The anti-immune dengue subgenomic flaviviral RNA is found in vesicles in mosquito saliva and associated with increased infectivity

Mosquito transmission of dengue viruses to humans starts with infection of skin resident cells at the biting site. There is great interest in identifying transmission-enhancing factors in mosquito saliva in order to counteract them. Here we report the discovery of high levels of subgenomic flaviviral RNA (sfRNA) in dengue virus 2-infected mosquito saliva. We show that salivary sfRNA is protected in detergent-sensitive, protease-resistant compartments. Furthermore, we show that incubation with mosquito saliva containing higher sfRNA levels results in higher virus infectivity in human cells. Since sfRNA potently inhibits innate immunity in human cells, we posit that sfRNA in mosquito saliva is present in extracellular vesicles that deliver it to cells at the biting site to inhibit innate immunity and enhance dengue virus transmission.


Introduction 37
Half of the world's human population is at risk of infection with dengue viruses (DENV) 38 transmitted by Aedes aegypti mosquitoes (Pierson and Diamond, 2020). Since DENV have 39 evolved to harness mosquito biology to enhance transmission to humans, understanding the 40 transmission cycle of DENV is important for controlling the spread of this disease. Recently, it 41 has become clear that mosquito saliva enhances infection in human cells (Wichit et al., 2016), 42 mouse models (Cox et al., 2012), and non-human primates (McCracken et al., 2020). The 43 mechanistic underpinings for this enhancement are beginning to be understood (Jin et al., 2018;44 Sun et al., 2020), and here we present evidence for a novel potential transmission enhancer, the 45 subgenomic flaviviral RNA (sfRNA). 46 sfRNA is a non-coding RNA (ncRNA) produced from partial degradation of the flaviviral 47 RNA genome (gRNA) by host 5'-3' exoribonucleases that stall at secondary structures in the 3' 48 untranslated region (UTR) (Pijlman et al., 2008). Data from many groups demonstrate that 49 sfRNA downregulates immune responses in both human and insect cells (Bidet and 50 Garcia-Blanco, 2014;Slonchak and Khromykh, 2018). For instance, DENV serotype 2 (DENV2) 51 sfRNA inhibits the innate immune response by limiting interferon (IFN) expression (Manokaran 52 et al., 2015) and translation of IFN stimulated gene products (Bidet et al., 2014), promoting 53 DENV2 propagation in human cells. Moreover, Zika virus (ZIKV) sfRNA can modulate mRNA 54 decay and splicing and likely limit the effective response of cells to viral infection (Michalski et 55 al., 2019). In DENV2-infected mosquitoes sfRNA is expressed at high levels in salivary glands 56 and downregulates expression of Toll pathway-associated genes Rel1a and CecG (Pompon et al., 57 2017). In mosquitoes infected with ZIKV, sfRNA suppresses apoptosis in mosquito tissues and 58 this enhances virus dissemination and accumulation in saliva (Slonchak et al., 2020). Strong 59 sequential Proteinase K (PK) and RNase A/T1 treatments in saliva samples. To prevent PK from 106 digesting the nucleases, we inhibited the former with PMSF, which was dissolved in DMSO (see 107 Fig. 2 legend). In vitro-transcribed sfRNA1 added to uninfected saliva was digested by RNase 108 regardless of PK treatment (Fig. 2D). In infected saliva, we noted partial digestion of sfRNA by 109 RNase in the absence of PK treatment and this could be attributed to DMSO, which was used to 110 control for PMSF (compare Fig. 2B & 2E) and is known to destabilize some membranes (He et 111 al., 2012). It should be noted that gRNA was not sensitive to RNase regardless of treatment (Fig. 112 2F). This suggests that gRNA, unlike sfRNA, is packaged in a 113 which is expected for DENV virions (Lok et al., 2012). Importantly, sfRNA in infected saliva 114 was digested by RNase to the same extent whether or not samples were pretreated with PK ( Fig.  115   2E). This result suggested that sfRNA in saliva is not protected from nuclease degradation by 116

proteins. 117
While there is ample evidence that sfRNA has potent anti-immune action (Slonchak and 118 Khromykh, 2018), we tested how sfRNA levels influenced saliva-mediated infections of human 119 cells. To do this we took advantage of our observations that salivary sfRNA levels varied 120 between independent mosquito infections. We collected pools of saliva from mosquitoes infected 121 with three DENV2 strains PR6452, PR1940 and PR9963, and identified those that had similar 122 levels of gRNA copies but substantially different sfRNA levels and therefore had either low or 123 high sfRNA:gRNA ratios (Table 1). In four independent experiments, one with PR6452, two 124 with PR1940 and one with PR9963, we compared infection of human Huh-7 cells with saliva 125 with low or high sfRNA: gRNA ratios (Table 1). For each comparison between low and high 126 sfRNA: gRNA ratio saliva pools we infected Huh-7 cells with the equivalent number of viral 127 genomes in the same total volume of saliva, which is known to impact infectivity (Wichit et al., 128 2016). To achieve this we diluted infected saliva samples with saliva from uninfected mosquitoes 129 as described in Table 1. As a measure of infection, we quantified foci of viral genome replication 130 detected by immunofluorescence with an anti-dsRNA antibody (Fig. 3). The geometric mean of 131 replication foci per cell was 34.7 when cells were inoculated with PR6452-infected saliva with a 132 lower sfRNA:gRNA ratio vs 64.1 for saliva with a higher ratio (Table 1 and Fig. 3B, C). Very 133 similar results were obtained with PR1940-infected saliva pools. In repeat 1, lower 134 sfRNA:gRNA ratio resulted in a geometric mean of 48.5 foci per cell, whereas higher ratio 135 produced 78.8 foci per well (Table 1  Taken together, our results indicate that infection-enhancing sfRNA in infected saliva is 141 located in the lumen of detergent-and DMSO-sensitive vesicles. While it is formally possible 142 that sfRNA is packaged in DENV2 virions in the saliva, as proposed in a recent study (Syenina 143 et al., 2020), we do not favor this conclusion based on the differential DMSO sensitivity 144 observed for gRNA and sfRNA, and our previous data (Bidet et al., 2014). EVs are well known 145 vehicles of coding and noncoding RNAs, including viral RNAs (Nolte et al., 2016). In fact, EVs 146 derived from DENV infected mosquito cells have been isolated, characterized and shown to 147 contain DENV gRNA fragments (Reyes-Ruiz et al., 2019;Vora et al., 2018). Furthermore, EVs 148 derived from arthropod cells in culture have been shown to transmit viral genomes to human 149 keratinocyte-like HaCaT cells (Zhou et al., 2018). Based on these considerations, we propose 150 that sfRNA in saliva is present in EVs, which can deliver their cargo to cells in the same location 151 and at the same time as virus is deposited in the biting site (Fig. 4). sfRNA has been shown to 152 have potent anti-immune function in human cells (Bidet and Garcia-Blanco, 2014;Slonchak and 153 Khromykh, 2018) and here we show that higher levels of sfRNA in saliva correlate with 154 enhanced DENV2 infectivity; therefore, we posit that sfRNA delivered to human skin cells 155 during biting will dampen the local innate immune response and give the virus an advantage that 156 favors mosquito transmission. 157 The role of EVs in the battle between pathogens and hosts is not unprecedented. The 158 nematode Heligmosomoides polygyrus uses EVs to deliver miRNAs and other transactivators to 159 suppress Type 2 innate immunity (Buck et al., 2014). Similarly, hosts can transfer small RNAs to 160 silence virulence genes of pathogens (Cai et al., 2018). In this report, we propose that a viral 161 pathogen harnesses the biology of its invertebrate vector to increase transmission to a human 162 host. 163 164

Mosquito oral infection 180
Three-to five-day-old female mosquitoes were starved for 17 h and offered a blood meal 181 containing 40 % volume of washed erythrocytes from specific-pathogen-free pig's blood (PWG 182 Genetics), 5 % of 10 mM ATP (ThermoFisher Scientific), 5 % human serum (SigmaAldrich), 50 183 % of RPMI media and 10 7 plaque forming unit (pfu) per ml in all experiments. Blood meal titer 184 was validated using plaque assay. Mosquitoes were exposed to the blood meal for 1.5 h and fully 185 engorged females were selected and maintained with water and 10 % sugar solution until 186 analysis at 10 days and 15 days post blood meal in NGC and PR strains, respectively. 187 188

Northern blots 189
Northern blots were conducted using NorthernMax Kit (Ambion) with modifications to 190 manufacturer's protocol as described (Bischoff et al., 2004;Filomatori et al., 2017;Pompon et 191 al., 2017). Total RNA from 50 orally-infected mosquitoes was extracted using RNAzol RT 192  Table 1) and standard curves as detailed (Pompon et al., 213 2017). 214 To calculate the limit of detection (LoD), gRNA and 3'UTR/ sfRNA1 fragments 215 encompassing the qPCR targets and generated for the standard curves were 10-or 2-fold serially 216 diluted from 1.2 x 10 7 to 1.9 copies for gRNA and from 6 x 10 8 to 117 copies for sfRNA1. Four 217 technical repeats per dilution of three independent repeats were quantified using one-step 218 RT-qPCR. Fractions of positive replicates per copies of gRNA or 3'UTR/ sfRNA1 were plotted 219 against a sigmoid curve to identify LoD at 95% confidence (Forootan et al., 2017). 220 sfRNA1 copy number was calculated by subtracting the number of sfRNA1 and 3'UTR 221 fragments (estimated using the 3'UTR/ sfRNA1 primers) to the number of gRNA fragments 222 (estimated using the envelope primers). For infected samples that contained detectable amount of 223 sfRNA, sfRNA: gRNA ratio was calculated by dividing the number of sfRNA over the number

RNase resistance assay after detergent or proteinase treatments 241
Twenty saliva collected from orally-infected or uninfected mosquitoes were pooled and divided 242 into four equal volume. Each subset was used to test one of the four combinations of RNase with 243 either Triton X-100 or proteinase K (PK) treatments. In vitro transcribed sfRNA1 generated for 244 standard curve (see above) was added to each uninfected saliva subgroup. Samples were treated 245 with 0.1 % Triton X-100 (Sigma) at room temperature for 30 min before adding 3 µl of RNase 246 A/T1 (ThermoFisher Scientific) at 37 °C for 30 min. Samples were treated with 0.25 µg of PK 247 (ThermoFisher Scientific) at room temperature for 30 min and with 2 mM phenylmethylsulfonyl 248 fluoride (PMSF) (Sigma) at room temperature for 10 min to inhibit PK before adding 3 µl of 249 RNase A/T1 at 37 °C for 30 min. As PMSF was dissolved in DMSO (Sigma), 1% DMSO was 250 added as control when PMSF was not added. RNA was extracted using QIAamp viral RNA mini 251 kit (Qiagen) and sfRNA and gRNA were quantified. 252 253

Infection of human cells by incubation with infected mosquito saliva 254
3,000 Huh-7 cells were seeded in each well of a 12-well removable chamber slide (ibidi GmbH) 255 and grown in 200 µl of complete media. Pools of saliva from 40 mosquitoes orally-infected with 256 10 7 pfu of DENV2 per ml of blood as detailed above were collected in DMEM medium 257 supplemented with 0.5 mM ATP from independent oral infections. Media was removed from 258 Huh-7 cells and cells were incubated with mosquito saliva pools for two hours at 37°C at which 259 point saliva inoculum was removed and replaced with 200 µl of DMEM supplemented with 2% 260 FBS (See Table 1 Table 2) that processed the background at 50 pixels, adjusted the threshold to 269 remove the noise signal, and quantified particle size larger than 0.08 µm 2 .    Whole mosquitoes were collected at 10 days post oral infection with DENV2 NGC strain. 507 sfRNA sequences were obtained by RNA circularization, directed expansion of circularized 508 templates, cloning of expected-size amplicons and sequencing. Out of 27 clones sequenced for 509 each sfRNA1 to sfRNA4 expected-sizes, we found sequences corresponding to sfRNA1 and 510 sfRNA3, and one corresponding to a decay intermediate (gc370-4). 511