Combining Banker Plants To Achieve High Pest Control Eciency In Multi-Pest, Multi-Natural Enemy Cropping Systems

Banker plants increase biological pest control by supporting populations of non-pest arthropod species used as alternative hosts or prey by natural enemies. Due to the specicity of trophic interactions, banker plants may not eciently promote natural enemies with different ecologies. Yet in most cropping systems different pest species are present together and require different biocontrol agents to eciently control them. In the present study, we tested the combined use of two banker plants and their associated prey / host to enhance populations of the specialist parasitoid Encarsia formosa targeting the main tomato pest Bemisia tabaci, and a polyphagous ladybird Propylea japonica targeting the secondary pest Myzus persicae in tomato crops. In a laboratory and a greenhouse experiment, we measured the abundances of these four species using the Ricinus communis – Trialeurodes ricini banker plant system alone, in combination with the Glycines max – Megoura japonica system, or in absence of banker plants. We found that the rst banker plant system enhanced populations of E. formosa, resulting in higher control of B. tabaci populations and the suppression of their outbreak in both our laboratory and greenhouse experiment. Conversely, abundances of P. japonica were not affected by this rst system, but were signicantly increased when the second was present. This resulted in high control of M. persicae populations and the suppression of their early and late outbreaks. Our study demonstrates the potential for combined banker plants to provide long-term, sustainable control of multiple pests by their target natural enemies in complex agroecosystems.


Introduction
With the rise of integrated pest management (IPM), practical applications of biological control agents releases have become increasingly complex to improve pest control with reduced used of chemical pesticides (Kogan 1998 Zang et al. 2021). Starting from simple pairs with one natural enemy species targeting one pest species, applications have been progressively improved in releasing combinations of multiple species of biocontrol agents to achieve sustainable control of multiple pest populations simultaneously and throughout the cropping season (Barbosa 1998;Heimpel & Mills 2008). Such e cacy is achieved notably through the precise monitoring of pest population dynamics in situ (Tan et al. 2016). Multi-species releases should also involve natural enemy species with limited niche competition (Liu et  Yet such systems are intrinsically complex to develop since many factors may in uence species interactions and the e cacy of pest control, including biocontrol agent tness, their relative proportion, introducing times and developmental stages (Huffaker et al. 1971). The successful colonization of crop systems, and the establishment of stable populations by released biocontrol agents is even more challenging and may heavily rely on means to support populations (Bianchi et al. 2006). This is especially true since biocontrol agents are most often released before pest populations reach high densities to prevent pest outbreak (Albajes et al. 2000). Therefore the development of strategies supporting biocontrol agent populations could be useful in extending the adoption of multi-species releases.
The use of banker plants has been increasingly investigated and developed in the context of conservation biological control (Frank 2010; Gurr et al. 2017). Banker plants are non-crop plants supporting populations of herbivorous species which do not attack adjacent crops (Parolin et al. 2012), and which may be used as alternative prey or hosts by natural enemies targeting the main pest species in cropping systems. Hence, banker plants can be used in combination with crops and serve as entry points for the inoculative release of biocontrol agents. By providing shelter and alternative prey / hosts, banker plants can enable the early colonization of adjacent crops by natural enemies  and the establishment of their populations when main prey are scarce (Yano et al. 2018). Plant species supporting populations of alternative prey / hosts in the same family or functional group than the main pest species are well suited to be used as banker plants (Laurenz & Meyhofer 2021). For instance, the non-crop oat species Avena sativa supporting populations of the alternative aphid species Metopolophium dirhodum resulted in an increased suppression of the main pest Myzus persicae by its parasitoid Aphidoletes aphidimyza in chilly crop (Capsium annuum; Hansen 1983). Similarly, the noncrop papaya plant Carica papaya supporting populations of the alternative white y species Trialeurodes variabilis successfully increased the suppression of the main pest Bemisia tabaci by the ladybird Delphastus pusillus in many vegetable crops (Osborne et al. 1991). These non-crop plants and alternative prey could also supply su cient food resources and space to help biocontrol agents maintain their populations after the suppression of the main pest species and thereby avoid their escape from the target agroecosystem, which is common in practical biological control application and may cause severe loss (Huang et al. 2011). Such effect could in turn increase the suppression of secondary outbreaks of pest populations (Zheng et al. 2017 Monticelli et al. 2021). Hence, the density of banker plants and alternative host / prey could severely impact the balance between the target agroecosystem and the banker plant system (Orrock et al. 2010). Most studies investigating the suitability of species as banker plants focused on a single plant species supporting a single alternative herbivorous prey species (Andorno & Lopez, 2014). Conversely in practical applications relying on multispecies releases of biocontrol agents, many functional non-crop plants have been studied for their potential to provide alternative food resources and shelter to these multiple natural enemies, but they have seldom been evaluated as constituents of a  Aparicio et al. 2020). Therefore, the use of distinct banker plant systems supporting E. formosa and P. japonica could reduce the risk for intraguild predation and help enhance the control of the B. tabaci -M. persicae pest complex in greenhouse tomato crops.
In the present study, we tested the role of combined banker plant applications in enhancing pest control in a multi-pest, multi-biocontrol agent system in a laboratory and a greenhouse experiments. Using tomato crops, we measured the abundance of the pest species B. tabaci as main tomato pest and M. persicae as secondary pest, and of the introduced biocontrol agents E. formosa and P. japonica. We evaluated the potential of the two banker plant systems in combination: (i) R. communis supporting populations of the white y T. ricini, itself parasitized by by E. formosa and (ii) G. max supporting populations of the aphid M. japonica, itself preyed upon by P. japonica (Table 1). We asked: How e cient is the long-term control of the two pest populations by the two natural enemies (1) In absence of banker plants ? (2) When one banker plant system is provided ? And (3)  All colonies were maintained for a year (from May 2019 to June 2020) and had a mixed population structure with both adults and nymphs present simultaneously. Plants were replaced every 5-7 days when needed. About 3-4 and 12-15 generations completed before the 2019 laboratory experiment and the 2020 greenhouse experiment, respectively.

Natural Enemies
Encarsia formosa and P. japonica were bought from Hengshui Tianyi Biocontrol Co., Ltd, Hengshui, Hebei, China. Commercial E. formosa were provided at the pupal stage parasitizing B. tabaci white ies at the 3rd -4th nymphal stage and with a single parasitoid wasp per nymph, as evidenced by a visible black dot inside the nymph. White ies were packaged in groups of 1,000 nymphs on one tomato leaf. The newly emerged viable E. formosa females were collected with a vacuum pipe and maintained in Petri dishes at 10-15 wasps per dish, in which honey (Baihua 75%; Bejing Aojinda Honey Co., Ltd.) was provided for feeding. Wasps were used in the experiments about 1-2 days later. Commercial P. japonica were provided as 4th instar larvae, packaged in groups of 30 larvae in a tetrahedron shape paper container lled with sawdust to reduce cannibalism. The larvae were kept in plastic boxes (30 per box, 15 × 15 × 20 cm) in the insectary and fed with daily supplied arti cial diet microcapsules (Tan et al. 2015) for the experiments.

Laboratory experiment
In 2019 in a laboratory experiment, we evaluated the potential for the combined use of the R. communis -B. tabaci and the G. max -M. japonica banker plant systems to enhance pest control by the combination of the specialist E. formosa and the generalist P. japonica biocontrol agents (Table 1). We tested four treatments: (a) control with pest species B. tabaci and M. persicae on tomato plants, but no biocontrol agents and no banker plant systems; (b) control with the two pest species and the two biocontrol agents but no banker plant systems; (c) one-banker plant system with the two pest species and the two biocontrol agents and with the R. communis -B. tabaci banker plant system; and (d) two-banker plant system with the two pest species, the two biocontrol agents, and the two banker plant systems (R. communis -B. tabacis and G. max -M. japonica).
In a cage in the insectary (made of hollow aluminium pipes of diameter 1.5 cm and 40-mesh fabric net walls; 1.8 × 1.2 × 1.5 m; Fig. 1), we placed 15 tomato plants and four banker plants infested with their respective alternative host / prey beforehand. In each cage, two R. communis banker plants were placed in A and C. In the two-banker plant system cages, two G. max plants were also added in B and D (Fig. 1), while in control cages no banker plant was provided. Ten days prior to the start of the experiment, tomato and R. communis plants were infested with B. tabaci and T. ricini white ies, respectively: 10 pairs of adults were enclosed on each of the ve plant leaves with a 40-mesh net. The adults were removed after they had laid eggs two days later. Then, eight days later the number of 3rd -4th instar nymphs was adjusted to a 100 per tomato plant and 200 hundreds per R. communis plant by gently removing excess nymphs with a brush. Tomato plants were further infested with 30 M. persicae 3rd -4th instar aphid nymphs per tomato plant (taken directly from rearing cages) while G. max were infested with 250 M. japonica aphid nymphs each. Finally, one day after aphid infestations, 20 parasitoid female adults were introduced on each R. communis plant and ve ladybird 4th instar larvae on each G. max plant if present, or per cage if the banker plants were absent and except in the control -no natural enemy treatment. Ten replicates per treatment (control, no biocontrol agent / control with biocontrol agents / one-banker plant system / two-banker plant system) were produced, with a total of 40 cages. The densities of insects were chosen based on a pilot experiment. We visually observed and counted the number of individuals of each species (B. tabaci, M. japonica, E. formosa and P. japonica) of all developmental stages (except eggs) on six randomly selected tomato plants per cage starting one week after the release of biocontrol agents and every Monday from June 3rd to September 30th 2019. We calculated the total number of insects of each species per six plants by summing numbers counted in each of the six plants so as to obtain one value per cage and per week.

Greenhouse experiment
In 2020 we performed a greenhouse experiment to estimate the potential for the combined banker plant systems to be used in practical biocontrol greenhouse applications involving multiple pest and biocontrol agent species. We tested the same four treatments as in our 2019 laboratory experiment (control, no biocontrol agent / control with biocontrol agents / one-banker plant system / two-banker plant system). We used four glass greenhouses located in the Noah Organic Farm (100 × 14 m, height 4.2 m). Each greenhouse was composed of ten independent chambers (12 × 12 m) isolated with plastic membranes preventing arthropod movement between chambers. In each chamber, we transplanted 56 tomato plants in seven rows with eight plants per row (Fig. 2). We transplanted two R. communis plants in each of four random points among the nine blue points shown in Fig. 2 in each chamber. In the two-banker plant system chambers, we also added two G. max plants in each of four other randomly selected points. In the control chambers, no banker plant nor associated alternative prey / host was provided. Two days after transplantation, 14 tomato plants randomly selected in each chamber were infested with B. tabaci adults and the R. communis banker plants were infested with T. ricini adults, following the same method as in the laboratory experiment. After three days -white ies nymphs had emerged by then -another 14 randomly selected tomato plants were infested with 30 3rd -4th M. persicae aphid nymphs per plant, and G. max plants were infested with 250 M. japonica aphid nymphs each. After two more days, 20 parasitoid female adults were released on each of the eight R. communis plants, and ve ladybird 4th instar larvae were released on each of the eight G. max plants per cage if the banker plants were present, or in four of the nine blue points randomly selected if the banker plants were absent. In the control -no natural enemy treatment, no biocontrol agents were released. Each treatment was replicated in ten chambers, randomly selected across the four greenhouses. The insect densities in greenhouses were chosen based on authors' preliminary surveys.
We monitored population dynamics of the two targeted pest species (B. tabaci and M. persicae) and of the two introduced biocontrol agents (E. formosa and P. japonica) starting one week after biocontrol agent releases and every Monday from June 1st to September 28th 2020. At each sampling date, we selected randomly eight plants per chamber and we inspected all plant parts and counted all insect individuals (except eggs). We calculated the total number of insects of each species per eight plants by summing numbers counted in each of the eight plants, so as to obtain one value per greenhouse chamber and per week. The environmental conditions inside greenhouses followed seasonal trends (June: 26.9 ± 0.4°C, 32.2 ± 1.9 % RH; July: 26.3 ± 0.4°C, 41.6 ± 1.6 % RH; August: 26.2 ± 0.4°C, 36.5 ± 1.9 % RH; September: 21.4 ± 0.5°C, 24.8 ± 1.5 % RH).

Statistical analyses
All statistical analysis were performed with R Core Team (2020) version 3.6.3. We analysed the impact of the treatment (control -no natural enemy / control -with natural enemies / one-banker plant system / two-banker plant system) on the number of insects per six plants in the laboratory experiment independently for each insect pest (B. tabaci vs. M. persicae) and natural enemy (E. formosa vs. P. japonica) species because of large differences in densities and population dynamics (Fig. 3). We used Generalized Linear Models (GLMMs) with a negative binomial distribution recommended for count data with overdispersion (function 'glmer.nb', library 'lme4'; Bates et al. 2015). The treatment was used as xed effect, while the cage and the week were used as random effects to account for repeated measures through time, and patterns of insect population dynamics, respectively. The signi cance of the treatment was tested with an anova based on a χ 2 test. Model validity was veri ed a posteriori (functions 'simulateResiduals' and testDispersion', library 'DHARMa'; Hartig 2020). To assess whether means across treatments were signi cantly different, we performed a post hoc comparison of means by computing estimated marginal means (function 'emmeans', library 'emmeans'; Lenth 2021). Statistical tests used for the greenhouse experiment were identical to those described for the laboratory experiment.

Laboratory experiment
The presence of natural enemies and of banker plants signi cantly affected the population densities of all four species (Table 2, Fig. 3). The presence of natural enemies prevented the early pest outbreaks of B. tabaci (populations 1.6 times lower in average: mean ± SE of individuals per six plants: control -no natural enemy 777 ± 29, control -with natural enemies 485 ± 7; Table 3, Fig. 3A) and of M. persicae (populations 1.4 times lower in average: control -no natural enemy 368 ± 10, control -with natural enemies 257 ± 7; Table 3, Fig. 3B) in Weeks 2-8. Adding one banker plant type (R. communis -T. ricini) caused a further signi cant reduction in B. tabaci numbers (297 ± 6, 0.6 times lower) compared with the control -with natural enemies systems, but this was not true for M. persicae (256 ± 7). Finally, adding a second banker plant type (G. max -M. japonica) caused a further signi cant reduction in M. persicae numbers (169 ± 5, 1.5 times lower) compared with the control -with natural enemies systems or the onebanker plant systems, and thanks to the suppression of the second aphid population outbreak in Weeks 10-14. Conversely in B. tabaci, no second outbreak was observed, and numbers were not different between the two-banker plant systems (293 ± 9) and the one-banker plant systems.
Populations of E. formosa were enhanced by the presence of the R. communis -T. ricini banker plant systems with a marked increase in numbers per six plants at Week 5 (Fig. 3C) resulting in a 1.2 times increase in average in the one-banker plant systems (35 ± 1) compared with the control -with natural enemies systems (21 ± 1; Table 3, Fig. 3C). However, adding a second banker plant type had no signi cant effect on E. formosa numbers compared (37 ± 2) with one banker plant type only. Finally, P. japonica numbers per six plants were not signi cantly different in presence (5.2 ± 0.2) or in absence of one banker plant type (5.0 ± 0.2) but they were 3.2 times higher in the two-banker plant systems (16 ± 1; Fig. 3D, Table 3).

Greenhouse experiment
Similar to the laboratory experiment, the presence of banker plant systems in the greenhouse experiment signi cantly impacted the population densities of all four species (Fig. 4; Table 2). Like in the laboratory experiment, the presence of natural enemies prevented the rst pest outbreak in Weeks 2-8, with B. tabaci numbers per eight plants 1.6 times lower in the control -with natural enemies systems (672 ± 11) compared with the control -no natural enemy systems (1,068 ± 22; Fig. 4A, Table 4), and M. persicae numbers 1.3 times lower in the control -with natural enemies systems (182 ± 6) compared with the control -no natural enemy systems (250 ± 6; Fig. 4B, Table 4). Also, adding one banker plant system (R. communis -T. ricini) caused a further signi cant reduction in B. tabaci numbers (428 ± 6, 1.6 times lower) compared with the control -with natural enemies systems, but this was not true in M. persicae (180 ± 6). Finally, adding a second banker plant system (G. max -M. japonica) resulted in a further decrease in M. persicae numbers (105 ± 3, 1.7 times lower) compared with the control -with natural enemies systems, again thanks to the suppression of the second aphid population outbreak in Weeks 10-14. Conversely and consistent with the laboratory experiment, there was no second population outbreak in B. tabaci populations, and their numbers in the two-banker plant systems (406 ± 7) were not different than in the one-banker plant systems.
Consistent with the laboratory experiment, E. formosa populations were enhanced by the presence of the R. communis -T. ricini banker plant system, with populations increasing from Week 3 and 2.1 times higher in average in the one-banker plant systems (mean ± Se of individuals per eight plants: 145 ± 7) compared with the control -with natural enemies systems (67 ± 3; Fig. 4C, Table 4). However, their numbers were not different in the two-banker plant systems (142 ± 6) compared with the one-banker plant systems. Finally, P. japonica numbers per eight plants were not different between the one-banker plant systems (41 ± 2) and the control -with natural enemies systems (42 ± 2) but they were signi cantly increased and 2.2 times higher in the two-banker plant systems (92 ± 4) compared with the one-banker plant systems, thanks to a continuous increase in population densities from Week 3 to 13 (Fig. 4D, Table 4).

Discussion
Conservation biological control has been widely proposed as the solution to increase biodiversity and the colonization and tness of natural enemies within cropping systems notably via the introduction of noncrop functional plants (Naranjo et al. 2015). These solutions could decrease the input costs of biocontrol agent releases, increase pest control effectiveness and limit non-target impact risks during biological control processes (Stiling & Simberloff 2000). In particular, the use of banker plants could be an economically sustainable solution to enhance the populations of natural enemies via banker plants supporting alternative arthropod prey or host populations (Huang et al. 2011). Yet with the increased complexity in biocontrol application systems, the use of single banker plant systems in such complex cropping systems may not be e cient to support multiple released biocontrol agent species. In the present study, we tested the potential for a combination of two banker plant systems to promote the simultaneous control of two major tomato pests by two biocontrol agents, in a laboratory and a greenhouse experiment. We showed that combined banker plant systems increased the control of both the main tomato pest Bemisia tabaci by Encarsia formosa and the secondary pest Myzus persicae by Propylea japonica in both laboratory and greenhouse settings. Notably, adding the Ricinus communis -Trialeurodes ricini banker plant system allowed the suppression of the early population outbreak in B. tabaci, while adding the Glycines max -Megoura japonica system allowed the suppression of the two population outbreaks in M. persicae.
As expected, we found that the R. communis -T. ricini banker plant system promoted the populations of E. formosa, resulting in an increased control of B. tabaci populations, while only the use of the G. max -M. japonica system promoted P. japonica populations and resulted in an increased control of the secondary pest M. persicae. This proves the e cacy of the chosen banker plant systems to promote these target biocontrol agent species, but also their strong speci city to the target biocontrol agents. This is because E. formosa is specialized on white ies, and hence cannot use aphids as alternative hosts (Tao et al. 2018). Similarly, the predatory ladybird P. japonica, although polyphagous (Yang et al. 2014), primarily feeds on aphids and has a reduced tness when fed on B. tabaci (Liu et al. 2008). This would explain that the R. communis -T. ricini system did not enhance P. japonica populations resulting in poor control of M persicae. Indeed the densities of P. japonica populations were too low to prevent the late population outbreak of M. persicae, suppressed in the two-banker plant systems only, in which P. japonica populations reached twice as high densities. The limited overlap between natural enemies supported by distinct banker plant systems could be bene cial to avoid the disruption of pest control. Such disruption could occur if natural enemies compete for the same pest resource and this could notably cause population outbreaks of secondary pest species (Bhattacharyya & Sinha, 2009). Both in the laboratory and greenhouse experiments, we observed niche partitioning between the main tomato pest B. tabaci and the secondary tomato pest M. persicae in absence of natural enemies. Aphid populations rose rapidly in Weeks 2-6 and then decreased in Weeks 7-10 under high summer temperatures, while B. tabaci populations rose slower to peak in Week 10. However once B. tabaci populations decreased in late summer, aphid populations broke out again. Such dynamics are common in tomato cropping systems (Lu et al. 2004;Ap et al. 2019). In the banker plant systems, the alternative host T. ricini and prey M. japonica did not occupy the niche left empty by the control of B. tabaci and M. persicae populations as they preferred their primary host plants R. communis and G. max, respectively. Therefore they did not damage tomato crops, showing their suitability as banker plant systems for pest control in commercial greenhouse crops.
In the present study, we deliberately released E. formosa at higher densities than P. japonica, and proportionally to the densities of their respective target pests B. tabaci and M. persicae. These pest densities simulated real conditions commonly found in commercial greenhouse cropping systems, and in response multi-species releases of biocontrol agents are often asymmetric Ma et al., 2018). The specialist natural enemy is released at higher densities to control the main pest species, while the generalist natural enemy is released at lower density to provide the suppression of the populations of secondary pest species throughout the cropping season (Garey & Ru é 1987; Fonseca et al. 2020).
One key aspect in the industrial development of banker plant systems is to adjust the balance between the densities of the main and alternative hosts / prey (Frank 2010). Since they are targeted by a shared . Hence, adding too many banker plant units may cause natural enemies to reduce pressure on main pests and resulting in poor control. Conversely, adding too few banker plant units may be insu cient to enhance natural enemy populations, resulting in too low densities and poor establishment of their populations and therefore a poor pest control in the long term. Future research in banker plant systems should focus on developing exible systems to accurately adjust the density of alternative prey / hosts and improve the sustainable control of arthropod complex by using multi-species biocontrol agent releases. Finally, it could be useful to investigate the potential for dual functions of banker plants, such as supporting populations of alternative prey / hosts and providing alternative oral resources (Wäckers & van Rijn 2012). This would allow the design of IPM strategies via the support of populations of specialist and generalist biocontrol agent species in multi-species systems providing pest control over the short-term and the long-term.
In conclusion, our study showed that the combined use of multiple banker plant systems targeting distinct biocontrol agents signi cantly increased the control of both main and secondary pest populations. Banker plants were necessary to avoid pest outbreaks and to provide long-term pest control via the establishment of natural enemy populations at high densities.   Table 2. Impact of the treatment (control -no natural enemy / control -with natural enemies / onebanker plant system / two-banker plant system) on the abundances of the two pest species B. tabaci and M. persicae and of their respective biocontrol agents E. formosa and P. japonica in the laboratory and the greenhouse experiments.
'***': P < 0.001... Table 3. Comparisons of means between treatments in the number of individuals per six plants for each insect species in the laboratory experiment (emmeans: contrast estimates ± SE and associated P-value).

Page 19/22
'***': P < 0.001. Table 4. Comparisons of means between treatments in the number of individuals per six plants for each insect species in the greenhouse experiment (emmeans: contrast estimates ± SE and associated P-value).