Behavioral Responses of Bemisia Tabaci Cryptic Species Med to Three Host Plants and Their Volatiles

Bemisia tabaci Gennadius (Hemiptera: Aleyrodidae) is a worldwide pest that damages more than 900 host plant species. The infestation behavior of this pest is affected by the volatile organic compounds (volatiles) of different plants and their growth stage. We investigated the chemical constituents of the volatiles extracted from three plants (Gossypium hirsutum, Abutilon theophrasti and Ricinus communis) at different growth stages (pre-owering, orescence and fruiting) and their effects on the behavior of adult B. tabaci. The selectivity studies on three plants showed that the B. tabaci preferred piemarker. The olfactometer studies showed that growth periods of the three plants also affected the preference of B. tabaci. Volatiles of piemarker and cotton plant had different levels of attraction to adults during all stages. Volatile substances released by castor at stage of owering have a repellent effect on B. tabaci. In the plant VS plant combination ,the adults showed the strongest preference to volatiles from before and during anthesis of piemarker, followed by cotton, and then castor. A total of 23, 22 and 18 compounds were detected from volatiles of piemarker, cotton and castor, respectively, and proportions among the compounds changed during different stages of plant development. The olfactory responses of B. tabaci to volatile compounds showed that linalool and high concentration of leaf acetate had strong trapping effect on this pest, while 1-nonanal had signicant repellent effect at high concentration. This study indicates that different plants and their growth stage affects their attractiveness or repellency to B. tabaci adults which is mediated by plant constitutive and dynamic changes. The compounds obtained by analysis screening can be used as potential attractants or repellents to control Mediterranean (MED) B. tabaci.


Introduction
Bemisia tabaci Gennadius (Hemiptera: Aleyrodidae) is a worldwide pest which is known to damage more than 900 host species (Lee 2020). It was rst observed on tobacco in Greece in 1889 (Brown et al. 1995;Gennadius 1889). With the introduction of Euphorbia pulcherrima in China beginning at the end of last century, B. tabaci has gradually become an important agricultural pest in local area. It is a complex population with many cryptic species (at least MEAM1 (Middle East-Asia Minor I) and MED (Mediterranean), formerly biotypes B and Q, respectively) (Dickey et al. 2013). Although chemical control is still the main method for against B. tabaci currently, its widespread application has brought negative impacts on the environment and potential threat to human health (Liu et al. 2016). Therefore, many scienti c researchers now focus on how to use natural enemies, microorganisms and semiochemicals from plants to control this pest.
In the interaction between plants and insects, the latter mainly rely on olfaction to sense extrinsic environment, and the plant volatiles provide important clues for insects to locate hosts for food or oviposition (Bernays and Chapman 1994;Liu et al. 2019). The olfactory system of insects has high sensitivity and speci city for plant volatiles (de Bruyne and Baker 2008). Previous study has reported that the preference of herbivorous insects to host plants is due to the presence of certain one or several attractive components in the host plant volatiles or the absence of repellent components (Li et al. 2002) .
Piemarker, Abutilon theophrasti, as a Malvaceae family plant, is generally more attractive to pests than cotton (Chen et al. 2020;Liu et al. 2016). Castor Ricinus communis, containing ricin, its extract is mainly used to insecticide material around the world (Hua et al. 2013). However, recent studies and our eld research have shown that plants attract or repel insects only at a certain growth stage (Knolhoff and Heckel 2014;Luo et al. 2018;Seiter et al. 2013). This may be related to the change of volatiles at different growth stages, but few studies have been reported. Therefore, it is necessary to study the host localization behavior of MED B. tabaci and its chemical mechanism in order to provide theoretical basis for the application of behavioral control methods to control this pest, to reduce the economic losses caused by it. The purpose of this study was to select the companion plants and their optimal growth period with regulatory effects on B. tabaci, in order to establish an accurate push-pull strategy for B. tabaci on cotton.
In this study, we aimed to determine (i) whether cotton Gossypium hirsutum, piemarker and castor had a differential attractiveness to adult B. tabaci using leaf disc test and free-choice assays, and (ii) whether different growth stage of those three plants had a differential attractiveness to adult B. tabaci using olfactometer, and (iii) volatiles of the three host plants at different stage are collected and analyzed, as well as (iv) identify chemical volatiles which attract or repel the insect.

Plant growth
Cotton plant, Gossypium hirsutum SGK321 seeds were obtained from the Institute of Plant Protection, Chinese Academy of Agricultural Sciences. Piemarker plant Abutilon theophrasti seeds were from Wang Zheng seedling sales center of China and castor plant Ricinus communis seeds were from Shouguang wentian seed industry co., ltd. They were grown in pots (size:13cm in diameter) in a 2:1 mixture of seedling substrate and soil. All plants were nurtured in a greenhouse and tested when they reached ve true leaves. All plants were watered weekly. They were maintained at 28 ± 1 • C with 65-75% RH and a 16:8 (L:D) at a light intensity of 1400-1725 lux, no pesticides were applied to the pants before or during the experiments.

Insect rearing
The B. tabaci (cryptic species MED), were mass reared on cotton plant in mesh cages (60cm×60cm ×100cm). They were maintained at 28 ± 1 • C with 65-75% RH and a 16:8 h (L:D). The B. tabaci was regularly identi ed by mt DNA Co I gene sequence.

Free-choice test in cages
The preference test of B. tabaci adults were determined on different plants using the free-choice test experiment. Three pots with cotton, piemarker and castor plants were arranged randomly in each iron framed cage (60cm×60cm×100cm) covered with the ne mesh nylon net to keep white ies from escape.
Twenty adults were introduced in the cage. Five replicates per treatment were set. To observe the preference study, 20 adults were introduced in the cage uniformly. In order to ensure the free movement of white ies, ve paper plates with a diameter of 10cm and a height of 0.5cm were hung on the top of the cage, which were divided into ve equidistant sites: east, south, west, north and middle. The numbers of adults on each plant were recorded at 1 h, 2 h, and 24 h after release (Liu et al. 2020;Zhou et al. 2008).

Leaf disc test
Referring to the method of Zhou et al. (2008), a lter paper sprayed with water was placed at the bottom of the glass petri dish (16cm diameter). The leaves of the plants under test were cut into round leaves with a hole puncher (1cm diameter). Two circular plates were taken from each plant and placed at different intervals in a petri dish. Fifty B. tabaci which were starved for 4 hours already released into a petri dish, and host selection of B. tobaci was observed 2 hours later. The experiment was repeated for 5 times. In order to reduce the in uence of the light source on the test, the device is placed in separate environmental chambers with the 10000 lux intensity top light source at 28±3℃, 70%±5% RH.

Olfactory choice test with Y-Tube olfactometer
The behavioral response of B. tabaci to volatile compounds of three plants was evaluate by a Y-tube olfactometer bioassays, following the equipments and procedure as previously described by Akol et al. (2010) and Saad et al. (2015) with slightly modi cations. Y-tube via odor sources using a QC-3 atmospheric sampler (0.2-3L/min) (Beijing Municipal Institute of Labour Protection, China) for circulating the volatile organic compounds. The air ow was adjusted and measured by a glass rotameter.
The test was carried out in a dark room so that adults cannot receive visual cues from the plants. A 300 lux LED light was placed 0.5 m vertically above the Y-tube olfactory meter. The adult white ies were released within 1.0 cm base of the Y-tube for after 4 h starvation and their responses assessed for 3 min.
If it cross 1/3 of any of the two arms of the Y-tube branch and stayed a least of 30 seconds and did not return was considered as a positive-responsive individual. Otherwise, they were considered as nonresponsive insect. For each 10 adults detected, the two lateral branches of the Y-tube were inverted 180•, which was a repeat. Each olfactory test was repeated 5 times, and a total of 50 adults were tested. The test time is arranged from 8:00 to 18:00. During the test, the temperature of the darkroom used for the test is kept at 26℃and the humidity is about 65%.

Volatile compounds collection and analysis
Different plants released dynamic headspace volatiles were collected at pre-owering, orescence and fruiting. Similarly, the pot of plant was removed and the roots with soil were packed in tin foil. The plant was transferred individually to 10 L cylindrical glass chambers. Before trapping volatile, the activated charcoal-ltered air entered the chamber at 1.0 L/min with a vacuum pump for more than 30 min. Filtered out the air of the system and then the kept plant in the glass container for 1 h before trapping volatiles. During the volatile collection, activated charcoal puri ed air was pumped through Te on tubing into the system at a ow rate of 1 L/min through a glass tube lled 200 mg, 60-80 mesh Tenax TA adsorption column (Nazrul et al. 2017). The samples were cyclic collection for 4h and collected 4 plants (replicates) per treatment.
The trapped volatiles were extracted from the adsorbent tube using n-hexane (Sinopharm Chemical Reagent Co. Ltd., Shanghai, China) in 4 times, 100μL each time, and 7.03ug/mL n-octane (Aladdin Co. Ltd., Shanghai, China) was added to piemarker and cotton sample as an internal standard, and 5.87ug/mL methyl salicylate (Aladdin Co. Ltd., Shanghai, China) was added to castor sample as an internal standard. Samples were stored at -20 °C until they were analyzed through GC-coupled mass spectrometry (GC-MS).
For GC-MS, Trace ISQ (Thermo Fisher, USA) for detection, identi cation, and quanti cation of plant volatile components. The chromatographic column was DB-5 MS column (30 m×0.25 mm×0.25μm). A helium (99.999%) was used as a carrier gas with a ow rate of 1.0 ml min-1 with constant mode. Heating procedure: The column temperature was maintained to 50•C for 1 min, the oven temperature was increased from 50 to 150•C at a rate of 5•C min-1 and held for 2 min, and then from150 to 250•C at a rate of 10•C min-1 and held for 2 min.

The olfactory responses of B. tabaci to volatile compound
The volatiles compounds with signi cant changes in the of growth period of three plants and their standard samples was selected (Linalool, CAS: 78-70-6, Purity 95%; leaf acetate, CAS: 3681-71-8, Purity 98%; and 1-nonanal, CAS: 124-19-6, Purity 96%. All purchased from Aladdin Co. Ltd., Shanghai, China.) were diluted with n-hexane solution of 100, 10 and 1 (μL/mL), respectively. 20 μL of single volatile solution was accurately absorbed and dropped on 1.5cm × 2cm qualitative lter paper, and put into the bottle connected by one arm of the "Y" shaped tube. A qualitative lter paper dripping with 20 μL nhexane was put into the bottle on the other arm as a control. The test method is the same as olfactory selection test.

Statistical analyses
IBM SPSS 25.0 as used to conduct all statistical analyses. Among copy numbers (4 replicates), ANOVA with Tukey HSD post hoc multiple pairwise statistical comparison and Kruskale Wallis test (nonparametric test) was used for analyzing the proportion of adults per plant in free-choise test and leaf disc test..For the Y-tube olfactory test, the hypothesis was that the pest showed no preference (i.e., a 25:25 response) for each arm. Data produced from Y-tube olfactometer choice bioassays were analyzed by c 2 test. Those who did not make a choice on host plants were not included in the selection rate.
The compounds were identi ed by comparing the fragmentation patterns from the mass spectra with the databases of the library NIST 2014 (National Institute of Standards and Technology, Washington, DC, USA) database. Exploration of the GC data collected from different stage of piemarker, cotton and castor samples were preliminarily conducted by principal component analysis (PCA). Classi cation models using partial least square-discriminant analysis (PLS-DA) were generated and validated. Class modelling of three plants were nally performed by Soft Independent Model Class Analogy (SIMCA) respectively, to test the differences in the plant volatiles of different growth stage (Deconinck et al. 2018).

Results
Preference of B. tabaci to three plants In free-choice assay, B. tabaci adults showed a signi cant difference in number among different plant treatments (Fig.1). In general, piemarker plant attracted a signi cantly higher number of B. tabaci adults than cotton and castor after 1 h (F=11.165, P=0.01), 2h (F=23.681, P=0.001) and 24h (F=44.432, P 0.0001) release, and no statistical difference between castor and cotton. Post-release time in uenced the number of adults on different host plants. The number of B. tabaci on piemarker increased with the time, and increased from 50.5 % to 60.6% after 24 h release, while the number of white ies on castor decreased from 21.1% to 12.4% after 24 h release. However, stable B. tabaci population dynamics were detected in cotton, and the percentage of this pests on the host after 24, 48 and 72 h were 28.4%, 28.7% and 27.0%, respectively. The results of petri dish test are consistent with that of free-choice test. The selectivity of B. tabaci to piemarker (66.61%) was the highest, signi cantly higher than that of cotton (22.39%) and castor (11.00%) (Fig.2).

Response of B. tabaci to pre-owering volatiles of three plant
The statistical analysis indicated respond preferentially of B. tabaci to volatiles from piemarker and cotton at pre-owering stage compared to air. The attraction of piemarker and cotton to adults were the most signi cant (67.6%, c 2 =11.7, P 0.001) and extremely signi cant (64.6%, c 2 =7.96, P 0.01) respectively, than that of air ow. However, there was no obvious reaction between caster at pre-owering stage(46.4% c 2 =0.38, P 0.05) and air ow chosed by adults (Fig.3 a).

Response of B. tabaci to orescence volatiles of three plant
During owering, the attraction of piemarker and cotton to B. tabaci were signi cantly greater than that of air ow, but castor was more resistant to them. Compared with air, The selection rate of B. tabaci on piemarker was 63.3% (c 2 =6.56, P < 0.05), cotton 61.7% (c 2 =5, P < 0.05) and castor 34.7% (c 2 =8.76, P < 0.01) (Fig.3 b).
The B. tabaci responded preferentially to the volatiles emitted by piemarker during orescence stage in combination of plant vs plant. Between the combinations of piemarker and castor, the selection rate of B.
tabaci for the former and the latter was 64.1% and 35.9% (c 2 =7.36 P 0.01), respectively. When the combination of piemarker and cotton was selected, 64.5% of the adults chose piemarker and 35.5% chose cotton (c 2 =7.84 P 0.01). There were also signi cant differences between cotton and castor. Among them, 63.3% of B. tabaci responded favourably to volatiles from cotton, while only 36.7 % to castor (c 2 =6.56 P 0.05) (Fig.3 b).

Response of B. tabaci to fruiting volatiles of three plant
The B. tabaci responded preferentially to volatiles from piemarker and cotton when compared with to air, respectively, but no signi cant difference between castor and air (Fig.4 a). Compared with air, the selection rate of B. tabaci was 61.1% for piemarker (c 2 =4.5, P 0.05), 62.0% for cotton (c 2 =5.3, P 0.05) and 52.6% for castor (c 2 =4.5, P 0.05).

Qualitative analysis of three plant volatiles in different periods
The results showed that more than 30 different compounds were detected, including alcohols, aldehydes, esters, terpenes and other compounds (Fig.5). In total, 23 volatiles were identi ed from the piemarker plants, with 21 detected before owering, 22 at owering, and 19 at fruiting. Twenty-two compounds were tentatively identi ed from three growth stage volatiles released by cotton, at the pre-owering, orescence and fruiting volatiles compounds was 19, 21 and 19 kinds, respectively. A total of 18 compounds were detected in castor during the three stages, including 14 at pre-owering, 17 at owering and 14 at fruiting stage (Fig.5).

PLS-DA analysis of different periods volatiles of three plant
The partial least squares-discriminant analysis (PLS-DA) showed that the plant volatiles contents were clearly separated among pre-owering, orescence and fruiting of piemarker. The rst two signi cant PLS components explained 43.3% and 47.1% of the total variance, respectively (Fig. 6 a). In this model, the following seven volatile compounds linalool, unknow, ethyl caprylate, ethyl nonylate, butyl acrylate, 1,3,7ocimene, 3-hexadecanol with VIP values ≥1.0 contributed most to the separation among different periods volatiles of piemarker (Fig. 6 b).
The PLS-DA also showed a clear separation among pre-owering, orescence and fruiting of cotton. The rst two signi cant PLS components explained 46.3% and 49% of the total variance, respectively (Fig.7  a). The rst component and the second component showed a clear separation among three volatiles of different periods of cotton. In this model, the following nine volatile compounds leaf acetate, unknow, ethyl caprylate, naphthalene, ethyl nonanoate, nonane, 1,3-xylene, dodecyl aldehyde, octanal with VIP values ≥1.0 contributed most to the separation among different periods volatiles of cotton (Fig.7 b) The PLS-DA showed that the rst component and the second component showed a clear separation among pre-owering, orescence and fruiting volatiles of castor. The rst two signi cant PLS components explained 43.5% and 44.5% of the total variance, respectively (Fig. 8 a). In this model, the following six volatile compounds 1-nonanal, ethyl caprate, 2-butyl-1-octanol, butyl acrylate, 3hexadecanol, naphthalene with VIP values ≥1.0 contributed most to the separation among different periods volatiles of castor (Fig. 8 b) The olfactory responses of B. tabaci to standard samples of volatile compound The Figure (4 b) showed that among the three compounds, the B. tabaci responded clear preference to linalool with the concentration of 1, 10 and 100μL/mL, and the selection rates were 64.4% (c 2 =7.72, P < 0.01), 61.2% (c 2 =4.56, P < 0.05) and 60.4% (c 2 =3.92, P < 0.05), respectively. In addition, leaf acetate was also signi cantly attractive to this pest at 100μL/mL, and the selection rate was 60.9% (c 2 =4.32, P < 0.05). However, there was no signi cant difference between 1 and 10μL/mL (c 2 =0.02, P > 0.05; c 2 =0.28, P > 0.05). At concentration of 100μL/mL, 1-nonanal showed the obvious repellant against B. tabaci, the selection rate only 32.5% (c 2 =11.56, P < 0.001). And there was no signi cantly preference of adults to medium and low concentrations (c 2 =1.87, P > 0.05; c 2 =0.02, P > 0.05).

Discussion
In this study we investigated the response of B. tabaci to chemical cues released by different plants at their different growth stages. Results showed that the piemarker plants attracted B. tabaci adults, while the castor repelled against them. However, attraction/avoidance only occurred during the certain growth period of plants, which was closely related to the release of volatiles by plants.
It was found that B. tabaci showed a strong preference to piemarker in the free-choice test. Moreover, the selection rate of adults increased with time increase, and adults showed obvious host selectivity in the process of diffusion. In the leaf disc test, the selection rate of B. tabaci to piemarker reached 66.61%, which was consistent with the results of free-choice test. The B. tabaci showed obvious preference for piemarker in previous behavioral studies. For instance, Lin et al. (2006) found that B. tabaci signi cantly preferred piemarker in the piemarker vs air treatment with a selection rate of 68.1%, but showed no obvious preference when compared with cucumber or tobacco. Wang et al. (2016) also found that piemarker and cotton had strong attractiveness to B. tabaci among 13 host plants.
Our study indicates the physiological stage of the host plant in uences the host-foraging behavior of the white y. The "Y" olfactometer test demonstrated that in the plant VS air ow treatment, compared with cotton, piemarker is more attractive to B. tabaci in the pre-owering stage. In the plant VS plant treatment, at pre-owering and owering stage, most adults chose piemarker, followed by cotton, and then castor. but the preference of B. tabaci to the three plants was not signi cantly different during fruiting stage which suggests that the piemarker, as an attractor of pests, was more attractive at the pre-owering and owering stages. Mohammed et al. (2021) studied the in uences of the volatiles from different parts of brinjal plants on the behavior of adult Leucinodes orbonalis, and found that adults responded differently to the volatiles extracted from fruits, leaves, shoots and owers. In addition, we detected the castor had a repellent effect on B. tabaci at the owering stage. In the castor VS air ow treatment, only 35.9% of adults chose castor at the owering stage while there was no signi cant difference in avoidance (preference) of B. tabaci at pre-owering stage and fruiting stage. Luo et al. (2018) investigated the population of B.tabaci in the castor trap belt, its neighboring cotton elds and castor-free cotton elds, and found that there was no signi cant difference in the number of B.tabaci on the three areas between June and July. However, in August and September, the number of white ies on castor and cotton adjacent to castor was 1205 and 1580 per 100 plants, respectively, signi cantly lower than that of control cotton (4697 per 100 plants). It also proved that castor had repellent effects on B. tabaci at a certain period.
The change of composition and content of volatiles at different growth stages of plants may be responsible for the host-seeking behavior of white ies. The volatiles play a key role in the process of host recognition, which is an important clue for insects to identify and locate food and natural enemies (Renou et al. 2019). In this study, more than 30 different compounds were detected, including alcohols, aldehydes, esters and terpenes (23 in piemarker, 22 in cotton, and 18 in castor). This is almost consistent with previous reports (Chen 2013;Rizwangul 2018;Zhang et al. 2016). In addition, the proportion of some volatile compounds in the three plants changed signi cantly at different growth stages, which may be the reason for the difference in preference for B. tabaci at different growth stages.
The olfactory responses of B. tabaci to volatile compounds such as linalool, leaf acetate and 1-nonanal were determined. It found that linalool and high concentration of linalool acetate had strong trapping effects on cryptic species MED B. tabaci, while 1-nonanal had signi cant repellent effects at high concentration. The result indicates that concentration of the compounds is also an important factor affecting host selection. Therefore, it indicates that linalool, leaf acetate and 1-nonanal can be used as potential attractants or repellents to control B. tabaci.
In general, our research proves that different plants and their growth stages affect the attraction (repellent) to B. tabaci adults and that the attraction (repellent) cue is the volatiles emitted from preowering, orescence or fruiting plants. The three volatile compounds (linalool, leaf acetate and 1nonanal) could be used as potential attractants or repellents in future.   Attracting effects of different host plants on Bemisia tabaci at pre-owering(a) and owering(b) stage. The data in the histogram represents the percentage of individuals who acted in each treatment. *Signi cant difference (P 0.05), **extremely signi cant difference (P 0.01), *** most signi cant difference (P 0.001) and ns shows no signi cant difference(P 0.05), c 2 test.

Figure 4
Behavioral response of Bemisia tabaci to different host plants at fruiting stage(a), and to a single volatile componen(b). The data in the histogram represents the percentage of individuals who acted in each treatment. *Signi cant difference (P 0.05), and ns shows no Signi cant difference(P 0.05), c 2 test.