Honeydew-associated microbes elicit defense responses against brown planthopper in rice

The digestive waste product of sucking herbivores, known as honeydew, betrays the presence of brown planthopper to rice plants, due to the bacteria contained in the honeydew secretions.


Introduction
The rice brown planthopper (BPH), Nilaparvata lugens (Stål) (Hemiptera: Delphacidae), is one of the most serious pests of rice in Asia. It is alarming that a sharp increase in the frequency and extent of hopper outbreaks has been noticed since ~2002, resulting in gross losses of rice crop, particularly in China, Thailand, and Vietnam (Fujita et al., 2013). Rice damage (often called 'hopper burn' as opposed to viral symptoms) results from extensive feeding of hoppers, and extraction of phloem sap by means of a stylet protected by a salivary sheath (Wang et al., 2008). Saliva is secreted into phloem which facilitates hopper feeding and interferes with the plant's immune system (Petrova and Smith, 2015;Ji et al., 2017;Ye et al., 2017). Due to ingestion of assimilate-rich sap, large amounts of sticky feeding residue, known as honeydew (Auclair, 1963), are deposited on plants through feeding of compatible insects. Honeydew then supports growth of various microbes on the plant surface, leading to a sooty appearance of infested plants (Fujita et al., 2013;Supplementary Fig. S1A at JXB online).
The broadly accepted concept of plant-herbivore co-evolution assumes that plants attacked by a diverse community of insects and pathogens evolve an equally diverse set of immune responses (Spoel and Dong, 2008;Fürstenberg-Haag et al., 2013). This implies that plants should also evolve diverse recognition mechanisms to be able to utilize these different responses adequately. Indeed, similar to plant-pathogen interactions, where plant immunity is triggered by pathogen-or microbeassociated molecular patterns, hereafter referred as PAMPs or MAMPs, respectively (Jones and Dangl, 2006;Zipfel, 2014), several herbivore-associated molecular patterns (HAMPs) have also been found (Felton and Tumlison, 2008;Mithöfer and Boland, 2008;Acevedo et al., 2015). However, in contrast to PAMPs, no receptors for HAMPs have been reported so far. In general, HAMPs occur in oral secretions, saliva, oviposition fluids, and feces of insect herbivores (Rayapuram and Baldwin, 2007). In sucking herbivores, honeydew represents another potential source of HAMPs, but it has not been investigated in detail. However, as honeydew can only be produced by actively feeding insects, this information could be important for plants in order to distinguish compatible and incompatible sucking insect attacks.
Sucking herbivores, including cicadas, aphids, whiteflies, planthoppers, and leafhoppers, transmit viruses, mycoplasmalike organisms, and pathogenic bacteria to plants (Sugio et al., 2011;Deng et al., 2013;Whitfield et al., 2015). In addition to pathogens, they also harbor obligate and/or facultative microbial symbionts inside their body that can be externalized via insect secretions, namely saliva and honeydew (Douglas, 2015;Skidmore and Hansen, 2017). For example, a set of 11 putative proteins from Buchnera aphidicola was found in the saliva secreted by potato aphids (Macrosiphum euphorbiae) (Chaudhary et al., 2014). Pea aphid (Acyrthosiphon pisum) honeydew also contained proteins from the endosymbiotic bacteria and gut flora (Sabri et al., 2013). In addition, sucking herbivores may ingest, together with phloem sap, various microbes from plants (Bove and Garnier, 2002), and secrete them afterwards as part of the honeydew. Finally, insects carry specific and/or non-specific microbes on their surface (Toledo et al., 2011), which may contaminate and further amplify in the nutrient-rich honeydew secretions on plants (Fujita et al., 2013).
Widespread microbial presence in plant-herbivore interactions indicates an important role for microbes in plant defense against herbivores (Kaloshian and Walling, 2016;Schausberger, 2018). Chaperonin GroEL from Buchnera species, obligate mutualists and primary endosymbionts of aphids (Wilson et al., 2010), is secreted in saliva and activates plant immune responses (Chaudhary et al., 2014). In another example, the facultative symbiont Hamiltonella defensa of whiteflies (Bemisia tabaci), secreted through saliva, exploits the antagonistic relationship between salicylic acid (SA) and jasmonic acid (JA) signaling (Glazebrook, 2005) to suppress plant defenses, and thereby benefit the insect host (Su et al., 2015). Similar to sucking herbivores, chewing insect Colorado potato beetle (Leptinotarsa decemlineata) larvae utilize bacteria in their oral secretions to manipulate tomato (Solanum lycopersicum) defense (Chung et al., 2013a, b). Bacteria from the oral secretions of the fall armyworm (Spodoptera frugiperda) showed various effects on defense in tomato and maize (Acevedo et al., 2017), and bacteria orally secreted by the false potato beetle (Leptinotarsa juncta) triggered distinct defense responses in preferred and non-preferred host plants (Wang et al., 2016).
In our research, a strong elicitor activity associated with the honeydew of BPHs, and particularly its macromolecular fraction, was observed. We hypothesized that some of the honeydew-associated microbes could be perceived by rice that triggers stronger defense against BPHs (Alamgir et al., 2016). Reports that planthopper-elicited defenses in rice partially resemble those induced by pathogens further supported this idea (Zhou et al., 2009;Duan et al., 2014;Li et al., 2017). We then isolated seven strains of culturable microbes from BPH honeydew, and investigated their role in rice defense against BPHs. Simultaneously, we prepared microbe-suppressed BPHs to show that honeydew microbes, indeed, betray BPHs to rice, which then mounts stronger immune responses, namely accumulation of more phytoalexins, and potentially recruits more natural enemies of BPHs through the modulation of their volatile emissions.

Plant cultivation and insect rearing
Rice plants (Oryza sativa L. var. Nipponbare) were used for insect rearing and collection of honeydew. Seeds were germinated in a nutrient-rich soil Kumiai Ube Baido No.2 (MC Ferticom, http://www.mcferticom. jp/index.html). After 10 d, seedlings were transferred to larger pots with paddy field soil mixed in a 5:1 (v/v) ratio with nutrient-rich substrate as specified above. The plants were kept at 24-26 °C day/20-22 °C night temperatures and ambient humidity at a 14-16 h photoperiod in the growth room supplemented with both natural and fluorescent lights. A colony of Koshi (Kumamoto Prefecture, Japan) field-collected BPHs (N. lugens) has been maintained in the laboratory since 2014 on a constant supply of young rice seedlings produced from non-sterile rice seeds, collected annually in the field. Seedlings were germinated and grown in the growth room using commercial soil substrate for rice (OK Soil; ISEKI & CO., Ltd, http://www.iseki.co.jp/english/).

Honeydew collection
Ten adult BPH insects were enclosed in a clean 4×6 cm clip cage, which was then attached to a single young leaf of the 6-to 8-week-old rice plant (Fig.  1A, B) kept in the growth room under controlled temperature conditions. Typically, 96 h later, honeydew deposits on the clip cages as shown in Fig.  1C were carefully collected using a 20 μl pipette tip, with a small amount of sterile water as diluent, and transferred into sterile microcentrifuge tubes. An identical procedure was used to wash the surface of clip cages attached to plants without BPHs, which was then used for control treatments. All collected fractions were stored at -80 °C until used in experiments (cell assays, intact plant treatments, microbe isolations, and identifications).

Isolation and cultivation of BPH honeydew-associated microbes
Honeydew from BPHs was plated in dilution series on agar plates with culture medium, LB Broth, Lennox (Nacalai Tesque, Kyoto, Japan) or Gifu Anaerobic Medium (GAM; Nissui Pharmaceutical Co., Ltd, Tokyo, Japan), and incubated for 2 d at 28 °C. Single colonies from plates with distinct morphologies were purified through several passages, and isolated microbial strains were streaked onto a master plate for maintenance, identification, and characterization.

Elicitation of rice cells
Cultured cells were used to examine the effect of honeydew extracts and microbes on induced defense responses of rice. Cell suspensions were derived from mature rice seed embryos placed on a modified N6 callus-forming agar medium supplemented with 1 mg l -1 synthetic auxin (2,4-D). Cells were transferred and propagated in 300 ml conical flasks containing liquid N6 culture medium under constant shaking (120 rpm) at 25 °C, and passage-propagated weekly as described previously in Shinya et al. (2016). Before each experiment, 40 mg aliquots of freshly transferred cells were placed in a 24-well microtiter plate (Techno Plastic Products AG, Switzerland) and pre-incubated in 1 ml of new culture medium for 30 min to subdue initial stress responses. Equivalent amounts of honeydew or water from empty clip cage washes were directly applied to treatment and mock control groups, respectively. Bacterial isolates from honeydew, grown on LB plate for 2 d at 28 °C, were suspended in sterile water, adjusted to OD 600 =0.2, and then 2 μl aliquots were added to preincubated cells. Pure water and 10 nM chitin oligomer (GlcNAc) 8 were used as negative and positive control treatments, respectively . Treated cells were incubated at slow speed on a Taitec MicroMixer E-36 (Taitec Corporation, Saitama, Japan) for 24 h under growth conditions described for plants. Harvested cells, after removal of liquids by micropipette, were frozen in liquid nitrogen and stored at -80 °C until analysis.

Treatment of intact rice plants
Rice seeds (O. sativa var. Nipponbare) were germinated as described above, and cultivated for ~6 weeks, after which the last fully developed leaf (~20×1 cm) was used for treatments. Typically, 2 μl of concentrated honeydew collection (or the respective control solution) were applied on the leaf, and gently rubbed on the surface with fingers covered by a clean rubber glove. To mimic BPH herbivory that includes small piercing wounds, the last fully developed leaf was wounded with a fabric pattern wheel along the midvein, and wounds were immediately treated with 2 μl of concentrated honeydew, or the respective control solution. Representative microbial isolates were suspended as described for cell treatments but using 15% (w/v) sucrose in sterile water, and 2-5 μl aliquots were rubbed on intact or wounded leaves (sucrose was used as control). For real herbivory, 10 BPH adults were applied to the last fully developed leaf enclosed in 4×6 cm clip cages. Treated parts of the leaves were sampled at 24, 48, and 72 h time points after treatment, immediately frozen in liquid nitrogen, and kept at -80 °C until analysis.

Classification of BPH-associated microbes by MALDI-TOF/MS
Microbes isolated from honeydew were subjected to matrix-assisted laser desorption/ionization-time of flight/MS (MALDI-TOF/MS) analysis as described in Tani et al. (2012), with some modifications. Using a toothpick, bacterial cells were lifted from the master plates and spotted onto the MALDI steel target plate, and dried in air. Then, 2 µl of matrix solution (saturated solution of sinapinic acid in 50% acetonitrile and 2.5% trifluoroacetic acid) was overlaid onto each sample, and samples were allowed to dry in air. The laser shots were applied until the intensity (arbitrary unit) of the highest peak reached between 6000 and 10 000 (usually 300-1000 shots). Escherichia coli DH5α (a derivative of E. coli K12) was used as a standard to validate the method. The data were analyzed with MALDI BioTyper 3.0 software (Bruker Daltonics Inc.) to construct a main spectra projection (MSP) dendrogram based on spectra similarity using default program settings as described in Tani et al. (2015).

Identification of honeydew microbes by DNA sequencing
Representative isolates in the MSP dendrogram were subjected to 16S rRNA gene sequencing. Genomic DNA was extracted from isolated colonies of representative strains and direct PCR was used to amplify ~1.5kb 16S rRNA gene fragments from the genomic DNA using the fD1 and rD1 primer set as reported by Weisburg et al. (1991). The PCR conditions were 30 cycles of 30 s at 94 °C, 30 s at 52 °C, and 2 min at 72 °C, with subsequent final extension at 72 °C for 7 min. An exception applied to isolate 4-24, which only amplified at the annealing temperature of 54 °C. Direct sequencing of amplified PCR products with the original fD1 and rD1 primers was conducted using an ABI Prism BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Carlsbad, CA, USA) on a 3130/3130xl Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). Furthermore, selected PCR products were re-amplified, size-fractionated on an agarose gel, excised, purified, and cloned into the pGEM-T Easy vector system (Promega Corp., Madison, WI, USA) for high quality sequencing with vector-specific primers. Nucleotide sequences were analyzed using Genetyx ver. 11 (Genetyx Corp., Tokyo, Japan). Nucleotide BLAST and homology search against registered 16S rRNA gene nucleotide sequences in the EzBioCloud database (Yoon et al., 2017) was conducted to identify microbes. The EzBioCloud results were cross-referenced with the user-deposited nucleotide sequences of 16S rRNA genes in the DDBJ (www.ddbj.nig.ac.jp) and EMBL (www. ebi.ac.uk) databanks.

Antibiotic resistance of microbes and development of microbe-free BPHs
Antibiotic resistance for each microbial isolate was tested using 11 different commercially supplied antibiotics: ampicillin (50 µg ml -1 ), carbenicillin (50 µg ml -1 ), kanamycin (50 µg ml -1 ), chloramphenicol (20 µg ml -1 ), rifampicin (50 µg ml -1 ), tetracycline (15 µg ml -1 ), gentamicin (50 µg ml -1 ), nalidixic acid (30 µg ml -1 ), neomycin (10 µg ml -1 ), spectinomycin (50 µg ml -1 ), and streptomycin (50 µg ml -1 ). Microbial isolates were applied on LB plates with antibiotics using a sterile toothpick and examined after growing for 2 d at 28 °C. Antibiotics effective against microbes were combined and used for treatment of BPHs as shown in Supplementary  Fig. S10. A clean 50 ml plastic tube containing 20 ml of sterile water (control treatment) or antibiotic cocktail (test treatment) was used with a bunch of 10-15 cm long stem cuttings freshly removed from the 2-to 3 week-old rice seedlings. Stems were inserted via a 5 mm hole punctured in the lid, and cut surfaces were immediately submerged in sterile water or antibiotic cocktail. To allow collection of clean honeydew from waterand antibiotic-treated BPHs, a Parafilm (BEMIS, Neenah, WI, USA) layer was placed on top of each lid, closely surrounding the rice stems ( Supplementary Fig. S10). Honeydew released from BPH by gravity was collected from the Parafilm, and used to determine microbial presence on LB plates using serial dilutions. Seedlings and antibiotics were replaced daily for 4 d, after which the antibiotic concentration was decreased to one-fifth, and BPHs were allowed to partially recover for another 2 d on the plants. BPHs were then captured, after being briefly anesthetized with carbon dioxide, and placed in larger 9×9 cm clip cages for plant treatments with 4-9 biological replicates, using set-up and time periods as described in each experiment.

BPH performance tests
In experiments shown in Supplementary Fig. S11A, the youngest developed leaf on 6-week-old rice plants was wounded and 5 µl of crude or filtered (0.22 µm pore) honeydew was immediately applied by rubbing on the surface. BPHs (10 adults or 15 nymphs at 3-4 instars) were applied 24 h later to the upper part of the leaf inserted through a slit in a plastic cylinder (8 cm diameter, 25 cm length; Supplementary Fig. S11D). Dead BPHs were counted after 7 d. In experiments shown in Supplementary  Fig. S11B and C, 10 µl of crude or filtered honeydew was directly applied by rubbing on intact leaves. BPHs (15 nymphs at 3-4 instars) were applied to the plants after 3 d, as described above. Dead BPHs were counted at 2, 4, and 6 d after feeding exposure.

Quantification of secondary metabolites
The protocol described in Tanabe  air purge from the attached syringe, samples were slowly eluted with 1.5 ml of 100% methanol (Wako Pure Chemical Industries, Ltd, Osaka, Japan). After brief clearing of eluates in a microcentrifuge at maximum speed, 10 μl aliquots were measured on a triple quadrupole LC-MS/MS 6410 system (Agilent Technologies) equipped with a Zorbax SB-C18 column (50×2.1 mm ID, 1.8 μm, Agilent Technologies), essentially as described in Alamgir et al. (2016).

Collection of headspace volatile organic compounds (VOCs)
A headspace collection system was used as described previously in Sobhy et al. (2017). Independent rice plants in pots (4-6 weeks old) were treated on the last fully developed leaf with 10 BPHs, 2 μl or 5 μl of concentrated honeydew, or wounding combined with honeydew application. Control leaves were used without treatments but still covered with empty clip cages where appropriate. For each collection, plants grown in a pot covered in a plastic Ziploc bag to limit soil volatiles were inserted in an acrylic cylinder (67 cm high×10 cm ID), and each cylinder with inlet and outlet ports was flushed with air at ~0.75 l min −1 , passing through 10 cm glass traps (5 mm ID) containing Porapak Q sorbent (200 mg, Supelco Analytical, Bellefonte, PA, USA). In every experiment, a set of 12 cylinders was connected to a single vacuum pump ULVAC DAP-12S (ULVAC KIKO Inc., Japan), while an additional four traps were used for collection and detection of background air volatiles. The bottom part of each cylinder was placed in 5 cm of water for complete closure of the system. VOCs were collected in 24 h periods for up to 2-3 d, depending on the experiment. Samples were eluted from traps with 1 ml of dichloromethane, spiked with 400 ng of tetralin (1,2,3,4-tetrahydronaphthalene; Nacalai Tesque, Kyoto, Japan) as an internal standard, and analyzed on a 240 Ion Trap Mass Spectrometer coupled to an Agilent 7891A GC using settings and conditions described in Sobhy et al. (2017). MS data collected in the mass range m/z 40-300 were analyzed by Agilent Workstation 7.02. Tentative identifications were made by comparison of spectra with the NIST 2011 Mass Spectral Library and Software (US National Institute of Standards and Technology, USA), and confirmed by co-elution with authentic standards (linalool, limonene, and caryophyllene; Wako Pure Chemical Industries, Ltd). Linalool, limonene, and caryophyllene contents were quantified using external standard calibration curves.

Phytohormone quantifications
Rice leaves (30-100 mg fresh mass) treated with honeydew or isolated microbes were harvested and flash-frozen in liquid nitrogen. Phytohormone extraction and analysis was performed essentially as described by Fukumoto et al. (2013), using deuterium-labeled internal standards available for JA, jasmonoyl-l-isoleucine (JA-Ile), abscisic acid (ABA), and SA. OPDA (12-oxo-phytodienoic acid), for which the authentic deuterated standard was not available, was quantified using the structurally related internal standard d3-JA, and expressed as equivalent of the respective compound.

Statistical analyses
Statistical analyses (

Elicitation of rice secondary metabolism with honeydew
To determine the effects of the sucking pest's honeydew on rice defense, honeydew was collected from the BPHs kept on O. sativa L. var. Nipponbare plants, using a clip cage with 10 adult insects feeding on a leaf ( Fig. 1A-C). In in vivo bioassays, rice cultured cells treated with honeydew isolates showed a strong variation in cell and media color between control (empty-cage-wash) and individual honeydew isolates (Fig.  1D). We assumed that BPH secretions have triggered specific changes in rice secondary metabolism, and therefore harvested the cells and examined their secondary metabolite levels. Two rice phenolamides, p-coumaroylputrescine (CoP) and feruloylputrescine (FP), were strongly induced at 24 h after addition of 2 μl of crude BPH honeydew to 1 ml of cultivation medium with cells ( Fig. 1E; Supplementary Fig. S1). To confirm that these phytoalexins are also induced in intact plants, the last fully developed leaves of the 6-week-old rice seedlings were treated with 2 μl of crude honeydew, which was gently rubbed on the leaf surface. While metabolite induction was not observed at 24 h and 48 h post-treatment, FP and another herbivoryinduced leaf phytoalexin, feruloylagmatine (FA), accumulated more at 72 h post-treatment compared to control leaves (Fig 2;  Supplementary Fig. S2). Similar to honeydew treatments, the levels of CoP, FP, and FA were induced by BPH feeding after 72 h (insets in Fig. 2).

BPH honeydew induces rice VOCs
Limonene, linalool, and caryophyllene are typical volatiles released from rice plants, constitutively (limonene and caryophyllene) and/or after herbivory (linalool and caryophyllene) (Gomez et al., 2005;Mithöfer and Boland, 2012;Sobhy et al., 2017). When 6-week-old rice seedlings were exposed to feeding of BPHs, release of monoterpene linalool to the headspace was promoted, but monoterpene limonene and sesquiterpene caryophyllene levels remained unchanged. While direct rubbing of 2 μl or 5 μl of honeydew on the leaf did not induce significant changes in VOC emissions, linalool levels increased in a dosedependent manner after honeydew application (Fig. 3). Next, we made a series of small wounds to the leaves, using a serrated fabric pattern wheel, and immediately applied honeydew that resembled honeydew deposition on the BPH-pierced leaf surface that, under natural conditions, also includes mechanical abrasions, chewing damage, and necrotic lesions ( Supplementary  Fig. S1A). This time, significantly higher emissions of linalool and caryophyllene were observed (Fig. 3), suggesting that honeydew indeed amplifies rice indirect defense responses in wounded leaves, which may also apply to field-grown rice plants.

Filtered honeydew shows reduced activity in rice cells
The strong induction of rice phytoalexins in both rice cells and seedlings, and VOCs in the seedlings suggested that honeydew contains a dominant elicitor of rice defense. In order to size-fractionate honeydew, and separately examine elicitor activity in each fraction, we centrifuged honeydew after dilution in sterile water 1:10 (v/v) at 15 000 g for 15 min at 4 °C. When the supernatants were tested for elicitor activity, CoP and FP levels were highly reduced compared with application of crude (2 μl) and diluted (10 μl) honeydew, or 10 μl of re-suspended pellet (in 100 μl of sterile water) to rice cells ( Fig. 4A; Supplementary Fig. S3). Next, supernatants were passed through a 0.22 µm filter (Syringe-Filter 0.22 µm, TPP Techno Plastic Products AG, Switzerland). When cells were treated with 10 μl of the filtrate, CoP and FP levels were even more reduced compared with centrifuged honeydew treatments (Fig.  4A). Substantial loss of elicitor activity during centrifugation and/ or filtration indicated that honeydew from BPH contains a fairly large molecular size elicitor, such as microbial cells, which strongly alters phytoalexin profiles in rice. Such a putative microbial elicitor was stable through a heat treatment of honeydew at 100 °C for 20 min (Supplementary Fig. S4). Therefore, although we refer to elicitors in honeydew as microbes, the actual elicitors should be heat-stable molecular patterns derived from the bacterial cells, such as hairpin-like proteins, flagellin, or oligosaccharides.

BPH honeydew contains microbes with diverse antibiotic resistance
The BPH body contains microbial cells, both internally and externally, either of which may eventually appear in honeydew. Although obligate microbes cannot be easily cultured on artificial media, assuming that microbes in honeydew should survive and proliferate on rice leaves, we approached isolation of microbes from BPH honeydew using a simple type of cultivation medium (see the Materials and methods). A total of 84 microbial isolates were obtained from raw BPH honeydew samples collected on four different occasions (designated by prefixes 1-4 in isolate numbers) and maintained on LB medium for further experiments. At first, whole-cell MALDI-TOF/MS analysis was used to group BPH honeydew-associated microbial isolates into species, based on the similarity of their major protein fingerprints (Fig. 4Ba). In addition, antibiotic resistance was examined with 11 different antibiotics (Fig. 4Bb). Because patterns of antibiotic resistance correlated well with the visual appearance of bacterial colonies (Fig. 4Bc), and the dendrogram, isolates could be divided into seven major clusters, and representative isolates (2-06, 2-08, 2-19, 3-16, 4-05, 4-07, and 4-24) from each cluster were subjected to 16S rRNA gene sequencing. The sequences obtained were compared with EzBioCloud (Table 1), DDBJ, and EMBL databases (Supplementary Table S1). Isolates 2-06, 2-19, and 4-24 showed high sequence similarity (>99%) to those with accession numbers JQ975877, GU124492, and GU124498, respectively, that belong to reported endosymbionts of N. lugens Supplementary Table S1).

Some but not all honeydew microbes elicit phenolamide levels in rice
All seven isolates were examined for their potential to elicit rice defenses. Prominent color changes were observed in rice cells treated with the microbial isolates 2-19, 3-16, and 4-24 (Fig. 5A). After visual assays, metabolites were extracted and phytoalexin levels were determined as before. Cells treated with isolates 2-19, 3-16, and 4-24 showed highly increased levels of CoP and FP relative to water-treated controls (Fig.  5B). Induction levels were comparable with those of chitin oligomer (GlcNAc) 8 , a potent PAMP known to trigger defense in plants (Zipfel, 2014). The remaining isolates induced less color change in cells and, correspondingly, lower levels of induced phytoalexins were detected in the extracts.
In contrast to rice cells, metabolites were not significantly elevated by simple rubbing of bacterial suspensions in 15% sucrose (OD 600 =0.2) on the leaf surface ( Supplementary Fig.  S5). Similarly, microbes applied on the intact leaf surface did not induce significant changes in phytohormone levels, determined by LC-MS/MS ( Supplementary Fig. S6), although feeding of BPHs clearly elicited all major oxylipins (OPDA, JA, and JA-Ile), and other hormones (ABA and SA) in the leaves after insect attack ( Supplementary Fig. S7). Previously, we observed induction of VOCs after combined wounding and honeydew treatments (Fig. 3), suggesting that small wounds from probing and sucking, as well as environmental damage,  Fig. 4. Induction of phytoalexins in rice cells treated with raw and filtered BPH honeydew and bacterial isolates from honeydew. (A) Phenolamide contents in cells treated with raw or processed honeydew determined by LC-MS/MS after 24 h treatment. Chitin oligomer (GlcNAc) 8 was used as positive control. Different letters show statistically significant differences between treatments by ANOVA (P<0.05; Tukey HSD test). n=3; error bars=SE; FM, fresh mass. (B) (a) Biotyper-generated MSP dendrogram based on the protein mass spectra. Different tree colors suggest a species-level identification (isolates under a tree of the same color are likely to belong to the same species). (b) Antibiotic resistance of microbial symbionts isolated from the BPH honeydew. '+' is used for resistant, and '-' for sensitive colonies on LB plates. Antibiotics are Amp, ampicillin (50 µg ml -1 ); Car, carbenicillin (50 µg ml -1 ); Kan, kanamycin (50 µg ml -1 ); Chl, chloramphenical (20 µg ml -1 ); Rif, rifampicin (50 µg ml -1 ); Tet, tetracycline (15 µg ml -1 ); Gen, gentamicin (50 µg ml -1 ); Nal, nalidixic acid (30 µg ml -1 ); Neo, neomycin (10 µg ml -1 ); Spe, spectinomycin (50 µg ml -1 ); and Str, streptomycin (50 µg ml -1 ). (c) Images of the representative microbial isolates from each combined MSP/antibiotic resistance clade. The isolates were identified by search against the EzBioCloud database after amplification of 16S rRNA gene sequences from representative bacterial isolates using the fD1 and rD1 primer combinations as described in the Materials and methods. See Supplementary Table S1 for additional blast hits found in DDBJ/EMBL databanks. might be essential for microbes in honeydew to enter plant cells. We then selected the isolate 2-08, as one with low elicitor activity, and the isolate 4-24, as a strong phytoalexin inducer, to conduct the following set of combination treatments. Bacterial isolates were suspended in 15% sucrose and, immediately after wounding with a serrated pattern wheel to mimic BPH piercing, bacterial suspensions (2 µl) were applied to the wounds. A time course of phytoalexin accumulation showed that, apart from wound-induced increases, further elevation of CoP and FP levels occurred in leaves treated with isolate 4-24. This was similar to crude honeydew application on wounded leaves (Fig. 6). At phytohormone levels, JA and JA-Ile were enhanced by crude honeydew, as well as by isolate 4-24, especially at 24-48 h post-treatment compared with wounded leaves treated with 15% sucrose as control (Supplementary Figs S8,S9). Application of isolate 2-08 showed much less effect on phenolamides and phytohormones, with levels remaining close to those of wounded leaves treated with 15% sucrose.

Reduced phenolamide accumulation in microbe-suppressed BPHs
It is well known that microbial symbionts provide insects with essential functions, and their loss is thus detrimental to the host. This naturally obstructs unbiased tests of honeydew microbiota in the elicitation of plant defense. Nevertheless, we still attempted to obtain BPHs with suppressed microbial levels to support our findings.
Detailed information from antibiotic screening was used to design an antibiotic cocktail effective against all types of honeydew isolates, and potentially other undetermined microbes. In order to avoid further disturbance of BPHs by feeding on artificial diet, antibiotics were applied to stem cuttings prepared from young rice seedlings, and BPHs were allowed to feed on them (see Supplemental Fig. S10 for details). Method development and evaluation were carried out based on BPH mortality and residual microbe levels in the collected honeydew. For this purpose, a layer of clean Parafilm was placed under the leaves to collect honeydew from BPHs ( Supplementary Fig. S10). After multiple trials, BPHs were plant-fed during an optimized 6 d cycle on antibiotic cocktails containing tetracycline (250 µg ml -1 ), rifampicin (100 µg ml -1 ), and spectinomycin (250 µg ml -1 (see the Materials and methods and Supplementary Fig. S10 for details). After completion of each antibiotic treatment cycle, BPHs were transferred to clip cages (4-9 biological replicates) and attached to young rice leaves for 4 d to elicit plant defense responses. Each time, a group of identically treated BPHs, but kept on seedlings without antibiotics, was used for comparison. Only leaf parts directly exposed to BPHs were collected and examined for phytoalexin levels (Fig. 7A). In repeated trials, variable efficiency of microbe elimination in BPHs was observed (Fig. 7B). However, only when microbes were completely suppressed (bottom panel of Fig. 7B) did plants accumulate significantly less direct defense phytoalexins compared with microbe-containing BPHs (top panel in Fig.  7B), or insects that still retained the isolate 4-24 (large colonies on a plate in the central panel of Fig. 7B). Notably, isolate 4-24 showed multiple antibiotic resistances (Fig. 4B), and it was difficult to eliminate completely from BPHs. Taken together, initially observed elicitor activity of honeydew in rice, microbe effects on rice cells and plants, and antibiotic treatment results suggest that rice plants may use certain microbes from honeydew as important signals for defense. In support of this, BPH nymphs showed higher mortality between 2 d and 4 d of feeding exposure to crude honeydew-treated leaves compared with filtered honeydew application ( Supplementary Fig. S11).

Discussion
Plants perceive various signals from the environment, which inform them about changes, and potential danger, such as the presence of pathogens and herbivores (Gust et al., 2017). Accordingly, we show that microbes from honeydew of BPHs can effectively elicit rice defenses. The antibiotics fed to BPHs suppressed cultivable microbes in the honeydew, which in turn attenuated the inducible defenses of rice against BPHs. We propose a model in which microbes in honeydew are perceived by rice to amplify its defense, which acts in addition to elicitors found in the BPH salivary secretions, and mechanical damage of various sources.

Quest for sucking/piercing arthropod elicitors
In contrast to chewing herbivores, piercing/sucking arthropods, such as BPHs used in this study, are associated with a minimal level of mechanical injury (Walling, 2008), which can be further masked by effectors released from the plant intruders (Bos et al., 2010;Atamian et al., 2013;Rodriguez et al., 2014;Naessens et al., 2015;Villarroel et al., 2016). For example, salivary EF-hand calcium-binding protein NlSEF suppresses  Fig. 6. Accumulation of phytoalexins in rice leaves treated with microbial isolates from BPH honeydew. Phenolamide contents were determined in leaves after wounding with a serrated pattern wheel and applying microbial isolates suspended in 15% (w/v) sucrose adjusted to OD 600 =0.2. Sucrose and honeydew were used as negative and positive controls, respectively. Different letters show statistically significant differences between treatments by ANOVA (P<0.05; Tukey HSD test). n=4; error bars=SE; W+Suc, wounding with 15% sucrose; W+Suc+2-08, wounding with isolate 2-08 suspended in 15% sucrose; W+Suc+4-24, wounding with isolate 4-24 suspended in 15% sucrose; W+HD, wounding with raw BPH honeydew; W, wounding; Suc, sucrose; HD, honeydew; FM, fresh mass. defense responses in rice against BPHs . Another secreted salivary effector, endo-β-1,4-glucanase NlEG1, enables the stylet of BPHs to reach phloem cells by degrading cellulose in the protective plant cell walls . Reports by Ye et al. (2017) and Ji et al. (2017), and findings of Petrova and Smith (2015), suggest that BPHs are involved in reprograming of the rice transcriptome, leading to nutritional enhancements that benefit them. However, in spite of the effectors, rice still vigorously responds to feeding of BPHs by accumulation of phytoalexins (Alamgir et al., 2016;insets in Fig. 2), and activates its hormonal signaling ( Supplementary  Fig. S7). Therefore, one or more dominant elicitors from BPHs must be perceived by rice.
Recently, secreted mucin-like protein from the salivary sheets of BPHs was shown to trigger defense responses in rice, including cell death, expression of defense-related genes, and callose deposition (Shangguan et al., 2018). In our report, honeydew, and specifically honeydew-associated microbes, elicited chemical defenses, such as phytoalexins, and release of VOCs from the rice plants. While honeydew is rich in nutrients, and thus widely open to contaminations, the microbes can originate from various sources, including plant interiors (e.g. phloem-restricted bacteria), the BPH digestive tract (e.g. insect gut symbionts), or the environment. From the functional perspective, insect-associated symbionts appear to be particularly good targets for evolution of novel signals of defense in plants (Schausberger, 2018). These can be delivered to plants via honeydew as shown here, or via saliva secreted to phloem during BPH feeding.

Insect-associated microorganisms modulate plant defense
In this study, BPH honeydew contained seven strains of culturable bacteria (Table 1; Supplementary Table S1). From the BPH genome and its analysis, yeast-like symbionts (YLSs) also occur in BPHs which complements essential nutritional pathways in the planthopper (Chen, 1981;Tang et al., 2010;Xue et al., 2014). However, YLSs have not been found on LB and GAM media, possibly due to their non-culturable character, or lack of their secretion into honeydew, and their role in plant defense needs to be examined separately.

Microbial effects on direct and putative indirect defenses
We show that both honeydew and microbes induced phenolamide phytoalexins that serve as direct defense against BPHs in rice (Alamgir et al., 2016). VOCs are also rapidly elicited during herbivore attack to recruit natural enemies of herbivores (Aljbory and Chen, 2018), which constitutes a sophisticated system of indirect plant defense. Although various microbederived VOCs affect plant growth and defense (Bitas et al., 2013;Junker and Tholl, 2013;Liu and Zhang, 2015), much less is known about the microbe-induced production of volatiles in plants. Here, we found that VOCs from rice can be positively modulated by microbe-containing BPH honeydew application. As direct and indirect defense responses are well known to suppress insect performance ( (Mithöfer and Boland, 2012), phenolamides and VOCs elicited by honeydew (and its microbiota) in this study are expected to reduce the performance of BPHs adequately under natural conditions. Induction of two pathways, direct (phenolamides) and indirect (VOCs) defense, suggests that honeydew affects some early steps in the rice defense cascade, such as phytohormone levels, and/or downstream signal transduction mechanisms (Wasternack and Song, 2017). While we have already examined phytohormones, it is still necessary to investigate other components of plant defense signaling against herbivores, such as Ca 2+ -mediated responses (Arimura and Maffei, 2010), levels of reactive oxygen species (ROS) (Zebelo and Maffei, 2015;Shinya et al., 2016), or activity of mitogen-activated protein (MAP) kinases (Hettenhausen et al., 2015). Finally, transcription factors induced by honeydew, and its components, that directly link signaling to defense genes, should be taken into consideration (Woldemariam et al., 2011).

Hormonal pathways under attack
Defense responses against sucking insects, including BPHs, resemble pathogen-induced alterations (Walling, 2000;Zhou et al., 2009;Duan et al., 2014;Villarroel et al., 2016). Here, complexity arises from the antagonism in plant defense signaling against insects and pathogens, which is mediated by JA and SA, respectively (Glazebrook, 2005). For example, feeding of whiteflies induced SA, which in turn suppressed jasmonatemediated responses in Arabidopsis (Zhang et al., 2013). Some chewing insects use microbes and SA to subdue wound-and HAMP-elicited jasmonate signaling (Chung et al., 2013b). Similarly, honeydew from aphids suppressed JA signaling via the salicylate pathway in broad bean plants (Vicia faba), partly via SA contained in honeydew (Schwartzberg and Tumlinson, 2014). Whiteflies even glycosylate SA, and deploy this conjugate to counteract plant defense (VanDoorn et al., 2015). A whitefly-associated facultative symbiont Hamiltonella defense that occurs in salivary secretions also suppressed induced defense responses in tomato (Su et al., 2015).
Although microbes and SA are helping invaders to subdue plant defenses against herbivores, BPH honeydew and its microbes clearly worked as amplifiers of wound-induced rice defense. We also report that feeding of BPHs elicits high levels of JA, and its active form JA-Ile, after 1 d and 3 d, respectively, while SA levels only increased significantly after 4 d of BPH feeding. It can be assumed that rice is using more instant information from the honeydew to elicit JA signaling before substantial changes in SA levels can take place ( Supplementary Fig. S7). However, caution is needed as rice contains extraordinary amounts of constitutive SA (Supplemental Fig. S7), and it is not clear which fraction of total extracted SA is active in cell defense and metabolism.
Similar to our study, elimination of bacterial symbionts from herbivores to study hormonal responses was attempted before. Reduced microbial levels in herbivorous spider mites Tetranychus urticae affected pest performance but this was not linked to significant changes in JA-Ile levels in the host plants (Staudacher et al., 2017). Interestingly, OPDA precursors of JA were enhanced and suppressed in the presence of two microbes, Wolbachia and Spiroplasma, in the spider mites, respectively (Staudacher et al., 2017). In contrast, OPDA levels correlated with the increases of JA and JA-Ile in rice leaves infested with BPHs in our study. Not only BPHs but also honeydew and bacterial isolate 4-24 promoted JA and JA-Ile levels in the wounded leaves of rice (Supplementary Figs S8, S9). Somewhat in contradiction to our results, silencing of OsHI-LOX (lipoxygenase), a gene involved in JA biosynthesis in rice, suppressed resistance to chewing insects but induced tolerance responses to BPHs under laboratory/greenhouse conditions (Zhou et al., 2009). Perhaps future experiments under natural conditions, including access to tritrophic interactions, will support the positive role of JA in defense signaling against BPHs, and highlight the importance of microbe-and honeydew-promoted direct and indirect defense responses found in this study.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. BPH-infested plants in the field and honeydewinduced secondary metabolite contents in rice cells. Fig. S2. Phenolamide contents in rice seedlings treated with BPH honeydew. Fig. S3. Induction of phytoalexins in rice cells treated with honeydew fractions. Fig. S4. Induction of phytoalexins in rice cells treated with filtered and/or heated honeydew. Fig. S5. Accumulation of phytoalexins in intact rice plants treated with microbial isolates from BPH honeydew. Fig. S6. Accumulation of phytohormones in intact rice leaves treated with microbial isolates suspended in 15% (w/v) sucrose. Fig. S7. Phytohormone accumulation in rice leaves infested with BPH. Fig. S8. Accumulation of phytohormones in wounded rice leaves treated with BPH honeydew. Fig. S9. Accumulation of phytohormones in wounded rice leaves treated with microbial isolates suspended in sucrose. Fig. S10. Custom-design system for treatment of BPH adults with antibiotics. Fig. S11. Performance of BPH on crude and filtered honeydew-treated leaves. Table S1. Identification of microbes isolated from BPH honeydew in DDBJ/EMBL.