Efficient co-infection of three- and two-segmented RVFV reporter viruses in both mammalian and insect cells

In vertebrates, genetic complementation of virus particles with a distinct genome composition strictly relies on their ability to co-infect the same cell. To assess if mammalian and insect cells are susceptible to RVFV co-infection, we generated two recombinant three-segmented RVFV variants encoding either eGFP or mCherry2 in place of the NSs gene, using a T7 polymerase-based reverse genetics system (Fig 1A). Cells infected with the individual reporter viruses showed abundant expression of the respective fluorescent protein (Fig 1B), providing a suitable strategy to track virus infection and identify co-infected cells through the detection of co-localized fluorescent signals. Moreover, RVFV-eGFP and RVFV-mCherry2 replicated with almost identical growth kinetics, reaching high titers already 24 h post-infection and peaking at 48 h post-infection (Fig 1C). Upon simultaneous inoculation with RVFV-eGFP and RVFV-mCherry2, both BSR-T7/5 (hamster) and C6/36 (mosquito) cells were found to be susceptible to co-infection, as clearly evidenced by the co-localization of green and red fluorescent signal in a fraction of the cell population (Fig 1D and 1E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Co-infection of mammalian and insect cells with recombinant three-segmented RVFV reporter viruses.

(A) Schematic representation of the T7 polymerase-based reverse genetics system. Fluorescently marked variants of RVFV were generated by simultaneous transfection of BSR-T7/5 cells with the transcription plasmids pUC57_S, pUC57_M, and pUC57_L encoding the RVFV-35/74 S, M and L genome segments, respectively, in antigenomic-sense orientation. The pUC57_S plasmid additionally encoded for either eGFP or mCherry2 in place of the NSs gene. (B) Direct fluorescence detection of BSR-T7/5 cells infected with either RVFV-eGFP (green) or RVFV-mCherry2 (red) (MOI 0.5) at 24 h post-infection. (C) Growth kinetics of RVFV-eGFP and RVFV-mCherry2 after infection of BSR-T7/5 cells at an MOI of 0.01. Virus titers were determined with an end-point dilution assay (fluorescence microscopy readout). Dots represent means of biological replicates (n = 3) at each time point, and the shaded area represents the standard deviation. The dashed line indicates the limit of detection (10^1.8 TCID₅₀/mL). Source data are provided in S1 Data. (D) Schematic representation of the simultaneous infection of mammalian (BSR-T7/5) or insect (C6/36) cells with RVFV-eGFP and RVFV-mCherry2. (E) Direct fluorescence detection of BSR-T7/5 and C6/36 cells co-infected with RVFV-eGFP and RVFV-mCherry2 (MOI 0.5 for each virus) at 24 h (BSR-T7/5) or 72 h (C6/36) post-infection. Inset images are magnifications of a region of interest (indicated as a dashed box). Co-infected cells co-express eGFP (green) and mCherry2 (red) and thus appear yellow. Scale bars, 200 μm (inset images 100 μm). Illustrations in Fig 1A and 1D were created with BioRender.com.

https://doi.org/10.1371/journal.pbio.3001870.g001

Simultaneous infections with RVFV-eGFP and RVFV-mCherry2 allowed us to qualitatively confirm cell susceptibility to co-infection (Fig 1E). However, the fast spread of three-segmented RVFV impeded us from assessing accurately how often these co-infection events are actually taking place, since co-infections can occur over multiple infection cycles. To overcome this, we again used the T7 polymerase-based reverse genetics system to generate two non-spreading incomplete RVFV particle populations lacking the M segment and encoding a fluorescent protein (FP), either eGFP or mCherry2, in the S segment in place of the NSs gene (iRVFV-SL-FP) (Fig 2A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Quantitative assessment of mammalian cells co-infected with non-spreading two-segmented RVFV reporter viruses.

(A) Schematic representation of the reverse genetics system used to create two-segmented RVFV reporter viruses. Incomplete RVFV-SL particles were generated by co-transfection of BSR-T7/5 cells with the transcription plasmids pUC57_S, encoding either eGFP or mCherry2 in place of the NSs gene, pUC57_L, and the protein expression plasmid pCAGGS_NSmGnGc. (B) Immunofluorescence assay for detection of BSR-T7/5 cells infected with iRVFV-SL-eGFP or iRVFV-SL-mCherry2 (MOI 0.5) at 24 h post-infection. (C) Supernatants from cells primarily infected with the incomplete RVFV particles in (B) were passaged onto naive BSR-T7/5 cells. Cells were examined at 24 h post-infection for expression of eGFP (green) and mCherry2 (red) by fluorescence microscopy. Expression of the Gn glycoprotein (green or red, depending on the virus) was detected with rabbit polyclonal anti-Gn serum in combination with FITC-conjugated (green) or Alexa Fluor 568–conjugated (red) secondary antibodies. Cell nuclei (cyan) were visualized with DAPI. Scale bars, 50 μm. (D) Schematic representation of the simultaneous infection of mammalian (BSR-T7/5) cells with non-spreading iRVFV-SL-eGFP and iRVFV-SL-mCherry2 particles. Co-infections were done at increasing MOI (ranging from 0.1 to 2.5 for each virus). After 48 h, cells were examined by fluorescence microscopy and fixed for flow cytometry analysis. (E) Direct fluorescence detection of BSR-T7/5 cells co-infected (MOI of 0.5 for each virus) with the two-segmented incomplete particle populations. The inset image is a magnification of a region of interest (indicated as a dashed box). Co-infected cells co-express eGFP (green) and mCherry2 (red) and thus appear yellow. Scale bars, 200 μm (inset images 100 μm). (F) Cells expressing eGFP, mCherry2, or both were quantified by flow cytometry. Mock-infected cells and cells infected exclusively with only one population of incomplete particles (MOI 0.5) were used as controls. (G) Relationship between the fraction of infected (left) and co-infected (right) cells as a function of the MOI. Dots represent experimental data points. Dashed lines represent the predictions of a model based on the assumptions that genome segments are randomly packaged into virus particles and that host susceptibility is heterogeneous. The code required to reproduce the model predictions is provided as S1 File. Source data are provided in S1 Data. Illustrations in Fig 2A and 2D were created with BioRender.com.

https://doi.org/10.1371/journal.pbio.3001870.g002

The non-spreading nature of iRVFV-SL particles allowed us to assess the extent of co-infection after a single cycle of infection (without the generation of virus progeny). The generation of iRVFV-SL-eGFP has been previously reported by our group (initially termed RRPs) [24], but the generation of iRVFV-SL-mCherry2 is first reported as part of this work. Both particle populations are produced following complementation with an expression plasmid encoding the structural glycoproteins (Gn and Gc) normally encoded by the M segment. Despite the absence of the M segment, the S and L genome segments encoding for the nucleocapsid (N) protein and the RNA-dependent RNA polymerase (RdRp or L protein), respectively, are sufficient to support the replication and transcription of the S and L segments upon infection of naive cells. Importantly, because the M segment encoding for the Gn and Gc glycoproteins is absent in these naive cells, infection with iRVFV-SL particles does not lead to assembly and release of progeny virus. Direct fluorescence microscopy combined with an immunofluorescence assay to detect Gn clearly showed that cells infected with iRVFV-SL-eGFP or iRVFV-SL-mCherry2 have abundant expression of the respective FP but no expression of Gn (Fig 2B). Furthermore, passaging the supernatant of cells infected with iRVFV-SL-FP to naive cells did not result in the expression of eGFP, mCherry2, or Gn, confirming that these particles are incomplete and not able to spread due to the lack of the M genome segment (Fig 2C).

To quantify to what extent cells can be co-infected with the two different non-spreading fluorescent virus variants, we simultaneously infected BSR-T7/5 cells with iRVFV-SL-eGFP and iRVFV-SL-mCherry2 at increasing multiplicities of infection (MOIs, ranging from 0.1 to 2.5) (Fig 2D). Through direct detection of co-localized eGFP and mCherry2 expression, we confirmed that these particles also could co-infect BSR-T7/5 cells (Fig 2E). Next, using flow cytometry, we quantified the fraction of non-infected cells, singly-infected cells, and co-infected cells (Fig 2F). Mock-infected cells and cells infected with only one population of incomplete particles were the basis to gate the flow cytometry data. A clear double (eGFP and mCherry2)-positive cell population was identified after co-infection with both populations of incomplete particles.

Interestingly, at each MOI tested, the percentage of infected and co-infected cells closely resembled that of a mathematical model of infection (model C), which assumes that genome segments are randomly packaged into virus particles and that host cell susceptibility is heterogeneous among the cell population (Fig 2G and S1 File). Model selection based on the Akaike information criterion (AIC) indicated that model C is much better supported than two simpler models of infection (models A and B; S1 Table). Hence, model C was chosen for visualization here. The population of co-infected cells rose sharply with increasing MOI, showing a typical dose-response sigmoidal curve, suggesting no apparent mechanism leading to the exclusion of multiple particles entering the same cell in the present experimental setup. A reasonably good fit of the different models to the data suggests that co-infection is unrestricted upon simultaneous exposure to virus particles, as none of the models includes mechanisms of exclusion.

Successful generation of incomplete RVFV particles lacking the S genome segment

To study the potential role of incomplete particles in the bunyavirus life cycle, we needed at least two distinct populations of RVFV particles having incomplete but complementing genomes. Besides generating iRVFV-SL-FP particles, we also generated incomplete RVFV particles lacking the S segment (iRVFV-ML) by following a similar T7 polymerase-based reverse genetics strategy (Fig 3A). Complete M and L genome segments were encoded in antigenomic-sense orientation by transcription plasmids, and the N protein was provided by an expression plasmid. Due to the absence of the S segment in the rescued particles and, consequently, the unavailability of N protein, infections with iRVFV-ML particles did not result in any appreciable viral genome replication. Accordingly, neither N nor Gn was detected in an immunofluorescence assay of cells infected with iRVFV-ML (Fig 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Generation of incomplete RVFV particles lacking the S genome segment.

(A) Schematic representation of the reverse genetics system used to create incomplete RVFV-ML particles. iRVFV-ML particles were generated by co-transfection of BSR-T7/5 cells with the transcription plasmids pUC57_M, pUC57_L, and the protein expression plasmid pCAGGS_N. (B) Immunofluorescence assay for detection of BSR-T7/5 cells exposed to iRVFV-ML (MOI 0.5) at 24 h post-infection. Expression of the N protein (green) was detected with a mouse anti-N monoclonal hybridoma in combination with Alexa Fluor Plus 488–conjugated secondary antibodies. Expression of the Gn glycoprotein (red) was detected with rabbit polyclonal anti-Gn serum in combination with Alexa Fluor 568–conjugated secondary antibodies. Cell nuclei (cyan) were visualized with DAPI. Scale bars, 50 μm. (C) Segment-specific RT-qPCR quantification of the genomic composition of two batches of rescued iRVFV-ML particles. Stock preparations of RVFV-eGFP and iRVFV-SL-eGFP were included as reference for comparison. Bars show means with SD. Dots represent replicates (n = 3 samples). Source data are provided in S1 Data. Illustration in Fig 3A was created with BioRender.com.

https://doi.org/10.1371/journal.pbio.3001870.g003

Due to the non-replicating nature of RVFV particles lacking the S genome segment, the infectious titer of iRVFV-ML stocks could not be determined with a conventional virus titration assay. Therefore, we confirmed the genomic composition of rescued iRVFV-ML particles by quantifying the S, M, and L genome segments via quantitative reverse transcription PCR (RT-qPCR) (Fig 3C). In both batches of rescued iRVFV-ML particles, high copy numbers of only the M and L genome segments were detected, whereas the S segment was not detected. Similarly, in samples containing iRVFV-SL-eGFP particles, only the S and L genome segments were present at high copy numbers, and the M segment was not detected. All three genome segments (S, M, and L) were detected in samples containing the three-segmented RVFV-eGFP used as control.

Co-infection with complementing incomplete RVFV particles results in efficient virus replication and spread among mammalian and insect cells

Having generated two distinct populations of two-segmented RVFV particles with an incomplete but complementing genome, we next investigated whether exposing cells to a mixture of these two populations would result in genome complementation (SL + ML = SML) and subsequent reconstitution of three-segmented (SML) infectious virus (Fig 4A). Rescue of infectious virus following incubation of cells exclusively with virus populations with an incomplete set of genome segments would strongly suggest a role for incomplete particles in virus replication and spread. To this end, BSR-T7/5 and C6/36 cells were simultaneously inoculated with iRVFV-SL-eGFP and iRVFV-ML at different MOIs (ranging from 0.001 to 0.25 for each virus). As shown previously in Figs 2B and 3B, incubating cells with only iRVFV-SL-eGFP or iRVFV-ML particles did not result in productive infections. Inoculation with iRVFV-SL-eGFP particles resulted in the expression of eGFP in single cells without spreading to neighboring cells, and inoculation with iRVFV-ML particles did not result in the production of viral proteins (Fig 4B and 4C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Co-infection with complementing incomplete RVFV particles results in efficient virus replication and spread among mammalian and insect cells.

(A) Schematic representation of the individual or simultaneous exposure of mammalian (BSR-T7/5) and insect (C6/36) cells to non-spreading iRVFV-SL-eGFP and iRVFV-ML particles. If individual susceptible cells become co-infected by the two different incomplete particles, genome complementation could eventually allow virus replication and production of infectious progeny. (B and C) BSR-T7/5 cells (B) and C6/36 cells (C) were mock-infected, infected exclusively with one population of incomplete particles (MOI 0.1), or simultaneously infected with iRVFV-SL-eGFP and iRVFV-ML (MOI 0.1 for each virus). Cells infected with a three-segmented RVFV-eGFP (MOI 0.2) were used as a positive control. In all cases, infected cells were analyzed at 24 h (BSR-T7/5 cells) or 72 h (C6/36 cells) post-infection by following the expression of eGFP (green) via direct fluorescence microscopy and the expression of Gn (red) via an immunofluorescence assay in BSR-T7/5 cells or an immunoperoxidase monolayer assay in C6/36 cells. For detection of Gn in C6/36 cells, the immunoperoxidase monolayer assay was chosen because the fluorescent signal got drastically reduced upon fixation of the cells. Expression of Gn was detected with rabbit polyclonal anti-Gn serum in combination with Alexa Fluor 568–conjugated secondary antibodies (immunofluorescence assay) or with HRP-conjugated secondary antibodies (immunoperoxidase monolayer assay). Cell nuclei (cyan) were visualized with DAPI. Scale bars, 200 μm. (D) Infections were also carried out at different MOIs (ranging from 0.001 to 0.25 for each virus, S1 Fig) in triplicate. Green squares represent successful genome complementation leading to virus replication and spread. White squares represent absence of genome complementation. Source data are provided in S1 Data. (E) Supernatants from BSR-T7/5 and C6/36 co-infected cells or cells infected with RVFV-eGFP were passaged onto naive BSR-T7/5 cells. Cells were examined at 24 h post-infection combining direct fluorescence microscopy and immunofluorescence as described for panels in (B). Scale bars, 200 μm. Illustration in Fig 4A was created with BioRender.com.

https://doi.org/10.1371/journal.pbio.3001870.g004

Remarkably, upon simultaneous inoculation with iRVFV-SL-eGFP and iRVFV-ML, abundant expression of eGFP but also of Gn was observed within the same cells, suggesting that particles from both populations can co-infect individual cells and together provide a complete set of genome segments encoding all the viral proteins. The co-expression of eGFP and Gn was accompanied by the detection of clusters of infected neighboring cells and a rapid increase in the percentage of infected cells over time, as was also observed after infection with the three-segmented RVFV-eGFP strain (Figs 4B, 4C and S1). This result strongly suggests that co-infection with complementing incomplete particles allows the reconstitution of a three-segmented virus. Notably, complementation by incomplete particles was efficient in both BSR-T7/5 and C6/36 cells even at low MOIs (≥0.1 and ≥0.01, respectively) (Figs 4D and S1). Next, to confirm whether cells simultaneously exposed to the two different populations of incomplete particles produce infectious progeny able to spread efficiently in vitro, we passaged the supernatants of both BSR-T7/5 and C6/36 cells to naive cells. The results showed that early after infection (24 hpi), a high proportion of the cell population abundantly co-expressed eGFP and Gn, with clusters of positive cells rapidly expanding, suggesting that infectious virus was rescued (Fig 4E).

Besides tracking the course of infection at the cell population level, we also analyzed the intracellular genome content of individual infected BSR-T7/5 cells using a combined smFISH-immunofluorescence method [22] (Fig 5A). As expected, in cells exposed exclusively to iRVFV-SL-eGFP, we detected abundant copies of the S and L genome segments, while the M segment and the Gn glycoprotein were absent. No expression of eGFP or Gn was detected in cells exclusively exposed to iRVFV-ML particles, as the S genome segment is missing in this population, and thus no genome replication or transcription could take place. Contrary to cells exposed to only one population of incomplete particles, cells simultaneously exposed to both iRVFV-SL-eGFP and iRVFV-ML particles showed abundant copies of each of the three genome segments, similar to cells infected with the three-segmented RVFV-Clone 13 (Fig 5B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Visualization of viral genome segments in RVFV-infected cells and in immobilized RVFV virions from virus stocks and co-infection supernatants.

(A) Schematic representation of the single-molecule viral RNA FISH-immunofluorescence method used to visualize the intracellular viral ribonucleoprotein (vRNP) composition of infected cells and the intravirion genomic composition of virus stocks and culture supernatants (illustration based on [22]). (B) BSR-T7/5 cells were mock infected, infected exclusively with one population of incomplete particles (MOI 1), or simultaneously infected with iRVFV-SL-eGFP and iRVFV-ML (MOI 1 for each virus), and fixed at 16 h post-infection. Cells infected with a three-segmented RVFV-Clone 13 (MOI 1) were used as positive control. (C) RVFV virions from incomplete particle stocks and supernatants of co-infected BSR-T7/5 and C6/36 cells were immobilized on coverglass by incubation for 5 h at 28°C. The S segment (N gene, red), M segment (polyprotein gene, blue), and L segment (RdRp gene, yellow) were hybridized using probe sets labeled with CAL Fluor Red 610, Quasar 670, and Quasar 570, respectively. RVFV particles (green) were detected with antibody 4-D4 [47] targeting the Gn glycoprotein in combination with Alexa Fluor 488–conjugated secondary antibodies. Cell nuclei (cyan) were visualized with DAPI. Merge images show the overlay of five (for cells) or four (for virions) individual channels. Individual spots, each representing either a single vRNP or a virus particle, were detected and assessed for co-localization in ImageJ with the plugin ComDet. Colored circles display the spots detected in each channel and their co-localization in the merge image. Scale bars, 10 μm (B), 5 μm (C). Illustration in Fig 5A was created with BioRender.com.

https://doi.org/10.1371/journal.pbio.3001870.g005

Finally, we used our smFISH-immunofluorescence method to obtain additional insights into the intravirion composition of the incomplete particle populations and the progeny virions released upon co-infection with these incomplete particles (Fig 5A). Importantly, the M genome segment was not present in iRVFV-SL-eGFP particles, and the S genome segment was not present in iRVFV-ML particles, confirming the absence of the respective segment in each of the incomplete particle populations. Notably, just as in stocks of the three-segmented positive control RVFV-Clone 13, a fraction of the virions produced by co-infected cells was shown to contain the complete set of S, M, and L genome segments, ultimately confirming the reconstitution of infectious three-segmented virus upon co-infection with incomplete particles (Fig 5C).

Co-infection with complementing incomplete RVFV particles supports in vivo virus spread in mosquitoes

Having confirmed the ability of incomplete RVFV particles to generate a productive infection in vitro upon co-infection, we hypothesized that incomplete bunyavirus particles could also play a role in between-host virus transmission. To investigate this, groups of Aedes aegypti mosquitoes were fed with bovine blood meals spiked with diverse virus populations. Mosquitoes were exposed to a mock blood meal (group #1), a blood meal spiked with a single incomplete virus particle population (group #2 to iRVFV-SL-eGFP and group #3 to iRVFV-ML), a blood meal spiked with three-segmented RVFV-mCherry2 (group #4) or a blood meal spiked with a mixture of incomplete particles (iRVFV-SL-eGFP and iRVFV-ML, group #5). After incubation for 12 to 15 d post-feeding at 28°C, mosquitoes were analyzed for virus replication in their bodies and dissemination to the saliva (Fig 6A). To this end, body homogenates and saliva samples were tested by virus isolation on BSR-T7/5 cells, using the expression of virus-encoded eGFP or mCherry2 (depending on the virus) as a readout.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. RVFV incomplete particles complement upon co-infection and allow virus replication and spread in mosquitoes.

(A) Schematic representation of the experimental design. Five groups of Aedes aegypti mosquitoes (n = 15–62 per group) were fed a blood meal spiked with different RVFV preparations and housed at 28°C. After 12–15 d, mosquitoes were sedated with CO₂ and body and saliva samples were collected. RVFV infection of mosquito bodies and transmission to mosquito saliva were assessed via virus isolation with a fluorescence microscopy readout. (B) RVFV infectious titers in mosquito bodies. Dots represent individual mosquitoes and are color-coded gray for negative bodies, light red (group #4) or light green (group #5) for positive bodies, and solid red (group #4) or solid green (group #5) for both body and saliva virus-positive mosquitoes. The dashed line indicates the limit of detection (10^1.5 TCID₅₀/mL). The incidence (in absolute numbers and percentages) of RVFV infection (bodies) and transmission (saliva) is indicated at the bottom for each group. Data corresponding to mosquitoes from groups #4 and #5 derive from two independent experiments. Source data are provided in S1 Data. (C) Representative fluorescence microscopy images of BSR-T7/5 cells inoculated with virus-positive body and virus-positive saliva samples from groups #4 (RVFV-mCherry2) and #5 (iRVFV: SL-eGFP + ML) at 48 h post-infection. Merge images show the overlay of two individual channels (eGFP and mCherry2). Scale bars, 200 μm. Illustration in Fig 6A was created with BioRender.com.

https://doi.org/10.1371/journal.pbio.3001870.g006

All the mosquitoes fed on a mock blood meal (group #1, n = 15) or a blood meal spiked with only one population of incomplete virus particles (group #2, n = 25; and group #3, n = 20) were virus-negative in their bodies and saliva (Fig 6B). No virus spread was observed in these groups, as iRVFV-SL-eGFP and iRVFV-ML both lack one genome segment, either the M or the S segment. In group #4, 33 out of 52 (63%) mosquitoes fed on a blood meal spiked with three-segmented RVFV-mCherry2 were virus-positive in their bodies, confirming the high vector competence of the mosquitoes for RVFV. Eleven of these 33 mosquitoes were also virus-positive in their saliva, implying that RVFV-mCherry2 had replicated in the mosquito midgut, disseminated throughout the body, reached the salivary glands, and was excreted in saliva. Strikingly, 24 out of 62 (39%) mosquitoes in group #5, which were fed on a blood meal spiked with a mixture of iRVFV-SL-eGFP and iRVFV-ML particles, were found to have virus-infected bodies. Of those 24, three mosquitoes were virus-positive in their saliva (Fig 6B and 6C). These results show that genome complementation via co-infection with two populations of RVFV exclusively comprising incomplete particles can also occur in vivo, supporting virus spread in mosquitoes and presumably contributing to virus transmission between hosts.

Predicting the contribution of incomplete particles to virus spread

We showed that experimental co-infection with complementing two-segmented incomplete particles results in the rescue of spreading three-segmented virus both in vitro and in vivo. However, incomplete particles naturally co-exist within a mixed population of empty, incomplete, and complete particles. To know the relevance of incomplete virus particles for infection dynamics, it is important to measure to what extent incomplete particles actually contribute to virus spread. Ideally, this question should be addressed by comparing the performance of a natural virus population to a population in which infection by incomplete particles is blocked. Alternatively, this could be studied with experiments in which only complete virus particles are generated. As such experiments are technically not possible at present, we used a simulation model to compare RVFV infection dynamics in three different scenarios: (i) non-selective genome packaging without productive co-infection by incomplete virus particles; (ii) non-selective genome packaging with productive co-infection by incomplete virus particles; and (iii) selective genome packaging (Fig 7A and S2 File).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Predicting the contribution of incomplete particles to within-host virus spread.

(A) Predicted RVFV infection dynamics in mammalian and insect cells in three different scenarios: selective genome packaging, non-selective genome packaging without co-infection by incomplete particles, and non-selective genome packaging with co-infection and productive complementation by incomplete particles. The low number of cells in the model is representative of localized virus spread. Solid lines represent the mean of n = 1,000 simulations per scenario, and the light-colored dashed lines represent 20 individual simulations. pckg.: packaging. (B) Contribution of incomplete particles to virus spread in mammalian and insect cells as a function of the MOI in a model representative of non-selective genome packaging allowing co-infection by incomplete particles. Lines represent the mean of n = 500–10⁴ simulated cells, with a larger number of cells for the low MOI conditions where overall infection frequencies are low. (C) Contribution of incomplete particles to virus spread as a function of the MOI for starting inoculums with different frequencies of complete particles. Lines are color-coded in a gradient of dark yellow to dark red based on an increasing fraction of complete particles (ranging from 0.01 to 0.27) within the virus population. The code required to reproduce the plots of this figure is provided as S2 File.

https://doi.org/10.1371/journal.pbio.3001870.g007

In both the first and second scenarios, there is non-selective packaging, and the distribution of genome segments over virus particles approximately follows the empirical distribution. In the first scenario, the model assumes that the infection is driven exclusively by complete three-segmented particles, which only account for a small fraction of the particle population. By contrast, the second scenario allows the infection to progress by both complete particles and co-infecting incomplete particles, provided all segments are present in a single cell. Lastly, in the third scenario, it is assumed that all the particles are complete due to selective packaging, which serves as a reference point. Since there are differences in genome packaging efficiencies between mammalian and insect hosts [22], and, therefore, the exact ratio between empty, incomplete, and complete particles differs considerably depending on the host, we performed simulations with RVFV populations representative of those found in mammalian and insect hosts.

Model parameters were defined to represent the conditions of localized virus spread (i.e., mean-field simulations in a small number of cells, for instance, 100), and the model predicts how the number of infected cells changes over time. In the first scenario, the number of infections increases the slowest, being limited by the small number of complete virus particles. In the second scenario, when the model considers that incomplete particles can generate a productive infection through complementation, the number of infected cells increases more rapidly, highlighting the contribution of incomplete particles to virus spread. However, the number of infected cells does not increase as quickly as in the third scenario in which all virus particles are complete, suggesting there is still a cost to non-selective packaging (Fig 7A and S2 File). Importantly, the model is sensitive to the host in which the virus is replicating, suggesting that the contribution of incomplete particles to virus spread is greater in mammalian cells than in insect cells due to the smaller fraction of complete particles in mammalian-derived virus.

When we considered how model parameter values affect the predictions, we found that the main trends noted here were also found over different conditions; there is often an advantage for the virus that allows complementation through co-infection over the non-complementing one, while the virus with selective genome packaging outperforms both viruses with non-selective packaging. Under some conditions, the performance of the virus with non-selective packaging but that allows genome complementation resembles that of a virus that employs selective packaging (S1 Text and S2 and S3 Figs and S3 File). Overall, these results suggest that incomplete virus particles can contribute to the spread of RVFV over a broad range of conditions but that non-selective packaging is generally costly and leads to inferior performance compared to a virus with selective packaging.

In both cases shown in Fig 7A, the viruses employing non-selective packaging with or without complementation appear to perform the same initially. Later in the infection process, the complementing virus outperforms the non-complementing virus. As the MOI will increase over rounds of infection, these simulation results suggest an important role of the MOI in determining the relevance of incomplete particles to virus spread. We therefore explored the relationship between MOI and the contribution of incomplete particles more systematically. We calculated the fraction of infected cells comprising a complete set of genome segments as a result of co-infection with incomplete particles only, over a range of MOIs (Fig 7B and S2 File).

The simulations show that the MOI has a pronounced effect on the contribution of incomplete particles to virus spread, with a limited contribution at extreme MOIs. At low MOIs, there is a low probability of particles co-infecting the same cell, and, consequently, complementation events are rare, whereas at high MOIs, in most cells, infection will be initiated by at least one complete particle, and complementation of incomplete particles is not required for virus spread. Notably, at intermediate MOIs, infection initiated by complementing incomplete particles is more common and plays an important role in virus spread (Fig 7B and S2 File). Here again, the contribution of incomplete particles is estimated to be higher in the mammalian host than in the insect host, because insect-derived virus preparations contain a larger fraction of complete particles [22].

To gauge how general these model results are, we considered the relationship between MOI and the contribution of incomplete virus particles over a broad range of randomly selected genome segments frequencies for a three-segmented virus. Interestingly, the larger contribution of incomplete particles at intermediate MOIs appears to be a general feature over different virus particle compositions (Fig 7C and S2 File). In line with what we observed for the mammalian- and insect-derived virus, this contribution becomes less relevant when the fraction of complete particles becomes larger.

Cell lines

BSR-T7/5 cells (Golden hamster kidney, CVCL_RW96) stably expressing bacteriophage T7 RNA polymerase [41] were maintained in Glasgow minimum essential medium (G-MEM) supplemented with 5% fetal bovine serum (FBS), 1% antibiotic/antimycotic, 1% MEM non-essential amino acids (MEM NEAA), and 4% tryptose phosphate broth at 37°C and 5% CO₂. For stable maintenance of the cell line, the medium was supplemented with 1 mg/mL Geneticin (G-418 sulfate) every other passage. Vero E6 cells (African green monkey kidney, ATCC CRL-1586) were maintained in MEM supplemented with 5% FBS, 1% antibiotic/antimycotic, 1% MEM NEAA, and 2 mM L-glutamine at 37°C and 5% CO₂. C6/36 cells (Aedes albopictus larva, ATCC CRL-1660) were maintained in L-15 medium (Leibovitz) (Sigma-Aldrich) supplemented with 10% FBS, 1% antibiotic/antimycotic, 1% MEM NEAA, and 2% tryptose phosphate broth at 28°C. Cell culture media and supplements were purchased from Gibco, unless specified otherwise.

Viruses

Virus stocks of RVFV strain Clone 13 [42] were obtained following infection of Vero E6 cells at an MOI of 0.005. Recombinant fluorescently marked variants of RVFV strain 35/74 (accession numbers JF784386-88) [43], expressing either eGFP (RVFV-eGFP) or mCherry2 (RVFV-mCherry2), were generated by a pUC57 transcription plasmid-based reverse genetics system [24]. Briefly, BSR-T7/5 cells were seeded in 6-well cell culture plates at 1.5 to 2.0 × 10⁵ cells/well (2 mL volume) and incubated for 18 to 24 h to generate a sub-confluent monolayer. Cells were subsequently co-transfected with the plasmids pUC57_S-FP (where FP corresponds to either eGFP or mCherry2 in place of the NSs gene), pUC57_M and pUC57_L, encoding the RVFV-35/74 S, M, and L genome segments, respectively, in antigenomic-sense orientation. Transfections were performed using TransIT-LT1 transfection reagent (Mirus), with slight modifications to the manufacturer’s instructions: 3 to 5 h prior to transfection, the seeding supplemented G-MEM was removed and substituted by 1 mL of Opti-MEM I Reduced-Serum medium. For transfection, 2,500 ng total plasmid/well (equal amounts per plasmid) with a 1:4 plasmid to transfection reagent ratio were used. At 3 to 5 h post-transfection, 1 mL of supplemented G-MEM was added. At 3 d post-transfection, culture supernatants were harvested and clarified by low-speed centrifugation. High-titer virus stocks of RVFV-eGFP and RVFV-mCherry2 were obtained following infection of BSR-T7/5 cells at an MOI of 0.001.

Generation of incomplete virus particles

Incomplete RVFV particles lacking either the S segment (iRVFV-ML) or the M segment (iRVFV-SL-FP) were generated by reverse genetics [24]. For the production of iRVFV-ML, BSR-T7/5 cells were co-transfected with pUC57_M, pUC57_L, and the protein expression plasmid pCAGGS_N (encoding for a codon-optimized RVFV-35/74 nucleocapsid protein). For the production of iRVFV-SL-FP, BSR-T7/5 cells were co-transfected with pUC57_S-FP (where FP corresponds to eGFP or mCherry2 in place of the NSs gene), pUC57_L, and the protein expression plasmid pCAGGS_NSmGnGc (encoding for the RVFV-35/74 polyprotein). Transfections were performed using TransIT-LT1 transfection reagent (Mirus), following the protocol described for the generation of recombinant fluorescently marked RVFV-35/74 variants.

To obtain a high-titer stock of iRVFV-ML particles, the rescued supernatant from the transfection was concentrated by ultracentrifugation at 28,000 rpm, 4°C for 2 h (SW 32 Ti swinging-bucket rotor, Optima LE-80K Beckman Coulter) using a 25% w/v sucrose cushion. Furthermore, to obtain higher titers of iRVFV-SL-FP particles, cells initially transfected with the three plasmids were repeatedly transfected two additional times with the protein expression plasmid pCAGGS_NSmGnGc, resulting in a polyclonal cell line replicating the S and L viral genome segments. New stocks of iRVFV-SL-FP particles were generated by a final transfection of the polyclonal cell line with pCAGGS_NSmGnGc. The latter transfections were performed using jetPEI transfection reagent (Polyplus) with a 1:3 plasmid to transfection reagent ratio. At 1 d post-transfection, culture supernatants were harvested and clarified by low-speed centrifugation.

Virus titration

Infectious titers of rescued virus and virus stocks were determined in an end-point dilution assay in combination with either an immunoperoxidase monolayer assay (IPMA, as described below) or direct microscopy detection of FP expression. BSR-T7/5 (3 × 10⁴ cells/well) or C6/36 (6 × 10⁴ cells/well) monolayers were incubated with 10-fold serial dilutions (starting at 1:10) of the samples for 72 h at 37°C and 5% CO₂ (BSR-T7/5) or 28°C (C6/36). Samples were analyzed in triplicate or quadruplicate, and the titer was calculated as the median tissue culture infectious dose (TCID₅₀/mL) using the Spearman–Kärber method.

Growth curves of fluorescent virus variants

BSR-T7/5 monolayers (3 × 10⁶ cells/T75 flask) were infected with RVFV-eGFP or RVFV-mCherry2 at an MOI of 0.01. At 2 h post-infection, the inoculum was removed, cells were washed with PBS, and fresh medium was added. Cell culture supernatants were harvested at 0, 2, 12, 24, 48, and 72 h post-infection and were clarified by centrifugation at 2,500 rcf for 10 min. Virus titers of the clarified supernatants were determined with an end-point dilution assay (fluorescence microscopy readout) as described above. Growth curve determinations were performed with three biological replicates per time point.

Genome segment-specific quantitative RT-PCR

As a conventional end-point dilution assay is not suited for determining the titer of non-replicating particles, we estimated the titer of iRVFV-ML stocks through genome segment-specific quantification via RT-qPCR, followed by a comparison with the genome copies of iRVFV-SL-eGFP and three-segmented RVFV-eGFP preparations with known infectious titer determined with the end-point dilution assay. Total RNA extractions of 80 μL of virus stocks lyzed with 240 μL of TRIzol LS Reagent (Invitrogen) were performed in triplicate with the Direct-zol RNA MiniPrep kit (Zymo Research), largely according to the manufacturer’s instructions, except for a more thorough in-column DNase I treatment. Namely, lyzed preparations were treated with 60 units of DNase I for 30 min. The extended DNase I treatment ensured the complete removal of residual plasmid DNA from the transfection. Total RNA was eventually eluted in 25 μL of DNase/RNase-free water (Zymo Research). Subsequently, viral cDNA was synthesized with the SuperScript IV First-Strand Synthesis System for RT-PCR (Invitrogen) using 100 units of SuperScript IV reverse transcriptase and a combination of S, M, and L segment-specific primers (S2 Table). After the reverse transcription reaction, quantitative PCR amplifications were performed with the Power SYBR Green PCR Master Mix using 5 μL of 50-, 250-, or 500-fold diluted cDNA preparations in a total volume of 25 μL, in combination with a 7500 Fast Real-Time PCR System (Applied Biosystems). Fragments from the three viral genome segments were amplified using specific primers (S3 Table) under the following conditions: an initial denaturation step at 95°C for 10 min; 40 cycles of denaturation at 95°C for 15 s, annealing at 59°C for 30 s, and extension at 72°C for 36 s; and a single cycle of denaturation at 95°C for 15 s, annealing at 60°C for 1 min, denaturation at 95°C for 15 s, and annealing at 60°C for 15 s. Per sample, an additional reaction intended to detect residual plasmid DNA was carried out using primers designed to amplify a fragment of the ampicillin resistance gene (ampR) (S3 Table) present in pUC57 and pCAGGS plasmids used for generating the different RVFV variants. Data were acquired and analyzed with the 7500 Fast System software version 1.4. (Applied Biosystems). Genome copies of each viral segment were calculated by intrapolation of the respective standard curve prepared with 10-fold serial dilutions of the viral segment cloned in pUC57 plasmids starting at 0.1 ng/μL.

Flow cytometry

BSR-T7/5 cells were simultaneously co-infected with iRVFV-SL-eGFP and iRVFV-SL-mCherry2 at increasing MOIs (ranging from 0.1 to 2.5 for each virus). At 48 h post-infection, cells were trypsinized, spun down at 200 rcf for 5 min, fixed with 4% paraformaldehyde for 15 min, washed twice with PBS, and resuspended in sample buffer (PBS supplemented with 1% bovine serum albumin and 0.01% sodium azide). Mock-infected cells and cells infected exclusively with iRVFV-SL-eGFP or iRVFV-SL-mCherry2 at an MOI of 0.5 were used as negative and singly-infected controls, respectively. Flow cytometry was performed using a BD FACS Aria III (BD Biosciences) equipped with a standard laser and filter configuration. Cell subpopulations were categorized and quantified based on the expression of eGFP, mCherry2, both or none. Per sample, 50,000 events were recorded. Data were analyzed with FlowJo version 10.7.1. The gating strategy applied involved discriminating cells from debris, followed by selection of single cells and assessment of FP expression (S4 Fig).

Co-infections with incomplete virus particles

BSR-T7/5 (5 × 10⁴ cells/well) or C6/36 (1 × 10⁵ cells/well) monolayers seeded in 24-well cell culture plates were simultaneously co-infected with iRVFV-SL-eGFP and iRVFV-ML particles at increasing MOIs (ranging from 0.001 to 0.25 for each virus). At 2 h post-infection, the inoculum was removed, cells were washed with PBS, and fresh medium was added. At defined time points post-infection (varied per cell line and ranged from 24 h to 72 h), cells were examined for expression of eGFP via direct fluorescence microscopy. Expression of Gn in BSR-T7/5 and C6/36 infected cells was examined via an IPMA or an immunofluorescence assay (as described below). The expression of eGFP was followed over the course of the infection as an indicator of viral spread. Co-infections at the different MOIs were performed with three biological replicates. Cells infected exclusively with either iRVFV-SL-eGFP or iRVFV-ML particles were used as singly-infected controls, whereas cells infected with a three-segmented RVFV-eGFP (MOIs ranging from 0.002 to 0.5) were used as positive control for virus spread. Mock-infected samples were used as negative control.

Immunostainings

BSR-T7/5 and C6/36 cells infected with different RVFV variants were subjected to an IPMA or an immunofluorescence assay to detect the expression of the viral proteins Gn and N. At defined time points post-infection (varied per experiment), cells were fixed with 4% paraformaldehyde for 15 min, washed with PBS supplemented with 0.5% Tween 80 (PBST), and permeabilized with 1% Triton X-100 in PBS for 5 min. Next, samples were blocked with 5% horse serum in PBS and subsequently incubated in sequential steps with primary and secondary antibodies diluted in blocking solution (S4 Table). Gn was detected with a polyclonal serum from a Gn-immunized rabbit (1:500 dilution, Thermo Fisher) as primary antibody. For the IPMA, HRP-conjugated goat polyclonal anti-rabbit immunoglobulin (1:500 dilution, P0448 Dako) was used as secondary antibody. For the immunofluorescence assay, goat polyclonal anti-rabbit IgG labeled with FITC (1:250 dilution, sc-2012 Santa Cruz Biotechnology) or donkey polyclonal anti-rabbit IgG labeled with Alexa Fluor 568 (1:500 dilution, A10042 Invitrogen) were used as secondary antibodies. N was detected with a monoclonal mouse hybridoma (1:100 dilution, F1D11 CISA-INIA) as primary antibody and a goat polyclonal anti-mouse IgG labeled with Alexa Fluor Plus 488 (1:500 dilution, A32723 Invitrogen) as secondary antibody. Sample incubations with the blocking solution, primary and secondary antibodies were each for 1 h at 37°C. Plates were washed with PBST after permeabilization, between the addition of primary and secondary antibodies, and prior to staining. In the IPMA, a 0.2-mg/mL amino ethyl carbazole solution in 500 mM acetate buffer (pH 5.0), 88 mM H₂O₂ was added as substrate. In the immunofluorescence assay, cell nuclei were stained by incubation with 1 μg/mL DAPI in PBS for 5 min. Mock-infected samples and samples without the addition of primary antibodies were used as negative controls.

Single-molecule RNA FISH-immunofluorescence

Experiments were performed with slight modifications to the Stellaris protocol for simultaneous FISH-immunofluorescence in adherent cells (Biosearch Technologies) [44–46]. BSR-T7/5 monolayers (1.5 × 10⁴ cells/well) were seeded on CultureWell 16 removable chambered coverglass (Grace Bio-Labs) and allowed to attach for at least 2 h at 37°C and 5% CO₂. Cells were infected with the different RVFV variants at an MOI of 1. At 2 h post-infection, the inoculum was removed and the medium was refreshed. At 16 h post-infection, cells were fixed and permeabilized with a 3:1 mixture of methanol (Klinipath)-glacial acetic acid (Merck) for 10 min. Cells were subsequently washed twice with PBS and once with prehybridization buffer (10% deionized formamide [Millipore] in 2× concentrated SSC [Gibco]) for 5 min. Cells were then incubated for 4 to 16 h at 37°C with 100 μL/well of virus-specific FISH probe sets (S5 Table) and primary antibodies in hybridization buffer (10% deionized formamide, 10% dextran sulfate [Sigma-Aldrich], 2 mM vanadyl ribonucleoside complexes [VRC, Sigma-Aldrich] in 2× SSC). FISH probes were added at a final concentration of 250 nM. Gn was detected with hybridoma 4-D4 [47] supernatant (1:160 dilution) as primary antibody (S4 Table). Following hybridization and incubation with primary antibodies, cells were extensively washed at 37°C (twice with prehybridization buffer for 30 min and twice with 2× SSC for 15 min). Subsequently, cells were incubated with 100 μL/well of secondary antibody for 1 h at 37°C. A goat polyclonal anti-mouse IgG labeled with Alexa Fluor 488 (1:1,000 dilution, A-11001 Invitrogen) was used as secondary antibody (S4 Table). Next, cells were washed twice with 2× SSC, and nuclei were stained by incubation with 100 μL/well of 1 μg/mL DAPI in 2× SSC for 5 min. Finally, cells were washed with 2× SSC and submerged in VectaShield antifade mounting medium H-1000 (Vector Laboratories). For analysis of virus stocks, undiluted or 1:3 diluted virus stocks were added on CultureWell 16 removable chambered coverglass and virions were allowed to attach to the surface for 5 h at 28°C. The same procedure as described for adherent cells was followed from the fixation step onward. The specificity of the FISH probes and antibodies was reported previously [22]. Mock-infected samples and samples without primary antibodies were used as negative controls.

Image acquisition and analysis

Light microscopy images were acquired with a Leica Model DMi1 inverted microscope and 10× HI PLAN I or 20× HI PLAN I objectives. Fluorescence microscopy images were acquired with an inverted widefield fluorescence microscope Axio Observer 7 (ZEISS, Germany) using appropriate filters and a 10× EC Plan-NEOFLUAR objective or a 1.3 NA 100× EC Plan-NEOFLUAR oil objective in combination with an AxioCam MRm CCD camera. Exposure times were defined empirically and differed depending on the probe sets and fluorescent dyes. For the FISH-immunofluorescence assay, Z-stacked images of infected cells and immobilized virions were acquired with a fixed interval of 0.28 to 0.31 μm between slices. Raw images were deconvolved in standard mode using Huygens Professional version 21.04 (Scientific Volume Imaging B.V., the Netherlands). If required, raw images were Z-aligned in ZEN 2.6 Pro (ZEISS, Germany) before deconvolution. For analysis, 3D data were converted to maximum intensity projections using Z-project within ImageJ [48]. Detection, quantification and co-localization analyses of individual spots, each representing a single virion or vRNP, were performed in ImageJ in combination with the plugin ComDet version 0.5.0 (https://github.com/ekatrukha/ComDet). Spot detection thresholds for each channel were set empirically by individual examination of images. The threshold to define co-localized spots was set to a maximum distance of 3 to 4 pixels between the centers of the spots. For visualization purposes, image brightness and contrast were manually adjusted in ImageJ.

Mosquito experiment

Adult Aedes aegypti (Rockefeller strain) mosquitoes were reared in 30 × 30 × 30 cm cubic cages at 27 ± 1°C and a 12:12 h light:dark cycle, at the Laboratory of Entomology, Wageningen University & Research. Five groups of mosquitoes, housed independently in small cardboard buckets, were allowed to feed through a Parafilm M membrane for at least 1.5 h on 37°C-heated virus-spiked blood meals using a Hemotek PS5 membrane feeding system (Hemotek, United Kingdom). Virus-spiked blood meals were prepared by mixing virus stocks with bovine blood washed twice with DPBS in a 2:1 ratio (Table 1). Virus-blood mixtures were back-titrated as described above. After feeding, mosquitoes were anesthetized using a semi-permeable CO₂ pad. Engorged female mosquitoes were selected, placed into new cardboard buckets, and maintained at 28°C. Throughout the duration of the experiment, mosquitoes were provided with a cotton pad soaked in a 6% sucrose solution ad libitum, except for the day before feeding and the day before forced salivation, in which mosquitoes were provided only with water ad libitum. To evaluate whether feeding on a blood meal spiked with a mixture of incomplete virus populations can result in virus infection and transmission by mosquitoes, body and saliva samples were collected and tested by virus isolation. Briefly, 12 to 15 d after the blood meal, mosquitoes were anesthetized using a semi-permeable CO₂ pad, wings and legs were removed with forceps, and mosquitoes were forced to salivate by inserting their proboscis inside a 20 μL filter tip pre-filled with 7 μL of a 1:1 FBS-50% sucrose mixture for at least 1 h. After salivation, mosquito bodies were collected and stored at −80°C until further processing. Data corresponding to mosquitoes from groups #1, #2, and #3 derive from a single biological experiment. Data corresponding to mosquitoes from groups #4 and #5 derive from two independent experiments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description and mean titers of the virus stocks used for blood meals in the mosquito experiment.

https://doi.org/10.1371/journal.pbio.3001870.t001

Virus isolation

Saliva samples were directly added onto 3 × 10⁴ BSR-T7/5 cells/well previously seeded in 96-well plates (groups #1, #2, and #3) or used to prepare four serial 10-fold dilutions, starting at 1:10 (groups #4 and #5). In the latter case, 50 μL of each dilution were added onto 3 × 10⁴ BSR-T7/5 cells/well previously seeded in 96-well plates. Mosquito bodies were suspended in 100 μL of complete G-MEM supplemented with an extra 1% antibiotic/antimycotic, homogenized using a pellet pestle and handheld pellet pestle motor (Daigger Scientific, USA), and clarified by centrifugation at maximum speed. The supernatants were diluted with additional 200 μL of supplemented G-MEM and used to prepare eight serial dilutions (5-fold the first dilution and 10-fold the remaining dilutions). Of each dilution, 50 μL were added onto 3 × 10⁴ BSR-T7/5 cells/well previously seeded in 96-well plates. For both saliva and body samples, the expression of the respective FP in plaques of infected cells was monitored by fluorescence microscopy 24 to 72 h post-infection and used as the readout method.

Modeling the fraction of infected and co-infected cells

To model the relationship between MOI and the fraction of (co-)infected cells, we considered predictions of three models. Here we describe the models briefly. A complete description of the models and the code are provided as S1 File. All three models use the MOI as utilized in the experiment (λ) and the relative inoculum frequency f of the two variants, denoted with G for iRVFV-SL-eGFP and C for iRVFV-SL-mCherry2, to predict four frequencies: the fraction of uninfected cells, ; the two fractions of cells infected by one virus variant, and ; the fraction of co-infected cells, P(G⋂C). The sum of these four fractions is one, as they represent all possible infection outcomes in this setup. Note that for simplicity, we only plotted the total fraction of infected cells (i.e., ) and co-infected cells (P(G⋂C)) in the main figure (Fig 2G). As there were no indications to the contrary, we assumed f_G = f_C = 0.5 for all models.

Model A predicts the four fractions based on the MOI following a previously described model based on the Poisson distribution [49]. This model has no free parameters to fit and assumes each virus particle has a complete set of genome segments. The four fractions are: where λ_G = f_Gλ and λ_C = f_Cλ.

Models B and C are both more complex and were not solved analytically, but model predictions were determined by an iterative approach (with 10⁵ iterations of cellular infection performed for all model estimates). Model B introduces non-selective packaging of genome segments. For both virus variants, the number of infecting virus particles per cell (Λ_G or Λ_C) was drawn from a Poisson distribution with mean λ_G or λ_C (using the rpois function in R). However, here λ_G = ψf_Gλ and λ_C = ψf_Cλ, where the free parameter ψ is the probability of infection per administered MOI unit. We introduced this parameter because incomplete particles cannot infect on their own, potentially giving rise to a discrepancy between the administered and actual MOI. Next, we drew the number of copies of genome segment S for each virus variant present in that cell (θ_S,G or θ_S,C) from a binomial distribution with a probability of success set to 0.5 and a number of trials 3Λ_G or 3Λ_C (using the rbinom function in R). To justify these values for the binomial distribution, consider that (i) in the absence of quantitative information on the distribution of genome segments over virus particles for this particular experiment, we assume both segments (S and L) are present at equal frequencies in both virus populations (G and C) and (ii) that up to three genome segments can be loaded into a virus particle. The number of L genome segments present in a cell can then be determined by subtraction:

Based on the complete distribution of genome segments that has infected the cell, we can now determine the fate of the cell. : At least one genome segment (S, L, or both) is not present. : Both genome segments are present, but all copies of the S segment are from the G variant. : Both genome segments are present, but all copies of the S segment are from the C variant. P(G⋂C): Both genome segments are present, and S segments from the G and C variants are present, i.e., both fluorescent markers are present. To fit this model to data, the free parameter ψ needs to be estimated.

Model C is identical to model B, except that it adds heterogeneous susceptibility of host cells to RVFV. These differences in susceptibility are likely to arise for many reasons, for instance, differences in the cell growth phase. Heterogeneous susceptibility is introduced by allowing the probability of infection to follow a beta distribution over cells. The beta distribution is a versatile probability distribution that has been previously used for this purpose [50]. For this model, λ_G = ζψf_Gλ and λ_C = ζψf_Cλ, where ζ is a stochastic variable that follows a beta distribution over cells, with shape parameters α and β (using the rbeta function in R to draw a new value for each iteration). To fit this model to data, the free parameters ψ, α, and β need to be estimated.

To estimate model parameters for models B and C, we minimized the negative log likelihood (NLL) using a stochastic hill climbing algorithm, with 100 independent runs initiated from randomly selected starting points in a larger parameter space. The NLL was determined from the multinomial likelihood of the four cell fractions for each dose using the dmultinom function in R, and then summing NLLs over all MOI values used in the experiment. Model selection was performed using the AIC.

Modeling virus spread and the relationship between MOI and co-infection

We generated a simple simulation model of viral spread to consider the impact of non-selective packaging and infection by incomplete virus particles on within-host dynamics. Here we provide a brief summary. A complete description of the model and the code are provided as S2 File. We modeled the number of infected cells over a total of g_max viral generations, in a fixed number of κ cells. Each cell produces φ virus particles and the probability that a virus particle will infect a cell in the next round of replication is ρ. Virus particles are assumed to remain infectious for a short period of time and therefore can only infect cells in the subsequent round of infection. At generation g₀, we assumed that only a single cell is infected. In the next round of infection, the mean number of infecting virus particles per cell (i.e., the MOI) will be λ_g+1 = φρi_g/n_g, where i is the number of infected cells and n is the number of uninfected cells. The MOI follows a Poisson distribution over cells, with a random value being drawn for each of the n_g cells (using the rpois function in R).

Next, the genome segment content of each infecting virus particle must be determined. For a non-selective random packager, there are eight genome segment contents possible {RNA1, RNA2, RNA3}: {0,0,0}, {1,0,0}, {0,1,0}, {0,0,1}, {1,1,0}, {1,0,1}, {0,1,1}, and {1,1,1}, with 0 indicating a segment is absent and 1 indicating it is present in one or more copies. The frequency of these eight types was set to values approximating the empirical distribution to model a virus with non-selective packaging that allows co-infection (for instance, {0.5, 0.09, 0.09, 0.09, 0.06, 0.06, 0.06, 0.05} for mammalian cells). The sample function, set to sampling with replacement, was then used to determine the identity of each infecting virus particle, with the probabilities for each outcome weighted by the frequency of the virus particle type. Once the genome segment content of all virus particles infecting a cell is known, we can determine whether the cell will become productively infected and transition from n to i, which only occurs when three segment types are present.

To consider a virus employing non-selective packaging without co-infection using this model, we wanted to block the contribution of all incomplete virus particles to spread. We then simply reset the frequency of the eight virus particle types to {0.95, 0, 0, 0, 0, 0, 0, 0.05}, in effect making all incomplete virus particles empty and inert particles. To consider a virus employing a selective genome packaging strategy, we made all virus particles to have a complete set of genome segments. However, this raises a complication: Making all virus particles complete could require a greater total number of genome segments than used in the viruses with non-selective packaging (i.e., if only one copy of each segment is present, then the incomplete particles will require appreciably fewer genome segments to assemble). To make the comparison fair, we therefore determined the minimal total number of genome segments present for the viruses with non-selective packaging and used this number to cap the production of virus particles for the virus with selective packaging (i.e., the same number of genome segments is allocated only to complete particles). For example, in the case of the virus employing selective packaging in mammalian cells, the frequency of virus particle types became {0.74, 0, 0, 0, 0, 0, 0, 0.26}. We then ran the model and plotted the frequency of infected cells over time to determine performance. Note that we also considered the effect of free parameter values, as described in S1 Text and S2 and S3 Figs and S3 File.

To consider the relationship between MOI and the fraction of cells infected only by incomplete particles, we first considered the trends for the simulations in a fixed number of cells described above. However, if we fix the mean MOI, we can consider model predictions systematically over a wide range of MOI values. To test the generality of the observations obtained, we randomly drew values for the frequency of each segment type from a uniform distribution using the runif function. Then, we normalized these values by the sum of all drawn values and considered model predictions (S2 File). To generate reproducible predictions and to limit the computational resources needed, the number of iterations (i.e., inoculated cells simulated) was dependent on the MOI (λ): λ≤1, n = 10⁴; 1<λ≤10, n = 5 × 10³; λ>10, n = 500.

Data analysis and visualization

All model predictions were performed and plotted in R Statistical Software [51] version 4.2.1. Prism 9 (GraphPad Software) was used to generate graphs of all the remaining results. Sample size varied per experiment and is indicated in each figure legend.