Imd pathway-specific immune assays reveal NF-κB stimulation by viral RNA PAMPs in Aedes aegypti Aag2 cells

Background The mosquito Aedes aegypti is a major vector for the arthropod-borne viruses (arboviruses) chikungunya, dengue, yellow fever and Zika viruses. Vector immune responses pose a major barrier to arboviral transmission, and transgenic insects with altered immunity have been proposed as tools for reducing the global public health impact of arboviral diseases. However, a better understanding of virus-immune interactions is needed to progress the development of such transgenic insects. Although the NF-κB-regulated Toll and ‘immunodeficiency’ (Imd) pathways are increasingly thought to be antiviral, relevant pattern recognition receptors (PRRs) and pathogen-associated molecular patterns (PAMPs) remain poorly characterised in A. aegypti. Methodology/Principle findings We developed novel RT-qPCR and luciferase reporter assays to measure induction of the Toll and Imd pathways in the commonly used A. aegypti-derived Aag2 cell line. We thus determined that the Toll pathway is not inducible by exogenous stimulation with bacterial, viral or fungal stimuli in Aag2 cells under our experimental conditions. We used our Imd pathway-specific assays to demonstrate that the viral dsRNA mimic poly(I:C) is sensed by the Imd pathway, likely through intracellular and extracellular PRRs. The Imd pathway was also induced during infection with the model insect-specific virus cricket paralysis virus (CrPV). Conclusions/Significance Our demonstration that a general PAMP shared by many arboviruses is sensed by the Imd pathway paves the way for future studies to determine how viral RNA is sensed by mosquito PRRs at a molecular level. Our data also suggest that studies measuring inducible immune pathway activation through antimicrobial peptide (AMP) expression in Aag2 cells should be interpreted cautiously given that the Toll pathway is not responsive under all experimental conditions. With no antiviral therapies and few effective vaccines available to treat arboviral diseases, our findings provide new insights relevant to the development of transgenic mosquitoes as a means of reducing arbovirus transmission.


Methodology/Principle findings
We developed novel RT-qPCR and luciferase reporter assays to measure induction of the Toll and Imd pathways in the commonly used A. aegypti-derived Aag2 cell line. We thus determined that the Toll pathway is not inducible by exogenous stimulation with bacterial, viral or fungal stimuli in Aag2 cells under our experimental conditions. We used our Imd pathway-specific assays to demonstrate that the viral dsRNA mimic poly(I:C) is sensed by the Imd pathway, likely through intracellular and extracellular PRRs. The Imd pathway was also induced during infection with the model insect-specific virus cricket paralysis virus (CrPV).

Conclusions/Significance
Our demonstration that a general PAMP shared by many arboviruses is sensed by the Imd pathway paves the way for future studies to determine how viral RNA is sensed by mosquito PRRs at a molecular level. Our data also suggest that studies measuring inducible immune pathway activation through antimicrobial peptide (AMP) expression in Aag2 cells should be PLOS

Introduction
Arthropod-borne viruses (arboviruses) are causing an ever-increasing global disease burden, with many arboviruses emerging, re-emerging or at risk of future emergence into the human population [1]. The mosquito species Aedes aegypti is distributed globally across the tropics and subtropics and is the most important arboviral vector of human diseases, including those caused by chikungunya, dengue, yellow fever and Zika viruses [2][3][4]. Given the lack of effective vaccines or antiviral therapies against most arboviruses, vector control measures that block transmission cycles remain one of the most effective ways of preventing human disease. The ability of individual vector species and populations to transmit any given arbovirus is defined as 'vector competence' [5]. An improved understanding of the molecular virus-vector interactions underpinning vector competence would aid the development of genetically modified transmission-incompetent insects that could help reduce the global burden of arboviral disease. Mosquito immune responses have been shown to pose a major barrier to arbovirus transmission [6][7][8][9][10][11][12]. Therefore, manipulating immune responses has been proposed as an attractive approach for reducing the competence of A. aegypti for the broad range of arboviruses transmitted by this species [13].
Intracellular antiviral immune responses in mosquitoes include RNA interference (RNAi) pathways and inducible signalling cascades that result in the production of antimicrobial peptides (AMPs) and other gene products [14]. The expression of AMPs can be induced through the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB)-regulated Toll and 'immunodeficiency' (Imd) pathways (Fig 1A), as well as the Janus kinase-signal transducer and activator of transcription (Jak-STAT) pathway [14]. In the model insect Drosophila melanogaster, which is classified in the order Diptera (flies) along with mosquitoes, AMPs like defensins have classically been considered to be Toll pathway-regulated while cecropins were considered to be Imd pathway-regulated [14]. However, the regulation of AMPs has been shown to be more complex than this [15,16], with species-and tissue-specific AMP regulation and cross-talk between NF-κB pathways [17][18][19][20][21].
In A. aegypti, both the Toll and Imd pathways can be activated by arbovirus infection in vivo in a virus-specific manner [6][7][8][22][23][24]. Furthermore, NF-κB-regulated AMPs have been shown to possess antiviral activity [25,26]. Despite the growing evidence that NF-κB pathways play an antiviral role in mosquito vectors, a deep understanding of the viral pathogen-associated molecular patterns (PAMPs) and associated pattern recognition receptors (PRR) that stimulate NF-κB signalling is lacking. So far, NF-κB-regulated AMPs have been shown to be induced following detection of the dengue virus (DENV) envelope protein by complementlike thioester-containing proteins (TEPs) in combination with scavenger receptor C in A. aegypti in vivo [25]. In addition to this virus-specific PAMP-PRR combination, one may also expect more generally responsive PRRs to exist.
While vector competence studies can ultimately only be performed in vivo, cell culture models are invaluable tools for dissecting molecular virus-immune interactions in fine detail. Here, we developed highly specific RT-qPCR and luciferase reporter assays for measuring activation of the Imd pathway in the A. aegypti-derived Aag2 cell line. In the process, we demonstrated that Toll signalling is not activated by bacterial or viral PAMPs in Aag2 cells under our experimental conditions. Furthermore, the assays were used to show that the Imd pathway can be induced by viral infection and by poly(I:C), a mimic of viral dsRNA PAMPs. Our data suggest the existence of both cytoplasmic and cell membrane-localised PRRs capable of detecting viral RNA in A. aegypti cells.

Fig 1. Stimulation of genes regulated through the Imd pathway by bacteria and zymosan in Aag2 cells. (A)
Schematic of the A. aegypti Toll and Imd pathways. Throughout the manuscript, Toll pathway data is consistently presented in orange while Imd pathway data is presented in yellow. (i) The Toll pathway is induced when the transmembrane cytokine receptor TOLL binds to its dimeric cognate cytokine SPÄ TZLE, which itself is activated by cleavage of the inactive precursor pro-SPÄ TZLE following an extracellular proteolytic cascade initiated by PAMP detection. Subsequent TOLL dimerisation recruits the death domain-containing proteins myeloid differentiation primary response 88 (MYD88), TUBE and PELLE, resulting in phosphorylation and ubiquitin-mediated proteasomal degradation of the negative regulator CACTUS. This allows the nuclear translocation of the NF-κB-like transcription factor REL1A and its co-activator REL1B into the nucleus to induce the expression of AMPs and other immuneregulated genes. (ii) Imd pathway activation is associated with recruitment of the immunodeficiency (IMD) and Fasassociated death domain (FADD) proteins to activate the caspase DREDD (shown here following peptidoglycan (PG) sensing by the peptidoglycan recognition protein PGRP-LP). DREDD cleaves the transcription factor REL2 in a phosphorylation-dependent manner, allowing REL2's N-terminal NF-κB domain to translocate into the nucleus and activate gene expression, while the C-terminal inhibitor of κB (IκB) domain remains cytoplasmic. REL2 is phosphorylated after caspases cleave IMD, allowing IMD to bind the inhibitor of apoptosis (IAP) family protein IAP2, an E3 ligase, which recruits and activates the IκB kinase (IKK) complex via K63 ubiquitin conjugation. CASPAR negatively regulates Imd pathway activation by inhibiting REL2 cleavage. (B) Summary of published gene expression data for transgenic A. aegypti overexpressing REL1A or REL2 [35], or wild-type A. aegypti intrathoracically inoculated with CACTUS dsRNA (CAC -/-) [7]. Gene expression data are replotted as fold induction relative to experimental control. Only genes relevant to the current study are shown. Genes are grouped based on induction through the Toll pathway, or both the Toll and Imd pathways. Gene annotation and Vectorbase Gene ID indicated; '[?]' denotes hypothetical genes with no functional or protein identity annotation. (C-E) Fold induction (i) of genes measured by RT-qPCR in Aag2 cells stimulated for 24 h with 1,000 colony forming units (CFU)/cell of heat-inactivated L. monocytogenes (C) or E. coli (D), or 1 mg/ml zymosan (E). (ii) Percent gene induction in stimulated cells transfected with specified dsRNAs directed against REL1A (dsREL1A) or REL2 (dsREL2) relative to the non-silenced control ('-'), which was taken to be 100%. Two independent non-overlapping dsRNAs were tested for each gene; dsRNAs against GFP serve as a negative control. Log 2 scale; genes consistently ordered by fold induction for L. monocytogenes (Ci) and REL2-specific or non-specific ('not specific') induction. https://doi.org/10.1371/journal.pntd.0008524.g001

Viruses
Cricket paralysis virus (CrPV) was a kind gift from Raul Andino. CrPV stocks were prepared by infecting confluent S2 cells one day post-seeding at a multiplicity of infection (MOI) of 0.2 in culture medium containing 2% (v/v) FBS and harvesting the culture supernatant four days post-infection. The supernatant was clarified by centrifugation and used as stock virus for infections. To determine the viral stock titre, S2 cells were seeded at 1 x 10 6 cells/well in 12-well plates and infected one day post-seeding with serially diluted inoculum made up in PBS. After a one-hour incubation with gentle rocking at room temperature, an overlay of 0.8% (w/v) methyl cellulose (Sigma-Aldrich) in culture medium containing 2% FBS (v/v) was added. Cells were fixed after four days of culture at 28˚C in 4% (w/v) paraformaldehyde and stained in 1% (v/v) crystal violet solution made up in 20% (v/v) ethanol.

Bacteria and immune stimulation
Escherichia coli DH5α were purchased from ThermoFisher Scientific. Listeria monocytogenes and Staphylococcus aureus were kind gifts from Adolfo Garcia-Sastre and Flora Samaroo respectively (both Icahn School of Medicine at Mount Sinai, New York, NY USA). Bacteria were cultured overnight at 37˚C with shaking in Luria Bertani medium (E. coli) or brain heart infusion broth (L. monocytogenes, S. aureus) without antibiotics, and were titrated on corresponding agar medium in 6-well plates. Culture media were from Sigma-Aldrich. Bacterial cultures were washed in phosphate buffered saline (PBS), resuspended in PBS and heat-inactivated at 60˚C for 3 h (E.coli, L. monocytogenes) or 75˚C for 6 h (S. aureus).
Cells were stimulated one day post-seeding by replacing the culture medium with fresh medium containing bacteria, zymosan or poly(I:C) at concentrations indicated. Zymosan A was purchased from Sigma-Aldrich and resuspended in sterile PBS. Poly(I:C) transfections were performed using Lipofectamine RNAi-MAX (ThermoFisher Scientific) one day postseeding as per manufacturer's instructions. RT-qPCR assays specifically measuring Imd pathway activation in Aag2 cells. (A) Efficiency of gene silencing with two independent non-overlapping dsRNAs directed against REL1A (i), MYD88 (ii), REL2 (iii) or IMD (iv). Non-silenced cells transfected with no dsRNA ('-') taken as 100%; dsRNAs against GFP act as a negative control. (B-F) (i) Dose-dependent fold induction of indicated genes relative to unstimulated control in Aag2 cells 24 h poststimulation with 1,000, 100 or 10 CFU/cell heat-inactivated L. monocytogenes ('G+') or E. coli ('G-'), or 1, 0.1 or 0.01 mg/ml zymosan ('zym.'). Dashed line in (F) indicates one-fold (no) induction. (ii, iii) Relative induction of indicated genes following stimulation with 1,000 CFU/cell heat-inactivated bacteria or 1 mg/ml zymosan in Aag2 cells previously transfected with two independent dsRNAs directed against REL1A or REL2 (ii) or MYD88 or IMD (iii). Non-specific control dsRNA against GFP (dsGFP) taken as 100%; two independent non-overlapping dsRNAs were used for test gene knockdowns. Statistical significance for fold induction (i) or percent reduction (ii, iii) indicated; � P < 0.05; �� P < 0.01; ��� P < 0.001. Error bars indicate standard error of the mean.
The mCherry expression vector pKM7 was constructed by digesting pIEx-EGFP [31], a kind gift from Doug Brackney (The Connecticut Agricultural Experiment Station, New Haven, CT USA), with XhoI and NcoI and substituting the EGFP sequence for the mCherry sequence. pKM28 encodes A. aegypti REL1A with a C-terminal FLAG tag (GAT-TACAAGGATGACGATGACAAG, peptide sequence DYKDDDD), and was cloned by digesting pIEx-EGFP with XhoI and NcoI and replacing the EGFP sequence with the FLAG-tagged REL1A sequence (vectorbase.org reference AAEL007696). pKM5 encodes A. aegypti REL2 with an N-terminal FLAG tag, which was cloned in the same way as pKM28 using the REL2 (AAEL007624) sequence. There are three isoforms of REL2 in A. aegypti produced by alternative splicing [32]; we cloned the dominant full-length sequence (NCBI reference XM_001658467.1), not the 'Rel-type' lacking the C-terminal IκB domain or the 'IκB-type' lacking much of the N-terminus. To generate the plasmid expressing constitutively active N-terminally FLAG-tagged REL2 lacking its IκB domain (pKM33) we amplified the N-terminal sequence of REL2 up to the caspase cleavage site from pKM5 and cloned it into pIEx-EGFP as for the other expression vectors. The caspase cleavage site was determined to be 552 LETD 555 by reference to D. melanogaster RELISH and by using http:// www.casbase.org/casvm/index.html [33]. All A. aegypti gene sequences were PCR amplified from Aag2 cells. We also generated a control vector expressing just the FLAG tag (pKM3) by substituting the EGFP from pIEx-EGFP with the FLAG sequence following digestion with XhoI and NcoI.
All cloning was performed using In-Fusion cloning (Takara Biosciences, Mountain View, CA USA). Insert sequences were purchased from Sigma-Aldrich or Integrated DNA Technologies (Coralville, IA USA). All of the plasmids that we generated here have been deposited with Addgene with reference numbers 154152-154168.

RNAi gene silencing
Aag2 cells were seeded at 1 x 10 5 cells/well in 96-well plates and concurrently transfected at a final dsRNA concentration of 10 nM using Lipofectamine RNAi-MAX (ThermoFisher Scientific) as per manufacturer's instructions. dsRNA was prepared by PCR amplification with primers detailed in Table 1 using RNA purified from Aag2 cells as a template; EGFP dsRNA was prepared using plasmid pIEx-EGFP as a template. PCR amplicons were subjected to in vitro transcription using the MEGAshortscript in vitro transcription kit (ThermoFisher Scientific) and gel purified before storage in aliquots at -80˚C.

RT-qPCR
Total RNA was extracted from 1 x 10 5 cells using the Quick-RNA MiniPrep kit with Zymo-Spin IIICG columns (Zymo Research, Irvine, CA USA) as per manufacturer's instructions. cDNA was generated from RNA using the iScript reverse transcription supermix for RT-qPCR (Bio-Rad, Hercules, CA USA) as per the manufacturer's instructions. qPCR amplification and detection was performed using the Hot FIREPol EvaGreen qPCR supermix Plus, no ROX (Solis Biodyne, Tartu, Estonia) and primers detailed in Table 2 at 95˚C for 10 min, followed by 40 cycles of 95˚C for 15 s and 60˚C for 30 s, as per manufacturer's instructions on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Gene expression was normalised against the 40S ribosomal protein S7 (RPS7) mRNA.
CrPV RNA standards for absolute quantification of viral RNA were generated by PCR amplifying nucleotides 2,173-3,200 of the CrPV genomic RNA non-structural ORF from infected Aag2 cells using the following primers: prKM120F, TAATACGACTCACTATAG GGGTTCTCCTCTCTCCTGTCAAAG; prKM120R, TCACTAATAAATTTGTAAGCT TTGTTCTTAA. The forward primer contains the T7 promoter sequence at its 5' end to generate a positive-sense RNA standard. PCR was performed using the SuperScript III One-Step RT-PCR System with Platinum Taq DNA Polymerase (ThermoFisher Scientific) as per manufacturer's instructions with the addition of 2 μl MgSO 4 , 2.5 μl bovine serum albumin (BSA) (New England Biolabs, Ipswich, MA USA) and 2.5 μl dimethyl sulfoxide (DMSO) (Sigma-Aldrich) in a 25 μl reaction volume. Cycles were 55˚C for 30 min, 94˚C for 2 min, followed by 40 cycles of 94˚C for 30 s, 59˚C for 30 s, 68˚C for 4 min, and a final extension at 68˚C for 10 min. Gel purified cDNA was subjected to in vitro transcription using the MEGAscript T7 High Yield Transcription Kit (ThermoFisher Scientific) as per manufacturer's instructions. RNA standards were treated with the DNA-free DNA removal kit (ThermoFisher Scientific) as per manufacturer's instructions and gel purified before use.

Luciferase assays
Aag2 cells were seeded at 1 x 10 5 cells/well in 96-well plates and transfected at the time of seeding with 10 ng pKM50 (Renilla luciferase transfection control) and 50 ng/well of the respective firefly reporter plasmid using Lipofectamine 3000 transfection reagent (ThermoFisher Scientific) as per manufacturer's instructions. Where appropriate, 300 ng/well protein expression plasmid was also included. Cells were harvested in passive lysis buffer (Promega, Madison, WI USA) and incubated with gentle agitation for 20 min at room temperature. Firefly and Renilla luciferase activity were measured using a CLARIOstar plate reader (BMG Labtech, Aylesbury, UK) and home-made luciferase assay reagents as published [34].

Immune gene data mine and transcription factor binding site analysis
Genes overexpressed compared to wild-type mosquitoes in transgenic A. aegypti overexpressing REL1A or REL2 were taken from Zou et al. supplemental tables S1, S2, S3 and S4 [35]. Genes upregulated in A. aegypti intrathoracically injected with CACTUS dsRNA were taken from Xi et al. supplemental table S6 [7]. Data were pooled and at the time of the analysis gene identifiers were converted to A. aegypti strain Liverpool AGWG genome assembly AaegL3 (vectorbase.org) [36,37]. Throughout this manuscript we use the most up-to-date gene identifiers from gene set AaegL5.2 [36,38]. NF-κB binding sites are enriched in the 500 bp upstream of the 5'UTR in D. melanogaster [39]. Therefore, for genes with an annotated 5'UTR in the AaegL3 assembly, the 1,000 bp upstream and the first 100 bp downstream of the transcription start site were downloaded from vectorbase.org to provide maximum opportunity for identifying NF-κB binding sites. For AAEL003857 (DEFD), AAEL011764 (PPO10), AAEL026265 (CTLGA8) and AAEL029044 (CECE) the promoters were not annotated at the time of the analysis. For these genes, the 2,000 bp upstream and the 100 bp downstream of the start codon were downloaded from vectorbase.org, as our own analysis of 129 genes with annotated 5'UTRs found that this would capture the 500 bp upstream of the 5'UTR in 98% of cases. Sequences in FASTA format were analysed for the presence of predicted D. melanogaster NF-κB binding sites using MatInspector (genomatix.de, Genomatix, Munich, Germany). In addition, we searched for A. aegypti REL1A-and REL2-specific and dually responsive NF-κB binding sites defined by Shin et al. [29].

Data analysis and images
All data were analysed in Microsoft Excel (Microsoft Corporation, Redmund, WA USA). Statistical significance was tested using Student's t test assuming unequal variance. The Bonferroni correction was applied for multiple comparisons. All images were prepared in Adobe Illustrator (Adobe Systems, San Jose, CA USA). Images were cropped and annotated for clarity only.

Immune-regulated genes are activated via the Imd but not the Toll pathway in Aag2 cells
To facilitate studies into molecular virus-immune interactions, we set out to develop RT-qPCR assays to measure the activation of the Imd and Toll pathways in Aag2 cells. We selected genes previously shown to be upregulated in vivo in A. aegypti by endogenous activation of NF-κB signalling, either through the transgenic overexpression of the transcription factors REL1A (Toll pathway) or REL2 (Imd pathway), or by transient dsRNA silencing of the negative regulator CACTUS (Toll pathway) (Fig 1B) [7,35]. Published data on Imd pathway activation via transient dsRNA silencing of the negative regulator CASPAR were not included in the analysis, as the authors of the original studies acknowledged that there was little overlap with REL2 overexpression data [7,35]. While components within immune pathways can be used as markers of pathway activation, these tend to be induced to a lesser degree than downstream secreted effectors [40] and were therefore excluded. In total we selected fourteen genes from these data sets for validation in Aag2 cells, including one defensin (DEFD), two cecropins (CECD, CECE), two serpins (currently annotated simply as 'SRPN'), two TEPs (TEP3, TEP22), one prophenoloxidase (PPO10), one C-type lectin (galactose-binding) (CTLGA8), and four 'hypothetical' genes with no functional or identity annotation (AAEL001508, AAEL003616, AAEL015527, AAEL021027) (Fig 1B). This panel of genes was chosen from our data mine to represent a broad range of proteins with known immune functions; CECE and PPO10 were selected for their known immune functions despite not being represented in the published in vivo data. Throughout this manuscript we refer to all genes by their current annotation and gene identifiers on vectorbase.org (A. aegypti strain Liverpool AGWG, gene set AaegL5.2) [36,38]. We also checked for the presence of predicted NF-κB binding sites in the promoters of these genes (Table 3).
Next, we tested the induction of the selected genes by RT-qPCR following treatment with heat-inactivated Gram-positive (Listeria monocytogenes) or Gram-negative (Escherichia coli) bacteria or the fungal PAMP zymosan, which are classic stimuli of insect NF-κB pathways [14]. The genes were induced to a greater or lesser extent by each of the stimuli, with up to 300-fold induction seen for some genes (Fig 1Ci and 1Di and 1Ei). Although several other studies have measured the expression of various immune-regulated genes by RT-qPCR in Aag2 cells [40][41][42], only one previous study validated which specific immune pathways these genes were regulated by, with validation only performed for stimulation with heat-inactivated E. coli [42]. We therefore tested the impact that transient silencing of REL1A or REL2 with dsRNA had on the induction of our putative NF-κB-regulated genes following stimulation with our panel of heat-inactivated bacteria or zymosan. Of the genes tested, nine were activated to markedly lower levels with all stimuli following silencing of REL2, while no impact was observed following REL1A silencing (DEFD, CECD, CECE, AAEL001508, AAEL008364 (SRPN), AAEL015527, CTLGA8, PPO10, TEP22) (Fig  1Cii and 1Dii and 1Eii; knockdown efficiency shown in Fig 2Ai and 2Aiii). These genes thus appear to be regulated by the Imd pathway in Aag2 cells. Four genes (AAEL003616, AAEL008208 (SRPN), AAEL021027, TEP3) were not clearly regulated by either pathway.
To validate a subset of the apparently Imd pathway-regulated genes in more detail, we verified the dose-dependency of the induction of DEFD, CECD, CECE, CTLGA8 and PPO10 (Fig 2Bi,  2Ci, 2Di, 2Ei and 2Fi). Both Gram-positive and Gram-negative bacteria stimulated robust expression of these genes, with up to 200-to 300-fold induction for DEFD and CECD, and approximately 40-to 50-fold induction for CECE and CTLGA8. We consistently saw lower induction for PPO10, and lower induction of all genes with zymosan compared to heat-inactivated bacteria. To more extensively test the NF-κB pathways through which these genes were induced, we repeated the stimulation in the presence of dsRNA targeting REL1A, REL2, MYD88 or IMD, with MYD88 and IMD acting earlier in the Toll and Imd signalling cascades respectively compared to the transcription factors REL1A and REL2. Once again, we consistently found that all genes were induced solely through the Imd pathway for all stimuli tested (Fig 2). Our findings are in line with recently published data by Zhang et al. that also found little effect of transient REL1A or MYD88 silencing on the induction of immune genes by E. coli in Aag2 cells [42].

REL1A-specific transcription factor binding sites are non-responsive in Aag2 cells
Given that we and others [42] only tested a limited number of genes and that there could still feasibly exist other Toll pathway-inducible genes in Aag2 cells, we turned to luciferase reporters to further investigate the responsiveness of REL1A-specific transcription factor binding sites in Aag2 cells. NF-κB reporters previously used in Aag2 cells [23] were based on D. melanogaster promoters that may not faithfully recapitulate the regulation of A. aegypti promoters. We therefore set out to develop NF-κB reporters based on transcription factor binding sites previously validated in A. aegypti [29].
First, we constructed plasmids in which firefly luciferase expression is driven by eight copies of REL1A-or REL2-specific NF-κB binding sites upstream of the D. melanogaster minimal Gene identifiers (Gene IDs) are from A. aegypti strain Liverpool AGWG, gene set AaegL5.2 (vectorbase.org) [36,38].
AaREL1A, sites responsive to A. aegypti REL1A; AaREL2, sites responsive to A. aegypti REL2; AaREL1A + AaREL2, sites responsive to both REL1A and REL2 of A. aegypti [29]. DmNF-κB, sites responsive to D. melanogaster NF-κB. HSP70 promoter element [28], which is present in the promoter of many insect expression plasmids (Fig 3Ai). We co-transfected Aag2 cells with each reporter and plasmids encoding mCherry or FLAG-tagged A. aegypti REL1A, REL2 or a truncated constitutively active REL2 construct lacking the inhibitory IκB domain (Fig 3Aii). None of the reporter plasmids were activated by the mCherry negative control (Fig 3Aiii-v). While the REL1A-specific reporter (1A-70-luc) was robustly induced specifically by REL1A, the REL2-responsive reporter (2-70-luc) was induced both by the REL2-ΔIκB construct and by REL1A (Fig 3Aiv and v). Since the control reporter lacking NF-κB binding sites (70-luc) was also induced by REL1A overexpression (Fig 3Aiii), we initially presumed this to be due to the activation of the minimal DmHSP70 promoter by REL1A. Only the REL2-ΔIκB construct and not full-length REL2 activated the REL2-responsive reporter (2-70-luc) (Fig 3Av), presumably because the presence of the IκB domain prevented the nuclear translocation of full-length REL2 in the absence of exogenous immune stimulation.
To overcome the presumed non-specific activation of the REL2-responsive reporter (2-70-luc) by REL1A, we generated equivalent reporters lacking the minimal DmHSP70 promoter (Fig 3Bi). Although the control reporter lacking NF-κB binding sites (luc) no longer responded to REL1A overexpression, and the REL1A-specific reporter (1A-luc) remained specific to REL1A stimulation, the REL2-responsive reporter (2-luc) was still induced by both REL1A and REL2-ΔIκB overexpression (Fig 3B). We therefore conclude that the previously reported REL2-specific NF-κB binding sites are not REL2-specific in Aag2 cells. We note that our reporters lacking the minimal DmHSP70 promoter (Fig 3B) were less inducible than the original reporters (Fig 3A). We therefore used our original REL1A-specific reporter containing the minimal DmHSP70 promoter (1A-70-luc) to test its inducibility by heat-inactivated Grampositive (L. monocytogenes, Staphylococcus aureus) or Gram-negative (E. coli) bacteria or zymosan. In line with our RT-qPCR data (Figs 1 and 2), we observed no Toll pathway activation by any of the tested stimuli using this luciferase reporter (Fig 3C). Indeed, we tested reporter induction over a range of times post-stimulation to exclude the possibility that our previous experiments had missed a potentially transient Toll pathway activation.
Finally, we sought to exclude the possibility that our REL1A-specific reporters (1A-70-luc, 1A-luc) were non-responsive due to the lack of binding sites for other transcription factors required for co-stimulation of the Toll pathway. The D. melanogaster Metchnikowin (Mtk) promoter is regulated independently by both the Toll and Imd pathways [43,44] and contains four NF-κB binding sites along with binding sites for other transcription factors that co-stimulate NF-κB signalling [30]. We generated a firefly luciferase reporter driven by one copy of a previously defined 348 bp enhancer region from the D. melanogaster Mtk promoter (Fig 4Ai) [30]. We also generated additional reporters in which the NF-κB binding sites in the Mtk enhancer were replaced with the same REL1A-or REL2-specific NF-κB binding sites used in our other reporters (Fig 4Ai). We first confirmed that the wild-type Mtk reporter (Mtk-wt-luc) was activated by both REL1A and REL2-ΔIκB overexpression (Fig 4Aii and iii). Replacing the NF-κB binding sites with REL1A-specific NF-κB binding sites (Mtk-1A-1-luc) made the reporter REL1A-specific (Fig 4Aiv). In agreement with our previous data (Fig 3), the Mtk reporter containing only 'REL2-specific' NF-κB binding sites (Mtk-2-luc) was activated by both REL1A and the REL2-ΔIκB construct (Fig 4Av). Nevertheless, we used this reporter to confirm that our manipulation of the Mtk enhancer region did not have a notable impact on the activation of Imd pathway-responsive reporters by E. coli or zymosan treatment (Fig 4B). In contrast, substituting the NF-κB binding sites for exclusively REL1A-specific NF-κB binding sites (Mtk-1A-1-luc) caused the reporter to become non-responsive to any of the immune stimuli tested (Fig 4C). We furthermore determined that even a reporter containing five copies of the REL1A-specific Mtk enhancer (Mtk-1A-5-luc) still failed to respond to immune stimulation (Fig 4Ai and 4D).
In summary, we conclude that under our experimental conditions the Toll pathway in Aag2 cells is not responsive to stimulation with a number of well-established immune stimuli. Therefore, our system is exquisitely specific for the Imd pathway, allowing the interactions of viruses with the Imd pathway to be investigated with high confidence.

Activation of REL2-regulated genes by viral PAMPs in Aag2 cells
We next investigated whether the expression of REL2-regulated genes in A. aegypti cells is activated by stimulation with dsRNA, a PAMP commonly found in cells infected with RNA viruses [45]. Of the five genes that we validated as REL2-inducible (DEFD, CECD, CECE, CTLGA8, PPO10), all were induced in a dose-dependent manner 24 h post-transfection with poly(I:C) (Fig 5A). Given that dsRNA is likely to be detected through different PRRs than bacterial or fungal PAMPs, we verified that poly(I:C) signals through REL2 using our wild-type and REL1A-responsive Mtk promoter reporters (Fig 5B). Similarly to bacteria and zymosan (Fig 4C and 4D), poly(I:C) induced luciferase expression from the wild-type, but not the REL1A-responsive, reporter, indicating that poly(I:C) is not activating the Toll pathway under our experimental conditions. Interestingly, poly(I:C) stimulated the expression of REL2-regulated genes both when it was transiently transfected into Aag2 cells and when it was simply added to the culture medium ( Fig 5C). However, the levels of gene induction were up to twenty-fold greater during transient transfection. While we found that there was some variability in the levels of REL2-regulated gene induction between genes and experiments, the observation that poly(I:C) induces REL2-regulated genes itself was highly reproducible.
To establish whether REL2-regulated genes can also be induced during RNA virus infection, we tested a number of viruses for their ability to infect Aag2 cells and induce AMP expression. We found that the insect-specific dicistrovirus cricket paralysis virus (CrPV), which is often used as a model virus in immune studies in D. melanogaster, replicates with rapid growth kinetics in Aag2 cells (Fig 5D). We detected the accumulation of CrPV RNA both intracellularly and extracellularly at high and low MOI, peaking by 6 hours post-infection, suggesting that CrPV replicates in Aag2 cells and releases viral particles into the extracellular medium. Furthermore, we detected the induction of DEFD, CECD, CECE and CTLGA8, but not PPO10, during CrPV infection (Fig 5E).
VAGO has been reported to be a virus-inducible cytokine regulated through REL2 that is involved in activating innate immune signalling in D. melanogaster and mosquitoes [46][47][48]. There are two VAGO genes in A. aegypti, VAGO1 (AAEL000200) and VAGO2 (AAEL000165) [49]. Despite extensive experiments and the reproducible induction of other REL2-regulated immune genes, we failed to detect the induction of VAGO2 during CrPV infection or stimulation with poly(I:C) (Fig 5F). We were unable to detect the expression of VAGO1 in either stimulated or unstimulated Aag2 cells.

Discussion
We developed a comprehensive and extensively characterised set of RT-qPCR and firefly luciferase assays to measure signalling through the Imd pathway in Aag2 cells. Importantly, the specificity of our RT-qPCR assays was verified by silencing genes involved in early signalling events within the Toll and Imd pathways as well as by silencing the downstream transcription factors. In addition, our luciferase reporter assays are based on transcription factor binding sites with previously verified specificity in A. aegypti [29]. Although we were able to develop REL1A-specific luciferase reporters, our REL2-responsive reporters were all also activated by . While the published NF-κB binding sites we based our reporters on specifically bound either REL1A or REL2 in an electrophoretic mobility shift assay (EMSA) [29], their specificity in A. aegypti cells was not reported. Other regulatory factors may therefore modulate the specificity of these NF-κB binding sites in Aag2 cells. Nevertheless, our REL2-responsive luciferase reporters still provide a specific read-out for the Imd pathway in Aag2 cells because we were unable to stimulate the Toll pathway under our experimental conditions.

Non-responsiveness of the Toll pathway in Aag2 cells
In our hands, the Toll pathway was not induced by Gram-positive (L. monocytogenes, S. aureus) or Gram-negative (E. coli) bacteria, the fungal PAMP zymosan or the viral PAMP dsRNA in Aag2 cells. Our data expand on and are in agreement with Zhang et al., who made similar observations with E. coli using their RT-qPCR assays [42]. In addition, we and Zhang et al. found that even genes considered by some to be specific markers of the Toll pathway, such as defensins, appear to be regulated by the Imd pathway in Aag2 cells (Figs 1 and 2) [42]. Furthermore, although Gram-positive bacteria and fungi are often said to be sensed by the Toll pathway while Gram-negative bacteria are said to be sensed by the Imd pathway, we found that all tested immune stimuli were capable of activating the Imd pathway in Aag2 cells. It is becoming more widely accepted that the regulation of AMP expression in Diptera is more complex than a model whereby individual genes are induced by specific stimuli through one defined NF-κB pathway under all conditions. For example, in D. melanogaster, certain AMPs have been shown to be primarily regulated through the Toll pathway in the fat body and through the Imd pathway in epithelia [18]. Cross-talk between NF-κB pathways has also been reported [19], and microarray analysis of genes induced after septic injury with E. coli or Micrococcus luteus revealed that most AMPs can be regulated by both the Toll and Imd pathways [15]. Our findings therefore further emphasise the need to carefully validate immune assays in each experimental system before drawing conclusions about specific pathways.
Why the Toll pathway appears to be non-responsive in Aag2 cells under our experimental conditions is unclear. We verified that all known signalling components of the Toll pathway downstream of SPÄ TZLE are present in the published Aag2 reference genome (vectorbase. org) [50], and that there are no obvious mutations that would render pathway components non-functional. Of specific note, while the PELLE sequence in the Aag2 reference genome is suggestive of a premature stop codon, we used PCR and Sanger sequencing to verify that this potential premature stop codon is not present in the PELLE sequence of our Aag2 clone. We have noted such sequencing errors and misannotations in the A. aegypti and Aag2 reference genomes before [51,52]. While it is feasible that there is some real deficiency in the Toll pathway in Aag2 cells, we instead suspect that secreted proteins involved in the sensing of extracellular bacteria and activation of SPÄ TZLE do not accumulate to sufficient levels in cell culture to facilitate Toll pathway activation. Similarly, the co-expression of SPÄ TZLE alongside bacterial stimulation has been shown to be required for optimal Toll pathway activation in D. melanogaster S2 cells [43].

Stimulation of REL2-regulated genes by viral PAMPs
Others have reported the sensing of virus-specific PAMPs by NF-κB pathways in A. aegypti [25]. Furthermore, DICER2 has been shown to sense infection with various RNA viruses in D. melanogaster and Culex spp. and Aedes spp. mosquitoes separately from its role in the RNAi pathway to stimulate the induction of VAGO via REL2 [46][47][48][49]. We extend these findings by demonstrating that a wide range of REL2-regulated genes are also induced by the general viral PAMP dsRNA and by CrPV infection in Aag2 cells. Our data indicate that dsRNA can be sensed by intracellular PRRs, however whether this sensing occurs via DICER2 in the context of REL2-regulated gene expression remains to be determined. Furthermore, the stimulation of REL2-regulated genes by treatment as well as by transfection with poly(I:C) is suggestive of additional extracellular or endosomal PRRs capable of detecting viral RNA in A. aegypti. This would mirror the situation in vertebrate systems where the endosomal membrane-bound TLR3 senses dsRNA [53]. A caveat to this interpretation of our data is that D. melanogaster cell cultures have been shown to take up dsRNA without transfection reagent [54], although this has not been reported for mosquito cell cultures.
In contrast to our data showing that CrPV infection induces AMP expression in Aag2 cells, arboviruses, including the flavivirus DENV and alphavirus Sindbis virus (SINV), do not induce AMP expression in Aag2 cells [40,41]. This is likely because flaviviruses and alphaviruses have evolved mechanisms to suppress immune signalling in their vector species [40,55]. In contrast, infection of Aag2 cells with CrPV is not a native virus-host combination and we hypothesise that CrPV therefore does not have the capacity to block REL2-regulated immune gene induction in Aag2 cells.
We have been careful here to distinguish between immune signalling through REL2 and the 'classical' Imd pathway characterised as a response to stimulation with bacterial PAMPs. Recently, gene expression by the D. melanogaster REL2 homologue RELISH has been shown to be stimulated by viral DNA through cGAS-STING signalling [56,57]. We therefore feel it is important to distinguish between the likely numerous and distinct signalling cascades converging on the transcription factor REL2 that respond to infection with different pathogens.

Lack of VAGO induction by viral stimulation in Aag2 cells
We found no evidence for the induction of VAGO1 or VAGO2 in Aag2 cells following CrPV infection or stimulation with dsRNA. To our knowledge, VAGO induction in A. aegypti cell cultures has also not been reported by other groups. The A. aegypti VAGO promoter reportedly contains NF-κB binding sites [48], although it is unclear which VAGO gene this refers to. Other groups have observed the transient induction of VAGO1, but not VAGO2, in A. aegypti infected with yellow fever virus (YFV), but neither VAGO1 nor VAGO2 were induced by DENV or West Nile virus (WNV) [58]. [49]. In cell culture, the dicistrovirus Drosophila C virus (DCV) has been shown to induce VAGO in D. melanogaster S2 cells, while DENV induces VAGO in Aedes albopictus RML-12 cells and WNV induces VAGO in Culex quinquefasciatus Hsu cells [47][48][49]. Furthermore, VAGO1 is reportedly antiviral against DENV in Aag2 cells infected with Wolbachia [49]. It seems unlikely that Aag2 cells have a specific defect in factors required for VAGO induction given that we show here that other immune genes can be induced via REL2 by viral PAMPs. While VAGO expression may be stimulated in a virus-, cell type-or vector-specific manner, the overall picture from our and other's data suggests that VAGO induction in A. aegypti is not a robust or consistent general response to viral infection.
Overall, our highly specific and well characterised tools for studying the Imd pathway in Aag2 cells, and our observation that dsRNA can induce signalling through REL2, pave the way for future studies defining molecular virus-immune interactions in A. aegypti that could ultimately inform the development of transmission-incompetent vectors to reduce the global arboviral disease burden.