Does drought stress modify the effects of plant‐growth promoting rhizobacteria on an aboveground chewing herbivore?

Soil microbes have important effects on the interactions of plants with their environment, by promoting plant growth, inducing resistance to pests or by conferring tolerance to abiotic stress. However, their effects are variable and the factors responsible for this variation are mainly unknown. Our aim was to assess how drought stress modifies the effect of the nonpathogenic rhizobacterium Pseudomonas simiae WCS417r on plant growth and resistance against the generalist leaf‐chewing caterpillar Mamestra brassicae. We studied Arabidopsis thaliana Col‐0 plants, as well as mutants altered in the biosynthesis of the phytohormones jasmonic acid (JA) and abscisic acid (ABA). Caterpillars did not prefer rhizobacteria‐treated plants, independently of drought stress. Rhizobacteria colonization had a variable effect on caterpillar performance, which ranged from positive in one experiment to neutral in a second one. Drought had a consistent negative effect on herbivore performance; however, it did not modify the effect of rhizobacteria on herbivore performance. The effect of drought on herbivore performance was JA‐mediated (confirmed with the use of the dde2‐2 mutant), but it was still present in the ABA‐deficient mutant aba2‐1. Plant biomass was reduced by both drought and herbivory but it was enhanced by rhizobacterial colonization. Pseudomonas simiae WCS417r is able to promote plant growth even when plants are suffering herbivory. Nevertheless, the microbial effect on the herbivore is variable, independently of drought stress. To get the best possible outcome from the rhizobacteria‐plant mutualism it is important to understand which other factors may be responsible for its context‐dependency.


Introduction
In nature, plants are involved in negative interactions with e.g. pathogens and herbivorous insects, but also in positive Correspondence: Ana Pineda, Netherlands Institute of Ecology (NIOO-KNAW), P.O. Box 50, 6700 AB Wageningen, The Netherlands. Tel: +31 (0)317 473 627; email: a.pineda@nioo.knaw.nl ones such as those with beneficial microorganisms and beneficial insects (Dicke & Baldwin, 2010;Philippot et al., 2013). In addition, plants have to cope with abiotic stresses, such as drought. Plant growth is reduced by herbivory and drought and importantly, both stresses are predicted to increase in frequency and intensity in the coming years (Massad & Dyer, 2010;Zhao & Running, 2010). However, despite the importance of understanding plant responses to simultaneous biotic or abiotic stresses, these are only recently starting to be unraveled (Mattson & Haack, 1987;van Oosten et al., 2008;Tariq et al., 2013;Coolen et al., 2016;Davila Olivas et al., 2016;Nguyen et al., 2016;Pineda et al., 2016). Nonpathogenic soil microbes that live around plant roots may play an important role in plant responses to their biotic and abiotic environment, since they can induce plant tolerance to both types of stresses (Bennett et al., 2006;Sugio et al., 2014;Timmusk et al., 2014). In addition, they can promote plant growth (Paré et al., 2011;Zamioudis et al., 2013).
Importantly, the effect of root-associated microbes on herbivorous insects varies from positive (D'Alessandro et al., 2014;Megali et al., 2014) to negative (van Oosten et al., 2008;Pangesti et al., 2016a;Zebelo et al., 2016) and to date we do not fully understand the factors responsible for this variation. Biotic factors such as feeding guild and degree of specialization of the insects have been shown to affect the outcome of plant-microbe interactions. Rootassociated microorganisms have generally a positive or neutral effect on phloem feeders and specialist chewers and a negative effect on mesophyll feeders and generalist chewers (Koricheva et al., 2009;Pineda et al., 2010). In contrast, it is less well understood how abiotic factors modulate microbe-plant-insect interactions (Pineda et al., 2013). For instance, soil composition (ratio of sand and peat) and more specifically iron availability seem to be among the factors that affect the enhanced immunity that microbes confer to plants (Pangesti et al., 2014;Zamioudis et al., 2014). Drought was also shown to increase the direct effect that grass endophytes had on herbivores (Vidal, 1996). However, whether drought influences the plant-mediated effects of microbes on herbivores is not yet known and the underlying mechanisms have not been studied.
Plant responses to biotic and abiotic stresses are regulated by signal-transduction pathways under the control of several phytohormones. Feeding by piercing-sucking insects and biotrophic pathogens mainly activates salicylic acid (SA)-dependent responses, whereas jasmonic acid (JA) signaling is mainly activated upon damage by chewing insects and necrotrophic pathogens (Erb et al., 2012;Pieterse et al., 2012). In addition, abscisic acid (ABA) has been known for decades as key regulator of plant responses to abiotic stress, and especially to drought (Fujita et al., 2006;Yoshida et al., 2014). However in the last years it is becoming evident that ABA is also involved in plant responses to herbivory (Vos et al., 2013;Hillwig et al., 2016).
Plants fine-tune their defense responses to multiple stresses through crosstalk between phytohormonal signaling pathways (Pieterse et al., 2012). In Arabidopsis the MYC2-branch of the JA-signaling pathway is activated upon chewing insect feeding resulting in the transcription of the JA-responsive genes VSP1 and VSP2 (Bodenhausen & Reymond, 2007;Verhage et al., 2011;Vos et al., 2013). Interestingly, this MYC2-branch is positively regulated by ABA, and, for instance, the herbivoreinduced levels of VSP1 are not observed in the ABAdeficient mutant aba2-1 (Vos et al., 2013). However, how drought affects the cross-talk between ABA and JA signaling pathways and the consequences for herbivores still needs to be unraveled.
Besides its role in plant defenses against herbivorous insects, JA is also involved in induced systemic resistance that is mediated by beneficial rhizosphere bacteria (van der Ent et al., 2009;Pozo et al., 2015). Root-associated microorganisms can induce resistance in systemic tissues (ISR) against pathogens and herbivorous insects through a priming mechanism . Priming is characterized by a faster or stronger activation of plant defenses upon pathogen or insect attack, with lower fitness costs than induced resistance (Conrath et al., 2006;Martinez-Medina et al., 2016). Most cases of microbial-ISR require intact JA-and ET-signaling pathways . However, it is not fully understood whether a functional ABA signaling pathway is required for ISR.
The objective of our study was to assess whether drought modifies the effect of the rhizobacterium Pseudomonas simiae WCS417r (previously known as Pseudomonas fluorescens WCS417r) on Arabidopsis thaliana growth and defenses against the generalist insect herbivore Mamestra brassicae. In addition, we aimed at unraveling the role of JA and ABA in these interactions. The bacterial strain P. simiae WCS417r has been shown to promote plant growth, lateral root formation and root hair formation in Arabidopsis Col-0 plants grown in agar (Zamioudis et al., 2013). Therefore, we expected P. simiae WCS417r to have a positive effect on plant growth in potting soil as well (Pangesti et al., 2016b). However, since rhizobacteria-induced resistance against M. brassicae was variable in our previous studies (Pangesti et al., 2014;Pangesti et al., 2015;Pangesti et al., 2016b), we expected to observe again a variable effect. Additionally, based on previous evidence showing that (1) P. simiae WCS417rmediated ISR requires an intact JA signal-transduction pathway (Pangesti et al., 2014;Pieterse et al., 2014), (2) that drought induces ABA (Fujita et al., 2006;Yoshida et al., 2014), and (3) that ABA has a synergistic effect on the MYC2-branch of the JA pathway against chewing insects (Vos et al., 2013), we expected drought to strengthen the negative effect of rhizobacteria on herbivore performance.

Preparation of plants, microbes, and insects
Plants were grown from seeds in a mixture of potting soil and sand (1 : 1, v : v). We used the wild type Arabidopsis thaliana accession Columbia (Col-0). In addition we studied the mutant delayed-dehiscence2-2 (dde2-2), defective in the ALLENE OXIDE SYNTHASE gene (AOS), encoding one of the enzymes in the JA biosynthetic pathway (von Malek et al., 2002). Finally we used the ABA biosynthetic mutant aba2-1, whose gene product ABA2 is implicated in the conversion of xanthoxin to ABAaldehyde . The soil mixture was autoclaved twice at 121°C for 20 min with a 24 h interval. All seeds were sterilized by 3 h exposure to chlorine gas, and subsequently sown in Petri dishes on water-saturated filter paper and kept for 3 d at 4°C (van de Mortel et al., 2012).
The rifampicin-resistant nonpathogenic plant growthpromoting rhizobacterial strain Pseudomonas simiae WCS417r (P. s. WCS417r, previously known as Pseudomonas fluorescens WCS417r) (Berendsen et al., 2015) was used for induction of induced systemic resistance (ISR). Bacteria were grown on King's B (KB) medium agar plates containing rifampicin (25 μg/mL) (Pieterse et al., 1996) for 48 h at 28°C. Bacterial cells were collected, resuspended in 10 mmol/L MgSO 4 and adjusted to a cell density of 1 × 10 9 cfu/mL (OD 660 = 1.0). For the rhizobacterial treatment the soil was inoculated with 100 mL of the suspension of P. s. WCS417r per kilogram of soil. An equal amount of 10 mmol/L MgSO 4 was added to the soil for the control treatment. Seeds were sown on the inoculated soil in individual pots and 10 d after sowing, plants were re-inoculated with 1 mL/plant of a suspension with the same bacterial density as described above, and the control plants with 1 mL of 10 mmol/L MgSO 4 .
Plants were cultivated in a growth chamber under 8 h : 16 h light : dark at 21 ± 1°C and 60% ± 10% relative humidity (RH) and watered with 20 mL water 3 times a week, keeping the soil water content at 100%. From the second week onwards, plants were fertilized once a week with 10 mL Hyponex R per plant. One week before the bioassays, plants were assigned randomly to the drought treatment, previously named "high drought" (Pineda et al., 2016). In the control treatment, plants were watered 3 times per week. For the drought treatment plants were not watered during a week after which plants were rewatered to 60% of soil water content (Pineda et al., 2016).

Experiment 1: Effect of rhizobacterial colonization and drought on performance of a generalist insect and expression of marker genes
A. thaliana plants were randomly assigned to different treatments. Plants were treated with rhizobacteria or not, water-stressed or normally watered and inoculated with neonate larvae of Mamestra brassicae or not. All these combinations led to 8 treatments.
(a) Performance of the generalist herbivore M. brassicae and plant performance Three neonate larvae of M. brassicae were transferred to 5-week-old A. thaliana Col-0 plants using a fine paintbrush. All plants were individually confined in plastic containers covered with insect-proof mesh cloth and closed with elastic bands. Larvae were weighed to the nearest 0.001 mg using a microbalance (CP2P, Sartorius AG, Germany) 9 d after infestation. To address the effect of the treatments on plant performance, fresh aboveground plant biomass was measured using an analytical balance. A total of 20 infested plants per treatment were evaluated, whereas 10 plants per uninfested treatment were arranged to assess plant biomass. Bioassays were performed in a growth chamber under 21 ± 1°C, 60%-70% RH, 16 h : 8 h light:dark.
(b) Expression of defense-related genes The same treatments as described above were used to evaluate the transcript levels of several marker genes of the JA-and ABA-signaling pathways on wild-type Col-0 plants. A total of 12 plants per treatment were prepared, and fully expanded leaves of all the treatments were harvested at 24 h after infestation. For each treatment, 4 biological replicates were used, each consisting of 9 damaged leaves (3 leaves from 3 plants). Leaf samples were immediately frozen by immersion in liquid nitrogen and stored at -80°C for further RNA extraction. An extra set of plants was prepared to assess leaf damage 48 h after infestation, using transparent paper with square millimeter raster and counting by eye the number of mm² removed by the insect.
Total RNA was extracted and purified as indicated by the protocol RNA plant Kit (Bioline, London, UK). Measurement of RNA quality and synthesis of cDNA was performed as explained in Pangesti et al. (2014). Transcript levels of the JA/MYC2-regulated gene VEGETATIVE STORAGE PROTEIN2 (VSP2) and of the ABA and drought stress-responsive gene RAB18 were quantified. As reference genes, GLYCERALDEHYDE-3-PHOSPHATE DEHYDROGENASE (GAPDH) and F-BOX FAMILY PROTEIN1 (FBOX1) were used. Sequences of gene-specific primers for qRT-PCR are included in supplemental materials. The quality of each primer was determined before qRT-PCR analysis, which was completed following the same procedure as Pangesti et al. (2014). All qRT-PCR experiments were performed in duplicate (technical replicates) and average values were used in the analyses. Thermal cycling conditions were 95°C for 3 min, followed by 40 cycles of 95°C for 15 and 45 s at 60°C. A normalization factor was calculated by geometrically averaging the threshold cycle (Ct) values of the constitutively expressed genes GAPDH and F-BOX. Then the transcript level for each tested gene was calculated relative to the normalization factor using the 2 − Ct method (Livak & Schmittgen, 2001).

Experiment 2: Corroborating the role of JA/ABA phytohormone signaling pathways in the rhizobacteria-insect-drought interaction
To assess the role of JA-and ABA-signaling pathways on the effect of P. s. WCS417r and drought on the generalist herbivore M. brassicae, 3 neonate larvae were inoculated on A. thaliana mutants that are defective in the JA (dde2-2) or ABA (aba2-1) signaling pathways. All mutants used in the experiment have the Col-0 background, hence larvae fed on the wild-type Col-0 plants were used as a control to compare herbivore performance. Caterpillars were weighed at 9 d after infesting the plants as explained above. Additionally, leaf damage on Col-0 plants was assessed visually at 3 d postinfestation by using millimeter raster paper. Plants were divided into 2 groups (half of the replicates in each group), so the full experiment was done in 2 consecutive days following exactly the same methodology. A total of 20 infested plants were evaluated, whereas 10 plants per uninfested treatment were arranged to measure plant biomass.

Experiment 3: Preference of M. brassicae larvae
Two-choice experiments were performed by setting 2 plants in a 1 L open container, and placing 10 neonate larvae in the middle of the container. All experiments were conducted in a climatized room (21 ± 2°C) without windows, and artificial light was provided from above by 4 high-frequency fluorescent tubes (TL-D 58 W Philips, Eindhoven, The Netherlands) at an intensity of 60 ± 5 μmoL photons/m 2 /s (8 h : 16 h light : dark). The position of the containers and of the plants inside these, was randomized. The preference of the neonate larvae was assessed, by counting how many larvae were on each plant after 48 h. The following combinations of plants were used (14-16 replicates for each pair): (1) control versus rhizobacteria-treated plants; (2)

Statistical analysis
Residuals were first inspected to confirm that the assumptions of linear models were met. To analyze the effect of P. s. WCS417r and drought on herbivore performance, linear mixed models (LMM) were used. Plants were set as a random factor, whereas rhizobacteria and drought treatment were set as fixed factors. To evaluate the effect of P. s. WCS417r and drought on the relative expression of defense-related genes and on plant biomass, 3-way ANOVA was used. Herbivory, drought, and rhizobacterial colonization were set as fixed factors. In order to assess the effect of the treatments on leaf damage, 2-way ANOVA was used, setting drought and rhizobacteria as fixed factors. To analyze preference in the 2choice tests, we calculated a response ratio for each replicate ([#control]/[#control +#treatment]). This was tested against a null hypothesis value of 0.5 in a 1 sample ttest. Data were analyzed with Genstat (14th Edition, VSN Int., UK).

Rhizobacterial colonization does not affect larval preference but enhances larval performance independently of drought stress
To unravel the effect of drought and rhizobacterial treatments on the preference of a generalist herbivore, we performed several 2-choice experiments with A. thaliana Col-0 (wild type).
When plants were not colonized by rhizobacteria, caterpillars preferred control over drought-stressed plants (t 11 = 0.034). However, when plants were treated with rhizobacteria, caterpillars showed no preference (t 19 = 0.093). Caterpillars did not show any preference between control and rhizobacteria-treated plants under normal watering conditions (t 12 = 0.893) nor under drought conditions (t 17 = 0.524).
In a first performance experiment with Col-0 plants (Part a in Materials and methods section), we evaluated the effect of rhizobacteria, drought and the combination of both on the insect herbivore. M. brassicae caterpillars had a reduced performance on drought-treated plants at 9 d after infestation (LMM; F 1,74 = 8.73, P = 0.004; Fig. 1C). In contrast, rhizobacteria colonization had a positive effect on larval mass of the caterpillars (LMM; F 1,74 = 10.78, P = 0.002; Fig. 1C). This effect was independent of drought stress, since there was no interaction between the main factors drought and rhizobacteria treatment on larval performance (LMM; F 1,74 = 0.00, P = 0.984).

Drought negatively affects larval performance in a JA-dependent manner but independent of ABA, and does not affect the variable rhizobacterial effects
In a second performance experiment (Experiment 3 in Materials and methods section), we studied wild type Col-0 plants together with JA-and ABA-impaired mutant plants to assess the effect of these signaling pathways on the observed effects. Drought again negatively affected the performance of the generalist herbivore M. brassicae when feeding on wild type Col-0 plants at 9 d after infestation (LMM; F 1,163 = 8.88, P = 0.003; Fig. 2). The same negative effect was observed in the ABA-impaired mutant aba2-1 (LMM; F 1,159 = 23.96, P < 0.001; Fig. 2). However, there was no negative effect of drought on larval performance when the caterpillars fed from the JA-impaired mutant dde2-2 (LMM; F 1,192 = 1.28, P = 0.258; Fig. 2). In contrast to the first experiment, in this experiment rhizobacteria colonization of Col-0 plants had no effect on M. brassicae performance (LMM; F 1,165 = 0.38, P = 0.538; Fig. 2) nor did it modify the effect of drought (LMM, interaction; F 1,167 = 0.01, P = 0.926). A similar lack of effect was observed in the mutants dde2-2 and aba2-1 (LMM; aba2-1: F 1,158 = 3.66, P = 0.058; dde2-2: F 1,192 = 1.28, P = 0.258; Fig. 2), although in the mutant aba2-1 there was a trend towards rhizobacteria inducing resistance to M. brassicae. There was also no interaction between rhizobacteria and drought (LMM,

Herbivore-induced transcription of the JA-responsive gene VSP2 is neither primed by rhizobacterial colonization nor by drought stress
Herbivore feeding by the generalist insect M. brassicae induced expression of the defense-related gene VSP2 (3way ANOVA: F 1,31 = 39.03, P < 0.001; Fig. 4 left). However, P. s. WCS417r did not prime A. thaliana plants for an enhanced expression of VSP2, as the expression levels of VSP2 were the same in both control and P. s. WCS417r-treated plants (3-way ANOVA: F 1,31 = 2.04, P = 0.166; Fig. 4 left). Drought had no effect on the expression levels of VSP2 either (F 1,31 = 0.05, P = 0.82; Fig. 4 left). There was no interaction between any of the factors (data not shown).

Discussion
Root colonization with beneficial microbes has a differential effect on insects that ranges from positive to negative and to date we do not fully understand the modulators of this variability (Partida-Martinez & Heil, 2011;Pineda et al., 2013;Pangesti et al., 2014). Here we show that, contrary to our expectations, drought is not one of the factors modulating the outcome of the interaction between A. thaliana plants, the rhizobacterium P. simiae WCS417r, and the leaf herbivore M. brassicae. Variability in results is common when studying ecological interactions, especially when not all factors that influence these interactions are understood (Heil, 2014). In our previous studies, we observed a high consistency in a closed system of sterile plate assays, but not in an open system with soil (Pangesti et al., 2014;Pangesti et al., 2015;Pangesti et al., 2016a). In these studies, we always used the same population of A. thaliana Col-0 plants, of the herbivore M. brassicae, and the same rhizobacterial strain. Now, after excluding drought as main driver of variability in our system, we propose that the microbial communities (and factors affecting those) that are present in the soil in the open system are probably responsible for the variability we observed in our studies. More studies unraveling what biotic and abiotic factors cause this variability are needed to be able to predict the final outcome of microbe-plant-insect interactions.
Rhizobacterial colonization increased both plant growth and herbivore performance in the first experiment, independently of whether plants were suffering from drought stress. However, when we repeated the experiment rhizobacteria did not affect caterpillar performance. Despite the general pattern suggesting that generalist leaf chewers are negatively affected by beneficial microbes (Koricheva et al., 2009;Pineda et al., 2010), examples of microbes inducing susceptibility to generalist leaf chewers are accumulating (D'Alessandro et al., 2014;Megali et al., 2014;Pangesti et al., 2015). For example, larvae of the generalist insect Spodoptera littoralis had a higher mass and decreased mortality when feeding from tomato plants that were colonized by a mixture of beneficial bacteria, explained by a reduction of induced plant defenses (JA and the toxic glycoalkaloid tomatine) in microbe-treated plants (Megali et al., 2014). In contrast, most studies on A. thaliana have shown that plant growth-promoting rhizobacteria lead to an increase in the defensive compounds glucosinolates and an associated induced resistance against generalist leaf chewers (van de Mortel et al., 2012;Aziz et al., 2016;Pangesti et al., 2016a). Although we did not investigate plant chemistry in this study, based on those previous studies, the induced susceptibility is more likely to be a result of the microbialenhanced plant quality rather than to a suppression of defensive compounds.
The expression of the JA/ABA-responsive gene VSP2 and the drought-responsive gene RAB18 was upregulated after herbivore damage and drought respectively. However, P. s. WCS417r colonization did not prime the expression of either VSP2 or RAB18. Our result for VSP2 expression contrasts with previous findings. In other studies, root colonization by P. s. WCS417r primed A. thaliana plants for an enhanced expression of JA-responsive genes such as VSP2, LOX2, PDF1.2 and HEL after pathogen or insect challenge ( Pozo et al., 2008;van Oosten et al., 2008;Pangesti et al., 2014). Interestingly, all those studies also observed microbial-induced systemic resistance, whereas in our study we did not observe ISR. The logical question then, is whether whenever microbial-ISR is not observed, plants are also not primed for an enhanced defensive response. The fact that we did not observe priming of defense-related genes in the colonized plants might be linked with the slightly reduced leaf damage and consequently less induction on the rhizobacteriacolonized plants. Experiments with an inducer such as MeJA or mechanical damage plus oral secretions could shed more light on this aspect without incorporating differences in induction as a result of insect behavior.
Interestingly, the negative effect of drought on larval performance disappears in the JA-impaired mutant dde2-2 but not in the aba2-1 mutant, which confirms that the effect of drought on herbivores is JA-mediated and not ABA-mediated. The plant response to osmotic stress involves ABA-dependent and ABA-independent signaling (Yoshida et al., 2014), so our results suggest that the plant response to drought that negatively affects herbivores is mainly regulated by ABA-independent mechanisms. In contrast, JA has been proposed to be the core signal that regulates plant resistance to herbivory (Erb et al., 2012), and here we show that it is also a crucial hormone regulating plant resistance under drought stress. Regarding the role of JA and ABA in plant-rhizobacteriainsect interactions, we could not provide evidence for this because rhizobacteria did not affect the herbivore on the wild-type Col-0 plants when analyzing the mutants. Therefore, the question of whether a functional ABAmediated response is required to mount rhizobacteriamediated ISR still remains open.
Plant growth-promoting rhizobacteria are able to promote plant growth and plant defenses. Nevertheless, the outcome of the interaction is variable and drought does not seem to be one of the factors explaining this variability. We first hypothesized that microbes may "help" plants against insects only when plans "need help," drought being one scenario where plants would need such help. We could not provide support for this hypothesis here, a possible reason being that our system included a plant accession and a rhizobacterial strain originating from areas where drought is not common (Lamers et al., 1988;Anastasio et al., 2011). Further work with other systems that have co-evolved in areas suffering from drought stress may find completely different results. Studies addressing additional aspects such as the timing of the interaction or intensity of drought stress would shed light on how the modulators of microbe-plant-insect-interactions function. In addition, plant roots associate not only with 1 bacterium but with a diverse microbial community, the so-called microbiome. There is scant research on the ecological functions of the rhizosphere microbiome, especially on how it modulates the interactions of plants with their biotic and abiotic environment (Philippot et al., 2013). This fundamental knowledge will allow us to advance at using of microbial inoculants in sustainable agricultural practices.