An orally bioavailable broad-spectrum antiviral inhibits SARS-CoV-2 in human airway epithelial cell cultures and multiple coronaviruses in mice

Coronaviruses (CoVs) traffic frequently between species resulting in novel disease outbreaks, most recently exemplified by the newly emerged SARS-CoV-2, the causative agent of COVID-19. Herein, we show that the ribonucleoside analog β-D-N4-hydroxycytidine (NHC, EIDD-1931) has broad spectrum antiviral activity against SARS-CoV-2, MERS-CoV, SARS-CoV, and related zoonotic group 2b or 2c Bat-CoVs, as well as increased potency against a coronavirus bearing resistance mutations to the nucleoside analog inhibitor remdesivir. In mice infected with SARS-CoV or MERS-CoV, both prophylactic and therapeutic administration of EIDD-2801, an orally bioavailable NHC-prodrug (β-D-N4-hydroxycytidine-5′-isopropyl ester), improved pulmonary function, and reduced virus titer and body weight loss. Decreased MERS-CoV yields in vitro and in vivo were associated with increased transition mutation frequency in viral but not host cell RNA, supporting a mechanism of lethal mutagenesis in CoV. The potency of NHC/EIDD-2801 against multiple coronaviruses and oral bioavailability highlight its potential utility as an effective antiviral against SARS-CoV-2 and other future zoonotic coronaviruses.

INTRODUCTION
The genetically diverse Orthocoronavirinae (coronavirus, CoV) family circulates in many avian and mammalian spe- cies. Phylogenetically, CoVs are divided into 4 genera: alpha (group 1), beta (group 2), gamma (group 3) and delta (group 4). Three new human CoV have emerged in the past 20 years with severe acute respiratory syndrome CoV (SARS-CoV) in 2002, Middle East respiratory syndrome CoV (MERS-CoV) in 2012, and now SARS-CoV-2 in 2019 (1-3). In fact, all human CoV are thought to have emerged originally as zoonoses (4- 6). The ongoing SARS-CoV-2 pandemic (referred to as COVID-19, Coronavirus disease 2019) has caused over 500,000 infections and over 25,000 deaths in 199 countries.Like SARS- and MERS-CoV, the respiratory disease caused by SARS-CoV-2 can progress to acute lung injury (ALI), an end stage lung disease with limited treatment options and very poor prognoses(3, 7, 8). This emergence paradigm is not lim- ited to humans. A novel group 1 CoV called swine acute diar- rhea syndrome CoV (SADS-CoV) recently emerged from bats causing the loss of over 20,000 pigs in Guangdong Province, China (9). More alarmingly, many group 2 SARS-like and MERS-like coronaviruses are circulating in bat reservoir spe- cies that can use human receptors and replicate efficiently in primary human lung cells without adaptation(9-12). The presence of these “pre-epidemic” zoonotic strains foreshadow the emergence and epidemic potential of additional SARS-like and MERS-like viruses in the future. Given the diversity of CoV strains in zoonotic reservoirs and a penchant for emergence, broadly active antivirals are clearly needed for rapid response to new CoV outbreaks in humans and domes- ticated animals.Currently, there are no approved therapies specific for any human CoV. β-D-N4-hydroxycytidine (NHC, EIDD-1931) is orally bioavailable ribonucleoside analog with broad-spec- trum antiviral activity against various unrelated RNA viruses including influenza, Ebola, CoV, and Venezuelan equine en- cephalitis virus (VEEV)(13-16). For VEEV, the mechanism of action (MOA) for NHC has been shown to be through lethal mutagenesis where deleterious transition mutations accumu- late in viral RNA(14, 17). Thus, we sought to determine NHC’s breadth of antiviral activity against multiple emerging CoV, its mechanism of action for CoV and its efficacy in mouse models of CoV pathogenesis.

RESULTS
To determine whether NHC blocks the replication of highly pathogenic human CoV, we performed antiviral assays in cell lines with MERS-CoV and the newly emerging SARS-CoV-2. We first assessed the antiviral activity of NHC against MERS- CoV in the human lung epithelial cell line Calu-3 2B4 (“Calu3” cells). Using a recombinant MERS-CoV expressing nanolucif- erase (MERS-nLUC)(18), we measured virus replication in cultures exposed to a dose range of drug for 48hr. NHC was potently antiviral with an average half-maximum effective concentration (IC50) of 0.15 μM and no observed cytoxicity in similarly treated uninfected cultures across the dose range (50% cytotoxic concentration, CC50, >10 μM) (Fig. 1A). The therapeutic index for NHC was >100. Using a clinical isolate of SARS-CoV2 (2019-nCoV/USA-WA1/2020), we performed antiviral assays in African green monkey kidney (Vero) cells and found NHC was potently antiviral with an IC50 of 0.3 μM and CC50 of >10 μM (Fig. 1B). We then determined the antivi- ral activity of NHC against SARS-CoV-2 in the Calu-3 cells through the measurement of infectious virus production and viral genomes. We observed a dose-dependent reduction in virus titers (Fig. 1C) with an IC50 of 0.08 μM. Viral genomic RNA was quantitated in clarified supernatants by qRT-PCR (Fig. 1D). Like the effect on infectious titers, we found a dose- dependent reduction in viral genomic RNA and a similar cal- culated IC50 of 0.09 μM. Collectively, these data demonstrate that NHC is potently antiviral against two genetically distinct emerging CoV.NHC is highly active against SARS-CoV-2, MERS-CoV, and SARS-CoV in primary human airway epithelial cell culturesTo determine if NHC would be similarly antiviral in primaryhuman cells, we performed a series of studies in primary air- way epithelial (HAE) cell cultures. HAE model the architec- ture and cellular complexity of the conducting airway and are readily infected by multiple human and zoonotic CoV, includ- ing SARS- and MERS-CoV (19). We first assessed cytotoxicity of NHC in HAE treated with an extended dose range for 48hr using quantitative PCR of cell death-related gene transcripts as our metric. NHC treatment did not appreciably alter gene expression even at doses up to 100 μM (fig. S1).

We then sought to determine if NHC would inhibit clinical isolate SARS-CoV-2 replication in HAE. We observed a dose depend- ent reduction in SARS-CoV-2 infectious virus production (Fig. 2A). In MERS-CoV infected HAE, NHC substantially reduced virus production with maximal titer reduction of > 5 logs at 10 μM (average IC50 = 0.024 μM), which correlated with re- duced genomic (ORF1) and subgenomic (ORFN) RNA in paired samples (Fig. 2B). We observed similar trends in titer reduction (> 3 log at 10 μM, average IC50 = 0.14 μM) and in copies of genomic and subgenomic RNA in SARS-CoV- in- fected HAE (Fig. 2C). Thus, NHC was potently antiviral against SARS-CoV-2, MERS-CoV and SARS-CoV in primary human epithelial cell cultures without cytotoxicity.NHC is effective against remdesivir (RDV)-resistant vi- rus and multiple distinct zoonotic CoVCoV are taxonomically divided into multiple genogroups (al- pha, beta, gamma, delta) but human-infecting CoV are found in only the alpha and beta subgroups thus far (Fig. 3A). There is high sequence conservation in the RNA-dependent RNA polymerase (RdRp, nsp12) across CoV (Fig. 3A). For example, the RdRp of SARS-CoV-2 has 99.1% similarity and 96% amino acid identity to that of SARS-CoV (Fig. 3A). To gain insight into structural conservation of RdRp across the CoV family, we modeled the variation reflected in the RdRp dendrogram in Fig. 3A onto the structure of the SARS-CoV RdRp (20) (Fig. 3B). The core of the RdRp molecule and main structural mo- tifs that all RdRp harbor (Fig. 3B and fig. S2) are highly con- served among CoV including SARS-CoV-2. We previously reported that CoV resistance to another broad spectrum nu- cleoside analog, RDV, was mediated by RdRp residues F480L and V557L in a model coronavirus mouse hepatitis virus (MHV) and in SARS-CoV, resulting in a 5-fold shift in IC50 (Fig. 3C)(21).

Consequently, we tested whether RDV re- sistance mutations in MHV conferred cross-resistance to NHC. In fact, the two RDV resistance mutations, alone or to- gether, conferred increased sensitivity to inhibition by NHC (Fig. 3D). As our previous studies have demonstrated a high genetic barrier to NHC for VEEV, influenza and CoV (14-16), the lack of cross-resistance further suggests that NHC and RDV may select for exclusive and mutually sensitizing re- sistance pathways.To explore the breadth of antiviral efficacy against zoono- tic CoV, we performed antiviral assays in HAE with threezoonotic Bat-CoV: SHC014, HKU3, and HKU5. Closely related to the beta 2b SARS-CoV, Bat-CoV SHC014 is capable of rep- licating in human cells without adaptation(11), suggesting its potential for zoonotic emergence into humans. The more dis- tantly related SARS-like beta 2b CoV, recombinant Bat-CoV HKU3, has a modified receptor binding domain to facilitate growth in cell culture (22). Lastly, Bat-CoV HKU5 is a MERS- like beta 2c CoV (23). NHC diminished infectious virus pro- duction and the levels of genomic/subgenomic viral RNA in HAE in a dose-dependent manner for all three Bat-CoVs (Fig. 4). Therefore, the antiviral activity of NHC was not limited by natural amino acid variation in the RdRp, which among the group 2b and group 2c CoV can vary by almost 20%. Moreo- ver, these data suggest that if another SARS- or MERS-like virus were to spillover into humans in the future, they would likely be susceptible to the antiviral activity of NHC.NHC antiviral activity is associated with increased viral mutation ratesIt has recently been shown that NHC treatment increases the mutation rate in viral genomic RNA of RSV (24), VEEV (14), influenza (24), and our previous study used RNA seq to show that overall transition mutation frequency is increased dur- ing NHC treatment of MHV and MERS-CoV during infection in continuous cell lines(16).

We sought to determine if NHC would increase the mutation frequency during MERS-CoV in- fection in human primary HAE. Using MERS-CoV-infected HAE treated with either vehicle or a dose range of NHC or RDV, we show that both drugs reduced virus titers in a dose- dependent manner (Fig. 5A). We then used a highly-sensitive high-fidelity deep sequencing approach (Primer ID NGS), which uses barcoded degenerate primers and Illumina in- dexed libraries to determine accurate mutation rates on viral RNA production (25). Using this approach, we analyzed a 538 bp region of viral genomic RNA in nonstructural protein 15 (nsp15). The error rates (#mutations/10,000 bases) in vehicle- (0.01) or RDV- (0.01) treated cultures were very low. RDV is reported to act via chain termination of nascent viral RNA, and thus the low error rates in RDV-treated cultures are in line with the proposed MOA (26). In contrast, the error rate was significantly increased in NHC-treated MERS-CoV RNA in a dose-dependent manner (Two-way ANOVA with Dun- nett’s multiple comparison test; 10-fold increase at 10 μM, P< 0.0001 at 24 and 48 hpi; 5-fold increase at 1μM, P < 0.0001 at 24 hpi and P = 0.0015 at 48 hpi) (Fig. 5C). The magnitude of the error rate in NHC-treated cultures correlated with vi- rus titer reduction. At 48 hpi the respective error rate and virus titer was 0.015 and 3.96E+06 pfu/mL for vehicle treat- ment, 0.045 and 2.86E+04 pfu/mL with 1 μM NHC; and 0.090 and 1.5E+02 pfu/mL 10 μM NHC. Thus, with 1 μM NHC a 3- fold increase in error rate resulted in a 138-fold decrease in virus titer, while with 10 μM NHC a 6-fold increase in error rate resulted in a 26,000-fold decrease in virus titer.We then examined the mutational spectra induced by NHC, which can be incorporated into viral RNA as a substi- tution for either cytosine (C) or uracil (U).

RNA-mutagenic antivirals may incorporate in both nascent negative and pos- itive sense RNA during genome replication (Fig. 5D). Ade- nine-to-guanine (A-to-G) and uracil-to-cytosine (U-to-C) transitions were enriched in MERS-CoV genomic RNA in an NHC dose-dependent manner (Fig. 5E). Collectively, these data used high-fidelity sequence analysis to demonstrate a specific enrichment for A:G and C:U transitions in MERS- CoV RNA after NHC treatment of primary HAE cell cultures.Therapeutic EIDD-2801 reduces SARS-CoV replication and pathogenesisGiven the promising antiviral activity of NHC in vitro, we next evaluated its in vivo efficacy using EIDD-2801, an orally bioavailable prodrug of NHC (β-D-N4-hydroxycytidine-5′-iso- propyl ester), designed for improved in vivo pharmacokinet- ics and oral bioavailability in humans and non-human primates (15). Importantly, the plasma profiles of NHC and EIDD-2801 are similar in mice following oral delivery (15). We first performed a prophylactic dose escalation study in C57BL/6 mice where we orally administered vehicle (10% PEG, 2.5% Cremophor RH40 in water) or 50, 150, or 500 mg/kg EIDD-2801 2hr prior to intranasal infection with 5E+04 PFU of mouse-adapted SARS-CoV (SARS-MA15), and then vehicle or drug every 12 hours thereafter. Beginning on 3 days post-infection (dpi) and through the end of the study, body weight loss compared to vehicle treatment was signifi- cantly diminished (50 mg/kg) or prevented (150, 500 mg/kg) with EIDD-2801 prophylaxis (Two-way ANOVA with Dun- nett’s multiple comparison test, P < 0.0001) (fig. S3A). Lung hemorrhage was also significantly reduced 5 dpi with 500 mg/kg EIDD-2801 treatment (Kruskal-Wallis Test, P = 0.010, fig. S3B). Interestingly, there was a dose-dependent reduction in SARS-CoV lung titer (median titers: 50 mg/kg = 7E+03 pfu/mL, 150 mg/kg = 2.5E+03 pfu/mL, 500 mg/kg = 50 pfu/mL, vehicle = 6.5E+04 pfu/mL) with significant differ- ences (Kruskal-Wallis with Dunn’s multiple comparisons test) among the vehicle, 150 mg/kg (P = 0.03) and 500 mg/kg (P = 0.006) groups. Thus, prophylactic orally administered EIDD-2801 was robustly antiviral and able to prevent SARS- CoV replication and disease.

Since only the 500 mg/kg group significantly diminished weight loss, hemorrhage and reduced lung titer to near un- detectable levels, we tested this dose under therapeutic treat- ment conditions to determine if EIDD-2801 could improve the outcomes of an ongoing CoV infection. As a control, we initiated oral vehicle or EIDD-2801 2 hours prior to infection with 1E+04 pfu SARS-MA15. For therapeutic conditions, we initiated EIDD-2801 treatment 12, 24, or 48 hours after infec- tion. After initiating treatment, dosing for all groups was per- formed every 12 hours for the duration of the study. Bothprophylactic treatment initiated 2 hours prior to infection and therapeutic treatment initiated 12 hours after infection significantly (Two-way ANOVA with Tukey’s multiple com- parison test) prevented body weight loss following SARS-CoV infection on 2 dpi and thereafter (-2 hours: P = 0.0002 to<0.0001; +12 hours: P = 0.0289 to <0.0001) as compared to vehicle treated animals (Fig. 6A). Treatment initiated 24 hpi also significantly reduced body weight loss (3-5 dpi, P = 0.01 to <0.0001) although not to the same degree as the earlier treatment initiation groups. When initiated 48 hpi, body weight loss was only different from vehicle on 4 dpi (P = 0.037, Fig. 6A). Therapeutic EIDD-2801 significantly (Krus- kal-Wallis with Dunnett’s multiple comparison test) reduced lung hemorrhage when initiated up to 24 hours after infec- tion (-2, +12, and +24 hours P < 0.0001) mirroring the body weight loss phenotypes (Fig. 6B). Interestingly, all EIDD-2801 treated mice had significantly (Kruskal-Wallis with Dunnett’s multiple comparison test) reduced viral loads in the lungs even in the +48 hours group (All P < 0.0001, Fig. 6C), which experienced the least protection from body weight loss and lung hemorrhage.

We also measured pulmonary function via whole body plethysmography (WBP). In Fig. 6D, we show the WBP enhanced pause (PenH) metric, which is a surrogate marker for bronchoconstriction or pulmonary obstruction (27), was significantly (Two-way ANOVA with Dunnett’s mul- tiple comparison test) improved throughout the course of the study if treatment was initiated up to 12 hours after infection (-2 hours: 2d pi to 5 dpi, P < 0.0001 to 0.019, +12 hours: 2 dpito 5 dpi, P < 0.0001 to 0.0192) although the +24 hours group showed sporadic improvement as well (3 dpi P = 0.002) (Fig. 6D). Lastly, we blindly evaluated hematoxylin and eosin- stained lung tissue sections for histological features of ALI using two different and complementary scoring tools (18), which show that treatment initiated up to +12 hours signifi- cantly reduced ALI (Kruskal-Wallis with Dunn’s multiple comparison test) (American Thoracic Society (ATS) Lung In- jury Score: -2 hours P = 0.0004, +12 hours P = 0.0053, Diffuse alveolar damage (DAD) Score: -2 hours P = 0.0015, +12 hours P = 0.0004, Fig. 6E). Altogether, therapeutic EIDD-2801 was potently antiviral against SARS-CoV in vivo but the degree of clinical benefit was dependent on the time of initiation post- infection.EIDD-2801 prophylactic and therapeutic efficacy corre- lates with increased MERS-CoV mutation rateAfter obtaining promising in vivo efficacy data with SARS- CoV, we investigated whether EIDD-2801 would be effective against MERS-CoV. As the murine ortholog of the MERS-CoV receptor, dipeptidyl peptidase 4 (DPP4), does not support vi- ral binding and entry, all in vivo studies were performed in genetically modified mice encoding a murine DPP4 receptor encoding two human residues at positions 288 and 330 (hDPP4 288/330 mice)(18, 28). Similar to our SARS-CoV data,all doses of prophylactic EIDD-2801 (50, 150 and 500 mg/kg) protected hDPP4 288/330 mice (fig. S4) from significant body weight loss (Two-way ANOVA with Dunnett’s multiple com- parison test, P = 0.03 to < 0.0001), lung hemorrhage (Kruskal- Wallis with Dunn’s multiple comparison test, P = 0.01 to<0.0001), and virus replication which was undetectable (Kruskal-Wallis with Dunn’s multiple comparison test, P < 0.0001) regardless of drug dose following intranasal infection with 5E+04 PFU mouse-adapted MERS-CoV (fig. S4).

We then evaluated the therapeutic efficacy EIDD-2801 fol- lowing the promising results of our prophylactic studies. Sim- ilar to our SARS-CoV study, EIDD-2801 treatment administered before or 12 hours after intranasal mouse- adapted MERS-CoV infection (5E+04 PFU) prevented body weight loss from 2 through 6 dpi (Two-way ANOVA with Tukey’s multiple comparison test, Fig. 7A, P = 0.02 to<0.0001) and lung hemorrhage on 6 dpi (Kruskal-Wallis with Dunn’s multiple comparison test, P = 0.0004 to < 0.0001, Fig. 7B), but treatment initiated 24 or 48 hours did not offer sim- ilar protection. Unlike body weight loss and lung hemorrhage data which varied by treatment initiation time, virus lung ti- ter on 6 dpi was significantly reduced to the limit of detection in all treatment groups (Kruskal-Wallis with Dunn’s multiple comparison test, Fig. 7C, P < 0.0001). Interestingly, when vi- ral genomic RNA was quantified in paired samples of lung tissue, EIDD-2801 significantly reduced quantities of viral RNA (One-way ANOVA with Dunnett’s multiple comparison test, P <0.0001 to 0.017) in an initiation time-dependent man- ner for all groups except for +48 hours (Fig. 7D). The discrep- ancy among infectious titers and viral RNA suggests that accumulated mutations render the particles non-infectious and undetectable by plaque assay, consistent with the MOA. To gauge the effect of EIDD-2801 treatment on lung function, we assessed pulmonary function by WBP. Mirroring the body weight loss data, normal pulmonary function was only ob- served in groups where treatment was initiated prior to or 12 hours after infection (Two-way ANOVA with Tukey’s multiple comparison test, -2hr: P < 0.0001 3 dpi, P = 0.0002 4 dpi,+12hr: P < 0.0001 3 dpi, P = 0.0008 4 dpi, Fig. 7E). Collec- tively, these data demonstrate that NHC prodrug, EIDD-2801, robustly reduces MERS-CoV infectious titers, viral RNA, and pathogenesis under both prophylactic and early therapeutic conditions.

To study the molecular mechanisms associated with drug performance in vivo, we investigated the correlation between infectious virus production and EIDD-2801-mediated muta- genesis of MERS-CoV RNA under therapeutic treatment con- ditions. Using Primer ID NGS, we measured the mutation rates of both viral genomic RNA (non-structural protein 10, nsp10) and host interferon stimulated gene 15 (ISG15) mRNA, a highly up-regulated innate immune-related gene after MERS-CoV infection (Fig. 7F). Primer ID NGS measures themutational frequency in single RNA molecules, each of which are represented by a single template consensus sequence (TCS) (25). Viral TCS were significantly reduced (Two-way ANOVA with Tukey’s multiple comparison test, -2 hours P<0.0001, +12 hours P = 0.0001, +24 hours P = 0.02) in a treat- ment initiation time-dependent manner (Fig. 7G) similar to viral genomic RNA measured by qRT-PCR. In contrast, the numbers of ISG15 TCS were similar (P = 0.2 to 0.8) for all groups indicating that neither vehicle nor drug treatment sig- nificantly affected the levels of or mutated ISG15 mRNA tran- scripts (Fig. 7G). Similar to our TCS data in Fig. 6G, the total error rate in viral nsp10 was significantly increased (Two-way ANOVA with Tukey’s multiple comparison test) in groups where treatment was initiated prior to (-2 hours, median er- ror rate = 10.5 errors/10,000 bases, P < 0.0001) and up to 24 hours post infection (12 hours, median error rate = 8.2 er- rors/10,000 bases, P < 0.0001 ; +24 hours, median error rate= 5.4 errors/10,000 bases, P = 0.0003) but the error rates in ISG15 remained at baseline for all groups (Fig. 7H). In addi- tion, nucleotide transitions observed in MERS-CoV genomes in vitro, were also observed in vivo in groups where treatment was initiated prior to and up to 12 hours post infection (Two- way ANOVA with Tukey’s multiple comparison test, P = 0.0003 to < 0.0001) (Fig. 7I).

Importantly, these transitions were not observed in host ISG15 mRNA (Fig. 7I). Lastly, the EIDD-2801 dose-dependent mutagenesis of viral RNA corre- lated with an increase in codon change frequency, including stop codons, in mice where treatment was initiated 12 hours or before (Two-way ANOVA with Tukey’s multiple compari- son test, vehicle median = 3.4; -2hr median = 22.8, P = 0.0035;+12 hours median = 20.0, P = 0.0004, Fig. 7J). Thus, approxi- mately 20% of the mutations observed in the -2 hours and+12 hours groups resulted in a codon change and alteration of the nsp10 protein sequence. When extrapolating our re- sults from nsp10 to the entirety of the 30kb MERS-CoV ge- nome, EIDD-2801 likely causes between 15 (+24 hours treatment) and 30 (-2 hours treatment) mutations per ge- nome, 10-20% of which result in amino acid coding changes. Altogether, our data demonstrates that EIDD-2801-driven mutagenesis correlates well with the reductions in viral load, strongly suggestive of an error catastrophe-driven mecha- nism of action under therapeutic conditions.

DISCUSSION
In the past 20 years, three novel human coronaviruses have emerged (29, 30). The group 2b SARS-like CoV represent an existential and future threat to global health as evidenced by the emergence of SARS-CoV and SARS-CoV-2. Zoonotic SARS-like bat CoV strains can use human angiotensin-con- verting enzyme 2 (ACE2) receptors, grow well in primary hu- man airway cells, and vary by as much as 25% in key therapeutic and vaccine gene targets (11, 31). Thus, to address the current public health emergency of COVID-19 and to max- imize pandemic preparedness in the future, broad-based vac- cines and therapeutics, which are active against the higher risk RNA virus families prone to emergence, are desperately needed. Small molecule antivirals can exert their antiviral effect through multiple mechanisms including blocking viral entry, inhibiting a virally encoded enzyme, blocking virus particle formation, or targeting a host factor required for replication (32). Multiple direct acting antivirals are currently under evaluation in randomized control trials to treat COVID-19 in- cluding hydroxychloroquine, remdesivir, lopinavir/ri- tonavir(33-35). Here, we report the broad-spectrum antiviral activity of NHC and its orally bioavailable prodrug EIDD- 2801, against SARS-CoV, MERS-CoV, and the current pan- demic strain SARS-CoV-2 in primary human airway epithelial cells. In addition to CoV, NHC is broadly active against mul- tiple genetically distinct viruses including VEE, influenza A and B, Ebola, and Chikungunya viruses (13-16, 19, 21, 24, 36- 38). Here, we show that prophylactic and therapeutic EIDD- 2801 significantly reduced lung viral loads and improved pul- monary function in mouse models of both SARS- and MERS- CoV pathogenesis. Although the improvement in both SARS- and MERS-CoV outcomes diminished with the delay of treat- ment initiation time, it is important to note that the kinetics of disease in mice are compressed as compared to that in hu- mans. Whereas SARS- and MERS-CoV lung titers peak on 1-2 dpi in mice concurrent with the onset of clinical signs and notable damage to the lung epithelium, in humans this oc- curs 7-10 days after the onset of symptoms (19, 28, 39, 40).

Thus, in mice, the window within which to treat emerging CoV infection prior to peak replication is compressed (e.g., 24-48 hours). As with oseltamivir treatment for influenza which fails to provide a protective effect if administered >5 days after the onset of symptoms, the window in which to treat COVID-19 patients prior to peak virus replication is likely during the first week of symptoms when pharyngeal shedding is at its highest(41, 42). However, virus replication and shedding may continue for several weeks in the most se- vere COVID-19 patients(34). Thus, early intervention with an antiviral like EIDD-2801 is likely to provide the most clinical benefit although there may opportunities in severe patients where the duration of virus replication may be extended. Our current study is clearly limited by the lack of in vivo efficacy testing with SARS-CoV-2. Currently, robust mouse models that recapitulate the SARS-CoV-2 pathogenesis observed in humans do not yet exist due to a noted virus spike glycopro- tein and mouse ACE2 receptor incompatibility complicating the evaluation of medical countermeasures(43, 44). In addi- tion, SARS-CoV and MERS-CoV, SARS-CoV-2 disease severity increases with increasing age. Our studies are limited by the lack of drug efficacy testing in CoV aged mouse models that recapitulate the age-related increase in pathogenesis ob- served in humans(45). The data provided in this manuscript suggest that EIDD-2801 should be quickly evaluated in pri- mate models of human disease, using immediate models for MERS-CoV and SARS-CoV pathogenesis or newly described cynomolgus and rhesus macaque models for SARS-CoV-2 (46- 49).

For VEE and influenza, NHC/EIDD-2801 exerts its antivi- ral activity on the RNA-dependent RNA polymerase leading to error catastrophe by inducing an error rate of replication that surpasses the error threshold allowed to sustain a virus population (14, 15). This process occurs when NHC is incor- porated during RNA synthesis then subsequently misread thus increasing mutation rates. Therefore, for CoV, the NHC MOA would appear less likely to be affected by the RNA proofreading activity encoded by the nsp14 exonuclease func- tion that otherwise limits misincorperation (50). Here, we present data using Primer ID NGS to quantitate the fre- quency and identity of the mutational spectra in the MERS- CoV genome in both drug-treated primary human airway cells and in mice at single genome resolution. As CoV are pos- itive sense RNA viruses that replicate through a negative sense RNA intermediate, NHC incorporation as a C or a U can occur in both polarities of RNA. We found increased nu- cleotide transitions (A to G, G to A, C to U, U to C) consistent with those reported after influenza and VEE infections (14, 15). Under identical conditions, RDV did not alter the muta- tion rate in MERS-CoV genomic RNA, supporting its reported mechanism of action as a chain terminator of viral RNA syn- thesis (26). In primary human lung cell cultures and mice in- fected with MERS-CoV, the NHC mutation rates inversely correlated with a reduction in infectious virus. In addition, we found a positive correlation between increased mutation rates and the frequency of nonsynonymous mutations and the degree of therapeutic efficacy in mice. To explore the po- tential off-target effect in host mRNA which may contribute to drug toxicity, we also examined the impact of NHC treat- ment on ISG15 transcripts, a gene highly induced following MERS-CoV infection.

Although ISG15 transcripts are present in great abundance, an accumulation of mutations was not observed in ISG15 in this model even at 500 mg/kg dosing. These data also support previous studies using RNAseq to demonstrate that the model coronavirus MHV displayed in- creased mutation frequencies following NHC treatment in vitro (16). With regard to nucleic acid specificity, ribonucleo- tides are efficiently removed from eukaryotic cell DNA; there- fore, treating a viral infection with a mutagenic ribonucleoside analog should show a selectivity for incorpo- ration into the viral genome and not be efficient at being in- corporated into and inducing mutations into host cell DNA (51). All together, these data strongly support the notion that EIDD-2801 and its active nucleoside analog NHC exert their antiviral effect through the induction of error catastrophe in the targeted virus. While our data suggest that the MERS-CoV nsp14 proofreading activity appeared ineffective against NHC in vitro and EIDD-2801 in vivo, future studies should investi- gate the antiviral activity of NHC in the presence or absence of the nsp14 proofreading activity, as loss of this activity in- creased the sensitivity of MHV and SARS-CoV replication to RDV treatment (50). Together, our data support the continued development of EIDD-2801 as a potent broad spectrum antiviral that could be useful in treating contemporary, newly emerged and emerging coronavirus infections of the future.

The primary goal of this study was to determine the antiviral activity of the nucleoside analog NHC (EIDD-1931) against multiple emerging CoV in vitro and antiviral efficacy of its prodrug, EIDD-2801, in mouse models of CoV pathogenesis. Coupling cell lines and primary HAE cell cultures, we evalu- ated the antiviral activity of NHC against the three most re- cently emerged human CoV: SARS-CoV, MERS-CoV, and SARS-CoV-2. For both SARS-CoV and MERS-CoV, the data presented for HAE studies are representative of those from 2- 3 separate human donors. For SARS-CoV-2, the HAE were from a single human donor. We evaluated drug cytotoxicity in both Calu-3 2B4 and HAE cell cultures. Calu-3 and the SARS-CoV and MERS-CoV HAE studies were performed in biological triplicate. HAE studies with SARS-CoV-2 and the SARS- and MERS-like bat CoV were performed with two wells per condition. Drug effects were measured relative to vehicle controls in vitro and comparisons in vivo were performed to vehicle controls. We also aimed to determine the antiviral ef- ficacy of EIDD-2801 in mouse models of CoV pathogenesis. These studies were intended to provide the preclinical data to justify nonhuman primate studies and human clinical tri- als. Mice were age- and sex-matched and randomly assigned into groups before infection and treatment. Pathology was scored blinded by a board-certified veterinary pathologist. Primary data for all studies are provided in data file S1. Efficacy studies were performed in animal biosafety level 3 facilities at UNC Chapel Hill. All work was conducted under protocols approved by the Institutional Animal Care and Use Committee at UNC Chapel Hill (IACUC protocol #16-284) ac- cording to guidelines set by the Association for the Assess- ment and Accreditation of Laboratory Animal Care and the U.S. Department of Agriculture.

The parental compound β−D−N4-hydroxycytidine (NHC, all in vitro studies) and its prodrug EIDD-2801 (all in vivo stud- ies) was supplied by Emory University Institute for Drug Discovery (EIDD). NHC was supplied as a 10 mM stock in DMSO and EIDD-2801 as a solid and solubilized in vehicle containing 10% PEG400, 2.5% Cremophor RH40 in water (10/2.5/87.5%, all v/v) prior to use. RDV was solubilized in 100% DMSO and provided by Gilead Sciences, Inc as previ- ously described (18, 19). At UNC, the human lung epithelial cell line Calu-3 2B4 cells was maintained in Dulbecco’s modified Eagle’s medium (DMEM, Gibco), 20% fetal bovine serum (Hyclone) and 1x an- tibiotic/antimycotic (Gibco). At Vanderbilt University Medi- cal Center (VUMC), Calu-3 2B4 were propagated in DMEM supplemented with 20% FBS (Gibco), 100 U/ml penicillin and streptomycin (Gibco), and 0.25 μM amphotericin B (Corn- ing). At VUMC, VeroE6 cells were cultured in DMEM supple- mented with 10% FBS (Gibco), 100 U/ml penicillin and streptomycin (Gibco), and 0.25 μM amphotericin B (Corn- ing). At UNC, VeroE6 cells were cultured in DMEM supple- mented with 10% Fetal Clone II (Hyclone) and 1x antibiotic/antimycotic (Gibco). Murine delayed brain tumor (DBT) cells were maintained in DMEM supplemented with 10% FBS (Gibco), 100 U/ml penicillin and streptomycin (Gibco), and 0.25 μM amphotericin B (Corning). Primary hu- man airway epithelial (HAE) cell cultures were obtained from the Tissue Procurement and Cell Culture Core Laboratory in the Marsico Lung Institute/Cystic Fibrosis Research Center at UNC and are described more thoroughly below (52).

Except for SARS-CoV-2, all viruses used for these studies were derived from infectious clones and isolated as previously de- scribed (53). SARS-CoV-2 clinical isolate was obtained at VUMC and UNC from the CDC (2019-nCoV/USA-WA1/2020 strain, GenBank accession no. MN985325.1) and passaged twice in Vero E6 cells at each respective institution to create a passage 5 working stock (54). Virus strains for in vitro ex- periments include SARS-CoV expressing the green fluores- cent protein (GFP) in place of open reading frames 7a/b (ORF7a/b, SARS-GFP)(53), bat-spike receptor binding do- main (Bat-SRBD)(22), a chimeric CoV strain derived from the HKU3 SARS-like bat coronavirus genomic sequence that has the wild type (Urbani SARS-CoV strain) RBD in the HKU3 spike gene to allow for virus replication in non-human pri- mate cell lines and HAE cultures, SHC014 SARS-like bat coro- navirus (11), MERS-CoV expressing nanoluciferase in the place of ORF3 (MERS-nLUC)(19), and MERS-CoV expressing the red fluorescent protein gene in the place of ORF 5 (RFP, MERS-RFP)(55). The virus stock utilized for MERS-CoV in vivo studies was derived from a plaque-purified isolate of the mouse-adapted MERS-CoV p35C4 strain (56). The virus stock utilized for SARS-CoV in vivo studies was derived from the infectious clone of the mouse-adapted SARS-CoV MA15 (MA15) strain (57). All work with MHV was performed using the recombinant WT strain MHV-A59 (GenBank accession no. AY910861)(58). MERS-CoV nLUC in Calu-3: At 48 hours prior to infection, Calu-3 2B4 cells were plated in a 96-well black-walled clear bottom plate at 5×104 cells/well. A 10 mM stock of NHC was serially diluted in 100% DMSO in 3-fold increments to obtain a ten-point dilution series. MERS-nLUC was diluted in DMEM supplemented with 10% FBS, and 1% Antibiotic-Anti- mycotic to achieve a multiplicity of infection (MOI) of 0.08.

Cells were infected and concurrently treated with NHC in triplicate per drug dilution for 1hr, after which viral inoculum was aspirated, cultures were rinsed once and fresh medium containing drug or vehicle was added. At 48 hours post infec- tion, nanoluciferase expression as a surrogate for virus repli- cation was quantitated on a Spectramax plate reader (Molecular Devices) according to the manufacturer’s instruc- tions (Promega, NanoGlo). For the 100% inhibition control, diluted MERS-nLUC was exposed to short-wave UV light (UVP, LLC) for 6 min to inhibit the ability of the virus to rep- licate. For the 0% inhibition control, cells were infected in the presence of vehicle only. DMSO was kept constant in all con- ditions at 0.05%. Values from triplicate wells per condition were averaged and compared to controls to generate a per- cent inhibition value for each drug dilution. The IC50 value was defined as the concentration at which there was a 50% decrease in luciferase expression. Data were analyzed using GraphPad Prism 8.0. The IC50 values were calculated by non- linear regression analysis using the dose-response (variable slope) equation (four parameter logistic equation): Y = Bot- tom + (Top-Bottom)/(1+10^((LogIC50-X)*HillSlope)). To measure cell viability to determine if there was any NHC- induced cytotoxicity, Calu-3 2B4 cells were plated and treated with NHC only as described above. Cells were exposed to the same ten-point dilution series created for the in vitro efficacy studies. As above, 0.05% DMSO-treated cells served as the 0% cytotoxicity control. Wells without cells served as the 100% cytotoxic positive control. After 48 hours, cell viability was measured on a Spectramax (Molecular Devices) via Cell-Titer Glo Assay (Promega) according to the manufacturer’s proto- col. Similar data were obtained in three independent experi- ments.SARS-CoV-2 in Calu-3:Calu-3 2B4 cells were adsorbed with MOI 0.1 PFU/cell of SARS-CoV-2 (2019-nCoV/USA- WA1/2020 strain) at 37°C.

Plates were manually rocked every 10 min to redistribute the inoculum. After 30 min, virus inoc- ulum was removed, cells were washed with Phosphate buff- ered saline (PBS) once to remove unbound virus, medium containing NHC or vehicle control (DMSO) was added back onto the cells, and cells were incubated for 72 hours at 37°C. SARS-CoV-2 in Vero E6: Vero E6 cells were plated at 20,000 cells/well in a 96-well plate. 24hr later, mediumcontaining a dose response of NHC was added concurrent with SARS-CoV-2 (2019-nCoV/USA-WA1/2020 strain) at an MOI of 0.05. 48 hours post infection, cell viability was meas- ured by CellTiter Glo assay.SARS-CoV, MERS-CoV, and SARS-CoV-2 in HAE: Humantracheobronchial epithelial cells provided by Dr. Scott Ran- dell were obtained from airway specimens resected from pa- tients undergoing surgery under University of North Carolina Institutional Review Board-approved protocols (#03-1396) by the Cystic Fibrosis Center Tissue Culture Core. Primary cells were expanded to generate passage 1 cells and passage 2 cells were plated at a density of 250,000 cells per well on Transwell-COL (12mm diameter) supports (Corning). Human airway epithelium cultures (HAE) were generated by provi- sion of an air-liquid interface for 6 to 8 weeks to form well- differentiated, polarized cultures that resembled in vivo pseu- dostratified mucociliary epithelium (59).· At 48 hours prior to infection the apical surface of the culture was washed with 500 μL PBS for 1.5 hours at 37°C and the cultures moved into fresh air liquid interface (ALI) media. Immediately prior to infection, apical surfaces were washed twice to remove accu- mulated mucus with 500 μL of PBS with each wash lasting 30 min at 37°C and HAE cultures were moved into ALI media containing various concentrations of NHC ranging from 10 μM to 0.0016 μM as indicated for each experiment (final % DMSO < 0.05%).

Upon removing the second PBS wash, 200 μL of viral inoculum (SARS-GFP, MERS-RFP or 2019- nCoV/USA-WA1/2020 strain) at an MOI of 0.5 was added to the apical surface and HAE cultures were incubated for 2 hours at 37°C. Viral inoculum was then removed, and the ap- ical surface of the cultures were washed three times with 500μL PBS and then incubated at 37°C until 48 hours post infection (hpi). For all HAE cultures, infectious virus pro- duced was collected by washing the apical surface of the cul- ture with 100 μL PBS. Apical washes were stored at -80°C until analysis and titered by plaque assay as previously de- scribed (19).qRT-PCR approach to assess cytotoxicityTotal RNA was isolated using the Zymo Direct-zol RNA Min- iPrep Kit (Zymo Research Corp.) according to the manufac- turer’s directions. Cells were treated with 1μM staurosporine (Sigma-Aldrich) as a positive control. First-strand cDNA was generated using Superscript III reverse transcriptase (Life Technologies). For quantification of cellular markers of tox- icity/apoptosis, real-time PCR was performed using commer- cially validated TaqMan-based primer-probe sets (table S1) and TaqMan Universal PCR Mix (Life Technologies). Results were then normalized as described above.MERS-CoV genomic RNA qRT-PCRMouse lungs were stored in RNAlater (ThermoFisher) at - 80°C until processed via homogenization in TRIzol (Invitro- gen). Total RNA was isolated using Direct-zol RNA MiniPrepkit (Zymo Research). Previously published TaqMan primers were synthesized by Integrated DNA Technologies (IDT) to quantify MERS genomic RNA (targeting orf1a: Forward: 5′- GCACATCTGTGGTTCTCCTCTCT-3′, Probe (6- FAM/ZEN/IBFQ): 5′- TGCTCCAACAGTTACAC-3′, Reverse: 5′-AAGCCCAGGCCCTACTATTAGC)(60). qRT-PCR was per-formed using 100ng total RNA compared to an RNA standard curve using TaqMan Fast Virus 1-Step Master Mix (Ther- moFisher) on a Quant Studio 3 (Applied Biosystems).Cell supernatants were harvested in TRIzol LS reagent (Invi- trogen), and RNA was purified following phase separation by chloroform as recommended by the manufacturer. The RNA in the aqueous phase was collected and further purified using PureLink RNA Mini Kits (Invitrogen) according to the man- ufacturer’s protocol.

Viral RNA was quantified using one-step quantitative reverse transcription PCR (qRT-PCR) on a StepOnePlus Real-Time PCR system (Applied Biosystems) by TaqMan Fast Virus 1-Step Master Mix chemistry (Applied Bi- osystems). SARS-CoV-2 N gene RNA was amplified using for- ward (5′-GACCCCAAAATCAGCGAAAT) and reverse (5′- TCTGGTTACTGCCAGTTGAATCTG) primers and probe (5′- FAM-ACCCCGCATTACGTTTGGTGGACC-BHQ1) designed bythe United States Centers for Disease Control and Prevention (oligonucleotides produced by IDT, cat# 10006606). Copy numbers were interpolated from a standard curve produced with dilutions of N gene RNA. Briefly, SARS-CoV-2-N positive control plasmid DNA (IDT, cat# 10006625) was amplified us- ing forward (5′- TAATACGACTCACTATAGGGATGTCTGATAATGGACCCCA)and reverse (5′- TTAGGCCTGAGTTGAGTCAG) primers, re- sulting in a 1280 nucleotide fragment containing a T7 pro- moter. The PCR product was purified by column (Promega) and in vitro transcribed using the mMESSAGE mMACHINE T7 Transcription Kit (Invitrogen) according to the manufac- turer’s protocol. Transcribed RNA was purified using RNeasy mini kit (Qiagen) according to the manufacturer’s protocol, and serial 10-fold dilutions were quantified as described above.Primer ID NGS is designed to specifically identify and remove RT-PCR mutations, while facilitating highly accurate se- quence determination of single RNA molecules, because each cDNA is created with a barcoded degenerate primer (N10, 410 combinations) from which Illumina indexed libraries are made. We used a multiplexed Primer ID library prep ap- proach and MiSeq sequencing to investigate the presence of mutations in the viral genomes and murine mRNA. We de- signed cDNA primers targeting multiple regions on the viral genome and murine mRNA, each with a block of randomnucleotides (11 bp long) as the Primer ID (25, 61) (table S2). Viral RNA was extracted using QIAamp viral RNA kit.

A pre- amplification titration of templates was performed to esti- mate the amount of template to use. We used SuperScript III to make cDNA with multiplexed cDNA primers based on the regions to be sequenced. We used 41R_PID11 for the pilot se- quencing and titration determination. For the MERS-CoV se- quencing, we multiplexed nsp10_PID11, nsp12_PID11 and nsp14_PID11 for the cDNA reaction; for the murine mRNA sequencing, we used mixed primers of nsp10_PID11, ifit3_PID11, isg15_PID11. After bead purification, we ampli- fied the cDNA with a mixture of forward primers (based on the described schemes) and a universal reverse primer, fol- lowed by another round of PCR to incorporate Illumina se- quencing adaptors and barcodes in the amplicons. After gel- purification and quantification, we pooled 24 libraries for MiSeq 300 base paired-end sequencing. The TCS pipeline ver- sion 1.38 (https://github.com/SwanstromLab/PID) was used to process the Primer ID sequencing data and construct tem- plate consensus sequences (TCSs) to represent each individ- ual input templates, and the sequences of each region in the pool was de-multiplexed. The RUBY package viral_seq ver- sion 1.0.6 (https://rubygems.org/gems/viral_seq) was used to calculate the mutation rate at each position. NCBI SRA Ac- cession numbers for sequence data are as follows: PRJNA613261 (Fig. 5) and PRJNA613454 (Fig. 7).We performed 4 mouse studies to evaluate the in vivo efficacy of the NHC prodrug (EIDD-2801). First, we performed prophylactic dose escalation studies for both SARS- and MERS-CoV to determine the most efficacious dose of EIDD- 2801 per virus. For SARS-CoV, in cohorts of equivalent num- bers of male and female 20-29 week old SPF C57BL/6J (Stock 000664 Jackson Labs) mice (n = 10/dose group), we admin- istered vehicle (10% PEG, 2.5% Cremophor RH40 in water) or 50, 150 or 500 mg/kg EIDD-2801 by oral gavage 2 hours prior to intranasal infection with 1E+04 PFU mouse-adapted SARS-CoV strain MA15 in 50μl.

Mice were anaesthetized with a mixture of ketamine/xylazine prior to intranasal infection. Vehicle or drug was administered every 12hr for the remain- der of the study. Body weight and pulmonary function by whole body plethysmography were measured daily. On 5 dpi, animals were sacrificed by isoflurane overdose, lungs were scored for lung hemorrhage, and the inferior right lobe was frozen at −80°C for viral titration via plaque assay. Briefly, 500,000 Vero E6 cells/well were seeded in 6-well plates. The following day, medium was removed and serial dilutions of clarified lung homogenate were added per plate (10−1 to 10−6 dilutions) and incubated at 37°C for 1hr after which wells were overlaid with 1X DMEM, 5% Fetal Clone 2 serum, 1X antibiotic/antimycotic, 0.8% agarose. Two days after, plaques were enumerated to generate a plaque/ml value. Lunghemorrhage is a gross pathological phenotype readily ob- served by the naked eye driven by the degree of virus replica- tion where the coloration of the lung changes from pink to dark red (62, 63). The large left lobe was placed in 10% buff- ered formalin and stored at 4°C for 1-3 weeks until histologi- cal sectioning and analysis. For MERS-CoV, the prophylactic dose escalation studies we performed similarly as done for SARS-CoV with our recently developed a mouse model for MERS-CoV, which has a humanized DPP4 receptor (hDPP4)(28). We performed all in vivo studies with EIDD-2801 in equivalent numbers of 10-14 week old female and male C57BL/6J hDPP4 mice. Second, we intranasally infected mice with 5E+04 PFU mouse-adapted MERS-CoV strain M35C4 in 50μl. Third, to titer lungs by plaque assay, Vero CCL81 cells were used and plaques were enumerated 3 dpi.To determine the time at which therapeutic administra- tion of EIDD-2801 would fail to improve outcomes with SARS-CoV or MERS-CoV infection, we performed therapeutic efficacy studies in mice where we initiated treatment 2 hours prior to infection or 12, 24 or 48 hours after infection. As 500 mg/kg provided the most complete protection from disease in prophylactic SARS-CoV studies, this dose was used for both therapeutic efficacy studies.

Vehicle or EIDD-2801 was given via oral gavage twice daily following initiation of treatment. For both SARS-CoV and MERS-CoV, the infectious dose for the therapeutic studies and the mouse strains were the same as that used in the prophylactic studies. The numbers of mice per group for the SARS-CoV studies were as follows: Vehicle (n = 10), -2 hours (n = 10), +12 hours (n = 10), +24 hours (n = 10), +48 hours (n = 10). The numbers of mice per group for the MERS-CoV therapeutic studies were as follows: Vehicle (n = 9), -2 hours (n = 9), +12 hours (n = 9), +24 hours (n = 7),+48 hours (n = 10). As described above, each day mouse body weight and pulmonary function were quantitated. On 5 dpi for SARS-CoV and 6 dpi for MERS-CoV, animals were hu- manely sacrificed and tissues were harvested and analyzed as described above. In addition, for the MERS-CoV study, lung tissue was harvested and stored in RNAlater (Thermo Fisher) at -80°C and then thawed, homogenized in Trizol reagent (Invitrogen) and total RNA was isolated using a Direct-zol RNA MiniPrep kit (Zymo Research). This total RNA was then used for Primer ID and qRT-PCR.Whole body plethysmographyPulmonary function was monitored once daily via whole- body plethysmography (Buxco Respiratory Solutions, DSI Inc.). Mice intended for this analysis were randomly chosen prior to the initiation of the study. Briefly, after a 30-min ac- climation time in the plethysmograph, data for 11 parameters was recorded every 2 s for 5 min.Two different and complementary quantitative histologic tools were used to determine if antiviral treatmentsdiminished the histopathologic features associated with lung injury. Both analyses and scoring were performed by a EIDD-1931 Board Certified Veterinary Pathologist who was blinded to the treat- ment groups.