Apis mellifera L. is the main pollinator of cultivated plants. With the increased emphasis on organic agriculture, the use of botanical insecticides has also increased. However, the effects of these products on bees remain to be determined. In this study, we aimed at assessing the acute toxicity and sublethal behavioral effects of botanical insecticides such as andiroba oil, citronella oil, eucalyptus oil, garlic extract, neem oil, and rotenone on honey bees, A. mellifera . Only andiroba oil demonstrated no lethality to A. mellifera adult workers. However, andiroba oil, garlic extract, and neem oil demonstrated an acute toxicity to bee larvae. Except for eucalyptus oil, larvae fed with syrup containing the other insecticides led to the development of lower body mass in adult workers. All these botanical insecticides were repellent to A. mellifera adult workers. In addition, the eucalyptus oil, garlic extract, neem oil, and rotenone decreased the rate of walking activity in adult workers. Our results demonstrate the potential acute toxicity and sublethal effects of botanical insecticides on honey bees and, thereby, provide evidence of the importance of assessing the risks of the side effects of biopesticides, often touted as environmentally friendly, to nontarget organisms such as pollinators.
Pollinators are essential for obtaining high yields in most cultivated crops. The honey bee, Apis mellifera L., is the most widely used insect for crop pollination ( Garibaldi et al. 2013 ), and their pollination services yield substantial economic benefits for the agricultural production ( Leonhardt et al. 2013 ). However, in recent years, the bee population has decreased throughout the world ( Lebuhn et al. 2013 ). Several factors have been considered as the potential causes of this decline, for instance, the use of organosynthetic pesticides ( Van der Sluijs et al. 2013 ).
Bee larvae and adults might be killed or suffer various sublethal effects when placed in contact with pollen and nectar contaminated with insecticides ( Desneux et al. 2007 , Koskor et al. 2009 ). Among the sublethal effects caused by insecticides are problems with the growth, development, reproduction, and behavior ( Desneux et al. 2007 ). The behavioral changes might include interferences with orientation, movement, and communication activity of the honey bees ( Desneux et al. 2007 , Belzunces et al. 2012 ). The effects of insecticides on the growth and development of honey bees can be defined as pupal malformation and failure of adult emergence ( Koskor et al. 2009 , Belzunces et al. 2012 ).
Specifically, the problems caused by organosynthetic insecticides on the environment and nontarget organisms such as bees ( Biondi et al. 2012 , Decourtye et al. 2013 ) has stimulated the use of natural products as an alternative pest control strategy, mainly in developing countries ( Copping and Menn 2000 , Duke et al. 2010 ). In general, these natural products have a lower persistence in the environment and, therefore, are considered environmentally and toxicologically safer than several of the currently used organosynthetic pesticides ( Copping et al. 2000 , Duke et al. 2010 ). In addition to lethal effects ( Vendramini et al. 2012 , Hsu et al. 2013 ), natural products, such as botanical insecticides, might have repellent effects, inhibit oviposition, and change the feeding and hormonal systems of several arthropod pests ( Copping et al. 2000 , Isman 2006 ). These characteristics made the use of these biopesticides attractive for pest management ( Copping et al. 2000 , Isman 2006 , Duke et al. 2010 ).
Plant-based insecticides such as neem (i.e., azadirachtin) have a range of commercial formulations that exhibit good efficacy against more than 400 insect species ( Ntalli and Menkissoglu-Spiroudi 2011 ) and mites ( Flamini 2003 ). Other botanical insecticides such as the andiroba oil, citronella oil, eucalyptus oil, garlic extract, and rotenone have also been recommended for pest control ( Moreira et al. 2007 , Mureithi 2008 ). However, the authorization to use these products remains controversial worldwide. Although the activity of these botanical insecticides on agricultural pests is well established, their effects on nontarget organisms such as honey bees need to be studied more comprehensively. Therefore, this study aimed at assessing the acute toxicity and behavioral sublethal effects of several botanical insecticides, namely andiroba oil, citronella oil, eucalyptus oil, garlic extract, neem oil, and rotenone on adult and larvae of the Africanized bee A. mellifera .
Materials and Methods
Botanical Insecticides and Honey Bees
The insecticides used in this study are among the main botanical insecticides used to control pests in Brazil ( Moreira et al. 2007 ). Six botanical insecticides were tested, and distilled water was used as the control in all experiments. The botanical insecticides and the concentrations (ml/l of water) used were the following: andiroba oil (10.00), citronella oil (10.00), eucalyptus oil (10.00), garlic extract (0.30; Natualho), neem oil (2.00; Natuneem), and rotenone (5.00; Rotenat CE). We used the insecticides at the concentrations recommended by the manufacturer (Natural Rural Industria e Comercio de Produtos Organicos e de Controle Biologico LTDA, Araraquara, São Paulo State, Brazil).
The adult workers and larvae of A. mellifera used in this study were collected from the hives of the apiary at the Federal University of Viçosa (UFV) in Viçosa (20° 48′45′′ S, 42° 56′15′′ W, altitude 600 m), Minas Gerais State, Brazil. Before the experiments, the hives were not exposed to any chemical treatments, including those generally used to control parasitic mites. To assess the effects of the botanical insecticides on A. mellifera , four bioassays were performed with different evaluation purposes, namely 1) and 2) acute toxicity of botanical insecticides on adult workers and larvae of A. mellifera , 3) repellent effects of botanical insecticides against A. mellifera workers, and 4) effects of botanical insecticides on the walking activity of A. mellifera adult workers.
Acute Toxicity Bioassays.
Effects on Adult Workers
An adult workers’ mortality bioassay was performed by using a completely randomized design, with 7 treatments and 10 replicates. Each replicate experiment included one Petri dish (9 cm diameter by 2 cm height) containing five adult workers. The randomization was performed by assigning the adult workers and treatments to Petri dishes (i.e., experimental units) at random. The bottom of each Petri dish was covered with 9-cm diameter filter paper (Whatman number 1) containing 1 ml of aqueous insecticidal solution at the recommended concentration. In the control dishes, a solution of distilled water with 0.1% Tween 80 was added to the filter paper. The volume of 1 ml of solution was used because this volume was applied to leaf disks of the same area as that of the filter paper. In all the aqueous solutions, the Tween 80 emulsifier was added at 0.1% concentration to ensure the complete solubility of the insecticidal solution in water. The aqueous solutions of insecticides were adequately mixed, homogenized, and applied over the total surface of the paper disk with the use of a micropipette to ensure a uniform distribution over the entire disk area.
After the insecticide applications, the Petri dishes were placed in a fume hood (Marconi, MA055, 1,850 m 3 of air h −1 ) for 5 min to dry the filter paper. Two plastic containers (1.5 cm diameter by 1.0 cm height) were placed in each Petri dish and contained food (85% sugar and 15% honey, w/w) and distilled water. The two containers were refilled when the bee mortality was evaluated, and the bees were allowed to feed freely throughout the evaluations. Five adult workers of A. mellifera were added to each Petri dish. These dishes were placed in an incubator at 25 ± 0.5°C with a relative humidity of 75 ± 5% and a photophase of 12 h. The adult mortality was assessed after 72 h of exposure to the treatments by prodding the insects with a fine hairbrush. The adults were considered dead if they were unable to move after this stimulation. The exposure period of 72 h was chosen because bees in the control treatments showed the minimum survival during this period (≥80%, which is the minimum recommended survival rate for a preliminary lethal effect bioassay; Galdino et al. 2011 ). The adult mortality rates in each replicate were recorded and calculated.
Effects on Worker Larvae
A larval development/mortality bioassay was performed by using a completely randomized experimental design with five replicates. Each replicate consisted of 10 1-d-old larvae of A. mellifera placed in a Petri dish (9 cm in diameter by 2 cm in height). Each larva was placed into of a polyethylene dome (9.8 mm height by 5.6 mm base diameter by 8.8 mm top diameter) containing a semisynthetic diet, water, and yeast extract. The polyethylene domes were inserted in 10 circular perforations created in a 10-mm-thick Styrofoam disk placed at the bottom of the Petri dish. One-day-old larvae were obtained from a brood comb of A. mellifera colonies maintained at the experimental apiary, which was marked and inspected daily to verify the presence of newly hatched larvae. In general, larvae hatch from eggs 66–93 h after oviposition ( Collins 2004 ). Newly hatched larvae were removed from the hives and immediately transferred to their polyethylene domes.
A semisynthetic larval diet that contained 49% royal jelly, 36.3% water, 6.8% d-fructose, 6.8% d-glucose, and 1.1% yeast extract was prepared according to the methods described by Rembold and Lackner (1981) and a modified version of the methods described by Silva et al. (2005) . The sugar and yeast extract were dissolved in water and filtered through a Millipore membrane (0.22 μm diameter pore). The royal jelly produced in the experimental apiary was added to the filtrate. The diet was homogenized and stored at 5°C in 10-ml glass vials covered with aluminum foil. Larvae were fed daily with 4, 15, 25, 50, and 70 μl of the diet for the first, second, third, fourth, and fifth days of larval development, respectively, based on the methodology described by Silva et al. (2005) . In addition, the larvae were fed a dietary insecticide only on the third day. According to Peng and Jay (1977), only after the third day of larval development, do the worker larvae begin to feed directly on food collected in the field (i.e., pollen and honey). Therefore, on the third day, each larva was provided 25 μl of a diet containing the mixture of 1.0 g royal jelly, 0.136 g d-fructose, 0.136 g d-glucose, 0.022 g yeast extract, and 0.73 ml of the botanical insecticides at the recommended concentration for pest control. In the control treatments, larvae were fed a diet without insecticides. This procedure of larval exposure to the botanical insecticides through the diet was adopted to verify the possible effects of these insecticides on the larvae in case of contamination with pesticide residues present in the pollen and nectar collected in the fields. Larvae were kept within the artificial domes and remained in constant contact with the diet during the feeding period.
After the larvae feeding period (i.e., after the fifth day of larval development), the larvae were removed from the polyethylene domes and placed in another Petri dish containing a double layer of cheesecloth at the bottom. This procedure allowed the larvae to pupate. The area within each Petri dish was divided into 10 compartments (of 2 cm height) with a white bond paper to individualize each larva. The Petri dishes containing the larvae were maintained in an incubator at 34°C, with a relative humidity of 96% and in the dark throughout the bioassay period to simulate the microclimate of the inside of the hive ( Becher et al. 2010 ).
The numbers of living individuals and their stage of development were assessed daily since they were newly emerged larvae until they became adults. Dead bees in each Petri dish were removed. At the end of the bioassay, the body masses of the surviving adult workers were measured and the larvae mortality rates were calculated.
A larval survival bioassay was performed in a similar manner as the larval developmental mortality bioassay. However, in this bioassay, each larva constituted one replicate. For each treatment, the survival of 50 larvae was monitored every day during the development of newly emerged (i.e., 1-d old) larvae to adulthood. The survival percentage at the end of each developmental stage (larva and pupa) was calculated and used for the survival analyses.
Behavioral Sublethal Effects Bioassays.
Forager Responses to Dietary Botanical Insecticides
A free choice test was conducted in an open apiary to assess the forager responses to dietary botanical insecticides. Foragers were allowed to select from an array of vials with sugar syrup containing different amounts of insecticides. Each glass vial (100 ml) (5 cm diameter by 8.8 cm height) contained 20 ml of sugar syrup (40% sucrose and 60% water, w/v; Malerbo Souza et al. 2003 ). Each botanical insecticide was added to the syrup at the concentration recommended for pest control. The control treatment contained only syrup. The experiment was performed using a randomized complete block design with 18 replicates. Each block (i.e., replicates) consisted of a wooden table (2 m length by 1.5 m width by 1 m height) with seven glass vials arranged in a circle of 1 m in diameter. The glass vials were spaced at a distance of 10 cm, and the table was 30 m away from the hives. The bees that fed on each vial were counted and collected in a 1.5 liter plastic container sealed with a lid. Forager choices were observed for 10 min for each replicate period. A new block replicate was set up 15 min after the end of each observation period. During the experiment, the average temperature was 24.7°C, and the air relative humidity was 83%.
Walking Activity Bioassay of Adult Workers
The effects of the botanical insecticides on the walking activity of adult worker bees were evaluated in a Petri dish arena assay based on the methods described by Tomé et al. (2012 , 2015 ) and Cordeiro et al. (2010) . The experimental conditions in this experiment were similar to the ones in the mortality assessment method. However, in this experiment, only one adult worker of A. mellifera was placed in each Petri dish. A total of 14 individuals (i.e., replicates) were used for each treatment. No mortality was observed during the exposure time (10 min) for any treatment. Each insect was left in an untreated arena for 24 h before the bioassay for acclimation purposes ( Xavier et al. 2010 ).
Each Petri dish was placed in a videotrack system (Version 3.0; ViewPoint Life Sciences Inc., Montreal, Canada) to evaluate the walking activity of the adult workers following the methods adapted from Xavier et al. (2010) . This system included a moviecamera attached to a microcomputer with software that recorded the walking activity of the individual bees. The measurements taken by the tracking system included the distance walked and the total ambulatory time (i.e., walking time) in each arena. The walking speed (cm/s) was calculated by dividing the distance walked by the time spent walking. The observation time for each replicate was 10 min.
Adult worker and larval mortalities (acute toxicity bioassays), adult worker body masses (bioassay of acute toxicity to larvae), the numbers of visiting bees (feeding preference bioassay), and the walking rates of bees (walking activity bioassay) were compared across diet treatments with a one-way analysis of variance (PROC ANOVA; SAS Institute 2008 ). The treatment means were compared with a Scott–Knott test at P < 0.05 ( Scott and Knott 1974 ). The normality and homoscedasticity of the data were evaluated (PROC UNIVARIATE; SAS Institute 2008 ), and no data transformation was required for the analyses. To assess the insecticide toxicity, survival curves of larva and pupa workers over a period of time were estimated using the LIFETEST procedure of SAS ( SAS Institute 2008 ). Survival curves were obtained through Kaplan–Meier estimators generated from the percentage of adult, larva, and pupa bees surviving from the newly hatched larvae until the end of the bioassay.
Acute Toxicity of Botanical Insecticides to A. mellifera
Adult workers ( F6;28 = 18.74, P < 0.001) and worker larvae ( F6;28 = 3.79, P = 0.010) experienced significantly different mortality rates across the dietary insecticide treatments. Adult workers that ingested citronella oil, eucalyptus oil, garlic extract, neem oil, or rotenone suffered from 42% to 60% higher mortality rates than workers fed with uncontaminated control diets. Only the andiroba oil did not increase the adult worker mortality. Worker larvae exposed to dietary andiroba oil, garlic extract, and neem oil experienced an increased mortality compared with workers fed on control diets. In contrast, the citronella oil, eucalyptus oil, and rotenone showed no significant acute toxicity on worker bee larvae ( Fig. 1 ).
The survival data analysis of A. mellifera adult workers originated from larvae fed on a diet with botanical insecticide residues showed indicated significant differences among the treatments (log-rank test, χ 2 = 56.35, df = 6, P < 0.001). The survival of A. mellifera workers was lower for bees raised from larvae fed on a diet treated with andiroba oil, garlic extract, and neem oil. The mortality of adult workers raised from larvae fed on a diet treated with these insecticides increased over a period of time ( Fig. 2 ). The survival curves of A. mellifera worker pupae raised from larvae fed on a diet treated with andiroba oil, garlic extract, and neem oil were similar (log-rank test, χ 2 = 0.37, df = 2, P = 0.833). The andiroba oil and garlic extract induced the mortality of larvae bee during pupation and adult emergence. The neem oil caused mortality at the beginning of the larval stages and at the adult emergence stage ( Fig. 2 ).
Adult workers raised from larvae fed on a diet with insecticide ( F6;21 = 6.86, P < 0.001) experienced significantly different body masses. Larvae fed on a diet containing andiroba oil, citronella oil, garlic extract, neem oil, or rotenone emerged as adult workers with lower body masses than the controls bees. In contrast, worker larvae fed on a diet with eucalyptus oil emerged with body masses similar to those of bees in the control treatment ( Fig. 3 ).
Walking Activity and Forager Preference of Adult Workers
The number of forager visits to the sugar syrup were significantly different between the treatments ( F6;102 = 13.54, P < 0.001). More foragers visited the unsupplemented control sugar syrup vials than the sugar syrup vials containing any of the insecticides. The sugar syrups with citronella and eucalyptus oils received the lowest number of visits. In the sugar sources with andiroba oil, garlic extract, neem oil, or rotenone, the number of visiting bees was intermediated between the unsupplemented control and the citronella and eucalyptus oil vials ( Fig. 4 a).
The walking speed of adult worker bees was significantly different between the treatments ( F6;91 = 5.50, P < 0.001). The adult workers moved slowly across the surfaces treated with eucalyptus oil, garlic extract, neem oil, or rotenone compared with the control bees. The speeds of adult workers on surfaces treated with andiroba oil and citronella oil were similar to those of the control bees ( Fig. 4 b).
The use of botanical insecticides has recently been promoted as an alternative pest control method, especially in crop systems where the conventional synthetic insecticides have limited use, such as in agroecological farming and organic agricultural systems ( Isman 2006 , Duke et al. 2010 ). However, botanical insecticides might cause adverse effects to nontarget organisms such as bees ( Melathopoulos et al. 2000 , Koskor et al. 2009 , Xavier et al. 2010 ). In this study, we confirmed that several botanical insecticides widely used for crop protection against pests (e.g., Neem) and as other less used for this purpose caused lethal and sublethal effects on larvae and adult honey bees A. mellifera at the recommended concentration for pest control . The garlic extract and neem oil induced toxicity to both larvae and adult worker bees. The citronella oil, eucalyptus oil, and rotenone demonstrated toxicity only to adult worker bees, but the andiroba oil was toxic to larvae only.
Most studies on the effects of insecticides on bees have focused on adult workers. However, all the developmental stages and castes might be potentially affected by residues of insecticides ( Tomé et al. 2012 ). In this study, we found that the acute toxicity of botanical insecticides on adults and larvae of A. mellifera workers increased with the exposure time of the insecticides through body contact in adults or ingestion of contaminated larval diets. This result indicates that the acute toxicity of the botanical insecticides on A. mellifera might have a delayed effect; however, a faster action of these insecticides can be achieved with an increase in the exposure time of the bee to the insecticide. Furthermore, bees can be contaminated with botanical insecticides during their repeated foraging visits to flowers in sprayed fields ( Koch and Weiber 1997 ) or through contact with residues of insecticides used to control parasitic mites of honey bees accumulated in the brood nests ( Melathopoulos et al. 2000 , González-Gómez et al. 2012 ). Another set of results indicate that once these insecticides reach the hives, they can exert toxic effects on the larvae bees. Thus, to preserve the A. mellifera population, it is necessary to reduce the exposure of bees to botanical insecticides, which can be achieved through the application of insecticides using the principles of ecological selectivity ( Bacci et al. 2009 ). Ecological selectivity (i.e., the selective use of pesticides) is based on tactics such as critical selection, dosage reduction, timing adjustment, and selective placement on the plant or surrounding environments as well as application techniques of broad-spectrum insecticides ( Hull and Beers 1985 ). In this context, the insecticide should be applied when the bees have lower foraging rates to the crops (i.e., late afternoon; Joshi and Joshi 2010 ). Another practice that can minimize the impact of botanical insecticides is the closure of the hive entrance and the use of artificial feeding on days when pesticides are sprayed to prevent the contact of bees with these products ( Riedl et al. 2006 ).
In this study, we noted the effects of botanical pesticides such as garlic extract and neem oil, which had been previously described as “safe” for honey bees ( Naumann and Isman 1996 , Riedl et al. 2006 ). We found that these two insecticides resulted in an acute toxicity to both larvae and adult workers of A. mellifera , which indicates that their use should be avoided in crops during the flowering stage when the plants are visited by bees. Although Melathopoulos et al. (2000) did not observe negative effects of neem on adult honey bees, they observed that this insecticide reduced the amount of larvae in colonies and, at sublethal doses, different malformations occurred when the bees emerged from the cocoons. Rembold et al. (1980) found that neem oil extracts were acutely toxic to immature honey bees. In addition, Efrom et al. (2012) found a significant increase in the mortality of adult workers A. mellifera with an increased exposure time of the bees to different concentrations of neem oil. These same authors observed a significant mortality of A. mellifera due to rotenone. Similarly, Xavier et al. (2010) observed a high acute toxicity of citronella oil to adults of the indigenous stingless bee Tetragonisca angustula .
The death of A. mellifera pupae exposed to andiroba oil, garlic extract, and neem oil might be due to the effects of these insecticides on the insects’ ecdysis ( Riba et al. 2003 , Pineda et al. 2009 , Mhazo et al. 2011 ). This hypothesis was corroborated by our finding of deformities in the majority of killed pupae. Melathopoulos et al. (2000) observed that neem oil formulations might have a negative effect on the development of A. mellifera larvae. Another sublethal effect observed on larvae that fed on diets containing andiroba oil, citronella oil, garlic extract, neem oil, and rotenone was the emergence of adult workers with lower body mass. This observation can be attributed to these insecticides, which reduced the food conversion rates in the larvae ( Savarimuthu et al. 2004 ).
Adult workers visited syrups that contained any of the botanical insecticides less often than the control syrups (i.e., syrups containing only sugar and water), which indicates that insecticide-containing syrups have a deterrent effect on foraging A. mellifera . In agreement with our results, Koskor et al. (2009) observed that a chronic treatment with sublethal doses of neem insecticides affected the pollen foraging of bumble bees. The repellent effects of garlic and citronella extracts on A. mellifera were also observed by Nicodemo and Nogueira Couto (2004) . However, Naumann et al. (1994) found that although foragers were deterred from feeding on sugar solutions with extremely low concentrations of azadirachtin, no significant reduction was noted in the foraging bees in canola fields sprayed with neem insecticide. Likewise, we observed the repellent effects of botanical insecticides on honey bee foragers. However, additional field experiments to assess the foraging behavior of bees are necessary to understand the repellent effects of insecticides on A. mellifera . The use of repellent compounds might be exploited to reduce the chance of accidental pesticide poisoning in honey bees. Repellent products might be included in prepared solutions with organosynthetic pesticides to reduce the contact of bees with these pesticides ( Solomon and Hooker 1989 , Belzunces et al. 2012 ). However, it is important to emphasize that these repellent products should not exert acute toxicity on honey bees. Otherwise, if the repellent effects last for a long time, they can affect the foraging activity and generate a state of starvation or nutritional deficiency at the individual or colony ( Belzunces et al. 2012 ). Furthermore, insecticides with repellent effects decrease the number of bees foraging on flowers, which might lead not only to inadequate nectar and pollen gathering but also to deficient crop pollination.
In addition to causing an acute toxicity to the bees, the eucalyptus oil, garlic extract, neem oil, and rotenone also decreased the walking activities of A. mellifera adult workers. These lower walking activities might result in greater contact with toxic pesticide residues, further increasing their toxic effects because the lethal effect of a substance is proportional to its amount and the time of contact with the body ( Corso and Gazzoni 1998 ). Another consequence of the walking behavioral change of adult workers is the impairment of activities both inside the hive and during foraging ( Tomé et al. 2012 ).
In conclusion, we demonstrated that several botanical insecticides, which are often touted as safe and environmentally friendly, might generate acute toxicity and sublethal effects on honey bees. The garlic extract and neem oil showed an acute toxicity to both adult and larva workers. Instead, the other insecticides showed an acute toxicity to either adults or larvae of A. mellifera . All the botanical insecticides changed the foraging behavior by being repellent to adult workers. In addition, the eucalyptus oil, garlic extract, neem oil, and rotenone decreased the rate of walking activity of adult workers. Therefore, the use of botanical insecticides for controlling insect pests on crops should be exercised with caution. Likewise, the potential risks of the side effects of these pesticides to nontarget organisms such as pollinators should be evaluated.
We thank the National Council of Technological and Scientific Development (CNPq), the Coordination for the Improvement of Higher Education Personnel (CAPES), the Minas Gerais State Research Foundation (FAPEMIG) and the Brazilian Agricultural Research Corporation (EMBRAPA-Café) for scholarships and resources.