 2009; Compant et al., 2011). Agrobacterium spp. or A. vitis often causes galls to develop at the crown of the vine; hence, the name crown gall (CG), but can also induce galls on the perennial stems where wounds are inflicted as a result of grafting or injury by freezing temperatures or farm implements (Burr and Otten, 1999). Interestingly, the natural occurrence of CG on young green grapevine shoots has not been documented. CG tumors first appear in early summer as soft masses of disorganized cells which are creamy white or light green in color. In autumn, they become dry and wood-like and turn brown to black, hence the origin of the alternative name “black knot of grapevine” (Süle and Burr, 1998). Bark cracking and peeling may be associated with gall development. Profuse gall development may cause girdling of the trunk which prevents the exchange of nutrients between root and shoot systems thus leading to the reduced vigor of CG-affected vines.

The causal agent of CG tumor disease is commonly referred to as Agrobacterium vitis which was recently reclassified to the genus Allorhizobium based on whole genome phylogeny (Mousavi et al., 2014; Gan et al., 2018). Although less common, some strains of tumor-inducing (Ti) plasmid-harboring Agrobacterium tumefaciens can also cause CG (Pu and Goodman, 1992; Abdellatif et al., 2013). The Ti plasmid (pTi) encodes genes for the processing, transfer, and stable insertion of the transfer-DNA (T-DNA) from pTi to the plant nuclear genome. The constitutive expression of T-DNA oncogenes in the plant genome results in the overproduction of plant hormones cytokinin and auxin which causes the unregulated proliferation of undifferentiated plant cells which manifest themselves as the tumorous outgrowth of CG. Additional T-DNA-encoded genes produce enzymes that synthesize novel low molecular weight class of compounds known as opines (Bevan and Chilton, 1982; Chilton et al., 1982; Zambryski et al., 1989). The specific opine-type produced in the tumor is characteristic of the virulent Agrobacterium/Allorhizobium strain. Common opines found in CG include octopine, nopaline, and vitopine. These metabolites produced by the transformed plant tumor cells are almost exclusively metabolized as an energy source by the virulent agrobacteria that have induced the CG (Dessaux and Faure, 2018; Kuzmanović et al., 2018). A physiologically active CG expands the assortment of nutrients utilizable by bacteria and fosters a rich niche for plant-associated bacteria to colonize, grow, and form complex ecological interactions (Barton et al., 2018; Dessaux and Faure, 2018). In one study, 138 culturable bacterial colonies were isolated representing distinct morphological groups from eight grapevine CG tumors that produced octopine, nopaline, or vitopine (E. Szegedi, unpublished data). All isolates were non-fluorescent on King’s B medium indicating that none of them were fluorescent Pseudomonas species. On the basis of their morphological and physiological characters they could be allocated into three groups: (i) A. vitis type colonies (85), (ii) A. tumefaciens type colonies (8), and (iii) unidentified isolates which formed yellow colonies (45) (E. Szegedi, unpublished data). One of the unidentified yellow isolates (named Rr-2-17) was shown to accumulate large amounts of acyl-homoserine lactone quorum sensing signal molecules which can activate the traR promoter (used by the pTi for activation of pTi conjugation). This isolate was identified as a Novosphingobium sp. by full-length 16S rRNA gene sequencing and then verified via comparative genomics (Gan et al., 2009, 2012a). Further characterization of Novosphingobium sp. Rr2-17 showed the influence of the stringent response regulator, rsh, on the accumulation of the acyl-homoserine lactone quorum sensing signal (Gan et al., 2009).

Recent grapevine microbiota studies used Illumina 16S rRNA gene amplicon sequencing to investigate microbial communities associated with the grapevine organs such as leaf, fruit, cane, and root as well as soil surrounding the roots (Abdellatif et al., 2013; Pinto et al., 2014; Gilbert et al., 2014b; Belda et al., 2017; Manici et al., 2017; Morgan et al., 2017; Alaimo et al., 2018; Marasco et al., 2018; Wei et al., 2018). These studies provided intriguing insights into the effects of various environmental factors on the structure on the grapevine microbiota. Studies that focused on the microbiota of grapevine CG remain scarce with one of the first being conducted on samples collected from a single location in Germany across a temporal gradient (Faist et al., 2016). Using OTU clustering approach, Faist et al. (2016) showed that an OTU classified as A. vitis was the most common OTU in the CG affected graft unions followed by two OTUs belonging to Enterobacter and Pseudomonas with varying relative abundance from season to season. Recently, the amplicon sequence variant (ASV) approach is gradually gaining popularity as it can determine real biological sequences at single nucleotide resolution albeit at the expense of higher false positives (Callahan et al., 2017; Nearing et al., 2018). The CG microbiota is an excellent model to test the utility of this approach since members of the genus Agrobacterium are known to exhibit strikingly high interspecies 16S rRNA gene similarity (Gan and Savka, 2018).

In this study, we investigated the CG microbiota of grapevines from seven different vineyards located in Hungary, Japan, Tunisia, and the United States through Illumina amplicon sequencing of the 16S V3–V4 rRNA gene region. First, we implemented a new primer design to limit amplification of the grape chloroplast 16S rRNA gene, thereby substantially improving the representation of reads belonging to bacterial 16S rRNA. Second, we implemented an ASVs method to improve recovery of Agrobacterium 16S rRNA gene sequences with ecological implications by only removing noise from sequencing instead of clustering sequences based on similarity into operational taxonomic unit (OTU). Third, we provide evidence that CG tumors have a small core microbiota and that the microbiota structure is dependent on the sampling site and/or climate. Finally, we show that the abundance of A. vitis in the CG is positively correlated with the abundance of at least three non-A. vitis groups, e.g., Novosphingobium sp., Xanthomonas sp., and Microbacteriaceae sp., suggesting the presence of a microbial CG “hub” in the CG tumor environment.

To define the microbiota of CG across, we sampled 73 CGs from various grape cultivars located in seven vineyards across four continents of the Northern Hemisphere. In contrast to the work done by Faist et al. (2016) who limited sampling to a single vineyard, our study provides a more comprehensive insight into the CG microbiota. Chloroplast contamination was dramatically reduced in this study using a modified reverse primer, albeit at the expense of reduced primer coverage for a few phyla which are not commonly plant associated (Hirsch and Mauchline, 2012; Turner et al., 2013; Schlaeppi and Bulgarelli, 2015). Similarly leveraging on primer mismatch to host plastid genome, a recent study has recommended the use of standard and unmodified primer pair “799F-1391R” targeting the V5–V7 region that was shown to dramatically reduce co-amplification of poplar plastid (Beckers et al., 2016). Given that a majority of Illumina-based microbiota studies target the V3, V4, or V3–V4 region (Klindworth et al., 2013; Gilbert et al., 2014a), protocol modifications directly aimed at overcoming co-amplification of host DNA may be beneficial. An attractive and potentially more cost-effective approach for future large-scale studies would be to design blocking primers that are complementary to the host chloroplast and mitochondrial rRNA gene (Hanshew et al., 2013; Beckers et al., 2016). A blocking primer will contain C3 spacer at its 3′-end that prevents extension during PCR when included into the standard PCR mix at an equal or higher concentration than the standard 16S primers (Arenz et al., 2015).

It is also worth noting that a majority of the sampling in this global study was performed between June and September, a critical period during which glucose concentration in the grape is most variable (Bauer et al., 1994). This may explain the variability in the microbiota composition across sample sites. It is also possible that the use of different DNA extraction methods may have contributed to the difference observed. However, given that the sampling was performed at multiple distant geographic regions at different time, the effect of such spatial and temporal variations on microbiota composition should outweigh the effect of variance in the DNA extraction method employed. Although examining the effect of extraction methodology on CG microbiota composition is beyond the scope of this study, similar studies have been conducted on non-CG samples with the general consensus that inter-sample variation always outweighed the variation in extraction method (Wesolowska-Andersen et al., 2014; Sinha et al., 2017).

The colony forming unit (CFU) of A. vitis from CG tumors of Riesling and Müller-Thurgau grapevine cultivars in Germany was previously shown to decrease dramatically after the month of June (Faist et al., 2016). Similarly, A. vitis was largely absent among Hungarian samples sampled in September 2014 but not July 2013 based on A. vitis PCR detection and amplicon sequencing assays. The absence of A. vitis in a majority of the opine-containing Hungarian CG samples suggests while opine serves as an attractant to A. vitis, there are additional biotic and/or abiotic factor(s) that can influence the population dynamics of A. vitis in CG tumors (Bhattacharya et al., 2010). On the other hand, high ASV3 to ASV2 ratio, indicating a strong dominance of Agrobacterium spp. over A. vitis was observed only in the Tunisian CGs. Agrobacterium spp. members were previously shown to be dominant in Tunisian soils based on culture-based method, an observation that was hypothesized to be due to the climatic and soil conditions of the country (Bouri et al., 2016). The diversity of Agrobacterium populations in Tunisian vineyards seems to be restricted to the genomic species G4, G7, and G9. Since Agrobacterium genomic species G9 was represented by ASV44 instead of ASV3, 16S rRNA reads mapping to ASV3 in the Tunisian CG tumors may originate from Agrobacterium spp. belonging to genomic species GS4 and/or GS7. The lack of Tunisian tumorigenic strains associated with genomic species G7 as determined by PCR-based detection of Ti-plasmid (VCF3-VCR3 primers; virC) lends support toward the affiliation of ASV3 in Tunisian CG tumors to Agrobacterium spp. genomic species G4 (Bouri et al., 2016). It is important to note that such a correlation cannot be applied to non-Tunisian samples that are lacking background microbial genetic data.

In addition to A. vitis (OTU_0003), Faist et al. (2016) also identified two additional OTUs (OTU_0005, Pseudomonas sp.; OTU_0008, Enterobacter sp.) showing high abundance and prevalence in tissues sampled from vines containing CG tumors during spring and autumn. Although our study did indeed demonstrate the prevalence and substantial abundance of A. vitis-linked ASV in a majority of the CG tumor tissue samples, no ASVs corresponding to Pseudomonas could be found in the core microbiota. However, we do observe a significant enrichment of Pseudomonas-linked ASVs in the Hungarian 2013 CG tissue samples, which may suggest the positive association of members from this genus during early CG formation possibly due to the ability to utilize opine compounds as a growth substrate (Bell et al., 1990). For example, the genome of P. kilonensis strain 1855-344, a strain which can catabolize octopine, has recently been shown to contain an octopine-catabolic operon named ooxAB (Eng et al., 2015). Interestingly, despite being present in the core microbiota, none of the Enterobacteriaceae-linked ASVs formed a significant co-occurrence with A. vitis (ASV2). However, several Enterobacteriaceae-linked ASVs were shown to be differentially abundant among samples from New York and Tunisia. The occurrence of members of the Enterobacteriaceae has recently been associated with the native microbiota of viticultural regions, and has been proposed to impart distinct chemical composition and sensory characteristics of regional wines (Bokulich et al., 2016). It has been suggested that plant-associated members of Enterobacteriaceae also play a part in the complex interactions among the environmental, temporal, plant-genetics, human, and other factors which influence grapevine growth and development collectively referred as “terroir” (Fischer et al., 1999; López-Rituerto et al., 2012). Within the family Enterobacteriaceae, members from the genera Pantoea and Erwinia are common plant inhabitants. However, the evolutionary relationships of members from these two genera could not be confidently established using the nucleotide sequence of the 16S rRNA gene alone. This uncertainty may be responsible for our failure to assign several Enterobacteriaceae-linked ASVs to the genus level. Since Erwinia and Pantoea spp. are readily culturable on agar medium, it may be more appropriate to infer their roles in the CG microbiota using a culture-based approach followed by biochemical characterization, whole genome sequencing, and comparative genomics (Hong et al., 2012; Gan et al., 2014; Walterson and Stavrinides, 2015).

The strong co-occurrence of Xanthomonas and Novosphingobium with A. vitis as observed in this study is consistent with the affiliation of their members with grapevine and more specifically with CG tumor tissues. For example, an AHL-producing Novosphingobium sp. strain Rr 2-17 was isolated from a CG tumor and its quorum sensing AHL-signal production was shown via transposon mutagenesis to be regulated by the stringent response gene, rsh (Gan et al., 2009). The TraR protein associated with conjugal transfer of Ti plasmid reacts strongly to the AHL signals generated by strain Rr 2-17, suggesting that it may play a role in amplifying the quorum sensing signal required to disseminate Ti plasmid (via conjugal plasmid transfer) among Agrobacterium, Allorhizobium, and potentially additional compatible strains in the tumor environments (Zhu and Winans, 2001; Zhang et al., 2002; Lowe et al., 2009). In silico identification of quorum sensing synthase gene from Novosphingobium genomes indicates most members exhibit the genomic potential to produce AHL (Gan et al., 2013, 2015, 2016). Known for their ability to mineralize complex aromatic compounds, it is possible that Novosphingobium spp. might harbor homologous genes associated with the catabolism of opine compounds or intermediates which have a structural resemblance to oxygenase-cleaved aromatic compound (Wilcke, 2000; Gan et al., 2011, 2012b; Mallick et al., 2011). Xanthomonas spp. also have been reported to be commonly present on grapevine leaves and bark, but absent from grapes or grapevine-associated soils as determined through culture-dependent and culture-independent approaches (Martins et al., 2013). The strong co-occurrence of Xanthomonas with A. vitis in CG tumors can be explained by the presence of high concentrations of lignin, cellulose, N-glycosylated proteins, and other cell wall precursors as well as cell wall degradation products found in bark on developing CG tumors (Blanvillain et al., 2007; Boulanger et al., 2014).

The identification and delineation of agrobacteria even based on full-length 16S rRNA gene sequences has been problematic as indicated by polyphyletic clustering pattern and low bootstrap support values, as observed in various phylogenetic trees even with high taxon sampling. The lack of informative sites is presumably due to the slower evolutionary rate of 16S rRNA among agrobacteria (Costechareyre et al., 2010). Despite being able to resolve amplicon down to the single base resolution, the monophyletic clustering of ASV2 with multiple genomic species of Agrobacterium exhibiting diverse and distinct ecological niches strongly suggests that caution needs to be exercised when inferring ecological interactions from 16S rRNA dataset. An improved understanding of the microbial interactions within such an environment can be gained through whole metagenome shotgun sequencing approach coupled with the ProxiMeta^TM Hi-C metagenome deconvolution method that can link plasmids to their hosts (Press et al., 2017). However, the challenges associated with strong host (grapevine) gDNA contamination will need to be addressed through increased sequencing depth and/or the selective removal of host methylated gDNA (Feehery et al., 2013).

The widely used Illumina standard V3–V4 16S rRNA primers are not suitable for grapevine microbiota studies as they exhibit perfect match to the grapevine plastid 16S rRNA gene. We report a new pair of V3–V4 16S ribosomal RNA gene PCR primers which prevent the co-amplification of this gene region and used this primer pair to investigate the microbial community of 73 CG tissue samples collected from multiple distinct geographic regions. The CG microbial community is diverse and varies significantly across samples and vineyards.

Sequencing data generated from this study have been deposited into the NCBI Sequence Read Archive under the BioProject code PRJNA490446.

HG, MS, and ES conceived the project, designed the experiments, and drafted the manuscript. ES, RF, SC, LK, AK, and TB collected the samples and performed the DNA extraction. HG performed the amplicon sequencing and conducted the bioinformatic analysis. All authors edited and contributed to the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.