Derivatives of MOPS: promising scaffolds for SARS coronaviruses Macro domain‐targeted inhibition

Abstract

The severe acute respiratory syndrome coronavirus (SARS‐CoV/CoV‐2) genome encodes 16 non‐structural proteins (nsps), which coordinate cell remodeling, virus replication and participate in viral evasion. Notably, nsp3 contains a protein module termed Macro domain, which carries IFN antagonist activity that interferes with host innate immunity response. This domain is able to bind and hydrolyze ADP‐ribose derivatives. This activity is correlated to viral escape and thus makes Macro domains a valuable therapeutic target. In the present paper, we report a SARS‐CoV Macro domain structure in complex with a MOPS molecule. Based on our structural data, molecular docking was performed on a set of MOPS analogs in the ADP‐ribose binding pocket. We present an ELISA‐based assay to select hits based on the inhibition of recombinant SARS‐CoV/CoV‐2 Macro domain‐ADP‐ribose complex formation. Among the tested analogs, MOPSO and CAPSO are the more efficient in inhibiting ADP‐ribose‐binding. Structural analysis of these molecules in the ADP‐ribose pocket reveals potential interactions with amino acid residues involved in the coordination of ADP‐ribose. Overall, these findings suggest that MOPSO and CAPSO bear potential to be used as a scaffold for the design of Macro domain‐specific inhibitors.

The SARS‐CoV/CoV‐2 genome encodes non‐structural protein 3, which contains a ‘Macro domain’. This protein domain exhibits antagonistic activity against interferon signaling, which disrupts the host innate immune response and can thus mediate viral escape, highlighting its therapeutic potential. Here, we present the structure of the SARS‐CoV Macro domain in complex with a MOPS molecule and evaluate the potential inhibitory effect of MOPS analogs against the SARS‐CoV/CoV‐2 Macro domain.

Abbreviations

4‐PBAK

4‐phenylbutyric acid potassium salt

ADP

adenosine diphosphate

AMP

adenosine monophosphate

ARTD

ADP‐ribosyl transferase diphtheria‐like

ARTs

ADP‐ribosyl transferases

BSA

bovine serum albumin

CAPS

3‐(cyclohexylamino)‐1‐propanesulfonic acid

CAPSO

3‐(cyclohexylamino)‐2‐hydroxy‐1‐propanesulfonic acid

CHES

2‐(cyclohexylamino)ethanesulfonic acid

CHIKV

Chikungunya virus

DSF

differential scanning fluorimetry

ELISA

enzyme‐linked immunosorbent assay

EPPS

N‐(2‐hydroxyethyl)piperazine‐N′‐(3‐propanesulfonic acid)

ESRF

European Synchrotron Radiation Facility

GF

gel filtration

HEV

hepatitis E virus

hPARP1

human poly(ADP‐ribose) polymerase 1

HTRF

homogeneous time‐resolved fluorescence

IFN

interferon

IRFs

interferon regulatory factors

ITC

isothermal titration calorimetry

MAR

mono‐ADP‐ribosylation

MERS‐CoV

Middle East Respiratory Syndrome Coronavirus

MOPS

3‐(N‐morpholino)propanesulfonic acid

MOPSO

2‐hydroxy‐3‐(4‐morpholino)‐1‐propanesulfonic acid

NAD⁺

nicotinamide adenine dinucleotide (oxidized form)

NMR

nuclear magnetic resonance

nsps

non‐structural proteins

ORFs

open reading frames

PAR

poly‐ADP‐ribosylation

PARG

poly(ADP‐ribose) glycohydrolase

PARPs

poly(ADP‐ribosyl) polymerase

PBS

phosphate‐buffered saline

PTM

post‐translational modification

RNA

ribonucleic acid

RTC

replication/transcription complex

RT‐PCR

real‐time polymerase chain reaction

RuV

Rubella virus

SARS‐CoV/CoV‐2

severe acute respiratory syndrome coronavirus/coronavirus 2

SDS/PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

TAPS

N‐tris(hydroxymethyl)methyl‐3‐aminopropanesulfonic acid

TSA

thermal shift assay

Introduction

ADP ribosylation is a versatile post‐translational modification (PTM), involving the covalent attachment of one or several ADP‐ribose moieties to protein substrates, named mono(ADP‐ribosyl)ation or MARylation and poly(ADP‐ribosyl)ation or PARylation, respectively. The enzymes involved in ADP‐ribosylation dynamic are grouped into a superfamily termed ADP‐ribosyl transferases (ARTs), including the PARP proteins—where PARP historically stands for poly(ADP‐ribosyl) polymerase, even though they can have PARylation, MARylation, or no activities, and which have also been called the ARTD family due to their similarity to diphtheria toxin, for review see Ref. [1]. Members of this family are able to bind, add, or remove ADP‐ribose units. Nicotinamide adenine dinucleotide (NAD+) is both the substrate of the reaction and the donor of the ADP‐ribose unit to be linked to a variety of amino acids, such as Arg, Cys, His, Ser, Asp and Glu [1, 2]. This reversible PTM controls protein activation states and/or modulates protein–protein interactions by changing the target docking properties. It was initially reported in the case of chromatin remodeling and gene regulation processes [3]. In the last decades, the roles played by ADP ribosylation were extended to protein homeostasis regulation, neuronal development, cell signaling, cancer, inflammation, and innate immune response to viral infections, reviewed in Ref. [4, 5]. In the latter process, ADP ribosylation was shown to play a critical role in the activation of host players involved in viral RNA sensing, interferon (IFN) regulatory factors (IRFs) activation, and interferon‐stimulated genes induction [6].

Viruses evolved different strategies to take advantage of [7, 8] or counter [9, 10, 11] ADP‐ribosylation effects. In several viral families, including Coronaviridae, Togaviridae, and Hepeviridae, one of the evasion mechanisms relies on the presence of an ADP‐ribose binding and hydrolysis module named Macro domain, similar to those found on certain members of the ARTD family, namely PARP9, PARP14 and PARP15 [1]. Mutagenesis studies on viral Macro domains in animal and cell models established a link between viral Macro domain function, IFN susceptibility during infection, viral virulence, and even stress granule formation and disassembly during viral infection [12, 13, 14, 15, 16]. This pointed to the Macro domain as a target for an antiviral strategy, further considered during the COVID 19 pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2). The fast spread of SARS‐CoV‐2 variants permanently challenging the efficiency of available vaccines underlined the urgency of exploring novel potential antiviral targets in order to propose new therapeutic [17, 18]. Moreover, apart from SARS‐CoV‐2, two other coronaviruses caused human fatal respiratory disease in the last two decades, namely SARS‐CoV and the Middle East respiratory syndrome coronavirus (MERS‐CoV), with a mortality rate of ~ 10% and 36%, respectively [19]. The World Health Organization (WHO) [20] records annually new MERS‐CoV infection and death cases in Arabic peninsula, mainly due to the continued circulation of the virus in camelids reservoir [19], reminding us that the coronavirus threat is not yet over.

SARS‐CoV/CoV‐2 and MERS‐CoV belong to the β‐coronavirus genus of Coronaviridae family [21]. The 2/3 of the positive sense ~ 30 kb RNA genome encodes non‐structural ORFs (ORF1a and ORF1ab), giving rise to 16 non‐structural proteins (nsPs). Each nsP harbors several molecular motifs and enzymatic activities ensuring the replication of the viral genome, reviewed in [22]. In SARS‐CoV‐2, the Macro domain is part of a Macro‐like three tandem repeated modules (also named Mac 1 to 3) located in nsP3, the largest processed protein domain of the replication/transcription complex (RTC). Mac 2 and 3 seem to interact with nucleic acids, while Mac 1 harbors a functional ADP‐ribose binding and hydrolysis domain [23], which will be termed Macro domain onward. Structural data of several viral Macro domains are available [24, 25, 26, 27, 28, 29]. These studies established that viral Macro domains are structurally conserved and contain a central beta‐sheet core sandwiched by alpha helices (Fig. 1A), delimitating an L‐shaped cleft. The ADP‐ribose lying in the heart of the central core is stabilized by a network of weak interactions. In the case of SARS‐CoV‐2, the key residues implicated in this network include (a) an aspartate (Asp22), in the adenosine binding pocket, stabilizing the nucleobase; (b) an asparagine (Asn40), interacting with the distal ribose; (c) and a succession of glycine residues Gly45‐48, involved in ADP‐ribose phosphate groups steadying. The two latter elements are part of the catalytic pocket [25, 26, 29, 31, 32]. Moreover, recent studies, including those using solution NMR spectroscopy and neutron crystallography for the characterization of hepatitis E virus (HEV) and SARS‐CoV‐2 Macro domain, further stressed the structural flexibility of the ADP‐ribose binding pocket [32, 33, 34, 35, 36]. The study of Correy et al. [32] further specified the position of water molecules, questioning the still not fully elucidated reaction mechanism.

Fig. 1.

SARS‐CoV structure in complex with MOPS and secondary structure elements. (A) Sequence alignment of Macro domains from different viral genus members, including Coronavirus (SARS‐CoV (PDB: 2FAV), SARS‐CoV‐2 (PDB: 6WOJ) and MERS (PDB: 5HIH)), Alphavirus (VEEV (PDB: 3GQE)) and Rubivirus (Rubella virus (RuV) (PDB: 8P0C)). The alignment was generated using the espript software. Secondary structure elements are represented above the alignment for SARS‐CoV. Residues with strict identity are marked in red box. Residues interacting with MOPS molecule are highlighted with a green star above. (B) Molecular surface model of the MOPS SARS‐CoV Macro domain complex. (C) A closed‐up view of MOPS in the ADP‐ribose binding pocket. (D) MOPS and (E) ADP‐ribose interaction with the backbone residues in the of SARS‐CoV Macro domains ADP‐ribose binding pocket. Chemical structure of MOPS/ADP‐ribose and corresponding amino acids are exposed according to stick and balls model. The covalent bonds of the ligand are shown in purple and that of the amino acid residues in brown. The crescents with the bristles represent hydrophobic interactions. The hydrophobic atoms are colored black. Hydrogen bonds formed between residues and ligand are represented as green dashed lines with the bond length as numeric numbers. Surrounding residues, in the hydrophobic pocket, in contact with the molecule are displayed as red eyelash symbols. Diagrams were generated by employing ligplot+ [30].

Structural data paved the way for Macro domain inhibitor screening and design. Recent studies reporting these attempts highlighted different strategies, identifying promising molecules. Drug repurposing approaches revealed that the Remdesivir parent nucleoside GS‐441524 inhibits ADP‐ribose binding [37] and hydrolysis [38]. In parallel, a strategy based on docking coupled to high‐resolution crystallographic studies pointed out that pyrrolo‐pyrimidine scaffold could be used for the generation of inhibitors targeting the adenosine site [39]. This was corroborated by studies showing that through rational modifications of pyrrolo‐pyrimidine and pyrimidoindole derivatives, ADP‐ribose binding and hydrolysis inhibition could be enhanced, reaching low μm IC₅₀ values [40, 41]. Moreover, by combining the two aforementioned strategies, the screening of an FDA‐approved drug library revealed that binding inhibition could be achieved by targeting subsites outside the adenosine binding pocket. This was exemplified by the action of two β‐lactam antibiotics aztreonam and ceftazidime [42], opening new avenues for Macro domain inhibitors scaffold design.

In this study, we report a SARS‐CoV Macro domain structure in complex with a MOPS molecule and evaluate the potential inhibitory effect of MOPS analogs against SARS‐CoV/CoV‐2 Macro domains. The inhibition of ADP‐ribose binding was monitored by thermal shift assay (TSA), ELISA‐like, and dot blot assay, the latter using anti‐poly‐ADP‐ribose reagent antibodies. Molecular docking of selected MOPS analogs into the ADP‐ribose binding site was used to study the molecular basis of the observed inhibition.

Results

Overall structure of SARS‐CoV Macro domain in complex with MOPS

Structural studies on viral Macro domains have opened interesting research avenues for the development of antiviral tools. Crystallization studies of SARS‐CoV Macro domain, done originally by our group, pinpointed that ADP‐ribose is buried inside hydrophobic crevice. A hydrogen bond established between the N6 nitrogen of the adenosine moiety and the carboxyl group of Asp23 insures its leaning in the catalytic cleft. The phosphate groups extend along three conserved glycine residues (Gly47‐49). A narrow pocket accommodates the distal ribose, stabilized by the conserved Phe133 and Asn41 [24]. During our various attempts to understand the dynamic of the ADP‐ribose binding pocket of SARS‐CoV Macro domain using different ligands and crystallization conditions, we obtained crystals harboring a MOPS molecule positioned in the ADP‐ribose pocket (Fig. 1B–E). The overall shape of the Macro domain was similar to the previously published data, with central mixed β‐sheets (β1–β7) wedged between two layers of α‐helices (α1–3 and α4–6) (Fig. 1A). In these structures, the 2‐N‐morpholine ring and the sulfonic acid of MOPS take the place of the distal ribose and the distal phosphate, respectively. The morpholine ring and the sulfonic acid form hydrogen bonds with residues such as Gly48, Ser129, Gly131, and Phe133 (Fig. 1D,E), which are involved in the interaction with the diphosphate of ADP‐ribose. In addition, the key residues Asn41, Gly41, and Ile132, forming direct contact with the distal ribose [24, 28, 43], establish also hydrophobic association with MOPS structure (Fig. 1D,E). These structural observations prompted us to hypothesize that MOPS could be used as a scaffold to design Macro domain inhibitors. Therefore, MOPS analogs with various substitutions in the morpholine ring and the connecting alkyl chain (Fig. 2) were docked to the ADP‐ribose pocket and compared to the original placement of MOPS. The superimposition of free and ligand‐bound structural models suggests that the fold is flexible enough to adapt to various ligands, namely MOPS, MOPSO, CAPS, CAPSO, EPPS, 4‐PBAK, TAPS, and CHES. Hence, these molecules were selected as potential antagonist candidates for ADP‐ribose binding. Most of these molecules are structurally related and contain a morpholine ring and a sulfonic acid group.

Fig. 2.

Structure of MOPS analogs. (MOPS) 3‐(4‐Morpholino)‐1‐propanesulfonic acid. (MOPSO), 2‐Hydroxy‐3‐(4‐morpholino)‐1‐propanesulfonic acid. (CHES) 2‐(Cyclohexylamino)ethanesulfonic acid. (CAPS) 3‐(Cyclohexylamino)‐1‐propanesulfonic acid. (CAPSO) 3‐(Cyclohexylamino)‐2‐hydroxy‐1‐propanesulfonic acid. (EPPS) 3‐[4‐(2‐Hydroxyethyl)‐1‐piperazinyl]‐1‐propanesulfonic acid. (4‐PBAK) 4‐(Piperidin‐4‐yl) butanoic acid hydrochloride. (TAPS) 3‐[(Tris(hydroxymethyl) methyl)amino]‐1‐propanesulfonic acid.

Our hypothesis was strengthened by studies highlighting the inhibitory effect of sulfonic acids buffers on enzyme activity. It has been shown, for example, that MOPS can reduce up to 40% the activity of bovine adrenal tyrosine hydroxylase [44]. Moreover, MOPS is able to inhibit the activity of polyester hydrolases from the cutinase family at high concentrations [45]. 2‐(N‐morpholino) ethanesulfonic acid (MES) can also have inhibitory properties. It can notably inhibit metallo‐β‐lactamase from Bacteroides fragilis [46]. The resolution of the crystal structure of CRN‐4 from DEDDh family of exonucleases in complex with MES revealed a unique allosteric inhibition mechanism, inducing the shift out of the active site nearby histidine side chain upon MES binding, impending substrate binding and reactive ions positioning [47]. Recently, the crystal structure of SARS‐CoV‐2 Macro domain in complex with MES revealed that it binds into the ADP‐ribose clef. The interaction involves hydrogen bonding from Asn40 to the morpholine oxygen and the residues interacting with the proximal phosphate accommodate the sulfonic acid [29]. Moreover, in another study the 2‐N‐morpholinoethanesulfonyl moiety was incorporated into the linker of poly (ADP‐ribose) glycohydrolase (PARG) inhibitors to target both adenosyl and ribose pockets of the SARS‐CoV‐2 Macro domain. This repurposing of PARG inhibitor fragments allowed the sulfone functional group to strength the residues backbone contacts with α and β‐phosphates of ADP‐ribose [48]. The ability of the MOPS analogs to fit flexibly within the ADP‐ribose binding pocket supports its potential as a scaffold for designing Macro domain inhibitors.

ELISA‐like screen for PAR‐binding inhibitors

The search for Macro domain inhibitors prompted the development of methodological tools with the aim of testing potential inhibitory molecules. These tools include (a) fluorescent, luminescent or biotin‐based assays, where ADP‐ribose is connected through a hydrolysable link to the reporter molecule, which can be monitored after link hydrolysis by Macro domain [41, 49, 50, 51] and (b) biophysical methods including isothermal titration calorimetry (ITC), scanning fluorimetry (DSF) and homogeneous time‐resolved fluorescence (HTRF) [39, 40, 42]. An ELISA‐like assay was also reported to test ADP‐ribosylhydrolase activity of SARS‐CoV‐2 Macro domain [52]. Moreover, ELISAs were largely employed to screen inhibitors targeting Macro domain containing proteins. Either for ADP‐ribosylation prevention, using a coated substrate to be ADP‐ribosylated in the presence of biotinylated NAD⁺ donor to facilitate the detection of attached ADP‐ribose [48]; or ADP‐ribose hydrolysis, using an α‐NAD⁺ isomer mimicking a protein linked ADP‐ribose [53]. Here, we developed an ELISA assay based on the immune detection of PAR (in a free or protein bound form) to screen rapidly inhibitors affecting ADP‐ribose binding to Macro domain.

The ELISA‐like assay is performed by coating the Macro domain protein on an ELISA plate. After the blocking, the coated Macro domain is saturated with fixed concentrations of either free PAR or auto‐PARylated hPARP1 (PAR‐hPARP1) (Fig. 3A). Inhibition assessment is based on the ability of MOPS analogs to occupy the ADP‐ribose binding site in the Macro domain. Hence, in the presence of an inhibitor molecule, the access to the ADP‐ribose binding pocket is denied. There is no PAR binding and no signal is recorded (Fig. 3B). The concentration of Macro domain protein used for coating as well as free PAR or PAR‐hPARP 1 used in the assay were optimized in a concentration‐dependent manner (Fig. 3C). The assay validation was performed using Remdesivir nucleoside GS‐441524 as a positive control for inhibition (Fig. 3D). Subtraction of the non‐specific binding was achieved by using BSA as ligand. As seen in Fig. 3D, the addition of GS‐441524 leads to a 75% reduction in PAR binding. This result confirms previously reported data [37] and validate our ELISA‐like binding assay.

Fig. 3.

ELISA‐like assay for ADP‐ribose inhibitors high scale pre‐screen. (A, B) Cartoon representation of the ELISA assay. ELISA plates are precoated with Macro domain proteins, then incubated with free PAR or auto‐PARylated hPARP1. (A) Signal detection indicates the detection of PAR binding to Macro domain using anti‐PAR agent. (B) The absence of signal denotes the inhibition of PAR binding to Macro domain, in presence of an inhibitor molecule. (C) Assay optimization: Macro domain of SARS‐CoV/COV‐2 were precoated at various concentration on the ELISA plate before the addition of PAR and signal detection. BSA coating serves as negative control (D) Inhibition assay validation in the presence of 50 μm of Remdesivir's parent nucleoside (GS‐441524). See Materials and methods for detailed procedure. The data represent the combined results of three independent experiments done in triplicate, and expressed as mean ± standard deviation.

Once the key assay parameters of PAR binding were determined and the assay validated, using a known Macro domain inhibitor, the selected MOPS analogs were tested. For that end, coated Macro domains of SARS‐CoV and SARS‐CoV‐2 were incubated with the molecules of interest, namely MOPS, MOPSO, CAPS, CAPSO, EPPS, 4‐PBAK, TAPS, CHES, before PAR addition (Fig. 4). After PAR addition, the assay was pursued as previously described. The results of data analysis are presented as percentage of PAR binding inhibition (Fig. 4). One can notice that CAPS, EPPS, 4‐PBAK, TAPS, CHES have no effect on PAR binding in the case of SARS‐CoV (Fig. 4A) and SARS‐CoV‐2 (Fig. 4B). However, MOPS, MOPSO and CAPSO inhibit partially PAR binding at the various concentration tested. In the case of SARS‐CoV Macro domain, the inhibition of PAR binding starts at 50 μm of the tested molecule. At 250 μm of MOPS and MOPSO the reduction of PAR binding oscillates between 10% and 15%. CAPSO shows a stronger inhibitory effect, since it impedes PAR binding starting from 50 μm. Maximal inhibition, up to around 30%, is reached at 250 μm (Fig. 4A). SARS‐CoV‐2 Macro domain shows a slightly different profile, since it is not sensitive to MOPS even at high concentrations. MOPSO and CAPSO, however, are able to hamper PAR binding. The highest inhibition, up to 20%, is achieved with CAPSO starting from 250 μm (Fig. 4B). It is useful to note that in the case of PAR‐hPARP1 binding and inhibition assessment, the assay shows a little higher background (data not shown).

Fig. 4.

MOPS analogs screen of ADP‐ribose binding inhibition towards Macro domains of SARS‐CoV and SARS‐CoV‐2. MOPS analogs were incubated at increased contractions (50–750 μm) with precoated Macro domains before PAR addition and signal detection, using anti‐PAR agent. See Materials and methods for detailed procedure. Experiments were done in triplicate and repeated at least three times. Results are presented as percentage of PAR binding inhibition when compared to negative control (PAR) (without analog addition). Values correspond to means of three independent experiment ± standard deviation. (A) SARS‐CoV. (B) SARS‐CoV‐2.

MOPS analogs inhibit PAR binding in the context of auto‐ribosylated hPARP1

In order to confirm and assess the inhibitory effect of MOPSO and CAPSO on a PARylated substrate, a classical dot blot assay was performed, as described in Ref. [43]. Accordingly, Macro domains were spotted into a nitrocellulose membrane at serial dilutions and incubated with increased concentrations of MOPSO or CAPSO before the addition of PAR‐hPARP1. Following PAR‐hPARP1 addition, signals on the membrane were revealed using anti‐PAR‐binding reagent (Fig. 5). Starting from 250 pmoles of recombinant Macro domain protein, a dose‐dependent inhibition of PAR‐hPARP1 is observed. In the case of MOPSO, the inhibitory effect is more prominent on SARS‐CoV‐2 Macro domain compared to SARS‐CoV, with 63% and 30% inhibition, respectively, at 250 μm of the inhibitor molecule (Fig. 5A,B) CAPSO, however, seem to inhibit equally both Macro domains. At 500 μm of CAPSO, the remaining signal ranges from 16% to 20% at 500 pmoles of protein, for SARS‐CoV and SARS‐CoV‐2, respectively (Fig. 5C,D). The evident drop in signal intensity with increased concentration of CAPSO and MOPSO for both Macro domains correlates with the signal decrease observed in the ELISA assay for PAR binding. Thereby, these data validate ELISA results and reinforce its use for a rapid and easy on/off pre‐screen of potential Macro domain inhibitors.

Fig. 5.

MOPSO and CAPSO inhibit PAR‐hPARP1 binding to Macro domains of SARS‐CoV and SARS‐CoV‐2. Serial dilutions of Macro domain proteins were spotted on the dot blot membrane and incubated with increased concentrations of MOPSO or CAPSO before the addition of PAR‐hPARP1. (A, B) MOPSO; (C, D) CAPSO. Lower panels: Dot blot and graphic representation of PAR‐hPARP1 binding to SARS‐CoV (A, C) and SARS‐CoV‐2 (B, D). Upper Panels: Binding signal quantification, done using imagej software. Average and standard deviation were computed using the graphpad prism program. Quantitative values were normalized to the binding of PAR‐hPARP1 for different dilution of SARS‐CoV and SARS‐CoV‐2 Macro domains without any MOPS analog. The data are presented as the percent of PAR‐hPARP1 binding values obtained from three independent experiments.

MOPSO and CAPSO stabilize SARS‐CoV/CoV‐2 Macro domains

In order to evaluate a possible effect of MOPS analogs on Macro domain protein integrity, TSA experiments were carried out (Fig. 6). Hence, different dilutions of SARS‐CoV/CoV‐2 Macro domain were mixed with MOPS analogs at increased concentration (50–1000 μm). In the absence of inhibitors, the recorded melting temperatures are 47 °C and 48 °C, for SARS‐CoV and SARS‐CoV‐2 Macro domains, respectively; in accordance with previous reports [26, 54, 55, 56]. Among all the analogs tested, only MOPSO and CAPSO shift the melting temperature (∆T _m = 1 °C) of both Macro domains. In SARS‐CoV, the stabilization by CAPSO occurs between 100 and 500 μm (Fig. 6A), while in SARS‐CoV‐2 the stabilization is observed at 250 and 500 μm (Fig. 6B). In case of MOPSO, the shift of the melting temperature occurs between 250 and 700 μm for SARS‐CoV and SARS‐CoV‐2 (Fig. 6C,D, respectively). Thermal stability of both Macro domains increases with increasing analog concentrations until 500 μm of MOPSO and 700 μm of CAPSO. After these concentrations, melting temperature decreases, due probably to a reduction in hydrogen‐bound formation of MOPSO and CAPSO at higher concentrations, consistent with previously reported data for N‐morpholino buffering compounds [57]. TSA was largely employed and shown useful for the screen of such potential Macro domain binders as nucleotides, benzimidazoles, and steroids. Virdi et al. [56] screened the ability of about 2500 compounds to change the melting temperature of the SARS‐CoV‐2 Macro domain. They showed, using molecular docking studies, that some of these compounds, mainly nucleotides, can fit into the ADP‐ribose binding pocket, suggesting their possible use as scaffold for inhibitor design. TSA was also employed to assess promising molecules isolated through crystallographic screening and/or computational docking [39, 41, 42]. All these studies highlighted the fact that the adenosine binding site was highly targeted by potentially inhibitory molecules. Our TSA results confirm that MOPS analogs penetrate into the ADP‐ribose pocket without undermining protein integrity.

Fig. 6.

Effect of CAPSO and MOPSO on the thermal stability of SARS‐CoV and SARS‐CoV‐2 Macro domains. Normalized melting profiles of SARS‐CoV and SARS‐CoV‐2 Macro domains in presence of increasing contractions of CAPSO (A, B) and MOPSO (C, D) are recorded by TSA. The experiment was repeated at least three times. One representative experiment is shown.

Docking of MAPSO and CAPSO molecules into the SARS‐CoV Macro domain ADP‐ribose binding site

The structure of MOPSO and CAPSO are similar to that of CAPS and MOPS, except that they possess an additional hydroxyl (OH) group (Fig. 2). The modeling results obtained with these two compounds are different from the non‐hydroxylated analogs. The presence of the hydroxyl group (OH) seems to be important for the interaction within the ADP‐ribose binding pocket. In order to ascertain their position in ADP‐ribose binding pocket, molecular docking experiments were performed for SARS‐CoV Macro domain by positioning MOPSO and CAPSO in a similar arrangement as the MOPS ligand described above (Fig. 7). Since commercial MOPSO and CAPSO might consist of a racemic mixture, both enantiomers (2S) and (2R) were docked. Structures of protein‐ligand were selected according to computed binding energies (∆G) and fit compared to MOPS molecule (Fig. 1). Both MOPSO and CAPSO molecules are predicted to form hydrogen bonds or hydrophobic contacts with residues fixing the ADP‐ribose in the binding pocket, including the catalytic Asn41 and the glycine stretch Gly47–49 (Fig. 7A–D) [24, 43].

Fig. 7.

Molecular docking and amino acids interaction in SARS‐CoV Macro domain. Docking structures (left panels) and potentially amino acids interaction (right panels) in the ADP‐ribose binding cleft. (A) (2S)‐MOPSO. (B) (2R)‐MOPSO. (C) (2S)‐CAPSO and (D) (2R)‐CAPSO. Docking was performed by positioning each analog at MOPS's place using AutoDock Vina [58]. Chemical structure of (2S)/(2R)‐MOPSO or (2S)/(2R)‐CAPSO and corresponding amino acids are exposed according to stick and balls model. Covalent bonds of ligands are shown in purple and of amino acid residues in brown. The crescents with the bristles represent hydrophobic interactions. The hydrophobic atoms are colored in black. Hydrogen bonds formed between residues and ligand are represented as green dashed lines with the bond length as numeric numbers. Surrounding residues, in the hydrophobic pocket, in contact with the molecule are displayed as red eyelash symbols. Diagrams were generated by ligplot+ [30].

The sulfonic acid of the (2S)‐MOPSO enantiomer is expected to form hydrogen bonds with Ser129, Ala130, Gly131 and Phe133 residues (Fig. 7A), potentially interacting with the α/β‐phosphate of ADP‐ribose molecule. The other part of the molecule, the morpholine ring of (2S)‐MOPSO, might interact with the conserved glycine rich stretch (residues 47–48) (Fig. 7A), accommodating the diphosphate moiety and the distal ribose of ADP‐ribose [24, 43]. In the case of (2R)‐MOPSO, the hydroxyl group is predicted to face Ala40, in the well conserved NAAN motif (Fig. 1A) [30, 33] and Ala51 (Fig. 7B), both potentially establishing hydrophobic contacts with the distal ribose and forming the bottom of the catalytic site [24, 32].

In the case of CAPSO (Fig. 7C,D), in both the R and S conformation, the sulfonic acid may also form hydrogen bonds with Ser129, Gly131, Phe133, and Leu127 and faces Val50, accommodating the α‐phosphate of ADP‐ribose in the SARS‐CoV and SARS‐CoV‐2 binding cleft [17, 25, 26, 29, 31]. In addition, the hydroxyl group of CAPSO is predicted to interact with Ala39, Ala51, and Val50 as (2R)‐MOPSO (Fig. 7C,D).

Discussion

Structural date of SARS‐CoV Macro domains in the presence of a variety of ligands enable us to appreciate the dynamics and plasticity of the catalytic binding pocket. This malleability, in contrast to what was observed for CRN‐4 nuclease (see above), does not induce a major conformational alteration in the binding pocket itself [47]. In the case of SARS‐CoV‐2, amino acid side chain shifts were mainly triggered by nucleotide containing ligands, such as ADP, ADP‐ribose phosphate, or AMP. In the absence of ligand or when MES was used as ligand, the flexibly was assured by the glycine stretch [29]. The data we report for SARS‐CoV crystallization with MOPS support these assumptions. Moreover, TSA experiments for MOPSO and CAPSO sustain their occupancy of the ADP‐ribose binding pocket with additional interactions, compared to MOPS, prompting thermal stabilization. These data are in accordance to what was reported for various viral Macro domains in the presence ADP‐ribose, but also other adenosine derivatives including NAD, AMP, ADP, and ATP [34, 54, 55]. In Macro domain from bat coronavirus HKU4 Ala residues from NAAN motif were shown to interact, via a water molecule, with the O3″ of NAD⁺, when the latter is used as a ligand. This interaction is believed to stabilize the α‐phosphate and prompt the nucleophilic attack in the deMARylation reaction [33]. When CAPSO is used as a ligand, the hydroxyl group of CAPSO might interact with water molecule clusters around the binding cleft, enriching and/or displacing the water‐hydrogen bonding with binding pocket residues, affecting thereby the water molecule network around the oxyanion subsite. The latter, in the case of SARS‐CoV‐2 Macro domain, was shown to accommodate the C2 hydroxyl group of the proximal ribose of ADP‐ribose, facing and interacting with the backbone amines of Phe156 and Asp157 via bridging water molecules [32].

The oxyanion hole, adjacent to the adenosine subsite was also investigated as a target for inhibitory purposes, since it was shown to accommodate fragments from the Enamine Essential library as well as β‐lactam derived antibiotic Aztreonam, through the interaction with sulfone, carboxylate, isoxazole, succinimide and α‐keto acid containing moieties [39, 42]. Moreover, a recent study emphasized the importance of fragment‐oxyanion subsite interaction in improving the inhibitory potency of pyrrolo‐pyrimidine derivatives, targeting the adenosine subsite [40]. Another noticeable feature emerging from modeling analysis is the potential proximity and the likely interaction of the sulfonic group with Phe133 (Fig. 7A–D). A phenylalanine at that position is well conserved among viral Macro domains, including those of coronaviruses and many alphavirus species. Where it is absent, it is replaced by a tyrosine. This is the case of chikungunya (CHIKV), HEV and Rubella (RuV) viruses but also bacterial and PARP containing Macro domains [24, 33, 55]. Phe133 is part of the so‐called SAGIF motif (Fig. 1A,C), forming a loop facing the glycine stretch and delimitating the diphosphate passage [17, 24, 33]. The aromatic cycle at this position ensures the adaptability to different ligands and the plasticity of the ADP‐ribose pocket; as exemplified in several crystallographic studies comparing ADP‐ribose binding to that of other ligands such as NAD⁺, ADP, cAMP, ATP but also various potentially inhibitory molecules including the Remdesivir's parent nucleoside GS‐441524 [29, 34, 38, 40]. Moreover, this residue is believed to play a critical role in water molecules positioning [29, 32]. Hence, the modeling of MOPSO and CAPSO, exhibiting a promising inhibitory potential towards SARS‐CoV and SARS‐CoV‐2 Macro domains, highlighted that their binding to the ADP‐ribose pocket probably solicit the interaction with the key residues from the catalytic site accommodating the distal ribose and the diphosphate. A better direct contact with the residues in the ADP‐ribose binding pocket combined to the additional interactions brought by the hydroxyl group, which is likely important for reorienting the alkyl chain to accommodate the sulfonic acid, could explain the experimental results obtained with CAPSO and MOPSO molecules.

Despite the tremendous effort in the search and the design of inhibitory molecules against viral Macro domains, potent inhibitors are still missing. Inhibitory strategies targeting Macro domain containing enzymes were initially and mainly based on the use of ADP‐ribose analogs, competing with the natural substrate. This was exemplified for ARTD family enzymes but also for viral Macro domains. In CHIKV Macro domain, combining library and virtual screening to in vitro assays led to the identification of small promising active molecules, built on flavonoid or pyrimidone backbones. These molecules target ADP‐ribose binding site, by potentially interacting with Tyr114, Val33 and Asn24 residues, corresponding to Phe133, Val50 and Asn41 in SARS‐CoV [59, 60, 61, 62, 63]. In the study of Zhang et al. [64], the substitution of the phosphates and the distal ribose of ADP‐ribose with various compounds led to the identification of promising compounds, inhibiting the deMARylation activity of human MacroD1 at mm concentrations. Molecular docking of the selected active compounds, targeting the distal ribose binding site, suggested a stacking interaction with residue Phe272, corresponding to Ser129 of SARS‐CoV. It also pinpointed the possible involvement of residues Gly270 and Val271, outside the binding pocket and corresponding to Leu127 and Leu128 of SARS‐CoV, in the inhibition. Moreover, studies in the field explored different compound families and tend to orient the search towards the targeting of allosteric sites in combination with ADP‐ribose binding pocket blocking. Allosteric inhibitory molecules, built on a carbazole scaffold, targeting the Macro2 of PARP14 were shown capable of preventing ADP‐ribose binding. The co‐crystallization of one of these compounds, a methane sulfonamide analog, showed it to bind to an allosteric site adjacent to the ADP‐ribose pocket. The sulfonamide moiety enhanced the interaction by stablishing hydrogen bonding with I1132 and G1133 among others, corresponding to Leu127 and Ser129 in SARS‐CoV [65]. Hence, effort should be pursued in that orientation, combining the targeting of the ADP‐ribose pocket to allosteric sites for the design of Macro domain inhibitors.

Conclusions

Finally, we developed in the present study an efficient low‐cost ELISA‐like assay for the pre‐screen of Macro domain binding inhibitors. Based on the presence of a MOPS molecule in the crystal structure of SARS‐CoV Macro domain, we investigated the inhibitory potential of MOPS analogs on SARS‐CoV and SARS‐CoV‐2 Macro domains. We found that MOPSO and CAPSO display an inhibitory effect towards SARS‐CoV and SARS‐CoV‐2 Macro domain, prompting the use of MOPS as an interesting initial scaffold for inhibitor design.

Materials and methods

MOPS analogs

3‐(4‐Morpholino)‐1‐propanesulfonic acid (MOPS) was purchased from EUROMEDEX (EU0034‐C), 3‐[4‐(2‐Hydroxyethyl)‐1‐piperazinyl]‐1‐propanesulfonic acid (EPPS) was obtained from Thermo Fischer Scientific, Illkirch‐Graffenstaden, France (10213672). All the other analogs were purchased from Sigma‐Aldrich, Saint Quentin Fallavier, France including, 2‐Hydroxy‐3‐(4‐morpholino)‐1‐propanesulfonic acid (MOPSO) (M8389‐25G) 2‐(Cyclohexylamino)ethanesulfonic acid (CHES) (29311‐10G) 3‐(Cyclohexylamino)‐1‐propanesulfonic acid (CAPS) (C2632‐25G), 3‐(Cyclohexylamino)‐2‐hydroxy‐1‐propanesulfonic acid (CAPSO) (C2278‐25G), 4‐(Piperidin‐4‐yl) butanoic acid hydrochloride (4‐PBAK) (540838‐5G), and 3‐[(Tris(hydroxymethyl)methyl)amino]‐1‐propanesulfonic acid (TAPS) (T5130‐25G).

Cloning and protein expression and purification

Cloning, expression, and purification of SARS‐CoV/CoV‐2 Macro domains were conducted as described in Refs [24, 28, 43]. Briefly, nucleotide sequences corresponding to residues 182–355, GenBank #AY291315 (SARS‐CoV) and 207–375, NCBI accession YP_009725299.1 (SARS‐CoV‐2) were cloned into pET28 vector (Novagen) – TWIST Bioscience, San Francisco, Californie, USA after codon optimization. Recombinant proteins, containing a His‐tag at the N‐terminus, were expressed in competent Escherichia coli cells (C41 (DE3) plys), in Turbo Broth medium (Cat#0104 AthernaES). The induction was carried out with 50 μm of IPTG at 25 °C (OD of 0.6 at 600 nm). SARS‐CoV‐2 Macro domain was eluted from the Ni‐NTA column (GE Healthcare, Vélizy, France) with 50 mm Tris–HCl, pH 8, 150 mm NaCl, 5% Glycerol and 200 mm Imidazole, following a column wash with 50 mm Tris–HCl, pH 8, 5% Glycerol, 30 mm Imidazole and 1 m NaCl. The protein was further purified by gel filtration (GF) on a Superdex 75 column, after an overnight dialysis step against GF buffer (50 mm Tris–HCl pH 8, 150 mm NaCl, 5% glycerol). In the case of SARS‐CoV Macro domain, the protein was eluted from an affinity chromatography column with 50 mm Tris–HCl, pH 8, 150 mm NaCl, and 250 mm imidazole. The second purification step was carried on a preparative Superdex 200 GF column in 10 mm MOPS, 150 mm NaCl, pH 7.0. Purified protein was concentrated to 8 mg·mL⁻¹ using a Vivaspin 10 kDa concentrator (Vivascience, Palaiseau, France).

Crystallization

Crystallization of the SARS‐CoV Macro domain was performed as described in Ref. [24]. Briefly, crystals grew in 0.1 m imidazole, pH 7.9, 1.3 m sodium citrate using the hanging drop vapor diffusion method in Linbro plates by mixing 2 μL of protein solution (8 mg·mL⁻¹) with 1 μL of reservoir solution (Vtot = 0.5 mL).

Data collection and structure determination

Diffraction intensities were recorded on the id14eh1 beamline at the European Synchrotron Radiation facility (ESRF). Crystals diffracted to a 1.57 Å resolution and a native data set was collected on an ADSC‐444 at a wavelength of 0.9334 Å. The data were processed, integrated, scaled and merged with autoProc [66]. The structure was determined by molecular replacement with the program molrep using the native structure of the SARS‐CoV Macro domain (PDB number: 2FAV) [24] as a model. Refinement was performed using BUSTER [67], after COOT model building [68]. Processing and refinement statistics are provided in Table 1.

Table 1.

Crystallographic data and statistic for the SARS‐CoV Macro domain complexed with MOPS. Statistics for the highest‐resolution shell are shown in parentheses.

	SARS‐CoV Macro domain complexed with MOPS
Data processing
Wavelength (Å)	0.93340
Space group	P 21
a, b, c (Å)	37.28, 55.41, 108.67
α, β, γ (°)	90.00, 91.71, 90.00
Resolution range (Å)	36.21–1.57 (1.59–1.57)
Total no. of reflections	224 278 (6704)
No. of unique reflections	61 227 (2731)
Completeness (%)	99.2 (89.2)
Multiplicity	3.7 (2.5)
I/σ (I)	10.7 (1.9)
R meas	0.099 (0.608)
R merge	0.08 (0.48)
R‐pim	0.05 (0.37)
CC1/2	0.99 (0.71)
Wilson B‐factor (Å)	11.36
Structure solution and refinement
No. of reflections, working set	61 225 (5693)
No. of reflections, test set (%)	3051 (274), 5.0%
R work	0.17 (0.26)
R free	0.20 (0.28)
No. of non‐H atoms	4464
Protein	3921
Ligand	43
Water	524
Protein residues	517
R.M.S. deviations
Bonds (Å)	0.014
Angles (°)	1.60
Average B‐factors (Å2)	16.15
Protein	14.80
Ligand	19.27
Water	26.14
Ramachandran plot
Favored (%)	98.83
Allowed (%)	1.17
Outliers (%)	0.00
Clashcore	3.89
PDB code	8CB3

ELISA‐like assay

Purified recombinant SARS‐CoV/CoV‐2 Macro domains (at a 10 μg·mL⁻¹ concentration) were separately coated on a 96‐well microtiter ELISA plates (MTPs, costar EIA/RIA Plates, Thermo Scientific, Illkirch‐Graffenstaden, France) overnight at 4 °C. Non‐specific sites on the plates were blocked 1 h at 37 °C with coating buffer (0.05 m carbonate–bicarbonate buffer, pH 9.6), supplemented with 2% of BSA. The inhibition of PAR binding was assessed by adding increased concentrations (50–1000 μm) of the corresponding MOPS analog to the coated Macro domain for 2 h at RT. Remdesivir's parent nucleoside GS‐441524 (MedChem Express, Sollentuna, Sweden) was used as a positive control for ADP‐ribose binding inhibition. At the end of the incubation period, purified free PAR (100 nm PAR, cat# 4336‐100‐01; Trevigen, Montluçon, France) or auto‐ADP‐ribosylated human PARP1 (hPARP1, cat# SRP0192; Sigma‐Aldrich), prepared as described in Ref. [42], were added to the plate for 2 h at room temperature (RT) with shaking. After this step, anti‐poly‐ADP‐ribose binding reagent (cat# MABE1031; Millipore, Guyancourt, France) diluted 1 : 2000 in blocking buffer was added to the wells. Following 1 h incubation at RT, the second antibody anti‐rabbit IgG (Dako, Les Ulis, France) diluted 1 : 2000 was dispense to each well and incubated 1 h at 37 °C. The plates were washed three times with PBS‐Tween after each step. The PAR–antibody complex was revealed by the addition of o‐phenylenediamine solution (cat# P8287; Sigma, Saint Quentin Fallavier, France). The reaction was stopped with 3 m sulfuric acid. Optical densities of formed complexes were measured at 492 nm in the microplate reader (TECAN). BSA was used as a negative control for ADP‐ribose binding. All experiments were performed in triplicate and the data was normalized to the value of PAR‐binding without inhibitor. Results are presented as means ± standard deviation.

Dot blot assay

The assay was carried out as described in Ref. [43]. Briefly, SARS‐CoV/CoV‐2 Macro domain proteins were spotted at various dilutions (750, 500 and 250 pmoles) on a nitrocellulose membrane (Schleicher & Schuell BioScience), using Minifold II dot blot apparatus. Bovine serum albumin (BSA) was used as negative control for binding assessment. The membrane was blocked for 1 h at RT with dot blot buffer (10 mm Tris pH 8, 150 mm NaCl and 0.05% of Tween 20), supplemented with 4% skimmed milk. The blocked membranes were incubated with 50–1000 μm of each tested compound selected by ELISA and TSA for 1 h at RT under agitation. The membrane was then subjected to three washes in dot blot buffer, then incubated with PAR‐hPARP1 for 1 h at RT under agitation. The unbound material was removed by three extensive washes with dot blot buffer. Macro domain‐PAR‐hPARP1 binding was revealed using anti‐PAR binding reagent (cat# MABE1031; Millipore) as described in Ref. [43]. Experiments were repeated three times and the data was normalized to the value of PAR‐binding without inhibitor. Results are presented as means ± standard deviation.

Thermal shift assay

The stability of SARS‐CoV/CoV‐2 Macro domains, in presence of the compounds of interest, was evaluated using protein thermal shift assays, as described in Ref. [43]. To that end, MOPS derivatives were added at increased concentrations (50–1000 μm) to MicroAmp^® Fast Optical 96‐well reaction plates (Applied Biosystems), containing 2 μm of Macro domain proteins in 20 mm HEPES, pH 8, and 150 mm NaCl. Protein thermal shift dye kit (Applied Biosystems, Thermo Scientific, Illkirch‐Graffenstaden, France) was used according to manufacturer's protocol. The reaction was conducted in a 7500 FastReal‐Time PCR System (Applied Biosystems). The first incubation step was carried at 25 °C for 2 min. Then the temperature was raised by 0.5 °C per min to 95 °C. Melting curves were normalized using graphpad prism, and the melting temperatures (T _m) were determined by the inflection point parameter.

Molecular docking of the MOPS analogs to the SARS‐CoV macro domain

Molecular docking of MOPS analogues into the ADP‐ribose pocket of the SARS‐CoV Macro domains was carried out using AutoDock Vina [58], as described in Ref. [69]. MOPS and its analogs were extracted from the PubChem database [70] as SMILE, then Mol2 and PDB files format of the ligands were generated by grade2 server insuring best geometry charge description and minimization [67]. MOPS ligand and water molecules were removed from the structure of the SARS‐CoV Macro domain. The docking grid (dimensions 16 × 10 × 11) was focused on the center of the ADP‐ribose pocket [70]. To rank the docking results, the binding energies (∆G) and the precision pose on the MOPS place were considered. Docking poses were visualized and analyzed using ucsf chimera [71].

Conflict of interest

The authors declare no conflict of interest.

Author contributions

OOG contributed to conceptualization, investigation, methodology, software, writing – original draft, review and editing. KA contributed to resources, writing – review and editing. BM contributed to methodology, writing – review and editing. BC contributed to resources, writing – review and editing. FF contributed to conceptualization, investigation, resources, software, writing – review and editing. NR contributed to conceptualization, investigation, methodology, project administration, resources, software, writing – original draft, writing –review and editing.

Peer review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/febs.70039.

Data availability statement

Crystallographic data are available through the PDB accession number (PDB: 8CB3). Biochemical data supporting our findings are present within the text. Any additional raw data are available upon request from the authors.