Experimental infection of bumblebees with honeybee-associated viruses: no direct fitness costs but potential future threats to novel wild bee hosts

Pathogen spillover represents an important cause of biodiversity decline. For wild bee species such as bumblebees, many of which are in decline, correlational data point towards viral spillover from managed honeybees as a potential cause. Yet, impacts of these viruses on wild bees are rarely evaluated. Here, in a series of highly controlled laboratory infection assays with well-characterized viral inocula, we show that three viral types isolated from honeybees (deformed wing virus genotype A, deformed wing virus genotype B and black queen cell virus) readily replicate within hosts of the bumblebee Bombus terrestris. Impacts of these honeybee-derived viruses - either injected or fed - on the mortality of B. terrestris workers were, however, negligible and probably dependent on host condition. Our results highlight the potential threat of viral spillover from honeybees to novel wild bee species, though they also underscore the importance of additional studies on this and other wild bee species under field-realistic conditions to evaluate whether pathogen spillover has a negative impact on wild bee individuals and population fitness.

AT, 0000-0002-2667-7306; TS, 0000-0003-1315-7042; ST, 0000-0002-3675-9583; RJP, 0000-0003-2517-1351 Pathogen spillover represents an important cause of biodiversity decline. For wild bee species such as bumblebees, many of which are in decline, correlational data point towards viral spillover from managed honeybees as a potential cause. Yet, impacts of these viruses on wild bees are rarely evaluated. Here, in a series of highly controlled laboratory infection assays with well-characterized viral inocula, we show that three viral types isolated from honeybees (deformed wing virus genotype A, deformed wing virus genotype B and black queen cell virus) readily replicate within hosts of the bumblebee Bombus terrestris. Impacts of these honeybee-derived viruses -either injected or fed -on the mortality of B. terrestris workers were, however, negligible and probably dependent on host condition. Our results highlight the potential threat of viral spillover from honeybees to novel wild bee species, though they also underscore the importance of additional studies on this and other wild bee species under field-realistic conditions to evaluate whether pathogen spillover has a negative impact on wild bee individuals and population fitness. so as to determine the capacity of the virus to replicate in a novel host, as well as by feeding, representing the more likely natural route of infection in the field [41]. These experiments were carried out under ad libitum food conditions. However, fitness costs when responding to an immune challenge may be dependent on host nutritional state, and have been shown for bumblebees when diet was restricted [46]. We, therefore, complemented our investigation with an experiment under starvation conditions.

Source of bees
Commercial B. terrestris colonies (Koppert B.V., Berkel en Rodenrijs, The Netherlands) were kept in an incubator at 30°C and 50% relative humidity with ad libitum 50% (w/v) sucrose solution. Every 2-3 days, they were fed with fresh-frozen honeybee pollen pellets (Imkerei Schachtner, Schardenberg, Austria) that had been freshly defrosted. Pollen was UV-irradiated before use to destroy pathogens. Honeybees for experiments and for generating viral inocula were taken from our local apiary (University of Halle, Germany), originally purchased as the subspecies Apis mellifera carnica, as is typical for beekeeping in the region. To check that bumblebees (12 source colonies: labelled B1-B12) and honeybees (2 source colonies, labelled 5.1 and G) as well as the fresh-frozen pollen pellets were devoid of viral pathogens, we tested them by real-time quantitative PCR (qPCR) for seven common honeybee viral targets and three Microsporidia (electronic supplementary material). Bumblebee and honeybee colonies were largely free of virus (electronic supplementary material, table S1), pollen was devoid of virus and Microsporidia were not detected.

Propagation of viral inocula
To propagate DWV-A and DWV-B for experimental inocula, we used the inocula from Tehel et al. [47]. Our BQCV inoculum was prepared by propagating the BQCV inoculum of Doublet et al. [48]. Viral propagation in honeybee pupae and absolute quantification of virus followed precisely methods in Tehel et al. [47]. We always generated the correct virus inoculum from the original inoculum, which was devoid of other viruses (electronic supplementary material, figure S1).
Inocula containing only DWV-A, only DWV-B or only BQCV at known concentrations were aliquoted and stored at -80°C for use in experiments, as was the control inoculum devoid of virus. For each virus, a single inoculum derived from one preparation was used for all experiments with Bombus and Apis. Ultradeep next-generation sequencing (NGS) on an Illumina platform confirmed the identity of our DWV-A and DWV-B inocula (see [47] for consensus sequences and the pipeline used to assemble them from NGS data as well as BioProject ID PRJNA515220 for the original NGS source files).

Experimental inoculation 2.3.1. Honeybees: injected with inoculum, satiated
We initially ensured that viral inocula were viable by injecting them into honeybee workers.
Freshly eclosed workers were cooled to 4°C and then injected laterally between the second and third tergite with 10 7 viral genome equivalents (or, as control, virus-free inoculum), a quantity sufficient to ensure 100% infection of adults [42], using a Hamilton syringe (hypodermic needle outer diameter: 0.235 mm). To avoid cross-contamination, syringes were cleaned after each use, and different syringes were used for each inoculum (DWV-A, DWV-B, BQCV) and for the control inoculum devoid of virus. The 249 individually injected honeybees were randomly assigned to injection treatments, held in groups of 20-22 in autoclaved metal cages (10 × 10 × 6 cm) independent of their source colony but with bees of the same treatment per individual cage in an incubator (30°C), fed ad libitum with 50% (weight/volume) sucrose solution and monitored daily till death, as in McMahon et al. [42]. At 10 days post-inoculation (d.p.i.), one bee per cage was removed to quantify viral titre.

Bumblebees: general handling
Viral inocula were tested in freshly emerged B. terrestris workers as follows. Firstly, we marked all workers in our 12 B. terrestris colonies. Colonies were checked daily and unmarked, newly emerged workers were transferred to autoclaved metal cages (10 × 10 × 6 cm), fed ad libitum with 50% (w/v) sucrose solution royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 7: 200480 and held in an incubator at 30°C. On the next day (i.e. 24-48 h after eclosion), workers were inoculated with virus (or control solution), either by injection or orally by feeding, and then kept in groups of 5-10 of the same treatment per cage. In an experiment, the number of bees per cage was constant (± one bee) for every treatment within any 1 day of infection. This procedure was repeated across 25 days to allow for sufficient replication per experiment.

Bumblebees: fed inoculum, satiated
Inoculation of B. terrestris workers by feeding was designed to test the likely route of viral spillover from honeybees at flowers in the field. Freshly emerged (24-48 h after eclosion) bumblebee workers were individually fed with 10 9 viral genome equivalents or the equivalent control solution devoid of virus (electronic supplementary material), a quantity inducing an acute infection [48,49]. Then bees were transferred to a new, autoclaved metal cage in small groups (5-10 bees per cage, grouped according to treatment). In total, 512 bees from five source colonies were evenly distributed between all four treatments (DWV-A, DWV-B, BQCV, control; 128 bumblebees per treatment) and were randomly assigned to cages independent of source colony. They were monitored daily for mortality. One bee per cage was removed at 18-25 d.p.i. to quantify viral titre. In a preliminary trial following the identical protocol as described above, we quantified viral titres at 10 and 20 d.p.i., but found no significant difference among them or with bees tested at 18-25 d.p.i. (electronic supplementary material, figure S2).

Bumblebees: injected with inoculum, satiated
Inoculation by injection was designed to test whether B. terrestris is a competent host for each virus. To inject workers, they were cooled on ice till immobile. Viral inoculation then followed that for honeybees; B. terrestris workers were then transferred to autoclaved metal cages in small groups (five to seven bees per cage). Bees were randomly assigned to cages independent of their four source colonies but grouped according to treatment per cage, resulting in n = 404 bees that were recorded daily for mortality (ca 100 bees per treatment: DWV-A, DWV-B, BQCV, control). One bee per cage was removed at 10 d.p.i. to quantify viral titre.

Bumblebees: injected with inoculum, starved
As B. terrestris workers did not exhibit elevated mortality over controls following viral inoculation under benign laboratory conditions with ad libitum food (see Results), we ran an additional experiment in which we removed their food to determine whether viral inocula induced mortality under non-benign, starvation conditions. Bees from three colonies were collected over a 14 day period as they eclosed, held in autoclaved metal cages and individually injected as described above. To control statistically for effects of age, bees of approximately the same age were held in the same cage. All bees were injected on the same day. At 13 d.p.i., after the virus had time to replicate, bees were individually transferred to a plastic cup covered with netting, devoid of sucrose solution but with a small cotton wool ball soaked in water, held at 30°C and checked every hour for mortality (electronic supplementary material, figure S3).
At death, bee size was estimated because size might determine the ability to survive under starvation [50] (electronic supplementary material, figure S6). Viral titre was quantified in a subset of bees collected at 13 d.p.i. In total, 326 B. terrestris were inoculated by injection in this experiment, of which 194 survived till 13 d.p.i. and, therefore, entered the starvation part of the experiment.

Viral titres
To quantify viral titres in adult worker bees arising from inoculation experiments, we crushed one whole honeybee or one bumblebee abdomen in 500 µl of 0.5 M PPB (pH 8.0) using a plastic pestle, of which 100 µl were used for RNA isolation. Absolute quantification of viral titre followed methods used for viral inocula described in Tehel et al. [47] (electronic supplementary material), including all positive and negative controls.

Statistics
All analyses were performed in R v. 3.5.1 (R Core Team). We used generalized linear models (GLMs) with a quasi-Poisson error distribution to test for the effect of treatment or experiment on viral titre.
Survivorship of experimentally inoculated bees was analysed using the Cox proportional hazards models with the R package coxme [51,52]. 'Cage' was used as a random factor in all analyses and 'round of infection' as a random factor for B. terrestris experiments in which an experiment was initiated across royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 7: 200480 multiple days. To assess the significance of predictors, statistical models including all predictors were compared with null (intercept only) or reduced models (for those with multiple predictors) using likelihood ratio (LR) tests. Pairwise comparisons between factor levels of a significant predictor were performed using pairwise post hoc tests, adjusting the family-wise error rate according to the method of Bonferroni ( package multcomp, [53]). For the experiments with bumblebees under satiated conditions (inoculated by injection and by feeding), survival models retained 'cage' and date or 'round of infection' (for B. terrestris experiments in which an experiment was initiated across multiple days) as random factors and treatment as a fixed factor. For the Bombus experiment under starvation conditions, 'cage' was again retained as a random factor, and treatment together with bee age and bee size entered as fixed factors. The median survival was calculated using the Survfit function in survival. In all survival analyses, bees that died within 1 day (24 h) post-inoculation were eliminated from subsequent analyses as death was probably a consequence of physical damage by injection per se rather than the inoculum.

Bumblebees: injected with inoculum, satiated
In contrast with honeybees, bumblebees injected with viral inocula and fed ad libitum did not die any faster than controls (Cox proportional hazards BQCV: Exp.   Though smaller worker bumblebees lived longer than larger workers (Cox proportional hazards: Exp. (β) = 1.665, p = 0.03), bee size did not differ between treatments (electronic supplementary material, figure S6) and bumblebee size did not differentially impact mortality across treatments (electronic supplementary material, table S2).

Viral titres across experiments
All three viruses replicated to higher titres in A. mellifera than B. terrestris. Inoculation of honeybees by injection led to three orders of magnitude higher viral titre (ca 3 × 10 13 viral genome equivalents per bee at 10 d.p.i.) than the equivalent inoculation by injection of bumblebees (ca 4 × 10 10

Discussion
Here, we show that B. terrestris is a competent host for BQCV, DWV-A and DWV-B, suggesting that spillover from honeybees is a potential threat for this and probably other wild bee species. We did not, though, observe impacts of these viruses on bumblebee mortality under laboratory conditions. Furthermore, all royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 7: 200480 three viruses replicated to higher titres in honeybees than in bumblebees, which is surprising, given that honeybees are generally smaller than bumblebees. The higher viral titre per honeybee suggests that these viruses are locally adapted to A. mellifera, which is probably their reservoir host. BQCV, DWV-A and DWV-B have been frequently detected in bumblebees (Bombus spp.; [19,24,26,27,29,39,40] and other wild bee species collected from the field (reviewed in [25]), as well as in other insect species associated with honeybees or the flowers they visit (e.g. [54][55][56][57][58]). Moreover, the negative strand of these (+)ssRNA viruses has also been detected in Bombus spp. and other wild bee species [19,[26][27][28] as evidence that virus is actively replicating inside these non-Apis hosts. Here, we have been able to show unequivocally that all three viruses can replicate to high titres in B. terrestris. Additional studies on other non-Apis bees, including non-commercial B. terrestris, as well as with other honeybee viruses are needed to understand the extent of their host tropism across wild bee species. It will also be important to determine how virulence evolves after a viral jump to a new wild bee species as this is central to disease emergence in the new host [59].
Under benign conditions of the laboratory, we found that BQCV, DWV-A and DWV-B were not virulent (i.e. did not reduce host fitness, sensu [60]). Gusachenko et al. [61] have recently reported similar findings for DWV-A and DWV-B. These results are surprising because DWV has been associated with field-collected Bombus spp. exhibiting clinical symptoms (deformed wings; [23]), which is typical of honeybees when infected by DWV in the pupal stage [47]. Also, when fed [19] or injected [45] into B. terrestris workers, DWV has been previously shown to reduce B. terrestris lifespan. Differences between former studies and ours may reflect the genetic background of the host; Fürst et al. [19] and Graystock et al. [45] employed B. terrestris from a different commercial source to our study (though [61] used the same source as [20]). Alternatively, it may reflect the source of virus; Fürst et al. [19] used a mixed DWV-A/DWV-B inoculum and Graystock et al. [45] used DWV isolated from B. terrestris, whereas we used DWV-A and DWV-B isolated from A. mellifera. Recombination between DWV-A and DWV-B deserves greater attention as a source of virulent virus that may impact both honeybees and bumblebees [44], as does the extent of local adaptation of DWV to a host species.
Another facet of virulence may be the size of the host in relation to viral titre. Honeybee workers are generally smaller than those of bumblebees and, in our experiments, we inoculated each host species with the same viral titre. A direct relationship between host size and inoculum titre could, therefore, account for the higher mortality of honeybees versus bumblebees that we observed. However, viral titres were actually higher in honeybees than bumblebees, arguing against a relationship between host size and inoculum titre that is constant across host bee species. Furthermore, viral titre seems to asymptote after several days in each host species, high in honeybees [47, this study] and lower in bumblebees (electronic supplementary material, figure S2), suggesting that initial viral inoculum size is not related to ensuing viral titre in a host. The relationship between viral titre and host mortality nevertheless deserves greater attention, not only within but also across host species.
Not even under stressful, starvation conditions did we detect a marked effect of either BQCV, DWV-A or DWV-B in reducing B. terrestris longevity in the laboratory. Condition-dependent virulence of honeybee viruses in Bombus spp. hosts has been seen for slow bee paralysis virus infecting B. terrestris, in which longevity was compromised only when hosts were starved [62], and for other bumblebee pathogens such as Crithidia bombi [63,64]. We, therefore, urge caution in the interpretation of our result that viral virulence was non-existent in B. terrestris. Laboratory conditions may underestimate the impact of honeybee virus spilling over into wild bees in the field, where hosts may be exposed to far harsher environmental conditions and limited resources, e.g. [65]. Insecticides have been highlighted as playing a role in insect, including Bombus spp., decline [6,11], with sublethal impacts of novel classes of insecticide on colony fitness [66,67]. Sublethal doses of insecticide can interact with pathogens to elevate host honeybee mortality [68][69][70], and may represent another condition-dependent factor for bumblebees and other wild bee species that exacerbates the impact on them of viral spillover from honeybees. Fieldrealistic experimental paradigms are now needed to reveal the role of viral spillover for the individual, colony and population fitness of wild bee species as well as additional experiments examining other response variables than mere mortality, e.g. offspring production, pupal development and foraging efficiency. Changes in sublethal parameters like these could decrease the success of a social bee colony enormously. Furthermore, our non-benign scenario (starvation) may have been too stressful to allow expression of condition-dependent virulence; use of more natural levels of stress, as may be typically experienced by bees in the field, is warranted to reveal condition-dependent virulence.
We found that viral titres were lower and the impact on host mortality was non-existent when BQCV, DWV-A or DWV-B was injected into B. terrestris versus injected into A. mellifera. These results suggest that virus may be locally adapted to its host, and that A. mellifera may be the reservoir host for all royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 7: 200480 three viruses. The immediate impact of viral spillover from honeybees to bumblebees and other wild bee species might then indeed be low, as we found under our benign laboratory conditions. But transmission from bumblebee to bumblebee could lead to local adaption of a virus to a Bombus host, with unknown consequences of pathogen spill-back from bumblebees and other wild bee species to honeybees if viral adaptation to the novel host (Bombus) trades off with a loss of virulence in the original host (Apis) [71][72][73]. The speed with which local adaptation to a novel host occurs, its relationship to virulence and whether it results in a loss of viral fitness or virulence in the reservoir host will help determine the impact of viral pathogen spillover for the entire bee pollinator community [74].
It is unsurprising that we found inoculation by injection to lead to higher viral titres than by oral inoculation of bumblebees. Injection of a pathogen into the insect haemocoel gives the pathogen access to the entire host body tissue, whereas oral infection initially gives it access to the gut alone. The former route of transmission, injection into the haemocoel, through V. destructor host feeding is thought to account for the huge increase in viral prevalence and intensity of infection of DWV in honeybees [75,76]. In support of this view, injection of another honeybee virus, Israeli acute paralysis virus, into B. terrestris led to systemic infection and rapid host death, whereas oral infection led to infection of the host gut in a dose-dependent manner and with more limited impact on host health [49].
That BQCV was extremely virulent in our honeybee assay is at first sight surprising because BQCV is widespread and highly prevalent in honeybee populations [24,26,29,30,77]. Both Retschnig et al. [78] and Doublet et al. [70] found no effect of feeding BQCV on adult honeybee mortality, suggesting it is a benign pathogen, though lethal when fed to queen [79], drone [80] and worker [70] pupae. The high virulence of BQCV in honeybees that we here and others [81] have observed is probably due to it having been injected into hosts. From epidemiological theory, pathogen prevalence is often inversely related to virulence in insect host populations [82]. To explain its high prevalence in honeybees, we suggest that BQCV is rather benign when infecting adult A. mellifera workers through its typical faecal-oral route of transmission.
The Western honeybee is the dominant flower visitor across most terrestrial ecosystems of the world [83]. Dominant species in a community often disproportionately influence pathogen transmission and dynamics [84] through their central role in contact networks [85], exacerbated in the case of A. mellifera because it is probably the reservoir host of BQCV, DWV-A, DWV-B. Though we recorded little to no virulence of these viruses on B. terrestris under laboratory conditions, their impact on this and other bee species (and other flower visitors) under field-realistic conditions should be the focus of future studies to evaluate the role of viral spillover in wild bee decline.
Data accessibility. All data supporting the findings of this study are available from the Dryad Data Repository: https:// doi.org/doi:10.5061/dryad.fxpnvx0nt [86].