Protease-activatable biosensors of SARS-CoV-2 infection for cell-1 based drug , neutralisation and virological assays 2 3

15 The world is in the grip of a severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) 16 pandemic, and there is an urgent unmet clinical need for effective antiviral therapies. Many 17 inhibitors of viral enzymes identified in vitro have limited efficacy against viral replication in 18 cells, but conventional plaque assays are impractical for high-throughput screens. In this 19 study, we therefore engineer cell-based biosensors of SARS-CoV-2 infection. Our assays 20 exploit the cleavage of specific oligopeptide linkers by SARS-CoV-2 Main or Papain-like 21 proteases, leading to the activation of green fluorescent protein (GFP) or firefly luciferase22 based reporters. First, we characterise these biosensors in cells using recombinant viral 23 proteases. Next, we confirm their ability to detect endogenous viral protease expression during 24 infection with wildtype SARS-CoV-2. Finally, we develop a sensitive luminescent reporter cell 25 line, confirm that it accurately quantitates infectious SARS-CoV-2 virus, and demonstrate its 26 utility for drug screening and titration of neutralising antibodies. 27


Detection of recombinant SARS-CoV-2 protease activity in cells 112
To test whether these biosensors could detect SARS-CoV-2 protease activity in cells, we co- concentrations of either drug (Figure 2A-C). Conversely, GC373 did not inhibit PLP2-FlipGFP 152 or TEV-FlipGFP activation by PLPro c.d or TEV protease, respectively (Figure 2D-G). 153 Next, we evaluated a panel of compounds previously reported to have activity against MPro 154 in vitro, but which failed to inhibit viral replication in cells (Jin et al., 2020a). In each case, we 155 used the highest concentration tolerated by HEK293Ts for 24 h. Unlike GC373 and GC376, 156 none of these compounds inhibited MPro activity in cells (Figure 2-figure supplement 1A). 157 Finally, we tested the HIV-1 protease inhibitors lopinavir and ritonavir (Figure 2-figure  158 supplement 1B). These compounds have been reported to have variable activity against 159 MPro activity in vitro, but have failed to improve outcomes in clinical trials (Cao et al., 2020;160 Horby et al., 2020; WHO, 2020). They marginally inhibited MPro activity in cells, but only when 161 used at high doses, sufficient to impact transfection efficiency, reflected by a decreased 162 number of BFP+ cells (Figure 2-figure supplement 1C). In summary, these data support the 163 principle that cell-based assays for antiviral compounds correlate better with activity against 164 viral replication than in vitro assays. 165

Activation of FlipGFP-based reporters by wildtype SARS-CoV-2 infection 166
To test whether our FlipGFP-based reporters could be activated by viral protease expression 167 during SARS-CoV-2 infection, we made use of a permissive HEK293T cell line over- Accordingly, HEK293T-ACE2 cells were plated in chambered coverslips, transfected with the 176 indicated reporter constructs, then infected after 12 h with SARS-CoV-2. After a further 24 h 177 incubation, reporter activation was quantitated as the ratio of FlipGFP/mCherry fluorescence 178 in spike-positive syncytia, compared with uninfected cells. As expected, we observed a 179 consistent increase in FlipGFP fluorescence in infected cells transfected with either the Opt3c-180 FlipGFP or PLP2-FlipGFP reporter, but not infected cells transfected with non-cleavable 181 controls ( Figure 3A-B). 182 Whilst confirming that our FlipGFP reporters could be activated by SARS-CoV-2 infection, the 183 magnitude of effect was, however, markedly reduced compared with the over-expression of 184 recombinant viral proteases (compare Figure 1D and Figure 1F with Figure 3B). This likely 185 reflects lower levels of protease expression during viral infection, compared with over-186 expression of recombinant proteases ( Figure 3C). In addition, it is possible that the 187 localisation of proteases and/or presence of other viral proteins and endogenous polyprotein 188 substrates during authentic viral infection reduces their likelihood of encountering reporter 189 molecules. Taken together, these data provided proof-of-concept that protease-activatable 190 biosensors may be exploited to signal SARS-CoV-2 infection, but failed to demonstrate a 191 usable window for high-throughput experiments. 192

Quantitation of SARS-CoV-2 infection using luciferase-based reporters 214
To test whether our luciferase-based reporters could be activated by SARS-CoV-2 infection, 215 we transfected HEK293T-ACE2 cells with the indicated reporter constructs, then infected after 216 12 h with SARS-CoV-2. After a further 24 h incubation, reporter activation was quantitated as 217 the ratio of FFluc/Rluc luminescence. Compared with uninfected cells, we observed a marked 218 increase in FFluc luminescence in infected cells transfected with either the 30F-Opt3 or 30F-219 PLP2 reporter, but not infected cells transfected with non-cleavable controls (Figure 5A-B). 220 In the context of viral infection, the 30F-PLP2 reporter showed a better dynamic range (29-221 fold increase in luminescence) than the 30F-Opt3c reporter (9.3-fold increase in 222 luminescence). As expected based on sequence conservation (Figure 1-figure   To test this cell line in a scalable format, we seeded HEK293T-ACE2-30F-PLP2 cells in 96-260 well plates, infected them with wild-type SARS-CoV-2, then measured FFluc and Rluc 261 luminescence after 24 h. As expected, we saw no difference in Rluc signal between mock-262 infected and SARS-CoV-2-infected cells ( Figure 7A). Conversely, a dramatic increase in 263  in that context, compared with over-expression of recombinant proteases (26-fold to 3-fold for 296 Opt3c/MPro reporter, and 12-fold to 3-fold for PLP2/PLPro reporter). Contributing factors likely 297 include (but may not be limited to): lower protease expression levels plus/minus reduced 298 accessibility during viral infection; background FlipGFP fluorescence; and the absence of therefore unlikely to be used to detect viral replication in practice. 301 To overcome these limitations, we therefore generated luciferase-based reporters. We 302 envisioned at least three advantages of this approach: first, luciferase-based assays are 303 typically highly sensitive; second, since the first step in measuring luciferase activity involves 304 lysing cells, the readout is not affected by syncytia formation during viral infection; and third, 305 luciferase-based assays may be readily adapted to high-throughput platforms. Accordingly, 306 compared with FlipGFP, the window for luciferase-based reporter activation was greatly 307 increased in assays of recombinant proteases (139-fold versus 26-fold for Opt3c/MPro 308 reporters, and 74-fold versus 12-fold for PLP2/PLPro reporters) and viral infection (12-fold vs. 309

3-fold for Opt3c/MPro reporters, and 29-fold versus 3-fold for PLP2/PLPro reporters). 310
Interestingly, whilst our luciferase-based Opt3c/MPro reporter displayed a higher sensitivity 311 for recombinant protease, our PLP2/PLPro reporter performed markedly better during SARS-312 CoV-2 infection. This may reflect inversion of the expression levels of MPro and PLPro in the 313 different settings, or differential accessibility of the reporters (substrates) to viral proteases in 314 the context of viral infection. Since PLPro of SARS-CoV has deubiquitylation (DUB) activity, 315 and is able to regulate whole cell levels of ubiquitylation during viral infection, we suspect that 316 PLPro of SARS-CoV-2 may also be able to readily access a wide variety of cellular proteins, 317 including cytosolic reporters (Yan and Wu, 2021). Either way, the use of a PLPro (rather than 318 MPro) reporter was critical for the optimisation of our system. 319 By further reengineering the 30F-PLP2 reporter for lentiviral expression, we have generated 320 the first stable, luminescent reporter cell line for SARS-CoV-2 infection. This simplifies the 321 experimental workflow, and allows facile assays in a 96-well plate format. It is therefore ideally 322 suited to high-throughput screens of candidate antiviral compounds or therapeutic antibodies, 323 and/or large-scale serological surveys for neutralising activity against authentic virus. In fact, 324 whilst the mechanism of reporter activation is different, our HEK293T-ACE2-30F-PLP2 cells 325 are conceptually similar to TZM-bl reporter cells for HIV infection, and we anticipate a similar 326 format, and generating stable reporter cell lines in other models of SARS-CoV-2 infection, 328 such as human airway epithelial cells. 329 Compared with reverse-engineered fluorescent or luminescent reporter viruses, a key 330 advantage of our luminescent reporter cell line is the potential to detect a range of clinical 331 SARS-CoV-2 isolates, including emerging variants of concern. The assessment of the ability 332 of these variants to escape from natural or vaccine-induced immunity has been complicated 333 by variability in the pseudotype assays commonly used by different laboratories (Altmann et 334 al., 2021). Beyond SARS-CoV-2, the PLP2 cleavage sequence is highly conserved in SARS-335 CoV, and partially conserved in MERS-CoV. This suggests that the Papain-like proteases of 336 these viruses will also be able to activate our 30F-PLP2 biosensor, and we confirmed this in 337 the case of recombinant SARS-CoV PLPro. Our biosensors therefore offer a standardised 338 "off-the-shelf" solution for the quantitation of authentic betacoronavirus replication and 339 fetal calf serum (FCS), 100 units/ml penicillin, and 0.1 mg/ml streptomycin at 37 C in 5% CO2. 348 All cells were regularly screened and confirmed to be mycoplasma negative (Lonza MycoAlert 349 and IDEXX BioAnalytics). FlipGFP plasmid (generated as above) was used as a template for a PCR reaction with the 366 indicated primers (Supplementary file 1), then assembled with EcoRI-linearized pcDNA3.1 367 as above. 368 Luciferase-based reporters 369 oligonucleotides encoding the indicated cleavage sequences (Supplementary file 1) as 372

above. 373
To generate a lentiviral expression vector for the 30F-PLP2 reporter, the pCMV-intron-30F-374 PLP2-pA-pSV40-hRluc reporter cassette was first amplified from the pGloSensor-30F PLP2 375 plasmid using the indicated primers (Supplementary file 1), then assembled with pHRSIN- In brief, images were acquired using a Cellomics ArrayScan XTI high-throughput imaging 480 platform (Thermo Fisher) using a 386 nm excitation/emission filter to detect DAPI-stained 481 nuclei and a 560 nm excitation/emission filter to detect AF594. Images were then analysed 482 with built-in high content HCS Studio software (by Thermo Fisher) using the Target Activation 483 application. For this, cellular objects were identified by applying overlays (masks) based on 484 DAPI intensity. Necessary steps to exclude non-cellular artefacts (large or small objects) were 485 activated based on average nuclei size. Additionally, background correction was performed 486 on both channels. The generated nuclei masks were then applied to the AF594 channel and 487 the threshold for AF594 staining was determined using stained mock-infected cells. Finally, 488 cells were considered infected if their AF594 signal was above this threshold. 42 fields were 489 scanned for each sample/condition to ensure the analysis of a sufficient number of cells. 490

Confocal microscopic analysis of FlipGFP-based reporters and spike protein 491 expression in SARS-CoV-2-infected cells 492
HEK293T-ACE2 cells were seeded at a density 9 x 10 4 cells/well of an 8-well µ-Slide (Ibidi, After 24 h incubation, media was aspirated from each well, and cells lysed with 50 µL/well 523 Dual-Glo Luciferase Buffer (Promega) diluted 1:1 with PBS + 1% NP-40, for 10 mins at room 524 temperature. Lysates were then transferred to opaque half-area 96-well plates, and reporter 525 activation quantitated as the ratio of firefly luciferase (FFluc)/Renilla luciferase (Rluc) activity 526 measured using the Dual-Glo kit (Promega) according to the manufacturer's instructions. In 527 brief, FFluc activity was first measured using a ClarioStar microplate reader. 25 µL Stop and 528 Glo Buffer and Substrate (Promega) was then added to each well. After incubation for 10 mins 529 at room temperature, Rluc activity was measured using the same ClarioStar microplate reader 530 and the ratio of FFluc/Rluc activity calculated for each condition. 531

Luminescent analysis of luciferase-based reporters in SARS-CoV-2-infected cells 532
HEK293T-ACE2 cells were reverse-transfected with plasmids and TransIT-293 at a ratio of 1 533 were dissociated with Accutase, combined with the transfection mix, and seeded at 9 x 10 4 537 cells/well. 538 The following morning, cells were infected with SARS-CoV-2 at MOI=1 (or for the titration 539 experiments, with the indicated volume of viral stock) and incubated for 24 h. Media was 540 aspirated from each well, and cells lysed with 50 µL/well of Dual-Glo Luciferase Buffer 541 (Promega) diluted 1:1 with PBS + 1% NP-40 for 10 mins at room temperature. Lysates were 542 then transferred to opaque half-area 96-well plates, and reporter activation quantitated as the 543 ratio of FFluc/Rluc activity as above. Where indicated, candidate antivirals were added to the 544 cells 1 h before infection with SARS-CoV-2. 545

Luminescent analysis of SARS-CoV-2-infected HEK293T-ACE2-30F-PLP2 reporter cells 546
µL/well of Dual-Glo Luciferase Buffer (Promega) diluted 1:1 with PBS + 1% NP-40 for 10 mins 550 at room temperature. Lysates were then transferred to opaque half-area 96-well plates, and 551 reporter activation quantitated as the ratio of FFluc/Rluc activity as above. For evaluation of 552 candidate antivirals, compounds were added to the cells 1 hr before infection with SARS-CoV-553 2. To measure SARS-CoV-2 neutralising activity, SARS-CoV-2 viral stock was pre-incubated 554 with serial dilutions of heat-inactivated sera (a kind gift from Ravi Gupta) for 2 h at 37°C, prior 555 to addition to the cells. 556

SARS-CoV-2, SARS-CoV, and MERS-CoV sequence logos 557
A SARS-CoV-2 genomic alignment file was retrieved from the GISAID database 558  Having manually inspected the genomic alignments to identify the regions of interest, we used 571 the 'extractalign' function in the European Molecular Biology Open Software Suite (EMBOSS) 572 (Rice et al., 2000). In the case of the GISAID alignment, we removed entries that were (one per region of interest and four per alignment inspected) were conceptually translated 575 using the 'transeq' function in EMBOSS. The resulting amino acid sequences were used as 576 input for the WebLogo application (Crooks et al., 2004;Schneider and Stephens, 1990). 577

Statistical analysis 578
General data manipulation was conducted using Microsoft Excel, and statistical analysis using 579