key: cord-0919726-knpctp4p
authors: Driouich, Jean-Sélim; Cochin, Maxime; Lingas, Guillaume; Moureau, Grégory; Touret, Franck; Petit, Paul Rémi; Piorkowski, Géraldine; Barthélémy, Karine; Coutard, Bruno; Guedj, Jérémie; de Lamballerie, Xavier; Solas, Caroline; Nougairède, Antoine
title: Favipiravir and severe acute respiratory syndrome coronavirus 2 in hamster model
date: 2020-07-07
journal: bioRxiv
DOI: 10.1101/2020.07.07.191775
sha: 99e95c7116ffa95d798ac6e49674ff073df8591a
doc_id: 919726
cord_uid: knpctp4p

There is a need for safe and effective antiviral molecules with which to combat COVID-19 pandemics. Recently, in vitro inhibitory activity of favipiravir against SARS-CoV-2 was reported. Here, we used a Syrian hamster model to explore the pharmacokinetics of this molecule and its in vivo efficacy against SARS-CoV-2. Results revealed that high doses (700-1400mg/kg/day) significantly reduced virus replication in the lungs accompanied by clinical alleviation of the disease. However, these high doses were associated with significant toxicity in hamsters. Favipiravir pharmacokinetics displayed non-linear increase in plasma exposure between the doses and good lung penetration. Analysis of viral genomes in vivo showed that favipiravir induced a mutagenic effect. Whilst the plasma trough concentrations observed in this study were comparable with those previously found during human clinical trials, this potential toxicity requires further investigation to assess whether a tolerable dosing regimen can be found in humans that effectively reduces virus replication.

In March 2020, the World Health Organization declared coronavirus disease 2019 (COVID-19) a 26 pandemic (WHO, 2020) . The COVID-19 outbreak was originally identified in Wuhan, China, in 27 December 2019 and spread rapidly around the world within a few months. The severe acute 28 respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of COVID-19, belongs to the 29 Coronaviridae family and is closely related to the SARS-CoV which emerged in China in 2002 (Zhu et 30 al., 2020) . After an incubation period of about 5 days, disease onset usually begins with an influenza-31 like syndrome associated with high virus replication in respiratory tracts (Huang et al., 2020 , He et al., 32 2020 . In some patients, a late acute respiratory distress syndrome, associated with high levels of 33 inflammatory proteins, occurs within one to two weeks (Huang et al., 2020) . As of 7 July 2020, more 34 than 11.6 million cases of COVID-19 have resulted in more than 538,000 deaths (Dong et al., 2020) . In 35 the face of this ongoing pandemic and its unprecedented repercussions, not only on human health but 36 also on society, ecology and economy, there is an urgent need for effective infection prevention and 37 control measures. 38

Whilst host-directed and immune-based therapies could prove useful for the clinical management of 39 critically ill patients, the availability of safe and effective antiviral molecules would represent an 40 important step towards fighting the current pandemic. As conventional drug development is a slow 41 process, repurposing of drugs already approved for any indication was extensively explored and led to 42 the implementation of many clinical trials for the treatment of COVID-19 (Mercorelli et al., 2018) . 43

However, the development of effective antiviral drugs for the treatment of COVID-19, should, as much 44 as possible, rely on robust pre-clinical in vivo data, not only on efficacy generated in vitro. Accordingly, 45 rapid implementation of rodent and non-human primate animal models should help to assess more 46 finely the potential safety and efficacy of drug candidates and to determine appropriated dose 47 regimens in humans (Chan et al., 2020 , Rockx et al., 2020 . 48

Favipiravir (6-fluoro-3-hydroxypyrazine-2-carboxamine) is an anti-influenza drug approved in Japan 49 that has shown broad-spectrum antiviral activity against a variety of other RNA viruses (Guedj et al., 50 2018 , Yamada et al., 2019 , Segura Guerrero et al., 2018 , Tani et al., 2018 , Jochmans et al., 2016 , 51 Takahashi et al., 2003 , Rosenke et al., 2018 . Favipiravir is a prodrug that is metabolized intracellularly 52 into its active ribonucleoside 5'-triphosphate form that acts as a nucleotide analogue to selectively 53 inhibit RNA-dependent RNA polymerase and induce lethal mutagenesis (Baranovich et al., 2013 , 54 Sangawa et al., 2013 . Recently, several studies reported in vitro inhibitory activity of favipiravir against 55 SARS-CoV-2 with 50% effective concentrations (EC50) ranging from 62 to >500µM (10 to >78µg/mL) 56 (Wang et al., 2020 , Jeon et al., 2020 , Shannon et al., 2020 . Based on these results, more than 20 clinical 57 trials on the management of COVID-19 by favipiravir are in progress (https://clinicaltrials.gov/). In the present study, a Syrian hamster model (Mesocricetus auratus) was implemented to explore the in vivo 59 safety and efficacy and the pharmacokinetics (PK) of several dosing regimens of favipiravir. 60

In vitro efficacy of favipiravir 62 Using VeroE6 cells and an antiviral assay based on reduction of cytopathic effect (CPE), we recorded 63 EC50 and EC90 of 32 and 52.5 µg/mL using a multiplicity of infection (MOI) of 0.001, 70.0 and >78.5µg/mL 64 with an MOI of 0.01 ( Figure S1 ) in accordance with previous studies (Wang et al., 2020 , Jeon et al., 65 2020 , Shannon et al., 2020 . Infectious titer reductions (fold change in comparison with untreated cells) 66

were ≥2 with 19.6µg/mL of favipiravir and ranged between 11 and 342 with 78.5µg/mL. Using CaCo2 67 cells, which do not exhibit CPE with SARS-CoV-2 BavPat1 strain, infectious titer reductions were around 68 5 with 19.6µg/mL of favipiravir and ranged between 144 and 7721 with 78.5µg/mL of the drug. 50% 69 cytotoxic concentrations (CC50) in VeroE6 and CaCo2 cells were >78.5µg/mL. 70

Infection of Syrian hamsters with SARS-CoV-2

Following Chan et al., we implemented a hamster model to study the efficacy of antiviral compounds 72 (Chan et al., 2020) . Firstly, we intranasally infected four-week-old female Syrian hamsters with 10 6 73 TCID50 of virus. Groups of two animals were sacrificed 2, 3, 4 and 7 days post-infection (dpi). Viral 74 replication was quantified in sacrificed animals by RT-qPCR in organs (lungs, brain, liver, small/large 75 bowel, kidney, spleen and heart) and plasma. Viral loads in lungs peaked at 2 dpi, remained elevated 76 until 4 dpi and dramatically decreased at 7 dpi ( Figure 1a ). Viral loads in plasma peaked at 3 dpi and 77 viral replication was detected in the large bowel at 2 dpi (Figure 1b and Table S1 ). No viral RNA was 78 detected in almost all the other samples tested (Table S1 ). Subsequently, we infected animals with two 79 lower doses of virus (10 5 and 10 4 TCID50). Viral RNA was quantified in lungs, large bowel and plasma 80 from sacrificed animals 2, 3, 4 and 7 dpi (Figure 1a and 1b) . Viral loads in lungs peaked at 2 and 3 dpi 81 with doses of 10 5 and 10 4 TCID50 respectively. Maximum viral loads in lungs of animals infected with 82 each dose of virus were comparable. Viral RNA yields in plasma and large bowel followed a similar 83 trend but with more variability, with this two lower doses. In addition, clinical monitoring of animals 84 showed no marked symptoms of infection but significant weight losses from 3 dpi when compared to 85 animals intranasally inoculated with sodium chloride 0.9% (Figure 1c) . 86 87 Figure 1 : Implementation of hamster model

Hamsters were intranasally infected with 10 6 , 10 5 or 10 4 TCID50 of virus. Viral replication was quantified using 89 an RT-qPCR assay. a Lung viral RNA yields. b Plasmatic viral loads. c Clinical course of the disease. Normalized 90 weight at day n was calculated as follows: (% of initial weight of the animal at day n)/(mean % of initial weight 91 for mock-infected animals at day n). Data represent mean ±SD (details in Table S1 ).

In vivo efficacy of favipiravir 93 To assess the efficacy of favipiravir, hamsters received the drug, intraperitoneally, three times a day 94 (TID). We used three doses of favipiravir: 18.75, 37.5 and 75mg/day (corresponding to 340±36, 670±42 95 and 1390±126 mg/kg/day respectively). 96

In a first set of experiments, treatment was initiated at day of infection (preemptive antiviral therapy) 97 and ended at 2 dpi. We infected groups of 6 animals intranasally with three doses of virus (10 6 , 10 5 and 98 10 4 TCID50) and viral replication was measured in lungs and plasma at 3 dpi ( Figure 2a ). When analysis 99 of virus replication in clarified lung homogenates was based on infectious titers (as measured using 100 TCID50 assay), an inverse relationship was observed between infectious titers and the dose of 101 favipiravir ( Figure 2b ). This trend was even more important when low doses of virus were used to infect 102 animals. At each dose of virus, mean infectious titers for groups of animals treated with 75mg/day TID 103 were significantly lower than those observed with untreated groups (p≤0.0001): reduction of infectious 104 titers ranged between 1.9 and 3.7 log10. For animals infected with 10 5 or 10 4 TCID50, significant 105 infectious titer reductions of around 0.8 log10 were also observed with the dose of 37,5mg/day TID 106 (p≤0.038). Drug 90% and 99% effective doses (ED90 and ED99) were estimated based on these results 107 and ranged between 31-42mg/day and 53-70mg/day respectively (Table 2 ). When analysis of virus 108 replication in clarified lung homogenates were assessed on viral RNA yields (as measured using 109 quantitative real time RT-PCR assay), significant differences with groups of untreated animals, ranging 110 between 0.7 and 2.5 log10, were observed only with the higher dose of favipiravir (p≤0.012). Once 111 again, this difference was more noticeable with lower doses of virus ( Figure 2b ). Since we found higher 112 reductions of infectious titers than those observed with viral RNA yields, we estimated the relative 113 infectivity of viral particle (i.e. the ratio of the number of infectious particles over the number of viral 114 RNA molecules). Decreased infectivity was observed in all treated groups of animals. These differences were always significant with the higher dose of favipiravir (p≤0.031) and were significant with the dose 116 of 37.5mg/day TID for animals infected with 10 5 or 10 4 TCID50 of virus (p≤0.041). We then measured 117 plasmatic viral loads using quantitative real time RT-PCR assay and found, with the higher dose of 118 favipiravir and the groups of animals infected with 10 6 or 10 4 TCID50, significant reductions of 2.1 and 119 2.62 log10, respectively (p≤0.022) (Figure 2b ). 120 

10 5 or 10 4 TCID50 of virus. Lung infectious titers (measured using a TCID50 assay) and viral RNA yields were 124 (measured using an RT-qPCR assay) expressed in TCID50/copy of ɣ-actine gene and viral genome copies/copy of ɣ-actine gene respectively. Relative lung viral particle infectivities were calculated as follows: ratio of lung 126 infectious titer over viral RNA yields. Plasmatic viral loads (measured using an RT-qPCR assay) are expressed in 127 viral genome copies/mL of plasma (the dotted line indicates the detection threshold of the assay). Data represent 128 mean ±SD. ****, ***, ** and * symbols indicate that the average value for the group is significantly lower than 129 that of the untreated group with a p-value <0.0001, ranging between 0.0001-0.001, 0.001-0.01 and 0.01-0.05 130 respectively (details in Table S2 and S3). Dose-response curves are presented in Figure S2 .

In a second set of experiments, we assessed, over a period of 7 days, the impact of treatment on the 133 clinical course of the disease using weight loss as the primary criterion ( Figure 3a ). Beforehand, we 134 evaluated the toxicity of the three doses of favipiravir with groups of four non-infected animals treated 135 from day 0 to day 3 (Figure 3b ). High toxicity was observed with the dose of 75mg/day TID with 136 significant weight loss noticed from the first day of treatment (Table S4) . We also found a constant, 137 but moderate, toxicity with the dose of 37.5mg/day TID that was significant at day 4 and 5 only. No 138 toxicity was detected with the lower dose of favipiravir. To assess if the toxicity observed with the 139 highest dose of favipiravir was exacerbated by the infection, we compared weight losses of infected 140 and non-infected animals treated with the dose of 75mg/day TID. Regardless of the dose of virus, no 141 significant difference was observed at 1, 2 and 3 dpi ( Figure S3 ). After this evaluation of favipiravir 142 toxicity, we intranasally infected groups of 10 animals with two doses of virus (10 5 or 10 4 TCID50). 143

Treatment with a dose of 37.5mg/day TID was initiated on the day of infection (preemptive antiviral 144 therapy) and ended at 3 dpi ( Figure 3a ). With both doses of virus, treatment was associated with 145 clinical alleviation of the disease (Figure 3c -d). With the dose of 10 5 TCID50, mean weights of treated 146 animals were significantly higher than those of untreated animals at 5 and 6 dpi (p≤0.031). Similar 147 observations were made with the dose of 10 4 TCID50 at 5, 6 and 7 dpi (p<0.0001). 148 

Normalized weight at day n was calculated as follows: (% of initial weight of the animal at day n)/(mean % of 154 initial weight for mock-infected animals at day n). Data represent mean ±SD. **** and * symbols indicate a 155 significant difference between treated and untreated animals with a p-value <0.0001 and ranging between 0.01-156 0.05 respectively (details in Table S2 and S4).

In a third set of experiments, treatment was started one day before infection (preventive antiviral 158 therapy) and ended at 2 dpi. We intranasally infected groups of 6 animals with 10 4 TCID50 of virus and 159 viral replication was measured in lungs and plasma at 3 dpi ( Figure 4a ). Once again, an inverse 160 relationship was observed between lung infectious titers and the dose of favipiravir (Figure 4b ). Mean 161 infectious titers for groups of animals treated with 37.5 and 75mg/day TID were significantly lower 162 than those observed with untreated groups (p≤0.002). Of note, undetectable infectious titers were 163 found for all animals treated with the higher dose. Estimated ED90 and ED99 were 35 and 47mg/day 164 respectively (Table 2 ). Significant reductions of viral RNA yields of 0.9 and 3.3 log10, were observed with 165 animals treated with 37.5 and 75mg/day TID respectively (p≤0.023). Resulting infectivity of viral 166 particle was decreased, with a significant reduction only for the higher dose of favipiravir (p=0.005). 167

Finally, we found significantly reduced plasmatic viral loads with animals treated with 37.5 and 168 75mg/day TID (p≤0.005). 169 177 ****, ** and * symbols indicate that the average value for the group is significantly different from that of the 178 untreated group with a p-value <0.0001, ranging between 0.001-0.01 and 0.01-0.05 respectively (details in Table   179 S2 and S3).

Favipiravir pharmacokinetics (PK) in a hamster model 181 We first assessed the PK and lung distribution of favipiravir in a subgroup of uninfected animals. Groups 182 of animals were treated respectively with a single dose of favipiravir administrated intraperitoneally: 183 6.25mg, 12.5 mg and 25 mg. In each dose group, we sacrificed 3 animals at specific time points post-184 treatment (0.5, 1, 5 or 8 hours) for determination of favipiravir in plasma. Drug concentration in lung 185 tissue was determined at 0.5 and 5 hours post-treatment. Subsequently, we assessed the favipiravir 186 concentration after multiple dose in animals intranasally infected with 10 5 TCID50 of virus. Groups of 9 187 animals received the three doses evaluated for 3 days (Figure 2a ): 18.75mg/day, 37.5mg/day and 188 75mg/day TID and were sacrificed at 12-hours after the last treatment dose. Favipiravir was quantified 189 in plasma (n=9) and lung tissue (n=3). 190

Results are presented in Table 3 and Figure S4 . The single dose PK analysis showed that the maximum 191 concentration of favipiravir was observed at 0.5 hour at all doses, then plasma drug concentrations 192 decreased exponentially to reach concentrations below 10 µg/ml at 12 hours. Favipiravir PK exhibited 193 a non-linear increase in concentration between the doses. After multiple doses, trough concentrations 194 (12 hours) of favipiravir also exhibited a non-linear increase between doses. The extrapolated 12 hours 195 post-treatment concentrations after a single dose were calculated in order to determine the accumulation ratio. Accumulation ratios were respectively 6, 16 and 21 at the 3 doses, confirming the 197 non-proportional increase between doses. The average concentration after single dose administration 198 over 0 to 12-hour intervals was calculated and the respective values obtained were 10.1 µg/mL, 38.7 199 µg/mL and 100.5 µg/mL for the 3 favipiravir doses. 200

Favipiravir lung concentrations were 1.6 to 2.7-fold lower than in plasma for both administration of 201 single and multiple doses. After a single dose, the mean lung to plasma ratio ranged from 0.37 to 0.62 202 according to the time post-treatment and was similar between the 3 doses of favipiravir at 0.5 hours. 203 A high ratio 5 hours post-treatment was observed at the highest dose (25 mg) with an increase by a 204 factor 1.6 to 1.8 compared with the lower doses. After multiple doses, the lung penetration of 205 favipiravir was confirmed with a mean lung to plasma ratio ranging from 0.35 to 0.44. Favipiravir was 206 not detected in the lungs at the lowest dose (18.75 mg/day). 207 To understand which genomic modifications accompanied favipiravir treatment, direct complete 215 genome sequencing of clarified lung homogenates from animals intranasally infected with 10 6 TCID50 216 of virus and treated with the two highest doses of drug (preemptive antiviral therapy; Figure 2 ) was 217 performed. Data were generated by next generation sequencing from lung samples of four animals 218 per group (untreated, 37.5mg/day TID and 75mg/day TID). The mean sequencing coverage for each 219 sample ranged from 10,991 to 37,991 reads per genomic position and we subjected substitutions with 220 a frequency ≥1% to further analysis. The genetic variability in virus stock was also analyzed: 14 221 nucleotide polymorphisms were detected of which 5 recorded a mutation frequency higher than 10% 222 (Table S6) . 223

In order to study the mutagenic effect of favipiravir, we used the consensus sequence from virus stock 224 as reference and all the mutations simultaneously detected in a lung sample and in virus stock were 225 not considered in the further analysis (1 to 4 mutations per sample, see Table S6 ). Overall, no majority 226 mutations were detected (mutation frequency >50%), mutations were distributed throughout the 227 whole genome and almost all of them exhibited a frequency lower than 10% (Figure 5a and 5b) . sub-populations were detected in two independent animals. Notably, 18 of these shared mutations 242 were detected only with treated animals, 14 of them being non-synonymous (Table S8) is significantly different from that of the untreated group with a p-value ranging between 0.001-0.01 and 0.01-257 0.05 respectively (details in details in Table S6 and S7). e Association between lung infectious titers (measured 258 using a TCID50 assay) and frequency of non synonymous, synonymous and G→A mutations. Each dot represent 259 data from a given animal.

In the current study, we used a hamster model to assess efficacy of the favipiravir against SARS-CoV-262 2. Following infection, viral RNA was mainly detected in lungs, blood, and, to a lesser extent, in the 263 large bowel. Peak of viral replication was observed at 2-3 dpi followed by observation of significant 264 weight losses, in line with recently reported investigations that involved 6-10 weeks old hamsters 265 (Kaptein et al., 2020 , Chan et al., 2020 . Clinically, the main symptom was weight loss, observed from 266 the first day of infection and followed by recovery at 6dpi. This confirmed that the in vivo model, with 267 younger animals (4 weeks-old), is suitable for preclinical evaluation of antiviral compounds against 268 SARS-CoV-2. 269

Using a preemptive strategy, we demonstrated that doses of favipiravir of around 700-1400mg/kg/day 270 TID reduced viral replication in lungs of infected animals and allowed clinical alleviation of the disease. 271

Reduction of viral replication was greater when estimated on the basis of infectious titers than on total 272 viral RNA as previously observed in non-human primates treated with Remdesivir (Williamson et al., 273 2020) . However, the effective doses of favipiravir were higher than those usually used in rodent models 274 (≈100-400mg/kg/day) (Sidwell et al., 2007 , Smither et al., 2014 , Julander et al., 2009 , Tani et al., 2018 , 275 Oestereich et al., 2016 , Yamada et al., 2019 . This can be correlated with the high favipiravir EC50 found 276 in vitro for SARS-CoV-2. Moreover, effective doses were associated with significant toxicity in our 277 hamster model. This observed toxicity reflected only the adverse effects of favipiravir and was not 278 exacerbated during SARS-CoV-2 infection. Indeed, similar weight losses were measured among 279 infected and non-infected animals treated with the highest dose of favipiravir at 1, 2 and 3dpi. 280

In the present study, reduction of viral replication was correlated with the dose of favipiravir 281 administrated and inversely correlated with the dose of virus inoculated. In a recent study, favipiravir 282 administrated per os twice daily (loading dose of 600mg/kg/day followed by 300mg/kg/day) revealed 283 a mild reduction of lung viral RNA yields using a similar hamster model with high doses of virus (2x10 6 284 TCID50) (Kaptein et al., 2020) . These results are in accordance with ours at the lower dose of favipiravir 285 (around 340mg/kg/day TID). 286

By characterizing the dose response curve, we estimated that the dose required to reduce by 90% 287 (ED90) the level of infectious titers in lungs is in the range of 570-780mg/kg/day. In the most favourable 288 situation, where high doses were used as a preemptive therapy, favipiravir led to undetectable viral 289 replication in lung and plasma. These results showed that the use of high doses of favipiravir could 290 expand its in vivo spectrum against RNA viruses. 291

With influenza viruses, favipiravir acts as a nucleotide analogue. It is metabolized intracellularly to its 292 active form and incorporated into nascent viral RNA strands. This inhibits RNA strand extension and induces abnormal levels of mutation accumulation into the viral genome (Baranovich et al., 2013 , 294 Sangawa et al., 2013 . Recently, it was shown in vitro that favipiravir has a similar mechanism of action 295 with SARS-CoV-2 through a combination of chain termination, reduced RNA synthesis and lethal 296 mutagenesis (Shannon et al., 2020) . Our genomic analysis confirmed the mutagenic effect of favipiravir 297 in vivo. Indeed, we found that favipiravir treatment induced appearance of a large number of G→A 298 and C→U mutations into viral genomes. This was associated to a decrease of viral infectivity probably 299 because alteration of the genomic RNA disturb the replication capacity. Similar findings were described 300 in vitro and in vivo with other RNA viruses (Baranovich et al., 2013 , Guedj et al., 2018 , Escribano-301 Romero et al., 2017 , Arias et al., 2014 . Of note, we also observed a strong inverse association between 302 infectious titers in lungs and the proportion of non-synonymous mutations detected in viral 303 populations. Because random non-synonymous mutations are more deleterious than synonymous 304 mutations (Cuevas et al., 2012) , this suggests that they were randomly distributed over the three 305 positions of the codons and that no compensatory mechanism was triggered by the virus to eliminate 306 them (i.e. negative selection). Finally, the inverse correlation between lung infections titers and the 307 frequency of G→A substitutions showed that an increased proportion of these mutations beyond an 308 error threshold might be expected to cause lethal mutagenesis. 309

Genomic analyses revealed that 18 mutations detected in viral sub-populations were shared only with 310 treated animals. Two of them were located in the nsp14 coding region involved in the proof-reading 311 activity of the viral RNA polymerisation (Eckerle et al., 2007 , Ferron et al., 2018 . However, they were 312 located in the N7 MTase domain involved in viral RNA capping (Chen et al., 2013 , Ma et al., 2015 . By 313 comparison, resistance mutations selected against Remdesivir in β-coronavirus murine hepatitis virus 314 model were obtained in the RdRP (nsp12) coding sequence (Agostini et al., 2018) . Further 315 investigations are needed to assess the impact of these mutations on the antiviral effect of favipiravir. 316

Favipiravir PK in our hamster model displayed a non-linear increase in plasma exposure between the 317 doses as already reported in nonhuman primates (Madelain et al., 2017) . The observed favipiravir 318 concentration versus time profiles were in agreement with previous results of a PK study performed 319 in 7-8 week-old hamsters orally treated with a single dose of 100mg/kg of favipiravir (Gowen et al., 320 2015) . The maximum plasma drug concentration occurred at 0.5 h after oral administration, earlier 321 than in humans, and then decreased rapidly in agreement with its short half-life (Madelain et al., 2016) . 322

After repeated doses, plasma exposure confirmed non-linear PK over the entire range of doses, further 323 emphasized by accumulation ratios. The important accumulation observed at the highest dose could 324 explain in part the toxicity observed in hamsters at this dose. Favipiravir undergoes an important 325 hepatic metabolism mainly by aldehyde oxidase producing an inactive M1 metabolite and inhibits 326 aldehyde oxidase activity in a concentration-and time-dependent manner. These properties explain the self-inhibition of its own metabolism as observed in our study in which the highest dose of 328 favipiravir led to a greater increase in favipiravir concentrations (Madelain et al., 2020) . 329

A good penetration of favipiravir in lungs was observed with lung/plasma ratios ranging from 35 to 330 44% after repeated doses, consistent with its physicochemical properties. Lung exposure was also in 331 accordance with previous studies (Gowen et al., 2015) . 332

How clinically realistic are these results? To address this question we compared the drug 333 concentrations obtained in the hamster model with those obtained in patients. In 2016, a clinical trial 334 evaluated the use of favipiravir in Ebola infected patients (Sissoko et al., 2016) . The dose used in Ebola 335 infected patients was 6000mg on day 0 followed by 1200mg BID for 9 days. The median trough 336 concentrations of favipiravir at Day 2 and Day 4 were 46.1 and 25.9µg/mL, respectively. This is within 337 the range observed here in hamsters treated with the highest dose (around 1400mg/kg/day), with a 338 mean trough concentration of 29.9µg/mL. However, additional investigations are required to 339 determine whether or not similar favipiravir plasma exposure in SARS-COV-2 infected patients are 340 associated with antiviral activity. The major differences in PK between hamster and humans, and the 341 toxicity observed at the highest doses in our animal model limits the extrapolation of our results. 342 Therefore, whether safe dosing regimens in humans may achieve similar plasma exposure and 343 recapitulate the profound effect on viral replication is unknown. Further, the intracellular 344 concentration of the active metabolite was not determined and which parameter of the drug 345 pharmacokinetics best drives the antiviral effect remains to be established. 346

In summary, this study establishes that high doses of favipiravir are associated with antiviral activity 347 against SARS-CoV-2 infection in a hamster model. The better antiviral efficacy was observed using a 348 preventive strategy, suggesting that favipiravir could be more appropriate for a prophylactic use. Our 349 results should be interpreted with caution because high doses of favipiravir were associated with signs 350 of toxicity in our model. It is required to determine if a tolerable dosing regimen could generate similar 351 exposure in non-human primates, associated with significant antiviral activity, before testing a high 352 dose regimen in COVID-19 patients. Furthermore, subsequent studies should determine if an increased 353 antiviral efficacy can be reached using favipiravir in association with other effective antiviral drugs, 354 since this strategy may enable to reduce the dosing regimen of favipiravir. Finally, this work reinforces 355 the need for rapid development of animal models to confirm in vivo efficacy of antiviral compounds 356 and accordingly, to determine appropriate dose regimens in humans before starting clinical trials.

We thank Laurence Thirion (UVE; Marseille) for providing RT-qPCR systems . We thank Camille Placidi 359 (UVE; Marseille) for her technical contribution. We also thank Pr. Ernest A. Gould (UVE; Marseille) for 360 his careful reading of the manuscript and English language editing. We thank Pr Drosten and Pr Drexler 361 for providing the SARS-CoV-2 strain through the European Research infrastructure EVA GLOBAL. This 362 work was supported by the Fondation de France "call FLASH COVID-19", project TAMAC, by "Institut 363 national de la santé et de la recherche médicale" through the REACTing (REsearch and ACTion targeting 364 emerging infectious diseases) initiative ("Preuve de concept pour la production rapide In vitro determination of EC50, EC90, CC50 and infectious titer reductions 391 One day prior to infection, 5×10 4 VeroE6 cells were seeded in 96-well culture plates (5×10 4 cells/well 392 in 100µL of 2.5% FBS medium (assay medium). The next day, seven 2-fold serial dilutions of favipiravir 393 (Courtesy of Toyama-Chemical; 0.61µg/mL to 78.5µg/mL, in triplicate) were added (25µL/well, in assay 394 medium). Eight virus control wells were supplemented with 25µL of assay medium and eight cell 395 controls were supplemented with 50µL of assay medium. After 15 min, 25µL of virus suspension, 396 diluted in assay medium, was added to the wells at an MOI of 0.01 or 0.001 (except for cell controls). 397

Three days after infection, cell supernatant media were collected to perform TCID50 assay (at 398 concentration of 78.5, 39.3, 19.6µg/mL), as described below, in order to calculate infectious titer 399 reductions and cell viability was assessed using CellTiter-Blue reagent (Promega) following 400 manufacturer's intructions. Fluorescence (560/590nm) was recorded with a Tecan Infinite 200Pro 401 machine (Tecan). The 50% and 90% effective concentrations (EC50, EC90) were determined using 402 logarithmic interpolation (% of inhibition were calculated as follows: (ODsample-ODvirus control)/(ODcell control-403 ODvirus control)). For the evaluation of CC50 (the concentration that induced 50% cytoxicity), the same 404 culture conditions were set as for the determination of the EC50, without addition of the virus, then 405 cell viability was measured using CellTiter Blue (Promega). CC50 was determined using logarithmic 406 interpolation. 407 workbench software v.20 (Qiagen). A de novo contig was also produced to ensure that the consensus 509 sequence was not affected by the reference sequence. Mutation frequency for each position was 510 calculated as the number of reads with a mutation compared to the reference divided by the total 511 number of reads at that site. Only substitutions with a frequency of at least 1% were taken into account 512 for the analysis (Table S6) . 513

We conducted a nonlinear regression of infectious viral load against dose, using an Emax model, giving 515 = 0 × (1 − ( + 50 )) with 0 being infectious viral load of untreated animals. We estimated 516 50 the dose required to decrease viral load by 50%, using a coefficient to account for the high 517 sigmoidicity of the relation between dose and titers. coefficient was chosen as the one maximizing 518 likelihood of the model. We extrapolated the 90 and 99 using 90 = √9 × 50 and 99 = 519 √99 × 50 , as well as their 95% confidence interval using the delta method. 520 

Activity of T-705 in a 599 hamster model of yellow fever virus infection in comparison with that of a chemically related 600 compound, T-1106

Antiviral treatment of SARS-CoV-2-606 infected hamsters reveals a weak effect of favipiravir and a complete lack of effect for 607 hydroxychloroquine

Structural basis and functional analysis of the SARS coronavirus nsp14-nsp10 complex

Favipiravir 613 Pharmacokinetics in Nonhuman Primates and Insights for Future Efficacy Studies of

Hemorrhagic Fever Viruses

Modeling Favipiravir Antiviral Efficacy Against Emerging Viruses: From Animal 618 Studies to Clinical Trials

Ebola Virus Infection: Review of the Pharmacokinetic and Pharmacodynamic 621 Properties of Drugs Considered for Testing in Human Efficacy Trials

Drug Repurposing for Viral Infectious Diseases: How 624

Far Are We?

RNA and DNA bacteriophages as molecular diagnosis controls in clinical 627 virology: a comprehensive study of more than 45,000 routine PCR tests

Efficacy of Favipiravir Alone and 630 in Combination With Ribavirin in a Lethal, Immunocompetent Mouse Model of Lassa Fever

A SIMPLE METHOD OF ESTIMATING FIFTY PER CENT ENDPOINTS

Comparative pathogenesis of COVID-19, MERS, and 639 SARS in a nonhuman primate model

Use of Favipiravir to Treat Lassa Virus Infection in Macaques

Mechanism of action of T-705 ribosyl triphosphate against influenza virus RNA 646 polymerase

Usutu virus replication and delays disease progression in an infection model in mice

Favipiravir strikes 652 the SARS-CoV-2 at its Achilles heel, the RNA polymerase

Efficacy of orally administered T-705 on lethal avian influenza A (H5N1) 655 virus infections in mice

Treatment with Favipiravir for Ebola Virus Disease (the JIKI Trial): A Historically Controlled

Single-Arm Proof-of-Concept Trial in Guinea

Post-675 exposure efficacy of oral T-705 (Favipiravir) against inhalational Ebola virus infection in a 676 mouse model

In vitro and in vivo activities of T-705 and oseltamivir against 679 influenza virus

Therapeutic effects of favipiravir against 682 severe fever with thrombocytopenia syndrome virus infection in a lethal mouse model: Dose-683 efficacy studies upon oral administration

Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-686 nCoV) in vitro

World Health Organization. WHO Director-General's opening remarks at the media 688 briefing on COVID-19 -11

Clinical benefit of remdesivir in rhesus 693 macaques infected with SARS-CoV-2

Reevaluation of the efficacy of favipiravir against rabies 696 virus using in vivo imaging analysis

A Novel Coronavirus from Patients with Pneumonia in 700

In vivo experiments were approved by the local ethical committee (C2EA-14) and the French 410 'Ministère de l'Enseignement Supérieur, de la Recherche et de l'Innovation' (APAFIS#23975) and 411 performed in accordance with the French national guidelines and the European legislation covering 412 the use of animals for scientific purposes. All experiments were conducted in BSL 3 laboratory. 413Animal handling 414 Three-week-old female Syrian hamsters were provided by Janvier Labs. Animals were maintained in 415ISOcage P -Bioexclusion System (Techniplast) with unlimited access to water/food and 14h/10h 416 light/dark cycle. Animals were weighed and monitored daily for the duration of the study to detect the 417 appearance of any clinical signs of illness/suffering. Virus inoculation was performed under general 418 anesthesia (isoflurane). Organs and blood were collected after euthanasia (cervical dislocation) which 419 was also realized under general anesthesia (isofluorane). 420

Anesthetized animals (four-week-old) were intranasally infected with 50µL containing 10 6 , 10 5 or 422 10 4 TCID50 of virus in 0.9% sodium chloride solution). The mock group was intranasally inoculated with 423 50µL of 0.9% sodium chloride solution. 424

Hamster were intra-peritoneally inoculated with different doses of favipiravir. Control group were 426 intra-peritoneally inoculated with a 0.9% sodium chloride solution. 427Organ collection 428 Organs were first washed in 10mL of 0.9% sodium chloride solution and then transferred to a 2mL or 429 50mL tube containing respectively 1mL (small/large bowel pieces, kidney, spleen and heart) or 10mL 430 (lungs, brain and liver) of 0.9% sodium chloride solution and 3mm glass beads. They were crushed 431 using a the Tissue Lyser machine (Retsch MM400) for 5min at 30 cycles/s and then centrifuged 5min à 432 1200g. Supernatant media were transferred to a 2mL tube, centrifuged 10 min at 16,200g and stored 433 at -80°C. One milliliter of blood was harvested in a 2mL tube containing 100µL of 0.5M EDTA 434 (ThermoFischer Scientific). Blood was centrifuged for 10 min at 16,200g and stored at -80°C. 435 Quantitative real-time RT-PCR (RT-qPCR) assays

To avoid contamination, all experiments were conducted in a molecular biology laboratory that is 437 specifically designed for clinical diagnosis using molecular techniques, and which includes separate 438 laboratories dedicated to perform each step of the procedure. Prior to PCR amplification, RNA 439 extraction was performed using the QIAamp 96 DNA kit and the Qiacube HT kit and the Qiacube HT 440 (both from Qiagen) following the manufacturer's instructions. Shortly, 100 µl of organ clarified 441 homogenates, spiked with 10µL of internal control (bacteriophage MS2) (Ninove et al., 2011) , were 442 transferred into an S-block containing the recommended volumes of VXL, proteinase K and RNA carrier. 443 RT-qPCR (SARS-CoV-2 and MS2 viral genome detection) were performed with the Express one step qPCR Universal kit (ThermoFisher Scientific) using 3.5µL of RNA and 6.5µL of RT-qPCR mix that contains 445 250nmol of each primer and 75nmol of probe. Amplification was performed with the QuantStudio 12K 446Flex Real-Time PCR System (ThermoFisher Scientific) using the following conditions: 50°C for 10min, 447 95°C for 20s, followed by 40 cycles of 95°C for 3s, 60°C for 30s. qPCR (ɣ-actine gene detection) was 448 perfomed under the same condition as RT-qPCR with the following modifications: we used the Express 449 one step qPCR Universal kit (ThermoFisher Scientific) and the 50°C step of the amplification cycle was 450 removed. Primers and probes sequences used to detect SARS-CoV-2, MS2 and ɣ-actine are described 451in Table S9 . 452Tissue-culture infectious dose 50 (TCID50) assay

To determine infectious titers, 96-well culture plates containing confluent VeroE6 cells were 454 inoculated with 150μL per well of serial dilutions of each sample (four-fold or ten-fold dilutions when 455 analyzing lung clarified homogenates or cell supernatant media respectively). Each dilution was 456 performed in sextuplicate. Plates were incubated for 4 days and then read for the absence or presence 457 of cytopathic effect in each well. Infectious titers were estimated using the method described by Reed 458 & Muench (REED and MUENCH, 1938) . 459Favipiravir pharmacokinetics 460 Animal handling, hamster infections and favipiravir administrations were performed as described 461 above. A piece of left lung was first washed in 10mL of sodium chloride 0.9% solution, blotted with 462 filter paper, weighed and then transferred to a 2mL tube containing 1mL of 0.9% sodium chloride 463 solution and 3mm glass beads. It was crushed using the Tissue Lyser machine (Retsch MM400) during 464 10min at 30 cycles/s and then centrifuged 5min à 1200g. Supernatant media were transferred to 2mL 465 tubes, centrifuged 10 min at 16,200g and stored at -80°C. One milliliter of blood was harvested in a 466 2mL tube containing 100µL of 0.5M EDTA (ThermoFischer Scientific). Blood was centrifuged for 10 min 467 at 16,200g and stored at -80°C. 468Quantification of favipiravir in plasma and lung tissues was performed by a validated sensitive and 469 selective validated high-performance liquid chromatography coupled with tandem mass spectrometry 470 method (UPLC-TQD, Waters, USA) with a lower limit of quantification of 0.1 µg/mL. Precision and 471 accuracy of the 3 quality control samples (QCs) were within 15% over the calibration range (0.5 µg/mL 472 to 100 µg/mL) (Bekegnran et al., submitted). Favipiravir was extracted by a simple protein precipitation 473 method, using acetonitrile for plasma and ice-cold acetonitrile for clarified lung homogenates. Briefly, 474 50 µL of samples matrix was added to 500µL of acetonitrile solution containing the internal standard 475 (favipiravir-13C,15N, Alsachim), then vortexed for 2min followed by centrifugation for 10min at 4°C. 476The supernatant medium was evaporated and the dry residues were then transferred to 96-well plates 477 and 50 µL was injected. To assess the selectivity and specificity of the method and matrix effect, blank 478 plasma and tissues homogenates from 2 control animals (uninfected and untreated) were processed 479 at each run. Moreover, the same control samples spiked with favipiravir concentration equivalent to 480 the QCs (0.75, 50 and 80 µg/mL) were also processed and compared to the QCs samples. (Table S10) long. Libraries were built by adding barcodes, for sample identification, and primers using AB Library 500Builder System (ThermoFisher Scientific). To pool equimolarly the barcoded samples a quantification 501 step by real time PCR using Ion Library TaqMan Quantitation Kit (ThermoFisher Scientific) was 502 performed. Then, emulsion PCR from pools and loading on 530 chip was performed using the 503 automated Ion Chef instrument (ThermoFisher Scientific). Sequencing was performed using the S5 Ion 504 torrent technology v5.12 (ThermoFisher Scientific) following manufacturer's instructions. Consensus 505 sequence was obtained after trimming of reads (reads with quality score <0.99, and length <100pb 506 were removed and the 30 first and 30 last nucleotides were removed from the reads). Mapping of the 507 reads on a reference (determine following blast of De Novo contigs) was done using CLC genomics 548