Release from natural enemies mitigates inbreeding depression in native and invasive Silene latifolia populations

Abstract Inbreeding and enemy infestation are common in plants and can synergistically reduce their performance. This inbreeding ×environment (I × E) interaction may be of particular importance for the success of plant invasions if introduced populations experience a release from attack by natural enemies relative to their native conspecifics. Here, we investigate whether inbreeding affects plant infestation damage, whether inbreeding depression in growth and reproduction is mitigated by enemy release, and whether this effect is more pronounced in invasive than native plant populations. We used the invader Silene latifolia and its natural enemies as a study system. We performed two generations of experimental out‐ and inbreeding within eight native (European) and eight invasive (North American) populations under controlled conditions using field‐collected seeds. Subsequently, we exposed the offspring to an enemy exclusion and inclusion treatment in a common garden in the species’ native range to assess the interactive effects of population origin (range), breeding treatment, and enemy treatment on infestation damage, growth, and reproduction. Inbreeding increased flower and leaf infestation damage in plants from both ranges, but had opposing effects on fruit damage in native versus invasive plants. Inbreeding significantly reduced plant fitness; whereby, inbreeding depression in fruit number was higher in enemy inclusions than exclusions. This effect was equally pronounced in populations from both distribution ranges. Moreover, the magnitude of inbreeding depression in fruit number was lower in invasive than native populations. These results support that inbreeding has the potential to reduce plant defenses in S. latifolia, which magnifies inbreeding depression in the presence of enemies. However, future studies are necessary to further explore whether enemy release in the invaded habitat has actually decreased inbreeding depression and thus facilitated the persistence of inbred founder populations and invasion success.


| INTRODUC TI ON
Understanding the forces that promote or prevent species range expansions remains a challenging goal in ecology (Barrett, 2015).
During invasion of a new range, populations can be simultaneously exposed to increased inbreeding following founder effects  and to substantial alterations in the biotic and abiotic environment (Catford, Jansson, & Nilsson, 2009).
In addition, inbreeding × herbivory interactions may provide a hitherto underappreciated explanation for invasion success in the face of repeated founder effects. During range expansion, plants often escape from their coevolved herbivores and pathogens, while host switching by native enemies in the introduced range mostly occurs with some time delay (Colautti, Ricciardi, Grigorovich, & MacIsaac, 2004;Dietz & Edwards, 2006;Mitchell, Blumenthal, Jarošík, Puckett, & Pyšek, 2010;Mitchell & Power, 2003). Hence, enemy attack is specifically reduced during initial introduction and towards the leading edge of range expansion, where inbreeding rates in plant populations are highest .
Enemy release may mitigate inbreeding depression in these founding plant populations, increase their persistence, and thus foster plant invasion success. Studies quantifying inbreeding depression in native and introduced plant populations in the presence versus absence of their native natural enemies are a first step to test this assumption.
Such studies can also yield information on how genetic differentiation among plant populations impacts the outcome of I × E interactions, which may help to explain reported inconsistency in their effects on plant fitness (Fox & Reed, 2011;Sandner & Matthies, 2016).
During invasions, plant species often evolve changes in performance traits (e.g., increased growth, reproductive output, and competitive ability) and chemical traits (reduced defense against specialists, increased defense against generalists, changes in inducibility and constitutive amounts of defense compounds, and increased allelopathy) (Agrawal et al., 2015;Joshi & Vrieling, 2005;Uesugi & Kessler, 2013).
This divergence can arise either from adaptive responses to changes in the selective regime for natural enemies and various other environmental factors (Atwood & Meyerson, 2011;Colautti & Barrett, 2013) or from genetic drift (Keller & Taylor, 2008;Lachmuth, Durka, & Schurr, 2011). Both adaptive and nonadaptive genetic differentiation may likely also have altered the genetic architecture underlying inbreeding depression and its dependency on herbivory, specifically through differences in the accumulation and purging (i.e., negative selection of deleterious recessive mutations in inbred populations) of genetic load in defense-related traits under past population bottlenecks : Lower herbivory pressure in invading plant populations may have lead to the accumulation of genetic load in defense traits, whereas high herbivore pressure in the native range may have lead to purging. If such range-dependent purging occurred under past population bottlenecks, this may magnify the effects of recent inbreeding × herbivory interactions on plant fitness in invasive relative to native populations.
Here, we investigate the combined effects of inbreeding and enemy infestation on the performance of native and invasive populations of Silene latifolia Poir. (Caryophyllaceae). During the invasive expansion from Eurasia to North America, the plant species experienced events conducive to the expression of I × E interactions: Introduced plants escaped their natural enemies (Wolfe, 2002) and experienced severe population bottlenecks (Keller, Gilbert, Fields, & Taylor, 2012;Taylor & Keller, 2007) as well as high inbreeding levels in founder populations (Fields & Taylor, 2014;Richards, 2000).
Moreover, Schrieber, Schweiger, Kröner, and Müller (2018) demonstrated that inbreeding diminishes metabolic responses to herbivory in populations from both distribution ranges. Finally, invasive populations evolved differences in enemy susceptibility and performance (Blair & Wolfe, 2004;Keller, Sowell, Neiman, Wolfe, & Taylor, 2009;Wolfe, Elzinga, & Biere, 2004), making S. latifolia ideally suited for examining the impact of genetic differentiation on the outcomes of I × E interactions. We conducted experimental in-and outbreeding within native and invasive S. latifolia populations, exposed the offspring to the absence and presence of native natural enemies, and measured traits related to infestation damage (inverse measure of defense), growth, and reproduction to address the following hypotheses: (a) Inbred plants incur higher infestation damage than outbreds. (b) Plant growth and reproduction are lower in inbreds than outbreds (inbreeding depression) and reduced in the presence as compared to the absence of natural enemies. (c) Inbreeding depression in growth and reproduction is stronger in the presence of natural enemies than in their absence (I × E interaction).
(d) The effects of inbreeding on infestation damage are stronger in invasive than native plants, which magnifies I × E interaction effects on growth and reproduction in invasive populations.

| Study system
Silene latifolia is a short-lived perennial herb mainly distributed across ruderal habitats. The plant is dioecious and produces sexually dimorphic flowers pollinated by insects. Females develop large numbers of capsules containing several hundred seeds, which lack a specific dispersal syndrome and are thus mainly dispersed passively and by human activities. Limited seed dispersal and restricted pollen transfer among neighboring plants can lead to restricted gene flow and the formation of kin-structured patches within populations (McCauley, 1994(McCauley, , 1997). These characteristics have been shown to result in high levels of biparental inbreeding in small, isolated, or recently founded S. latifolia populations (Fields & Taylor, 2014;Richards, 2000).
(Mycrobotryaceae)-a systemic sterilizing fungus; and Brachycaudus lychnidis L. (Aphididae)-an aphid that causes flowers to abort due to phloem feeding (Wolfe, 2002). Moreover, native populations are attacked by various leaf-and flower-feeding generalist herbivores, including slugs (mainly Arion lusitanicus Mabille (Arionidae)), beetles, thrips, caterpillars (often Mamestra brassicae L. (Noctuidae)), and leaf miners as well as by several generalist rust and mildew fungi . In the invaded range (North America), H. bicruris is completely absent (Wolfe, 2002), the occurrence of M. violaceum is locally restricted to a small region in Virginia (Antonovics, Hood, Thrall, Abrams, & Duthie, 2003), and the abundance of aphids as well as leaf-and flower-feeding generalists is very low relative to the native range (Wolfe, 2002). As a result of adaptive responses to changes in the selective regime concerning enemy attack and climate as well as of genetic drift effects, invasive S. latifolia populations exhibit higher growth, reproduction, and susceptibility to enemy infestation than native populations (Blair & Wolfe, 2004;Keller et al., 2009;Wolfe et al., 2004). A tradeoff between growth/reproduction and enemy susceptibility was not detected in this species .

| Field sampling and experimental setup
We collected open-pollinated seeds from eight native and eight invasive S. latifolia populations (Supporting Information Figure S1, Table S2). Sampling in the native range focused on regions thought to be the source of introduced populations (broadly, eastern and western Europe), while sampling in the invasive range comprised the geographic regions of initial introduction and early expansion (eastern North America), as identified by Taylor and Keller (2007) and Keller et al. (2012). Within each population, we sampled one capsule (maternal family) from each of five different female plants that were equally distributed over the population area and spatially separated from each other as far as possible (min. 6 m for one female pair in smallest population and ≥10 m for all remaining pairs). Using these field-collected families, we conducted two generations of experimental inbreeding and outbreeding within all native and invasive populations under controlled greenhouse conditions. The offspring were exposed to the absence and presence of natural enemies in a common garden in the species' native range. Data for the outbred plants from this experiment have previously been used to investigate adaptive and nonadaptive differentiation in growth, reproduction, and enemy susceptibility between the native and invaded range ).

| Experimental inbreeding and outbreeding
For the parental generation, we germinated ten seeds from each of the five field-collected families in 0.8 mM gibberellic acid in a germination chamber (16-hr light at 25°C, 8-hr dark at 13°C). After 6 days, the seedlings were planted into pots and transferred to the greenhouse (16-hr light at 25°C, 8-hr dark at 13°C) where they received weekly fertilization (Kamasol Brilliant Rot, Compo Expert, Münster, GE). After 7 weeks, we randomly chose one male and one female plant per family for the crossings. Each female received pollen from a sib male belonging to the same family (inbreeding), and pollen from a male belonging to a different family within the same population (outbreeding) at distinct flowers (Supporting Information Figure S3).
For the second generation, we randomly chose one capsule per combination and propagated the F1 plants from its seeds as described for the parental generation. Female inbred offspring received pollen from an inbred male from the same family, while female outbred offspring received pollen from an outbred male from a different family with respect to the relationships created in the first generation (Supporting Information Figure S3). For our breeding design, we decided against an independent pairing of partners or reciprocal crosses over two generations, since these approaches create bias either because they yield many more inbred than outbred lines (independent pairing) or because they do not use the same initial (P generation) gene pool for inbreeding and outbreeding, as more field-sampled plants are involved in creating the outbred lines than in creating the inbred lines (reciprocal crossing).
We lost seven of the 160 population × family × breeding treatment combinations due to lack of germination, high mortality, lack of flowering, or production of sterile flowers in both inbred and outbred families during the propagation of the F1 generation. Consequently, we obtained a total of 153 population × family × breeding treatment combinations for the F2-generation, which were used for the enemy release experiment.

| Enemy release experiment
We exposed native and invasive, inbred and outbred S. latifolia 11.878°E, alt: 116 m). The planting area was densely covered by a diverse plant community of grasses and forbs including a patchy population of S. latifolia that was infested by all of the above-mentioned specialist and generalist enemies. In the common garden, we established four vegetation-free belts, which comprised four 5 × 6.5 m plots, respectively (∑ = 16 plots) (Supporting Information Figure S4). Each plot included all native and invasive populations represented by two to three maternal families each with one inbred and one outbred individual. As such, the five families within each population were split between two plots (plot pair), which together comprised all of the 153 population × family × breeding treatment combinations. Each plot pair was replicated an additional seven times. While populations and families were planted randomly within the plots, the range and breeding treatments were uniformly distributed according to a fixed scheme (Supporting Information Figure   S4) in order to reduce confounding plot edge effects. Plots within pairs and plot pair repetitions were randomly distributed across the experimental area. We experimentally excluded natural enemies in eight of the plots (enemy exclusions) over a period of three months (Supporting Information Figure S4). For this purpose, we used slug fences coated with a gastropod deterrent (Schneckenabwehrpaste, Irka, Mietingen, GE), as well as a molluscicide (Limex, Celaflor), systemic insecticides (alternating between Calypso and Confidor, Bayer, Leverkusen, GE), and a systemic universal fungicide (Baycor M, Bayer, Leverkusen, GE), which were applied in a two-week cycle in accordance with the manufacturers' instructions. The remaining eight plots (enemy inclusions) were not treated with pesticides and therefore extensively colonized by specialist and generalist herbivores two weeks after the experiment was set up. The removal of vegetation, however, deterred A. lusitanicus from entering the inclusion plots, so we equipped them with slug fences whose impassable sides were turned toward the plot interior and introduced 15 A. lusitanicus individuals to each plot. This corresponds to the average number of slugs in four 5 × 6.5 m patches of undisturbed vegetation close to the experimental plots recorded at dusk on a humid day. We adjusted the number of slugs within each inclusion plot to 15 three times a week. The infection with specialist and generalist fungi remained low in all inclusion plots for the entire experimental period.
All plots were weeded weekly and watered when necessary during the experiment.
After three months of exposure to or protection from natural enemies, we collected data on defense-related traits in the enemy inclusion plots. We collected leaves at similar stages of development to determine trichome density in a 5 × 5 mm area away from the main vein and at the broadest section of the leaf.     n.e. (572) n.e. (571) n.e. (282) 0.544 (1192) 0.176 (1128) n.e. (1192) n.e. (579) Note. Performance responses are presented with the applied model type, error distribution, and link function. The table presents parameter estimates for each of the fixed-effect terms retained in the respective minimal adequate mixed-effects model (main effects presented with second factor level in parenthesis) with their levels of significance ( n.s. for the entire nonindependence arising from the individuals' relatedness. However, we consider the above-mentioned caveats that would have arisen from bias in reciprocal or independent pairings more severe.

| Statistical analysis
All models were fitted with a maximum likelihood approach. We validated the chosen model types, link functions, and data transformations by assuring that all (G)LMMs exhibit variance homogeneity and normal distribution of residuals via visual inspection of model checking plots (Zuur, Ieno, Walker, Saviliev, & Smith, 2009).
Moreover, GLMMs were tested for under-and overdispersion (R package: blemco, Korner-Nievergelt et al., 2015). The GLMMs for leaf damage and number of flowers were overdispersed and consequently complemented by an observational-level random factor in order to improve the model fit and avoid biased parameter estimates (Harrison, 2014(Harrison, , 2015

| Interactive effects of range and breeding treatment on defense-related traits
The density of leaf trichomes was not significantly influenced by range, breeding treatment, the interaction range × breeding treatment, or one of the covariates (Table 1, Figure 1a). The proportion of damaged leaves was significantly related to range and breeding treatment (Table 1). Invasive plants experienced more leaf damage compared to native plants (p = 0.02, χ 2 = 5.39), and inbred plants from both distribution ranges suffered stronger from leaf infestation compared to outbreds (p < 0.001, χ 2 = 41.69) (Figure 1b). The proportion of damaged flowers depended significantly on range, breeding treatment, and the covariate sex (Table 1). Flower infestation was higher for invasive than native (p = 0.01, χ 2 = 6.79), inbred than outbred (p < 0.001, χ 2 = 40.98) (Figure 1c), and male than female plants (p = 0.02, χ 2 = 5.22). The proportion of damaged fruits was significantly influenced by the interaction range × breeding treatment (p = 0.04, χ 2 = 4.12). Here, invasive plants received generally more fruit damage than native plants and fruit infestation was higher on inbred than outbred native plants but lower on inbred than outbred invasive plants (Figure 1d). Tukey post hoc comparisons among outbreds and inbreds within both ranges demonstrated that the inbreeding effect was not significant within the native (p = 0.64) and invasive range (p = 0.18).

| Interactive effects of range, breeding treatment, and enemy treatment on fitnessrelated traits
The aboveground biomass of experimental plants was significantly related to the interaction range × enemy treatment, to breeding treatment, and to plant sex (

| Population variation in I × E interaction effects on fitness-related traits
The magnitude of inbreeding depression (

| Inbreeding increases infestation damage in native and invasive plant populations
In accordance with our hypothesis, inbred S. latifolia plants from both distribution ranges for the most part incurred higher infestation damage from natural enemies in the common garden than outbreds (Figure 1b (Leimu, Kloss, & Fischer, 2008).
In contrast to leaf and flower damage, fruit damage was significantly contrarily affected by inbreeding in native and invasive populations ( Figure 1, Table 1). The proportion of fruits infested by H. bicruris was slightly higher in inbred than outbred native plants, but lower in inbreds than outbreds within invasive populations ( Figure 1d). Although the breeding effects within each range were nonsignificant, this finding highlights that genetic differentiation and demographic disequilibrium can synergistically shape the attractiveness of S. latifolia to H. bicruris, which is a complex trait composed of flower morphology, number, and size as well as the composition of floral volatiles (Dötterl et al., 2006;Dötterl, Jürgens, Wolfe, & Biere, 2009;Elzinga & Bernasconi, 2009). The attractiveness of S. latifolia to herbivores was shaped not only by inbreeding but also by plant sex. Males received significantly more flower damage than females (Table 1), likely because their higher flower number (Table 1) attracted more specialist aphids and generalist chewing-biting herbivores.

| Enemy release mitigates inbreeding depression in native and invasive plant populations
Both inbreeding and enemy infestation reduced the growth and reproduction of S. latifolia in native and invasive populations, whereby inbreeding had a pronounced effect and enemy infestation a rather weak effect ( Figure 2). As hypothesized, the effects of breeding and enemy treatment were not purely additive. The magnitude of inbreeding depression was independent of the enemy treatment for biomass, corolla diameter, and flower number (Figure 2a,b,c), but significantly lower in enemy exclusions than inclusions for fruit number ( Figure 2d).
While some studies found that herbivory increases inbreeding depression in multiple traits related to both growth and reproduction (Campbell et al., 2013;Carr & Eubanks, 2002), other studies also observed that I × E interactions only affect late live history traits very closely linked to reproductive success (Bello-Bedoy & Núñez-Farfán, 2011;Schou, Loeschcke, & Kristensen, 2015). The latter can occur, since the investment in reproduction by the end of a growing season is highly dependent on an individual's cumulative performance and thus on the cumulative effects of inbreeding and stress (natural enemies) on performance throughout the season (Orr, 2009) and stressful conditions should thus ideally quantify seed output, viability, and germination as well as demographic rates in order to parameterize models that estimate population growth and spread rates (Normand, Zimmermann, Schurr, & Lischke, 2014;Schultz, Eckberg, Berg, Louda, & Miller, 2017).

| Have I × E interactions contributed to invasion success in S. latifolia?
In contrast to our expectation, inbreeding effects on damage and I × E interaction effects on fitness were not more strongly pronounced in invasive than native populations, but rather equal in their magnitude in both ranges (Figures 1, 2)  . The absence of these differences can be explained with two alternative evolutionary scenarios.
First, it is not only the relaxation from selection, but also low natural degrees of inbreeding in the history of a population that can lead to the accumulation of genetic load in specific traits (Leimu et al., 2008;. Natural inbreeding exposes deleterious recessive mutations to negative selection. As a consequence, the frequency of these mutations within populations can rapidly decrease (i.e., purging of genetic load). If inbreeding levels are low, recessive mutations are masked in the heterozygous state, can be passed to the next generation, and thus accumulate in the population gene pool (Crnokrak & Barrett, 2002 (Table 1). This observation has also been made in previous studies on S. latifolia (Blair & Wolfe, 2004;Schrieber et al., 2018Wolfe et al., 2004) and suggests that invasive populations evolved increased tolerance of enemy infestation. The evolution of increased tolerance during range expansion has been observed in several other plant species (Abhilasha & Joshi, 2009;Zou, Rogers, & Siemann, 2007) and is assumed to arise from shifts in the natural enemy community, that is, reduced attack by specialists and increased attack by generalist (Fornoni, 2011).
Both of the alternative evolutionary scenarios outlined above are supported by our data, and they are mutually nonexclusive.
This highlights that the combined effects of inbreeding and enemy infestation depend on the population history for selection by herbivores (differences in herbivore abundance and species composition) as well as the population history for inbreeding (frequency and intensity). Both of these factors can cause strong variation in inbreeding effects among distribution ranges (Figures 1d, 2d) and among populations within distribution ranges (Figure 3

| CON CLUS I ON S AND PER S PEC TIVE S
Our findings demonstrate that enemy release can mitigate inbreeding depression in plant populations. This supports the idea that I × E interactions have the potential to contribute to the successful establishment and expansion of introduced populations. On the other hand, I × E interactions might hamper the colonization of novel habitats that exhibit increased stress levels relative to a species' native source habitat (Hufbauer et al., 2013;Rosche, Hensen, & Lachmuth, 2017). Furthermore, our data illustrate that the inbreeding effects on an organism's interaction with its environment are shaped by the evolutionary histories of populations. As the native and invaded range of a species can differ systematically in the stress regimes they experience, ongoing invasions provide ideal study systems for investigating the effects of evolutionary differentiation on the outcomes of I × E interactions, and how, in turn, the different outcomes may alter the evolutionary trajectories of invasive populations. Studies addressing these issues hold implications that extend far beyond invasive model species. I × E interactions may potentially shape the dynamics of natural populations whenever they are simultaneously exposed to habitat change and increased inbreeding rates following founder effects or population size reductions. These conditions occur not only during species range expansions, but also during range shifts and retractions in the course of global change (Colautti et al., 2017).

ACK N OWLED G EM ENTS
This study was financially supported by the German Academic

DATA ACCE SS I B I LIT Y
All data that support this article have been deposited in Dryad (https://doi.org/10.5061/dryad.fb851dn).