Defective viral genomes from chikungunya virus are broad-spectrum antivirals and prevent virus dissemination in mosquitoes

Defective viral genomes (DVGs) are truncated and/or rearranged viral genomes produced during virus replication. Described in many RNA virus families, some of them have interfering activity on their parental virus and/or strong immunostimulatory potential, and are being considered in antiviral approaches. Chikungunya virus (CHIKV) is an alphavirus transmitted by Aedes spp. that infected millions of humans in the last 15 years. Here, we describe the DVGs arising during CHIKV infection in vitro in mammalian and mosquito cells, and in vivo in experimentally infected Aedes aegypti mosquitoes. We combined experimental and computational approaches to select DVG candidates most likely to have inhibitory activity and showed that, indeed, they strongly interfere with CHIKV replication both in mammalian and mosquito cells. We further demonstrated that some DVGs present broad-spectrum activity, inhibiting several CHIKV strains and other alphaviruses. Finally, we showed that pre-treating Aedes aegypti with DVGs prevented viral dissemination in vivo.

Introduction Chikungunya virus (CHIKV) is a positive stranded RNA virus belonging to the alphavirus genus (Togaviridae family) that is responsible for chikungunya fever, a dengue-like syndrome with severe joint pain. This arthropod-borne virus (arbovirus) transmitted by Aedes spp. mosquitoes re-emerged in the last 15 years, posing a serious public health threat by causing two major worldwide epidemics with close to 8 million cases [1][2][3] and other frequent localized outbreaks. Its emergence in the Indian Ocean islands in [2005][2006] was caused by the Indian Ocean lineage strain (CHIKV-IOL) derived from the Eastern-Central-Southern African lineage [1][2][3]; while the Caribbean strain originated from the Asian lineage and caused the outbreak in the Caribbean states and the Americas in 2013-2014 [2,3]. Because there are currently no CHIKV treatment or licensed vaccines available, vector control strategy remains a main axis of prevention and outbreak control.
Along with high mutation rates, recombination is another main driving force of RNA virus evolution. Non-homologous recombination can give rise to truncated and/or rearranged viral genomes called defective viral genomes (DVGs). First described in influenza virus in 1954 by Von Magnus [4], DVGs have since been described in all viral families [5,6], generally when virus is passaged at high multiplicity of infection (MOI) conditions that favor the appearance of defective genomes requiring helper function from full-length virus. Because they lack part of their genome or its encoded functions, DVGs must co-infect cells with their parental virus, in order to take advantage of the proteins encoded by the full-length virus. Hence, DVGs that hijack the replication machinery or use proteins encoded by the parental virus, may compete with wild-type virus for resources, which can result in inhibition of the parental virus [7,8]. These types of DVGs are often referred to as defective interfering particles. Furthermore, many DVGs have a strong, immunostimulatory potential both in mammals and invertebrates [7,[9][10][11]. DVGs were identified in sera from patients suffering from acute dengue virus infection, but their pathophysiological role remains unknown [12]. In humans, their presence correlates with milder disease and better outcome in influenza virus [13] and respiratory syncytial virus infections [10]. All these reasons explain the recent renewed interest in using DVGs as antiviral therapy [14,15].
Only a few studies cover alphavirus DVGs. In Sindbis virus, a low-fidelity polymerase was shown to recombine at higher rate, overproducing DVGs that correlated with interference [16]. In invertebrates, alphavirus DVGs (Sindbis virus and CHIKV) were shown to be a template for the viral DNA form, which is key in modulating the antiviral immune response and establishing persistent infections in insects [9]. However, no studies have focused on developing antivirals derived from alphavirus DVGs.
In this work, we document and characterize naturally occurring DVGs bearing deletions, arising during CHIKV infection in vitro and in vivo. From these, we selected the DVGs based on their frequency and recurrence between replicates and/or their ability to be carried over multiple passages. We show that natural DVGs have a strong antiviral activity on CHIKV both in mammalian and mosquito cells in vitro, and demonstrate that interfering DVGs can be broad-spectrum against other alphaviruses. Finally, we show that DVGs can be introduced into the mosquito vector prior to infection to inhibit arbovirus dissemination.

Generation of defective viral genomes (DVG) by in vitro high MOI passage
In order to capture DVGs containing deletions as they arise in cell culture, the Caribbean strain of CHIKV (CHIKV Carib) was serially passaged in triplicate at high MOI in mammalian (Vero, Huh7) and mosquito (Aag2, U4.4) cells. RNA was extracted from the clarified supernatant of each replicate at each passage and RNA deep sequencing was performed to identify DVGs. As expected, the amount of DVGs between the first and last passages increased by between 1 and 5 orders of magnitude in all cell types, and the rates and oscillations differed with each cell type ( Fig 1A).

Characterization of the DVGs generated in vitro
Next, we pooled the data from each cell line to identify potential deletion hotspots in the different cell types. The resulting heat map (Fig 1B) reveals the frequency at which each nucleotide position was deleted (regardless of the total length of the deletion), where the x-axis indicates the nucleotide position along the full CHIKV genome. Each row represents a different cell type. We observed clear differences in regions with higher deletion frequencies between cell types, suggesting that the same viral strain generates different deletion variants depending on the host environment. In Vero and Huh7 cells, deletions were abundant in the region spanning the non-structural proteins nsP1-nsP2, yet the hotspot profile was shorter in HuH7 compared to Vero cells.  1B). Taken together, these results suggest that both the virus strain, as well as the cellular environment, influence the generation and maintenance of DVGs during passaging in vitro.
To visualize the different DVGs in these viral populations, the specific start (x axis) and stop (y axis) breakpoints of individual DVGs with deletions of >100 nucleotides were plotted ( Fig 1C). The analysis revealed predominant DVGs forming three clusters (cluster A, B and C) based on the location and the size of the deletion that was observed in all cell types, while a fourth cluster (cluster D) was more prominent in mosquito cells. Cluster A presented relatively small deletions of approximately 2500 nt, between the nsP1 and nsP2 genes. Cluster C presented large deletions of over 9000 nt from the end of nsP2 to the beginning of the 3'UTR, and for cluster B, an even larger deletion spanning nsP1 to the 3'UTR. In Aedes mosquito cells (Aag2 and U4.4), another cluster D was present with deletions of approximately 5000nt between nsP3 and the 3'UTR.

Generation of chikungunya DVGs in mosquitoes in vivo
To see if DVGs were also generated during in vivo mosquito infection, we fed Aedes aegypti mosquitoes an infected blood meal containing 10 6 PFU/ml of the Caribbean strain of CHIKV. After allowing the infection to disseminate throughout the mosquitoes over seven days, we sacrificed and dissected ten of them to collect individual organs: midgut, abdominal wall (body), thorax, heads and legs/wings (Fig 2A). RNA was extracted and sequenced, and the data were analyzed using the BBmap pipeline. Similar DVG patterns to the data generated in Aag2 cells ( Fig 1C) were observed, with four distinct clusters ( Fig 2B). One mosquito was not infected. In two mosquitoes out of the nine remaining (Fig 2C), we found the same DVG (named CM1) in several organs (midgut, body, head and legs/wings for one mosquito and midgut, body and head for the other mosquito). No other mosquitoes, nor the viral stock used to infect them,

PLOS PATHOGENS
Defective viral genomes from chikungunya virus contained this DVG-CM1. These results indicate that DVGs are readily generated in the mosquito host, and that they share similar deletion profiles as what was observed in cell culture.

Chikungunya DVG candidates are non replicative
In the previous sections we uncovered all of the possible DVGs generated during virus infection. Our goal was to down-select, from the hundreds of individual DVGs, those that would most likely and efficiently compete with wild-type virus. In other words, to identify which of the total DVGs would be the most potent defective interfering particles. Our rationale was that such DVGs would occur more frequently at high MOI passage, and would increase to higher frequency over time as the DVG competes with wildtype virus for resources. Thus, we selected DVGs based on: having high frequency, being maintained throughout the passage series and occurring in several replicates.
In addition to DVG-CM1, carrying a deletion from nucleotide 2695 (middle of nsP2) to nucleotide 5358 (end of nsP3), we selected and cloned another 19 DVGs, from the cell passage deep sequencing data of CHIKV Carib (described above) and CHIKV IOL, along with representatives from all four clusters ( Fig 3A). We named DVGs according to their virus strain of origin (C for CHIKV Carib and I for CHIKV IOL) and the cell in which they were generated (V for Vero, H for Huh7, A for Aag2 and U for U4.4, M for in vivo mosquito). Their genomic composition is shown in Fig 3A and their exact deletions are listed in Table 1. Three additional candidates did not belong to any defined cluster, but were present at high frequency in several replicates and were maintained throughout passages (Table 1).
To confirm that the selected DVGs were indeed defective genomes, unable to self-replicate in absence of virus, we transfected in vitro transcribed RNA of each DVG or wild-type virus in 293T cells. We harvested each DVG-or control-transfected cells at 8, 20, 28, and 44 hours post transfection, extracted RNA from the cells and performed an RT-qPCR. Contrary to CHIKV Carib full-length virus RNA that had an increasing RNA copy number throughout the timepoints, all DVG RNA copy numbers decreased over time, reflecting the progressive degradation of transfected RNA, and demonstrating that none of the candidate DVGs could selfreplicate (S1 Fig).

Chikungunya DVGs interfere with wild-type virus in vitro
We next tested the interfering activity of the 20 DVGs derived from either mammalian cell culture ( Fig 3B) or mosquito host environments ( Fig 3C) in mammalian cells. We chose 293T cells for their high transfectability, and a CHIKV Carib virus expressing Gaussia luciferase under a sub-genomic promoter (CHIKV Carib-Gluc) as the target virus for rapid quantification. To test the DVGs that were derived from mammalian or mosquito cell culture, 293T cells were transfected with a mix of in vitro transcribed RNA corresponding to one DVG of interest (or a control RNA) and the CHIKV Carib-Gluc full length virus RNA at a 1:1 or 10:1 molar ratio. Supernatants were harvested at 48 hours to measure luciferase activity as a surrogate measure for wild-type virus replication, since titers and luminescence correlated (S2 Fig). At a 1:1 molar ratio, most DVGs did not reduce wild-type virus luciferase expression, but transfection of IV1 and IH1 showed a modest decrease (p< 0,05). However, increasing the amount of DVG RNA to 10 times that of CHIKV Carib-Gluc RNA significantly inhibited virus replication in nearly every case by 1 to 3 orders of magnitude (Fig 3B and 3C). Of note, the smallest DVGs with the largest deletions (CH2 and CH3 from cluster B) had no, or little, interference activity at the highest DVG:CHIKV Carib-Gluc molar ratio, with a decrease of less than 1 log ( Fig 3B). These results show that DVGs derived from both mammalian or mosquito cell culture can inhibit virus replication in mammalian cells when introduced exogenously.

Defective viral genomes from chikungunya virus
To determine whether these mammalian and mosquito cell-derived DVGs could also inhibit virus replication in mosquito cells, we repeated the experiment in Aedes albopictus cells (U4.4). As for 293T cells, no interference was observed at a 1:1 molar ratio; but considerable inhibition with decreases of 1 to 2 log of luciferase activity was observed at 10:1 ratios for all of the DVGs that previously inhibited virus in mammalian cells, except for IV4. As seen in 293T cells, no interference was observed for DVG CV3 in mosquito cells (p>0,05) (Fig 3D), which along with CH3, were the only two of 20 DVG candidates to fail to inhibit virus in any condition. On the other hand, most of the mosquito cell-derived DVGs that could inhibit virus in 293T cells completely lost their interference activity in mosquito cells; only DVG CA2, CA3 and CM1 significantly reduced luciferase activity by 1 to 2 log in U4.4 cells at a molar ratio of 10:1 (Fig 3E).

Chikungunya DVGs can be broad-spectrum inhibitors
The CHIKV Indian Ocean lineage belongs to the East, South and Central African (ECSA) genotype and has approximately 7% nucleotide divergence with the CHIKV Caribbean strain that belongs to the Asian genotype [17]. In the previous experiments, DVGs derived from either the Indian Ocean Lineage or Caribbean strains were tested against the Caribbean strain virus expressing Gaussia luciferase. Importantly, all of the CHIKV IOL-derived DVGs inhibited the Caribbean strain, showing that DVGs can act broadly within the same virus species

PLOS PATHOGENS
Defective viral genomes from chikungunya virus (Fig 3B-3E). To extend these results, we repeated the same experiment in 293T using CHIKV IOL strain as the target virus. In this case, we measured inhibition by virus titer instead of luminescence. Most DVGs resulted in reduced virus titers, either modestly or significantly, whether they were identified from mammalian ( Fig 4A) or mosquito (Fig 4B) environments. We next investigated whether CHIKV DVGs could inhibit even more broadly within the alphavirus genus. We carried out the same experiment in 293T cells with O'nyong'nyong virus (ONNV, CHIKV's closest relative) or Sindbis virus (SINV, a very distant relative) as the targets. While the inhibitory effect was lost for most DVGs, some DVGs still significantly inhibited O'nyong'nyong virus (CV2 and CH1) (Fig 4C). Of the mosquito-derived DVGs, CM1 maintained inhibitory activity with close to no detectable O'nyong'nyong virus titers (Fig 4D). When tested against the distantly related Sindbis virus, interference activity was lost for all DVGs, but for CM1 that inhibited Sindbis virus by 2 logs (Fig 4E and 4F). These results confirm that some DVGs derived from one alphavirus can have inhibitory activity on different strains, including closely, and in some cases more distantly related viruses from the same genus.

Chikungunya DVGs block viral dissemination in vivo, in Aedes aegypti mosquitoes
After demonstrating that DVGs can be used to limit/inhibit infection in vitro, we tested if their interference activity could be used to block infection or dissemination in the mosquito host.
To do so, we injected purified RNA of the CV4, IH1 or CM1 DVGs, a control RNA, or PBS into Aedes aegypti mosquitoes 2 days prior to feeding them with a blood meal containing the CHIKV Carib-Gluc virus. Five days post infection, mosquitoes were sacrificed and midguts were dissected from the rest of the carcass (Fig 5A). Virus replication was measured in each mosquito by quantifying luciferase activity. Replication in the midgut was similar in all mosquitoes regardless of whether interfering DVG candidates were present or not. However, replication was significantly reduced in CV4-and CM1-treated mosquito carcasses (Fig 5B). Virus in midguts (representing infection) and carcass (a proxy for viral dissemination) was then quantified, and classified as positive or negative, depending on whether infectious viruses were detected or not (limit of detection 30 PFU per organ). All groups had a similar proportion of positive midguts (between 81 and 91,7%) (Fig 5C), confirming that the number of infected mosquitoes following bloodmeal feeding was similar for each group. On the other hand, virus dissemination was significantly impacted by the presence of DVGs: mosquitoes injected with either CV4, IH1 or CM1 DVGs had lower dissemination rates compared to PBS injected mosquitoes (16,7%, 47,6% and 41,7% of total infected mosquitoes in each group) (Fig 5C). These results confirm that DVGs can impact viral replication and dissemination in vivo, in the mosquito host.

Discussion
Described in all RNA virus families [5,6] and often considered a waste product of viral replication, defective interfering particles, and more broadly speaking DVGs, have recently garnered attention for their possible use as antivirals [6,15,[18][19][20][21][22]. Indeed, the presence of DVGs in natural human infections correlates with milder disease in clinical studies on respiratory syncytial virus and influenza virus [10,13]; and influenza defective interfering particles protect mice against lethal challenge [18,19,23]. In this study, we aimed to characterize, as widely as possible, the DVGs bearing deletions that arise during CHIKV infection in vitro in mammalian and mosquito environments and in vivo in its mosquito host. We show that chikungunya DVGs arise in all environments, in vitro and in vivo, but their type and abundance can depend on the host and cell type, and on the virus strain, highlighting that both cellular environment

PLOS PATHOGENS
Defective viral genomes from chikungunya virus and the viral genome bear determinants of DVG generation. For example, cluster D was strongly represented in both Aedes spp. mosquito cells, even though the deletion did not occur at the exact same position, but it was rare in mammalian cells (Fig 1C).
To find DVGs that could be useful as antivirals, we selected the most redundant DVGs, that arose in several conditions and were propagated over passages, which we hypothesized to be indicative of higher fitness and the ability to be packaged by wild-type virus, which are important characteristics of interfering DVGs. When used in higher quantity relative to wild-type virus, most candidates showed promising inhibiting capacity. However, our analysis could not identify a single or clear reason why one DVG candidate was better than another. Of note, because smaller genomes should be replicated more quickly [7,8], the conventional dogma proposes that the smallest DVGs would most efficiently hijack the parental virus replication machinery and outcompete wild-type virus. However, in our work, the smallest DVGs

PLOS PATHOGENS
Defective viral genomes from chikungunya virus identified (two DVGs from cluster B) had no or very little interference activity. This observation implies that strict competition for the replicase complex and replication speed is not the only mechanism by which DVGs interfere with parental virus. It is possible that the DVGs described here also have immunostimulatory activity as has been described for other viruses [7,[9][10][11][24][25][26][27][28][29]. A specific sequence or RNA structure might render one DVG more immunostimulatory. Additionally, they may more effectively compete for structural proteins or host ressources [8,30].
Of the DVGs generated in Aedes aegypti in vivo, CM1 attracted our attention because it could cross bottlenecks. During a natural viral infection, after an infectious blood meal, the midgut is the first organ infected before the virus reaches the hemocoel to disseminate to all other organs [31]. Previous work has shown that virus exit from the midgut is the main population bottleneck during mosquito infection with a drastic reduction in population size, followed by egress from the salivary glands, a mandatory step for viral transmission to the mammalian host through infected saliva [31][32][33][34][35]. CM1 was newly generated in the midguts of two independent mosquitoes and found to disseminate to the body wall, head and legs/wings of these same mosquitoes. The possibility that CM1 was generated de novo in other organs is unlikely because this DVG did not appear in any organs of any other mosquitoes that had not generated it in the midgut. It is not clear how or why CM1 is able to cross the midgut bottleneck. One possibility is that it belongs to a collective infectious viral unit, a structure that simultaneously contains and transports multiple viral genomes to a single cell, such as polyploid virions, aggregates of virions or virion-containing lipids vesicles [36,37]. The possibility of collective infectious unit containing DVGs has already been proposed. Indeed, Leeks et al. modeled that if the viral population contains many DVGs, the collective infectious unit could become very large. Nonetheless, the presence of interfering DVGs would disfavor transmission within the collective infectious unit compared to free virions [38]. While CM1 had an inhibiting activity at high molar ratio compared to the parental virus, its frequency in mosquito samples was very low (4 to 176 reads per million).
A valuable characteristic of some of these interfering DVGs is their broad-spectrum activity, with inhibition not only on other CHIKV strains but also on other alphaviruses. This cross reactivity had already been reported with influenza and Sendai virus, since their DVGs were shown to be efficient vaccines or vaccine adjuvants in mammalian models not only against the virus from they were derived, but also against unrelated viruses [15,[18][19][20][21][22]. From a therapeutic point of view, this is of particular interest since the risk of worldwide dissemination of arthritogenic alphaviruses is well accepted [3,[39][40][41][42][43], especially since no antivirals or vaccines against any of these viruses are currently licensed. In this context, any treatment (antiviral or vaccine) that could work across a viral genus or family would be helpful in facing outbreaks to come.
It will be important in the future to test if chikungunya DVGs could function as antivirals in mammalian in vivo models as well; however, a proper delivery system still needs to be developed. In previous work on influenza A and Sendai DVGs, authors isolated DVGs by ultra-centrifugation of a mix of wildtype virus and DVG on sucrose gradients [15,[18][19][20][21][22]. However, we could not successfully separate chikungunya DVGs from wild-type virus using these methods. New approaches, such as delivering RNA or DNA in nanoparticles or nanostructured lipids [44,45], or attempting to package DVGs in virus-like particles (VLPs) could be explored. Another major hurdle to overcome is the choice of in vivo model. Wild-type C57BL/6 could be used, but as mice develop only a transient, mild infection, it would be difficult to demonstrate an inhibitory effect for even the best DVG candidates. While the most commonly used mouse models rely on interferon α/β knockouts [46,47], the induction of innate immunity would be excluded, which may be an important mechanism of action in chikungunya DVGs [10,[24][25][26][27]29].
Finally, when injected in Aedes aegypti mosquitoes two days before an infectious blood meal, three DVGs significantly reduced viral dissemination from the mosquito midgut, thereby shifting the status of mosquitoes from being competent to incompetent vectors for CHIKV spread. This result is a proof of principle that DVGs might be useful tools in controlling CHIKV infection in the vector population. As releasing incompetent mosquitoes in the wild during an arbovirus outbreak has already been shown to be an effective epidemic control strategy [48,49], and despite the need for a simpler delivery system, it is tempting to propose that interfering DVGs could be used as a control strategy not only for CHIKV, but for any arbovirus. Furthermore, interfering DVGs are presumably safe antiviral tools because they are inert molecules that cannot self-replicate, and are only active when wild-type virus is present.
In conclusion, this work describes the different types of deleted DVGs generated during CHIKV infection in both vertebrate and invertebrate environments in vitro and in vivo in the mosquito vector. Moreover, we identified criteria to down-select the best defective interfering particle candidates able to inhibit wild-type CHIKV in vitro in both vertebrate and invertebrate hosts. An interesting observation is the broad-spectrum activity of some interfering DVGs able to interact with related alphaviruses. Finally, we show that pre-exposure to a DVG can modulate viral dissemination in mosquitoes in vivo. These results strengthen the idea that defective interfering particles might be a useful therapeutic tool for CHIKV infection as well as an efficient vector control strategy.

Ethics statement
Human blood used to feed mosquitoes was obtained from healthy volunteer donors who gave formal written consent at the ICAReB biobanking platform (

Virus
The viral stocks were generated from CHIKV infectious clones derived from the Indian Ocean lineage, ECSA genotype (IOL; described in [50]) or from the Caribbean strain, Asian genotype (Carib; described in [51]). To test DVG interference activity, a CHIKV infectious clone containing the Gaussia luciferase gene under a subgenomic promoter was used (obtained from Andres Merits). In vitro transcription (IVT) with SP6 mMESSAGE mMACHINE kit (Invitrogen) was performed on Not I linearised plasmids prior to transfection in BHK using lipofectamine 2000 (Invitrogen). After one passage in Vero cells, the stocks were titered and kept at -80˚C before use.

Defective viral genomes from chikungunya virus
The O'nyong'nyong virus (ONNV) infectious clone, under an SP6 promoter, was obtained from Andres Merits. Sindbis virus was generated from the pTR339 infectious clone described in [16].

Plaque assay
Viral titration was performed on confluent Vero cells plated in 24-well plates, 24 hours before infection. Ten-fold dilutions were performed in DMEM alone and transferred onto Vero cells. After allowing infection, DMEM with 2% FCS, 1% P/S and 0,8% agarose was added on top of cells. Three days post infection, cells were fixed with 4% formalin (Sigma), and plaques were manually counted after staining with 0,2% crystal violet (Sigma).

Viral passages
Cells were seeded in 24-well plates to reach approximately 80% confluency the next day. For passage 1, virus was diluted in PBS to obtain a multiplicity of infection of 5 PFU/cell (high MOI). Cells were incubated with the viral inoculum at 37˚C for 1 hour. Following virus adsorption, the inoculum was removed, the infected cells were washed with PBS, and replaced with the appropriate cell culture medium containing 2% (v/v) FCS. At 48-72 hour post infection, supernatant was harvested and clarified by centrifugation (12 000 x g, 5 min). The following passages were performed blindly, using a high volume (300 μl) of the clarified supernatant from previous passage to infect naïve cells followed by the same procedure. A total of 10 passages were performed. Each passage was titered by plaque assay to determine at which passages DVG accumulation may have resulted in interference. At least three replicates were performed per cell type.

Sequencing
RNA of 100 ul of each sample supernatant was extracted using TRIzol reagent (Invitrogen) or ZR viral RNA kit (Zymo) following the manufacturer's protocol. RNA was eluted in 15-30 μL nuclease-free water. After quantification using Quant-iT RNA assay kit (Thermo Fisher Scientific), RNA libraries were prepared with NEBNext Ultra II RNA Library kit (Illumina). Multiplex oligos (Illumina) were used during library preparation. The quality of the libraries was verified using a High Sensitivity DNA Chip (Agilent) and quantified using the Quant-iT DNA assay kit (Thermo Fisher Scientific). Sequencing of the libraries (diluted to 1 nM) was performed on a NextSeq sequencer (Illumina) with a NextSeq 500 Mid Output kit v2 (Illumina) (151 cycles).

Next generation sequencing data analysis
All analyses were performed using BBTools suite (Bushnell B.-sourceforge.net/projects/ bbmap/). First, fastq files generated from sample sequencing were trimmed for low-quality bases and adaptors using BBDuk. Then, alignment was performed using BBmap and the appropriate CHIKV reference sequence (Carib-GenBank accession no. LN898104.1, IOL-GenBank accession no. AM258994). BBMap's variant caller CallVariants was used to report deletion events, and the overall DVG frequency per sample was calculated as the sum of the number of junction read counts (n) corresponding to each DVG and normalised as 5/3 × n/N, where N denotes the 98th percentile of the coverage per position. Multiplying by 5/3 aims to correct for the detection limit. Indeed, even if reads are 150 nucleotides long, aligned portions are usually only 30 to 120 nucleotides long. This implies that a read starting 30 or less nucleotides upstream of the breakpoint will be aligned on the right side but not on the left side of the

PLOS PATHOGENS
Defective viral genomes from chikungunya virus breakpoint, and thus, will not be considered as a DVG. Consequently, we are missing (30+30)/ 150 = 60/150 = 2/5 of the deletions, and the counts must be multiplied by 5/3 as a correction factor. Heatmaps illustrate the deletion score per nucleotide position based on deletion events removing that particular position. Specifically, scores were computed as the sum of the number of reads per million reads (RPM) supporting the deletion of a specific nucleotide position. For plotting start/stop breakpoints, deletions with lengths below 10 nucleotides were discarded.

Cloning selected defective genomes
Defective genomes selected from passages were cloned in the CHIKV infectious clones (under SP6 promoter) corresponding to their strain, using the previously described In Vivo Assembly (IVA) method [52] or using In-Fusion reagent (Takara Bio Reagent). Primers were designed using SnapGene software, with a melting temperature of 60˚C, and obtained from Integrated DNA Technology (IDT). 50 μl PCR using either Phusion high fidelity DNA polymerase (Thermo Fisher) or Q5 DNA polymerase (NEB) was carried out with a melting temperature of 57˚C for Phusion enzyme or 65˚C for Q5 polymerase. 18 cycles were carried out and the PCR products were then Dpn I (Thermo Fisher)-treated for at least 2 hours to remove the plasmid template and purified (Macherey Nagel PCR and gel purification kit). When IVA technique was used, 2ul of the PCR products were directly transformed in XL10-Gold extra competent cells (Stratagene) according to supplied protocol. When In-Fusion reaction was needed, PCR products were ligated using In-Fusion reagent (Takara Bio Reagent) following supplier instructions and 1ul was transformed in Turbo cells (NEB).

Luciferase assay to test DVG interference activity
IVTs were performed using the SP6 mMESSAGE mMACHINE kit (Invitrogen) from Not I linearized infectious clones of DVG and CHIKV Carib-GLuc (see above). The internal control of the IVT kit (pTRI-Xef) was also in vitro transcribed and used as the control RNA in all experiments. RNA production was quantified by Qubit RNA TS Assay kit (Thermo Fisher) and diluted to be at the indicated molar ratios compared to the Carib-GLuc CHIKV. Mixed DVG and full genome RNA were transfected in 293T or U4.4 cells seeded in 96-well plates with TransIT-mRNA transfection kit (Mirus) following the supplier's protocol (25 ng of Carib-GLuc CHIKV per well). The medium was changed 4 hours post transfection to avoid cellular toxicity. 48 hours after transfection, supernatant was collected and mixed with coelenterazine (supplied by Y. Janin, Institut Pasteur) at a final concentration of 0,05μM and luminescence was measured on a Tecan Infinite 200 microplate reader.
When tested with non-luciferase expressing viruses, the procedure was followed the same way but the supernatant was titered by plaque assay 48 hours post transfection.

RT-qPCR to test for DVG self-replication
50ng of DVG or CHIKV Carib IVT was transfected in 293T cells (seeded the day before in 48-well plates), in triplicate, as described above. 8 hours post transfection, supernatant was removed, and cells were washed 3 times with PBS before adding fresh medium. At 8, 20, 28 and 44 hours post-transfection, 200ul of lysis buffer from ZR 96 viral RNA kit (Zymo) was added on the cells after supernatant removal and stored at -20˚C until all time points were collected. Cellular RNA was extracted with the ZR 96 viral RNA kit and eluted in 15 μl.

Mosquitoes
Aedes aegypti female mosquitoes belonging to the 7th generation from wild mosquitoes collected in Kamphaeng Phet Province (Thailand) were grown at 28˚C, 70% relative humidity and a 12 hour light: 12 hour dark cycle, and fed with 10% sucrose.

Mosquito infections
A day before blood feeding, 6 to 10-day old females were selected and starved for a day. A blood meal containing 10 6 PFU/ml of CHIKV was offered to a pool of 60 females through a membrane feeding system (Hemotek Ltd) set at 37˚C with a piece of desalted pig intestine as the membrane. Following the blood meal, fully engorged females were selected and incubated at 28˚C, 70% relative humidity and under a 12 hour light: 12 hour dark cycle with permanent access to 10% sucrose.
After 7 days, the mosquitoes were sacrificed and dissected to collect heads, thorax, midgut, legs and wings (legs/wings), and abdominal wall (body). Each organ was ground in 300 ul of L15 supplemented with 2% FBS with Qiagen TissuLyser 2 machine, then clarified by centrifugation before titration and RNA extraction with TriZol reagent for RNA deep sequencing.

Mosquito injections
One hundred and fifty nanograms of RNA produced by in vitro transcription (as described above) and purified by phenol-chloroform was mixed with Leibovitz's L-15 medium and Cellfectin II Reagent (Thermo Fisher) following the supplier's instructions. For each condition, forty 6 to 8 day-old females were injected intra-thoracically with 300nl of this mix with Nanoject III Nanoliter Injector (Drummond scientific company). Two days later, after a night of starvation, mosquitoes were fed with an infectious blood meal of 10 6 PFU/ml of CHIKV Carib-GLuc as described above. After 5 days, mosquitoes were sacrificed; midgut and carcass were dissected and ground as mentioned before. Each sample was then tested for luciferase activity and titered by plaque assay. In vitro transcribed RNA of each DVG candidate (derived from mosquito cells) was co-transfected with CHIKV Carib-luciferase RNA at 1:1 or 10:1 molar ratio in U4.4 cells. After 48 hours, samples were titered and luciferase activity was measured (Fig 3E). (n = 3, �� p<0,01, ���� p<0,0001, ns non-significant compared to wild-type after Dunnett's multiple comparisons on one-way ANOVA test). (EPS)

PLOS PATHOGENS
Defective viral genomes from chikungunya virus