Foraging behaviour of Scymnus syriacus (Coleoptera: Coccinellidae) provided with Aphis spiraecola and Aphis gossypii (Hemiptera: Aphididae) as prey: Functional response and prey preference

Scymnus syriacus Marseul is one of the most important biocontrol agents of aphids in citrus orchards. Understanding the effi ciency and host specifi city of natural enemies can help improve their effectiveness as biological control agents of particular pests. In this study, the functional response, preference and switching behaviour of different stages of S. syriacus feeding on Aphis spiraecola Patch and Aphis gossypii Glover, on orange leaves, were studied under laboratory conditions of 27 ± 2°C, 65 ± 5% RH and a photoperiod of 16L : 8D. The results indicate that 4th instar larvae and adult males and females of this predator exhibit a type II functional response when offered different numbers (3, 5, 7, 10, 20, 30, 40, 60 and 80) of both A.spiraecola and A. gossypii. Adult females had signifi cantly higher attack rates (a) on both species of aphids. Handling times (Th) of adult male predators on A. spiraecola and those of 4th instar larvae on A. gossypii were shorter than that of the other stages. In no-choice preference experiments, adult females fed equally on both species of aphid, but the 4th instar larvae preferred A. gossypii. In switching experiments, prey selection was evaluated by simultaneously presenting different ratios (30 : 50, 40 : 40 and 50 : 30) of the two aphids, A. spiraecola : A. gossypii, to the predator. The 4th instar larvae and adult females of S. syriacus showed higher preference for A. gossypii at each ratio offered (except adult females offered the 50 : 30 ratio). Based on their foraging behaviour 4th instar larvae and adult females of S. syriacus are suitable for use in integrated management of citrus aphids.


INTRODUCTION
Citrus is a semi-tropical fruit crop with global availability and a popular component of human diets, which is grown in more than 140 countries (Liu et al., 2012). The most common citrus fruit with commercial importance are oranges, lemons, limes, grapefruit and tangerines (Liu et al., 2012;Loghmanpour Zarini et al., 2013). There are many species of aphids that can cause economic damage in citrus orchards (Rajabi, 1986). The citrus aphid, Aphis spiraecola Patch and the melon aphid, Aphis gossypii Glover (Hemiptera: Aphididae) are among the aphids that attack citrus trees, especially in the north of Iran (Soroushmehr et al., 2008;Tena & Garcia-Mari, 2011;Alizadeh Kafeshani et al., 2018). Adults and immatures of these aphids can cause direct damage, especially to young citrus plantations by feeding on phloem sap. During feeding they inject saliva, which is phytotoxic and adversely affects the physiology of the trees. Furthermore, growth of sooty mould on aphid honeydew on leaves hinders their photosynthesis (Blackman & Eastop, 2000;Garzo et al., 2002). Both of these species of aphids also cause indirect damage by transmitting important viruses (Blackman & Eastop, 2000), such as The stock culture of the predator was maintained on orange leaves (C. sine nsis cv Thomson) infested with A. spiraecola or A. gossypii in plastic containers (9 × 10 × 21 cm) for three generations. In experiments that involved S. syriacus and one aphid species (functional response and no choice experiments), S. syriacus was reared on that aphid species. In experiments that involved S. syriacus and both aphid species (switching and choice experiments), the adaptation of S. syriacus to one species was avoided by rearing it on a mixed diet of an equal number of both species of aphids.
The exp eriments were conducted in Petri dishes (9 cm in diameter). Each dish had a hole (2 cm in diameter) in the lid, which was covered by fi ne-mesh net for ventilation and its bottom lined with a layer of solidifi ed agar solution (2%) as a substrate to prevent orange leaves (C. sinensis cv. Thomson; 16.5 cm 2 ) from desiccating. The Petri dishes were sealed with parafi lm to prevent the escape of insects. All insect cultures and experiments were carried out in a growth chamber at 27 ± 2°C, 65 ± 5% RH and a photoperiod of 16L: 8D.

Functional response experiment
This experiment was conducted with 4 th instar larvae and adults (males and females less th an 24 h old) of S. syri acus and a mixture of 3 rd and 4 th instar nymphs of A. spiraecola or A. gossypii. The predators were starved for 24 h prior to the experiments by placing them individually in Petri dishes devoid of prey. Different numbers (3, 5, 7, 10, 20, 30, 40, 60, and 80) of aphids were separately offered to the different stages of the predator. After 24 h, the number of aphids eaten by the predator was recorded. For each prey density and predator stage, the exp eriment was replicated 10 times without prey replacement.

Prey preference
In this study, we preferentially selected the 4 th instar larvae and adult females of S. syriacus to study their preference and switching between A. spiraecola and A. gossypii, because they are more voracious than adult males (Omkar & Kumar, 2013).

No-choice experiment
In the no-choice preference assay, 1-day-old 4 th instar larvae and females of S. syriacus, which were starved for 24 h, were used. A mixtu re of eighty 3 rd and 4 th instar nymphs of A. spiraecola or A. gossypii were exposed to 4 th instar larvae and adult females of S. syriacus. After 24 h, predators were removed from the arenas and the number of prey consumed was recorded. The experiment was replicated 10 times.
Independent t-test was used to test the signifi cant difference between the predation rates of the predator of two species of prey (SPSS, 2007).

Choice experiment and switching
Switching behaviour of 1-day-old 4 th instar larvae and adult females of S. syriacus was investigated at different densities of A. spiraecola or A. gossypii (mixture of 3 rd and 4 th instar nymphs). The predators were starved for 24 h before the test. A. spir aecola : A. gossypii ratios presented to the predators were 30 : 50, 40 : 40 and 50 : 30. After 24 h, each predator was removed from the arena and the number of prey consumed was recorded. There were 10 replicates for each ratio of prey and developmental stage of predator.

Statistical analyses
Statistical analysis of functional responses followed Juliano's (2001) approach. First, the type of functional response was determined using a logistic regression of the proportion of prey eaten For the successful biological control of pests, it is necessary to understand the foraging behaviour of the natural enemies in terms of their functional response, prey preference and switching between prey (Heidarian et al., 2012;Rezaei et al., 2019). Functional response is the number of prey eaten per predator as a function of prey density (Jervis & Kidd, 1996). Holling (1959) categorized the functional response into three different types: a linear relationship between prey density and the number of prey consumed (type I), a curvilinear relationship (type II), and a sigmoidal relationship (type III). Other types of functional response, such as a dome-shaped response (type IV) and negative exponential response (type V) were reported by other researchers (Luck, 1985;Sabelis, 1992). Among these fi ve types of functional response, type II and type III have received the most attention. Two important parameters: attack rate (a) and handling time (T h ) can be evaluated using functional response models. Natural enemies with high attack rate (a) and low handling time (T h ) are considered to be the most effective biocontrol agents (Bayoumy, 2011;Tazerouni et al., 2016).
Generalist predators may show prey preferences when foraging in patches containing more than one species of prey patches (Jaworski et al., 2013). Knowledge of the prey preferences of a predator is important for determining its predatory effi ciency in situations in which various pests are present (Enkegaard et al., 2001). The tendency of a predator to choose one prey species over another may change as the relative densities of the species change (Murdoch, 1969). This behavioural trait is known as switching. Switching has a crucial role in extending the persistence of predator-prey systems (Van Baalen et al., 2001). In positive switching, a predator's preference for a specifi c prey increases as a function of the abundance of the prey, which could infl uence the stability of the prey populations. In negative switching, the predator's preference for a prey declines as a function of the abundance of this prey, and may destabilize trophic interactions by increasing the mortality of the species of prey whose density is decreasing (Murdoch, 1969;Chesson, 1984;van Baalen et al., 2001). Not all predators show switching behaviour (Chesson, 1984).
Since different types of prey have a critical effect on the foraging behaviour of predators, the present study was designed to evaluate the interactions of S. syriacus with two species of aphids, A. spiraecola and A. gossypii. Thus, in the current research the functional response, prey preference and switching of S. syriacus between A. spiraecola and A. gossypii reared on the leaves of orange, Ci trus sinensis L., were investigated in order to assess the potential efficiency of this predator for controlling the abundance of aphids on citrus trees as a part of an IPM program.

MAT ERIALS AND METHODS
The initial populations of S. syriacus and the two species of prey, A. gossypii and A. spiraecola, were collected from unsprayed citrus orc hards near Gaemshahr (36°27´47˝N, 52°51´36˝E), Mazandaran province, Iran. The aphids were reared on the buds of young orange trees for several generations.
(N e ) as a function of prey offered (N 0 ). This is done by fitting a polynomial function: where P 0 , P 1 , P 2 and P 3 are the intercept, linear, quadratic and cubic coeffi cients, respectively. These coeffi cients were estimated using the method of maximum likelihood. If P 1 < 0, the proportion of prey eaten declines monotonically with the initial number of prey offered, which indicates a type II functional response (Juliano, 2001). If P 1 > 0 and P 2 < 0, the proportion of prey eaten is initially positively dependent on the density and consequently indicates a type III functional response. After determining the type of functional response, the handling time (T h ) and attack rate (a) were calculated using Rogers random predator equation (Rogers, 1972) as follows: where N e is the number of prey eaten, N 0 is the initial prey density, a is the attack rate, T h is the handling time and T is the duration of the experiment. Pairwise comparisons of parameters of the functional responses for different larval instars or type of prey were performed using the indicator variable method (Juliano, 2001) as follows: where j is an indicator variable that has a value of 0 for the fi rst data set and 1 for the second data set. The parameters D a and D Th estimate the differences between the data sets being compared for the values of the parameters a and T h , respectively. In other words, the handling t ime for one stage is T h , and th at for another stage is T h + D Th . Testing for a signifi cant difference in handling times between two stages is accomplished by testing the null hypothesis that D Th includes 0 (Juliano, 2001).
The maximum predation rate (T/T h ), which is the maximum number of prey that can be consumed by a predator during 24 h, was calculated using the estimated T h (Hassell, 2000).
Prey preferences exhibited by 4 th instar larvae and adult females of S. syriacus were analysed by calculating Manly's index of preference (Manly, 1974). The index was calculated using the following formula: where β 1 (Beta = Man ly's preference index) is the preference for a particular prey, e 1 and e 2 are the number of type 1 and 2 prey remaining at the end of the experiment, respectively, and A 1 and A 2 are the number of type 1 and 2 prey offered, respectively. The value of the index falls between 0 and 1. An index close to 1 indicates strong preference for prey type 1 and a value equal to 0.5 indicates that the predator selects prey randomly and does not show any preference. Paired t-test was used to test the significance of Manly's index (SPSS, 2007). It is appropriate to use this index for assays in which prey species are offered simultaneously and are not replaced during the experiment (Manly et al., 1972). To compare the number of prey eaten by each stage of predator at the mentioned ratios, one-way ANOVA followed by Tukey (p < 0.05) test was used (SPSS, 2007). Paired t-test was used to compare the number of prey eaten by each stage of predator in each combination. To test the signifi cant difference between observed (number of prey eaten) and expected values for each prey ratio, Chi-square test was used. Furthermore, to evaluate the capacity of S. syriacus to switch between A. spiraecola and A. gossypii, the β-values were fitted to Manly's linear model for frequency dependent selection (Manly et al., 1972;Manly, 1973) as well as to non-linear models.

Functional respon se
The results of lo gistic regression analy ses of 4 th instar larvae and adult males and females of S. syriacus to di fferent numbers of A. spiraecola and A. gossypii are presented in Table 1. In all cases, the linear coeffi cient was negative (P 1 < 0), indicating a type II functional response. The results revealed a decline in the proportion of the number of each prey consumed by S. syriacus of th e initial prey density (N e / N 0 ) with increase in prey density (Fig. 1).
The attack rate ( a) and handling time (T h ) of 4 th instar larvae and adult males and females of S. syriacus were different for the two species of prey. Attack rate (a) of the different stages of S. syriacus fed on A. spiraec ola varied from 0.0525 to 0.0925 h -1 and the estimated handling time (T h ) ranged from 0.4165 to 0.4985 h (Table 2). Based on the asymptotic 95% confi dence interval for D a , there were signifi cant differences between the attack rate (a) of adult males and adult females, and of adult females and 4 th instar larvae (Table 3). It shows that the attack rate (a) of adult females of S. syriacus on A. spiraecola is higher than that of 4 th instar larvae and adult males (Table 3). Furthermore, the handling time (T h ) of adult males on A. spiraecola was signifi cantly shorter than that of adult females (Table 3).
Attack rates (a) of different stages of S. syriacus fed on A. gossypii ranged from 0.0328 to 0.1242 h -1 and their handling time (T h ) varied from 0.2365 to 0.5362 h ( Table 2). The as ymptotic 95 % confi dence interval for D a and D Th in pairwise comparison of various stages of S. syriacus fed on A. gossypii did not include 0, which means that there were signifi cant differences between the attack rate (a) and handling time (T h ) of various stages of S. syriacus (Tabl e 3). Based on the results, adult females and 4 th instar larvae of S. syriacus had the highest and lowest attack rate (a) on A. gossypii, respectively (Table 2). This trend also occurred in the values of handling time (T h ) ( Table 2).
Based on the results of pairwise comparisons, there were signifi cant differences between the attack rate (a) of each stage of S. syriacus on the two types of prey (Table 3). The attack rates of 4 th instar larvae were highest on A. spiraecola , wh ereas those of adult males and adult females were highest on A. gossypii (Tabl e 2). Furthermore, the asymptotic 95% confi dence interval for D Th of 4 th instar larvae on the two types of prey did not include 0, which means that there were signifi cant differences between the handling times (T h ) of 4 th instar larvae of S. syriacus when feeding on A. spiraecola and A. gossypii (Table 3). Based on the results, handling times on A. gossyp ii were lower ( Table 2).
The highest maxim um predation rate (T / T h ) was recorded for adult males feeding on A. spiraecola and for 4 th instar larvae feeding on A. gossypii ( Table 2). The val ues of the coeffi cient of determination (R 2 ) indicated that the random predator e quation adequately described the functional response of the different stages of S. syriacus feedi ng on both species of prey (Table 2).

No-choice pre ference
The mean number o f prey consumed by 4 th instar larvae and adult females of S. syriacus when provided with A. spiraecola or A. gossypii in no-choice experiment is illustrated in Fig. 2. There was a signi fi cant difference between the numbe r of prey consumed by 4 th instar larvae (t = -2.961; df = 18; p = 0.008). A. gossypii was m ore preferre d than A. spiraecola by 4 th instar lar vae. Furthermore, there was no signifi cant difference in the number of prey consumed by adult females (t = -0.161; df = 18; p = 0.874). Based on these results, the adult females of S. syriacus consumed signifi cantly more A. spiraecola tha n the 4 th instar larvae (t = -5.762; df = 18; p < 0.0001). However, the consumption of both adult females and 4 th instar larvae of S. syriacus on A. gossypii did not differ signifi cantly (t = -0.809; df = 18; p = 0. 429).

Choice experiment and switching
The mean numbers of prey consumed by 4 th instar larvae and adult females of S. syriacus at different ratios of A. spiraecola and A. gossypii are s hown in Table 4. According to the results, 4 th instar larvae consumed signifi cantly more A. spiraecola an d A. gossy pii at th e 50 : 30 and 40 : 40 (A. spiraecola : A. gossypii) ratios. The highest numbe rs of A. spiraecola and A. gossypii consumed by adult females were recorded at the 50 : 30 and 30 : 50 (A. spiraecola:A. gossypii) ratios. At each ratio ( ex cept 50 : 30 for adult females), A. gossypii was signifi cantly more preferred by both 4 th instar larvae and adult females compared to A. spiraecola (Table 4).
When both A. spiraecola and A. gossypii were offered in equal numbers to the 4 th instar larvae and adult females of S. syriacus, the Manly's preferenc e index (β 1 ) was significantly in favour of A. gossypii (Table 5). For adu lt femal es, the index for other ratios was signifi cantly higher for A. gossypii than A. spiraecola (Table 5). For 4 th instar larvae, the preference index for the 50 : 30 ratio was higher for A. spiraecola and for the 30 : 50 ratio there was no signifi cant preference for either of the two species of aphid (Table 5).

Functional response
The results of functional response studies can determine the success or failure of biocontrol agents in inundative releases (Waage & Greathead, 1988; Heidarian et al., Fig. 2. The mean number of aphids consumed by 4 th instar larvae and adult females of Scymnus syriacus provided with Aphis spiraecola and Aphis gossypii for 24 h in the no-choice experiment. Different letters indicate signifi cant difference between the numbers of aphids consumed by each stage (p < 0.05; t test).  Farhadi et al., 2010). Based on the results, 4 th instar larvae and adult males and females of S. syriacus exhi bit a type II functional response to varying densities of A. spiraecola and A. gossypii, whereby they kill a n increasingly higher propor tion of prey as prey density declines. Type II functional response may in part refl ect the suitability of a predator for regulating orchard populations of both species of aphids (Bayoumy & Awadalla, 2018). This type of response is also reported for S. syriacus preying on A. spiraecola (Sahragard, 1988) Sabaghi et al., 2011a). This reveals that the species of prey and host plant may affect the functional response of S. syriacus (De Clercq et al., 2000;Sarmento et al., 2007).

T able 2. Attack rate (a), handling time (T h ), and maximum predation rate (T/T h ) of 4 th instar larvae and adults of
To determine the magnitude of the responses, the attack rate (a) and handling time (T h ) were estimated. The attack rate (a) and handling time (T h ) of 4 th instar larvae and adult males and females of S. syriacus were different when feeding on the two species of prey, which implies that different stages of the predator with similar responses do not necessarily attack or handle both types of prey in a similar way (Bayoumy & Awadalla, 2018). Attack rate (a) is the rate at which a predator searches for prey. According to the results, the attack rates (a) of adult females foraging for both species of aphids were signifi cantly the highest. Similarly, Sabaghi et al. (2011b) report that adult females of S. syriacus exhibit high attack rates (a) when feeding on A. fabae. A higher attack rate (a) means that higher numbers of prey are eaten by predators in a given time interval  & Kidd, 1996). The larger size of females and their high-energy requirements for producing eggs means that adult females need to attack and consume more prey than other stages (Hodek & Honek 1996). However, the lower theoretical predation rate (T/T h ) of females recorded in this study are due to their higher handling time (T h ).
Handling time (T h ) is the total time spent by a predator in catching, subduing, killing, eating and digesting prey (Veeravel & Baskaran, 1997). In this study, adult males and 4 th instar larvae were signifi cantly faster in subduing and digesting A. spiraeco la and A. gossypii, respectively. Sometimes, predators that have a short handling time (T h ) may not have a high attack rate (a) (Bayoumy & Awadalla, 2018). So, the inverse relationships recorded in this study between handling time (T h ) and attack rate (a) in different stages of S. syriacus may be due to partial consumption, especially at high prey densities (Hodek & Honek, 1996;Aljetlawi et al., 2004). That is, subduing and digesting prey did not impede the predator from further searching for or handling prey. Partial consumption can result in a greater reduction in pest abundance. Pairwise comparisons revealed that adult females attacked A. gossypii more than A. spiraecola. Based on the lower handling time (T h ) of 4 th instar larvae whe n provided with A. gossypii, this stage has higher rates of maximum predation (T / T h ). This is because of the high energy demands of its subsequent non-feeding pupal stage (Hodek & Honek, 1996). In addition, the shorter handling time (T h ) for A. gossypii may be related to its smaller size compared to A. spiraecola (pers. observ.), which makes it more vulnerable to predation (Heydari et al., 2011). It is concluded that 4 th instar larvae of S. syriacus are likel y candidates for use in the biocontrol of A. gossypii. Furthermore, adult females could be more effi cient if employed against infestations of both aphids in citrus orchards.

No-choice preference
Prey preference is critical for understanding the population dynamics of predators and their ability to control the abundance of their prey. In addition, in order to optimize the mass rearing of natural enemies for use as biocontrol agents, it is important to know their preferred prey/host (Jervis & Kidd, 1996) Nutritional quality of aphids for the survival, development, weight gain, egg maturation and oviposition of predators has an important role in the prey preferences of predators (Provost et al., 2006;Jessie et al., 2015). Physiological and morphological status of host plant, predator characteristics such as age, size, hunger level, genetic traits, encounter rate and environmental conditions may also affect the prey preferences of predators (Thompson, 1975;Ferran & Dixon, 1993;Dixon 1998;Al-Zyoud & Sengonca, 2004;Provost et al., 2006;Banihashemi et al., 2017). It is possible that these traits infl uenced the selection of prey by S. syriacus in our study. However, the different ability of 4 th instar larvae and adult females of S. syriacus in handling prey, as mentioned above, as well as different physiological characteristics of aphids may affect the prey preference of the different stages of predators.

Choice experiment and switching
Switching behaviour is a good strategy for coping with prey scarcity (Hassanzadeh Avval et al., 2018). S. syriacus 4 th instar larvae and adult females preferred A. gossypii, when two species of aphids were offered in equal proportions. However, this predator, especially the adult females, did not switch their preference from A. gossypii to A. spiraecola when the abundance of A. spiraecola was increased, indicating a density independent preference. Likewise, working with two coccinellids (Coccinella septempunctata L. and Harmonia axyridis Pallas) and two species of prey (Aphis citricola Van der Goot and Tetranychus urticae Koch), Lucas et al. (1997) does not report switching behaviour as both coccinellids preferred A. citricola. The preference for a particular species implies that the predator choose that prey independently of its abundance or accessibility (Keshavarz et al., 2015).
Generally, predators forage for food effi ciently and tend to increase their nutritional gain while decreasing the costs and risks associated with predation (Stephens & Krebs, 1986). Therefore, when a predator encounters two types of prey it chooses the one most likely to maximize its fi tness (higher survival, fecundity, etc.), and net energy gain (Stephens & Krebs, 1986). In this regard, large prey are more likely to be selected by predators because of their higherenergy content (Soares et al., 2004). The two species of aphids used in our experiments differed in terms of their size, i.e. A. spiraecola is the larger of the two. However, S. syriacus preferred the smaller species (i.e. A. gossypii) in the current study. Similarly, Keshavarz et al. (2015) report that the aphidophagous ladybird, Propylea quatuordecimpunctata L. prefers A. gossypii over A. fabae despite the larger size of the latter. According to Holling (1961), predators spend more time capturing and killing large than small prey, which may decrease their capture efficiency and consumption rate. Schmidt et al. (2012) report that, even when prey size is controlled for, prey quality affects the predator's preference. Other factors such as aphid mobility, its escape behaviour (walking away, kicking), defensive secretion from abdominal cornicles, vulnerability to predators, the aphid's physical structure, such as thickness of its cuticle, length of appendages, overall body size and duration of the experiment may infl uence the acceptance of particular aphids by predators (Lawton et al., 1974;Crawley, 1992;Eubanks & Denno, 2000;Nedved & Salvucci, 2008). Therefore, it is necessary to determine whether and to what extent these features determined S. syriacus preference for A. gossypii over A. spiraecola in this study.

CONCLUSION
The results of the curre nt study extend our knowledge on the functional response and prey preference of one of the most important predators of citrus aphids and provide a better understanding of prey-predator interactions. This laboratory study revealed that the 4 th instar larvae of S. syriacus were more effi cient predators of A. gossypii and adult females more effi cient predators of both A. spiraecola and A. gossypii. However, we believe that the 4 th instar larvae and adult females of S. syriacus could act synergistically in controlling the population abundance of both of these aphids in citrus orchards, because 4 th instar larvae usually remain longer in patches of aphids, and thus can be very efficient in controlling the abundance of aphids in each patch, while the adults usually fl y and disperse actively and thus can find new patches of aphids (Hodek & Honek, 1996). Therefore, further fi eld-based studies are needed in order to provide more accurate estimates of the biocontr ol potential of S. syriacus in citrus orchards.