Comparative lipidomic and proteomic analysis reveals species‐specific differences in midgut composition and insecticide absorption between Apis mellifera and Bombus terrestris

Abstract

BACKGROUND

The importance of bee pollinators for agricultural and ecological reasons is widely recognised and the declines in pollinator populations have been partly attributed to insecticide exposure in addition to habitat loss, climate change, pathogens and parasites. While metabolic detoxification of insecticides in bees is well‐studied, less is known about species‐specific differences in absorption. This study investigates the link between midgut composition and patterns of absorption in two key pollinator species, the Western honeybee Apis mellifera and the buff‐tailed bumblebee Bombus terrestris.

RESULTS

Comprehensive lipidomic and proteomic analysis revealed large‐scale differences in the gut composition between the two species, with B. terrestris showing a higher ratio of sphingomyelins to phosphatidylcholines compared to A. mellifera. Ex vivo Ussing chamber experiments and in vivo feeding assays demonstrated that the A. mellifera displayed higher midgut permeability for the two insecticides imidacloprid and chlorantraniliprole when compared to B. terrestris. The observed structural variation in midgut composition correlates with the differing absorption rates, suggesting a potential mechanism for species‐specific variations in insecticide toxicity.

CONCLUSION

The finding that A. mellifera midguts display a higher rate of insecticidal absorption provides a novel insight into the role of absorption in bee toxicology. This also has implications for the development of more selective insecticides and improved risk assessment strategies for non‐target organisms. © 2025 The Author(s). Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

Lipidomic and proteomic analysis revealed large‐scale differences in the midgut composition of the honeybee Apis mellifera and the buff‐tailed bumblebee Bombus terrestris. These differences correlated to increased insecticide absorption in the bumblebee when using an ex vivo Ussing chamber and subsequently in vivo feeding assay (created in Biorender).

1. INTRODUCTION

Pollinators play a vital role in the agricultural sector, with the global economic value of pollination estimated at EUR 153 billion. ¹ As the global population increases, there is a greater demand for food production and, consequently, for crop‐pollinating species. ¹ While pollination can occur mechanistically via water or wind, ² , ³ many plants rely on animal pollinators, including bats, birds and various insect species, including bees. ⁴

Among insect pollinators, social bees, in particular honeybees and bumblebees, are ranked amongst the top crop pollinators. ⁵ The Western honeybee, Apis mellifera, is perhaps the most important, being the most frequent single species of pollinator for crops worldwide. ⁶ A. mellifera's significance is underscored by its ability to increase yield in 96% of animal‐pollinated crops ⁷ in addition to its effectiveness in both natural and managed agricultural systems. ⁸ , ⁹

Complementing the honeybee's role, the buff‐tailed bumblebee is another key insect pollinator that has been mass produced in Europe for agricultural pollination for three decades. ¹⁰ Bumblebees, with approximately 260 wild species worldwide, ¹¹ exhibit unique capabilities that distinguish them from honeybees. They display adaptability to both open field and greenhouse environments, ¹¹ can perform buzz‐pollination for efficient pollen extraction from specific flowers, ¹² and have a longer relative tongue length, enabling pollination of flowers with deeper corollas. ¹³ , ¹⁴ These slightly differing ecological niches make both honeybees and bumblebees vital for crop production.

Despite their importance, declines in insect pollinator species have been widely reported. ¹⁵ Bumblebees are particularly threatened, with approximately 21% of species in Europe declining. ¹¹ While global honeybee colony numbers have increased, in Europe overwintering colony loss has reached 25%, ¹⁶ a statistic which increases to 59% in North America. ¹⁷ The reason behind this phenomenon is complex, with multiple factors suggested, including pathogens (fungi, viruses, bacteria and parasites), ¹¹ , ¹⁸ climate change, ¹⁹ , ²⁰ habitat loss ²¹ , ²² and agrochemical use. ²³ , ²⁴

Increasing agricultural intensification has been identified as a potential factor in declining bee health, both directly through exposure and indirectly due to the loss of food sources, therefore the interactions between agrochemicals and bee pollinators are of particular interest. The mechanism by which bees detoxify xenobiotics and therefore reduce toxicity has been described previously. ²⁵ , ²⁶ , ²⁷ , ²⁸ For many agrochemicals, this detoxification is mediated by cytochrome P450s, found predominantly, but not exclusively, in the CYP9Q‐family. In A. mellifera, CYP9Q1, CYP9Q2, CYP9Q3 and CYP6AQ1 have been shown to metabolise various insecticides, ²⁵ , ²⁶ , ²⁸ with the orthologues CYP9Q4, CYP9Q5 and CYP9Q6 showing similar detoxification patterns in Bombus terrestris. ²⁶ , ²⁷ Additionally, the CYP336 family has been identified as containing key detoxifiers of alkaloids, with orthologues in both A. mellifera and B. terrestris shown to metabolise natural insecticidal compounds such as nicotine. ²⁹

Despite this understanding of metabolic processes, some insecticides exhibit differing levels of toxicity between these two species that cannot be explained by metabolism alone. One area that remains poorly understood in relation to insecticide activity is absorption. Most insecticidal compounds target proteins within the insect body, ³⁰ often expressed in nerve or muscle tissue located behind the gut or cuticular barrier. ³¹ Consequently, orally ingested compounds must be absorbed across the midgut to reach their target and exert toxicity. ³¹

Previous studies have examined the lipidomics and proteomics of different bee species, including A. mellifera ³² , ³³ and B. terrestris. ³⁴ , ³⁵ However, there has been little focus on the differences in gut composition between the two species. Understanding these differences could be crucial because they may contribute to variations in insecticide absorption and, subsequently, toxicity.

In addition to investigating the structure of the gut itself, it is also of interest to monitor how insecticidal compounds are absorbed from the midgut lumen into the haemolymph. One method designed to monitor this absorption is the Ussing chamber, ³⁶ commonly used in pharmacology, which measures the transport of small molecules, nutrients and ions across epithelial tissues. ³⁷ While this system has been developed in large Lepidopteran species and Orthoptera, ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ it has not yet been applied to bee species.

In this study, we sought to address these knowledge gaps by comparing the gut composition of A. mellifera and B. terrestris using complementary proteomic and lipidomic analysis. We then examined the functional implications of these compositional differences using an Ussing chamber ³⁶ to monitor absorption of different insecticides across the midgut. The insecticides used had differing modes of action, which demonstrate systemicity to varying degrees and thus may be transported through the plant to pollen and nectar of bee attractive crops, even when applied outside the flowering period. Finally, we correlated this ex vivo data with an in vivo feeding assay to elucidate the relationship between gut composition and insecticide absorption.

2. METHODS

2.1. Lipidomics

Adult A. mellifera worker bees used in the study were removed from a queen‐right healthy colony, managed according to BBKA guidelines, located in Jealott's Hill, UK. Adult worker bees from B. terrestris Audax colonies (Fargro Ltd., Arundel, UK) were reared at 35 °C and 50% relative humidity.

A. mellifera and B. terrestris were anaesthetised on ice, then the midguts dissected out and washed in 0.1 M NaCl to remove contaminating tissue. Samples of five midguts were pooled with four biological replicates per species.

Each sample was homogenised for 30 s with 400 μL of IPA using FastPrep24 Classic with a ceramic ball and lysing matrix A (MP Biomedicals, California, US). The homogenate was centrifuged at 16 000 × g for 10 min and the resulting supernatant evaporated to dryness using a Genevac EZ‐2 (BioPharma Group, Winchester, UK). The sample was reconstituted in 250 μL of 5 μg mL⁻¹ d62‐dipalmitoyl phosphatidylcholine (CK Isotopes Ltd, Leicester, UK) in acetonitrile:isopropanol:water (65:30:5) and filtered using a nylon syringe filer (Crawford Scientific, Lanarkshire, UK).

Samples were analysed by liquid chromatography (LC)‐high resolution accurate mass–mass spectrometry (MS) using a Vanquish LC coupled to a Q Exactive Plus Orbitrap mass spectrometer (Thermo Fisher Scientific, Massachusetts, US) as described previously. ⁴² , ⁴³

The chromatographic separation used a Kinetex C18 2.6 μm, 150 × 2.1 mm column (Thermo) using a 30 min gradient and 10 μL injection volume. The LC gradient used 10 mM ammonium formate, 0.1% formic acid in acetonitrile:water (60:40) and line B1: 10 mM ammonium formate and 0.1% formic acid in isopropanol:acetonitrile (90:10) as the A and B mobile phases, respectively. Initial LC conditions were 68% mobile phase A, 32% mobile phase B, with a flow rate of 260 μL min⁻¹ and a column temperature of 45 °C. The LC gradient was as follows: 0.0 → 1 min (32% B), 1 → 4 min (32 → 45% B), 4 → 5 min (45 → 52% B), 5 → 8 min (52 → 58% B), 8 → 11 min (58 → 66% B), 11 → 14 min (66 → 70% B), 14 → 18 min (70 → 75% B), 18 → 21 min (75 → 97% B), 21 → 25 min (97% B), 25 → 25.1 min (97 → 32% B), 25.1 → 30 min (32% B).

The MS analysis was performed with heated electrospray ionisation using the following source conditions: sheath gas flow 25 arbitrary units, auxiliary gas flow 15 arbitrary units, auxiliary gas heater temperature 350 °C, spray voltage 3.5 kV (+), −3.5 kV (−), capillary temperature 250 °C, S‐lens radio frequency (RF) level 50.0 arb. units. High‐resolution full scan (70 000 resolution), all ion fragmentation (70 000 resolution) and top 10 data dependant MS2 (17 500 resolution) experiments were collected covering a mass range of 120–1800 m/z. Samples were analysed first in positive polarity, then in a separate batch in negative polarity.

Data quality was assessed using a quality control sample created by pooling equal volumes of aliquots from all samples that were injected periodically across the samples. In addition, the intensity of the internal standard (d62‐dipalmitoyl phosphatidylcholine) was monitored throughout the run to assess variability, an extraction blank was prepared and mixed reference standards covering metabolites from a wide range of lipid classes were used to perform system suitability tests prior to the run and provide confidence of annotation and allow semi‐quantitation of the analysed lipids.

Peak picking and annotation of lipid data was performed in LipidSearch 4.2.29 (Thermo Fisher Scientific, Massachusetts, US). Lipid features were then semi‐quantified using a relevant lipid of known concentration in the mixed reference standard (single point calibration) to allow comparative quantitation of midgut lipids across lipid classes. The following standards (Merck Sigma Aldrich, Gillingham, UK) were used: linoleic acid (25 μg mL⁻¹) for fatty acyl lipids, PC(18:2(9Z,12Z)/18:2(9Z,12Z)) (10 μg mL⁻¹) for phosphatidylcholine, phosphatidylinositol and related lipids, LysoPE(16:0/0:0) (1 μg mL⁻¹) for phosphatidylethanolamine and related lipids, SM(d18:1/16:0) (10 μg mL⁻¹) for sphingolipids and coenzyme Q9 (25 μg mL⁻¹) for coenzymes. Features were combined and the data table formatted using in‐house KNIME (Konstanz Information Miner) workflows. ⁴⁴ Data were assessed for data quality and general trends using principal component analysis within SIMCA v14.1. ⁴⁵

2.2. Proteomics

The intestinal tracts of A. mellifera and B. terrestris, reared as according to Section 2.1, were removed and the midguts separated from the rectum and Malphigian tubules. Dissected midguts were washed with nuclease‐free water, frozen in liquid nitrogen and stored at −80 °C until protein extraction. Eight midguts from each species were pooled and 600 μL of ice cold 15% w/v trichloroacetic acid, 1% w/v dithiothreitol added. A Fastprep was used to disrupt the midgut tissue, with four pulses of 20 s, 5 m s⁻¹ with Lysing matrix A (MP‐Biomedicals, California, US). This was performed with cooling on ice between pulses. After 30 min on ice with brief sonication, samples were centrifuged at 16 000 × g for 15 min at 4 °C, as described previously. ⁴⁶ The resultant protein pellet was washed with ice‐cold acetone, resuspended in RIPA buffer (Thermo Fisher Scientific, Massachusetts, US) and subject to vortexing and brief pulses of sonication to ensure solubilisation. Protein concentration was determined using a bicinchoninic acid assay. ⁴⁷ Samples containing 50 μg of protein extract were reduced with 10 mM dithiothreitol at 37 °C for 30 min, followed by alkylation with 40 mM iodoacetamide in the dark at room temperature for 30 min. Single‐pot solid‐phase‐enhanced sample preparation was used for sample cleanup and digestion. A volume of 5 μL of MaqReSyn hydroxyl beads (20 mg mL⁻¹) combined with each sample and acetonitrile was added to a final concentration of 90% v/v to promote protein binding to the beads and the mixture incubated for 10 min. Beads were immobilised on a magnetic rack and the supernatant discarded. Proteins bound to the beads were washed with acetonitrile to remove contaminants. For digestion, beads were resuspended in 20 μL of 50 mM triethylammonium bicarbonate, pH 8.0, and sequencing‐grade trypsin (Promega) was added at an enzyme‐to‐substrate ratio of 1:25 w/w. Digestion was performed at 37 °C for 18 h with shaking at 800 rpm. After digestion, beads were magnetised and the peptide‐containing supernatant was collected. Peptides were lyophilised and subsequently reconstituted in 0.1% formic acid for LC–MS/MS analysis.

2.2.1. Liquid chromatography

Peptide separation was performed using a nanoElute high‐performance liquid chromatography system (Bruker Daltronics, Massachusetts, US) coupled online with a TimsTOF HT mass spectrometer. Peptides (500 ng) were loaded onto a 25 cm × 75 μm internal diameter Aurora Series column (IonOpticks, Melbourne, Australia) maintained at 50 °C. The mobile phases consisted of (A) 0.1% formic acid in water and (B) 0.1% formic acid in acetonitrile. A linear gradient was applied from 3% to 35% B over 30 min at a flow rate of 300 nL min⁻¹, followed by a ramp to 90% B over 2 min and re‐equilibration at 3% B for 3 min. The total run time/sample was 40 min.

2.2.2. Mass spectrometry

Mass spectrometry data were acquired on a timsTOF HT instrument (Bruker Daltronics, Massachusetts, US) operated in data‐independent acquisition mode with dia‐PASEF. The ion mobility range was set from 0.85 to 1.27 Vs cm⁻², with a ramp time of 100 ms. The mass range was scanned from 100 to 1700 m/z. The dia‐PASEF method included 21 windows with isolation widths of 25 m/z, covering the mass range of 475–1000 m/z. Collision energy was ramped linearly from 20 eV at 0.6 Vs cm⁻² as a function of ion mobility. The duty cycle was optimised for 0.95 s, ensuring sufficient sampling of chromatographic peaks.

2.2.3. Data processing

Raw data were processed using directDIA workflow in Spectronaut (Biognosys, v18.0, Schlieren, Switzerland) with default settings. Databases were imported from UniProt A. mellifera UP000005203, 20 240 830, 19 054 entries and B. terrestris UP000835206, 20 241 028, 19 625 entries. Default settings including background subtraction factory settings for peptide identification and Q‐value cutoff of 0.01 (1% False Discovery Rate) at both peptide and protein levels. Fixed modifications were set as carbamidomethylation of cysteine (+57.021 Da) and variable modifications, including oxidation of methionine (+15.995 Da). Integrated barycentric absolute quantification (iBAQ) was selected for protein abundance calculation. The intensities of all identified peptides belonging to a protein were summed and divided by the number of theoretically observable tryptic peptides, yielding the iBAQ value.

The proteins identified from each midgut proteome were used to carry out a functional gene ontology (GO) enrichment analysis. The datasets were analysed using g:Profiler ⁴⁸ , which ran the genes through the respective species database. The proteins identified in each species with an iBAQ of >20 were included in the analysis and were run against all annotated genes with a significance threshold of 0.05. The GOs matched were molecular function, cellular component and biological processes in addition to the Kyoto Encyclopedia of Genes and Genomes (KEGG) biological pathway.

2.3. Ussing chamber

Bees were reared and sampled as described in Section 2.1. The midgut of the bee was removed by isolating the abdomen and making a lateral incision on the dorsal side. The intestinal tract was subsequently separated and the midgut detached. The midgut tissue was opened longitudinally and mounted on Ussing chamber sliders with an aperture of 0.031 cm² (P2407; Physiologic Instruments, Florida, US). The extracted midgut was washed using 200 μL of luminal buffer to remove food contents and any remaining peritrophic membrane or connective tissues were detached.

The midgut‐mounted slider was inserted between the luminal and haemolymph chambers (P2400; Physiologic Instruments, Florida, US), then 2 mL of luminal buffer (5 mmol L⁻¹ CaCl₂, 24 mmol L⁻¹ MgSO₄, 190 mmol L⁻¹ sucrose, 20 mmol L⁻¹ potassium gluconate, 5 mmol L⁻¹ CAPS, pH 5) and haemolymph buffer (5 mmol L⁻¹ CaCl₂, 24 mmol L⁻¹ MgSO₄, 190 mmol L⁻¹ sucrose, 20 mmol L⁻¹ potassium gluconate, 5 mmol L⁻¹ Tris, pH 6) were added to each respective chamber. The apparatus was heated to 35 °C to mimic internal temperature. ⁴⁹ Oxygen was continuously bubbled through both chambers to ensure gut tissue remained oxygenated as described previously. Next, 50 μg of Amaranth dye (Sigma Aldrich) was added to the luminal chamber before commencement of experiments to ensure gut integrity, with presence of the dye in the haemolymph chamber resulting in termination of the experiment. The haemolymph chamber was sampled at 0, 10 20, 30, 60 and 120 min, as described previously, ³⁸ by the removal of 50 μL of the chamber contents, which was subsequently replaced with fresh haemolymph buffer. The assay was run for a total of 2 h, within the limit of tissue degradation, and experiments were conducted with a minimum of four Ussing chamber runs. Midgut tissue used in the run was subsequently macerated in 200 μL of acetonitrile, centrifuged at 3000 × g and the supernatant analysed for compound retention.

2.4. Ussing chamber analysis

Samples were analysed using a Thermo Vanquish ultra‐high performance liquid chromatography system connected to a Thermo Q‐exactive Focus mass spectrometer.

All samples were analysed using Waters (Wilmslow, UK) Acquity BEH C17, 1.7 μM, 2.1 × 50 mm columns, using a 2‐min gradient method and a 10 μL injection volume. The LC gradient used water with 0.2% (v/v) formic acid (Optima; Thermo Fisher Scientific, Massachusetts, US) and acetonitrile (HPLC grade; VWR) as the A and B mobile phases, respectively. Initial LC conditions were 95% mobile phase A and 5% mobile phase B, with a flow rate of 700 μL min⁻¹ and a column temperature of 40 °C. The LC gradient was as follows: 0.0 → 0.1 min (5% B), 0.1 → 1.1 min (95% B), 1.1 → 1.5 min (95% B), 1.5 → 1.6 min (95 → 5% B), 1.6 → 2.0 min (5% B).

The MS analysis was performed with electrospray ionisation in positive ionisation mode using the following source conditions: source gas flow 58 arb. units, auxiliary gas flow 16 arb. units, auxiliary gas heater temperature 480 °C, spray voltage −3.5 kV, capillary temperature 300 °C, S‐lens RF level 50.0 arb. units. Analysis was performed using a full‐scan MS¹ method with 35 000 mass resolution at 200 m/z and a scan window from 100 to 900 m/z.

Data analysis was performed using RStudio ⁵⁰ and visualised with the ggplot2 ⁵¹ and ggsci ⁵² packages.

2.5. Bee feeding assay

Adult bees were anaesthetised using CO₂ and inserted into 1.5‐mL Eppendorf tubes with the end removed. Bees were force‐fed 2 ng of each compound as described previously. ⁵³ , ⁵⁴ Briefly, the bees were secured with parafilm, ensuring immobilisation but still allowing for feeding to take place, and left for 2 h to starve prior to exposure. Compounds were initially dissolved in dimethyl sulfoxide (DMSO) to a stock concentration of 1000 ppm and subsequently dissolved to a concentration of 2 ppm in 20% sucrose. A 1‐μL droplet of the solution was added to an inoculation loop, which was presented to the immobilised bee. Once the feeding response was stimulated, the inoculation loop was visually monitored to ensure that the droplet was entirely ingested.

Post‐feeding, the bees were left for 1 h to allow for the compound to reach the gut. The head and thorax were removed and the intact intestinal track extracted. The intestinal tract and remaining body segment were macerated separately in 500 μL of acetonitrile. The supernatant was removed and analysed for the desired compound using LC–MS, as described in section 2.4.

3. RESULTS

3.1. Lipidomics

A comprehensive lipidomic analysis of B. terrestris and A. mellifera revealed approximately 1500 lipid features across all major lipid classes. Principal component analysis illustrated substantial differences in the midgut lipid profiles of the two species (Fig. 1(A)). Quality control samples clustered tightly on the scores plot, indicating high analytical data quality. The loadings plot (Fig. 1(B)) demonstrated an even distribution of lipids for both species, suggesting that each possesses a unique midgut lipid composition. Certain lipid classes showed species‐specific enrichment, with B. terrestris exhibiting higher levels of coenzymes and O‐acyl‐ω‐hydroxy fatty acids (OAHFA). Both species displayed differential composition of the fatty acyl composition of triglycerides and glycerophospholipids.

Figure 1.

Principal component analysis for the lipidomics analysis of Apis mellifera and Bombus terrestris midgut samples. (A) Scores plot showing clustering of each species with A. mellifera denoted in red and B. terrestris in blue. (B) Loadings plot with each point indicating a lipid, coloured by annotated lipid class. DG, diaglyceride; FA, fatty acid; MG, monoglyceride; PA, phosphatidic acid; PC, phosphatidycholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; Pl, phosphatidylinositol; PS, phosphatidylserine; SM, sphingomyelin; TG, triglyceride.

The most abundant lipid classes in both species were phosphatidylcholine and sphingomyelin (SM), but their relative proportions differed significantly between species (Fig. 2). In B. terrestris, sphingomyelins were dominant, comprising approximately 60% of the total identified lipids, whereas in A. mellifera, phosphatidylcholines were dominant, accounting for 54% of total identified lipids. Other lipid classes identified included ceramides, OAHFAs, phosphatidylethanolamines (PEs) and triglycerides (TG). These minor classes collectively contributed less than 2% to the total midgut lipid composition. Species‐specific differences in minor lipid classes were observed, with A. mellifera showing a higher percentage of TGs, while B. terrestris exhibited higher percentages of OAHFAs, ceramides and PEs.

Figure 2.

The percentage contributions of the summed semi‐quantified lipids in each chemical class to the total lipidome identified in the lipidomics analysis of midgut samples of Apis mellifera (red) and Bombu terrestris (blue).

Figure 3 illustrates the species‐specific distribution of individual lipid species within the major lipid classes. A. mellifera displayed a high percentage of lysophosphatidylcholines (LPCs), particularly LPC(18:0) and LPC(18:1), and SM(d34:1). In contrast, B. terrestris showed a high percentage of sphingomyelins with dihydroxy sphingoid bases, particularly SM(d34:1), SM(d36:1) and SM(d38:1) as well as LPC(18:1). These findings highlight the distinct lipid compositions of midgut membranes in these two bee species, which may contribute to differences in their physiological functions such as passive absorption and subsequently responses to environmental stresses.

Figure 3.

Percentage contribution of key lipids to overall lipidomic composition for Apis mellifera and Bombus terrestris. Individual lipids contributing to >1% of the total were selected. The main lipid classes represented are O‐acyl‐ω‐hydroxy fatty acids (yellow), (lyso)phosphatidylcholine (light blue), (lyso)phosphatidylethanolamines (dark blue), sphingomyelins (orange) and triglycerides (green).

3.2. Proteomics

Mass spectrometry analysis of extracted and washed midguts from A. mellifera and B. terrestris yielded comprehensive proteomic profiles. A total of 2646 unique proteins in B. terrestris and 3336 proteins in A. mellifera were identified. Figure 4 illustrates the relative abundance of these proteins revealing a similar range spanning approximately five orders of magnitude in both species.

Figure 4.

Quantitative proteomic analysis of Apis mellifera and Bombus terrestris midgut proteins, ordered by integrated barycentric absolute quantification (iBAQ) intensity. All proteins identified are displayed in pink with the key enzyme groups, esterases, glutathione‐S‐transferases, UDP‐transferases and cytochrome‐P450s, highlighted in red. Proteins annotated as transporters, including ATP‐binding cassette transporters are denoted in green.

Due to their involvement in absorption, distribution, metabolism and excretion processes, certain proteins were of specific interest, such as detoxification enzymes and small molecule transporters including esterases, gluthathione‐S‐transferases (GSTs), UDP‐glucoronosyltransferases (UGTs), cytochrome‐P450s and ATP‐binding cassette (ABC) transporters. The relative abundance of these proteins is highlighted in Fig. 4.

The proteomic analysis of the A. mellifera midgut fraction revealed the presence of six cytochrome‐P450s (CYP9Q1, CYP9Q2, CYP9Q3, CYP336A1, CYP6AS13 and CYP6A15). Notably the CYP9Q subclade (CYP9Q1–3) and CYP336A1 have been previously identified as key xenobiotic metabolisers. ²⁵ , ²⁷ , ⁵⁵ , ⁵⁶ , ⁵⁷ Among these, CYP9Q1 showed the highest abundance, with an iBAQ value exceeding 450. The midgut of B. terrestris identified seven P450s (CYP6BD1, CYP9R1, CYP9P1, CYP6AS10, CYP6AS13, CYP6AQ27 and CYP336A24). Similar subclades (CYP6AS and CYP336) were found in both species.

Analysis of other detoxification enzymes revealed similar numbers in both species. A. mellifera exhibited three GSTs, four UGTs and seven esterases, while B. terrestris showed four GSTs, six UGTs and 12 esterases. The esterases included sphingomyelin, phosphodiesterases and thioesterases among others. Interestingly, the most abundant detoxification enzyme in both bee species was a GST. In A. mellifera C3VMN1 showed an iBAQ of 2333.9 while in B. terrestris LOC100650722 had an iBAQ of 22 584.

Multiple families of small molecule transporters were identified in both species, including metal (zinc and sodium) transporters, monocarboxylate transporters, organic cation transporters, amino acid transporters and ATP‐binding cassette (ABC) transporters. ABC transporters, known for their role in absorption, ⁵⁸ were found in both species, with nine ABCS identified in A. mellifera and seven in B. terrestris. Notably, the proteins in this family were found in greater abundance in A. mellifera, with the most abundant ABC transporter giving an iBAQ of 135.4, relative to 3.5 in B. terrestris.

GO term enrichment analysis of the gut proteomes revealed key overrepresented processes in both bee species (Fig. 5). Among the biological processes identified, several were common to both A. mellifera and B. terrestris, including biosynthetic processes, cellular localisation and macromolecule processing. Multiple metabolism‐related terms were also highlighted, such as generation of precursor metabolites, nucleoside phosphate metabolic processes and purine‐containing compound metabolic processes. Examination of the cellular component GO terms revealed, as expected, many similarly over‐represented processes in both datasets. These primarily involved organelles (particularly intracellular organelles), proteins (protein‐containing complexes) and ribosomes (ribosomal subunits and ribonucleoprotein complexes). The over‐represented molecular function terms correlated with previously identified process, notably those related to translation regulation. Terms associated with ribosomal constituents and RNA binding were also prominently represented.

Figure 5.

Summary of the Gene Ontology (GO) term processes identified as most significant by GProfiler. Significant GO terms in the proteome of the midguts from Bombus terrestris and Apis mellifera classified by their enriched pathway, biological process, cellular component and molecular function.

3.3. Ussing chamber

An ex vivo Ussing chamber assay was used to monitor absorption of insecticides across the midgut of A. mellifera and B. terrestris. The concentration of compounds detected in the haemolymph chamber was quantified to determine insecticide passage through the midgut. Additionally, the midguts used in the chamber were extracted and macerated to analyse compound sequestration within the tissue.

Marked differences in imidacloprid absorption were observed between the two species (Fig. 6). In B. terrestris, no imidacloprid was detected in the haemolymph chamber until 30 min into the assay. After 2 h, 0.02 μg of imidacloprid had crossed into the haemolymph chamber. In contrast, A. mellifera exhibited significantly higher absorption, with 0.18 μg of imidacloprid detected, representing a 9‐fold increase compared to B. terrestris. Analysis of the extracted midguts revealed higher imidacloprid sequestration in B. terrestris (0.0056 μg) compared to the midgut of A. mellifera (0.0044 μg).

Figure 6.

Ussing chamber results for Apis mellifera and Bombus terrestris midguts when incubated with the diamide insecticide chlorantraniliprole. The concentration of imidacloprid (ppm) recorded in the haemolymph chamber of the Ussing chamber over time (min). Error bars denote standard error.

The diamide chlorantraniliprole showed a similar pattern of absorption to imidacloprid between the two bee species (Fig. 6), with a higher proportion crossing the midgut of A. mellifera. After 2 h, 1.08 μg of chlorantraniliprole was detected in the A. mellifera haemolymph chamber compared to 0.06 μg for B. terrestris, an 18‐fold difference. Consistent with the imidacloprid results, analysis of the midgut tissues showed greater chlorantraniliprole sequestration in B. terrestris compared to A. mellifera.

Although both compounds showed higher rates of absorption in A. mellifera relative to B. terrestris, there were some notable differences. While chlorantraniliprole moved across the gut at a higher rate than imidacloprid in both species, in A. mellifera this was a 6‐fold difference whereas in B. terrestris it was only 3‐fold higher.

Following the Ussing chamber experiments, the gut tissues were analysed for parent compound sequestration (Table 1) but also potential metabolites to assess not only absorption but also metabolic activity. LC–MS analysis revealed the presence of metabolites in both species. For imidacloprid, a +16 mass unit metabolite was detected in gut tissues of both B. terrestris and A. mellifera. This metabolite corresponds to the hydroxylated form of imidacloprid, suggesting similar phase I oxidative metabolism in both bee species. In the case of chlorantraniliprole, a +16 metabolite was also identified in B. terrestris gut tissues, again indicating oxidative metabolism similar to that observed with imidacloprid. Interestingly, this metabolite was not detected in the A. mellifera gut tissue.

Table 1.

The mass of the compound sequestered (μg) for two different insecticides, imidacloprid and chlorantraniliprole, in the midguts of two bee species Apis mellifera and Bombus terrestris at the end of an Ussing chamber run.

Species	Imidacloprid	Chlorantraniliprole
Apis mellifera	0.0044 (0.00023)	0.0036 (0.0011)
Bombus terrestris	0.0056 (0.0010)	0.0062 (0.0014)

Bracketed values represent standard error.

3.4. Feeding assay

The in vivo feeding assay was conducted to validate the ex vivo Ussing chamber results and assess the translocation of insecticides across the gut barrier in living bees. A. mellifera and B. terrestris were immobilised and fed 2 ng of either chlorantraniliprole or imidacloprid and left for 1 h, after which the insects were dissected to quantify the amount of compound that had traversed the gut and entered the body cavity.

For chlorantraniliprole, a marked difference in absorption was observed between the two species (Figs 7 and 8). In A. mellifera, approximately 60% of the ingested compound was detected in the body fraction, indicating successful gut translocation. In contrast, only 13% of chlorantraniliprole was found in the body fraction of B. terrestris, suggesting a significantly lower absorption rate.

Figure 7.

Ussing chamber results for Apis mellifera and Bombus terrestris midguts when incubated with the diamide insecticide chlorantraniliprole or imidacloprid. The concentration of compound (μg) moving from the luminal to the haemolymph chamber of the Ussing chamber is recorded over time (mins). Error bars denote standard error.

Figure 8.

Summary of an in vivo feeding assay for the diamide chlorantraniliprole and the neonicotinoid imidacloprid in Apis mellifera and Bombus terrestris. Bars represent the percentage of total concentration of insecticide detected in the body cavity of the bee relative to the gut fraction after 1 h. Error bars denote standard error.

A similar interspecies difference was observed with imidacloprid, albeit with lower absorption rates. In A. mellifera, 8.5% of the ingested imidacloprid was detected in the body fraction, compared to only 1.16% in B. terrestris. These findings corroborate the patterns observed in the ex vivo Ussing chamber experiments, demonstrating consistently higher gut permeability in A. mellifera compared to B. terrestris for both tested compounds. Moreover, the compounds reveal a compound‐specific effect, with chlorantraniliprole exhibiting greater gut translocation than imidacloprid in both species. This was 4.6‐fold higher in A. mellifera and 11.2‐fold higher in B. terrestris.

4. DISCUSSION

Absorption plays a crucial role in the distribution of insecticides within an insect's body and has been proposed as a potential mechanism for selectivity between pest and non‐target organisms. ³¹ , ⁵⁹ , ⁶⁰ While extensive research has demonstrated the importance of metabolism in insecticide selectivity and toxicity in bees, ²⁷ , ²⁸ , ⁵⁵ , ⁶¹ the role of absorption has received little attention. This study presents data revealing significant structural differences between the midguts of two social bee species, A. mellifera and B. terrestris, and the link between insecticide absorption.

Lipidomic analysis uncovered numerous differences in midgut composition between the two species, with the ratio of sphingomyelin to phosphatidylcholines being particularly noteworthy. The importance of these two phospholipids has been well documented because they constitute more than 50% of phospholipids in mammalian membranes ⁶² and exhibit markedly different dynamic properties. ⁶³ Sphingomyelins, characterised by highly saturated acyl chains, trans double bond and amide bonds in their hydrophilic regions, exert a strong rigidifying effect. ⁶⁴ In contrast, phosphatidylcholines form highly fluid lipid regions. ⁶³ In mammalian systems, an increased sphingomyelin content correlates with decreased permeability of small molecules with no specific transport system. ⁶⁵ Moreover, sphingomyelin‐rich membranes display reduced water permeability compared to those dominated by phosphatidylcholine, indicating tight packing of molecules. ⁶⁶

The higher ratio of sphingomyelins to phosphatidylcholines in the midgut lipidomic profile of B. terrestris suggests lower permeability to water‐soluble compounds compared to A. mellifera. This correlates with the effect observed in our ex vivo and in vivo assays, which demonstrated lower absorption rates of both imidacloprid and chlorantraniliprole B. terrestris.

In addition to the lipidomic analysis, we also identified several relevant protein families through proteomic analysis of the two bee species. In insects, small molecules can pass through the gut membrane by two main mechanisms of passive diffusion: transcellular (through the epithelial cells themselves) or paracellular (around the cells and through septate junctions). ³¹ However, additional routes of absorption often involve transporter proteins, which can aid small molecules, particularly those which struggle to diffuse across the membrane passively due to their physical chemical properties. ⁶⁷ The proteomic analysis revealed the presence of multiple small molecule transporters in both bee species, including the well‐documented ATP‐binding cassette transporter family. ⁵⁸ These proteins facilitate the movement of small molecules across epithelial cells by actively hydrolysing ATP to pump the molecules in or out of the cell. ⁶⁸ Members of this protein transporter family were found in higher abundance in A. mellifera than B. terrestris, suggesting a higher absorption efficiency, something which was reinforced by the ex vivo and in vivo assays.

The role of metabolism, particularly by the cytochrome‐P450 CYP9Q clade, is well documented in these species. ²⁵ , ²⁶ , ²⁷ , ²⁸ , ⁵⁵ While the A. mellifera proteins CYP9Q1–3 were identified in this study, their B. terrestris orthologues were not detected, possibly due to insufficient quantities. Members of the CYP336 clade, also implicated in xenobiotic metabolism, ²⁹ , ⁵⁷ were found in the midgut proteome of both bee species. Notably, a +16 metabolite of both imidacloprid and chlorantraniliprole, indicative of cytochrome‐P450 hydroxylation, ⁶⁹ was identified in the Ussing chamber midgut samples. This again suggests that, while the CYP9Qx proteins were not identified in the B. terrestris proteomic analysis, they are present at a level not detected by the proteomics. This observation corroborates previous work on key P450s in these species capable of low‐level imidacloprid and chlorantraniliprole metabolism ²⁶ , ⁵⁵ while also confirming the continued viability of the membranes during these assays.

While direct correlations between ex vivo assays and toxicity data are complicated to establish, insecticide sensitivity is known to vary between these bee species. ⁷⁰ For instance, differences in sensitivity to chlorantraniliprole have been reported for certain formulations, with long‐term exposure effects seen in bumblebees. ⁷¹ , ⁷² Bee toxicity patterns have been closely linked to metabolic capacity, as exemplified by the alfalfa leafcutter bee Megachile rotundata, which lacks the CYP9Q subfamily and consequently exhibits 170 to 2500‐fold higher sensitivity to many insecticides than other bee pollinators. ⁵⁶ However, our study is the first to link differences in bee toxicity to midgut absorption, a critical process for neurologically targeting insecticides. ³¹

This work demonstrates significant variability in insecticide absorption between A. mellifera and B. terrestris, with A. mellifera showing higher midgut permeability for the tested compounds. While increased permeability could result in higher toxicity due to greater target accessibility, ⁷³ the implications of lower absorption rates are more nuanced. Compounds absorbed more slowly may be sequestered and released gradually, potentially resulting in higher persistence due to slower clearance from the insect. This phenomenon is reminiscent of the pharmacological profile of metformin, which exhibits slow, dose‐dependent absorption along the gastrointestinal tract in humans. ⁷⁴ This prolonged retention and gradual release of insecticides may explain the chronic effects observed in these bee species. ⁷⁵ , ⁷⁶ , ⁷⁷ The evolutionary rationale for this increased permeability is not yet known. However, the contrasting lifecycles of these two species may offer some explanation. Honeybee workers must overwinter in colder climates, ⁷⁸ making rapid nutrient absorption from limited food resources critical for survival. In contrast, B. terrestris workers do not overwinter ⁷⁹ therefore do not require this digestive mechanism and this potentially explains their lower rate of gut absorption. This balance between rapid absorption and prolonged persistence highlights the complex relationship between midgut permeability and overall insecticide activity, underscoring the need for comprehensive toxicological assessments that consider both acute and chronic exposure scenarios.

5. CONCLUSION

In conclusion, we present a comprehensive lipidomic and proteomic midgut profile of the bee species A. mellifera and B. terrestris that identifies key structural differences in the gut. These differences correlate with the higher midgut permeability of A. mellifera to both imidacloprid and chlorantraniliprole, potentially explaining some of the varying toxicological profiles observed in these bee species. Our findings have significant implications for applied research as future insecticides can be screened for midgut absorption, potentially leading to the development of more selective compounds and improved risk assessment strategies for non‐target organisms.

CONFLICT OF INTEREST

All authors are employed by the agrochemical company Syngenta, a manufacturer of pesticides.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.