key: cord-0907832-t4dw5rg1
authors: Liu, Cheng; Boland, Sandro; Scholle, Michael D.; Bardiot, Dorothee; Marchand, Arnaud; Chaltin, Patrick; Blatt, Lawrence M.; Beigelman, Leonid; Symons, Julian A.; Raboisson, Pierre; Gurard-Levin, Zachary A.; Vandyck, Koen; Deval, Jerome
title: Dual Inhibition of SARS-CoV-2 and Human Rhinovirus with Protease Inhibitors in Clinical Development
date: 2021-01-27
journal: Antiviral Res
DOI: 10.1016/j.antiviral.2021.105020
sha: 2750f3bfb4a1b10d77705e90808fdc75528f8fda
doc_id: 907832
cord_uid: t4dw5rg1

The 3-chymotrypsin-like cysteine protease (3CLpro) of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is considered a major target for the discovery of direct antiviral agents. We previously reported the evaluation of SARS-CoV-2 3CLpro inhibitors in a novel self-assembled monolayer desorption ionization mass spectrometry (SAMDI-MS) enzymatic assay (Gurard-Levin, Liu et al., 2020). The assay was further improved by adding the rhinovirus HRV3C protease to the same well as the SARS-CoV-2 3CLpro enzyme. High substrate specificity for each enzyme allowed the proteases to be combined in a single assay reaction without interfering with their individual activities. This novel duplex assay was used to profile a diverse set of reference protease inhibitors. The protease inhibitors were grouped into three categories based on their relative potency against 3CLpro and HRV3C including those that are: equipotent against 3CLpro and HRV3C (GC376 and calpain inhibitor II), selective for 3CLpro (PF-00835231, calpain inhibitor XII, boceprevir), and selective for HRV3C (rupintrivir). Structural analysis showed that the combination of minimal interactions, conformational flexibility, and limited bulk allows GC376 and calpain inhibitor II to potently inhibit both enzymes. In contrast, bulkier compounds interacting more tightly with pockets P2, P3, and P4 due to optimization for a specific target display a more selective inhibition profile. Consistently, the most selective viral protease inhibitors were relatively weak inhibitors of human cathepsin L. Taken together, these results can guide the design of cysteine protease inhibitors that are either virus-specific or retain a broad antiviral spectrum against coronaviruses and rhinoviruses.

specificity for each enzyme allowed the proteases to be combined in a single assay reaction 23 without interfering with their individual activities. This novel duplex assay was used to profile a 24 diverse set of reference protease inhibitors. The protease inhibitors were grouped into three 25 categories based on their relative potency against 3CLpro and HRV3C including those that are: 26 equipotent against 3CLpro and HRV3C (GC376 and calpain inhibitor II), selective for 3CLpro 27 (PF-00835231, calpain inhibitor XII, boceprevir), and selective for HRV3C (rupintrivir). 28

Structural analysis showed that the combination of minimal interactions, conformational 29 flexibility, and limited bulk allows GC376 and calpain inhibitor II to potently inhibit both 30 enzymes. In contrast, bulkier compounds interacting more tightly with pockets P2, P3, and P4 31 due to optimization for a specific target display a more selective inhibition profile. Consistently, 32 the most selective viral protease inhibitors were relatively weak inhibitors of human cathepsin L. 33

Taken together, these results can guide the design of cysteine protease inhibitors that are either 34 virus-specific or retain a broad antiviral spectrum against coronaviruses and rhinoviruses. 35 standards. The samples were purified by washing the SAMDI arrays with deionized ultrafiltered 126 water (50 µL / spot) and dried with compressed air. A matrix comprising alpha-cyano cinnamic 127 acid in 80% acetonitrile:20% aqueous ammonium citrate (10 mg/mL final) was applied in an 128 automated format by dispensing 50 nL to each spot in the array. SAMDI-MS was performed 129 using reflector-positive mode on an AB Sciex TOF-TOF 5800 System (AB Sciex, Framingham, 130 MA) with 400 shots / spot analyzed in a random raster sampling. For data analysis, area under 131 the curve (AUC) peaks for the product and internal standard were calculated using the TOF/TOF 132 Series Explorer (AB Sciex) and the amount of product formed was calculated using the equation 133 (AUC product / AUC internal standard ). Negative controls were pre-quenched with 0.5% formic acid. 134

Assay robustness was determined by Z-Factor, calculated using the equation; 135 1 -(3(σ pos +σ neg )/(µ pos -µ neg )) 136

where σ is the standard deviation and µ is the average conversion of the positive and negative 137 control wells. 138

Assays were performed in 20 µL volume in 384-well small-volume non-binding microtiter plates 140 (Greiner Bio-One; Monroe, NC) at ambient temperature in the assay buffer (20 mM sodium 141 acetate, 1 mM EDTA, 0.005% Brij-35 and 5 mM DTT pH5.5 To enable a duplex readout of multiple enzyme activities using SAMDI-MS, the enzyme product 162 masses must be mass resolvable. Therefore, peptide substrates were designed for 3CLpro and 163 HRV3C to include their respective canonical cleavage sites so that the anticipated cleavage 164 products would be peptides with distinct masses. Each peptide also features a biotin moiety that 165 enables the specific immobilization of the peptide to neutravidin-presenting SAMs following a 166 homogenous reaction in a 384-well plate for MS analysis (Gurard-Levin, 2020). To determine 167 the product masses, 3CLpro and HRV3C were incubated with their cognate substrates in separate 168 reactions and the products analyzed by SAMDI-MS. In the presence of enzyme, the SAMDI-MS 169 respectively, corresponding to each protease product (Supplementary Figure 1) . The distinct 171 masses of the products support their potential application in a duplex SAMDI-MS assay. 172

To verify selectivity of the two enzymes for their respective substrate, each enzyme was 173 incubated over a range of concentrations of 0.78-25 nM in the presence of each substrate at 10 174 µM individually and the reactions were monitored over time. Product yields were quantitatively 175 assessed in a SAMDI-MS assay, using internal standard peptides introduced after the reaction. 176

The standards differ from the products in mass but feature similar ionization efficiencies 177 (Gurard-Levin, 2020), enabling a ratiometric analysis to determine yield using the equation 178 AUC prod / (AUC Int Stnd + AUC prod ) ( Figure 1A) . In the presence of their cognate enzyme, product 179 formation accumulated with increasing enzyme concentration ( Figure 1B) . Importantly, the 180 HRV3C substrate is not active towards the 3CLpro enzyme, and the 3CLpro substrate exhibits 181 only minimal activity toward HRV3C at 25 nM ( Figure 1C) . Below 25 nM, the HRV3C enzyme 182 specifically cleaves its substrate. The selectivity of each enzyme for its substrate provides further 183 support for their potential application in a duplex SAMDI-MS assay. 184

Before duplexing, the kinetic parameters of each reaction were characterized individually to 186 ensure optimal assay conditions for characterizing inhibitors. Starting with a buffer recently 187 optimized for the 3CLpro protease (Gurard-Levin, 2020), the K M and maximum velocity (V max ) 188 of each peptide substrate was measured in parallel using SAMDI-MS. Concentrations of the two 189 peptides were titrated between 1.5 and 200 µM and the product monitored over time. Plotting the 190 initial velocity of the linear portion versus the substrate concentration revealed Menten curves for the two enzymes. The SAMDI-MS assay measured an apparent K M of 192 16.43 µM and a V max of 0.29 µM/min for 3CLpro and a K M of 10.76 µM and V max of 193 0.21 µM/min for HRV3C, resulting in a k cat /K M of 5.9 min -1 µM -1 for 3CLpro and 3.2 min -1 µM -1 194 for HRV3C ( and 6 nM HRV3C generate suitable amounts of product in a 30-minute reaction. Importantly, 206 each enzyme also exhibits specificity for its corresponding substrate ( Figure 2C) . 207

The robustness of each enzyme in the duplex assay is critical to quantitatively measure 209 compound potency and selectivity. To assess the robustness of the duplex SAMDI-MS protease 210 assay, reactions were miniaturized to 6 µL in a 384-plate format using the optimized conditions 211 with 3 nM 3CLpro and 6 nM HRV3C. The presence of positive and negative controls permitted 212 calculation of a Z-(or Z'-)factor (a measure of robustness), which considers the mean and 213 standard deviation of positive and negative controls. In this experiment, negative control 214 reactions were pre-quenched with 0.5% formic acid (final concentration). The SAMDI-MS 215 uniformity experiment to measure robustness, the data support a robust assay with Z'-factors of 217 0.77 and 0.70 for 3CLpro and HRV3C, respectively ( Figure 3B) . 218

Next, we evaluated a panel of 3CLpro and HRV3C reported inhibitors in the duplex SAMDI-MS 220 assay (Figure 4) . In the assay, GC376 was equally potent in its inhibition of 3CLpro and 221 HRV3C, with IC 50 values of 2.3 and 5.4 nM, respectively ( Figure 5) . Likewise, calpain inhibitor 222 II displayed a similar degree of inhibition against the two enzymes. PF-00835231 was about 223 1000-fold more potent against 3CLpro than HRV3C. Similarly, both calpain inhibitor XII and 224 boceprevir displayed a marked preference for 3CLpro inhibition over HRV3C. In contrast, 225 rupintrivir was selective for HRV3C, with no observed inhibition of 3CLpro. The inhibition 226 potency results obtained in the duplex SAMDI-MS assay were confirmed in the FRET assay 227 format (Supplementary Table 1 ). In summary, our results distinguish three categories of 228 protease inhibitors based on their relative potency against 3CLpro and HRV3C including those 229 that are: equipotent against 3CLpro and HRV3C (GC376 and calpain inhibitor II), selective for 230 3CLpro (PF-00835231, calpain inhibitor XII, boceprevir), and selective for HRV3C (rupintrivir). 231

To evaluate the selectivity of protease inhibitors beyond viral targets, the same six compounds 232 were tested against human cathepsin L. The most potent inhibitors in this assay were GC376 and 233 calpain inhibitor II and XII, with IC 50 values below 1 nM ( Table 1) . PF-00835231 was a 234 moderate inhibitor of human cathepsin L (IC 50 ~500 nM). The two commercially approved 235 antiviral protease inhibitors rupintrivir and boceprevir only weakly inhibited the human enzyme 236 (IC 50 >1 µM). Taken together, these results indicate that virus-specific protease inhibitors were 237 J o u r n a l P r e -p r o o f relatively weak inhibitors of human cathepsin L, whereas molecules with broad antiviral 238 spectrum GC376 and calpain inhibitor II tended to also inhibit the human enzyme. 239

Comparison of 3CLpro and HRV3C Active Sites 240 GC376, calpain inhibitors II and XII, PF-00835231 and boceprevir have been co-crystallized 241 with 3CLpro, while rupintrivir was not (in line with its lack of activity). However, rupintrivir 242 was co-crystallized with HRV3C protease (PDB 1CQQ & 6KU8), its intended target (Figure  243 6A). HRV3C protease and SARS-CoV-2 3C-like protease display a comparable fold, allowing 244 superimposition of the corresponding crystal structures through structural homology methods, 245 e.g., SHEBA (Supplementary Fig 3) . Nonetheless, the identity between the two proteins 246 remains limited (<15%). The primary sequence of HRV3C protease is also notably shorter; 247 missing the C-terminal dimerization domain typical of coronaviral 3C-like proteases. As a result, 248 the proteases differ markedly around positions P2, P3, and P4. The difference is apparent when 249 displaying the molecular surfaces of the proteins around the bound inhibitors, revealing a more 250 open binding site in HRV3C around those positions (Figure 6A and B) . GC376 inhibits 3CLpro 251 by forming a covalent reversible adduct between the catalytic Cys145 and its aldehyde moiety. 252

This functional group represents one of the simplest warheads that can be used to target Cys 253 residues. As a result, GC376 does not display any functional group that could interact-254 favorably or not-with any of the sub-pockets corresponding to positions P1' and beyond. On 255 the other side of its structure, GC376 is capped by a [(benzyloxy)carbonyl]amino moiety 256 allowing some conformational flexibility. Interestingly, GC376 has been co-crystallized multiple 257 times with SARS-CoV-2 3CLpro (PDB IDs 7C6U, 7CBT, 7C8U, 7BRR, 6WTT, 6WTK, 258 6WTJ). While the carbonyl of the [(benzyloxy)carbonyl]amino moiety consistently forms an H-259 bond interaction with the backbone of Glu166 and can therefore be considered a real "anchor 260 J o u r n a l P r e -p r o o f point," the benzyl group adopts multiple conformations, suggesting that only limited interactions 261 are made (Figure 6C ). In the majority of cases, the benzyl group is "folded" toward sub-pocket 262 P3 and the pyrrolidinone moiety occupying sub-pocket P1. The corresponding region (P3) can 263 essentially be considered solvent-exposed in both 3CLpro and HRV3C protease. Similarly, 264 calpain inhibitor II binds 3CLpro through a minimal aldehyde moiety and positions an isobutyl 265 (Leucine) side chain into the P3 region, making only limited contacts. Compared with GC376, 266 calpain inhibitor II forms an additional H-bond through its acetamide moiety, but lacks the bulky 267 5-methyl-1,2-oxazol-3-yl moiety found in rupintrivir. In summary, the combination of minimal 268 interactions (essentially with P1 and P2), conformational flexibility, and limited bulk allows 269 GC376 and calpain inhibitor II to inhibit both 3CLpro and HRV3C proteases, despite the low 270 sequence identity between them. In contrast, compounds interacting more tightly with pockets 271 P2, P3, and P4 and/or optimized for a specific target display a more selective inhibition profile. activity (k cat /K m = 5.9 min -1 µM -1 for native 3CLpro vs. 0.058 min -1 µM -1 for MBP-3CLpro) 282 (Figure 2A) . We evaluated the antiviral spectrum of a set of reference protease inhibitors 283 J o u r n a l P r e -p r o o f including PF-00835231 and GC376, as well as calpain inhibitor II and XII, boceprevir, and 284 rupintrivir (Figure 4) . There again, changing the enzyme source to the native construct resulted 285 in a 5-10-fold increase of inhibition potency for GC376 as well as calpain inhibitors II and XII 286 ( Figure 5) . This increased potency can be explained by the lower enzyme concentration used for 287 the native construct (3nM) vs. MBP-3CLpro (125 nM). Importantly, our 3CLpro IC 50 values are 288 in good agreement with previous reports for all compounds including the clinical candidate PF-289 00835231 (Boras, Jones et al. 2020 , Hoffman, Kania et al. 2020 , Ma, Sacco et al. 2020 . Our 290 results also reveal that GC376 and PF-00835231, the two most studied protease inhibitors for the 291 treatment of SARS-CoV-2, have very distinct selectivity profiles. GC376 is a very broad 292 cysteine protease inhibitor with no selectivity for 3CLpro over HRV3C and cathepsin L. On the 293 other hand, PF-00835231 inhibits SARS-CoV-2 3CLpro about 100-and 500-fold more potently 294 than cathepsin L and HRV 3C, respectively ( Table 1) . Since cathepsin L is a target for protease 295 inhibitors blocking SARS-CoV-2 cellular entry, targeting both 3CLpro and cathepsin L with a 296 single molecule could potentially have a synergistic antiviral effect at least in cell culture 297 systems (Liu, Luo et al. 2020 , Mellott, Tseng et al. 2020 . However, the antiviral benefit of 298 inhibiting cathepsin L alone or together with 3CLpro in the context of COVID-19 still needs to 299 be demonstrated in animal models and in the clinic. 300

One of the objectives of our study was to define the molecular bases for the selectivity of 301 cysteine protease inhibitors. When binding HRV3C protease, the selective inhibitor rupintrivir 302 positions a 4-fluorophenyl moiety in a P2 "groove" that is partially exposed, in contrast to the 303 more enclosed sub-pocket found in SARS-CoV-2 3CLpro. Such a 4-fluorophenyl moiety was 304 recently reported to decrease the activity of structurally related compounds against SARS-CoV-2 305 3CLpro (Zhang, Lin et al. 2020) . Rupintrivir also binds in an extended conformation that would 306 J o u r n a l P r e -p r o o f not be accepted by SARS-CoV-2 3CLpro. In particular, the [(5-methyl-1,2-oxazol-3-307 yl)formamido] moiety, which makes several favorable contacts in HRV3C protease, would 308 sterically conflict with the amino-acid section connecting the catalytic and dimerization domains 309 of SARS-CoV-2 3CLpro. The other selective inhibitor boceprevir displays a rigid and 310 hydrophobic 3,3-dimethylbutanoyl-6,6-dimethyl-3-azabicyclo[3.1.0]hexan-2-yl moiety, which 311 occupies P2 instead of the more usual leucine residue found in other 3CLpro inhibitors as in 312 consensus substrates. This elaborate group makes multiple hydrophobic contacts with two 313 methionine residues (M49, M165), which are absent from HRV3C protease (one replaced by 314 glycine, one having no corresponding residue; yielding the more open binding site around P2). 315

Both selective inhibitors boceprevir and PF-00835231 form multiple contacts with the amino-316 acid section connecting the catalytic and dimerization domains of SARS-CoV-2 3CLpro, which 317 is missing from HRV3C protease. Of note, the 4-methoxy-indole group found in PF-00835231 318 results from a dedicated optimization of the compound series against 3CLpro (Hoffman, Kania et 319 al. 2020). The lack of corresponding contacts in HRV3C protease explains the selectivity of this 320 inhibitor toward 3CLpro. On the other hand, the simple warhead of GC376 and its lack of 321 targeted functional groups explain its ability to potently inhibit not only SARS-CoV-2 3CLpro 322 and HRV3C protease, but also human cathepsin L ( Figure 5 and Table 1) . 323

Can a protease inhibitor potently target coronaviruses and picornaviruses without inhibiting 324 human proteases? Our structural analyses indicate that compounds optimized to interact tightly 325 with specific pockets of their protein target site tend to display a more selective inhibition 326 profile. Therefore, optimizing a protease inhibitor for broad antiviral spectrum against 327 coronaviruses and picornaviruses without also inhibiting the human counterparts might be 328 challenging, but theoretically achievable. The inhibition of human proteases by clinical stage PF-329 J o u r n a l P r e -p r o o f 00835231, boceprevir, and rupintrivir has been described and these compounds are relatively 330 selective (Dragovich, Prins et al. 1999 , Howe and Venkatraman 2013 , Hoffman, Kania et al. 331 2020 . This selectivity profile is consistent with our own results showing preferred inhibition for 332 one viral protease ( Table 1) . Calpain inhibitor XII, while most active on calpain I, is active vs. 333 calpain II and cathepsin B and L (Li, Ortega-Vilain et al. 1996, Sacco, Ma et al. 2020) . 334 Therefore, our observation that calpain inhibitor XII does not inhibit HRV3C does not mean that 335 this compound is a selective protease inhibitor ( Table 1) . Also calpain inhibitor II is described to 336 inhibit human cysteine proteases calpain I, II, cathepsin L and cathepsin B (Sasaki, Kishi et al. 337 1990) . Furthermore, while inhibitor GC376 does contain a glutamine surrogate in the P1-position 338 and a consensus leucine on P2, this is not considered sufficient to render selectivity for viral 339 proteases, especially as the released active GC373 is a reactive aldehyde. GC373 is known to 340 inhibit cathepsin B with an IC 50 value of 9 nM (Swisher, Prior et al. 2015) . These data are 341 consistent with a very recent study showing that GC373 and GC376 inhibit cathepsin L 342 (https://doi.org/10.1101/2020.11.21.392753). In summary, our 3CLpro/HRV3C duplex assay 343 provides a convenient way to rapidly and systematically measure the antiviral selectivity of 344 clinically relevant protease inhibitors. Coupled with cathepsin L testing, our analyses indicate 345 that dual inhibition of SARS-CoV-2 and human rhinovirus proteases can be achieved with 346 molecules like GC376 and calpain inhibitor II, but that this broad antiviral spectrum also extends 347 to human proteases such as cathepsin L. The potential benefit of inhibiting 3CLpro together with 348 cathepsin L for the treatment of COVID-19 will require further investigation. structures of SARS-CoV-2 3CL-pro with bound GC376 (7C6U green, 7CBT cyan, 7C8U 379 magenta, 6WTT tan, 7BRR light grey, 6WTK blue, 6WTJ orange). The terminal benzyl group of 380 CG376 is found in multiple conformations, but is in most cases oriented toward the solvent-381 exposed sub-pocket P3 and the pyrrolidinone moiety occupying sub-pocket P1. Closer contact 382 with the protein is only suggested for structure 6WTT; wherein the benzyl group is oriented 383 toward P4. 384 385 386 J o u r n a l P r e -p r o o f 

Koen Vandyck, and Jerome Deval are current employees of Aligos Therapeutics. Sandro Boland, Dorothee Bardiot, Arnaud Marchand, and Patrick Chaltin are current employees of Cistim