Diesel exhaust particles alter gut microbiome and gene expression in the bumblebee Bombus terrestris

Abstract Insect decline is a major threat to ecosystems around the world as they provide many important functions, such as pollination or pest control. Pollution is one of the main reasons for the decline, alongside changes in land use, global warming, and invasive species. While negative impacts of pesticides are well‐studied, there is still a lack of knowledge about the effects of other anthropogenic pollutants, such as airborne particulate matter, on insects. To address this, we exposed workers of the bumblebee Bombus terrestris to sublethal doses of diesel exhaust particles (DEPs) and brake dust, orally or via air. After 7 days, we looked at the composition of the gut microbiome and tracked changes in gene expression. While there were no changes in the other treatments, oral DEP exposure significantly altered the structure of the gut microbiome. In particular, the core bacterium Snodgrassella had a decreased abundance in the DEP treatment. Similarly, transcriptome analysis revealed changes in gene expression after oral DEP exposure, but not in the other treatments. The changes are related to metabolism and signal transduction, which indicates a general stress response. Taken together, our results suggest potential health effects of DEP exposure on insects, here shown in bumblebees, as gut dysbiosis may increase the susceptibility of bumblebees to pathogens, while a general stress response may lower available energy resources. Those effects may exacerbate under natural conditions where insects face a multiple‐stressor environment.


| INTRODUC TI ON
Global biodiversity loss is one of the major challenges humanity currently faces (Díaz et al., 2006;Dirzo et al., 2014). Especially the rapid decline in insects is cause for concern, as they provide or contribute to many important ecosystem functions such as pollination, nutrient cycling, pest control, and linking trophic levels (Cardoso et al., 2020;Noriega et al., 2018). Pollution is one of the major reasons for the decline alongside intensification of land use, climate change, and invasive species (Miličić et al., 2021;Sánchez-Bayo & Wyckhuys, 2019).
Pesticides harm insects on many different levels ranging from subtle changes in the gut microbiome over behavioral changes to increased mortality (Desneux et al., 2007;Motta et al., 2018;Ndakidemi et al., 2016). Other anthropogenic pollutants might also contribute to the observed declines in insects, but their impacts are often less wellstudied (Cameron & Sadd, 2020;Feldhaar & Otti, 2020;Sánchez-Bayo & Wyckhuys, 2019). Airborne particulate matter deriving from traffic or industrial processes has become ubiquitous in the environment (Gieré & Querol, 2010;Zereini & Wiseman, 2010). While the harmful effects on mammals, in particular humans, have been intensively studied, research investigating the impact on insects remains scarce (Kim et al., 2015;Valavanidis et al., 2008). Insects can encounter these pollutants in various ways, for example, by foraging in contaminated areas, consuming contaminated food, or direct deposition on the insect's cuticle (Feldhaar & Otti, 2020;Łukowski et al., 2018;Negri et al., 2015). The airborne particulate matter might enter an insect's body via oral ingestion or the tracheal system (Feldhaar & Otti, 2020;Negri et al., 2015). Social insects might be at an increased risk, as pollutants are transferred to and stored in their nests, which could lead to a higher exposure to conspecifics and the brood (Feldhaar & Otti, 2020;Hladun et al., 2016).
Vehicle brake dust and diesel exhaust particles (DEPs) are major classes of airborne particulate matter deriving from traffic released into the environment (Hamilton & Hartnett, 2013;Harrison et al., 2012;Rönkkö & Timonen, 2019). Brake dust particles contain various metals and phenolic compounds, depending on the brake lining used (Iijima et al., 2007;Thorpe & Harrison, 2008). Exposure of different invertebrate species to such particles showed mixed effects. Particulate matter contamination in soil did not affect colony founding in the ant Lasius niger (Seidenath et al., 2021). However, soilfeeding earthworms (Eisenia fetida) showed a strongly increased mortality when exposed to soil spiked with brake dust particles (Holzinger et al., 2022). DEPs have a different composition than brake dust. They are composed of an elemental carbon core with adsorbed organic compounds, such as polycyclic aromatic hydrocarbons (PAHs), and traces of metals and other elements (Greim, 2019;Wichmann, 2007).
Exposure to high doses of diesel exhaust particles (1 and 2 g/L) in food over a period of 7 days reduced survival in Bombus terrestris workers compared to controls by nearly 50 percent (Hüftlein et al., 2023).
Many classical ecotoxicology approaches focus on the effect of a substance on mortality, growth, or reproduction. However, pollutants can also have more subtle sublethal effects on insects, which may have severe consequences in the long term (Straub et al., 2020).
Direct sublethal effects include changes in physiology such as stress reactions or detoxification processes. By interacting with microorganisms inside the insect's body, oral exposure to pollutants may indirectly affect insect health.

While many functions of the bacterial symbionts in bumblebees
have been proposed, only very few have been demonstrated in experiments (Hammer et al., 2021;Zhang & Zheng, 2022). The gut microbiome of bumblebees may be important for detoxification as microbiota-free individuals had lower survival when exposed to toxic concentrations of selenate (Rothman et al., 2019). Moreover, resistance to the common trypanosomatid parasite Crithidia bombi is higher in bumblebees with an intact microbiome compared to microbiota-free individuals (Koch & Schmid-Hempel, 2011b). When infected with C. bombi the outcome varies with host microbiota composition rather than genotype .
Examining the effects of anthropogenic pollutants, such as airborne particulate matter, on the gut microbiome is an important tool for assessing their risk for insect health (Duperron et al., 2020). Even with a conserved gut microbiome, the relative abundance of core bacteria and the presence of other microorganisms will vary with age, diet, and changing environmental parameters Kwong & Moran, 2016). Different pollutants affect the microbial composition of bee guts. In honeybee workers, pesticides or antibiotics change the relative and absolute abundance of core gut microbiota species (DeGrandi-Hoffman et al., 2017;Motta et al., 2018;Raymann et al., 2017). An array of environmental toxicants, such as cadmium, copper, selenate, and hydrogen peroxide, alter the gut microbiome of Bombus impatiens at field-realistic concentrations (Rothman et al., 2020). These shifts in the microbial community may affect bumblebee health. Intestinal dysbiosis, compositional and functional alteration of the microbiome, is associated with various diseases and health problems in humans and vertebrates (DeGruttola et al., 2016;Levy et al., 2017;Shreiner et al., 2015). In insects, dysbiosis negatively affects reproductive fitness, immunity, and resistance to pathogens (Ami et al., 2010;Daisley et al., 2020;Raymann et al., 2017).
Transcriptome analysis is a sensitive tool to characterize sublethal effects of potentially harmful substances on a molecular and cellular level (Prat & Degli-Esposti, 2019;Schirmer et al., 2010).
Changes in gene expression help to identify biological processes, such as stress responses and detoxification processes, at an early stage. Exposure to different pollutants have been shown to induce changes in gene expression in several insect species. Mosquitos (Aedes aegypti) exposed to anthropogenic pollutants (insecticides, PAHs) increased the expression of genes related to detoxification, respiration, and cuticular proteins (David et al., 2010). Fireflies (Luciola leii) showed a similar response when exposed to benzo(a) pyrene, a widespread PAH . In different bee species, the neonicotinoids imidacloprid, thiamethoxan, and clothianidin induce an upregulation of metabolic, immune, and stress response genes (Aufauvre et al., 2014;Bebane et al., 2019;Christen et al., 2018;Colgan et al., 2019;Gao et al., 2020;Shi et al., 2017).
The expression of genes related to detoxification was higher in honeybees (A. mellifera) exposed to heavy metals than in controls (Al Naggar et al., 2020;Gizaw et al., 2020;Zhang et al., 2018).
In contrast to pesticides, the effects of other environmental pollutants, such as particulate matter, on gene expression in bees as well as their gut microbiome are largely unclear. To address this knowledge gap, we exposed workers of the buff-tailed bumblebee Bombus terrestris to airborne particulate matter deriving from traffic and investigated changes in the gut microbiome and gene expression. Bumblebees were fed sugar water spiked with sublethal concentrations of brake dust or diesel exhaust particles (DEPs). Adding to this oral exposure, one group of bumblebees was exposed to DEPs via air to enable potential uptake in the tracheal system. We expect changes in the composition of the gut microbial community, as previous research showed changes due to different metals in a closely related Bombus species (Rothman et al., 2020). Moreover, we expect changes in the expression of detoxification and metabolic genes, indicating an increased stress level, as the toxic compounds in the particulate matter may interfere with bumblebee physiology.

| Dose selection
The data on airborne particulate matter in terrestrial environments is sparse as it is difficult to quantify and identify the origin.
Evidence for high levels of input of airborne particulate matter are often revealed only after it has settled, for example, by analyzing soil samples. Unnaturally high amounts of specific metals could be attributed to external resources such as brake dust (Alsbou & Al-Khashman, 2018;Peikertova & Filip, 2016). Isotopic analyses of urban soils in Arizona revealed up to 0.54% (w/w) as soot carbon black presumably produced by burning fossil fuels (Hamilton & Hartnett, 2013). While bees are contaminated by airborne particulate matter in the wild, we have no data or modeling on the uptake of these particles (Negri et al., 2015). In previous experiments, chronic oral DEP exposure over 7 days reduced survival of bumblebees when exposed to concentrations of 1 g/L and more (Hüftlein et al., 2023). Oral exposure to brake dust particles reduced survival

| Experimental procedure
At the beginning of the experiment, adult workers from the four colonies were randomly assigned to one of six treatments. Control: fed with sugar water only (50% Apiinvert) (n = 56); Solvent control: fed with sugar water spiked with 0.02% (v/v) of the emulsifier Tween20 (n = 56); Brake dust: fed with sugar water spiked with 0.02% (v/v) of the emulsifier Tween20 and 0.4 g/L brake dust particles (n = 56); DEP: fed with sugar water spiked with 0.02% (v/v) of the emulsifier Tween20 and 0.4 g/L diesel exhaust particles (n = 56); Flight control: fed with sugar water (50% Apiinvert) and allowed to fly once per day in a plastic box (7 × 7 × 5 cm, EMSA, Emsdetten, Germany) for 3 min (n = 24); DEP flight: fed with sugar water (50% Apiinvert) and allowed to fly once per day for 3 min in a plastic box (7 × 7 × 5 cm, EMSA, Emsdetten, Germany) that contained 1.5 (±0.1) mg of diesel exhaust particles (n = 24).
The experiment was conducted in a climate chamber at 26°C and 70% humidity under a constant 12:12 h light: dark cycle. Bumblebees were kept in Nicot cages (Nicotplast SAS, Maisod, France) connected to a 12 mL syringe (B. Braun SE, Melsungen, Germany) with the tip cut off, that contained 2 mL of the respective feeding solution (ad libitum).
Every day the syringes were replaced with fresh ones to prevent molding or bacterial growth in the food. The exposure lasted for 7 days. At the end of the experiment, the animals were frozen at −20°C.
Within a week after the end of the experiment, we randomly selected twelve (three workers per colony) bumblebees per treatment for transcriptome analysis (N = 72). Additionally, for the control, solvent control, brake dust, and DEP treatment, we randomly selected 20 bumblebees (five workers per colony) for microbiome analysis (N = 80), respectively.

| Generation and collection of diesel exhaust particles (DEPs)
Diesel exhaust particles were collected from a four-cylinder diesel engine (OM 651, Daimler AG, Stuttgart, Germany) during a repeating cycle of transient and stationary operating points, resembling an inner-city driving scenario with stop-and-go intervals. The engine was operated on a test bench with a water-cooled eddy-current brake as previously described in Zöllner (2019). DEP samples were collected by an electrostatic precipitator (OekoTube Inside, Mels-Plons, Switzerland). A fast response differential mobility particulate spectrometer DMS500 (Combustion, Cambridge, England) was applied to measure submicron particle size distributions of raw exhaust samples. Depending on engine load and speed during the inner-city cycle, solid particles showed a median diameter between 52.1 ± 1.8 nm and 101.9 ± 1.7 nm. DEP composition was characterized by thermogravimetric analysis (TGA, STA 449 F5 Jupiter, Netzsch-Gerätebau GmbH, Selb, Germany). A fraction of 72.2% ± 1.1% of the DEP mass was attributed to elemental carbon, 23.2% ± 0.9% w/w to organic fractions, and 4.6% ± 0.7% w/w to inorganic matter. Quantification of PAHs revealed concentrations of 444 ppm for pyrene, 220 ppm for phenanthrene, and 107 ppm for fluoranthene.
The elemental composition of the DEP samples was analyzed by inductively coupled plasma optical emission spectrometry (ICP-OES, Optima 7300 DV, PerkinElmer Inc., Waltham, United States of America) and interpreted according to Zöllner (2019). It showed fractions of calcium (1.63% w/w), zinc (0.53% w/w), and phosphorus (0.50% w/w) that can be traced back to diesel fuel and lubrication oil.

| Bumblebee gut microbiome analysis
Prior to dissection bumblebees were defrosted and rinsed in 70% ethanol, 90% ethanol, and twice in ultrapure water. We placed each bumblebee on an autoclaved square of aluminum foil (5 × 5 cm) and opened the abdomen with sterilized tweezers and scissors. After carefully separating the midgut and hindgut from the crop and transferring it to an Eppendorf tube, we snap-froze the gut in liquid nitrogen. All samples were stored at −80°C until further processing.

| PCR amplification and sequencing of 16S rDNA fragments
Metagenomic DNA of bumblebee gut samples was purified using the
Forward reads of 16S rDNA fragments (R1 reads) were analyzed using the QIIME2 microbiome analysis package (ver. 2021.11; Bolyen et al., 2019). Unless indicated otherwise, all analysis tools were used as plugins of the QIIME2 package. The respective parameters used along the analysis steps are readily accessible by provenance information in the QIIME2 data files (available as Appendix S1). In brief, the following analysis steps were performed: Demultiplexed reads were trimmed for 16S primer sequences (plugin cutadapt; Martin, 2011), denoised, dereplicated, and chimera-checked (plugin DADA2; Callahan et al., 2016) resulting in amplified sequence variants (ASVs). Rare ASVs were filtered using the median frequency (=6) of ASVs over all samples.
Taxonomic classification of ASVs was performed (plugin featureclassifier; Bokulich et al., 2018) using the prefitted sklearn-based taxonomy classifiers based on the SILVA reference database (ver. 138.1; Quast et al., 2013;Yilmaz et al., 2014). ASVs that could not be taxonomically assigned at any taxonomic level ('unassigned') as well as samples with less than 3900 reads in total were removed prior to subsequent analysis steps. Alpha diversity metrics, such as Shannon diversity index, Faith's phylogenetic diversity, Pielou's evenness, and observed ASVs, were obtained using the QIIME2's 'core-metrics-phylogenetic' workflow (plugin diversity), rarefied to 3900 reads per sample. To assess the overall effects of treatment and colony origin on microbial composition we performed permutational multivariate analysis of variance ADONIS from the R package vegan (Oksanen et al., 2022) in Qiime2. To find significant differences in α-diversity we fitted generalized linear mixed models (GLMMs) with treatment as fixed factor and colony as random factor using the function glmmTMB from the package glmmTMB (Brooks et al., 2017). We checked model assumptions using model diagnostic test plots, that is, qqplot and residual vs. predicted plot from the package DHARMa (Hartig, 2022).
We then produced statistics with the function Anova() from the package car (Fox & Weisberg, 2019) to calculate p-values for differences between treatments. For significant treatment effects, we ran pairwise comparisons using Tukey HSD post-hoc test with Benjamini-Hochberg correction from the package multcomp (Hothorn et al., 2008). Differential abundance of the rarefied data we analyzed using the package DESeq2 with a negative binomial distribution, a significance level cutoff of FDR < 0.01, replacement of outliers turned off, and cooksCutoff turned off (Love et al., 2014). Compositional differential abundance analysis was performed using Aldex2 (plugin aldex2; Fernandes et al., 2013).

| Transcriptome analysis of whole bumblebee abdomens
Bumblebees were defrosted and rinsed in 70% ethanol, 90% ethanol, and twice in ultrapure water prior to dissection. The abdomen was cut off with sterile scissors, placed in an Eppendorf tube, and snap-frozen in liquid nitrogen. All samples were stored at −80°C until further processing.

| RNA sequencing
Total RNA was prepared from abdomen samples using the RNeasy

| Differential expression analysis
RNA-Seq reads were further analyzed using the OmicsBox bioinformatics platform (v. 2.0.36, www.biobam.com). Unless indicated otherwise, all tools used for differential expression analyses are accessible within the OmicsBox platform. RNA-Seq reads were preprocessed by Trimmomatic (details see Appendix S1: RNAseq_1_trimmomatic_report) (Bolger et al., 2014) to remove sequencing adapters, low-quality sequences, and short reads from the dataset. The quality-trimmed reads were mapped to the B.
A gene-specific count table was created from the mapping files using HTseq (Anders et al., 2015) and differentially expressed genes were identified by edgeR (Robinson et al., 2010), respectively. Functional annotation of the B. terrestris genome was based on annotation release v. 102 (available in gff3 format from metaz oa.ensem bl.org). Since 4975 of the 12,008 genes did not contain any functional annotation, the functional annotation workflow of the OmicsBox platform was used to update the published annotation with additional information. In brief, the coding sequences of unannotated genes were used to extract functional annotations from refseq_protein database (www.ncbi.nlm.nih.gov) and InterProScan (www.ebi.ac.uk). These we then fed into the GO mapping and annotation tools of the pipeline and finally merged to the existing functional annotations. To assess the overall effects of treatment and colony origin on gene expression we performed permutational multivariate analysis of variance ADONIS from the R package vegan (Oksanen et al., 2022) in Qiime2.

| Effect of pollutants on the bumblebee gut microbiome
Amplicon sequencing of the bacterial 16S rDNA fragments yielded a total of 2,425,928 raw reads. After quality filtering and removal of unassigned sequences, we also removed samples with a sampling depth below 3900 reads (n = 7), all from DEP treatment, to ensure adequate sampling depth (13 DEP replicate samples remained in the analysis). In the remaining samples we obtained 1,856,025 16S rDNA gene sequences with a mean of 25,425 reads per sample (n = 73), corresponding to 468 amplicon sequence variants (ASVs).
Sample-based rarefaction curves suggest a sufficient sequencing depth for a representative coverage of the microbiome as most of the samples reach a plateau ( Figure A1). ADONIS analysis revealed a significant effect of treatment on microbiome composition (R 2 = 0.423, p < .001). There was no significant effect of colony origin (R 2 = 0.001, p = .946) on microbiome composition.

| Taxa abundance
On the genus level, the most common bacterial taxa (>1% in at least one treatment) were: Gilliamella, Snodgrassella, Lactobacillus, Asaia, Bombiscardovia, Methylorubrum, and Bombilactobacillus. The relative abundance of the most common genera for each sample shows a different microbial composition in the DEP treatment compared to the other treatment groups (Figure 1).
While the relative abundance of ASVs did not differ between control, solvent control, and brake dust, DEP treatment had 16 differentially abundant ASVs compared to the control, according to DESeq2 (Figure 2, Table A1). Eleven ASVs had a higher abundance in the DEP treatment than control. Five ASVs had reduced abundance in comparison to the control treatment. A more conservative approach to identify differential abundance is ALDEx2, which revealed five ASVs with significantly altered abundance in the DEP treatment compared to the control: Snodgrassella 1 + 2, Neisseriacae, Lactobacillus bombicola, and Bombiscardovia (Table A2).

| β-diversity of the gut microbiome
The community composition of the bumblebee gut microbiome differed between treatments indicated by significant differences between the robust Aitchison distances (Overall PERMANOVA F I G U R E 1 Relative abundance of the most common bacterial genera for each sample. Samples are arranged according to treatment.

| Effect of pollutants on bumblebee gene expression
In the transcriptome analysis, we focused only on biologically relevant comparisons of treatments to prevent unnecessary inflation of reported results. We compared control vs. solvent control, control vs. DEP, control vs. brake dust, and flight control vs. DEP flight. The analysis for differently expressed genes (DEGs) revealed differences between our treatments. In total, 324 genes were differentially expressed in the DEP treatment compared to the control (lowcount gene filter settings: CPM Filter = 1, samples reaching CPM Filter = 2). 165 genes were upregulated (LogFC > 1) and 159 genes downregulated (LogFC < −1), respectively (Table A3, Figure A2). In the brake dust treatment only one gene (lipase 3) was differentially expressed (upregulated) in comparison to the control. In the solvent control, there were no differentially expressed genes compared to the control. In the DEP flight treatment, we found no differentially expressed genes in comparison to the flight control. ADONIS analysis revealed a significant effect of treatment on gene expression (R 2 = 0.279, p = .002). There was no significant effect of colony origin (R 2 = 0.031, p = .054) on gene expression.
The variation in gene expression of bumblebee workers is clearly distinct between the control and the DEP treatment ( Figure 5). The F I G U R E 2 Log 2 fold change in relative abundance of ASVs in the DEP treatment in comparison to the control. Cutoff for inclusion of ASVs in this plot was FDR (=p adj ) < .01. Colors represent most specific taxonomic label.  The functional network analysis based on -Score ≥ 0.4 for differentially expressed GO terms with FDR ≤0.05 in the DEP treatment shows clustering to specific functional groups ( Figure A6a).
Upregulated functions are related to phosphorylation, regulation of metabolic process, guanyl nucleotide binding, and signal transduction ( Figure A6b). Downregulated functions are related to mitochondria, lipid metabolic processes, the endoplasmic reticulum, and phospholipid biosynthetic processes ( Figure A6c).

| DISCUSS ION
In this study, we found that oral exposure to diesel exhaust particles (DEPs) changes the gut microbiome and gene expression of bumblebee workers, while DEP exposure via air did not. Brake dust, the second pollutant we tested via oral exposure, did not induce changes in the gut microbiome or gene expression in the bumblebee workers.
While the composition of the microbial gut community in control, solvent control, and brake dust exposure treatment was similar, we detected major shifts in the DEP treatment. This raises several interesting questions: (1) (Patel et al., 2020;Sun et al., 2021).
Also, shifts in the microbial gut community due to PAH exposure have been reported in different animals, such as fish, sea cucumbers, or potworms (Enchytraeidae) (DeBofsky et al., 2020(DeBofsky et al., , 2021Ding et al., 2020;Quintanilla-Mena et al., 2021;Zhao et al., 2019). Therefore, we suspect PAHs to be the leading cause of changes in the bumblebee gut microbiome in our study. However, the large amount of elemental carbon in DEPs, may itself provide another explanation. The DEPs may function like activated carbon with its large surface-area-to-volume ratio and may adsorb microbes that are then discharged by excretion (Naka et al., 2001;Rivera-Utrilla et al., 2001;Wichmann, 2007). Even though activated carbon has no direct negative impact, constant adsorption and discharge might disrupt the bacterial community resulting in the compositional and quantitative changes similar to those observed in our study.
The bacterium Snodgrassella, one of the dominant core bacteria in undisturbed gut microbiomes of bumblebees (Hammer et al., 2021), is nearly absent after the DEP exposure. Snodgrassella, together with Gilliamella, forms a biofilm coating the inner wall of the ileum (Hammer et al., 2021;Martinson et al., 2012). Both host and symbionts could profit from this biofilm formation as it prevents bacteria from washout and enables the formation of a syntrophic network (Kwong et al., 2014;Powell et al., 2016;Zhang & Zheng, 2022).
Additionally, the biofilm could protect the host against gut parasites, such as C. bombi, which need to attach to the gut wall to persist (Koch et al., 2019;Näpflin & Schmid-Hempel, 2018). However, the mutualistic relationship between the microbes seems to be disrupted by DEP exposition, as Snodgrassella abundance is extremely diminished.
In contrast, Gilliamella increases in relative abundance after DEP exposure. This indicates that Gilliamella may be able to form a biofilm independently from Snodgrassella. A relatively simple explanation for the higher relative abundance of Gilliamella might be that the reduction of Snodgrassella leaves Gilliamella as the only dominant bacterium in the gut, and therefore, Gilliamella might thrive better or fill the void. Snodgrassella seems especially prone to pollutants, as Rothman et al. (2020) already reported a decrease in its relative abundance after exposure of bees to copper, selenate, or glyphosate. Additionally, we found an unknown bacterium from the family Neisseriaceae, the same family to which also Snodgrassella belongs, having a lower relative abundance after DEP exposure. If this is a consistent result, it might indicate a general susceptibility of this family to DEPs.
The higher abundance of Asaia in the DEP treatment was driven by two samples, in which Asaia dominates the bacterial community with relative abundances of 99% and 67%, respectively. Asaia is a flower-associated acetic acid bacterium, which is commonly found in the gut of members of different insect orders, such as Hemiptera, Diptera, and Hymenoptera (Bassene et al., 2020;Crotti et al., 2009;Kautz et al., 2013). It can dominate the gut microbiome of Anopheles mosquitos, which is why it is considered a potential tool in malaria control (Capone et al., 2013;Favia et al., 2008). While there have been reports of Asaia in bumblebees, the dominance of Asaia in some of the DEP samples is rather uncommon (Bosmans et al., 2018).
DEPs might disrupt the natural microbiome community opening the door for opportunistic bacteria such as Asaia (Favia et al., 2007).
Even though we kept the bumblebees in this experiment indoors throughout their lives, Asaia bacteria may derive from pollen fed to the bumblebees before the start of the experiment.
We detected an interesting pattern in the genus Lactobacillus, one of the core gut bacteria of bumblebees (Hammer et al., 2021).
While the species L. bombicola, a bumblebee-associated bacterium, has a lower abundance after DEP exposure, the abundance of the honeybee-associated L. apis increases. Again, the disruption of the original microbiome caused by DEPs might explain that foreign bacteria can establish themselves in the microbiome. As the pollen fed to the bumblebees before the experiment was collected by honeybees, it could be the source of L. apis.
The DEP-induced changes in the gut microbiome may affect bumblebee health, as core bacteria could prevent infections by parasites.
The abundance of Gilliamella, Lactobacillus, and Snodgrassella is negatively correlated with the parasites Crithidia and Nosema, while noncore bacteria are more abundant in infected bumblebees (Cariveau et al., 2014;Mockler et al., 2018). The biofilm formation of Snodgrassella and Gilliamella may form a physical barrier to the trypanosome C. bombi, which needs to attach to the ileum wall to persist (Koch et al., 2019, Näpflin & Schmid-Hempel, 2018. The disruption of this biofilm and the higher abundance of noncore bacteria, such as Asaia, may increase the parasite susceptibility of bumblebees exposed to DEPs. The transcriptome analysis revealed significant changes in gene expression after oral exposure of bumblebees to a sublethal dose of DEPs. In total, 165 genes were upregulated, and 159 genes were downregulated. GO enrichment analysis and network analysis indicate that these changes could be related to a general stress response against pollutants. While upregulated GO terms involve many metabolic and catabolic processes, downregulated GO terms include metabolic and biosynthetic processes. DEP exposure might deplete stored reserves causing the observed changes as a consequence of higher energetic costs. Changes in metabolism seem to be a typical reaction to pollutants in insects, which seems reasonable as they often interfere with biochemical processes. Transcriptional changes in bumblebees and honeybees exposed to sublethal doses of neonicotinoids are mainly linked to metabolic processes (Bebane et al., 2019;Colgan et al., 2019;Gao et al., 2020;Shi et al., 2017). Exposure to heavy metals or PAHs induces similar changes in spiders, mosquitos, moths, and fireflies (Chen et al., 2021;David et al., 2010;Li et al., 2016;Zhang et al., 2019Zhang et al., , 2021. Even though the changes differ in detail, certain processes seem commonly involved in the response to pollutants. Consistent with our findings, exposure to insecticides or PAHs affects mitochondrial functioning, an important part of the insect energy metabolism (Colgan et al., 2019;Zhang et al., 2019Zhang et al., , 2021. This supports the idea of increased energy demand caused by pollutants (Beyers et al., 1999;Calow, 1991). We also observed an upregulation of signal transduction in our study, similar to observations in honeybees and fireflies exposed to Imidacloprid and the PAH benzo(a)pyrene, respectively (Gao et al., 2020;Zhang et al., 2019Zhang et al., , 2021. Typically, chemical stressors, such as PAHs, insecticides, and heavy metals, affect genes associated with detoxification processes and drug metabolism (Chen et al., 2021;David et al., 2010;Gizaw et al., 2020;Zhang et al., 2019).
However, in our study, we did not find any differentially expressed detoxification-related genes. Possibly the number of PAHs attached to the DEPs was not enough to trigger a reaction that would lead to a measurable increase in detoxification. Overall, the observed changes in gene expression after oral DEP exposure of bumblebees resemble a general stress response to pollutants.
As microbiome and gene expression of bumblebees significantly changed after oral DEP the question arises if and how these systems might affect each other. Metabolic changes may be caused by the DEP-induced changes in the gut microbiome, which can potentially alter the type and amount of metabolites provided to the host (Douglas, 2018). Moreover, insect immunity might be dependent on gut microbiome. In honeybees the native gut microbiome stimulates immune gene expression, inducing the production of antimicrobial peptides (Kwong et al., 2017). The function and the mechanistic underpinning of this interaction is not entirely clear, but the host might regulate the microbiota in this way. However, host health might also benefit from this interaction by priming the immune system against future infections. Pollutants altering the gut microbiome might thereby jeopardize insect health. This could explain the increased mortality in honeybees with altered gut microbiome due to antibiotic exposure (Raymann et al., 2017).
In contrast to oral exposure, we did not find any effect on gene expression after exposure of bumblebees to DEPs via the air. To cause changes, DEPs need to enter the tracheal system or attach to sensory organs, such as the antennae. The exposure of bumblebees for 3 min per day may not have been enough to affect them. Particles on the antennae may have been removed quickly by cleaning behavior and the spiracles seem to be an effective protective barrier against the uptake of particles into the tracheae (Harrison, 2009;Schönitzer, 1986). Thus, our results should be taken with care because probably only very few particles entered the tracheal system of the bumblebees.
Unlike DEPs, oral exposure to brake dust particles did not affect the gut microbial community or the gene expression of the bumblebees. However, some concerns remain about the experimental procedure. For one, we did not use brake dust from a real braking scenario, but rather artificially milled brake pads. Dust derived from them may have different physicochemical properties. Milled brake dust particles have a much higher mean particle size than DEPs (10 μm vs. 0.01 μm). As we defined treatment concentration per weight, these different physical properties lead to big differences in the particle counts of the treatment solutions, that is, solutions with brake dust contained far fewer particles than those with DEPs. Moreover, large brake dust particles tend to sink to the bottom of the feeding syringes, which might have reduced the particle uptake. While brake dust seems not to affect the bumblebees, further studies are needed to address the indicated limitations of the present study.
Another problem that needs to be addressed is how the doses used in this study relate to field-realistic concentrations encountered by bumblebees. With the still often vague knowledge of origin and quantity of airborne fine particulate matter present in terrestrial habitats, we know even less about their potential uptake by insects. Contamination of bee products is documented, but there is a need for realistic modeling of encounter rate of insects with airborne particulate matter (Conti & Botrè, 2001). The doses used in this study are presumably higher than those encountered naturally. However, our experimental setup does not include other stressors bees have to face in the wild, such as parasites, limited food availability, or abiotic factors such as drought or heat stress. Bumblebees may be able to compensate for facing one stressor but will eventually be overstrained by multiple stressors.
Taken together, the results from our microbiome and transcriptome analysis indicate potential consequences for insect health, here shown in bumblebees, after oral DEP exposure. Gut dysbiosis may increase the susceptibility of bumblebees to pathogens, while a general stress response may lower available energetic resources.
This highlights the potential role of airborne particulate matter such as DEPs as a driver of insect declines.

DATA AVA I L A B I L I T Y S TAT E M E N T
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. The microbiome and RNA-Seq sequencing data were deposited at NCBI's Sequence

F I G U R E A 2
Differential expression of genes in the DEP treatment in comparison to the control. Blue dots represent significantly downregulated genes, red dots represent significantly upregulated genes. The horizontal red line marks a −log 10 (FDR = 0.05). The two vertical red lines mark a log 2 FC of −1 and 1, respectively.

F I G U R E A 3
Non-metric multidimensional scaling plot based on the log 2 fold changes (FC) between control and solvent control. The axes of the nMDS plot represent dimensional reductions of genes expression visualizing the variability of the transcriptional changes for each treatment. Each point represents one sample, colored according to the respective treatment.

F I G U R E A 4
Non-metric multidimensional scaling plot based on the log 2 fold changes (FC) between control and brake dust treatment. The axes of the nMDS plot represent dimensional reductions of genes expression visualizing the variability of the transcriptional changes for each treatment. Each point represents one sample, colored according to the respective treatment.
F I G U R E A 5 Non-metric multidimensional scaling plot based on the log 2 fold changes (FC) between flight control and DEP flight treatment. The axes of the nMDS plot represent dimensional reductions of genes expression visualizing the variability of the transcriptional changes for each treatment. Each point represents one sample, colored according to the respective treatment.

F I G U R E A 6
Network analysis of enriched gene terms and functional groups in the DEP treatment based on Kappa-Score ≥ 0.4 for GOs with FDR ≤ 0.05 using the ClueGo and CluePedia plugins of Cytoscape. (a) Functionally grouped network of upregulated (red) and downregulated (blue) gene ontologies. (b) pie chart with functional groups, including specific terms upregulated in the DEP treatment. (c) pie chart with functional groups, including specific terms downregulated in the DEP treatment. The area covered by each group represents the relative number of GO terms within each group. The most significant term each group is labelled.

TA B L E A 3
Differentially expressed genes in the DEP treatment compared to the control (low-count gene filter settings: CPM Filter = 1, samples reaching CPM Filter = 2).