Wolbachia enhances insect‐specific flavivirus infection in Aedes aegypti mosquitoes

Abstract Mosquitoes transmit a diverse group of human flaviviruses including West Nile, dengue, yellow fever, and Zika viruses. Mosquitoes are also naturally infected with insect‐specific flaviviruses (ISFs), a subgroup of the family not capable of infecting vertebrates. Although ISFs are not medically important, they are capable of altering the mosquito's susceptibility to flaviviruses and may alter host fitness. Wolbachia is an endosymbiotic bacterium of insects that when present in mosquitoes limits the replication of co‐infecting pathogens, including flaviviruses. Artificially created Wolbachia‐infected Aedes aegypti mosquitoes are being released into the wild in a series of trials around the globe with the hope of interrupting dengue and Zika virus transmission from mosquitoes to humans. Our work investigated the effect of Wolbachia on ISF infection in wild‐caught Ae. aegypti mosquitoes from field release zones. All field mosquitoes were screened for the presence of ISFs using general degenerate flavivirus primers and their PCR amplicons sequenced. ISFs were found to be common and widely distributed in Ae. aegypti populations. Field mosquitoes consistently had higher ISF infection rates and viral loads compared to laboratory colony material indicating that environmental conditions may modulate ISF infection in Ae. aegypti. Surprisingly, higher ISF infection rates and loads were found in Wolbachia‐infected mosquitoes compared to the Wolbachia‐free mosquitoes. Our findings demonstrate that the symbiont is capable of manipulating the mosquito virome and that Wolbachia‐mediated viral inhibition is not universal for flaviviruses. This may have implications for the Wolbachia‐based DENV control strategy if ISFs confer fitness effects or alter mosquito susceptibility to other flaviviruses.

Dengue fever caused by DENV is a severely debilitating disease with 40% of the world's population at risk of infection and an estimated 300 new infections reported yearly (Bhatt et al., 2013).
A newly emerging threat to global health is Zika fever caused by the ZIKV with outbreaks reported in both tropical and subtropical regions (Pujhari & Rasgon, 2016;Younger, 2016).
Mosquitoes are also known to naturally harbor flaviviruses that are incapable of infecting humans or other vertebrate animals and are therefore known as insect-specific flaviviruses (ISFs).
Cell fusing agent virus was the first ISF to be discovered in Ae.
This suggests that the presence of CxFV may make mosquitoes more susceptible to WNV (Newman et al., 2011). Conflicting results were also observed in other studies where CxFV did not have an effect on replication, transmission, or dissemination of WNV in Culex quinquefaciatus (Kent, Crabtree, & Miller, 2010).

Wolbachia pipientis is an endosymbiotic bacterium of insects
that is currently being developed as a biocontrol agent (Iturbe-Ormaetxe, Walker, & LO'Neill, 2011;Moreira et al., 2009;Walker et al., 2011). Wolbachia is naturally present in about 52% of arthropods and is maternally inherited (Weinart, Arauju-Jnr, Ahamed, & Welch, 2015). The symbiont manipulates host reproduction, such that the eggs of Wolbachia-free females do not hatch when they have been fertilized by a Wolbachia-infected male (Serbus, Casper-Lindley, Landmann, & Sullivan, 2008). This phenomenon, referred to as cytoplasmic incompatibility (CI), leads to the spread and invasion of Wolbachia into wild populations. Another desirable phenotype of Wolbachia is its ability to inhibit infection of the host with other pathogens (Bian, Xu, Lu, Xie, & Xi, 2010;Bian et al., 2013;Frentiu et al., 2014;Walker et al., 2011). This "pathogen blocking" by Wolbachia was first observed in Drosophila melanogaster (Hedges, Brownlie, O'Neill, & Johnson, 2008;Teixeira, Ferreira, & Ashburner, 2008) where flies infected with Drosophila C virus (DCV) and cricket paralysis virus and accumulated virus at a slower rate leading to higher survival rates compared to Wolbachia-free controls (Hedges et al., 2008).
Although largely consistent, there are some reports of Wolbachia enhancing rather than preventing pathogen co-infection including Plasmodium and WNV within Anopheles gambiae and Culex tarsalis, respectively (Dodson et al., 2014;Hughes, Vega-Rodriguez, Xue, & Rasgon, 2012). In both of these instances, however, the mosquitoes were only transiently infected with Wolbachia via artificial microinjection and so may not be representative of insects with germline tissue infections (Joubert & O'Neill, 2017). Several vectors naturally infected with Wolbachia have also exhibited increased susceptibility to pathogens. This has been shown in Cx. pipiens and Spodoptera exempta (African armyworm moth) with increased susceptibility to Plasmodium and nucleopolyhedrovirus (double-stranded DNA virus), respectively (Graham, Grzywacz, Mushobozi, & Wilson, 2012;Zele et al., 2014). In contrast, the natively infected Ae. albopictus exhibits reduced susceptibility and transmission of DENV (Mousson et al., 2012). These studies suggest that Wolbachia-mediated pathogen blocking may depend on several factors that are influenced by specific Wolbachia strain and pathogen-host interactions including history of association.
The mechanistic basis of Wolbachia-mediated pathogen blocking is still not well understood (Terradas & McGraw, 2017). Currently, pathogen blocking has been partly attributed to the ability of Wolbachia to increase the innate immune responses of the host, thereby making it resist subsequent pathogen infection (Bian et al., 2010;Pan et al., 2012;Rances, Ye, Woolfit, McGraw, & O'Neill, 2012). It has also been hypothesized that competition between Wolbachia and pathogens for key host resources such as lipids (Caragata et al., 2013) and intracellular space (Moreira et al., 2009) may underpin blocking. This may be particularly relevant for viruses that require lipids for attachment and entry into host cells and for replication (Lu, Cassese, & Kielian, 1999;Mackenzie, Khromykh, & Parton, 2007). Most recently, there is some evidence primarily from Drosophila that Wolbachia may be modifying host cellular structures or organelles rendering them less hospitable to viral replication (Rainey et al., 2016;White et al., 2017). A range of studies also point to a correlation between Wolbachia densities and the strength of blocking (Frentiu et al., 2010;Osborne, Iturbe-Ormaetxe, Brownlie, O'Neill, & Johnson, 2012), a trend that would be expected with any of the above explanations for blocking.
While Wolbachia appears to shift the composition of the microbiome in mosquitoes (Audsley, Seleznev, Joubert, O'Neill, & McGraw, 2018), little is known about its effects on ISFs. There is also little known about the effects of ISFs on mosquito health. If these infections affect survival or reproduction, Wolbachia-infected insects in the field may receive an advantage in carrying the symbiont. For example, native viruses in Drosophila such as DCV and cricket paralysis virus reduce host fitness and Wolbachia infections are hence beneficial (Hedges et al., 2008). A survey in wild populations of D. melanogaster demonstrated that Wolbachia infection was not associated with changes in the diversity of native viruses in the insect (Webster et al., 2015). Wolbachia infections in Ae. aegypti differ significantly from those found in D. melanogaster however, exhibiting higher symbiont loads (McGraw, Merritt, Droller, & O'Neill, 2002;McMeniman et al., 2009;Moreira et al., 2009;Walker et al., 2011), broader tissue distributions (Moreira et al., 2009;Walker et al., 2011), greater activation of the immune response (McGraw et al., 2002;McMeniman et al., 2009;Moreira et al., 2009;Walker et al., 2011), and greater fitness costs (McMeniman, Hughes, & O'Neill, 2011;Min & Benzer, 1997). These discrepancies may result from different periods of association/evolutionary history, long (~5,000 years) in the case of D. melanogaster (Richardson et al., 2012) and short (<10 years) in the case of the newly infected Ae. aegypti .
Understanding the fitness consequences of Wolbachia for Ae.
aegypti is necessary to effectively model the long-term stability and success of the symbiont as a biocontrol agent in wild populations.
Our work focused on determining if Wolbachia-mediated viral blocking extends to naturally occurring flaviviruses in mosquitoes.
We sampled Wolbachia-infected mosquitoes from field release sites in Cairns, Australia, and symbiont-free mosquitoes from nearby control areas outside of the Wolbachia release zone. Using flavivirus general degenerate primers, we amplified the NS5 region of the virus genome and sequenced the PCR amplicons of individual positive mosquitoes using Miseq Illumina sequencing. We further screened laboratory colonies and field mosquitoes using primers designed specifically for several of the ISF sequences. We found that ISFs are common and widely distributed in Ae. aegypti populations with infection rates and abundance consistently higher in field mosquitoes compared to the laboratory colonies. This possibly indicates that variations in environmental conditions could be playing a role in controlling ISF infection in Ae. aegypti. Unexpectedly, we found that Wolbachia enhanced ISF infection rates and loads in Ae. aegypti demonstrating that the antivirus effect associated with Wolbachia is not common to all flaviviruses. These findings may have implications for Wolbachia-DENV control if ISFs affect host fitness or play a role in mosquito susceptibility to flaviviruses.

| Mosquito sampling
Wolbachia-infected and uninfected (wild-type) Ae. aegypti mosquitoes were sampled from three different communities in Cairns, Australia. The Wolbachia-infected mosquitoes were sampled in 2013 from two sites where wMel mosquitoes were released in 2011 (Hoffmann et al., 2011) and 2013 (Ritchie, 2014). These are Gordonvale (GV) and Parramatta Park (PP), respectively. Wild-type mosquitoes were sampled from Holloways Beach (HB) that is outside the original release zone. BG-sentinel mosquito traps (Biogen, Germany) were set randomly in these areas, and adult mosquitoes were collected overnight. Ae. aegypti mosquitoes were morphologically identified and placed in vials containing 80% ethanol. A total of 95 individual mosquitoes were assessed across the three collection sites (39 from GV, 21 from PP, and 35 from HB) for the presence of ISFs.

| Screening for insect-specific flaviviruses
RNA and DNA were simultaneously extracted from each mosquito using the TRIzol ® method from Invitrogen (Life technologies, Carlsbad, CA, USA). The DNA was used to screen for the presence of Wolbachia infection via qPCR as previously described (Frentiu et al., 2014). Two samples each from PP and GV were found to be Wolbachia negative and were excluded from further analysis. All HB samples were confirmed to be Wolbachia negative as expected. The RNA was DNase-treated to remove genomic DNA contamination using DNase 1 recombinant RNase-free (Roche, Germany). Reverse transcription of RNA to cDNA and the PCR amplification of the NS5 region using general degenerate flavivirus primers were carried out following the protocol of Sanchez-Seco et al. (2005). Briefly, reverse transcription of RNA to cDNA and subsequent first-round amplification were carried out using 1 μg of RNA in the Access RT-PCR System (Promega, Madison, WI, USA). Reverse transcription controls that did not include enzyme were included in each run to rule out genomic DNA contamination. One microliter of the first-round amplification was then used for the second round of nested PCR. All PCR products were run in a C1000™Thermal Cycler (Bio-Rad, CA, USA). PCR products (143 bp) were analyzed using gel electrophoresis on a 2% Agarose gel (Sigma, Life Science, USA) stained with RedSafe™ (iNtRON Biotechnology). Products were then visualized on the Quantum gel documentation system (Fisher Biotec).

| Sequencing of PCR products
A subset of six individual mosquitoes from each of the three sites were selected for further processing for sequencing. To ensure a good representation of an area, the samples were selected from different traps that were not in close proximity. One microliter of the second-round PCR product of each of the selected individual samples was used as the template for a 5-cycle amplification with primers barcoded with Illumina sequence adapters (Berry, Ben Mahfoudh, Wagner, & Loy, 2011). The PCR amplicons were analyzed using gel electrophoresis as described above. The amplicons were then excised and gel extracted using QIAquick gel extraction kit (Qiagen, Germany) following manufactures instructions. Extracted samples were then paired-end sequenced using MiSeq at Ramaciotti sequencing center, NSW, Australia.

| Processing and clustering of sequences
All sequences were processed with cutadapt (Martin, 2011) as the very first step, to remove degenerate primers used for the PCR amplicons. Cutadapt was run with the following settings; minimum overlap 10 nucleotides, minimum read length: 1 nucleotide (this is mainly to allow downstream R1 and R2 merging), and number of attempts to trim a primer was set to 2. To classify sequences into Operational Taxonomical Units (OTUs), vsearch (Rognes, Flouri, Nichols, Quince, & Mahe, 2016) was used to merge reads with number of mismatches set to two nucleotides, the number of allowed N's set to 0, and the minimum overlap set to 32 bases. This was followed by filtering reads based on expected error of 1. Identical sequences were then collapsed into a single sequence (dereplication) and then clustered using 97% identity. Contingency table of cluster counts was subsequently generated using usearch (Edgar, 2010).

The OTUs were finally imported into the flavivirus Database under
The Virus Pathogen Resource (www.viprbrc.org) and the Basic Local Alignment Search Tool (BLAST) used to find the closest match or hit of each OTU.

| Phylogenetic reconstruction
Individual OTUs were aligned with their best hits using the multiple sequence comparison by log-expectation (MUSCLE) tool (Edgar, 2004) provided by The European Bioinformatics Institute (EMBL-EBI). The aligned sequences were manually trimmed and then imported into Phylogeny.fr together with other common flaviviruses (Dereeper, Audic, Claverie, & Blanc, 2010;Dereeper et al., 2008).
The one click mode of Phylogeny.fr that uses MUSCLE for sequence alignment and maximum likelihood (PhyML) for tree building with aLRT (approximate likelihood-ratio test) statistical test for branch support values, and TreeDyn for tree drawing was used for the phylogeny tree (Anisimova & Gascuel, 2006;Chevenet, Brun, Banuls, Jacq, & Christen, 2006;Guindon & Gascuel, 2003).

| Screening of field mosquitoes using OTU-specific primers
Primers (Table 1) were designed for 7 ISF OTUs that were selected based on their abundance and diversity in the sequenced samples as well as their phylogenetic positions. All primers were designed using the Primer3 tool in The Virus Pathogen Resource database (viprbrc.org). Quantitative PCR using SYBR Green (Roche, Applied Science, Switzerland) in a LightCycler480 (Roche, Applied Science, Switzerland) was then used to validate the presence and abundance of OTUs in all mosquitoes sampled from the field. This was performed using 1 μl of the first-round amplification, 5 μl of 5X SYBR Green master mix, and 0.5 μl of 10 mmol/L each of forward and reverse primers in a total volume of 10 μl. The cycling conditions were pre-incubation at 95°C for 5 min, 45 amplification cycles of 95°C for 10 s, 60°C for 10 s, and 72°C for 10 s followed by a melting curve at 95°C for 5 s, 65°C for 1 min, and a continuous acquisition mode at 97°C. The housekeeping gene RPS17 (Cook et al., 2006;Thellin et al., 1999) was used to normalize virus abundance.

| Screening of laboratory mosquitoes
Wolbachia-infected and wild-type lines maintained in the laboratory were screened for ISFs to ascertain whether they showed similar Switzerland) using 1.5 μl of cDNA, 5 μl of 5X SYBR Green master mix, and 0.5 μl of 10 mmol/L each of forward and reverse OTUspecific primers in a total volume of 10 μl. The cycling conditions were as above.

| Statistical analysis
To determine whether there was an association between A Mann-Whitney test in GraphPad Prism (version 6) was used to analyze differences in ISF abundance between the wild-type and wMel mosquitoes.

| Wolbachia infection is associated with higher rates of ISF infection as measured by ISF generalist primers in PCR
We observed high ISF infection rates in the all field-collected samples as measured by PCR; 100% and 95% for GV (Gordonvale) and PP (Parramatta Park), respectively, and 74% for HB (Holloways Beach). If we test for the effect of Wolbachia in dictating infection frequency, we see that it is significant (Wald = 7.80; df = 1; p = .005).
Our findings suggest that ISFs are a common feature of the mosquito virome and that Wolbachia may be enhancing the frequency of infection in the field.

| The ISFs include well-characterized viruses as well as what appear to be novel viruses
To identify specific ISFs in the Ae. aegypti field populations, we sequenced a subset of samples that were ISF positive for both the wild-type (n = 6) and wMel (n = 12) mosquitoes. Following clustering analysis, a total of 26 "unique" ISF OTUs were identified in the sequenced samples (Figure 1). A list of all the OTUs and their sequences can be found in Table S1. Four OTUs (1, 2, 3, and 13) were very similar (>80%, Table 2) to the previously described ISFs Kamiti river virus, Cell fusing agent, and CbaAr4001.
This group also forms a strongly supported (95%) phylogenetic cluster ( Figure 1). OTU2's closest relative (75% bootstrap support) was Cell fusing agent. The majority of the OTUs, however, had low similarity (<50%) to known ISFs and OTUs 25 and 26, which had no match (Table 2). This lack of close relatives is recapitulated in the poor resolution within the phylogeny (Figure 1).

| Wolbachia is often associated with higher ISF frequencies
All but five OTUs (16, 20, 21, 23, and 25) were fixed in wildtype and wMel-infected field populations based on sequence analysis. In the group of viruses not fixed, we found a significant effect of Wolbachia on infection frequency (Wald = 5.49, df = 1, p = .019), suggesting that Wolbachia was associated with higher rates of infection in the sequenced samples ( Figure 2). We then tested whether these same trends were also present in the total set of samples (sequenced and not, n = 93) from the field using RT-qPCR primers designed specifically for seven of the OTUs (Table 1). OTUs 1-3 were selected as they are closely related to well-characterized ISFs (Figure 1, Table 2) and because they were In each case of significance, OTU frequencies were higher in wMel mosquitoes (Figure 3a). OTUs 16 and 21 that were previously not found in the sequenced wild-type samples were detected through RT-qPCR. This may be due to the sensitivity cutoff employed with the sequence data whereby we excluded OTUs with <10 sequence reads. Lastly, we then determined if these differences were also seen in laboratory lines of wMel-infected and wild-type mosquitoes. Unlike in the field, the rates of infection appear similar (Wald = 1.177; p = .27) between the two lines ( Figure 3b).

| Wolbachia infection is associated with differences in abundance of ISFs
To determine whether the presence of Wolbachia had an effect on the abundance of ISFs, we compared the viral load of the seven selected OTUs between the wild-type and wMel mosquitoes under field and laboratory conditions by RT-qPCR. Even though Wolbachia reduced the load of OTU2 in the field (Figure 4a) having a higher load (p = .017) compared to wMel mosquitoes, the opposite effect was observed in the laboratory where the wMel mosquitoes had a significantly higher load (p < .0001) than that in the wild type. In the field, OTU20 (p = .0009) and OTU21 (p < .0001) were significantly more abundant in wMel mosquitoes compared to the wild type (Figure 4b,c). This effect was not observed in the laboratory lines with no significant differences observed in loads of OTU20 (p = .17) and OTU21 (p = .91) between wMel and wildtype mosquitoes (Figure 4e,f). There were no significant differences between loads of OTU1 (p = .94) and OTU3 (p = .63) in wMel and wild-type mosquitoes in the field. In the laboratory lines however, wMel mosquitoes consistently had a higher abundance of OTU1 (p < .0001) and OTU3 (p < .0001) compared to the wild type. In both the field and laboratory mosquito lines, there were no differences in the loads of OTU16 (p > .05) and the OTU25 (p > .05) between Wolbachia-infected and wild-type mosquitoes ( Figure S1). In summary, regardless of the mosquito line, ISF loads varied considerably between field and laboratory environments with the former consistently harboring higher ISF density than the latter. This suggests that environmental conditions and differences in host genetic background may influence abundance of ISFs. Our findings demonstrate that in general, Wolbachia does not inhibit ISF loads in Ae. aegypti mosquitoes and in some cases may enhance them. It also suggests that the effect of Wolbachia on ISFs is virus-specific and environmental conditions may influence this effect, such that the laboratory environment may not be predictive of the field.

| D ISCUSS I ON
Despite large-scale field releases of Wolbachia-infected Ae. aegypti mosquitoes (Hoffmann et al., 2011;McGraw & O'Neill, 2013;Ritchie, 2014), the antivirus or blocking effect of the symbiont on the naturally occurring ISFs in Ae. aegypti mosquitoes is currently not known. This work therefore examined whether wMel Wolbachia-infected mosquitoes sampled from field release sites in Australia and in laboratory populations exhibit symbiontassociated changes in their ISFs. Generally, we found that in the

F I G U R E 2 ISF infection rates for OTUs not fixed in both WT
and Wolbachia-infected mosquitoes from the field as determined by sequencing. Across these 5 OTUs, wMel mosquitoes exhibited higher infection rates (p = .019) field wMel mosquitoes had both higher ISF infection rates and abundances compared to wild-type mosquitoes. We should point out that the number of field populations tested is small and so the findings could be the result of environmental or stochastic factors.
Mosquitoes from other field release sites around the globe should be profiled to determine whether these relationships are robust and generalizable. In the laboratory where Ae. aegypti are reared under optimal conditions, there was no difference in ISF infection rates between wild-type and wMel mosquitoes. However, wMel mosquitoes did exhibit higher loads of ISFs compared to the wild type. These findings were unexpected given that Wolbachia has been extensively shown to inhibit flaviviruses of medical importance in Ae. aegypti (Bian et al., 2010;van den Hurk et al., 2012;Moreira et al., 2009;Walker et al., 2011).
A total of 26 OTUs were observed in the sequenced field samples with all mosquitoes harboring multiple, concurrent infections.
Some of these OTUs were closely related to each other indicating they are likely variants of the same virus. Very few of the OTUs clustered with characterized ISFs present in the databases. This confirms observations in invertebrates (Shi et al., 2016), and more specifically in Drosophila (Webster et al., 2015), that there are novel insectspecific viruses that are yet to be classified. Our study is limited by sequencing a short fragment (143 bp) that probably affects the ability to have unambiguous matches and could explain why some OTUs matched to more than one virus in the database.
Environmental variables therefore need to be tested empirically to establish their effects on ISF infection rates and load in Ae. aegypti mosquitoes. It is also possible that age could contribute to variation in ISF infection (Bolling et al., 2012), but the sampling of adults from wild populations did not allow for age control.
It was unexpected to observe a higher ISF infection rate in wMel increase susceptibility of the DNA virus nucleopolyhedrovirus in the African armyworm, S. exempta (Graham et al., 2012). In laboratory lines, Wolbachia does not influence ISF infection rates suggesting that population genetic variation and differences in environmental conditions between the laboratory and field could be influencing Wolbachia interaction with ISFs. This hypothesis is not supported by Kho et al. (2016) and Caragata et al. (2013) (Hughes et al., 2012) but inhibits P. falciparum (Hughes, Koga, Xue, Fukatsu, & Rasgon, 2011) in An. gambiae.
The effect of the Wolbachia-ISF relationship on viruses of medical importance has not been examined in mosquitoes. It is possible that ISF enhancement by Wolbachia may not have an effect on arboviruses (Crockett et al., 2012;Kent et al., 2010). Hobson- Peters et al., 2013), further strengthening the pathogen blocking effect of Wolbachia. Given that ISF is common and widely distributed in Ae. aegypti mosquitoes, this could be advantageous to the current Wolbachia-dengue control strategy. More concerning is the possibility of ISF enhancement resulting in increased susceptibility of mosquitoes for arboviruses as was reported in a different study for WNV (Newman et al., 2011). This would have serious consequences for the current Wolbachia-DENV control strategy as Wolbachia-infected mosquitoes will facilitate arbovirus proliferation instead of limiting them. Finally, our findings point to the need to carefully examine environmental conditions before embarking on Wolbachia-Ae. aegypti field releases as the Wolbachia-pathogen effects observed in the laboratory may not be representative of the field.

ACK N OWLED G M ENTS
The authors would like to thank the Cairns Eliminate Dengue team at James Cook University for assistance with sample collection.
The research was supported by an ARC Discovery Grant to EAM (DP160100588).

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
EAM conceptualized and supervised the project. CK provided field samples. HEA did the experimental work. KT, HEA, EAM, RH, and DP analyzed data. The manuscript was written by EAM and HEA.