Glutamic Acid Reshapes The Phytobiome To Protect Plants Against Pathogens

Background: The physiology and growth of plants are strongly inuenced by their associated microbiomes. Conversely, the composition of the phytobiome is exible, responding to the state of the host and raising the possibility that it can be engineered to benet the plant. However, technology for engineering the structure of the microbiome is not yet available. Results: Here we show that glutamic acid reshapes the plant microbial community and enriches populations of Streptomyces, a functional core microbe, both above and below ground, in strawberry and tomato. Upon application of glutamic acid, the population size of Streptomyces increased dramatically in the anthosphere and the rhizosphere. At the same time, diseases caused by species of Fusarium were signicantly reduced in both habitats. Plant resistance-related genes were not activated, suggesting that glutamic acid modulates the microbiome community directly, rather than activating the host’s own protective mechanisms. Conclusions: Much is known about the structure of plant-associated microbial communities, but little has been learned about how the community composition and complexity are controlled. Our results demonstrate that the microbiome community can be engineered and unlock the mode of action of glutamic acid.

Because plant microbiomes play a critical role in plant development and health [13][14][15], it is reasonable that maintenance of a healthy microbiome would promote growth and crop yield in agricultural systems [16,17]. Our understanding of the plant-associated microbial community has expanded in recent years to include less abundant or even unculturable taxa, resulting in awareness of interactions with an everincreasing microbial diversity and recognition that plants do not exhibit normal growth and physiology without interactions with their associated microbes [18][19][20]. This ecological and functional integration of the plant and its microbiome is encompassed within the holobiont, the assemblage of the host and the other species living in or around it, which together form a discrete ecological unit. Moreover, complex microbial populations reside in association with all plant tissues, implying that the initial phases of colonization, as well as subsequent microbe-microbe interactions, selectively in uence the structure of the microbiome [19,[21][22][23] and that the core microbial community has a vital role in the overall microbiome stability and the tness of the host [24,25].
Trends in plant microbiome studies have approached microbiome engineering with the goal of improving plant health and productivity [26] via either top-down or bottom-up approaches [25]. The top-down approach refers to manipulation of environmental and physicochemical conditions to select the desired biological process [9,27]. This top-down design involves relatively macro-scale processes resulting in microbiome engineering. Conversely, bottom-up approaches link molecular and biochemical characteristics and relatively micro-scale processes with precise mechanisms in the interaction [27].
Consequently, these approaches are exible enough to be applied to microbiome-associated phenotypes [27]. In humans or animals, they can be facilitated by prebiotics, a terminology invented in 1995 by Gibson and Roberfroid [28]. Such prebiotics may selectively in uence the gut microbiome [29][30][31][32] or move to other organs through the blood, directly in uencing animal health [33]. In botanical systems, a similar role is played by biostimulators [34], de ned as substances applied to plants with the aim to enhance nutrition e ciency, abiotic stress tolerance and/or crop quality traits, regardless of nutrient content [35,36]. In recent years, the terminology has been extended in scope from screening substances to understanding their mode-of-action [37,38]. Thus, humic acid, fulvic acid, and seaweed extract were identi ed as biostimulators that enhanced tolerance against abiotic stress, promoted plant growth, and improved soil quality [35,37]. Such biostimulators may contain plant hormone-like compounds or activate hormone activities as their mode-of-action but so far, the mechanisms underpinning biostimulator function are poorly understood [37].
We previously observed collapse of the anthosphere microbial community structure coincident with the aging of strawberry plants [6]. In particular, the loss of diversity and reduction in population density of Streptomyces globisporus SP6C4, a core microbe, was negatively correlated with onset of two major anthosphere diseases, gray mold (Botrytis cinerea) which includes brown spots on ower petals, and blossom blight (Cladosporioides sp.), which appears as fuzzy gray mycelium on ower pistils and stamens. We then hypothesized that a speci c plant metabolite could be amended to rebuild the microbial community structure to maintain the health of the plant. Here, we propose that glutamic acid con gures the microbial community and modulates the abundance of S. globisporus SP6C4.

Microbiome collapse and disease
The microbial community in strawberry owers shifts throughout the growing season from one of high diversity (weeks 0-12) to one of low diversity (weeks 14-24), a pattern coincident with the loss of S. globisporus SP6C4, which we consider to be a core member of the ower microbial community [6]. Here, we recalculated strawberry ower microbial population data to identify the top 10 OTUs and the diversity of the microbial community throughout the growing season (Fig. 1a). In contrast to the incidence of gray mold disease, which increased from weeks 14-24, diversity in the anthosphere community during weeks 1-12 collapsed on week 14 at a time coincident with the onset of disease in plants exposed to the pathogen (Fig. 1b,c). These patterns indicated that the microbial community structure changed with the age of the plant or stage of blossoming, and that the collapse of the community included loss of the core microbial population.
Effect of plant exudates on the core microbe Petal and ovary samples were analyzed for amino acids, organic acids and soluble sugars ( Fig. 1d and Additional le 1: Figure S1). Amino acids in the petals did not differ signi cantly throughout periods of low and high disease incidence except for glutamic acid and proline, the content of which decreased signi cantly during the period of high gray mold disease incidence. In contrast, the content of the 23 amino acids in the ower ovary did not vary regardless of disease incidence (Additional le 1: Figure   S1a). Unlike the situation with amino acids, the content of soluble sugars and organic acids in petals and ovaries did not differ between periods of low and high disease (Additional le 1: Figure S1b,c,d). These ndings suggest that glutamic acid and proline have a key role in maintaining microbial diversity and the density of the core bacterium SP6C4 in the anthosphere. Biolog plates PM1 and PMB3 were used to identify carbon (PM1) and nitrogen (PMB3) sources in uencing the core strain. S. globisporus SP6C4 grew equally well on the carbon substrates (Additional le 1: Figure S2a and Additional le 2: Table S1), but growth on nitrogen substrates increased markedly on L-tyrosine, L-proline, L-aspartic acid, L-cysteine, agmatine, and L-glutamic acid (OD 590 ≥ 0.4) (Additional le 1: Figure S2b and Additional le 2: Table S2).
However, only L-glutamic acid in uenced the growth of S. globisporus SP6C4 in both the amino acid analysis of owers and the Biolog plates. Bacterial growth was further evaluated with the four amino acids L-glutamic acid, L-proline, aspartic acid, and L-tyrosine, with basal medium and L-asparagine as negative controls and L-tryptophan as a positive control. SP6C4 did not grow on unamended basal medium, but growth at OD 600 was 0.6 for L-asparagine, 0.7 for aspartic acid, 0.8 for L-proline, 0.8 for Ltryptophan and 1.3 for L-glutamic acid (Fig. 1e). As additional assessment, growth on each amino acid was evaluated at 0.02%, 0.2% and 2% ( Fig. 1f and Additional le 1: Figure S2c,d) in basal medium. Regardless of concentration, L-glutamic acid had the greatest effect on the growth of SP6C4.

Effect of glutamic acid on anthosphere diseases
To investigate the effect of L-glutamic acid on the occurrence of gray mold and blossom blight diseases and the density of strain SP6C4 in strawberry owers, disease incidence (DI) was evaluated at two-week intervals. At the same time, L-glutamic acid and L-asparagine were sprayed three times, at two-week intervals, from week 4 to week 8 in a strawberry greenhouse (Additional le 1: Figure S3). Gray mold DI from week 0 to week 4 remained relatively low (10 -16%) regardless of treatment. At 6 weeks, the untreated control presented a DI of 16.6%; the DI in the L-asparagine-treated plot was 16%, and the DI in the plot treated with L-glutamic acid was signi cantly lower, at 12.4%. At week 8, the DI in the untreated control increased to 34% but that in the plot sprayed with L-glutamic acid was maintained below 17% (Additional le 1: Figure S4a-c). DI values for blossom blight presented even greater differences among the treatments. At 8 weeks, 35% of owers in the untreated control and 36% of those treated with Lasparagine developed disease symptoms, whereas fewer than 11% showed symptoms of blossom blight in the L-glutamic acid-treated plots (Additional le 1: Figure S4d-f).
Glutamic acid restructured the anthosphere microbiome We next evaluated whether L-glutamic acid can modulate the structure of the microbial community. Sequencing analyses were performed on an Illumina MiSeq platform resulting in a total of 3,307,450 reads (Additional le 2: Table S3 and Additional le 3: Table S4) and 162 operational taxonomic units (OTUs) in the strawberry ower. Microbial diversity over 8 weeks in the untreated control did not vary signi cantly, whereas owers treated with L-asparagine showed increased diversity and those sprayed with L-glutamic acid had signi cantly lower alpha diversity at 6 and 8 weeks ( Fig. 2a and Additional le 1: Figure S5a,b). We then used the Silva database to identify community members responsible for the shift in overall community structure. The heatmap and tree present the relative abundance of the most common OTUs (Fig. 2b). Enterobacteriaceae from weeks 2-8 had high relative abundance in the untreated control (93%, 92%, 94%, and 75%), in the L-asparagine treatment (96%, 81%, 56%, and 94%), and from weeks 2-4, in the L-glutamic acid treatment (99.8%, and 73.54%). The Pseudomonadaceae had the second most abundant OTUs followed by Moraxellaceae (Fig. 2b,c and Additional le 1: Figure S5c). Enrichment of Streptomycetaceae occurred only in the L-glutamic acid treatment during weeks 6 and 8, accounting for 99.98% and 99.99% of the community. These results are indicated by a change in the color of the heatmap from purple to yellow (Fig. 2b). Compared to the relative abundance of OTUs calculated as log2 ratios in Metacoder, Streptomycetaceae had a log2 ratio value of 3 only upon treatment with L-glutamic acid ( Fig. 3a and Additional le 1: Figure S6).
The effect of the amino acids on the structure of the microbial community and suppression of plant diseases was analyzed with principal coordinate analysis (PCoA, beta diversity), non-metric multidimensional scaling (NMDS, Bray-Curtis distance method), and 3D plots of the disease incidence enrichment of the core microbe. PCoA data clearly distinguished groups, one that consisted of the untreated control and the L-asparagine treated samples (weeks 2 and 4) and the other, comprised of Lglutamic acid-treated owers (weeks 6 and 8) (Fig. 2d). At the OTU level, PCoA analysis presented 11 OTUs with sequence identity to S. globisporus SP6C4 of greater than 98%. The microbial communities of L-glutamic acid-treated ower samples (weeks 6 and 8) were also clearly distinguished (Fig. 2e).
Anthosphere microbial community structures were affected by the occurrence of gray mold and blossom blight (Fig. 3b). The dispersion in NMDS indicated that microbial community structure was co-related with patterns of disease occurrence, and especially that occurrence was suppressed by treatment with Lglutamic acid. A 3D plot with three vectors and colored circles indicating the abundance of Streptomyces OTUs showed that the population of the core microbe represented more than 90% of the microbial community in L-glutamic acid-treated samples on weeks 6 and 8 (Fig. 3c). The population of SP6C4 was determined by qPCR with primers for the SP6C4-speci c lanM lantipeptide biosynthesis gene [39]. On owers sprayed with L-glutamic acid, the density of SP6C4 was greater than 10 5 copy/g of ower, but lanM gene copies at 8 weeks on the untreated control and those sprayed with L-asparagine had fewer than10 4 copy/g of ower (Additional le 1: Figure S4g). These results were transformed as a bubble plot that veri ed that L-glutamic acid increased the population size of strain SP6C4 on strawberry owers and was responsible for the low disease incidence but high lanM copy number from week 4 to week 8 compared to the untreated and L-asparagine controls (Additional le 1: Figure S4h). We also generated a PICRUSt2 pro le of the metabolic pathways present in the microbiome community of treated owers by using the KEGG database. The results showed that treatment with L-glutamic acid enhanced the abundance of Streptomyces pathways such as for biosynthesis of the type II polyketide backbone, sesquiterpenoids, neomycin, nonribosomal peptide siderophore, degradation of glycosaminoglycan and glycan, metabolism of glutamic acid, glutamatergic synapse, and limonene and pinene pathways (Fig. 3d-f and Additional le 1: Figure S7).

Effect of glutamic acid on soil-borne disease
The anthosphere has a very simple microbiome with only a few species in the community. To extend our results to a more complex microbiome, we tested the effect of glutamic acid against Fusarium oxysporum f. sp. lycopersici (FOL), the causal pathogen of Fusarium wilt disease of tomato. To evaluate whether SP6C4 and L-glutamic acid could in uence the occurrence of this soil-borne disease, we established seven experimental treatments: untreated control, S. globisporus SP6C4, L-glutamic acid, FOL, SP6C4+FOL, L-glutamic acid+FOL and SP6C4+L-glutamic acid+FOL. Disease severity was evaluated seven weeks later on a scale of 0-5 (Additional le 1: Figure S8a-c). Plants treated with FOL alone presented severe disease symptoms at week 6 and all of them were dead at 7 weeks, but the other plants had disease indices of less than 2 even at 7 weeks (Additional le 1: Figure S8c).
qRT-PCR of the lanM gene in the tomato rhizosphere revealed that plants treated with SP6C4 or even Lglutamic acid alone had more than 10 6 lanM gene copies per 100 ng of soil DNA regardless of the presence of the pathogen. However, the untreated control and FOL only-treated plants had fewer than 10 2 lanM gene copies, the background level (Additional le 1: Figure S8d). Taken together, the results showed that SP6C4 or L-glutamic acid successfully suppressed Fusarium wilt disease in tomato and enhanced the population density of microbes carrying the lanM gene in the rhizosphere. Additionally, at 7 weeks, both shoot length and shoot weight were signi cantly reduced by treatment with FOL, but the damage was lessened by strain SP6C4 or L-glutamic acid (Additional le 1: Figure S8e,f).

Rhizobiome composition was shifted by glutamic acid
To investigate changes in the microbial community structure over time, tomato rhizosphere samples were collected 1, 3, 7, and 10 weeks after treatment except for the FOL only treated samples, none of which survived for 10 weeks. Total sequencing read numbers were 5,863,545 (Additional le 2: Table S5 and Additional le 4: Table S6) and the number of OTUs counted was 3,247. All sequences were compared with taxa in the Greengenes database at a similarity cut-off value of ≤98%, and the top 10 OTUs present in greatest relative abundance were visualized at the phylum level (Additional le 1: Figure S9a). Proteobacteria, Actinobacteria, Firmicutes and Bacteroidetes represented more than 5% of the community. Heatmap analysis revealed changes in the rhizosphere microbial community structure, which was divided into three distinguishable clusters (Fig. 4a). The rst cluster was enriched in Bacillaceae by treatment with strain SP6C4 or SP6C4+FOL (Group I; GR I); treatment with L-glutamic acid led to enrichment of Burkholderiaceae (Group II; GR II); and the third cluster included unenriched members of the community (Fig. 4a). Separation of L-glutamic acid and SP6C4-enriched taxa from other members of the community was visualized in an NMDS plot (Additional le 1: Figure S9b).
Co-occurrence patterns and relationships within the tomato rhizosphere microbiome among the six different treatments described above were analyzed by using Spearman's algorithm (R version 3.4.4). to create a rank of co-occurrence network pattern (Spearman's ρ > 0.8) that represented either positive or negative relationships among the community members. Streptomycetaceae, Burkholderiaceae and Bacillaceae were selected as keystone taxa in the positive relationship. In the negative clusters, Caulobacteraceae, Chitinophagaceae, Devosiaceae, Rhizobiaceae and Xanthobacteraceae were identi ed as important taxa (Additional le 1: Figure S10). Based on the network results, we analyzed the abundance of SP6C4 in both the positive and the negative clusters. At week 3 before inoculation with FOL, communities of the untreated control and those after treatment with FOL had more negative clusters than positive relationships but the nding was not signi cant (Fig. 4b).These results indicated a limitation to microbial composition analysis with only two clusters. Therefore, we also analyzed the relative abundance (RA) of the keystone taxa in both positive (Streptomycetaceae, Bacillaceae, and Burkholderiaceae) and negative (Caulobacteraceae, and Chitinophagaceae) clusters. Before treatment with FOL, Streptomycetaceae were present at 40% RA in the FOL+SP6C4 treatment. The RA of Bacillaceae was 60% in the SP6C4 treated rhizosphere and that of Burkholderiaceae was the greatest (60%) in the rhizosphere of L-glutamic acid-treated plants. Interestingly, in the untreated control plants, keystone taxa in the negative clusters were the most abundant microbes (Caulobacteraceae, 60% and Chitinophagaceae, 45%) (Fig. 4c). Taken together, the microbial community structure in the tomato rhizosphere was affected by introduction of strain SP6C4 or L-glutamic acid. The introduction of the core microbe, SP6C4 enriched Bacillaceae, and drenching with L-glutamic acid increased the density of Burkholderiaceae in the rhizosphere. The ndings indicated that SP6C4 and L-glutamic acid have different modulating effects on the rhizosphere microbiome community. The disease indices after FOL treatment alone reached 3; plants treated with FOL+L-glutamic acid+SP6C4 had a disease index of 2, and plants treated with FOL+antibiotics developed a disease index of 1.6. However, disease development in plants treated with L-glutamic acid+FOL, L-glutamic acid alone, or antibiotics alone did not differ signi cantly from the untreated plants (Additional le 1: Figure S11).

Glutamic acid does not activate ISR in strawberry or tomato
We wondered whether L-glutamic acid inhibited disease occurrence by activating the plant's induced systemic resistance (ISR) response in either strawberry or tomato. For strawberry, a total of 7 treatments was examined including an untreated control, pathogen (Botrytis cinerea) only, L-glutamic acid, antibiotics, L-glutamic acid with antibiotics, antibiotics with the pathogen, and L-glutamic acid, antibiotics, and pathogen (Fig. 5a). The pathogen only treatment showed 100% disease incidence, but with L-glutamic acid and the pathogen, infection was less than 50% (Additional le 1: Figure S12a,b). As expected, the population of the SP6C4 was increased signi cantly in treatments with L-glutamic acid only and L-glutamic acid with the pathogen (10 5 lanM gene copy per g of ower) (Fig. 5b). The in uence of Lglutamic acid on activation of plant ISR genes was evaluated by qRT-PCR. The jasmonic acid (JA) related genes LOX2 and PR10 were not expressed in the untreated control or in treatments with L-glutamic acid and L-glutamic acid with the pathogen. However, LOX2 and PR10 were highly expressed in the pathogen only, antibiotics with the pathogen, and L-glutamic acid plus antibiotics and pathogen treatments ( Fig. 5d and Additional le 1: Figure S13a). Expression of salicylic acid (SA) related genes (PR1 and PR2) showed a pattern similar to that of JA related gene expression; the pathogen, antibiotics with pathogen, and Lglutamic acid, antibiotics and pathogen treatments triggered expression of the PR1 and PR2 genes. These SA related genes were not expressed in the untreated control or in treatments with L-glutamic acid or L-glutamic acid with pathogen ( Fig. 5e and Additional le 1: Figure S13a).
The population size of the SP6C4 as measured by qRT-PCR with lanM in the rhizosphere soil of the Lglutamic acid only and L-glutamic acid+FOL-treated plants was 10 5 gene copies/g of rhizosphere soil. However, the untreated control, FOL, antibiotics, antibiotics+FOL, and L-glutamic acid+antibiotic+FOL treated plants showed signi cantly lower density of the core microbe, with only 10 3 lanM gene copies/g of rhizosphere soil (Fig. 5c). Collectively, we interpret these results to indicate that glutamic acid increased density of Streptomyces, the functional core microbe, which suppressed the fungal pathogen. In tomato, ISR and PAMP-related gene expression was evaluated relative to that of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH). JA related genes (Tomlex A, Tomlex C and PINII) were expressed after treatment of the seedlings with FOL, antibiotics, antibiotics+FOL and L-glutamic acid+ antibiotics+FOL, but not in the untreated control or after treatment with L-glutamic acid, and Lglutamic acid+FOL ( Fig. 5f and Additional le 1: Figure S13b). Among SA-related genes (PR1b1, PR-P2 and SAMT), only PR1b1 was expressed, and only in plants treated with FOL, antibiotics, FOL+antibiotics, and FOL+L-glutamic acid+antibiotics but not in the untreated control or in plants treated with L-glutamic acid or L-glutamic acid+FOL ( Fig. 5g and Additional le 1: Figure S13c). The ET related gene ERF1 was not expressed in any of the treatments (Additional le 1: Figure S13d) and the PAMP response gene Pti5 showed response patterns similar to those of the JA related genes (Additional le 1: Figure S13e). Taken together, these ndings suggest rst, that plant ISR was not activated by L-glutamic acid; second, that Lglutamic acid recon gured the anthosphere and rhizosphere microbiome communities; and third, that the engineered microbiomes protected the plant from the pathogens.

Discussion
That microbes from the soil play a critical role in plant health has been known and investigated for well over century [40], but it is only recently that the assemblage of the microbiota selected by the plant from the environment (i.e., the microbiome), together with the the host, has been recognized as an ecological unit, the 'holobiont' [41,42]. Collectively, the microbiome extends the genetic and physiological capacity of the host, contributing to its growth and well-being by providing ecological services and protection from biotic and abiotic stresses.
Much as the animal gut microbiome can be in uenced by probiotics, diet [32,43,44] or prebiotics [45][46][47] with the potential to engineer its composition or activity, so also is the structure of plant microbiome community responsive to the types and amounts of metabolites present in plant exudates secreted into the rhizosphere [48,49]. The plant developmental stage and genotype in uence the microbiome community structure as well as the root architecture and chemistry, which have a signi cant impact on microbiome composition [50]. The quality and quantity of root exudate directly impacts rhizosphere microbiome assembly.
We have shown here that glutamic acid, either secreted by the plant or added exogenously, functions as a prebiotic and plays a key tool in a bottom-up model of plant microbiome engineering [25,51] built around Streptomyces as a core member of the microbial community. In both the anthosphere of strawberry, with its very simple microbiome, and in the complex rhizosphere microbiome of tomato, glutamic acid initiated a cascade resulting in recon guration of the microbiome and enrichment of Streptomyces in the community. Of note is that as a consequence of this process, both foliar and root pathogens were controlled. on unrelated plant species. It is not uncommon for chemicals applied to plants to induce systemic resistance, but with glutamic acid the effect was not due to the induction of resistance through either the ethylene/jasmonic acid or the salicylic acid pathway. We were surprised by the extent of modi cation of the two microbiomes by the addition of a single chemical. However, some substrates are preferentially metabolized by microbes [45] and can selectively engineer the composition or activity of entire microbial communities, in uencing the health of the host [46,52], much as when diet affects the composition of the gut microbiome [29,32,44,53]. Our results indicate that glutamic acid functions directly as a link to the microbiome; it directly affected the microbiome community structure and engineered it to suppress disease incidence. Moreover, the fact that glutamic acid did not activate host plant resistance mechanisms suggests that it may provide insight into evolutionary and functional relationships between the plant and its microbiome. Glutamic acid, in particular, is metabolized by Streptomyces as sole source of carbon and nitrogen, favoring vigorous growth [54], which may help to explain its effect on the plant-associated communities we observed in this study. How did the relationship of the host and the microbiome evolve? Perhaps the answer can be found in the biological function of the microbiome. Plants are constantly exposed to changing environmental forces that also act to shape the microbiome, but at the same time, the microbiome community structure is exible and capable of buffering the impact of the environment on the host. In our experimental system, the community structure and abundance of the phytobiome were in uenced by the amount of glutamic acid available via exudates or by exogenousdelivery. Given that glutamic acid is naturally present in host exudates, it would seem that plants already have the potential to engineer protective microbiomes themselves. Thus, with better understanding of the relationship between plant exudates and microbiome assembly, it may be possible to develop crops that can recruit their own microbiota to better withstand pathogen attack.

Conclusions
While our results indicate clearly that glutamic acid is a powerful mediator of the structure of the plantassociated microbiome, much remains to be determined about how it interfaces with the complex metabolic and signaling exchanges among microbes and their plant hosts. There has been considerable progress in recent years towards elucidating the structure and function of plant-associated microbial communities, but new approaches are needed to reveal how the composition of the community and its function are controlled. Based on the results of this study, we propose that glutamic acid con gures the microbial community and modulates the composition of a core microbiome that bene ts the plant by in uencing such agronomic metrics as crop quality and yield.

Strawberry petal and ovary samples for exudate analysis
For amino acid (AA) and organic acid (OA) analysis, 100 µg of freeze-dried petal or ovary was mixed with 10 mL distilled water, sonicated for 30 min, and ltered through a hydrophilic PVDF membrane (Millipore Durapore®, 0.45 μm, 25 mm diameter) by centrifugation at 1,077 g for 5 min. AAs and OAs in the samples were analyzed by gas chromatography-mass spectrometry (GC−MS) using an Agilent 6890 N gas chromatograph interfaced with an Agilent 5975B mass-selective detector (70 eV, electron impact mode) equipped with an Ultra-2 (5% phenyl-95% methylpolysiloxane bonded phase; 25 m ´ 0.20 mm i.d., 0.11 μm lm thickness) cross-linked capillary column (Agilent Technologies, Palo Alto, CA, USA). The temperatures of the injector, interface, and ion source were 260, 300, and 230°C, respectively. Helium was used as the carrier gas at a ow rate of 0.5 mL min -1 in the constant ow mode. Samples were loaded in the split-injection mode (10:1); the oven temperature for AA pro ling was initially set at 120°C (2 min), rose rst to 240°C at 5°C min -1 then to 300°C (3 min) at 30°C min -1 . The oven temperature for OA analysis was initially 100°C (2 min), rose rst to 240°C at 5°C min -1 , and then to 300°C (5 min) at 30°C min -1 . The mass range scanned was 50-600 u at a rate of 0.99 scans per sec. In the selected ion monitoring (SIM) mode, three characteristic ions for each AA and OA were used for peak identi cation and quanti cation.
Amino acid and organic acid pro ling and pattern recognition AA analysis was performed by using a previous method [57,58]. Brie y, 0.5 ml aliquots from the petal or ovary were adjusted to pH ≥ 12 with 5.0 M NaOH and diluted with 0.5 mL distilled water and 0.1 mg of norvaline as internal standard. A two-phase ethoxycarbonylation (EOC) reaction was immediately conducted in the aqueous phase. The reaction mixture was then acidi ed (pH ≤ 2.0) with 10.0% sulfuric acid, saturated with sodium chloride, and subjected to extraction sequentially with diethyl ether (3. Carbon source analysis of strawberry owers Soluble sugars including glucose, fructose, maltose, ra nose, and sucrose were analyzed as described by Yoon et al. [59]. Flower samples (0.1 g) were homogenized in glass tubes with 6 mL of HPLC grade ethanol (80%), and incubated at 65°C for 20 min. The supernatant fraction was collected after centrifugation at 3500 rpm for 10 min and the process was carried out three times. The pooled extracts were ltered through a 0.45 µm syringe lter and then concentrated under nitrogen. Sugar content was determined with an Agilent 1100 high performance liquid chromatograph (HPLC) with a refractive index detector (Agilent Tech., Germany) after baseline resolution of a column (ZORBX, 4.6 X 150 mm, 5 mm particle size, Agilent Tech) at a ow rate of 1 mL/min. Samples (20 µL) were injected with 75% acetonitrile and sugar content was calculated with an internal standard.

Carbon and nitrogen source utilization
Streptomyces globisporus SP6C4 was grown on MS medium (20 g mannitol, 20 g soya, 20 g agar per L) at 30°C for 5 days. A single colony was streaked on a fresh plate and mature spores were recovered after 10 days with a sterilized cotton ball and 1 mL of ddH 2 O. After ltration, the spore concentration was adjusted to an OD 600 nm of 2.0, mixed with 0.2 % carrageenan stock solution, and incubated, 100 µL per well, in sealed plates (PM1-carbon sources and PM3B-nitrogen sources) (Biolog, Bremen, Germany) at 28°C for 2 days. Then 10 µL of Biolog redox dye was added to each well and the intensity of color change was monitored at OD 590 nm every 30 min for 3 hours with a Synergy H1 Hybrid Multi-Mode microplate reader (BioTek, Winooski, VT, US) [60].

Disease incidence of gray mold and blossom blight and qPCR of lanM
The incidence of gray mold and blossom blight caused by Botrytis cinerea and Cladosporium spp., respectively, was expressed as the percentage of infected plants in a greenhouse of 9 plots, each with 100 strawberry plants. Early symptoms of gray mold included brown spots on ower petals and were followed by gray conidia covering owers and fruits [61]. Blossom blight appeared as gray fungal growth on ower pistils and stamens and as infected, malformed or misshapen fruits [62]. Differences in disease incidence among an untreated control and treatments with 2% glutamic acid or L-asparagine were analyzed by followed by the paired ANOVA and compared for mean separation with the untreated plots with Tukey's HSD (P = 0.05).
To determine whether the population size of the core microbe S. globisporus SP6C4 increased in response to the amino acid treatments, microbial DNA from the ower anthosphere was extracted and the SP6C4-speci c marker gene lanM was quanti ed by qPCR with F and R primers as described by Kim et al. [39]. qPCR reactions in SYBR Green ® TOYOBO master mix included denaturation at 98°C for 5 min followed by 40 cycles of denaturation at 98°C for 30 sec, annealing at 59°C for 30 sec and elongation at 72°C for 45 sec with a CFX Connect™ Optics Module Real-Time PCR System (Bio-Rad, Hercules, CA, USA).

Microbial community analysis of strawberry owers
Flowers were collected from a 660 m 2 greenhouse with 15 plots of 1.5 X 3 m 2 , each with 100 strawberry plants. Each of three treatments (untreated control, L-glutamic acid or L-asparagine at a nal concentration of 2%, pH 6.5) in ve randomly arranged replicate plots was sprayed for 1 min per plot (Sprayer: HP-2010, Korea, 1.5 L discharge capacity min -1 ) at two-week intervals during January and February, 2018. Five samples per plot, each with 3 to 5 owers, were collected into 50-mL Falcon tubes at two-week intervals from December, 2017 through February, 2018, chilled on ice to preserve microbial communities, and transported to the laboratory for sequence analysis. on an Illumina MiSeq platform (Illumina Inc., San Diego, CA, USA). Sequences of 300 bp or more and nucleotide quality scores >30 were recovered after screening with the DADA2 package in R (version 1.14).
The Silva database (http://www.arb-silva.de/) for OTU clustering was used to assign taxonomy of OTUs and alpha diversity with a taxonomic classi cation similarity cutoff of ≤ 97%, principal coordinate analysis (PCoA) and nonmetric multidimensional scaling (NMDS), and OTU bars were visualized with ggplot2 (R, version 3.4.4). Superheat (version 0.1.0) was used to generate heatmaps and OTU abundance was calculated with Metacoder (version 0.3.3) and PICRUSt2 (version 2.1.4 beta). Accession numbers for all sequencing data were recorded in GenBank (Additional le 2: Table S7).
Effect on gray mold incidence of antibiotic and amino acid treatments to engineer the microbiome community Strawberry seedlings (cv. Meahyang) were stored at -2°C for one month for vernalization and then transferred to plastic pots (10 cm diameter). Twelve days after planting, each plant had 5-7 owers. Then, L-glutamic acid (5 µg/mL) and the antibiotics erythromycin and clindamycin (10 µg/mL each, to inhibit Streptomyces, Research Products International, Mt. Prospect, IL, USA) [63][64][65][66] were applied with a sprayer. Three days later, freshly grown conidia of B. cinerea were collected with a cheese cloth lter and sprayed at 10 5 cfu/mL on the owers. The seven treatments of 5 plants each included an untreated control, pathogen only (B. cinerea), L-glutamic acid only, antibiotics only, Glu + pathogen, antibiotics + pathogen, and Glu + antibiotics + pathogen. All plants were maintained in a growth chamber with a daytime temperature of 25°C ± 3; a nighttime temperature of 15°C ± 3; and relative humidity of 85%. Seven weeks later, disease incidence was scored on 30 owers (10 independent replicates) and the lanM gene was quanti ed on 1 g of owers (n = 3 to 5 owers).
For qRT-PCR, RNA was extracted from the ower samples using the plant RNA single-step extraction method [67,68]. Each sample (100 ± 0.5 mg) was added to a 2-mL tube of lysing matrix E (Fast DNA TM Spin Kit for Soil DNA extraction, MP Biomedicals) with 1 mL of TRIzol ® Reagent (Invitrogen) and homogenized with a FastPrep-24 instrument (MP Biomedicals) for 1 min. Four jasmonic acid (JA-) and salicylic acid (SA-) related ISR marker genes with the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene as a standard housekeeping gene [69][70][71] were detected by qRT-PCR. For reactions, 1 µg of RNA was used as template to synthesize cDNA with a TOYOBO ReverTra Ace ® qPCR RT Kit (Toyobo Co., Osaka, Japan). After cDNA synthesis, 20 µL of the products were diluted 1:5 with RNase-free water and 4 mL was mixed with 25 μL of SYBR Green ® TOYOBO master mix, 1 μL of each forward and reverse primer, and 16 μL of HPLC grade H 2 O. The PCR program included an initial denaturation at 98°C for 1 min, followed by denaturation at 98°C for 30 sec, annealing as indicated in Additional le 2: Table S8 and 60°C for 30 sec, and elongation at 72°C for 45 sec for 40 cycles. qRT-PCR was performed with a CFX Connect™ Optics Module Real-Time PCR System (Bio-Rad, USA). All primer information is presented in Additional le 2: Table S8. The experiment was conducted with three technical replications.
Fusarium wilt disease suppression on tomato by strain SP6C4 and with L-glutamic acid For assays of Fusarium wilt disease control, tomato plants (cv. Heinze) were maintained in a plant growth chamber for 4 weeks. Conditions included a 16 hr day cycle at 27 ± 2°C and an 8 hr night cycle at 20 ± 2°C, both at 80% relative humidity. Seed was sterilized in 1.5% NaOCl for 30 min with gentle shaking and washed 3 times with ddH 2 O. The seeds were germinated on damp cotton in a Petri dish (9-cm, diam.) for 3 days at 4°C and then transferred to plastic pots (10-cm, diam.) with autoclaved nursery soil. After 5 days' germination, 10 mL of L-glutamic acid (5 µg/mL) and 10 5 cfu/mL of Fusarium oxysporum f. sp. lycopersici (FOL) chlamydospores (10 mL) were drenched into the soil. Images of stem and leaf growth were captured two weeks later, at the early vegetative stage and at six weeks, (late vegetative stage). Shoot length, shoot fresh weight and disease indexes were scored weekly at 6 levels: (0, no symptoms; 1, slight yellowing of the lower leaves; 2, moderate yellowing of the entire plant; 3, wilted plant; 4, plants severely stunted or browning; 5, plants dead). All treatments had 3 biological replications and the mean ± SE of the results was calculated by one-way ANOVA in R (version 3.4.4.).
For sequencing analysis of rhizosphere populations, growth conditions of tomato plants were as described above. Tomato seedlings were grown in sterilized soil for 10 weeks and then 10 mL of Lglutamic acid (5 µg/mL) was applied by drenching 3 times at 3-day intervals between weeks 2 and 3. Strain SP6C4 was cultured in TSB broth containing 20% sucrose and 1% mannitol for sporulation. The harvested spores were washed four times with deionized, distilled H 2 O and the pellet was suspended in 50 mL (OD 595nm 0.7 ± 0.05) of 0.1 % Hoagland solution containing 0.1% methylcellulose (MC) and inoculated into the soil four weeks after planting. Seven days later, FOL chlamydospore stock (10 5 cfu/mL) was inoculated into the soil. The disease index was scored every 5 days for 20 days using the ve-grade scale above. At 10 weeks, the rhizosphere soil of 3 replicate plants was pooled for DNA extraction and sequencing. The seven treatments included an untreated control, FOL alone, L-glutamic acid (Glu) alone, SP6C4 alone, FOL + Glu, FOL + SP6C4, and FOL + Glu + SP6C4. Rhizosphere soil (0.5 g) was added to lysing matrix E and DNA was extracted using a FastDNA Spin Kit (MP Bio). The DNA was suspended in 50 mL of DES buffer and the tubes were stored at -20 °C for sequencing and lanM gene qRT-PCR. For sequencing, 200 ng of DNA was precipitated with ethanol and the V4 region of 16S rRNA was ampli ed with primers 515F forward (5'-gtgycagcmgccgcggtaa-3') and 806R reverse (5'ggactacnvgggtwtctaat-3'). PCR products were subjected to Illumina MiSeq 250-bp paired-end sequencing at Macrogen (Daejeon, Korea). For 16S rRNA gene-based bacterial community analysis, the data were trimmed of low quality reads (< 30 minimum quality score) and primer sequences by using the DADA2 package in R (version 1.14), quality ltered, and processed according to Greengenes data base (https://greengenes.secondgenome.com) with a taxonomic classi cation similarity cutoff of ≤ 98%. The most dominant OTUs were shown by NMDS and OTU bars using ggplot2 in the R package (R, version 3.4.4). Other visualizations were made using superheat (version 0.1.0) for heatmaps, OTU abundance was calculated with NOI-seq (version 3.10) and co-occurrence was calculated with Spearman's method.
All sequencing data and GenBank accession numbers were recorded in Additional le 2: Table S7.
Rhizosphere microbial community engineering with L-glutamic acid and suppression of Fusarium wilt disease in tomato Seedlings of tomato (cv. Heinze) were maintained in a growth chamber under day/night conditions of 16 h light and 8 h dark. Temperature during the light cycle was 25°C ± 2 and was increased after inoculation of FOL to 28°C ± 2 to enhance pathogen infection. The dark phase temperature was held at 20°C ±2 and humidity was no greater than 85%. Seeds were sown in autoclaved nursery soil and irrigated with 0.1% Hoagland's solution. After 12 days, the seedlings were treated with a 10 mL mixture of L-glutamic acid (5 µg/mL) and antibiotics (erythromycin and clindamycin, each at 10 µg/mL). FOL inoculation was performed with a chlamydospore stock solution (10 5 cfu/mL) at 15 days. At 6 weeks, the rhizosphere soil was collected to calculate lanM copy number and the expression of ISR-related genes by qRT-PCR (Additional le 2: Table S8). At the nal of sampling time (week 6), the wilt disease index was determined for 5 independent plants as 6 levels: 0, no symptoms; 1, slight yellowing of the lower leaves; 2, moderate yellowing of the entire plant; 3, wilted plant; 4, plants severely stunted or browning; 5, plants dead.
Rhizosphere samples consisting of 150 mg of soil closely adhered to roots were added to lysing matrix E tube (MP biomedicals), lysed in 1-mL of TRIzol ® Reagent (Invitrogen) and homogenized with a FastPrep-24 kit by a RNA single-step extraction method. The extracted RNA was cleaned with a spin column (RNeasy kit, Qiagen, Hidden, Germany). Ten μL of DNase and 70 μL of RDD buffer (RNeasy kit, Qiagen) were added on the column and incubated at ambient temperature for 15 min and then the column was washed with 350 μL of RW1 buffer and 500 μL of RPE buffer (RNeasy kit) at 8000 x g for 15 sec. The column was transferred to a new tube (1.5-mL) and incubated on ice for 1 min. For elution, 15 μL of RNase-free water was added and the column was centrifuged at 8000 x g for 1 min. For qRT-PCR, 1 μg of total RNA and oligo dT primers were used with a ReverTra Ace ® qPCR RT kit (Toyobo). qRT-PCR was performed with 4 µl of cDNA, 16 mL RNase free water, 25 μL of SYBR Green ® master mix (QPK-201T, Japan), and ΔΔ Ct values were calculated for JA (tomlex A, tomlex C, PINII), SA (SAMT, PR1b1, PR-P2), ET (erf) and PAMP; activated at pathogen infection (pti5)-related genes and actin and tubulin housekeeping genes [72][73][74]. All primer information is presented in Additional le 2: Table S8.

Statistical analyses
All data except for sequence analyses were analyzed by ANOVA and t-test. Comparisons were used to demonstrate differences among mean values with Tukey's HSD and graphs were visualized by ggplots

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availabilityofdata and materials
Sequencing data for L-glutamic acid-treated ower samples have been deposited in GenBank under SAR accession number SRR11355399 [https://www.ncbi.nlm.nih.gov/sra/SRR11355399] and all other GenBank data in Additional le 2: Table S7. All data are available in the manuscript the supplementary materials and analyses of microbial community composition were carried out with R program (version 3.4.4). The source code of R for data analyses is available on GitHub at https://github.com/ekfks0125/2020_Kim.git.

Competing Interests
The authors declare that they have no competing interests.