Host-Plant Selection Behavior of Ophraella communa, a Biocontrol Agent of the Invasive Common Ragweed Ambrosia artemisiifolia

Simple Summary Ophraella communa is an effective biocontrol agent against the invasive common ragweed Ambrosia artemisiifolia. However, whether some closely related non-target plants can become alternative host plant species of O. communa in China remains unclear. Although extensive host-plant selection tests have been used to ensure the host specificity of O. communa in other countries, some doubts remain. In this study, we conducted a series of choice experiments in outdoor cages and open fields to determine the preference of O. communa for A. artemisiifolia and three non-target plant species: sunflower (Helianthus annuus), cocklebur (Xanthium sibiricum), and giant ragweed (Ambrosia trifida). The results showed that this beetle poses no threat to the biosafety of H. anunuus or A. trifida and exhibits a robust dispersal capacity to find and feed on A. artemisiifolia. However, in the future, we should be aware that X. sibiricum has the potential to be an alternative host plant for O. communa. Abstract Understanding the host-selection behavior of herbivorous insects is important to clarify their efficacy and safety as biocontrol agents. To explore the host-plant selection of the beetle Ophraella communa, a natural enemy of the alien invasive common ragweed (Ambrosia artemisiifolia), we conducted a series of outdoor choice experiments in cages in 2010 and in open fields in 2010 and 2011 to determine the preference of O. communa for A. artemisiifolia and three non-target plant species: sunflower (Helianthus annuus), cocklebur (Xanthium sibiricum), and giant ragweed (Ambrosia trifida). In the outdoor cage experiment, no eggs were found on sunflowers, and O. communa adults rapidly moved from sunflowers to the other three plant species. Instead, adults preferred to lay eggs on A. artemisiifolia, followed by X. sibiricum and A. trifida, although very few eggs were observed on A. trifida. Observing the host-plant selection of O. communa in an open sunflower field, we found that O. communa adults always chose A. artemisiifolia for feeding and egg laying. Although several adults (<0.02 adults/plant) stayed on H. annuus, no feeding or oviposition were observed, and adults quickly transferred to A. artemisiifolia. In 2010 and 2011, 3 egg masses (96 eggs) were observed on sunflowers, but they failed to hatch or develop into adults. In addition, some O. communa adults crossed the barrier formed by H. annuus to feed and oviposit on A. artemisiifolia planted in the periphery, and persisted in patches of different densities. Additionally, only 10% of O. communa adults chose to feed and oviposit on the X. sibiricum barrier. These findings suggest that O. communa poses no threat to the biosafety of H. anunuus and A. trifida and exhibits a robust dispersal capacity to find and feed on A. artemisiifolia. However, X. sibiricum has the potential to be an alternative host plant for O. communa.


Introduction
Exploring and understanding host specificity is an important part of the methodology for selecting biological control agents [1][2][3], and the concept of insect behavior has been widely applied to improve testing for host specificity [4][5][6][7][8]. During the long-term evolutionary process, herbivorous insects have developed a series of specialized behavioral strategies to distinguish between host and non-host plants [9], and multiple mechanisms have been proposed to underlie host-plant selection [10][11][12]. For example, the "preferenceperformance hypothesis" predicts that female insects evolve to oviposit on hosts on which their offspring will fare best [13,14]. To maximize overall fitness, herbivorous insects must assess host-plant quality, both between and within species, and locate and select the most suitable host for feeding and larval development [15][16][17].
Although many agents used for the biological control of weeds exhibit extreme host specificity, the endogenous conditions of the insect and test arena may cause host-plant selection behavior to become more labile, thus affecting the host-plant range [18,19]. There are numerous types of host specificity tests, including choice tests [20], non-choice tests [20,21], cage tests [22], and open field choice tests [21]. The selection of the test, and even the distribution pattern of the test plants, affects the host specificity results. For example, Microthrix inconspicuella is a potential agent for the control of the polygonaceous weed Emex australis, and, under caged quarantine conditions, the larvae of this moth have been found to feed on apples, a rosaceous crop [23]. However, under field conditions or when the larvae are contained in large sleeve cages, apple foliage is not attacked [22]. Therefore, the outcomes of host specificity tests typically vary under different test designs, owing to behavioral factors.
Several behavioral factors influence test results, including sequential behavioral responses during host-plant selection [24][25][26], experience and learning [27][28][29][30][31], and timedependent effects [32]. These factors may lead to two types of false results. A false positive result refers to an attack during the test but no potential for attack under field conditions, whereas a false negative result occurs when a plant species is not attacked during the test but might be attacked in the field. For example, if non-target plants are near the host plant, they may be more prone to attack, and insects may habituate to and accept non-host plants through repeated contact, thus leading to a false positive result. False results may lead to the rejection of potential biological control agents that might be adequately host-specific, or to the release of candidate agents that may attack non-host plants in the field.
To minimize the potential for false results, many test methods have been designed, including the use of large arenas [33], natural arenas [4], open field testing [34], and behavior-based host-selection tests, which should indicate whether a plant is susceptible to feeding or oviposition by a biological control agent under any set of field conditions [1,5]. In this study, we experimentally analyzed under field conditions the host-selection behavior of a potential biological control agent, Ophraella communa (Coleoptera: Chrysomelidae), against the common ragweed Ambrosia artemisiifolia (Asteraceae).
Ophraella communa has been found to be an effective agent for the biological control of common ragweed, a widespread and harmful invasive alien weed [35,36], and has achieved great success in China [37]. It is an oligophagous leaf beetle that feeds on plants of the Asteraceae family. Several studies have focused on its host range, and it has been reported to attack cockleburs (Iva axillaris, Xanthium strumarium, X. canadense, and X. italicum), giant ragweed (Ambrosia trifida), sunflower (Helianthus annuus), feverfew (Parthenium hysterophorus), and Jerusalem artichoke (H. tuberosus) (Asteraceae) [2,35,36,38]. Watanabe [36,[38][39][40]. Therefore, this beetle was rejected for release as a biocontrol agent for ragweed because of the possible damage to crops in Australia [41]. Although extensive host-plant selection tests have been used to ensure the host specificity of O. communa [2,[42][43][44], some doubts remain, such as whether cockleburs can become an alternative host-plant species and the host-plant range expansion of O. communa in China is unclear. The risk of attack by O. communa and the subsequent level of damage that might occur in sunflower crops under field conditions remain unknown [19,36,45].
Host and non-host plants often coexist under natural conditions, and the "physical obstruction hypothesis" describes the situation in which host plants are effectively hidden by large or tall non-host plants [13], which are usually used to protect crops from pest infestations in the field. Similarly, a biological control agent may have difficulty locating a targeted invasive host when the plant coexists with larger or taller non-host plants. In China and Europe, A. artemisiifolia has become a major agricultural weed, especially spring-sown crops, such as sunflower and maize [46][47][48]. Sunflower is a large and tall plant that can easily act as a barrier to biological control. Therefore, when O. communa is used to control A. artemisiifolia in sunflower cultivation, it is unclear whether the weed can hide among the crops, leading to a reduction in biological control efficiency.
Understanding the characteristics of the host-plant selection behavior of O. communa is important for better prediction and evaluation of its safety and efficacy as a biological control agent. It is also important to determine whether cockleburs can become alternative host-plant species in China, what is the risk of attack by O. communa on sunflower, and the control efficacy of A. artemisiifolia under field conditions. Therefore, in this study, outdoor cage and open field tests were performed to investigate the host-selection behavior of this beetle in the hope of answering these questions. The A. artemisiifolia, X. strumarium, and A. trifida seeds were sown in individual seed trays with sterilized nutritional soil (Langfang Dingxin Seedling Company, Langfang, China) and individually transplanted into plastic pots (15 cm in diameter and 10 cm in height) with loamy clay soil at the three-to four-leaf stage. H. annuus seeds were sown directly in the same plastic pots. The seedlings were placed in an unheated and naturally lit greenhouse at the Langfang Experimental Station of the Institute of Plant Protection, Chinese Academy of Agricultural Sciences (LF Station, IPP, CAAS) in Langfang City, Hebei Province (39 • 30 42" N, 116 • 36 07" E) and watered every four days. All individuals of each species were used in the experiments when they were approximately 20-30 cm in height.

Insect Culture
Ophraella communa pupae were collected from IPP and HAAS and used to construct colonies on A. artemisiifolia plants at the LF Station, IPP, and CAAS. The O. communa population was maintained in an unheated greenhouse under a 16 h L:8 h D photoperiod at 26 ± 1 • C and 70 ± 10% relative humidity (RH) and was used for the experiments after three generations.

Distribution and Oviposition Preference Behavior of O. communa Adults on Four Different Coexisting Plant Species in Outdoor Cages
The experiments were conducted in outdoor cages (6.5 × 24.5 m) at LF Station, IPP, CAAS. Five sample plots (2.5 × 4.5 m) were regularly arranged in the field, and each plot was covered with a single mesh cage (2 m in height) on 2 July 2010 ( Figure 1A). Two plants of each of the four tested species (approximately 30-40 cm in height) were transplanted into the above cages on 8 July 2010. The planting patterns are shown in Figure 1B. O. communa adults at 2 days of age were randomly collected from the greenhouse, and on 15 July 2010,

Coexisting Plant Species in Outdoor Cages
The experiments were conducted in outdoor cages (6.5 × 24.5 m) at LF Station, IPP, CAAS. Five sample plots (2.5 × 4.5 m) were regularly arranged in the field, and each plot was covered with a single mesh cage (2 m in height) on 2 July 2010 ( Figure 1A). Two plants of each of the four tested species (approximately 30-40 cm in height) were transplanted into the above cages on 8 July 2010. The planting patterns are shown in Figure 1B. O. communa adults at 2 days of age were randomly collected from the greenhouse, and on 15 July 2010, 10 pairs (female: male = 1:1) were released on each sunflower plant in each cage. From the day after release to the 5th day, the numbers of adults and eggs on each plant in each cage were counted daily; the observations were performed every other day until the 27th day after release (one O. communa generation). Five replicates were performed for each experiment.

Host-Plant Selection Behavior of O. communa on Regularly Distributed Ragweed Patches in Sunflower Plots
The experiments were conducted in an open field (45 × 70 m) at the LF Station, IPP, and CAAS in 2010 and 2011. Six sample plots (20 × 20 m) were prepared, and X. strumarium was planted among different barrier bands ( Figure 2A). The intercropping patterns of A. artemisiifolia and H. annuus are shown in Figure 2B. In each plot, 20 A. artemisiifolia plants were evenly planted in the center of each plot (shaded area in the figure with a 1 m radius), and 24 sunflowers were planted in a homocentric ring with a 3 m radius to create the sunflower barrier. Twenty-seven and forty-five sunflowers were planted in homocentric rings with radii of 6 and 9 m, respectively. One, two, and three A. artemisiifolia plants per

Host-Plant Selection Behavior of O. communa on Regularly Distributed Ragweed Patches in Sunflower Plots
The experiments were conducted in an open field (45 × 70 m) at the LF Station, IPP, and CAAS in 2010 and 2011. Six sample plots (20 × 20 m) were prepared, and X. strumarium was planted among different barrier bands ( Figure 2A). The intercropping patterns of A. artemisiifolia and H. annuus are shown in Figure 2B. In each plot, 20 A. artemisiifolia plants were evenly planted in the center of each plot (shaded area in the figure with a 1 m radius), and 24 sunflowers were planted in a homocentric ring with a 3 m radius to create the sunflower barrier. Twenty-seven and forty-five sunflowers were planted in homocentric rings with radii of 6 and 9 m, respectively. One, two, and three A. artemisiifolia plants per cluster were evenly intercropped at intervals of three sunflowers in the 6 m radius homocentric ring and at intervals of four sunflowers in the 9 m homocentric ring. Sunflower and ragweed seedlings were planted 80 cm apart. O. communa adults at 2 days of age were randomly collected from the laboratory culture; on July 4 of both years, 40 pairs were released on A. artemisiifolia in the center of each plot. After three days, visual sampling was used to count the numbers of O. communa adults, eggs, larvae, and pupae on A. artemisiifolia and sunflower plants in each plot, and observations were performed every six days until the sunflower fruit ripened on September 26. Six replicates were performed for each experiment and continued for two years. and ragweed seedlings were planted 80 cm apart. O. communa adults at 2 days of age were randomly collected from the laboratory culture; on July 4 of both years, 40 pairs were released on A. artemisiifolia in the center of each plot. After three days, visual sampling was used to count the numbers of O. communa adults, eggs, larvae, and pupae on A. artemisiifolia and sunflower plants in each plot, and observations were performed every six days until the sunflower fruit ripened on September 26. Six replicates were performed for each experiment and continued for two years.

Data Analysis
Statistical analyses were performed using the SAS system for Windows V8. The experimental data were checked for normality and homoscedasticity, and if required, were arcsine square-root or log-transformed before analysis. In the outdoor cages experiment, three-way ANOVA followed by the Tukey test (p values ≤ 0.05) was performed to compare the data on O. communa distribution (adults and eggs) considering the effects of plant species, days after release and cages (blocks) and their interactions. In the open field experiment, preliminary analyses indicated no significant effects of year and blocks (plots). Therefore, a three-way ANOVA followed by the Tukey test (p values ≤ 0.05) was used to test for the effects of plant species, distance from center, and ragweed cluster density on the cumulative densities of the different O. communa developmental stages.

Distribution of O. communa Adults on Four Different Coexisting Plant Species in Outdoor Cages
In the outdoor cages experiment, the results of three-way ANOVA indicated that only plant species had a significant effect on dynamics of O. communa adults, whereas days after release, cages and their interactions had no significant effects (Supplementary  Table S1). O. communa adults released on H. annuus moved rapidly to the other three plant species, and there was a significantly lower number of O. communa adults on sunflower compared to the other plant species (Figure 3). After release, approximately 70% of the beetles were observed on the four different plant species, approximately 32.5% and 25% moved to A. artemisiifolia and X. sibiricum, respectively, approximately 2% moved to A. cluster were evenly intercropped at intervals of three sunflowers in the 6 m radius homocentric ring and at intervals of four sunflowers in the 9 m homocentric ring. Sunflower and ragweed seedlings were planted 80 cm apart. O. communa adults at 2 days of age were randomly collected from the laboratory culture; on July 4 of both years, 40 pairs were released on A. artemisiifolia in the center of each plot. After three days, visual sampling was used to count the numbers of O. communa adults, eggs, larvae, and pupae on A. artemisiifolia and sunflower plants in each plot, and observations were performed every six days until the sunflower fruit ripened on September 26. Six replicates were performed for each experiment and continued for two years.

Data Analysis
Statistical analyses were performed using the SAS system for Windows V8. The experimental data were checked for normality and homoscedasticity, and if required, were arcsine square-root or log-transformed before analysis. In the outdoor cages experiment, three-way ANOVA followed by the Tukey test (p values ≤ 0.05) was performed to compare the data on O. communa distribution (adults and eggs) considering the effects of plant species, days after release and cages (blocks) and their interactions. In the open field experiment, preliminary analyses indicated no significant effects of year and blocks (plots). Therefore, a three-way ANOVA followed by the Tukey test (p values ≤ 0.05) was used to test for the effects of plant species, distance from center, and ragweed cluster density on the cumulative densities of the different O. communa developmental stages.

Distribution of O. communa Adults on Four Different Coexisting Plant Species in Outdoor Cages
In the outdoor cages experiment, the results of three-way ANOVA indicated that only plant species had a significant effect on dynamics of O. communa adults, whereas days after release, cages and their interactions had no significant effects (Supplementary  Table S1). O. communa adults released on H. annuus moved rapidly to the other three plant species, and there was a significantly lower number of O. communa adults on sunflower compared to the other plant species (Figure 3). After release, approximately 70% of the beetles were observed on the four different plant species, approximately 32.5% and 25% moved to A. artemisiifolia and X. sibiricum, respectively, approximately 2% moved to A.
" shows the sites and densities per cluster of common ragweed planted in the 6 m and 9 m radius homocentric rings in each plot.

Data Analysis
Statistical analyses were performed using the SAS system for Windows V8. The experimental data were checked for normality and homoscedasticity, and if required, were arcsine square-root or log-transformed before analysis. In the outdoor cages experiment, three-way ANOVA followed by the Tukey test (p values ≤ 0.05) was performed to compare the data on O. communa distribution (adults and eggs) considering the effects of plant species, days after release and cages (blocks) and their interactions. In the open field experiment, preliminary analyses indicated no significant effects of year and blocks (plots). Therefore, a three-way ANOVA followed by the Tukey test (p values ≤ 0.05) was used to test for the effects of plant species, distance from center, and ragweed cluster density on the cumulative densities of the different O. communa developmental stages.

Distribution of O. communa Adults on Four Different Coexisting Plant Species in Outdoor Cages
In the outdoor cages experiment, the results of three-way ANOVA indicated that only plant species had a significant effect on dynamics of O. communa adults, whereas days after release, cages and their interactions had no significant effects (Supplementary Table S1). O. communa adults released on H. annuus moved rapidly to the other three plant species, and there was a significantly lower number of O. communa adults on sunflower compared to the other plant species (Figure 3). After release, approximately 70% of the beetles were observed on the four different plant species, approximately 32.5% and 25% moved to A. artemisiifolia and X. sibiricum, respectively, approximately 2% moved to A. trifida (two adults on one plant), and approximately 10% remained on the sunflower plants. On the third day after release, most adult beetles were found to feed on A. artemisiifolia and X. sibiricum. Only one adult remained on a single sunflower plant, and several tiny feeding spots were observed; however, this area was negligible compared to the entire leaf area. After five days, no adult beetles were found on the sunflowers, but the population of adult O. communa remained high in A. artemisiifolia and X. sibiricum. One or two adults occasionally fed on A. trifida.
plants. On the third day after release, most adult beetles were found to feed on A. artemisiifolia and X. sibiricum. Only one adult remained on a single sunflower plant, and several tiny feeding spots were observed; however, this area was negligible compared to the entire leaf area. After five days, no adult beetles were found on the sunflowers, but the population of adult O. communa remained high in A. artemisiifolia and X. sibiricum. One or two adults occasionally fed on A. trifida.

Oviposition Preference Behavior of O. communa Adults on Four Different Coexisting Plant Species in Outdoor Cages
In the outdoor cages experiment, the results of the three-way ANOVA showed significant effects of the plant, the day, and their interaction on the dynamics of O. communa egg deposition (Supplementary Table S1). By tracking the movement of O. communa adults among the four tested plant species, we found that they preferred to lay eggs on A. artemisiifolia followed by X. sibiricum. Very few eggs (<60) were observed on one A. trifida plant and no eggs were found on H. annuus plants during the entire survey period. The oviposition of O. communa on A. artemisiifolia showed a significant peak of 623.0 eggs per plant on 24 July 2010, which was significantly higher than that on X. sibiricum (146.2 eggs per cage) and A. trifida (9 eggs per cage) ( Figure 4).

Oviposition Preference Behavior of O. communa Adults on Four Different Coexisting Plant Species in Outdoor Cages
In the outdoor cages experiment, the results of the three-way ANOVA showed significant effects of the plant, the day, and their interaction on the dynamics of O. communa egg deposition (Supplementary Table S1). By tracking the movement of O. communa adults among the four tested plant species, we found that they preferred to lay eggs on A. artemisiifolia followed by X. sibiricum. Very few eggs (<60) were observed on one A. trifida plant and no eggs were found on H. annuus plants during the entire survey period. The oviposition of O. communa on A. artemisiifolia showed a significant peak of 623.0 eggs per plant on 24 July 2010, which was significantly higher than that on X. sibiricum (146.2 eggs per cage) and A. trifida (9 eggs per cage) (Figure 4).

Host-Plant Selection Behavior of O. communa on Regularly Distributed Ragweed Patches in Sunflower Plots
In the open field experiment, the results of the three-way ANOVA indicated that only plant species had a significant effect on the number of O. communa individuals in different developmental stages, whereas distance, density, and the interactions between the three

Host-Plant Selection Behavior of O. communa on Regularly Distributed Ragweed Patches in Sunflower Plots
In the open field experiment, the results of the three-way ANOVA indicated that only plant species had a significant effect on the number of O. communa individuals in different developmental stages, whereas distance, density, and the interactions between the three factors had no significant effects (Supplementary Table S2). In 2010 and 2011, there were significant differences in the number of O. communa individuals at different developmental stages in A. artemisiifolia compared with H. annuus ( Figure 5). In both years, the number of O. communa adults on common ragweed was significantly higher than that on sunflower (Figure 5a,b). The number of eggs laid also showed consistency (Figure 5c,d). Very few eggs were found on sunflowers, and all died during development. Moreover, the number of larvae (Figure 5e,f) and pupae (Figure 5g,h) on sunflower was significantly lower and was close to zero.  In addition, based on two years of observation and records, the O. communa adults were mainly found feeding and/or ovipositing on A. artemisiifolia planted in the center of each plot during the early period after release, and very few adults were found on A. artemisiifolia planted in homocentric rings with radii of 6 m and 9 m (Supplementary Figure S1a,b). During the entire survey period, from July to September, the O. communa population completed two generations on A. artemisiifolia planted in the center and one generation in both homocentric rings. By September (60 days after release), almost all A. artemisiifolia planted in the center had died, and adult O. communa had moved to A. artemisiifolia planted in both homocentric rings to feed and oviposit (Supplementary Figure S1).

Discussion
The host-plant selection behavior of herbivorous insects is complex. When larval and/or adult insects encounter target or non-target plants, the morphology and chemical properties of the plant surfaces are first evaluated by the contact receptors (antennae, mouthparts, ovipositors) of the insects, and the inner chemical characteristics of the plants are assessed to determine whether they are acceptable or antagonistic [49]. In non-choice tests, herbivorous insects are typically confined to only one test plant species; therefore, they tend to have a broader host range than in choice tests [5,20,50]. Host range overestimation may lead to the rejection of candidate biological control agents that are adequately hostspecific under field conditions [4]. The risk of O. communa feeding on sunflower is negligible because the leaf beetle is occasionally found on H. annuus when all A. artemisiifolia plants are defoliated near the sunflower field. If the beetle feeds only on sunflower, the number of offspring will be reduced and the beetle cannot survive [42]. To date, there has been a debate on whether O. communa can feed on and damage H. annuus even though host specificity tests have been conducted for nearly 30 years.
In our field cage test, several tiny feeding spots from adult O. communa were found on sunflower leaves. However, those adults left the sunflower plant in the next survey (four days after release) and did not feed or oviposit on the sunflower thereafter. In the open field investigation, adult O. communa released on A. artemisiifolia in the center of the plot primarily fed and oviposited there. As O. communa spread to the periphery, several adults were occasionally found on sunflowers, but no feeding or oviposition behavior was observed. Our results demonstrate that adult O. communa are averse to sunflower compared with ragweed. "Preference-performance hypothesis", also known as the "mother knows best" hypothesis, predicts that females prefer a host that assures the greatest fitness of their offspring [14,51,52]. In our study, no O. communa eggs were found on sunflowers in the cage test, but three egg masses were found on sunflowers in open sunflower fields in 2010 and 2011. However, only one egg mass hatched, and all larvae died during development. These results support the conclusion that sunflower is an unsuitable host plant for O. communa offspring and are consistent with findings from previous studies carried out in Canada [42] and China [43,44,53]. In addition, it is worth noting that we should be alert to the possibility of individuals dispersing from outside the field into the experimental plots in the open field experiment, because ragweed leaf beetles are known to disperse over long distances. Yamanaka et al. [54] found that, with the passage of time, O. communa spills over to adjacent locations at roughly the one-beetle-generation time scale according to the "resource concentration hypothesis" and "reaction-diffusion theory". In addition, herbivorous insects can find their host plants over long distances to feed and oviposit, even though the host plants are hidden in a range of other plants and plant volatile organic compounds play an important role in the host location process. Insects rely on a powerful olfactory system, with olfactory receptor neurons able to identify volatiles cues, made by specific key compounds or specific blends emitted from suitable host plants [55,56]. For example, diterpene hydrocarbons released by the seedlings of brassicaceous hosts Brassica oleracea and Brassica napus species, alone or in combination with one or more minor compounds, are key vectors for host localization by Bagrada hilaris [57,58].
In many herbivorous species, female adults avoid reproduction in places where their offspring are at a high risk of predation [59][60][61][62]. In this study, many natural enemies, such as ladybeetles (Harmonia axyridis and Coccinella septempunctata), lacewings (Chrysopa spp.), and Pentatomidae, were observed on sunflower leaves (data not shown). This study confirmed that O. communa is safe for use as a biological control agent to control ragweed, based on its host-selection behavior in an open field experiment. However, when all common ragweed plants are completely eradicated or defoliated from the local population, the leaf beetle O. communa of suboptimal alternative host plants (A. trifida and H. annuus) should not be ignored. It has been reported that O. communa began feeding on A. trifida after all A. artemisiifolia plants were defoliated under field conditions in Japan [63].
In our choice test using closed cages in the field, most O. communa adults (25%) moved rapidly from H. annuus to X. sibiricum, in addition to the host-plant A. artemisiifolia, to feed and lay eggs, and several were found feeding and oviposition on X. canadense but not on H. annuus. Additionally, when A. artemisiifolia died in October, many O. communa adults moved to X. canadense to prepare for overwintering (unpublished data). In Japan, O. communa can also be found feeding on X. canadense, and adults in the field have been found to move to X. canadense overwinter after the death of A. artemisiifolia in late summer [39]. Our results indicated that X. sibiricum may be a suitable host-plant species for this beetle. This result is consistent with those of Cao et al. [43] and Liu et al. [64], who suggested that X. sibiricum could be used as a lower-ranked host plant next to the target weed. In China, X. canadense is a common weed in cultivated fields, especially in soybean, tobacco, and sunflower fields [65][66][67]. Therefore, O. communa may be used to control X. canadense in China in the future.
In its native range, O. communa does not utilize A. trifida as a host plant [36,54,68,69], and, in our study, although A. trifida was attacked by several adults, the damage level was very low. Therefore, the beetle cannot effectively control A. trifida, and there have been no reports on the use of O. communa to effectively control A. trifida in China. However, this beetle has been reported to feed extensively on A. trifida in fields throughout the Japanese islands [35,36,63]. These results indicate the expansion of the host range of occasionally introduced O. communa, which may be the result of the co-evolution of herbivorous insects and host plants [19,69,70].
In our study, after the adult O. communa that were released on A. artemisiifolia plants in the center of the sunflower field completed one generation (approximately 30 days), they initiated a search for suitable host plants across the sunflower barrier (planted in homocentric rings with a radius of 3 m), and the number of O. communa in the peripheral population was supplemented by the population in the center. This result indicated that O. communa has a robust capacity to find A. artemisiifolia for feeding. The "resource concentration hypothesis" predicts that specialist herbivorous insects are more likely to find and stay longer on host plants growing in dense or nearly pure contexts [71][72][73]. In our study, the cumulative densities of O. communa feeding or remaining on A. artemisiifolia did not differ among plant clusters of different densities, which does not support the "resource concentration hypothesis". This result was consistent with the observations of Yamanaka et al. [54]. Insects with high dispersal abilities may not be limited by patch borders. Hence, their densities per plant did not differ among host-plant patches of different sizes [74]. In addition, if patches are closer together, insects may move more easily between them, thus diminishing differences in density [75,76]. O. communa has been shown to rapidly disperse after introduction into a new area [77][78][79], and our results further support this high dispersal ability. Certainly, the proximity of the A. artemisiifolia clusters (< 4.5 m) may have resulted in a lack of difference in the densities of O. communa in plants. In this study, we used well-established plants as test plants (all 135 individuals of each species were used for the experiments when they were approximately 20-30 cm in height). However, sunflower was sown; thus, seeding and younger sunflower plants were exposed to O. communa and these might be more susceptible to feeding and oviposition. More experiments are needed to confirm this in the future. Finally, how can we better predict the long-term benefits and risks of ragweed biology control? We advocate research on host specificity and population differentiation before the release of biocontrol agents to promote the development of improved biological control under changing global conditions [3,19,80].

Conclusions
In summary, by observing the host-plant selection behavior of O. communa, we conclude that this beetle poses no threat to the biosafety of H. annuus. In addition, X. sibiricum has the potential to become an alternative host plant for O. communa in the future; however, it cannot efficiently control A. trifida in China. In the open field study, some O. communa adults crossed the barrier formed by H. annuus to feed and lay eggs on A. artemisiifolia planted in the periphery, and the spatial interactions between A. artemisiifolia and O. communa did not support the "resource concentration hypothesis". We conclude that O. communa has a robust dispersal capacity to find and feed on A. artemisiifolia.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/insects14040334/s1, Figure S1: Occurrence and frequency of Ophraella communa individuals in different developmental stages on A. artemisiifolia planted at different distances; Table S1: Three-way ANOVA of the effects of four plant species, days after release, and cages (blocks) on Ophraella communa distribution (adults and eggs) in outdoor cages; Table S2: Three-way ANOVA of the effects of plant species, distance from center, and ragweed cluster density on the number of O. communa individuals in different developmental stages on A. artemisiifolia and H. annuus planted.