Patterns of viral prevalence and load show independent circulation of some honey bee viruses in wild bees Authors: Galen Cobb, Marley E Iredale, Marta L Wayne, Bryony C Bonning Published: 2026 ABSTRACT AbstractHoney bee (Apis mellifera) – associated viruses are increasingly detected in wild bees, but the extent and direction of transmission between wild bees and honey bees remain unclear. We surveyed wild bees from 23 genera alongside sympatric A. mellifera across seasons in north-central Florida, USA, quantifying prevalence, viral load, and co-occurrence of four common viruses: Black queen cell virus (BQCV), Deformed wing virus (DWV), Israeli acute paralysis virus (IAPV), and Sacbrood virus (SBV). At least one virus was detected in 98% of honey bees and 58% of wild bees. DWV and BQCV dominated overall, but IAPV occurred more frequently, and at higher levels in wild bees, particularly Halictidae than in honey bees. Seasonal dynamics differed markedly between groups: IAPV and SBV persisted year-round in wild bees but were undetectable in honey bees in spring. Co-infections were common, particularly in honey bees, and showed phylogenetic structuring among wild bee families. Therefore, while BQCV and DWV in wild bees likely reflects spillover from managed bees, IAPV and possibly SBV may circulate independently within wild bee communities. These results challenge assumptions of unidirectional pathogen flow from honey bees to wild bees, highlighting the value of seasonal sampling and virus quantification for understanding transmission.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-57579-y.Keywords: Apis mellifera, Virus quantification, Spillover, Wild bees, Multiplex RT-qPCRSubject terms: Ecology, Ecology, Microbiology, Zoology Introduction Due to its economic and cultural importance, the estern honey bee, Apis mellifera, has been the focal species for studying viruses in bees, i.e., Apoidea. Of the more than 70 viruses known to infect bees, most were first identified and characterized in A. mellifera 1. These viruses can cause severe clinical signs, including brood mortality, physical deformities, paralysis/behavioral abnormalities, and queen infertility 1–4. Further, most colonies are infected with multiple viruses, which are implicated in pathogen-induced colony losses5. Testing for these viruses outside of honey bees (Apis spp.) has been focused on commercialized pollinators, particularly bumble bees (Bombus spp.) and other species common to agroecosystems6–10. Few studies have demonstrated concrete fitness impacts that could lead to the type of declines reported for wild bees11,12. Surveys in wild bees have included species from across five of the seven taxonomic families, with one untested taxonomic family, Strenotitidae, being endemic to Australia11,13,14. Viruses are commonly detected in wild species, but prevalence varies taxonomically by host and by virus. Questions of how and to what degree these viruses enter populations of wild bee species remain. For this study, we focused on four bee viruses that cause severe disease in honey bees: Black queen cell virus (BQCV; Triatovirus nigereginacellulae), Deformed wing virus (DWV; Iflavirus aladeformis, all variants), Israeli acute paralysis virus (IAPV; Aparavirus israelense), and Sacbrood virus (SBV; Iflavirus sacbroodi). These viruses have single-stranded, positive sense RNA genomes and are classified in the families Iflaviridae (DWV and SBV)15, and Dicistroviridae (BQCV and IAPV)16. These viruses have been found on bee-collected pollen4,17,18 raising the possibility of environmental transmission. While there has been a proliferation of work examining the prevalence of these viruses, relatively few studies have quantified viral load in non-Apis bees, e.g.6,10,11. Viral load (genome equivalents or copies), while an imperfect proxy, can provide indirect evidence of infection and virulence. Correlations between viral load and clinical signs in honey bees have been found for DWV, SBV, and Acute bee paralysis virus (ABPV; closely related to IAPV), but not all viruses show this correlation. Pathogenicity and viral load showed no correlation for BQCV19. Additionally, latent infections, or infections with no clinical signs, are common for these viruses9. Irrespective of infection status, viral load can provide insight into transmission. Higher levels likely correspond with an increased chance of transmission through the release of more virions into the environment via feces17. In honey bees, transmission of these viruses can be direct (horizontal and vertical), indirect (environmental), and vectored (via the parasitic mite, Varroa destructor)20. Virus spillover from honey bees has been put forward as the source of virus in wild bee populations. The spillover hypothesis outlines a strictly environmental route, in which these pathogens are being transmitted from honey bee populations to wild pollinators through fecal or salivary contamination of shared floral resources21. High prevalence in honey bees compared with lower prevalence in wild bees has been cited as creating the potential for spillover to wild bees22–24. Indirect evidence for interspecific transmission is growing. Pollen samples taken from foraging honey bees tested positive for DWV and BQCV, even when collected from uninfected individuals17. Further, pollen contaminated with DWV can remain infectious, causing clinical signs in infection-naïve honey bees upon injection25. Flowers contaminated with the feces of infected honey bees have tested positive for virus and bumble bees that visited contaminated flowers tested positive for acquired infections26,27. Such spillover to wild bees via environmental transmission has also been demonstrated for fungal and microsporidian bee pathogens28,29. Far less attention has been given to the possibility of transmission among wild species outside of honey bee interactions, but transmission among wild species would allow viruses to persist during periods of waning infections in honey bees. The wide variation in viral prevalence and levels between seasons is well established19,30. The goal for this study was to increase our understanding of the prevalence and levels of four common honey bee - associated viruses in wild bees. To this end, we (1) expanded the taxonomic breadth of testing within wild bee families, (2) quantified viral loads to contextualize detection data (taking advantage of multiplex RT-qPCR31), and (3) compared patterns of viral co-occurrence. We sought to address the ways in which viral prevalence and load differ between wild bees and sympatric honey bees, including variation by season. Results Prevalence and viral loads A total of 323 bees, representing 23 genera (including 56 honey bees and 267 wild bees), were collected and analyzed (Table 1). At least one virus was detected in 65% of tested bees. DWV and BQCV had the highest prevalence at 49.8% and 34.1%, respectively. IAPV and SBV had lower prevalence at 18.6% and 17.0%, respectively. Viral detection was higher in honey bees (98% positive) than wild bees (58% positive). All four viruses were detected in at least 15 of the 23 genera, confirming the broad distribution of these viruses among bee species. Prevalence was higher in honey bees than wild bees for BQCV, DWV, and SBV, but IAPV showed higher prevalence in wild bees than honey bees (Fig. 1A). Interestingly, this was the first detection of any of these viruses in Mellitidae. However, only one species from this family is found in north-central Florida, and only five individuals were caught from this species, so Mellitidae was excluded from comparisons at the family level. Details of viral detection and viral copies per bee are provided in Table S-1. Table 1Percent prevalence for each virus in honey bees, wild bees (by family and overall), and overall across the year. N BQCVDWVIAPVSBVHoney bees560.820.840.130.20All wild bees2670.240.430.200.17Andrenidae180.110.330.110.0Apidae600.480.570.230.18Halictidae1310.190.360.200.18Megachilidae530.090.450.150.15Melittidae50.600.200.600.20All species3230.340.500.190.17N, number of individuals tested. Percent prevalence for each virus in honey bees, wild bees (by family and overall), and overall across the year. N, number of individuals tested. Viral prevalence (a) and the distribution of log viral load (b) for each virus grouped by honey bees and wild bees. Viral load was measured in genome equivalents (GEs) per µg of total RNA. The error bars in (a) represent 1.96 x standard error in site-specific prevalence. Median viral loads (genome copies/µg RNA) showed a similar pattern to prevalence (Fig. 1B). Overall viral loads were higher in honey bees. BQCV and DWV showed higher levels in honey bees, while SBV showed no difference in virus levels between the two groups of bees. IAPV was present at higher levels in wild bees. Higher IAPV loads were clustered among wild bee genera in the family Halictidae (Fig. 2). Log viral load of IAPV grouped by genera. Colors represent the taxonomic families of the genera (in order of appearance from left to right: light blue, Andrenidae; pink, Apidae; yellow, Halictidae; green, Megachilidae; dark blue, Melittidae). Viral load was measured in genome equivalents (GEs) per µg of total RNA. Seasonal pattern of detection There was greater viral prevalence in the fall than spring with positivity rates of 79.9% and 52.1%, respectively. Median viral loads (genome copies/µg RNA) were consistent across seasons in wild bees, with the exception of IAPV, which showed higher median viral loads in the spring (Fig. 3). SBV and IAPV showed little to no detection in honey bees in the spring, which differed strongly from the seasonal prevalence of both viruses in wild bees (Fig. 3). Log viral load for honey bees and wild bees grouped by season. Fall values are presented in orange and spring values are presented in green. Viral load was measured in genome equivalents (GEs) per µg of total RNA. In honey bees, both BQCV and DWV viral loads (mean ± the standard deviation and maximum viral loads) increased across spring sampling dates and spiked late May to early June (A. mellifera Spring BQCV: VLmean = 5112 ± 20,023, VLmax = 90,091 on 6/1/22; A. mellifera Spring DWV: VLmean = 8,539 ± 33,443, VLmax = 133,941 on 6/1/22). Throughout the fall, DWV loads oscillated with high peaks in viral load (A. mellifera Fall DWV: VLmean = 19,899 ± 56,869, VLmax = 262,569 on 10/25/22). Meanwhile, BQCV started at lower levels that slowly increased through the season but never returned to the levels observed in late spring (A. mellifera Fall BQCV: VLmean = 420 ± 502, VLmax = 2,239 on 10/25/22). IAPV and SBV were not present in honey bees during much of the spring season but appeared before and during fall sampling with levels similar to those seen in wild bees (A. mellifera IAPV: VLmean = 24.7 ± 16; A. mellifera SBV: VLmean = 69.6 ± 47.3). All four viruses were found in wild bees across sampling days in both seasons. In contrast to the honey bees, viral loads in wild bees stayed relatively stable across seasons (BQCV: VLmean = 141 ± 262; DWV: VLmean = 97.1 ± 116; IAPV: VLmean = 86.2 ± 200; SBV: VLmean = 67.8 ± 103). Viral loads in wild bees showed no obvious association with observed abundance of foraging honey bees (Fig. 4). Timeline of sampling overlaid with the average daily score of actively foraging honey bees and the log median viral load of each virus found in honey bees and wild bees. Grey bars mark the sampling days. Green bars mark the average daily score of observed honey bee foraging. Markers (yellow circles for honey bees and blue triangles for wild bees) indicate the log median GEs for each virus. Co-occurrence across taxa Co-occurrences of viruses were common. Slightly over half of all positive bees (56.2%) were positive for more than one virus (Table 2). Co-occurrence was higher in honey bees than wild bees (78.2% to 48.4%, respectively). All possible combinations of the four viruses were observed. However, the relative frequencies of viral combinations were dissimilar between wild bee families and honey bees (Fig. 5). Honey bees showed the highest frequency of co-occurrence (78.2%), which was dominated by BQCV-DWV combinations. A similar pattern was observed in wild Apidae, the most closely related bees to honey bees, although only 71.4% were positive for co-occurrence. As analyzed by permutation (see details in Methods section, Viral Co-occurrence and Testing Spillover Assumptions), all wild bee families showed an enrichment for individuals with no detected viruses, which was not observed in honey bees (Fig. 6). All wild bee families showed significant divergence from the viral co-occurrence patterns observed in honey bees, with the greatest divergence occurring in more taxonomically distant families (i.e., Megachilidae, Halictidae, and Andrenidae). Table 2Percent of detections that were positive for either a single virus or multiple, co-occurring viruses in honey bees, wild bees (by family and overall), and overall across the year. N Single VirusCo-occurring VirusesHoney bees550.220.78All wild1550.520.48Andrenidae80.620.38Apidae420.290.71Halictidae720.580.42Megachilidae300.700.30Melittidae30.001.00All species2100.440.56N, number of individuals that tested positive for at least one virus. Percent of detections that were positive for either a single virus or multiple, co-occurring viruses in honey bees, wild bees (by family and overall), and overall across the year. N, number of individuals that tested positive for at least one virus. Co-occurrence patterns of viruses across bee families: frequencies of viral incidence classes among families. Cool colors indicate the number of viruses found present in a sample and warm colors indicate the specific viruses found. Co-occurrence patterns of viruses across bee families: relative enrichment in frequency of viral incidence classes in wild bees as compared to frequencies within honey bees. Viral incidence classes, listed on the y-axis, represent all possible combinations of presence/absence for the four viruses. Each combination is denoted by the first initials of present viruses (B, BQCV; D, DWV; I, IAPV; S, SBV; BD, BQCV + DWV; BI, BQCV + IAPV, etc.). The color of each cell represents the degree of over- or under-enrichment with blue indicating over-enrichment and yellow indicating under-enrichment. Number reported is the prevalence of each viral incidence class as observed in each family. Enrichment = observed frequency – average permuted frequency. Discussion The motivation for this study was to gain insight into the prevalence and extent of infection of wild bees by four viruses that commonly infect the honey bee, A. mellifera. This study confirms two common findings describing wild bee encounters with these four viruses: (1) all four viruses are present in wild bee species and, (2) co-occurrence of viruses in individuals is ubiquitous. Determining the impact of viruses on the overall health of wild bees is a complex endeavor. This requires investigation into the multiple processes of virus and host interaction, including encounter, infection (i.e., with viral replication), pathogenicity/virulence, and transmission, and further how these processes scale from within and between individuals to across populations and communities. While these processes can be sequential, building upon one another, there can also be barriers between processes, requiring analysis of the extent to which each process occurs and affects host populations. The distinct nature of seasonal patterns of prevalence and viral loads between wild bees and honey bees provides potential insight into transmission of these viruses in wild bees. Some of the results contrasted with the spillover hypothesis and suggest at least one of these viruses may be circulating among wild bees independent of spillover events from honey bees. All four viruses were detected in wild bees, but apart from IAPV, prevalence was lower than in sympatric honey bees. This pattern is congruent with many studies that found generally higher viral prevalence in honey bees than wild bees for DWV, BQCV, and SBV13,14,24,32. Consistent with our findings, bee community surveys regularly observe that wild bee taxa have higher IAPV prevalence than honey bees11,13. However, the taxon dominant for IAPV varies between studies, e.g., Halictus in this study, Andrenidae11, and Eucera from Apidae13. Among wild bees (excluding Melittidae), prevalence was highest in Apidae, the family to which honey bees belong. Comparison of prevalence between studies is difficult due to inconsistent practices between studies, such as pooling of individuals within taxa prior to testing10,13,18 or a focus on select taxa6,22,26,33–35. Additionally, the majority of wild bees are solitary and can therefore be difficult to robustly sample. Both Dolezal et al. 11 and this study, however, tested individuals and analyzed wild bees at the level of family, making comparisons more reliable. Viral loads found in wild bees were comparable to those found in other populations/species11. Wild bees generally had lower viral loads than sympatric honey bees for DWV, BQCV, and SBV, as observed previously10,11. As with prevalence, IAPV was the single exception. Wild bees, particularly the genera in Halictidae, showed higher median viral levels of IAPV than sympatric honey bees. The combination of higher prevalence and higher viral load suggests that IAPV may be infecting and circulating among wild bees, independent of spillover from honey bees. However, the full implications of higher or lower viral loads on infections among wild bees are unclear. A large challenge to contextualizing viral load data is a lack of information on the relationship between viral load and infection/virulence11. For infection, an important consideration is the distinction between detection of virus and demonstration of replication. Detection may reflect true infection, that is, active viral replication within the host; or it may reflect contamination either externally or within the gut (consumed virions)34. External contamination was controlled in this study through a rigorous washing protocol. However, while individuals may encounter viruses regularly in their environment, compatibility filters mediated through host-virus interactions can prevent exposed individuals from developing infection. Few techniques testing for viral replication have been developed for use with non-Apis species; those that have include the gold standard - detection of negative-strand RNA (for positive-sense RNA viruses only)36 and viral-derived small interfering RNAs (siRNA)37. Given that siRNA can be generated from hairpin loop RNA structures and viral sequences integrated into the host genome38, the use of viral-derived siRNAs is not indicative of virus replication unless there is coverage of the entire genome. In the absence of these replication proxies, however, viral load may still serve as indirect evidence of infection when detected at levels exceeding typical environmental exposure. Mazzei et al. 25, tested samples of contaminated pollen collected from flowers visited by honey bees infected with DWV and found on average fairly low viral contamination (< 1000 copy number/20 mg of pollen). For reference, honey bees, which are generally classified as a medium-sized bee, consume only a few milligrams of pollen in a given feeding event. Even accounting for variation in pollen contamination across flower species25 and differences in feeding behavior between wild bee species and honey bees, exposure levels for wild bees are likely to be fairly low. Nevertheless, for all four viruses, viral loads observed in some individuals of wild bees from this study were 100 times greater (e.g., > 100,000 copy number/bee) than the likely levels of ingestion. Moreover, all viral loads were at similar levels to those observed in experimentally infected wild bees (species: Megachile rotundata and Colletes inaequalis11). Testing for stronger proxies of virus replication36 will be the next step in establishing infection competency across these diverse bee species. Although higher viral loads are more likely to reflect true infection, the consequences of such infections – as informed by artificial infection studies – remain largely unknown. Wild bees tested in this study showed a highly inflated proportion of individuals with no detected viruses compared with sympatric honey bees. Higher rates of non-detection in wild bees could be the result of potential barriers to encounter or compatibility. Differences in life histories and parasitic associates may drive variation in the encounter rates for honey bees vs. wild bees. For example, while honey bees are a social species, the majority of wild bees are solitary. The confined space of a hive shared with many individuals, social behaviors (such as trophallaxis, the mouth-to-mouth sharing of food), and their association with the parasitic vector, V. destructor, make honey bees vulnerable to high rates of viral transmission20. The presence or absence of V. destructor and its population level in a colony are especially important to viral dynamics in honey bees39. For wild bees, whose primary route of exposure is predicted to be contact with contaminated floral parts or pollen, encountering and consuming a quantity of virus sufficient for detection seems less probable. These differences could lead to lower overall rates of detection in wild bees. However, fecal-oral transmission may not be the sole route of transmission in wild bees. There are several parasites of wild bees, including many mites other than V. destructor40, that could act as vectors. Post-encounter differences in barriers to infection or host-mediated responses may affect detection rates, for example, if virions decay more quickly in some species than others. Compatibility filters may lead to host specificity and thus different clearance rates between wild bees and honey bees. Species with generally higher clearance rates may show greater rates of non-detection despite similar encounter rates. Alternatively, a higher proportion of individuals with no detected viruses may be the result of missing a portion of infected individuals during collection. All bees tested in this survey were wild caught, implying that they cannot be so adversely affected by the virus(es) as to restrict flying or floral visitation. Differences in tolerance, as well as in virulence, between infected honey bees and wild bees may lead to differences in the prevalence of virus observed in active individuals. Tolerance is the relationship between viral load and impact on host fitness41,42. Wild bees may be more susceptible to severe disease from the viruses at loads that do not adversely affect honey bees. If this were the case, infected wild bees would be more likely to develop severe clinical signs that limit their foraging behavior than honey bees (even at comparable viral loads) and thus have a reduced presence among bees collected for this study. Both the barriers to infection and difference in tolerance hypotheses are additionally supported by the lower mean viral loads and absence of extremely high levels observed in the wild bees that tested positive for BQCV, DWV, and SBV. However, viral load data alone are insufficient for fully characterizing within-host virulence. The virulence of these viruses in wild bee species is still vastly understudied. Direct infection of individuals from non-Apid species with a viral inoculum containing DWV, IAPV, and SBV resulted in few observable clinical signs from the treatment, and mortality did not appear to be affected despite clear increases in load for all three viruses following exposure11. Virus levels in infected bees, however, stayed within the range of viral loads observed in wild-caught wild bees. This relatively small-scale study (two species only; n = 18 for Megachile rotundata and n = 30 for Colletes inaequalis) may not fully capture the variation of pathogenicity in individual hosts, and the use of an inoculum mixture would not control for interactions between viruses (see discussion of viral co-occurrence below). In honey bees, pathogenicity/virulence varies greatly between colonies as well as between individuals within a colony. Šimenc et al.19, despite using a protocol that pooled individuals by colony, demonstrated a large range in per-bee levels in samples from both affected and healthy honey bee colonies. They also found significant differences in mean viral load for affected vs. healthy honey bees for four viruses (ABPV, DWV, SBV, and Chronic bee paralysis virus), but overlapping 95% confidence intervals between the two groups for all but SBV. BQCV showed no significant difference in mean load between healthy and affected colonies. This suggests that virulence for these viruses is strongly mediated by host responses in individual honey bees and that detection may often not signify disease. In this study, we saw large differences between mean and median viral loads for honey bees for BQCV and DWV due to the presence of extreme outliers in individual viral loads (500 thousand-fold and 100 million-fold above median for highest copy numbers, respectively) even in the absence of observable clinical signs, further confirming the loose relationship between viral load and disease. Curiously, viral load means and ranges in our wild bees more closely matched those observed in healthy honey bee colonies, as described previously19. However, more information on host-specific virulence is needed for any robust conclusions. While more research is certainly needed, preliminary evidence suggests a lower impact of these viruses on wild bees, particularly those less related to honey bees. Artificial infection studies are required to address whether wild bees are less susceptible to virus-induced disease. Further, factors known to contribute to higher virulence, including viral co-occurrence and route of transmission to wild bees, warrant further investigation. Co-occurrence of viruses is often detected in bees6,43. Demonstration of co-occurrence is important because it provides the opportunity for viral interactions that can impact infection. Viruses can interact synergistically or antagonistically, promoting or inhibiting infection by other viruses. In the case of synergistic interactions, this can lead to coinfections (simultaneous infection by two viruses) or superinfections (a secondary infection after an initially established infection). Evidence from several studies suggests DWV and SBV may interact synergistically in honey bees, with some strains of DWV facilitating SBV replication44–47. A similar helper effect may be occurring between DWV and BQCV in honey bees46 Information on the potential interactions between viruses in wild bee species is lacking. If strains are shared between honey bees and wild bees (which is likely for DWV)9,18,34, similar interactions may be occurring in wild bees. Our results showed different patterns of co-occurrence between wild bees and sympatric honey bees. BQCV and DWV were commonly found in tandem in honey bees, consistent with the extremely high prevalence of these two viruses20,24,37. However, the same combination was found in significantly lower frequencies in wild bee families. Interestingly, frequencies among the wild bee families had a phylogenetic pattern. For example, wild Apid bees, which belong to the same family as Apis mellifera, showed higher frequency for this virus combination compared with wild bees from more distantly related bee families. This observation is in line with the phylogenetic distance effect observed in many host-shift systems48,49. The phylogenetic distance effect describes the phenomenon of decreasing transmissibility or competence with increasing phylogenetic distance between donor and recipient hosts. This phenomenon provides useful hypotheses for testing our understanding of spillover dynamics between honey bees and wild bees. Spillover events in which viruses are transmitted from a reservoir in honey bees to wild bees have become the primary hypothesis explaining the detection of these viruses in wild bees. Spillover has large consequences for management and conservation interventions as it positions honey bees as the source of these diseases for wild bees, implicating honey bee management as a viable control option for protecting wild bees from viruses. The findings from this study show mixed support for this hypothesis. Lower prevalence in wild bees compared with honey bees for three of the viruses in this study (BQCV, DWV, and SBV) could be consistent with spillover events being the dominant mode of viral exposure, as could lower levels, given that viral load and transmission are expected to be positively correlated. However, other evidence for SBV points to circulation of virus during times with no detection in A. mellifera. Spillover, which by definition requires interaction between species, would be affected by the seasonal patterns of viruses in those species. In honey bees, viruses often show distinct patterns of seasonality, either in prevalence or in viral load19,50. In these honey bee populations, SBV was detected almost exclusively in the fall. This trend may correlate with the population dynamics of V. destructor, which peaks regionally in the fall in correspondence with the brood dynamics of honey bee colonies51, and the role of V. destructor as a vector of honey bee - associated viruses20. This contrasts with detection in wild bees, in which SBV was found consistently across seasons, including during times when honey bees were not collected (mid-spring). In contrast to these other viruses, IAPV was consistently present as indicated by viral prevalence, levels, and seasonal patterns for viral circulation among wild bees. Higher peak virus loads and higher prevalence in wild bees suggest IAPV may have greater transmissibility within wild bees than DWV, BQCV, or SBV. IAPV was the only virus to show seasonal differences in viral loads among wild bees, with higher levels in the spring, which contrasted with the lack of detection of IAPV in honey bees caught during the spring. Misaligned seasonality between wild bees and honey bees for SBV, and particularly IAPV, contradicts the spillover of these viruses from honey bees to wild bees. Possible circulation of IAPV and SBV among wild bees, in addition to spillover events from honey bees, should be considered. In conclusion, this study provides novel insights into the dynamics of honey bee – associated viruses in wild bee species in Florida using a methodology that can be applied in other pollinator communities to help advance our understanding of the pathogens that can potentially negatively impact these crucial species. Materials and methods Specimen collection In the spring and fall of 2022, wild bees and honey bees were hand-netted from flowers at five sites around Alachua County in north-central Florida. The sites were spread across three natural areas located in Alachua County: Kanapaha Prairie (KP; 2 sites), Paynes Prairie Preserve State Park (PP; 2 sites), and San Felasco Hammock Preserve State Park (SF; 1 site). These sites are representative of habitats typical of natural areas in the region and relatively remote from sources of agriculturally managed honey bees. Activity levels of foraging honey bees were scored at each site visit (0 = no honey bees observed, 1 = 1–10, 2 = 11–50, 3 = 51–100, 4 = 100+). For each specimen collected, the taxonomic identities of both the bee and its corresponding plant were determined as precisely as possible. Due to challenges in species-level identification for some wild bees that would compromise the specimens for RNA extraction (e.g., identification requiring dissection or extensive comparison with type specimens), only bees that could be easily identified to the level of genus or species were included. Upon collection, netted bees were stored in individual vials on dry ice until transferred to a -80 °F freezer, where they were stored until RNA extraction. RNA extraction Total RNA was extracted from each bee specimen using the Zymo Direct-zol RNA Purification kit (Zymo Research, Irvine, CA, USA). Each specimen was cleaned with a dry sterile cotton swab to disrupt clumped pollen and then subjected to a minimum of three washes in chilled 70% ethanol prior to extraction. The bee was assigned to a size class (i.e., small, medium, large, and extra-large) and then processed with a corresponding amount of TRIzol Reagent (Thermo Fisher Scientific, Waltham, MA, USA), (S: 400 µL, M: 800 µL, L: 1000 µL, XL: 1000 µL x 2). The entire body of each specimen was processed with the exception of extra-large bees from the genera Bombus, Svastra, and Xylocopa. Due to the limiting size of the tube used in the Precellys Evolution tissue homogenizer (Bertin Instruments, Montigny-le-Bretonneux, France), these specimens were sliced in half vertically prior to homogenization in TRIzol and the corresponding supernatants were combined prior to the RNA wash and purification. Elution volumes also varied by size class (S: 50 µL, M-XL: 100 µL). After extraction, total RNA concentration and purity were quantified using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific Inc., Lafayette, CO). Multiplex RT-qPCR viral identification Multiplex RT-qPCR was performed to simultaneously identify and quantify all four viruses (i.e., BQCV, DWV, IAPV, SBV) using a BioRad CFX96 Touch Thermal Cycler (BioRad, Laboratories, Hercules, CA, USA) with 5 channel capabilities. The protocol and method validation was as described previously31. This methodology was developed to allow for efficient detection of multiple RNA viruses using small amounts of sample. Multiplex standard curves were generated during method validation using in vitro transcribed (IVT) RNA of the targeted sequences for all viruses and applied to subsequent runs. Targeted sequences for the four viruses of interest were taken from a universal standard reference plasmid that contains sequences for seven common honey bee viruses43. Each sequence was ligated into individual plasmids with a T7 promoter for in vitro transcription. Transcription was performed using the Biosearch Technologies Ampliscribe T7-Flash transcription kit (LGC Biosearch Technologies, Petaluma, CA, USA). Primers and probes were designed to match target sequences (Table S-2). PrimeTime probes from Integrated DNA Technologies (Coralville, IA, USA) were used with double quenchers for BQCV, DWV, and SBV. Primer and probe sequences and their properties are as previously reported31. RT-qPCR was conducted in 10 µL reactions with 4 µL of RNA (400 ng to 1 µg) per reaction. Each sample contained RNA from an individual bee; no samples were pooled. Samples that contained less than 100 ng/µL of total RNA after extraction were excluded from analysis. Positive controls included diluted IVT RNA of all four viruses in equal concentration. Non-template controls were also included on every plate. All samples were run in triplicate. Viral detection and quantification of viral load Positive detections were determined by cut-offs in the end RFU and Cq values based on the standard curves for each virus31. To reduce the probability of false detections, end RFU values had to fall above the lowest RFU value observed in the standard curve to ensure that bees identified as positive were showing amplification consistent with known positive controls. Cq values were restricted on the lower bound to ≥ 10 and on the upper bound to either ≤ 35 or below the highest Cq value in the standard curve (if < 35). Virus levels (genome copies/µg RNA) were calculated using the standard curves for all positive bees through the inverse.predict function of the R package “chemCal” 52. Virus levels were then standardized based on the total amount of RNA in each RT-qPCR reaction. Standard curves were validated using IVT RNA at concentrations ranging from 10 to 107 genome equivalents (GEs) for DWV and IAPV and 100–107 GEs for BQCV and SBV. Reported viral loads above 107 GEs are projected values. Viral prevalence was calculated for each virus in both honey bees and wild bees. Median viral loads were calculated for each of the four viruses in both honey bees and wild bees at multiple timescales: across the year, within collection season (spring, March-June; fall, September-October), and within day of collection. Viral co-occurrence and testing of spillover assumptions Co-occurrence was examined by calculating the frequencies of all possible combinations of presence/absence for the four viruses (16 combinations). These co-occurrence frequencies were examined within honey bees, as well as five families of wild bees. We tested the null hypothesis that wild bee viral co-occurrence frequencies reflect the co-occurrence frequencies in honey bees due to infection of wild bees by honey bees, i.e. spillover from honey bees, by using a permutation test. All else being equal, viral co-occurrence frequencies in the wild bees should reflect co-occurrence frequencies found in the honey bees at a given site. Assumptions behind this test include no differences between honey bees and wild bees, efficacy of transmission, host resistance and tolerance, within-host competition, etc. Nevertheless, this simplified model is useful because it provides an introductory test of a base null. Expected frequencies were permuted by randomly assigning positive/negative status to individuals of the wild bee host families (and honey bees separately) caught at each site at the frequencies observed for honey bees within each site. The permutation analysis did account for spatial structure in transmission patterns. Each permutation was repeated 1000 times. Enrichment scores were then calculated (observed frequency, mean permuted frequency) to quantify the difference between permuted frequencies and observed frequencies of the co-occurrence classes. P-values were calculated as the percent of permuted values that were more extreme than the observed value. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Acknowledgements The authors thank Matthew Henriques for assistance with RNA extractions. Author contributions Conceptualization: GC, MW; Methodology: GC, MI, MW, BB; Investigation: GC, MI; Visualization: GC; Writing: GC; Editing: all authors; Funding Acquisition: GC, BB; Supervision: MW, BB. Funding This study was supported in part through the scientific research funding program at experiment.com (to GC). Data availability All data associated with this study are included in the published article and Supplementary Information. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Data Availability Statement All data associated with this study are included in the published article and Supplementary Information.