Detoxification of plant defensive glucosinolates by an herbivorous caterpillar is beneficial to its endoparasitic wasp

Plant chemical defences impact not only herbivores, but also organisms in higher trophic levels that prey on or parasitize herbivores. While herbivorous insects can often detoxify plant chemicals ingested from suitable host plants, how such detoxification affects endoparasitoids that use these herbivores as hosts is largely unknown. Here, we used transformed plants to experimentally manipulate the major detoxification reaction used by Plutella xylostella (diamondback moth) to deactivate the glucosinolate defences of its Brassicaceae host plants. We then assessed the developmental, metabolic, immune, and reproductive consequences of this genetic manipulation on the herbivore as well as its hymenopteran endoparasitoid Diadegma semiclausum. Inhibition of P. xylostella glucosinolate metabolism by plant‐mediated RNA interference increased the accumulation of the principal glucosinolate activation products, the toxic isothiocyanates, in the herbivore, with negative effects on its growth. Although the endoparasitoid manipulated the excretion of toxins by its insect host to its own advantage, the inhibition of herbivore glucosinolate detoxification slowed endoparasitoid development, impaired its reproduction, and suppressed the expression of genes of a parasitoid‐symbiotic polydnavirus that aids parasitism. Therefore, the detoxification of plant glucosinolates by an herbivore lowers its toxicity as a host and benefits the parasitoid D. semiclausum at multiple levels.


Multitrophic interactions involving plants, insect herbivores, and their
antagonists are ubiquitous in terrestrial ecosystems and underpin our understanding of the structure and function of ecological communities (Stam et al., 2014). Most plants in nature are attacked by insect herbivores, and high infestations can severely damage plant tissues and thus reduce plant fitness (Johnson, Lajeunesse, & Agrawal, 2006). To prevent or reduce attack, plants employ an array of strategies to reduce herbivory, including the production of a wide assortment of toxic, repellent, antidigestive, and antinutritive chemical defences (Mithöfer & Boland, 2012). Plant chemical defences can also traverse trophic levels, moving up the food chain to affect not only the consuming herbivores but subsequently also herbivore predators (Hartmann, 2004; rectly modify proteins and interfere with intracellular redox homeostasis (Brown & Hampton, 2011). ITCs also typically react quickly with the intracellular nucleophile glutathione (GSH) leading to its depletion (Jeschke, Gershenzon, & Vassão, 2016a). The presence of glucosinolates in plants has fueled selection for a suite of mechanisms in herbivores that mitigate or circumvent exposure to toxic glucosinolate hydrolysis products (Jeschke et al., 2016b), including behavioural adjustments, detoxification, rapid excretion processes and sequestration (Winde & Wittstock, 2011). Larvae of Plutella xylostella (the diamondback moth, Lepidoptera: Plutellidae), a specialized herbivore that is a devastating pest of brassicaceous crops (Furlong, Wright, & Dosdall, 2013;Zalucki et al., 2012), can feed without ill effects on glucosinolate-containing plants due to the activity of glucosinolate sulphatases (GSS). These enzymes are abundant in the digestive system of this insect, and desulphate glucosinolates preventing myrosinase-catalysed hydrolysis and ITC formation (Ratzka, Vogel, Kliebenstein, Mitchell-Olds, & Kroymann, 2002). Sun et al. (2019) showed that genetic disruption of P. xylostella desulphation led to the increased formation of ITCs, which caused steep declines in larval growth, survival and reproduction, demonstrating desulphation to be a genuine detoxification strategy in this herbivore.
At higher trophic levels, plant defences can act indirectly by reducing the quality of herbivores available as prey or hosts. Alternatively, direct exposure to defensive chemicals ingested by the prey or host can negatively affect the growth and development of parasitoids and predators Kaplan, Carrillo, Garvey, & Ode, 2016;Lampert, Zangerl, Berenbaum, & Ode, 2011;Ode, 2019).
When the common green lacewing Chrysoperla carnea fed on P. xylostella in which glucosinolate desulphation was blocked by RNA interference, its development was slowed by consuming these ITCcontaining prey (Sun et al., 2019). Nevertheless, C. carnea suffered no reproductive effects, perhaps due to its ability to metabolize ITCs via a general detoxification pathway. However, it is not clear how other predators or parasitoids cope with plant defences such as ITCs in their prey or host.
The solitary endoparasitoid Diadegma semiclausum (Hymenoptera: Ichneumonidae) is an important natural enemy of P. xylostella, and is frequently used in biocontrol programmes against this pest (Furlong et al., 2013;Li, Eigenbrode, Stringam, & Thiagarajah, 2000). Young larvae of this parasitoid develop by feeding on the haemolymph of the caterpillar host until the parasitoid larva reaches its final instar.
At this point, it starts to feed on all tissues indiscriminately and eats the host completely, until it pupates in the puparium made by the host caterpillar before death. During all phases of parasitoid growth, the host caterpillar keeps feeding on the plant, thereby exposing the parasitoid larvae continuously to glucosinolates and their metabolites. Previous research has shown that D. semiclausum development is influenced by the brassicaceous species on which their P. xylostella hosts were reared, possibly due to interspecific differences in glucosinolate profiles (Dosdall, Zalucki, Tansey, & Furlong, 2011;Gols et al., 2008). However, whether glucosinolates, ITCs or other glucosinolate metabolites benefit or harm endoparasitoids of P. xylostella such as D. semiclausum, is unknown. Moreover, how parasitism by D. semiclausum influences P. xylostella glucosinolate metabolism and development is also not known.
Here, we manipulated the detoxification capacity of P. xylostella larvae in order to examine the effects of plant glucosinolate defences on the interactions between P. xylostella and D. semiclausum. We hypothesized that blocking the desulphation activity responsible for glucosinolate detoxification would lead to the accumulation of ITCs in host tissues, affecting in turn the development of the endoparasitoid. Additionally, we also examined the effects of blocking glucosinolate desulphation on the immune responses of the herbivore against parasitism.

| Experimental overview
Plant-mediated RNAi was used to silence detoxification-related genes in the herbivore P. xylostella. Arabidopsis thaliana Columbia-0 (Col-0) plants, which naturally contain glucosinolates, were engineered to produce dsRNA targeting the Pxgss genes encoding glucosinolate sulphatases (GSS) responsible for glucosinolate detoxification in the herbivore. While it is unclear whether P. xylostella is a natural herbivore of wild A. thaliana, this plant has been shown to serve as a suitable host for P. xylostella, supporting similar growth as on cultivated Brassica crops (Barker, Poppy, & Payne, 2007). Arabidopsis thaliana mutants deficient in the production of glucosinolates were also engineered for RNAi and were used as toxin-free controls. Through the combination of these treatments (presence or absence of glucosinolate defences in the plant and presence or absence of glucosinolate detoxification in the herbivore), we investigated the effects of glucosinolate ingestion and its detoxification by the specialist herbivore both on the herbivore and on the parasitoid D. semiclausum. Insect parameters measured included growth and development, survival, protein and lipid contents, and herbivore immune phenoloxidase (PO) activity, while the movement of glucosinolate metabolites among trophic levels was quantified by targeted HPLC-MS/MS analyses. Finally, quantitative real-time PCR (qPCR) was conducted to examine the effects of glucosinolate-derived ITCs on the expression of selected herbivore and parasitoid genes. These procedures are described in detail in the following subsections.

| Plants and insects
Arabidopsis thaliana Col-0 accession plants with wild type-glucosinolate levels (obtained from the Arabidopsis stock center) and transgenic myb28myb29 knockout mutants (without aliphatic glucosinolates, obtained from Daniel J. Kliebenstein, University of California Davis) (Sønderby et al., 2007)

| Plutella xylostella Pxgss gene silencing by plantmediated RNAi
Plant mediated RNAi was used to silence the expression of Pxgss genes in the midgut of P. xylostella larvae, as described in Sun et al. (2019). In short, we constructed a tobacco rattle virus-based dsRNA producing system (Pxgss-RNAi construct) targeting Pxgss in P. xylostella larvae. An empty vector construct was used as a negative control. Both the Pxgss-RNAi construct and empty vector construct were transformed into A. thaliana Col-0 wild-type and myb28myb29 double knockout mutant plants via Agrobacterium tumefaciens (strain GV3101). We had previously determined that plant transformation with these constructs had no effects on plant morphology, glucosinolate profile, and levels of other defensive secondary metabolites (Sun et al., 2019).

| Measurements of P. xylostella growth and development
Plutella xylostella larvae were fed ad libitum on either empty vectortransformed or Pxgss RNAi-transformed A. thaliana plants (from both Col-0 and myb28myb29 backgrounds) after hatching in a controlled environmental chamber under the conditions described above. The percentage of larval pupation, which occurred from 7-10 days post hatching, was recorded in the four treatment groups relative to the initial numbers of hatched larvae (approximately 120 larvae per treatment). In the late fourth-instar stage, P. xylostella larvae were collected in 1.5 ml Eppendorf tubes to measure larval weights (20 replicates per treatment) as well as soluble protein (five replicates per treatment, three larvae were pooled in one sample) and lipid content (five replicates per treatment, three larvae pooled) after freeze-drying (ALPHA 1-4 LD plus freeze dryer, Martin Christ, Osterode am Harz, Germany).
Protein and lipid content were measured to assess the nutritional status of the insects after Pxgss-silencing and glucosinolate ingestion. Fresh and dry bodyweights were measured using a microbalance (XP26, Mettler Toledo, Gießen, Germany). Soluble protein and lipid content were measured as described below in 2.8.

| Diadegma semiclausum parasitism
Approximately three hundred P. xylostella second-instar larvae of each of the four treatments, Pxgss-silenced or unsilenced feeding on either A. thaliana Col-0 or myb28myb29 plants, were individually exposed to D. semiclausum to be parasitized once. Female and male D. semiclausum adults were sexually mature and maintained together for a few days to promote mating before being exposed to P. xylostella. After parasitism, P. xylostella larvae were returned to their original food plant until they developed into prepupae. Before the parasitoid larva enters pupation, it excretes a meconium containing waste products, which is present as a black pellet inside its cocoon. The cocoons were kept individually in 5 ml amber glass vials with a cotton cover until emergence. Subsequently, newly emerged adults of D. semiclausum were sexed and transferred to 1.5 ml Eppendorf tubes, and immediately frozen in liquid nitrogen for further analyses. As part of these experiments, unparasitized P. xylostella fed on either A. thaliana Col-0 or myb-28myb29 plants were raised under the same conditions (21°C, 70% relative humidity, and a 16:8 hr light:dark photoperiod). The detailed experimental time line for D. semiclausum parasitism of P. xylostella is shown in Figure S1.

| Diadegma semiclausum larval development duration, adult emergence and adult bodyweight
To determine the physiological effects of glucosinolates ingested by host larvae on the parasitoid in the absence of RNAi (Figure 1), second-instar P. xylostella larvae were parasitized by female D. semiclausum as described above and allowed to feed ad libitum on either A. thaliana Col-0 or myb28myb29 plants (one plant per treatment, approximately 50-100 larvae for each of the two treatments). Plants were changed every day to ensure that sufficient food was available. This experiment was repeated three times independently (see Supporting Information file). Then, the percentages of successful D. semiclausum adult emergence were calculated relative to the number of P. xylostella larvae parasitized. Emerged D. semiclausum adults were sexed, freeze-dried and weighed (approximately 20 males and 10 females were obtained per treatment).
To determine the physiological effects of silencing Pxgss on D. semiclausum (Figure 2g-j), a second experimental setup was used.
Silenced and nonsilenced second-instar P. xylostella larvae feeding on either A. thaliana Col-0 or myb28myb29 plants (approximately 600 larvae per treatment, 2,400 larvae in total, with 100 larvae per plant in a single cage, plants exchanged daily to ensure food availability) were parasitized by female D. semiclausum adults. First, we recorded the emergence of D. semiclausum adults (12-16 days post parasitism) from only a subset of the insects (two cages per group, i.e., around 120-180 surviving parasitized P. xylostella per treatment, see Supporting Information file). Next, D. semiclausum adults emerging from all groups (all six cages per treatment) were sexed, immediately frozen in liquid nitrogen, freeze-dried and weighed (40 male replicates and 10 female replicates were collected per treatment). Soluble protein and lipid contents in these adults were measured as described below in 2.8. Only males were used for this chemical analysis, as the number of female wasps was insufficient for reliable measurement.

| 4MSOB-ITC complementation experiment
In leaves of A. thaliana Col-0 plants grown for these experiments, 4-methylsulphinylbutyl glucosinolate (4MSOB) represents over 70% of the aliphatic glucosinolates (Brown, Tokuhisa, Reichelt, & Gershenzon, 2003). To determine whether the elevated 4-methylsulphinylbutyl isothiocyanate (4MSOB-ITC) concentrations in Pxgsssilenced P. xylostella caused the D. semiclausum phenotypes observed, complementation experiments were conducted by infiltrating 4MSOB-ITC into leaves of myb28myb29 plants lacking aliphatic glucosinolates as described. A 0.3 µl quantity of 800 µM 4MSOB-ITC per mg fresh leaf was injected using a needleless syringe (Katagiri, Thilmony, & He, 2002) to mimic the natural content of damaged leaves (Sun et al., 2019). Leaves infiltrated with solvent (0.4% aqueous ethanol) served as negative controls. Diadegma semiclausum female adults were allowed to parasitize P. xylostella larvae (approximately 600 per treatment) continuously feeding on these leaves, and D. semiclausum adult emergence percentage, body dry weight, soluble protein and lipid contents were measured. First, the percentage of D. semiclausum adult emergence (13-16 days post parasitization) from approximately 200 successfully parasitized P. xylostella per treatment was determined.
Then, emerged D. semiclausum adults were sexed, immediately frozen in liquid nitrogen, freeze-dried and weighed (30 male replicates and 10 female replicates per treatment). Third, soluble protein and lipid content in adults was measured as described below in 2.8.
Homogenized samples were centrifuged at 13,000 g for 20 min at 4°C to separate undissolved particles. Clear supernatants were transferred to 1.5 ml Eppendorf tubes, and 20 μl of each sample were used to measure protein concentration (Serva Electrophoresis, Heidelberg, Germany) according to the manufacturer's instructions. Soluble protein content was normalized by insect dry weight.
Lipid content was determined following a previously published protocol (Jeschke et al., 2016a). Weighed and pulverized dried body samples (approximately 5 mg for P. xylostella larvae, and approximately 2.5 mg for D. semiclausum adults, five replicates per treatment) were extracted two times with 1 ml of 2:1 chloroform:methanol by vortexing for 30 s. After centrifugation at 13,000 g for 20 min at 4°C and careful removal of the lipid-containing solvent, the remaining powder was dried at 80°C for 48 hr and weighed to calculate the proportion of lipid present.

| Immune phenoloxidase (PO) activity
PO activity is part of a critical host immune defence reaction that promotes melanization during the encapsulation response against parasitoids (Strand & Pech, 1995). To compare PO activity in the haemolymph of nonsilenced and Pxgss-silenced P. xylostella fourth-instar larvae feeding on either A. thaliana Col-0 or myb28myb29 mutant plants, "spontaneous PO" activity assays were performed as described by Barthel et al. (2016), without a protease pretreatment to activate pro-PO in the samples. Haemolymph of eight larvae (1 μl from each larva) was pooled into one sample and 10 samples per treatment were assessed for their PO activities. Haemolymph was extracted by puncturing the larvae with a sterile hypodermic needle and immediately pipetting the "bleeding" haemolymph into 1.5 ml amber safe-lock Eppendorf tubes with 200 μl ice-cold sodium cacodylate solution (0.01 M Na-cacodylate, 0.005 M CaCl 2 in Milli-Q water). The haemolymph mixture was directly frozen in liquid nitrogen and stored at −80°C until measurement. To measure PO activity, frozen samples were thawed on ice then centrifuged at 4°C and 2,800 g for 15 min to remove cell debris. An aliquot (100 μl) of the resulting supernatant (10 replicates per treatment, one measurement per replicate) was transferred to a 96 well polystyrene plate (VWR International, Darmstadt, Germany) and subsequently mixed with 200 μl of 3 mM L-DOPA (Sigma-Aldrich, Munich, Germany, freshly prepared and covered with silver foil) per single well, and 100 μl sodium cacodylate solution treated in the same way was measured as a negative control (eight replicates). Absorbance at 490 nm was measured once per minute for 45 min at 30°C in a Multiskan Spectrum multiplate reader (Thermo-Electron, Waltham, MA, USA). The changes in absorbance from 15-26 min of the 45 min measurements were confirmed to be linear and were used to calculate the PO activity (1U = 0.001 AU/min, V max expressed as mU). Data were obtained with SkanIt Software for Multiskan Spectrum version 2.1 (Thermo-Electron).

| Metabolic analyses
To analyse how glucosinolates were metabolized by D. semiclausum, third-instar D. semiclausum larvae at 7-8 days post parasitism (three larvae pooled per sample) were collected from P. xylostella using a dissecting microscope and washed in TE buffer (Tris-EDTA buffer, pH 8) to remove residual P. xylostella haemolymph. The remaining carcass of P. xylostella was also collected (approximately 3 mg per sample). Macherey-Nagel, Düren, Germany) with mobile phase A (0.2% formic acid in milliQ water) and mobile phase B (acetonitrile). 4MSOB-ITC and its conjugates were analysed by loading samples onto a Agilent XDB-C18 column (50 × 4.6 mm × 1.8 μm, Agilent Technologies, Waldbronn, Germany) with mobile phase A (0.05% formic acid in milliQ water) and mobile phase B (acetonitrile). The elution gradient profiles were previously described in Sun et al. (2019); MS parameters for 4MSOB, desulpho-4MSOB, 4MSOB-ITC and its conjugates were also described before (Gloss et al., 2014;Malka et al., 2016); Analyst 1.5 software (Applied Biosystems Sciex, Germany) was used for data acquisition and processing. Quantification of individual compounds was achieved by external calibration curves (the external standards are given in Table S1).
For these measurements, fourth-instar larvae were individually pooled into TRIzol reagent (Invitrogen, Waltham, MA, USA) in 1.5 ml Eppendorf tubes, and then kept at 4°C before RNA isolation. Total RNA was iso-  Table S2.

| Arabidopsis glucosinolate content does not influence the development of the parasitoid Diadegma semiclausum in Plutella xylostella hosts with nonsilenced detoxification
To examine how the development of D. semiclausum is affected by the glucosinolate content in the diet of its P. xylostella host, D. semiclausum was allowed to parasitize P. xylostella second-instar larvae ( Figure S1) that were fed ad libitum on Arabidopsis thaliana plants either containing (wild-type Col-0) or lacking (myb28myb29 mutant) aliphatic glucosinolates. Aliphatic glucosinolates comprise 70%-80% of the total foliar glucosinolates in A. thaliana Col-0 plants (Brown et al., 2003), and are the only glucosinolate class in this plant that forms stable ITCs after hydrolysis by plant myrosinases. Adult emergence of D. semiclausum was similar irrespective of the glucosinolate content in the diet of its host, with approximately 40% emergence success in both groups (Figure 1a). The body masses of D. semiclausum male and female adults were also not affected by the glucosinolate content in the diet of their host (Figure 1b).

| Blocking glucosinolate detoxification affects the development and physiology of Plutella xylostella caterpillars
In order to explore how D. semiclausum is affected by plant toxins in its herbivorous host P. xylostella, we used RNAi targeting the Pxgss genes that encode glucosinolate sulphatases (GSSs) in the herbivore, to block glucosinolate detoxification. GSSs desulphate plant glucosinolates in the larval midgut, forming nontoxic desulpho-glucosinolates that are not capable of being activated by plant myrosinases to form toxic glucosinolate hydrolysis products (Ratzka et al., 2002). Suppression of Pxgss expression had previously been shown to cause reduced GSS activity and increased concentrations of the toxic ITCs resulting from hydrolysis of glucosinolates in P. xylostella larvae, causing negative F I G U R E 1 Plant glucosinolate content has little impact on the development of Diadegma semiclausum in Plutella xylostella hosts. D. semiclausum females were allowed to parasitize P. xylostella larvae feeding on either A. thaliana Col-0 or myb28myb29 plants (with or without aliphatic glucosinolates, respectively), and the following variables were measured. (a) Diadegma semiclausum adult emergence percentage (χ 2 = 0.001, p = .976, n = 82 and 80, respectively); and (b) D. semiclausum adult dry bodyweight (sex, F 1,62 = 20.356, p ≤ .0001; plant, F 1,62 = 1.020, p = .317; sex × plant, F 1,62 = 0.084, p = .772; n = 27, 18, 7 and 14 represent the respective numbers of replicates in each of the treatments presented in the graph, from left to right) were not affected by aliphatic glucosinolate content. Bars denote means and the interval is the SE. Significant differences (p ≤ .05) were determined by a two proportions z-test in (a). and Tukey HSD tests in conjunction with two-way ANOVA in (b) fitness effects (Sun et al., 2019). Here, we examined the effect of Pxgss silencing on P. xylostella growth, development and chemical composition in more detail. The pupation of Pxgss-silenced P. xylostella larvae was delayed in comparison to nonsilenced controls, but only on food plants containing aliphatic glucosinolates ( Figure 2b). Silencing of Pxgss did not affect the growth of caterpillars in terms of late fourth-instar larval biomass, irrespective of whether the food plant contained glucosinolates or not (Figure 2c). Glucosinolates in leaf tissues did affect larval metabolism since feeding on glucosinolate-containing plants led to an approximately 40% reduction in soluble protein levels (Figure 2d), while lipid levels were not affected (Figure 2e). In order to assess how Pxgss silencing and exposure to ITCs affect general larval immune responses, we measured the activity of phenoloxidase (PO) in the larval haemolymph. PO activity is part of a critical host immune defence reaction that promotes melanization during the encapsulation response against parasitoids (Strand & Pech, 1995).
However, PO activity was not affected by Pxgss silencing or dietary glucosinolate ingestion (Figure 2f), suggesting that exposure to ITCs does not impair this aspect of P. xylostella immunity.

| Blocking glucosinolate detoxification in the host caterpillar negatively affects development of the endoparasitoid D. semiclausum
We determined whether development of the endoparasitoid D. semiclausum was affected by the metabolism of glucosinolates in its herbivorous host P. xylostella. The emergence of D. semiclausum parasitizing gss-silenced P. xylostella larvae that had been fed on glucosinolate-con- comprising about 70% of the total aliphatic glucosinolate pool (Brown et al., 2003). D. semiclausum developing from P. xylostella larvae that had fed on 4MSOB-ITC-infiltrated plants had delayed development ( Figure S2a), lower adult emergence success ( Figure S2a), higher adult weights ( Figure S2b) and higher adult lipid content ( Figure S2d) compared to D. semiclausum parasitizing larvae that had fed on myb-28myb29 leaves infiltrated only with solvent (negative control).

| Glucosinolate metabolites are transferred from P. xylostella hosts to D. semiclausum larvae
Younger larvae of D. semiclausum primarily feed on the haemolymph of their hosts, but later instars consume almost all tissues just before they complete larval development and pupate ( Figure S1). Therefore, these larvae will inevitably encounter glucosinolates or their metabolites while developing in a P. xylostella host that feeds on glucosinolate-containing plant tissues. D. semiclausum larvae developing in gss-silenced P. xylostella feeding on glucosinolatecontaining plants encountered higher amounts of 4MSOB-ITC and reduced amounts of desulpho-4-methylsulphinylbutyl glucosinolate (desulpho-4MSOB) in host tissues than those developing on nonsilenced P. xylostella larvae (Figure 3a,b). In addition, higher amounts of the known 4MSOB-ITC mercapturic acid pathway conjugates (ITC-GSH, ITC-CG and ITC-Cys; Figure S3a

| Parasitism by D. semiclausum alters glucosinolate metabolism and excretion in the host P. xylostella
The increases in levels of 4MSOB-ITC and its conjugates in Pxgsssilenced P. xylostella carcasses after D. semiclausum parasitism were much lower in magnitude than the decreases in desulpho-4MSOB ( Figure 3). To learn more about the efflux of 4MSOB metabolites, we directly compared the quantities of 4MSOB metabolites present in haemolymph, frass, and pupae of parasitized and nonparasitized P. xylostella larvae. Although the concentrations of 4MSOB-ITC were significantly higher in the haemolymph of gss-silenced than in nonsilenced P. xylostella larvae, parasitism reduced 4MSOB-ITC haemolymph levels in Pxgss-silenced larvae by 87% (Figure 4a).
Conversely, parasitism significantly increased 4MSOB-ITC levels in the frass of Pxgss-silenced larvae compared to unparasitized silenced controls (Figure 4b). At the prepupal stage, parasitism decreased 4MSOB-ITC in P. xylostella to nearly undetectable levels, while the pupae of unparasitized larvae still had measurable 4MSOB-ITC content (Figure 4c). Accordingly, 4MSOB-ITC conjugates were more abundant in the frass and prepupae of parasitized P. xylostella, compared to nonparasitized larvae, although these conjugates were found in lower amounts than 4MSOB-ITC itself ( Figure S3).
Although Pxgss silencing successfully reduced the formation of the 4MSOB detoxification product desulpho-4MSOB in the host, the haemolymph of parasitized P. xylostella larvae contained approximately 75% less desulpho-4MSOB than that of nonparasitized larvae, for both Pxgss-silenced and nonsilenced larvae (Figure 4d). Similarly, the frass from parasitized Pxgss-silenced P. xylostella larvae contained about 85% less desulpho-4MSOB than frass from nonparasitized silenced larvae, whereas desulpho-4MSOB in nonsilenced larvae was not affected by parasitism (Figure 4e). However, the prepupae of both Pxgss-silenced and unsilenced larvae parasitized by D. semiclausum contained 17.3and 47.9-fold higher concentrations of desulpho-4MSOB, respectively, than nonparasitized P. xylostella pupae (Figure 4f). Therefore, larvae of D. semiclausum appear to influence the excretion of 4MSOB-ITC by the host and to absorb desulpho-glucosinolates when parasitizing P. xylostella, and these compounds are retained in the prepupae (Figure 4g). into the haemolymph of their hosts and cause immune suppression, resulting in lower rates of encapsulation of the developing parasitoids (Beckage, 2011;Webb et al., 2006) as well as changes in the phenology of the host (Harvey, 2005). PDVs associated with ichneumonid parasitoids like D. semiclausum are called ichnoviruses, and produce proteins important in infection, such as vankyrins and viral annexins (Tanaka et al., 2007), to reduce the rates of encapsulation of the developing parasitoids in host haemolymph. To determine if higher ITC levels might affect the expression of symbiotic PDV-related genes upon parasitism, the expression of three well-studied viral-related gene transcripts (Etebari et al., 2011), vankyrin1, vankyrin2 and viral innexin1, was measured by qRT-PCR. Expression of these three genes was reduced by 80%, 70% and 62%, respectively, upon D. semiclausum parasitism in Pxgss-silenced hosts feeding on Col-0 plants (with aliphatic glucosinolates) compared to nonsilenced hosts feeding on Col-0 plants or either silenced or nonsilenced hosts feeding on myb-28myb29 plants (without aliphatic glucosinolates) ( Figure 5).

| Parasitism by D. semiclausum reduces P. xylostella ecdysone receptor (EcR) expression
To further explore how gss silencing influences parasitism of P. xylostella by D. semiclausum, we measured the expression of a P. xylostella ecdysone-related gene (EcR), which is induced by ecdysone to control larval development and pupation (Israni & Rajam, 2017). EcR transcripts were reduced 30%-70% in P. xylostella fourth-instar larvae upon successful parasitism by D. semiclausum ( Figure S4). However, Pxgss silencing and the glucosinolate content of the P. xylostella diet had no significant effect on expression.
In untransformed P. xylostella where the glucosinolate detoxification system was functional, D. semiclausum tolerated a range of glucosinolate content in the herbivore diet (Figure 1). This suggests that previous reports on changes in D. semiclausum developmental variables, such as cocoon and adult weight and duration of development (Dosdall et al., 2011;Kahuthia-Gathu, Löhr, & Poehling, 2008), are probably not caused by alterations in the glucosinolate content of the food plant of the host herbivore. Although we employed genetically F I G U R E 3 Metabolites of 4MSOB are present in Diadegma semiclausum parasitizing P. xylostella larvae fed on A. thaliana Col-0 plants. D. semiclausum was allowed to parasitize nonsilenced and Pxgss-silenced P. xylostella larvae feeding on either A. thaliana Col-0 (with aliphatic glucosinolates) or myb28myb29 (without aliphatic glucosinolates) plants. (a) 4MSOB-ITC (host carcass, p ≤ .05; parasitoid larva, p ≤ .05; meconium, p ≤ .01; n = 5); (b) Desulpho-4MSOB (host carcass, p = .151; parasitoid larva, p ≤ .05; meconium, p = .056; n = 5 in all bars): and (c) 4MSOB-ITC conjugates (host carcass, p ≤ .01; parasitoid larva, p ≤ .05; meconium, p ≤ .01; n = 5) were quantified in the carcass of P. xylostella prepupae, third-instar larvae of D. semiclausum, meconium left in the cocoon and adults of D. semiclausum, in which D. semiclausum parasitized either nonsilenced (black bars) or Pxgss-silenced (grey bars) P. xylostella. The general mercapturic acid pathway is shown in Figure S3a. 4MSOB-ITC-GSH: 4MSOB-ITC-glutathione conjugate; 4MSOB-ITC-CG: 4MSOB-ITC-cysteinylglycine conjugate; and 4MSOB-ITC-Cys: 4MSOB-ITC-cysteine conjugate; concentrations are shown in Figure S3b as stacked bars. 4MSOB and its metabolites were nearly undetectable in P. xylostella larvae fed on myb28myb29 plants and are not shown in the graphs. Coloured objects depict the parts being analysed. Significant differences (p ≤ .05) between means (±SE) were determined by Mann-Whitney Wilcoxon tests in a-c. separately conducted for each tissue modified plant lines in this study to block a specialized detoxification reaction of an herbivore, which increased the accumulation of toxic isothiocyanate products in its body, this is also reflective of natural situations. Brassicaceae plants are also subject to herbivory by numerous species of generalist herbivores without specialized detoxification pathways. Several lepidopteran species are known to produce and accumulate large quantities of isothiocyanates when feeding on glucosinolate-containing plants (Jeschke et al., 2017), and so infesting parasitoids would encounter a similar situation to that in sulphatase-silenced P. xylostella.
How plant defences affect endoparasitoid performance is poorly understood despite the fact that endoparasitoids constitute a very abundant group of enemies of insect herbivores Harvey, van Dam, Raaijmakers, Bullock, & Gols, 2011;Ode, 2006). Since the physiology of endoparasitoids is very tightly coupled with that of their hosts, these insects could be very susceptible to plant toxin content in the host diet and the extent of detoxification by the host. Accordingly, a few studies have shown that increased toxin levels in the host diet can actually lead to reduced parasitoid performance (Barbosa et al., 1991;Garvey, Creighton, & Kaplan, 2020). The tolerance of natural enemies of insect herbivores to allelochemicals in host or prey tissues can also depend on the degree of specialization of the natural enemy on their host or prey. For example, while low concentrations of the furanocoumarin xanthotoxin did not affect the specialist parasitoid Copidosoma sosares in its interaction with its specialist herbivore host Depressaria pastinacella, exposure of the parasitoid Copidosoma floridanum (which has a much broader host range) to this compound in the haemolymph of one of its polyphagous hosts reduced survival and offspring production (Lampert et al., 2011). Plant chemical defences can also affect predators. For instance, the wolf spider Camptocosa parallela shows lower preference for tobacco hornworm (Manduca sexta) prey containing higher nicotine levels (Kumar, Pandit, Steppuhn, & Baldwin, 2014), and this predator prefers prey that detoxify nicotine (Kumar, Rathi, Schöttner, Baldwin, & Pandit, 2014). Likewise, predators such as the ladybug Adalia bipunctata are deterred by the glucosinolate-sequestering cabbage aphid Brevicoryne brassicae (Kazana et al., 2007). The larval development of the lacewing C. carnea is reduced when preying on Pxgss-silenced P. xylostella (Sun et al., 2019). The greater physiological intimacy between parasitoids and their hosts compared to predators and their prey may have resulted in parasitoids of insect herbivores being better adapted to the presence of plant defence metabolites in their hosts than predators. Furthermore, parasitoids developing in hosts that contain plants toxins can themselves coopt these plant defences against their own antagonists such as hyperparasitoids (Bowers, 2003;van Nouhuys, Reudler, Biere, & Harvey, 2012) offering them the opportunity to exploit enemy free space, a concept that is usually restricted to insect herbivores (Murphy, Lill, Bowers, & Singer, 2014). Thus, the presence of toxins in their host may not only help protect parasitoids indirectly (avoiding predation of their host), but also directly against their own enemies in the fourth-trophic level (Murphy et al., 2014).

| Herbivore detoxification can modulate the immune response to endoparasitoids
Numerous factors can influence the outcome of insect herbivoreendoparasitoid interactions. We hypothesized that plant defences could influence the immune response of the herbivore host against the parasitoid. Adult female parasitoids in several subfamilies of the Ichneumonidae and Braconidae (e.g., Campopleginae, Microgastrinae) inject symbiotic polydnaviruses (PDVs) into the host haemolymph, which produce proteins that alter host growth (Harvey, 2005) and disrupt cellular and humoral immune responses leading to overall immune suppression (Hasegawa, Erickson, Hersh, & Turnbull, 2017). PDV vankyrins are homologues of the Drosophila melanogaster NF-κβ transcription factor inhibitor Iκβ (Kroemer & Webb, 2005, 2006. These proteins are thought to protect parasitoids from the cellular immune system of the host by suppressing NF-κβ signaling cascades (Tian, Zhang, & Wang, 2007), which blocks blood cell formation and the cellular encapsulation response against parasitoids (Gueguen, Kalamarz, Ramroop, Uribe, & Govind, 2013). PDV innexins are homologues of insect innexins, which form gap junctions between the cytoplasm of insect cells and so play crucial roles in cellular immune responses (Hasegawa & Turnbull, 2014;Turnbull, Volkoff, Webb, & Phelan, 2005). Viral innexins can perturb the physiological functions of native insect innexins (Hasegawa et al., 2017). Here, we observed that the expression of the PDV genes vankyrin1, vankyrin2 and viral innexin1 was suppressed when glucosinolate detoxification was blocked in P. xylostella hosts feeding on glucosinolate-containing plants ( Figure 5). This may have contributed to the lower emergence rate of D. semiclausum adults from these hosts (Figure 2g). Interestingly, D. semiclausum adults that emerged successfully (but belatedly; Figure 2g) from Pxgsssilenced P. xylostella larvae feeding on Col-0 plants were slightly heavier than the D. semiclausum adults from the other treatments ( Figure 2h). These wasps also had a higher lipid content (Figure 2j) than those developing from P. xylostella control hosts, but did not differ in protein content (Figure 2i). At the herbivore level, dietary ITCs are known to increase larval lipid content while decreasing protein content, due to the imbalance in amino acid metabolism caused by ITC detoxification, which depletes cysteine levels (Jeschke et al., 2016a).

| Endoparasitoid performance is enhanced by modification of host gene expression and alteration of ingestion and distribution of plant defences
Parasitism can influence the physiology and behaviour of herbivores in a multitude of ways. We demonstrated here that D. semiclausum parasitism alters P. xylostella ecdysone receptor (EcR) expression in a way that inhibits host pupation. However, this inhibition happened independently of the glucosinolate content of the host diet ( Figure S4). Such manipulation of host development is crucial for completion of the endoparasitoid life cycle. Another example involving EcR concerns a symbiotic bracovirus of the endoparasitoid Cotesia vestalis, which produces a miRNA that arrests host growth by altering the expression of the host EcR gene (Wang et al., 2018).
Parasitism can sometimes induce herbivores to ingest plant toxins to improve survival of the host (Bruce, 2014) or that of the parasitoid (Pashalidou et al., 2015). For instance, parasitized Grammia incorrupta (woolly bear caterpillars) engage in self-medication, increasing their ingestion of plant pyrrolizidine alkaloid toxins in response to endoparasitic tachinid flies resulting in improved herbivore survival (Singer, Mace, & Bernays, 2009). Parasitization can also affect the feeding habits of Pieris rapae larvae (Van Der Meijden & Klinkhamer, 2000), and M. sexta larvae parasitized by Cotesia congregata decrease their feeding in the last stages of parasitoid development by induced anorexia, which actually seems to benefit the parasitoid (Adamo, Linn, & Beckage, 1997).
Moreover, parasitism can also affect the metabolism of plant toxins by the herbivorous host. For example, parasitism of webworm larvae (D. pastinacella) by C. sosares lowers furanocoumarin detoxification rates (per unit of larval weight) in the host, potentially increasing the haemolymph concentrations of these toxins (McGovern et al., 2006). In the present study, parasitism by D. semiclausum did not increase glucosinolate ingestion by its host, but did result in increased excretion of ITCs. When desulphation was blocked by Pxgss gene silencing, the hydrolysis of glucosinolates by plant myrosinases resulted in increased concentrations of toxic ITCs in haemolymph and frass (Figure 4a,b), but these were lower in parasitized hosts than in nonparasitized ones. Therefore, D. semiclausum appears to alter the distribution of toxic plant defence metabolites in P. xylostella to limit its exposure while it feeds on herbivore tissues. The mechanisms used by this parasitoid to manipulate ITC distribution remain to be determined.

| Herbivores that detoxify defences and parasitoids that benefit from detoxification: Consequences for plant protection
The benefits of P. xylostella glucosinolate detoxification for both the herbivore and its parasitoid D. semiclausum are of considerable relevance to the plant as well. Plants that are attacked by herbivores with the ability to detoxify their major defence compounds face a dilemma, since producing increased concentrations of defences will probably not be an effective countermeasure. One strategy is to switch resources to the production of greater amounts of other defence compounds in its arsenal (Koricheva, Nykanen, & Gianoli, 2004), a tactic employed by A. thaliana (Burow et al., 2009) and other Brassicaceae (Kuchernig, Burow, & Wittstock, 2012), especially when fed upon by Pieris rapae, a herbivore that also F I G U R E 5 Expression of D. semiclausum symbiotic polydnavirus (PDV)-related genes is suppressed in Pxgss-silenced P. xylostella larvae. (a) vankyrin1 (Χ 2 = 13.704, df = 3, p ≤ .01; n = 10, 8, 10 and 10, respectively); (b) vankyrin2 (Χ 2 = 13.299, df = 3, p ≤ .01; n = 10, 8, 10 and 10, respectively); and (c) viral innexin1 (Χ 2 = 16.105, df = 3, p ≤ .01; n = 10, 8, 10 and 10, respectively) gene transcripts in D. semiclausum-parasitized P. xylostella fourth-instar larvae feeding on either A. thaliana Col-0 or myb28myb29 plants. Significant differences (p ≤ .05) between medians were determined by Kruskal-Wallis tests with Dunn's post hoc tests in a-c detoxifies glucosinolates (Wittstock et al., 2004). Herbivory by P. rapae alters the major route of glucosinolate activation in A. thaliana so that instead of isothiocyanates, nitriles are formed (Burow et al., 2009), with these compounds deterring P. rapae oviposition (Mumm et al., 2008). Such diversion of defensive resources to other products of the same pathway, or even the activation of separate biosynthetic pathways can therefore help plants protect themselves from attack by a well-adapted herbivore.
Another plant strategy to combat an herbivore with a strong capacity to detoxify chemical defences is to recruit natural enemies of herbivores (Gols et al., 2015). Plant volatiles released in response to herbivore damage have been demonstrated to attract herbivore enemies (Clavijo McCormick, Unsicker, & Gershenzon, 2012;Hare, 2011). While this can increase the fitness of plants attacked by herbivores (Gols et al., 2015;van Loon, de Boer, & Dicke, 2000), evidence showing that this happens under natural conditions is scarce (Clavijo McCormick et al., 2012). In the Brassicaceae, A. thaliana and other species emit volatile glucosinolate hydrolysis products and other volatile metabolites that attract herbivore predators and parasitoids (Bruce, 2014;Hopkins, van Dam, & van Loon, 2009;Mumm et al., 2008). For example, the predator C. carnea and the parasitoid C. plutellae are both significantly attracted by allyl isothiocyanate present in the frass of P. xylostella larvae feeding on cabbage (Reddy, Holopainen, & Guerrero, 2002). 3-Butenyl-ITC is attractive to Diaeretiella rapae, a parasitoid that predominantly attacks Brassicaceae-feeding aphids (Blande, Pickett, & Poppy, 2007). As one of the world's most destructive pests of Brassicaceae plants (Talekar & Shelton, 1993), P. xylostella has developed significant resistance to most synthetic pesticides, as well as to modern biological pesticides like Bacillus thuringiensis (Bt) toxins (Li, Feng, Liu, You, & Furlong, 2016). Thus increased use of natural enemies is being explored to reduce P. xylostella damage (Furlong et al., 2013;Sarfraz, Keddie, & Dosdall, 2005). Our current results suggest that the manipulation of herbivore metabolism could be useful in such an effort. On the one hand, plants that inhibit herbivore detoxification should suffer less damage due to the decreased performance of herbivores. On the other hand, in the case of the P. xylostella -crucifer interaction, the higher ITC levels in the frass of silenced P. xylostella caterpillars may attract more P. xylostella (Pivnick, Jarvis, & Slater, 1994;Renwick, Haribal, Gouinguené, & Städler, 2006) as well as predators and parasitoids, potentially making these plants useful as "dead-end trap crops". However, the effects of such modifications on the populations of herbivore enemies are nuanced and species-specific, and the ecological ramifications of such an approach in both natural and agricultural settings need further research. In this study, we demonstrated that P. xylostella detoxification of glucosinolates enhances the performance of D. semiclausum. Enhanced parasitoid performance may positively impact parasitoid population dynamics and enhance their recruitment by plants, which may ultimately have a positive effect on plant fitness.
In conclusion, the desulphation of plant glucosinolates by the specialist herbivore P. xylostella prevents the formation of toxic ITCs and thus increases insect growth, survival and reproductive success. Here we show that this detoxification reaction also benefits a representative of the next trophic level, D. semiclausum, a widespread endoparasitoid of P. xylostella caterpillars, by increasing adult emergence and decreasing development time. Future research on other endoparasitoids is needed to determine if the overall susceptibility of D. semiclausum to plant defences in its herbivore host is a general trait of this group. Glucosinolate desulphation also appears to facilitate the action of symbiotic polydnaviruses of D. semiclausum that suppress host immunity, but more work is needed to understand the mechanisms responsible for improved parasitoid performance. Daniel G. Vassão https://orcid.org/0000-0001-8455-9298