Thiamethoxam exposure in honey bees deregulates short ORF gene expression in the brain and compromises the immune response

Maximizing crop yields heavily relies on the use of agrochemicals to control insect pests. One of the most widely used insecticides are neonicotinoids. Here, we analysed the impact of sub-lethal chronic long-term exposure to the neonicotinoid Thiamethoxam on gene expression and alternative splicing in brains of Africanized honey bees Apis mellifera. Our results reveal a small number of differentially regulated genes showing a concentration dependent response to two low doses of chronic, 10-day Thiamethoxam exposure. Unexpectedly, most of these genes have no annotated function, but encode short Open Reading Frames (sORFs), a characteristic feature of anti-microbial peptides. Likewise, we find that Thiamethoxam exposure sensitizes bees to infection by non-pathogenic bacteria Bacillus badius and Ochrobactrum anthropi. Moreover, infection with estimated single Serratia marcescens kills bees arguing that Varroa mites may essentially contribute to colony collapse by penetrating the cuticle to spread this pathogen. Our results implicate an altered immune response to Thiamethoxam exposure compromising the immune response leading to a decline in bee populations.


INTRODUCTION
The western honey bee Apis mellifera is valued as a highly beneficial species for the human society for producing honey, but most importantly for its critical role in the provision of managed pollination services that contribute to increased yields of many food crops to sustaining global food security. Globally, crop productivity is also enhanced by continuous use of pesticides, often on highly simplified agroecosystems. There is a need to find a balance between the necessity of pesticide application and the unintended effects this has on pollinating insects, including honey bees, where agrochemicals are one of the factors implicated in contributing to their declining health and abundance [1][2][3][4] .
Compared to organophosphate pesticides, neonicotinoids are relatively safe for vertebrates and have been one of the most important protection agents against insect pests outside the EU 5,6 . Neonicotinoids are applied to seeds and will systemically spread to the leaves of growing plants to kill leave eating pests.
Applying neonicotinoids to plant seeds dramatically reduces side effects from spraying, but still retains toxic burden for pollinators as neonicotinoids are also found in pollen and nectar 6 .
Neonicotinoids act as agonists, competing with the neurotransmitter acetylcholine in binding to nicotinic acetylcholine receptors (nAChR) [7][8][9] . The increased toxicity for insects is thought to be caused by the characteristically high nAChR density within insect nervous systems [reviewed in 10 ]. Uptake of sublethal levels of some types of neonicotinoids affects the honey bee brain, in that foraging behaviour, navigation, communication, learning and memory is impaired [reviewed in 11 ] [12][13][14] .
Understanding the risk for honey bees requires detailed knowledge of cellular and molecular effects that result from the exposure to an insecticide in order to mitigate negative effects or refine the target specificity towards pest species.
Changes in gene expression and processing of RNAs, including alternative splicing, are one of the options available to an organism to respond to environmental perturbations 15, 16 . Sub-lethal exposure of xenobiotics can induce modulation of splicing reactions [17][18][19] .
Neonicotinoid exposure has been linked to a decline in bee health including a reduction of immune competence 20 ;; 21-23 . Insects do not have antibodies and rely on the innate immune system to fight microbial infections. The best insight on insect immunity stems from studies in the fruit fly Drosophila melanogaster, where cellular and humoral immune responses have been identified 24,25 . The cellular response is mediated by three types of hematopoetic cell lineages 24,26,27 , while the humoral immune system can be split into Toll and Imd pathways [reviewed in 28 ]. The Tollpathway is triggered following an immune challenge by Gram-positive bacteria and fungi, ultimately leading to expression of antimicrobial peptides (AMPs) that are then secreted from the fat body into the hemolymph 24,29,30 . In contrast, the Immune deficiency (Imd) pathway leads to expression of a different set of AMPs after a Gramnegative bacterial infection activates pattern recognition receptors and a complex intracellular signaling cascade 24,25,29 . AMPs are short peptides of 10-100 amino acids and are characterised by an evolutionary dynamism among insect species 31 .
We have previously shown that worker-bee larvae in colonies contaminated with the neonicotinoid imidacloprid, have altered expression of genes belonging to lipidcarbohydrate-mitochondrial metabolic networks 32 . We have further demonstrated that sub-lethal levels of Thiamethoxam, which is another member of the 5 neonicotinoid-class of insecticides, can cause impairment in the midgut and brain of the Africanized Apis mellifera (AHB), as well as contribute to a reduction in honey bee lifespan [33][34][35][36] .
Here, we analysed the effects of chronic sub-lethal Thiamethoxam exposure on gene expression and alternative splicing in the brain of worker honey bees. Our results reveal a small number of 52 differentially regulated genes showing a concentration dependent response to two low doses of chronic, 10-day Thiamethoxam exposure.
Unexpectedly, most of these genes have no annotated function, but the vast majority are characterized by encoding short Open Reading Frames (sORFs). Half of these genes are predicted to encode antimicrobial peptides suggesting a compromised immune response after Thiamethoxam exposure. Indeed, challenge of Thiamethoxam exposed bees by infection with bacteria dramatically decreased viability. Hence, our results suggest that Thiamethoxam sensitizes bees to bacterial infection by compromising an immune response mediated by anti-microbial peptides.

Toxicity assays
For chronic exposure to the neonicotinoid Thiamethoxam (Sigma), Africanized Apis mellifera bees were marked with a colour pen at eclosion and collected 15 days later from the hives in the apiary of the Biosciences Institute, UNESP, Rio Claro-SP (Brazil). Bees were then kept in small cages in an incubator at 32° C at relative humidity of 70 % for a 10 day duration of Thiamethoxam exposure. For each experimental group bees were collected from three different hives and kept in groups of 20 individuals. Bees were fed ad libidum with a solution containing water and inverted sugar (1:1) for the control and for toxin exposure Thiamethoxam was added at field relevant concentrations of 2 ng/ml and 5 ng/ml, respectively. Ten days after the start of the exposure, bees were cold-anaesthetised, brains were dissected and RNA was extracted.
RNA extraction, Illumina sequencing and analysis of differential gene expression Total RNA was extracted from five dissected brains per sample by cracking the brain in liquid nitrogen, brief homogenization with a pistil in 50 µl of Tri-reagent (SIGMA) before the volume was increased to 500 µl and proceeded according to the manufacturer's instructions. Total RNA was then stored in 70% ethanol and shipped at ambient temperature from Brazil to the UK.
Total RNA was treated with DNase I (Ambion) and stranded libraries for Illumina sequencing were prepared after poly(A) selection from total RNA (1 μg) with the TruSeq stranded mRNA kit (Illumina) using random primers for reverse transcription according to the manufacturer's instructions. Pooled indexed libraries were sequenced on an Illumina HiSeq2500 to yield 28-41 million paired-end 125 bp reads for three controls, three low and three high Thiamethoxam exposed brain samples.

Reverse transcription quantitative polymerase chain reaction (RT-qPCR)
Reverse transcription was done with Superscript II (Invitrogen) as previously described 42 . PCR was done as described and PCR products were analysed on ethidium bromide stained 3% agarose gels 43 . Neuronal genes used as reference for gene expression comparison were bee erect wing (ewg) and Amyloid Precursor 44 . To validate cDNAs obtained from RT the following primers were used to amplify bee ewg with AM ewgG and actR1 (TTTGTCCCATGCCAACCAT), and Appl with primers AM appl F and AM appl R2, as described previously 32,46 .

Bacterial infection assays
For infection assays, Apis mellifera from the University of Birmingham Winterbourne garden hives were used and injected with Bacillus badius, Serratia marescens strains SY13 and 274, Serratia fonitcola and Ochrobactrum anthropi as detailed below.S.
fonitcola and O. anthropi were isolated from honey bee guts by plating the gut content on LB agar plates incubated at 30º C. Individual colonies were then grown up in LB media and glycerol stocks frozen. Bacterial species were identified by colony PCR and ribosomal 16S sequencing: A colony was picked from an LB agar plate with a yellow tip and placed into 10 µl TE in a PCR tube and heated to 94º C for 5 min.
The PCR mix was added adjusting the MgCl concentration to 1.5 mM. PCR was done for 20-40 cycles with 54º C annealing for 40 sec and 1 min extension at 72º C. Forager honey bees for infection assays were collected from colonies of the Winterbourne Garden of the University of Birmingham (UK) and kept as described 43 .
Injections were done as described 43 . Bacteria for injections were freshly plated and grown overnight on LB plates. Then a single colony was used to inoculate a 5 ml LB Statistics was done in GraphPad prism using Anova followed by Tukey-Kramer post hoc test.

Chronic Thiamethoxam exposure alters expression of a subset of genes in honey bee brains
Here, we assessed whether the neonicotinoid Thiamethoxam alters gene expression in the brain of Africanized honey bees Apis mellifera at field relevant concentrations after long-term exposure. For these experiments age-matched, control sugar-fed bees (sample A) were compared to bees exposed to 2 ng/ml (low dose, LD, sample B) and 5 ng/ml (high dose, HD, sample C) Thiamethoxam contained in the sugar solution given for feeding for 10 days. To enhance robustness against genetic variation, we chose worker honey bees from three unrelated hives. To validate these results from the Illumina sequencing, we performed RT-qPCR for three of the dose-responsive genes (GB41923, a putative Sodium chloride cotransporter, and GB48969, GB40669, two genes with unknown function). For all three genes we detected an expression difference upon Thiamethoxam exposure (Supplemental Fig 1). We also validated and confirmed differential expression of bubblegum, encoding a very long-chain acyl-CoA synthetase, which has been found down-regulated in honey bee larvae exposed to the neonicotinoid imidaclorid Alternative splicing has been suggested as a mechanism to adapt gene expression to environmental changes 18,47 . Accordingly, we analysed the RNA-seq data for changes in alternative splicing (Supplemental table 2). Although this analysis revealed significant differences, upon manual inspection of sequencing traces using Integrated Genome Viewer (IGV) 39 , we did not further consider this, because the analysis process mostly selected genes with complex splicing patterns that showed no obvious differences in the number of sequence reads over alternatively spliced gene sections in IGV (data not shown).

Dose-responsive expression following Thiamethoxam exposure occurs mostly in genes encoding uncharacterized ORFs
Next, we categorized the genes responsive to the dose of Thiamethoxam according to their functions, taking into account known functions for orthologues in Drosophila and functions deduced from annotated protein domains retrieved by BLAST analysis.
This analysis revealed that 59 % (22 from 37) and 73 % (11 from 15) of up- or downregulated genes, respectively, had neither clear orthologues in Drosophila nor any recognizable protein domains that would indicate a biological function (Fig 2A).
Among up-regulated genes 14 % (5 from 37) were assigned roles in cellular signaling, which potentially could be linked to altered neuronal function, such as olfactory and taste perception, and as structural components (cytoskeleton), and 8 % (3 from 37) of genes were assigned functions in regulation of gene expression by transcriptional regulation (Fig 2A). The very high number of genes with unknown function is unexpected compared with Drosophila as in this organism the fraction of genes with unknown functions in gene expression studies is about 20 % 48,49 .

Most Thiamethoxam dose-responsive genes in the brain encode short proteins
When analyzing the set of dose-responsive, differentially expressed genes for the presence of annotated proteins and protein domains to assign functions, we noticed that many of the genes with unknown function coded for short ORFs. For Thiamethoxam induced differentially up-and down-regulated genes 73 % (27 from 37) and 87 % (13 form 15) encode for genes with ORFs of 250 aa or shorter, respectively ( Fig  2B).
Since these short ORFs could encode anti-microbial peptides, we used a machine learning algorithm to predict whether they encode anti-microbial peptides 41 . From the 13 code for antimicrobial peptides, of which 52 % (11 genes) and 55% (6 genes) were from differentially up- and down-regulated genes, respectively (Fig  2B).

Thiamethoxam sensitizes worker bees to bacterial infection
Thiamethoxam-induced alteration of anti-microbial peptide gene expression could disturb the immune response. To test this hypothesis we adopted a procedure initially developed for Drosophila, where viability is monitored after injection of bacteria. In this assay, non-pathogenic bacteria are efficiently cleared by anti-microbial peptides with little impact on viability 50 . To test efficient clearance of non-pathogenic bacteria in bees, we used Bacillus badius, a non-pathogenic bacterium commonly found in the environment (Fig  3A). In addition, we isolated two bacterial species from worker bee were injected into bees, viability was not affected, indicating that the bee immune system is able to clear the infection efficiently (Fig  3A  and  B). Injection of S. fonticola at these concentrations was lethal. In contrast, injection of sub-lethal doses of Thiamethoxam together with either B. badius and O. antrhopi resulted in lethality ( Fig   3A  and  B). Intriguingly, diluting either Thiamethoxam or injected bacteria 10 fold still dramatically decreased viability 48 h after injections.
We also assessed the effect of Serratia marescens, a known bacterial pathogen of bees 53 . Injection of S. marescens killed bees even at a very low bacterial load, where we estimate that only one bacterium was present in the injected volume ( Fig   4).

DISCUSSION
A key finding of our analysis of brain transcriptomes of worker honey bees is the highly enriched fraction of dose-responsive, uncharacterised genes encoding short open reading frames (sORFs). Strikingly, the majority of these dose-responsive sORFs are predicted to encode antimicrobial peptides. Rather than being non-coding RNAs, sORFs have only recently been recognized to encode functional peptides [54][55][56] some of which play important roles during development 57 . Differential regulation of sORFs was not reported in recent whole-transcriptome studies to evaluate the effects of neonicotinoid exposure in brains of honey bees or bumblebees 19,58,59 . Likewise, we did not find an overlap of the differentially expressed genes. The main difference to our study is that we analysed changes upon long-term low dose chronic exposure, while these studies used shorter term exposures. However, we note that other studies infecting bees with viruses of Nosema revealed an overrepresentation of genes with known function in immune and defence processes, where expression is altered upon exposure to different types of neonicotinoids 19,21,58-60 . Since we did not detect differential expression of known immune genes, the bees in our study were not infected with these known pathogens, but we cannot exclude an impact of other microbiota that are non-pathogenic in healthy bees.
Bees, including Apis mellifera are characterised by their limited set of canonical immune genes, compared to non-social insects, such as the fruit fly D. melanogaster [61][62][63] . Currently, only six antimicrobial-peptide genes comprising four gene families have been described in honey bees 63 . In contrast, Drosophila has 20 antimicrobial-peptide genes comprising eight gene families 24,63 . From these genes, only defensin is conserved between honey bees and Drosophila, consistent with the idea that antimicrobial peptides evolve fast to adapt to species-specific environmental conditions 31 . Given the low number of known antimicrobial peptides in bees, it is conceivable that new, as yet uncharacterised genes encoding antimicrobial peptides are evolving. Alternatively, the short ORFs that we found to be differentially expressed upon chronic low dose Thiamethoxam exposure in the brain might form a basal immune defence in the brain, similar to the antimicrobial environment present in saliva containing many antimicrobial peptides in many animal species. However, antimicrobial peptides have also been identified in having role in learning and memory in Drosophila 64 .
Various xenobiotics have been shown to alter the gut microbiome 65,66 . Our results are consistent with previous findings where neonicotinoid exposure adversely affects insect immunity 66,67 . The immune challenge by non-pathogenic bacteria was welltolerated by worker bees, while the Thiamethoxam-exposed worker bees succumbed by the same bacterial challenge demonstrating that pesticide exposure weakens immunity of honey bees.
We note that pathogens can enter the hemolymph through punctures inflicted by is required to change its splicing pattern 72 . We previously also analysed alternative splicing in bee elav and Xbp-1 genes as potential markers for defects in synapse formation and the stress response upon high dose acute Thiamethoxam exposure, but also low dose chronic exposure did not affect their splicing 43,48,49 . Potentially, alternative splicing might change only in a few cells in the brain eluding detection without single cell analysis or could in addition also be more subtle, requiring more replicates for detection of significantly altered alternative splicing changes 18 .   Venn diagrams indicating the number of differentially expressed genes between control bees and bees exposed to a low dose (left) and high dose (right) of Thiamethoxam that were up- (A) or down-regulated (B). Genes that responded to the injected dose are indicated by a white number on dark background).