Plant‐phenotypic changes induced by parasitoid ichnoviruses enhance the performance of both unparasitized and parasitized caterpillars

Abstract There is increasing awareness that interactions between plants and insects can be mediated by microbial symbionts. Nonetheless, evidence showing that symbionts associated with organisms beyond the second trophic level affect plant‐insect interactions are restricted to a few cases belonging to parasitoid‐associated bracoviruses. Insect parasitoids harbour a wide array of symbionts which, like bracoviruses, can be injected into their herbivorous hosts to manipulate their physiology and behaviour. Yet, the function of these symbionts in plant‐based trophic webs remains largely overlooked. Here, we provide the first evidence of a parasitoid‐associated symbiont belonging to the group of ichnoviruses which affects the strength of plant‐insect interactions. A comparative proteomic analysis shows that, upon parasitoid injection of calyx fluid containing ichnovirus particles, the composition of salivary glands of caterpillars changes both qualitatively (presence of two viral‐encoded proteins) and quantitatively (abundance of several caterpillar‐resident enzymes, including elicitors such as glucose oxidase). In turn, plant phenotypic changes triggered by the altered composition of caterpillar oral secretions affect the performance of herbivores. Ichnovirus manipulation of plant responses to herbivory leads to benefits for their parasitoid partners in terms of reduced developmental time within the parasitized caterpillar. Interestingly, plant‐mediated ichnovirus‐induced effects also enhance the performances of unparasitized herbivores which in natural conditions may feed alongside parasitized ones. We discuss these findings in the context of ecological costs imposed to the plant by the viral symbiont of the parasitoid. Our results provide intriguing novel findings about the role played by carnivore‐associated symbionts on plant‐insect‐parasitoid systems and underline the importance of placing mutualistic associations in an ecological perspective.


| INTRODUC TI ON
Plants are at the basis of most terrestrial food webs and interact with various organisms in nature, including herbivorous and carnivorous insects (Schoonhoven et al., 2005;Stam et al., 2014;Turlings & Erb, 2018). In recent years, there has been a rapidly growing body of evidence showing that plant-insect interactions can be mediated by a large variety of microbial symbionts acting as "hidden" players (Mason et al., 2019;Pineda et al., 2017;Shikano et al., 2017). For instance, beneficial soil microbes such as plant growth-promoting bacteria and fungi can enhance plant resistance against herbivore attack (Pineda et al., 2010;Pozo & Azcón-Aguilar, 2007). On the other hand, microbes associated with herbivorous insects may help their hosts to exploit plants with unbalanced nutritional value (Douglas, 2015) or to counteract plant defences (Chung et al., 2013).
Carnivorous organisms such as insect parasitoids also harbour a diversity of symbionts including bacteria, fungi and a wide array of viruses influencing plant-insect interactions (Dicke et al., 2020). Yet, the effects of carnivore-associated symbionts are far less investigated within a plant-insect interaction perspective compared with plant-and herbivore-associated symbionts .
Polydnaviruses are abundant and unique symbionts associated with larval endoparasitoids of the Ichneumonoidea (Braconidae and Ichneumonidae). Virus particles are produced exclusively in the calyx region of the ovary of parasitoid females from a proviral template. They are stored in the wasp oviducts and then injected by parasitoid females into a caterpillar host. Injection of polydnaviruses allows survival of parasitoid offspring within the herbivorous host by impairing host immune response and by altering host development and metabolism (Beckage, 2012;Burke & Strand, 2014;Lu et al., 2010;Strand et al., 2006;Webb et al., 2006). Two genera of polydnaviruses are defined, the Bracoviruses associated with braconid wasps and the Ichnoviruses associated with ichneumonid wasps (Francki et al., 1991).
To date, the only available information showing that polydnaviruses affect plant responses to herbivory is restricted to two plantherbivore parasitoid systems, both of which focused on bracoviruses (Microplitis croceipes bracovirus McBV, Cotesia glomerata bracovirus CgBV) Tan et al., 2018;Zhu et al., 2018).

Interestingly bracovirus manipulations in the caterpillars have been
shown to change the composition of herbivore oral secretions (saliva and/or regurgitate), which often contain the elicitors exploited by plants to recognize insect damage (Bonaventure, 2012;Bonaventure et al., 2011;Rivera-Vega et al., 2017). Indeed, the activity of glucose oxidase and β-glucosidase is reduced in oral secretions of Helicoverpa zea and Pieris brassicae caterpillars injected with bracoviruses isolated from their respective endoparasitoid species, M. croceipes and C. glomerata. The consequence of bracovirus-induced manipulations of caterpillar oral secretion is that plants downregulate defencerelated genes and reduce their chemical defences Tan et al., 2018). In turn, bracovirus-mediated changes in plant quality benefit the growth of the caterpillar in which the parasitoid larvae develop, thus increasing the fitness of the offspring of the bracovirus-injecting parasitoid female (Tan et al., 2018). These discoveries have opened a novel scenario in plant-insect interactions showing that, although third-trophic level symbionts do not come directly in contact with the plant, they can still induce changes in plant phenotype, mediated by their effects on the infected caterpillars.
In addition to bracoviruses, other groups of parasitoid viruses are likely to affect plant-insect interaction, since herbivory by parasitized caterpillars induces specific plant responses and parasitoid identity typically override the identity of the herbivore attacker (Cusumano et al., 2019;Poelman, Zheng et al., 2011;Zhu et al., 2015). We are only now starting to explore the diversity of parasitoid symbionts and the underlying mechanisms behind their interactions with the plants Dicke et al., 2020;Shikano et al., 2017). From a molecular perspective, our understanding about the way parasitoid-associated viruses manipulate caterpillar oral secretions is largely incomplete. Even though transcripts for a specific subset of bracoviruses genes have been detected in a host salivary gland (Bitra et al., 2011), it remains unclear if viral-encoded proteins are actually present in caterpillar oral secretions. Furthermore, we are not aware of the full range of viral-induced manipulations in caterpillar salivary glands, which caterpillar-encoded proteins are upregulated and which ones are downregulated upon viral infection. The ecological effects that parasitoid-associated viruses induce at the plant-insect interface require further investigation as well. For example, an interesting but unexplored hypothesis is that other unparasitized herbivores also benefit from the increase in plant quality induced by parasitoidassociated symbionts. The most common ecologically relevant scenario to test such hypothesis is to investigate whether unparasitized caterpillars take advantage from feeding on plants concurrently attacked by parasitized caterpillars (which thus are infected by parasitoid-associated symbionts). In fact, parasitism levels in nature rarely reach 100% which means that usually not all herbivores feeding on a plant are parasitized.
The aim of the current study was to investigate for the first time whether ichnoviruses, which are associated with thousands of parasitoid species, affects the proteome of caterpillar oral secretions with cascading plant-mediated interactions between parasitized and unparasitized caterpillars. Using as model species the solitary parasitoid Hyposoter didymator which carries H. didymator ichnovirus (HdIV), we investigated the role of ichnoviruses as hidden players in the interaction between corn (Zea mais) and the fall armyworm Spodoptera frugiperda. We experimentally manipulated the phenotype of herbivores by isolating calyx fluid (containing viral particles) from wasp females and injecting it into caterpillars subsequently feeding on corn plants. We specifically investigated: (1) whether ichnoviruses are responsible for the majority of the changes occurring in the salivary glands of naturally parasitized caterpillars by using a comparative proteomic approach, (2) whether the performance of unparasitized caterpillars increased when feeding on plants previously treated with insect saliva collected from caterpillars that had been injected with parasitoid calyx fluid containing the ichnovirus and (3) whether parasitoid offspring benefitted from plant-mediated ichnovirus-induced manipulation.

| Plants and insects
Corn plants (line B73 HT) were obtained from organic seeds at the Diascope experimental research station (INRA,Mauguio,France,43°36'37"N,3°58'35"E). Plants were grown in plastic pots (7 × 8 cm) in a climatic chamber at 25 ± 2°C, 60% RH and 16:8 h (L:D) and used for the experiments when they were two weeks old with four fully developed leaves. Plants were allowed to acclimatize under laboratory conditions at the insect quarantine platform PIQ (University of Montpellier, DGIMI laboratory) 3-5 days before the experiments.
The corn strain of Spodoptera frugiperda was maintained at 24 ± 2°C, 65% RH and 16:8 h (L:D) on a semisynthetic corn-based diet (Poitout et al.,1972). The parasitoid Hyposoter didymator was maintained on S. frugiperda larvae in the same abiotic conditions, using second to third instar larvae for parasitism. S. frugiperda is not naturally present in France and is considered as a quarantine pest.
Consequently, experiments described hereafter were conducted in a confined environment at the DGIMI insect quarantine platform (PIQ).

| Isolation of ichnovirus particles from H. didymator and injection into S. frugiperda caterpillars
Calyx fluid (containing the ichnovirus particles) was extracted from H. didymator wasps anaesthetized on ice and dissected in phosphatebuffered saline (PBS, 1x, pH 7.4 Fischer Life Technology) under a light microscope. The calyx region of the ovaries was collected in 250 µl PCR tubes. The volume was adjusted with PBS to reach the desired concentration in female equivalents (f.e.) as described in Dorémus et al. (2013) (for example, ovaries from 30 wasps pooled in 60 µl of PBS for injection of 100 nl containing 0.05 f.e./caterpillar). A concentration of 0.05 f.e. was selected based on preliminary investigations that showed a consistent effect on the phenotype of injected caterpillars (i.e., a reduction in the weight of injected caterpillars two days post injection compared to PBS-injected caterpillars, Doremus et al., 2013). Calyx tissues were disrupted by several passages through a 20 µl micropipette cone. Tubes containing the disrupted biological material were centrifuged for 5 min at 5000 rpm and then supernatant containing the calyx extracts was stored on ice until injections into third instar S. frugiperda caterpillars were carried out (as described below). Viral purification by centrifugation has been shown to have similar effects on caterpillar physiology as other purification techniques such as filtration or using a gradient (Beckage et al., 1994). Presence of viral particles in calyx extracts was confirmed by negative staining and observation under an electron microscope Zeiss EM10CR at 80 kV.  (Dorémus et al., 2013).
After microinjections, the caterpillars that recovered within 1 h were allowed to feed on corn plants for two days before using them for experiments (proteomic investigations of caterpillar salivary glands and performance bioassays). This time window was selected because two days post injection (p.i.) the parasitoid progeny in parasitized caterpillars is still at the egg stage, so it is possible to exclude any physiological effect on caterpillar phenotype due to the feeding by wasp larvae.

| Salivary gland dissection, SDS-PAGE and quantitative proteomic analyses
To investigate whether different injection treatments affect the full protein composition of caterpillar salivary glands, we carried out comparative proteomic investigations. Labial salivary glands from two days p.i caterpillars (treatments CF, PBS and PAR) were processed as described by Celorio-Mancera et al. (2012). Briefly, glands were first dissected in cold PBS, then rinsed with cold PBS and placed in a droplet of 20 µl of cold PBS per pair of glands on a Petri dish kept on ice. Glands were cut in half inside the droplet, and subsequently transferred along with the buffer solution to 1.5 ml Eppendorf tubes. Glands of five individuals were pooled as a biological replicate. The samples were centrifuged for 5 min at 17,530 g at 4°C. The supernatant was transferred to new Eppendorf tubes, and protein concentration in the samples was quantified by Bradford spectrophotometric assay (Bradford, 1976). A preliminary assessment of the complexity of the salivary protein profiles between PAR-, PBS-and CF-treated caterpillars was obtained by separating the proteins using 4%-12% SDS-PAGE precast mini gels (Biorad).
Search parameters were default parameters with slight modification.
Briefly, first search precursor mass tolerance was set to 20 ppm, and main search was set (after recalibration) to 6 ppm. A maximum of two missed-cleavages was allowed. Search was performed allowing variable modifications: Oxidation (Met), Acetyl (Nterm) and with one fixed modification: Carbamidomethyl (Cys). False discovery rate (FDR) was set to 0.01 for peptide and proteins, and minimal peptide length to seven. Quantification was also performed using maxquant with standard parameters. Graphical representation and statistical analysis were done using perseus (Tyanova et al., 2016, v1.6.10.43) using standard workflow (reverse and contaminant entries removing, filtering based on number of valid value: at least three in one group, and then imputation using "Replace missing values from normal distribution" tool from perseus). All t tests were performed using a FDR of 5% and s0 of 0.1 (Tusher et al., 2001).

| Performance of parasitized and unparasitized caterpillars
To evaluate if plant induction with salivary gland extracts from differently injected caterpillars (PAR, CF or PBS) affects the performance of the herbivores, we carried out a relative growth rate experiment.
Using salivary gland extracts for plant induction treatments allows Each plant was damaged on the first fully expanded leaf using a pattern wheel. The wheel was rolled over the leaf surface on each side of the midrib, two lines in parallel (length 3 cm distance between each other of 0.75 cm) creating a c. 2.25 cm 2 area with 20 tiny holes (~0.5 mm 2 ). A total of 20 µl of salivary gland extract, prepared as described before, from differently injected S. frugiperda caterpillars was applied to the tiny holes on these mechanically damaged leaves. There were two groups of third instar caterpillars (parasitized or unparasitized) which were feeding on (1)   F I G U R E 3 (Left) Total number of proteins found in salivary glands of Spodoptera frugiperda caterpillars: bars represent proteins significantly more abundant (red), less abundant (green) or not significantly different (white) in the virus-infected (CF) and parasitized (PAR) treatments compared with saline-injected controls (PBS). (Right) direct comparisons of the proteomic changes between CF and PAR treatments in the subset of proteins that displayed a significant increase in abundance (top venn diagram) or a significant decrease in abundance (lower venn diagram) in the previous comparison with PBS. Light grey colour indicates proteins shared in the CF and PAR treatments; medium grey colour indicates unique proteins of CF treatment; dark grey colour unique proteins of the PAR treatment [Colour figure can be viewed at wileyonlinelibrary.com] TA B L E 1 List of the subset of proteins detected in the salivary glands of Spodoptera frugiperda caterpillars with decreased (green) or increased (red) abundance in the presence of HdIV virions (i.e., both treatments with calyx fluid-injected and parasitized caterpillars) compared to saline-injected controls (PBS). Proteins shown in this list have a cutoff score in terms of intensity (log2 LFQ differences) >|1| and a predicted signal peptide. Numbers ahead of protein IDs refer to numbers indicated in Figure 2 Protein were carried out. The bioassays were performed in blocks, with the parasitism status of the feeding herbivore as the block factor.
Treatments were randomized within each block.
Twenty-four hours after application of salivary gland extract, the treated leaf was collected for the caterpillar feeding bioassay.
Parasitized and unparasitized caterpillars were weighed and then fed on the treated corn leaves in plastic tubes closed with cotton wool and lined with 2% agar to keep leaves moist. Forty-eight hours later, caterpillars were reweighed and relative growth rate was calculated as follows: (final weight − initial weight)/(average weight).

| Performance of parasitoids
To determine if the performance of parasitized caterpillars also influences development of the endoparasitoids, we conducted a parasitoid performance experiment. Third instar S. frugiperda caterpillars were parasitized by H. didymator and fed leaves from plants treated with salivary gland extract from PBS-injected unparasitized caterpillars (PBS) or CF-injected caterpillars (which are infected with HdIV) as described above. Twenty-four hours after treatment, the treated leaf was collected and placed in a plastic tube closed with cotton wool and lined with 2% agar to keep leaves moist. Treated leaves were replaced every other day to keep food fresh until parasitoid cocoon formation.
To assess the performance of the wasps we recorded: (1) developmental time (time from wasp parasitization until cocoon formation), (2) developmental mortality (proportion of wasps that yield a cocoon out of the total number of parasitized caterpillars), and (3) cocoon weight (recorded the second day after its formation). and analysed with ANOVA. Developmental time data of parasitoids were analysed with a general linear model (GLM) with gamma error distribution and inverse link function. Mortality data were analysed with a GLM with binomial error distribution and logit link function.

| Statistical analysis
Cocoon weight data were normally distributed and analysed with linear models. Significance of the factors in the GLMs was assessed using likelihood ratio tests (LRTs) (Crawley, 2007). Significance levels for factors in the linear model were derived directly from F-tests.
The adequacy of the statistical models was assessed with residual plots (Crawley, 2007). ANOVA and GLMs have been carried out with r statistical software (R Development Core Team, 2013).  less abundant, and 137 out of 291). As expected from the high correlation between PAR and CF samples ( Figure S1), only minor variations were observed between these two treatments: all protein entries except one (GSSPFG00005192001-PA, Table S1) had levels similarly affected in either PAR or CF samples ( Figure S2).

| Nature of S. frugiperda salivary gland proteins affected by parasitism or HdIV infection
Out of the 624 proteins differentially expressed by either parasitism or calyx fluid injection, a subset of 335 S. frugiperda entries corresponded to proteins with levels significantly affected with at least a two-fold change (log2Dif>|1|). From that, 112 proteins were more abundant in saliva from PAR-or HdIV-infected caterpillars and 223 proteins were more abundant in saliva from PBSinjected control caterpillars (Table S1) Table 1). To this list of potential effectors in insect-plant interactions, we also added an entry corresponding to glucose oxidase (GOX; GSSPFG00008369001-PA), which despite the absence of a predicted signal peptide, is described in the literature as a major lepidopteran salivary protein and a herbivory-associated elicitor (Chen & Mao, 2020;Rivera-Vega et al., 2017).
The proteome of salivary glands from parasitized or calyx fluidinjected caterpillars contain two related viral proteins, GlyPro1 and GlyPro2, that both belong to the glycine and proline rich protein family. This finding strongly suggests that the salivary gland tissue is actually infected by the parasitoid symbiont HdIV. Among the proteins whose levels are increased in PAR and CF samples compared to PBS (Table 1), we found five putative lipases, several hydrolases, the two prophenoloxidases, and a nucleotidase (apyrase). Three serine protease inhibitors seem also affected by parasitism or calyx fluid injection as well as a mucin-like protein and a fibrohexamerin-like protein.
Conversely, proteins less abundant in PAR and CF samples include a number of enzymes belonging to various classes. The most strongly affected appear to be five different carboxylesterases TA B L E 2 Performances of Hyposoter didymator parasitoids that developed into Spodoptera frugiperda caterpillars feeding on corn leaves previously induced with salivary gland extract from caterpillars injected with phosphate-buffered saline (PBS) or caterpillars injected with calix fluid (CF) containing HdIV virions. Developmental time (days) is recorded from oviposition to cocoon formation; Cocoon weight (mg) is recorded the second day after its formation as fresh weight; developmental mortality (%) is calculated as the proportion of wasps that yield a cocoon out of the total number of parasitized caterpillars. For each performance determinant, different letters indicate significant differences between treatments (GLM, p < .05)

PBS CF
Developmental time (days) 7.89 ± 0.12 a 7.26 ± 0.11 b Developmental mortality (%) 20.03 ± 0.51 a 19.81 ± 0.45 a Cocoon weight (mg) 30.00 ± 7.34 a 30.77 ± 8.49 a F I G U R E 4 Relative growth rate of unparasitized (a) and parasitized (b) Spodoptera frugiperda caterpillars scored 48 h after feeding on corn leaves either undamaged (UD) or induced with salivary gland extract from: caterpillars injected with phosphate-buffered saline (PBS); caterpillars parasitized by Hyposoter didymator (PAR); and caterpillars injected with calix fluid (containing virions) isolated from the parasitoid H. didymator (CF). Different letters above bars indicate significant differences among treatments (GLM, p < .05) with similarity with insect juvenile hormone esterase, a UDPglucuronosyltransferase, two entries matching with peroxidases, and a protein disulphide-isomerase. GOX is also significantly less abundant in both PAR (log2Dif of −5.3) and CF (log2Dif of −3.3) samples compared to PBS controls (Figure 2).

| Performance of unparasitized caterpillars
The relative growth rate of unparasitized S. frugiperda caterpillars was significantly affected by the leaves induced with different types of caterpillar saliva that were offered as food (ANOVA, F = 18.372, df = 3,63, p < .001). Caterpillars allowed to feed on leaves that were previously induced with saliva from unparasitized herbivores (treatment PBS) displayed reduced relative growth rates compared with caterpillars feeding on leaves induced with saliva from parasitized or virus-infected caterpillars (treatments PAR and CF, respectively).
No differences in relative growth rate were found when caterpillars were feeding on leaves previously induced with saliva obtained from either PAR or CF treated caterpillars (Figure 4a).

| Performance of parasitized caterpillars
Similarly, the growth rate of parasitized caterpillars was significantly affected by the leaves induced with different types of caterpillar saliva that was offered as food (ANOVA, F = 19.375, df = 3,64, p < .001).
Parasitized caterpillars feeding on leaves previously induced with saliva from unparasitized caterpillars (treatment PBS) showed the lowest relative grow rates. Again, no differences were found in relative growth rates between caterpillars feeding on leaves previously induced with saliva from either parasitized or virus-infected caterpillars (treatments PAR and CF, respectively) ( Figure 4b).

| Performance of parasitoids
Parasitoid larvae developed significantly faster when their host caterpillars were feeding on leaves previously induced with salivary gland extract from CF-injected caterpillars compared with PBSinjected unparasitized caterpillars (χ 2 = 14.744, df = 1, p < .001) (  Ode et al., 2016;Poelman, Gols et al., 2011;Poelman, Zheng et al., 2011;Tan et al., 2019Tan et al., , 2020. While it was previously assumed that parasitoid larvae growing within the herbivore body were responsible for the specific responses of plants to feeding by parasitized caterpillars (Poelman, Zheng et al., 2011), it is now acknowledged that parasitoid-associated symbionts can be the real hidden driving forces mediating such complex interactions (Dicke et al., 2020;Shikano et al., 2017). Here, we report the first molecular and ecological evidence that ichnovirus infection affects plant-insect interactions, increasing the awareness that such parasitoid-associated symbionts have a much more extended phenotype than was previously thought.
From a mechanistic perspective, it has been hypothesized that parasitoid-associated viruses could interact directly or indirectly with the plant . A direct interaction would occur when viral-encoded proteins come in contact with the plant tissues. This is a fascinating hypothesis which is based on the evidence that bracovirus genes are expressed in salivary glands (Bitra et al., 2011;Zhu et al., 2018), suggesting that virus-encoded proteins could be produced in the insect saliva and released into the plant during caterpillar feeding. Polydnaviruses could also act indirectly when viral injection in the caterpillar haemolymph induces physiological changes which alter the biochemical composition of caterpillar salivary glands. Evidence for the indirect mechanism of action is available for two bracoviruses, as targeted approaches have shown that the activity of enzymes known to activate plant defences is reduced after virus injection (Tan et al., 2018;Zhu et al., 2018).
Our results demonstrate that an ichnovirus may affect insectplant interactions both directly and indirectly. In our proteomic analyses, we found two virus-encoded proteins present in salivary glands of S. frugiperda caterpillars naturally parasitized by H. didymator or injected with calyx fluid containing HdIV. Both are related glycine-proline rich proteins, encoded by the same HdIV-specific gene family, and known to be highly expressed and secreted in parasitized hosts (Volkoff et al., 1999). The presence of "alien" proteins in caterpillar salivary glands is indicative of qualitative changes that could be used by the plant to recognize whether the herbivore attacker has been parasitized or not and tailor the defences accordingly. There is increasing evidence showing that plants reduce their defences when attacked by caterpillars carrying polydnavirusesassociated parasitoids Tan et al., 2018); thus it is possible to argue that viral "alien" proteins, which represent  (Tian et al., 2012), it appears to suppress defences in tobacco (Musser et al., 2002) and its effect in corn remains unclear (Louis et al., 2013). In tomato, the bracovirus associated with M. croceipes (McBV) manipulates plant responses to herbivory by decreasing GOX activity (Tan et al., 2018). The ichnovirus associated with H. didymator (HdIV) could act in a similar way, although it remains to be investigated if performances of S. frugiperda caterpillars are affected by exogenous application of GOX in corn. Apyrase is an ATP-hydrolyzing enzyme previously described in S. frugiperda  and Helicoverpa zea saliva (Wu et al., 2012).
Application of H. zea apyrase to wounded tomato leaves was shown to downregulate plant defences (Wu et al., 2012). Although the effect of apyrase on corn still remains to be analysed, the observed increase in apyrase levels in HdIV-infected S. frugiperda salivary glands may contribute to a decrease in plant defences induced by herbivory.
Our results also indicate that parasitism or virus injection af-  (Tan et al., 2018). It is well known that plant nutritional quality can indirectly impact parasitoid fitness via effects on the herbivore host (Ode, 2006). Interestingly we found an increase in performance not only for parasitized caterpillars but also for unparasitized caterpillars which thus take advantage of plant-mediated Yet dissimilarities between the two tri-trophic systems make comparisons challenging, especially at the plant level as on study focused on tomato-bracovirus interactions and our study investigated cornichnovirus interactions. Thus more experimental evidence is needed to conclude that bracoviruses achieve stronger plant-mediated benefits for their symbiotic partners compared with ichoviruses.
Polydnaviruses are the most intensively studied mutualistic symbionts of parasitoids, yet research has generally been restricted to their role in host-parasitoid interactions (Edson et al., 1981;Lu et al., 2010;Shelby & Webb, 1999;Strand et al., 2006;Strand & Burke, 2013;Webb et al., 2006). As a consequence, the effects of polydnaviruses on tissues such as hemocytes (involved in insect immunity) and fat bodies (involved in general metabolism) have been intensively studied while we know very little about the role played by such viruses in tissues like salivary glands and midgut which are important at the plant-insect interface. Studies that will investigate the temporal patterns and tissue specificity of polydnavirus infection during the whole parasitoid development inside the herbivore host will be particularly informative for understanding how plant-insect interactions are shaped by parasitoid symbionts. By extending the study of polydnaviruses at the plant level, novel positive and negative effects for their symbiotic parasitoid partners have been discovered.
While some research has shown that top-down manipulation of plant quality increases parasitoid fitness indirectly (Tan et al., 2018), other research has unraveled surprising ecological costs as well . For example, polydnaviruses initiate an interaction network across four trophic levels which trigger changes in herbivoreinduced plant volatiles attracting insect hyperparasitoids . Interestingly the extended phenotype of polydnaviruses can also reach other plant-associated insects. If polydnaviruses enhance the performance of unparasitized herbivores feeding on the plant, as shown in this study, then there could be negative effects at the plant level. Future research should be undertaken in order to unravel what are the overall consequences of top-down effects induced by parasitoid-associated viruses for the plant fitness. Placing microbial mutualistic symbioses in a community context is thus crucial in order to fully understand the "hidden" role that polydnaviruses play in plant-based food webs.

ACK N OWLED G EM ENTS
We

CO N FLI C T O F I NTE R E S T
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that supports the findings of this study have been made available in the Dryad Digital Repository at the following citation: .