Abstract
Interactions between the insect immune system and RNA viruses have been best studied in Drosophila, where RNA interference, NF-kB and JAK-STAT pathways underlie antiviral immunity. In response to these immune mechanisms, insect viruses have convergently evolved suppressors of RNA interference that act by diverse mechanisms to permit viral replication. However, interactions between the insect immune system and DNA viruses have received less attention, primarily because few Drosophila-infecting DNA virus isolates are available. Here, we use a recently-isolated DNA virus of Drosophila melanogaster, Kallithea virus, to probe known antiviral immune responses and virus evasion tactics in the context of DNA virus infection. We find that fly mutants for RNA interference and Immune deficiency (Imd), but not Toll, pathways are more susceptible to Kallithea virus infection. We identify the Kallithea virus-encoded protein gp83 as a potent inhibitor of Toll signalling, strongly suggesting that Toll mediates antiviral responses during Kallithea virus infection, but that it is suppressed by the virus. Further, we find that Kallithea gp83 inhibits Toll signalling either through NF-kB transcription factor regulation, or transcriptionally. Together, these results provide a broad description of known antiviral pathways in the context of DNA virus infection and identify the first Toll pathway inhibitor in a Drosophila virus, extending the known diversity of insect virus-encoded immune inhibitors.
Introduction
Innate antiviral immunity in insects has been best studied in response to RNA virus infections of Drosophila melanogaster. Antiviral immune mechanisms that target RNA viruses include RNA-mediated defences such as RNA interference (RNAi) and RNA decay pathways, cellular defences such as apoptosis, phagocytosis, and autophagy, and transcriptional responses. The latter are primarily mediated by Janus kinase/signal transducers and activators of transcription (JAK-STAT) and Nuclear factor kB (NF-kB) pathways (reviewed in Merkling and van Rij, 2013; Bronkhorst and van Rij, 2014; Lamiable and Imler, 2014; Xu and Cherry, 2014; Palmer, Varghese and van Rij, 2018).
The insect response to DNA viruses is less well studied, but RNAi and apoptosis have demonstrated antiviral activity (Clem, 2001; Bronkhorst et al., 2012; Kemp et al., 2013) and the JAK-STAT pathway is active during infection, possibly mediating a tolerance response (West and Silverman, 2018). Baculovirus, nudivirus, and iridovirus infections of Drosophila all give rise to virus-derived small interfering RNA (vsiRNAs), which regulate DNA virus gene expression (Bronkhorst et al., 2012; Jayachandran, Hussain and Asgari, 2012; Kemp et al., 2013; Webster et al., 2015) and mutants for RNAi effectors Dicer-2 (Dcr-2) and Argonaute-2 (AGO2) are hypersensitive to Invertebrate iridescent virus 6 (IIV6; an iridovirus) infection. This suggests that RNAi is also an important defence against DNA viruses, and IIV6 correspondingly encodes a suppressor of RNAi (Bronkhorst et al., 2012, 2014). Virus-encoded suppressors of apoptosis are also widespread in DNA viruses, acting through binding and inhibition of cellular caspases (e.g. p35), or stabilization of cellular inhibitors of apoptosis (e.g. IAP gene family; Bump et al., 1995; Xue and Robert Horvitz, 1995; Byers, Vandergaast and Friesen, 2016). In contrast, the contribution of transcriptional responses, such as the NF-kB pathways, to DNA viruses has not yet been elucidated.
There are two NF-kB pathways in Drosophila: Toll and Imd, which primarily function in antibacterial (Toll: gram-positive, Imd: gram-negative) and antifungal (Toll) defense, although both provide protection against some RNA viruses (reviewed in Valanne, Wang and Ramet, 2011; Merkling and van Rij, 2013; Lamiable and Imler, 2014; Myllymaki, Valanne and Ramet, 2014; Palmer, Varghese and van Rij, 2018). Toll and Imd pathways are activated following recognition of a pathogen-associated molecular pattern (PAMP; e.g. bacterial peptidoglycan), leading to the phosphorylation and degradation of the inhibitor of kappa B (IkB; encoded by cactus for Toll signalling, and by the relish C-terminus in Imd signalling) (reviewed in Valanne, Wang and Ramet, 2011; Myllymaki, Valanne and Ramet, 2014). Under non-signalling conditions, IkB sequesters NF-kB transcription factors in the cytoplasm. These transcription factors are encoded by dorsal (dl) and Dorsal immune-related factor (Dif) in Toll signalling, and relish (rel) in Imd signalling, and all translocate to the nucleus to induce gene expression following ΙκΒ degradation (reviewed in Valanne, Wang and Ramet, 2011; Myllymaki, Valanne and Ramet, 2014). Although the mechanism by which Toll and Imd recognise RNA viruses is unclear, both are active and provide immunity against some viral infections in insects, most likely through induction of antiviral effector responses. For example, Toll is broadly antiviral against RNA viruses such as Drosophila C virus, Nora virus, and Flock House virus in Drosophila during orally acquired, but not systemic infections, and in Aedes mosquitoes against dengue virus (Zambon et al., 2005; Xi, Ramirez and Dimopoulos, 2008; Ramirez and Dimopoulos, 2010; Ferreira et al., 2014). Additionally, Imd is antiviral against a subset of viruses in Drosophila, such as Cricket Paralysis virus, Drosophila C virus, and Sindbis virus and in Aedes cell culture against the alphaviruses Semliki Forest virus and O’nyong’nyong virus (Fragkoudis et al., 2008; Avadhanula et al., 2009; Costa et al., 2009; Waldock, Olson and Christophides, 2012; Sansone et al., 2015).
Although the effect of NF-kB signalling on DNA virus infection in insects has not been directly tested, polydnaviruses, ascoviruses, baculoviruses, and entomopoxviruses have acquired suppressors of NF-kB signalling by horizontal gene transfer, providing indirect evidence for anti-DNA virus NF-kB signalling (Thoetkiattikul, Beck and Strand, 2005; Lamiable et al., 2016). First, a ‘polydnavirus’ encoded in the genome of the Braconid parasitoid wasp Microplitis demolitor has acquired homologs of ΙκΒ, some of which inhibit Dif and rel by direct binding (Thoetkiattikul, Beck and Strand, 2005). However, this is a domesticated endogenous viral element that forms a component of the wasp venom, and as these ΙκΒ homologues are not found in related nudiviruses, baculoviruses, or hytrosaviruses, it seems likely they were acquired to inhibit anti-parasitoid immune responses in the host of the parasitoid wasp, rather than the antiviral immune response of the wasp itself (Bitra, Suderman and Strand, 2012; Herniou et al., 2013). Second, homologs of diedel, which encode a cytokine that inhibits the Imd pathway in Drosophila, are similarly found in ascoviruses, baculoviruses, entomopoxviruses, and Leptopilina spp. polydnavirus venom, likely through independent horizontal transfer from arthropod hosts (Lamiable et al., 2016). Virus-encoded diedel phenocopies fly-encoded diedel, suggesting that viral diedel has retained an Imd-suppressive function, and that the Imd pathway likely interacts with these DNA viruses (Lamiable et al., 2016). However, it is still unclear whether antiviral Toll signalling is targeted by insect virus-encoded immune suppressors, and whether these hijacked host pathway inhibitors represent a subset of a greater diversity of NF-kB immune inhibitors or reflect evasion of virus-specific immune mechanisms.
The recent isolation of Kallithea virus (KV; Webster et al., 2015; Palmer et al., 2018), a nudivirus that naturally infects Drosophila melanogaster at high prevalence in the wild, provides a tractable system to study host-DNA virus interactions and to identify immune evasion strategies in DNA viruses. Moreover, because some virus-encoded immune suppressors have been found to be highly host-specific, the use of native host-virus pairs is vital to our understanding of viral immune evasion (e.g. Parisien, Lau and Horvath, 2002; Mariani et al., 2003; Goffinet et al., 2009; Rajsbaum et al., 2012; van Mierlo et al., 2014; Stabell et al., 2018). Here, we use this system to describe the interaction between antiviral immune pathways and a DNA virus in Drosophila. Using mutant fly lines, we find that the RNAi and Imd pathways mediate antiviral protection against KV in vivo, but that abrogation of Toll signalling has no effect on virus replication. Through re-analysis of previous RNA-sequencing data, we observe a broad downregulation of NF-kB responsive antimicrobial peptides following infection, and perform a small-scale screen for KV-encoded immune inhibitors. We identify viral protein gp83 as having a complex interaction with NF-kB signalling, leading to induction of Imd signalling but potent suppression of Toll signalling. This suppression acts directly through, or downstream of, NF-kB transcription factors. Finally, through deletions of conserved protein regions and analysis of the related Drosophila innubila nudivirus (DiNV) gp83 ortholog, we show that the immunosuppressive activity of gp83 against D. melanogaster NF-kB signalling is conserved.
Materials and Methods
Fly strains, virus growth, and mortality experiments
All fly lines were maintained and crossed on standard cornmeal medium at 25 °C. Viral titre and mortality were measured following KV infection in two control lines (w1118 and Oregon R) and in mutant lines compromised in the following immune signalling pathways: RNAi (Dcr-2L811fsx (Lee et al.,2004) and AGO2414 Okamura et al., 2004)), Toll (spz4 (Weber et al., 2007), dl1 (Nusslein-Volhard, 1979), Dif1 (Rutschmann et al., 2000), and pll2/pll21 trans-heterzygotes (Anderson and Nüsslein-Volhard, 1984; Hecht and Anderson, 1993)), and Imd (rele20 (Hedengren et al., 1999) and imd10191 (Pham et al., 2007)).
For mortality assays, 100 female flies of each genotype were injected with 50 nL of either KV suspension (105 ID50, as described in (Palmer et al., 2018)) or chloroform-treated KV suspension (which inactivates KV through the destruction of the membrane) and transferred in groups of 10 to sucrose agar vials. The number of surviving flies was recorded on alternate days, and each group of flies was transferred to fresh food each week. Per-day mortality was analysed as a binomial response variable with the Bayesian generalised linear mixed modelling R package, MCMCglmm (Hadfield, 2010), with days post-inoculation (dpi), dpi2 (to allow for non-linear changes in mortality), and genotype as fixed effects, and vial as a random effect, as described previously (Palmer et al., 2018). All confidence intervals are reported as 95% highest posterior density (HPD) intervals.
Viral titre was measured in each line after intra-abdominal injection of 50 nL of KV suspension. Infected female flies of each line (n=50) were transferred to 10 sucrose agar vials in groups of 5, and 5 vials of each genotype were homogenised in Trizol at 5 and 10 dpi. For RNAi mutants, flies were also assayed at 3 dpi. DNA was extracted by phenol-chloroform precipitation and viral titre estimated by quantitative PCR relative to host genomic DNA, using previously described primers (rpl32; Palmer et al., 2018)). Log-transformed viral titre was analysed as a Gaussian response variable using MCMCglmm (Hadfield, 2010), with genotype, dpi, and genotype-by-dpi interactions as fixed effects. Titre in RNAi and NF-kB mutants were assayed in separate experiments, and therefore analysed independently. We took a statistical approach to account for the impact of differing genetic backgrounds between mutant lines, using the range of KV titres seen previously across 120 different natural genetic backgrounds from the Drosophila Genetic Reference Panel (Palmer et al., 2018). Specifically, considering w1118 and Oregon R as controls and mutants of each pathway as the ‘experimental’ group, a null distribution of effect sizes expected only from differences in genetic background was created by randomly choosing two DGRP lines to serve as controls and additional DGRP lines reflecting the mutant lines used in each pathway. For each null draw, the same model was fitted as described above, the absolute value of the effect size was recorded, and this was repeated 1000 times to obtain a distribution. If the average effect size associated with mutants in a pathway was greater than the highest 5% of effect sizes, we concluded that the observed differences in KV titre were due to mutations in the tested pathway.
Cell culture and virus propagation
S2 cells were cultured at 25 °C in Schneider’s Drosophila Medium with 10% heat-inactivated fetal bovine serum and 50 U/mL penicillin and 50 ug/mL streptomycin (Life technologies). KV was purified from flies 10 days after initial infection as previously described (Palmer et al., 2018).
Cloning
We selected 10 KV genes identified as highly expressed at three dpi (Palmer et al., 2018) to screen for KV-encoded immune suppressors. These were gp23, gp43, gp72, gp83, ACH96233.1-like, ACH96143.1-like, putative protein 1, putative protein 12, putative protein 15, putative serine protease (corresponding to GenBank accession numbers AKH40365.1, AKH40394.1, AKH40346.1, AKH40369.1, AKH40392.1, AKH40340.1, AQN78560.1, AKH40392.1, AKH40404.1, and AQN78556.1). Each KV gene was amplified using the Qiagen Long Range PCR kit as per the manufacturer’s instructions, with primers that introduced restriction sites and the Drosophila Kozak sequence (restriction enzymes and primers used in Supplementary Table 1), and cloned into a pAc5.1 vector (Invitrogen) with a V5-His C-terminal tag. The KV gene gp83 was also cloned into pAc5.1 vector with GFP instead of V5-His to introduce a C-terminal GFP tag. Deletion constructs for gp83 were created by separately amplifying 2 segments of gp83 with primers that span the desired deletion and performing a second PCR reaction with these segments as a template, and the forward and reverse primers from the 5’ and 3’ segments, respectively (Supplementary Table 1; gp83Δ1: CGLIECSELLRDRLCSKL deletion; gp83Δ2: WSDRLNLI deletion). The resulting amplicons with deletions were cloned as described above. The gp83 gene from DiNV (Unckless, 2011; Hill and Unckless, 2017) was also cloned as above (Supplementary Table 1).
Additionally, Toll pathway components pll, tube, cact, Dif, and dl were cloned into the pAc5.1 vector, as described above (Supplementary Table 1). Other Toll and Imd pathway constructs have been described before, including pAc5.1-Tolllrr (Tauszig et al., 2000), pAc5.1-dl-GFP (Li and Dijkers, 2015), pMT-PGRP-LCx (Kaneko et al., 2006), pAc5.1-rel-GFP (Foley and O’Farrell, 2004), and the firefly luciferase (FLuc) reporter plasmids with promoter sequences from Drosomycin (Drs), Diptericin (Dpt), and Attacin-A (Att-A) (Tauszig et al., 2000) or with 10X STAT binding sites (Baeg, Zhou and Perrimon, 2005).
Transfection and RNAi Knockdown in S2 cells
S2 cells were transfected using Effectene transfection reagent, as per the manufacturer’s instructions. Double-stranded RNA (dsRNA) was synthesized against cactus, gp83, FLuc, renilla luciferase (RLuc), and GFP to knockdown these genes in S2 cells. Primers with flanking T7 sequences were used to amplify regions of each gene (Supplementary Table 1) and dsRNA was synthesized from the resulting PCR products with T7 RNA polymerase and purified using GenElute Total RNA mini kit (Qiagen) (e.g. van Cleef et al., 2011).
Immune suppression assays
The 10 cloned KV genes were tested for their ability to suppress RNAi, JAK-STAT, Toll, or Imd activity. RNAi suppression assays were performed as described previously (van Cleef et al., 2011). Briefly, 5×104 S2 cells were seeded in a 96-well plate and 24 hours later transfected with 33 ng of pMT-FLuc, 33 ng pMT-Rluc, and 33 ng of either pAc5.1 empty vector or the pAc5.1 expression plasmid encoding a KV gene. Two days later, 400 ng of either GFP or GL3 dsRNA was added to each well, and CuSO4 was added 8 hours later to a final concentration of 500 μM to induce expression of the luciferase reporters. RLuc and FLuc luciferase activity were measured using the Dual Luciferase Assay Kit (Promega).
For JAK-STAT immunosuppression assays, 5×104 S2 cells were seeded in a 96-well plate and transfected 24 hours later with 30 ng of 10XSTAT-FLuc, 20 ng pAc5.1-Rluc, and 50 ng of either pAc5.1 empty vector or the pAc5.1 expression plasmid encoding a KV gene. Luciferase activity was measured at 48 hours following transfection.
For NF-kB immunosuppression assays, a plasmid encoding the Imd receptor PGRP-LC (isoform x; pMT-PGRP-LCx) (Werner et al., 2003; Kaneko et al., 2006) or a constitutively active Toll construct lacking the extracellular leucine-rich repeat domain, pAc5.1-Tolllrr (Tauszig et al., 2000) was transfected alongside each KV gene, and a NF-kB-responsive FLuc reporter containing the either the Dpt (Imd) or Drs (Toll) promoter sequence (Tauszig et al., 2000). For Toll immune suppression assays, 5×104 S2 cells were seeded in 96-well plates and 24 hours later transfected with 50 ng of either empty pAc5.1 vector or a pAc5.1 KV gene expression construct, 20 ng of either pAc5.1 or pAc5.1-Tolllrr, 10 ng of Drs-FLuc, and 10 ng pAc5.1-Rluc. Imd immune suppression assays were performed in the same manner, except that pMT, pMT-PGRP-LCx, and Dpt-FLuc were substituted for pAc5.1, pAc5.1-Tolllrr, and Drs-FLuc, respectively, and CuSO4 was added immediately following transfection. Analogous experiments were performed using pAc5.1-dl, pAc5.1-Dif, and pAc5.1-pll instead of pAc5.1-Tolllrr, or by transfecting 5 ng of cact dsRNA. In the latter case, 70 ng of KV gene expression construct was transfected instead of 50 ng. RLuc and FLuc activity were assayed 48 hours after transfection.
Immunosuppression assays were also performed using KV-infected cells. 5×104 cells were seeded in 96-well plates, followed by the immediate addition of 5 μL of either KV suspension (103 ID50) or chloroform-treated KV, and transfected the next day. For RNAi suppression assays with KV, 50 ng pMT-RLuc, 50 ng pMT-FLuc (van Cleef et al., 2011), and 5 ng of either GFP or GL3 dsRNA were transfected 2 dpi and CuSO4 added 8 hours later. To measure JAK-STAT activity following KV infection, 70 ng of 10XSTAT-FLuc and 30 ng pAc5.1-Rluc (Van Cleef et al., 2014) were transfected. For Toll suppression assays, 70 ng of either pAc5.1 or pAc5.1-Tolllrr, 20 ng of Drs-FLuc, and 10 ng pAc-RLuc were transfected. Finally, to measure Imd activity following KV infection, 70 ng of either pMT or pMT-PGRP-LCx, 20 ng of Dpt-FLuc, and 10 ng pAc-RLuc were transfected, and CuSO4 was added immediately following transfection. Luciferase activity was measured at 4 dpi.
The R package MCMglmm was used to determine significance in immune suppression assays, with the RLuc-normalised FLuc values as a Gaussian response variable. In the original screen for immune suppressors, any experimental induction of an immune pathway was treated as a fixed effect (e.g. addition of dsRNA against FLuc in the RNAi suppression assay, PGRP-LC overexpression in the Imd suppression assay, and Tolllrr in the Toll suppression assay), each KV gene was treated as a random effect, and the interaction between KV gene and the induced experimental change to signalling output was treated as a random effect. In subsequent NF-kB suppression experiments, where the only tested KV gene was gp83, gp83 and the interaction between gp83 and overexpression of NF-kB receptors were treated as fixed effects. Likewise, when immune suppression experiments were carried out with KV-infected cells instead of over-expressing KV genes, KV infection status, the induction of an immune pathway, and the interaction between these were treated as fixed effects.
Immunoprecipitation and western blotting
To test whether gp83 directly interacted with dl, 2×106 cells were seeded in 6-well plates and transfected with 150 ng of either pAc5.1 empty vector, pAc5.1 encoding V5-tagged gp83, or V5-tagged cact alongside 150 ng of the expression plasmid (pAc5.1) encoding GFP or GFP-tagged dl. Two days post-transfection, two wells per treatment were resuspended in lysis buffer (0.1% NP-40, 30 mM Hepes-KOH, 150 mM NaCl, 2mM MgOAc) supplemented with cOmplete protease inhibitor cocktail (Roche) and 5 mM DTT, and disrupted 30 times through a 25-gauge needle. After 10 minutes incubation on ice, cell debris was pelleted by centrifuging at 16,000xg for 30 minutes and supernatant was either stored as an input control or collected and incubated for 5 hours at 4 °C with magnetic control beads. Binding control beads were removed and the resulting supernatant was incubated with GFP-trap magnetic beads (Chromotek) overnight at 4 °C. Beads were washed 3 times in lysis buffer and 3 times in 25 mM Tris-HCl, 150 mM NaCl solution, and protein complexes eluted by boiling 10 minutes at 95 °C in Laemmli buffer.
Whole cellular protein extracts were prepared by heating S2 cells for 10 min at 95 °C in Laemmli buffer. Whole cellular extracts or immunoprecipitated proteins were separated on a 12% SDS-PAGE gel and transferred to a nitrocellulose membrane. Non-specific binding was blocked with blocking solution (phosphate buffered saline with 0.1% Triton-X (PBT) and 5% dry milk). Proteins of interest were probed with primary antibody diluted in blocking solution overnight at 4 °C, and visualized with an hour incubation of secondary antibody in blocking solution. Membranes were washed 3 times in PBT before and after each step. Antibodies that were used include mouse anti-dl (1:100 dilution, Developmental Studies Hybridoma Bank), mouse anti-V5 (1:1000 dilution, Invitrogen), rat anti-tub-α (1:1000 dilution, SanBio), and rabbit anti-GFP (1:1500 dilution, abcam ab6556) as primary antibodies, and goat anti-mouse IR-Dye 680 (1:15,000 dilution, LI-COR), goat anti-rat IR-Dye 800 (1:15,000 dilution, LI-COR), goat anti-rabbit IR-Dye 800 (1:15,000, LI-COR). An Odyssey Infrared Imager (LI-COR) was used to image blots.
Mass spectrometry
S2 cells were seeded (106) and co-transfected with pCoBLAST and pAc5.1-gp83gfp plasmid at a 1:19 ratio (125 ng and 2.38 μg, respectively). Medium was replaced 3 hours post-transfection, and again at 48 hours post-transfection with medium supplemented with blasticidin (20 μg/mL). Another 48 hours later, cells were refreshed with medium containing 4 μg/mL blasticidin, which was thereafter replaced every 3-4 days with medium containing 4 μg/mL blasticidin, resulting in a polyclonal cell line.
For mass spectrometry, wild-type S2 cells or S2 cells stably expressing GP83gfp were lysed in 50mM Tris-HCl (pH 7.8), 150mM NaCl, 1% NP-40, 0.5mM DTT, 10% glycerol and protease inhibitor cocktail (Roche). Approximately 4 mg of protein lysate was subjected to GFP-affinity purification using 7.5 μL GFP-trap beads (Chromotek) for approximately 1.5 hours at 4 °C. Beads were washed twice in lysis buffer, twice in PBS containing 1% NP-40, and three times in PBS, followed by on-bead trypsin digestion as described previously (Smits et al., 2013). Afterwards, tryptic peptides were acidified and desalted using Stagetips, eluted, and brought onto an EASY-nLC 1000 Liquid Chromatograph (Thermo Scientific). Mass spectra were recorded on a QExactive mass spectrometer (Thermo Scientific) and MS and MS2 data were recorded using TOP10 data-dependent acquisition. Maxquant (v1.5.1.0) was used to analyse raw data, using recommended settings (Cox and Mann, 2008). LFQ, IBAQ, and match between runs were enabled. The peptides were mapped to Drosophila melanogaster proteins (UniProt June 2017) and contaminants and reverse hits were filtered with Perseus (v1.3.0.4) (Tyanova et al., 2016). Missing values were imputed, assuming a normal distribution, and significance determined by a t-test on log-transformed LFQ-values between wild-type and gp83-expressing S2 cells.
Immunofluorescence microscopy
5×105 S2 cells were seeded in 12-well plates with glass coverslips in each well. Cells were transfected with 100 ng of pAc5.1 or pAc5.1-gp83-V5 and 100 ng of pAc5.1-dl-GFP. Two days after transfection, cells were fixed with 4% paraformaldehyde, washed twice in PBS, once with PBT, and blocked with PBT with 10% goat serum. Cells were stained by incubation with mouse anti-V5 (1:400, Invitrogen) for one hour, followed by fluorophore-containing goat anti-mouse secondary antibody (1:400, AlexaFluor) with 10 ug/mL Hoechst for one hour. Finally, cells were washed twice in PBT and twice in PBS, mounted on slides with Fluoromount-G (eBiosciences), and imaged with an Olympus FluoView FV1000. Fluorescence was measured in whole cells, or separately in the cytoplasm and nuclei by outlining the region of interest in Fiji (Schindelin et al., 2012) to calculate the mean fluorescence.
Results and Discussion
RNAi and Imd pathways are antiviral against KV in vivo
The RNAi pathway provides antiviral activity against the DNA virus IIV6, and KV-derived vsiRNAs are produced upon infection of adult naturally-infected Drosophila (Bronkhorst et al., 2012, 2014; Webster et al., 2015). However, the contribution of Imd and Toll pathways to anti-DNA virus immunity have not been described. We used fly lines mutant for RNAi, Imd, and Toll pathway components to assess whether these pathways fulfil an antiviral function during KV infection. First, we infected mutants for RNAi genes Dcr-2 and AGO2 with KV, and measured viral titre and mortality following infection. Following KV infection, both Dcr-2 and AGO2 mutants exhibited significantly greater KV titres at 3 dpi, with KV titre 78-fold greater in Dcr-2 mutants (95% HPD intervals: 18-281 fold; MCMCp < 0.001) and 55-fold greater in AGO2 mutants (13-237 fold, MCMCp < 0.001; Figure 1). However, the increased KV replication in RNAi mutants was not sustained at later infection timepoints. At 5 dpi, Dcr-2 mutants did not have significantly different KV titre from the controls (MCMCp = 0.22), and KV had a slightly diminished advantage in AGO2 mutants (12-fold increase; 2.5-43 fold, MCMCp < 0.001; Figure 1). By 10 dpi, there was no significant difference between viral titre in control flies and either Dcr-2 mutants (MCMCp = 0.43) or AGO2 mutants (MCMCp = 0.7). Therefore, either the antiviral effect of RNAi is short-lived (for example, a viral suppressor of RNAi may eventually be expressed in vivo), other immune pathways take over as the dominant antiviral force, or KV negatively regulates its own replication or saturates a resource. Nevertheless, despite the similar titres during late infection, there was still a significant increase in KV-induced mortality in Dcr-2 and AGO2 mutants, where 70% of control flies were alive at 19 dpi, compared to 25% in Dcr-2 mutants (MCMCp < 0.001) and 38% in AGO2 mutants (Figure 1), possibly due to early host damage or increased expression of virulence factors throughout infection (e.g. Jayachandran, Hussain and Asgari, 2012). These results extend the antiviral role of the RNAi pathway to KV infection.
We next infected Imd and Toll pathway mutants with KV and assessed KV DNA levels by qPCR at 5 and 10 dpi. We found Imd pathway mutants had significantly greater viral titre as compared to two control lines, with imd mutants having 6-fold greater KV titre at 5 and 10 dpi (2.7-13.7 fold, MCMCp<0.001), and rel having 8-fold greater viral titre at 5 and 10 dpi (3.1-15.9 fold, MCMCp < 0.001; Figure 1). Because the Imd effect spans 5 and 10 dpi, and we have previously measured KV titre in 125 inbred lines of the Drosophila Genetic Reference Panel at 8 dpi (Mackay et al., 2012; Palmer et al., 2018), we attempted to account for genetic background by comparing the average effect of Imd mutants to the distribution of effects consistent with natural variation in the genetic background. This analysis indicated that the increased titre observed in Imd mutants is unlikely to be due to genetic background (p = 0.01), if the variation in the DGRP is representative of the variation expected between lab-maintained fly lines. We also infected flies mutant for the Toll pathway components spz, pll, Dif, and dl. Viral titre was unchanged in most Toll pathway mutants compared to controls, except in Dif mutants (MCMCp = 0.02; Figure 1), where some vials remained uninfected. Further, the pathway-level effect of Toll mutants was within the expected distribution of effects caused by differences in genetic background, even when Dif mutants were excluded (p = 0.28). We concluded that the Imd pathway is antiviral against KV, but that abrogation of Toll function has no effect on KV growth. This could indicate that Toll is not antiviral against this DNA virus, or that it is efficiently suppressed by virus infection. The latter is consistent with our observation that genes encoding antimicrobial peptides are generally downregulated in KV-infected flies compared to uninfected controls (Figure 1), and we therefore explored the capability of KV to suppress innate immune pathways in cell culture.
KV growth in cell culture
KV growth in D. melanogaster cell culture has not previously been described. We found variation in the ability of KV to infect five commonly-used cell lines, and KV grew well in S2, S2R+, and DL2 cells, but poorly in Kc167 and Dm-BG3-c2 cells (Figure S1). In S2 cells, which we used for further analyses, KV was released into the medium at substantial levels starting from 3 dpi (Figure S1). Therefore, in all subsequent experiments, we assayed cells at 4 dpi, assuming that a high proportion of cells would be infected at this timepoint. We did not observe any overt cytopathic effects of KV-infected cells within 14 days of infection. However, when KV-infected cells were passaged at 7 dpi, we observed larger (MCMCp < 0.001) and fewer (MCMCp < 0.001) cells, likely due to a decrease in cell proliferation (Figure S1).
KV inhibits JAK-STAT and Toll, and induces Imd signalling in cell culture
We used previously established luciferase reporter-based assays to describe the effect of KV infection on RNAi, JAK-STAT, Toll, and Imd pathways. To determine if KV suppresses RNAi, we measured the RNAi silencing efficiency of cells inoculated with KV or chloroform-inactivated KV (hereafter referred to as mock-treated) by co-transfecting expression plasmids encoding FLuc with either GFP dsRNA or FLuc dsRNA. In both mock and KV-treated cells, FLuc dsRNA caused a 95% reduction in FLuc activity compared with GFP dsRNA treated cells, indicating KV infection does not inhibit RNAi in cell culture (MCMCp = 0.9; Figure 2). Many viruses studied in Drosophila encode a suppressor of RNAi (e.g. Li, Li and Ding, 2002; Van Rij et al., 2006; Nayak et al., 2010; van Mierlo et al., 2012, 2014; Bronkhorst et al., 2014; Van Cleef et al., 2014), and therefore the absence of KV-induced RNAi suppression is somewhat surprising. It is possible that KV-RNAi interactions are different in the cell types that are naturally infected by KV, and that our inability to observe RNAi suppressive activity is a limitation of the cell culture model. Alternatively, if KV transmission does not occur until later stages of infection, there may be limited selective pressure to evade RNAi, as RNAi mutants and control flies have similar titres during late infection.
The JAK-STAT pathway has an antiviral role during Drosophila C virus infection (Dostert et al., 2005) and mediates tolerance to the DNA virus IIV6, evidenced by upregulation of vir-1 and the Turandot (Tot) genes (West and Silverman, 2018). However, previous transcriptional profiling did not identify differential expression of these genes following infection with KV (Palmer et al., 2018). We assessed JAK-STAT activity in mock and KV-treated cells with a FLuc reporter containing ten STAT binding sites (Baeg, Zhou and Perrimon, 2005). This reporter is endogenously active in S2 cells (Baeg, Zhou and Perrimon, 2005), but KV infection led to a 58% reduction in STAT-mediated FLuc activity (37-74%, MCMCp < 0.001; Figure 2), indicating that JAK-STAT is down-regulated or inhibited following KV infection. However, in addition to mediating a transcriptional immune response, the JAK-STAT pathway is involved in cell proliferation (Arbouzova and Zeidler, 2006), which also decreases following KV infection in cell culture (Figure S1), making cause and effect difficult to distinguish.
We also assayed the effect of KV on Toll and Imd signalling. However, these pathways are not constitutively active in S2 cells. To measure KV suppression of these pathways, we therefore co-transfected Tolllrr (a Toll receptor lacking the leucine-rich repeat extracellular domain) or PGRP-LC (an Imd pathway receptor) with Drs or Dpt luciferase reporters to artificially induce signalling of Toll and Imd pathways, respectively. Transfection of Tolllrr increased Drs-Fluc by 243-fold (MCMCp < 0.001), consistent with previous reports (Tauszig et al., 2000). However, KV infection reduced the maximum level of Tolllrr-mediated Drs activity by 81% (38-93%, MCMCp < 0.001; Figure 2), indicating KV can inhibit Toll signalling. Over-expression of PGRP-LC led to a 4-fold increase in Dpt-FLuc (3-5 fold, MCMCp < 0.001). In contrast to the effect on Toll signalling, KV infection led to a 3.6-fold increase (2.6-4.8 fold, MCMCp < 0.001) in Dpt-FLuc, which additively increased Dpt-FLuc when PGRP-LC overexpressing cells were infected with KV (17-fold increase compared to Imd-inactive, mock-treated cells; 12-23 fold, Figure 2). This suggests that KV infection in S2 cell culture suppresses Toll signalling but induces Imd signalling.
KV-encoded gp83 modifies NF-kB signalling during infection
The immunosuppressive function of nudivirus genes has not previously been explored. To identify KV-encoded immune inhibitors, we cloned 10 KV genes that are highly expressed at 3 dpi in adult flies (Palmer et al., 2018) and performed immune suppression assays for RNAi, JAK-STAT, Toll, and Imd pathways. We were unable to identify KV-encoded suppressors of RNAi or JAK-STAT among these 10 genes, although we confirmed that Cricket Paralysis Virus protein 1A potently suppressed RNAi in these assays (Nayak et al., 2010); MCMCp = 0.006; Figure 2). However, we found that gp83— a KV gene encoding no recognisable protein domains—significantly reduced Tolllrr-induced Drs-FLuc expression (Figure 2). In this experiment, Tolllrr overexpression induced Drs-FLuc by 24-fold (8-66 fold), but by only 1.9-fold (0.3-8 fold; MCMCp = 0.02) when gp83 was co-expressed. We confirmed that this was not an artefact of the luciferase reporter-based assay by repeating the experiment with qPCR of endogenous Drs as a readout, where gp83 overexpression potently reduced Tolllrr-induced Drs expression (MCMCp < 0.001; Figure S2). We further found that overexpression of gp83 caused a 5-fold (1.5-18 fold) increase in Imd-mediated Dpt-FLuc expression, with or without PGRP-LC overexpression (MCMCp = 0.008; Figure 2).
We next aimed to confirm that the interactions between the transfected KV gene gp83 and NF-kB pathways are representative of the function of gp83 during KV infection. Therefore, we silenced gp83 during KV infection using dsRNA, and measured associated changes in Toll, Imd, and JAK-STAT signalling. Co-transfection of gp83 with independent dsRNAs targeting gp83 completely reversed inhibition of Drs-FLuc compared with transfection of GFP dsRNA, indicating that these dsRNAs effectively silence gp83 (MCMCp < 0.001 for both dsRNAs; Figure S3). KV infection had no effect on Drs-FLuc (MCMCp = 0.26), but inhibited Tolllrr-induced signalling (MCMCp < 0.001), as reported above (Figure 2). However, knockdown of gp83 during KV infection led to increased Drs-FLuc in Toll-inactive cells (MCMCp = 0.004) and cells expressing Tolllrr (MCMCp < 0.001; Figure 3). Likewise, knockdown of gp83 in KV-infected cells caused a modest decrease in Dpt-FLuc in Imd-inactive cells (MCMCp = 0.03), and this effect was much stronger in PGRP-LC overexpressing cells (MCMCp = 0.006; Figure 3). Consistent with a specific interaction with NF-kB signalling, gp83 knockdown had no effect on the ability of KV to suppress JAK-STAT signalling (MCMCp = 0.63; Figure 3).
The immunosuppressive function of gp83 on Toll signalling is consistent with the observed downregulation of AMPs following KV infection in vivo and substantiates the hypothesis that Toll is antiviral and suppressed during infection. However, the induction of antiviral Imd signalling by a single viral protein is unexpected, and it is unclear why KV might not have evolved to avoid or supress Imd activation as seen for in other insect-infecting DNA viruses (Lamiable et al., 2016). Assuming Imd activation is detrimental to virus transmission, this could indicate a trade-off between suppression of Toll and activation of Imd, or that gp83 is directly recognised by the fly immune system. We conclude that KV-encoded gp83 is involved in mediating complex interactions with NF-kB signalling, including suppression of Toll signalling and induction of Imd signalling.
Immune suppression by gp83 occurs downstream of Toll transcription factors
Previously described polydnavirus-encoded Toll pathway inhibitors imitate ΙκΒ, blocking the nuclear entry of NF-kB transcription factors (Thoetkiattikul et al, 2005). Although the precise mechanism of interaction between gp83 and Toll signalling is unknown, suppression of Tolllrr-induced signalling indicates that gp83 also interferes with intracellular Toll signalling. We therefore performed genetic interaction experiments to narrow down the point in the Toll signalling pathway at which gp83 acts. Drs-FLuc was greatly increased by overexpressing pll (240-fold [131-414] induction of Drs-FLuc), silencing cact (75-fold [33-161] induction of Drs-FLuc), and overexpressing Dif (563-fold [317-1002] induction of Drs-FLuc;) or dl (459-fold [257-778] induction of Drs-FLuc;). Co-overexpression of gp83 potently reduced Drs-FLuc in each of these scenarios (MCMCp < 0.001 for each) - pll/gp83 co-overexpression led to a 0.55-fold change in Drs-FLuc (0.31-0.99 fold), cactdsRNA/gp83 led to a 1.73-fold change in Drs-FLuc (0.75-3.5 fold), Dif/gp83 led to a 0.86-fold change in Drs-FLuc (0.5-1.5 fold), and dl/gp83 led to a 1.5-fold change in Drs-FLuc (0.9-2.5 fold) relative to baseline Drs-FLuc expression (Figure 4). Additionally, V5 staining of gp83V5 revealed that gp83 is a nuclear protein (Figure 4E). Together, these results indicate that gp83 either inhibits NF-kB transcription factors, or acts downstream of them.
Virus-encoded inhibitors of NF-kB in mammals have been reported to operate by promoting degradation of NF-kB transcription factors, blocking NF-kB access to the nucleus, or interfering with transcriptional co-activators to evade the interferon response (reviewed in Zhao et al., 2015). Using co-immunoprecipitation and subsequent western blotting or mass spectrometry, we tested whether gp83 directly binds dl. Following immunoprecipitation of GFP-tagged dl, we were able to detect cact as an interacting positive control, but not gp83 (Figure S4). Additionally, we created an S2 cell line stably expressing GFP-tagged gp83, immunoprecipitated gp83gfp, and performed mass spectrometry on interacting partners. We identified 19 Drosophila melanogaster proteins, including 4 nuclear proteins (Nipped-B, Brf, Mlf, Ulp1), that were enriched in the gp83 immunoprecipitate (log2 fold enrichment > 2.5; FDR < 0.1; Figure S4). While we did not identify known downstream NF-kB pathway factors, the extracellular Toll ligand spz was among those enriched, despite the nuclear localization of gp83. However, dsRNA knockdown of spz did not rescue the immunosuppressive effect of gp83, indicating this interaction may not occur in live cells, or that it is not required for gp83 to inhibit Toll signalling. Further, knockdown of a subset of the enriched genes, including 3 of the 4 identified nuclear proteins, was unable to rescue the gp83 immunosuppressive effect (Figure S4), suggesting that gp83 may not form stable complexes with host proteins to interfere with NF-kB signalling (Figure S4).
Although we did not detect a direct association between dl and gp83, we observed a reduction in dl protein levels upon gp83 overexpression that is not dependent on Toll signalling (Figure S5). We quantified this effect by transfecting either GFP or GFP-tagged dl, with or without gp83, and measuring fluorescence by confocal microscopy. We found that while gp83 caused a 53% reduction in GFP levels, possibly due to a dl binding site in the actin 5C promoter of this construct (Zehavi et al., 2014); 42-62%, MCMCp < 0.001), it caused a significantly greater reduction in dlgfp (73% reduction; 66-78%, MCMCp < 0.001). However, KV infection did not decrease dl protein levels, indicating this may not be the primary mechanism by which KV inhibits Toll signalling (Figure S5). Instead, we hypothesize that gp83 interferes with the access of dl to either the nucleus or NF-kB binding sites, which indirectly affects dl localization and results in increased turnover. We prefer the latter explanation, that gp83 directly interferes with the Toll pathway transcriptional response, because overexpression of gp83 simultaneously induced the Dpt reporter (Figure 2) and reduced dl-responsive promoters (Drs-FLuc and Act5C-GFP; Figure 2, Figure S5). These observations implicate gp83 in regulating transcription at diverse loci responsive to both dl and rel, and suggest an interaction between gp83 and NF-kB-responsive genes.
Immunosuppressive function of gp83 depends on conserved residues and is conserved in other nudiviruses
Conflict between the host immune system and virus-encoded immune inhibitors may result in an evolutionary arms race, leading to recurrent positive selection and eventual host specialization (e.g. Obbard et al., 2009; Sawyer and Elde, 2012; Brockhurst et al., 2014). Consistent with this, some immune inhibitors are only effective against their native host species, thereby defining the viral host range (e.g. Parisien, Lau and Horvath, 2002; Mariani et al., 2003; Goffinet et al., 2009; Rajsbaum et al., 2012; van Mierlo et al., 2014; Stabell et al., 2018). We tested whether the immunosuppressive function of gp83 is conserved, and whether gp83 acts in a species-specific manner. The gp83 locus is absent from nudiviruses distantly related to KV, such as Heliothis zea nudivirus 1 (HzNV1), Tipula oleracea nudivirus (ToNV) and Peneaus monodon nudivirus (PmNV), but gp83 homologs are found in the more closely related Gryllus bimaculatus nudivirus (GrBNV), Nilaparvata lugens endogenous nudivirus (NlENV), Oryctes rhinoceros nudivirus (OrNV), Drosophila innubila nudivirus (DiNV), Tomelloso virus (TV), Mauternbach virus (MV), and Esparto virus (EV; Figure 5). Although gp83 lacks recognisable protein domains, several regions are strongly conserved among these nudiviruses, suggesting functional conservation (Figure 5). To test whether gp83 function depends on these conserved domains, we made two gp83 deletion constructs (gp83Δ1 and gp83 Δ2) that respectively remove conserved regions of 18 and 8 amino acids without substantially altering protein stability, and transfected these alongside Tolllrr with the Drs-FLuc reporter. Although detectable by western blotting, gp83 Δ1 (MCMCp = 0.67) or gp83 Δ2 (MCMCp = 0.79) were unable to inhibit Toll signalling, indicating that these conserved residues are important for the immunosuppressive function of gp83. To test whether gp83 function is conserved among viruses, we cloned gp83 from DiNV and performed Toll immunosuppression assays. The gp83 homolog from DiNV was able to completely inhibit D. melanogaster Toll signalling (MCMCp < 0.001), despite only 57% amino acid identity with KV gp83, demonstrating that the immunosuppressive function of gp83 is conserved in other nudiviruses and that it is not highly host-specific. This observation suggests that the Toll-gp83 interaction may not be a hotspot of antagonistic ‘arms race’ coevolution and has not led to specialization of DiNV gp83 to the D. innubila immune system at the expense of its ability to function in D. melanogaster. This could be because gp83 has relatively few direct interactions with host proteins (Figure S4), and may instead interact directly with transcription factor binding sites which are under high constraint, and therefore unable to evolve resistance to the immunosuppressive effect of gp83 (Nitta et al., 2015).
Conclusions
In this study we investigated the role of known anti-RNA viral immune pathways in the context of DNA virus infection, including RNAi, JAK-STAT, Imd, and Toll pathways. Our data support an antiviral role for RNAi and Imd against KV, consistent with previously-described antiviral RNAi against IIV6 and DNA virus-encoded suppressors of Imd (Bronkhorst et al., 2012; Kemp et al., 2013; Lamiable et al., 2016). Furthermore, we identified gp83 as a KV-encoded Toll suppressor that acts downstream of NF-kB transcription factor release of ΙκΒ, strongly suggesting that Toll signalling can be antiviral during DNA virus infection in insects. The immunosuppressive effect of gp83 is conserved in other nudiviruses, and has not evolved host-specificity in DiNV, indicating that the Toll-gp83 interaction is unlikely to be a hotspot of reciprocal host-virus adaptation and that other KV genes may be more important in determining host range.
Acknowledgements
We thank Maria-Carla Saleh, Marc Dionne, François Leulier, David Finnegan, and Bruno Lemaitre for kindly sharing RNAi, Toll, and Imd pathway mutant fly lines. We thank Pascale Dijkers, Jean-Luc Imler, Neal Silverman, Edan Foley, and Norbert Perrimon (Addgene plasmid # 37393) for kindly sharing Toll, Imd, and JAK-STAT constructs. We thank Rob Unckless for sharing a DiNV DNA sample with us. We thank the Ruth Steward and the Developmental Studies Hybridoma Bank for making the dorsal antibody available. WHP is supported by the Darwin Trust of Edinburgh and by an EMBO Short-Term Fellowship in RvR’s laboratory (Grant Number 7095). Work in RvR’s laboratory is supported by a European Research Council Consolidator Grant under the European Union’s Seventh Framework Programme (grant number ERC CoG 615680) and a VICI grant from the Netherlands Organization for Scientific Research (grant number 016.VICI.170.090).