Multiple endogenous seed-born bacteria recovered rice growth disruption caused by Burkholderia glumae

Burkholderia glumae is a causal agent of bacterial grain and seedling rot in rice, and is a threat to stable global food supply. The virulence of B. glumae was suppressed when it was inoculated on budding seed rather than on non-budding seed. To clarify the phenomena, pathogen titer inside the rice plant was measured by serial dilution plating of lysates from budding rice seedlings. Surprisingly, morphologically different types of colonies were observed on the plates. These ‘contaminated’ rice seed-born bacteria (RSB) were identified by sequencing 16S rRNA genes as three strains of Pseudomonas putida (RSB1, RSB10, RSB15) and Stenotrophomonas maltophilia (RSB2). All bacteria and B. glumae were simultaneously inoculated onto rice seeds, and all three P. putida RSBs suppressed the growth disruption caused by B. glumae, whereas RSB2 had no effect. Thus, the virulence was synergistically suppressed when co-treated with RSBs. The effect could be dependent on the high biofilm formation ability of RSB2. By comprehensive microbiota analysis, endogenous rice flora were changed by RSBs treatment. These results suggest the possibility of novel pathogen control through pre-treatment with endogenous beneficial microorganisms. The method would contribute substantially to the implementation of sustainable agriculture stated in Sustainable Development Goals of United Nations.

Burkholderia glumae is a seed and soil-borne plant pathogen that causes bacterial seedling rot and bacterial grain rot of rice 1,2 . This is one of the most important diseases affecting rice production worldwide, particularly in East Asia, Southeast Asia, North America, and South America 3 . Because the optimal temperature range for the growth of B. glumae is 30-35 °C 4 , and production of toxoflavin is enhanced at high temperatures around 37 °C 5 , rapid multiplication may occur more frequently in tropical and semi-tropical countries, and the disease is likely to spread more rapidly with intense global warming. Shew et al. estimated the economic loss for a 1 °C increase in the Mid-South United States as a $112 million USD annual decrease in consumer surplus in that specific area of US, and a loss of production equivalent to feeding 2.17 million people 6 .
In Japan, most rice seeds are sown in nursery boxes, and seedlings are moved to nursery beds before transplanting to the rice field. Because nursery boxes are maintained under relatively high temperatures (28-30 °C) to promote simultaneous germination, seedling rot tends to occur in case seeds had been infected 7,8 . Sometimes apparently healthy seeds but contaminated with B. glumae are sown and transplanted, causing eventual development of grain rot in the field 3,9,10 . Recently, one of the measures to controlling plant disease is breeding tolerant rice varieties 11 , and some virulent B. glumae strains have been analyzed genetically to identify putative virulence genes. However, a rice cultivar resistant to B. glumae is not commercially available at this time. Other major means for the control of B. glumae in Japan 12 is seed treatment with oxolinic acid, a quinoline derivative. However, the occurrence of strains naturally resistant to oxolinic acid has been a serious limitation to this method of disease control 12,13 .
Endophytes can be defined as microbial communities (especially bacteria and fungi) that are found inside plant tissues without causing any apparent harm to the host. The presence of seed-borne bacteria has been documented throughout seed maturing stages of rice 14 and in the endosphere of mature rice seeds 15 . From a rice plant which was grown in the field, Bertani et al. isolated a total of 1,318 putative bacterial endophytes and made a working collection of 229 isolates 16 . In another study, a total of 4,155 bacterial operational taxonomic showed that seeds of rice naturally harbor bacterial endophytes that play key roles in modulation of seedling development 18 . Also, some endophytes were shown to have biocontrol ability and plant-growth-promoting functions 19 . However, the function and ecology of the great majority of bacterial endophytes inside the host plant is unknown. In this study, multiple rice seed born bacteria (RSB) were identified which showed an ability to suppress the virulence of B. glumae. Furthermore, the distribution of endogenous bacteria in plants that had been treated with the pathogen and RSBs changed dramatically. The result indicated a possible novel pathogen control technology mediated through endogenous microbiota. This will be also a nice model to clarify the mechanism of multiplication of opportunistic pathogens such as B. glumae.

Results
Pregerminated seedlings are more tolerant to B. glumae infection. The virulence of B. glumae on both pregerminated rice seeds and non-pregerminated seeds was tested. As shown in Fig. 1A,B, severe growth disruption was observed when non-pregerminated seeds were treated with B. glumae 301682. In contrast, the growth of pregerminated seeds treated with the pathogen was unaffected. A similar trend was shown in the case of B. glumae 301169, although it showed lower virulence than 301682 infection. A water-treated control showed normal growth. To quantify the difference in virulence, bacterial titer in the infected seedlings was analyzed (Fig. S1). Surprisingly, the number of bacteria in pregerminated seedlings was higher than in nonpregerminated seed from 1 through 4 days post inoculation (dpi). Furthermore, bacteria were also detected in the non-inoculated seed. This showed that the virulence would not simply be correlated to the population size of endogenous bacteria.
Isolation and identification of the contaminated endophyte. The shape and color of bacteria grown from pregerminated seeds varied ( Fig. 2A,B). In contrast, the bacteria from non-pregerminated seedlings were homogenous (Fig. 2C,D). Thus, it was speculated that the pregerminated seeds carried endophytic bacteria other than the inoculated pathogen. Supporting this, bacteria were also detected in seedlings treated with water only (Fig. S1). Furthermore, bacteria in the pregerminated seedlings were detected from 1 dpi, while bacteria from non-pregerminated seeds were detected only from 3 dpi. This strongly suggested that seedborne endophytic bacteria that multiplied during the pre-germination period, before pathogen treatment, control the pathogen growth in the plant.
Next, endophytic bacteria were isolated from each colony derived from endophytic tissue of 4 dpi plants (Fig. 2E). Three isolates of P. putida and one of S. maltophilia were identified by their 16S rRNA gene sequence   S4). As shown in Fig. 2F-I, the three P. putida showed different hues and morphologies, and thus were named RSB1, RSB10, and RSB15. The S. maltophilia was named RSB2.

Biocontrol effects of the isolated bacteria.
To test the hypothesis that bacteria which multiplied during germination attenuated the virulence of B. glumae, isolated RSBs and B. glumae were co-inoculated on nonpregerminated seeds (Fig. 3). The virulence was indexed by measuring root and leaf length of infected seedlings. Each P. putida RSB treatment allowed root and leaf elongation in the presence of pathogen. In contrast, RSB2 did not protect against B. glumae. Thus, P. putida RSBs have the ability to suppress the pathogenicity caused by B. glumae. Next, a competitive inhibition assay was done with each RSB paired against different B. glumae strains. Strong inhibition zone (halo) was observed in the case of 301682 (base)-RSB1 (spot), 301169 (base)-RSB2 (spot), and mild halo was observed in the case of 301169 (base)-RSB1 (spot), as shown Fig. 4. This strongly suggested that RSB1 was competitive against the pathogen. Notably, RSB1 colonies multiplied to almost 20 mm diameter on the B. glumae 301169 plate, whereas size of the colonies on 301682 was under 10 mm diameter, even at 7 days after dropping bacterial suspension on the plate (Figs. S5, S6). This also indicated that RSB1 multiplied faster than RSB10 and RSB15. RSB2 was competitive against the weak pathogen but less so against the strong pathogen. A swarming motility test (Fig. 5A) revealed high mobility of P. putida, especially RSB1 and RSB10, while RSB2 had no mobility. Biofilm formation of P. putida RSBs was almost two times that of B. glumae, whereas RSB2 produced approximately 7 times more biofilm than B. glumae (Fig. 5B). The effects of RSBs on plant growth were tested (Fig. S7), demonstrating that RSB15 slightly but significantly promoted root growth.
Cooperative effects of the endophyte. Based on these contrasting characteristics of RSBs, it was speculated that both P. putida and S. maltophilia applications might strengthen the ability of P. putida to protect. Therefore, a combined application of each of the P. putida RSBs as well as the S. maltophilia RSB2 was tested. As shown in Fig. 6, Figs. S8, S9, the P. putida RSBs and RSB2 coordinately protect rice seedlings from B. glumae infection based on leaf length and disease index analysis. Notably, the combined application of RSB2 and RSB15 resulted in 100% germination (Fig. 6, Figs. S8, S9). To clarify the effects of high biofilm formation activity of RSB2 on the synergistic effects, biofilm formation mutants were screened. From approximately 1,000 UV random mutants, 20 mutants were selected by a first screen in microplates. Following a second screening in test tubes, two enhanced (#UV-17, 28) and two reduced (#UV-45, 60) biofilm formation mutants were selected (Fig. S10). The virulence of B. glumae supplemented with RSB2 and RSB15 mutants was tested (Fig. S11). It was revealed that one of enhanced biofilm formation mutants, #UV-17 showed the highest biofilm formation and slightly accelerated the protection effects. Furthermore, two of reduced biofilm mutants prevented the effects. The other enhanced mutant, #UV-28, did not show any significant change, and biofilm formation seemed insufficient to enhance protection. Overall, the biofilm formation activity of RSB2 was likely to be one the factors coordinating protection effects.  To determine the distribution of endophytes and the pathogen in developing seedlings, culturable bacteria numbers of several parts of the seedling (leaf, seed and root) were counted over time distinguished based on the time each colony appeared and the colony morphology (Fig. S12). Putative B. glumae colonies which appeared at 66 h with yellow pigment were detected mainly in the leaf parts but also both in seeds and roots. www.nature.com/scientificreports/ By ITS analysis, using exactly the same DNA that was used for 16S rRNA gene sequencing (Fig. S13), the drastic change seen in 16S rRNA gene abundances was not observed, largely because Simplicillium accounted for approximately 90% of the out relative abundance.

Discussion
Seeds can act as vectors for transgenerational transmission of endophytes 20 . In this study, rice seed endophytes unexpectedly multiplied during the pregermination period (germination before pathogen inoculation), a normal period in rice cultivation. Interestingly, pregermination fortifies the rice plant against B. glumae infection. RSBs were eventually isolated as endogenous microbes that grew in the seed following artificial inoculation of the pathogen. These results strongly suggested that rice seed endophytes might control disease severity. In particular, the seed associated pathogen B. glumae seemed to be greatly influenced by endophytes. Furthermore, treatment with the endophytes greatly influenced other members of the endogenous microbiota. To my knowledge, there have been no reports that endophytes control B. glumae virulence. Four potential biocontrol agents for the plant endophytic pathogen B. glumae were isolated and identified. Several questions remain. Are these bacteria common in rice plants? In which part of the rice seedlings are they distributed? Without comparing whole genome sequences, it is hard to determine to what extent these bacteria are similar to previously discovered P. putida and S. maltophilia detected in rice seeds or seedlings. P. putida was found in the rhizosphere of rice seedlings cultivated in neutral-pH soil in the Netherland, while it was found in shoots of rice seedlings cultivated in acidic soil 21 . In the same experiment, S. maltophilia was detected in rhizosphere, roots, and shoots of rice seedlings cultivated in both soil types 21 . In maturing rice seeds on a rice plant cultivated in a paddy field in Japan, only S. maltophilia was found at the seed surface at the early maturing stage 22 . In contrast, during analysis of endophytes of rice seedlings by the same group, using the same rice cultivar, neither P. putida nor S. maltophilia was detected 15 . These results indicated that flora of rice might be dependent on habitat, environment, and cultivar. The seed Nipponbare used in this study was collected at a paddy field in Tsukuba, Japan in 2014 and preserved in a refrigerator. Thus, remarkably, the endophytes may have been established in the seeds during the cultivation process and remained dormant for five years in the seeds! In this study, seedlings collected for experiments had not been surface sterilized, so RSBs possibly existed on the plant surface or rhizosphere. However, since it was demonstrated that the supplementation of RSBs suppressed the virulence (Fig. 3), it is highly likely those bacteria multiplied inside the plant. In fact, from the NGS analysis of the microbes in the seedlings, the genera Pseudomonas and Stenotrophomonas were detected abundantly from only the correspondingly RSB-treated seedlings (Fig. 7). Also, putative B. glumae were detected from the leaf by culture-dependent analysis (Fig. S12). This strongly suggested that the pathogen preferentially colonized the   23 . Further understanding of intra-plant localization may elucidate how the pathogen gains virulence despite coexisting with the endophytes in planta. Unexpectedly, the relative abundance of the Burkholderia OTUs did not change due to RSB15 treatment (Fig. 7). Furthermore, the relative abundance of Burkholderia OTUs was lower in the case of B. glumae and RSB2 treatment, which showed intense growth disruption. This indicated that the bacterial titer might not corelate with the virulence. However, the occupancy of the Burkholderia OTU relative to other OTUs reduced from 37 to 22%. Therefore, the disease symptoms of B. glumae could be due to the balance of endophytes. In contrast to the NGS analysis, putative B. glumae was abundant in the case of culture-dependent analysis (Fig. S12). This inconsistent results due to the different time points of  www.nature.com/scientificreports/ sample preparation that 4 dpi plant was used for NGS analysis while 10 dpi plant was used for culture-dependent analysis and also the limitation of NGS analysis that counting all cell included dead cell. In previous reports, various strains of P. putida have been isolated and shown to have biocontrol activity against not only bacteria, but also fungi or nematodes [24][25][26] . Along with these P. putida strains that showed antimicrobial activity 27,28 , RSB1 and RSB2 isolated in this report also showed antibacterial activity against the pathogen (Fig. 4). It is known that B. glumae secrets a toxin called toxoflavin that can inhibit the growth of bacteria in the genera Escherichia, Shigella, Micrococcus, and Bacillus 29,30 . Growth of RSB1, RSB10 and RSB2 on a plate embedded with the high toxoflavin secreting strain B. glumae 301682 was inhibited compared to growth on a plate embedded with the lower toxoflavin-secreting strain B. glumae 301169 (Figs. S5, S6).
Some P. putida strains also showed resistance to broad-spectrum antibiotics 31 . P. putida and S. maltophilia collected from ranch soil rhizospheres showed organophosphate-degrading features 32 . Metallo-β-lactamase, responsible for resistance to antibiotics, is coded in the genome of S. maltophilia 32 . RSBs were resistant to kanamycin, ampicillin and chloramphenicol, and possibly others (data not shown). Therefore, perhaps those strains were resistant to the multiple antibiotics because of the antibiotic mechanisms such as RND efflux pumps. Interestingly, on the chloramphenicol plate, no colonies were found for any bacteria other than the four strains identified in the study. In contrast, a variety of microbes, and 10 times more bacteria, were found on chloramphenicolunsupplemented plates (data not shown). This suggested that antibiotic resistant microbes were screened by the plate used for growth curves, and it fortuitously and coincidentally served as a biocontrol bacteria selection plate. Acidovorax and Paenibacillus, that were detected by NGS (Fig. 7) but not present on the chloramphenicol plate, possibly do not have chloramphenicol resistance, and therefore would be ignored in the culture-based experiment.
Each P. putida RSBs derived from infected rice seedlings showed different morphology and characteristics. The 16S rRNA gene sequence of RSB1 was distinct from that of RSB10 or RSB15 (Fig. S3) and also in phylogenetic analysis (Fig. S4). Interestingly, RSB1 showed highly-competitive activity against the more pathogenic strain, B. glumae 301682 (Fig. 4). Although the 16S rRNA gene sequences of RSB10 and RSB15 were highly similar (Fig. S3), motility was very different (Fig. 5). Motility has been shown to affect endophytic colonization and bacterial mobility within host plants 19 . However, since RSB15, which showed low motility, had protective effects against B. glumae infection, motility might be not required for the protection. RSB15 also promoted root elongation (Fig. S7), and thus it is plausible that the strain also secretes a plant hormone such as phytohormone indole-3-acetic acid (IAA).
To my knowledge, there are no reports showing cooperative effects of P. putida and S. maltophilia for the control of a plant pathogen. Interestingly, cooperative work was reported between S. maltophilia and P. aeruginosa during polymicrobial airway infections 33 . In that study, microscopic analysis of biofilms formed in vitro revealed that S. maltophilia formed well-integrated biofilms with P. aeruginosa. Because the low biofilm formation strain weakened the cooperative work with RSB15 (Fig. S11), a molecule that promotes biofilm formation might accelerate the cooperation. Since application of RSB2 alone with B. glumae showed no protective effects, biofilm formed by RSB2 at normal densities was clearly not sufficient. Through the production of biofilms, RSB2 strains could easily adhere to the surfaces and aid in RSB2′s transmission. For S. maltophilia in human settings, the biofilm can also promote pathogen antibiotic tolerance, rendering some of the therapeutic options ineffective and causing setbacks in the selection of an appropriate treatment. This is consistent with the hypothesis that acceleration of biofilm formation of RSB15 with the help of RSB2 might protect plants from toxoflavin secreted by B. glumae. Further studies visualizing RSBs on site within biofilms will clarify the point. Altering the order of application of RSBs, for example, by including pre-treatment with RSB2 prior to other RSBs or pathogen, might enhance effects and deserves further investigation. It was also reported that members of the S. maltophilia and B. cepacia species complexes are typically multidrug resistant and excellent biofilm producers 32 . These phenotypes are controlled by quorum sensing systems from the diffusible signal factor (DSF) family. Resolving how DSF affects the interaction of B. glumae and RSBs would be an important future direction. High biofilm formation could be essential for the synergistic role of RSB2, and could be useful to promote effects by other P. putida that have been described in other studies or that are commercially available microbial inocula.
Many strains of S. maltophilia were isolated from various plants such as potato 34 , cucumber 35 , wheat 36 , and poplar 37 , and some of them showed plant growth promoting and disease suppressive effects. Most of those strains were reported to produce the phytohormone IAA [34][35][36][37] . However, the RSB2 isolated in this study was not shown to have any plant growth promoting (Fig. S7) nor disease suppression effects (Fig. 3) in the experimental conditions tested. Other forms of inoculation, such as soil treatment 34 , might be valuable to test.
Great numbers of strains belonging to P. putida and S. maltophilia are highly versatile and adapted to diverse habitats such as soil, water, the plant rhizosphere and endosphere, and occasionally as opportunistic pathogens in immune compromised patients. Thus, it would be essential to evaluate the environmental and human health impact prior to adoption as biocontrol material in the field. Many S. maltophilia described that are responsible for a variety of infections in both humans and animals. By the comparative whole genome analysis of 375 unique S. maltophilia from various origins such as humans, other animals, and the environment, it was determined that bacteria originating from environment were distinct from those isolated from humans or other animals 38 . This indicates that RSB2 is unlikely to gain pathogenicity towards humans or other animals. Further genomic analysis of RSB2 will be needed to be confirm that it indeed belongs to the group of environmental strains that cannot infect humans and other animals.
Since RSBs multiply easily and rapidly in either liquid LB (data not shown) or NA solid medium (Fig. 2), those could likely be easily cultured on a commercial scale. Seed treatment before sowing might save cultivation costs compared to spraying bacteria on the plant or pretreating field soil. Furthermore, RSBs could also increase protection even on the most resistant rice variety which has been developed. Although the enhanced protection ability of RSBs (Fig. 6), those rice still showed disease phenotype (Figs. S8, S9) www.nature.com/scientificreports/ of the effective concentration or inoculation methods are necessary. In addition, improvement of diagnostic tools for detection at early stages of rice development, or even at the seed stage, would be helpful to diminish the incidence of B. glumae. Further molecular analysis of the interaction between the pathogen and endophytes, including RSBs, will lead to novel factors that can be useful to detect traces of B. glumae in seeds or in planta. Current global climate change may cause an increase in new or previously negligible diseases. Enhancement of agricultural production to feed rapidly an increasing population worldwide, paired with the development of sustainable agriculture, is essential. Characterization and application of endophytes to control plant pathogens is an effective and sustainable technique. Development of novel biocontrol agents derived from rice plants may aid in achieving these goals. Moreover, the development of novel technology for appropriate disease control using endophytes tailored to the actual disease, plant species, and environment could reduce current agricultural overuse of pesticides or fertilizer.

Materials and methods
Plant material and pathogenic organism. Rice 'Nipponbare' seeds harvested in the test field of NARO in Tsukuba city Japan and kept at 4 °C for more than two years were used. The pathogens B. glumae MAFF301682 (B. glumae 301682) and B. glumae MAFF301169 (B. glumae 301169) were provided by Genetic Resource center, NARO, Tsukuba. Spontaneously grown bacteria on Luria-Bertani (LB) medium supplemented with chloramphenicol were isolated and used for further inoculation as chloramphenicol resistant strains.
Bacterial inoculation to rice seed. Burkholderia glumae strains were grown at 30 °C on LB medium supplemented with 1% agar and 10 μg/ml chloramphenicol for 3 days at 28 ± 2 °C and collected with sterilized distilled water (SDW) for inoculation. Rice endophytic bacteria isolated in the experiment RSBs were grown at 30 °C on LB medium supplemented with 1% agar and 10 μg/ml chloramphenicol for 24 h at 28 ± 2 °C and collected with SDW for inoculation. Approximately 20-25 rice seeds sterilized by soaking in Antiformin (available chlorine 5%) (Wako, Tokyo, Japan) for 10 min. The seeds were then rinsed with SDW several times. For pregermination, the sterilized seeds were soaked in sterilized water at 30 °C for 2 days in the dark. The pre-germinated or sterilized non-germinated rice seeds were soaked in 10 ml suspension of bacteria(s) adjusted to the indicated OD 600 and held under a vacuum for 1 min. Following agitation for 10 min under normal air pressure, the inoculated seeds were dried on paper and sown in sterilized soil (Bonsol No. 2, Sumitomo Kagaku Kougyo, Japan). The inoculated seeds were incubated in a growth chamber at 28 °C under a 14-h photo period and 24 °C under 10-h dark period with 100% humidity. Disease rate was evaluated by the length of the leaf shelf and root and disease index at the indicated time after bacterial inoculation.
Bacterial growth experiment in rice plant. Three whole plant samples were collected and washed in SDW three times and smashed by micro-smash (Bio-rad) with two stainless steel beads (4.8 mm) in a 2 ml sample tube with 1 ml SDW. Serial dilutions were prepared from the smashed suspension, and 100 μl aliquots from each dilution was spread on chloramphenicol-containing medium and incubated for 3 days at 28 ± 2 °C. Morphologically distinct bacterial colonies were each picked and spread on a new LB medium for further purification. The purified isolates were preserved in 20% glycerol solution at − 80 °C.

Molecular characterization of bacterial isolates. Bacterial DNA was obtained using the DNeasy
Blood &Tissue Kit (QIAGEN). PCR for 16S rRNA gene amplification was performed by using the bacterialspecific primers, 27F (5′-AAG GAG GGG ATC CAG CCG CA-3′) and 1492R (5′-GTG CCA GCA GCC GCGG -3′). PCR amplifications were performed with KOD plus polymerase as described before 39 . The PCR product was purified using Wizard PCR Preps DNA Purification System (Promega, Madison, WI, USA). Purified doublestranded PCR fragments were directly sequenced with Big Dye Terminator Cycle sequencing kits (Applied Biosystems, Forster City, CA, USA). Biofilm formation. Biofilm formation was measured using the method previously reported 40 . Briefly, an overnight culture of bacteria (OD 600 = 0.4) was diluted 1:100 in fresh sterile LB broth and 3 ml aliquots were removed to 13 ml sterilized tube (polypropylene) and incubated without agitation for 48 h at 28 °C. Planktonic bacteria were discarded, and the biofilms, which formed on the tube wall, were washed three times with distilled water. Into each tube, 4 ml of 0.1% crystal violet (FUJIFILM Wako Pure Chemical Co., Osaka, Japan) were added, and the tubes were incubated for 15 min at 30 °C. Excess crystal violet was then discarded, and stained biofilms were washed three times with 6 ml of distilled water. Finally, 70% ethanol was added to the stained biofilms, and the OD at 590 nm was read to assess the strength of biofilm formation. For small-scale biofilm mutant screening, 10 μl overnight culture of each independent bacteria was transferred to fresh 90 μl LB media in 96-well microtiter plates and incubated without agitation for 48 h at 30 °C. Biofilm cells were stained with 0.1% crystal violet and washed, the stain remaining in the cells was solubilized with 70% ethanol, and the optical density at 590 nm was determined.
Swarming motility. Bacteria swarming was done following a previous report 41 with some modifications.
Briefly, 1 ml of an overnight culture of each strain was adjusted to OD 600 3.0, centrifuged and washed twice with SDW, and suspended in 100 μl SDW. To inoculate the plates, 5 μl bacteria suspension was spotted in the center. Plates consisted of modified M9 medium 41  www.nature.com/scientificreports/ Competitive inhibition test. One ml of an overnight culture of each strain was adjusted to OD 600 3.0, centrifuged and washed twice with SDW, and suspended in 100 μl SDW. For making a bacterial embedded plates, 800 μl of a "base" bacterial suspension was added to 20 ml of LB agar pre-cooled to 42 °C, gently mixed, and then poured into one square petri dish. Two μl of each "spot" suspensions were spotted on "base" bacteria embedded plates. Area of inhibition ling (halo) and colony size was measured 7 days after dropping.

Generation of mutants by UV-mutagenesis.
Five ml of overnight culture of RSB2 was collected and resuspended into SDW. The bacterial suspension was diluted to 10 6 CFU/ml and plated onto LB plates. After 90 min at 30 °C of preculture, UV (Germicidal f15t8/UVB) was irradiated in 1 min in the hood. Each colony that had emerged on the plate after 2 days at 30 °C was picked as a mutant candidate. The number of colonies on the UV irradiated plate was half of that of the non-irradiated plate. Around 1,000 mutant candidates were applied to the first step screening using the mini-scale biofilm formation test described above in a microtiter plate. By the first step of screening, 20 mutants were picked as biofilm formation mutants. Among these candidates, 2 of the mutants that enhanced biofilm formation and 2 of the mutants that suppressed biofilm formation selected for further experiments.

NGS analysis.
To assess the structure of the microbial community in the rice seedlings, 16S rDNA gene and ITS amplicon sequencing using MiSeq was performed. Four days after treating non-pregerminated seed with B. glumae (OD = 0.0004) and RSBs (OD = 0.04), twenty-five seedlings collected three times and mixed all of them and ground with liquid-nitrogen, and DNA was extracted for 3 sets (technical replications) of each ground sample using the DNeasy Plant Mini Kit (Qiagen, USA) according to the manufacturer's instructions. The extracted DNA was adjusted with distilled water to a concentration of 30 ng/μl in a total volume of 50 μl. The DNA concentration was fluorometrically determined using Qubit Assay Kits (Thermo Fisher Scientific Inc., Waltham, MA, USA) and a Nanophotometer (Implen GmbH, Munich, Germany). Sequence analysis of the rice samples using a MiSeq system (Illumina, Inc., San Diego, CA, USA) was performed by Fasmac Co., Ltd (Atsugi, Japan) as described before 42  . Analysis of the sequencing results included trimming of the primer region using Fastx toolkit, joining of the forward and reverse reads using FLASH, and quality filtering with sickle tool. The 97% identity OTU clustering and chimera filtering were performed using UCHIME (USEARCH package v8.0.1623) in QIIME, version 1.9.0. These data were then used to assign taxonomy against the Greengenes 13_8 database with a 97% similarity threshold using the UCLUST v1.2.22q in the assign taxonomy script of QIIME. Raw sequence counts were normalized by converting them to relative abundances through total sum scaling; for each sample, the sequence counts for each taxa was divided by the total number of sequence counts for that sample, and the result was converted to a percentage.
DNA deposition. Culture-independent MiSeq sequence reads of the 16S rRNA gene and ITS are available in the DDBJ sequence read archive (DRA): PRJDB10214.

Statistical analysis.
Statistical analyses were carried out by use of R software version 3.6.1 43 . Tukey's HSD test was done using library (multcomp 44 ) and command cld to show significance.