Banker plants promote functional biodiversity and decrease populations of the cabbage whitefly Aleyrodes proletella

In this study, potential banker plant systems against the cabbage whitefly Aleyrodes proletella Linnaeus (Hemiptera: Aleyrodidae) were developed under controlled conditions. The two most promising banker plant systems, that is, the parasitoid Encarsia tricolor Förster (Hymenoptera: Aphelinidae) either with Aleyrodes lonicerae Walker on European columbine (columbine system) or with Trialeurodes vaporariorum Westwood (Hemiptera: Aleyrodidae) on Hokkaido squash (pumpkin system), were further evaluated in the field. Although the pumpkin system produced three times more parasitoids than the columbine system, both banker plants led to an 1.5‐fold increase in A. proletella parasitism rates. However, only the pumpkin system increased the abundance of syrphid larvae on cabbage by 61.5% and reduced A. proletella populations on average by 4.4%–25.8% depending on the respective assessment date. In conclusion, the pumpkin system revealed to be a promising (supplementary) control measure against A. proletella. Options for further improvement and standardization of the pumpkin system as well as a potential implementation in cabbage production are discussed.

Especially organic cabbage producers lack efficient control measures against A. proletella. Although conventional control with chemical insecticides has proven high efficacies against A. proletella under controlled and field conditions, there is an increasing demand for insecticide-free and ecologically sustainable pest management (Kovaříková et al., 2017;Richter & Hirthe, 2014). Alternatives to conventional insecticides are therefore desperately needed. Besides cropping of resistant cabbage plants (Hondelmann et al., 2020), cultural control measures like fine crop cover netting can lead to a remarkable reduction of A. proletella populations (Saucke et al., 2011). However, this effect often disappears later in the growing season; crop cover nets are labour intensive, may facilitate aphid pests and hinder colonization by natural enemies (Ludwig & Meyhöfer, 2016).
Biological control strategies exploiting natural enemies pose another non-chemical alternative to reduce A. proletella populations on cabbage crops.
Well-timed release of E. tricolor may serve a solution. Although mass release by hand is promising under semi-field conditions on caged cabbage plants, results under field conditions are variable (Saucke et al., 2011;Springate, 2016). A release of natural enemies on banker plants can be more effective than hand-release (Kidane et al., 2018;Pickett et al., 2004). Banker plants are a biological control method for an early and continuous release of natural enemy populations in a crop. A banker plant system consists of a plant species deliberately infested with herbivores which serve as alternative hosts/prey for natural enemies of the target pest. Alternative hosts/ prey as well as shelter and reproduction habitats are permanently provided even in absence of the pest.
Banker plant systems are well adopted in greenhouse crops against aphids, whiteflies and other pests (Huang et al., 2011).
The aim of this study is to develop and evaluate different banker plant systems with E. tricolor as natural enemy against A. proletella under controlled conditions and in the field.

| Insects and plants
All insects used in this study derived from rearings established at the Section Phytomedicine, Institute of Horticultural Production Systems, Leibniz Universität Hannover, Germany. The whiteflies, that is, A. lonicerae, A. proletella and T. vaporariorum, were reared on Aegopodium podagraria (ground elder), Brassica oleracea var. gemmifera (Brussels sprouts) and Cucurbita maxima 'Uchiki Kuri' (Hokkaido squash), respectively. Recently emerged females (wings translucent and not yet fully expanded) and males were collected from the respective rearing to be used in the experiment. Aleyrodes proletella on Brussels sprouts was used as rearing system for E. tricolor. Mated E.
tricolor females for experiments were produced as follows: leaves with parasitized whitefly puparia (dark) were placed in a plastic box with a translucent, perforated lid and incubated at room temperature for 48 hr. About 20 emerged females and five males were then transferred to a snap-cap vial (height: 80 mm, diameter: 30 mm) with a perforated lid and honey droplets as food source. They were kept together for another 4 days. The 5-6 days old females were then expected to be mated and were used for experiments. Ground elder was propagated from rhizomes, all other experimental plants from seeds. Plants were grown in pots (diameter: 120 mm) under greenhouse conditions before being used for experiments when three to six true leaves were fully expanded.   (Bährmann, 2002;Huldén, 1986;Mound & Halsey, 1978). These herbaceous plants are compatible with the environmental conditions on Central European farmland and are neither known as economically relevant pest weeds nor as host plants for A. proletella (Ellenberg, 1979;Mound & Halsey, 1978;Weber & Gut, 2005). The performance of A. proletella was evaluated on the seven most suitable host plants for A. lonicerae and T. vaporariorum and on Brussels sprouts as control treatment. Therefore, a recently emerged whitefly female and three conspecific males were transferred to a leaf cage (Ø 140 mm), which was fixed to the first fully expanded true leaf of a plant (Rechner et al., 2017). The leaf cage was built of a Petri dish with a gauze lid; a small opening (sealed with cotton wool) on one side for the leaf petiole enables to insert an entire living leaf or a leaflet. A construction of three wooden sticks glued to a Petri dish lid held the cage stable in the appropriate position and a clip on each wooden stick enabled to adjust the height of the cage. The males were removed after 14 days. Female mortality was evaluated every 2-3 days to compare its longevity on the different plant species. In an interval of 14 days, females were shortly immobilized with carbon dioxide and gently transferred to another plant of the same age. Deposited eggs (sum of hatched and not hatched eggs) were counted and further observed to determine egg-adult development times and nymphal mortality. Six replicates were conducted of each treatment. Experiments were performed under controlled conditions at 16 hr day (light, 25 ± 1°C, 61 ± 5% r.

| Parasitoid performance
The performance of E. tricolor was investigated on the following four whitefly-plant combinations: A. lonicerae on the two most suitable host plants based on the whitefly's performance, T. vaporariorum on Hokkaido squash and A. proletella on Brussels sprouts (control). Therefore, 30 adult whiteflies were caged to the underside of the youngest fully expanded leaf for 48 hr. The offspring developed into the secoond and third nymphal stage before being gently reduced to 25 individuals per leaf with the help of a dissection needle. A 5-6 days old mated E. tricolor female was then caged to the whitefly nymphs and allowed to deposit eggs for 24 hr. Parasitoid offspring that turned into pupae (dark whitefly nymph) were individually transferred to gel capsules and further observed until adult emergence.
The numbers of pupal and adult offspring as well as developmental times were evaluated daily. To investigate parasitoid fitness, the head widths of emerged E. tricolor adults as well as length and width of the respective whitefly nymphs' exuviae were determined under the microscope (van Lenteren et al., 1976;Williams, 1995). The areal host size was calculated from the exuviae length and width by expecting an ellipse host shape applying the following formula: areal host size = π (length/2) (width/2). Each treatment was replicated 14-21 times and all E. tricolor females within one treatment derived from different leaves. Experimental conditions were the same as described before.

| Field evaluation
Two banker plant systems were evaluated under field conditions, that is, European columbine with A. lonicerae and E. tricolor (columbine system), and Hokkaido squash with T. vaporariorum and E. tricolor (pumpkin system). An experimental plot consisted of two areas of Brussels sprouts plants (each 4 m × 2.4 m, 0.6 m between rows, 0.5 m between plants). The area between the Brussels sprouts (2.4 m × 2 m) was planted with the columbine system (0.25 m between rows and plants), the pumpkin system (0.75 m between rows and plants) or covered with mulch foil to prevent vegetation (control treatment). Plots were arranged in a randomized block design with six replicates and 14-17 m distance between plots. The space between the plots was covered with grass, which was kept short by regular mowing.
The banker plant species were grown separately under two gauze tents in the greenhouse. When one to three true leaves were fully expanded, European columbine and Hokkaido squash plants were evenly infested with 3,000 females of the respective alterna-
Under controlled conditions, the whitefly performance (longevity, fecundity, developmental success) on selected plant species was evaluated by applying Kruskal-Wallis rank sum tests with Dunn's post hoc and Holm-adjusted p-values (R package 'PMCMRplus'; Pohlert, 2018). Egg-adult development times of the parasitoid E. tricolor were analysed with a linear model (lm) followed by an analysis of variance (ANOVA function) and multiple comparisons of means after Tukey, if applicable (package 'multcomp'; Hothorn et al., 2008).

The nonparametric Kruskal-Wallis test with Dunn's post hoc and
Holm-adjusted p-values was applied to analyse the head width of E.
tricolor offspring, as well as the number of pupal and adult offspring and the size of the whitefly host. The relationship between the head width of E. tricolor and the host size was analysed by Spearman's rank correlation after assessment of linearity (scatter plot) and bivariate normal distribution ('MVN' package; Korkmaz et al., 2014).
For the field experiment, data collected on the 28.09.2015 were excluded from statistical analysis due to mistakes in data collection.
The numbers of parasitized whitefly puparia on banker plants were summed per plot before determining differences between the two banker plant systems with a two-sided Wilcoxon rank sum test ('PMCMRplus' package; Pohlert, 2018). Generalized linear mixed-effects models (glmer) fit by maximum likelihood were applied to compare the banker plant systems with each other and with the control treatment in terms of parasitism rates as well as numbers of herbivores, E. tricolor adults and predators on cabbage plants (response variables; package 'lme4'; Bates et al., 2015). Count data were fitted with negative binomial models (glmer.nb) and a log link function to account for overdispersion in count data (Hilbe, 2011). Dispersion parameters were determined with package 'blmeco' (Korner-Nievergelt et al., 2015). A binomial distribution with a logit link function was used in case of parasitism rates. Explanatory variables were the treatment (pumpkin system, columbine system and control), assessment date, distance to the banker plants (1.5 and 4 m), position to the banker plants (upwind and downwind) and block (one to six).
An identification number (plot ID) was assigned to each of the 18 plots. Plot ID was taken as random effect to account for temporal non-independence of repeatedly collected data from the same plots. proletella parasitism rate on cabbage plants 14d later while controlling for assessment date and blocks (package 'ppcor'; Kim, 2015).

| Performance of alternative hosts
The longevity of the alternative hosts on the evaluated host plants

| Performance of the cabbage whitefly
Longevity of A. proletella differed between plant species (χ 2 (7, N = 48) = 15.02, p = .036). Aleyrodes proletella survived most days on Brussels sprouts (control; 46 ± 12 days), but it died earlier on Hokkaido squash (p = .046). No differences were detected between the seven potential banker plant species selected based on the alternative host performances (all p > .05); longevity on these plants ranged between 9 ± 2 days on Hokkaido squash and 45 ± 13 days on C. persicifolia.
Additionally, hosts from AL/GU were larger than ones from AL/AV (p < .001). Hosts from AP/BO did not differ in size from hosts on AL/AV and AL/GU (both p > .05). Parasitoid head widths correlated positively with the size of the hosts they emerged from (r s = .71, p < .001; Figure 2).

| Field evaluation
The sum of parasitized puparia over the entire growing season was three times higher on the pumpkin system (1,626 ± 266 individuals m −2 ) compared to the columbine system (546 ± 134 individuals m −2 ; U = 0, p = .002). A higher parasitoid production by the pumpkin system was determined on the 3rd, 17th (peak) and 31st August 2015 (p = .030, p = .002 and p = .002, respectively). There was no difference on the other assessment dates (p > .05). The number of parasitized whitefly puparia on pumpkin and European columbine correlated positively with the parasitism rate of A. proletella on cabbage 14 days later (r s = .82, p < .001 and r s = .51, p < .001, respectively; Figure 3).
The banker plants affected A. proletella parasitism rates on cabbage (χ 2 (2, N = 700) = 41.92, p < .001). The pumpkin system (22.7 ± 1.2%) as well as the columbine system (22.5 ± 1.2%) increased the average whitefly parsitism rate on cabbage plants compared to the control without banker plants (14.9 ± 0.9%; both p < .001; Figure 4). Parasitism rates were 1.4-fold to 2.8-fold (pumpkin system) and 1.3-fold to 4.3-fold (columbine system) higher than the control depending on the assessment date. The direction and distance of the cabbage plants to the banker plants did not affect these differences between the treatments (both p > .05). However, pumpkin as banker plant led to higher whitefly parasitism rates on cabbage in 1.5 m distance (24.4 ± 1.8%) than in 4 m distance (21.0 ± 1.7%; p = .001) and on the downwind side (26.1 ± 1.9%) compared to the upwind side (19.3 ± 1.5%; p < .001). No differences in terms of whitefly parasitism rate neither between the two distances nor between the two directions were determined in European columbine or control treatment (all p > .05). Finally, A. proletella infestations differed between treatments (χ 2 (2, N = 1,008) = 9.99, p < .01). The pumpkin system (on average 9.26 ± 0.74 puparia per leaf) decreased A. proletella numbers by 4.4%-25.8% depending on assessment date (on average 17.3%) compared to the control (on average 11.19 ± 0.87 puparia per leaf; p = .005; Figure 4). There were no other differences between treatments, distances or directions in terms of whitefly infestation (all p > .05). Harvested pumpkins from the pumpkin system yielded on average 321 ± 12 dt/ha.

| D ISCUSS I ON
This study evaluated the pumpkin system as the most promising annual banker plant system against A. proletella under field conditions.
It produced more parasitoids than other banker plants, facilitated populations of parasitoids and syrphid larvae on cabbage, increased A. proletella parasitism rates and finally decreased infestation by A.

proletella.
The marketability of pumpkin is another economic advantage over the other evaluated banker plant systems. Pumpkin plants were able to tolerate the deliberate infestation with T. vaporariorum. The yield of the pumpkin system in this study (321 ± 12 dt/ ha) was comparable with the yield reported in literature for uninfested plants of the same Hokkaido squash variety 'Uchiki Kuri', which ranges between 300 and 325 dt/ha (Hirthe & Heinze, 2007).
Same was the case for cantaloupe melon production by cantaloupe banker plants preinfested with B. argentifolii and Er. hayati (Goolsby & Ciomperlik, 1999). Exploiting pumpkin as banker plants may therefore lead to multiple economic and ecological benefits for producers and environment.
Wind and distance affected A. proletella parasitism rates only with the pumpkin system, but not with the other treatments. A reason may be that the pumpkin plants were taller and therefore more wind exposed than the European columbine plants. Thus, E. tricolor adults on the pumpkin leaves were more likely to be wind spread since small flying insects like whitefly parasitoids mainly show a windborne dispersal (Kristensen et al., 2013;Ludwig et al., 2018;2019).
This suggests an installation of the pumpkin system upwind from the cabbage crop to achieve maximum parasitism rates of A. proletella. However, this study does not reveal the distance limit of the pumpkin system, because the parasitism rates were still enhanced at the maximum investigated distance of 4 m from the banker plants.
More research is needed to determine the most effective distance between strips of banker plants as well as the optimal strip size (ratio banker plants: cabbage crop). Future studies also need to investigate the optimal ratio between uninfested and preinfested pumpkin plants taking into account pest control services on the one hand and production costs for banker plants on the other hand.
A commercial production of pumpkin banker plants needs to ensure high-quality and standardized products. There are several options to optimize the production process. For instance, increasing the initial parasitism on planted banker plants certainly will further enhance parasitism rates of A. proletella especially at the beginning of the growing season ( Figure 3). An increased parasitism on banker plants may also lead to the production of more males, which are necessary for a long-term maintenance of a stable and effective E. tricolor population. Another option is the use of larger whitefly hosts like A. proletella or A. lonicerae in the rearing of E.
tricolor. Larger hosts will lead to larger and fitter E. tricolor females that deposit their eggs on the banker plants during production ( Figure 2; Hora et al., 1995;Luo & Liu, 2011;Williams, 1995). This F I G U R E 4 Parasitism rate of Aleyrodes proletella and number of A. proletella nymphs (N3/N4) on cabbage plants (mean ± SE) with (columbine system, pumpkin system) and without banker plants (control). Different superscript letters indicate significant differences between treatments in terms of parasitism rate (lower-case letters) and whitefly infestation (upper-case letters; p ≤ .05; glmer, binomial and negative binomial distribution, respectively) may increase the quantity and quality of E. tricolor on the planted pumpkin banker plants which will lead to higher parasitism rates of A. proletella on cabbage (Figure 3).
Another natural enemy as alternative or in addition to E. tricolor in the banker plant system could also improve the impact on A. proletella populations. For instance, a combined use of E. tricolor and whitefly predators like C. arcuatus or syrphid larvae in a banker plant system may lead to additive or synergistic effects (Schultz et al., 2009). Furthermore, the whitefly parasitoid E. inaron may be a potentially better alternative to E. tricolor.
Encarsia inaron used to be highly abundant in at least certain parts of Europe up to the 1950s (Butler, 1938;Stein, 1958), but has almost disappeared as parasitoid of A. proletella since that time (Gumovsky, 2005;Laurenz et al., 2019;Springate, 2017). The displacement of E. inaron may have facilitated the outbreakes of this whitefly pest in recent decades (Williams, 1996), because E. inaron may perform better on primary hosts (159 eggs per female on Siphoninus phillyreae) than E. tricolor (85 eggs per female on A. proletella; Gould et al., 1995;Williams, 1995 (Brady & White, 2012). Reasons may be the relatively fast development to the adult stage or the small size of Trialeurodes and Bemisia nymphs, which may result in insufficient time and limited nutritional resources for E. tricolor to complete its development.
The quantity of alternative hosts/prey seemed to be more important for the success of a banker plant system than the quality.
Aleyrodes lonicerae was a qualitatively better host for E. tricolor than T. vaporariorum in terms of reproduction, development and fitness of offspring (see 3.3 Parasitoid performance). However, the latter developed much higher population sizes than A. lonicerae under field conditions. This high availability of alternative hosts/prey on the pumpkin system let not only to a higher parasitoid production and increased parasitism rates of A. proletella, but also to an increase in syrphid larvae on the cabbage plants and finally to decreased A. proletella populations. These contradicting results between laboratory and field also underline the importance of and the need for more solid field studies.
The impact of general predators like syrphid larvae on A. proletella populations may often been underestimated. In this study, both banker plant systems evaluated in the field increased parasitism rates of A. proletella to a similar extent (Figure 4). However, only the pumpkin system led to higher numbers of syrphid larvae on the cabbage plants as well as a decrease in A. proletella infestation. This suggests that predation by syrphid larvae can have a significant impact on the population size of A. proletella. Therefore, more research is desired to develop and implement respective measures like flower strips, potentially as a combined strategy with banker plants, to further promote syrphids (Laurenz & Meyhöfer, 2016).
Perennial banker plants are another option to permanently increase the local abundance of natural enemies in order to promote biological control services. Therefore, perennial host plants of nonpest whitefly species could be installed in field margins as shelter, overwintering and reproduction habitat for alternative host/prey and natural enemies in the agricultural landscape (Gurr et al., 2017).
Potential candidates are some of the here investigated herbaceous host plants of A. lonicerae (e.g. A. podagraria, G. urbanum, A. vulgaris, F. vesca) or even woody plants like Lonicera spp. for A. lonicerae, Fraxinus spp. for S. phillyreae or Viburnum spp. for Aleurotuba jelinekii (Evans, 2007b;Mound & Halsey, 1978;Pickett & Wall, 2003). More Generally more field studies are needed to confirm the results under different conditions (e.g. climate, A. proletella infestation levels).

ACK N OWLED G EM ENTS
We thank Lisa Hildebrandt, André Brun, Florian Wulf, Timo Michel, Birgit Milde, Serafine Herrmann and Johannes Specht for their support during data collection and crop cultivation. Open access funding enabled and organized by Projekt DEAL.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest and confirm that there are no disputes over the ownership of the data presented and all contributions have been attributed appropriately.

Sebastian Laurenz conducted experiments, analysed data statisti-
cally and wrote the manuscript. Rainer Meyhöfer secured funding.
Both authors conceived research as well as read and approved the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in LUH-Projekt Seafile at https://doi.org/10.25835/ 0017692 (Laurenz & Meyhöfer, 2020).