In insects, InR, ilp, and their downstream signaling pathways have multiple evolutionarily conserved functions including larval development, regulation of feeding and immune response regulation (Grönke et al., 2010; Darby and Lazzaro, 2023). To demonstrate the importance of IIS in D. melanogaster during WNV-NY99 infection we knocked-down InR by RNAi and used mutant flies deficient in ilp7. Previous studies have found that these genes were similarly important for survival against the attenuated WNV-Kun strain (Ahlers et al., 2019). Unlike other fruit fly ilps, ilp7 is conserved in mosquitoes (Grönke et al., 2010). We infected control and InR knockdown or ilp⁷ mutant D. melanogaster with 2,000 PFU of WNV-NY99 by intrathoracic injection and monitored survival for 30 days. These results show that the knockdown of InR significantly increased the susceptibility of flies to WNV-NY99 infection ( Figure 1A ). Similarly, mutant flies lacking ilp⁷ were more likely to succumb to WNV-NY99 infection ( Figure 1B ).

To show the importance of insulin signaling during WNV-NY99 infection in insect cells, we treated multiple mosquito cell lines and D. melanogaster S2 cells with 1.7 μM bovine insulin prior to WNV infection (Esmann, 1963; Kang et al., 2008; Surachetpong et al., 2009; Xu et al., 2013; Ahlers et al., 2019). Similar to previous findings using WNV-Kun (Ahlers et al., 2019), we found that insulin treatment reduced WNV-NY99 viral load throughout infection ( Figures 1C–G ). We showed that the antiviral activity of IIS-mediated immunity is functional in mosquito cell lines from Culex ( Figure 1C ) and Aedes species ( Figures 1D, E ), as well as Drosophila S2 cells ( Figure 1F ). We next tested if ingested insulin reduced WNV-NY99 infection in adult D. melanogaster. Fruit fly food was supplemented with 10 μM bovine insulin or vehicle control. By 10 dpi, fruit flies in the insulin-fed group had lower levels of WNV compared to control flies ( Figure 1G ). Taken together these data show that insulin signaling reduces levels of the pathogenic strain of WNV in adult fruit flies and insect cells, supporting our previous results using WNV-Kun (Ahlers et al., 2019).

The effects of immunity on inhibiting virus growth can induce the development of viral evasion strategies (Tenthorey et al., 2022). For instance, the receptor binding domain (RBD) of SARS-CoV-2 was reported to develop escape mutations when the virus was cultured with anti-RBD antibodies (Greaney et al., 2021). Similarly, the envelope protein of HIV is known to develop mutations in conserved sites targeted by therapeutic broadly neutralizing antibodies (Dingens et al., 2017). Individual immune pathways exert unique selective pressures in comparison to these pathways working in concert (Mongelli et al., 2022). We hypothesized that serial passage of WNV during insulin stimulation would select for virus sequence variants, since insulin stimulates JAK/STAT immunity while suppressing the RNAi response (Ahlers et al., 2019). WNV infection alone was not sufficient to stimulate a JAK/STAT response (Ahlers et al., 2019). We performed serial passage of WNV-NY99 in Hsu cells under sustained insulin-mediated selection, with cells infected with virus from the previous passage for 10 passages. Cells were primed with media containing 1.7 μM bovine insulin or vehicle control 24 hours prior to infection with virus. Virus load was analyzed at each passage. We observed that insulin treatment significantly reduced the virus load in Hsu cells after 8-10 passages ( Figure 2 ). Viral titers under insulin treatment at early passages 1-3 and 5-7 were also signifcantly lower than titers in control cells at passages 8-10. These results suggested that the effects of insulin were greater as passage number increased, suggesting that treatment may be affecting the virus.

To test if sustained insulin-mediated selection was associated with changes in the WNV-NY99 genome, we performed Sanger sequencing of WNV genes E, NS2B, NS3, and NS5 in virus collected from infected cells from passages 5 and 9 ( Table 1 ). Previous studies have shown that these genes are prone to mutation in other circumstances of selection (Jerzak et al., 2007; Ebel et al., 2011). The proteins coded by these viral genes play important functions during the establishment of the disease, specifically virus entry, genome replication, and immunity, and they interact with the PI3K/Akt/ERK signaling axis (Scherbik and Brinton, 2010; Albentosa-González et al., 2021; Xie et al., 2021) that is downstream of insulin receptor signaling. The E protein is an important structural protein that is a major target of the host immune response (Arjona et al., 2007). The proper conformations of the E protein in many WNV variants, including NY99, require an N-linked glycosylation at residues 154-156. A study examining this glycosylation site reported that mice infected with WNV-NY99 mutants lacking this motif were less likely to develop the neuroinvasive disease (Shirato et al., 2004). The NS3 protein is a multi-functional molecule and plays many important roles in WNV infections. It has helicase activity, nucleotide triphosphatase activity and in conjunction with NS2B, it acts as a serine protease (Aleshin et al., 2007). This protease cleaves the other non-structural proteins from the viral polyprotein during infection. The NS2B/NS3 protease of dengue virus, another flavivirus, is known to disrupt IFN type I production (Setoh et al., 2017). The NS5 protein of WNV-NY99 is a potent antagonist of type I interferon activity. Research comparing the NS5 proteins of WNV-NY99 and WNV-Kun showed that the NY99 protein was 30-fold more effective at suppressing IFN-mediated JAK/STAT signaling (Laurent-Rolle et al., 2010). With Sanger sequencing, we examined sequences only for E, NS2B, NS3, and NS5. Future studies with RNA sequencing of the WNV genome will reveal mutations associated with other viral gene products.

In passage 5 and 9 virus, we did not detect any mutations in the NS2B and NS3 genes compared to the original stock genome. However, we did detect mutations in the envelope and NS5 genes in replicates of insulin-treated Hsu cells, with a single synonymous C to T transition at position 549 of the envelope gene that persisted from passage 5 to 9. Mutations were not observed in replicates of control cells nor did any mutation persist in these cells from passage 5 to 9. We calculated the codon usage as a fraction of frequency of amino acid (SNP) per thousand over frequency of amino acid (wild-type) per thousand in Cx. quinquefasciatus and H. sapiens. Synonymous codons are not used equally within organisms and the abundance tRNAs for their specific codons can affect translation rate, causing ribosome pausing while the less abundant tRNAs are recruited. This could also affect the timing and pattern of protein folding (Jiang et al., 2023). Codon usage fractions less than one represent mutant codons that are less abundant in the organism while values greater than one represent mutant codons that are more abundant (Jiang et al., 2023). Codon usage fractions were calculated using values from the Kasuza codon usage database (www.kazusa.or.jp/codon) (Nakamura et al., 2000). For example, we found that the codon frequency for envelope C1603A and NS5 G1861A mutations was ~2-5-times lower than wild-type ( Table 1 , columns 8-9), suggesting decreased tRNA codon availability for the transcript variant that could affect translation of the WNV polypeptide in both H. sapiens and Cx. quinquefasciatus. Additionally, we found that the codon frequency for the persistent envelope C549U mutation was 22% higher than wild-type in Cx. quinquefasciatus, suggesting that there is increased tRNA availability for this codon variant. Increased tRNA availability could also lead to increased translation of the WNV polypeptide and counteract insulin-mediated antiviral effects. Together, these results suggest that serial passage of WNV-NY99 in Cx. quinquefasciatus cells under insulin pressure induces mutations in the WNV genome that could affect virus replication under treatment. RNA sequencing in future studies could be used to determine proportional abundance of mutants found within a viral population.

We next examined human cellular and adult mosquito responses to control WNV or insulin-selected WNV. Given that flaviviruses stimulate Akt phosphorylation to promote cell survival and abrogate pro-apoptotic signaling (Dunn and Connor, 2012), we infected human liver HepG2 cells and adult Cx. quinquefasciatus with passaged WNV and used western blotting to quantify Akt activation (phosphorylation) in infected cells ( Figure 3 ) and mosquitoes ( Figure 4 ). This cell model is relevant based on our earlier work with HepG2 cells showing that endothelin signaling and associated Akt activation are induced by insulin stimulation to control WNV replication (Trammell et al., 2023). Furthermore, insulin receptor signaling in mosquitoes also activates Akt (Ahlers et al., 2019; Trammell et al., 2022). We compared Akt phosphorylation between HepG2 cells infected with control or insulin immune pressured (IIP) WNV. In passages 2 and 3 there was no difference in Akt phosphorylation between the treatment groups at 24 and 72 hpi. However, at passages 7 and 8, while HepG2 cells infected with control WNV displayed Akt phosphorylation at 24 hpi, cells infected with IIP WNV displayed lower Akt phosphorylation at 24 hpi ( Figure 3 ). In adult Cx. quinquefasciatus, there was reduced Akt phosphorylation at 3 and 10 dpi following infection with IIP WNV from passage 7 compared to control WNV. However, there was no difference in Akt phosphorylation following mosquito infection using control or IIP WNV from passage 2 ( Figure 4 ). Together, these results suggest that the insulin-induced WNV mutations due to serial passage affect Akt phosphorylation in human HepG2 cells and adult Cx. quinquefasciatus. It would be valuable to test at which passage WNV loses its ability to activate Akt in cells and mosquitoes. Future studies could also examine how insulin in Cx. quinquefasciatus bloodmeals affects replication of control WNV-NY99 and IIP WNV-NY99, as well as how these viruses affect mortality in D. melanogaster.

Our data from Hsu cells showed that serial passage of WNV-NY99 under insulin-mediated selection was associated with mutations in the WNV envelope and NS5 genes and altered host responses. Accordingly, we sought to examine host responses to infection under insulin selection in Cx. quinquefasciatus, the primary vector of WNV in North America (Rochlin et al., 2019). We infected female mosquitoes with WNV-Kun in the presence or absence of 170 pM bovine insulin in the bloodmeal (Ahlers et al., 2019). At 10 dpi, total RNA was collected from groups of five mosquitoes in each treatment group for RNA sequencing. Using these data, we performed Gene Set Enrichment Analysis (GSEA) to analyze gene expression patterns. Gene ontology (GO) categories related to host defense and immunity (Reid et al., 2018) were more enriched in WNV-Kun infected mosquitoes provisioned with 170 pM insulin relative to controls. For example, insulin fed mosquitoes showed significant enrichment of genes in the serine-type peptidase GO category ( Figure 5A ). Genes within this category have been linked to antiviral responses in mosquitoes against dengue virus (DENV). Multiple flaviviruses show antagonistic activity against the expression of a specific serine protease (Colpitts et al., 2011). Further, the oxidoreductase category includes genes that reduce Zika virus load in mosquitoes. This GO category is highly enriched in insulin-fed mosquitoes during WNV-Kun infection (Bottino-Rojas et al., 2018) as well as a similar GO category, response to oxidative stress, in D. melanogaster S2 cells treated with insulin during WNV-Kun infection (Trammell et al., 2023). We also analyzed the RNAseq data using a gene set derived from Cx. quinquefasciatus infected with WNV-NY99 (Paradkar et al., 2015) to generate a heatmap representing the fold change in gene expression compared to uninfected controls. We observed that a majority (60%) of genes were upregulated in infected mosquitoes that were blood fed 170 pM of insulin compared to vehicle control ( Figures 5B, C ; Supplementary Table 1 ).

We next used these RNAseq data from WNV-Kun-infected Cx. quinquefasciatus that were fed insulin or buffer control to map the data to the consensus sequence for WNV-Kun (GenBank Accession KX394396.1). We identified a non-synonymous mutation in the NS3 gene (K84I) of WNV-Kun following a single virus passage in insulin-treated mosquitoes ( Figure 6 ). The mutation was in the N-terminal protease domain close to ADP binding sites in the C-terminal helicase domain. Molecular dynamics simulations predicted that the K84I mutation affects ADP binding free energy (ΔΔG_bind,ADP) and the folding free energy of the NS3-ADP complex (ΔΔG_fold,ADP). We observed a small change in binding affinity (ΔΔG_bind,ADP = +3.8 kJ/mol) and observed a significant change in stability of the NS3-ADP complex (ΔΔG_fold,ADP = +46.5 kJ/mol). The K84 residue forms an H-bond with residue E43. A prior study examining the effects of amino acid changes on the stability of protein complexes suggested that ΔΔG values of ±8.4 kJ/mol and beyond represent a significant change. This value was determined by comparing free energy change predictions by simulations against experimental values (Patel et al., 2021). The K84I mutation lost this H-bond interaction which might affect the stability of the structure. This mutation was not detected in virus from infected control mosquitoes.

In summary, we demonstrated that IIS-mediated selection is associated with mutations in the WNV-NY99 genome both in vitro and in vivo. Additionally, insulin enhances the expression of immune genes that are predicted to contribute to host defenses against WNV. These promising findings encourage further investigation of the insulin-dependent antiviral pathways in the mosquito vector and how these shape the evolution of the WNV genome.

Baby hamster kidney-21 (BHK-21) (ATCC, CCL-10) and HepG2 (ATCC, HB-8065) cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and maintained at 37°C in 5% CO₂. Culex quinquefasciatus Hsu cells were cultured in L-15 media supplemented with 10% FBS and 10% triphosphate buffer. These cells were maintained at 28°C in 5% CO₂. West Nile virus-Kunjin strain MRM16 (WNV-Kun) was gifted by R. Tesh and propagated in Vero cells. West Nile virus strain 385-99 (WNV-NY99) was obtained from BEI Resources, NIAID, NIH (NR-158) and propagated in Vero cells. All experiments with a specific virus type utilized the same stock.

Knockdown of InR was achieved by crossing y¹v¹;P{TRiP.JF01482}attP2 (InR dsRNA cassette, Bloomington Drosophila Stock Center (BDSC) 31037) with y¹w*;P{Act5C-GAL4}25FO1/CyO,y⁺ (Actin driver, BDSC 4144). Sibling offspring not expressing the actin driver were used as controls. Ilp7¹ mutant flies (w¹¹¹⁸;TI{w^+mW.hs=TI}Ilp7¹ , BDSC 30887) were compared to w¹¹¹⁸ controls (BDSC 5905). Oregon-R-C flies (BDSC 5) were also used. Flies are negative for Wolbachia infection. Flies were kept on standard cornmeal food (Genesee Scientific #66–112) at a temperature of 25°C with 65% relative humidity and a 12-hour light/dark cycle. The adult flies used for experiments were female and aged between 2 to 7 days post-eclosion. Bovine insulin (Sigma 10516) was incorporated into the food to achieve a final concentration of 10 μM. Flies were intrathoracically injected with 2,000 PFU of WNV-NY99 or PBS vehicle control, as previously described (Ahlers et al., 2019).

Virus replication in cells and flies was measured using a standard plaque assay on BHK-21 cells as previously described (Mead et al., 2024). Hsu cell culture supernatant from three biological replicates per condition was collected at 5 days post-infection and stored at -80°C. At 5 and 10 days post-infection, three sets of five flies were collected from each treatment condition. Flies were homogenized in 50 μl Phosphate-Buffered Saline (PBS) and stored at -80°C. Cell culture supernatant or fly homogenate was serially diluted in DMEM with 2% FBS and plated on a 12-well plate of BHK21 cells at 1.5 × 10⁵ cells/well. Plates were incubated for 2 h at 37°C/5% CO₂ and rocked every 15 min. Wells were overlayed with 4% low melting point agarose (Invitrogen 16520050) for a final concentration of 0.75% agarose and 4% FBS in DMEM. Plates were incubated for four days at 37°C/5% CO₂ before visualization with 0.1% crystal violet (Fisher 548-62-9).

Hsu cells were seeded onto a 6-well plate at a density of 10⁶ cells per well and incubated overnight. At 24 hours post-seeding, the cell media in each well was aspirated and replaced with either 0 μM or 1.7 μM insulin media, depending on the treatment group. After 24 hours of incubation in the chosen media, the cells were infected with WNV at MOI 0.001 PFU/cell. The plates were incubated with virus for 2 hours before aspirating the infection media and replaced with 2% FBS L-15 media. Infected Hsu plates were incubated for 5 days before collection of cell media. Cells adhered to the plate were lysed in 300 μL of Trizol and stored at -80°C to preserve RNA.

HepG2 cells were seeded onto a 24-well plate at a density of 7.5 x 10⁵ cells per well and incubated for 48 hours. After the incubation period, the cells were infected with WNV at MOI 0.001 PFU/cell. The plates were incubated with virus for 1 hour before aspirating the infection media and replaced with 2% FBS DMEM media.

Protein lysate was collected by applying 100 μL of RIPA buffer (Ahlers et al., 2019) to each well of infected HEPG2 cells or by homogenizing five mosquitoes in 100 μL RIPA buffer. The concentration of protein lysate was assessed by BCA assay. 5 μg of protein was diluted in 2x Laemmli and heated to 95°C for 5 minutes. The prepared samples were analyzed by SDS-PAGE using 10% acrylamide gel followed by transfer to a PVDF membrane. Blots used to target Akt (1:2,000) (Cell Signaling 4691) were blocked in 5% non-fat dairy milk (NFDM) in Tris-buffered saline and 0.1% Tween (TBST) before application of primary antibodies. Blots used to target P-Akt (1:1,000; Cell Signaling 4060) or WNV envelope (1:250, Novus Biologicals NBP2-52709) were blocked in 5% BSA in TBST before application of primary antibodies. Blots were incubated with primary antibodies overnight at 4°C followed by 2-hour incubation with HRP-conjugated secondary anti-rabbit or -mouse IgG-HRP conjugate antibodies (1:10,000; Promega W401B, W402B) at room temperature. For actin blotting (1:10,000; Sigma A2066), PVDF membranes were stripped with gentle rocking for 30 minutes. The procedure for actin blotting was similar to Akt blots. Densitometry was performed using ImageJ.

RNA was isolated (ZymoResearch R2050) from three biological replicates of infected Hsu cells at passages 5 and 9 ( Figure 2 ) and 1 μg of RNA was used to prepare cDNA (BioRad 170-8891). The cDNA was used as the template for various PCR reactions targeting the 4 genes of interest - E, NS2B, NS3 and NS5 ( Table 2 ).

The genes coding for NS3 and NS5 were amplified in two parts by overlapping primer pairs. The primers targeting these genes were selected from a phylogenetic study of WNV-NY99 in Arizona which used an array of primers to sequence the whole virus genome (Hepp et al., 2018). The products from these PCR reactions were purified (NEB T1030S). The purified products were sent to Eurofins for Sanger sequencing. Forward and reverse sequencing reactions were prepared for the genes of each biological replicate. The forward and reverse sequence files were used to assemble complete gene sequences using the ‘Assemble contigs’ function in SnapGene (v7.1.1). These complete sequences were aligned to the genome of the virus used for the first passage and analyzed for mutations.

Female Cx. quinquefasciatus (BEI Resources NR-43025) were reared as described previously (Kauffman et al., 2017), with modifications. Briefly, mosquitoes were maintained under a 12 h light-dark cycle and at 28°C and 80% humidity. Adult mosquitoes were housed in a 1-gallon polypropylene Nalgene container with mesh screening and maintained on 10% sucrose-soaked cotton balls. For egg production, the adult females (~7 days old) were allowed to feed on defibrinated chicken blood overnight using an artificial mosquito feeder (Chemglass Life Sciences). Mosquitoes were provided with oviposition substrates, clear cups containing distilled water with soaked filter paper, 3 days following the blood meal. The collected egg rafts were washed into larval flats with distilled water and allowed to hatch in the same pans, with daily supplementation with fish food until pupation. Experimental mosquitoes (7 to 15 days old) were transferred to a 2.5 L polypropylene Nalgene container with mesh screening at least 24 h prior to the beginning of an experiment.

WNV-NY99 or -Kun was mixed with defibrinated chicken blood, resulting in a final concentration of 10⁷ PFU/mL. For WNV-Kun infections, the bloodmeal was supplemented with HEPES buffer or 170 pM insulin, as described previously (Ahlers et al., 2019). Female Cx. quinquefasciatus, bred for three generations, deprived of sucrose for 12 h, and then fed using an artificial mosquito feeder (Chemglass Life Sciences). Females were kept in 2.5 L cartons with constant access to 10% sucrose, and collected at 3 or 10 dpi for analysis. Females that did not feed were excluded from further analysis.

Mosquito tissues were lysed with gentle disruption in Trizol Reagent (ThermoFisher 15596), RNA was isolated by column purification (ZymoResearch R2050) and DNA was removed (ThermoFisher 18068). Following RNA extraction, library preparation and RNAseq was performed as previously described (Trammell et al., 2023). RNA sequencing was performed at the Spokane Genomics CORE at Washington State University-Spokane in Spokane, WA, USA.

Gene Set Enrichment Analysis (GSEA) was performed using RNAseq datasets and their associated p values as previously described (Subramanian et al., 2005). The GSEA used transcript levels for each experimental condition normalized to 0 µM insulin from the RNAseq analysis. GO categories were selected based on GSEA p value (p < 0.05).

The NS3 of WNV is made up of two domains: a helicase and a protease. The crystal structure of WNV-Kun helicase is available (PDB id-2QEQ) without bound form of the ADP nucleotide. To generate the full NS3 protein, a homology model of WNV-Kun helicase-protease was generated using the structure of NS3 protease-helicase from dengue virus (PDB ID-2WHX) using the Prime-structure prediction wizard of Schrödinger (2020-4). The conformation of nucleotide ADP was transferred to the homology model from NS3 protease-helicase from dengue virus and loops of the homology model were refined using Prime Module of Schrödinger (2020-4).

Desmond, a molecular dynamics (MD) engine of Schrödinger (2020-4), was used to execute the MD calculations for predicting the stability of WNV-Kun helicase-protease-ADP complex. Two separate MD simulations were performed: 1) WNV-Kun helicase-protease-ADP complex, 2) Mutant K84I WNV-Kun helicase-protease-ADP complex. The complexes were embedded in the TIP3P water model of orthorhombic box and were neutralized with the counter ions and physiological salt concentration kept at 0.15 M of NaCl. The system was specified on periodic boundary conditions, the particle mesh Ewald (PME) method for electrostatics, a 10 Å cutoff for Lennard–Jones interactions and SHAKE algorithm for limiting movement of all covalent bonds involving hydrogen atoms. OPLS3e forcefield was used to prepare the systems.

The energy was minimized using hybrid method of steepest descent (10 steps) and the LBFGS (limited-memory Broyden–Fletcher–Goldfarb–Shanno) algorithm with a maximum of 2000 steps with solute restrained, followed by similar energy minimization for 2000 steps without solute restraints. Then 12 ps simulation was carried out in NVT (Constant volume and temperature) ensemble for restraining nonhydrogen solute atoms at 10 K temperature was repeated and then followed by 12 ps simulation in the NPT (constant pressure and temperature) ensemble of temperature 10 K for restraining nonhydrogen solute atoms. About 24 ps simulation in the NPT ensemble were restrained with solute nonhydrogen atoms at temperature 300 K and 24 ps simulation in the NPT ensemble at temperature 300 K with no restraints. Berendsen thermostats and barostats were used to control the temperatures and pressures during the initial simulations. The relaxed system was subjected to simulation time of 50 ns with a time step of 2 fs in the NPT ensemble using a Nose–Hoover thermostat at 300 K and Martyna–Tobias–Klein barostats at 1.01 bar pressure. Every trajectory was recorded with a time interval of 100 ps.

MM-GBSA was calculated for MD trajectories of both complexes 1) WNV-Kun helicase-protease-ADP complex, 2) Mutant K84I WNV-Kun helicase-protease-ADP complex using the Prime module of Schrödinger (2020-4) to calculate binding affinity (ΔΔG_bind,ADP) of ADP. The difference between the former and latter provided the binding affinity change (ΔΔG_bind,ADP) of ADP due to the K84I mutation. Residue scanning of Schrödinger (2020-4) was performed on snapshots of MD trajectory of WNV-Kun helicase-protease-ADP complex, to check the stability (ΔΔG_fold,ADP) of the structure due to mutation K84I.

Results shown are representative of at least three independent experiments. Data points in dot plots and bar graphs represent a biological replicate of an individual well of cells ( Figures 1C–F , 2 ), a pool of five flies ( Figure 1G ), or an immunoblot ( Figures 3 , 4 ). Statistical analyses were completed using GraphPad Prism. Two-tailed unpaired t-tests assuming unequal variance were utilized to compare normally distributed pairwise quantitative data. Two-way ANOVA with Šidák correction was used to compare multivariate data. All error bars represent standard error of the mean. Survival curves ( Figures 1A, B ) represent three replicate experiments per condition pooled together and analyzed by the log-rank (Mantel-Cox) test using GraphPad Prism to determine P values between infected genotypes.