Identification of a Dual Inhibitor of Secreted Phospholipase A2 (GIIA sPLA2) and SARS-CoV-2 Main Protease

The development of novel agents to combat COVID-19 is of high importance. SARS-CoV-2 main protease (Mpro) is a highly attractive target for the development of novel antivirals and a variety of inhibitors have already been developed. Accumulating evidence on the pathobiology of COVID-19 has shown that lipids and lipid metabolizing enzymes are critically involved in the severity of the infection. The purpose of the present study was to identify an inhibitor able to simultaneously inhibit both SARS-CoV-2 Mpro and phospholipase A2 (PLA2), an enzyme which plays a significant role in inflammatory diseases. Evaluating several PLA2 inhibitors, we demonstrate that the previously known potent inhibitor of Group IIA secretory PLA2, GK241, may also weakly inhibit SARS-CoV-2 Mpro. Molecular mechanics docking and molecular dynamics calculations shed light on the interactions between GK241 and SARS-CoV-2 Mpro. 2-Oxoamide GK241 may represent a lead molecular structure for the development of dual PLA2 and SARS-CoV-2 Mpro inhibitors.


Introduction
With more than 450 million cases of infected people and 6 million casualties globally, the discovery of efficient agents to treat COVID-19, which is caused by the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), is an unmet need [1]. Pioneering studies showed that enzymes, such as RNA-dependent RNA polymerase (RdRp) and SARS-CoV-2 main protease (M pro ), are attractive targets for the development of novel antiviral agents [2,3].
Studies aiming to understand the pathobiology of COVID-19 have also demonstrated the involvement of lipids and lipid metabolizing enzymes in this potentially lethal infection [4,5]. Phospholipases A 2 (PLA 2 s) are enzymes which catalyze the hydrolysis of membrane glycerophospholipids, releasing free fatty acids (FFAs) and lysophospholipids and initiating arachidonic acid (AA) cascade and promotion of inflammation [6][7][8]. Proteomics studies on SARS-CoV-2 infected cells have revealed alterations of proteins linked to the inflammatory response due to the viral infection [9]. The expression of two PLA 2 s, namely cytosolic PLA 2 (GIVA cPLA 2 ) and secreted PLA 2 (GIIA sPLA 2 ) was notably differentiated after 24 h of infection [9]. Plasma metabolomic and lipidomic studies associated with COVID-19 showed elevated levels of FFAs and reduction in phosphatidylcholines (PCs), which indicated increased enzymatic activity of PLA 2 s [10]. Large-scale plasma analysis has revealed that lipids are strongly involved in the response to infection [11]. The concentrations of oleic acid (OA, C18:1) and AA (C20:4) were directly correlated to the severity of the disease in COVID-19 patients who required admission to an intensive care unit [11]. Most recently, an independent cohort study has demonstrated elevated levels of GIIA sPLA 2 in the plasma of deceased patients in comparison to patients with severe or mild COVID-19, indicating that GIIA sPLA 2 is associated with increased mortality due to COVID-19 [12]. This study highlights the importance of GIIA sPLA 2 , establishing this enzyme as a factor that leads to severe COVID-19 morbidity and mortality, and suggesting it as a therapeutic target to prevent COVID-19 mortality.
Previously described data [4,5,[9][10][11][12], as well as data reported in recent review articles summarizing the role of PLA 2 s in inflammatory diseases [13,14], prompted us to explore if we could identify a small-molecule inhibitor able to simultaneously inhibit PLA 2 and SARS-CoV- 2 M pro . It would be advantageous, if we could target two enzymes with a dual inhibitor able to simultaneously block virus replication by inhibiting SARS-CoV-2 M pro and regulate the inflammatory response by inhibiting PLA 2 . In the present study, we focus on 2-oxoamide (also known as α-ketoamide) small molecules as appropriate agents to inhibit both PLA 2 and SARS-CoV-2 M pro and we describe the first dual inhibitor of GIIA sPLA 2 and SARS-CoV-2 M pro .

Design of Inhibitors
In 2020, Hilgenfeld and coworkers demonstrated that 2-oxoamides are potent inhibitors of SARS-CoV- 2 M pro [15] and reported the X-ray structure of SARS-CoV-2 M pro in complex with the 2-oxoamide inhibitor 1 (Figure 1) [15]. SARS-CoV-2 M pro is a cysteine protease and its key cysteine residue may attack small-molecule inhibitors containing either a reactive carbonyl group or a Michael acceptor functionality [15][16][17][18][19][20][21]. In previous years, we have designed and synthesized a variety of 2-oxoamides as inhibitors of PLA 2 s. More specifically, we have developed 2-oxoamides (2), which are based on non-natural δor γ-amino acids and selectively inhibit GIVA cPLA 2 [22,23], while 2-oxoamides (3) based on natural α-amino acids selectively inhibit GIIA sPLA 2 [24]. after 24 h of infection [9]. Plasma metabolomic and lipidomic studies associated with COVID-19 showed elevated levels of FFAs and reduction in phosphatidylcholines (PCs), which indicated increased enzymatic activity of PLA2s [10]. Large-scale plasma analysis has revealed that lipids are strongly involved in the response to infection [11]. The concentrations of oleic acid (OA, C18:1) and AA (C20:4) were directly correlated to the severity of the disease in COVID-19 patients who required admission to an intensive care unit [11]. Most recently, an independent cohort study has demonstrated elevated levels of GIIA sPLA2 in the plasma of deceased patients in comparison to patients with severe or mild COVID-19, indicating that GIIA sPLA2 is associated with increased mortality due to COVID-19 [12]. This study highlights the importance of GIIA sPLA2, establishing this enzyme as a factor that leads to severe COVID-19 morbidity and mortality, and suggesting it as a therapeutic target to prevent COVID-19 mortality.
Previously described data [4,5,[9][10][11][12], as well as data reported in recent review articles summarizing the role of PLA2s in inflammatory diseases [13,14], prompted us to explore if we could identify a small-molecule inhibitor able to simultaneously inhibit PLA2 and SARS-CoV-2 M pro . It would be advantageous, if we could target two enzymes with a dual inhibitor able to simultaneously block virus replication by inhibiting SARS-CoV- 2 M pro and regulate the inflammatory response by inhibiting PLA2. In the present study, we focus on 2-oxoamide (also known as α-ketoamide) small molecules as appropriate agents to inhibit both PLA2 and SARS-CoV-2 M pro and we describe the first dual inhibitor of GIIA sPLA2 and SARS-CoV-2 M pro .

Inhibition of SARS-CoV-2 M pro by 2-Oxoamide PLA2 Inhibitors and Analogs
The inhibitory potency of the known PLA2 inhibitors and all of the new 2-oxoamides synthesized against SARS-CoV-2 M pro was assessed by determining the extent of enzyme inhibition (% inhibition); the results are summarized in Tables 1 and 2 Amine 5b was synthesized from carbobenzoxy-L-valinol by protection of the hydroxyl group, using tert-butyldimethylsilyl chloride, as described in [27], and then removal of the Cbz group. Amine 5c was synthesized from tert-butyloxycarbonyl-L-valinol (9) by protection of the hydroxyl group, followed by removal of the Boc group (Scheme 3). Amine 5b was synthesized from carbobenzoxy-L-valinol by protection of the hydroxyl group, using tert-butyldimethylsilyl chloride, as described in [27], and then removal of the Cbz group. Amine 5c was synthesized from tert-butyloxycarbonyl-L-valinol (9) by protection of the hydroxyl group, followed by removal of the Boc group (Scheme 3).
The results for the in vitro inhibition of SARS-CoV-2 M pro by the analogs of GK241 are summarized in Table 2. Conversion of the free carboxyl to the corresponding amide 7d or to a hydroxymethyl-protected group 7c, 7b resulted in abolishment of the inhibitory potency (entries 2-4, Table 2), indicating that a free carboxyl group was necessary for the inhibition. When the long chain of GK241 was replaced by a shorter one (reduction  Table 2).
When the valine residue of GK241 was replaced by alanine, 8a (entry 7, Table 2) was found to inhibit SARS-CoV-2 M pro by 65.49% at 40 µM, while the conversion of the free carboxyl group to an ester (7a, entry 1, Table 2) again led to the abolishment of the inhibitory potency. The derivatives 8b and 7f, based on an Ala-Ala dipeptide, presented almost no activity (entries 8 and 6, Table 2). Finally, compound 7e, based on a glutamine surrogate, did not present any activity (entry 5, Table 2).

Molecular Mechanics Docking and Molecular Dynamics Calculations
To obtain a better insight into the interactions between SARS-CoV-2 M pro and the most active compound GK241, we applied molecular mechanics docking and molecular dynamics (MD) calculations. More specifically, to simulate the specific interaction, compound GK241 was subjected to covalent docking calculations as implemented in the Maestro Schroedinger suite by creating a covalent bond between Cys145 and the 2-carbonyl carbon of the 2-oxoamide moiety. The procedure transformed carbonyl to an sp 3 carbon atom and the adjacent oxygen to a hydroxyl group. The crystal structure of SARS-CoV-2 M pro protein PDB 6Y2F [15] was adequately prepared for simulations by adding the two missing flexible residues E47 and D48, as this loop was near the binding cavity and initial calculations showed that it could interact with the ligand. Moreover, initial calculations showed that the formation of the covalent bond could result in both R and S configurations for the 2-carbon atom, which is not surprising, since there are examples in the literature reporting inhibitors that may lead to both R and S configurations [28]. Thus, docking calculations were finally performed to collect 100 structures for each 2-carbon atom configuration. These structures were further ranked according to the ligand binding energy calculated using the MM-GBSA approach, ranging between −58.69 to −10.66 kcal/mol. Common structural characteristic of the generated structures for both the R and S configuration groups was the tendency to orient the long aliphatic chain to S2, S3, S4 protease clefts (Figure 2A,B), while differences were mainly observed in the conformation of the covalent bond formatted between Cys145 and the 2-oxoamide carbonyl, as expressed by the dihedral angle Cβ cys145 -S cys145 -C2-O2. The lowest energy structures of up to 3 kcal relative binding energy were grouped according to common conformational characteristics of the ligand; representative structures are shown in Figure 2C and Figure S1 (Supplementary Materials). The resulting docking structures appeared to sample the available conformational space by orienting the valine moiety carboxylate (structures 3 and 5) or isopropyl (structures 2 and 4) to the S1 cavity, while in structures 1 and 6 the -OH group or the initial part of the aliphatic chain were oriented in this cleft. Another interesting structural characteristic was that only structures 2, 5 and 7 interacted with H41 specifically through the -OH group. These seven selected structures were further subjected to 50 nsec MD calculations to assess the stability of the protein-ligand interaction. The calculated binding energies and mean RMSD for both protein and ligand are summarized in Table S1 (Supplementary Materials) along with major interactions between the ligand and specific protein residues. Visual inspection of the MD trajectories and conformational flexibility of both the protein and the ligand showed that, in most of the structures, the long aliphatic chain had the tendency to widely explore the conformational space, probably inducing protein conformational changes. The RMSF calculated for the protein Cα showed that the loop between residues 45-50, in particular, exhibited increased flexibility ( Figure S2, Supplementary Materials) which appeared to be related to interactions with part of the long aliphatic chain. A relative flexibility was also observed for residues 185-192, specifically in the case of structure 1. Among the seven structures, the most stable during MD simulations was found to be structure 7, and, thus, MD simulations were extended to 150 nsec to validate the initial observation ( Figure 2D). In structure 7, the 2-carbon atom exhibited an S configuration and both the -OH and amide moieties aligned, in general, very well with the ketoamide crystal structure reported by Hilgenfeld and coworkers [15] (Figure 2A,B). The -OH group interacted with H41 for 50% of the MD sampled structures, while the valine side-chain occupied the S1 cavity with the carboxylate interacting with Q166 and the amide NH forming an H-bond with the T26 backbone carbonyl during the whole MD simulation, both directly and through a water molecule. Both protein and ligand RMSD appeared to be low during the whole MD simulation and the protein RMSF validated the stability of all parts of the protein (Figure 2C,D).
( Figure S2, Supplementary Materials) which appeared to be related to interactions with part of the long aliphatic chain. A relative flexibility was also observed for residues 185-192, specifically in the case of structure 1. Among the seven structures, the most stable during MD simulations was found to be structure 7, and, thus, MD simulations were extended to 150 nsec to validate the initial observation ( Figure 2D). In structure 7, the 2carbon atom exhibited an S configuration and both the -OH and amide moieties aligned, in general, very well with the ketoamide crystal structure reported by Hilgenfeld and coworkers [15] (Figure 2A,B). The -OH group interacted with H41 for 50% of the MD sampled structures, while the valine side-chain occupied the S1′ cavity with the carboxylate interacting with Q166 and the amide NH forming an H-bond with the T26 backbone carbonyl during the whole MD simulation, both directly and through a water molecule. Both protein and ligand RMSD appeared to be low during the whole MD simulation and the protein RMSF validated the stability of all parts of the protein (Figure 2C,D). In order to further rationalize the experimental results, the derivatives 8d and 8a were subjected to covalent docking following the same procedure as above, resulting in 100 structures for each of the R and S 2-carbon configurations. The calculated MM-GBSA interaction energies were higher in both cases than those observed for GK241, ranging between −50.91 and −8.35 kcal/mol for 8d and −56.99 and −25.98 kcal/mol for 8a. For comparison with GK241, we selected structures, similar to structure 7, specifically having an In order to further rationalize the experimental results, the derivatives 8d and 8a were subjected to covalent docking following the same procedure as above, resulting in 100 structures for each of the R and S 2-carbon configurations. The calculated MM-GBSA interaction energies were higher in both cases than those observed for GK241, ranging between −50.91 and −8.35 kcal/mol for 8d and −56.99 and −25.98 kcal/mol for 8a. For comparison with GK241, we selected structures, similar to structure 7, specifically having an S configuration and forming an H bond between the -OH and residue H41, which are presented in Figure S3 (Supplementary Materials) along with the crystal structure and structure 7 for comparison. Concerning the derivative 8d, a major difference observed was that the 10-carbon-atom aliphatic chain was always oriented differently covering part of the S1 and S2 cavities, as shown in Figure S3A (Supplementary Materials). This major difference in the conformation of the aliphatic side-chain reflected a major difference in the MM-GBSA calculated binding energy of~4 kcal compared to GK241 structure 7 and was in agreement with the experimental results showing no activity at 40 µM concentration. On the other hand, derivative 8a showed major similarities with GK241 as far as the aliphatic chain was concerned, mainly occupying the same part of the enzyme active site. However, the alanine moiety of 8a occupied the S1 cavity differently to GK241 as the N-Cα ala bond adopted a different conformation compared to GK241, resulting in a different orientation of the methyl group of the alanine moiety compared to the corresponding valine isopropyl. These differences can explain the small differences observed in the experimental activity of these derivatives.

Discussion
PLA 2 s are a superfamily of enzymes [6][7][8] which are involved in almost any inflammatory disease [6][7][8]13,14,29]. In humans, three PLA 2 types, represented by GIIA sPLA 2 , GIVA cPLA 2 and GVIA iPLA 2 , are of high medicinal interest and have been targets for the development of small-molecule synthetic inhibitors [13]. Among the various classes of synthetic inhibitors, 2-oxoamides constitute a class of compounds whose members can selectively inhibit either GIVA cPLA 2 or GIIA sPLA 2 . The results of the present study showed that the selective 2-oxamide inhibitors of GIVA cPLA 2 AX109 and AX074 [22,23] did not exhibit any appreciable inhibition of SARS-CoV- 2 M pro . Similarly, the selective pentafluoroethyl ketone inhibitor of GVIA iPLA 2 GK187 [25] did not show appreciable inhibition of SARS-CoV-2 M pro . On the contrary, the potent 2-oxamide inhibitor of GIIA sPLA 2 (IC 50 143 nM) [24] was found to inhibit SARS-CoV-2 M pro with an IC 50 value of 24 µM. Given that GIIA sPLA 2 has most recently been recognized as a factor contributing to the severity and mortality of COVID-19 [12,30], this finding is of high importance.
Peptide and peptide-mimetic 2-oxoamides have been identified as potent inhibitors of SARS-CoV-2 M pro and their interaction with the catalytic site of the cysteine protease SARS-CoV-2 M pro has been defined by determining the X-ray structure of the enzyme-inhibitor complex [15]. Inhibitors of SARS-CoV-2 M pro have attracted high interest as candidate antiviral drugs [31,32], and, recently, the inhibitor PF-07321332 (nirmatrelvir) has received emergency approval by the Food and Drug Administration (FDA).
Inflammation is a critical factor in COVID-19 [5,33], and, consequently, agents able to combat virus replication, and, at the same time, regulate inflammation, could offer a new approach for the treatment of COVID-19. Since GIIA sPLA 2 is associated with increased mortality by COVID-19 [12], and lipid mediators arising from the activity of PLA 2 have been correlated with severe SARS-CoV-2 infection in humans [34], a therapeutic compound able to simultaneously inhibit both SARS-CoV-2 M pro and GIIA sPLA 2 would be of great value, as it would significantly reduce the risk of COVID-19 mortality. For the first time, a dual inhibitor of GIIA sPLA 2 and SARS-CoV-2 M pro is identified. GK241 shows weak inhibitory activity against SARS-CoV-2 M pro compared to other known 2-oxoamide SARS-CoV-2 M pro inhibitors; however, it may represent a basis for the development of a new class of potent dual-action inhibitors.

General Chemistry Methods
Forced-flow chromatography on Merck ® (Merck, Darmstadt, Germany) Kieselgel 60 F 254 230-400 mesh was used for the purification of the products, while aluminumbacked silica plates (0.2 mm, 60 F 254 ) were used for thin-layer chromatography (TLC). The visualization of the developed chromatograms was performed by fluorescence quenching using phosphomolybdic acid, ninhydrin or potassium permanganate stains. The melting points were determined on a Buchi ® 530 apparatus (Buchi, Flawil, Switzerland) and were uncorrected. Specific rotations were measured on an AA-65 series (Optical Activity Ltd., Bury, UK) polarimeter. 1

General Procedure of Deprotection of tert-Butyl Esters to Carboxylic Acids 8a-c
To a stirred solution of tert-butyl ester 7a,f,g (1 mmol) in dry CH 2 Cl 2 (1 mL), TFA (1 mL) was added and the reaction mixture was left stirring for 3 hrs. After removal of the solvent, the residue was diluted in diethyl ether and precipitation by petroleum ether (40-60 • C) and filtration afforded the desired product.    To a flame-dried flask, under argon, NaH 60% (1.3 mmol, 52 mg) and dry N,Ndimethylformamide (DMF) (1.3 mL) were added. A solution of alcohol 9 (1.0 mmol, 203 mg) in dry DMF (0.7 mL) was added dropwise at 0 • C. The reaction mixture was stirred for 30 min at 0 • C and then benzyl bromide (1.1 mmol, 0.13 mL) was added dropwise. The reaction mixture was left stirring for 16 hrs at room temperature. Upon completion, the reaction mixture was quenched with a saturated aqueous solution of NH 4 Cl (1 mL), H 2 O (3 mL) was added, the aqueous layer was extracted with ethyl acetate (2 × 10 mL) and the combined organic layers were washed with H 2 O (15 mL). The organic layer was collected, and after drying over Na 2 SO 4 , the solvent was removed under reduced pressure.

Enzyme Assay
The enzyme inhibition assay was performed as previously described [15]. A buffer composed of 20 mM Tris, 100 mM NaCl, 1 mM EDTA, pH 7.3 was used for the enzyme inhibition assay. For the determination of the inhibition rate, 0.5 µM of SARS-CoV-2 M pro was incubated with 40 µM or 100 µM of 2-oxoamide in the buffer at 37 • C for 10 min. The FRET substrate was then added to each well at a final concentration of 10 µM and a final total volume of 50 µL, to initiate the reaction. The GraphPad Prism 6.0 software (GraphPad) was used for the calculation of % inhibition rate. Measurements of the inhibition rate for the compounds were performed in triplicate and are presented as mean ± SD.

Protein Preparation
Docking calculations were performed using the SARS-CoV-2 M pro crystallographic structure in complex with the covalent α-ketoamide inhibitor 13b (PDB ID: 6Y2F) [15]. Preparation and minimization of M pro , using the Protein Preparation Wizard tool within the Maestro Schrodinger suite, were performed to ensure structural correctness. Hydrogen addition, bond orders and steric clashes correction, water molecules and HetAtoms deletion, charge optimization and restrained minimization supported by the OPLS3 force field were achieved [35]. Moreover, addition of the missing residues E47 and D48 was performed using the Crosslink Proteins tool, in the Maestro Schrodinger suite. The inhibitor GK241 and analogs 8a and 8d were prepared for docking using the LigPrep tool, in the Maestro Schrodinger suite [36].

Covalent Docking
Covalent docking is a multiple step process, that is designed upon Schroedinger's Glide and Prime, capable to determine ligands activity against a protein target taking into account both non-covalent interactions and covalent bond formation. Covalent docking calculations were carried out using the Covalent Docking application, implemented in Maestro Schrodinger suite. Initially, pose selection was carried out using non-covalent docking simulations (Glide) and positional constraints. Specifically, ligand docking was performed in a mutated binding site. The reactive residue was transformed to alanine and the ligand warhead (the ligand moiety able to form covalent bond) was docked closely to the catalytic residue avoiding unfavorable clashes. Subsequently, the mutation was reversed, and receptor sampling was performed. Covalent bond formation was achieved based on geometric criteria and structural optimization. The following step involved both minimization of protein-ligand complexes in vacuum, and clustering of the optimized poses. This early selection was used as a basis for the further minimization, scoring and