key: cord-0895799-vfilvut1
authors: Chan, Shiu-Wan; Shafi, Talha; Ford, Robert C.
title: Kite-shaped molecules block SARS-CoV-2 cell entry at a post-attachment step
date: 2021-05-31
journal: bioRxiv
DOI: 10.1101/2021.05.29.446272
sha: e1d7aa1d38c0940e16dc3877b43b4becab48dffe
doc_id: 895799
cord_uid: vfilvut1

Anti-viral small molecules are currently lacking for treating coronavirus infection. The long development timescales for such drugs are a major problem, but could be shortened by repurposing existing drugs. We therefore screened a small library of FDA-approved compounds for potential severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) antivirals using a pseudovirus system that allows a sensitive read-out of infectivity. A group of structurally-related compounds, showing moderate inhibitory activity with IC50 values in the 1-5µM range, were identified. Further studies demonstrated that these ‘kite-shaped’ molecules were surprisingly specific for SARS-CoV and SARS-CoV-2 and that they acted early in the entry steps of the viral infectious cycle, but did not affect virus attachment to the cells. Moreover the compounds were able to prevent infection in both kidney- and lung-derived human cell lines. The structural homology of the hits allowed the production of a well-defined pharmacophore that was found to be highly accurate in predicting the anti-viral activity of the compounds in the screen. We discuss the prospects of repurposing these existing drugs for treating current and future coronavirus outbreaks.

of structurally-related compounds, showing moderate inhibitory activity with IC50 values in 23 the 1-5 M range, were identified. Further studies demonstrated that these 'kite-shaped' 24 molecules were surprisingly specific for SARS-CoV and SARS-CoV-2 and that they acted 25 early in the entry steps of the viral infectious cycle, but did not affect virus attachment to the 26 cells. Moreover the compounds were able to prevent infection in both kidney-and lung-27 derived human cell lines. The structural homology of the hits allowed the production of a 28 well-defined pharmacophore that was found to be highly accurate in predicting the anti-viral 29 activity of the compounds in the screen. We discuss the prospects of repurposing these 30 existing drugs for treating current and future coronavirus outbreaks. long term that could require new vaccine development. This is particularly problematic with 43 RNA viruses with high mutation rates, especially for vaccines that target the spike protein that 44 is on the outside of the virion and thus subject to continuous selective pressure (2). For example 45 the vaccine strain initially observed all over the world was replaced by the D614G spike variant 46 in February, 2020 (3 months after the pandemic was announced) and other new variants are 47 sweeping through the world currently (3-5) (6). Hence it seems appropriate to look for further 48 control measures for the virus (7). Missing from the arsenal of effective measures to date has 49 been an effective anti-viral therapy that could be administered prior to, or after acquiring the 50 virus. So far, drug therapy has been limited to attempts to reduce the most life-threatening 51 symptoms of the infection that arise due to over-stimulation of the immune response (8). We 52 therefore lack a first line anti-viral defence to add to the current toolkit (9). Such first-line 53 treatments would allow more time to develop new vaccines, could improve therapeutics and 54 might be employed as prophylactics in those who cannot be vaccinated or do not respond well 55 to vaccine. Such drugs, would ideally target conserved steps in the viral life cycle, be broad-56 spectrum and therefore generally applicable to COVID variants of the present and future (9).

In order to re-purpose drugs for fast-tracking COVID-19 prophylaxis and treatments, we 107 undertook screening of two libraries of FDA-approved drugs and natural products from 108 APExBIO (28) by using MLV pseudotyped with the SARS-CoV-2 S protein (29), with the 109 goal of targeting the major attachment and entry steps (Fig.1) . The spike protein is derived 110 from the Wuhan-Hu-1 SARS-CoV-2 and has been codon optimized for mammalian 111 expression (12). To find a suitable human cell type for the screening of drugs inhibiting viral and SARS-CoV-2-S pseudovirus infectivity, but not to the same high level as 293T cells, 130 which were therefore employed for initial drug screening. Compared to 293T cells not 131 overexpressing ACE2, infectivity of SARS-CoV-2-S pseudovirus in 293T-ACE2 cells was 132 about 100x higher (Fig.3a,d) . Quantitation by Western blot estimated the number of ACE2 133 receptors to be at least ten times higher in 293T-ACE2 cells than in the untransfected 293T 134 cells with no loss of ACE2 expression in late passaged cells (P16) compared to early 135 passaged cells (P4) (Fig.3b,d) . The number of trimeric spike proteins present on the surface 136 of the pseudovirus in each infection experiment was also estimated (Fig. 3c,d) . The data 137 imply that the 293T-ACE2 cells' ACE2 receptors will outnumber spike protein in the 138 pseudovirus infection experiments (Fig.3b, c, d) , which is likely to be representative of the in-139 vivo situation, especially at early stages of infection. Furthermore the quantitation confirmed 140 that the concentrations of proteins in the assays described below were well below the drug 141 concentrations used. This was important to allow for the possibility of full inhibition by any 142 given drug that was working by blocking the ACE2-Spike interaction and hence virus 143 attachment to the cell. 146 An initial screen of a library of 1363 FDA-approved drugs was carried out using the above-147 mentioned cell line. The drugs at 10M were incubated with the cells after dilution of the 148 drugs into cell growth media from (predominantly) DMSO-solubilised stock solutions or 149 (occasionally) ethanol-based or water-based stocks. Any cytotoxicity effects of the drugs at 150 this concentration were controlled for using an XTT cell viability assay. After screening one-151 third of the compounds it became apparent that there was a prevalence of inhibitory activity 152 found within a class of molecules that displayed a similar structure and a shape reminiscent 153 of a traditional Chinese Kite. These had a well-conserved tri-cyclic core structure (forming 154 the sail of the kite) and a more variable extension from the central 6-or 7-membered ring 155 (forming the tail of the kite). We, therefore, selected 61 kite-shaped molecules from the two 156 libraries. Five that were cytotoxic were excluded at this stage; the remaining molecules 157 showed a range of activity against pseudovrius infectivity. Supplementary Table S1 158 summarises the experimental data for the kite-shaped molecules. share common MLV post-entry steps but differ in receptor recognition and entry mechanisms 165 (18,33). We also included the 14 th ranked hit, trimipramine, because it had previously been 166 identified as a specific SARS-CoV entry blocker and an inhibitor of SARS-CoV-2 infection 167 (26). All the selected kite-shaped molecules showed effects similar to those of 168 hydroxychloroquine, a known blocker of SARS-CoV-2 entry, by specifically inhibiting 169 infectivity of the SARS-CoV-2-S, SARS-CoV-S, MERS-CoV-S but not VSV-G pseudotyped 170 viruses, suggesting that the nine kite-shaped molecules target entry steps specific to these 171 three coronaviruses (Fig.4a) . The kite-shaped molecules inhibited SARS-CoV-S and SARS-

CoV-2-S pseudoviruses equally well and the inhibition was 1.2 to 8-fold higher than that for 173 MERS-CoV-S pseudovirus, suggesting that they may, in addition, discriminate between 174 different receptor-mediated pathways for viral entry. Although low levels of inhibition of 175 VSV-G pseudovirus infectivity were indicated for a few of the nine compounds, this may be 176 due to some cytotoxicity at 10M, as suggested by the correlation between % inhibition and 177 % viability (Fig.4b) . Alternatively, it could be due to inhibition of common post-attachment 178 pathways in late endosome/lysosome. Hydroxychloroquine, a lysosomotropic agent, did not 179 inhibit VSV-G pseudovirus infection, where fusion takes place at the early endosome stage 180 (17). In contrast, the reverse transcriptase inhibitor, tenofovir disoproxil fumarate, 181 completely inhibited infection of all pseudoviruses. These results imply that the kite-shaped 182 molecules target the SARS-CoV-2 entry steps specifically, rather than any post-entry steps 183 mediated by the MLV or the reporter.

Although the A549 cells transfected with ACE2 showed less propensity for infectivity than 185 the HEK293T cell line (hence noisier luciferase readouts), a test of eight of the above- To deduce the efficacy of the kite-shaped molecules, we generated dose-response curves 198 using a range of drug concentrations from 10M to 0.05M and using hydroxychloroquine 199 for comparison (Fig.5a) . The kite-shaped molecules displayed IC50 values from 1.9M to 200 4.7M, compared to 0.7M for hydroxychloroquine (Fig.5b) . All the drugs showed de 201 minimis cytotoxicity at 5M and even at the highest drug concentration employed, the cells 202 generally displayed a viability above 76% (chlorprothixene and chlorpromazine are 203 exceptions with cell viability at 54% and 68%, respectively (Fig.4b, Fig. 5 ). Since we diluted 204 the drugs in cell growth medium, a confounding effect on the determination of IC50 could be 205 water solubility of the drugs. Most of the drugs are readily water soluble in their charged 206 state (pizotifen malate is the exception), but they may partition into the cell membrane via 207 their uncharged forms which will have very low water solubility (Supplementary Table S2 ).

The similarity in the overall structure of the kite-shaped molecules allowed the generation of 209 a pharmacophore ( Figure 6 ). Pharmacophores for tricyclic antidepressants (TCAs) have To study the mechanisms of inhibition in greater detail, we undertook a time-of-addition 223 experiment in which drugs were added at different time-points during infection ( Fig.7a) with 224 the hypothesis that the time-point(s) may distinguish different step(s) that may be inhibited 225 by the drugs. As before, we studied the top nine hits together with trimipramine. The kite-226 shaped molecules were able to inhibit infection when added during the entry step (with and 227 without 1h pre-incubation), to similar extents to the full-treatment (Fig.7b) . Most of the kite-228 shaped molecules did not inhibit infectivity when only added 1 and 2 hours post-infection 229 (hpi). This observation was similar for the entry blocker, hydroxychloroquine, suggesting 230 that the kite-shaped molecules are also entry blockers. The exceptions were chlorprothixene 231 and chlorpromazine, which were still able to reduce infectivity to 45% and 35%, respectively, 232 when added at 1hpi. However, these reductions in infectivity were still much lower than the 233 complete inhibition of infectivity observed for the reverse transcriptase inhibitor, tenofovir.

Altogether, these results suggest that the kite-shaped molecules inhibit a SARS-CoV-2-235 specific entry step.

Kite-shaped molecules inhibit a post-attachment step 238 To further delineate the entry step that is inhibited by the kite-shaped molecules, we 239 undertook temperature shift experiments to distinguish between attachment and post-240 attachment steps that were assumed to proceed (virus attachment) -or not proceed (virus 241 entry, post attachment) -at the low temperature (4 o C) employed in the first experiment 242 (Fig.8a ). With this assay, hydroxychloroquine did not inhibit attachment, in agreement with 243 its main role as a post-attachment entry blocker (Fig.8b) . Most of the kite-shaped molecules 244 reduced infectivity to 48-71% at the attachment step apart from chlorprothixene which did 245 not inhibit virus attachment. Thioridazine reduced infectivity to 28% in the attachment assay.

These data suggest that all the kite-shaped molecules may reduce attachment to some extent, 247 although it should be acknowledged that this interpretation of the data depends on complete 248 removal of the added drugs at the wash step, which may be dependent on water solubility at Chlorprothixene, one of the top hits in this study, was identified in a repurposing study 273 screening 8,810 drugs that were either FDA-approved or investigational (36).

Methotrimeprazine and piperacetazine also emerged as top hits from that study, and these two 275 compounds share the basic kite-shaped structure of the TCAs. Evidence from observations of 276 patient populations has suggested there was a lower incidence of symptomatic and severe 277 SARS-CoV-2 problems in psychiatric patients (37), and this report was followed up with an CoV-2 infectivity in the lungs: ACE2 is highly expressed in airway cells but SLC6a19 is not.

Airway-expressed homologs of SLC6a19, such as SLC6a14, SLC6a15 and SLC6a20 may be 332 considered as possible replacements for SLC6a19 in the lungs. However for these proteins 333 there is currently no evidence for any direct interaction with ACE2. One of the candidates, 334 SLC6a15, has been reported to interact directly with several SARS-Cov-2 proteins including 335 the M membrane glycoprotein (57,58), and like ACE2, it appears to carry a C-terminal PDZ-336 binding motif which therefore offers a potential route for an interaction bridged by an 337 unknown PDZ protein. If SLC6a19 is being replaced by a homolog in the airways, then it 338 could be argued that the most likely candidate is SLC6a15.

The above hypothesis is supported by our time-of-addition and temperature shift experiments, CoV-S pseudoviruses, suggests that they target a pathway shared by the three viruses but 344 may, in addition, discriminate between different receptor-mediated pathways for viral entry.

Although the kite-shaped molecules generally share a common mode of action, they may also 346 possess unique mechanisms of inhibition. Whereas most of the kite-shaped molecules 347 displayed a similar inhibitory pattern of infectivity, chlorprothixene and chlorpromazine 348 showed some discrepancies. Both inhibited the VSV-G pseudovirus to a greater extent than 349 the other drug hits. Both showed some degree of inhibition when added at 1hpi and 2hpi.

Whereas all the other kite-shaped molecules displayed some degree of inhibition of 351 attachment, chlorprothixene did not inhibit attachment. Although some of these 352 discrepancies could be accounted for by the relative toxicity of these two drugs, off-target 353 effects and water insolubility of chlorprothixene, we cannot exclude the possibility that they 354 are results of possession of unique targets. Chlorpromazine has been known to inhibit 355 clathrin-mediated endocytosis, which is utilized by the VSV to enter cells, so it is not 356 surprising that it will inhibit VSV-G pseudovirus infection to some extent (59) (60). propose whether our kite-shaped drug hits are direct-acting antivirals and/or host-targeting 366 agents. Further work will be required to identify the common and unique modes of action of 367 our drug hits in order to facilitate the formulation of a drug cocktail.

Although it is well recognized that SARS-CoV-2 infects lungs, gut and eyes, increasing 369 evidence suggest liver and kidney tropism with kidney predicted to be the most susceptible Endosomal fusion is the major entry pathway for SARS-CoV-2 in ACE2-deficient cells (20).

It is, therefore, of paramount importance for an anti-viral regime to be able to target both In conclusion, our study has generated a class of kite-shaped molecules that target a 386 potentially conserved post-attachment step of SARS-CoV-2 cell entry which could inform 387 clinical trials in the current crisis. We have also created a pharmocophore that will allow for 388 improvement in drug design as a broad-spectrum antiviral for future pandemics. There have with only drugs without pseudovirus, in duplicate, to test for drug cytotoxicity. After 430 incubation for 37 o C, 5%CO2 for 48h, they were tested for luciferase activity for % infectivity 431 relative to the infected, solvent controls and for % viability relative to the solvent controls.

For the generation of concentration curves, serial dilutions of drugs were titred, in duplicates 433 and IC50 was calculated using Prism9 (GraphPad). Cells seeded on 96-well plates were pre-cooled on ice for 1h (see Fig.8a ). For the attachment 460 assay, drugs were diluted into pre-cooled viruses on ice to 10M just before infection and the 461 100l virus-drug mix was added to each well for 1h. After 1h, the wells were washed 3x The protein bands were quantified against a standard curve of recombinant spike protein.

The spike protein in the SARS-CoV-2-S pseudovirus has been cleaved to yield the S1 Post-entry assay involved no drug pre-treatment and infection in the absence of drug.

Following virus wash-off, drug was added at 1 hour post-infection (hpi) or 2hpi for the rest of 897 48h. (b) Infectivity was measured as luciferase activity and expressed as % infectivity to 898 infected, own solvent control (dimethylsulphoxide, ethanol or water) at the same time-point.

Data are presented as mean +/-SD of two repeats. The protein bands were quantified against a standard curve of recombinant spike protein.

The spike protein in the SARS-CoV-2-S pseudovirus has been cleaved to yield the S1 952 subunit. The inset shows the same blot at higher contrast for clarity. Low exposure blot was 

Current and Future Direct-Acting Antivirals Against COVID-19. Front 562 Microbiol 11

Angiotensin-565 converting enzyme 2 is a functional receptor for the SARS coronavirus

Dipeptidyl peptidase 4 is a functional receptor for 569 the emerging human coronavirus-EMC

Pharmacophore 624 modeling for hERG channel facilitation

Drug Repurposing Screen for 628 Compounds Inhibiting the Cytopathic Effect of SARS-CoV-2

Repurposing chlorpromazine to treat COVID-19: The 631 reCoVery study

634 (2021) Inhibition of the replication of SARS-CoV-2 in human cells by the FDA-approved drug 635 chlorpromazine

Several FDA-Approved Drugs Effectively Inhibit 638 SARS-CoV-2 Infection in vitro

Pharmacological Inhibition of Acid Sphingomyelinase Prevents Uptake of SARS-CoV-2 by 644 Epithelial Cells

Identification of Potent and Safe Antiviral Therapeutic 647 Candidates Against SARS-CoV-2

Tricyclic Antidepressants. in StatPearls, Treasure 649 Island (FL)

Metrics 651 other than potency reveal systematic variation in responses to cancer drugs

Pharmacokinetics of 654 chlorpromazine and key metabolites

Hydroxychloroquine relative bioavailability: within 656 subject reproducibility

X-ray structure of dopamine transporter 658 elucidates antidepressant mechanism

Structure 661 of the human histamine H1 receptor complex with doxepin

X-ray structures and mechanism of the 663 human serotonin transporter

Allosteric binding site in a Cys-loop 666 receptor ligand-binding domain unveiled in the crystal structure of ELIC in complex with 667 chlorpromazine

Crystal structure-based discovery of a novel synthesized PARP1 inhibitor (OL-1) with 670 apoptosis-inducing mechanisms in triple-negative

Ebola Virus Glycoprotein Complexes with Tricyclic Antidepressant and Antipsychotic Drugs

Ebola virus and severe acute respiratory syndrome 676 coronavirus display late cell entry kinetics: evidence that transport to NPC1+ endolysosomes 677 is a rate-defining step

Identification of novel inhibitors of the amino acid transporter B(0) AT1 (SLC6A19), a 680 potential target to induce protein restriction and to treat type 2 diabetes

Mutations in SLC6A19, encoding B0AT1, cause Hartnup disorder

Structural basis for action by diverse antidepressants on biogenic amine transporters

Structural basis for the 691 recognition of SARS-CoV-2 by full-length human ACE2

Comparative host-coronavirus protein interaction 720 networks reveal pan-viral disease mechanisms

725 (2021) Multilevel proteomics reveals host perturbations by SARS-CoV-2 and SARS-CoV. 726

The use of inhibitors to study endocytic 729 pathways of gene carriers: optimization and pitfalls

Role of clathrin-mediated 731 endocytosis during vesicular stomatitis virus entry into host cells

Identification of novel macropinocytosis 734 inhibitors using a rational screen of Food and Drug Administration-approved drugs

A Single-Cell RNA Expression Map of Human 737 Coronavirus Entry Factors

739 (2020) Expression of SARS-CoV-2 receptor ACE2 and TMPRSS2 in human primary 740 conjunctival and pterygium cell lines and in mouse cornea

Enhanced isolation of SARS-CoV-2 by TMPRSS2-744 expressing cells

2020) Data, Reagents, Assays and Merits of 747 Proteomics for SARS-CoV-2 Research and Testing

Influenza Virus

Severe acute respiratory syndrome 750 (SARS): breath-taking progress

The 2014 Ebola virus disease outbreak in West Africa

Zika virus

Outbreak of avian influenza A(H5N1) virus infection in Hong Kong

An Overview of Current Knowledge 757 of Deadly CoVs and Their Interface with Innate Immunity

759 (2021) Novel Canine Coronavirus Isolated from a Hospitalized Pneumonia Patient, East 760 Malaysia

Hepatitis C virus envelope proteins regulate CHOP via 762 induction of the unfolded protein response

Effects of hepatitis C virus envelope glycoprotein 764 unfolded protein response activation on translation and transcription

Hepatitis C Virus Envelope Protein E1 767 Binds PERK and Represses the Unfolded Protein Response

769 (2006) Cap-dependent and hepatitis C virus internal ribosome entry site-mediated 770 translation are modulated by phosphorylation of eIF2 alpha under oxidative stress

Hydrogen peroxide induces La cytoplasmic shuttling and increases 773 hepatitis C virus internal ribosome entry site-dependent translation

Zika Virus Induces an Atypical Tripartite Unfolded 776 Protein Response with Sustained Sensor and Transient Effector Activation and a Blunted BiP 777 Response mSphere 6 778 79

AutoPH4: An Automated 780 Method for Generating Pharmacophore Models from Protein Binding Pockets

Development and 783 validation of a genetic algorithm for flexible docking11Edited by

PubChem in 2021: new data 787 content and improved web interfaces

Structural basis for action by diverse antidepressants on biogenic amine transporters

Structural basis for the 792 recognition of SARS-CoV-2 by full-length human ACE2

Intact Viral Particle Counts 794 Measured by Flow Virometry Provide Insight into the Infectivity and Genome Packaging 795 Efficiency of Moloney Murine Leukemia Virus

We sincerely thank François-Loïc Cosset

Cornell) for the MLV pseudovirus system

LLC) for the Huh-7 cell line. We also thank colleagues for donating reagents

The following reagents were obtained through BEI 802

NIH: human embryonic kidney Cells (HEK-293T) expressing human 803 angiotensin-converting enzyme 2, HEK-293T-hACE2 cell line, NR-52511

NR-53522; polyclonal anti-SARS-related coronavirus 2 spike glycoprotein (IgG, Rabbit)

NR-52947; spike glycoprotein (stabilized) from SARS-related coronavirus 2

The following reagent was contributed by David Veesler for distribution through BEI 809

810 time-of-addition experiment. (a) Schematic of time-of-addition experiment. Full-time 1000 treatment involved 1h drug pre-treatment and 1h infection in the presence of drug followed 1001 by drug and virus wash-off and addition of fresh drug for the rest of 48h. Entry assay 1002 involved 1h infection in the presence of drug with and without 1h drug pre-treatment

Post-entry assay involved no drug pre-treatment and infection in the absence of drug

Following virus wash-off, drug was added at 1 hour post-infection (hpi) or 2hpi for the rest of 1006 48h. (b) Infectivity was measured as luciferase activity and expressed as % infectivity to 1007 infected

Data are presented as mean +/-SD of two repeats. own solvent control (dimethylsulphoxide, ethanol or water). Data are presented as mean +/-1024 SD of four repeats for attachment assays and three repeats for penetration assays

Infectivity was measured as luciferase activity and expressed as % infectivity to infected,