A cyclic peptide inhibitor of the SARS-CoV-2 main protease

This paper presents the design and study of a first-in-class cyclic peptide inhibitor against the SARS-CoV-2 main protease (Mpro). The cyclic peptide inhibitor is designed to mimic the conformation of a substrate at a C-terminal autolytic cleavage site of Mpro. The cyclic peptide contains a [4-(2-aminoethyl)phenyl]-acetic acid (AEPA) linker that is designed to enforce a conformation that mimics a peptide substrate of Mpro. In vitro evaluation of the cyclic peptide inhibitor reveals that the inhibitor exhibits modest activity against Mpro and does not appear to be cleaved by the enzyme. Conformational searching predicts that the cyclic peptide inhibitor is fairly rigid, adopting a favorable conformation for binding to the active site of Mpro. Computational docking to the SARS-CoV-2 Mpro suggests that the cyclic peptide inhibitor can bind the active site of Mpro in the predicted manner. Molecular dynamics simulations provide further insights into how the cyclic peptide inhibitor may bind the active site of Mpro. Although the activity of the cyclic peptide inhibitor is modest, its design and study lays the groundwork for the development of additional cyclic peptide inhibitors against Mpro with improved activities.


Introduction
Cyclic peptides are an important class of drugs. The binding affinities and specificities of cyclic peptides rival those of small molecule drugs and biomacromolecule drugs [1e6]. Cyclic peptides often exhibit enhanced conformational stability and improved biological activity compared to linear analogues [7,8]. This conformational rigidity increases resistance to degradation by endogenous proteases, leading to increased plasma stability [9]. Peptide cyclization can also facilitate passage of the peptide through a cell membrane [10,11]. Although cyclic peptides often do not meet common physicochemical guidelines for drug-like cellular uptake and bioavailability, such as Lipinski's Rules of 5, a variety of cyclic peptide drugs are active against intracellular targets [12e18].
Secondary structure and backbone rigidity often impart a greater propensity for cyclic peptides to cross the plasma membrane [19e26]. The cyclic undecapeptide immunosuppressant drug cyclosporine A is an archetypal example for cell-permeable cyclic peptides. Despite its large size (MW ¼ 1202 Da), cyclosporine A is orally bioavailable and passively crosses the plasma membrane. Cyclosporine A achieves its remarkable permeability and bioavailability by adopting a rigid secondary structure comprising two bstrands connected by a b-turn to form an antiparallel b-sheet. The amide backbone of cyclosporine A is also highly methylated, and the remaining unmethylated NH groups are intramolecularly hydrogen-bonded. Cyclic peptide inhibitors that target viral proteins, such as the SARS-CoV-2 main protease (M pro ), may offer promise as antiviral drugs with pharmacological properties similar to cyclosporine A. M pro is one of the best-characterized drug targets for coronaviruses [27e40]. M pro cleaves the initially translated viral polyprotein into its component proteins within cells infected by SARS-CoV-2. Cleavage generally occurs immediately after a Gln residue, and the Gln residue is typically preceded by a hydrophobic residue, most often Leu. The residue that follows the Gln is often a small amino acid, such as Ser, Ala, or Asn. M pro autolytically cleaves itself from the polyprotein [41]. Inhibiting M pro activity slows or halts viral replication, offering the promise of improved clinical outcomes for COVID-19 and other coronavirus diseases. Furthermore, there are no known human proteases with similar cleavage specificity to M pro , suggesting that it should be possible to develop inhibitors that target M pro without off-target toxicity.
Peptide-based inhibitors previously developed to target the SARS-CoV M pro have effectively been repurposed and modified to target the SARS-CoV-2 M pro d N3 from Jin et al., 13b from Zhang et al., and 11a and 11b from Dai et al. [16,42,43] These inhibitors block SARS-CoV-2 replication in cell-based studies, making them promising antiviral drug candidates. While the M pro inhibitors N3, 13b, 11a, and 11b have shown promise against inhibiting SARS-CoV-2 replication, additional M pro inhibitors will most likely be needed for their improved properties or to be used in combination therapies [44].
In this paper, we describe the design and in vitro evaluation of UCI-1 (University of California, Irvine Coronavirus Inhibitor-1), a first-in-class cyclic peptide that we hypothesized would inhibit the SARS-CoV-2 M pro , which is required for viral replication (Fig. 1). UCI-1 is a cyclic pentapeptide containing four amino acids from a M pro substrate constrained in a macrocycle linked together with a [4-(2-aminoethyl)phenyl]-acetic acid (AEPA) group to create a paracyclophane. This rigidified macrocycle is designed to mimic the conformation of a C-terminal autolytic cleavage site of a naturally occurring M pro substrate. Evaluation of UCI-1 in an in vitro M pro inhibition assay reveals that UCI-1 inhibits the SARS-CoV-2 M pro at mid-micromolar concentrations. LC/MS analysis indicates that UCI-1 resists cleavage by M pro , despite containing a scissile amide bond. Furthermore, UCI-1 is found to be non-toxic toward human embryonic kidney cells at concentrations that inhibit M pro [45].

Results
Design of UCI-1. We designed the cyclic peptide inhibitor UCI-1 based on the crystal structure of an inactive SARS-CoV M pro (C145A) variant with a 10 amino-acid C-terminal extension corresponding to the C-terminal prosequence of M pro (PDB 5B6O) ("M pro 316 ", Fig. 2) [46]. The SARS-CoV M pro amino acid sequence is 96% identical to the SARS-CoV-2 M pro amino sequence, and the three-dimensional structure of the SARS-CoV-2 M pro is highly similar to the structure of the SARS-CoV M pro [47]. In the M pro 316 crystal structure, C-terminal residues 301e310 (SGVTFQGKFK) extend into and complex with the active site of another M pro 316 molecule in an adjacent asymmetric unit (Fig. 2 inset). This complex reveals how the P2eP1eP1 0 -P2 0 -P3 0 positions (residues 305e309, FQGKF) of the C-terminal autolytic cleavage site fit into the active site of M pro 316 . We designed UCI-1 to mimic the conformation that the P2eP1eP1 0 -P2 0 -P3 0 residues adopt in the active site of M pro 316 . In the active site of M pro 316 , these residues adopt a "kinked" conformation in which the phenyl group of Phe309 at the P3 0 position points toward the backbone of Phe305 at the P2 position (Fig. 2 inset). To mimic this conformation, we envisioned linking the phenyl group of Phe309 to the backbone of Phe305 to create a macrocycle. To realize this design, we used the molecular visualization software PyMOL (version 2.2.2, Schr€ odinger) to build a model of the envisioned cyclic peptide by modifying Phe305 and Phe309 in the active site of M pro 316 (Fig. 3). In PyMOL, we deleted residues 301e304 to expose the amino group on Phe305; we also deleted residue 310 and the carbonyl of Phe309. We then connected the para position of Phe309 to the amino group of Phe305 with a CH 2 CO group to create a macrocycle. The newly created amino acid derived from Phe309 thus constitutes the amino acid AEPA.
We recognized that upon linking Phe309 and Phe305 as described above, Phe305 and Gln306 were poised to form a b-turn in which the carbonyl group of AEPA hydrogen bonds with the amino group of Gly307. We envisioned that b-turn formation in the cyclic peptide inhibitor would promote rigidity of the cyclic scaffold. To introduce additional conformational rigidity to the macrocycle, we also mutated Gly307 to serine, which is the most common residue at the P1' position among the 11 known SARS-CoV-2 M pro cleavage sites. The resulting cyclic peptide inhibitor UCI-1 was then further studied as described below. In an accompanying paper, we present a tutorial on the design process using free software [48]. Synthesis of UCI-1. We synthesized UCI-1 by Fmoc-based solidphase peptide synthesis of the protected linear peptide H 2 N-FQSK-AEPA-COOH on 2-chlorotrityl chloride resin, followed by cleavage of the linear protected peptide from the resin and subsequent solution-phase macrocyclization and global deprotection (Scheme 1). UCI-1 was purified using reverse-phase HPLC. The synthesis and purification proceeded smoothly on a 0.1 mmol scale and yielded 22 mg of purified UCI-1 as the TFA salt. Detailed procedures for synthesis of UCI-1 are given in the Materials and Methods section.
Enzyme inhibition assay. To evaluate whether UCI-1 inhibits the SARS-CoV-2 M pro , we used a fluorescence-based M pro inhibition assay with native M pro (Life Sensors) and the fluorogenic M pro substrate K(Dabcyl)-TSAVLQSGFRKM-E(EDANS)-NH 2 [28,29,49]. For the inhibition assay, we pre-incubated M pro with varying concentrations of UCI-1 (23.4e750 mM) in assay buffer for 30 min. We then added the fluorogenic M pro substrate and monitored M pro activity in a continuous kinetic assay with 360 nm excitation and 460 nm emission. In each well, the concentration of M pro was 0.1 mM and the concentration of the fluorogenic substrate was 50 mM. Fig. S1 shows the data from the continuous kinetic assay.
Initial rates for M pro activity in the presence or absence of UCI-1 were obtained by fitting a straight line to the linear portions of the curves from the continuous kinetic assay. The % activity was calculated by comparing the initial rate at each concentration to the initial rate in the absence of inhibitor. Non-linear regression was then used to fit the data and estimate an IC 50 of ca. 160 mM for UCI-1 ( Fig. 5A).
To better understand the significance of inhibition of M pro by UCI-1, we synthesized and tested two control peptides derived from UCI-1 d the acyclic analog of UCI-1 "peptide-1a" (Fig. 5D) and the diastereomer "peptide-1b" (Fig. 5E) d that were predicted to not inhibit M pro . Peptide-1a exhibits little or no inhibition at concentrations at or below 750 mM (Fig. 5B). This finding indicates that the cyclic structure of UCI-1 is important for its activity. There is a slight reduction in rate of cleavage of the fluorogenic substrate upon addition of 1500 mM peptide-1a. This reduction in rate may reflect either slight inhibition of M pro or that the acyclic peptide acts as a competitive substrate at high concentrations. Peptide-1b exhibits little or no inhibition at concentrations at or below 1500 mM (Fig. 5C). The inactivity of peptide-1b suggests that the shape and stereochemistry of UCI-1 are critical for fitting into and blocking the active site of M pro , and that inhibition by UCI-1 does not simply result from its hydrophobicity and cyclic structure.
Conformational analysis, docking, and molecular dynamics simulations of UCI-1. To better understand the three-dimensional structure of UCI-1 and how it could interact with the active site of the SARS-CoV-2 M pro , we performed conformational analysis and docking studies. Conformational searching of UCI-1 (MacroModel with the MMFFs force field and GB/SA water) revealed that UCI-1 adopts a global minimum energy conformation that resembles the kinked conformation that residues 305e309 adopt in the active site of M pro 316 (Figs. 2 and 4A). In the global minimum energy conformation, the AEPA residue acts as a rigid spacer, with Phe, Gln, Ser, and Lys forming a bridge. As we had envisioned, the Phe and Gln residues adopt a b-turn conformation, with Phe at the iþ1 position and Gln at the iþ2 position. The Phe side chain is well situated to fit into the S2 pocket, and the Gln side chain is well situated to fit into the S1 pocket. The Ser, Lys, and AEPA residues, in turn, are poised to occupy the S1 0 , S2 0 , and S3' pockets.
The cyclic peptide appears to be particularly rigid. In the conformational search, the peptide backbones of most of the conformers adopt the conformation described above, differing only in side chain geometry and the type of b-turn formed by Phe305 and Gln306 (Fig. 4B). To evaluate how UCI-1 might fit in the active site of M pro , we docked UCI-1 with the X-ray crystallographic structure of the SARS-CoV-2 M pro PDB 6YB7 using Autodock Vina [50]. This initial docking study revealed that UCI-1 fits into the active site of the SARS-CoV-2 M pro in the envisioned manner (Fig. 4C).
To further evaluate how UCI-1 may achieve inhibition of M pro , we performed a complementary docking experiment on UCI-1, and then performed a molecular dynamics simulation on the docked structure. In this study, we first docked UCI-1 with an equilibrated version of the X-ray crystallographic structure of the SARS-CoV-2 M pro dimer PDB 6Y2E, which has previously been simulated in explicit water [51]. Next, we chose a low-energy configuration of UCI-1 bound to an equilibrated crystal structure of the M pro dimer PDB 6Y2E that best matched the docked structure observed in the initial docking experiment with PDB 6YB7. Finally, we performed 10 ns of molecular dynamics simulation of the docked structure using the CHARMM36 force field and TIP3P water at 310 K. This procedure was then repeated for peptide-1a and peptide-1b to gain insight into why these control peptides do not substantially inhibit M pro .
In the UCI-1 docking experiment, the side chains of UCI-1 occupy the active site M pro in the envisioned manner, where Phe occupies the S2 pocket, Gln occupies the S1 pocket, Ser occupies the S1 0 pocket, Lys occupies the S2 0 pocket, and the AEPA group occupies the S3 0 pocket (Fig. S2). Over the course of the simulation, UCI-1 stays bound to the active site and the amino acids maintain their positions in the pockets of the active sitedonly the Lys side chain appears to leave its respective pocket over the 10 ns simulation (Fig. S3). In contrast, the control peptides explore more of the active site and their side chains do not remain in their respective pockets as frequently. Notably, the D-Phe side chain of peptide-1b does not stay in the S2 pocket, and at times travels towards the S3' pocket, causing the peptide to lift away from the active site. The linear control, peptide-1a, does not appear to bind well in any of the pockets.
Stability of UCI-1 toward M pro . The amide bond between residues at positions P1 and P1' of UCI-1 has the potential to be cleaved by M pro , because residues at these positions correspond to the cleavage site of M pro substrates. To determine whether M pro cleaves UCI-1, we used LC/MS to analyze UCI-1 in the well solution from the 96well plate of the enzyme inhibition assay after 4 h. To aid in LC/MS identification of the cleavage product of UCI-1, we synthesized peptide-1c, which would correspond to the authentic UCI-1 M pro cleavage product. We then spiked the well solution with peptide-1c to identify and determine the limits of detection of the putative cleavage product.
UCI-1 is not appreciably cleaved by M pro (Fig. S2). LC/MS analysis of 5 mL of the 187.5 mM UCI-1 enzyme inhibition assay well solution spiked with 20 mM peptide-1c shows a peak at 2.2 min that corresponds to peptide-1c (Fig. S2B). The peak at 2.2 min is absent in the LC/MS trace acquired before spiking the well solution with peptide-1c (Fig. S2A), indicating that there is little to no cleavage of UCI-1 by M pro . The ion current at 2.2 min in the LC/MS trace acquired before spiking with peptide-1c shows no significant evidence of the UCI-1 cleavage product (Fig. S2C), whereas the ion current at 2.2 min in the LC/MS trace acquired after spiking with peptide-1b shows the mass of peptide-1c (Fig. S2D), providing additional evidence that UCI-1 resists cleavage by M pro . These findings confirm that UCI-1 acts as an inhibitor and not a competitive substrate. The resistance of UCI-1 to cleavage by M pro is consistent with the observations by others that cyclic peptides often resist proteolytic cleavage [52e57]. Cytotoxicity of UCI-1. To evaluate whether UCI-1 is cytotoxic, we exposed human embryonic kidney (HEK-293) cells to varying concentrations of UCI-1 (0e256 mM) for 72 h, and then assessed cell death using a lactate dehydrogenase (LDH) release assay. At the highest concentration evaluated, UCI-1 elicits little or no LDH release from HEK-293 cells, indicating that UCI-1 is not cytotoxic at concentrations up to 256 mM (Fig. S3).

Discussion
UCI-1 is noteworthy, because it represents a direct creation of a cyclic peptide inhibitor of M pro from the crystal structure of a linear peptide substrate. UCI-1 is created from the linear peptide substrate by the addition of only a single methylene group to the para position of Phe309 that is then linked to the backbone carbonyl of Phe305 to create a rigid peptide macrocycle linked together by AEPA. In synthesizing UCI-1, the cyclization proceeds smoothly, without formation of detectable oligomers, uncyclized linear peptide, or side products. The apparent rigidity of UCI-1 demonstrates an emergent property of the AEPA linker, whereby a short strand of amino acids can span across the AEPA linker to create a macrocycle that adopts a b-turn structure and a doughnut-like shape. To further evaluate the utility of the AEPA linker, we designed a homologue of UCI-1 in which we replaced the rigid AEPA linker with a flexible tri-alanine linker (tri-Ala-UCI-1) and performed conformational searching of this homologue (MacroModel with the MMFFs force field and GB/ SA water). Like UCI-1, tri-Ala-UCI-1 is a 21-membered macrocycle.
Conformational searching of tri-Ala-UCI-1 reveals that the global minimum energy conformation of tri-Ala-UCI-1 adopts a compact closed structure, rather than an open structure (Fig. 6). While UCI-1 contains an open ring with a stabilizing b-turn structure and a "kinked" conformation that resembles the conformation of the linear peptide M pro substrate, tri-Ala-UCI-1 lacks these features. Because the AEPA amino acid replaces a tripeptide and enforces an open structure, it is a useful template for inducing a well-defined conformation in the spanning tetrapeptide [58,59].
UCI-1 is a paracyclophane containing four a-amino acids, in addition to AEPA. Conformationally constrained paracyclophane macrocycles containing AEPA and two, three, or five a-amino acids can also be envisioned. Molecular modeling of homologous AEPA macrocycles derived from UCI-1 with two, three, and five amino acids demonstrates that all the macrocycles adopt doughnut-like open structures in which the amino acids are extended above the AEPA linker (Fig. 7). In additional modeling studies, we have found that AEPA macrocycles that contain amino acids other than glycine adopt this open doughnut-like structure. Macrocycles containing only two a-amino acids are strained to the point of bending the AEPA unit out of planarity d a common property of smaller paracyclophanes (Fig. 7A) [60]. Nevertheless, we were able to synthesize cyclo(AEPA-FQ) and found that the cyclization proceeded smoothly, without the formation of detectable oligomers, uncyclized linear peptide, or side products.

Conclusion
In spite of the rapid development of vaccines to combat COVID-19, antiviral drugs are still needed. Antiviral drugs that slow or halt viral replication can lead to a shortened time to recovery from COVID-19, offering the promise of improved mortality rates and alleviating the tremendous strain experienced by hospitals during the COVID-19 pandemic [61]. Antiviral drugs could also be used as first line of defense in a future coronavirus outbreak or pandemic, or even as a prevention.
The design of the UCI-1 demonstrates that cyclic peptides that mimic the conformation of linear peptide substrates of M pro can be developed as inhibitors against M pro . Almost all of the M pro inhibitors that have been reported thus far are linear peptides that contain a "warhead" that forms a covalent bond with the active cysteine of M pro . While the activity of UCI-1 is modest compared to other known M pro inhibitors, UCI-1 lays the groundwork for developing additional cyclic peptide inhibitor analogues with improved activity against M pro . Design and development of the next generation of UCI-1 analogues will likely produce better inhibitors, and we anticipate that the studies described here will pave the way for the design of more potent cyclic peptide inhibitors of M pro with potential to be developed into drugs. We also anticipate that the AEPA unit will prove useful in the design of other conformationally constrained cyclic peptides with potential pharmacological applications.
The Fmoc-AEPA-2-chlorotrityl resin was transferred to a peptide synthesis coupling vessel and subjected to cycles of peptide coupling with Fmoc-protected amino acid building blocks. The linear peptide was synthesized from the C-terminus to the N-terminus. Each coupling cycle consisted of i. Fmoc-deprotection with 20% (v/v) piperidine in DMF for 5 min (2x), ii. washing with DMF (3x), iii. coupling of the amino acid (5 equiv) in the presence of HCTU (4.5 equiv) and 20% (v/v) N-methylmorpholine (2,4,6collidine) in DMF for 10e20 min. iv. washing with DMF (3x). After coupling of the last amino acid, the terminal Fmoc group was removed with 20% (v/v) piperidine in DMF. The resin was transferred from the coupling vessel to a Bio-Rad Poly-Prep chromatography column.
The linear peptide was cleaved from the resin by agitating the resin for 1 h with a solution of 1,1,1,3,3,3-hexafluoroisopropanol (HFIP) in CH 2 Cl 2 (1:4, 7 mL) [62]. The suspension was filtered and the filtrate was collected in a 250-mL round-bottomed flask. The resin was washed with additional HFIP in CH 2 Cl 2 (1:4, 7 mL) and then with CH 2 Cl 2 (2 Â 10 mL). The combined filtrates were concentrated by rotary evaporation to give a white solid. The white  Synthesis of peptide-1a. Rink amide AM resin (300 mg, 0.68 mmol/g) was added to a peptide synthesis coupling vessel. The resin was suspended in dry DMF (10 mL) and allowed to swell for 30 min. The solution was drained from the resin and the Fmoc group was removed with 20% (v/v) piperidine in DMF for 20 min. The solution was drained and washed with DMF (5x). Fmoc-AEPA-OH (0.50 equiv, 40.8 mg, 0.102 mmol), HATU (0.5 equiv), and HOAt (0.5 equiv) in 20% (v/v) 2,4,6-collidine in dry DMF (8 mL) was then added to the resin mixed for 12 h. The solution was then drained, washed with DMF (3x) and a mixture of acetic anhydride/pyridine (3:2, 10 mL) was added. The resin was mixed for 15 min to cap the unreacted resin sites. The resin was then washed with DMF (3x). This procedure yielded 0.17 mmol/g of loaded resin.
The Fmoc-AEPA-Rink amide AM resin was subjected to cycles of peptide coupling with Fmoc-protected amino acid building blocks as described above, ending with cleavage of the terminal Fmoc. The resin was then washed with CH 2 Cl 2 (3x) and then dried by pushing N 2 gas through the Poly-Prep column. The peptide was cleaved from the resin and globally deprotected by mixing the dried resin with TFA/triisopropylsilane (TIPS)/H 2 O (18:1:1, 10 mL) and gently rocking for 2.5 h. The peptide was drained into a glass beaker, precipitated in cold ether, and subjected to purification by RP-HPLC as described above. The synthesis of peptide-1a yielded 17 mg (40.4% yield based on resin loading) of the peptide as the TFA salt.
Synthesis of peptide-1c. 2-Chlorotrityl chloride resin (300 mg, 1.6 mmol/g) was added to a Bio-Rad Poly-Prep chromatography column (10 mL). The resin was suspended in dry CH 2 Cl 2 (10 mL) and allowed to swell for 30 min. The solution was drained from the resin and a solution of Fmoc-Gln(Trt)-OH (0.50 equiv, 146.57 mg, 0.24 mmol) in 6% (v/v) 2,4,6-collidine in dry CH 2 Cl 2 (8 mL) was added immediately and the suspension was gently agitated for 12 h. The solution was then drained and a mixture of CH 2 Cl 2 /MeOH/N,Ndiisopropylethylamine (DIPEA) (17:2:1, 10 mL) was added immediately. The mixture was gently agitated for 1 h to cap the unreacted 2-chlorotrityl chloride resin sites. The resin was then washed with dry CH 2 Cl 2 (2x) and dried by passing nitrogen through the vessel. This procedure yielded 0.36 mmol/g of loaded resin.
The Fmoc-Gln(Trt)-2-chlorotrityl resin was subjected to cycles of peptide coupling with Fmoc-protected amino acid building blocks as described above, ending with cleavage of the terminal Fmoc. The resin was then washed with CH 2 Cl 2 (3x) and then dried by pushing N 2 gas through the Poly-Prep column. The peptide was cleaved from the resin and globally deprotected by mixing the dried resin with TFA/triisopropylsilane (TIPS)/H 2 O (18:1:1, 10 mL) and gently rocking for 2.5 h. The peptide was drained into a glass beaker, precipitated in cold ether, and subjected to purification by RP-HPLC as described above. The synthesis of peptide-1c yielded 13 mg (13.6% yield based on resin loading)of the peptide as the TFA salt.
Synthesis of H 2 NeK(Dabcyl)-TSAVLQSGFRKM-E(EDANS)-NH 2 . Rink amide AM resin (300 mg, 0.68 mmol/g) was added to a peptide synthesis coupling vessel. The resin was suspended in dry DMF (10 mL) and allowed to swell for 30 min. The solution was drained from the resin and the Fmoc group was removed with 20% (v/v) piperidine in DMF for 20 min. The solution was drained and washed with DMF (5x). Fmoc-Glu(EDANS)-OH (1.5 equiv, 189 mg, 0.306 mmol), HATU (1.5 equiv), and HOAt (1.5 equiv) in 20% (v/v) 2,4,6-collidine in dry DMF (8 mL) was then added to the resin mixed for 12 h. The solution was then drained, washed with DMF (3x) and a mixture of acetic anhydride/pyridine (3:2, 10 mL) was added. The resin was mixed for 15 min to cap the unreacted resin sites. The resin was then washed with DMF (3x). This procedure yielded 0.48 mmol/g of loaded resin.
Fmoc-Glu(EDANS)-Rink amide AM resin was subjected to cycles of peptide coupling with Fmoc-protected amino acid building blocks as described above. For coupling Fmoc-Lys(Dabcyl)-OH, Fmoc-Lys(Dabcyl)-OH (2 equiv, 180.3 mg, 0.408 mmol), HATU (2 equiv), and HOAt (2 equiv) in 20% (v/v) 2,4,6-collidine in dry DMF was added to the resin and mixed for 24 h. The resin was then washed with DMF (3x) and the terminal Fmoc was removed with 20% (v/v) piperidine in DMF for 20 min. The resin was then washed with DMF (3x) followed by CH 2 Cl 2 (3x) and then dried by pushing N 2 gas through the Poly-Prep column. The peptide was cleaved from the resin and globally deprotected by mixing the dried resin with TFA/triisopropylsilane (TIPS)/H 2 O (18:1:1, 20 mL) in a round bottom flask and stirring for 2.5 h. The peptide was drained into a glass beaker, precipitated in cold ether, and subjected to purification by RP-HPLC as described above. The synthesis of H 2 NeK(Dabcyl)-TSAVLQSGFRKM-E(EDANS)-NH 2 yielded 42 mg (12.6% based on resin loading) of the peptide as the TFA salt.
Enzyme inhibition assay. A proprietary assay buffer containing detergent was used for the inhibition assays (BPS Bioscience, catalog # 79956). The substrate with the cleavage site of M pro (indicated by the arrow, Y), H 2 NeK(Dabcyl)-TSAVLQYSGFRKM-E(EDANS)-NH 2 , was used in the fluorescence resonance energy transfer (FRET) based continuous kinetic assay, using a black-walled 96-well plate. The emergence of EDANS fluorescence due to the cleavage of the substrate by M pro was monitored at 460 nm, with excitation at 360 nm, using a Varioskan LUX fluorescence spectrophotometer (Thermo Fisher) using the top-read mode at 37 C. Stock solutions (20 mg/mL) of UCI-1, peptide-1a, and peptide-1b were prepared gravimetrically by dissolving the peptides in deionized H 2 O and then diluted in assay buffer to create 10X solutions for each concentration in a serial dilution. For the determination of the IC 50  LDH release assay. The LDH release assay was performed using the Pierce LDH Cytotoxicity Assay Kit from Thermo Scientific. Experiments were performed in replicates of five, and an additional 10 wells were used for controls. Cells were cultured in the inner 60 wells (rows BeG, columns 2e11) of the 96-well plate. DMEM:F12 media (100 mL) was added to the outer wells (rows A and H and columns 1 and 12), in order to ensure the greatest reproducibility of data generated from the inner wells. A 10-mg/mL stock solution of UCI-1 was prepared gravimetrically in sterile deionized H 2  Docking of UCI-1 to SARS-CoV-2 M pro . The model of the SARS-CoV-2 M pro was generated as follows: Starting coordinates of SARS-CoV-2 M pro were generated from the SARS-CoV-2 M pro crystallographic structure (PDB 6YB7) using PyMOL and were saved as a new PDB file. In PyMOL, the dimethyl sulfoxide molecule that sits in the active site of 6YB7 was deleted. A minimum energy (global minimum) structure of UCI-1 was generated by conformational searching in MacroModel (MMFFs force field in GB/SA water). iBabel was used to convert the PDB file generated in MacroModel into a PDBQT file prior to docking. Docking was performed using AutoDock Tools and AutoDock Vina. In AutoDock Tools, a grid was chosen to encompass the active site of SARS-CoV-2 M pro in the size of 25x25 Â 25 Å and with the coordinates x ¼ 9.250, y ¼ À5.944, z ¼ 18.944. SARS CoV-2 M pro was treated as a rigid receptor in these calculations. The lowest energy cluster, as determined by AutoDock Vina, was chosen to represent the docking model shown in Fig. 7C.
Molecular Dynamics Simulation. The initial SARS-CoV-2 M pro model that was used for MD simulations was generated from the final frame of the published wild type M pro trajectory [50], which was originally seeded with PDB structure 6Y2E (see reference for details). Initial peptide conformations were obtained by docking UCI-1, peptide-1a, and peptide-1b respectively to the A chain active site of the initial M pro model using AutoDock Vina [49]; the search box in each case was given a radius equal to 10 times the peptide's radius of gyration, centered on residue M165 (chosen because of its central location in the conserved substrate binding region, as confirmed by both visual inspection of the structure and the analysis of Jin et al. (2020)) [31], and an exhaustiveness of 32 was employed. Out of the 25 minimum energy poses, the pose most closely resembling the conformation found when docking against PDB 6YB7 was retained for use as the initial condition. CHARMM representations of the peptide models were created using CGenFF version 3.01 [63,64,65,66]. The combined peptide/dimer structures were solvated in TIP3P water [67] using VMD version 1.9.3; [68] a cubic water box with minimum margin of 10 Å from the protein complex on each side was employed, with NaCl added to achieve neutral charge. Each complex was subsequently minimized for 10,000 iterations using namd 2.13 [69] with the CHARMM 36 m force field [70], followed by a 10 ps simulation period for box adjustment. (All simulations were performed using periodic boundary conditions with an NpT ensemble at 310 K and 1 atm, Langevin dynamics with a damping coefficient of 1/ps and a Nos e-Hoover Langevin piston 71,72 were respectively employed for temperature and pressure control. Rigid bonds were used only for water molecules, and a 1 fs step size was employed.) Following adjustment, each complex was simulated for 10 ns, with 1,000 frames being retained for subsequent analysis (1/ps). Visualization and analysis of the resulting trajectories was performed with VMD.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.