Engineered ZIKV strains recapitulate the phenotypes of natural isolates in mosquitoes
To understand the viral genetic factors impacting mosquito transmissibility, we selected, because of their distinct levels in transmission efficiency, recent ZIKV strains from Senegal and Thailand, representing the African and Asian lineages, respectively. To minimize the impact of natural genetic variability, we used reverse genetics to transform the original virus isolates, iSenegal and iThailand, into engineered strains, which we designated as rSenegal and rThailand, respectively (Supplementary Fig. S1A). To assess whether the strains generated by reverse genetics recapitulated the phenotypes of their natural counterparts, we exposed Ae. aegypti mosquitoes from Colombia to each of these viruses via infectious blood meals (Supplementary Fig. S1B). We chose a mosquito colony from Colombia to ensure consistency with our previous study, in which we found a significant difference in transmission efficiency between the Senegal and Thailand strains using the same mosquito colony. At 7 and 14 days post blood meal, we measured the prevalence of infection and systemic viral dissemination by RT-PCR and evaluated ZIKV transmission potential by detecting the presence of infectious virus in salivary secretions (Supplementary Fig. S2, Supplementary Table S1). Both rSenegal and rThailand exhibited a similar prevalence of infection, systemic dissemination, and viral presence in saliva compared to their natural counterparts. In line with our earlier study, both iSenegal and rSenegal demonstrated significantly higher dissemination prevalence and greater transmission efficiency than iThailand and rThailand across the tested time periods (Supplementary Fig. S2B, C). These findings confirm the comparability of the ZIKV strains generated by reverse genetics to the natural isolates in terms of their infection dynamics in mosquitoes. Consequently, we used rSenegal and rThailand as parental strains for all subsequent experiments (Supplementary Fig. S1B).
We first examined the growth kinetics of rSenegal and rThailand strains in mosquito (C6/36) cells in vitro, due to their ease of manipulation and robustness. Although this cell line is known to be deficient in the RNA interference (RNAi) antiviral response, a previous study has shown a limited impact of the RNAi response on ZIKV infection in Ae. aegypti-derived cells. Over the four-day time course, the rSenegal strain consistently showed higher titers compared to rThailand (Fig. 1A). To understand why these strains produced different levels of infectious particles, we performed several functional assays. We evaluated viral attachment by measuring membrane-bound viral RNA levels after one-hour viral attachment on ice. This experiment revealed that the rThailand strain had higher attachment efficiency than the rSenegal strain (Fig. 1B), thus failing to explain the superior growth kinetics of the rSenegal strain. We also analyzed viral internalization efficiency by measuring relative intracellular viral RNA at 3 h post infection (h.p.i.) normalized to the attached virus particles (Fig. 1C). The validity of the assay was verified with Dynasore, an inhibitor of clathrin-mediated endocytosis, which is considered the primary internalization pathway of orthoflaviviruses. Internalization efficiency was lower under Dynasore treatment (Fig. 1C, left panel) and was significantly higher for the rSenegal strain than for the rThailand strain (Fig. 1C, right panel), suggesting that differences in internalization efficiency contribute to the higher infectious particle production of the rSenegal strain. Next, we monitored the replication of ZIKV genomic and antigenomic RNA over 24 h using strand-specific RT-qPCR (Fig. 1D). Both strains showed detectable levels of ZIKV antigenomic RNA from 12 h.p.i., with higher replication efficiency in the rSenegal-infected cells detected at 18 h.p.i. (Fig. 1D, left panel). The relative viral genomic RNA level was also significantly higher in rSenegal-infected cells from 15 to 24 h.p.i. (Fig. 1D, right panel). This result indicates that a higher efficiency of viral genome replication by the rSenegal strain may also contribute to the higher production of infectious virus particles. We also explored the ratio of infectious virus titers to viral RNA copies to assess the release of non-infectious immature or defective virus particles (Fig. 1E). Although this ratio increased over time for both strains, it was significantly lower in rSenegal-infected cells at 48 h.p.i., suggesting a slightly higher proportion of immature or defective virus particles produced by the rSenegal strain. By monitoring the decay of infectious virus titers over time at 28 °C, we found that there was no detectable difference in the stability of infectious virus particles between the rSenegal and rThailand strains (Fig. 1F). Finally, we observed that the rSenegal strain significantly reduced cellular ATP levels compared to the rThailand strain at 96 h.p.i. (Fig. 1G). Together, these results indicate that the more efficient virus particle production of the rSenegal strain in mosquito cells reflects enhanced virus internalization and viral genome replication efficiency, despite increased cell toxicity.
To investigate the genetic determinants responsible for differences in mosquito transmissibility between the rSenegal and rThailand strains, we analyzed the full-length genome sequences of these strains (Supplementary Fig. S1C). Since comparative sequence analysis identified 102 amino-acid differences and 1216 nucleotide differences without specific clustering, we constructed a first set of chimeric viruses by swapping SPs, nSPs, and UTRs between the parental strains by reverse genetics (Fig. 2A). We observed significant differences in plaque size on a mammalian (Vero E6) cell monolayer between the chimeric viruses. Adding the rSenegal SPs into the rThailand backbone increased plaque size, whereas adding the rThailand SPs into the rSenegal backbone decreased it (Fig. 2B). The rThailand nSPs introduced into the rSenegal backbone led to smaller plaques but there was no detectable effect of the reciprocal replacement. In contrast, introducing rSenegal UTRs into the rThailand backbone reduced plaque size, while doing so in the opposite direction showed no significant change (Fig. 2B). These results suggest that SPs, nSPs, and UTRs, influence plaque size, although their effects vary depending on the backbone strain. We then assessed the growth kinetics of the chimeric viruses in mosquito cells. Chimeric viruses using the rSenegal strain as a backbone showed that introducing either SPs or nSPs from the rThailand strain significantly reduced virus titers compared to the parental rSenegal strain (Fig. 2C, left panel). For chimeric viruses using the rThailand strain as a backbone, replacement of the nSPs increased virus titers, particularly from 24 to 72 h.p.i. (Fig. 2C, right panel). These results indicate that swapping either SPs or nSPs influence growth kinetics but with a different magnitude depending on the backbone strain, whereas UTRs do not contribute significantly. To assess whether minor ZIKV genetic variants might influence the growth kinetics of chimeric viruses, we performed deep sequencing to estimate the frequency of minor variants in both the viral inoculum and in cells collected at 72 h.p.i. We examined three different chimeric viruses alongside the parental rSenegal strain (Supplementary Table S2). All minor variants detected exhibited similar frequencies in the inoculum and in the cells collected at 72 h.p.i., with less than a 10% change. These results do not support an influence of minor viral variants on the growth kinetics.
To identify the genomic regions underlying the distinct growth kinetics of the parental strains in mosquito cells, we examined the internalization efficiency of chimeric viruses. The only replacement that significantly influenced internalization efficiency relative to the parental strain was substituting rThailand SPs in the rSenegal strain (Fig. 2D). We next compared the replication efficiency of chimeric viruses by measuring the production of ZIKV antigenomic and genomic RNAs. Swapping nSPs between parental strains resulted in significant differences in the production of both viral antigenomic RNA (Fig. 2E, left panel) and genomic RNA (Fig. 2E, right panel), indicating that nSPs play a crucial role in viral genome replication. Replacing the SPs from the rThailand strain also decreased the production of viral antigenomic RNA (Fig. 2E, left panel). Finally, we found significant changes in cellular ATP levels following replacement of nSPs in both directions at 96 h.p.i. (Fig. 2F). Taken together, these findings show that differences in growth kinetics between the rSenegal and rThailand strains primarily reflect the effect of SPs on viral internalization and the effect on nSPs on viral genome replication. The nSPs from the rSenegal strain also lead to higher cell toxicity than the nSPs from the rThailand strain.
To investigate the viral genetic basis of transmissibility in mosquitoes in vivo, we first assessed the infectivity of the chimeric viruses when delivered through an infectious blood meal. We orally exposed Ae. aegypti mosquitoes from Colombia to differing virus doses as outlined in Supplementary Table S1 and estimated the 50% oral infectious dose (OID) for each chimeric virus from the dose-response curves (Supplementary Fig. S3). The rSenegal strain had a significantly lower OID estimate than the rThailand strain and introducing either the SPs or the nSPs from the rThailand strain into the rSenegal strain increased the OID estimates whereas the rThailand UTRs did not have a detectable impact (Supplementary Fig. S3B). Conversely, introducing the SPs, nSPs, or UTRs from the rSenegal strain into the rThailand strain did not significantly change the OID estimates. These findings suggest that both SPs and nSPs are required to confer the rSenegal strain a higher infectivity in mosquitoes relative to the rThailand strain. They also show that most chimeric viruses achieve 80-100% infection prevalence when the blood meal titer is >10 PFU/ml. Given this, we chose 2 × 10 PFU/ml as the standard oral infectious dose for subsequent experiments.
To assess the mosquito transmissibility of the chimeric viruses, we exposed Ae. aegypti mosquitoes from Colombia to each of the viruses via infectious blood meals containing 2 × 10 PFU/ml. Actual blood meal titers varied from 1.3 to 4.0 × 10 PFU/ml, and this variation was factored into our statistical analysis (Supplementary Table S3). At 7, 10, and 14 days post blood meal, we detected midgut infection and systemic viral dissemination by RT-PCR and evaluated transmission potential by detecting the presence of infectious ZIKV in salivary secretions (Supplementary Fig. S1B). At least 20 mosquitoes per time point were tested for each virus (Supplementary Table S1). We define infection prevalence as the proportion of blood-fed mosquitoes with a virus-positive body, dissemination prevalence as the proportion of infected mosquitoes with a virus-positive head, and transmission prevalence as the proportion of mosquitoes with a virus-positive head releasing infectious virus in their saliva (Supplementary Fig. S1B). Overall transmissibility is encapsulated in transmission efficiency, which is defined as the proportion of blood-fed mosquitoes with infectious saliva. As expected from the infectivity experiment (Supplementary Fig. S3A), the infection prevalence was 80-100% across viruses and time points (Fig. 3A). Dissemination prevalence was also 80-100% across viruses and time points (Fig. 3B). Both infection prevalence and dissemination prevalence were significantly influenced by the virus (Supplementary Table S3), however the magnitude of these differences was modest, ranging on average from 83.3% to 99.0% and from 84.5% to 96.8%, respectively (Fig. 3A, B). Transmission prevalence significantly increased over time and ranged from 0% to ∼50% across viruses and time points (Fig. 3C). It is possible that the observed increase in transmission prevalence over time reflects an increasing proportion of saliva samples exceeding the detection threshold of the assay, but this would remain biologically meaningful because it would indicate an increase in the infectious titer of saliva samples. Notably, the rSenegal strain resulted in significantly higher transmission prevalence than the rThailand strain. Introducing either the SPs or the nSPs from the rThailand strain into the rSenegal strain decreased transmission prevalence, whereas the rThailand UTRs did not have a detectable impact. Conversely, only the SPs from the rSenegal strain increased transmission prevalence when introduced into the rThailand strain. These findings indicate that differences in transmission prevalence between the parental strains are primarily determined by the SPs and, to a lesser extent, the nSPs.
To narrow down the genome regions responsible for differences in mosquito transmissibility between the rSenegal and rThailand strains, we constructed a second set of chimeric viruses by substituting each viral gene in the rSenegal strain with the corresponding gene from the rThailand strain (Supplementary Fig. S4A), and a third set by substituting each viral gene in the rThailand strain with the corresponding rSenegal gene (Supplementary Fig. S4B). Introducing specific genes such as prM, E, or NS1 from the rThailand strain into the rSenegal strain resulted in smaller plaques on Vero E6 cells, while replacement of NS1 and NS4 led to larger plaques (Supplementary Fig. S4C). Conversely, all chimeric viruses of the third set formed smaller plaques than the parental rThailand strain, indicating an asymmetric influence on plaque size (Supplementary Fig. S4D). We then compared the growth kinetics of the new sets of chimeric viruses in mosquito cells with that of the parental strains. In the second set, all replacements significantly reduced infectious viral titers except C and NS1 (Supplementary Fig. S4E). In the third set, replacement of C, E, or NS5 transiently increased virus titers at 48 and 72 h.p.i. but resulted in lower titers than rThailand at 96 h.p.i., while replacing NS2 led to a significantly slower growth (Supplementary Fig. S4F). Taken together, these results show that substitutions of E or NS5 most significantly impact virus growth kinetics in mosquito cells, but no single-gene substitution significantly alters the efficiency of infectious particle production in both directions.
To assess how individual viral genes affect mosquito transmissibility, we exposed mosquitoes from Colombia to the second and third sets of chimeric viruses via infectious blood meals containing 2 × 10 PFU/ml (Supplementary Fig. S1B). Actual blood meal titers varied from 0.7 to 2.5 × 10 PFU/ml and from 1.2 to 2.1 × 10 PFU/ml for the second and third sets, respectively, and this variation was factored into our statistical analysis (Supplementary Table S3). At least 10 and 16 mosquitoes per time point were tested for each virus for the second and third sets, respectively (Supplementary Table S1). Although we observed slight differences, likely due to variations in mosquito generations and experimental conditions, we confirmed that the confidence intervals of the parental strains overlapped across all experiments. Virtually all mosquitoes exposed to the second set of chimeric viruses were infected and had a disseminated infection, with no significant variation among viruses (Fig. 3D, E). Transmission prevalence increased over time, reaching 29-71% on day 14 (Fig. 3F), with a marginally significant effect of the virus (Supplementary Table S3). Following exposure to the third set of chimeric viruses, both infection prevalence and dissemination prevalence were significantly influenced by the virus (Supplementary Table S3), however the magnitude of these differences was modest, ranging on average from 78.7% to 98.5% and from 79.6% to 95.5%, respectively (Fig. 3G, H). Transmission prevalence significantly increased over time but was not influenced by the virus (Fig. 3I; Supplementary Table S3). Together, these results show that no single viral gene is solely responsible for the observed variation in transmission prevalence in mosquitoes from Colombia.
To investigate if the observed differences in transmission prevalence between chimeric viruses (Fig. 3C) were specific to the mosquitoes from Colombia, we tested the first set of chimeric viruses in two other mosquito colonies with different genetic backgrounds and contrasting levels of ZIKV susceptibility. ZIKV susceptibility is known to be significantly higher in globally invasive populations of Ae. aegypti found outside Africa than in native African populations. We chose a ZIKV-resistant mosquito colony from Uganda representative of the native African populations, and a colony from Cape Verde with admixed genetic ancestry and an intermediate level of ZIKV susceptibility. We challenged mosquitoes from Uganda and Cape Verde with infectious blood meals containing 1 × 10 PFU/ml and 2 × 10 PFU/ml, respectively, to maximize infection prevalence in these relatively resistant colonies. Actual blood meal titers varied from 3.0 to 8.0 × 10 PFU/ml and from 0.7 to 4.0 × 10 PFU/ml for the mosquitoes from Uganda and Cape Verde, respectively, and this variation was factored into our statistical analysis (Supplementary Table S4). In both mosquito colonies, the virus significantly influenced transmission efficiency (Supplementary Table S4). The rSenegal strain resulted in significantly higher transmission efficiency than the rThailand strain and swapping either SPs or nSPs -- but not UTRs -- between the parental strains resulted in intermediate transmission efficiencies that were consistent with the patterns previously observed with the mosquitoes from Colombia (Supplementary Fig. S5). Overall, the influence of SPs on transmission efficiency was more pronounced than the influence of nSPs. These results indicate that the viral genetic determinants of ZIKV transmission efficiency are largely independent of the mosquito genetic background.
To understand the mechanisms behind the varying mosquito transmissibility of chimeric viruses in mosquitoes, we developed a stochastic model of in vivo viral dynamics that could reproduce the qualitative outcomes of oral exposure to the chimeric viruses, capturing patterns of infection, dissemination, and transmission. Following an infectious blood meal, the virus first infects and replicates within the midgut epithelial cells, then 'escapes' from the midgut and disseminates via the hemocoel to other tissues. It ultimately reaches the salivary glands, where it is released in the saliva and transmitted to the next host. We aimed to use the stochastic model to pinpoint the key processes within mosquitoes that are most likely influenced by differences among the chimeric viruses. The output of our logistic model of in vitro viral growth kinetics (Fig. 4A) was consistent with our time-course analyses of infectious particle production using the first set of chimeric viruses when the maximum titer supported by the cell culture (carrying capacity, k) was fixed at 10 PFU/ml, the starting virus concentration (s) was either 10 or 10 PFU/ml, and the growth rate (r) was either 0.035 or 0.05 h, respectively (Supplementary Fig. S6A). We applied this viral growth model to capture in vivo mosquito infection dynamics, incorporating additional parameters: the probability of midgut infection (β), representing the likelihood that the virus successfully infects the midgut; the escape rate (λ), denoting how quickly the virus transfers between different tissues; and the blood meal clearance rate (μ), indicating how rapidly the virus is lost within the ingested blood (Fig. 4B). Our sensitivity analyses showed that within the ranges of the parameter values used, the probability of midgut infection (β) had the strongest effect on the proportion of simulations with midgut infection, followed by the blood meal clearance rate (μ) (Supplementary Fig. S7). In scenarios where the starting virus concentration was 10 PFU/ml, all simulations resulted in successful infections when β ranged from 10 to 10 and μ varied between 0.014 and 0.08 h. We subsequently used these parameter values in sensitivity analyses to evaluate viral dissemination and potential transmission (Supplementary Fig. S8). We found that by using values of β between 10 and 10 (Supplementary Fig. S6B), we could simulate the dose-response curves of midgut infection by the first set of chimeric viruses (Supplementary Fig. S3). These results suggest that differences in the dose-response curves of midgut infection across the chimeric viruses likely reflect variations in β, encompassing all processes from viral attachment, internalization, and replication in midgut cells.
When we exposed mosquitoes from Colombia to the first set of chimeric viruses, we observed significant variation in transmission prevalence between viruses (Fig. 3C). Our sensitivity analysis showed that the growth rate (r) was the most influential parameter, at least within the ranges of parameter values used, on viral dissemination in the hemocoel on day 7 (Supplementary Fig. S8A) and transmission potential (salivary gland infection) on day 10 post infectious blood meal (Supplementary Fig. S8B). There was a relatively narrow range of the growth rate (r) values (between 0.05 and 0.1 h) where the proportion of simulations went from zero to one for dissemination on day 7, and for transmission potential on day 10. Our analysis of potential causes for the observed differences in transmission dynamics between chimeric viruses focused on variations in r and λ between tissues. We explored alternative model parameterizations for r and λ by using tissue-specific parameter values and by introducing Gamma distributions to account for between-mosquito variation. Of the five hypothesized scenarios (Table 1), only one -- featuring a lower hemocoel-to-salivary gland escape rate (λ) compared with the midgut-to-hemocoel escape rate (λ) and between-mosquito variation in this parameter -- resulted in 100% prevalence of infection and dissemination but less than 60% transmission prevalence (Fig. 4C and scenario five, Table 1). The other scenarios primarily shifted the transmission prevalence curve to the right. Manipulating the variance of the Gamma distribution of λ reproduced the differences in transmission prevalence observed between the chimeric viruses (Fig. 4D), supporting the conclusion that these differences likely reflect variation in the hemocoel-to-salivary gland escape rate.