Bacterial phytopathogen infection disrupts belowground plant indirect defense mediated by tritrophic cascade

Abstract Plants can defend themselves against herbivores through activation of defensive pathways and attraction of third‐trophic‐level predators and parasites. Trophic cascades that mediate interactions in the phytobiome are part of a larger dynamic including the pathogens of the plant itself, which are known to greatly influence plant defenses. As such, we investigated the impact of a phloem‐limited bacterial pathogen, Candidatus Liberibacter asiaticus (CLas), in cultivated citrus rootstock on a well‐studied belowground tritrophic interaction involving the attraction of an entomopathogenic nematode (EPN), Steinernema diaprepesi, to their root‐feeding insect hosts, Diaprepes abbreviatus larvae. Using belowground olfactometers, we show how CLas infection interferes with this belowground interaction by similarly inducing the release of a C12 terpene, pregeijerene, and disconnecting the association of the terpene with insect presence. D. abbreviatus larvae that were not feeding but in the presence of a CLas‐infected plant were more likely to be infected by EPN than those near uninfected plants. Furthermore, nonfeeding larvae associated with CLas‐infected plants were just as likely to be infected by EPN as those near noninfected plants with D. abbreviatus larval damage. Larvae of two weevil species, D. abbreviatus and Pachnaeus litus, were also more attracted to plants with infection than to uninfected plants. D. abbreviatus larvae were most active when exposed to pregeijerene at a concentration of 0.1 μg/μl. We attribute this attraction to CLas‐infected plants to the same signal previously thought to be a herbivore‐induced plant volatile specifically induced by root‐feeding insects, pregeijerene, by assessing volatiles collected from the roots of infected plants and uninfected plants with and without feeding D. abbreviatus. Synthesis. Phytopathogens can influence the structuring of soil communities extending to the third trophic level. Field populations of EPN may be less effective at host‐finding using pregeijerene as a cue in citrus grove agroecosystems with high presence of CLas infection.


| INTRODUCTION
Plants, as immobile organisms, are often exposed to multiple stressors.
Stressors can be herbivores, pathogens, parasites and abiotic conditions or some combination of those occurring either simultaneously or sequentially both above-and belowground. Plants cannot actively avoid stressors but can feature innate or induced defensive phenotypes. Enhanced pest tolerance over the lifetime of a plant can depend on physical structures to prevent attacks and internally the production of defensive secondary compounds in response to stressor exposure.
Plant secondary compounds are downstream products of molecular pathway activation and can mediate plant interactions with their biotic and abiotic environments (Ozawa, Arimura, Takabayashi, Shimoda, & Nishioka, 2000;Pieterse & Dicke, 2007). With specific regard to insect herbivores which feed on the plant, many of these compounds are toxic and repellant, but when released as volatiles can also be utilized by herbivore predators and parasites to locate hosts (Mumm & Dicke, 2010b). Based on the metabolic cost implied in the production of secondary metabolites, many studies have investigated variation in production by classifying their modes of production as "constitutive" or "inducible" despite evidence that these defenses are not mutually exclusive (e.g., Zangerl & Berenbaum, 1990). There has been a larger focus on induced chemistry likely because it represents a clear response to attack.
Throughout the plant, induced responses can be produced locally at the feeding site or systemically (Bezemer & Van Dam, 2005; Halitschke, Kessler, Sardanelli, & Denno, 2008). The production of induced defenses is activated and regulated by phytohormone pathways such as jasmonic acid, salicylic acid, and ethylene (Pieterse & Dicke, 2007). Phytohormones regulate a suite of induced plant responses, which include the production herbivore-induced plant volatiles (HIPVs; Walling, 2000). Pathways that produce HIPVs are highly conserved among plants and are often activated differentially depending on the type of damage incurred by plants (Zipfel, 2014). For example, insect feeding guild (piercing and sucking versus chewing) influences which pathways are activated (Walling, 2000;Wu & Baldwin, 2009;Zarate, Kempema, & Walling, 2007).
Plants also affect the phytobiome by changing the surrounding volatile profile both above-and belowground (Dicke, 2016). The ability of plants to facilitate attraction of herbivore predators and parasites with HIPVs has been shown in multiple plant species. Predators and parasites can exhibit top-down control of insect herbivore populations above-and belowground which can regulate the overall impact of herbivory on plants (van Dam, Qiu, Hordijk, Vet, & Jansen, 2010;Dicke & Baldwin, 2010;Kessler & Baldwin, 2000;Preisser, Dugaw, Dennis, & Strong, 2006). Changes in volatile profile due to insect feeding can influence plant-beneficial tritrophic interactions by attracting tertiary predators and parasites of insect herbivores which is known as indirect defense (Mumm & Dicke, 2010a;Price et al., 1980). This has been widely investigated in aboveground environments (e.g., De Moraes, Mescher, & Tumlinson, 2001;Meiners & Hilker, 1997) and more recently in belowground environments (Ali, Alborn, & Stelinski, 2010;Rasmann et al., 2005).
There has been growing interest in exploiting such multitrophic interactions for biological pest control in agriculture (Bommarco, Kleijn, & Potts, 2013). For belowground pest species, the interest has been in exploiting the attraction of entomopathogenic nematodes (EPN) to damaged plants for location and infection of root-feeding herbivores. This tritrophic interaction has been shown in multiple systems (Aratchige, Lesna, & Sabelis, 2004;van Tol et al., 2001). Studies using maize and citrus have demonstrated this interaction both in the laboratory (Ali et al., 2010;Rasmann et al., 2005) and directly in agroecosystems Degenhardt et al., 2009). However, a central unknown of manipulating a natural environment is how the interactive effects of multiple plant stressors impact belowground tritrophic interactions. Agricultural systems, while disturbed, are useful model systems for investigating the effects of multiple stressors on plants and their related interactions because of the redundancy of pests and pathogens and, thus, the predictability of the species occurring in these environments. We used the citrus huanglongbing (HLB) pathosystem to investigate the effects of a phloem-limited bacterial infection on a belowground tritrophic interaction mediated by a HIPV, pregeijerene.
Huanglongbing is caused by the gram-negative bacterium, Candidatus Liberibacter asiaticus (CLas), which is transmitted by the Asian citrus psyllid (ACP; Diaphorina citri). HLB is the most damaging disease of cultivated citrus worldwide (da Graca et al., 2016). The vector is attracted to infected trees following induced changes in tree volatile production, which may promote spread of disease (Mann et al., 2012;Mauck, Moraes, & Mescher, 2016). CLas is quickly moved throughout the plant via phloem translocation; the roots have the third highest accumulation of the bacteria of all plant tissues (Tatineni et al., 2008). This suggests that activation of defense pathways may occur throughout the plant, including in the roots, which may influence HIPVs released from the plant upon secondary attack.
In the citrus rhizosphere, the C12 terpene compound, pregeijerene, attracts soil-dwelling infective juveniles (IJs) of multiple EPN species to feeding D. abbreviatus larvae. These larvae chew on the plant's roots, which stimulates an induced release of pregeijerene into the soil . EPNs are attuned to HIPVs as cues in their specific plant system for host-finding through exposure over time (Hiltpold, Baroni, Toepfer, Kuhlmann, & Turlings, 2010;Rasmann, Ali, Helder, & van der Putten, 2012;Willett, Alborn, Duncan, & Stelinski, 2015). We hypothesized that CLas infection may impact belowground release of HIPVs and investigated how this may impact EPN response to feeding D. abbreviates larvae in the citrus root zone.

| Insects
Noninfected adult D. citri used to feed on plants as a damage treatment were obtained from a laboratory culture at the University of Florida, Citrus Research and Education Center (Lake Alfred, USA). The culture was established from field populations in Polk Co., Florida, USA (28.09N, 81.99W), before the local discovery of HLB in Florida.
D. citri colonies were kept in double-screened, 3.764-m secure enclosures in an air-conditioned glasshouse at 27-28°C, 60%-65% RH, and L14:D10 photoperiod. Bimonthly testing of randomly sampled adults and plants by quantitative PCR (qPCR) assays was conducted to confirm that psyllids and plants in this culture were uninfected by CLas.
Diaprepes abbreviatus and Pachnaeus litus larvae were obtained from a laboratory colony maintained at University of Florida's Citrus Research and Education Center (CREC) in Lake Alfred, FL. This culture was periodically supplemented from a larger colony maintained at the Division of Plant Industry Sterile Fly Facility in Gainesville, FL. Larvae were reared on a commercially prepared diet (Bio-Serv, Inc., Frenchtown, NJ, USA) using procedures described by Lapointe and Shapiro (1999). Larvae used in experiments were from third to sixth instars.

| Nematodes
The entomopathogenic nematode, S. diaprepesi HK3, was initially isolated from D. abbreviatus larvae in Florida commercial citrus orchards.
S. diaprepesi was reared in larvae of the greater wax moth, Galleria mellonella, at approximately 25°C according to procedures described in Kaya and Stock (1997). Infective juveniles emerged from insect cadavers into emergence traps and were then transferred to and stored in shallow water in tissue culture flasks at 15°C in an incubator for up to 2 weeks before use in experiments. times. Plants referred to as damaged were exposed to larval feeding for 48 hr prior to assays.

| Effect of CLas infection on subterranean hostseeking behavior of herbivores and their parasites
In order to unravel the interactive effect of CLas infection in plants on belowground first-and second-level trophic interactions, a series of experiments was performed using glass, soil-filled olfactometers described in detail by Ali, Alborn, and Stelinski (2011). In brief, six-arm olfactometers (Ali et al., 2011) were modified into four-choice chambers by closing off two of the six arms. Glass olfactometers were further modified by removing all mesh screens to allow entomopathogenic nematodes access to all parts of the olfactometer. Treatment plants were placed in the distal portion of all arms of the olfactometer along with a single fifth-instar D. abbreviatus larva in certain treatments. Aliquots consisting of either 30 μl pentane (blank control) or various solutions of pregeijerene (described below) were placed on filter paper and added to the root zone in each of the two arms. The olfactometer was then filled with washed autoclaved silica sand adjusted to 10% moisture by volume.

| Weevils: Diaprepes abbreviatus and Pachnaeus litus
Root herbivores, D. abbreviatus and P. litus, were used to assess changes in belowground herbivore assays in response to plants with CLas infection and/or herbivore feeding damage. A single weevil larva of either species was released and allowed 24 hr to choose between the two treatments. At the end of 24 hr, choice was designated by the presence of the larva in the chamber at either end of the olfactometer. To exclude directional bias or possible inherent preference between plants, the initial olfactometer treatment compared beetle response between two undamaged and uninfected plants for both species.

| Entomopathogenic nematode: Steinernema diaprepesi
To assess the interactive effect of CLas infection and insect damage on the attraction of EPN to their insect hosts, 2,500 S. diaprepesi infective juveniles (IJs) were released into a two-choice setting containing various sets of two treatments. Nematodes were allowed to forage for 24 hr before the olfactometer was disassembled. Nematodes were extracted from the substrate over 48 hr using Baermann funnels and counted using a dissection microscope. The treatments investigated mirrored those tested with D. abbreviatus larvae. The treatments over-

| Infestation and infection
Six-arm olfactometers were modified to create a four-choice design by closing off access to two of the six arms, as described in Ali et al. (2010), and only attaching four outer chambers containing treatments, as described above. Treatments were placed randomly among the four arms. D. abbreviatus larvae were placed into stainless steel screen cages (1-cm 2 cubes, 225 mesh) that prevented feeding on roots but allowed entry of nematodes; there was one caged larva per arm. At the onset of the experiment, 2,500 S. diaprepesi IJs were released into the central chamber and allowed to forage for 48 hr. Two experimental designs were replicated 20 times each. In both experiments, the caged larva in each treatment arm was combined with a treatment applied to the plant or a blank. The first experiment consisted of four treatments paired with a caged larva: an uninfected plant, a plant with CLas infection, and two blanks with nothing added to these treatment arms except for the caged larvae. In the second experiment, nematode response was tested to a CLas-infected plant as compared with an uninfected plant that was actively damaged by D. abbreviatus larval feeding. Two additional blank (negative control) treatment arms completed the four possible directions nematodes could forage. Nematodes were collected and counted as described above. Also, caged larvae were collected to determine infestation by EPNs as measured by subsequently emerging IJs and larval mortality as described in Ali et al. (2012).

| Dose-response
A dose-response curve was developed to determine the most active dosage of pregeijerene to attract D. abbreviatus larvae under our olfactometer experiential conditions. Pregeijerene was obtained as described in Ali et al. (2011). In the olfactometer, the larvae were given a choice between a dichloromethane blank and one of five logarithmically doses of pregeijerene dissolved in this solvent. Volatiles were released from filter paper as described in Ali et al. (2011). Larvae were scored as making a choice when they moved into either of the chambers containing treatments in the olfactometer 48 hr after release. Larvae were collected by disassembling olfactometers and sifting removed sand in the radiating treatment arms.

| Induced release of pregeijerene
Volatiles were collected from the following plants to assess the presence or absence of pregeijerene: an HLB-infected plant, an uninfected plant damaged by D. abbreviates larval root feeding as described above, a plant with both CLas infection and D. abbreviatus damage, and for comparison, an uninfected and uninfected plant, and volatile samples of the substrate itself. In each case, the presence of pregeijerene was verified with an authentic standard (CAS 020082-17-1).
Each treatment was replicated five times.
Baseline volatile production was determined by initially sampling the volatiles from the roots of both plants three days before treatment was applied for comparison. On day 4, plants were infested with one larva at the root zone. Noninfested plants were not exposed to weevils during this period.
Plants were potted in sand-filled glass root-zone chambers as previously described in Ali et al. (2011). Seedlings were given 3 days to adjust to their sand-filled chambers. Infested plants were subjected to an additional 3 days of feeding by weevil larvae. The belowground chambers of each infestation type were simultaneously sampled for three subsequent days after infestation. Root feeding damage by beetle larvae was confirmed visually as described in Ali et al. (2011).

| Statistical analyses
In olfactometer experiments where two-choice responses of beetle larvae were investigated, McNemar's tests were used to analyze data.
Data from two-choice experiments with entomopathogenic nematode response as the dependent variable were converted to percent response. These data were then analyzed using paired t tests after data were analyzed for normality. Both McNemar's test and the t tests

| D. abbreviatus infestation and infection
More D. abbreviatus larvae were infected by EPNs when in the presence of CLas-infected than uninfected plants (χ 2 = 9.68; df = 3; p = .0079; Figure 6a). Nematode infection of D. abbreviatus larvae was similar in the presence of plants infected with CLas that received no additional feeding damage from D. abbreviatus and those that were uninfected by pathogen, but that were actively infested with larval D. abbreviatus; however, both of these treatments resulted in greater EPN infection of D. abbreviatus larvae than blank controls (χ 2 = 15.16; df = 3; p = .0041; Figure 6b).

| Optimal dosage of pregeijerene for attraction of D. abbreviatus
Attraction of D. abbreviatus to pregeijerene did not increase proportionally in a stepwise pattern with log dosage (χ 2 = 0.5880; df = 4; p = .4430).

| Induced release of pregeijerene
As shown previously (Ali et al., 2010), pregeijerene release from plant roots damaged by D. abbreviatus larvae was confirmed (Table 1).
Pregeijerene was also emitted from plants infected with CLas regardless of additional D. abbreviatus larval damage (Table 1). α-Santalene and α-Z-bergamotene were the only two other volatiles detected in Diaprepes-damaged and CLas-infected plants (data not shown); however, these do not affect EPN behavior (Ali et al., 2010).

| DISCUSSION
We investigated how multiple stressors impact plant response and possibly influence the structure of plant-associated communities.

Previous investigations of belowground tritrophic interactions in-
volving entomopathogenic nematodes (EPNs) used undamaged, disease-free plants in various artificial environments where multiple stressors were least likely to be present. The intention of these investigations was to isolate the interaction of interest, which has been extremely important in unraveling these multitrophic interactions Filgueiras, Willett, Junior, & Pareja, 2016;Rasmann et al., 2005). However, multiple members of the phytobiome contribute to the fluctuation, be it disruption or amplification, of a trophic cascade in a natural environment (Dicke, 2016). The experiments presented here show how a subtropical tree's response to infection with a bacterial pathogen, CLas, affects interactions in the phytobiome extending to the third trophic level, eliciting what was previously thought to be solely an HIPV response to root damage in citrus (Ali et al., 2011).
Both root-feeding herbivores and EPNs were more attracted to CLas-infected roots than to healthy plants in this investigation (Figures 1 and 4). Also, nonfeeding D. abbreviatus larvae were more likely to be infected with EPNs when occurring adjacent to CLasinfected tree roots than roots from uninfected trees ( Figure 6). Based on our results, this attraction appears to be linked to the inducible release of pregeijerene from infected plant roots (Table 1). This cue has been previously shown to mediate EPN host location in this system, affecting the behavior of a broad spectrum of EPN species (Ali et al., 2010(Ali et al., , 2011. The results presented in this study indicate that this HIPV is also released upon plant stress unrelated to insect feeding.
Here, we discuss the implications of this finding with regard to belowground tritrophic interactions and the possible causes of "redundant" plant response to distinct stressors.
Information transmitted via chemicals in the soil depends on the ability of the receiver to select relevant information among the noise (Jablonka, 2002). EPNs use many cues to forage for their hosts in the soil environment; cue is used here because of mounting evidence of the adaptability of EPNs to quickly learn from infochemicals Willett et al., 2015). The most reliable of these hostlocation cues is emission specific and related to the feeding or absolute presence of potential host insects (Turlings, Hiltpold, & Rasmann, 2012). HIPVs are released in response to insect feeding and are used by EPNs as host-location cues in the soil; this has been shown in multiple systems. However, in ancestral hybrid rootstocks, pregeijerene is released constitutively rather than as an induced response observed in cultivated rootstocks (Ali et al., 2011). Furthermore, in nature citrus does not occur in monoculture and disease incidence is likely infrequent (Liu et al., 2012 ( F I G U R E 7 Response of Diaprepes abbreviatus larvae to five logarithmically increasing doses of pregeijerene. A two-choice olfactometer was used to assess response of larvae. Experiments with each dose were replicated 30 times. The mean number of larvae moving toward the blank was 1.8. The optimal dose for D. abbreviatus attraction was 0.1 μg/μl (z = 3.011; df = 4; p = .0026). Total numbers of larvae per treatment without an asterisk did not differ significantly (α = .05) Alternatively, given that the herbivores were also attracted to infected plants, the redundancy of this cue may be counterbalanced in citrus groves with high CLas infection through increased dependence on pregeijerene-alternative cues relating to an insect's location such as carbon dioxide or scents from insect frass. Furthermore, it is possible that nematodes responded to other signature cues released by infected trees, in addition to pregeijerene, which were undetected here.
Also, differential attraction to this cue based on relative ratios of pregeijerene to its degraded form, geijerene (Ali et al., 2010), may indicate presence of the herbivore over a time course following initiation of herbivory, which is a hypothesis that warrants further testing.
Consistent with aboveground studies demonstrating vector attraction to infected plants (Mann et al., 2012;Mauck, De Moraes, & Mescher, 2010) and investigations with Arabidopsis indicating benefit to the herbivore from bacterial infection of the plant (Cui et al., 2002;Groen et al., 2013), the root herbivore species investigated here were also more attracted to infected plants than to uninfected plants. This response appeared to be mediated by pregeijerene (Table 1), which was attractive alone to larval weevils, when presented at an optimal dosage ( Figure 7). We postulate that the root herbivores are attracted to pathogen-infested plants because such plants are weakened by bacterial infection. The CLas pathogen is phloem-limited (da Graca et al., 2016) and is known to specifically accumulate in the roots following inoculation into the plant by the psyllid vector (Johnson, Wu, Bright, & Graham, 2013). It is also possible that there is a nutritional benefit to herbivores that feed on HLB-infected plants given accumulation of callose in the phloem of diseased trees (Kim, Sagaram, Burns, Li, & Wang, 2009) and/or weakened defense response (Nwugo, Duan, & Lin, 2013). This may impact the phytobiome by resulting in development of higher quality insect hosts for EPNs to generate greater numbers of progeny, as occurs with infected cadavers (Barbercheck, Wang, & Hirsh, 1995;Hazir et al., 2016). It may eventually become detrimental to EPN populations due to a coinciding increased incidence of nematophagous species (Duncan et al., 2007) if these members of the community become more associated with pathogeninfected than uninfected plants ( Figure 6). This hypothesis deserves future investigation.
Why would the plant coincidentally release the same volatile organic compound in response to a bacterial pathogen as occurs due to herbivore feeding? It is possible the compound serves other purposes in the root system not explored here, such as antimicrobial activity (e.g., Huang et al., 2012). The lack of specific recognition of CLas infection versus herbivore feeding by the plant may be caused by citrus cultivation practices, or pregeijerene release may be a more general stress response by citrus than previously thought. The use of pregeijerene as an HIPV host-seeking cue by EPNs may have evolved following cultivation of citrus in monoculture 360 years ago (Liu et al., 2012). The inducible release of pregeijerene following herbivore damage in the commercial hybrid rootstock investigated here and constitutive release from the hybrid's parent (Ali et al., 2011) (Goldschmidt, 2014;Heil et al., 2012). There is evidence, however, that shoot defenses appear to vary in a rootstock-dependent manner (Agut, Gamir, Jaques, & Flors, 2016). Further investigation is warranted to determine how various scion and rootstock combinations may differentially respond to herbivore damage versus bacterial pathogen infection.
Plants express traits that can attract indirect defenses, such as EPNs, which in turn affects other members of their community. The result of these cascading interactions may benefit the plant so as to limit damage from herbivory (Mumm & Dicke, 2010a). It is reasonable to assume that selection pressure should favor these traits and they may evolve in agroecosystems much more rapidly than in nature as plants are forced into monoculture; redundancy of trait expression throughout an area may allow for quicker attuning to the defense cry by third-trophic-level predators and parasites (Dicke & Baldwin, 2010 HIPV. It is therefore unclear whether the plant utilizes this as a specific cue for defense against herbivore feeding or whether it is an incidental cue released without the "signaling" specificity that was previously postulated (Ali et al., 2010).
In conclusion, it is becoming clear that at least in the citrus model system, EPNs respond to many induced plant cues (Filgueiras et al., 2016). Their adaptability may be due to behavioral plasticity guided by both intra-and interspecific communication and short-and long-term memory (Willett et al., 2015). EPNs are therefore able to tailor their behavioral responses to dynamic changes in the subterranean environment and respond to ephemeral cues associated with their hosts rather than relying upon "signature" cues previously thought to be consistently associated with their hosts. Future research should continue to investigate the effect of stressor complexes to seek patterns within these multilayered interactions.

This work was funded by a University of Florida Research Foundation
Professorship award to LLS.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
LLS and KSP-S collected the data. KSP-S provided materials. MJR and LLS wrote the manuscript. MJR and XM analyzed the data. All authors contributed to study design and interpretation.