Microbiome of the wasp Vespula pensylvanica in native and invasive populations, and associations with Moku virus

Invasive species present a worldwide concern as competition and pathogen reservoirs for native species. Specifically, the invasive social wasp, Vespula pensylvanica, is native to western North America and has become naturalized in Hawaii, where it exerts pressures on native arthropod communities as a competitor and predator. As invasive species may alter the microbial and disease ecology of their introduced ranges, there is a need to understand the microbiomes and virology of social wasps. We used 16S rRNA gene sequencing to characterize the microbiome of V. pensylvanica samples pooled by colony across two geographically distinct ranges and found that wasps generally associate with taxa within the bacterial genera Fructobacillus, Fructilactobacillus, Lactococcus, Leuconostoc, and Zymobacter, and likely associate with environmentally-acquired bacteria. Furthermore, V. pensylvanica harbors—and in some cases were dominated by—many endosymbionts including Wolbachia, Sodalis, Arsenophonus, and Rickettsia, and were found to contain bee-associated taxa, likely due to scavenging on or predation upon honey bees. Next, we used reverse-transcriptase quantitative PCR to assay colony-level infection intensity for Moku virus (family: Iflaviridae), a recently-described disease that is known to infect multiple Hymenopteran species. While Moku virus was prevalent and in high titer, it did not associate with microbial diversity, indicating that the microbiome may not directly interact with Moku virus in V. pensylvanica in meaningful ways. Collectively, our results suggest that the invasive social wasp V. pensylvanica associates with a simple microbiome, may be infected with putative endosymbionts, likely acquires bacterial taxa from the environment and diet, and is often infected with Moku virus. Our results suggest that V. pensylvanica, like other invasive social insects, has the potential to act as a reservoir for bacteria pathogenic to other pollinators, though this requires experimental demonstration.

Introduction Sacbrood Virus (SBV), Black Queen Cell Virus (BQCV), Kashmir Bee Virus (KBV), and others known to contribute to colony loss and impaired fitness (reviewed in [39]). While these are often thought of as "honey bee viruses," infections in social wasps have been identified indicating the potential for cross-infection between insect species [28,[40][41][42][43][44]. Invasive species such as V. pensylvanica may spread diseases to new ranges and novel hosts, or conversely, be infected by endemic diseases that are commonly found in newly sympatric hosts [28,40,44]. In order to study the effects of emerging infectious diseases on insect populations, we surveyed V. pensylvanica colonies for Moku virus-a recently-described single-stranded positive-sense RNA virus (family: Iflaviridae) originally described in wasp colonies in Hawaii [45]. Aside from Hawaii, Moku virus has been found in Vespula spp. from several geographical locations, and has been found in honey bees and their associated Varroa mites [28,45,46]. While Moku virus is apparently widespread and may infect multiple insect species, the transmission potential, virulence, and effects on insect populations of Moku infections are currently unknown, although it is hypothesized that Vespula spp. are the natural reservoirs of this disease [45]. Previous work in our study population of V. pensylvanica suggests that Moku virus loads are strongly bimodal at the colony level, that colonies with high loads have reduced longevity in some years, and colonies that are at higher densities have higher Moku loads [47].
As the microbial ecology and effects of Moku virus on V. pensylvanica are largely unknown, we conducted an exploratory study using 16S rRNA gene sequencing and RT-qPCR on these invasive social wasps at the colony level. We investigated three lines of inquiry to better understand the microbial communities of our study organism: First, does V. pensylvanica associate with a defined microbiome, is there any evidence of environmental transmission of microbes, and are their microbiomes similar in the native and invasive range? Second, does this wasp species associate with endosymbiotic bacteria? Third, are there associations between Moku virus and the wasp microbiome?

Field sites and collections
Fieldwork was conducted under a permit from the National Park Service (HAVO-2016-SCI-0050). We collected adult worker wasps from the entrances of colonies found along Hilina Pali Road and at Kipuka Kahali'i in Hawaii Volcanoes National Park (HAVO) on the Big Island of Hawaii and the campus of the University of California, Riverside (UCR) in Riverside, CA. Colonies were discovered by following foragers back to the nest site, or by noticing characteristic foraging traffic at the colony entrance. We collected between 20 and 40 adult worker wasps from each of 40 colonies at HAVO between August 25-27, 2017, as well as from 13 colonies at UCR between October 6-26, 2017 (see S1 Table for nest site coordinates and sampling dates). We collected V. pensylvanica workers from nest entrances using a net or a portable vacuum. While collecting, we attempted not to disturb the colony and thus primarily collected foragers, though in some cases we likely collected guard wasps as well. We flash-froze the collected adults alive in a liquid nitrogen dry shipper, and then transferred to a -80˚C freezer upon return to the lab until processing. Thirty-seven colonies sampled in Hawaii are included in a companion study on colony longevity in V. pensylvanica [47]. Three of the Hawaii colonies were likely perennial colonies that had survived from the previous year, given their size, and none of the monitored colonies in Hawaii survived the next winter.

DNA extractions and library preparation
We extracted DNA by first pooling 20 wasps per colony (N = 53 colonies). We homogenized the samples by adding still-frozen insects to 15 mL tubes with grinding balls using a Geno/ Grinder (SPEX SamplePrep, Metuchen, NJ), with the block pre-chilled with liquid nitrogen. We then quickly transferred the homogenate to new sample tubes and re-froze at -80˚C until DNA or RNA extraction. Ultimately, we extracted DNA from a~20 mg aliquot of thawed homogenate using a Qiagen DNeasy kit (Qiagen, Valencia, CA) by following the manufacturer's protocol plus an overnight incubation at 55˚C.
We  [52,53], a unique barcode sequence, and Illumina adapter sequence through two rounds of PCR. We normalized the resulting libraries with a SequalPrep Normalization kit (ThermoFisher, Waltham, MA), then pooled the libraries and performed a final cleanup with a PureLink PCR Purification kit (Invitrogen, Carlsbad, CA). Finally, we sequenced the libraries with a V3 Reagent Kit with 2 X 300 cycles on an Illumina MiSeq Sequencer in the UC Riverside Institute for Integrated Genome Biology. We also ran blank samples to control for reagent contamination that we prepared and sequenced in the same way as regular samples. Raw sequence data are available on the NCBI Sequence Read Archive under accession number PRJNA707052.

Quantification of Moku viral titer
We used data reported in another study on Moku Virus and colony longevity [47] and repeat the methods here for clarity. Briefly, we quantified Moku virus titer using reverse-transcription quantitative PCR (RT-qPCR) of viral RNA. For each colony, we extracted RNA from a~20 mg aliquot of re-frozen wasp homogenate from 20 pooled workers (the homogenization procedure is described above) using 1 ml of TRIsure (Bioline Inc., Taunton, MA) following the manufacturer's protocols. We then quantified RNA for each sample using a Qubit spectrophotometer (ThermoFisher Scientific, Waltham, MA), and normalized RNA concentration to 2ng/ul. We then amplified Moku virus RNA using the primers MVF and MVR [46], and amplified Vespula eIF3 [54] as a reference gene. We used a BioRad CFX Real Time PCR machine (BioRad, Hercules, CA), and Luna OneStep RT-qPCR kits (New England Biolabs, Ipswich, MA) according to manufacturer instructions, with 10ul reaction volumes, 10ng of RNA per reaction, and an annealing temperature of 60˚C. Melt curves were checked to verify a single PCR product. We quantified reaction efficiencies alongside samples, and always obtained efficiencies between 95 and 105% for the 10-fold dilution series spanning the range of observed sample Cq values. We ran all samples in duplicate, and averaged Cq values. We calculated relative Moku Virus titer as-(Cq moku -Cq eIF3 ) for each sample, resulting in an index with higher values corresponding to greater viral titer on a log scale. Because colony virus titer was strongly bimodal in a larger dataset including two years of data [47], we used a titer of 7 as a threshold and classified colonies as either high-titer or low-titer.

Bioinformatics and statistics
We processed the 16S rRNA gene libraries with the QIIME2 pipeline (v2019.7) [55] by first trimming adapters, sequencing primers, and low-quality ends from the reads, then used DADA2 [56] to quality filter the reads, remove singletons, and bin sequences into Amplicon Sequence Variants (ASVs; 100% identical sequence reads) using the default parameters, followed by reagent contamination removal with the R package "decontam" [57]. We used the QIIME2 q2-feature classifier [58] to assign taxonomy to the ASVs with the SILVA database trained to the 799-1115 region of the 16S rRNA gene [59] (as well as local BLAST searches against the NCBI 16S Microbial database), and generated sequence alignments with MAFFT [60], then tabulated ASV counts in a table. We used this table to calculate diversity metrics, and tested the statistical significance of alpha diversity (Shannon Diversity Index) with Kruskal-Wallis tests in QIIME2, and beta diversity (Bray-Curtis dissimilarities) through Adonis PERMANOVA (999 permutations) with the R package "vegan" [61] in R v. 3.5.1 [62]. We visualized the Shannon diversities through boxplots, proportional abundance of bacterial taxa as a stacked bar plot, and beta diversity through Non-metric Multidimensional Scaling (NMDS) with the R package "ggplot2" [63]. We indicated shared and unique ASVs through Venn diagrams with the BEG Venn Diagram tool (http://bioinformatics.psb.ugent.be), and plotted heatmaps with "pheatmap" [64]. We used ANCOM [65] to test for differentially abundant taxa between locations, and used mantel tests in "vegan" to test for correlations between bacterial diversity and Moku virus titer.

Results
We obtained 1,033,410 quality-filtered (average quality score Q38) 16S rRNA gene sequences with an average of 19,498 reads per sample from two separate populations (Total N = 53; "Riverside" N = 13; "Hawaii" N = 40) that clustered into 2,943 unique ASVs (S1 File). Through rarefaction analyses, we determined that we had acceptable ASV coverage at a read depth of 2,232 reads (S1 Fig . We compared the Shannon diversity of our samples, but did not find a significant difference in alpha diversity between the two populations (H = 2.62, P = 0.11, Fig 1). Most of the ASVs were unique to each population, with 872 (29.6% of ASVs, 12.0% of reads) only found in Riverside samples, 1,930 (65.6% of ASVs, 40.3% of reads) only found in Hawaiian samples, and 141 (4.8% of ASVs, 47.7% of reads) found in both populations (Fig 1).
We compared the microbiomes of wasp colonies with low Moku virus titer (N = 7) or high Moku virus titer (N = 26) in our Hawaiian samples (we did not detect Moku virus in any Riverside samples). We found no significant effect of viral titer on either the alpha (H = 2.95, P = 0.09) or beta diversity (F = 0.99, R 2 = 0.03, P = 0.45) in our samples, and did not find any significantly differentially abundant ASVs between high-titer and low-titer colonies. We also ran mantel tests using Spearman correlations to test relationships between wasps' alpha (Shannon diversity index) and beta (Bray-Curtis dissimilarities) microbial diversity with Moku virus titer, but Moku virus titer was not significantly correlated with microbial alpha or beta diversity in our samples (alpha: ρ = -0.20, P = 0.32; beta: ρ = 0.07, P = 0.41).

Discussion
Our results suggest that the invasive wasp V. pensylvanica associates with a simple microbiome consisting largely of lactic acid bacteria and Zymobacter, along with significant associations of endosymbiotic bacteria. Notably, by comparing wasp colonies from two geographically-distinct locations, we also show that social wasp-associated microbial communities may contain environmentally-associated microbes similar to other Hymenoptera [17,18,29,50], and as

PLOS ONE
The microbiome of Vespula pensylvanica and associations with Moku virus mentioned above, wasp colonies appear to harbor some identical microbial taxa across their range. Similar to our findings, previous work has shown that some social Vespa species also possess simple microbiomes and associate with prey bees' bacterial symbionts [25]. In our small-scale study, we found that the microbiomes of V. pensylvanica colonies are unaffected by Moku infection, and a larger sample size would be useful to confirm this result. Coupled with the fact that Moku virus is transmissible to honey bees [45,46], multispecies infection interactions may occur between wasps and other insects. More research is needed to understand how ecological invasions affect insect populations and their associated microbes.

PLOS ONE
The microbiome of Vespula pensylvanica and associations with Moku virus The microbial communities of wasp colonies in our study were somewhat different based on their geographical location, suggesting that environmental exposure affects which bacteria they associate with. For example, Fructobacillus spp. and Leuconostoc spp.-taxa known to

PLOS ONE
The microbiome of Vespula pensylvanica and associations with Moku virus associate with hymenopterans and flowers [17,51,66]-were significantly different between Hawaii and Riverside samples, and may be potential symbionts for insects or plants. Our results may also be affected by uneven sampling, which may have caused us to find more total ASVs in wasps from Hawaii simply due to a greater number of these samples. Interestingly, while environment apparently plays a significant role in bacterial inoculation, the wasps seem to associate with taxa conserved across geographic distances, consisting largely of members of the Lactobacillaceae family along with an ASV of Zymobacter palmae. While we are unable to conclude that this microbiome is maintained through generations, previous work has shown that other social wasps consistently harbored Z. palmae, and seemed to possess a small microbial community [25], although we recognize that we sequenced a different region of the 16S rRNA gene, potentially limiting direct comparisons with other amplicon studies. Likewise, as we analyzed pooled, whole wasps, we may be detecting bacteria present on the wasps and only reflect their surroundings, not a potential symbiosis. Similarly, we found taxa corresponding to honey bee gut microbiota suggesting that wasps were preying upon honey bees and were exposed to their bacteria, confirming previous results [25]. While this agrees with previous studies, the honey bee core bacteria were not seen ubiquitously throughout our samples, suggesting that the bacteria are not colonizing the wasps and are likely being detected only as DNA presence, not live cells. This result was not entirely surprising, as many members of the honey bee core gut bacteria are coevolved, specialized taxa that are unlikely to colonize other species [67]. As there is interest in the ability of wasps to vector diseases, assaying the specific anatomical locations and viability of wasp-and bee-associated bacteria present in V. pensylvanica would help uncover more interactions between invasive and native wasps.
Wasps in our study harbored a variety of putatively endosymbiotic bacteria-especially those collected in Hawaii. These endosymbionts included bacteria in the genera Arsenophonus, Rickettsia, Sodalis, and Wolbachia, which have been found in other wasp species [44,68,69], but to the best of our knowledge, we are the first to find Rickettsia and Sodalis spp. in Vespula species. Interestingly, several of our samples' microbiomes were dominated by either Arsenophonus or Rickettsia implying that these bacteria are likely able to either outcompete other environmentally-acquired taxa or conserved microbes or reach such high densities that they dominate the sequencing results. While our results are promising, we are unable to rule out sampling/sequencing error as a source of the peculiarly high proportion of endosymbionts, and as our samples were pooled, we may simply be detecting a colony-wide infection rather than individual wasps' microbes. We are also unable to track the transmission mode of the endosymbionts, especially as some are known to have complex epidemiology [70]. Additionally, in other samples, endosymbiont infections were either not detected or in lower proportional abundance as part of a more complex microbial community, especially when comparing the much lower levels in wasps collected in Riverside versus Hawaii. These two apparently conflicting results suggest that endosymbionts are not ubiquitous in V. pensylvanica colonies, and as the wasp colonies otherwise appeared normal, may not be virulent or beneficial to colonies. As our data are compositional, we are unable to assess the bacterial load of the endosymbionts at this time, as infection titer is likely important in understanding infection dynamics [71]. Suggestively, colony-level Arsenophonus titer (quantified with qPCR) was correlated with the proximity of wasp colonies to honey bee hives at the Hawaii field site [47], which could imply spillover from honey bees, although this remains to be tested. We suggest that manipulation studies be conducted with endosymbiotic bacteria to probe deeper into both individual-and colony-level effects on wasps in multiple generations.
Neither Moku virus presence nor titer affected the wasps' microbiomes. Moku virus is a recently-discovered virus that is known to infect V. pensylvanica and honey bees, and has been previously detected in Hawaii, the United Kingdom, and Belgium [45, 46]-although the virulence, infectivity, and range of this virus is currently unknown. Moku virus titer predicted Vespula colony longevity in 2016 at our Hawaii field site, but not in 2017 when the samples in this study were collected, suggesting a variable role of this virus in wasp colony dynamics [47]. Moku virus presence and titer were not associated with microbiome diversity, indicating that there may not be cross-talk between the microbiome and this particular viral infection in social wasps. Likewise, as a non-phage virus, Moku likely does not directly infect bacteria [45]. Many other non-viral hymenopteran diseases are known to interact with the microbiome [72,73], and in bees the gut microbiome is heavily involved in immune function and defense against parasites [73][74][75][76]. Even though Moku virus did not affect the microbiome, this virus can still infect honey bees, which indicates the potential for introduced wasps to transmit disease to nontarget hosts [44,77].
In summary, our data suggest that the social wasp V. pensylvanica possesses a simple microbiome mainly composed of lactic acid bacteria and putatively environmentally-acquired taxa along with several species of endosymbionts. Furthermore, we show that these invasive wasps likely maintain some of these core bacteria across geographically distinct regions, indicating potential symbioses. We also found that wasp colony microbiomes are not affected by Moku virus presence, despite high titers, which supports the hypothesis of V. pensylvanica being a natural reservoir for this disease, along with a potential vector to other insects. We suggest that future studies examine the physiological effects of Moku virus on wasps and other Hymenoptera, and the potential for invasive wasps to change the microbial and viral ecology of their introduced ranges. We are also interested in assaying the contributions of wasps' microbes to colony fitness and nutrition and subsequent work should be conducted to understand the microbial ecology of social wasps.  Table. Latitude and longitude coordinates and sampling dates for each sample. "R" denotes "Riverside" and "H" denotes Hawaii sampling locations. (DOCX)