Screening assay generation
Screening assays are based on the replication machinery of vesicular stomatitis virus (VSV). Two-component systems were based on replication-incompetent VSV∆L-dsRed and VSV∆P-dsRed27 as the viral components and transduced 293T (American Type Culture Collection, Manassas, VA) or BHK21 (American Type Culture Collection, Manassas, VA) cells. 293T cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) (Lonza, Basel, Switzerland) supplemented with 10% FCS (Invitrogen, Carlsbad, California, USA), 1% P/S (PAA Laboratories, Pasching, Austria), 2% glutamine (PAA Laboratories), 1x sodium pyruvate (Gibco, Carlsbad, California, USA), 1x non-essential amino acids (Gibco). BHK21 cells were cultured in Glasgow minimum essential medium (GMEM) (Lonza) supplemented with 10% fetal calf serum (FCS), 5% tryptose phosphate broth, 100 units/ml penicillin and 0.1 mg/ml streptomycin (P/S) (Gibco).
First, regulatable constructs were cloned into VSV-GFP48,49. A VSV-GFP variant coding 3CLpro-Off was cloned by Gibson assembly (NEB, Ipswich, USA)50. The intergenic region between GFP and L was removed by restriction enzyme digestion of sites close to the end of GFP (MscI) and after the beginning of L (HpaI). Missing GFP and L parts were PCR amplified and overlapping sequences to the vector and SARS-CoV-2 protease inserted by primer pairs GFP-34bp-before-MscI-for / GFP-prot-rev and prot-L-for / L-33bp-after-HpaI-rev, respectively (Table 1). The SARS-CoV-2 protease sequence was retrieved from cDNA of purified virus isolates. The sequence of the SARS-CoV-2 protease corresponds to the Wuhan-1 isolate sequence (NCBI Reference Sequence: NC_045512.2, Supplementary Table 1). An additional restriction site (BbvCI) was introduced in the N-terminal protease recognition sequence to facilitate further cloning. We used this BbvCI (NEB) site together with NheI (NEB) to remove GFP and introduce a firefly luciferase, thereby generating VSV-Luc-SARS-Prot-Off with primers IGR (intergenic region)-luciferase for and luciferase-cut1-rev (Table 1).
A VSV-GFP variant coding 3CLpro-On was also cloned via Gibson assembly. VSV-GFP was digested with BstZ17l and XbaI (NEB). Vector and SARS-CoV-2 protease overlapping fragments were PCR amplified with primers N-35nt-before-BstZ17l-for and P-GGSG-rev for upstream sequences of N and P that were omitted in the digestion and GGSG-P-for and P-35nt-after-XbaI-rev for downstream P sequence (Table 1). The SARS-CoV-2 protease was amplified from cDNA with primers (GGSG)3-prot-for and (GGSG)3-prot-rev (Table 1).
VSV∆P-Luciferase/Luc was cloned by digestion of the VSV∆P-dsRed plasmid with SmaI and NotI and Gibson assembly with a PCR on a firefly luciferase plasmid with primers ∆P-Luc-for and -rev.
Two-component system lentiviruses used in this study originate from blasticidin resistance encoding pLenti CMVie-IRES-BlastR (Addgene accession: #119863). An additional variant resistance gene was cloned into pLenti CMVie-IRES-BlastR, namely hygromycin (pLenti CMVie-IRES-HygroR). First, the blasticidin resistance was omitted by digestion of pLenti CMVie-IRES-BlastR with MscI and NotI (NEB). Then, hygromycin resistance was added by Gibson assembly of resistance genes plus vector overlapping sequences introduced by PCR.
Lentiviruses encoding regulatable 3CLpro-Off and -On switches were generated with Gibson assembly as follows. For 3CLpro-Off, first an L protein blasticidin lentivirus was cloned with primers LV-L-for and L-LV-rev (Table 2) to facilitate the cloning of the large fusion construct. This plasmid was then digested with NheI and HpaI. GFP, the SARS-CoV-2 protease and the N-terminal part of L omitted by HpaI digestion were replaced by a PCR on the full VSV-3CLpro-Off-GFP plasmid with primers LV-GFP-for and L-rev (Table 2). 3CLpro-On was amplified fully from VSV-3CLpro-On-GFP plasmid with primers LV-P-for and P-LV-rev (Table 2). Both constructs were deposited in GenBank with accession numbers ON262564 (3CLpro-Off blasticidin) and ON262565 (3CLpro-On hygromycin).
Hygromycin-resistance-based lentivirus vectors encoding variable proteases were cloned via Gibson assembly. First, proteases and adjacent cleavage sites were codon-optimized and synthesized by Integrated DNA Technologies (USA). For integration into inhibition-on constructs, they were amplified and extended with a flexible (GGSG)3-linker sequence with primer pairs GGSG-*virus*-On-for and *virus*-GGSG-On-rev (Table 2). The (GGSG)3-linker sequence was used for a fusion PCR with N- and C-terminal phosphoprotein fragments generated by PCRs on previous phosphoprotein constructs with primer pairs LV-P-for with P-GGSG-rev and GGSG-P-for with P-LV-rev. The phosphoprotein fragments containing GGSG sequences and (GGSG)3-linker sequence extended proteases were joined with a fusion PCR with primers LV-P-for and P-LV-rev. These fusion PCRs were then ligated into a hygromycin-resistance vector digested with NheI and PacI.
N-terminal glutamine to asparagine mutants of variable proteases were generated via Gibson assembly. Forward primer containing a (GGSG)3-linker sequence and a Q-to-N mutation were designed for all proteases (Table 2). PCRs on existing constructs of variable proteases were performed with primers *virus*-On-N-term-QtoN-rev was paired and P-LV-rev. These fragments spanning the respective protease and the C-terminal phosphoprotein fragment were fused to the N-terminal phosphoprotein sequence via a fusion PCR and then ligated into a hygromycin-resistance vector digested with NheI and PacI.
Lentiviral transduction
Lentiviruses were generated by CaPO4 transfection of lentiviral plasmids together with Gag-Pro-Pol and VSV glycoprotein51. Lentivirus containing supernatants were harvested 24 and 48 h after changing transfection medium containing chloroquine (~ 12 h after CaPO4 transfection) and pooled. Pooled supernatants were used to perform spin-infection of 4 × 105 293T or BHK21 cells per well in a 6-well plate at 1000 g and 37 °C with medium containing 8 µg/ml polybrene (Sigma, St. Louis, US)52. Two days after transduction, cells were split selected either by 12 µg/ml for 293T and 12 µg/ml for BHK21 blasticidin (InvivoGen, France) and 400 µg/ml for 293T and 600 µg/ml for BHK21 hygromycin (InvivoGen, France).
Screening assay with FluoroSpot read-out
Compounds were screened in a 96-well format. Ten thousand 293T or BHK21 cells expressing either a regulatable construct or native VSV proteins were seeded per well (cell number can be adjusted up to 20.000 cells per well for toxic compounds). Four hours after seeding, compounds and virus (multiplicity of infection, MOI: 0.01 of VSV-∆P-RFP, MOI: 0.1 of VSV-∆L-RFP) were added to wells. After 40 h, supernatants were removed, and fluorescent spots counted in a Fluoro/ImmunoSpot counter (CTL Europe GmbH, Bonn, Germany) with the manufacturer-provided software CTL switchboard 2.7.2. 90% of each well area was scanned concentrically to exclude reflection from the well edges, and counts were normalized to the full area. Automatic fiber exclusion was applied while scanning. The excitation wavelength for RFP was 570 nm, the D_F_R triple band filter was used to collect fluorescence. Manual quality control for residual fibers was also performed. In parallel, plates with the same compound treatment scheme were incubated with 20 µL of 5 mg/ml Thiazolyl Blue Tetrazolium Bromide / MTT (SIGMA) for 4 h, then lysed with 0.1 g sodium dodecyl sulfate/ml 0.01 M HCl over night with gentle shaking. MTT absorbance was measured at 550 nm (main absorbance) and 655 (base absorbance to substract).
Alternatively, spot counts were performed with a BZ-X810 All-in-One fluorescence microscope from Keyence (Ōsaka, Japan) or a Cytation|1 Imaging reader from BioTek (Vermont, USA). Exemplary read-outs are displayed in Supplementary Fig. 7.
Screening assay with fluorescence-activated cell scanning (FACS) read-out
Transiently transfected 293T cells: 4 × 105 cells per 6-well were seeded 1 day prior to transfection. Plasmids were transfected with TransIT®-LT1 transfection kit (Mirus Bio LLC, Madison, WI, USA) with 800 ng plasmid DNA and the recommended 1: 3 µg-DNA: µl-reagent ratio. Six to twelve hours after transfection, 293T cells were split and 15.000 cells seeded in 96-well plates in 50 µl medium. Compound and virus (MOI 0.1) were added in 50 µl to reach desired concentrations. After 2 days, cells were detached with 0.05 % Trypsin-EDTA (Gibco) and transferred to a 96-well round-bottom plate (TPP Techno Plastic Products AG, Switzerland) for automatic sampling by fluorescence-activated cell scanning (BD LSRFortessa X-20). Stably transduced BHK21 cells: fifteen-thousand cells were seeded and immediately treated with compound doses and VSV-∆P-RFP at MOI 0.01. After 2 days, cells were treated as described above and measured via FACS. Gates for live cells and singlets were applied before division in fluorescent and non-fluorescent 293T and BHK21 cells (Supplementary Figs. 5 and 6). Both mean fluorescence intensity (MFI) of singlets and dsRed fluorescent events of P1 are used to quantify read-outs. MFI read-outs were more sensitive and were therefore used for transiently transfected cells, which generally showed a weaker response than stably transduced cells. Stably transduced cells treated with toxic compounds showed autofluorescence. Hence, MFI was less useful to distinguish actual signal from artefacts. Thus, we chose dsRed events to display compound cross-comparisons.
Screening assay with luciferase read-out
Transiently transfected 293T cells and stably transduced BHK21 cells were prepared as described above in FACS screening. After 2 days, BHK21 cells were lysed with Bright-Glo™ (Promega, Madison, USA) and measured with a SPARK bioluminescence reader from Tecan (Grödig, Austria). 293T cells were prepared with VivoGlo™ luciferin (Promega) and measured with a GloMax Explorer (Promega).
Statistics and reproducibility
Reproducibility
Sample sizes were chosen empirically based on experience from our previous studies. At least two biologically independent replicates were performed per condition. Biologically independent replicates meaning distinct wells with the same conditions, not only multiple measurements of the same well. Although inter-assay variability of overall signal was observed (see also Supplementary Fig. 1g, h), intra-assay comparison distinguished compound potency reliably.
Z-factor
To validate basic metrics of the assay, we calculated Z-factors for each 3CLpro gain-of-function (On – N – C – N,C) expressing cell line. We observed that the most of the calculated Z-factor values are within the interval indicated by Zhang et al.53. for an excellent assay. In brief, threshold values for both positive and negative control values were calculated as described in the following. The negative threshold value is equal to the mean signal of the negative controls plus three times their standard deviation (SD), while the positive threshold value is equal to the mean signal of the positive controls minus three times their SD. Then, the difference between the thresholds has been calculated and defined as S, ‘separation band’. We subsequently computed the absolute difference between the two means, defined as ‘dynamic range’, R. Ultimately, we calculated the Z-factor as the product of S/R. These results suggest the reliability of this cell-based assay (Table 3).
Half maximal effective concentration (EC50)
Given a cellular system as basis of EC50 calculation, we expect the dynamic range of the assay to be greater than in a biochemical assay, where the amount of enzyme is stable. In an excess of potent compounds, protease-viral fusion proteins are constantly renewed and cells continue growth. Therefore, virus replicates continuously and produces more read-out through excess compound. At lower concentrations, compound molecules are depleted and signal plateaus.
Effective concentrations were calculated for 3CLpro-On N-term FACS data
GC376: 15.04 ± 7.58 SD (95% confidence interval 11.54–19.69); boceprevir: 20.83 ± 5.98 SD (95% confidence interval 17.74–24.44); PF-00835231: 11.10 ± 2.95 SD (95% confidence interval 10.53–11.71); nirmatrelvir: 10.83 ± 6.65 SD (95% confidence interval 8.532–13.72) (Fig. 4c). EC50 calculations were performed with Graphpad Prism 8 according to the recommendations of the software’s user guide. For each compound, the highest signal was defined as 100%, and values of all concentrations and replicates were normalized to this value. For every compound, we performed a nonlinear regression analysis, in particular the agonist-versus-response function. For EC50 the constrain (defined in Graphpad Prism 8 as the constant F) was set to 50. Top and bottom values were defined as 0 and 100, respectively. The standard deviation was calculated with following formula:
$${SD}=sqrt{frac{sum ({{residual}}^{2})}{n-K}}$$
(1)
Residuals are defined as the vertical distance of the point from the fit or curve. K is the number of parameters fit by regression.
Immunoblotting
Samples for immunoblotting were collected from 293T cells expressing VSV-L. We used this replication-supporting cell line for the expression of sufficient fusion protein (GFP-3CLpro-L) in the non-active condition (+ PI) of VSV-3CLpro-Off. SDS-PAGEs of protein lysates were performed under reducing conditions, on an 8% polyacrylamide gel for VSV-GFP and VSV-3CLpro-Off constructs. Gels were run for 90 min at 100 Volt. Proteins were transferred to 0.45-µm nitrocellulose membranes (Whatman, Dassel, Germany) by using a tank blotting system (Bio-Rad, Hercules, CA, US). The blotting time was 80 min. Blotting buffer contained 15 % methanol. The membranes were blocked overnight with 1x PBS containing 5% skim milk and 0.1% Tween 20 (PBSTM). GFP was stained by a mouse antibody (clones 7.1 and 13.1; Roche, Basel, Switzerland) diluted 1:1000 in PBSTM and a rabbit GFP/YFP antibody produced by Stephan Geley (conc.: 0.3 mg/mL) diluted in 1:2000 in PBSTM. β-Actin was stained with a monoclonal mouse anti-Actin antibody, A5441-.5 ML from SIGMA diluted 1:1000 in PBSTM. Blots were not stripped prior to β-Actin staining. PageRulerTM Prestained Protein ladder 26616 (ThermoFisher, Massachusetts, USA) was used as marker.
Fluorescence microscopy
105 BHK21 VSV-L expressing cells were seeded in glass-bottom dishes with four chambers (ibidi GmbH, Gräfelfing, Germany) 8 h before infection. Cells were infected with an MOI of 1. Single images were acquired up to 16 h after infection at 37 °C using a 63X/NA1.4 objective on an automated live cell imaging Zeiss Axiovert 200 M microscopy equipped with a Sola light engine LED light source (Lumencor, Visitron Systems GmbH, Puchheim, Germany), a pco.edge 4.2 scMOS camera (PCO AG, Kelheim, Germany), controlled by VisiView software (Visitron). Exposure times were 200 ms for GFP and 10 ms for phase contrast.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
