1 Introduction

The honey bee (Apis mellifera) is one of the most economically important insects for humans. Besides providing six different products (honey, pollen, wax, royal jelly, venom, and propolis), the honey bee plays a significant role in plant pollination, including various domestic crops (Aizen et al. 2009). The well-publicized mortality of honey bee populations recorded in the last decade (Bacandritsos et al. 2010; Johnson et al. 2010) predicted a looming pollination crisis that may threaten worldwide food security (Withgott 1999; Kremen and Ricketts 2000). This is especially true since the value of crops pollinated by bees was estimated at US$14.6 billion in the USA alone in 2000 (Morse and Calderone 2000).

Despite the ongoing debates regarding the causes of honey bee decline around the world, a majority of the scientific community considers this decline to be a multifactorial phenomenon (VanEngelsdorp et al. 2010; Wu et al. 2012). This means that no single factor—such as pathogens and parasites (Fries 2010; Dainat et al. 2012), insufficient experience in honey bee hive management, or excessive use of pesticides (Johnson et al. 2010; Van der Sluijs et al. 2013)—is individually causing honey bee decline or mortality, but the contribution of various factors acting together are responsible.

Some widely used pesticides, such as neonicotinoids, are known to be extremely toxic for bees (Williamson et al. 2014; Chakrabarti et al. 2015) and can, even at sublethal doses, significantly decrease honey bee performance such as brood production, weight gain, disease resistance, and trigger disorders in colony dynamics and labor partition (Mackenzie and Winston 1989). These insecticides are valuable for pest control in agriculture, increasing the crop production and providing a worldwide food security (Potts et al. 2010; Godfray et al. 2014). Pesticide residues at various concentrations were identified in honey bee products and hives (Lambert et al. 2013).

The equivocal results obtained from studies on honey bee in-field pesticide toxicity and exposure, tolerated level of exposure, and effect on honey bee health (Blacquière et al. 2012; Henry et al. 2015) have raised new questions and debates. It is believed that in order to ameliorate these problems and find sustainable solutions, closer collaborations and information exchange between pesticides companies, the scientific community, agriculturists, and beekeepers is strongly needed.

In this study, we are particularly interested in testing whether the use of agricultural pesticides has in-field impact on forager bees’ activity and survival. Forager bees are more susceptible to in-field exposure to pesticides during their foraging flights compared to in-hive nurse bees. It has been shown that forager behavior, orientation, communication dances, and return flights are all highly affected by sublethal insecticide doses (Vandame et al. 1995; Williamson et al. 2014). Neonicotinoids, in particular, can impair the olfactory memory and learning capacity of honey bees (Decourtye et al. 2005; Aliouane et al. 2009; Williamson and Wright 2013; Williamson et al. 2014) and alter the homing behavior of forager bees (Fischer et al. 2014). Besides tracking in-field forager bee mortality in our current study, we have tested both recovered foragers from the field and bees that were fed imidacloprid in the lab to determine the variability in the activity of five genes (AChE, CAT, GstD1, GstS3, and GstS4) as potential biomarkers of exposure to oxidative stressors (Lionetto et al. 2003; Durou et al. 2007). Some of these genes are part of the known antioxidant gene families in honey bees (Corona and Robinson 2006), while others are known to be linked to pesticide sublethal toxicity in both honey bees (Boily et al. 2013; Alburaki et al. 2015) and bumble bees (Samson-Robert et al. 2015).

2 Materials and Methods

2.1 Honey bee colonies

This experiment was conducted on four honey bee colonies. In May 2015, four new colony divisions, equal in size, were made from existing, overwintered colonies of standard stock. Four sister Carniolan (Apis mellifera carnica) queens, artificially inseminated, were purchased from a commercial queen producer at the same period and introduced into these four new colonies. The new queens and divisions were housed in new ten-frame Langstroth hives, and a board (1.5 × 1.5 m) was placed in front of each hive to help observe and collect any bee mortality at the front entrance. When an unusual number of dead bees (∼50 or more) were observed on those boards, dead bees were collected and sent for pesticide residue analysis as described below.

2.2 Location and landscape

In order to assess forager survival and/or stress related to probable pesticide in-field exposure, four different landscape locations were carefully chosen. First, candidate locations were selected, and geographical information system (GIS) studies were conducted on a 2.5-km radius from each candidate location. The GIS analysis was conducted in order to assess the agricultural (AG) areas or crop fields available for honey bees within a typical foraging distance for honey bees (Seeley 2010). The GIS study was performed using Esri ArcGIS® software (Redlands, CA).

2.3 Marked and recovered bees

One beehive was assigned to provide a source capped brood in our studies. This assigned hive was not part of this study and served only as a source for capped brood frames. One day prior to the date of introducing bees into our experimental hives, three capped worker brood frames, which were close to worker eclosion, were kept for 24 h in an incubator at 35 °C and 53% relative humidity. One-day-old bees that hatched the next day were randomly divided into groups of 50–60 bees, and they were collectively marked by a dot on their dorsal tergite using a collective marking box designed after (Alburaki 1990).

Three hundred marked bees were introduced in each of the four experimental hives on four different dates (Table I). In addition, a control set of 300 1-day-old marked bees were randomly selected on each marking date and stored at −80 °C. Introduced bees were left in hives for 20–22 days, except for the first date (42 days). Marked bees were then manually recovered from the hives and killed on dry ice and stored at −80 °C (Table I). Heads of both recovered bees (mostly foragers) and the control bees were subsequently dissected and separated for subsequent gene expression analyses, while the rest of their bodies were sent for pesticide residue analysis.

Table I Timetable and results of the main experimental procedures of the field study, showing the introduction and collection dates of the marked honey bees
Full size table

2.4 Honey bee samples fed in the lab

One-day-old bees were used in three different treatment groups (B, C, D) and a control group (A) as described in Table II. Capped brood frames were taken from three hives and placed in an incubator at 35 °C and 53% RH, and monitored for 48 h. During that time, emerging bees were placed in Plexiglas® cages (100 bees/cage) and fed ad libitum with water, a protein patty, and sugar syrup. The sugar syrup provided for cages B, C, and D contained, respectively, different imidacloprid concentrations of 5, 20, and 100 ng/g. Imidacloprid was chosen as being the largest used insecticide in the world (Jeschke et al. 2011; Stoner and Eitzer 2012). Honey bees of the control group (A) were fed with regular sugar syrup that contained no imidacloprid (Table II). At the end of the experiment (7 weeks), 40 live bees were sampled from each group, killed on dry ice, and stored at −80 °C for further molecular analysis.

Table II Laboratory feeding procedures conducted on the four caged honey bee groups (three treatments and one control)
Full size table

2.5 Landscape analysis

In order to test the presence or absence of pesticide residues that forager bees may have encountered in the major available crops, whole flowers of the surrounding crops were sampled for chemical analysis. During the flowering period of each crop (corn, cotton, soybean, and sorghum), flowers from the closest four to six major crop fields surrounding each location were randomly sampled at 15–20 spots per field (Table III). Flower samples of each crop type and location were pooled and stored at −80 °C for chemical analysis. Honey bee and vegetation samples were shipped on dry ice for pesticide residue detection.

Table III Background, location, and distribution of the experimental colonies
Full size table

2.6 Visited crops and flowers

Marked foragers of each location are expected to forage in the surrounding crop fields and gather pollen and nectar from the available sources. Nevertheless, in order to precisely determine the visited crops and flowers without perturbing the experimental hives with pollen traps, pollen were trapped at eight time points from neighbor hives in each apiary and were subsequently identified.

2.7 Gene expression

Total RNA was extracted from the recovered foragers’ brain using TRIzol® Reagent protocol from Invitrogen (Chomczynski 1993) with some modifications. Brains of 20 or 10 foragers that were randomly selected from each sample were removed from dry ice, rapidly dissected, and immediately added separately to 1 mL TRIzol with 5 mg of acid-washed glass beads and gently mixed for 2 min. Two hundred microliters of chloroform was added, and the total mixture was incubated at room temperature for 15 min followed by a centrifugation at 10,200 rpm for 15 min at 4 °C.

Three hundred microliters of the supernatant was transferred to a fresh tube, and the tube was washed with 500 μL each of isopropanol and incubated for 15 min at room temperature, followed by centrifugation at 10,200 rpm for 15 min at 4 °C. The pellet was subsequently washed with 1 mL 75% ethanol and centrifuged at 10,200 rpm for 15 min at 4 °C. Finally, the RNA pellet was well dried and 60 μL of nuclease-free water was added. RNA extractions were nanodropped (Thermo Scientific NanoDrop 2000/2000c Spectrophotometers) for RNA quantity and quality and were diluted to 200 ng/μL and stored at −80 °C.

Two-step reverse transcription quantitative PCR (RT-qPCR) was used to quantify the genetic expression of five different genes in the bee brains. One microgram of RNA was used as a template for cDNA synthesis using BioBasic High Reverse Transcriptase kits and random hexamer primers. RT-qPCR was performed, in triplicate, on a BioRad CFX384 using LifeTechnologies PowerUP SYBR Green master mix. The studied genes were as follows: (1) acetylcholinesterase (AChE), (2) catalase (CAT), (3) glutathione-S-transferase D1 (GstD1), (4) glutathione-S-transferase S3 (GstS3), and (5) glutathione-S-transferase S4 (GstS4) (Corona and Robinson 2006; Williamson et al. 2013; Alburaki et al. 2015). Primers were designed based on the full sequences of these genes available from the NCBI database under accession numbers: (1) AB181702.1, (2) NM001178069.1, (3) NM001178028.1, (4) XM006572044.1, and (5) NM001142656.2. The studied genes were normalized using GeNorm (Vandesompele et al. 2002) in all the RT-qPCR runs using a set of five reference genes (28S, Actin, CamIIk, GAPDH, and E2F) known to be accurate and stable in honey bee tissues (Scharlaken et al. 2008). Primers’ sequences of both reference and studied genes are available in the “Data accessibility” section.

2.8 Detection of pesticide residues

Pesticide residues were quantified in the recovered foragers as well as in the flowers of the corn, cotton, soybean, and sorghum using liquid chromatography-mass spectrometry (LC-MS) (Barnett et al. 2007; Walorczyk and Gnusowski 2009). All the chemical analyses for pesticide residue detection were processed at the USDA National Scientific Laboratories in Gastonia, NC. A comprehensive chemical analysis that included 174 chemical active substances was run for each sample, and positive results were reported in the text. Complete analytical reports can be found in the DOI in the “Data accessibility” section.

2.9 Statistical analysis

Statistical analyses and figure generation were carried out and generated in the R environment (R Core Team 2011). Variables of this study included (1) number of recovered foragers (forager mortality); (2) AChE, CAT, GstD1, GstS3, and GstS4 expression; and (3) percentage of AG in the total foraging area. Data were treated either per location (four groups) to study the landscape effects on honey bee foragers or per AG area (two groups) to explore the putative impacts of the exposure to pesticides on the honey bee colonies. Variables were first tested for normality using the Shapiro-Wilk test and were normalized using a log transformation. When a simple comparison of two variables was needed, the Welsh two-sample t test was used at a 95% confidence interval. Analysis of variance (ANOVA) was also performed to quantify the difference between variables regarding the treatment and time at a 95% confidence level. Principal component analyses (PCAs) were carried out using “Devtools” Package as well as “FactoMine” to calculate the percentage of the variability expressed in three-dimensional space. Correlations between genetic expressions of the studied genes for both caged bees and field-recovered foragers were performed using the R libraries “Performance Analytics” and “Corrplot.”

3 Results

3.1 Landscape study

The locations and their agricultural classification based on the GIS were as follows: Jackson (low AG area with urban activity), Milan (high AG area), Yum-Yum (very high AG area), and Chickasaw (a natural park that contains no agricultural activity; non-AG area) (Figure 1, Table III). Within a 2.5-km-radius foraging distance, honey bees had access to each location (Jackson, Milan, Yum-Yum, and Chickasaw) to 19, 55, 71, and 5% agricultural area or crop fields, respectively (Figure 1). The remaining landscapes consisted of forest, woodland, open water, and urban areas (such as buildings and roads). Therefore, the highest AG area was located in Yum-Yum (71%), followed by Milan (55%). The Jackson location was a relatively low AG area (19%) but had a substantial urban component (46%). Chickasaw was considered the control treatment of this study with only 5% AG area. In some cases, and in order to evaluate the total exposure to pesticides, results were exposed by two landscape groups: AG and non-AG groups.

Figure 1.
figure 1

Geographical location of the four studied colonies and locations in western Tennessee, USA. The GIS study to determine the landscape nature at each location was conducted on a 2.5-km radius. The landscape classifications included Jackson (low AG area with human use, 19%), Milan (high AG area, 55%), Yum-Yum (very high AG area, 71%), and Chickasaw (non-AG area, 5%). One honey bee colony was placed in each location.

Full size image

3.2 Recovered foragers

Recovered forager and/or bees varied among dates and locations (Table II). On the first collection date (24 July), when marked bees were left for 42 days in their respective hives, no marked bees were recovered except for the Chickasaw location where only 16 bees were recovered (Table I, Figure 2). The highest number of recovered bees was observed in Chickasaw on the last date (16 September), in which 275 of 300 bees (91.6%) were recovered (Table I). No significant differences were observed in the total recovered bees whether per locations (Jackson, Milan, Yum-Yum, Chickasaw) or per treatment (AG area and non-AG area) (Figure 2). However, regardless the treatment factor, there was a large difference in the number of recovered bees among dates (P < 0.001).

Figure 2.
figure 2

Percentage of the recovered honey bee foragers per date from each location (Jackson, Milan, Yum-Yum, and Chickasaw) as well as the total number of overtime-recovered foragers per location and landscape (AG and non-AG areas).

Full size image

3.3 Gene expression

Different gene expression patterns were identified in recovered foragers’ brains from each location for each studied gene (Figure 3). No significant differences were recorded for the first three genes (AChE, CAT, GstD1) among bees of the studied locations or between the studied locations and control bees (CT). However, the expression of both GstS3 and GstS4 were significantly higher (P < 0.05) between CT and both C and M for GstS3 as well as between CT and all locations (C, J, M, and Y) (P < 0.001) for GstS4. AChE expression negatively correlated with both CAT (r = −0.84) and GstS4 (r = −0.88) and positively with GstS3 (r = 0.94) (Figure 3). For bees fed in the lab with sugar water tainted with imidacloprid, none of the five genes showed significant differences among the four groups of bees (A, B, C, and D), except for GstD1 (P < 0.05) with no paired-groups significance (Figure 4). Principal component analyses showed differences among recovered foragers from the four locations on axes 1 and 2 (Figure 5) with respect to the five studied genes. The variability expressed on both axes 1 and 2 were 45.95 and 30.60%, respectively (Figure 5). Samples from Milan and Jackson showed more resemblance to each other than to bees from the other locations (Chickasaw and Yum-Yum) with respect to the gene expression of the five studied genes.

Figure 3.
figure 3

Gene expression results of the five studied genes (AChE, CAT, GstD1, GstS3, and GstS4) performed by RT-qPCR on the recovered forgers’ brains. Codes of X-axis are as follows: “C,” Chickasaw; “CT,” control; “J,” Jackson; “M,” Milan; and “Y,” Yum-Yum. P values are *P < 0.05, ***P < 0.001 and are calculated based on ANOVA for overtime expression at a 95% confidential level. The correlation matrix with the correlation coefficient values of the five studied genes are also provided.

Full size image
Figure 4.
figure 4

Gene expression results of the five studied genes (AChE, CAT, GstD1, GstS3, and GstS4) performed by RT-qPCR on honey bee brains fed in the lab with different imidacloprid concentrations. A, B, C, and D are the four studied groups detailed in Table III. ANOVA shows no significant differences among the studied groups for any of the studied gene except for GstD1 (P < 0.05). The correlation matrix shows no significant correlation among the studied genes.

Full size image
Figure 5.
figure 5

Principal component analysis (PCA) of the five studied genes of the recovered foragers expressed on axes 1, 2, and 3. Individuals and variable factor maps are provided along with the variable percentages expressed on the first two axes. Differentiation of three distinct groups (AG and non-AG areas) is clarified with eclipses on the PCA individual factor map. Axes 1 and 2, respectively, absorbed 46.0% and 30.6% of the dataset variability.

Full size image

3.4 Pesticide residues

Recovered foragers

The results of the recovered foragers showed some residues of different pesticides (Table IV). On the first collected date, neonicotinoids were only identified on foragers at the Jackson location (1 ng/g imidacloprid). Milan’s foragers contained 10.6 ng/g azoxytrobin, a fungicide, on the same collected date. Cyhalothrin (5.4 ng/g), a synthetic pyrethroid insecticide, was identified on foragers of Yum-Yum on the second date (9 September). No pesticides were detected in Chickasaw foragers (non-AG area) at any time or any forager on the third sampling date (October 6) (Table IV).

Table IV Results of the pesticide residue detection performed by LC-MS for recovered foragers of each location, control bees, flowers of corn, soybean, cotton, and sorghum
Full size table

Crop flowers

Various pesticide residues were found at different concentration in crop flowers collected from crop fields surrounding our apiaries. High concentrations of neonicotinoids were recorded on cotton flowers at both the Milan and Yum-Yum locations at 25.3 ng/g (imidacloprid) and 23.7 ng/g (thiamethoxam), respectively (Table IV). Significant concentrations of acephate on cotton were detected in Jackson (309 ng/g) and Milan (4190 ng/g) as well as other insecticides such as bifenthrin and methamidophos (Table IV). Concentrations of imidacloprid (5.3 and 2.4 ng/g) were also recorded on the soybean flowers of Jackson and Yum-Yum, respectively. On sorghum, various concentrations of cyhalothrin, bifenthrin, and spinosad (all insecticides) were identified at the Jackson and Milan locations. Other pesticides detected at low concentrations are detailed in Table IV. Identification of the trapped pollen collected at eight different time points from hives neighboring our experimental colonies clearly evidenced that foragers intensively visited and collected soybean, sorghum, and corn pollen (data not shown).

4 Discussion

This study examined how agricultural pesticides used extensively in southern US field crops may contribute to in-field forager mortality of honey bees. The activities of the AChE, a neural enzyme, along with three other detoxification enzymes (GstD1, GstS3, and GstS4) and the CAT, were studied in bee brains in order to characterize physiological stresses resulting from exposure to pesticides. The four experimental locations (Jackson, Milan, Yum-Yum, and Chickasaw) were carefully chosen based on landscape structures and potential exposure to pesticides. As organic crop production is rare in western Tennessee, a non-cropping location (Chickasaw) was placed in a non-AG area in a Tennessee state park where little or no agriculture occurred within bee foraging distance.

On the first collection date (24 July), when bees were left in-hive for 42 days, only 16 marked foragers were recovered from Chickasaw and none were recovered from the other locations. In order to increase the number of the recovered foragers from all locations, we shortened the period between introduction and collection of bees to 20–22 days (Table I). Although the number of foragers that were recovered varied among locations, no significant differences were observed among locations or between treatment categories (AG and non-AG areas) (Figure 2). This indicates that foraging in AG or non-AG fields did not have significant impact on in-field forager mortality. This conclusion is encouraging to beekeepers but might not be valid under all environmental conditions. Nonetheless, the concentration of some insecticides found on the flowers of the measured crops indicate that foraging honey bees could be exposed to high concentrations of those pesticides (Table IV). For instance, concentrations of imidacloprid and thiamethoxam (25.3 and 23.7 ng/g) that theoretically exceed the bee oral LD50 were identified in cotton flowers of Milan and Yum-Yum locations, respectively, whereas no neonicotinoids were detected in Jackson’s cotton flowers (Table IV). Acephate was also recorded at elevated concentrations in the cotton flowers of Jackson (309 ng/g) and extremely high concentrations (4190 ng/g) in Yum-Yum. Other various concentrations of insecticides were detected in the cotton and sorghum flowers of the three AG locations. Low concentrations of imidacloprid were reported on soybean flowers of both Jackson and Yum-Yum locations, while no neonicotinoids were found in corn and sorghum flowers (Table IV). The neonicotinoid residues in our study are much higher than what has been reported in a recent study conducted in western Tennessee (Stewart et al. 2014), but this study was targeting seed treatment effects whereas our current study would be influenced by foliar-applied insecticides. The recovered foragers seemed not to have encountered the pesticides detected in the crops’ flowers, as no high concentrations were detected in those bees (Table IV). The absence of significant pesticide residues in the recovered foragers likely explains the similarity in foragers’ mortality between AG and non-AG areas, as no link was established between the foragers and the elevated concentrations of pesticides found in the crop flowers (Figure 2).

For recovered foragers, none of the antioxidant genes showed significant differences among the locations, which suggests that the foraging environment (as quantified as percentage AG area) is not involved in enhancing honey bee physiological stress or the detoxification process. Interestingly, expression of both the GstS3 and GstS4 genes were higher in control bees that were killed at 1-day-old (Figure 3). This finding leads to conclude that the expression of the GST genes are most probably related to the bee age, as described in (Słowińska et al. 2015), rather than exposure to pesticides (Figure 3).

In order to assess any interference that might have biased the gene expressions of the recovered foragers (e.g., environmental conditions), we tested these genes on bees fed with imidacloprid in the lab. Surprisingly, none of those genes including the AChE significantly differed from the control group (A) that was not fed with imidacloprid (Figure 4 and Table II). On the one hand, the genes of the GST family are known to be upregulated in response to abiotic stresses, such as cold, heat, and insecticides (Yan et al. 2013). On the other hand, AChE response to imidacloprid concentrations are similar to what has been found by Badiou-Bénéteau et al. (2012), in which bees fed in the lab with thiamethoxam lethal doses showed no AChE response, although the same authors suggested that AChE expression could be considered a robust biomarker for other insecticide exposure in living bees (Badiou et al. 2008). Previous field and laboratory studies investigating the neonicotinoid impact on honey bee survival (more precisely on the AChE activity) reported elevated levels of AChE activity in honey bee brains when subjected to thiamethoxam or clothianidin (Boily et al. 2013; Alburaki et al. 2015). Our contradictory results regarding the relationship between AChE activity and exposure to pesticides suggest a complex physiological response of honey bees to abiotic stressors. Thus, beside the level of expression, some other key factors such as the slope of the response and the short-term and long-term response magnitudes to stressors, should be considered to accurately quantify gene expression (Chechik and Koller 2009).

The overall expression of the five studied genes on the recovered foragers showed an interesting separation of the three groups of locations (Figure 5). The highest similarity in gene response was found to be between foragers of the Milan and Jackson locations. Chickasaw foragers (non-AG area), however, are clearly well separated from both other groups as well as Yum-Yum foragers (AG area) (Figure 5). It is hard to conclude at this point that these differences were a result of different treatment factors, especially because low pesticide levels were detected in recovered foragers (Table IV and Figure 5). Our analyses suggest that a wise option would be to select candidate groups of genes that exhibit similar response to particular molecules in order to find more appropriate honey bee biomarkers for exposure to pesticides.