Tetranychus evansi spider mite populations suppress tomato defenses to varying degrees

Abstract Plant defense suppression is an offensive strategy of herbivores, in which they manipulate plant physiological processes to increase their performance. Paradoxically, defense suppression does not always benefit the defense‐suppressing herbivores, because lowered plant defenses can also enhance the performance of competing herbivores and can expose herbivores to increased predation. Suppression of plant defense may therefore entail considerable ecological costs depending on the presence of competitors and natural enemies in a community. Hence, we hypothesize that the optimal magnitude of suppression differs among locations. To investigate this, we studied defense suppression across populations of Tetranychus evansi spider mites, a herbivore from South America that is an invasive pest of solanaceous plants including cultivated tomato, Solanum lycopersicum, in other parts of the world. We measured the level of expression of defense marker genes in tomato plants after infestation with mites from eleven different T. evansi populations. These populations were chosen across a range of native (South American) and non‐native (other continents) environments and from different host plant species. We found significant variation at three out of four defense marker genes, demonstrating that T. evansi populations suppress jasmonic acid‐ and salicylic acid‐dependent plant signaling pathways to varying degrees. While we found no indication that this variation in defense suppression was explained by differences in host plant species, invasive populations tended to suppress plant defense to a smaller extent than native populations. This may reflect either the genetic lineage of T. evansi—as all invasive populations we studied belong to one linage and both native populations to another—or the absence of specialized natural enemies in invasive T. evansi populations.


| INTRODUC TI ON
Plants and herbivores share a 420 million year history of antagonistic coevolution (Labandeira, 1998). Over this time, these adversaries have been in an arms race of adaptations and counter-adaptations.
Understanding why defense suppression is a successful herbivore offense strategy requires insight into its evolutionary costs and benefits (Blaazer et al., 2018). A benefit of defense suppression for herbivores is that it prevents expression of plant defense, which would otherwise have resulted in reduced herbivore performance Musser et al., 2002). At the same time, however, defense suppression creates a hospitable, nutritious plant (i.e., a "public good," Rankin, Bargum, & Kokko, 2007) from which competitors and natural enemies can also benefit (Alba, Glas, Schimmel, & Kant, 2011;Ataide et al., 2016;Kant et al., 2015). Such biotic interactions introduce new costs. A prime example of such costs can be found in Tetranychus evansi spider mites (Blaazer et al., 2018). By suppressing tomato defense, T. evansi increase not only their own performance but also that of competing Tetranychus spp. spider mites Godinho, Janssen, Dias, Cruz, & Magalhães, 2016;Sarmento, Lemos, Bleeker, et al., 2011). In addition, defense suppression by T. evansi exposes their offspring to increased predation by Phytoseiulus longipes predatory mites, possibly due to reduced transfer of defensive compounds from plants through spider mites into their eggs (Ataide et al., 2016). Defense suppression may therefore entail considerable costs depending on the biotic environment in which it is employed (Glas et al., 2014;Schimmel, Ataide, Chafi, et al., 2017).
To assess the role of biotic interactions in the evolution of defense suppression, it is pivotal to quantify variation in defense suppression across different biotic environments. In Leptopilina boulardi parasitoid wasps, for example, genotypes from different localities were found to suppress the immune system of their Drosophila hosts to varying degrees, depending on the abundance of specific host species (Dupas & Boscaro, 1999). Yet, in herbivores, variation in defense suppression has hitherto been investigated only scarcely . Here, we aim to quantify variation in the magnitude of suppression of plant defense among populations of the defense-suppressing spider mite T. evansi and secondarily to explore whether differences relate to characteristics of their biotic environments.
Specifically, we assessed if variation was explained by the host plant from which populations were sampled and by the presence or absence of specialized natural enemies from native and non-native environments.
The tomato red spider mite T. evansi Baker and Pritchard (Acari: Tetranychidae) is a herbivorous spider mite from South America, feeding mainly from solanaceous host plants such as tomato, potato, and eggplant (Migeon & Dorkeld, 2018;Navajas, Moraes, Auger, & Migeon, 2013). Spider mites are cell content-feeders, piercing plant parenchyma cells with their stylets, sucking up the contents, and leaving behind empty cells that are visible as white feeding scars (Bensoussan et al., 2016). T. evansi occurs mostly in tropical, subtropical, and Mediterranean climates, and can reach fast intrinsic rates of population increase due to its short generation time (<15 days), especially at high temperatures (Bonato, 1999;Gotoh et al., 2010).
Over the last twenty years, T. evansi has become invasive in many areas with subtropical and Mediterranean climates, such as sub-Sahara Africa, the Mediterranean region, and East Asia . In its invasive range, T. evansi replaces T. urticae and other indigenous spider mite species as the dominant species in spider mite communities and colonizes new host plant species (Azandémè-Hounmalon et al., 2015;Ferragut, Garzón-Luque, & Pekas, 2013).
The mechanism of defense suppression by T. evansi has mostly been studied on cultivated tomato plants, Solanum lycopersicum.
In tomato, the induced defense response against spider mites is orchestrated by the plant hormones JA and SA Li, Williams, Loh, Lee, & Howe, 2002). These hormones set in motion internal signaling cascades leading to the production of defense-associated compounds such as proteinase inhibitors (PIs) and polyphenol oxidases (PPO; Arnaiz et al., 2018;Kant et al., 2004;Martel et al., 2015). T. evansi suppress tomato defense downstream of plant hormone accumulation, such that the plant's expression of defense-associated genes is downregulated to the benefit of the herbivore Ataide et al., 2016;Sarmento, Lemos, Bleeker, et al., 2011). In the course of the infestation, such suppression can temporarily result in expression levels of defense genes at or below the plant's housekeeping levels (Sarmento, Lemos, Bleeker, et al., 2011) but for most of the time in an intermediate level of induction Schimmel et al., 2018). This suppression is mediated by secreted salivary effector proteins Villarroel et al., 2016) that restrain the defense response to levels of induction low enough for the mite to tolerate (Ataide et al., 2016), independent of herbivore-associated bacteria . Expression levels of defense-associated plant genes therefore are an adequate measure of defense suppression by T. evansi, as long as timing is standardized and a benchmark treatment for defense induction is included Sarmento, Lemos, Bleeker, et al., 2011;Schimmel, Ataide, Chafi, et al., 2017).
The lack of natural enemies in areas where T. evansi is invasive may therefore reduce such ecological costs, allowing T. evansi to suppress plant defense more strongly as an evolutionary consequence of reduced antagonistic pleiotropy (Cooper & Lenski, 2000).
We sampled T. evansi mites from eleven locations across its native and invasive range (Figure 1), and measured their magnitude of defense suppression with two approaches. First, we measured the expression of a reporter gene for defense induction in pLAP-A1:GUS tomato plants. In these plants, the promoter of the β-glucuronidase (GUS) reporter gene is fused to the JA-dependent promoter of the plant defense-associated gene leucine aminopeptidase A1 (LAP-A1; Chao, Gu, Pautot, Bray, & Walling, 1999). When plant defense is induced, LAP-A1 is activated, and thereby also the GUS reporter, of which its activity can be determined in a fluorimetric assay (Jefferson, Kavanagh, & Bevan, 1987). Because these assays were more variable than expected, we also measured the level of gene expression the defense-associated tomato genes LAP-A1, polyphenol oxidase-D (PPO-D), proteinase inhibitor IIc (PI-IIc), and pathogenesis-related protein 1a (PR-1a) through quantitative reverse transcription-polymerase chain reaction (qRT-PCR). T. evansi suppresses these genes in tomato Sarmento, Lemos, Bleeker, et al., 2011;Schimmel, Ataide, Chafi, et al., 2017). In all experiments, we included a treatment where tomato plants were infested with a defense-inducing T. urticae genotype as a benchmark for defense induction. For those genes where we observed different levels of expression among populations, we investigated if this variation was affected by host plant species or geographical range (invasive or native). We also verified that variation was not due to the identity of their host plants in the laboratory, or the time that populations had been maintained there. Last, to investigate the possibility that differences in suppression could be explained by differences in genetic lineage (Boubou et al., 2012;Gotoh et al., 2009), we sequenced a part of the mitochondrial cytochrome oxidase subunit 1 gene (CO1).

| Spider mite populations and culture conditions
We obtained eleven T. evansi populations that had been collected by other research groups from several locations across South America, Europe, Africa, and Asia ( Figure 1, Table 1). Because a change in host plant can have drastic consequences for genetic and phenotypic variation within a population (Dermauw et al., 2013;Magalhães, Blanchet, Egas, & Olivieri, 2009;Wybouw et al., 2015), we maintained them on the same host plant as they had been on in the research group from which we obtained these populations, that is, ached S. nigrum leaves or S. lycopersicum leaflets. We placed leaves and leaflets with their abaxial side facing upwards on wet cotton wool in open plastic trays in a controlled environment (25°C; 16:8 hr light: dark photoperiod; 60% relative humidity). We grew plants in a greenhouse (25:18°C; 16:8 hr photoperiod; 50%-60% relative humidity) for 4-5 weeks before leaves were used to feed the mite cultures.

| Infestation treatments and sampling
We obtained pLAP-A1:GUS seeds from Linda Walling (University of California, Riverside, USA). In these plants, the promoter of the β-glucuronidase (GUS) reporter gene is fused to the JA-dependent promoter of the plant defense-associated gene leucine aminopeptidase A1 (LAP-A1; Chao et al., 1999). When plant defense is induced, LAP-A1 is activated, and thereby also the GUS reporter, of which its activity can be determined in a fluorimetric This population was referred to as "BP" by Gotoh et al. (2009Gotoh et al. ( , 2010, and as "Vicoça-1" by Alba et al. (2015). We choose to use the latter name, because it was first collected in Viçosa and described by Sarmento et al. (2007). b Population Sde Eliyahu-1 was collected from potato (S. tuberosum) and eggplant (S. melongena) in a field where also tomato was grown, but none of the tomato plants were infested with T. evansi. We reasoned that this lack of preference for tomato could potentially be caused by a different defense suppression phenotype, which we preferred not to select against in laboratory cultures. However, we did not have potato or eggplant leaves available at the time this population arrived, and therefore chose the tomato mutant defenseless-1 (def-1), which does not accumulate JA after spider mite feeding (Li et al., 2002), as a host. If the lack of preference for tomato in this population was caused by an inability to suppress JA-dependent tomato defense, then we prevented selection for more potent suppressors by using def-1 host plants.

| Leaf damage quantification
We quantified leaf chlorotic spots as measure of leaf damage. Even though these spots contain not only cells that collapsed after they were emptied by the mite but also neighboring cells that collapsed later without being eaten, they are routinely used as a proxy for mite feeding intensity (Bensoussan et al., 2016). We quantified the damaged area of each infested leaflet using ImageJ version 1.49 (Rasband, 2016). We transformed RGB-colored scans of damaged leaflets to black and white images using the Type tool and distinguished damaged from nondamaged leaf area by applying a color threshold typical for spider mite leaf damage using the Adjust Threshold tool. After this step, leaf damage appears as black spots while undamaged leaf surface was white. The background was automatically transformed to dark and ignored during the measurements. We then selected the damaged area within the leaf with the Selection tool and measured damaged leaf area in mm 2 by using the Analyze Particles tool. We averaged leaf areas across the three damaged leaflets into one value per plant. Each scan included a piece of millimeter paper to ensure accurate scaling of leaf size and damaged surface area.
Plants usually respond in a dose-dependent manner to spider mite damage (Gols, Roosjen, Dijkman, & Dicke, 2003;Horiuchi et al., 2003) and herbivory in general (Agrawal, 2004;Niinemets, Kännaste, & Copolovici, 2013). Therefore, we normalized our measurements of tomato gene expression (GUS assays and qRT-PCR measurements) to the absolute amount of leaf damage, to correct for variation due to differences in damaged tissue. Normalization is appropriate, because previous work showed that defense gene induction correlates with damage ; note that as a result of the plant's wound response (Wasternack et al., 2006), induction of jasmonate defenses progresses with mite feeding damage .
We also present non-normalized averages to allow comparison to the uninfested control treatments which have no feeding damage.

| Protein extraction and total protein quantification
We ground frozen leaf material in 15 ml tubes by vortexing for 15 s while using two slim metal rods to crush the leaflets. We repeated this step four times. We then transferred leaf material to 2-ml Eppendorf tubes and manually ground it to fine powder using a sterile pestle for 15 s and repeated manual grinding three times.
During both grinding methods, we kept our samples frozen and afterward stored them at −80°C. We extracted total protein by adding 300 µl extraction buffer (50 mM NaPO 4 (pH 7.

| RNA extraction and cDNA synthesis
Of all treatments (Table 2), we extracted total plant RNA from ground, frozen leaf tissue using the hot phenol method of Verwoerd, Dekker, and Hoekema (1989 TA B L E 2 Treatment details and sample size dNTPs, 0.5 μl RT), synthesized cDNA at 42°C for 60 min, and inactivated the RT enzyme at 70°C for 10 min. We diluted the resulting cDNA solutions 5 times.

| Gene expression assays (qRT-PCR)
To investigate the degree to which the T. evansi populations suppressed tomato defense, we measured expression of the defenseassociated tomato genes LAP-A1, polyphenol oxidase-D (PPO-D), proteinase inhibitor IIc (PI-IIc), and pathogenesis-related protein 1a (PR-1a) in all treatments (Table 2). We used the tomato actin gene as a housekeeping reference (Løvdal & Lillo, 2009 We performed qRT-PCR on an ABI 7500 Real-Time PCR system (Applied Biosystems, Foster City, USA) and prepared samples such that all genes for the same samples were run on the same plate, in duplo. The PCR program, quality control, and calculation of relative expression are explained in Appendix Note S1. We normalized relative expression to absolute feeding damage through dividing it by the damaged area in mm 2 .

| Statistics
We used R v3. as a fixed factor, as well as plant stage (categorical, 2 levels). We expressed the time that populations had been cultured in laboratory environments in an estimated number of generations, assuming a generation time of 14 days at 25°C (Bonato, 1999). We square root-transformed relative transcript abundance to meet model assumptions and assessed the significance of terms using approximate F tests with a Kenward-Roger approximation.

| CO1 sequencing
To determine the genetic lineage (Boubou et al., 2012;Gotoh et al., 2009) to which the T. evansi populations used in this study belong, we sequenced a part of the mitochondrial CO1 gene (Appendix Note S2). CO1 sequences were deposited in GenBank under accession numbers MT019694-MT019820 (Table S2).

| RE SULTS
To investigate variation in defense suppression among the T. evansi populations, we first quantified differences in feeding damage and assessed the magnitude of the JA-responses via measuring GUS activity in pLAP-1A:GUS plants. We found significantly different amounts of damage (7-28 mm 2 of leaf tissue per leaflet) among populations (F 10,112 = 3.99, p < .001, Figure S1). When normalized for feeding damage, GUS activity was highly variable but not significantly different among populations (F 10,112 = 0.78, p = .644, Figure   S2). We also observed low levels of fluorescence in some of the control treatments ( Figure S2B), which could be an indication of enzymatic activity in the absence of GUS. Possibly, the activity of tomato glycosidases other than GUS introduced some background variability in our measurements (Gu, Pautot, Holzer, & Walling, 1996).
To obtain more specific insight into the activation of tomato de- The T. urticae genotype Santpoort-2, our benchmark treatment for defense induction, induced higher expression than any of the T. evansi populations for all marker genes (all pairwise comparisons p < .05), except for LAP-A1 expression, which was similar between tomatoes infested with SC and Santpoort-2 (p = 1.00).
Although T. evansi was previously found to sometimes suppress tomato defense expression significantly below control levels (Godinho et al., 2016;de Oliveira, Pallini, & Janssen, 2016;Sarmento, Lemos, Bleeker, et al., 2011), we found expression levels to be similar to the levels in control plants or to be slightly higher for PPO-D, PI-IIc, and PR-1a and to be significantly higher for LAP-A1 ( Figure S3).
To further explore the observed variation in defense suppression among T. evansi populations, we assessed the correlation between marker gene expression levels and either geographical range, host plant, or time in culture. We found that invasive populations tended to suppress tomato defense less strongly than native populations, and this pattern was significant for the level of PR-1a expression ( Figure 3a). Expression levels did not correlate with the host plant species from which the T. evansi populations had been collected ( Figure 3b). Likewise, expression levels were similar among plants infested with T. evansi populations cultured on S. lycopersicum or S. nigrum ( Figure 3c) and did not correlate with the time that populations had been maintained in laboratory environments (Figure 3d).
To determine to which of the two genetically differentiated T. evansi lineages (Boubou et al., 2012;Gotoh et al., 2009)

| D ISCUSS I ON
Multiple arthropod herbivore species suppress the defenses of their host plants to prevent exposure to harmful plant defense and enhance herbivore performance Musser et al., 2002). Lowered plant defenses, however, may also increase the performance of competing herbivores and promote predation (Ataide et al., 2016;Glas et al., 2014;Kant et al., 2008;Sarmento, Lemos, Bleeker, et al., 2011;Schimmel et al., 2018;Schimmel, Ataide, Chafi, et al., 2017;. Biotic interactions among defense-suppressing herbivores and competitors or predators may therefore give rise to ecological costs associated with defense suppression and may vary among locations. To understand the role of biotic interactions in the evolution of defense suppression, it is necessary to quantify variation in defense suppression across different biotic environments. The purpose of this study was to investigate intraspecific variation in defense suppression among T. evansi populations from eleven locations and secondarily to explore if suppression differed across host plant species and native or non-native ranges. We found significant variation in expression of the JA-responsive marker genes LAP-A1 and PPO-D, and at the SA-dependent locus PR-1a (Figure 2). This shows that T. evansi populations suppress the two hormonal signaling pathways that regulate tomato defense expression against spider mites to varying degrees. Although the effect size of this variation was small relative to the magnitude of induction by the T. urticae genotype Santpoort-2, small differences in defense gene expression still can correlate with significant differences in spider mite performance .   (Cooper & Lenski, 2000), or it may erode through genetic drift (Halligan & Keightley, 2009). To obtain more insight into the F I G U R E 3 Variation in expression of defense-associated tomato genes compared between native and invasive populations (a), among field host plant species (b), laboratory host plant species (c), and depending on the time the populations have been cultured in the laboratory (d). Gene expression of LAP-A1, PPO-D, PI-IIc, and PR-1a was measured using qRT-PCR and expressed in transcript abundance relative to that of actin and corrected for differences in feeding damage. Details of statistical tests for differences in relative transcript abundance are given in the upper corners of each graph. In panels a-c, values were normalized to the lowest treatment median. Thick lines indicate treatment median, boxes encompass data from first to third quartile, whiskers indicate fences (nearest observed value ≥ first or ≤ third quartile ± 1.5 box height), and circles indicate outliers. In panel d, values were normalized to the smallest individual relative expression. Circles indicate data points and lines indicate linear model predictions for relative transcript abundance over time in culture effect of enemy release (Colautti, Ricciardi, Grigorovich, & MacIsaac, 2004;Jeffries & Lawton, 1984) on T. evansi defense suppression, future research could investigate which kind and which amounts of volatiles P. longipes needs to detect T. evansi-infested tomato plants (Sarmento, Lemos, Bleeker, et al., 2011). Insight into how suppression of plant defense affects the recruitment and performance of other natural enemies, such as N. floridiana fungi (Elliot et al., 2000;Hountondji, Sabelis, Hanna, & Janssen, 2005), also awaits further study.
Geographical range and genetic lineage are completely collinear variables in our dataset, and we cannot disentangle their effects on variation in defense suppression among T. evansi populations.
evansi lineages are partly reproductively isolated (Gotoh et al., 2009;Knegt et al., 2017). Differentiation between these lineages likely preceded invasion of areas outside South America, but among the invasive populations lineage I is more prevalent than lineage II (Boubou et al., 2012;Meynard, Migeon, & Navajas, 2013). Previous studies have found that lineage I tolerates colder temperatures than lineage II (Migeon, Auger, Hufbauer, & Navajas, 2015) and has higher expression of digestive proteases (Santamaría et al., 2018). Our results complement these findings by showing that lineage II tends to suppress tomato defenses more strongly than lineage I (Figure 3a), as all our invasive T. evansi populations belonged to lineage I and both native populations to lineage II. Therefore, an equally possible explanation for the observed trend in defense suppression between native and invasive populations is that already in South America differences among the habitats of the two T. evansi lineages, such as different abundance of competitors and predators, selected for different levels of defense suppression. Future work could confirm this hypothesis by characterizing more T. evansi populations from their native South American habitats.
The four Solanum host plant species used in this study vary in their defensive metabolites and may therefore harbor different arthropod communities (Cipollini & Levey, 1997;Girard et al., 2007;Hartl et al., 2010;Jared et al., 2016;Nohara et al., 2007;Tingey, 1984). Because the costs of defense suppression by T. evansi depend on biotic interactions with competitors and predators in these communities, we hypothesized that this variation could select T. evansi to suppress plant defenses to different degrees. However, we found no indication that host plant species identity explained variation in defense suppression among T. evansi populations (Figure 3b). Some of our mite populations were maintained on a different host plant than tomato, and these mites were therefore confronted with a new host F I G U R E 4 Phylogenetic relationships between Tetranychus evansi populations based on mitochondrial CO1 gene sequences (868 bp). Relationships were inferred using the maximum likelihood method and the general time-reversible model plus substitution rate variation among sites. Nucleotide positions with gaps or missing data (8.8%) were excluded. Branch support based on 5,000 bootstraps is indicated above each node. Populations of which CO1 sequences were collected in this study are shown in black font along with their sample size, and reference sequences are indicated by their GenBank accession numbers and shown in gray. The naming of the two differentiated T. evansi lineages is as in Boubou et al. (2012) plant during our experimental assay. In theory, this could cause extra effects on gene expression in the tomato plant other than due to differences in damage; however, we do account for the factor "host plant" in our statistical analysis. Therefore, the significant differences we find are supported even when effects of placing mites on a novel host plant were present. Future work could aim to characterize arthropod communities on these host plants in nature, to be able to assess their interactions with T. evansi and their potential effects on defense suppression in more detail.
The tomato genes assayed in this study constitute marker genes of tomato defense induction. This does not imply direct causal relationships between their gene products and spider mite performance. Although expression of PI genes and PI activity, for example, increase upon infestation with defense-inducing T. urticae (Ataide et al., 2016;Godinho et al., 2016;de Oliveira et al., 2016;Sarmento, Lemos, Bleeker, et al., 2011), and a weak negative correlation between PI activity and T. urticae (but not T. evansi) performance was observed (de Oliveira et al., 2016), the efficiency of these compounds as digestive inhibitors has been questioned because spider mite guts may lack their enzymatic targets (Arnaiz et al., 2018;Santamaría et al., 2012). Similarly, plant PPOs have been hypothesized to react with plant phenolic compounds in the herbivore gut after ingestion to produce quinones, which subsequently damage enzymes, membranes, and DNA (Constabel & Barbehenn, 2008), thus decreasing herbivore performance. However, because these processes might not be effective in spider mite guts due to their acidity (Erban & Hubert, 2010;Martel et al., 2015), the defensive role of PPOs against spider mites also awaits experimental confirmation. Since it is not known which tomato genes have a causal relationship with spider mite performance, these defense marker genes may paint an incomplete quantitative picture, and possibly we overlook relevant defenses with different induction and suppression kinetics. It would for example be interesting to also investigate the accumulation of steroidal glycoalkaloids, since these correlate with resistance of nightshades to T. evansi (Jared et al., 2016). Nevertheless, because T. evansi was previously shown to be sensitive to the magnitude of JA-defenses (Ataide et al., 2016), while PPO-D and PI-IIc have been shown to be reliable markers for the magnitude of this defense , our results must be largely relevant.

Suppression of plant defense by herbivorous arthropods is an
intriguing phenomenon due to its complex ecological consequences . Biotic interactions with competitors and natural enemies may shape the costs associated with defense suppression (Ataide et al., 2016;Glas et al., 2014;Sarmento, Lemos, Bleeker, et al., 2011;Sarmento, Lemos, Dias, et al., 2011), and we found variation in defense suppression among T. evansi populations from various locations, potentially related to their varying biotic environments.
Notably, however, T. evansi is not helpless against biotic threats. In response to the presence of competing T. urticae mites, T. evansi increases its web production to secure feeding sites (Sarmento, Lemos, Dias, et al., 2011) and increases fecundity to promote population growth (Schimmel, Ataide, Chafi, et al., 2017). Additionally, T. evansi males actively interfere with the reproduction of T. urticae females (Clemente, Rodrigues, Ponce, Varela, & Magalhães, 2016;Clemente et al., 2018;Sato, Alba, & Sabelis, 2014;Sato, Staudacher, & Sabelis, 2016). Moreover, in the presence of cues associated with P. longipes, T. evansi females choose to more often oviposit in their web, where their eggs are less prone to predation by P. longipes than on the leaf surface (Lemos et al., 2010). Although these traits may also entail costs (e.g., web production), they provide protection against competitors and natural enemies and thus "buffer" (Frank, 2007) T. evansi against the negative biotic consequences of defense suppression (Blaazer et al., 2018). In this context, future research could investigate if the degree to which T. evansi populations engage into such buffering behavior correlates with the variation in defense suppression observed in this study, because this may point toward ecological costs of defense suppression.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
All authors designed the research. BK and TTM: experimentation.
BK and TTM: analysis of the data. BK: writing-the manuscript.
All authors contributed to and approved the final version of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available at Dryad under https://doi.org/10.5061/dryad. np5hq bzpw (Knegt, Meijer, Kant, Kiers, & Egas, 2020