Ambrosia artemisiifolia (Asterales: Asteraceae), common ragweed, is an annual monoecious outcrossing plant native to North America, accidentally introduced in Asia, Australia and Europe, where it became an invasive alien species. It has significant negative effects on human health due to its highly allergenic pollen and on the yield in spring-sown crops, like sunflower and beets (Essl et al. 2015; Mouttet et al. 2018; Müller-Schärer et al. 2014; Schaffner et al. 2020).

The oligophagous leaf beetle O. communa has been used in China as a BCA since 2001, which is a natural enemy of Ambrosia in its native region in North America. It was accidentally introduced in Europe and first discovered in 2013 in Northern Italy and Southern Switzerland (Müller-Schärer et al. 2014). Ophraella development is composed of an egg stage, three larval instars (L1, L2 and L3 instar are used in the following text), a pupal stage and finally the adult stage (Suppl. material 1: Fig. S1; Zhou et al. 2010b). Larval and pupal development takes 7–17 days and 6–12 days, respectively (Lommen et al. 2017b; Zhou et al. 2010b). Each Ophraella stage can live and develop on Ambrosia (Yamazaki et al. 2000), with 4–5 generations in Northern Italy (Müller-Schärer et al. 2014), but up to 6–7 generations per year in southern China (Chen et al. 2013)

In order to use genotypes, we assumed to be genetically most distinct, we used Ambrosia seeds from 11 mother plants (half-sibs) from 11 regions in three continents (Table 1, Suppl. material 1: Fig. S2) and Ophraella from eight couples (full-sibs) from eight regions in three continents (Table 1, Suppl. material 1: Fig. S2). Ambrosia seeds were collected between 2013 and 2016 and conserved in paper envelopes in the cold chamber at the University of Fribourg (constant temperature: ~ 5 °C). Ophraella were collected in the field in 2017 and 2018 (cf. Table 1) and reared on Ambrosia plants grown from seeds of a mixture of Italian populations in the quarantine facility at the University of Fribourg (22 ± 5 °C, 16:8 L:D cycle) for either two generations (Chinese locations) or one generation (all other locations) before their use in the experiment. This was done to reduce environmental maternal effects and to align the Ophraella development of the various collections with the phenology of the Ambrosia cohorts. The permit to import A. artemisiifolia and O. communa into our quarantine facility at the University of Fribourg was issued by the Swiss Federal Office for the Environment (permit number A130598-3).

To characterize the bioclimatic conditions of the sampled Ambrosia and Ophraella locations, interpolated GIS data were extracted for 19 climate factors from WorldClim collected over 30 years at 5 minutes spatial resolution (https://www.worldclim.org/). To illustrate eco-climatic variation among the different sampling regions, we performed a principal component analysis (PCA) and compared the similarity with a Mantel test, using 999 permutations and pairwise tests between organisms. Both sampled Ambrosia and Ophraella genotypes spread over a large eco-climatic range, but when the two species were superposed, the ellipses do not differ between the two species, indicating a high eco-climatic correspondence between the plant and herbivore populations sampled (Suppl. material 1: Fig. S3).

In 2018, seeds were germinated in Petri-dishes on double thickness moistened filter paper in the growing chamber (19 ± 5 °C, 14:10 L:D cycle). On the day of germination, seeds were transferred in trays filled with commercial soil (Proter + Pro type 4, Fenaco Genossenschaft, Switzerland, containing 150 mg/l N, 350 mg/l P₂O₅, 800 mg/l K₂O, salt content < 3 KCl et pH = 6.2 (CaCl₂) until they reached six leaves. Plants were then transplanted into 1 L pots filled with the commercial soil, vermiculite (Vermica AG, Bözen, Switzerland) and sand (2:2:1) and kept in the greenhouse of the University of Fribourg (25 ± 5 °C under a 16:8 L:D cycle). Every second day, new leaf (leaves > 5 mm long) pairs (nodes) were counted to account for leaf age when later used in the experiment. In order to have leaves of the same age available (cf. below) for our tests, we repeatedly produced new cohorts of the plant genotypes. Plants were watered equally with 250 ml every second day.

Eight locations of Ophraella were reared on Italian Ambrosia plants in cages in the quarantine facility of the University of Fribourg. Each day, pupae from the cages were isolated in Petri-dishes on a clean filter paper. After emergence, several virgin Ophraella couples were randomly formed within each location. Each couple was isolated in Petri-dishes (Ø90 mm) with filter paper and fed with a fresh Ambrosia leaf from rearing plants by inserting the petiole in a wet floral foam. A single couple per location was selected for the experiment that laid enough eggs, to test full-siblings representing one Ophraella genotype (see Suppl. material 1: Fig. S4 for details). Each day, the selected couples were checked for new egg batches. Leaves with eggs were collected and isolated and a new leaf was added in the Petri-dish. Isolated egg batches were checked each day for egg hatching and L1 instar larvae were transferred with a paintbrush on the leaf from a specific Ambrosia genotype. Each larva was fed on 2 leaves (10–20 days old) during its development, the first leaf for L1 – L2 instar, and the second leaf for L3 instar to pupal instar, with both leaves coming from the same plant genotype. We opted for three replicates for all paired combinations of the 11 Ambrosia and 8 Ophraella genotypes by first transferring each one L1 instar larvae on three test leaves of the same plant genotype. In case an L1 instar died before developing to L2, which might have been caused by injuring the small larvae during the transfer on the leaves, we started a new test directly with a L2 instar larva from the same egg batch and a new leaf from the same plant individual. A few interactions could not be tested due to the lack of Ambrosia plants or Ophraella eggs. In total, we performed 233 tests by successfully transferring L1 and an additional 52 tests by transferring L2, resulting in a total of 285 tests.

Larvae were checked daily for mortality and instar. Leaf area consumed during L2 and L3 larval instar was measured by comparing the leaf area before and after feeding. We scanned the leaf at the beginning and end of the L2 (leaf one) and at the beginning and end of L3 (leaf 2) using ImageJ software (v1.51k) to measure the leaf area. The difference between the two measurements equals the leaf area consumed by the larvae. The sex of newly emerged adults was determined with a binocular microscope and beetles were conserved in the freezer (-20 °C). Adult fresh weight and oven-dried weight (60 °C for 24 h) of each adult was measured using a Microbalance (Mettler MT-5, Mettler-Toledo, Inc., Columbus, OH, USA) with a resolution of 1 mg. The dry weight of each individual was then subtracted from its fresh weight to calculate the percentage of water (Zhou et al. 2011), as

$\%\text{ water content }=\frac{\text{ fresh weight-dry weight }}{\text{ fresh weight }}$.

We built 11 × 8 matrices representing Ambrosia-Ophraella genotype interactions for the different performance traits of Ophraella larvae and adults and for the leaf consumption, with x-axis for the Ambrosia and the y-axis for the Ophraella genotypes. Effects of plant and herbivore genotypes and their interactions on Ophraella performance were assessed using linear mixed-effects models and generalized linear mixed-effects models (LMM/GLMM) and fit using the lmer/glmer function obtained from the R package lme4 that uses maximum likelihood to estimate the model parameters (Bates et al. 2014). In the models, the plant and the herbivore genotype and their interactions were included as fixed effects and leaf node level and sex (for adult biomass only) as random effects. Normality of the residuals of all models was assessed using QQ-plots. For the survival and the developmental time, GLMM with Binomial distribution and with Poisson distribution, respectively, was used; and LMM for leaf consumption, adult dry weight and adult water content. Mixed-effect regression models were used to analyze the correlation between dry biomass of Ophraella adults and total leaf area they consumed. Plant and herbivore genotype and their interactions were also included as fixed effects, leaf node level and sex as random effects. If a significant interaction term was detected, we analyzed each of the sex separately. We adjusted p-values using the Bonferroni–Holm method to correct for type 1 error. All statistical analyses were run with R version 3.6.1 (2019).

We found significant effects of Ophraella genotypes on L1, L2 and L3 survival (c² ≥ 15.14, P_adj ≤ 0.05; Table 2), but not on pupal survival (c² = 10.83, P_adj = 0.16), whereas no significant effects were found for Ambrosia genotypes on L1, L2, L3 and pupal survival (c² ≤ 15.7, P_adj ≥ 0.62; Table 2). There were no significant effects of Ambrosia-Ophraella genotype interactions (c² ≤ 70.23, P_adj ≥ 0.26). In general, we found significant effects of Ophraella genotype (c² = 26.20, p < 0.001), but no effects of Ambrosia genotype and Ambrosia-Ophraella genotype interactions (c² = 7.37, P_adj = 1 and c² = 65.92, P_adj = 0.47; respectively) on adult emergence (Fig. 2); with the lowest emergence of Ophraella genotype from CN-GX (Fig. 2). European Ophraella genotypes (CH and Italy) had the best survival performance until adult, whereas no CN-GX larvae reached adult emergence (Table 2 and Suppl. material 1: Fig. S5).

Figure 2.

Effect of Ambrosia artemisiifolia and Ophraella communa genotype on Ophraella adult emergence (survival up to adults; up left), death stage (stage when died; up right), total developmental time (L1 to adult, gray cells indicate tests without adults emergence); bottom left), and on leaf area eaten during L2 and L3 instar (bottom right). Blank cells represent missing data. Dark-colored horizontal line means that the interaction is driven by the Ophraella genotype, dark-colored vertical line means that the interaction is driven by the Ambrosia genotype.

There were significant effects of Ophraella genotype on developmental time of L2, L3 and pupae (c² ≥ 19.05, P_adj ≤ 0.008; Table 2), but not on L1 developmental time (c² = 4.17, P_adj = 0.76). Ambrosia genotype did not affect any developmental stage (c² ≤ 8.51, P_adj = 1; Table 2), nor did the Ambrosia-Ophraella genotype interactions (c² ≤ 28.11, P_adj = 1). In general, total developmental time in days from L1 instar till adult emergence showed a significant difference among Ophraella genotypes (c² = 13.29, P_adj = 0.05), but not among Ambrosia genotypes and for their interaction (c² = 1.4, P_adj = 1 and c² = 4.29, P_adj = 1; respectively) (Fig. 2), with the CN-HN genotype showing the fastest developmental time (Table 2 and Suppl. material 1: Fig. S6).

Both Ophraella and Ambrosia genotype affected the amount of leaf area eaten from L2 to pupae (c² ≥ 24.79, P_adj ≤ 0.01; Table 2, Fig. 2), without a significant interaction term (c² = 43.99, P_adj = 0.26; Table 2). CN-GX Ophraella genotype consumed the lowest and European Ophraella genotypes (CH and Italy) the largest amount of leaf area, and the Ambrosia genotypes from Germany and France were consumed the least (Suppl. material 1: Fig. S7).

We found significant effects of Ophraella genotype on adult dry weight (c² = 36.10, P_adj < 0.001) (Suppl. material 1: Fig. S7, Table 2) and marginally on water content (c² = 13.80, P_adj = 0.06), while no effect was found on these traits by Ambrosia genotype (c² ≤ 13.58, P_ad ≥ 0.62; Table 2). Ambrosia-Ophraella genotype interactions were significant for Ophraella dry weight (c² = 62.10, P_ad = 0.02), but not for water content (c² = 34.11, P_ad = 0.52; Suppl. material 1: Fig. S7). Ophraella females were significantly larger than males (t = -3.765, P < 0.001; Suppl. material 1: Fig. S8). We further found a significant correlation between beetle dry biomass and total Ambrosia leaf area consumed (R² = 0.07, P = 0.004; Suppl. material 1: Fig. S8).

We selected our Ambrosia and Ophraella genotypes from different continents and locations to reach a high genetic diversity for our tests. This genetic diversity also reflects the observed high within population genetic diversity found in European Ambrosia populations targeted for biocontrol (Genton et al. 2005; McGoey et al. 2019; van Boheemen et al. 2019). Our results of Ophraella performance provide strong evidence that the Ophraella genotype is driving the plant-herbivore interactions studied. We found significant effects of the different Ophraella genotypes on 11 of the 13 traits measured, including herbivore survival, developmental time, adult weight and food consumption. Higher survival of pupae than of larvae is in line with previous findings (Zhou et al. 2010a; Zhou et al. 2011). Moreover, seven traits were only significantly affected by Ophraella genotype including L1 and L3 survival, adult emergence, developmental time and relative water content of adults. Previous studies on Ophraella showed that a higher survival and faster developmental time would result in a higher population increase and more efficient control of Ambrosia (Augustinus et al. 2020). Furthermore, larger body size in insects is generally linked to greater fecundity and access to mates (Blanckenhorn 2000) and is thus expected to further enhance population build-up and expansion in O. communa (Chen et al. 2014). The Ophraella genotype from Linxiang, China (CN-HN) clearly showed the fastst developmental time on all Ambrosia genotypes, followed by the three European genotypes, which did slightly less well for this trait, but similarly well for survival and leaf area consumed. This might indicate that the studied three genotypes from Europe could be genetically closely related, such as when deriving from a single introduction event, but this needs to be further verified. Their overall good performance on most Ambrosia genotypes tested is so far a good sign for a successful biocontrol outcome in Europe.

With regard to outcomes for biocontrol management using Ophraella to suppress Ambrosia populations, our findings thus follow scenario 2 outlined in Fig. 1, indicating an optimal situation at least for a short- to mid-term biocontrol management. This result could be due to the shorter generation time of the antagonist as compared to its host plant, leading to a more virulent antagonist genotype (Kaltz et al. 1999), or due to its oligophagous nature, which allows Ophraella to deal with a large diversity of plant defense chemicals (Ali and Agrawal 2012). The fact that the Ambrosia genotype only differed for total leaf area consumed by the larvae, but did not influence survival and developmental time, or the weight of the herbivore, may indicate that Ophraella can compensate for observed differences in secondary plant metabolites in Ambrosia (Fukano and Yahara 2012; Sun and Roderick 2019; Wan et al. 2019) by adapting their feeding rate (Müller et al. 2006). Significant Ambrosia-Ophraella G × G interactions were found only for two out of 13 variates measured, i.e., L2 survival and dry weight of adults.

We are aware that our data on the beetle development and growth are considerably better than the data collected from the plant, i.e., that one effect was tested more thoroughly than the other. Our findings also might have been influenced by using cut leaves, known to elicit induced defense mechanisms (Beck et al. 2014), but this may hardly have changed our overall findings of the Ophraella genotype driving herbivore performance. Furthermore, we also acknowledge that there might be various trade-offs with larval performance that will affect the level of impact on the Ambrosia genotypes, such as host preference for oviposition, plant growth rate, regrowth capacity and tolerance (Zytynska and Preziosi 2011).

In its introduced range, the level of genetic variation of a plant invader can vary from a single genotype as for Rubus alceifolius in La Reunion and Mauritius (Amsellem et al. 2001) and location-specific haplotypes of the weed fern Lygodium microphyllum (Goolsby et al. 2006), to a few biotypes such as for the Asteraceen Chondrilla juncea in Australia (Espiau et al. 1997; Gaskin et al. 2013) and up to a high within population genetic variation as found in Lantana camara in South Africa (Mukwevho et al. 2017) and Ambrosia artemisiifolia populations in Europe, where the within population variation even exceeds the level reported from their native range (van Boheemen et al. 2017). It would thus be helpful to know early in a biocontrol project, if a few genotypes or a single BCA biotype could efficiently control the target weed population, or whether multiple genotypes and populations of a BCA are needed, which would involve more time and money.

Based on these settings, we can distinguish three scenarios, with greatly different outcomes for a biocontrol management success (Fig. 1). The first scenario designates the situation where some specific plant genotypes are resistant or tolerant to all antagonist genotypes tested (Fig. 1; Underwood and Rausher 2000). In this case, the plant genotype drives the plant-antagonist interaction, such as when using insect- or pathogen-resistant/tolerant crop cultivars (Moreau et al. 2006; Scott et al. 2010). Genotypes of a plant invader in the introduced range may originate from an area with heavy attack by the BCA that was subsequently introduced to the same location. Such a long-evolved association may have resulted in a homeostatic relationship with little impact of the BCA (cf. Hokkanen and Pimentel 1989), but more recent evidence for such an outcome is yet missing. Furthermore, resistance or tolerance of the plant invader to the BCA could also arise through new, potentially transgressive genotypes resulting from admixtures after multiple introductions (e.g., for A. artemisiifolia, Genton et al. 2005; van Boheemen et al. 2017; for Silene latifolia, Wolfe et al. 2007), or through interspecific hybridization (e.g., for Tamarix spp., Gaskin and Kazmer 2009; and for Fallopia spp., Gammon et al. 2007; Krebs et al. 2010) in the introduced range. However, we are not aware of cases from the weed biocontrol literature that such genotypes of the plant invader were resistant or tolerant to introduced BCA, which leaves the scenario 1 presently rather theoretical. In the second scenario, the antagonist genotype drives the plant-antagonist interaction, i.e., when specific antagonist genotypes affect all or at least most of the plant genotypes, including the dominant ones (Fig. 1; Lommen et al. 2017a; Roderick et al. 2012; Wajnberg 2004). This is expected to result in an unstable co-occurrence pattern with an effective and at least initially sustainable biocontrol (Lommen et al. 2017a). This scenario allows selecting genotypes or biotypes of more effective BCAs, which presently are mainly used in augmentative biocontrol (Lommen et al. 2017a; Szűcs et al. 2012). Thirdly, specific plant genotype by antagonist genotype interactions in a population are expected to result in an overall co-occurrence by maintaining the genetic diversity of both players (Sasaki 2000). In a biological management setting, this would entail the introduction of a suite of antagonist genotypes to reach overall control (Campanella et al. 2009; Goolsby et al. 2006).

The present study illustrates a conceptual approach on the G by G interaction between a BCA and its target plant by assessing whether the response variables are better explained by the genetic variability of the BCA or the plant. Investigating the genetic structure both among and within populations of the plant invader and the BCA remain the important first steps in developing a successful weed biocontrol project. Should such investigations reveal distinct genetic variability, especially within the BCA, we advocate to integrate bioassays as outlined in this study during the pre-release phase of a biocontrol program. This will give a first indication of the probability for an at least short- to mid-term efficacy and sustainability when introducing a potential BCA, and on where to find the most efficient agent genotypes.