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

Honey bee colony losses are multifactorial and have intensified in recent decades, threatening pollination services and food security. Stressors include parasites (e.g., Varroa destructor), pathogens (e.g., deformed wing virus, DWV), pesticides (notably neonicotinoids), nutrition, and environmental conditions. Although laboratory studies document lethal and sublethal pesticide effects, field studies have produced contradictory results, sometimes showing no detectable harm or even country-specific outcomes. The study aims to mechanistically explain these discrepancies by analyzing honey bee health as a complex system of interacting stressors. The authors hypothesize that system-level properties arising from interaction architecture, particularly feedbacks, can shape outcomes and account for divergent field observations, and they test model-derived predictions in vivo.

Previous research shows neonicotinoids can affect bee navigation, immunity, and reproduction under laboratory conditions, but field results are inconsistent, with some large-scale trials reporting no adverse effects and others showing negative, context-dependent impacts. Explanations have invoked colony buffering capacity and variability in foraging and exposure contexts. Regulatory responses have differed internationally (EU bans vs ongoing reviews elsewhere), reflecting scientific uncertainty. The literature also documents DWV’s global spread via Varroa, pesticide–pathogen interactions (e.g., clothianidin impairing immunity and promoting viral replication), and fungicide-insecticide synergies that can elevate toxicity. Systems-based and structural analysis approaches have been proposed to assess multi-stressor effects when parameters are uncertain.

The study combines structural systems analysis with laboratory validation and colony-level modeling.

• Conceptual and mathematical model: The authors construct a qualitative interaction network influencing honey bee health, encompassing toxic compounds (e.g., neonicotinoids), parasites (Varroa), pathogens (DWV), nutrition (sugar, pollen), and temperature stress. They formalize the system with ODEs for four state variables: x_HB (bee health), x_TC (toxic stress), x_VA (parasite stress), x_VI (pathogen stress), with external inputs for sugar, pollen, absolute temperature deviation, and sub-optimal temperature. Functions encode monotone positive/negative influences, self-regulation, and potential immune suppression (parameter ε) by pathogens.
• Structural analysis: Using community (Jacobian) matrices, sign patterns, and the concept of monotonicity, they derive a parameter-independent structural influence matrix that captures net steady-state effects of persistent inputs. They analyze equilibria and stability via bifurcation theory to assess conditions for monostability vs bistability, focusing on the effect of an immune-suppressing pathogen (DWV) that introduces a positive feedback loop between virus load and impaired immunity.
• In vivo validation: Over six summers, they conducted standardized caged-bee survival experiments. Two designs: (1) Diachronic comparison of bees collected early season (low DWV prevalence) vs late season (high DWV prevalence), confirmed by qRT-PCR (Ct<30 positive). (2) Synchronic manipulation: larvae were fed DWV (1000 genome copies per bee) or not. Additional stressor challenges included 50 ppm nicotine in sugar solution and sub-optimal temperature (32 °C vs 34.5 °C control). Survival was monitored daily; distributions of lifespans and interquartile ranges (IQRs) were compared.
• Colony-level modeling: The authors adapt a published compartment model with foragers (F) and hive bees (H) by splitting hive bees into younger (Y) and older (O) classes, introducing premature mortality η in younger bees to reflect individual-level early deaths. They analyze how η shifts the critical forager death rate threshold m for colony collapse.
• Data, code, and analyses: Viral detection by qRT-PCR with specified primers; experiments replicated 3–13 times; custom Mathematica and MATLAB codes publicly available (BeeStability).

• Structural influence: Any added stressor has a net negative effect on bee health at steady state, independent of parameter values.
• Bistability from immune suppression: Without immune suppression, the system is monostable with a single globally stable equilibrium whose health level depends on stressor intensity. With an immune-suppressing pathogen (e.g., DWV), a positive feedback emerges that can produce three equilibria (two stable), i.e., bistability. Under intermediate stressor levels, similar initial conditions can lead to markedly different outcomes (high vs low bee health), explaining variable responses to the same exposure.
• Experimental validation: Presence of DWV reduces median survival and broadens longevity distributions, consistent with bistability. • Early-season (low DWV) controls: median survival ≈ 23 days; IQR 6. • Late-season (high DWV): shorter median survival and broader distribution; IQR ≈ 10. • Controlled viral dosing: control median 18 days, IQR 5; DWV-treated median 10 days, IQR 12. • Under nicotine (50 ppm) or low temperature (32 °C), DWV presence increased variability: IQRs increased from 3–7 (early/low DWV) to 8–16 (late/high DWV), with more early deaths yet some long survivors.
• Colony implications: Introducing premature mortality of younger hive bees (η>0) lowers the critical forager death rate m beyond which the only stable equilibrium is colony collapse. Thus, early individual mortality (as observed under immune suppression) makes colonies more vulnerable, shifting the collapse threshold to lower forager mortality rates.

The findings mechanistically reconcile contradictory field results on pesticide impacts. In systems lacking strong immune suppression (low DWV/Varroa prevalence), colonies may buffer pesticide stress, yielding no apparent harm at field-realistic exposures. Where immune-suppressing pathogens are prevalent, bistability means the same pesticide exposure can precipitate early deaths for some bees, elevating colony collapse risk, while others may fare normally, generating inconsistent outcomes across contexts. The systems approach demonstrates that interaction architecture and feedbacks, rather than individual effect magnitudes, govern observed variability. The results support multi-stressor, systems-based risk assessments and suggest that prevalence of DWV/Varroa is a critical contextual factor when interpreting field studies and regulatory evidence. The authors also note that while pesticide upregulation of detoxification genes is typically homeostatic, fungicide–insecticide interactions that impair detoxification could, if confirmed mechanistically, generate dynamics analogous to immune suppression.

This work shows that an immune-suppressing pathogen (DWV) introduces a positive feedback leading to bistability in honey bee health, causing divergent outcomes under similar exposures and explaining inconsistent field results for pesticides. Laboratory survival data confirm broader and bimodal-like longevity patterns in the presence of DWV, and colony modeling indicates that early individual mortality lowers the threshold for colony collapse. The systems-based, parameter-independent analysis provides a robust framework for understanding multi-stressor impacts and informs pesticide risk assessment beyond single-stressor paradigms. Future research should: (1) quantify context-dependent transitions between equilibria and hysteresis; (2) integrate additional stressors (nutrition quality/quantity, multiple pesticide mixtures) into structural analyses; (3) empirically test whether anti-detoxification interactions can induce feedbacks analogous to immune suppression; and (4) incorporate pathogen prevalence metrics (DWV/Varroa) into field study design and regulatory evaluations.

• The mathematical model is qualitative and parameter-independent; while robust to parameter uncertainty, it abstracts many stressors into coarse variables (e.g., a single toxicant), limiting quantitative prediction.
• Laboratory validations used caged bees and survival as a proxy for health; generalization to field colonies requires caution.
• Seasonal DWV prevalence was used as a proxy for infection status; although supported by qRT-PCR sampling, individual-level initial conditions remained uncontrolled.
• The hypothesized pesticide anti-detoxification feedback is speculative and not demonstrated here.
• The colony model extension assumes specific structure and mortality patterns; other demographic processes and environmental feedbacks were not included.