key: cord-0299496-voaz1li4
authors: Shi, Yunlong; Zeida, Ari; Edwards, Caitlin E.; Mallory, Michael L.; Sastre, Santiago; Machado, Matías R.; Pickles, Raymond J.; Fu, Ling; Liu, Keke; Yang, Jing; Baric, Ralph S.; Boucher, Richard C.; Radi, Rafael; Carroll, Kate S.
title: Thiol-based mucolytics exhibit antiviral activity against SARS-CoV-2 through allosteric disulfide disruption in the spike glycoprotein
date: 2021-07-01
journal: bioRxiv
DOI: 10.1101/2021.07.01.450701
sha: 1b83d60f12227d1000933e05abd689d70a38dd6d
doc_id: 299496
cord_uid: voaz1li4

Small molecule therapeutics targeting severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) have lagged far behind the development of vaccines in the fight to control the COVID-19 pandemic. Here, we show that thiol-based mucolytic agents, P2119 and P2165, potently inhibit infection by human coronaviruses, including SARS-CoV-2, and decrease the binding of spike glycoprotein to its receptor, angiotensin-converting enzyme 2 (ACE2). Proteomics and reactive cysteine profiling link the antiviral activity of repurposed mucolytic agents to the reduction of key disulfides, specifically, by disruption of the Cys379–Cys432 and Cys391–Cys525 pairs distal to the receptor binding motif (RBM) in the receptor binding domain (RBD) of the spike glycoprotein. Computational analyses provide insight into conformation changes that occur when these disulfides break or form, consistent with an allosteric role, and indicate that P2119/P2165 target a conserved hydrophobic binding pocket in the RBD with the benzyl thiol warhead pointed directly towards Cys432. These collective findings establish the vulnerability of human coronaviruses to repurposed thiol-based mucolytics and lay the groundwork for developing these compounds as a potential treatment, preventative and/or adjuvant against infection.

Control of pandemics requires rapid and sensitive testing, widely available vaccines, and effective 26 therapeutic agents for the infected. For the current SARS-CoV-2 pandemic, impressive progress 27 in the development and testing of vaccines has been achieved. However, the development of 28 effective SARS-CoV-2 antiviral therapeutics has lagged. Multiple strategies have been employed 29 to identify effective anti-SARS-CoV-2 therapeutic agents. Classic antiviral strategies that focused 30 on viral replication and propagation, e.g., protease and RNA polymerase inhibitors, have received 31 much attention but have produced limited clinical efficacy to date. 1-5 Many therapeutic strategies 32 have focused on the intricate SARS-CoV-2 cellular entry processes that include interactions of 33 the spike glycoprotein's receptor-binding domain (RBD) with the angiotensin-converting enzyme 34 2 (ACE2), cell surface proteases, and complex fusion events. 6, 7 The development of antiviral 35 agents that target the receptor binding and entry processes requires precise knowledge of the 36 dynamic structures of the viral elements that mediate the binding and entry of virus into target 37 cells. 8,9 38 39 Disulfide bond formation is central to the dynamic structure of many viral receptor binding and 40 entry/fusion proteins. 10 The role of disulfide bonds in cognate receptor binding proteins, i.e., spike 41 proteins, has been widely studied in coronaviruses, including mouse hepatitis virus (MHV), CoV, and SARS-CoV-2. 11-13 Viral disulfides are initially formed in the endoplasmic reticulum (ER). 43

These bonds, important for both virus binding and fusion, are further stabilized by the oxidizing 44 extracellular milieu. 10,13 The SARS-CoV-2 RBD contains a four disulfide pairs: 1) Cys480-Cys488 45 situated at the ACE2 binding surface, and 2) Cys336-Cys361, Cys379-Cys432, Cys391-Cys525 46 to stabilize the β sheet structure. 8 The position of these disulfides in RBD crystal structures, has 47 led to speculation that reduction of these bonds may have untapped therapeutic utility. 14-18 3 However, the precise function of these disulfide pairs cannot be read from structure alone, and 49 cysteine reactivity mapping has not been performed to investigate this hypothesis. 50 51 Thiol-based mucolytic agents, often administrated orally or as aerosols, are currently in use as 52 therapeutics for pulmonary diseases, e.g., cystic fibrosis. [19] [20] [21] Of these agents, preclinical 53 compounds P2119 and P2165 (Figure 1a) , originally identified from a library of glycosylated thiols 54 and advanced as mucolytics, 22,23 are unique as they are restricted to the extracellular space, i.e., 55 not membrane permeant, P2119 and P2165 also exhibit a greater therapeutic index compared to 56 clinically available therapeutics, like N-acetylcysteine (NAC) 23 which is taken up by the cell and 57 metabolized to hydrogen sulfide (H2S) 24 a second messenger that initiates pleiotropic changes in 58 myriad targets. 25 P2119 and P2165 also afford an opportunity to compare the intrinsic potency of 59 a monothiol versus a dithiol compound, and their associated mechanisms of action ( Figure 1b) . 60 61 Here, we show that thiol-based mucolytic agents, P2119 and P2165, potently inhibit infection by 62 human coronaviruses, including SARS-CoV-2, and decrease the binding of spike glycoprotein to 63 its receptor, ACE2. The antiviral activity of repurposed mucolytic agents is linked to the reduction 64 of disulfides in the RBD of spike glycoprotein. Proteomics and reactive cysteine mapping shows 65 that the Cys480-Cys488 pair, located in the loop region of the ACE2 binding surface, is not 66 susceptible to alkylation during live cell infection, establishing the stability of this disulfide in a 67 native setting. By contrast, Cys432 and Cys525, which form disulfides with C379 and Cys391, 68 respectively, are identified as hyper-reactive cysteines that form semi-stable disulfides. Molecular 69 docking analysis provides insight into the targeting of Cys432 and Cys525 by mucolytics. These 70 simulations predict that reduction of these three disulfides control ACE2 binding by triggering 71 conformational changes in the RBD. The latter finding suggests that Cys379-Cys432 and 72 Thiol-based mucolytic agents have potent antiviral activity against human coronaviruses. 90

The antiviral activities of P2119 and P2165 were first evaluated in a recombinant infectious clone 91 of SARS-CoV-2 virus produced in human nasal epithelial (HNE) cells (Figure 1c ). 26 Exposure of 92 SARS-CoV-2 to monothiol P2119 or dithiol P2165 mucolytics at 30 mM (note: this concentration 93 was chosen based on mucolytic activity and effective surface airway deposition in mouse lungs 22 ) 94 resulted in respective >3 log and >6 log reductions in viral titer. To contextualize these findings, 95

we compared dose-dependent virucidal activity of NAC, the gold-standard for approved care in 96 thiol-based molecule mucolytics, and P2165 against SARS-CoV-2 ( Figure 1d ). Dose-dependent 97 inhibition by P2165 was noted at concentrations over 3 mM (1-3 logs), while NAC was ineffective 98 as a virucidal agent at concentrations up to 100 mM. Analogous virucidal activities were observed 99 for the closely related coronavirus SARS-CoV (Figure 1e ), which shares ~75% identity with 100 SARS-CoV-2 spike protein and three of four conserved disulfides in the RBD. 27 Coronavirus 101 NL63-CoV 28 also utilizes ACE2 as its receptor 29 and has RBD disulfides that are essential for 102 protein stability and/or receptor binding. 30 Exposure of NL63-CoV to P2119 (10 or 30 mM) or 103 P2165 (10 mM) and infection of LLC-MK2 cells showed similar reduction of viral titer (>3-4 logs; 104 Figure S1a ). Finally, since glycosylation of the spike protein likely varies with cell type 31 the activity 105 of thiol-based mucolytics was also tested with SARS-CoV-2 propagated in Vero E6 cells ( Figure  106 S1b). Here, P2119 reduced SARS-CoV-2 titers by ~1 and 3 logs at 10 and 30 mM, respectively. 107 P2165 reduced virus titers by ~2 and 4 logs at 10 and 30 mM, respectively. These collective data 108 indicate that P2119 and P2165 have potent antiviral activity irrespective of the cell type producing 109 SARS-CoV-2, that the dithiol is a more effective virucidal agent compared to the monothiol, and 110 hint at a common mechanism of action for these thiol-based mucolytics. Thiol-based mucolytics inhibit SARS-CoV-2 spike binding to human ACE2 receptor. P2119 126 and P2165 are p-methoxybenzyl thiols conjugated to glucose (P2119) or mannose (P2165) 7 monomers. The sugar units impart hydrophilicity and block diffusion of mucolytic compounds into 128 the cell where viral enzymes, e.g., proteases and mRNA polymerase, hijack host machinery. 129

Since P2119 and P2165 are membrane impermeant and reduction of titer can be observed when 130 the virus is exposed to these compounds prior to cell infection, we hypothesized that P2119 and 131 P2165 act by inhibiting viral entry. To examine this possibility, we assessed in vitro binding 132 between recombinant spike receptor binding domain (RBD) from SARS-CoV-2 and immobilized 133 human ACE2. In this workflow, spike RBD was exposed to thiol-based mucolytics, subjected to 134 spin gel filtration for small-molecule removal, and tested for ACE2 receptor binding. Mapping mucolytic-sensitive disulfides in human and SARS-CoV-2 cysteinomes in native 169

virus. Based on the promising antiviral studies above, we next sought to identify disulfides that 170 exhibit sensitivity to thiol-based mucolytics in extracellular human and SARS-CoV-2 proteomes. 171

Supernatants from cultured HNE cells infected with SARS-CoV-2 were exposed to P2119 or 172 P2165, subjected to differential alkylation redox proteomics, and quenched by urea (Figure 3a  173 and Table 1 , Supporting Information). Cysteine-containing peptides that undergo alkylation by 9 iodoacetamide (IAM) after reduction with P2119 or P2165 represent targets of these thiol-based 175 mucolytics ("mucolytic-sensitive Cys"). In addition, monothiol compounds, like P2119, can form 176 mixed disulfides with cysteines in protein targets ("mucolytic-modified Cys"). whereas dithiols, like 177 P2165, are designed to undergo rapid thiol-disulfide exchange to minimize stable covalent 178 modification by the mucolytic agent. CoV-2 peptides identified as "mucolytic-sensitive", the majority were localized to spike protein, 191

i.e., 10 out of 15 (67%) peptides after P2165 treatment, 13 out of 19 (68%) peptides after P2119 192 treatment, and 9 out of 12 (75%) P2119-modified peptides. RBD Cys361, Cys391, Cys432 and 193

Cys525 were susceptible to covalent modification by P2119 (Figure 3d Mapping mucolytic-sensitive disulfides and reactive cysteines in recombinant SARS-CoV-212 2 RBD. The goal of subsequent proteomics was to identify "mucolytic-sensitive" disulfides and 213 "hyper-reactive" cysteines in recombinant SARS-CoV-2 RBD (Figure 4a ,b and Table 1 , 214

Supporting Information) and cross-reference these data to that obtained with SARS-CoV-2. Mass 215 spectrometry (MS) analysis of vehicle-treated RBD identified peptides containing Cys342, 216 Cys361, Cys391, and Cys525, in contrast, while exposure to P2165, P2119 or TCEP control 217 identified peptides containing seven of eight RBD cysteines (Figure 4c) i.e., the Cys336 peptide 218 11 was not found. These mapping data and existing structures therefore suggest that Cys480-219

Cys488 forms a stable disulfide in recombinant RBD expressed and purified from human HEK293 220 cells. Notably, this disulfide pair is susceptible to reduction in truncated protein, in contrast to full-221 length spike harvested from infected cells. The remaining three disulfides Cys336-Cys361, 222

Cys379-Cys432, Cys391-Cys525 exist as an ensemble of free and disulfide-bonded states. 223

To gain further insight into cysteines reactivity in recombinant RBD, we applied a reactivity-based 225 approach using two different chemical probes for alkylation (Figure 4d ). In these studies, RBD 226 was first alkylated with a very low concentration of iodo-N-(prop-2-yn-1-yl)acetamide (IPM; 0.1 227 mM) to detect low pKa or "hyper-reactive" cysteines. 33 Spin gel filtration was used to remove 228 excess IPM and then peptides were incubated with dithiothreitol (DTT). After reduction, nascent 229 thiols were alkylated with a high concentration of IAM (4 mM). MS analysis of peptides derived 230 from these samples revealed that only Cys379, Cys432 and Cys525 were alkylated with IPM. 231

Peptides containing these cysteines also exhibited more intense signals after DTT reduction-IAM 232 alkylation, consistent with powerful thiolate nucleophilicity at these three sites (Figure 4d ). By 233 contrast, Cys361, Cys391, Cys480 and Cys488 were only alkylated after reduction with DTT. The 234 reactivities of Cys379, Cys432, and Cys525 were further differentiated using an acrylamide-235 derivatized electrophile, which is less reactive than IAM and IPM. This compound underwent 236

Michael addition with Cys379 and Cys525, but not Cys432 (Figure 4d directly involved in RBD-ACE2 interaction 8,35,36 and these residues have been parsed into three 256 distinct "contact regions" (CR1-3; Figure 5a ). 37 To dissect the relative contributions of 257 experimentally observed disulfides to dynamics in CR1-3, we conducted MD simulation of the 258 RBD in different redox states. The difference in eigenvector centrality 37 between native and 259 reduced systems on a per residue basis is depicted in Figure 5b . This analysis facilitates the 260 identification of residues and regions that present significant changes from the mean dynamic 261 13 behavior in the native state post-disulfide reduction. The resulting data indicate that reduction of 262 the Cys379-Cys432 disulfide bond leads to important dynamic changes in several residues of 263 the interaction region, particularly in CR1. Reduction of the Cys391-Cys525 disulfide was 264 associated with fewer changes overall, but, some residues at CR2 and CR3 were also perturbed. 265

These data were further supported by clustering trajectories using a structure similarity criterion 266 (Figure 5c,d) . The RBM region in native RBD was able to access conformations that differed 267 significantly from a reference structure of this domain in complex with ACE2. Reduction of the 268 Cys379-Cys432 disulfide resulted in structure ensembles that differed greatly from the reference 269 structure. Even though reduction of Cys391-Cys525 generated structure ensembles comparable 270 to native RBD in terms of a gross structural descriptor, the observed conformations were quite 271 different. A greater tendency to "shrink" the RBM region approaching CR1 and CR3 was 272 observed, which may have a significant impact on ACE2 recognition; a similar trend was observed 273 for Cys480-Cys488. Although this disulfide pair is located at the RBM, its reduction yielded 274 structure ensembles resembling not only native RBD, but also a subset exploring quite different 275 RBM conformations ( Figure S3 ). In summary, MD simulations suggest that both Cys379-Cys432 276 and Cys391-Cys525 disulfides, experimentally confirmed by MS as targeted by mucolytic agents 277 P2119 and P2165, play a major role in RBD dynamics. From a mechanistic standpoint, reduction 278 of Cys379-Cys432 and Cys391-Cys525 disulfide pairs appears to impact RBD binding to ACE2, 279 despite being structurally distant from the interaction region. 5c,d). Small differences in benzyl orientations were observed, but close distances were 322 consistently maintained between corresponding sulfur atoms ( Although several thiol-containing compounds have been shown to inhibit viral receptor binding in 330 vitro, 44,45 they lack potency e.g., NAC, glutathione (GSH), or are cytotoxic e.g., DTT, TCEP. In this 331 study, we have repurposed mucolytic thiol-based reductants P2119 and P2165 as antiviral agents 332 for human coronaviruses, including SARS-CoV-2. The virucidal effects of these compounds can 333 be attributed to the intrinsic reducing ability of the thiol group, independent of the ability to form 334 stable covalent linkages. Based on their location of action i.e., extracellular, the ability of P2119 335 and P2165 to inhibit RBD-ACE2 binding in vitro, and proteomic cysteine site-reactivity mapping 336 in SARS-CoV-2, the simplest explanation for the potent antiviral activity of these mucolytics is that 337 they efficiently reduce key disulfides in the RBD of the SARS-CoV-2 spike protein. The 338 observation that most mapped cysteine modifications occurred on the spike protein supports our 339 proposed mechanism of action for P2119 and P2165. Of particular importance is the reduction of 340

Cys379-Cys432 and Cys391-Cys525, which were identified as "mucolytic-sensitive" in both 341 native spike protein and recombinant RBD. Despite the susceptibility of the Cys480-Cys488 pair 342 to reduction in recombinant RBD, this disulfide was not "in-play" in native virus during infection. 343

In essence, the Cys480-Cys488 disulfide is a conserved, thermodynamically stable disulfide, 344 while Cys379-Cys432 and Cys391-Cys525 disulfides are in dynamic equilibrium with their thiol 345 states and thus, are more sensitive to changes in redox poise. 346 347 MD simulations carried out in this study indicate significant conformational changes in the RBM 348 concomitant with reduction of Cys379-Cys432 and Cys391-Cys525 disulfides, even though they 349 are distal to the ACE2 recognition motif. These RBD disulfides appear to be allosteric, a class of 350 disulfides distinct from structural or catalytic disulfides that control protein function by triggering 351 conformational change when they break and/or form. 46 for SARS-CoV-2 to enter host cells via the ACE2 receptor and may be the cause behind the age-379 dependent severity of COVID-19. From this perspective, thiol-based mucolytics P2119 and P2165 380 may derive antiviral activity from changes in extracellular redox poise in two ways: 1) direct 381 reduction of disulfides on redox-sensitive target proteins required for viral entry e.g, RBD Cys379-382

Cys432 and Cys391-Cys525, and 2) indirectly by restoration of intracellular glutathione levels. 383

Finally, thiol-based mucolytics manage mucus hypersecretion via disruption of intermolecular 384 mucin disulfides, with concomitant reduction in airway inflammation and infection. 23 

Improved Inhaled Mucolytic to Treat Airway Muco-Obstructive Diseases

Functions as a Fast-Acting Antioxidant by Triggering Intracellular H(2)S and Sulfane 533

Hydrogen Sulfide and Cell Signaling

Genetics Reveals a Variable Infection Gradient in the Respiratory Tract

Coronavirus of Probable Bat Origin

Systematic Assembly of a Full-Length Infectious Clone of Human Coronavirus NL63

Human Coronavirus NL63 Employs the Severe Acute Respiratory Syndrome 548

Coronavirus Receptor for Cellular Entry

Crystal Structure of NL63 Respiratory Coronavirus

Receptor-Binding Domain Complexed with Its Human Receptor

Impact of Host Cell Line Choice on Glycan Profile

Molecular dynamics simulations related to SARS-Cov-2

Systematic and Quantitative 560 Assessment of Hydrogen Peroxide Reactivity With Cysteines Across Human 561

SARS-CoV-2 S Protein:ACE2 Interaction Reveals Novel Allosteric Targets

Basis of Receptor Recognition by SARS-CoV-2

Deep Mutational Scanning of SARS-CoV-2 Receptor Binding Domain 573

Reveals Constraints on Folding and ACE2 Binding

Enhanced Receptor Binding of SARS-CoV-2 through 576 Networks of Hydrogen-Bonding and Hydrophobic Interactions

Binding Pocket in the Locked Structure of SARS-CoV-2 Spike Protein

Bat and Pangolin Coronavirus Spike Glycoprotein Structures Provide Insights into 585 SARS-CoV-2 Evolution

Retinoids and Steroids as Ligands of the Free Fatty Acid Pocket 589 of the SARS-CoV-2 Spike Protein

A

Protein Closed Conformation Stabilizers: Epidemiological and Computational 595

Controlled SARS-CoV-2 Binding and Cellular Entry

Uncovering Cryptic Pockets in the SARS-CoV-2 Spike 601

Spike Protein Disulfide Disruption as a Potential Treatment for SARS-604

CoV-2. bioRxiv 2021

Disruption of Disulfides within RBD of SARS-CoV-2 Spike Protein Prevents Fusion 608 and Represents a Target for Viral Entry Inhibition by Registered Drugs

Allosteric Disulfides: Sophisticated Molecular Structures 611

Enabling Flexible Protein Regulation

Antibody Structures Inform Therapeutic Strategies

Structural and Functional 620 Ramifications of Antigenic Drift in Recent SARS-CoV-2 Variants

Structural Impact 624 28 on SARS-CoV-2 Spike Protein by D614G Substitution

The Effect of the D614G Substitution on the 628

Structure of the Spike Glycoprotein of SARS-CoV-2

Spike Mutation D614G Alters 632 SARS-CoV-2 Fitness

Decline of the Extracellular Thiol-Disulfide Balance and Its Role in SARS-CoV-2