Exposure to low doses of pesticides induces an immune response and the production of nitric oxide in honeybees

Honeybees are essential pollinators of many agricultural crops and wild plants. However, the number of managed bee colonies has declined in some regions of the world over the last few decades, probably caused by a combination of factors including parasites, pathogens and pesticides. Exposure to these diverse biotic and abiotic stressors is likely to trigger immune responses and stress pathways that affect the health of individual honeybees and hence their contribution to colony survival. We therefore investigated the effects of an orally administered bacterial pathogen (Pseudomonas entomophila) and low-dose xenobiotic pesticides on honeybee survival and intestinal immune responses. We observed stressor-dependent effects on the mean lifespan, along with the induction of genes encoding the antimicrobial peptide abaecin and the detoxification factor cytochrome P450 monooxygenase CYP9E2. The pesticides also triggered the immediate induction of a nitric oxide synthase gene followed by the delayed upregulation of catalase, which was not observed in response to the pathogen. Honeybees therefore appear to produce nitric oxide as a specific defense response when exposed to xenobiotic stimuli. The immunity-related and stress-response genes we tested may provide useful stressor-dependent markers for ecotoxicological assessment in honeybee colonies.


Results
The lifespan of honeybees is affected in a stressor-dependent manner. We investigated the effect of entomopathogens and pesticides on the lifespan of honeybees by supplementing the standard feeding solution with the bacterium Pseudomonas entomophila, the herbicide pendimethalin, the insecticide thiacloprid, or one of two fungicides (fludioxonil and dimoxystrobin). Honeybees orally infected with P. entomophila generally died sooner than controls fed on sugar syrup (Fig. 1). In six of the nine assays, we observed significant differences between the infected group and control (Fig. S1a) with mean LT 50 values of 10 and 20 days, respectively (Fig. 1a). A similar effect was observed for honeybees exposed to pendimethalin, with mean LT 50 values of 10 days in the exposed group compared to 23 days in the control (Fig. 1b). However, some exposed individuals significantly outlived the controls, surviving up to 31 days after treatment. As expected, thiacloprid also reduced the honeybee lifespan, with LT 50 values of 14 days compared to 20 days for the controls (Fig. 1c). In contrast, exposure to the non-systemic fungicide fludioxonil had a complex effect, with some assays indicating that the treated group survived for a significantly shorter period (LT 50 = 13 days) than controls (LT 50 = 18 days) and others showing a significantly longer survival period (LT 50 = 27 days) than controls (Fig. S1d), leading to a non-significant difference in the LT 50 between the fludioxonil-treated and control groups overall (Fig. 1d). Interestingly, the strobilurin fungicide dimoxystrobin appeared to extend the honeybee lifespan from a mean LT 50 of 15 days in the control to 21 days in the treated group (Fig. 1e).
Immunity-related and stress-related genes are induced by pesticide challenge. To evaluate the effects of low pesticide doses on the immunity and stress responses of honeybees, we analyzed the expression of 17 marker genes encoding AMPs, detoxification enzymes and redox factors by qRT-PCR, using samples of bee gut tissues dissected 1, 3, 6, 24 and 48 h after pesticide ingestion. Similarly, marker gene expression was tested 1, 3, 6 and 24 h post-infection with P. entomophila (Fig. 2). We observed the induction of AMPs (abaecin, apisimin, defensins 1 and 2, and hymenoptaecin) in response to most of the stressors at most of the sampling time points. Although the overall induction was weak to moderately high (1.5-fold to 100-fold), the gene encoding the AMP abaecin was consistently and strongly induced (> 10,000-fold) in response to the bacteria and the pesticides, peaking at the early time points (1-3 h). The gene encoding the AMP hymenoptaecin was also induced (1.5fold to 1000-fold), with the strongest induction (up to 1000-fold) in response to thiacloprid and P. entomophila. Interestingly, a third AMP gene (encoding apisimin) was repressed or weakly induced (1.5-fold to tenfold) in response to P. entomophila but moderately induced (≤ 1000-fold) in response to pendimethalin at the earlier time points. The inhibitory Toll pathway gene cactus-2 showed weak induction at some individual time points in response to the pesticides, mostly with values below 1.5-fold. However, cactus-2 was weakly (≤ tenfold) but consistently induced throughout the time course in response to P. entomophila.
Several CYP genes were weakly (≤ tenfold) or moderately induced (≤ 1000-fold), but cyp9e2 was strongly induced (up to > 10,000-fold) by P. entomophila and the pesticides at the early time points, indicating a role in the immediate response to these stressors. Two genes encoding UDP-glucuronosyltransferases (UGTs) were strongly induced (≤ 10,000-fold) by P. entomophila but only moderately induced (generally ≤ 1,000-fold) by the pesticides. The hopscotch gene encoding a tyrosine kinase in the JAK/STAT pathway was only weakly induced (≤ tenfold) regardless of the treatment. The Nos gene was moderately (≤ 1000-fold) or strongly (≤ 10,000-fold) induced by all the pesticides after 1 h, but only weakly (≤ tenfold) induced in response to P. entomophila. Similarly, the gene encoding catalase was strongly upregulated (> 10,000-fold) after 6 h, but only in response to the pesticides. The Duox gene encoding dual oxidase was minimally induced by all the stressors.
The analysis of gene expression therefore revealed three sets of genes strongly induced by the stressors we tested-one set of genes (principally abaecin and cyp9e2) induced by the pesticides and the bacterial pathogen, another (principally the UGT genes) induced strongly by the pathogen but only weakly or moderately by the pesticides, and a third (principally Nos and catalase) induced by the pesticides alone, with a significant delay between the immediate expression of Nos and the subsequent expression of catalase.
Free radicals show distinct accumulation profiles in the honeybee gut. The almost immediate upregulation of Nos by abiotic stress followed by the delayed induction of catalase suggested that free radi- www.nature.com/scientificreports/ cals may generally be involved in the stress response to the pesticide treatments. We therefore used a general indicator for oxidative stress, CM-H 2 DCFDA, to quantify the free radical levels at four time points after each pesticide treatment (and following exposure to P. entomophila) compared to untreated controls. After 1 h, the fluorescence signal was lower than the control in all five treatments, significantly so in the case of fludioxonil and pendimethalin (Fig. 3a). After 3 h, the fluorescence signal was significantly higher in the P. entomophila and pendimethalin treatments compared to the control, and we observed a smaller but still significant increase for the fludioxonil treatment (Fig. 3b). After 6 h, we observed no significant difference between the five treatments and untreated control (Fig. 3c). After 24 h, we once again observed significant differences between the control and the treatments with P. entomophila, thiacloprid and pendimethalin (Fig. 3d).

Discussion
The impact of various biotic and abiotic stressors on honeybees is not well understood. We therefore tested the effect of one entomopathogen and low doses of four pesticides on honeybee survival as well as the expression of immunity-related and stress-response genes. Accordingly, we exposed adult worker bees orally to the entomopathogenic bacterium P. entomophila, the fungicides fludioxonil and dimoxystrobin, the herbicide pendimethalin, and the insecticidal neonicotinoid thiacloprid. We found that survival depended on the individual stressors, but that the effect of the treatment was not always the same in replicate tests, arguably reflecting the different ages of the individual bees obtained from the www.nature.com/scientificreports/ same hive [42][43][44][45] . However, using bees from a single hive reduced the variability of the assays because the individuals faced the same environmental influences and shared the same resources, in contrast to bees from different hives. Nevertheless, the honeybee lifespan was reduced by the insecticide thiacloprid and, in the majority of tests, by the entomopathogenic bacterium and the herbicide pendimethalin. This is supported by previous ecotoxicological studies that showed a moderate to low effect of pendimethalin on honeybee health 46 . Interestingly, dimoxystrobin generally increased the honeybee lifespan, whereas fludioxonil showed different effects in different test replicates and the overall effect on survival was nonsignificant. We did not control for behavioral adaption regarding food selection and quality, so fungicide intoxication may have been compensated by natural detoxification 30 , or whether the reproduction compensated for individual losses, as previously suggested 47 . The oral route was used in this study to expose honeybees to low doses of each xenobiotic, resulting in pesticide concentrations similar to those previously found as residues in bee bread. Therefore, the shorter lifespan of honeybees exposed to pendimethalin and thiacloprid was not surprising, nor was it necessarily anticipated. Although thiacloprid targets the nicotinic acetylcholine receptor 48 , the effect of low doses on honeybee survival is inconclusive 24,[49][50][51] . In contrast, the extended lifespan of insects in response to low doses of toxins and free radicals has been described as an evolutionary adaption (hormesis), boosting the expression of genes that protect cells from stress 52 . This may have contributed to the prolonged lifespan following exposure to dimoxystrobin. Despite the increased mortality caused by exposure to fungicides combined with insecticides 53 , the effects of single fungicides, including dimoxystrobin, on bee mortality have not been characterized in detail. The analysis of gene expression levels provided insight into the effects of the entomopathogen and different xenobiotic stressors on the honeybee immune system. P. entomophila has previously been shown to induce an immune response in honeybees under laboratory conditions 54 . As expected, the bacterium was well suited as a representative biotic stressor, not only reducing the honeybee lifespan but also inducing many of the immunityrelated and stress-dependent marker genes. P. entomophila induced most of the AMP genes, as well as cyp9e2 and both UGT-2C1 and UGT-2B15, but the redox-defense genes Nos and Duox, or the catalase gene were only weakly upregulated, even though the corresponding products are essential for antimicrobial activity 55 . The response to the entomopathogen differed qualitatively from the response to the pesticides, but all four pesticides induced similar responses indicating the activation of a conserved mechanism to counter the stress imposed by xenobiotics. We observed the strong induction of genes encoding the AMP abaecin, CYP9E2, NOS and catalase. The hymenoptaecin gene was strongly induced by P. entomophila and the insecticide and to a lesser extent by the other pesticides. Abaecin and hymenoptaecin were previously shown to work synergistically, with the combined antibacterial activity greater than the sum of each component's activity when presented alone 35 . This may indicate a specific synergistic response to thiacloprid and P. entomophila, although the strong expression of abaecin in response to all treatments suggests that abaecin may play a universal, stressor-independent role in defense. The two main functions of AMPs are the recognition of pathogens via PAMPs such as LPS and peptidoglycans, and the metabolism of xenobiotics 56 . The stressor-independent induction of abaecin suggests that this AMP is involved in both activities.
Invertebrate humoral defense involves stressor recognition followed by elimination, facilitated by the activation of AMPs and the production of toxic superoxide anions and hydrogen peroxide 32,57 . Although the production and segregation of ROS and RNS primarily involves the hemocytes and fat body 58 , these reactive species are also known to confer antimicrobial activity in the gut epithelium 32,59 . Interestingly, Duox was only moderately upregulated in the gut (if at all) regardless of the stressor. In D. melanogaster, dual oxygenase is the most important factor in the initiation of an immune response against invading microbes 60,61 , and the neonicotinoid imidacloprid specifically interferes with this pathway 62 . In contrast, we found that Nos expression was strongly and immediately induced in response to the pesticides, peaking within 1-3 h in most cases. In the case of thiacloprid exposure, even stronger Nos induction was detected after 6 h, correlating with the catalase expression peak, and possibly indicating the specificity (hence higher toxicity) of the insecticide. The defense against xenobiotics therefore appears to activate RNS rather than ROS. Highly-reactive NO, produced by the oxidation of arginine to citrulline by NOS 63 , is considered a key effector in the defense responses of invertebrates by interacting with ROS such as superoxide anions and hydrogen peroxide 59 , as well as signaling for the induction of AMPs 64,65 . ROS and RNS intermediates react to form other cytotoxic compounds such as peroxynitrite with a synergistic mode of action 38,66 . Although the fluorescent dye CM-H 2 DCFDA generally indicated oxidative stress with the moderate accumulation of ROS after 3 h, the potential contribution of the gut microbiome cannot be ruled out, and the specific reactive molecules could not be identified. Further experiments are required to specifically detect the nitrogen-derived compounds we assume are responsible for the observed effect. The weak induction of Nos and Duox by the entomopathogen P. entomophila aligns with previous reports showing that this bacterium can inhibit Duox expression 54 , possibly reflecting an evolutionary strategy to inhibit ROS production based on uracil sensing 67 . It is unclear whether P. entomophila achieves the suppression of insect defenses by directly modulating redox-related genes that were not tested in our experiments, or indirectly by, for example, influencing the composition of the gut microbiome. Nevertheless, the upregulation of CYP and UGT genes suggests that a strong detoxification response is induced by the entomopathogen, indicating the presence of free radicals. Despite the weak upregulation of Nos, the production of NO by hemocytes to facilitate an immune response in the gut is nevertheless possible 68,69 .
ROS and RNS are useful as an immediate response against stressors, but their persistence is likely to damage host cells 70 . Accordingly, they are removed by protective antioxidant enzymes such as catalase and detoxification enzymes such as CYPs and UGTs, which bind molecular oxygen and other cytotoxic compounds to directly form non-toxic water or water-soluble products in insects and mammals 41,71 . We observed the strong upregulation of catalase following the induction of Nos, suggesting the role of catalase is to clear up RNS produced by NOS. Further experiments are required to measure free radical levels and to determine which compounds are removed by catalase. The catalase gene was moderately upregulated in response to P. entomophila, coinciding www.nature.com/scientificreports/ with the minimal induction of Nos and Duox. In contrast, UGT-2C1 and UGT-2B15 were induced more strongly by the entomopathogen than the pesticides. Most of the CYP genes were moderately induced, but cyp9e2 was upregulated 100-10,000-fold after 1 h for all stressors, indicating a detoxification function that is not restricted to bacterial infections 72 . Indeed, CYP9E2 has been shown to metabolize thiacloprid efficiently in honeybees, whereas other CYPs cannot fulfil this function 73 . Moreover, a recent analysis of the sublethal effects of air pollution, a chemically complex stressor, also showed that cyp9e2 was strongly upregulated in honeybee heart tissue 74 .
Our data indicate that biotic and abiotic stressors induce the preferred expression of genes encoding UGTs and catalase, respectively, whereas CYP9E2 appears to fulfil a universal detoxification function. The ability of dimoxystrobin to disrupt the mitochondrial respiratory chain in fungi may explain the particularly strong induction of cyp9e2 (> 10,000-fold) by this xenobiotic. Interestingly, our gene expression data were not homogeneous at the various sampling time points, possibly reflecting the widely spaced sampling intervals but also the fact that our insects were collected from a working hive rather than synchronously bred in the laboratory. Age and developmental stage may influence the potency of stress responses and immunity in bees 43,44 . Regardless of the stressor, the primary stress response of eukaryotic cells relies on the immediate activation of defense signaling molecules such as ROS [75][76][77] . However, the production of these volatile compounds in response to xenobiotics, followed by their elimination, is a dynamic cell state that may also explain the results of our time course experiments. In addition, our gene expression data clearly indicate the induction of AMP genes in response to stress. In contrast to the increase in AMP gene expression we observed, previous studies in honeybees and masonbees showed that low doses of neonicotinoids cause the depletion of hemocytes, resulting in limited antimicrobial activity [78][79][80] . Moreover, we did not observe significant upregulation (> 1-10 for only single timepoints) of the Toll inhibitor cactus-2 following xenobiotic exposure. Assumably, the weak induction still allowed an active immune response, a hypothesis that would require further studies with Toll pathway components. In contrast, RNA interference experiments have shown that clothianidin induces relish, encoding an inhibitor of the NF-κB signaling pathway, thus promoting the replication of DWV 81 . All these studies involved animals under controlled laboratory conditions, undoubtedly facilitating the collection of more homogenous data, but ruling out the environment as a potential factor responsible for the "natural priming" of the immune system. In contrast, we administered different stressors that, with the exception of thiacloprid, do not primarily target insects, using individuals collected directly from one hive. These insects therefore shared the same resources and were exposed to the same environmental conditions. Different studies of the same stressors have often reached dissimilar conclusions, probably influenced by factors that are not controlled in the experiment, such as the age of individual insects 82 . Similarly, the level of toxin residues in bees exposed to imidacloprid differed greatly between animals that were fed in groups of 10 and those fed individually 83 . Such deviations may explain the heterogeneity in some of our results (e.g., the replicates in the survival experiments) and confirm the significant effect of environmental conditions on honeybee health. Honeybees are continuously exposed to multiple anthropogenic stressors as well as natural changes that affect their resistance in ways that are not fully understood 5,84 .
We have shown that P. entomophila (a biotic stressor) and low doses of abiotic stressors induce specific immunity-related and stress-response genes in the intestine of adult honeybees, whereas other key candidate genes such as Duox and the JAK/STAT pathway component hopscotch do not appear to be involved. The key players in the detoxification of stressors that are not designed specifically to target honeybees are NOS, which promotes the release of NO, and the AMP abaecin. We also found that the entomopathogen P. entomophila strongly induces genes encoding UGTs but induces Duox only weakly. Our results therefore indicate that the intestinal immune response can differentiate between biotic and abiotic stressors, but distinct xenobiotic stressors appear to trigger the same host responses. Our study helps to unravel the molecular mechanisms underlying xenobiotic stress responses in the honeybee gut, revealing the activation of NOS in the intestinal epithelium, in turn triggering a NO-mediated defense response followed by the activation of catalase to minimize self-inflicted damage. The most responsive marker genes we tested, such as cyp9e2 and catalase, may provide useful biomarkers for ecotoxicological assays in honeybee populations 85,86 .

Methods
Beekeeping. Western honeybees (Apis mellifera carnica, Pollmann 1879) were obtained from a beehive located near the Institute of Insect Biotechnology in Giessen, Germany (50° 34′ 05.8′′ N, 8° 40′ 18.6′′ E). For the survival experiments, the hive was opened and female bees of all ages were collected from the frames. For the immune response experiments, forager bees were collected from the hive entrance to ensure the immune system was already primed by their natural environment 87 . www.nature.com/scientificreports/ Every day, surviving bees were counted and provided with fresh food and water. Bees were kept in the dark at 26° C and 60% relative humidity. Survival assays were carried out three times per treatment and control. The R libraries survival and survminer 88 were used to fit Kaplan-Meier curves to the survival data and plot the statistical analysis and mean LT 50 for Fig. 1. Statistical values for each single assay (Fig. S1) were determined using GraphPad Prism version 9.0.0 (GraphPad Software, La Jolla, CA, USA). Survival assays in which the control animals survived less than 15 days were not considered for statistical analysis. To determine significant differences between replicates, we carried out a log-rank (Mantel-Cox) test with subsequent Bonferroni correction.
Analysis of gene expression. For each treatment and sampling time point, nine bees in three biological and three technical replicates were provided with 70 μl of sugar syrup. Fed animals were kept individually at 26 °C and 60% relative humidity in the dark. Guts were dissected 1, 3, 6, 24 and 48 h after feeding. Dissected midguts were placed in 2-ml screw cap tubes (Sarstedt, Nümbrecht, Germany) and homogenized using a FastPrep-24 homogenizer (MP Biomedicals, Santa Ana, CA, USA). RNA was extracted using TRIzol reagent according to the supplier's instructions (Thermo Fisher Scientific, Waltham, MA, USA) and the RNA purity and quantity were measured using a NanoDrop spectrophotometer (VWR, Radnor, PA, USA). The RNA was reverse transcribed using the iScript system (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer's instructions. Gene expression levels were determined by quantitative real-time PCR (qRT-PCR) using primers designed with Primer3 and BLAST (Table S1). The honeybee eIF3-S8 gene encoding eukaryotic translation initiation factor 3 subunit C was used as the internal reference. Samples were amplified in 96-well plates on a StepOnePlus Real-Time PCR system (Applied Biosystems, Foster City, CA, USA) using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific). Melting curves and primer tests were used to evaluate primer specificity. The cycling conditions are summarized in Table S1. Absolute expression ΔCt values were determined by normalization to the reference gene. The relative expression levels of candidate genes were calculated using the 2 −ΔΔCt method. Values > 1.5 were defined as an increase of gene expression. Gene expression data were analyzed using StepOne v2.3 (Applied Biosystems) and Microsoft Excel 2010 (Microsoft Corporation, Redmond, WA, USA). Comparisons between treatments were based on the means of three biological and three technical replicates and were used to generate a heat map.

Quantification of ROS.
For each treatment and sampling time point, nine bees in three biological and three technical replicates were provided with 70 μl of sugar syrup. Fed animals were kept individually at 26 °C and 60% humidity in the dark. Guts were dissected 1, 3, 6 and 24 h after feeding and stained for 20 min with 10 µM CM-H 2 DCFDA (Thermo Fisher Scientific) in PBS containing 2 mg/ml of the catalase inhibitor 3-amino-1,2,4triazole (MilliporeSigma). Gut tissues were homogenized using a FastPrep-24 homogenizer and the fluorescence intensity was measured using the bottom optics of a CLARIOstar microplate reader (BMG Labtech, Ortenberg, Germany) with excitation at 485 nm, emission at 538 nm and a gain of 1800. Data were analyzed using Microsoft Excel 2010 and GraphPad Prism version 9.0.0.