4′-Fluorouridine is an oral antiviral that blocks respiratory syncytial virus and SARS-CoV-2 replication

Description

The COVID-19 experience has highlighted the need for orallybioavailable broad-spectrum antivirals that could be quickly deployed against newly emerging viral pathogens. Remdesivir-a direct-acting broad-spectrum antiviral-is still the only small molecule therapeutic approved for use against SARS-CoV-2 infection in the United States, but it requires intravenous administration. The ensuing restriction to hospitalized patients compromises its clinical effect as treatment is initiated too late in the infection cycle (1). We have demonstrated efficacy of orally available EIDD-2801 (molnupiravir) against influenza viruses in human organoid models and ferrets (2), and subsequent animal and human data showed that antiviral efficacy of molnupiravir extends to SARS-CoV-2 in vivo (3,4). Molnupiravir acts by inducing lethal viral mutagenesis after incorporation into viral genomic RNA of influenza viruses (2) and betacoronaviruses (5). The drug was recently approved in the United Kingdom and is currently considered for emergency use authorization against COVID-19 in the United States. However, even with this accelerated development timeline, molnupiravir only became available to patients nearly two years into the pandemic. To have a substantial effect on a mounting pandemic, an antiviral must be approved for human use before a new pathogen emerges, making the case for the development of broad-spectrum antivirals.
We have identified RSV disease as a viable primary indication for a candidate broad-spectrum antiviral, on the basis of the unmet major health threat imposed by RSV and well-established protocols for clinical trials of anti-RSV therapeutics. RSV infections are responsible for over 58,000 hospitalizations of children below 5 years of age in the United States annually, and approximately 177,000 hospitalizations of adults above the age of 65 (6)(7)(8)(9). Despite this major health and economic burden, no therapeutics have been licensed specifically for treatment of RSV disease (10). Anti-RSV drug discovery efforts have increasingly focused on inhibiting the viral RNA-dependent RNA polymerase (RdRP) complex (11). The core polymerase machinery comprises the large (L) polymerase protein, its obligatory cofactor, the phosphoprotein (P), and the encapsidated negative-sense RNA genome (11). Allosteric inhibitors of RSV L have potent activity as seen, for instance, with the experimental drug candidates AVG-233 (12) and inhaled PC786 (13).
In search of a drug that is active against RSV and SARS-CoV-2, is orally available, and acts through a distinct mechanism of activity (MOA) from molnupiravir, we explored 4′-fluorine substitutions in a series of analogs of the molnupiravir parent molecule N 4 -hydroxycytidine (NHC) (14). The focus on 4′-fluorine ribose substitutions was motivated by the small atomic radius and strong stereo-electronic effect of fluorine that can influence backbone conformation flexibility, which may lead to improved selectivity indices, increased lipophilicity, and greater metabolic stability (15). A synthetic intermediate in the approach to 4′fluoro-N 4 -hydroxycytidine (compound 5 in fig. S1) was deprotected to provide 4′-FIU (Fig. 1A), which emerged as broadly active antiviral when biotested. 4′-Fluorouridine is an oral antiviral that blocks respiratory syncytial virus and SARS-CoV-2 replication (Page numbers not final at time of first release) 2

4′-FlU is a broad-spectrum mononegavirus inhibitor with high SI
Following the approach of using RSV disease as a primary indication to advance a novel candidate broad-spectrum antiviral, we first assessed activity of 4′-FlU against a recombinant RSV A2-line19F (recRSV A2-L19F) (16) and clinical RSV isolates on immortalized HEp-2 cells. The compound showed potent dose-dependent activity against all RSV strains tested, returning half-maximal effective concentrations (EC50 values) ranging from 0.61 to 1.2 μM ( Fig. 1B and table S1). This cell culture potency was on par with the previously reported anti-RSV activity of NHC ( fig. S2). Global metabolic activity of established human and animal cell lines (HEp-2, MDCK, BHK-T7, and BEAS-2B) exposed to up to 500 μM of 4′-FlU remained unaltered, indicating that the antiviral effect is as a result of cytotoxicity ( Fig. 1C and table S2). When glucose was replaced with galactose as a carbohydrate source to link cell metabolic activity strictly to mitochondrial oxidation (17), we determined a half-maximal cytotoxic concentration (CC50) of 4′-FlU of 250 μM (Fig. 1C and table S2).
When tested on disease-relevant primary human airway epithelial cells (HAE) derived from two different donors (Fig. 1D), 4′-FlU showed ≥17-fold increased anti-RSV potency relative to that on HEp-2 cells, but unchanged low cytotoxicity (CC50 169 μM) ( We next explored the 4′-FlU indication spectrum. We assessed a panel of negative-sense RNA viruses of the paramyxovirus and rhabdovirus families, including measles virus (MeV), human parainfluenza virus type 3 (HPIV3), Sendai virus (SeV), vesicular stomatitis virus (VSV), and rabies virus (RabV) that, like RSV, belong to the mononegavirus order, and found that the compound demonstrated submicromolar active concentrations ( Fig. 1F and table S1). Testing a representative of phylogenetically distant positive-sense RNA viruses, the betacoronavirus SARS-CoV-2 was also sensitive to 4′-FlU, with EC50 values ranging from 0.2 to 0.6 μM against isolates of different lineages ( Fig. 1G and table S1).
At initial mechanistic characterization, 4′-FlU inhibited RSV and paramyxovirus RdRP complex activity in cell-based minireplicon systems ( Fig. 1H and table S1). The RdRP activity of Nipah virus (NiV)-a highly pathogenic zoonotic paramyxovirus with pandemic potential (18)-was also efficiently inhibited by 4′-FlU in an NiV minireplicon reporter assay. The antiviral effect of 4′-FlU was dose-dependently reversed by addition of an excess of exogenous pyrimidines (cytidine and uridine) but not purines to the cultured cells, which is consistent with competitive inhibition of RdRP activity (2,19) (Fig. 1I).

Incorporation of 4′-FlU by RSV and SARS-CoV-2 RdRP causes sequence-modulated transcriptional stalling
To characterize the molecular MOA of 4′-FlU, we purified recombinant RSV L and P proteins expressed in insect cells ( Fig. 2A) and determined performance of the bioactive 5′-triphosphate form of 4′-FlU (4′-FlU-TP) within in vitro primer extension assays (20) (Fig.  2B). In the presence of radio-labeled ATP and an increasing amount of UTP, RSV RdRP complexes elongated the primer until reaching a G in third position on the template strand, and continued further upon addition of CTP ( Fig. 2C) ( fig. S4 and data S1). Replacing UTP with 4′-FlU-TP resulted in efficient primer extension up to the third nucleotide, confirming that RSV RdRP recognizes and incorporates 4′-FlU in place of UTP (Fig. 2C). Incorporation kinetics (21) showed only a moderate reduction in substrate affinity for 4′-FlU-TP compared with UTP ( Fig. 2D). Further addition of CTP to the reaction mix resulted in limited elongation rather than the expected full-length product, which suggested delayed polymerase stalling by incorporated 4′-FlU (fig .  S4 and data S1). Direct side-by-side comparison with GS-443902-the active metabolite of remdesivir and a "delayed polymerase stalling" inhibitor well-characterized for SARS-CoV-2along with RSV and other RNA viruses (21,22), corroborated this antiviral effect of 4′FlU-TP ( Fig. 2E and data S1).
When a modified template coded for incorporation of only a single UTP ( Fig. 2F and data S1), primers elongated preferentially to position i+3 after 4′-FlU-TP, whereas the efficiency of full elongation was strongly reduced compared with extension in the presence of UTP. However, repositioning the incorporation site further downstream in the template triggered immediate polymerase stalling at position i ( fig. S5), indicating template sequence dependence of the inhibitory effect. Transcription stalling at i or i+3 were also observed after multiple 4′-FlU incorporations: an AxAxxx template ( Fig. 2G) and direct tandem incorporations through an AAxxAx template (fig. S5) caused stalling at position i, whereas increasing spacer length between the incorporated uridines shifted preferential stalling to i+3 ( fig. S5). This variable delayed polymerase stalling within one to four nucleotides of the incorporation site was equally prominent when we examined de novo initiation of RNA synthesis at the promoter with a synthetic native RSV promoter sequence rather than extension of primertemplate pairs ( fig. S6).

4′-FlU is rapidly anabolized, metabolically stable, and potently antiviral in disease-relevant well-differentiated HAE cultures
Quantitation of 4′-FlU and its anabolites in primary HAE cells (Fig.  3A) demonstrated rapid intracellular accumulation of 4′-FlU, reaching a level of 3.42 nmol/million cells in the first hour of exposure (Fig. 3B). Anabolism to bioactive 4′-FlU-TP was efficient, resulting in concentrations of 10.38 nmol/million cells at peak (4 hours after exposure start) and 1.31 nmol/million cells at plateau (24 hours). The anabolite was metabolically stable, remaining present in sustained concentrations of approximately 1 nmol/million cells over a 6-hour monitoring period, corresponding to an extrapolated half-life of 9.7 hours (Fig. 3C).
To explore efficacy in a disease-relevant human tissue model, we cultured the HAEs at the air-liquid interface, inducing the formation of a well-differentiated 3D airway epithelium that included ciliated and mucus-producing cells (25) (Fig. 3D). Adding 4′-FlU to the basolateral chamber of the transwells after apical infection of the epithelium with RSV potently reduced apical virus shedding with an EC50 of 55 nM (Fig. 3E). Overall titer reduction spanned nearly four orders of magnitude, ranging from 3.86×10 4 TCID50 in control cells to 78.18 TCID50 at 5 μM basolateral 4′-FlU, approaching the level of detection.
Confocal microscopy validated formation of a pseudostratified organization of the epithelium with tight junctions in the airway epithelium tissue model (Fig. 3F), visualized efficient RSV replication in vehicle-treated tissue models (Fig. 3G), and confirmed near-sterilizing antiviral efficacy in the presence of 50 μM basolateral 4′-FlU ( Fig. 3H and figs. S8 and S9). Under the latter conditions, positive staining for RSV antigen was rarely detected.

4′-FlU is orally efficacious in a therapeutic dosing regimen in a small-animal model of RSV infection
To test 4′-FlU efficacy in vivo, we employed the mouse model of RSV infection (supplementary text), challenging animals with recRSV-A2-L19F, which efficiently replicates in mice (16). In a dose-to-failure study, we infected BALB/cJ mice intranasally and initiated once-daily oral treatment two hours after infection at 0.2, 1, or 5 mg 4′-FlU/kg body weight. Treatment at all dose levels resulted in a statistically significant reduction in lung virus load compared with vehicle-treated animals (Fig. 4A). The antiviral effect was dose dependent and approached nearly two orders of magnitude at the 5 mg/kg dose. Consistent with high metabolic stability in HAEs, a twice-daily dosing regimen did not significantly enhance efficacy (fig. S10). Becuase animal appearance, body weight, temperature (fig. S11), and relative lymphocyte and platelet counts (Fig. 4B and fig. S12) were unchanged in the 5 mg/kg group compared with vehicle-treated animals, we selected this dose for further studies.
For a longitudinal assessment of therapeutic benefit, we employed an in vivo imaging system (IVIS) with a red-shifted luciferase (26) expressing a RSV reporter virus generated for this study. This assay allows a noninvasive spatial appreciation of intrahost viral dissemination. Daily imaging (Fig. 4C and fig. S13) revealed considerable reduction of bioluminescence intensity in lungs of 4′-FlU-treated animals at day 5 post-infection, corresponding to peak viral replication, independent of whether treatment was initiated 24 hours before or 1 hour after infection (Fig. 4D). This IVIS profile is consistent with reduced viral replication and ameliorated viral pneumonia in treated animals.
To probe the therapeutic window of 4′-FlU, we initiated treatment at 2, 12, 24, 36, and 48 hours after infection. All treatment groups showed a statistically significant reduction of lung virus burden compared with vehicle-treated animals, but effect size was dependent on the time of treatment initiation ( Fig. 4E and  fig. S14). On the basis of our experience with therapeutic intervention with related respiratory RNA viruses that cause lethal disease (25), we require a reduction of lung virus load of at least one order of magnitude. With this constraint, the therapeutic window of 4′-FlU extended to 24 hours after infection in mice.

4′-FlU is effective against SARS-CoV-2 in HAE and the ferret model
To test activity against SARS-CoV-2 in the human airway organoids, we first confirmed that the WA1 isolate replicated efficiently in HAEs of all donors tested (Fig. 5, A to C, and fig. S15). Treatment of infected organoids with basolateral 4′-FlU dose-dependently reduced apical virus shedding, albeit with a limited maximal effect size of approximately two orders of magnitude at 50 μM (Fig. 5D). Confocal microscopy revealed that the epithelium was largely devoid of SARS-CoV-2 nucleocapsid protein under these conditions (Fig. 5E), with only sporadic staining detectable in a small subset of ciliated cells ( Fig. 5F and fig. S15).
To probe for a corresponding antiviral effect in vivo, we determined efficacy of oral 4′-FlU against an early pandemic isolate (WA1) and VoC alpha, gamma, and delta in the ferret model (27), which recapitulates hallmarks of uncomplicated human infection (3). For dose level selection in ferrets, we determined single oral dose ferret pharmacokinetic (PK) profiles of 4′-FlU. When administered at 15 or 50 mg/kg, peak plasma concentrations (Cmax) of 4′-FlU reached 34.8 and 63.3 μM, respectively, and overall exposure was 154 ± 27.6 and 413.1 ± 78.1 hours×nmol/ml, respectively, revealing good oral dose-proportionality ( Fig. 6A and table  S3). On the basis of this PK performance, we selected once-daily dosing at 20 mg/kg body weight for efficacy tests (Fig. 6B).
Intranasal infection of ferrets with 1×10 5 PFU of each isolate resulted in rapid viral shedding into the upper respiratory tract, which plateaued in vehicle-treated animals 48 to 60 hours after infection (Fig. 6C). Therapeutic treatment with 4′-FlU initiated 12 hours after infection reduced virus burden in nasal lavages by approximately three orders of magnitude (WA1) to <50 PFU/ml within 12 hours of treatment onset. All three VoC were highly sensitive to 4′-FlU, remaining below the level of detection 36 to 48 hours after onset of oral treatment. Viral titers in nasal turbinate tissue extracted 4 days after infection (Fig. 6D) and associated viral RNA copy numbers (fig. S16) correlated with this reduction in shed virus load. Shedding of infectious particles ceased completely in all animals after 2.5 days of treatment (3 days post-infection).

Conclusions
This study identifies and characterizes the ribonucleoside analog 4′-FlU, which potently inhibits pathogens of different clinicallyrelevant negative and positive-sense RNA virus families. The compound causes delayed stalling of RSV and SARS-CoV-2 polymerases within in vitro RdRP assays, reminiscent of the antiviral effect of remdesivir (28,29). However, 4′-FlU can also trigger immediate RdRP stalling depending on sequence context, suggesting steric hindrance of polymerase advance or of accommodating the next incoming nucleotide as the underlying MOA. We cannot exclude that additional effects further enhance the antiviral effect in cellula as proposed for other nucleoside analogs (30). Slightly lower sensitivity of SARS-CoV-2 to 4′-FlU compared with RSV could be a result of the exonuclease activity of the coronavirus polymerase, which can eliminate ribonucleoside analogs (31,32). Alternatively, coronavirus RdRP may have a greater capacity to tolerate the compound, because SARS-CoV-2 RdRP showed a higher tendency than RSV polymerase to advance after 4′-FlU-TP incorporation in the RdRP assays, which do not contain exonuclease functionality.
Once-daily oral administration to mice and ferrets significantly reduced the burden of RSV and SARS-CoV-2, respectively, when treatment was initiated up to 24 (RSV) or 12 (SARS-CoV-2) hours after infection. Because RSV (33) and SARS-CoV-2 (34) host invasion is slower in humans, these data outline a viable therapeutic window for human treatment. Equally potent activity against SARS-CoV-2 VoC alpha, gamma, and delta demonstrated broad anti-coronavirus efficacy of 4′-FlU, building confidence that the compound will remain active against future VoC that may be increasingly less responsive to spike-targeting vaccines or antibody therapeutics. Formal tolerability studies are pending, but 4′-FlU was well tolerated by the human organoid models and efficacious in murids and mustelids. Blood analysis of treated mice uncovered no antiproliferative effect of 4′-FlU on the hematopoietic system. These results establish 4′-FlU as a broad-spectrum orally efficacious inhibitor of major RNA viruses, making it a promising therapeutic option for RSV disease and COVID-19, and a muchneeded contributor to improving pandemic preparedness. editing: J.S. and R.K.P. Competing interests: G.R.B. and G.R.P. are coinventors on patent WO 2019/1736002 covering composition of matter and use of EIDD-2749 and its analogs as an antiviral treatment. This study could affect their personal financial status. All other authors declare that they have no competing interests. Data and materials availability: All data are available in the main text or the supplementary materials. Materials and methods are available as supplementary materials at the Science website. Transfer of EIDD-2749 material to other institutions for research purposes is covered by MTAs from Emory University. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/. This license does not apply to figures/photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material.