An early suitability assessment of two exotic Ophraella species ( Coleoptera : Chrysomelidae ) for biological control of invasive ragweed in Europe

Classical biological control is an important means of managing the increasing threat of invasive plants. It constitutes the introduction of natural enemies from the native range of the target plant into the invaded area. This method may be the only cost-effective solution to control the rapidly expanding common ragweed, Ambrosia artemisiifolia, in non-crop habitats in Europe. Therefore, candidate biocontrol agents urgently need to be assessed for their suitability for ragweed control in Europe. A previous literature review prioritized the host-specifi c leaf beetle Ophraella slobodkini as a candidate agent for ragweed control in Europe, whereas it rejected its oligophagous congener O. communa. Meanwhile, O. communa was accidentally introduced and became established south of the European Alps, and we show here that it is expanding its European range. We then present a short version of the traditional pre-release risk-benefi t assessment for these two candidate agents to facilitate fast decision-making about further research efforts. We selected two complementary tests that can be conducted relatively rapidly and inform about essential risks and benefi ts. We conducted a comparative no-choice juvenile performance assay using leaves of ragweed and sunfl ower, the most important non-target plant, in Petri dishes in climatic conditions similar to that in the current European range of O. communa. This informs on the fundamental host range and potential for increasing abundance on these host plants. The results confi rm that O. slobodkini does not survive on, and is hence unlikely to cause severe damage to sunfl ower, while O. communa can survive but develops more slowly on sunfl ower than on ragweed. In parallel, our species distribution models predict no suitable area for the establishment of O. slobodkini in Europe, while O. communa is likely to expand its current range to include a maximum of 18% of the European ragweed distribution. Based on this early assessment, the prioritization and further assessment of O. slobodkini seem unwarranted whereas the results urgently advocate further risk-benefi t analysis of O. communa. Having revealed that most of the European area colonized by ragweed is unlikely to be suitable for these species of Ophraella we suggest the use of such relatively short and cheap preliminary assessment to prioritise other candidate agents or strains for these areas.


INTRODUCTION
Invasions by alien plant species are increasing in frequency and severity (Lambdon et al., 2008;Early et al., 2016).The deployment of classical biological control (BC) may be an important, and often the only, option for managing areas that are too large or too sensitive to be subjected to chemical or mechanical control (van Wilgen et al., 2013;Seastedt, 2015).In classical weed BC specialised antagonists, often from the plant's native range, are released into the invaded range.The introduced antagonists should establish stable populations and provide sustainable control by suppressing the abundance of the target plant below economic or ecological thresholds, while being of limited ecological risk in the introduced area (Müller-Schärer & It originates from warm temperate and subtropical climates in North America (LeSage, 1986).The larvae as well as the adults feed preferentially on A. artemisiifolia (Yamazaki et al., 2000).If densities of the beetle are high they can cause early mortality before plants become reproductive, or, if plants manage to survive until reproduction, reduce their seed production (Zhou et al., 2014).The beetle is capable of unassisted dispersal over hundreds of kilometres per year (Moriya & Shiyake, 2001).Given its potential impact on common ragweed, O. communa had already been mentioned as the most promising candidate BC agent for Europe by Kiss (2007).It was, however, not regarded as a candidate by Gerber et al. (2011) for Europe and had earlier been rejected as a candidate for Australia because of its wider fundamental host range including sunfl ower (Palmer & Goeden, 1991).Feeding on sunfl ower and a few other plant species from the same tribe (Heliantheae) has indeed been reported based on results from no-choice tests in the laboratory (Palmer & Goeden, 1991;Cao et al., 2011), and from fi eld surveys after its accidental introduction into Japan (Yamazaki et al., 2000).Although release has not been petitioned for to date in Europe, O. communa was detected in Europe in 2013: it was found to have established south of the European Alps, in the Swiss canton of Ticino and several regions of Northern Italy where it caused severe defoliation of ragweed (Müller-Schärer et al., 2014).Since then, O. communa has continued to spread (Fig. 1A) (see also Augustinus et al., 2015).Recent aerobiological studies in Northern Italy indicate that it has already contributed to reduced levels of airborne ragweed pollen since 2013 (Bonini et al., 2015a, b).An evaluation of its suitability as BC agent in Europe is urgently needed to help local and national authorities decide on how to react to its presence (Müller-Schärer et al., 2014).In both species of Ophraella, gravid females prefer to feed and oviposit on A. artemisiifolia, and the three larval stages also feed on this species until they pupate on their host or neighbouring plants (Futuyma, 1990).Both species are multivoltine and the adults overwinter on other plant species (Futuyma, 1990;Watanabe & Hirai, 2004).
Delivering a complete pre-release risk-benefi t analysis such as needed for a petition for the release of a weed BC agent can take up to ten years and is expensive (Moran et al., 2005).This is partially due to currently common riskadverse "precautionary" approach towards the introduction of exotic species (Sheppard, 2003).The risk assessment in weed BC has historically mainly focussed on the direct risks for non-target plants (Müller-Schärer & Schaffner, 2008), inferred from protocolled experiments on host specifi city (Briese, 2004;Sheppard et al., 2005;Barratt et al., 2010).Nowadays, rigorous host-specifi city testing on tens of non-target species is often demanded, which alone can take up to ten years.The assessment of benefi ts remains a scientifi c challenge, as the success of control agents in BC programs is highly context-dependent (van Klinken & Raghu, 2006).Spatially-explicit demographic models can help to predict the long-term effect of BC at the population level, but this requires detailed data on the demography, success (Briese, 2004).The prioritized candidate agents are then subjected to an extended pre-release assessment of the potential direct and indirect risks (e.g. for non-target species) and benefi ts (e.g.impact on target), including rigorous experimentation and modelling (Louda et al., 2003;Thomas & Reid, 2007).The results are needed in order to petition for them to be released into the invaded environment, which is usually evaluated by the national competent authorities of the countries concerned.The regulation in Europe is currently still complex because each country has different laws, which are not designed for the introduction of benefi cials (Shaw, 2008;Shaw et al., 2011), but the new EU Regulation 1143/2014 on Invasive Alien Species may stimulate adaptation and harmonisation of the regulations (Shaw et al., 2016).
Common ragweed, Ambrosia artemisiifolia L. (Asteraceae), originates from North America (Bassett & Crompton, 1975) and has become a widespread invader on several other continents, including Europe.This annual monoecious plant is a pioneer species that can grow in a wide range of habitats and climates (Essl et al., 2015).As it is an important aeroallergen and crop weed it causes serious harm to our society.Common ragweed's spread (Essl et al., 2015) and impact (Hamaoui-Laguel et al., 2015) are expected to increase under climate change.While herbicide treatments and mechanical controls have been developed and implemented as short-term measures (Buttenschøn et al., 2009), they do not provide suffi cient control in the long term, and are not suitable or too costly in most habitats invaded by common ragweed.Biological control may be the only cost-effective solution providing long-term control.This method has indeed already been successfully implemented on other continents.In China successful ragweed control is provided by the stem-galling moth Epiblema strenuana (Walker) and the leaf-feeding beetle Ophraella communa LeSage (Zhou et al., 2011a, b), and in Australia by E. strenuana and the leaf-feeding beetle Zygogramma bicolorata Pallister (Palmer & Goeden, 1991).There is, therefore, an urgent need to investigate the option of classical BC for ragweed control in Europe.
An earlier literature review prioritized six exotic specialised insect species as candidates for classical BC of ragweed in Europe (Gerber et al., 2011).The chrysomelid beetle Ophraella slobodkini Futuyma was among them, mainly because of its reported very narrow fundamental host range.Species of the family Chrysomelidae have so far been most successful in BC programs of invasive weeds worldwide compared to other insect taxa (Clewley et al., 2012).The species itself was only described in 1991, and mainly occurs in Florida, in a subtropical climate (Futuyma, 1991).It has never been used as a BC agent against ragweed, and little is known about its potential impact.A risk-benefi t assessment of this species was therefore considered highly desirable (Gerber et al., 2011).Its congener O. communa, in contrast, has already performed well as an inundative BC agent of ragweed in crop fi elds in Canada (Teshler et al., 2002) and more recently in an inoculative approach in China (Huang et al., 2011;Zhou et al., 2011a, b).spread and density-dependent interactions of the species involved and has rarely been done (Raghu et al., 2006;Barratt et al., 2010;Hahn et al., 2012).Researchers as well as authorities are likely to have to decide quickly about investing in ragweed BC agents in Europe.In this paper, we present a short version of the traditional pre-release risk-benefi t assessment to quickly evaluate if the candidate agents O. slobodkini and O. communa are worth being subjected to further in-depth risk-benefi t assessments.

MATERIAL AND METHODS
Designing a rapid preliminary risk-benefi t assessment for Ophraella spp.
We selected two different aspects of the traditional risk-benefi t assessment that are crucial for determining the suitability of a BC agent, which can also be relatively easily and quickly assessed.The fi rst aspect is the capability of larvae to complete their development to the adult stage on a host plant, and the speed with which they complete their development.Larval performance is a major focus of host specifi city testing next to female oviposition preference, but is easier to measure than female behaviour.From the wide range of possible tests, we chose a no-choice laboratory experiment, which is the most stringent non-target host plant risk assessment possible.It specifi es the genetically determined fundamental host range, i.e. the range of hosts on which a herbivore is able to complete its life cycle, or at least parts thereof (van Klinken & Heard, 2000;van Klinken & Edwards, 2002).If they cannot complete their development, herbivores will fail to build up populations on that species.The underlying assumption is that phylogenetically more closely related species are at higher risk (Wapshere, 1974).The speed of development is also informative, since faster herbivore development allows faster population build-up, which is likely to result in high densities of populations of the herbivore and large effects.The laboratory setting allowed the mimicking of European temperature regimes.It is known that the host range of Ophraella is restricted to the family Asteraceae (Futuyma & McCafferty, 1990).We limited the test to a European variety of the sunfl ower, Helianthus annuus L., the non-target species most at risk due to its close phylogenetic relationship to A. artemisiifolia (both are in the tribe Heliantheae) and its economic importance.
The second aspect we selected is the capacity of the agent to establish in the geographic target area, a prerequisite for successful biocontrol.The percentage of the predicted geographic range of the target weed potentially covered by the agent provides an indication of the potential extent of BC success (Kriticos, 2003) and of the geographical area at risk of non-target effects.This percentage cover can be assessed by modelling the predicted suitable range of both the target weed and the BC agent in the area of release.Species distribution models (SDMs) are a common tool used to predict species distributions based on environmental similarities.Generally, SDMs use data on species-level biogeography and ecology, relate these to environmental space, and map likely species distributions, under current or possible future conditions (Booth et al., 2014).The specifi c modelling methodology chosen depends on the system studied and the availability of suitable data.We used a climatic suitability modelling approach to make predictions about the climatic suitability of the target area for ragweed and both candidate species of Ophraella (van Klinken & Raghu, 2006), based on the similarity between the climate in the target area and the climate of present worldwide occurrences of these species (Elith et al., 2006).This method is often used to assess the potential geographic ranges of introduced alien species (Broennimann & Guisan, 2008;Peterson et al., 2008;Villemant et al., 2011) since occurrence data can usually be retrieved from the literature and climatic data are available online.
We argue that combining the two aspects is essential for a robust preliminary evaluation.The juvenile performance assay allows the risk for a particular species of non-target host plant to be assessed under similar conditions to those in the target area, but such experiment can only include a limited number of host plants and set of environmental conditions.Climatic suitability modelling, in contrast, refl ects a more integrated ecological approach, identifying the potential geographic area where the candidate agents, and hence associated risks and benefi ts, may occur.

Material
A permission to import and work with the exotic organisms in the greenhouse and/or quarantine facilities of the University of Fribourg, Switzerland, was issued by the Swiss Federal Offi ce for the Environment (A130598-3).Plants of A. artemisiifolia of mixed origins (European and USA) and sunfl owers (Helianthus annuus PR64H42, a variety grown notably in Ticino, Switzerland, where O. communa is established) were grown from seeds in a greenhouse (16L : 8D photoperiod with minimum temperatures of 25/17°C).We collected ca.150 adults of O. slobodkini in Florida, USA, in June 2014, that were then imported and maintained in culture on ragweed in the quarantine facilities (14L : 10D photoperiod with 25/16 ± 1.5°C).About 150 egg batches of O. communa were directly collected from wild ragweed in Abbi- ategrasso and Magnago, northern Italy, in early April 2015, and brought into the same facilities for use in the experiment.From each species of beetle, single egg batches were checked daily for hatching in April 2015.

Experimental procedure
We measured juvenile performance of the beetles in a no-choice experiment by placing fi rst instar larvae on detached leaves in Petri dishes.This takes up less space and requires fewer plants, allowing for more replicates, than when using entire plants.We acknowledge that detached leaves might affect larval performance through an altered physiology (Smith & Beck, 2013), but our focus here is a comparative one.We assessed each species of beetle on ragweed or sunfl ower under two different European temperature regimes in a full factorial design with 15 replicates.Hence, we had eight experimental groups (two beetle species × two host plants × two temperatures) with 15 larvae each, using 60 larvae of each beetle species in total.We maximised the likelihood of host plant acceptance by avoiding competition and providing fresh food ad libitum.The experiment started in mid-April 2015.Both Ophraella species originate from climates that are hotter than most European climate and it is unknown how European climates affect their performance.Therefore, we used two experimental conditions representing conditions in the European area currently colonized by O. communa.We used four incubators (Panasonic Healthcare Co, MIR-154-PE), two of which were set to "cool" condition, corresponding to spring conditions in the currently colonized range (14L : 10D at 19.6/11.6 ± 0.3°C), and two to "warm" condition, corresponding to summer conditions in that area (16L : 8D at 26.0/17.5 ± 0.3°C).Leaves were detached from ragweed and sunfl ower plants, rinsed with sterilized water and then individually placed onto moist fi lter paper in a Petri dish.Within 24 h after hatching, the fi rst instar larvae (L1) were randomly assigned to one of the four experimental treatments (one of the host plants and one of the climatic conditions), and then individually transferred into the middle of the leaf in their own Petri dish with a fi ne brush.Dishes were closed with parafi lm to prevent the larvae from escaping and keep the conditions inside moist.They were placed into one of the incubators of the allocated climatic condition and regularly moved within and between incubators with the same climatic condition.Leaves were replaced twice a week, or more frequently if they became unsuitable.The fi lter paper was also replaced regularly to avoid fungus proliferating.All larvae used hatched within one week from 9 egg batches of O. communa and 10 of O. slobodkini.Of each batch, on average 6.3 ± 3.5 larvae were used.To avoid bias by egg batch origin, each of the eight experimental groups included larvae from 6-8 different egg batches of each species of beetle.Survival and development of the transferred larvae was monitored daily.Moulting to the next instar was checked by recording shed larval skins and the sex of eclosing adults was visually determined (Guo et al., 2010).The experiment stopped when all larvae had either died or developed into adults.

Statistical analysis
The statistical analysis using R (version 3.1.2,R Core Team, 2014) involved making linear mixed effects models of the total development time from L1 to the adult stage using the lmer function in the package "lme4" (Bates et al., 2015).We defi ned a full model (including beetle species, plant species and climatic conditions, and all their relevant interactions as fi xed effects, and egg batch identity as a random effect on the intercept) with a Gaussian error distribution using an identity link function.We fi tted the model using the Restricted Maximum Likelihood (REML) method and visually inspected the residual plots.We then similarly formulated 12 additional simpler models, by deleting all possible combinations of fi xed factors and their interactions, but always keeping the random effect.We then fi tted all these 13 models using the Maximum Likelihood method in order to obtain likelihood estimates.We ordered the models according to their likelihood, using the modifi ed Akaike Information Criterion (AICc), which penalises the inclusion of extra parameters, to adjust for the relatively small sample size (Burnham & Anderson, 2002), and calculated the evidence ratio of the best model compared to each of the lower-ranked models (Burnham et al., 2011).We rejected all models with ΔAICc > 6 compared to the best model (this is a conservative threshold, Richards, 2005), and additionally, to adhere to the principle of parsimony, those with ΔAICc values < 6 but including more parameters than the best model.We fi nally used the (single) remaining model with an REML fi t to obtain estimates and confi dence intervals of development time for each of the experimental groups.We assessed the goodness of fi t calculating the conditional R 2 , which takes random variance into account, following Nakagawa & Schielzeth (2013).

Species occurrences and bioclimatic data
We collected all occurrences recorded for A. artemisiifolia, O. slobodkini and O. communa worldwide (see Supplementary Information 2 for sources and collection methods).We used WORLDCLIM to obtain climate data on current conditions at a spatial resolution of 5 min (www.worldclim.org/bioclim,Hijmans et al., 2005).To derive a set of predictors that are critical for survival and physiological functions for each of the species, we evaluated the importance of these 19 climatic variables and the derived variable growing degree days [GDD; using GDD > 8°C for A. artemisiifolia based on this threshold for the closely related sunfl ower (Sadras & Hall, 1988), and GDD > 10°C for the beetles, because this is the most common lower threshold for development for insects (Pruess, 1983)].Response curves reveal how the model's predictions change as the bioclimatic variables varied and those that did not affect the predictions were deleted.The climatic variables eventually chosen for each species are listed in the Supplementary Information 2, Table S3.

Statistical modelling
Duplicated occurrences within a raster pixel were removed.As only occurrences were available, we generated 10,000 random pseudo-absences to fi ll the absence component in the models (Barbet-Massin et al., 2012).Current climates were projected for each species using the Biomod2 framework (Thuiller et al., 2009(Thuiller et al., , 2013) ) and a combination of different modelling techniques to fi nd the optimal solution for dealing with the inherent uncertainty of each of these models (Araujo & New, 2007) and the discrepancies resulting from using different techniques.The Biomod2 platform accounts for inter-model variability by fi tting ensembles of forecasts and analysing the resulting range of uncertainties using bounding box, consensus and probabilistic methodologies (Araujo & New, 2007;Thuiller et al., 2009).For each of the species, generalized linear models (GLM), generalized boosting models and maximum entropy models (MAXENT) were calibrated on a random sample of the initial data (80%), and then tested using the remaining data (20%) using both the receiver-operating characteristic curve and true kill statistic (Pearce & Ferrier, 2000).To balance under-fi tting with overfi tting predictions, we simplifi ed the model based on the Bayesian Information Criterion for GLM and used the same setting of the regularization parameter β in MAXENT for all feature types (β was caculated based on the number of presence data we used in the model) (Merow et al., 2014).We then estimated the response curves.The above tech-niques were chosen because they reveal the most effective SDMs (Elith et al., 2006).The whole training-evaluation procedure was repeated 25 times for each model, using a different set of initial data in each replicate, to ensure robustness of the predictions and provide uncertainty estimates (Broennimann & Guisan, 2008).We also measured the non-analogous environments between the study area used to calibrate the SDMs and the projected area for the SDMs, and removed all non-analogous climates in the fi nal prediction to avoid extrapolation of models beyond training conditions (Mesgaran et al., 2014).The predictive power of the models was tested using the area under the receiver-operator characteristic function (area under the curve criteria, AUC) for evaluation (Pearce & Ferrier, 2000).Given the high AUC values (> 0.8) (van Proosdij et al., 2016) of each modelling approach for all species, the predictions of the models are robust.In order to visually overlay the insects distributions on that of A. artemisiifolia, the probabilities of presence of the insects were then binary transformed into presences and absences for each raster pixel (see details in Sun et al., 2017).

Juvenile performance
Not a single O. slobodkini larva survived on sunfl ower and all but one died in the fi rst larval stage.In all the other experimental groups, nearly half of the 15 larvae developed into adults, regardless of beetle species, host plant or temperature (Fig. 2).Whereas mortality on sunfl ower occurred mainly in the fi rst larval stage, mortality on ragweed occurred in all stages (Fig. 2).In total 12 males, 11 females, and one adult of unknown sex were recorded for O. communa, and 5, 7 and 1 for O. slobodkini, respectively, which do not deviate signifi cantly from the expected 1 : 1 sex ratio.Both sexes occurred in all treatments in which adults developed.
The best statistical model for the total development time from L1 to the adult stage included in addition to the random effect of egg batch identity, the main effects of species of beetle, plant and conditions, and the interaction between species of beetle and conditions, and explained 96.5% of the total variance (see details of all models in Supplementary Information 1, Table S1).The fastest development of about 3 weeks was recorded for O. communa on ragweed in "warm" conditions, whereas O. slobodkini took nearly 4 weeks (24% longer than O. communa) (Fig. 3).Develop-ment time approximately doubled in the cool conditions in which that of O. slobodkini was 32% longer than that of O. communa.Ophraella communa developed signifi cantly slower (ca.15% longer) on sunfl ower than on ragweed in both warm and cool conditions.The development times of each of the different stages were strongly correlated with the total development time (Supplementary Information 1, Table S2), and the proportion of the time spent in each stage was similar in all the groups (Supplementary Information 1, Fig. S1).

Species distribution model
The occurrence data yielded 5018, 541, and 36 worldwide raster pixels for the presence of A. artemisiifolia, O. communa and O. slobodkini, respectively.The evaluation of climatic variables resulted in the selection of GDD, along with 2-8 different bioclimatic variables for each of the species (Supplementary Information 2, Table S3).Fig. 1B presents the projected climatic suitability for O. communa, overlain on the predicted range of A. artemisiifolia.Ophraella communa is predicted to be able to colonize a much larger area than currently (Fig. 1A), extending to the Mediterranean and even north of the Alps, and, in addition, areas around the Black Sea (Fig. 1B).Together, this would include 18% of the predicted range of A. artemisiifolia under current climatic conditions.In contrast, the climate in Europe seems to be unsuitable for Ophraella slobodkini, since there is no geographic area predicted for this species, within or outside the predicted range of A. artemisiifolia.

DISCUSSION
The results of our juvenile performance assay corroborate earlier studies that O. slobodkini poses less of a risk to sunfl ower than O. communa (Gerber et al., 2011).Since none of the larvae of O. slobodkini survived on sunfl ower, we predict that O. slobodkini is unable to build up a population on this variety.If eggs are laid on sunfl ower, damage  Values recorded in each of the two conditions are estimated development times and their 5-95% confi dence intervals based on the best fi tting linear mixed model, whereas the connecting lines only serve to visualise the interactions (there is an interaction between beetle species and condition, but not between plant species and condition).from larvae will be negligible, as only one out of 30 larvae reached the second instar.Paynter et al. (2015) recently report that the ratio of survival on non-target versus target plants in no-choice tests is a good estimate of the probability of non-target plants being attacked in the fi eld.Our results indicate that the probability would be below 5% (e.g. a spill-over effect of adult beetles reaching high densities on the target plant).Because O. communa survived on the sunfl ower variety used, this sunfl ower was included in the fundamental host range of O. communa.The probability of survival of larvae of O. communa was as high on this sunfl ower as on ragweed, which contrasts with previous studies in which a very low percentage survival is recorded on sunfl ower compared to ragweed (Dernovici et al., 2006;Cao et al., 2011).The probability of survival is likely to vary with sunfl ower variety, age and environmental conditions, even though we recorded no effect of temperature on survival.Since the realized host range in the fi eld is usually a subset of the fundamental range (van Klinken & Heard, 2000), this result requires further investigation under European fi eld conditions (e.g.Briese et al., 2002;Briese, 2005;Sheppard et al., 2005).
The slower development of O. communa on sunfl ower compared to ragweed, in combination with our personal observations that more tissue of sunfl ower was consumed, strongly indicate that sunfl ower is a less suitable food source for the larvae than ragweed.This is also likely to result in a slower build-up of beetle populations on sunfl ower than on ragweed.This is in line with other developmental studies, as Dernovici et al. (2006) predict that O. communa will not increase in abundance on sunfl ower based on lifetable parameters recorded in a laboratory.Several studies indicate that O. communa females prefer A. artemisiifolia for oviposition over other plants, and oviposition on sunfl ower growing close to ragweed is indeed only very rarely recorded in the fi eld (Dernovici et al., 2006;Cao et al., 2011;Zhou et al., 2011c).Importantly, Ophraella communa is not reported as a pest of sunfl ower in its native North American range (Palmer & Goeden, 1991).Taken together, despite the ability of O. communa larvae to survive on the European sunfl ower variety tested under European conditions, the risk for sunfl ower is slight, but this urgently needs to be confi rmed in the fi eld in Europe.
The performance of O. slobodkini on ragweed under European conditions was poorer than that of O. communa in the juvenile performance assay.This indicates that O. slobodkini is less well adapted to developing under European conditions, which is not surprising given that the climate in its native range is hotter than that of O. communa (Futuyma, 1991).The climatic suitability modelling resulted in even more distinct patterns for these two species of beetles.According to the model predictions based on climate, no region in Europe is currently suitable for O. slobodkini.We acknowledge the uncertainty of the model due to the few records of the occurrence of this species (van Proosdij et al., 2016) in a small geographic area (in and around Florida).It is also possible that historical records of Ophraella occurrence from elsewhere were misidentifi ed and include the recently recognized O. slobodkini and that this species is more widespread in North America.The current range of Ophraella communa in Europe, in contrast, is likely to increase in all directions by hundreds of kilometres.The model indicates also that its potential range is restricted to the 50° latitude and further expansion west and east is limited to only maximum 18% of the potential ragweed distribution in Europe (Fig. 1B).

Directions for future research on agent selection and testing
Despite our confi rmation that O. slobodkini will not develop on sunfl ower, we reject the prioritisation of this species as a BC agent for use in Europe since our distribution model indicates that no part of Europe is currently suitable for this species.As a consequence, we recommend focusing the research effort on other species.In contrast, we advocate further risk-benefi t analysis of the formerly rejected O. communa.The juvenile performance experiment revealed that this species can develop under European conditions, confi rms there is a potential risk of it attacking sunfl ower, and indicates that this species is potentially capable of building up populations on ragweed.The distribution model highlights the need for an assessment for an area larger than that currently colonized by O. communa and that the ongoing expansion of this species highlights the urgency of such an assessment.Follow-up studies on O. communa should weigh potential benefi ts against potential risks, and compare these with the impact of A. artemisiifolia in the absence of BC agents (van Wilgen et al., 2013).Therefore, more extensive host-range tests including oviposition and feeding tests under choice and no-choice conditions with various developmental stages and varieties of sunfl ower, and other plant species, both in the laboratory and in the fi eld, are needed.Impact studies should assess its potential effect on A. artemisiifolia population growth and pollen production in relation to beetle density, plant phenology and climate.
Since it is very likely that O. communa will spread to the other climatically suitable areas in Europe (Fig. 1B) (Pratt & Center, 2012), our results may also be used to alert authorities in the specifi ed areas to the potential future establishment of O. communa in their area.Based on our fi ndings, the French Agency for Food, Environmental and Occupational Health & Safety has already created an Expert Group to conduct a Pest Risk Analysis (Anses, 2015) and recently also a Benefi t Analysis for O. communa in France, where the species does not currently occur.
Our results indicate no establishment or lower performance of O. communa in the cooler areas in Europe.Natural spread to such areas, however, might be facilitated by future changes in climate (Sun et al., 2017), or the species may quickly adapt to current conditions.Evidence for rapid adaptation in the fi eld comes from Japan, where O. communa adapted to the shorter day length in more northern areas by a shift in the critical day length inducing diapause (Tanaka et al., 2014).Limited performance may be partially overcome by increasing densities by augmentative releases early in the season (as is practiced in China, Zheng et al., 2011), or by developing cold-resistant strains by selective breeding of suitable genotypes or through acclimation (Zhou et al., 2013).Alternatively, other agents may be needed for these areas.

CONCLUSION
We have shown that this rapid preliminary risk-benefi t assessment proved useful in shifting research focus from the formerly prioritized candidate O. slobodkini to the formerly rejected and less specifi c congener O. communa for ragweed control in Europe.It also helped to identify the need for other BC agents for use in the cooler parts of Europe.We strongly advocate the use of a rapid procedure (but adapted to the species under study) for further use in the prioritization of other BC candidates of ragweed in Europe.

Fig. 1 .
Fig. 1.Map showing the presence in Europe of O. communa recorded between 2013 and 2016 as circles in different colours for different years (A), which shows that the range of this species is increasing.Projected climatic suitability for O. communa in Europe under current climatic conditions is indicated in brown and covers 18% of the projected area suitable for A. artemisiifolia indicated by different colours, from unsuitable in light red via medium suitability in yellow, to suitable in green (B, modifi ed from Sun et al., 2017).

Fig. 3 .
Fig. 3. Interaction plot depicting the development time from L1 to adults for all of the 38 adults that developed of the two species of ragweed leaf beetles (O.communa solid lines, O. slobodkini dashed line) on ragweed (dark) and sunfl ower (light) under two different conditions (cool, 14L : 10D at 19.6/11.6°C;warm, 16L : 8D at 26.0/17.5°C).Values recorded in each of the two conditions are estimated development times and their 5-95% confi dence intervals based on the best fi tting linear mixed model, whereas the connecting lines only serve to visualise the interactions (there is an interaction between beetle species and condition, but not between plant species and condition).

Table S2 .
Pearson correlation coeffi cient between the development time of each of the larval stages L1-L3, the pupal stage and total time spent developing (from L1 to adult), for the 38 adults that emerged.