RESULTS

Change in virus composition before and after exposure to the environment

Overall, the most prevalent viruses in the bumble bee colonies were DWV-A, DWV-B, BQCV, LSV, ARV-1, which are all honey bee associated viruses . In addition, several bumble bee associated viruses were abundant: BcDV, CBV and MV 1 . Total viral load across all viruses (log scale) increased on average by 1.34±2.21 (mean±SD) from the beginning to the end of the field exposure (day 0 = 3.67±1.47; day 45 = 5.00±1.61; lm: estimate±SE = 1.335±0.368, p<0.001), while total virus richness on average increased by a factor of 1.46±1.75 from day 0 to day 45 (day 0 = 3.14±1.03; day 45 = 4.60±1.31; lm: 1.457±0.282, p<0.001). Among the 9 screened viruses, only 1 to 5 were present at day 0, whereas at day 45 between 2 and 7 viruses were present among the colonies (Fig. 2).
The generally large increase in viral richness from day 0 to day 45 was driven by the appearance of new and mostly bumble bee specific viruses, such as CBV (lm: 2.852±0.467, p<0.001) or the MV 1 (lm: 2.291±0.404, p<0.001) as well as the DBV (lm: 3.006±0.309, p<0.001). At day 0 colonies had either viral loads dominated by the BcDV (significant increase; lm: 0.804±0.376, p = 0.036) or a combination of two honey bee viruses: DWV-B (significant decrease; lm: -0.990±0.363, p = 0.008) and BQCV (significant increase; lm: 0.639±0.198, p = 0.002; Fig. 2, SI Fig. S2).

Influences of initial colony viral loads on colony development

Among all measured colony development parameters, only the moth infestation index was related to the initial viral loads of viruses, with increasing loads of DWV-B colonies were related to higher moth infestation rates at the end of the field exposure (estimate±SE=0.210±0.081, p=0.012), while increasing BcDV loads led to fewer moth infestations (-0.218±0.079, p=0.008; SI Table S3).

Change in viral patterns in relation to landscape structure

Total Viral Load Change – The mean distance among forest patches (1250 m radius) had a marginally significant and positive relationship with the total viral load change in bumble bee colonies, while mean distance among agricultural patches showed a large positive and significant relationship (Table 2A, Fig. 3a), meaning that bumble bee colonies located in more isolated (or less connected) agricultural patches had a higher increase in viral loads during the field exposure (SI Fig. S3a).
Virus Richness Change – Habitat diversity (300 m), a measure of landscape heterogeneity, was significantly negatively related to virus richness change, showing that bumble bee colonies placed in areas with fewer different habitat types (lower heterogeneity) were infected with a higher number of different viruses after field exposure (Table 2B, Fig. 3b). On the other hand, forest area (100 m) was significantly positively related to virus richness change; with increasing forest cover within the close surroundings of a colony, the number of viruses infecting the colony increased (Table 2B; SI Fig. S3b, c).
Virus Turnover – The area covered by forests (100 m) had the strongest effect on viral turnover. Colonies in areas with more forest showed a higher change in their virus composition (Table 2C). However, with increasing distances (decreasing connectivity) among forest patches (400 m), virus turnover also increased significantly. Also, fragmentation of vegetated vineyards (400 m) significantly influenced the turnover, such that the viral composition changed less for colonies in areas with higher fragmentation (Table 2C; SI Fig. S4a-c).
Virus Appearance – Habitat diversity and forest area both were significantly related to virus appearance, with habitat diversity (300 m) again showing a negative effect, such that colonies in areas with a lower number of different habitat types were more frequently infected with viruses that were not yet present at the beginning of the field exposure. Increasing forest area within the wider surrounding of a colony (500 m) led to the appearance of new virus infections (Table 2D; SI Fig. S4d, e).
Overall, among the different spatial scales tested (ranging from 100 m to 1500 m), the predictors included in the final models had their strongest influence on the different response variables within smaller scales (100 m to maximum 500 m), indicating that the pattern of viral infections in bumble bee colonies influenced by the landscape is influenced by local rather than meso- or landscape scale heterogeneity.
Regarding the responses of colony development parameters to the landscape and field structure, we found very similar responses as for the viral patterns with generally positive relationships with agricultural or residential areas and habitat diversity but negative links to forest cover and connectivity. For detailed results see SI section 3.