SARS-CoV-2 3CL-protease inhibitors derived from ML300: investigation of P1 and replacements of the 1,2,3-benzotriazole

Starting from compound 5 (CCF0058981), a structure-based optimization of the P1 subsite was performed against the severe acute respiratory syndrome coronavirus (SARS-CoV-2) main protease (3CLpro). Inhibitor 5 and the compounds disclosed bind to 3CLpro using a non-covalent mode of action that utilize a His163 H-bond interaction in the S1 subpocket. In an effort to examine more structurally diverse P1 groups a number of azoles and heterocycles were designed. Several azole ring systems and replacements, including C-linked azoles, with similar or enhanced potency relative to 5 were discovered (28, 29, and 30) with demonstrated IC50 values less than 100 nM. In addition, pyridyl and isoquinoline P1 groups were successful as P1 replacements leading to 3-methyl pyridyl 36 (IC50 = 85 nM) and isoquinoline 27 (IC50 = 26 nM). High resolution X-ray crystal structures of these inhibitors were utilized to confirm binding orientation and guide optimization. These findings have implications towards antiviral development and preparedness to combat SARS-like zoonotic coronavirus outbreaks.


Introduction
The Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has been one of the greatest health threats in a generation with nearly 7 million deaths worldwide since Coronavirus disease (COVID-19) took hold [1]. Prior short-lived and overall much less deadly outbreaks include SARS-CoV (2003) and MERS (2012) [2]. Despite promising vaccination efforts and status, antiviral therapeutics are considered essential to reduce symptoms in infected patients, prevent hospitalizations, protect high risk immunocompromised populations, and an obligatory investment to enhance preparedness for future outbreaks. Among the approaches to target SARS-CoV-2, inhibitors of the chymotrypsin like main protease (3CL pro ) have been a subject of intense focus due to the vital role it plays in viral replication [3].
Peptidic covalent inhibitors have been a major focus, which dates back to the initial efforts after SARS-CoV-1 in 2003 [4]. In 2021, P zer received emergency use authorization for the use of their covalent SARS-CoV-2 3CL pro inhibitor, PF-07321332 or nirmatrelvir (Figure 1, 1) for the treatment of acute disease.
As impressive and welcome this early success has been to clinical care major efforts continue toward the development of improved potent inhibitors in order address limitations of this therapy. For example, nirmatrelvir requires co-dosing with ritonavir to boost its exposure to sustain plasma concentrations su ciently above the cellular EC 90 in order to provide adequate target coverage. Drawbacks of related peptide-based covalent inhibitors include low membrane permeability and the required reactivity derisking inherent from the covalent warhead [5][6][7]. Thus, next generation inhibitors have the opportunity to offer signi cant improvement provided they can achieve desirable PK and potency. These parameters must be set with a high bar in mind in order to achieve clinically meaningful antiviral e cacy and a desirable safety pro le for potential prophylactic applications [8].
With a renewed focus on such efforts a number non-peptidic, non-covalent 3CL pro inhibitors have appeared in the last two years to offer new starting points for coronavirus antiviral efforts [9][10][11][12]. The structure of one clinical candidate by Shionogi in partnership with Hokkaido University, [13] known as Ensitrelvir (2), was approved in Japan in November of 2022. In preclinical models of antiviral activity 2 demonstrated an EC 50 of 370 nM in a TMPRSS2 VeroE6 model of SARS-CoV-2 infection. In 2021 a promising perampanel-derived non-covalent series of inhibitors of 3CL pro exempli ed by 3 was disclosed by the Jorgensen and Anderson groups [10,14], achieving an antiviral EC 50 of 980 nM in a SARS-CoV2 plaque reduction assay. In early 2022, the Carlsson team published excellent work disclosing the broadspectrum inhibitor 4 derived from an ultralarge virtual screening campaign which utilized MLPCN probe ML188 as a starting ligand [15]. In SARS-CoV-2 infected Huh7 cells in a CPE-based assay format compound 4 displayed antiviral activity comparable to that of 1 with an EC 50 of 110 nM. Non-covalent 3CL pro inhibitors 3 and 4 bear a related 3-pyridyl or isoquinoline ring system as a P1 group, the former rst identi ed from SARS-CoV-1 probe ML188 [9] and then later by the COVID Moonshot program [16].
Previously, our group described the development of the benzotriazole based P1 inhibitor 5 [17] from a structure-guided optimization of ML300. In a SARS-CoV-2 enzymatic assay 5 inhibited SARS-CoV-2 3CL pro with an IC 50 of 68 nM and in a VeroE6 based plaque reduction assay afforded an EC 50 of 558 nM.
Despite initial promising ndings for lead compound 5 there were several shortcomings to be addressed including the modest cellular activity. In addition, the series suffered high metabolic clearance and CYP inhibition. Metabolite identi cation studies highlighted metabolic soft-spots on the N-benzylic portion of the molecule and up to one-third abundance of metabolism occurring within the P1 benzotriazole ring system [17]. Herein, we describe ndings from a more extensive optimization of the P1 and interactions within the S1 sub-pocket, including the rst X-ray structures of carbon-linked heterocycle based inhibitors within the ML300 series bound to SARS-CoV-2 3CL pro as replacements of the N-linked 1,2,3-benzotriazole moiety. \ We began by examining the crystal structure of the SARS-CoV-2 3CL pro :ML300 complex (PDB: 7LME), and utilized molecular docking and structure-guided design to incorporate replacements of the 1,2,3benzotriazole into P1 [17]. A hydrogen bonding interaction with His163 is a key feature of ML300 and other published sub-micromolar 3CL pro inhibitors, both covalent and non-covalent [17,9,18,11,10] which de nes the S1 subpocket. The removal of atoms necessary to achieve this anchoring interaction is signi cantly detrimental to binding a nity. As shown in Figure 2, this initial effort led to pyridine 7.
Compound 7 as described previously [17], represents a 10-fold reduction in potency versus the benzotriazole 6. In contrast, in human microsomes and S9 fractions pyridine congener 7 demonstrated a two-fold reduction in intrinsic clearance metabolism, albeit still high scaled to hepatic blood ow. To con rm the predicted orientation and binding pose of 7 and understand approaches to improve potency we obtained a high resolution X-ray structure bound to SARS-CoV-2 3CL pro . Indeed, inhibitor 7 occupies a similar orientation and presentation of the P2 c and P2 sp groups compared to prior inhibitors in the series, utilizinga pyrazole and cholorophenyl ring, respectively. Importantly, the pyridine nitrogen acts a suitable hydrogen bond acceptor towards His163 with an interatomic distance of 2.8 angstroms. Other putative hydrogen bond interactions include the P2 c pyrazole nitrogens within 3.0 angstroms of Cys44 backbone carbonyl and Thr25 side chain. In addition, there is a highly conserved 3.0 angstrom interaction between with the Glu166 backbone NH and the anilido amide oxygen. Since 7 displayed an encouraging yet nonoptimal pro le (MW 402, LE = 0.27) and a crystal structure bound to 3CL pro was in hand, we performed a series of docking studies to design and prioritize analogues that incorporate a hydrogen bond acceptor to mimic His163 interaction and potential secondary interactions at the N-termini loop of the oxy anion hole region (Leu141-Ser144). respectively. Initial exploratory SAR in the S1 pocket involved replacing P1 heterocycles with pyrazine, 9, resulting in a >2-fold loss in potency at 6.56 µM and the pyridazine, 10, which maintained micromolar activity, within 4-fold of parent compound 7. Replacement of a 6-membered heterocycle with a 5membered 1,2,3-triazole, 11, and thiazoles 12, 13 and 14 showed no improvement vs. 7. Compound 11 is particularly surprising considering the 1,2,3-benzotriazole 6 has an IC 50 of 270 nM representing a 30-fold loss relative to 11.This observation may in part be due to a combination of an electronic effect of the bicyclic ring and/or based on the structural data a space lling property of the extended ring within the P1 where van der Waals contacts with backbone loop residues Lue141 and Asn142 engage in favorable contacts. Heterocycles substituted with either a small methyl (15), or larger phenyl group (16 and 17) resulted in either a signi cant 20-fold loss or were immeasurable. Finally, saturation of the heterocycle in the form of lactam 18, a key motif utilized in peptidomimetic 3CL pro inhibitor discovery programs, most notably by P zer in nirmatrelvir [18], also resulted in no measurable inhibition when tested in racemic form at 100 mM as the top concentration.
One common challenge observed during optimization of 3CL pro inhibition has been differential SAR as a result of the exibility within loop regions of the active site. This is particularly noted within ML300 series which undergoes a reorganization of the S2 subpocket to accommodate both the N-benyzl P2 sp and the azole P2 c moiety [17]. In an effort to nd more texture in SAR to suitably assess P1 replacements we replaced the pyrazole of 7 with an often more potent imidazole and further addition of a metauoro substitution on the P2 sp phenyl ring to afford 19 (Table 2). This amalgamation of improvements leads to an overall ~8-fold improvement of IC 50 to 0.15 µM whilst also improving ligand e ciency [19][20][21]. With 19 as a sub-micromolar starting point containing the C-4 imidazole as the P2c group we revisited prior P1 modi cations ( Table 2, left column). In addition, in context of the 3-uoro-5-chloro substitution a number of bicyclic P1's ( Table 2, right column) were examined, including a number of benzotriazolyl P1's to further improve inhibitory activity relative to 5. In light of the comparable potency of 27 and 28 we obtained a high resolution X-ray structure of isoquinoline 27 to con rm it's orientation within the S1 pocket. Indeed, a number of potent non-covalent inhibitors bearing an isoquinoline have been leveraged to target SARS-CoV-2 3CL pro [15,16]. Shown below in Figure 4 is an overlay comparing the orientation of 27 and 7.
The co-crystal complex of 27 ( Figure 3) reveals essentially identical positioning of the isoquinoline nitrogen relative to the pyridyl nitrogen of 7 as expected. There are obvious differences within the P2 c and the P2 sp , including a chlorine halogen preference deeper within the P2 sp when the ring system is no longer substituted with uorine. This is the subject of a more in depth disclosure beyond the scope of this report.
As discussed above and as revealed from the structures disclosed the benzo ring of the isquinoline appears to modestly impact electronics and/or provide a subtle steric bulk interaction with Asn142. In order to improve LipE of 27 relative to 19 we asked whether we could modulate interactions in this P1 groove near Asn142 with simpler substitutions. With this in mind, a handful of substituted pyridines were designed in the context of the study in Table 2 to assess whether potency could be maintained or improved while favorably impacting calculated physicochemical properties.
Interestingly, introduction of a uorine in the 3-position of the pyridine (35) resulted in a loss in primary potency in comparison to both the unsubstituted pyridine 19 and the isoquinoline 27. However, exchanging the 3-uoro to a 3-methyl 36 led to 2-fold improvement in the IC 50, 217 nM to 84 nM, respectively, but also a much improved LE and LipE. Unfortunately, this trend is not maintained with larger 3-cyclopropylpyrine derivative 37 as both LE and LipE trend in opposite direction. Finally, we introduced 4substituted pyridines (38-40) which resulted primarily in higher IC 50 values in comparison to the unsubstituted pyridine 19; however, simple methyl derivative 38 was equipotent to 19. Overall ndings from this limited study suggests SAR around 3 and 4-position of the P1 pyridyl ring system is narrow in scope, consistent with the structural information.
The ndings from this P1 study offer new directions to further optimize antiviral activity and metabolic stability to develop nanomolar inhibitors against SARS-CoV-2 and future variants. Details from these efforts will be the subject of a future disclosure. In summary this study led to an enhanced understanding of P1 structural diversity and SAR within the ML300 series. In addition, molecules were identi ed with good potency, Clinical agents and pre-clinical non-covalent SARS-CoV-2 3CL pro inhibitors: 1 (Nirmatrelvir)[8], 2 (Ensitrelivr) [13], 3 [10], 4 [15], and ML300 derived 5 (CCF0058981) [17]. Supplementary Files