Multiple orthoflaviviruses secrete sfRNA in mosquito saliva to promote transmission by inhibiting MDA5-mediated early interferon response

Abstract

Numerous orthoflaviviruses transmitted through the bites of different mosquito species infect more than 500 million people annually. Skin infection at the bite site represents a critical and conserved step in transmission and a deeper understanding of this process will promote the design of broad-spectrum interventions. Here, we identify and characterize a transmission-enhancing viral factor in mosquito saliva that is shared across orthoflaviviruses. Saliva from West Nile virus-infected Culex and Zika virus-infected Aedes contains a viral non-coding RNA, subgenomic flaviviral RNA (sfRNA), within lipid vesicles distinct from virions. Higher concentration of sfRNA in infectious saliva positively correlates with infection intensity in human cells and skin explants. Early sfRNA delivery into transmission-relevant skin cell types and human skin explant demonstrate that sfRNA is responsible for the infection enhancement. Co-inoculation of sfRNA in a mouse model of transmission enhanced skin infection and worsened disease severity, supporting the role of salivary sfRNA as a transmission-enhancer. Mechanistically, salivary sfRNA attenuates early interferon response in human skin cells and skin explants by disrupting MDA5 signaling. Our results, derived from two distinct orthoflaviviruses and supported by prior studies, establish salivary sfRNA as a pan-orthoflavivirus transmission-enhancing factor driven by a conserved viral non-coding RNA.

Factors in mosquito saliva are required for efficient Orthoflavivirus transmission. The authors report a viral RNA fragment in saliva that reduce early innate immune response to enhance transmission.

Introduction

Numerous mosquito-borne orthoflaviviruses cumulatively infect half a billion people annually, resulting in an estimated 150,000 fatalities and economic losses exceeding €9 billion¹. Given the wide geographic distribution of Aedes and Culex mosquito vectors, nearly the entire human population faces the risk of infection^2,3. Furthermore, predictive models considering increasing urbanization, global transportation, and climate change anticipate exacerbation and broadening of the associated public health, economic, and social burdens⁴. More concerning is the very likely emergence of yet-unknown mosquito-borne orthoflaviviruses in the near future, further complicating the landscape of orthoflavivirus diseases that necessitate targeted interventions¹. However, safe and effective interventions to protect against orthoflaviviral diseases are lacking. While vector mosquito control stands as the most widely implemented strategy, its effectiveness in preventing epidemics is only moderate⁵. Scaling vector control approaches to cover extensive territories is also challenging, and the efficacy is hindered by the rapid development of insecticide resistance⁶. Additionally, there is no curative treatment⁷, and the available vaccines pose notable safety concerns^8,9. A promising approach to confront the rising threat posed by multiple orthoflaviviruses is the design of pan-orthoflavivirus interventions, achieved through the identification of targets that are conserved across orthoflaviviruses.

The transmission of mosquito-borne orthoflaviviruses occurs during a mosquito bite¹⁰ (although alternative transmission routes such as sexual and maternal transmissions are possible¹), when infectious virions present in the saliva are deposited in the epidermis and dermis^11,12. Subsequent productive infection of the skin is essential for transmission^13,14. However, infection initiation is hindered by a potent antiviral response in the skin^15,16. Viral double-stranded RNA (dsRNA), an intermediate of viral replication, is detected by retinoic acid-inducible gene I (RIG-I), melanoma differentiation-associated gene 5 (MDA5) or toll-like receptor 3 (TLR3)¹⁷. These three viral RNA sensing branches separately and synergistically activate biochemical cascades that converge onto the phosphorylation of interferon regulatory factor-3 (IRF3) and/or IRF7, which trigger the expression of type I interferon (IFN). IFN then triggers an antiviral state by inducing the expression of IFN-stimulated genes (ISGs) with anti-orthoflaviviral properties^18,19. Robust evidence supports that mosquito saliva modulates the antiviral innate immune response in the skin to promote orthoflavivirus infection^13,20–22. Nonetheless, identification of these anti-immune transmission-enhancing factors remains partial.

Recently, we discovered a mosquito saliva factor that augments saliva infectivity for dengue virus (DENV), the most prevalent Orthoflavivirus²³. DENV secretes a non-coding subgenomic orthoflaviviral RNA (sfRNA) with anti-immune properties^24–26 within salivary lipid exosomes to enhance saliva infectivity²⁷. Nonetheless, our prior study did not provide evidence of an effect on transmission and lacked a mechanistic understanding. SfRNA biogenesis results from the incomplete degradation of orthoflaviviral genomic RNA (gRNA) by 5’-to-3’ host exoribonucleases, which stall on 3’UTR structures, called exoribonuclease-resistant RNA (xrRNA)²⁸. Intriguingly, the process of sfRNA biogenesis is conserved across orthoflaviviruses^29–32, raising the hypothesis that salivary sfRNA secretion occurs for multiple orthoflaviviruses.

West Nile virus (WNV) and Zika virus (ZIKV) rank among the most important orthoflaviviruses in terms of global public health burden. WNV is the most widely distributed Orthoflavivirus³³ and elicits symptoms potentially resulting in severe neurological manifestations leading to mortality or persistent sequelae³⁴. Since 2000, WNV has infected an estimated 7 million people, causing 2700 deaths in the USA³⁵, and is annually responsible for over 100 fatalities in the EU³⁶. ZIKV, acknowledged as a pathogen of international concern by the World Health Organization (WHO) in 2016, has infected more than 2 million people since its emergence^37,38. While ZIKV symptoms typically present as mild, the infection poses substantial risks of congenital Zika syndrome in newborns of infected mothers and Guillain-Barré syndrome in infected adults. Importantly, WNV and ZIKV exhibit distinct transmission dynamics with WNV transmitted by various Culex mosquitoes, notably Cx. quinquefasciatus³⁹, whereas ZIKV is primarily disseminated by Aedes mosquitoes, predominantly Ae. aegypti^40,41.

Here, using WNV and ZIKV as Orthoflavivirus models, we investigated whether sfRNA secretion in mosquito saliva represents a pan-orthoflavivirus mechanism for enhancing transmission and elucidated how salivary sfRNA increases transmission. First, we detected sfRNA in the saliva of Culex mosquitoes infected with WNV and Aedes mosquitoes infected with ZIKV. Second, we found that salivary sfRNA from WNV-infected Culex and ZIKV-infected Aedes is packaged within lipid vesicles. Third, combining approaches with transmission-relevant human skin cell types, human skin explants and a mouse model of transmission, we demonstrated that salivary sfRNA enhances Orthoflavivirus transmission. Finally, we showed that early sfRNA exposure suppresses the initial IFN response by disrupting MDA5-mediated signaling.

Results

SfRNA is secreted in saliva from WNV-infected Culex and ZIKV-infected Aedes mosquitoes

WNV can produce up to four sfRNA species by stalling host exoribonucleases at distinct xrRNA structures⁴². To identify the sfRNA species in Cx. quinquefasciatus, we performed Northern blot (NB) on mosquitoes orally infected with WNV and observed a major band above 500 nt (Fig. 1a), as previously reported in Cx. pipiens mosquitoes⁴³ and corresponding to the predicted sfRNA1 species at 528 nt. We also noted a fainter band above 400 nt, which does not align with any predicted xrRNA, but was previously observed in monkey Vero cells⁴³. We then collected saliva and salivary glands (SG)—the site of saliva production—from Culex mosquitoes 10 days post oral infection and using absolute quantification (Supplementary Fig. 1a, b) detected WNV gRNA in 80% of SG with a geometric mean of 1.95 × 10³ copies per infected sample and in 60% of saliva with a geometric mean of 1.97 × 10² copies per infected sample (Fig. 1b). Using an optimized subtractive RT-qPCR protocol for absolute quantification of sfRNA (Supplementary Note; Supplementary Fig. 2a–d and Supplementary Fig. 3a, b), we detected sfRNA in 93% of gRNA-positive SG with a geometric mean of 1.96 × 10⁶ per infected sample and in 100% of gRNA-positive saliva with a geometric mean of 4.15 × 10⁵ copies per infected sample (Fig. 1c). To normalize sfRNA to the infection level (estimated by gRNA level), we calculated the ratio of sfRNA:gRNA and observed ratios in SG and saliva at 1005 and 1431, respectively (Fig. 1d). To gain deeper insights into sfRNA secretion mechanism, we further analysed the data of salivary sfRNA secretion. In SG, we observed a positive correlation between sfRNA and gRNA copies (Fig. 1e), indicating a relationship between sfRNA and its precursor in this organ, coherent with sfRNA biogenesis occuring within the SG. In contrast, in saliva, we identified a negative correlation between sfRNA and gRNA levels (Fig. 1f). These opposing correlations between gRNA and sfRNA in the SG and saliva are consistent with the two viral RNAs being secreted through distinct mechanisms. By analyzing paired SG and saliva samples from the same mosquito, we further examined correlations between the two compartments. The quantities of gRNA and sfRNA in saliva remained relatively constant and showed no significant association with their levels in the SG (Supplementary Fig. 4a, b). Collectively, our findings provide evidence that WNV secretes sfRNA into Culex mosquito saliva and underscore the pivotal role of the SG in the production of expectorated sfRNA.

Fig. 1. sfRNA is secreted in saliva from WNV-infected Culex and ZIKV-infected Aedes mosquitoes.

a Detection of WNV sfRNA in orally infected Culex mosquitoes by Northern blot. Ld., RNA ladder; Ctrl., in vitro transcribed folded WNV sfRNA1; Inf., WNV-infected Culex mosquitoes; Uninf., mock-infected Culex mosquitoes. b–d Quantification of gRNA (b), sfRNA (c), and the ratio of sfRNA:gRNA (d) in salivary glands (SG) and saliva from Culex mosquitoes orally infected with WNV. n = 20 SG biological replicates and 20 saliva biological replicates. e, f Correlations between gRNA and sfRNA quantity in SG (e) and saliva (f) from Culex orally infected with WNV. g–i Quantification of gRNA (g), sfRNA (h), and the ratio of sfRNA:gRNA (i) in SG and saliva from Culex mosquitoes micro-injected with WNV. n = 20 SG biological replicates and 20 saliva biological replicates. j Direct detection of WNV sfRNA using SL-RT-PCR in saliva from WNV micro-injected mosquitoes. Number for infected and uninfected saliva indicate sample number. k–m Quantification of gRNA (k), sfRNA (l), and the ratio of sfRNA:gRNA (m) in SG and saliva from Aedes mosquitoes micro-injected with ZIKV. Each point represents one pair of SG or saliva collected from one mosquito. n = 29 SG biological replicates and 29 saliva biological replicates. b–d, g–i, k–m Lines show geometric means ± 95% C.I. Dotted lines indicate LoD. sfRNA detection rate was calculated among gRNA-positive samples. e, f p values result from linear regression analysis. b–d, g–i, k, l p-values according to two-sided T-test. Souce data are provided as a Source data file.

To further support the role of the SG, we bypassed midgut infection by inoculating WNV into the thorax of Culex mosquitoes and collected SG and saliva at 7 days post-infection (dpi), i.e., earlier than after oral infection due to the shortened extrinsic incubation period. Direct injection resulted in 100% of SG and saliva samples positive for gRNA, with geometric means of 1.57 × 10⁷ and 2.72 × 10⁴ copies per sample, respectively (Fig. 1g). SfRNA was detected in 100% of infected SG, with a geometric mean of 1.40 × 10⁸ copies per sample, and in 80% of infected saliva, with a geometric mean of 2.09 × 10⁵ copies per sample (Fig. 1h). This corresponded to sfRNA:gRNA ratios at 8.9 in SG and 9.5 in saliva (Fig. 1i). As with samples from orally-infected mosquitoes, we observed a strong positive correlation between gRNA and sfRNA in SG (Supplementary Fig. 5a); however, the correlation was maintained in saliva (Supplementary Fig. 5b), likely reflecting conditions associated with micro-injection. Collectively, these results indicate that sfRNA secretion is an intrinsic feature of salivary gland infection rather than dependent on the route of viral entry.

To ascertain the secretion of sfRNA in saliva, we applied two additional methods. First, we conducted an NB on a pool of hundreds of mosquito saliva samples and detected a band of the expected size, but very faint (Supplementary Fig. 6). The inherent limits of minute volumes of saliva and the low resolution of NB are probably the causes of this unsatisfactory detection. Second, we designed a direct molecular detection assay for sfRNA by applying splinted-ligation reverse-transcription PCR (SL-RT-PCR), as for mRNA decay quantification⁴⁴. This method overcomes a major limitation in the field, where direct detection of sfRNA in low-input biological samples has remained challenging. Briefly, a DNA splint complementary to both an anchor RNA and the sfRNA sequences brings the two RNA molecules into close proximity, enabling T4 DNA ligase-mediated ligation of the monophosphorylated 5’ end sfRNA to the anchor RNA. After splint DNA degradation, the resulting chimeric RNA fragment is specifically RT-PCR-amplified using primers annealing onto the anchor RNA and sfRNA sequences. We validated the detection method with a synthetic 5’-end monophosphorylated fragment of sfRNA, amplifying a band of 98 bp as expected, and showed the absence of amplification without the sfRNA target or the splint fragment (Fig. 1j, lanes 1–3), although amplification occurred without ligase (Lane 4), indicating ligase-independent ligation. We further validated the specificity of the SL-RT-PCR assay by showing a lack of amplification when the synthetic monophosphorylated template encompassed the sfRNA target plus 10 nucleotides upstream of the 5’-end (Supplementary Fig. 7). We then applied the SL-RT-PCR to four pools of 10 saliva (samples 1–4) from WNV-infected Culex mosquitoes, containing 9.7 × 10⁴, 1.45 × 10⁶, 4.4 × 10⁵ and 2.25 × 10⁵ sfRNA copies for samples 1–4, respectively, as determined by substractive RT-qPCR method (Supplementary Fig. 3) and detected sfRNA in all samples (Fig. 1j, lanes 5–8). Notably, band intensity from the SL-RT-PCR correlated with sfRNA quantities determined with subtractive RT-qPCR. We repeated the molecular assay on the saliva pool sample 1 but omitting the splint fragment to confirm specificity (Fig. 1j, lane 9). Finally, three pools of 10 saliva samples from uninfected mosquitoes did not result in amplification (Fig. 1j, lanes 10–12), confirming the specificity of the direct molecular detection of sfRNA and thereby the presence of sfRNA in saliva.

To determine whether sfRNA secretion in saliva is conserved among orthoflaviviruses transmitted by different mosquito genera, we inoculated Aedes mosquitoes with ZIKV and collected SG and saliva at 7 dpi. In whole mosquitoes, NB showed the presence of a single sfRNA band at the expected size (Supplementary Fig. 8), as previously detected in Ae. aegypti⁴⁵. ZIKV gRNA was detected in 100% of SG and in 100% of saliva samples, with geometric means of 1.00 × 10⁷ and 1.63 × 10⁴ copies per sample, respectively (Fig. 1k; Supplementary Fig. 9a, b). We quantified ZIKV sfRNA using primers that annealed to both previously identified sfRNA species in Ae. aegypti⁴⁵ and an optimized sfRNA absolute quantification method (Supplementary Note; Supplementary Fig. 10a–c and Supplementary Fig. 11a, b). ZIKV sfRNA was detected in 100% of infected SG, with a geometric mean of 3.22 × 10⁸ copies per sample, and in 100% of infected saliva, with a geometric mean of 2.47 × 10⁵ copies per sample (Fig. 1l). This resulted in sfRNA:gRNA ratios of 32.2 in SG and 26.1 in saliva (Fig. 1m). As for WNV-infected Culex, ZIKV gRNA and sfRNA correlations were different between SG and saliva with a weaker correlation in saliva (Supplementary Fig. 12a, b). Collectively, the findings demonstrate that ZIKV, another Orthoflavivirus, secretes sfRNA in the saliva of Aedes mosquitoes, indicating that sfRNA secretion in saliva is a conserved mechanism among orthoflaviviruses vectored by different mosquito species.

Salivary sfRNAs from WNV-infected Culex and ZIKV-infected Aedes are within lipid vesicles

In our previous study, we provided biochemical and microscopic evidence that DENV sfRNA is expectorated within lipid vesicles distinct from virus particles²³. To determine whether this phenomenon is shared by other orthoflaviviruses, we assessed the nuclease resistance of salivary sfRNA from WNV-inoculated Culex mosquitoes following detergent-based disruption of lipid membranes (Fig. 2a). We showed that micrococcal nuclease (MNase) degraded in vitro transcribed WNV sfRNA (Fig. 2b) (RNase A/T1 was inefficient in degrading WNV sfRNA) and a viral RNA control fragment (Supplementary Fig. 13a) in the presence of Triton X-100 detergent and uninfected Culex saliva. Consistent with salivary gRNA and sfRNA being enclosed within lipid particles, neither of the viral RNAs were degraded by MNase unless pre-treated with Triton X-100 (Fig. 2c, d). To account for variability in the initial amounts of sfRNA and gRNA between the saliva pools, we normalized the RNA quantities within each saliva pool and observed consistent trends (Supplementary Fig. 14a–c). We then reasoned that if sfRNA and gRNA are enclosed in separate lipid particles, they would exhibit different sensitivities to the detergent treatment. Leveraging the inherent variability between biological replicates—potentially due to mechanical stresses from pipetting and liquid handling⁴⁶—we found a lack of correlation between sfRNA and gRNA levels after detergent and MNase treatments (Fig. 2e). This observation suggests that sfRNA is encapsulated in distinct lipid particles, separate from those containing gRNA.

Fig. 2. sfRNA in WNV-infected Culex saliva and ZIKV-infected Aedes saliva is protected from nuclease degradation by a detergent-sensitive layer.

a sfRNA inside a lipid bilayer is sensitized to nuclease degradation by Triton X-100 treatment. b–d Sensitivity to Triton X-100 and Micrococcal nuclease (MNase) treatment for in vitro-transcribed WNV sfRNA diluted in uninfected Culex saliva (b), for salivary WNV sfRNA (c), and for salivary WNV gRNA (d). Saliva was collected from WNV-inoculated Culex mosquitoes. n for (b) = 4 biological replicates. n for (c) and (d) = 9 biological replicates. e Correlation between sfRNA and gRNA quantities in WNV-infected Culex saliva after Triton X-100 and MNase treatments. f–h Sensitivity to Triton X-100 and RNase A/T1 treatment for in vitro-transcribed ZIKV sfRNA diluted in uninfected Aedes saliva (f), for salivary ZIKV sfRNA (g), and for salivary ZIKV gRNA (h). Saliva was collected from ZIKV-inoculated Aedes mosquitoes. n for (f) = 6 biological replicates. n for (g) and (h) = 7 biological replicates. i Correlation between sfRNA and gRNA quantities in ZIKV-infected Aedes saliva after Triton X-100 and RNase A/T1 treatments. Boxplots indicate median and lower and higher quartiles. Dots indicate saliva pool repeats. The dot colors indicate the different pools of saliva. u, uninfected saliva; i, infected saliva from the corresponding mosquito. b–d, f–h p-values according to post hoc Fisher’s LSD test. e, i p values result from linear regression analysis. Souce data are provided as a Source data file.

We repeated the nuclease resistance assay using saliva from ZIKV-inoculated Aedes mosquitoes. After confirming the degradation of in vitro-transcribed ZIKV sfRNA (Fig. 2f) and viral control RNA (Supplementary Fig. 13b) by RNase A/T1 (RNase), we observed that salivary sfRNA was partially protected from RNase degradation unless pre-treated with a detergent (Fig. 2g). Surprisingly, ZIKV sfRNA and gRNA in saliva were partially degraded by RNase both with and without detergent pre-treatment (Fig. 2h), potentially due to unintended physical stress on the virions during the one-day-long process from saliva collection to RNase treatment. Nonetheless, leveraging this unexpected sensitization, we observed a discrepancy between sfRNA and gRNA quantities (Fig. 2g, h), consistent with the two RNA fragments being enclosed in different particles. Similar trends were observed after normalizing RNA quantities within saliva pools (Supplementary Fig. 14d–f). Finally, we noted the absence of correlation between salivary sfRNA and gRNA in samples treated with both detergent and RNase (Fig. 2i)—especially clear for two samples with detectable level of sfRNA and non-detectable level of gRNA. Altogether, the analysis of saliva from two different mosquito genera infected with different orthoflaviviruses suggest the conservation of sfRNA secretion within salivary lipid vesicles across multiple orthoflaviviruses.

Salivary sfRNA increases WNV transmission

To elucidate the role of salivary sfRNA in transmission, we infected Culex mosquitoes with WNV by oral feeding, which introduces greater variability in the salivary sfRNA:gRNA ratio compared to thoracic inoculation (Fig. 1). Saliva pools were collected and categorized based on their sfRNA:gRNA ratio into low (ratio < 15), moderate (15 < ratio < 350), or high (350 < ratio) sfRNA concentrations (Fig. 3a; Supplementary Table 1). To validate the use of the saliva-collection media as an inoculum, we determined that Erioglaucine, used as a proxy for salivation in the saliva-collection media, did not affect cell viability (Supplementary Fig. 15). To support the normalization of saliva inoculum dose based on gRNA quantities, we showed that gRNA quantities were proportional to saliva infectiousness as measured by particle forming unit (PFU) (R² = 0.98; p < 0.001; Supplementary Fig. 16a) and that sfRNA concentration was not associated with the gRNA-to-PFU (PFU) ratio (Supplementary Fig. 16b), a measure for infectivity per virion. Saliva from the different sfRNA concentration groups were then used to infect Huh7 human hepatocyte cells, using volumes containing the same number of gRNA copies (Supplementary Table 1). Total saliva volumes were normalized with uninfected saliva. Quantification of WNV infection at 24 and 48 h post-infection (hpi) showed a progressive increase in infection with ascending categories of salivary sfRNA concentration (Fig. 3b, c) and a positive correlation with sfRNA:gRNA ratio (Fig. 3d). We repeated the saliva infection experiments with human skin explants (Fig. 3a). Injection of saliva containing higher sfRNA concentrations (Supplementary Table 2) led to increased WNV infection at 24 hpi (Fig. 3e, f).

Fig. 3. SfRNA in saliva increases WNV transmission.

a Pools of saliva were collected, categorized according to sfRNA:gRNA ratio and used to inoculate cells or human skin explants. b, c WNV infection in Huh7 cells at 24 (b) and 48 hpi (c) with low (L), moderate (M) and high (H) concentration of salivary sfRNA. Infection levels were normalized to L within time points. n per category = 2 biological replicates for 24 h and 4 biological replicates for 48 h. d Correlation between infection intensity (i.e., WNV gRNA fold change) in Huh7 and sfRNA concentration (i.e., sfRNA:gRNA ratio) in the corresponding saliva used for infection. e WNV infection in skin explants at 24 h post injection with low and high salivary sfRNA concentration. n = 3 biological replicates. f Correlation between infection intensity in skin explants and sfRNA concentration in the corresponding saliva used for infection. g Cells and skin explants were transfected and injected, respectively, with sfRNA or Control RNA (Ctl.) prior WNV infection. h–j WNV infection at MOI 0.005 in Huh7 (h), HFF1 (i), and U937 (j) cells at 24 and 48 hpi with WNV post sfRNA transfection. n for Huh7 at 24 hpi = 9 biological replicates and at 48 hpi = 6 biological replicates. n for HFF1 at 24 hpi = 6 biological replicates and at 48 hpi = 9 biological replicates. n for U937 at 24 and 48 hpi = 6 biological replicates. k WNV infection in skin explants at 24 hpi with WNV post sfRNA injection. n = 4 biological replicates. l Mice were intradermally co-injected with WNV and sfRNA or Ctl. RNA. SfRNA alone was injected as control. Skin biopsies were collected at day 1 and clinical signs and survival were recorded daily. n for skin biopsies = 9 biological replicates from 3 mice. n for clinical sign evaluation = 14 biological mouse replicates. m WNV infection in mouse skin at 24 h post injection. n = 9 biological replicates. n, o Clinical signs (n) and survival (o) for mice co-injected with WNV and either sfRNA or Ctl. RNA. n = 8 biological replicates. b, c, e Bars show mean ± s.e.m. h–j, k, m Bars show geometric mean ± 95% C.I. n Lines show mean ± s.e.m. Dots indicate repeats. b, c, h–j Different letters show significant differences (p < 0.05) according to post hoc Fisher’s LSD test. e, k p-values according to two-sided T-test. n p-values according to mixed-effect general linear model. o p-value according to Gehan-Breslow-Wilcoxon survival analysis. Souce data are provided as a Source data file.

We next investigated whether early presence of sfRNA enhances infection. First, we transfected Huh7 cells with in vitro-transcribed sfRNA shortly before WNV infection (Fig. 3g). Intracellular levels of sfRNA or a size-matched viral RNA control were comparable after transfection, averaging ~10⁸ copies per 250,000 cells (Supplementary Fig. 17a). To achieve a sfRNA:gRNA ratio approximating that observed in mosquito saliva (Fig. 1d), we employed a low inoculum (MOI 0.005), corresponding to 1250 PFU. Based on a gRNA:PFU ratio of 100—commonly observed in virus preparations from mosquito cells⁴⁷—this inoculum was estimated to contain 125,000 gRNA copies, resulting in a high sfRNA:gRNA ratio of 800. We observed that sfRNA transfection increased WNV infection at 24 and 48 hpi (Fig. 3h). Second, as orthoflaviviruses are thought to initially infect skin fibroblasts and monocytes following a mosquito bite^10,48, we repeated WNV infection post sfRNA transfection in HFF1 fibroblasts and U937 monocytes. While intracellular levels of transfected sfRNA and control RNA were equivalent (Supplementary Fig. 17b, c), infection was increased after sfRNA transfection at 24 and 48 hpi in both cell types, although the increase was lower in U937 (Fig. 3i, j). To rule out biases associated with low MOI—where few cells are initially infected—and the use of high sfRNA:gRNA ratio, we repeated the experiments with a higher MOI of 0.5 in all three cell types, resulting in an estimated very low sfRNA:gRNA ratio of 8. Notably, the pro-viral function of sfRNA was maintained under these higher inoculum conditions (Supplementary Fig. 18a–c). Third, we injected a mixture of WNV with either sfRNA or the control RNA into human skin explants (Fig. 3g) and observed that co-injection with sfRNA enhanced infection at 24 hpi (Fig. 3k).

Finally, we used a mouse model of transmission to evaluate the effect of sfRNA on infection and disease outcomes. Mice were intradermally injected with a small volume containing WNV and either sfRNA or the control RNA (Fig. 3l). As a non-infected control, mice were injected with sfRNA alone. At one day post injection (dpi), infection at the injection site in the skin was increased by co-injection with sfRNA (Fig. 3m). At 4 dpi, we confirmed successful systemic infection by detecting RNAemia (Supplementary Fig. 19), although this alone does not inform about disease severity^49–51. Over a longer observation period, we found that sfRNA co-injection aggravated disease severity (Fig. 3n), increased weight loss (Supplementary Fig. 20) and reduced mouse survival (Fig. 3o). Altogether, these results—obtained using infectious saliva with variable sfRNA concentrations and in vitro-transcribed sfRNA to mimic salivary delivery in in vitro, ex vivo and in vivo models—demonstrate that salivary sfRNA enhances transmission, worsening disease outcomes.

Salivary sfRNA mitigates the IFN response

To decipher how WNV sfRNA impacts the innate immune response, we quantified expression of IFN-β and three ISGs, namely CXCL10, MX1, and IFI6, in the Huh7 cell samples previously infected with saliva containing different sfRNA concentrations (same samples as in Fig. 3b–d). At 24 hpi, we observed a robust induction of IFN-β and the three ISGs in all samples treated with infectious saliva compared to mock-infected control (Fig. 4a–d). However, the levels of innate immune gene induction were similar across saliva samples irrespective of salivary sfRNA concentrations. Since cells infected with saliva containing higher sfRNA concentrations had higher viral loads (Fig. 3b) and previous studies have shown a correlation between viral load and intensity of IFN induction^52,53, these findings suggest that higher salivary sfRNA dampened the innate immune response at 24 hpi. In contrast, at 48 hpi, the expression of all four IFN-related genes correlated with the infection levels (Fig. 4a–d).

Fig. 4. Salivary sfRNA inhibits IFN responses.

a–d Expression of IFN-β (a), CXCL10 (b), MX1 (c), and IFI6 (d) in Huh7 cells at 24 and 48 h post infection (hpi) with WNV-infected Culex saliva containing different concentrations of sfRNA. Category of sfRNA concentration; L, low; M, moderate; and H, high. n per category, 2 biological replicates. The dotted line indicates the mock-infected values. e–h Expression of IFN-β (e), CXCL10 (f), MX1 (g), and IFI6 (h) in Huh7 cells at 24 and 48 hpi with WNV at MOI 0.005 post sfRNA transfection. n = 6 biological replicates. i–l Expression of IFN-β (i), CXCL10 (j), MX1 (k), and IFI6 (l) in HFF1 cells at 24 and 48 hpi with WNV at MOI 0.005 post sfRNA transfection. n = 6 biological replicates. m–p Expression of IFN-β (m), CXCL10 (n), MX1 (o), and IFI6 (p) in U937 cells at 24 and 48 hpi with WNV at MOI 0.005 post sfRNA transfection. n = 6 biological replicates. q–t Expression of IFN-β (q), CXCL10 (r), MX1 (s), and IFI6 (t) in skin explants at 24 hpi with WNV post sfRNA injection. Ctl., RNA control. n = 4 biological replicates. Bars show mean ± sem. Repeats are represented by dots. a–p Different letters show significant differences (p < 0.05) according to post hoc Fisher’s LSD test. q–t p-values according to two-sided T-test. Souce data are provided as a Source data file.

To determine whether sfRNA inhibits the innate immune response, we quantified the expression of the IFN-related genes in hepatocyte Huh7, fibroblast HFF1, and monocyte U937 cells that were transfected with sfRNA or the control RNA prior to WNV infection at MOI 0.005 and 0.5 (same samples as in Fig. 3h–j and Supplementary Fig. 18a–c). In Huh7 cells, sfRNA transfection inhibited the expression of IFN-β and IFI6 at 24 and 48 hpi for both inocula, while CXCL10 and MX1 expressions were diminished at 48 hpi with the low inoculum only (Fig. 4e–h; Supplementary Fig. 18d–g). In HFF1 cells, under low MOI infection, sfRNA strongly suppressed IFN-β expression at 24 hpi and reduced CXCL10 expression at both 24 and 48 hpi (Fig. 4i–l). Under high MOI conditions, all four markers of IFN response were downregulated at 48 hpi (Supplementary Fig. 18h–k). In U937 cells, sfRNA transfection under low MOI did not inhibit the expression of the IFN-related genes (Fig. 4m–p); however, all four IFN markers were downregulated at 48 hpi, with IFN-β also reduced at 24 hpi under high MOI conditions (Supplementary Fig. 18l–o). The lack of inhibition of certain IFN-related genes may reflect differences in the kinetics or intensity of gene expression among the different cell types.

Finally, we quantified the expression of IFN-related genes in human skin explants co-injected with WNV and either sfRNA or control RNA (same samples as in Fig. 3k). SfRNA moderately reduced IFN-β expression and significantly diminished CXCL10, MX1, and IFI6 expressions at 24 hpi (Fig. 4q–t). Together, these results demonstrate that salivary sfRNA dampens innate immune activation in the skin.

SfRNA alters MDA5 signaling of interferon

To determine whether the early presence of sfRNA increases infection by altering the IFN response, we chemically inhibited IFN induction using the TBK1/IKKε inhibitor MRT67307, and infected Huh7 cells with WNV following transfection with either sfRNA or control RNA (Fig. 5a). Successful inhibition of the IFN response was confirmed by the lack of induction for IFN-β, CXCL10, MX1, and IFI6 expressions following control RNA transfection and infection (Supplementary Fig. 21a–d). Supporting the hypothesis that sfRNA enhances infection by altering the IFN response, we observed that inhibition of IFN induction abolished sfRNA-mediated enhancement of infection at 24 and 48 hpi (Fig. 5b). The main pattern recognition receptors triggered by orthoflavivirus infection are RIG-I, MDA5 and TLR3⁵⁴. Since Huh7 cells used here do not express TLR3⁵⁵, we tested the role of RIG-I by using RIG-I-deficient Huh7.5 cells⁵⁶ (Fig. 5a). In Huh7.5 cells, the sfRNA-mediated enhancement of WNV infection was maintained at 24 and 48 hpi (Fig. 5c). Next, we silenced MDA5 in Huh7.5 to avoid any confounding effect from RIG-I (Supplementary Fig. 22a, b). Strikingly, while sfRNA-mediated enhancement of infection was observed following control siRNA transfection, the infection enhancement was abolished at 24 and 48 hpi when MDA5 was silenced (Fig. 5d). Together, these findings indicate that early presence of sfRNA promotes infection by altering MDA5 downstream signaling.

Fig. 5. sfRNA alters MDA-5 antiviral interferon response.

a Scheme of RIG-I and MDA5-mediated antiviral interferon response. The methods used to disrupt the antiviral response are written in red. b–d WNV gRNA fold change post sfRNA transfection at 24 and 48 hpi in Huh7 cells treated with MRT67307 (b), in RIG-I-deficient Huh7.5 cells (c), and in Huh7.5 cells upon MDA5 silencing (d). n for (b) and (d) = 6 biological replicates. n for (c) = 9 biological replicates. Ctl., RNA control. Bars show geometric mean ± 95% C.I. Repeats are indicated by dots. Different letters show significant differences (p < 0.05) according to post hoc Fisher’s LSD test. Souce data are provided as a Source data file.

Discussion

The incomplete understanding of orthoflavivirus transmission by mosquitoes hampers the identification of pan-orthoflavivirus targets that could be harnessed for broad-spectrum strategies to simultaneously protect against the multiple orthoflavivirus-related health threats. In this study, we identified and characterized sfRNA in mosquito saliva as a transmission-enhancing factor conserved across orthoflaviviruses. Using two phylogenetically distant orthoflaviviruses (WNV and ZIKV), belonging to distinct phylogenetic groups⁵⁷ and vectored by different genera of mosquitoes (Culex and Aedes mosquitoes), we showed that both orthoflaviviruses secrete sfRNA in salivary lipid particles. Combined with our previous discovery of DENV sfRNA in Aedes salivary vesicles²³ and our mechanistic characterization of sfRNA loading into salivary exosomes²⁷, the current study implies that sfRNA secretion in mosquito saliva is conserved across multiple, if not all, orthoflavivirus-mosquito species systems. We then designed a robust experimental approach to assess the function of salivary sfRNA in infection by infecting human cells and human skin explants with infectious mosquito saliva. This unprecedented dataset revealed that higher salivary sfRNA concentrations correlate with enhanced infection in human skin. We further demonstrated that sfRNA is responsible for the infection enhancement by reproducing this effect through the supplementation of in vitro-transcribed sfRNA in transmission-relevant cell types and human skin explants. To extrapolate from the in vitro and ex vivo findings and provide in vivo evidence for the role of salivary sfRNA in transmission, we used a mouse model of transmission. We established that sfRNA co-delivered with orthoflavivirus in the skin increases transmission and aggravates disease severity. Finally, we elucidated the mechanism by which the early presence of sfRNA enhances infection, demonstrating that sfRNA suppresses the early innate immune response and increases WNV infection by altering MDA5-mediated IFN induction. Altogether, our study establishes that multiple orthoflaviviruses secrete sfRNA in mosquito saliva to promote bite-initiated skin infection by inhibiting the innate immune response, thereby enhancing transmission.

Our results show that salivary sfRNA is shielded from nucleases by a detergent-sensitive layer, indicating the presence of a protective lipid membrane. These detergent-sensitive compartments could be either virions enclosed by a lipid bilayer envelope or extracellular vesicles (EVs). EVs are non-replicative, cell-derived membranous particles secreted into the extracellular space by most cells⁵⁷. We previously detected EVs in saliva of Aedes mosquitoes²³, while others observed EVs from mosquito cells^58,59. Multiple lines of evidence support that sfRNA is secreted inside EVs rather than virions, which by definition contain gRNA. Firstly, orthoflavivirus virions are spherical structures with a diameter of approximately 55 nm, which is too small to accommodate both gRNA and sfRNA^60–62. Secondly, we observed a divergence between sfRNA and gRNA quantities in SG and saliva, suggesting that sfRNA and gRNA produced in SG are not secreted through the same pathway. Thirdly, we showed that the particles containing sfRNA and gRNA differ in their sensitivity to nuclease degradation. Fourthly, the cytosolic origin of the EV lumen^57,63 would allow the loading of sfRNA. Indeed, sfRNA biogenesis, which results from gRNA degradation, occurs in the cytosol within processing bodies (PB)⁶⁴, a process supported in mosquitoes by the interaction of sfRNA with PB proteins such as Staufen^28,65. In contrast, virions are assembled in the ER and secreted via the trans-Golgi network⁷. Overall, several viruses exploit EVs to secrete diverse types of viral RNA fragments^66–69 and our integrated datasets with DENV, WNV and ZIKV in Aedes and Culex mosquitoes suggest that orthoflaviviruses load sfRNA into mosquito EVs for salivary secretion. A study we published during the revision further support the secretion of sfRNA inside salivary exosomes²⁷.

We observed that salivary sfRNA favors bite-initiated skin infection by mitigating the early innate immune response. During the initial stages of orthoflavivirus infection, the innate immune system mounts a potent antiviral response mediated by IFN and involving ISGs¹⁵. RIG-I and MDA5 are the two main viral RNA recognition receptors responsible for IFN activation via the IRF3/IRF7 phosphorylation signaling cascade⁷⁰. Loss-of-function studies for RIG-I and MDA5 have shown increased infection for WNV, ZIKV and DENV, establishing RIG-I and MDA5-mediated responses as major barriers to orthoflavivirus infection^17,71–74. We demonstrated that the infection enhancement by the early presence of WNV sfRNA was abrogated when MDA5 signaling was silenced, whereas RIG-I elimination did not affect sfRNA’s impact on infection. Our results indicate that salivary WNV sfRNA dampens the IFN response by mitigating MDA5 signaling in Huh7 cells, although the molecular target of sfRNA may vary across cell types. Consistent with sfRNA altering IFN upstream of IRF3/IRF7, one previous study reported that WNV sfRNA modulates IFN activation upstream of IRF3/IRF7 in murine embryonic fibroblasts (MEF)⁷⁵ and our findings show that chemical inhibition of IRF3/IRF7 activation abrogated the sfRNA effect. Alternatively, ZIKV sfRNA inhibits the IFN response by stabilizing NS5 inhibition of STAT1 phosphorylation—a step downstream of IRF3/IRF7⁷⁶. DENV sfRNA inhibits RIG-I activation by binding TRIM25 to prevent RIG-I ubiquitylation²⁶ and interacts with G3BP1, G3BP2 and Caprin1 to reduce ISG translation⁷⁷. Despite differences in mechanisms, sfRNA from multiple orthoflaviviruses share the ability to inhibit the innate immune response²⁴, pointing to adaptive convergence in flavivirus evolution. However, our mechanistic contribution to WNV sfRNA function differs from previous studies as we evaluated the early presence of sfRNA prior to infection establishment and the production of other viral IFN inhibitors⁷.

Salivary sfRNA emerges as the first transmission-enhancing strategy linked to an orthoflaviviral component. Our study provides evidence that co-delivery of sfRNA at the bite site increases skin infection and exacerbates disease severity. Previous studies have underscored the importance of skin infection in transmission^10,14, and our findings showing the potent immune inhibition of sfRNA in fibroblasts—the primary dermal cell types—emphasize the critical role of orthoflavivirus replication in stromal cells before viral migration through myeloid cells^25,78. Based on our data and those of others, we propose a model in which the early presence of sfRNA in saliva prevents the initial activation of innate immune response to facilitate the establishment of skin infection and amplify subsequent viral dissemination. Our findings suggest that this model is valid for mosquito-borne orthoflaviviruses.

We acknowledge several limitations of the study. While we propose that sfRNA secretion in salivary vesicles is a shared characteristic among orthoflaviviruses, our evidence is derived from the examination of only three orthoflavivirus species vectored by two distinct mosquito genera. Given that there are over 70 known orthoflavivirus species⁷⁹, including other highly pathogenic ones transmitted by mosquito genera not investigated here—such as for yellow fever virus, which is transmitted by Haemagogus mosquitoes⁸⁰—our findings may not universally apply to all mosquito-borne orthoflaviviruses. Although sfRNA biogenesis and its anti-immune functions are conserved across all orthoflavivirus groups tested to date^24,28,81, it remains to be determined whether salivary sfRNA secretion contributes to transmission enhancement for all mosquito-borne orthoflaviviruses. Additionally, our characterization of the lipidic vesicles containing sfRNA is limited. The small volume of mosquito saliva restricts the extent of biochemical and biophysical analyses typically performed on EVs³². Moreover, isolating sfRNA-containing vesicles from virions presents challenges due to the physical and chemical similarities between virions and lipidic vesicles⁶⁸.

Methods

Ethics statements

All the research comply with relevant ethical regulations. Animal protocols were approved by the national ethical committee from the Ministry of Higher Education and Research in France (permission numbers: 43466; 31273) and IRD ethical committee. Ethical approval for the use of human skin explant was provided by the French Ministry of Higher Education and Research to Biopredic (permission number: AV-2023-5431). Informed consent was requested.

Cells and viruses

Aedes albopictus C6/36 and Human hepatocyte Huh7 cells were obtained from Cell Lines Services GmbH, and Huh7.5 cells were kindly provided by Paul D. Biensiasz (The Rockefeller University, New York, NY, USA). Vero (CCL-81), HFF1 (SCRC-1041), and U937 (CRL-1593.2) cells were obtained from ATCC. Mammalian cell lines were grown in Dulbecco’s modified Eagle medium (DMEM) (Gibco), supplemented with 5% heat-inactivated fetal bovine serum (FBS) (Eurobio)—except for HFF-1 which was grown with 15% FBS—and 1% penicillin/streptomycin mix (Invitrogen) at 37 °C with 5% CO₂. Mosquito cells were grown in Roswell Park Memorial Institute (RPMI) media supplemented with 1% non-essential amino-acids (ThermoFisher Scientific), 10% FBS, and 1% penicillin/streptomycin mix at 28 °C with 5% CO₂.

WNV strain IS-98-ST1 (or Stork 98) was isolated from a stork in Israel in 1998⁸² and obtained from Dr. Philippe Desprès, Centre de Ressources Biologiques, Institut Pasteur, Paris. ZIKV strain H/PF13 was collected from human serum in French Polynesia in 2013⁸³ and obtained from the European Virus Archive-Global (EVAg). Viruses were amplified in C6/36 cells, titrated in BHK21 cells as described⁴⁰ and stored in aliquots at −70 °C.

Mosquitoes

Experiments with Culex mosquitoes were carried out using the Culex quinquefasciatus SLAB strain, collected in California, USA⁸⁴, and obtained from the Institut des Sciences de l’Evolution in Montpellier, France. Experiments with Aedes mosquitoes were carried out using the Aedes aegypti Bora-Bora strain collected in French Polynesia in 1980⁸⁵. Eggs from Cx. quinquefasciatus were hatched and reared at 25 ± 1 °C, whereas eggs from Ae. aegypti were reared at 27 ± 1 °C. Both species were reared at 70 ± 5% relative humidity and 12 h:12 h day:night. Larvae were distributed in plastic trays at a density of 200 individuals per tray and fed half a tablet of concentrated yeast and 1 g of TetraMin (Tetra) on the day of hatching, and then 1.5 g of TetraMin every two days until pupation. Pupae were transferred to cages and supplied with 10% sugar solution and water ad libitum.

Mice

Five- to seven-week-old C57BL/6J male mice (Mus musculus) were purchased from Charles-River (France) and housed in ventilated cages in NexGen Mouse 500 (Allentown; Serial number: 1304A0078) in the biosafety level 3 animal facility at MIVEGEC-IRD, Montpellier, France. 34 mice were maintained with a 17 h:7 h light/dark cycle, 53-57 % humidity, 20–24 °C temperature and provided with an irradiation-sterilized mouse diet (A03, SAFE, France) and sterilized water ad libitum. Upon arrival, mice were let to rest for one week before experiments.

Mosquito oral infection

Four-day-old female Cx. quinquefasciatus were starved for 12 h, and offered an infectious blood meal for 60 min at sunset time, using the Hemotek feeding system (Discovery Workshops) with chicken skin (obtained from spring chicken purchased in supermarket). The artificial blood meal contained WNV at 8 × 10⁵ plaque forming unit (PFU)/ml, 50% volume of washed erythrocytes from rabbit’s blood (animals housed in the BSL2 VectoPole animal facility, authorization number: H3417221), 25 mM ATP (ThermoFisher Scientific), 5% FBS (Eurobio-scientific) and RPMI to complete the volume to 2.5 ml. Engorged females were selected and maintained in the rearing conditions for 10 days before analysis.

Mosquito inoculation

Five-day-old cold-anesthetized female Cx. quinquefasciatus or Ae. aegypti were intrathoracically inoculated with 69 nl containing 100 PFU of WNV or 50 PFU of ZIKV, respectively, using Nanoject II (Drummond Scientific Company) and needles made from 1.14 mm O.D. glass capillaries × 1.75″ length (Drummond). Mosquitoes were then maintained in rearing conditions for 10 days before analysis.

Mosquito salivary glands dissection and saliva collection

Cold-anesthetized mosquitoes had their wings and legs removed before inserting individual proboscises into 20 µl filter-sterile tips containing 10 µl of RPMI (for RNase resistance assay) or of DMEM (for human cell infection with WNV) with 2% 25 mM Erioglaucine (Sigma-Aldrich) for 30 min at 25 °C. Mosquitoes with a blue abdomen (color of Erioglaucine) were considered to have salivated and the corresponding media were collected either individually or as pools. Salivary glands were dissected after salivation, labeled to identify the associated saliva and homogenized using Fast-Prep bead bitter homogenizer (MP) with glass beads.

sfRNA production

Full length in vitro-transcribed sfRNAs from WNV and ZIKV were generated by amplifying the sfRNA sequence from virus cDNA using the following primer pair with T7-tagged forward: for WNV, 5’-TAATACGACTCACTATAGGGAGTCAGGCCGGGAAGTTCC-3’ and 5’-AGATCCTGTGTTCTCGCACC-3’; for ZIKV, 5’-TAATACGACTCACTATAGGGTCTTAATGTTGTCAGGCCTGCTA-3’ and 5’-CCGCTATTCGGCGATCTGT-3’. Amplicons were reverse transcribed using Megascript T7 kit (Ambion), monophosphorylated at the 5’ end using RNA 5’ Polyphosphatase (Lucigen), extracted in RNase-free water (Invitrogen) using E.Z.N.A. Total RNA kit I (Omega) to remove the polyphosphatase and folded by slowly reducing the temperature from 95 to 4 °C with 5 mM of Mg²⁺, which regulates sfRNA folding⁸⁶. sfRNA quantity was estimated using nanodrop (ThermoFisher Scientific) and used to calculate copy number.

As control, a same-size RNA fragment (i.e., 483 nt) corresponding to a part of the dengue virus NS2 gene was in vitro-transcribed using the primers 5’-TAATACGACTCACTATAGGGGCAGCTGGACTACTCTTGAG-3’, 5’-GGTCCTGTCATGGGAATGTC-3’²⁸, monophosphorylated and folded as for sfRNA.

Northern blot

Northern blot was conducted using NorthernMax Kit (Ambion) with modifications to manufacturer’s protocol as described²³. Total RNA from 100 infected mosquitoes or from saliva samples from about 300 infected mosquitoes was extracted using TRIzol reagent (Invitrogen) and separated on a denaturing gel with 5% Acrylamide/Bisacrylamide 19:1 (Starlab) and 8 M Urea (Invitrogen). Biotinylated single-stranded RNA ladder (Kerafast) was also loaded on the gel. RNA was transferred onto a Hybond-N+ nylon membrane (Merck) using Trans-Blot Turbo (Bio-Rad) at constant 1.3 A for 30 min with Voltage ˂25 V. The membrane was UV-crosslinked, pre-hybridized and hybridized overnight with a biotin-16-dUTP (Roche) labeled dsDNA probe (1 mg) generated from the primers used to amplify the qPCR sfRNA targets for WNV and the primers used to generate full-length sfRNA for ZIKV. After washes, the membrane was blocked using Odyssey Blocking Buffer (LI-COR) and stained with IRDYE 800cw streptavidin (LI-COR) in TBS. Pictures were taken with ChemiDoc (Bio-Rad).

Absolute quantification of gRNA copies

Total RNA was extracted using E.Z.N.A. Total RNA kit I (Omega Bio-Tek). gRNA for WNV was quantified using iTaq Universal SYBR Green one-step RT-qPCR kit (Bio-Rad) with primers: 5’-ATTCGGGAGGAGACGTGGTA-3’ and 5’-CAGCCGCCAACATCAACAAA-3’; in LighCycler 96 thermocycler (Roche) with the following thermal profile: 50 °C for 10 min, 95 °C for 2 min, and 40 cycles at 95 °C for 15 s, 60 °C for 15 s and 72 °C for 20 s. gRNA for ZIKV was quantified using iTaq Universal Probes one-step RT-qPCR kit (Bio-Rad) with primers: 5’-TTGGTCATGATACTGCTGATTGC-3’ and 5’-CCTTCCACAAAGTCCCTATTGC-3’ and probe: 5’-CGGCATACAGCATCAGGTGCATAGGAG-3’; in LighCycler 96 (Roche) with the following thermal profile: 50 °C for 15 min, 95 °C for 2 min, and 45 cycles at 95 °C for 10 s and 60 °C for 30 s.

Absolute quantification for both gRNAs was obtained by generating standard equations using in vitro-transcribed RNA qPCR targets. RNA targets for WNV and ZIKV were produced by amplifying virus cDNA with primer pairs: 5’-TAATACGACTCACTATAGGGATTCGGGAGGAGACGTGGTA-3’/5’-CAGCCGCCAACATCAACAAA-3’, and 5’-TAATACGACTCACTATAGTTGGTCATGATACTGCTGATTGC-3’/5’-CCTTCCACAAAGTCCCTATTGC-3’, respectively. Amplicons were reverse transcribed using Megascript T7 kit (Ambion), extracted using RNeasy Mini Kit (Qiagen) and quantified using nanodrop (ThermoFisher Scientific) to calculate copy number. Serial dilutions were used to establish absolute standard equations. Using the gRNA templates, the limit of detection (LoD) at 95% was determined by calculating fractions of detected samples in three replicates of serial dilutions⁸⁷.

gRNA detection rate was calculated as the proportion of samples with detectable amount of gRNA among all samples.

Absolute quantification of sfRNA copies

sfRNA and 3’UTR were jointly quantified in total RNA extracts from E.Z.N.A. Total RNA kit I using the following primers for WNV: 5’-AGTTGAGTAGACGGTGCTGC-3’ and 5’-CCGTAGCGTGGTCTGACATT-3’, and for ZIKV: 5’-GCTGGGAAAGACCAGAGACT-3’ and 5’-CTATTCGGCGATCTGTGCCT-3’. Two different qPCR conditions were applied for each pair of primers. First, quantification for both viruses was conducted using iTaq Universal SYBR Green one-step RT-qPCR kit (Bio-Rad) in LightCycler with the following thermal profile: 50 °C for 10 min, 95 °C for 2 min, and 40 cycles at 95 °C for 10 s and 60 °C for 25 s. Second, quantification with WNV primers was conducted in the same conditions but for the reverse transcription step set at 60 °C for 20 min. Additionally, quantification with ZIKV primers was conducted using a two-step RT-qPCR. RT was conducted with the reverse primer and M-MLV enzyme (Promega) with a denaturation/annealing step at 70 °C and following manufacturer’s conditions with RT at 37 °C. qPCR was then performed in AriaMx thermocycler (Agilent) using EvaGreen qPCR MixPlus (Euromedex) with the following thermal profile: 95 °C for 15 min, and 40 cycles at 95 °C for 15 s, 60 °C for 20 s and 72 °C for 20 s.

Absolute quantification and LoD for sfRNA for WNV and ZIKV were obtained as described above for gRNA but using dilutions of sfRNA copies.

sfRNA copy number was calculated by subtracting the number of 3’UTR + sfRNA copies to the number of gRNA copies. For samples that contained detectable amount of gRNA, sfRNA:gRNA ratio was calculated by dividing the number of sfRNA over the number of gRNA. sfRNA detection rate was calculated as the number of samples with detectable amount of sfRNA only among gRNA-positive samples.

Splinted-ligation reverse-transcription PCR (SL-RT-PCR)

Two µl of RNA extract were mixed with 20 pmol of DNA splint oligo (5’-GCTGATGGCGATGAATGAACACTGCGTTTGCTGGCTTTGATGGAAAGTCAGGCCGGGAAGTTCCCGCCACCG-3’), 30 pmol of RNA anchor (5’-GCUGAUGGCGAUGAAUGAACACUGCGUUUGCUGGCUUUGAUG-3) in 8 µL total volume reaction using DEPC-treated water (ThermoFisher). Annealing was conducted at 95 °C for 5 min, 70 °C for 5 min, 60 °C for 5 min, 42 °C for 5 min, and 25 °C for 5 min. Ligation was conducted with 200 U of T4 DNA ligase (NEB) in 1X T4 DNA ligase buffer and 20 U of RNase OUT (Thermo Scientific) at 16 °C for 18 h. DNA digestion was performed by incubation with 5 U of RNase-free DNase I (Thermo Scientific) in 1X DNase buffer with MgCl2⁺ at 37 °C for 3 h. RT-PCR amplification was performed using one-step RT-PCR (Bio-Rad) and 300 nM of RNA anchor-matching forward (5’-GCTGATGGCGATGAATGAACACTGC-3’) and sfRNA-matching reverse (5’-GCAGGCAGCACCGTCTACT-3’) primers in total volume of 10 µl in LightCycler 96 (Roche) with the following thermal profile: 50 °C for 10 min, 95 °C for 1 min, followed by 40 cycles of 95 °C for 10 s, 60 °C for 25 s. Amplicons were run on 3.5% agarose gel in 0.5X TAE buffer, stained with GelRed, and visualized under UV exposure (Vilber).

As control, a 5’ end fragment of WNV sfRNA (5’-GAAAGUCAGGCCGGGAAGUUCCCGCCACCGGAAGUUGAGUAGACGGUGCUGCCUGC-3’) synthetized by Eurofins (EU) was monophosphorylated with 5 U of T4 ploynucleotide kinase (NEB), 1X Buffer, 1 mM ATP in 25 µl total reaction at 37 °C for 30 min and purified using EZNA Total RNA kit I (Omega). The sfRNA template plus 10 nucleotides was similarly produced and included the following nucleotides at the 5’-end: AUUUUGAGGA. Absolute quantification of RNA control. The same-size RNA fragment was quantified using one-step RT-qPCR with conditions identical to gRNA but using the following primers: 5’-TGGCAGTGGTGTCCGTTTCC-3’ and 5’-ATGCTCACCATCCCGACTGC-3’.

Absolute quantification was obtained by generating standard equations using in vitro-transcribed RNA qPCR targets of the same-size RNA fragment as detailed above.

RNase resistance assay

10⁷ copies of in vitro-transcribed sfRNA or of viral RNA control were mixed with 10 saliva samples from uninfected mosquitoes and subjected to RNAse resistance assay. Pools of 40 saliva samples from intrathoracically-inoculated mosquitoes were divided into 4 equal volumes and subjected to the different conditions of the RNAse treatment assay. Volumes were adjusted to 100 µl with PBS.

For WNV, samples were supplemented with 10 µl of 0.1% Triton X-100 (Sigma) for 30 min at 4 °C, before adding 0.5 µl of Micrococcal nuclease (MNase) (Thermo Scientific) and 10 µl of buffer (50 mM Tris-HCl pH 8; 10 mM CaCl₂ final concentration) for 1 h at 37 °C. Controls for Triton X-100 and RNase were added the same volume of PBS instead and maintained at the corresponding temperatures. RNA was then extracted using QIAamp viral RNA kit (Qiagen).

For ZIKV, the same treatments were applied except that 3 µL of RNase A/T1 (ThermoFisher Scientific) was added for 30 min at 37 °C instead of MNase.

Virus titration of saliva

100,000 Vero cells per well of 24-well plates were incubated with 125 μl of 10-fold serial dilutions of saliva-containing media for 1 h at 37 °C with 5% CO₂. After removing the inoculum, 1 ml of sterile 2% CarboxylMethyl Cellulose (CMC, Sigma-Aldrich) in DMEM supplemented with 2% FBS and 1% P/S was added. Four days later, cells were fixed with 4% paraformaldehyde (Sigma-Aldrich) for 20 min, covered with crystal violet solution (VWR) for 45 min, and rinsed. Plaques were counted using binocular microscope and plaque forming unit (PFU) per ml was calculated using the formula: number of plaques × 8 × dilution factor.

gRNA:PFU ratio was calculated by dividing the gRNA copy number quantified as detailed above to the PFU number within the same saliva.

Infection of human cells with WNV infectious Culex saliva

Pools of 10 saliva samples from Culex mosquitoes infected by oral feeding with 10⁸ PFU/ml of WNV were collected and stored at −70 °C. gRNA and sfRNA were quantified from a frozen aliquot of each pool. 5000 Huh7 cells per well of a 96-well flat-bottomed plate were inoculated with 1000 gRNA copies of the different saliva pools for 2 h. Total volume of the infected saliva was homogenized to 50 µl by adding uninfected saliva from Culex female mosquitoes. After infection, inoculum was removed and cells were supplemented with 2% FBS DMEM and collected at 24 and 48 h post inoculation by adding TRK lysis buffer and total RNA was extracted using E.Z.N.A. total RNA kit I.

Injection of WNV infectious Culex saliva into human skin explants

Five mm diameter Human skin explants obtained from Biopredic International (France) were decontaminated by washing in 25 mM HEPES (Gibco), 400 U/ml P/S and 10 µg/ml amphotericin B (Biowest) in DMEM. Explants were maintained individually in 8 µm cell culture insert (Thincert) with 200 µl of two-week skin culture medium w/o animal components (MIL215C, Biopredic) at 37 °C, 5% CO₂. Explants were inoculated with 760 gRNA copies of saliva pools. Total volume of the infected saliva was homogenized to 30 µl by adding uninfected saliva from Culex female mosquitoes. At 24 h post-injection, explants were cut into small pieces and, homogenized using the bead bitter homogenizer and glass beads before RNA extraction as detailed for mouse skin above.

Relative quantification of WNV gRNA and IFN-related genes in human cells and skin explants

RNA was reverse transcribed using the PrimeScript RT Reagent Kit (Perfect RealTime, Takara Bio Inc.). Real-time PCR reaction was performed in duplicate using Takyon ROX SYBR MasterMix blue dTTP (Eurogentec) on an Applied Biosystems QuantStudio 5 (Thermo Fisher Scientific) in 384-well plates. Transcripts were quantified using the following program: 3 min at 95 °C followed by 35 cycles of 15 s at 95 °C, 20 s at 60 °C, and 20 s at 72 °C. Values for each transcript were normalized to the geometric means of Ct values of 4 different housekeeping genes (RPL13A, ACTB, B2M, and GAPDH) using the 2^-∆∆Ct method. Primers used for quantification of transcripts by real-time qPCR were: RPL13A 5’-AACAGCTCATGAGGCTACGG-3’ and 5’-TGGGTCTTGAGGACCTCTGT-3’, ACTB 5’-CTGGAACGGTGAAGGTGACA-3’ and 5’-AAGGGACTTCCTGTAACAATGCA-3’, B2M 5’-TGCTGTCTCCATGTTTGATGTATCT-3’ and 5’-TCTCTGCTCCCCACCTCTAAGT-3’, GAPDH 5’-TGCACCACCAACTGCTTAGC-3’ and 5’-GGCATGGACTGTGGTCATGAG-3’, IFN-β 5’-TGCTCTCCTGTTGTGCTTCTC-3’ and 5’-CAAGCCTCCCATTCAATTGCC-3’, CXCL10 5’-CGCTGTACCTGCATCAGCAT-3’ and 5’-GCAATGATCTCAACACGTGGAC-3’, Mx1 5’-AAGCTGATCCGCCTCCACTT-3’ and 5’-TGCAATGCACCCCTGTATACC-3’, IFI6 5’-GGGTGGAGGCAGGTAAGAAA-3’ and 5’-GACGGCCATGAAGGTCAGG-3’. Primers for WNV gRNA were as described above.

Infection of human cells post sfRNA transfection

2 × 10⁵ Huh7, HFF1, U937, or Huh7.5 cells were transfected with 10¹⁰ copies of either monophosphorylated-folded-sfRNA or the control RNA using TransIT-mRNA Transfection Kit (Mirus) for 2 h. The transfection media was removed and cells were immediately infected with WNV at a multiplicity of infection, MOI, of 0.005 or 0.5 for 1 h. Cells were then maintained in 2% FBS media for 24 and 48 h before extracting RNA using E.Z.N.A. total RNA kit I.

Co-injection of sfRNA and WNV in human skin explants

Five mm diameter human skin explants as detailed above were inoculated with 10² PFU of WNV mixed with 10⁹ copies of monophosphorylated-folded-sfRNA or the control RNA in a total volume of 4 µl using a Nanoject II injector (Drummond) and glass needles. At 24 h post injection, explants were cut into small pieces and, homogenized using the bead bitter homogenizer and glass beads before RNA extraction as detailed for mouse skin above.

Co-injection of sfRNA and WNV in mice

Six- to eight-week-old mice were shaved with the animal trimmer (VITIVA MINI, BIOSEB) on the lower back one day prior to injection to limit shaving-induced inflammation. Mice anesthetized by intraperitoneal injection of 0.2 ml/mouse of 10 mg/ml of ketamine (Imalgène 1000, Boehringer Ingelheim Animal Health) and 1 mg/ml of xylazine (Rompun 2%, Elanco GmbH) were intradermally inoculated with 10³ PFU of WNV mixed with 10¹⁰ copies of monophosphorylated-folded-sfRNA or the same number of the control RNA. The total volume of inoculum was 4 µl. As an additional control, other mice were only injected with monophosphorylated-folded-sfRNA.

At 24 h post-injection, skin biopsies from the injection site were collected and homogenized in 340 µL of TRK lysis buffer using Fast-Prep bead bitter homogenizer (MP) set at 1.4 m/s for 60 s with 1 nm glass beads. Total RNA was extracted using E.Z.N.A. total RNA kit I. At 4 days post injection, blood samples were collected via mandibular puncture in another set of mice. RNA was extracted using QIamp viral RNA kit (Qiagen) and WNV gRNA per ml of blood was absolutely quantified after evaluating blood volume by pipetting. Clinical signs and weight were registered daily to determine a clinical score (CS) and calculate the percentage of weight loss until 14 days post-injection. CS of 0 was assigned to healthy mice; 1 for mice with ruffled fur, lethargy, hunched posture, no paresis, normal gait; 2 for mice with altered gait, limited movement in 1 hind limb; 3 for lack of movement, paralysis in 1 or both hind limbs; 4 for moribund mice; and 5 for dead mice⁸⁸. Mice were euthanized under anesthesia if they displayed neurological symptoms, severe distress, or weight loss exceeding 20%, or reached 12 days post-injection.

Chemical inhibition of interferon signaling

Media of 2 × 10⁵ Huh7 cells was supplemented with 10 µM of MRT67307 (MedChemExpress) diluted in water for 1 h, then during transfection of monophosphorylated-folded-sfRNA or the control RNA, and infection with WNV as described above.

MDA-5 silencing

2 × 10⁵ Huh7.5 cells were transfected overnight with 5 pmol of multiplex siRNAs (SI05130608, Qiagen) using Lipofectamine RNAiMax (ThermoFisherScientific). As a control, Allstars multiplex siRNAs (SI91027281, Qiagen) were similarly transfected. 48 h later, cells were transfected with monophosphorylated-folded-sfRNA or the RNA control and subsequently infected with WNV as described above.

Western blot

2 × 10⁵ Huh7.5 cells were lysed with 100 µl of RIPA (Biotech) supplemented with protease inhibitor cocktail (cOmplete, Roche). After centrifugation at 14,000 × g for 15 min at 4 °C, proteins were quantified in the supernatant using Qubit protein assay (ThermoFisherScientific). Normalized quantities of protein extracts were separated on NuPAGE 4–12% acrylamide gel (Invitrogen) and transferred onto 0.2 µm Nitrocellulose membrane using a Trans-Blot Turbo Transfer System (Bio-Rad). Membranes were blocked with 5% skimmed milk (ThermoScientific), incubated with 1:4000 anti-MDA5 antibody (21775-1-AP, Proteintech) or 1:400 anti-Actin (MA5-11869, Invitrogen) in blocking solution, washed with 0.05% tween-20 PBS, incubated with 1:500 anti-rabbit IgG HRP-linked antibody (Cell Signaling) in blocking solution, and washed. Blots were revealed using SuperSignal West Pico PLUS (ThermoScientific) and observed with ChemiDoc Imaging System (BioRad).

Statistical analysis

Comparison of gRNA and sfRNA copies, relative gRNA fold changes and IFN-related genes were conducted using two-sided T-test or post hoc Fisher’s LSD test. Absolute number and relative values of gRNA and sfRNA copies were log-transformed before statistical analysis. Correlation analysis was performed using linear regression analysis. Survival analysis were performed with Gehan-Breslow-Wilcoxon test. The statistical analyses were performed using GraphPad Prism v9.0.

The progression of clinical scores in mice was analyzed using a general linear mixed model with a negative binomial distribution to account for overdispersed data. The progression of weight change in mice was analyzed using a general linear mixed model with an inverse Gaussian distribution and log link function, appropriate for non-negative, right-skewed data. The model assessed the effects of days post-infection and experimental groups (WNV + sfRNA vs. WNV + Ctl.), including random intercepts for individual mice to account for variability across subjects. For pairwise comparisons between groups, post hoc analyses were conducted using estimated marginal means (emmeans) with Tukey’s adjustment for multiple comparisons. All analyses were conducted in R (version 4.3.3) using the lme4⁸⁹, glmmTMB⁹⁰, and emmeans packages^91,92.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: I.S.P. and J.P.; Formal analysis: I.S.P., H.M., and J.P.; Funding acquisition: J.P.; Investigation: I.S.P., H.M., L.B., K.J., L.P., N.S., M.T., J.Z., F.R., Z.R., Q.N., F.R.C., S.R., S.F., C.M., C.G., B.S., W.S., and E.F.M.; Methodology: I.S.P., H.M., A.G., D.R.S., R.H., D.M., S.N., and J.P.; Project administration: J.P.; Supervision: O.M. and J.P.; Visualization: I.S.P. and J.P.; Writing—original draft: I.S.P. and J.P.; Writing—review & editing: all authors.

Peer review

Peer review information

Nature Communications thanks Louis Lambrechts who co-reviewed with Shiho Torii, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

PhD scholarships for I.S.P. was provided by the Institut Méditerranéen Hospitalier (IHU, Marseille), and for F.R.C. and F.R. by the French Ministry of Research and Higher Education (CBS2 doctoral school, Montpellier). Post-doctoral fellowships from Fondation pour la Recherche Médicale to H.M. (SPF202110013925) and to E.F.M. (ARF202309017577). Support for the research was provided by the French Agence Nationale pour la Recherche (ANR-20-CE15-0006) and by the EU (HORIZON-HLTH-2023-DISEASE-03 #101137006) to J.P.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77288-4.