key: cord-0837433-4k9m9xe4
authors: Blanchard, Jan E.; Elowe, Nadine H.; Huitema, Carly; Fortin, Pascal D.; Cechetto, Jonathan D.; Eltis, Lindsay D.; Brown, Eric D.
title: High-Throughput Screening Identifies Inhibitors of the SARS Coronavirus Main Proteinase
date: 2004-10-15
journal: Chem Biol
DOI: 10.1016/j.chembiol.2004.08.011
sha: 7cc6442a29dc229ab5c041ba3974ff0d02e21ffa
doc_id: 837433
cord_uid: 4k9m9xe4

The causative agent of severe acute respiratory syndrome (SARS) has been identified as a novel coronavirus, SARS-CoV. The main proteinase of SARS-CoV, 3CL(pro), is an attractive target for therapeutics against SARS owing to its fundamental role in viral replication. We sought to identify novel inhibitors of 3CL(pro) to advance the development of appropriate therapies in the treatment of SARS. 3CL(pro) was cloned, expressed, and purified from the Tor2 isolate. A quenched fluorescence resonance energy transfer assay was developed for 3CL(pro) to screen the proteinase against 50,000 drug-like small molecules on a fully automated system. The primary screen identified 572 hits; through a series of virtual and experimental filters, this number was reduced to five novel small molecules that show potent inhibitory activity (IC(50) = 0.5–7 μM) toward SARS-CoV 3CL(pro).

had been developed as a transition-state analog inhibitor for HIV-1 protease and had a K i of 0.6 M with 3CL pro [29] .

Although a number of existing drugs that could potentially act as inhibitors of 3CL pro were identified through virtual screening, other than ribavirin, it is not known how active these compounds actually are toward the protease. Additionally, in vitro screening campaigns illustrate that out of over 500 existing protease inhibitors, only two were identified as active toward 3CL pro . These findings illustrate the need to look toward the development of innovative inhibitors for this protease. One such study has recently identified unique keto-glutamine analogs as inhibitors of 3CL pro with IC 50 's in the low micromolar range (J.C. Vederas, submitted). In an effort to advance this initiative, we sought to identify novel small molecules that specifically target 3CL pro . In this paper, we describe the cloning, expression, and purification of 3CL pro from the SARS-CoV Tor2 isolate, and subsequent 

In order to screen SARS-CoV 3CL pro against a library of are spatially separated; subsequent fluorescence from the Abz functionality is not quenched and can be detected. 50 ,000 compounds, it was necessary to use an activity assay for the proteinase that could be adapted to an automated system. Ideally, such an assay would (1) work of kinetic data using the Abz-Tyr(NO 2 ) substrate, the over a time scale of several minutes at most to facilitate primary data were corrected for inner filter effect [31] . the rapid evaluation of thousands of potential inhibitors;

This adjustment was essential as it was observed that (2) be homogeneous for ease of automation; and (3) at even relatively low concentrations of substrate (2 M), be sensitive to minimize background signal from the the fluorescence of the cleaved product was signifipresence of the compounds tested. Quenched fluorescantly quenched by the nitrotyrosine moiety of the intact cence resonance energy transfer (FRET) assays have substrate. become a common tool to monitor proteinase activity The K m for the Abz-Tyr(NO 2 ) substrate with 3CL pro was [30] and fulfill the necessary criteria for use in an efficient 820 Ϯ 130 M (Figure 2A ). This value is comparable to and robust high-throughput screening (HTS) campaign. those seen recently for nonlabeled 11-mer peptides that The basis for such an assay is the modification of a also mimic the cleavage sites of the natural substrates peptide substrate to include a fluorescent label and of 3CL pro (0.286-1.94 mM) [17], implying that the labels quencher on opposing sides of the proteinase cleavage incorporated for FRET were not detrimental to the assosite (Figure 1 ). Using this substrate, the activity of a ciation between enzyme and substrate. The calculated proteinase can be measured directly without the need values of the apparent k cat and k cat /K m for the labeled for additional steps to purify and/or characterize the substrate were 1.01 Ϯ 0.09 min Ϫ1 and 1.2 Ϯ 0.2 mM Ϫ1 products. This type of assay is also beneficial as it permin Ϫ1 respectively, and were also similar to those remits enzymatic reactions to be monitored in real time to obtain accurate reaction rates. This aspect of the quenched-FRET assay is particularly attractive since, in evaluating the activity of a proteinase in terms of a rate, the potential for false negatives to occur in the screen due to the presence of inherent fluorescence from the compounds being tested is reduced.

Two fluorogenic peptide substrates were synthesized to develop a continuous assay of 3CL pro proteolytic activity based on FRET. The peptides were designed principally on the sequences flanking 3CL pro in the polyprotein and solubility considerations. A peptide incorporating the anthranilate-nitrotyrosine donor-acceptor pair (Abz- and was the preferred substrate for HTS. In the analysis also illustrates the quality of the screen in terms of the to be monomeric, whereas it is the dimeric form of the reproducibility between the duplicate sets of data. This enzyme that is believed to be catalytically competent.

figure also indicates that 3CL pro appears to have been For the purposes of screening, it is desirable to use the inhibited to some degree by a large number of comlowest concentration of substrate that yields a reliable, pounds. The cutoff to determine the hit threshold was reproducible signal. It was found that assays using 100 set to 50% residual enzyme activity in order to focus M of the fluorogenic peptide gave a strong signal that on the most potent inhibitors. was linear with varying 3CL pro concentration ( Figure 2B ).

Five hundred and seventy-two hits (1.1% hit rate) were This enabled us to screen at substrate concentrations identified from the primary screen of 3CL pro . For the well below K m , which is advantageous for identifying preliminary analyses of these hits, we opted to restrict inhibitors that compete with substrate for the enzyme the number of compounds entering secondary screenactive site [33] and are therefore more likely to be speing by a series of virtual and experimental filters. The cific toward 3CL pro . hits were first classified by recursive partitioning into The compound library used to screen 3CL pro contained 126 groups based on similarities between molecular 50,000 small molecules with an average molecular mass structure and activity in the primary screen [37, 38] . of 325 g/mol. This particular collection was screened Within each of these groups, the compound that strongly because of its high quality, diversity, drug-likeness [34] , inhibited 3CL pro was chosen as a representative candiresupply availability, and past success with other targets date for further study in secondary analyses. The dose- [35] . The campaign was run in duplicate in 384-well response curves of the 126 representative compounds microplate format on a fully automated system. were determined using standard assay conditions (100 The statistical parameter ZЈ, which is described as a M substrate, 1 M 3CL pro ) and 90 nM-100 M of tested measure of the quality of an HTS campaign [36] , is decompound. The dose-response relationship of 72 comfined as pounds resulted in the typical sigmoidal semilogarithmic curve associated with desirable inhibitors; the 54 com-

pounds that did not result in this typical curve were filtered out. To eliminate nonspecific inhibitors, the effect of the where cϩ , cϪ , cϩ , and cϪ are the standard deviations () and averages () of the high ( cϩ ) and low ( cϪ ) controls.

remaining 72 compounds on 3CL pro activity in the presence of bovine serum albumin (BSA) was determined. This value reflects both the error associated with the controls of a screen, as well as the size of the "hit win-If the degree of inhibition of 3CL pro by a compound is decreased in the presence of BSA, inhibition is not likely dow" (Figure 3) . A ZЈ of 0.5 or greater is indicative of a quality screen with a well-defined hit window [36] . The to be targeted exclusively toward 3CL pro . These socalled "promiscuous" inhibitors likely form aggregates ZЈ for the primary screen of SARS-CoV 3CL pro was 0.60 for each of the two replicates indicating that the assay in solution that nonspecifically absorb or adsorb the target enzyme [39] . Only three compounds showed any used to detect activity of 3CL pro was robust and amenable to HTS. The replicate plot of residual enzyme activity marked reduction in enzyme inhibition in the presence of BSA, with the greatest effect being a decrease of though this renders such compounds as unlikely drug candidates, they may be useful in studying the active only 20%; these compounds were removed from the prospective inhibitor pool. site architecture of 3CL pro . Of the remaining 69 candidate inhibitors, the inhibitory action of 5 of these was not The remaining 69 potential candidates were predominated by compounds with at least one potentially reac-significantly affected by the inclusion of 1 mM DTT; these were selected for further characterization. tive center (Table 1 ). The thiol side chain of the active site cysteine of 3CL pro is necessarily a good nucleophile

The dose-response curves for these five compounds with 3CL pro yielded IC 50 values of 0.5-7 M ( Figure 5 ). and therefore has the capacity to combine with a variety of electrophilic functionalities to form covalent adducts.

To evaluate the selectivity of each compound, we investigated their ability to inhibit four other proteinases with This reaction is the basis for several classes of existing peptidic inhibitors of cysteine proteases which include varying structural and mechanistic relatedness to SARS-CoV 3CL pro . The Hepatitis A virus (HAV) 3C pro , like all ␣-␤-unsaturated ketones, esters, and amides [40, 41] ( 1, 14) have long been recognized as being reactive toward thiols in general, titis C Nonstructural 3 proteinase (NS3 pro ) and chymotrypsin are serine proteases with the same two ␤-barrel and cysteine proteases in particular [43] . The two identified sulfur containing compounds (Table 1 , 13) are also fold as 3CL pro [13, 46]. Finally, papain is a paradigm cysteine proteinase with an active site Cys/His/Asp ca-capable of forming disulfide linkages with the catalytic thiol of 3CL pro to inactivate the enzyme [44] The re-talytic triad whose structural fold does not resemble that shared by chymotrypsin, NS3 pro , 3CL pro , and 3C pro . maining carbonyl functionalities and corresponding nitrogen-and sulfur-containing analogs listed in Table 1 , It was difficult in some cases to obtain a full doseresponse curve for every inhibitor with each of the pro-5-12, may also have enough electrophilic character to promote the nucleophilic attack of the active site thiol teinases due to interfering compound fluorescence and/ or compound insolubility under each of the assay condi-of 3CL pro . Due to the potential inherent reactivity of the 69 candidate inhibitors, we sought to identify and eliminate tions. In such cases, if the lack of data at higher compound concentrations precluded interpolation of a reli-those compounds whose high reactivity rendered them nonspecific toward the active site of 3CL pro by including able IC 50 value, the IC 50 is reported as a lower limit (Table  2) . Alternatively, those compounds for which enzyme 1,4-dithio-D,L-threitol (DTT) in the inhibition reaction. A reduction in the inhibition of 3CL pro by any candidate inhibition was not seen at the highest concentration tested, are indicated accordingly. It is important to note compound in the presence of DTT suggests that such a compound would probably react nonspecifically with that in the absence of full mechanistic analyses, the IC 50 values are intended to reveal profound differences in any biologically available thiol functionality [40, 41] . Al-Considering the mechanistic and structural relationships between 3CL pro and HAV 3C pro , and to a lesser extent NS3 pro , we were interested to learn that some of the compounds showed highest activity against these related enzymes. In addition to inhibiting 3CL pro with an IC 50 MAC-8120 and MAC-13985 also demonstrated interesting selectivity profiles with good activity against SARS-CoV 3CL pro (IC 50 values were 4.3 and 7 M, respectively) and no detectable inhibition of any of the other proteinases tested. These particular compounds, however, proved to be difficult in the assays of HAV 3C pro and NS3 pro due to background fluorescence, and could not be tested at concentrations greater than 5 M. Nevertheless, no detectable inhibition was seen at that concentration, indicating that an IC 50 for these molecules would necessarily be much higher.

The protease selectivity data, therefore, point to two molecules with apparent selectivity for 3CL pro , MAC-8120 and MAC-13985. These may well be the most promising leads for further mechanistic characterization and optimization efforts toward a protease-based antiviral for SARS-CoV. One molecule, MAC-30371, has an interesting selectivity for the picornaviral-like proteinases and could prove to be a viable lead for this family of proteolytic enzymes. yeast (IC 50 ϭ 56 M to 4 mM) [47] . This study examined the inhibition of the isomerase by a series of substituted potency to provide insight into the selectivity of the benzothiazoles, and indicated that MAC-8120 interinhibitors, and to aid in their prioritization as leads for acted with cysteine residues on the enzyme that were continuing studies. previously shown to be central in dimerization. This in-Compound MAC-22272 showed the least selectivity teraction, although as yet uncharacterized, was reported with respect to 3CL pro (IC 50 ϭ 2.6 M), as it was the only to perturb the contact interface of the subunits of the inhibitor capable of inactivating all of the proteinases homodimer to inactivate the enzyme. It is therefore quite studied, although the response with NS3 pro (IC 50 Ͼ 500 possible that MAC-8120 associates with the active site M) was significantly lower than with the other enzymes. cysteine of 3CL pro in a similar manner to disrupt its cata-MAC-5576 showed some selectivity toward those prolytic activity. The exact mechanism of inhibition of 3CL pro teinases with a chymotrypsin-like fold, as it was seen by MAC-8120 and the remaining secondary hits requires to inhibit 3CL pro (IC 50 ϭ 0.5 M), HAV 3C pro (IC 50 ϭ 0.5 further study and characterization in order to evaluate M), and chymotrypsin (IC 50 ϭ 13 M), although no effect their potential to act as leads for the development of was seen with NS3 pro , which also shares this structural novel therapeutics against SARS-CoV, and potentially, motif. Papain, which lacks this fold, was unaffected by other picornaviral-like proteinases. Additionally, the MAC-5576. An important difference between 3CL pro and core structure of these compounds can be chemically 3C pro is that the coronaviral enzyme is only functionally modified to improve key physical properties such as active as a dimer whereas 3C pro is active as a monomer potency and solubility. in vitro. The interaction between the 3CL pro subunits is It is clearly evident that the 3CL pro screening campaign sufficiently weak that disrupting the dimerization interhas generated a wealth of data that can be mined for a face has been proposed as a promising target for novel number of studies. While it was our intention to identify antiviral agents [32] . In view of the relatedness between novel small molecules that would lead to the develop-3CL pro and HAV 3C pro , it is therefore likely that of the five ment of therapeutics effective against SARS-CoV, the secondary hits, those that are also active toward HAV results generated by the screening campaign can be 3C pro do not interact with 3CL pro at its dimer interface, applied to other types of studies of 3CL pro and related proteinases including the picornaviral 3C pro . To this end, but at a site that is common to both proteinases. , and peptide and 3CL pro as above. Wells conpolymerase (Roche Applied Sciences) according to the manufacturtaining high and low controls, where neat DMSO was added instead er's conditions, using an annealing temperature of 50ЊC. The reof library compounds, were included on each microplate; buffer was sulting amplicon was digested using BsmI and HindIII, and was used added to the reaction instead of 3CL pro for low controls. Primary to replace the corresponding fragment in pT7HSP1. This yielded hits were defined as those compounds that reduced the activity of pT7HSP2, whose nucleotide sequence was confirmed as described 3CL pro to half of the average residual activity of the high controls. 

Old drugs for a new bug; influenza, HIV drugs covalent inhibition of papain by a peptide nitrile. 13 C NMR evidence for a thioimidate ester adduct

A major outbreak of

Cysteine proteases and Severe Acute Respiratory Syndrome in Hong Kong

Picornaviral 3C cysteine proteinases have a fold similar to chymotrypsin-like serine proteinases

Serpin mechansim of Hepatitis C virus Nonstructural 3 (NS3)

Outcome of coronaprotease inhibition

Rodvirus-associated severe acute respiratory syndrome using a standard treatment protocol

Cheng, products with NaCl as the inducer

A bacteriophage T7 RNA Small molecules targeting severe acute respiratory syndrome polymerase/promotor system for controlled exclusive expreshuman coronavirus

energy transfer (IFETS)

suring kinetic parameters with inner filter effect correction

Dissection study on the proteinase

Severe Acute Respiratory Syndrome 3C-like protease reveals the critical role of the extra domain in dimerization of the enzyme

Mechanistic considerations in highthroughput screening

Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings

High-throughput screening identifies novel inhibitors of Escherichia coli dihydrofolate reductase that are competitive with dihydrofolate

A simple statistical parameter for use in evaluation and validation of high-throughput screening assays

Topics in Applied Multivariate Analysis, D.M. Hawkins

Classification and Regression Trees

A common mechanism underlying promiscuous inhibitors from virtual and high-throughput screening

Structure-based design, synthesis, and biological evaluation of irreversible human rhinovirus 3C protease inhibitors. 1. Michael acceptor structure-activity studies

Structure-assisted design of mechanism-based irreversible inhibitors of human rhinovirus 3C protease with potent antiviral activity against multiple rhinovirus serotypes