A deeper understanding of system interactions can explain contradictory field results on pesticide impact on honey bees

Honeybee colony losses have dramatically increased, posing a threat to pollination services and food security. These losses are multifactorial, with interacting stressors like parasites, pathogens (e.g., DWV), agrochemicals (especially neonicotinoids), forage availability, and environmental conditions affecting bee health and colony stability. While laboratory studies clearly demonstrate neonicotinoid toxicity, field studies yield inconsistent results, with some showing no detectable negative effects. This inconsistency has been attributed to the buffering capacity of honeybee colonies, but the underlying mechanisms remain unclear. The variability in study contexts, including stress factors, nutrition, and foraging resources, complicates the issue. The lack of consistent scientific evidence has led to differing regulatory approaches across regions, despite similar situations. To address this knowledge gap, the researchers adopted a systems biology approach.

Existing literature highlights the multifaceted nature of honeybee decline, involving various interacting factors that affect bee health and colony stability. Several studies have shown the negative effects of neonicotinoid insecticides on bees in laboratory settings, impacting navigation, immunity, and reproduction. However, field studies have yielded conflicting results, with some reporting no significant negative effects on honeybee colonies near neonicotinoid-treated crops, while others observe both positive and negative impacts. This lack of consistent field-based results necessitates a deeper understanding of the complex interplay of factors influencing bee health. The buffering capacity of honeybee colonies has been suggested as a possible explanation for the inconsistent findings, but the mechanisms underlying this buffering remain to be elucidated.

The researchers developed a conceptual model of honeybee health influenced by multiple factors: *Varroa destructor* mites, DWV, toxic compounds (neonicotinoids), suboptimal temperatures, nectar, pollen, immune response, and detoxification system. This model was mathematically represented using a system of ordinary differential equations (ODEs), accounting for interactions among honeybee health, toxic compounds, parasites, and pathogens. External inputs like sugar, pollen, and temperature were also included. The model's structure was analyzed using a community matrix and structural influence matrix, examining parameter-independent system properties. The analysis focused on the signs of interactions to infer qualitative behaviors, circumventing the need for precise parameter values. The system's equilibria, representing stable states of bee health, were then studied, differentiating between cases with and without immune-suppression by DWV. In the case with immune-suppression, bifurcation theory was used. To validate the model's predictions, the researchers conducted laboratory experiments on caged honeybees over six years using a standardized method. Experiments were done on bees with and without DWV infection. The survival rates of bees exposed to nicotine (as a toxic compound representative) or suboptimal temperatures were compared. The presence of DWV and its effect on survival were assessed through qRT-PCR analysis. Additional experiments were carried out using a honeybee population model showing the effects of forager bee mortality on colony stability, modified to include the premature death of young bees.

The structural analysis revealed that any additional stressor negatively impacts honeybee health. However, analysis of system equilibria revealed a crucial insight: a pathogen that impairs the bee immune response (like DWV) can introduce a critical positive feedback loop and cause bistability. Bistability generates two stable equilibria: one with high bee health and one with low bee health. The existence of an intermediate unstable equilibrium means small variations in initial conditions can lead to vastly different outcomes: prolonged survival or premature death for the same stressor level. Laboratory experiments confirmed these predictions: in the presence of DWV, honeybee survival exhibited significantly higher variability, with some bees dying prematurely and others surviving much longer, even under identical exposure conditions to nicotine or suboptimal temperatures. The spread of DWV is facilitated by *V. destructor*, the mite vector, emphasizing the importance of considering multiple stressors. In a mathematical model simulating honeybee colonies, the inclusion of premature death of young bees, mirroring the effect of DWV, decreased the threshold for forager bee mortality causing colony collapse. This implies that the effect of DWV worsens the risk of colony failure.

The findings address the research question of contradictory field results by demonstrating that the complex interplay of stressors, particularly the presence of an immune-suppressing pathogen like DWV, significantly impacts the outcome of pesticide exposure. The bistability observed in the model and experiments highlights the limitations of single-stressor approaches in assessing pesticide risk. The study's findings suggest that the buffering capacity of honeybee colonies can prevent collapse under low DWV prevalence and limited other stressors. However, when DWV is prevalent, the bistability increases colony vulnerability to collapse. The data supports the conclusion that low DWV prevalence and/or mite infestations may explain the absence of adverse effects from neonicotinoids in some field studies. The study also suggests that pesticides impairing the bee's detoxification system could cause similar dynamic behaviour.

This research demonstrates that a systems biology approach, considering the interplay of multiple stressors and system architecture, is crucial for understanding honeybee health and the impact of pesticides. The observed bistability, driven by immune-suppressing pathogens, reveals why field studies produce inconsistent results. Future research should investigate other factors contributing to variability, expand the model to incorporate more stressors, and develop a more comprehensive, multi-stressor approach to pesticide risk assessment.

The study's laboratory setting might not fully capture the complexities of real-world field conditions. The model simplifies numerous interactions, and the choice of specific functions and parameter values could influence results. Further research should incorporate more stressors and refine model parameters for increased accuracy. The use of nicotine as a representative toxic compound warrants further investigation of other pesticides with different mechanisms of action.