Parallel shRNA and CRISPR-Cas9 screens enable antiviral drug target identification

Abstract

Broad-spectrum antiviral drugs targeting host processes could potentially treat a wide range of viruses while reducing the likelihood of emergent resistance. Despite great promise as therapeutics, such drugs remain largely elusive. Here we used parallel genome-wide high-coverage short hairpin RNA (shRNA) and clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 screens to identify the cellular target and mechanism of action of GSK983, a potent broad-spectrum antiviral with unexplained cytotoxicity. We found that GSK983 blocked cell proliferation and dengue virus replication by inhibiting the pyrimidine biosynthesis enzyme dihydroorotate dehydrogenase (DHODH). Guided by mechanistic insights from both genomic screens, we found that exogenous deoxycytidine markedly reduced GSK983 cytotoxicity but not antiviral activity, providing an attractive new approach to improve the therapeutic window of DHODH inhibitors against RNA viruses. Our results highlight the distinct advantages and limitations of each screening method for identifying drug targets, and demonstrate the utility of parallel knockdown and knockout screens for comprehensive probing of drug activity.

Figure 1

shRNA and CRISPR-Cas9 screens to identify the cellular target and mechanism of action of GSK983. (a) Structure of GSK983. (b) GSK983 dose response in K562 cells. Viable cells were counted by flow cytometry (FSC/SSC) following 72 h GSK983 treatment at the indicated concentration. Error bars represent ± standard deviation of 8 biological replicates from two independent experiments. Schematic representation of genome-wide shRNA (c) and CRISPR-Cas9 (d) screens. (e) Top ten hits from the shRNA and CRISPR-Cas9 screens in cellular and biological context. Circle size is proportional to MLE score absolute value. Square size is proportional to median fold-enrichment or disenrichment. (f) Comparative analysis of results from shRNA and CRISPR-Cas9 screens. Pyrimidine metabolism (orange), CoQ₁₀ biosynthesis (blue), regulation of mTORC1 activity (green). (g) Validation of selected top hit genes from the shRNA screen using a competitive growth assay. A total of 27 shRNAs were retested (6 targeting DHODH; 4 targeting CMPK1; 3 each targeting COQ2, PDSS2, PDSS1, and COQ10B; and 5 negative controls). Error bars represent ± standard deviation of log(mCherry enrichment ratio) values for all retested shRNAs targeting each gene. P values were calculated by Mann-Whitney U test. Validation of selected sensitizing (h) or protective (i) sgRNAs from the CRISPR-Cas9 screen using a competitive growth assay. Bars represent the average of two biological replicates.

Figure 2

GSK983 inhibits DHODH to block virus replication and cell proliferation. (a) Schematic representation of mammalian pyrimidine metabolism. Genes that appeared as strong sensitizing hits in the shRNA screen (CMPK1, DHODH, UCK2) and CRISPR-Cas9 screen (UCK2) are highlighted in yellow. (b) Dihydroorotic acid had no effect on GSK983-induced growth inhibition in K562 cells. (c) Orotic acid reversed GSK983-induced growth inhibition in K562 cells. For (b) and (c), viable cells were counted by flow cytometry (FSC/SSC) following 72 h treatment with 48 nM GSK983 or vehicle and the indicated concentration of (dihydro)orotic acid. Error bars represent ± standard deviation of 4 biological replicates. (d) GSK983 and analogues inhibited recombinant human DHODH in vitro. K_i values are averages of two independent K_i determinations at different inhibitor concentrations. The range between independently calculated K_i values for each inhibitor is shown in parentheses. IC₅₀ values for inhibition of episomal HPV-16 replication in cell-based antiviral assays are those reported by GlaxoSmithKline. (e) Structures of GSK983, 6Br-pF, 6Br-oTol, and GSK984.

Figure 3

Deoxycytidine (dC) reverses the anti-proliferative effect of GSK983 but not antiviral activity. Uridine (a) and deoxycytidine (b) largely reversed GSK983-induced growth inhibition in K562 cells. For (a) and (b), viable cells were counted by flow cytometry (FSC/SSC) following 72 h treatment with 48 nM GSK983 or vehicle and the indicated concentration of uridine or deoxycytidine. Error bars represent ± standard deviation of 4 biological replicates. (c) GSK983 inhibited replication of luciferase-expressing DENV in A549 cells (black). 1 mM uridine reversed antiviral activity (blue), while 1 mM deoxycytidine did not (red). Error bars represent ± standard deviation of 3 biological replicates. (d) Uridine (blue) and deoxycytidine (red) reversed GSK983-induced growth inhibition in A549 cells. Viable cells were counted by flow cytometry (FSC/SSC) following 72 h treatment with no exogenous pyrimidines (control), 1 mM uridine, or 1 mM deoxycytidine and the indicated concentration of GSK983. Error bars represent ± standard deviation of 3 biological replicates. (e) Ribonucleoside (uridine) salvage sustains both RNA virus replication and cellular DNA synthesis. (f) Deoxyribonucleoside (deoxycytidine) salvage sustains cellular DNA synthesis but not RNA virus replication. For (e) and (f), Pyr = pyrimidine, rNuc = ribonucleotides, dNuc = deoxyribonucleotides. (g) Deoxycytidine reversed GSK983-induced S phase cell cycle arrest in K562 cells. Following 24 h treatment with 48 nM GSK983, cells were treated with 10 μM 5-ethynyl-2′-deoxyuridine (EdU) for 2 h and fixed in 70% EtOH. Cells were stained with Azide-fluor 488 and 7-AAD and analyzed by flow cytometry. Flow cytometry plots depict one of three biological replicates.