Restriction of SARS-CoV-2 replication by targeting programmed −1 ribosomal frameshifting Yu Suna, Laura Abriolab, Rachel O. Niedererc, Savannah F. Pedersend, Mia M. Alfajaroe,f, Valter Silva Monteirof, Craig B. Wilene,f, Ya-Chi Hod, Wendy V. Gilbertc, Yulia V. Surovtsevab, Brett D. Lindenbachd,1, and Junjie U. Guoa,1 aDepartment of Neuroscience, Yale University School of Medicine, New Haven, CT 06520; bYale Center for Molecular Discovery, Yale University, West Haven, CT 06516; cDepartment of Molecular Biophysics and Biochemistry, Yale University School of Medicine, New Haven, CT 06520; dDepartment of Microbial Pathogenesis, Yale University School of Medicine, New Haven, CT 06520; eDepartment of Laboratory Medicine, Yale University School of Medicine, New Haven, CT 06520; and fDepartment of Immunobiology, Yale University School of Medicine, New Haven, CT 06520 Edited by Lynne E. Maquat, University of Rochester School of Medicine and Dentistry, Rochester, NY, and approved May 21, 2021 (received for review November 12, 2020) Translation of open reading frame 1b (ORF1b) in severe acute viral FSEs are attractive RNA targets for specific interference respiratory syndrome coronavirus 2 (SARS-CoV-2) requires a pro- with viral gene expression (11, 12). Indeed, mutations and drugs grammed −1 ribosomal frameshift (−1 PRF) promoted by an RNA that alter the efficiency of −1 PRF in HIV-1, which regulates pseudoknot. The extent to which SARS-CoV-2 replication may be Gag-Pol translation, have been shown to impede HIV-1 repli- sensitive to changes in −1 PRF efficiency is currently unknown. cation (11, 13, 14). In addition, an interferon-induced host pro- Through an unbiased, reporter-based high-throughput compound tein, Shiftless (SHFL), has been shown to interact with HIV-1 screen, we identified merafloxacin, a fluoroquinolone antibacterial, FSE, inhibit −1 PRF, and restrict HIV-1 replication (15), sug- as a −1 PRF inhibitor for SARS-CoV-2. Frameshift inhibition by mera- gesting that frameshift inhibition has become part of the host floxacin is robust to mutations within the pseudoknot region and is antiviral response. Several compounds have recently been shown similarly effective on −1 PRF of other betacoronaviruses. Consistent to modulate −1 PRF of SARS-CoV-2 to varying degrees with the essential role of −1 PRF in viral gene expression, meraflox- (16–18), although the specificity of these compounds and their acin impedes SARS-CoV-2 replication in Vero E6 cells, thereby provid- antiviral activity remain unclear. ing proof-of-principle for targeting −1PRFasaplausibleand Here, we show that merafloxacin, a fluoroquinolone com- effective antiviral strategy for SARS-CoV-2 and other coronaviruses. pound, specifically and robustly inhibits −1 PRF in betacor- BIOCHEMISTRY onaviruses including SARS-CoV-2. Inhibition of −1 PRF by translation | ribosomal frameshifting | RNA pseudoknot | coronavirus | merafloxacin impeded SARS-CoV-2 replication in Vero E6 cells, merafloxacin indicating that targeting −1 PRF represents a plausible antiviral strategy for SARS-CoV-2 and potentially other coronaviruses. evere acute respiratory syndrome coronavirus 2 (SARS-CoV- Results S2), the etiological agent of COVID-19, belongs to a family of zoonotic human coronaviruses. Upon the entry of SARS-CoV-2 A High-Throughput Screen Identifies −1 PRF Modulators for − into host cells, the first set of viral proteins are translated from SARS-CoV-2. To quantify 1 PRF efficiency in uninfected cells, B the long (>21-kb) open reading frame ORF1ab, which takes up we constructed a plasmid-based frameshifting reporter (Fig. 1 ), approximately two-thirds of the viral genome (Fig. 1A) (1, 2). replacing the stop codon of a firefly luciferase (FLuc) coding The ORF1ab-encoded polyprotein is subsequently processed into 16 individual nonstructural proteins (nsp) by two proteases, Significance PLpro/nsp3 and 3CLpro/nsp5. The 3′ half of ORF1ab, ORF1b, encodes a variety of enzymes critical for viral transcription and A large variety of RNA viruses, including the novel coronavirus replication, including an RNA-dependent RNA polymerase SARS-CoV-2, contain specific RNA structures that promote (RdRp/nsp12), an RNA helicase (Hel/nsp13), a proofreading programmed ribosomal frameshifting (PRF) to regulate viral exoribonuclease and N7-guanosine methyltransferase (ExoN/ gene expression. From a high-throughput compound screen, nsp14), an endonuclease (NendoU/nsp15), and a 2′-O-methyl- we identified a PRF inhibitor for SARS-CoV-2 and found that it transferase (nsp16). In all coronaviruses, translation of ORF1b substantially impeded viral replication in cultured cells. Inter- requires a programmed −1 ribosomal frameshift (−1 PRF) (3). estingly, the compound could target not only SARS-CoV-2 but When ribosomes arrive at the end of ORF1a, instead of con- also other coronaviruses that use similar RNA structures to tinuing elongation and soon terminating at an adjacent in-frame promote frameshifting. These results suggest targeting PRF is a stop codon, a subset of ribosomes backtrack by one nucleotide plausible, effective, and broad-spectrum antiviral strategy for and are repositioned in the −1 reading frame before continuing SARS-CoV-2 and other coronaviruses. the translation elongation cycle, thereby producing a full-length ORF1ab polyprotein. A −1 PRF region often contains two main Author contributions: Y.S., Y.V.S., B.D.L., and J.U.G. designed research; Y.S., L.A., R.O.N., components: a heptanucleotide slippery sequence (UUUAAAC S.F.P., M.M.A., and V.S.M. performed research; Y.S., L.A., R.O.N., S.F.P., M.M.A., V.S.M., in SARS-CoV-2) and a downstream stable secondary structure C.B.W., Y.-C.H., W.V.G., Y.V.S., B.D.L., and J.U.G. analyzed data; and B.D.L. and J.U.G. wrote the paper. acting as a frameshift-stimulatory element (FSE). The FSE Competing interest statement: Yale University has filed a provisional patent application secondary structures, most commonly being RNA pseudoknots related to this work titled “Compounds and Compositions for Disrupting Programmed (4, 5), are thought to facilitate −1 PRF in part by transiently Ribosomal Frameshifting.” pausing the incoming ribosome, allowing tRNAs to realign This article is a PNAS Direct Submission. within the slippery sequence. In betacoronaviruses, including This open access article is distributed under Creative Commons Attribution License 4.0 SARS-CoV-2, a three-stem pseudoknot (Fig. 1B) has been (CC BY). proposed to act as the productive conformation (3), although 1To whom correspondence may be addressed. Email: [email protected] or junjie. conformational dynamics of this region have recently been [email protected]. described (6–8). This article contains supporting information online at https://www.pnas.org/lookup/suppl/ In contrast to its wide adoption by RNA viruses, −1 PRF is doi:10.1073/pnas.2023051118/-/DCSupplemental. much less prevalent in cellular mRNAs (5, 9, 10). Therefore, Published June 14, 2021. PNAS 2021 Vol. 118 No. 26 e2023051118 https://doi.org/10.1073/pnas.2023051118 | 1of6 Downloaded by guest on September 30, 2021 A respectively. Based on a positive control construct in which FLuc ORF1a S ORF3-10 and RLuc were translated continuously without frameshifting, ORF1b we estimated the PRF efficiency to be ∼20% in HEK293T cells, FSE B consistent with previous reporter-based measurements (16, 20), FLuc RLuc(−1) C although substantially lower than ribosome footprint profiling- based measurements (21) (Discussion). Stem 1 U-G To make the PRF reporter more suited for high-content …UUUAAACGGGUUUGCGGUGUAAG C screening, we replaced FLuc and RLuc with mCherry and GFP •••••••••• | RLuc/FLuc ** ** ** ACG D C GAC-ACGGCGUGCCACAUUCUG-CCCGA 0.2 0.1 0 (Fig. 1 ), respectively. Consistent with the luciferase-based re- U ••• •••••• •• ••••• porter, mCherry-FSE-GFP(−1) reporter yielded reduced GFP A CUGAUGUCGU---A--U---ACAGGGCUUUU… GUA signals compared to the in-frame mCherry-GFP fusion construct Stem 3 Stem 2 without an FSE (Fig. 1 D and E). Shifting GFP back to the DEmCherry GFP No-FSE mCh-GFP fusion 0 frame after FSE [mCherry-FSE-GFP(0)] completely abolished mCh-FSE-GFP(−1) GFP signal (Fig. 1 D and E), consistent with the expectation that mCh-FSE- mCh-FSE- No-FSE mCh-FSE-GFP(0) GFP(0) GFP(−1) mCh-GFP fusion nonframeshifted ribosomes would terminate at an in-frame UAA stop codon within Stem 1 (Fig. 1B). To identify chemical modifiers of −1 PRF, we treated Percentofcells HEK293T cells in 384-well plates with each of 4,434 compounds 0 20 40 60 40 20 0 μ F A mCherry 0 0.3 0.6 0.9 1.2 at 10 M final concentration (Fig. 1 and Dataset S1 ), which GFP/mCherry ratio included 640 FDA-approved drugs, 1,600 compounds from the Pharmakon 1600 collection, and 1,872 compounds from a GFP G Tested-In-Human collection, and transfected cells with the Ivermectin mCherry-FSE-GFP(−1) reporter plasmids. Twenty-four hours F after transfection, total cell numbers, as well as mCherry and HEK293T Add Transfect Merafloxacin GFP signals in transfected (mCherry+) cells were quantified. cells compounds reporter Effect (%) Effect GFP/mCherry ratios were compared to the in-frame mCherry- Zʹ = 0.91-0.95 GFP fusion positive control as well as the mCherry-FSE-GFP(0) 24 hr 30 min 24 hr Automated imaging −60 −40 −20 0 20 negative control. Our high-content screening showed high ro- & quantification 0 2000 4000 bustness, with Z′ scores ranging between 0.91 and 0.95. Rank As expected, the vast majority of the tested compounds had Fig. 1. A high-throughput screen identifies SARS-CoV-2 PRF modulators. (A) little or no effect on the GFP/mCherry ratio (Fig. 1G). Screen Schematic illustration of the SARS-CoV-2 genome architecture, with the FSE actives were selected by using mean ± 3 SDs as cutoffs. False indicated. (B) Schematic illustration of the dual luciferase-based −1 PRF re- positives due to their intrinsic fluorescence (e.g., doxorubicin porter design.