The Effects of Four Insect Growth-Regulating (IGR) Insecticides on Honeybee (Apis mellifera L.) Colony Development, Queen Rearing and Drone Sperm Production
- First Online:
- Cite this article as:
- Thompson, H.M., Wilkins, S., Battersby, A.H. et al. Ecotoxicology (2005) 14: 757. doi:10.1007/s10646-005-0024-6
- 365 Views
This study assessed the effects of exposure to IGRs on the long-term development of the honeybee colony, viability of queens and sperm production in drones and integrated the data into a honeybee population model. Colonies treated with diflubenzuron resulted in a short-term reduction in the numbers of adult bees and brood. Colonies treated with fenoxycarb declined during the season earlier and started the season slower. The number of queens that successfully mated and laid eggs was affected in the fenoxycarb treatment group but there were no significant differences in the drone sperm counts between the colonies. An existing honeybee population model was modified to include exposure to IGRs. In the model, fenoxycarb reduced the winter size of the colony, with the greatest effects following a June or an August application. Assuming a ‘larvae per nurse bee’ ratio of 1.5 for brood rearing capability, the reduction in winter size of a colony following a fenoxycarb application was at its worst about 8%. However, even if only those bees reared within 2 weeks of the IGR being applied are subject to premature ageing, this might significantly reduce the size of over-wintering colonies, and increase the chance of the bee population dwindling and dying in late winter or early spring.
KeywordsIGRhoneybeepopulation modelsublethal effects
Insect growth regulating insecticides (IGRs) have been developed due to their high activity and selectivity against insects with inherently low toxicity to non-target wildlife. The role of insect hormones in reproduction has been extensively reviewed (De Fur et al., 1999; Tasei, 2001). Due to their mode of action these types of pesticides are likely to pose a greater hazard to larval stages than to adult insects. However, there are few studies on the effects of these types of compounds on the development and over-wintering ability of honeybee colonies or on their ability to produce viable drones and queens.
This study aimed to assess the effects of exposure to a juvenile hormone analogue, fenoxycarb, a chitin synthesis inhibitor, diflubenzuron, an ecdysteroid synthesis inhibitor, azadirachtin and an ecdysteroid analogue, tebufenozide, on honeybee colonies, drone sperm production and queen rearing. Both juvenile hormone analogues and chitin synthesis inhibitors are known to adversely affect honeybee brood through mortality and developmental abnormalities (for a review see Tasei, 2001) but have also been reported to have sublethal effects on adult bees. For example, juvenile hormone analogues have been reported to result in precocious foraging, reduced longevity, inhibition of hypopharyngeal glands and impaired vitellogenin synthesis in newly emerged workers (Jaycox et al., 1974; Gerig, 1975; Robinson, 1985; Pinto et al., 2000). Diflubenzuron dosing of newly emerged bees resulted in reduced weight gain, and suppressed development of the hypopharyngeal glands (Gupta and Chandel, 1995). However, the effects of a range of IGRs on the ability of drones to produce sperm or on the viability of queens have not been reported. IGRs may thus result in short-term effects on brood or longer-term sublethal effects. In this study a honeybee population model was also used to identify the most important factors in determining the over-winter size, and thus survival, of honeybee colonies after IGR exposure.
Materials and methods
Fenoxycarb was obtained as the formulation Insegar (250 g/l) and diflubenzuron as Dimilin Flo (480 g/l) as commercial formulations from UAP, Welburn, York. UK. Neither azadirachtin nor tebufenozid are registered for use as insecticides in the UK and therefore they were obtained as active ingredients from Sigma-Aldrich UK.
Effects on colonies and drones
Test solutions were prepared by diluting the IGRs with 50% w/v sucrose to a rate equivalent to their maximum application rate (0.6 kg/200 l Insegar (25% fenoxycarb), 0.3 l/500 l Dimilin Flo (48% diflubenzuron)) on flowering crops. As neither azadirachtin nor tebufenozide are registered in the UK they were applied at the maximum rate shown not to result in anti-feedant properties (Peng et al., 2000), 1 mg ai/l azadirachtin and the maximum application rate 160 mg ai/l tebufenozide (BCPC 2001). Negative control solutions were 50% w/v sucrose.
Standardised colonies of bees and brood were obtained from National Bee Unit, CSL, York. Colonies with a low incidence of minor brood disease (chalkbrood, sac brood and bald brood) and in which both American foulbrood (Paenibacillus larvae) and European foulbrood (Melissococcus pluton) were clinically absent were selected. Colonies that had been previously treated with a varroacide were not used within 4 weeks of treatment.
Test colonies were headed by queens of similar age and were housed in a single chamber wooden Smith hive. Each colony had a dead bee trap fitted to the front. Any drone pupae in the colonies were destroyed prior to treatment and drones were excluded by shaking the bees outside the colony (ensuring the queen remained within the brood chamber) and placing a queen excluder on the base of the brood chamber to prevent re-entry of adult drones. The test colonies were numbered and each test group labelled with the study number, and treatment. Colonies were located on the test site at CSL and allowed to fly freely. Dose groups were placed approximately 20 m apart from each other to minimise drifting.
The colonies were divided into groups of five colonies. On day 0 the groups were dosed. Groups of colonies were fed fenoxycarb, diflubenzuron, azadirachtin or tebufenozide in sucrose and a control group was fed untreated sucrose. Doses were administered by removing frames of stores from the colonies, and placing a glass beaker containing 500 ml treated or control sucrose (50% w/v aqueous solution) within the brood chamber; the beaker contained a cork float to allow access to the sucrose. After 1 week the beaker was removed, the volume of any remaining feed was recorded and the frames replaced.
On the day of dosing a single frame was selected from the centre of each colony and a plastic overlay sheets using to locate each cell, 100 brood cells containing eggs were selected. On day 17 after marking each cell the cell was uncapped and the pupa removed to observe any abnormalities. In addition another frame was selected on the day of dosing in which 300 cells were marked and their contents, (eggs, young larvae, old larvae, sealed, stores) were determined. The fate of the contents of the 300 cells was determined by reference to the previous age of the contents and the expected change over the week period. This procedure was repeated after 1, 2, 3 and 5 weeks for the 100 eggs and weekly for 6 weeks for the 300 marked cells. Percentage data was arcsine transformed prior to statistical analysis using ANOVA with treatment (p<0.05).
All colonies were assessed by determining counts of the number of combs covered by adults and brood, as well as any behavioural or physical abnormalities. These assessments were undertaken prior to the day of test item application, 2 weekly after the application until November and then monthly until a year after treatment. The data was analysed using a generalised linear model. Data on the levels of brood and bees in each colony were compared with controls during the spring after treatment to determine if there were effects on colony development. The data were normalised by comparison with colony size at time of treatment. The REML part of Genstat was used to fit correlated error structures. The best fitting model was one which modelled the covariance structure of the data with an antedependence structure of order 1. Use of this model implies that measurements on different colonies are independent but that successive measurements made on the same colony are correlated. Measurements that are at least two time units apart are taken to be independent given the intervening measurement. Unequal time intervals are allowed for in the analysis. The best transformation to achieve approximate stability of variance over treatments at a particular time was the square root, a transformation that is often useful for data that are based on areas. Since the model is fitted to the square root of the ratio all tests need to be done on that scale. The complex variance structure means that treatment means cannot be compared using the standard t-tests. Wald statistics were used for testing treatment or time differences of interest.
Adult drones emerging after treatment were prevented from leaving the colony by a queen excluder placed on the floor of the colony above the entrance. Drones were collected from the colonies 35–40 days after treatment. Semen was collected from mature drones into saline and the sperm counted using a haemocytometer. The statistical analysis of sperm count data was performed using ANOVA with treatment (p<0.05).
Honeybee queens were reared by grafting 1–2 day old worker larvae and placing them in rearer colonies. This allowed the productions of large numbers of sister queens. The queens were allowed to pupate in the rearer colonies and then each pupa was transferred to an Apidea containing a small number of adult workers. The Apidea was supplied with 400 ml fondant containing the IGR at a rate equivalent to the application rate (0.6 kg/200 l Insegar, 0.3 l/500 l Dimilin Flo, 1 mg ai/l azadirachtin, 160 mg ai/l tebufenozide) or untreated fondant. This enabled the queens to be fed on the treated sucrose from emergence. The fenoxycarb and diflubenzuron tests were repeated in two consecutive years with 12 queens used per treatment group and 12 queens used in the controls in each year. The azadirachtin and tebufenozide study was run only in the second year. Two to three weeks after emergence the number of mated queens was determined by recording egg production. Criteria of 60% of the control queens to be mated determined the test as valid. All mated queens were then introduced into larger colonies and when established (approx 2 weeks later) the queen caged onto a single frame for one day and the number of eggs laid recorded. The queen replacement in the colonies was too large to allow assessment of the egg production to be reliably assessed in the first year. Therefore, egg production assessments were undertaken for all treatments in the study undertaken in the second year. The statistical analysis of egg production data was performed using ANOVA with treatment p<0.05.
Honeybee population model
An existing honeybee population model (Wilkinson and Smith, 2002) was modified to include exposure to IGRs. The model simulates egg laying in the colony, brood development, and an adult honeybee population consisting of younger nurse bees and older forager bees. Adult honeybee workers normally change over from brood-rearing work to foraging work at around 3 weeks of age (Graham, 1992). In the model, the queen’s egg laying rate is determined by a variable curve function simulating seasonal variation. Adult bee mortality is also determined by a seasonally dependent function, set at a level to give a model population with a realistic seasonal cycle, and also a stable population from year to year. An extra feedback mechanism was added to the model so that the amount of brood that could be reared was limited by the number of nurse bees available, as might occur several weeks after a period of high brood mortality.
The impact of brood mortality after treatment with an IGR was modelled by reducing egg laying rate and the effects on population size at the end of December was recorded. Colony size at the end of the year is an important factor in determining the success of colony development the following spring, and colony production the following summer, which is known to affect over-winter survival. The month of IGR application was varied between model runs to determine the effects of the timing of application.
There is some evidence of sub-lethal effects of IGRs on honeybees (e.g. juvenile hormone effects on longevity). Sub-lethal effects were simulated in two ways: firstly by shortening the time that a nurse bee can produce brood food and rear brood before converting to forager status (premature ageing mechanism), and secondly by increasing the mortality of affected adults (shortened lifespan mechanism). The former was simulated by reducing the mean number of larvae that can be reared per nurse bee (the L/N ratio), and the latter by increasing mortality rates by a multiplication factor. Since fenoxycarb can persist for several months (Pesticide Manual 2002, max DT50 2.5 months) and the effects of fenoxycarb exposure have been reported for 2–3 months after treatment (Tasei, 2001) we initially simulated sub-lethal effects lasting 2.5 months from initial IGR application. Shorter effect periods were also then simulated to determine how important the IGR persistence might be for colony over-wintering size.
The colonies treated with control, diflubenzuron, azadirachtin and tebufenozide all consumed the 500 ml treated sucrose. Of the colonies treated with fenxoycarb, 1 consumed 300 ml, 2 consumed 350 ml and 2 consumed the 500 ml treatment, i.e. a mean of 400 ml was consumed. Based on a mean of 14950 bees/colony (range 12600–16800) 500 ml of treated sucrose equates to 0.033 ml/bee. Based on a mean 7366 brood cells per colony (at all stages of development) (range 5702–9576) 500 ml treated sucrose equates to 0.067 ml/brood cell. As these IGRs act on brood rather than on adult bees the calculated dose per brood cell (assuming all are treated in a single brood cycle), this would equate to 50 μg fenoxycarb/brood cell, 214 μg diflubenzuron/brood cell, 0.067 μg azadirachtin/brood cell and 10 μg tebufenozide/brood cell.
The mean sperm count in drones from controls was 2.3×106 (stdev. 1.4×106) sperm/ml, from diflubenzuron treated colonies was 2.7×106 (stdev. 1.8×106), from the fenoxycarb was 2.0×106 (stdev. 1.9×106), from azadirachtin treated colonies was 1.8×106 (stdev. 6×105) and from tebufenozid treated colonies was 1.2×106 (stdev. 5×105) sperm/ml. There were no significant differences in the sperm counts between the colonies but there were fewer drones per colony available for sampling in the diflubenzuron (mean 31) and fenoxycarb (mean 26) than in the control colonies (mean 39) due to the effects of the treatments. In addition, one colony in the diflubenzuron treatment and two colonies in the fenoxycarb treatment failed to provide any mature drones.
Colonies treated with diflubenzuron resulted in a short-term reduction in the numbers of adult bees and brood after treatment when compared with controls. There was no significant effect on development of brood the following spring but there did appear to be a slower increase in levels of brood when compared with controls.
Increasing mortality rates by 30% to simulate sub-lethal effects of shortened lifespan, rather than reduced brood rearing capability, gave a significantly smaller effect on winter size (Fig. 8c, compare with b). With the ‘shortened lifespan’ mechanism, a fenoxycarb application had the worse effect on winter size of a colony when applied in July. In order to simulate an effect with the ‘shortened lifespan’ mechanism as large as that with the ‘premature ageing’ mechanism (Fig. 8b), the mortality of affected adults had to be increased by 500%.
There are very few studies of the long-term impact of pesticides on honeybee colonies. This is a particular concern for IGRs, which are not acutely toxic to adult bees but affect brood and may also have sublethal effects on emerging adult bees, e.g. behaviour.
Tebufenozide (a dibenzylhydrazine) is marketed as a lepidopteran specific ecdysteroid analogue and this study confirmed that it has no apparent effects in honeybee colonies or the development of queens. The selectivity of tebufenozide is thought to be due to selective binding to the ecdysone receptors (Smaaghe et al., 1996). There are no broad-spectrum ecdysteroid analogues marketed as pesticides and therefore the risk posed by this class of IGRs to honeybees is currently low although the effects of other analogues should be reviewed to confirm the absence of adverse effects.
Diflubenzuron is a chitin synthesis inhibitor rather than an endocrine disrupter and has severe short-term effects on brood mortality but apparently no longer-term effects on colony viability. Gupta and Chandel (1995) showed exposure to diflubenzuron, resulted in reduced weight gain in newly emerged bees and suppressed the development of the hypopharangeal glands. There are also apparently effects on the number of eggs laid by treated queens presumably due to effects on the development of the ovaries or ovicidal effects.
Colony overwintering was also affected by azadirachtin, a plant allelochemical extracted from neem (Azadirachta indica) seeds, that had no apparent effects at the brood level, in fact, colony assessment showed increases in brood levels after treatment, or on queen development. However, 4 of the 5 colonies treated with azadirachtin failed to overwinter. This suggests that azadirachtin affects the exposed bees, as brood, sufficiently for them to fail to survive from September/October to the start of the brood rearing season in March without exerting detectable short-term effects. The effects of azadirachtin are thought to be due to effects on juvenile hormone and ecdysteroid titres through a blockage of morphogenetic peptide hormone release (e.g. PTTH, allototropins) and a direct effect on other tissues resulting in an overall loss of fitness (Mordue and Blackwell, 1993).
The most severe effects in this study were shown by fenoxycarb, a juvenile hormone analogue, which affected the colony viability not only in the short-term, by resulting in brood mortality, but also affects the ability of the colony to overwinter. Fenoxycarb also severely affected the ability of the colony to requeen itself as none of the treated queens were mated. Juvenile hormone levels are usually low and ecdysteroid levels high in laying queens (Robinson et al., 1991). In bumblebees, juvenile hormone (JH) is involved in caste determination with the critical age 5 days and prospective queen larvae have significantly higher levels of JH than workers (Cnaani et al., 2000). The effect on the ability of the honeybee colony to overwinter may be due to sublethal effects on the adult bees by inducing precocious foraging (Robinson and Ratnieks, 1987) or decreasing longevity. Studies have reported that older honeybees regulate behavioural development in younger bees (Huang et al., 1998). Bees reared in isolation from older bees had higher levels of juvenile hormone were more likely to become precocious foragers. JH is involved in the regulation of behavioural development in honeybees (Robinson and Vargo, 1997). JH titres increase with age, they are low in bees that work in the hive and high in bees that forage and are involve in colony defense. Treating bees with JH induces precocious foraging and removal of the corpora allata (the glands that produce JH) delays bees in developing into foragers (Huang et al., 1998). Thus, it appears likely that the exposure to the JH analogue fenoxycarb induced precocious foraging in the exposed individuals and reduced the number of nurse bees to rear brood. This is supported by the work of Jaycox et al. (1974) who showed that bees treated with a JH mimic could not develop their hypopharangeal glands and started to move out of the brood nest to guard the colony and forage and also reduced longevity by 39%. Rutz et al. (1974) also showed that application of a JH analogue resulted in regression of the hypopharangeal glands. Protein status appears to be a major determinant of the ability of bees to overwinter with the lipoprotein vitellogenin playing a significant role (Amdam and Omholt, 2002). Pyriproxyfen, a JH analogue impaired vitellogenin synethsis in the haemolymph (Pinto et al., 2000). Therefore, fenoxycarb may affect both the behaviour and the energy stores of the developing bees affecting the ability of the colony to prepare itself for winter and develop in the following spring.
The honeybee population model simulates effects of the brood mortality immediately following application of the chemical. However, the size of the effect is relatively small, and does not account for the more severe effects measured in the field, where fenoxycarb dosed colonies were 50% smaller than control colonies the following spring in both numbers of bees and levels of brood rearing. Despite various assumptions, the model has shown that there are two mechanisms that could realistically account for reduced colony size during the following winter, and hence a reduced likelihood of colonies surviving to the following spring. The first mechanism, premature ageing of affected adult bees and hence early loss of brood-rearing capabilities had a much larger effect than expected, though as yet there is no field data to indicate how much the L/N ratio might be affected in reality. The second mechanism, shortened lifespan of affected adult bees (due to onset of precocious foraging) had a smaller effect and so is perhaps less likely to be the explanation of observed field effects. Again, an important unknown factor is the extent by which mortality rate of adult bees in the field might be increased by IGRs such as fenoxycarb.
A significant finding from the model is that application of IGRs in spring and early summer could have substantial long-term effects on colony size and viability. Whereas, one might suppose that a colony would easily recover by the following winter, the model suggests this will not necessarily be the case. Sub-lethal effects such as premature ageing can have worse effects than massive brood mortality, as it severely reduces the ability to rear the next generation of nurse bees, and there is a knock-on effect. Also, although the effects were smaller during the later months such as August, September and October, late application of IGRs, although not substantially reducing the size of colonies entering winter, may have an adverse effect on the ability of the bee population to grow in early spring of the following year when it is vital for colony survival that there are over-wintered bees still in a physiological state to rear brood.
The model showed that even if only those bees reared within 2 weeks of the IGR being applied are subject to premature ageing, such relatively short-term persistence of IGRs might nevertheless significantly reduce the size of over-wintering colonies and increase the chance of the bee population dwindling and dying in late winter or early spring, as was shown for one fenoxycarb and four azadirachtin treated colonies.
The worst time of the year to apply an IGR varies in the model according to which mechanism is employed. To better understand the potential seriousness of these sub-lethal effects, it would be necessary to undertake field studies to determine, which is the prime mechanism of sub-lethal effects. If premature ageing is the prime mechanism then it would be useful to determine a measure of the effect, for example, how much the L/N ratio might be reduced. If increased mortality is the prime mechanism then likewise that should be quantified in the field. Furthermore, the persistence of sub-lethal effects may vary between chemicals. Results of such field studies could be fed back into the model to better predict the seriousness of sub-lethal IGR effects, and to which months IGRs might be applied to minimize longer-term impacts on honeybees.
This study has shown significant effects of some IGRs on the longer-term viability of honeybee colonies. Short-term effects on brood mortality may appear to recover within 4–6 weeks but there may also be longer-term sublethal effects on emerging adults and queens. This study has used high levels to determine whether effects can be detected (maximum application rate for fenoxycarb and diflubenzuron, maximum dose whilst ensuring uptake for azadirachtin) and the effects of foraging on crops treated at these levels (actual usage rate for azadirachtin) should be elucidated. It is important that sublethal effects on bees, both workers and queens, are not ignored in pesticide risk assessment. The effects of IGRs on populations on non-target invertebrates is dependent on their reproductive rate (Stark et al., 1997) and therefore these types of compounds may have a more significant impact of species with lower reproductive rates, such as bumblebees, which would be less affected by reduced lifespan, as they do not overwinter.
This work was funded by the UK Pesticides Safety Directorate, Defra (Project PN0936).