A DNA Virus of Drosophila

Little is known about the viruses infecting most species. Even in groups as well-studied as Drosophila, only a handful of viruses have been well-characterized. A viral metagenomic approach was used to explore viral diversity in 83 wild-caught Drosophila innubila, a mushroom feeding member of the quinaria group. A single fly that was injected with, and died from, Drosophila C Virus (DCV) was added to the sample as a control. Two-thirds of reads in the infected sample had DCV as the best BLAST hit, suggesting that the protocol developed is highly sensitive. In addition to the DCV hits, several sequences had Oryctes rhinoceros Nudivirus, a double-stranded DNA virus, as a best BLAST hit. The virus associated with these sequences was termed Drosophila innubila Nudivirus (DiNV). PCR screens of natural populations showed that DiNV was both common and widespread taxonomically and geographically. Electron microscopy confirms the presence of virions in fly fecal material similar in structure to other described Nudiviruses. In 2 species, D. innubila and D. falleni, the virus is associated with a severe (∼80–90%) loss of fecundity and significantly decreased lifespan.


Introduction
The advent of high-throughput DNA sequencing technology has facilitated the discovery and identification of microbes from environmental samples. Though most of the focus has been on metagenomics of microbial communities, leading to the detection of a huge diversity of bacteria and their related bacteriophages [1,2,3], viral metagenomic approaches have recently been used to explore viral diversity within individuals exhibiting symptoms ranging from obesity in humans to colony collapse disorder in honey bees to Shaking Mink Syndrome in mink [4,5,6,7].
There is growing appreciation for the important role of interactions among symbionts in host ecology and evolution [8,9,10,11,12]. In particular, the interaction between vertically and horizontally transmitted microbes and pathogens is the focus of much theoretical and empirical attention [13,14,15,16,17]. Wolbachia, probably the most common vertically transmitted endosymbiont among insects [18,19], has recently been found to confer resistance to certain RNA viruses in some hosts [14,15,20,21]. However, the importance of such virus protection in natural populations of Drosophila has not yet been explored. To investigate the relationship between insect hosts, endosymbiotic bacteria, and viruses, wild-caught Drosophila innubila females, about 1/3 of which are infected with Wolbachia [22], were screened for virus infection using a viral metagenomic approach.
D. innubila is a member of the mushroom-feeding quinaria group of Drosophila. They inhabit woodlands and forests of the Sky Islands of Mexico, Arizona, and New Mexico. Adults feed, mate and oviposit on mushrooms and larvae burrow through mushroom tissue, feeding on it prior to pupation. Species in the quinaria group are hosts to endosymbionts such as Wolbachia and Spiroplasma [23,24], parasitic nematodes [25], parasitoids [26] and mites (Emma Dietrich, personal communication). As for the vast majority of Drosophila species, very little is known about virus infection in natural populations.
Of the roughly dozen different virus identified in Drosophila [27,28], the most well-studied in natural populations is probably Drosophila Sigma Virus. Several studies have examined the frequency of Drosophila Sigma Virus infection in natural populations, indicating some degree of host specificity and infection frequencies ranging form absence to more than 70% [29,30,31]. Some work has also been done on host range of Drosophila C Virus, which infects several species from across the genus, with little host specificity [32]. However, many of these lines were maintained in the lab for several generations or from stock centers.
Reported here is the development and implementation of a new virus discovery protocol for Drosophila and other insects. This protocol revealed the presence of a new DNA virus that is both taxonomically and geographically widespread and is associated with significant mortality in at least two species of Drosophila. Wolbachia, however, appears to play no role in protection of D. innubila from the adverse effects of the virus.

Samples
Flies for the metagenomic survey were collected in the Chiricahua Mountains, as described in Unckless et al. [22], by sweep netting over mushroom baits in 2006 and 2007 near the Southwest Research Station, Portal, AZ. They were then shipped to the lab in Rochester, NY, where females were placed individually in vials and allowed to lay eggs for 6 days. These females were then dissected, their ovaries removed and screened for Wolbachia, and the rest of the carcass frozen at 280uC [22]. Forty-two Wolbachia-infected and forty-one uninfected carcasses, spanning 2 collection years, were selected for viral screening. Flies that produced few or no offspring were overrepresented, in order to increase the chance of including virus-infected flies. One fly that was injected with, and later died from, Drosophila C Virus (DCV) was added to the Wolbachia-infected sample to assess the efficiency of enrichment for viruses.

Virus enrichment and extraction
A protocol was developed to remove as much host nucleic acid as possible, leaving capsid-protected viral nucleic acids intact before extraction, production of cDNA libraries, and sequencing ( Figure S1). The protocol is a modification of several previously published protocols [4,33,34]. All flies for each sample were homogenized in 200 ml viral buffer [33], then centrifuged for 5 min. at RT at 25006g. The supernatant was then transferred to a new centrifuge tube. Genomic DNA was digested by adding 0.1 volumes of DNase I and reaction buffer (AMPD1-1KT, Sigma-Aldrich, USA) and incubating for 15 min. at RT. Sigma stop solution (included in DNase I kit) was added at 0.1 volume, and the solution was incubated at 70uC for 10 min. Genomic RNA was digested with 2 ml 0.02 mg/ml RNase A/T1 (ENN051, Fermentas, Glen Burnie, MD) incubated at 37uC for 3 hours. 1 ml Ribolock (EO0381, Fermentas, Glen Burnie, MD) was added to protect viral nucleic acids. Samples were enriched for viruses because, in any infected fly, the total RNA from a virus will be only a small fraction of the total RNA from the entire fly. In addition, because most flies are probably uninfected with viruses, pooling the flies for virus detection requires selective removal of nucleic acids of the host and resident microbes.
RNA was extracted using the E.Z.N.A. viral RNA extraction kit (R6874-01, Omega Bio-Tek, USA), which will also isolate DNA. Further sample preparation including library preparation (Rapid Library Preparation Method, Roche, Germany), nebulization and emulsion was performed at the Engencore sequencing facility (Columbia, SC). The 2 samples were bar-coded and run on 1/4 of a chip using a Roche 454 machine with Titanium chemistry.

Data analysis
Both raw reads and contigs (assembled at Engencore using the Roche/454 Life Sciences Newbler algorithm) were analyzed. All searches were performed locally using stand-alone BLAST+ [35] with a minimum E-value of 0.0001. Initially, each contig and read were searched against the RefSeq protein [36] database using BLASTx with the BLOSUM62 matrix and gap costs of 11 and 1 for opening and extension, respectively. Since the goal was in finding viruses, several additional searches were performed with restricted databases to enhance the sensitivity of the search. The restricted databases included all viruses in Viral RefSeq protein [36], 3 RNA virus databases (single-stranded RNA viruses, doublestranded RNA viruses and Drosophila C Virus), and 3 DNA virus databases (Baculoviruses, Nudiviruses, and Oryctes rhinoceros Nudivirus). All but the Viral RefSeq protein database were constructed de novo using NCBI's taxonomy browser. The Baculovirus, Nudivirus and Oryctes rhinoceros Nudivirus searches were added after identification of a putative Nudivirus in the initial searches (see below). Limiting the size of the database decreases the E-value of any particular match, increasing its significance, because the probability of a chance match decreases. For all searches, significant hits were characterized by parsing the BLAST output and accessing Genbank to identify genes and organisms for the hit. These scripts were written in PERL and utilized functions in BIOPERL [37]. Sequences with BLAST hits to Nudiviruses were deposited in Genbank, except those sequences shorter than 200 bp, which are presented in the online supplemental material (Material S1).

Survey of wild flies for DiNV infection
After discovery of a putative DNA virus (see results), several species of Drosophila were surveyed for infection with this virus. Flies were collected from Rochester, NY and the Southwest Research Station in 2009 and 2010. DNA from 7 Drosophila phalerata (4 females and 3 males) individuals collected in Munich, Germany was kindly provided by Kelly A. Dyer. In addition, D. innubila were collected in 2010 at the Southwest Research Station and immediately dissected and extracted on site to minimize possible horizontal transmission of the virus among flies. DNA was extracted using the Puregene DNA purification kit (QIAGEN, Valencia, CA). Flies were screened for the virus using standard insect COI primers (1490 and 2198) [38] as a control for extraction quality and newly developed primers (P47F: 59-TGAAACCA-GAATGACATATATAACGC and P47R: 59-TCGGTTTCTC-AATTAACTTGATAGC) for the P47 homolog found in the metagenomics search. For each species, the P47 locus from at least one individual was sequenced using BigDye Terminator v3.1 (#4337455, Applied Biosystems, Carlsbad, CA) and deposited in GenBank (Accession numbers JN44311-JN44330). These sequences were used to build a phylogenetic tree using PhyML [39] with the HKY85 model of substitution and 100 bootstrap replicates and Mr. Bayes [40] with the GTR+Gamma model with a chain length of 1,100,000 and a burn in of 100,000 generations. The P47 ortholog from OrNV was used as an outgroup.

Electron microscopy of virus particles
Because transmission of the DNA virus may be fecal-oral, as hypothesized for the closely related Oryctes rhinoceros Nudivirus [41], fly fecal material was scraped from the side of vials containing infected D. falleni and then PCR screened for the virus, using the methods described above. These samples were almost invariably positive for the virus, so fecal material was primarily used for imaging. To concentrate the virus on a microscope slide, a crowded vial of infected flies was inverted on a microscope slide and flies were allowed to defecate for 4 d. A small sample was scraped and PCR screened for the virus. An attempt was made to find virus particles in whole flies by dissecting out the digestive tract for imaging.
The glass slide with deposited feces was fixed in 0.1 M sodium cacodylate buffered 2.5% glutaraldehyde for 24 hours and postfixed in 1.0% buffered osmium tetroxide for 20 min. The slide was transitioned through a graded series of ethanol to 100% (63) and infiltrated with Spurr epoxy resin overnight. The next day, size 3 BEEM capsules were filled with fresh resin and inverted and placed onto the glass slide over the fecal matter. The slide was placed into a 600uC oven and allowed to polymerize overnight. The polymerized BEEM capsules were removed from the glass slide using the ''pop-off'' technique [42] which involves dipping several times into liquid nitrogen. The capsules containing the entrapped fecal material were trimmed with a razor blade to a small trapezoid and thin sectioned on a Reichert ultramicrotome using a diamond knife at 70 nm. The sections were placed onto 200 mesh copper grids and stained with aqueous uranyl acetate and lead citrate. The grids were examined using a Hitachi 7650 Transmission Electron Microscope and micrographs were captured using an attached Gatan Erlangshen 11 megapixel digital camera.

Fitness of infected flies
Wild-caught females were used to assess the survival and fecundity of flies as a function of infection with virus and Wolbachia. In September 2009, flies were captured near the Southwest Research Station as described above and transferred to sugar agar for transport to Rochester, NY, which took ,6 d. In Rochester, flies were placed on mushroom food (instant drosophila medium plus a piece of commercial Agaricus bisporus mushroom and a cotton roll) for egg laying. Females were then moved to new vials every other day until they died. The experiment lasted for 55 days, at which time no flies were laying fertilized eggs. Females were screened for the DNA virus and their offspring were counted as they emerged. In July and August 2010, the same protocol was followed for D. falleni. In this case, flies were collected around Rochester, NY and were established in culture the day they were collected, but the experiment was terminated after 10 days. As described above, flies from the 2010 D. innubila collection were dissected on site and mature eggs (stage 10 or later) in both ovaries were counted to assess fitness costs associated with the virus. All flies were PCR screened for the DNA virus and, as a control, insect COI.

Experimental infection of lab-reared flies
To directly assess the fitness consequences of viral infection, flies were injected with live virus and survival was monitored. Since cell culture for this virus has not yet been established, live virus was isolated as follows. Several wild D. innubila females were homogenized individually in viral buffer (see above) and centrifuged at low speed for one minute to remove large fly debris. The supernatant was then spun through a 0.45 mm filter (UFC30HV25, Millipore USA, Billerica, MA). 10 ml of the filtrate for each sample was used to PCR screen for the virus, using the DNA extraction and PCR protocols as described above. The filtrate from 4 flies identified by PCR as positive for DiNV infection were pooled as the virus-positive, and 4 flies screening negative for DiNV were pooled as the virus-negative control. Each sample was diluted 1:10 before being used for injection. A separate control, sterile virus buffer was also employed. Six-to eight-dayold male and female D. innubila and D. falleni were injected with about 200 nL of one of the three treatments using a Narishige IM300 microinjector (Narishige, Japan). Fly survival was monitored daily and flies were kept at low density (maximum = 10 per vial) and transferred to new food every other day until most flies injected with the virus-positive filtrate were dead. Both the viruspositive and virus-negative samples were examined using electron microscopy to be sure a) only DiNV was present in the viruspositive samples and b) no bacteria were present in either sample.

Vertical transmission
For a subset of wild-caught D. innubila mothers found to be infected with the DNA virus, offspring were PCR screened to assess vertical transmission of the virus. Offspring were frozen 2-6 d after emergence, and DNA was later extracted and screened as described above. A total of 27 daughters and 2 sons of were screened from 10 virus-infected females. The low offspring production of virus-infected females limited the sample size and male-killing by Wolbachia resulted in the highly skewed sex-ratio.

Viral metagenomics
A total of 1225 raw reads and 25 contigs (using 1154 raw reads) were recovered for the Wolbachia uninfected sample and 44,675 raw reads and 124 contigs (using 5939 raw reads) were recovered for the Wolbachia infected sample. Raw reads averaged 320.3 bp for the uninfected sample and 394.9 bp for the infected sample. Table 1 shows the types of hits for each search and sample. An E-value cutoff of 0.0001 was used for all searches. Hits to bacterial and eukaryotic sequences were overwhelmingly to ribosomal RNA genes (data not shown), which could be due to the high ratio of ribosomal RNA transcripts to gene-specific mRNA transcripts and/or because the ribosome somehow protects ribosomal RNA from RNase digestion.
The Wolbachia-infected sample was spiked with 1 fly that had been injected with Drosophila C virus (DCV). Two-thirds (26,688) of reads with significant hits in the infected group had DCV as the best hit, as opposed to less than 1% (9 reads) in the uninfected group, demonstrating both the sensitivity and specificity of the method. Figure S2 shows the distribution of hits across the DCV genome and simple calculations show that the average coverage is more than 12006.
All virus hits are summarized in Table 2. There were 27 unique virus hits, including 9 bacteriophage, and 12 double-stranded DNA, 3 single-stranded RNA, and 3 double-stranded RNA viruses. Sixteen virus families were represented, whose normal hosts include a variety of organisms.
Three viruses in the list stand out. First was DCV, which was expected to be present in large quantities. The second most common virus hit was to Cricket paralysis virus (CrPV), which is closely related to DCV and therefore may represent poor quality reads that were actually DCV. However, CrPV has a broad host range and can infect Drosophila melanogaster [43,44], so these reads may represent a real RNA virus, closely related to CrPV, that infects Drosophila innubila.
Finally, 29 reads (26 and 3 from the Wolbachia-uninfected and Wolbachia-infected samples, respectively) had Oryctes rhinoceros Nudivirus (OrNV), a double-stranded DNA virus, as the best hit. Narrowing the search database to just viruses did not increase the number of sequences hitting OrNV, but searches restricted to Nudiviruses increased this to 31 and 7 (see Table 3) from the Wolbachia-uninfected and infected samples, respectively, and searching against OrNV itself increased it to 30 and 3.
OrNV is a member of the small and unclassified Nudivirus group [45,46]. Nudiviruses have large genomes -OrNV is almost 130 kb -and are characterized by a rod-shaped virion. The search against all Nudivirus proteins found 16 unique Nudivirus proteins among the raw reads (Table 3). For 13 of these, the best hit was OrNV, representing about 10% of the 139 predicted proteins [41] for OrNV ( Figure S3). A total of 9540 bases hit OrNV yielding a genomic coverage of about 0.076. Two more sequences hit the closely related dsDNA virus family Baculoviridae and had marginally significant hits to OrNV when searched against only the OrNV genome. In accordance with proposed Nudivirus nomenclature [47], the above virus will be referred to as Drosophila innubila Nudivirus (DiNV). Table 3 shows the percent identity from BLAST hits between each DiNV sequence and the other Nudiviruses yielding matches. Because both sequence and gene conservation is so low in the Nudiviruses, a phylogenetic analysis of DiNV within the Nudiviruses is not possible with the current data. Based on the data in Table 3, it appears that DiNV is most closely related to OrNV. The remainder of the paper will focus on DiNV.

Survey of wild flies
DiNV virus is present in several Drosophila species, including members of 3 subgenera, from both northeastern and southwestern United States, but is not found in a small sample of D. phalerata from Europe (Table 4). The prevalence of infection varies among species and is most common within members of the quinaria group. If all species are included, males (36%) were infected significantly more often than females (25%; P = 0.005, FET). Within the 2010 collection of D. innubila, males were also significantly more likely to be infected than females (P = 0.0037, FET). A phylogenetic tree of P47 sequences from each species is presented in Figure S3. All isolates from the Arizona collection had identical sequences, while there was some variation in sequences from New York. The sequences from a Drosophila simulans male and a Drosophila melanogaster or simulans female (both from the subgenus Sophophora) were identical, but were nested within sequences from other members of the subgenus Drosophila.

Electron microscopy
The fecal material contained numerous virus particles morphologically similar to Baculoviruses and Nudiviruses (Figure 1). The capsid is approximately 120630 nm, with an envelope ,135 nm in diameter, making it small among the Nudiviruses [47]. No virus particles were observed from the digestive tracts of flies.    Figure 2B).

Experimental infection of lab-reared flies
After injection, fly survival was monitored daily until most flies injected with the virus-positive filtrate had died (after 33 days the experiment was stopped, at which point 3 male D. falleni injected with the virus were still alive). To assess survival after injection  There was a moderately significant interaction (P = 0.06) between sex and treatment with males injected with the viruspositive filtrate surviving several days longer than females (see Figure 3). Overall, D. falleni experienced higher mortality (P = 0.005) than D. innubila regardless of treatment or sex.

Vertical transmission
Of the 27 female offspring screened for infection only 1 was positive, while neither of the 2 male offspring was positive, for an overall ''vertical'' transmission rate of 0.034. Note that the single positive offspring may have contracted the virus periorally and therefore, effectively horizontally, making this an upper estimate of vertical transmission.

Discussion
Using viral metagenomics a putative Nudivirus was discovered to infect several species of Drosophila. The virus was termed Drosophila innubila Nudivirus. Though the host specificity of the virus is not yet known, the species name was retained for clarity. This may need to be revised later. The abbreviation of Drosophila innubila Nudivirus is problematic because conventions and practical concerns in the Drosophila community are at odds with those in the Nudivirus community. For example, with more than 2000 species of Drosophila, species abbreviations usually employ the first three letters of the species name, Dinn for Drosophila innubila. In addition to the Nudivirus described here, Drosophila is host to Nora Viruses and Noda Viruses, so using NV could be confusing. For those working on Drosophila, DinnNuV might be most informative, although quite cumbersome. Following Nudivirus conventions, the abbreviation would be DiNV. For the current manuscript, the simpler DiNV will be used. While several RNA viruses are well-characterized from D. melanogaster and its close relatives [27,28], DiNV is the first report of a DNA virus in Drosophila. As Drosophila are an important model for the study of the molecular biology and evolution of immunity, this discovery broadens the scope of host-pathogen interactions that can be studied in the genus.

Drosophila innubila Nudivirus
DiNV is similar in sequence to the double-stranded DNA viruses of the Nudivirus group, being most closely related to OrNV, which infects rhinoceros beetles. Supporting the conclusion that the virus discovered in D. innubila is a Nudivirus, electron microscopy revealed viral particles in the feces of D. innubila similar in fine structure to other described Nudiviruses.
DiNV is associated with greatly reduced survival and offspring production among wild-caught individuals of D. innubila and with greatly reduced offspring production in D. falleni. While this association does not prove that these viruses cause the reduced fitness, the data strongly suggest that DiNV is a highly pathogenic infection. Furthermore, the prevalence of DiNV infection in natural populations of D. innubila from the Chiricahua Mountains of Arizona was consistently high, around 40%, in 2 successive years, suggesting that this virus may cause a major reduction in mean absolute fitness within this species. The prevalence of infection was similarly high in other members of the quinaria group -D. munda and D. tenebrosa -in collections from the Chiricahua Mountains. The prevalence of infection within D. falleni, a quinaria group species common in the eastern North America, was much lower, around 3%. The frequency of infection in species outside the quinaria group was more sporadic. Darren Obbard has found a similar virus in the melanogaster group, but his nucleotide sequences are about 25% diverged from those found in this study (personal communication).
Microinjection of DiNV-positive filtrate into lab-reared flies further suggests that the virus is highly pathogenic. Most flies injected with the virus died within two weeks of injection. In both species, males survived longer than females, and in both sexes, D. falleni survived longer than D. innubila. The difference between species may be due to size differences in flies (D. falleni used in the experiment were larger than D. innubila and may therefore require a higher virus titer for the same pathogenic effect), or could reflect selection for increased virulence of the virus to its natural host, since the virus injected was from D. innubila. The difference between the sexes cannot be explained by size since males are smaller and survived longer. Interestingly, overall and in the 2010 D. innubila collection, males had higher rates of infection than females (see Table 4). This is consistent with the lower mortality in males observed in the lab.
Vertical transmission of DiNV is unlikely to be important in the population, since the rate of transmission from mothers to offspring in the laboratory culture was ,5% in D. innubila. The single instance of mother-offspring transmission may have mediated via a fecal-oral route, as this is the predominant mode of transmission in OrNV [41]. Further supporting fecal-oral transmission, virus genes were PCR-amplified and virus particles were found using electron microscopy in fly fecal material. Thus, viral infections in natural populations may result from horizontal transmission among adult flies and their offspring at their mushroom feeding and breeding sites.
Phylogenetic analysis of the partial viral P47 shows that the DiNV infecting D. innubila and D. falleni form closely related but genetically distinct clades. Thus, DiNV is a geographically widespread, prevalent, and pathogenic DNA virus for which members of the Drosophila quinaria group appear to be particularly important hosts.

A new protocol for virus discovery
One goal of the metagenomics survey was to show that the protocol could detect virus in a single fly. In a 40-fly sample spiked with a single fly infected with DCV, 2/3 of all reads had DCV as a best hit, demonstrating a high level of sensitivity of the protocol. The almost complete absence of such hits in the sample not spiked with DCV attests to the specificity of the method. In addition to DCV, our screen uncovered several putative viral sequences. The most abundant of these were assigned to DiNV.
Given the success in recovering DCV from the spiked sample, why weren't more viruses found? Most sequences found some hits, so although there could be some virus sequences in our dataset with no homologs in the Refseq database, most sequences were readily assignable. There are at least 3 other possibilities. First, viral capsids may vary in their ability to protect viral RNA from degradation by RNAse, allowing some viruses to go undetected. Second, some viruses may be rare in D. innubila. Given that Wolbachia may provide some protection against RNA virus infection, the prevalence of RNA virus infection may be driven down by Wolbachia in D innubila, making them harder to detect. Finally, rare and virulent viruses may not have been present in the sample of 80 flies used, since infection frequency and virulence are usually assumed to be negatively correlated.
Most well-studied viruses of Drosophila have minor fitness effects in flies that either inherit the virus vertically or contract it through feeding, but greater effects when flies are injected with the virus [28]. This lack of virulence is perhaps because those viruses that are well-studied were discovered in cell culture or laboratory stocks and are therefore by nature less virulent since stocks and cell lines with very virulent viruses would not last long. This ascertainment bias is lessened in surveys of natural populations since captured flies could be quite sick. While we do not know the natural route of infection for any Drosophila viruses (although some are at least partly vertically transmitted), it is probably safe to conclude that DiNV is not vertically transmitted and infection frequencies are high enough that infection via mites (which are found at relatively low frequency on D. innubila), the natural analog of microinjection, is unlikely. Therefore, DiNV appears to be a virus exhibiting high virulence without requiring a rather drastic injection to show such effects.
The discovery of a DNA virus that naturally infects Drosophila opens the way for study of host immune response to DNA virus infection in an easily cultured species. The system also lends itself to studies of host-pathogen coevolution between geographically isolated sister species and between semi-isolated Sky Island populations of D. innubila.

Supporting Information
Material S1 Sequences for short DiNV orthologs of other Nudiviruses. Note these are too short (,200bp)