Since the discovery of the first viral disease of honey bees (sacbrood, documented by Gershom Franklin White in 19131), viruses have become widely recognized as serious honey bee pathogens2,3,4,5,6,7,8. Recent advances in high-throughput sequencing have now led to the discovery of at least 74 different viruses in honey bees9. While many of these appear to be asymptomatic, several—including Lake Sinai virus (LSV), black queen cell virus (BQCV), sacbrood virus (SBV), deformed-wing virus (DWV) A, B, C, and D,10 acute bee paralysis virus (ABPV), Israeli acute paralysis virus (IAPV), Kashmir bee virus (KBV), chronic bee paralysis virus (CBPV), and slow bee paralysis virus (SBPV), among others—do cause disease or are linked to colony weakening. Symptoms may include failure to pupate, deformed wings at emergence, trembling or paralysis, and shortened lifespan11,12, and some viruses are associated with colony dwindling and winter losses6,7,8,11,13,14,15,16. Despite being linked to these adverse outcomes, we still have large gaps in our knowledge of the factors governing honey bee virus epidemiology17.

In the absence of an additional vector, viruses spread through the fecal–oral route, sexually (from drones to queens during mating), vertically (from queens to her progeny), and, to a lesser extent, through direct physical contact mediated by worn setae18 (reviewed in Yañez et al.18). Through these transmission routes, viruses were historically not problematic (although significant outbreaks did occasionally occur).4 However, the threat of viruses intensified after the host-jumping and subsequent global spread of the Varroa destructor mite (hereon Varroa), which can act as a mechanical and biological vector (reviewed elsewhere17,18,19,20,21). Greater virulence is observed when viruses are inoculated via injection compared to orally, which is attributed to the fact that injection circumvents the bee’s normal immune defenses encountered in the digestive tract18,22,23. As a result, viruses transmitted via Varroa feeding wounds tend to lead to more intense infections than those transmitted by ingestion, even when the mite is not acting as a biological vector24,25. While the relative importance of viruses to other factors influencing colony health is difficult to ascertain,3 it is likely to be cumulatively substantial especially when Varroa abundance is high19.

Varroa is not the only parasite to have been implicated in virus pathogenicity. Based on observations made in the early 1980s, Vairimorpha apis (formerly Nosema apis26, but see Bartolomé et al.27) is thought to exacerbate BQCV infections by causing damage to the bee’s gut epithelium, thereby facilitating viral infection28,29. This mechanism has not been substantiated further, and it is not clear why this association would exist for BQCV and not other viruses, but the explanation is consistent with the observation that V. apis and BQCV tend to covary30,31.

Several surveys broadly tracking virus prevalence over time have documented annual and seasonal trends, with highly regional distributions. In a survey conducted in the U.S., Traynor et al.32 identified a worrying trend of annual doubling of CBPV prevalence, increasing from 0 to 25% between 2009 and 2014, as well as seasonal prevalence patterns of an LSV variant (LSV-2; highest in spring and early summer), and ABPV (highest in winter). In southwest Germany, D’Alvise et al.33 found that BQCV and LSV peaked in the spring and summer, but contrary to Traynor et al. 32, ABPV peaked in July. BQCV trends in China34, France31, and Lithuania35 agree with those in Germany33, and prevalence of SBV in Lithuania and Canada tends to be higher in spring or summer compared to fall35,36. BQCV and DWV were the most prevalent viruses measured in the US32, China34, and France31, but SBV was similarly prevalent in Lithuania and France (where it tied with BQCV for second place)31,35. BQCV, LSV, and SBV were the most prevalent viruses in the 2017 Canadian National Honey Bee Health Survey37, whereas BQCV, DWV, and IAPV were most prevalent across provinces in an independent Canadian dataset (in which LSV was not measured)38. In contrast to all other countries discussed here, ABPV was the second most prevalent virus in Austria30.

These seasonal, annual, and regional patterns reported above and in additional surveys39,40,41,42,43 suggest that climatic variables or other factors that covary with climate (such as foraging activity, colony population size, or relationships with pathogens or parasites which themselves are affected by climate) could conceivably influence viral occurrence or intensity. While high temperatures negatively affect viral replication in honey bees kept in the laboratory44,45, few studies have analyzed the effects of climate17. Piot et al.46 found that in Europe, precipitation and temperature significantly predicted prevalence of viruses in the AKI-complex (ABPV, KBV, and IAPV) as well as DWV and SBPV in bumble bees and solitary bees, but not in honey bees. Prado et al.47 found that warmer winters were associated with lower virus abundance, perhaps mediated by elevated immune effectors. To our knowledge, these are the only studies rigorously testing linkages between climatic data and honey bee viruses.

Here, we used a multi-year, longitudinal dataset of abundances of nine viruses in honey bee colonies located in five Canadian provinces to describe regional and temporal patterns in intensity and occurrence, as well as test the hypothesis that climatic variables (temperature, precipitation, and wind speed) are meaningful predictors. Because viruses can be transmitted by the fecal–oral route, we hypothesized that high wind speed and precipitation would limit defecation flights and therefore be associated with higher viral intensity within colonies, but lower overall occurrence due to reduced transmission opportunities between colonies. We also hypothesized that temperature would positively influence viral intensity and occurrence indirectly via positive effects on colony population size and foraging activity (since larger, crowded populations have more opportunities for transmission within and between colonies via drifting and robbing). We also used data for two other parasites (Varroa and Vairimorpha spp.) to test if their abundance was also positively linked to viral intensity or occurrence as has been previously documented19,29. Finally, we discuss potential mechanisms through which relationships with climatic variables could arise.

Primary dataset

Overview of virus intensities, number of viruses per sample, and climatic variables

By conducting a new analysis of a dataset that was in part previously described48,49, we found clear regional trends in viral occurrence and intensities, with the most notable trends being (1) few viruses detected in Quebec (QC), (2) many viruses detected in British Columbia (BC), and (3) low virus intensities in samples from Ontario (ON) (Fig. 1a). We detected an overall average of 4.2 ± 1.2 (standard deviation) viruses per sample (of nine that were tested), with the largest mean number in BC (5.0), followed by southern Alberta (AB) (4.5), northern AB (4.3), QC near Lac St. Jean (QC (LSJ); 3.7), Manitoba (MB; 3.5), QC near Quebec City (QC (QC); 3.4), ON (3.3) and QC near Montreal (QC (Mon); 3.2) (Fig. 1b, c). We found significantly fewer viruses per sample in 2021 compared to 2020, but the magnitude of the difference is lower than the regional variation (coefficient =  − 0.4). As reported previously48, in this dataset, overall mean temperature was 15.1 °C (2.9–22.7 °C), mean precipitation was 2.0 mm (0.1–19.6 mm), and mean wind speed was 9.5 km/h (3.9–17.5 km/h). Temperature predicted the number of viruses per sample better than sampling date (Fig. 1d), but region and the region-by-date interaction were the most influential terms in the model. Varroa abundance did not consistently correlate with viral abundance (Fig. 1e) and including mite abundance as a covariate did not improve model fit (p = 0.36, χ2 = 0.83, ΔAIC = 1.17). See Table 1 for complete statistical reporting.