In this study, exposure to various concentrations of clothianidin left different signatures of microbiota dysbiosis in the three gut sections of honeybees. Changes of correlations reveal pathogen activity spikes correlated to different patterns of mutualist imbalance according to gut section. Such dysbiosis patterns were expected as gut sections are colonized by specific microbial communities (Fig. 1–3; Supplementary Fig.S1-S3), each microbial community differentially shaping the environmental conditions of each gut section . In the midgut and the rectum, a decrease of ASVs activity correlation was detected in all exposed groups to clothianidin. In the ileum, two types of variations occurred: significant correlations among ASVs increased under 0.1 and 1 ppb groups; and decreased for the 10 ppb group.
We expected that a clothianidin exposure gradient would significantly increase positive and negative correlations between pathogens [41, 67]; and a change of correlation type between core members . Despite disrupted microbial activity correlations, dominant core and non-core members were still active in all test groups, as reported in a previous study .
Our work highlights the importance of low activity taxa in the gut microbiota stability as illustrated by the loss of Devosia and Leifsonia; and exposed in previous studies [62, 69–72]. Despite their low bacterial activity, these taxa are characterized by high degree in the correlational networks: DG, CC and NC within the control midgut network (Supplementary Table S9).
To understand why the lowest clothianidin concentration induced the lowest honeybee survival, we investigated whether a specific signature of microbiota dysbiosis could be associated to the low clothianidin (0.1 ppb) treatment group. The loss of positive correlations with the increasing neonicotinoid concentrations along the gut section, is a landmark of gut dysbiosis . As stated above, the most extensive adverse impact in terms of correlational network structure, was recorded at 0.1 ppb (Fig. 3; Supplementary Fig.S1A,S2A). Noticeably, the gain of DG and the high CC and NC for Lawsonella, a human pathogen  supports its important central role inside the ileum (Supplementary Table S13-14), in addition to the important decrease of DG and NC in the rectum (Supplementary Table S17-18) for Bifidobacterium, Pediococcus and Commensalibacter: known as probiotics [15, 75, 76]. Then, numerous correlations occurred with low activity taxa in the midgut (Fig. 3) with the high NC (Supplementary Table S10) for Moraxella  suggesting a potential pervasive adverse effect on the overall network. Also, the majority of correlations (negative and/or positive depending the gut section) with two beneficial strains, Pediococcus and Lactobacillus involved potential pathogens such as Thiohalobacter  and Prevotella .
The significant impact of different pesticide concentrations on the overall network may depend on the microbial strain that first metabolizes the molecule, which influences the syntrophic exchange network. For instance, honeybee gut microbiota exposed to two fipronil concentrations did not respond similarly: the lowest concentration (0.25 µg/kg) affected Bifidobacterium sp. abundance, with no significant bees mortality increase, whereas the highest concentration (1 µg/kg) did not affect Bifidobacterium sp. abundance, but induced a significant bee mortality increase . Daisley et al.  showed in gnotobiotic Drosophila that the pesticide chlorpyrifos was more toxic than its metabolite (chlorpyrifos oxon). In our case, the final metabolites’ toxicity, according to the initial clothianidin concentration, could differentially impact the honeybee physiology, and in turn, survival. Interestingly, clothianidin quantification with LC-MS/MS suggests a slower degradation at 0.1 ppb (Supplementary Table S7-S8) which could result from a different clothianidin metabolization pathway, potentially translating into lower survival. Previous works reported clothianidin degradation by Flavobacterium and Pseudomonas sp.  and imidacloprid degradation by Leifsonia sp. . These taxa were mostly impacted at 0.1 ppb for Flavobacterium and Pseudomonas across all gut sections (Fig. 3; Supplementary Fig.S4B; Fig. 5A; Table S10;14;18) and for Leifsonia in the ileum and rectum (Fig.S4B; Fig. 5A; Supplementary Table S14, S18). These taxa are widely connected with the overall network (high NC). In addition, Flavobacterium and Pseudomonas played a central position inside the network (high CC).
Within all these three gut sections, we observed a gain of edges/significant correlations (positive and/or negative) for the genera Bifidobacterium, Frischella, Gilliamella, Lactobacillus, Parasaccharibacter and Snodgrassella. All these bees symbionts are known to be involved in either host immunity or maintaining a homeostatic microbiota [38, 81–84]. For example, Gilliamella apicola  and Lactobacillus  are responsible for short-chain fatty acid production, and their diminishing activity may likely alter the host’s immunity . Moreover, the functional complementarity between Snodgrassella alvi and G. apicola ensures homeostatic microbiota in the intestine ecosystem . Frischella perrara  and Parasaccharibacter spp.  were documented as important key factors in the immune system.
Investigating the local effect of clothianidin gradient on the gut microbiota structure, we found a gain of correlations (positive and/or negative) among low activity taxa. These taxa who showed to be involved in: 1) disturbed environment and/or in pathogens activites includes genera like Moraxella , Lawsonella , Thiohalobacter , Ralstonia , Leifsonia , Prevotella , Psychrobacter [39, 91] and Pseudaminobacter ; and 2) in healthy environment as probiotic and/or in antimicrobial activity. These low activity taxa are belonging to the genera Oceanobacillus , Blautia [94, 95], Bacillus [54, 96] and Lysinibacillus [93, 97].
Within the midgut, exposure to 0.1 ppb clothianidin (Fig. 3) was more harmful relatively to the other concentrations as supported by an increase of DG and CC; mainly impacting Lactobacillus genus activity (Supplementary Table S9-S10). Increase of these topological parameters (DG and CC) support how the Lactobacillus genus became centrally connected inside the network at 0.1 ppb. Lactobacillus genus is known in improving the immune system, resistance against pathogens [21, 40, 98, 99] and reduced pesticide toxicity  supporting how Lactobacillus spp. activity is favorable for the host.
Complementarily, the loss of connectivity (diminishing DG) for Bifidobacterium and Pediococcus (diminishing CC) (Supplementary Table S9-S10). supports clothianidin adverse impact on these genera known for their probiotic properties [15, 75, 76, 101, 102]. However, the increase of NC for Bifidobacterium suggests a likely favorable cooperation inside the network. Network node metrics suggest a pathogenic shift compensated by mutualistic correlation following exposure to clothianidin. Noticeably, clothianidin induced a shift from a positive to a negative correlation activity between Frischella (decreasing activity) and Pediococcus (increasing activity). This competition shift suggests a dysbiosis pattern .
Within the ileum, 1 ppb treatment (Fig. 4B) was more adverse relatively to the other concentrations, mainly targeting Bifidobacterium genus activity, supported by an increase of DG, CC and NC (Supplementary Table S13, S15).
Within the rectum, 10 ppb treatment (Fig. 5B) was more adverse relatively to the other concentrations, mainly impacting Frischella genus activity, with a gain of nine significant local correlations supported by an increase of DG and CC, and a loss of connectivity with the overall network, as supported by a decrease of NC (Supplementary Table S17, S20).
Taken together, pathogens, and correlations rise are not surprising as clothianidin induced negative correlations between core and non-core members within each gut section. Given that all these symbionts are either involved in host immunity and/or microbiota equilibrium [38, 81–84] and that clothianidin neonicotinoid could potentially disrupt the microbiota-immunity axis [21–23], our results suggest that clothianidin exposure induced dysbiosis is at least targeting, the microbiota-immunity axis.
More specifically, the impact due to pathogen invasion was variable along the gut section and gradient of exposure. Ralstonia genus impacted much more the midgut network (0.1 ppb) (Fig. 3) and the ileum (1 ppb) (Fig. 4B). The high CC and NC for Ralstonia endorse its DG and important NC (Supplementary Table S10, S15) within the overall network suggesting this strain exerts an important negative impact on microbiota structure following clothianidin exposure. Our results provide additional evidence from the lens of bacterial activity that honeybees exposed to neonicotinoids are more sensitive to microbial gut pathogens [33, 103, 104]. In addition, the dysbiosis patterns may suggest a shift from opportunistic to infectious symbionts by stressors , which could potentially result from an impairment from resistance to colonization (i.e. the microbiota resistance to pathogen invasion ).
Previous studies already highlighted fluctuations of honeybee microbiota. Application of coumaphos, tau-fluvalinate  and tetracycline  showed to increase Giamella apicola prevalence. Different experimental approaches may induce microbial composition variability  as observed in our work with G. apicola and Snodgrassella alvi. The authors [107, 108] used entire bee gut, sampled from hive, whereas we used gut sections, sampled from cages. Snodgrassella alvi abundance decreasing in the ileum is consistent with [109, 110] who reported the same observations in the bee gut after glyphosate exposure. Moreover, S. alvi is responsible in the upregulation of the gene’s expression related to antimicrobial peptide  and as honeybee gut biofilm pioneer, S. alvi destabilization may create an overall disbalance of gut microbiota. In our experiment, it is likely the decreasing Snodgrassella actiivty helped in adversely impacting the immune system of Apis mellifera.
We observed a decreasing in Frischella perrara activity, while exposure to other pesticides was variable as it either induced (nitenpyram)  or not (imidacloprid) an increase of F. perrara abundance . F. perrara is known to have a key role in honeybee immunity in limiting microbial resistance , therefore F. perrara disbalance might affect the honeybee gut immunity, leading to microbiota dysbiosis. Finally, the increasing activity of Bifidobacterium is in agreement with previous studies [19–21] that tested nitenpyram and thiacloprid exposure on honeybees.
Altogether, the initial pesticide concentration has differentially impacted the microbiota correlation network. For any given concentration, different members could first handle the molecule and thus determine the kind of metabolites excreted, and in turn determine its toxicity, which may not be necessarily proportional to the concentration. Further analysis on metabolites generated by Flavobacterium spp., Leifsonia spp. and Pseudomonas spp. grown in media with a clothianidin concentration gradient might give more explanations on their ability to modulate clothianidin toxicity on honeybee.
Finally, it is documented that honeybees are more attracted to contaminated than non-contaminated syrup . A dose-dependent attraction was observed for Nitenpyram where food consumption was negatively correlated with pesticide concentration: normal with 3–30 µL/L, low with 300 µL/L . In our case, honeybees exposed to 0.1 ppb consumed significantly more syrup, translating into higher clothianidin bioaccumulation in the 0.1 ppb group, relatively to the other groups (Fig. 2, Supplementary Table S6). Previous studies highlighted how nutritional stress shapes the gut microbiota composition [113, 114] inducing long term negative impacts on honeybee health . Therefore, it is very likely that higher bioaccumulated clothianidin in younger bees (with an immature microbiota) would have long-term health effect.