Cooperative interactions between seed-borne bacterial and air-borne fungal pathogens on rice

Bacterial-fungal interactions are widely found in distinct environments and contribute to ecosystem processes. Previous studies of these interactions have mostly been performed in soil, and only limited studies of aerial plant tissues have been conducted. Here we show that a seed-borne plant pathogenic bacterium, Burkholderia glumae (Bg), and an air-borne plant pathogenic fungus, Fusarium graminearum (Fg), interact to promote bacterial survival, bacterial and fungal dispersal, and disease progression on rice plants, despite the production of antifungal toxoflavin by Bg. We perform assays of toxoflavin sensitivity, RNA-seq analyses, lipid staining and measures of triacylglyceride content to show that triacylglycerides containing linolenic acid mediate resistance to reactive oxygen species that are generated in response to toxoflavin in Fg. As a result, Bg is able to physically attach to Fg to achieve rapid and expansive dispersal to enhance disease severity.

B acterial-fungal interactions (BFI) widely exist in distinct environments, such as on the human body, in food and in soil [1][2][3][4] . These interactions often have a physical or molecular basis 2,5 , and they can lead to numerous biological effects that vary from antagonism to cooperation 2,6-8 . Some of these effects involve metabolite exchange, signaling chemotaxis and physical communication, such as planktonic form, mixed biofilm and intrahyphal colonisation 1 . Symbiotic relationships constituted in particular niches can also result in the translocation of nutrients, resistance to environmental stress and pathogenic acquisition 6,7 .
Beneficial effects of BFI have previously been described for Burkholderia terrae and Lyophyllum sp. in soil 7 . For example, the presence of Lyophyllum sp. has been found to be essential for the colonisation and transportation of Burkholderia terrae in soil 9 , and the bacteria exhibited better survival in acidic soils with fungal exudates as nutrients 10,11 . Similarly, Burkholderia rhizoxinica lives within fungal hyphae and is crucial for fungal pathogenicity toward rice seedlings in soil 12,13 . To date, molecular and physical interactions between Burkholderia and fungi have been well studied in soil, while only limited studies of BFIs between bacteria and air-borne fungi have been conducted.
The seed-borne bacterial plant pathogen, Burkholderia glumae (Bg), is one of the causal agents of bacterial panicle blight in rice fields 14 . This bacterium produces toxoflavin, which is both a critical virulence factor and an antimicrobial that induces production of superoxides and hydrogen peroxide (H 2 O 2 ) 14,15 . Thus, Bg has the potential to monopolise rice grains by blocking the growth of saprophytic fungi and other pathogens 16,17 . However, we have observed that the air-borne fungal plant pathogen, Fusarium graminearum (Fg), is resistant to toxoflavin 18 .
Here we examined the potential for Bg and Fg to interact and co-exist, particularly regarding the mechanism by which Fg develops resistance to toxoflavin that is produced by Bg. This study was facilitated by the generation of toxoflavin-sensitive Fg mutant strains and experiments performed in planta. The data obtained demonstrate that the two pathogens cooperatively interact, and this provides Fg with an opportunity to produce more spores and toxins and it provides Bg with an opportunity to achieve aerial dispersal.

Results
Bg is often isolated with Fg from field-grown rice grains. Rice grains were collected from black-and glutinous-type rice crops at intervals of 7 days (d) from the flowering times. Both Fg and Bg were frequently observed on the collected grains ( Fig. 1a) and their coexistence frequencies were quantitated with a PCR-based assay that identified Fg and Bg as 420 bp and 138 bp products, respectively, (Fig. 1b). Sequencing of the PCR products was performed to confirm the detection of Bg and Fg (data not shown). Among the field-grown grains examined, 40% (range: 22-52%) carried Bg alone, 1% (range: 0-4%) carried Fg alone, 24% (range: 2-36%) carried both pathogens, and 35% (range: 8-56%) were pathogen-free (Fig. 1c). Thus, Bg did not monopolise the rice grains examined and Fg was able to cohabitate with Bg on the rice grains.
Isolation of Fg mutant strains sensitive to toxoflavin. To understand the mechanism by which Fg is resistant to toxoflavin, a transcription factor deletion mutant library of the wild-type (WT) Fg strain, GZ03639, was screened. Three toxoflavinsensitive mutants, ΔGzZC190, ΔGzbZIP005 and ΔGzC2H008 lacking FGSG_07589, FGSG_02939 and FGSG_01106 genes, respectively, were identified. ΔGzZC190 exhibited the highest sensitivity to toxoflavin, followed by ΔGzbZIP005 and ΔGzC2H008 (Fig. 2a, Supplementary Fig. 1). All three mutants and the WT strain were evaluated for 17 phenotypes, including vegetative growth, sexual development, conidiation, virulence, toxin production and stress responses, among others. These phenotypes remained unchanged in the ΔGzZC190 and ΔGzbZIP005 strains, except for sensitivity to toxoflavin. In contrast, ΔGzC2H008 exhibited pleiotropic phenotypic changes, including abnormal sexual development, loss of virulence, and reduced toxin production, consistent with previous results (http:// ftfd.snu.ac.kr/FgTFPD) 19 .
The TF genes, FGSG_07589, FGSG_02939, and FGSG_01106, contain Zn(II) 2 Cys 6 , bZIP, and C 2 H 2 zinc finger DNA-binding domains, respectively. FGSG_07589 and FGSG_02939 are less conserved compared to FGSG_01106 among ascomycete fungi, and FGSG_07589 is more variable than FGSG_02939 in the Fusarium genus ( Supplementary Fig. 2). FGSG_01106 is a homologue of arsenite resistance protein 2 (ARS2) and is required for a DNA damage response 20 , while the functions of the other two genes have not been studied. We mainly focused on FGSG_07589 because it is more unique to Fg and its deletion resulted in greater sensitivity to toxoflavin.
Vegetative growth of ΔGzZC190 and the WT strain were directly compared. The growth of ΔGzZC190 was normal in the absence of toxoflavin and was dramatically reduced in toxoflavinsupplemented media (Fig. 2a). Under the same conditions, growth of the WT strain was diminished, yet still robust (Fig. 2a).
To confirm the specificity of FGSG_07589 to a toxoflavin response, FGSG_07589 was amplified from the WT strain and reintroduced into ΔGzZC190 to generate a GzZC190-complemented strain (GzZC190c) (Supplementary Fig. 3). Toxoflavin sensitivity was subsequently restored in GzZC190c (Fig. 2). Furthermore, in minimal medium (MM) containing toxoflavin, ΔGzZC190 exhibited delayed spore germination, while normal spore germination was observed in the WT and GzZC190c strains ( Supplementary Fig. 4).
Transcriptome analyses of Fg strains in response to toxoflavin. RNA-seq analyses were performed to profile the gene expression patterns of WT, ΔGzZC190, ΔGzbZIP005, and ΔGzC2H008 strains to toxoflavin. First, we monitored the genes of FGSG_07589, FGSG_02939, and FGSG_01106 in the WT strain. Only FGSG_07589 was found to be highly induced by exposure to toxoflavin (Supplementary Table 1). With a ≥twofold change in expression selected as the criteria, 1062 genes were upregulated and 1216 genes were downregulated in the WT strain compared to ΔGzZC190 (P < 0.05). According to the same criteria, 579 genes were upregulated and 881 genes were downregulated in the WT strain compared to ΔGzbZIPH005; and 1064 genes were upregulated and 1053 genes were downregulated in the WT strain compared to ΔGzC2H008. For the differentially expressed genes (DEGs), a Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway-enrichment was performed. This analysis (with P < 0.05 and ≥5 gene count numbers) revealed that~22% of pathways in the WT strain were upregulated compared to ΔGzZC190 (Supplementary Table 2), while~75% of pathways in the WT strain were upregulated compared to the other two mutants (Supplementary Tables 3 and 4). These results suggest that a greater number of pathways were induced in ΔGzZC190 in response to toxoflavin compared to the WT strain. In particular, glycolysis/ gluconeogenesis pathways were upregulated, while alanine, aspartate and glutamate metabolism pathways were downregulated, in the WT strain compared to all three mutants. In contrast, diverse metabolic pathways, including amino acid, carbon, and lipid metabolism pathways, were very dynamic in all three mutants. Furthermore, among the pathways related to protection from reactive oxygen species, which is the main function of toxoflavin, a few lipid metabolism pathways were highly enriched. For example, fatty acid degradation was downregulated in the WT strain compared to ΔGzZC190, yet was upregulated in the WT strain compared to ΔGzC2H008. Meanwhile, upregulation of fatty acid metabolism was observed in the WT strain compared to the GzbZIP005 and ΔGzC2H008 strains. In depth analyses of the WT strain compared to ΔGzZC190 further showed that the greatest sensitivity to toxoflavin without any changes in tested phenotypes occurred in relation to an upregulation of genes involved in the fatty acid biosynthesis pathway ( Supplementary Fig. 5).
Lipid droplets are involved in toxoflavin resistance. Based on the transcriptome analysis data described above, and previous observations that lipid droplets (LDs) play an important role in tolerance to phytotoxins, fungicides and oxidative stress 21,22 , we focused on fatty acid biosynthesis as a possible mechanism by which Fg develops resistance to toxoflavin. Therefore, lipid staining was initially performed of the WT and ΔGzZC190 fungal cells. A dramatic increase in LD size was observed in the WT strain after toxoflavin treatment, while the LD size in the ΔGzZC190 strain appeared unaffected by toxoflavin treatment (Fig. 2b). Previously, oleate was shown to increase LD formation in Drosophila S2 cells 23 . Therefore, we supplemented the MM containing toxoflavin with unsaturated fatty acids including oleic acid, linoleic acid, or linolenic acid and repeated the experiment. Large LDs were observed in all treated models, even in the hyphae of ΔGzZC190, and the vegetative growth of ΔGzZC190 was restored to~70% that of the WT strain (Fig. 2c Fig. 6). As controls, LDs stained in Colletotrichum gloeosporioides (Cg) and Magnaporthe oryzae (Mo) hyphae were smaller than those of Fg (Fig. 2b). Moreover, when either oleic acid, linoleic acid or linolenic acid was added to Cg and Mo and the experiments were repeated, the toxoflavin sensitivity of Cg and Mo were unaffected (Fig. 2c, Supplementary Fig. 6). Furthermore, we also determined whether oleic acid, linoleic acid or linolenic acid enhance the resistance of WT to H 2 O 2 , since toxoflavin damages eukarytotic cells through the production of superoxide and H 2 O 2 24,25 . Supplementation restored vegetative growth of the WT, ΔGzZC190 and GzZC190c strains that was completely inhibited in the presence of 5 mM H 2 O 2 (Fig. 2c, Supplementary  Fig. 6).
Triacylglycerides mediate toxoflavin resistance. Due to the observed increase in LD size in the WT strain after toxoflavin treatment, triacylglyceride (TAG) content was measured. The goal was to obtain further insight into the relationship between increased LD size and toxoflavin-mediated ROS stress. LD has been shown to reduce cellular ROS stress via protection of unsaturated fatty acids 22,26,27 . Consequently, we compared the amounts of TAGs in the WT and ΔGzZC190 strains with and without toxoflavin treatment. The total amounts of TAGs were higher in the ΔGzZC190 strain than in the WT strain before toxoflavin treatment (ANOVA; P < 0.05). After toxoflavin treatment, there were no statistically significant differences in the total TAGs between the WT and ΔGzZC190 strains. A similar profile was observed for the amounts of TAGs containing oleic acid and the TAGs containing linoleic acid (Fig. 3a). Interestingly, the latter levels were reduced in the toxoflavin-treated ΔGzZC190 strain compared with the toxoflavin-treated WT strain (P < 0.05) and the untreated ΔGzZC190 (P < 0.01) strain (Fig. 3a, Supplementary Fig. 7). Correspondingly, genes involved in the biosynthesis of linolenic acid were found to be upregulated in the WT strain compared to the ΔGzZC190 strain after toxoflavin treatment, whereas most genes were downregulated in the toxoflavin-treated ΔGzZC190 strain compared to the untreated ΔGzZC190 strain (Fig. 3b). These results are consistent with the observed changes in the TAGs containing linolenic acid in response to toxoflavin.
Superoxide dismutase 1 mediates toxoflavin resistance. The vegetative growth of mutant strains carrying deletions of FgAP, FgSKN7, FgATF1, or FgELP1 has been associated with oxidative stress [28][29][30][31] . Here, the growth of these mutant strains was not suppressed on toxoflavin-supplemented media ( Supplementary  Fig. 8a). However, vegetative growth of the superoxide dismutase 1 mutant (Δsod1) 32 was inhibited similar to that of the ΔGzZC190 strain ( Supplementary Fig. 8a). Interestingly, catalase supplementation of the ΔGzZC190 strain did not restore vegetative growth that was inhibited by toxoflavin, yet it did restore vegetative growth that was inhibited by H 2 O 2 ( Supplementary  Fig. 8b). In ΔGzZC190, transcript levels of all five sod genes were upregulated. However, following toxoflavin treatment, all five genes were downregulated. The opposite profile was obtained in the WT strain, with the transcripts of all five sod genes being upregulated after toxoflavin treatment ( Supplementary Fig. 9a). Transcript levels of genes encoding catalases were also upregulated in the WT strain after toxoflavin treatment, yet a distinct pattern was not observed in the ΔGzZC190 strain ( Supplementary  Fig. 9b).
Bg enhances Fg spore production. To address the ecological significance of the interaction between Fg and Bg, we conducted several experiments to determine whether chemical or physical interactions exist between these two pathogens. Compared to the cultivation of Fg alone, co-cultivation resulted in~10-fold greater fungal spore production. Addition of a cell-free supernatant of a Bg (BGR1) culture to Fg (GZ03639) further increased spore production (Fig. 4a). In contrast, filtrates of Pseudomonas syringae (DC3000) and Bukholderia pyrrocinia (CH67) cultures did not enhance spore production (Fig. 4a). When toxoflavin alone or H 2 O 2 were added to GZ03639, spore production also increased, albeit less than that of the bacterial culture filtrate (Fig. 4b,  Supplementary Fig. 9a). There were three Fg field strains which also exhibited higher spore production with toxoflavin (Supplementary Fig. 10b). Gene expression profiling of the WT strain with toxoflavin correspondingly resulted in higher expression of 24 genes related to sporulation (Supplementary Table 5).
Toxoflavin increases production of deoxynivalenol by Fg. A plant's response to fungal infection includes an oxidative burst, and this in turn can induce the production of mycotoxins in fungi 33 . For example, ROS have been shown to trigger aflatoxin production in Aspergillus parasiticus 34 , and in Fg, production of a α-linolenic acid Correspondingly, the genes, Tri5 and Tri6, which encode proteins with functions important to DON production in Fg, were found to be upregulated in the presence of 1 mM H 2 O 2 or 80 mg l −1 toxoflavin in our studies (Fig. 4c). These conditions also resulted in greater production of DON (Fig. 4d).
Co-cultivation results in a physical attachment between Bg and Fg. To understand the benefit, if any, of greater fungal spore production to the bacterium, Bg, physical interactions between the pathogens were examined. In a nested plate system designed to monitor physical attachment 37 , both BGR1 and DC3000 grew well on a central plate embedded in media inoculated with Fg. However, only BGR1 was found outside of the central plate, and it was physically attached to fungal hyphae and spores (Fig. 5a, b). While the signalling molecules involved remain to be isolated, these results demonstrate that BGR1 is able to move towards GZ03639. A similar observation was made when a greater number of BGR1 cells were found to have moved into a capillary tube containing a culture filtrate of GZ03639, compared with other capillaries that contained pure medium or a culture filtrate of another rice fungal pathogen, M. oryzae (Fig. 5c). Furthermore, the Fg culture filtrates did not attract the DC3000 cells (Fig. 5c).

Co-inoculation increases disease severity and DON production.
A single inoculation of BGR1 or GZ03639 on rice heads in the late-flowering stage resulted in disease severities of 15 and 33%, respectively. In contrast, when both pathogens were co-inoculated, the disease severity was 82% and disease progression to the stem was observed (Fig. 6a, Supplementary Fig. 11). In addition, both Bg and Fg were detected in symptomatic grains of 34 out of 36 (94%) untreated rice plants that were adjacent to the coinoculated rice plants. DON production was also~twofold higher in the rice plants that were inoculated with both Fg and Bg compared with the rice plants that received a single-Fg inoculation (Fig. 6b). The Fg-only rice plants also had many hyphae, while spores were rarely observed. In contrast, both hyphae and spores were observed on the rice plants inoculated with both Fg and Bg, with Bg cells frequently present on the hyphae and spores (Fig. 6c).
Physical attachment of Bg to Fg protects Bg from UV stress. In addition to dispersion benefits, the Fg spores appear to provide protection for Bg from ultraviolet (UV) light-induced damage. When BGR1 cells and GZ03639 spores were exposed to UV germicidal lamps for 1 and 5 min, respectively, both pathogens were completely killed. In contrast, when BGR1 cells were cocultivated with GZ03639 spores and then were exposed to UV light for 5 min, colony-forming BGR1 was recovered, although the GZ03639 spores were completely unviable ( Supplementary  Fig. 12, Supplementary Table 6).

Discussion
Our efforts to identify a mechanism by which Fg is able to tolerate the production of toxoflavin by Bg on rice heads started with the characterisation of three Fg mutants carrying deleted TFs that exhibited sensitivity to toxoflavin, ΔGzZC190, ΔGzbZIP005 and ΔGzC2H008. To identify possible links among their genes in relation to toxoflavin resistance, transcriptome data were obtained for each, and for the parental wild-type strain, in the presence and absence of toxoflavin. Interestingly, many of the DEGs that were identified were enriched in pathways involving fatty acid degradation, fatty acid biosynthesis and fatty acid metabolism. Toxoflavin treatment was also found to induce the production of larger LDs in Fg cells. LDs have previously been shown to reduce cellular ROS stress via protection of unsaturated fatty acids 22,26,27 . In addition, LDs serve as the principal reservoirs for storing cellular energy and contribute to the repair of  23 , during Drosophila development LDs limit the levels of ROS 22 , and in fungi LDs trap phototoxins to serve as a resistance mechanism 21 . To further investigate the molecular details of the mechanism mediating toxoflavin resistance in Fg, the TAG content of the WT and ΔGzZC190 strains with and without toxoflavin treatment was investigated with LC-MS/MS. Toxoflavin treatment of ΔGzZC190 was found to reduce the levels of TAGs containing linolenic acid, and these results were consistent with the observation that genes associated with fatty acids containing α-linolenic acid were upregulated in response to toxoflavin (Fig. 3). Taken together, these results suggest that TAGs containing linolenic acid are protected within LDs, and the genes involved in the biosynthesis of linolenic acid are induced in response to ROS generated from toxoflavin. Further support for this model was provided by the treatment of ΔGzZC190 with linolenic acid (or linoleic acid) and the subsequent restoration of toxoflavin sensitivity. Thus, TAGs containing linolenic acid may play a role in resistance to ROS stress that is generated in response to toxoflavin. Furthermore, the previous observations that vegetative growth and sporulation in Aspergillus flavus were enhanced after treatment with autoxidated linolenic acid 38 suggests that linolenic acid may have additional roles in Fg condia formation after toxoflavin treatment. When vegetative growth of the superoxide dismutase 1 mutant (Δsod1) 32 was found to be similar to that of the ΔGzZC190 strain, all five sod genes of Fg were investigated and all of them were found to be downregulated following toxoflavin treatment (Supplementary Fig. 9). Further examination of these results in the present study suggests that superoxides that are produced following toxoflavin treatment may be converted into H 2 O 2 , which could be subsequently converted into water by catalases in the WT Fg strain. These observations, in combination with the previous observation that chlamydospore-like structures form and accumulate lipids that are associated with resistance to oxidative stress in Fg 39 , suggest that LDs containing linolenic acids have the potential to rapidly scavenge superoxides that are generated by toxoflavin to provide toxoflavin resistance.
ROS have been shown to play an important role in the cell differentiation of fungi and other eukaryotes 40,41 . In addition, transient increases in ROS levels have been shown to induce a transition between vegetative growth and asexual reproduction 42 . A model in which increased spore production by Fg is related to ROS production triggered by toxoflavin is supported by the present results, since both toxoflavin and H 2 O 2 were shown to cause a statistically significant increase in Fg spore production ( Supplementary Fig. 10), and 24 sporulation-related genes were highly upregulated by toxoflavin (Supplementary Table 5), including FGSG_01877 and FGSG_01915 which were identified as regulators for sporulation in Fg. Accordingly, when these two genes were deleted, spore production was decreased 43,44 . Conversely, when FGSG_01877 was overexpressed, a substantial increase in spore production was observed in Fg 43 . Thus, the present results indicate that toxoflavin enhances the expression of two regulators of sporulation (FGSG_01877 and FGSG_01915) and this increases spore production in Fg.
Bg can be present prior to the arrival of Fg spores on rice because Bg is seed-borne and can move to the rice panicle as the plant grows. We propose that the presence of Bg on rice is beneficial to Fg for the following reasons. First, Bg enhances production of DON by Fg, which aids in disease progression. Previously, ROS were found to enhance DON production in Fg 35,36 , and this observation is consistent with the observation that DON production increased following toxoflavin treatment. While DON production is not essential for the initial infection of Fg in wheat plants (yet it helps systemic colonisation in wheat spikelets 45 ), co-inoculation of Fg and Bg resulted in greater disease severity and higher DON production in infected rice heads (Fig. 6, Supplementary Fig. 11). Second, Bg can prevent the growth of other fungal competitors of Fg because other saprophytic and pathogenic fungi are sensitive to toxoflavin. Fungi must compete with various microorganisms to occupy an area, and on a plant surface there can be a diverse set of microorganisms. Toxoflavin exhibits a broad range of antibiotic properties that can affect eukaryotes 46 . Therefore, the presence of Bg can facilitate the occupation of rice heads by Fg since other microorganisms are sensitive to the production of toxoflavin. An interesting observation from the present study was that the physical attachment of Bg cells to Fg spores protected the Bg cells from UV light-induced damage. Specifically, the present results demonstrate that the attachment of bacteria to conidia resulted in a marked increase in resistance to UV stress compared to unattached bacteria. Physical attachment to Fg also facilitated aerial movement of the bacterium for its dispersal. Generally, the migration of bacteria is very limited and often requires water for dispersal. In contrast, the attachment of bacteria to fungal hyphae provides a more rapid and expansive route of dispersal. Correspondingly, recent reports have indicated that fungi are essential for the colonisation and dispersal of Burkholderia species 9,37,47 . Thus, cooperative interactions between Bg and Fg result in the dispersal of Bg, as well as protection of Bg from UV stress.
The identification of this BFI that accounts for the observed exacerbation of disease frequency and severity in rice fields provides an opportunity to target this association and attenuate its virulence. For example, ROS scavenging or lipid depletion could potentially interfere with this cooperative relationship to reduce disease severity. Moreover, further examination of the evolutionary history of the genes involved may lead to the identification of other systems that may have evolved cooperative interactions.

Methods
Strains and culture conditions. Fusarium graminearum (Fg) wild-type (WT) strain, GZ03639, and transcription factor deletion mutants (TFs) 19 derived from this WT strain were used in this study. TF complementation construct was generated by the double-joint (DJ) PCR method 48 . The primers used for genetic manipulation were listed in the Supplementary Table 7. Burkholderia pyrrocinia (CH67) and Pseudomonas syringae (DC3000) were used as negative controls. Fg field strains, SCK04, GWR12R5 and GWS16-4-7, were obtained from Soonchunhyang University (Cheonan, Korea), while other fungal and bacterial species were obtained from the Centre for Fungal Genetic Resources (Seoul National University, Seoul, Korea). The bacterial strains were cultured in lysogeny broth (LB) medium at 30°C with shaking at 200 r.p.m. for 24 h; the fungal strains were maintained on potato dextrose agar (PDA) at 25°C; and spores were induced in carboxymethyl cellulose (CMC) 49 . Spore germination was tested in minimal medium (MM). All strains used in this study were stored in 15% (v/v) glycerol at −70°C.
Detection of Bg and Fg from diseased rice grains. Rice grains with head blight symptoms were collected from a southern province in Korea (Gyeongnam) during August 2015 at intervals of 7 d beginning with the flowering times of black-and glutinous-type rice. Genomic DNA was extracted from 50 rice grains collected from each cultivar, and the presence of Bg or Fg was tested by conventional PCR with F. graminearum-specific primers (Fg16F; 5′-CTCCGGATATGTTGCGTC AA-3′, Fg16R; 5′-GGTAGGTATCCGACATGGCAA-3′) 50 and a B. glumae primer set designed for a rhs family gene (BG1F; 5′-CCGCGCTGTTCATGAGGGATAA-3′, BG1R; 5′-CGGGCGGAACGACGGTAAGT-3′) 51 . The predicted PCR product lengths were 420 and 138 bp, respectively. The PCR products were purified from gels for sequencing (Macrogen, Seoul, Korea). Relative transcript abundance was measured in reads per kilobase of exons per million mapped reads 52 . Differentially expressed genes (DEGs) were identified using the DEGseq software package in the R statistical environment 53 . In this analysis, the MA plot-based method was employed, and the filtering threshold included P < 0.05 and an absolute value ≥twofold change. All DEGs were annotated according to Gene Ontology (GO; http://geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.jp/kegg/) databases to establish functional categories. Enrichment of the functional categories was performed by using a hypergeometric distribution 54 . P-values <0.01 and <0.05 were considered to indicate significance in the GO and KEGG analyses, respectively. The network system of the fatty acid biosynthesis pathway was constructed with Cytoscape software (www.cytoscape.org/).
Lipid body staining and oleate test. Mycelia were incubated in MM containing 80 mg l −1 toxoflavin and 1% sodium oleate, 1% linoleic acid, or 1% linolenic (Sigma-Aldrich) at 25°C with shaking at 200 r.p.m. for 24 h. Lipid droplets in fungal cells were stained with a Nile Red solution consisting of 50 mM Tris/maleate buffer (pH 7.5), 20 mg ml −1 polyvinylpyrrolidone, and 2.5 μg ml −1 Nile Red Oxazone (Sigma-Aldrich). The samples were incubated for 15 min at room temperature and were washed two times with phosphate-buffered saline. Fluorescence emitted by the lipid droplets was observed with a Carl Zeiss confocal microscope (excitation at 490 nm, LSM 510). To determine whether vegetative growth that was affected by toxoflavin could be complemented with exogenous sodium oleate, 1% sodium oleate was added into MM agar supplemented with 80 mg l −1 toxoflavin or 5 mM H 2 O 2 . Each fungal strain was incubated at 25°C for 4 d. Triacylglyceride analysis. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analyses were performed with a triple quadrupole mass spectrometer (LC-MS 8040; Shimadzu, Kyoto, Japan) coupled to a Nexera2 LC system (Shimadzu) using a Kinetex C18 column (100 × 2.1 mm, 2.6 μm; Phenomenex, Torrance, CA, USA). Mobile phase A was 10 mM ammonium acetate in water: methanol (1:9, v/v) and mobile phase B was 10 mM ammonium acetate in methanol:isopropanol (1:1, v/v). To achieve the best separation of TAG, a gradient elution was conducted as follows: 30% B (0 min), 95% B (15-20 min), and 30% B (20-25 min). The flow rate was 200 μl min −1 . Quantitation was performed with selected reaction monitoring of the (M + NH 4 + ) ion and related product ions for each TAG 57,58 .
Quantitative real-time PCR (qRT-PCR). To compare the transcript levels of genes, WT and ΔGzZC190 were grown in 50 ml of complete medium (CM) at 25°C with shaking at 200 r.p.m. for 72 h. The filtered mycelia were incubated in 50 ml of MM containing toxoflavin (80 mg l −1 ) for 24 h at 200 r.p.m. and total RNA was extracted using NucleoSpin RNA Plant Kit (MACHEREY-NAGEL GmbH & Co. KG, Düren, Germany). The amounts of transcripts were quantified by qRT-PCR with specific primers (Supplementary Fig. 7). Quantification values were automatically determined using Bio-Rad CFX Manager, version 3.1, and the threshold cycle (Ct) values were determined. The experiment was repeated twice, with three replicates for each repeat, and final Ct values are presented as an average 59 . To determine whether vegetative growth affected by toxoflavin can be complemented with exogenous catalase (Sigma-Aldrich), 1000 unit was added into MM agar supplemented with 80 mg l −1 toxoflavin or 5 mM H 2 O 2 , and each fungal strain was incubated at 25°C for 4 d.
Induction of Fg spore production. Asexual spores (10 5 ml −1 ) of Fg were inoculated in 50 ml of potato dextrose broth (PDB) and incubated at 25°C on a rotary shaker at 200 r.p.m. After 24 h, mycelia were collected by filtration through Miracloth (Calbiochem) and were washed twice with sterile distilled water. Freshly collected mycelia were then inoculated in 20 ml CMC containing 10, 30, 50 or 80 mg l −1 toxoflavin and incubated at 25°C with shaking at 200 r.p.m. After 24 h, the numbers of spores produced were counted with a haemocytometer.
Transcription of Tri genes and trichothecene production. To test the effect of toxoflavin on trichothecene biosynthesis, mRNA levels of the trichothecene biosynthesis genes, Tri5 (FGSG_03537, trichodiene synthase) and Tri6 (FGSG_03536, transcription factor for trichothecene synthesis), as well as trichothecene production, were analysed. Briefly, each strain was incubated in 50 ml GYEP (10 g of glucose, 1 g of yeast extract, and 1 g of peptone in 1 l) 35 supplemented with or without toxoflavin (80 mg l −1 ) at 25°C on a rotary shaker (200 r.p.m.) for 5 d. Total RNA was purified from the mycelia using a NucleoSpin RNA Plant Kit (MACHEREY-NAGEL GmbH & Co. KG), according to the manufacturer's protocol. The amounts of Tri5 and Tri6 transcripts were quantified by qRT-PCR with specific primers for each gene 60 . Concentrations of DON in the culture filtrates were quantified with an enzyme-linked immunosorbent assay kit (CUSABIO, College Park, MD, USA), according to the manufacturer's instructions 61 .
Attachment assay of bacteria on fungal hyphae and spores. Dispersal and attachment of bacteria on fungal hyphae and spores was monitored 37 with a slight modification. Briefly, sterile tissue culture plates (35 × 10 mm) were used as a physical barrier for bacterial cells during the pouring of LB or MM agar into Petri dishes (90 × 15 mm) up to 5 mm below the edge of the smaller plate. Selected fungi and Burkholderia strains were inoculated onto the inner side of the smaller plate, so that only bacterial cells which could attach and/or disperse would be detected on the outside of the smaller plate. Different regions of the cultures were observed under a field emission scanning electron microscope (JSM-6700F) after preparation 62 .
Assay of Bg attraction to Fg. An attraction assay was performed according to the method used for P. syringae 63 , with a slight modification. Briefly, bacteria cells (OD 600 of 0.1) were grown in LB at 30°C and then were diluted 10 2 -, 10 3 -and 10 4fold. An aliquot (100 μl) of each diluent was then added to the wells of a 96-well plate. Fungal spores (10 5 ml −1 ) were incubated in PDB for 3 d, and the culture was filtered through a 0.2-μm syringe filter. Capillary tubes filled with the filtrate were placed into the 96-well plate containing serial dilutions of the bacterial cells, and the plate was subsequently incubated at 30°C for 3 h. Fungal filtrates in the capillary tubes were allowed to spread on LB agar plates at 30°C. After 24 h, the numbers of colonies that grew were counted and recorded. The experiment was repeated twice with three replicates per sample.
UV light stress test. Bacterial cells, fungal spores and fungal spores with bacterial cells were exposed to a germicidal UV light (40 W; G40T10) for 0, 1, 3, and 5 min, respectively. To induce attachment of the bacterial cells to the spores, Fg and Bg were co-incubated on LB agar at 30°C for 10 d. The spores were then collected by filtration through Miracloth and diluted with distilled water to 10 2 ml −1 . Each diluent (100 μl) was subsequently spread on an LB agar plate and incubated at 30°C . After 24 h, the plates were exposed to UV light as indicated above and the numbers of bacterial colonies present were counted and recorded.
Disease severity following co-infection of Bg and Fg. Briefly, rice (Oryza sativa L. cv. Dongjin) heads were dipped into suspensions of Bg (10 7 cells per ml) or Fg (10 6 spores per ml) for 1 min and then were individually sealed in plastic bags for 72 h. The infected plants were then placed in a greenhouse and rice grains exhibiting blight symptoms were counted 14 d after inoculation. Disease severity was calculated based on the number of diseased grains per rice head. For the DON analysis, 40 spikes were dried, pulverised, and then analysed by an enzyme-linked immunosorbent assay method.
Data availability. The RNA-seq data have been deposited in the NCBI Gene Expression Omnibus database with accession code GSE104623. Other relevant data supporting the findings of the study are available in this published article and its Supplementary Information files, or from the corresponding authors upon request.