Ae. aegypti Ago2, Dcr2, and TSN genes were identified by homology to D. melanogaster and the An. gambiae orthologs. These genes are present as single copy 1:1:1 orthologous trios. Ae. aegypti orthologs encode conserved protein domains that identify them as RNAi components (Figure 1). The major catalytic enzyme, Ae aegypti Ago2, shares 42.7% amino acid (a.a.) identity with the An. gambiae ortholog and 33.2% identity with the D. melanogaster ortholog (Additional File 1A). Similarly, Ae. aegypti Dcr2 shares 51.3% a.a. identity with An. gambiae Dcr2 and 31.1% with D. melanogaster (Additional File 1B).Ae aegypti TSN shows more conservation across species, with 69.7% and 63% a.a. identity, respectively (Additional File 1C). These levels of similarity are common among proteins of these species 20.

To determine whether Ae. aegypti RNAi components are required for defense against SINV TR339-eGFP, Dcr2, Ago2, and TSN were silenced by injection of dsRNA 3 days prior to an infectious bloodmeal. We predicted that dsRNA injection would transiently silence expression of these genes, because pre-existing RISC proteins should be available to initiate the silencing process. Although both TSN and Dcr2 transcripts showed evidence of transient silencing in midguts following dsRNA injection, there was no detectable reduction in Ago2 mRNA levels by qRT-PCR (Figure 2A). However, there was a reduction in Ago2 116 kDa protein levels, compared to non-specific dsRNA-injected controls, in unfed midguts just prior to an infectious bloodmeal and at 2 dpi (Figure 2B). The silencing was no longer evident by 3 dpi. On average, Ago2 protein levels were reduced to 49% of controls (+/- 16%) immediately prior to the virus meal and 22% of controls (+/- 17%) at 2 dpi. Previous reports showed that dsRNA effectively silences most targeted genes in the midgut of several different mosquito species for >7 days post-injection 193637. In the current study, silencing of RNAi components is more transient.

To determine if Ago2 dsRNA-injections altered virus replication, viral (+) and (-) RNA strands were detected by two-step qRT-PCR of TR339-eGFP-infected mosquito midgut pools. TR339-eGFP (+) strand RNAs showed a transient increase over beta-galactosidase (βGAL)-injected controls at 1 dpi (Figure 2C). The transient increase showed a high variance, probably due to variability in the absolute number of ingested viral particles per mosquito. Importantly, although viral copy number varied from experiment to experiment, in each replicate, Ago2 dsRNA-treated mosquito midguts showed higher viral copy number than bGAL-treated. In contrast, no changes were observed in the copy number of TR339-eGFP (-) strand RNA (data not shown).

With the knowledge that RNAi component silencing requires a functional RISC, we predicted that enhancement of viral titers would be transient. Indeed, Ago2 silencing resulted in transient increases in TR339-eGFP infection rates in whole mosquitoes (χ² = 5.6, P = 0.02) and increased median viral titers (Mann-Whitney U test, P = 2e-15) (Figure 2D, Table 1). Viral eGFP fluorescence in midguts of Ago2 -silenced mosquitoes was higher than controls at 4 dpi (χ² = 16.9, p = 4e-5); however, enhanced viral infection and replication were no longer evident by 7 dpi (Figure 2E; Table 1). Together, these results indicate the requirement for Ago2 in anti-viral defense during TR339-eGFP infection.

A similar pattern of enhanced viral infection was observed following Dcr2 dsRNA treatment. Transient Dcr2 silencing resulted in temporary increases in TR339-eGFP infection rates compared to controls (χ² = 13.0, P = 0.0003) (Figure 2D), however increases in eGFP fluorescence were not significant (Table 1). At 4 dpi, the Dcr2 dsRNA-injected group also had significantly higher median titers than βGAL controls (Figure 2D) (Mann Whitney U test, P = 0.0001), and, by 7 dpi, the effect was abrogated (Figure 2E). As was seen with Ago2, there is a clear requirement for Dcr2 during anti-viral defense.

TSN silencing showed less striking enhancement of viral infection than either Ago2 or Dcr2. At 4 dpi, median viral titers of TSN-silenced mosquitoes were not significantly different from controls (Figure 2D); however infection rates were slightly higher (Mann Whitney U test, P = 0.03). By 7 dpi, the infection pattern, detected by eGFP fluorescence showed significantly higher dissemination (χ² = 5.2, p = 0.02), even though clearing of midgut infection was evident (Table 1). This fluorescence was primarily present in hemolymph with very little in peripheral tissues. Concurrently, viral titers were reduced to levels below those of controls (Figure 2E). Importantly, detection of eGFP in this system is indicative of viral protein production. Therefore, the eGFP fluorescence in hemocytes coupled with reduced viral titers indicates that only a small amount of live virus escaped the midgut. Alternatively, hemocytes became infected in the midgut and then migrated to other tissues. Nevertheless, the virus was limited in its ability to replicate further. All of the results suggest that the effects of silencing wane by 7 dpi. This could be due to the decay of exogenously-administered dsRNA, or transient stimulation of infection rates at 4 dpi might be overcome by a stimulated anti-viral defense pathway by 7 dpi.

SINV non-structural protein 3 (nsP3) participates in replication complex formation 3839. Injection of dsRNA derived from TR339 nsP3 showed that this treatment nearly ablated viral replication in 12 of 13 mosquitoes at both 4 and 7 dpi (Figure 2D, 2E, Table 1). Midgut eGFP fluorescence was significantly reduced in nsP3-targeted mosquitoes (χ² = 8.2, P = .00042) (Table 1). This evidence demonstrates the ability of virus RNA-derived dsRNA to trigger anti-viral RNAi in Ae aegypti.

Mutation of genes encoding RNAi pathway components in D. melanogaster resulted in increased mortality of flies infected with invertebrate viral pathogens 914. In addition, there is evidence that some arboviruses cause pathology in infected mosquito tissues 404142. To determine whether RNAi components prevent mortality during SINV infection, we examined survival curves of TR339-eGFP infected Ae. aegypti injected with Ago2, Dcr2 or TSN dsRNA. Results indicated that mortality was not increased over βGal dsRNA-injected controls (Additional File 2). The transient effects of the dsRNA injection or the weak replication of TR339-eGFP may have been factors in the lack of mortality. Therefore, the mortality study was repeated with Dcr2 silenced mosquitoes infected with MRE16-eGFP. Again, there was no significant enhancement of mosquito mortality (Additional File 3). Nevertheless, to confirm these preliminary findings, a gene knock-out system for mosquitoes would be required.

The production of sequence-specific small RNAs is a hallmark of RNAi defense. RNAi activity was confirmed during natural TR339-eGFP and MRE16-eGFP infections by identifying viral small RNAs in adult female mosquitoes (Figure 3). We used MRE16-eGFP whole genome capture probes, as these two viruses share about 75% nucleotide identity 43. TR339-eGFP infected mosquitoes accumulated far more small viral RNAs between 2 and 7 dpi than MRE16-eGFP infected mosquitoes. Small viral RNAs from MRE16-eGFP were barely detectable only at 2 dpi. These data are interesting in light of the differences in replication kinetics of these two viruses (Figure 4). They could indicate a correlation between replication efficiency in the mosquito and susceptibility to the anti-viral RNAi defense, or they could point to a viral suppressor of RNAi in the MRE16-eGFP virus 41044. It is also apparent that TR339-eGFP-derived (+) sense small RNAs are more abundant than (-) sense small RNA in mosquito tissues. This suggests that the (+) virus strand is targeted by RNAi more efficiently than the (-) virus strand or that diced (+) sense viral RNA is preferentially stabilized in the RISC and used as the guide strand. It could also suggest that the primary viral dsRNA trigger is (+) sense viral RNA secondary structure, rather than (+) and (-) dsRNA of the replicative intermediate.

The mosquito midgut is the initial site of arbovirus replication and thus is suspected to be the first site of anti-viral defense. Blood-feeding alone results in reduced midgut Dcr2 mRNA levels at 1 and 2 days post-feeding (Figure 4A) (Mann-Whitney U Test, P = 0.02 and P = 0.02, respectively). Although fluctuations were also observed in Ago2, Dcr2 and TSN mRNA levels at other timepoints, they were not significantly different from controls.

TR339-eGFP and MRE16-eGFP replicate and amplify in the midgut for 2 to 3 days prior to escape to other tissues 3335. Mosquitoes were orally infected with either MRE16-eGFP or TR339-eGFP, and RNAi component midgut transcript levels were determined relative to un-infected blood-fed controls (Figure 4B, 4D; Additional File 4). The mosquito's overall transcript pattern differed for each virus strain. During MRE16-eGFP infection, TSN transcript levels, but not Ago2 or Dcr2, were significantly enriched over controls at 1 and 3 dpi (Mann-Whitney U Test, P = 0.05 and P = 0.04, respectively), and highly variable at 2 dpi. In contrast, during TR339-eGFP infection, TSN transcript levels were not significantly enriched until 3 dpi, when a 5-fold enrichment was followed by a sharp decrease at 4 dpi (Mann-Whitney U Test, P = 0.04 and P = 0.04, respectively).

Enhancement of Ae. aegypti TSN transcripts was previously reported in microarray analyses of SINV MRE16 infection 5. The TSN transcript enrichment in Ae. aegypti during SINV infection contrasts with reports of TSN depletion in Drosophila during DCV infection 6. DCV has a single-stranded positive sense RNA genome, as does SINV. However, DCV, a picorna-like virus, is an insect-specific pathogen whereas SINV is an arbovirus, and by definition, infects both mosquitoes and vertebrates and establishes persistent, non-lethal infections in the insect vector.

Detection of (-) strand RNA served as evidence of viral replication in the samples used for RNAi component transcript analysis. Interestingly, viral replication markedly differed between the two SINV strains. MRE16-eGFP (-) RNA levels increased over the 7 day period, from an average of 276 copies per nanogram (ng) midgut RNA at 1 dpi to 670 copies at 7 dpi (Figure 4C). In contrast, TR339-eGFP (-) RNA was undetectable in 2 of 3 experimental replicates at 1 dpi and increased to an average high of 33 copies at 2 dpi. Over the remainder of the 7 day time course, (-) strand RNA levels remained low (Figure 4E). The steady increase in (-) strand SINV RNA in mosquitoes over the 7 day period differs from that observed in vertebrate cell culture. In vertebrate cells, (-) strand RNA synthesis ceases about 30 hr post-infection, due to the cleavage of nsP1-4 products shifting replication towards (+) strand synthesis 45. The continued increase in (-) strand RNA in midguts in contrast to that of vertebrate cell culture is probably due to the progressive cell-to-cell spread of virus through the tissue, whereas, in vitro cell culture infection is initiated by bathing all cells in virus.

Importantly, despite clear differences in MRE16-eGFP and TR339-eGFP growth characteristics, no statistically significant changes from controls were seen for Ago2 or Dcr2 mRNAs during SINV infection (Figure 4B, 4D). Further, enrichment of TSN transcripts occurred at different times following infection for each virus strain. Transcript enrichment occurred early during MRE16-eGFP infection concomitant with viral replication, whereas, TSN enrichment in TR339-eGFP infected midguts was delayed, possibly because viral replication remained at low levels. These results suggest that TSN may act as a sensor for the RNAi pathway.

The lack of statistically significant increases in Ago2 transcript levels during SINV infection led us to ask whether protein levels fluctuated. Figure 5 shows Ago2 protein profiles in MRE16-eGFP and TR339-eGFP infected midguts compared to unfed and un-infected blood-fed controls. Bloodfeeding alone reduces Ago2 protein levels compared to midguts from unfed mosquitoes, and these proteins re-accumulated over a time period concomitant with the expected time course of bloodmeal digestion, vitellogenesis, and a return to the unfed state.

Virus infection influenced Ago2 protein patterns in the midgut. Depletion of Ago2 protein occurred during either TR339-eGFP or MRE16-eGFP infection at 1 dpi; thereafter, Ago2 protein levels varied in a virus-dependent manner, and were distinct from the profile seen in un-infected controls. In MRE16-eGFP-infected midguts, Ago2 protein remained depleted at 2 dpi and increased over un-infected controls at 3 dpi. This enrichment corresponds to the concurrent, though statistically insignificant, increase in Ago2 transcripts (Figure 4B). In contrast, during TR339-eGFP infection, Ago2 protein levels were enriched and evident as multiple bands at 2 dpi, were roughly equivalent to bloodfed controls at 3 dpi, and reduced again at 4 dpi. By 7 dpi, protein profiles for all virus-infected midguts were similar to those of bloodfed controls (data not shown). When comparing natural fluctuations of Ago2 protein levels in Figure 4 with those from dsRNA-injected mosquitoes in Figure 2, it is apparent that injection of non-specific dsRNA results in depletion of Ago2 levels.

In summary, the low level of TR339-eGFP replication in midguts, coupled with a lack of Ago2 protein depletion and the detection of small viral RNAs, indicates that anti-viral RNAi defense is active during infection with this virus. The low steady state levels of (-) strand RNA detected by qRT-PCR do not reveal fluctuations that could occur due to RNA replication or degradation. Therefore, further analysis is required to determine whether these low levels are due to inefficient replication or an active RNAi response. A much different pattern was observed during MRE16-eGFP infection. MRE16-eGFP replicates to significantly higher copy number than TR339-eGFP while somehow preventing the accumulation of small viral RNAs indicative of effective RNAi defense. The specifics of this evasion or suppression mechanism are not understood and warrant further study. Importantly, these results underscore the different effects that can occur with different arboviruses and indicate that specific effects on the RNAi pathway need to be characterized separately for each virus system.

The differential effects of MRE16-eGFP and TR339-eGFP on RNAi pathway activity indicate a need to investigate the viral features that prevent or evoke RNAi pathway activity. A deeper understanding of these virus-specific effects will be important to defining the characteristics of an efficient arbovirus and the limitations of RNAi anti-viral defense in arbovirus vectors.

MRE16-eGFP and TR339-eGFP virus stocks were generated from SINV infectious cDNA clones with inserts of enhanced GFP (eGFP) driven by the second sub-genomic promoter, using standard methods 354647. MRE16-eGFP was passed once in baby hamster kidney (BHK-21) cells and twice in Aedes albopictus clone C6/36 cells. TR339 5'2J-eGFP was passed once in BHK-21 cells and three times in C6/36 cells.

For mosquito feedings, all viruses were diluted as indicated in de-fibrinated sheep blood and provided at 37°C in a water jacketed artificial feeder with parafilm membrane. An aliquot of each SINV stock was titered by viral plaque assay on Vero cells and the other aliquots were stored at -80°C and diluted prior to feeding. TR339-eGFP blood-meals contained approximately 3.3 × 10⁸ pfu/ml and MRE16 meals contained 2.2 × 10⁷ pfu/ml. Representative mosquitoes from each group were selected for dissection and eGFP detection with an Olympus epi-fluorescence microscope to confirm midgut infection with each batch of virus. Viral titers from individual mosquitoes were determined by plaque assays of filtered mosquito homogenates as previously described 29.

Colonized Ae. aegypti Rexville D- Higgs' White Eye (HWE) mosquitoes were reared at 28°C, 80% relative humidity, with a photoperiod of 14:10 (L:D). Adults were provided with a sugar source and water and held in the same conditions during the extrinsic incubation period.

dsRNA was transcribed from templates generated from cDNA clones following PCR amplification with 5' primer extensions bearing a T7 RNA polymerase promoter (Additional File 4) and the protocol described with the Megascript dsRNA kit (Ambion, Austin, TX). dsRNA was purified with phenol/chloroform extraction and diluted to a concentration of 1.0 μg/μl in phosphate-buffered saline (PBS). Adult female Ae. aegypti (2–3 days post emergence) were anesthetized with cold, held on ice and intrathoracically injected with approximately 500 nanograms dsRNA. Three days later, mosquitoes were harvested for protein extraction or fed a blood-meal containing 8 log pfu/ml TR339-eGFP.

Evidence of virus infection and dissemination was observed at 4 and 7 days post-infection (dpi) by dissecting mosquitoes in PBS and observing eGFP fluorescence of midguts and peripheral tissues by epi-fluorescence microscopy (Olympus, Center Valley, PA).

Pools of 5 midguts from three independent feedings were frozen at -80C prior to extraction of total RNA using Trizol Reagent (Invitrogen, Carlsbad, CA). Genomic DNA was removed using TurboDNase I (Ambion, Austin, TX) in a 100 μl reaction for 1 hour at 37°C and inactivated using DNase I inactivation beads. Samples were stored as ethanol slurries at -80°C in multiple aliquots. Amplicons were prepared in an Opticon 2 real-time PCR thermocycler (BioRad, Hercules, CA) in triplicate using SYBR Green One step qRT-PCR reagents (Invitrogen, Carlsbad, CA) in 20 μl reactions with 2.0 ng total RNA per well. Additional File 4 shows the qPCR primers used. Cycling parameters followed the manufacturer's recommendations and used a 57°C extension temperature. Relative transcript levels were determined using the comparative Ct analytical method with an actin reference standard and the blood-fed control as calibrator 48.

Two step amplification of (-) strand SINV RNA was performed according to the methods of Richardson 49. Briefly, 10 ng mosquito midgut total RNA were used for reverse transcription reactions. For MRE16-eGFP and TR339-eGFP, the forward primer (Additional File 4) was used to transcribe (-) strand RNA. A 2 μl aliquot from the reverse transcriptase reaction was used in quantitative real-time PCR using ABI Powerscript (Applied Biosystems, Foster City, CA). Viral copies were calculated using the standard curve method 49.

All midguts were harvested into PBS with protease inhibitor cocktail (P2714, Sigma, St Louis, MO), flash frozen in dry ice, stored at -80C, and processed simultaneously for immunoblot analysis. Pools of 10 adult female midguts were triturated in 1% sodium dodecyl sulfate, protease inhibitor cocktail, 20 micrograms/ml phenylmethylsulfonylfluoride (Sigma-Aldrich, St Louis, MO) diluted in PBS. Equivalent amounts of protein extracts (10 υg) were separated using standard denaturing conditions on a 4 to 15% gradient gel and transferred to PVDF membrane using a tank blotter (BioRad, Hercules, CA) and transfer buffer containing 25 mM Tris, pH 8.3, 192 mM glycine. Ago2 was detected using a 1:100 dilution of affinity purified rabbit anti- Ae. aegypti Ago2 antibody recognizing the peptide sequence C-YERMQIRTEIQDGHPMFFV (QCB, Hopkinton, MA). Antibody affinity purification followed methods recommended by the antibody service company (QCB, Hopkinton, MA). The corresponding genetic code is present only once in the Ae. aegypti genome 50. Multiple banding patterns were observed; the predicted molecular weight of Ago2 is 115 kilodaltons (kDa). The loading control was a 19 kDa anti-Ago2 antibody cross-reacting band that remained constant during all conditions. The secondary antibody was goat anti-rabbit IgG conjugated to horseradish peroxidase (Abcam, Cambridge, MA). Proteins were detected using ECL Plus (GE Healthcare, Piscataway, NJ) detection reagents and ECL Hyperfilm (GE Healthcare, Piscataway, NJ).

Mosquitoes were infected as described or fed blood alone and held for 2 or 7 days prior to storage in Trizol reagent (Invitrogen, Carlsbad, CA) at -80°C. Total RNA extraction followed manufacturer's recommendations. Small RNAs corresponding to 18–25 nucleotides were separated on a 7 M urea, 12% acrylamide gel, then excised and eluted. These small RNAs were treated with calf intestinal alkaline phosphatase to remove 5' phosphates and radiolabeled using T4 polynucleotide kinase and gamma-³²P ATP. Radiolabeled RNAs were mixed with pooled biotinylated viral probes, described below, in 10 mM Tris, pH 7.6 and denatured for 1 min at 95°C. Reaction mixtures were adjusted to 400 mM NaCl and incubated at 42°C for 4 hr to allow hybridization. Streptavidin agarose beads were added to the mixture, incubated at room temperature for 10 min to allow capture of the biotinylated probes, and washed 4 times in 10 mM Tris, pH 7.6, 400 mM NaCl. Radiolabeled small RNAs were eluted from the probes in 10 mM Tris, pH 7.6 at 95°C for 5 min. Eluted small RNAs were concentrated by ethanol precipitation, separated on a 5% acrylamide gel containing 7 M urea, and detected on a Typhoon phosphorimager (GE Biosciences).

Biotinylated, strand-specific, viral RNA hybrid selection probes were prepared by restriction endonuclease digestion of the infectious MRE16-eGFP cDNA clone plasmid, followed by PCR amplification of 3 to 5 kB portions of the viral genome (Additional File 4). Each of these was used as a template in transcription of biotinylated probes of (-) or (+) strand polarity. Probes representing the (-) or (+) viral strands were equally pooled to provide genome- or anti-genome-wide coverage in the hybridization step. TR339-eGFP was not used as a hybridization probe because of inefficiencies in biotinylated RNA probe.