A single A. mellifera colony, sourced from a local beekeeper, was brought to Xi’an Jiaotong Liverpool University in April 2016. Honey bee pupae with purple eyes were collected by opening capped brood cells. Most pupae were free from mites in the period April 20-23 2016, with only a small number infested with V. destructor. During May 20-23 2016, we collected further pupae and found that 33% of the capped brood cells were now infested with V. destructor and none of them contained T. mercedesae. When purple eyed-pupae were infested by V. destructor, both were sampled simultaneously and stored separately. A second A. mellifera colony infested with T. mercedesae was brought in October 2016. Purple eyed-pupae, with or without infesting T. mercedesae, were collected from November 1–8 2016, as described above (50% of capped brood cells contained T. mercedesae). V. destructor was absent in all of the opened capped brood cells.

Total RNA was isolated from individual pupae and mites using Total RNA Extraction Reagent (GeneSolution), according to the manufacturer’s instructions. Glycogen (1 μg) was added to facilitate isopropanol precipitation of the mite RNA samples. Reverse transcription (RT) reactions were carried out using 1 μL of total RNA, random primers (TOYOBO), ReverTra Ace (TOYOBO), and RNase Inhibitor (Beyotime). RNase H (Beyotime) was added to digest RNA in RNA/cDNA heteroduplex after cDNA synthesis. These RT products were used for subsequent RT-PCR and qRT-PCR. RT-PCR was conducted to assess DWV infection in sampled pupae and mites using KOD-FX (TOYOBO) and the DWV primer set #1 (Supplementary Table S1) using the conditions 2 min at 94°C, followed by 32 cycles of 10 s at 98°C, 20 s at 55°C, and 30 s at 68°C, with a final 3 min extension at 68°C. PCR products were analyzed on a 2% agarose gel. PCR targeting A. mellifera and T. mercedesae EF-1α, and V. destructor β-actin mRNAs (Supplementary Table S1), were used as controls to verify successful RT (Supplementary Figure S1).

We performed qRT-PCR using a Hieff^TM® qPCR SYBR Green Master Mix (Low Rox Plus, Yesen) and DWV primer set #2 (Supplementary Table S1) with a QuantStudio5 Real-Time PCR System (Thermo Fischer). To perform absolute quantification of DWV RNA, we first prepared standard curves that corresponded to DWV target RNA. Target DNA was prepared by PCR, followed by purification with a AxyPrep^TM PCR cleanup kit (Axygen). DNA concentration was measured using a Nanodrop 2000 spectrophotometer (Thermo Fischer) to calculate copy number using the formula Copy number = DNA concentration (ng/μL) × 6.02 × 10²³ (copies/mol)/length (bp) × 6.6 × 10¹¹ (ng/mol), in which 6.6 × 10¹¹ ng/mol is the average molecular mass of one base pair and 6.022 × 10²³ copies/mol is Avogadro’s number. Linear standard curves were then generated using target DNA of 10¹–10⁹ copy number per reaction, followed by plotting of Ct values against log copy number values. After RT, the copy number of target RNA in a sample was estimated using the standard curve. The amount of cDNA added to each reaction was normalized by the Ct values of either A. mellifera, V. destructor, or T. mercedesae 18S rRNA.

PCR products obtained by RT-PCR were purified and directly sequenced by Sanger sequencing. To sequence the 4 kb DWV RNA region encoding partial sequence for Lp, helicase, and all of the VP proteins, we performed PCR using DWV primer set #3 (Supplementary Table S1) using the conditions 2 min at 94 °C, followed by 35 cycles of 10 s at 98°C, 20 s at 55°C, and 3.5 min at 68°C, with a final 5 min extension at 68°C. These PCR products were sequenced as previously described, and categorized into DWV viral subtypes (A, B, and C).

Prior to phylogenetic analysis, we aligned nucleotide sequences using Mafft (v7.307) (Katoh and Standley, 2013) and then used Gblocks (v0.91b) (Castresana, 2000) to automatically eliminate divergent regions or gaps. The best fit nucleotide substitution models were determined for the alignments by Jmodeltest (v2.1.10) with parameters set to “-g 4 -i -f -AIC -BIC -a” (Darriba et al., 2012). Mrbayes (v3.2.3) (Ronquist and Huelsenbeck, 2003) was run with the parameters set to 1,000,000 (1 sample/100 generations) until split frequencies were below 0.01. The first 25% of samples for each run were designated as the burn-in. DWV type A strain (NC_004830.2), DWV type B strain (NC_006494.1), DWV type C strain (CEND01000001.1), and Sacbrood virus (NC_002066.1) were used as references.

In April 2016, we collected purple eyed-honey bee pupae (6–8 days after capping) from a single colony. In this initial screen, there were almost no mites but approximately 61% (11 out of 18) of the honey bee pupae were infected with DWV when analyzed using RT-PCR (Figure 1A for analysis of 10 pupae).

In a second collection from the same colony in May 2016, we found that V. destructor was more abundant and 94% (17 out of 18) of pupae without infesting mites in the capped brood cells were DWV positive (Figure 1B for analysis of 10 pupae) (P < 0.041; two-tailed Fisher’s exact test). We also examined pupae from capped brood cells that were infested with V. destructor and found that 100% were positive for DWV infection (30 out of 30). We compared the copy number of DWV from pupae in mite infested cells to pupae without direct mite infestation. As shown in Figure 1C, mite infested pupae had a higher DWV copy number compared to those without infestation.

In November 2016, we also collected purple eyed-honey bee pupae from a colony infested with T. mercedesae. In this colony, 65% (13 out of 20) of pupae in capped brood cells without infesting mites were positive for DWV (Figure 1D for analyzing 10 pupae). We again collected pupae in capped brood cells infested by T. mercedesae (100% positive for DWV, 31 pupae) and compared their DWV copy number to pupae without direct mite infestation. As shown in Figure 1E, mite infested pupae had a DWV higher copy number than those without infestation.

We next measured and compared DWV copy number in 23 honey bee pupa/infesting Varroa mite pairs isolated from the same capped brood cells. We found a linear correlation between DWV copy number in the pupa and infesting mite. Specifically, we observed that the samples grouped into two clusters that were either low or high DWV copy number (Figure 2A). There were more pairs of pupa/mite in the cluster with low copy number of DWV. We also performed the same analysis with 28 honey bee pupa/T. mercedesae capped brood cell pairs; a linear correlation and the presence of two copy number clusters was also found in this mite species (Figure 2B). In contrast to V. destructor above, there were more pairs of pupa/mite in the cluster with high copy number of DWV in case of T. mercedesae infestation.

In order to determine the strains of DWV isolates in our study, we directly sequenced RT-PCR products containing the partial sequences for VP2 and VP1, and full sequence of VP4, from 24 honey bee pupa/V. destructor pairs isolated from matched capped brood cells.

Examination of the Sanger sequencing electrograms demonstrated that both pupae and mites could be infected by either a single (Figure 3A) or multiple DWV variants (Figure 3B). For example, in the sequence shown in Figure 3A, single peaks are present at all positions in both forward and reverse sequencing reactions. However, in other samples, multiple peaks are present at equivalent positions in both forward and reverse sequencing reactions (Figure 3B), demonstrating that these multiple peaks are not sequencing artifacts. Nevertheless, it is still possible that deep transcriptome sequencing may reveal the presence of very low prevalence variant(s) in infected pupae or mites that appear infected with a single variant. In our analysis, we found that eight pairs of pupa and Varroa mite shared infection by a single DWV variant, although the variant observed differed between pairs. There were three pairs in which pupae were infected by a single variant and the mites infected by multiple variants. Conversely, there were three pairs in which pupae were infected by multiple variants and the mites a single variant. Finally, the remaining 10 pupa and mite pairs were infected by multiple variants in each (Table 1).

We also analyzed DWV sequences in 17 honey bee pupa/T. mercedesae pairs. We found that two pairs of pupa and mite shared infection by single DWV variants, although the viral variant again varied between the pairs. There was a single pupa and mite pair in which the pupa was infected by a single variant and the mites infected with multiple variants. There was also one pair in which the pupa was infected by multiple variants and the mites infected by a single variant. The remaining 13 pairs were infected by multiple variants in both pupae and mites (Table 2).

Phylogenetic relationships estimated using these sequence (Accession number: MF134306-MF134370), and published DWV type A, B, and C sequences, indicated that all the DWV isolates in our study were more related to DWV type A rather than types B or C (Figures 4A,B). Because most posterior probabilities in the trees were less than 1, we additionally sequenced a 4 kb region of the viral genome containing part of the Lp sequence, all of the VP genes, and part of the helicase gene, from 14 representative samples (V. destructor, T. mercedesae, mite-infested pupae, and infestation-free pupae) in total. A phylogenetic tree was then estimated from these data (Accession number: MF134371-MF134383). As shown in Figure 5, all isolates are distinct from DWV types B and C and cluster together with type A. Although 2C1 outgroups the other isolates, including the type A (Figure 5), the viral genome sequence encoding the entire polyprotein, as well as parts of the 5′ and 3′ UTRs (Accession number: MF036686) shows higher identity to type A (96%) rather than types B (84%) or C (80%).

As shown in Table 1, 19 honey bee pupa and infesting V. destructor pairs were previously classified into high and low DWV copy number clusters (Figure 2). Among these, the same DWV variant was present in both the #10 pupa/mite pair and #14 pupa (Figure 4). However, the #10 and #14 pairs were classified into high and low copy number clusters, respectively (Table 1). A similar pattern was shown in the #6 mite sample (high copy number) and #23 mite sample (low copy number). These results demonstrate that the same DWV type A variant can result in either low or high copy number in both honey bee pupae and infesting V. destructor. As all low copy number pairs of pupa and infesting T. mercedesae were infected by multiple DWV variants (Table 2), a similar characterization was not possible for this species.

Our results demonstrate that an increase in the V. destructor population of a colony enhances the prevalence of DWV infection in honey bee pupae, even in pupae without direct mite infestation. Furthermore, direct infestation of pupae with either V. destructor or T. mercedesae dramatically increased DWV copy number of the pupa. These results support previous reports that V. destructor and T. mercedesae function as vectors for DWV and promote transmission in honey bees (Shen et al., 2005; Yue and Genersch, 2005; Forsgren et al., 2009). Our study indicates that both mite species are equally effective at increasing DWV copy number in infested honey bee pupae.

These data raise the question as to the source of DWV present in honey bee pupae without direct mite infestation. DWV could be present in larval food (Yue and Genersch, 2005; Singh et al., 2010) and that copy number may increase if honey bee workers have high DWV copy number due to mite infestation. Furthermore, V. destructor and T. mercedesae carrying DWV may enter brood cells, feed on hemolymph, but then leave before the cell is capped. Vertical transmission of DWV from the queen may also occur and the mite infestation of drone would enhance this process by producing the semen containing high titer of DWV (Chen et al., 2006; Yue et al., 2007; de Miranda and Fries, 2008).

The same question as to the initial source of DWV also applies to virus found in V. destructor and T. mercedesae. DWV may again be vertically transmitted from the mother mites and this hypothesis is supported by our recent study demonstrating that T. mercedesae males and nymphs have high DWV copy numbers that are the same as females (Dong et al., 2017). In addition, DWV could be horizontally transmitted from infested pupae (for both V. destructor and T. mercedesae) or adult workers (for V. destructor only).

We have confirmed that there is a linear correlation between DWV copy number in honey bee pupae and the mite individual that infests them, supporting previous studies (Forsgren et al., 2009; Kielmanowicz et al., 2015). For both V. destructor and T. mercedesae, pupa/mite pairs clustered into two groups containing either low or high DWV copy number. These results suggest that DWV replication in honey bee pupae is positively regulated relative to the amount of DWV introduced by infesting mites. As different DWV variants were found in the honey bee pupa/mite pairs (Tables 1, 2), it seems that DWV does not simply exchange/flow between pupa and mite. We found that both honey bee pupae and mites can be infected by either a single variant or multiple DWV variants before the mite enters the brood cell containing the fifth instar larva. In fact, we found that 19 out of 20 pupae were infected by multiple DWV variants, despite no direct V. destructor infestation.

The fact that V. destructor infestation increases DWV copy number in honey bees was well established (Shen et al., 2005; Yue and Genersch, 2005) and four major mechanisms have been previously proposed, (1) V. destructor induces immune-suppression in honey bees, stimulating DWV replication (Yang and Cox-Foster, 2005), (2) Replication-active more virulent DWV strains are selected for by V. destructor infestation (Martin et al., 2012), (3) DWV induces immune-suppression in honey bees at a threshold copy number (Nazzi et al., 2012; Di Prisco et al., 2016), and (4) Direct injection of the mixed variants of DWV to hemolymph of honey bee results in replication of the selective variant to high copy number (Ryabov et al., 2014). Our results seem to be inconsistent with the first and second proposed mechanisms.

If mite-mediated immune-suppression in honey bees occurs irrespective of DWV copy number in infesting mites, the copy number found in the paired honey bee pupae would not correlate with that in the mites. However, we found a strong correlation in copy number between mites and pupae, suggesting a relationship. Furthermore, even though the DWV isolates in our study were classified as the virulent type A (Mordecai et al., 2016a,b), the same variants were present in low and high copy number clusters in both honey bee pupae and mites. The patterns of DWV infection with either single or multiple variants in honey bee pupae and infesting mites appear to be the same (Tables 1, 2), suggesting that V. destructor and T. mercedesae may not select for a specific type A variant. This observation is similar to previous report suggesting that a single type of virulent DWV is amplified in honey bee pupae, despite the infesting V. destructor contains a diverse population of DWV strains (Ryabov et al., 2014). However, this may not be the case if other types (B and C) co-exist (Mordecai et al., 2016a). Based on several reports, both types A and B can be equally virulent (McMahon et al., 2016; Mordecai et al., 2016a). Selection of specific virulent DWV types/variants may depend on the extent of replication in the mites, although the replication mechanism of DWV in honey bees and mites is not yet fully understood.

Our data are most consistent with the third proposed mechanism: that there is immune-suppression at a threshold copy number. This support derives from our observation that DWV replication in honey bee pupae appears to depend on the copy number of DWV introduced by infesting mites. Honey bee pupae may be able to suppress DWV replication if only low levels of virus are introduced by the mite, keeping DWV copy number in the pupae relatively low. Having two major clusters of the pupa/mite pairs with low and high (but not medium) copy number of DWV (Figure 2) also supports this hypothesis. Surprisingly, the same results were basically obtained with two different mite species, suggesting that the common mechanism is likely to operate. If this model is correct, the dominant DWV variant present in the mite should also be the dominate type in the pupa of each high DWV copy number pupa/mite pair. Our study suggests that this is the case for both V. destructor and T. mercedesae. For example, the same DWV variant was present in all pupa/V. destructor pairs if they were singly infected, irrespective of copy number. In pair #6, the single variant present in the mite dominated the multiple variants found in the honey bee pupa. Both pupae and V. destructor in the pairs of #8 and #16 were infected by multiple DWV variants, although the dominant variant was shared between them (Table 1). The same pattern also occurs in pairs of pupa/T. mercedesae (Table 2). These findings support the hypothesis that DWV replication in honey bee pupae is positively regulated relative to the amount of DWV introduced by infesting mites. This may be due to immune-suppression in the pupae induced by high DWV copy number (Nazzi et al., 2012; Di Prisco et al., 2016). Our results do not fit well to the fourth proposed hypothesis above. Many individuals of V. destructor and T. mercedesae were infected by multiple variants and most of the mite-infested honey bee pupae were infected by multiple variants at either low or high copy number (Tables 1, 2). This hypothesis may also depend on the copy number of DWV injected as the third proposed mechanism. Our discussion above is based on the results obtained with 23 pupa/V. destructor and 29 pupa/T. mercedesae pairs, and thus analyzing more pairs in multiple colonies will be necessary to strengthen our arguments.

As the copy number of DWV in V. destructor or T. mercedesae is the critical factor determining the extent of DWV replication in infested honey bees, it is important to understand the transmission and replication mechanisms of DWV in the mites as these may differ from those in honey bee. The DWV population is extremely heterogeneous, with many variants and recombinants between different types in both honey bees and mites, even in the single colony. This viral heterogeneity may emerge because there are multiple hosts for the virus (honey bee, V. destructor, and T. mercedesae). It is thus important to understand how actively the different variants and recombinants replicate in honey bee and mite.