CGM23 corresponds to a pan-coronavirus lipopeptide inhibitor potently inhibiting virion fusion

Summary

We developed a pan-coronavirus lipopeptide fusion inhibitor, CGM23, which binds to the highly conserved spike heptad repeat-1 domain, thus interrupting the formation of the six-helix bundle required for membrane fusion. In vitro, CGM23 potently inhibited infection by all human coronaviruses tested, including SARS-CoV, MERS-CoV, and SARS-CoV-2 and the seasonal coronaviruses. CGM23 is based on the amino acid sequence of SARS-CoV-2 spike protein. Computational modeling, 4-phenylbutanoic acid was appended to the N-terminus to mimic interactions with the hydrophobic pocket adjacent to the heptad repeat-2. A C-terminal addition of palmitic acid via a PEG linker enhanced CGM23’s fusion inhibition. In a wild-type mouse model infected with mouse-adapted SARS-CoV-2, intranasal CGM23 prevented lung infection when given 30 min before challenge. Furthermore, therapeutic CGM23 8 h post-inoculation significantly reduced viral loads in the lungs. In summary, CGM23 represents a promising pan-coronavirus inhibitor with both potential therapeutic and prophylactic activity in humans.

Graphical abstract

Highlights

• •CGM23 is an S2-targeting lipopeptide fusion inhibitor active against human coronaviruses
• •CGM23 shows broad antiviral activity across diverse human coronaviruses
• •CGM23 retains α-helicity and stability, with a 7-h half-life after intranasal delivery
• •Intranasal CGM23 provides prophylactic and therapeutic protection against SARS-CoV-2 in vivo

Pharmacology; Biochemistry; Virology

Introduction

Three major human outbreaks by different but highly pathogenic coronaviruses have occurred in the past 20 years: severe acute respiratory syndrome (SARS)¹ in 2003, middle east respiratory syndrome (MERS)² in 2012 and coronavirus disease 2019 (COVID-19)³ in 2019. In addition, four other coronaviruses with reduced pathogenicity have become endemic in humans and are referred to as the seasonal coronaviruses: HCoV-OC43, HCoV-HKU1, HCoV-NL63, and HCoV-229E.⁴ Coronaviruses are widely distributed in many mammals, including a notably large reservoir in bats. These bat viruses may “spill over” into the human population either by infecting an intermediate host or by directly jumping into humans due to genetic and structural conservation of their receptors.^5,6,7 The recent global COVID-19 pandemic caused by SARS-CoV-2 is a poignant example of the potential pathogenic consequences of coronavirus spillover in humans.^8,9 SARS-CoV-2 RNA vaccines proved pivotal in protecting the human population from COVID-19-associated hospitalization and death. However, the ability of these vaccines to inhibit infection became increasingly transient with the emergence of variants with immunoevasive properties, prompting recommendations for administration of follow-on boosters as frequently as every four months in high-risk individuals.^10,11 In particular, the Omicron family of variants has proven remarkably prone to evading vaccine-acquired immunity and causing breakthrough infections.^12,13 Two doses of the first generation SARS-CoV-2 spike vaccine mostly fail to suppress symptomatic infection by Omicron,¹⁴ and only modestly reduce the risk of long COVID.¹⁵ Thus, considering the repeated nature of coronavirus zoonoses and the vulnerability of humans to these viruses, development of inhibitors with pan-coronavirus activity is a high priority.

The fusion of virion and host cell membranes plays a key and highly conserved role in the SARS-CoV-2 life cycle.^16,17,18,19 Fusion is mediated by the spike protein decorating the surface of viral particles. Spike is partitioned into two subunits, S1 and S2. A receptor-binding domain (RBD) present in S1 engages specific membrane receptors such as ACE2, leading in turn to a major conformational change in S2 that results in the insertion of the fusion peptide into the plasma membrane.²⁰ The conformational change of the S2 subunit following the binding of S1 to ACE2 involves cleavage at the S2’ site by a host protease. TMPRSS2, a type II transmembrane serine protease, facilitates this cleavage and enables early viral entry at the plasma membrane.^21,22 Alternatively, SARS-CoV-2 entry can also occur after endocytosis (the late entry pathway) involving lysosomal cathepsin L/B cleavage of the S2 protein.^23,24,25 Fusion critically depends on the creation of the fusion peptide, a six-helix bundle (6-HB) resulting from the interaction of heptad repeat 2 (HR2) helices and heptad repeat 1 (HR1) domains in S2.¹⁶ While the RBD of S1 is often targeted by neutralizing antibodies, it undergoes frequent mutations rendering the antibodies inactive.²⁶ In sharp contrast, the sequence of the HR domains of S2 is highly conserved among all coronaviruses, and inhibitors targeting these components of the fusion apparatus are predicted to inhibit infection by a broad, perhaps full range of coronaviruses.²⁷

Peptide-based fusion inhibitors targeting the HR1 domain have been reported.^{28,29,30,31,32} The first fusion inhibitor, EK1, utilized peptide sequences from the HR2 domain of HCoV-OC43 and displayed broad anti-coronavirus activity.³³ Of note, the addition of cholesterol (EK1C4)³⁰ or palmitic acid (EK1-C16)²⁸ at its C-terminus sharply enhanced its antiviral activity.^28,30 The mechanism by which lipidation increases fusion inhibition remains uncertain,³⁴ but might involve peptide anchoring in the cell membrane. A structure-activity relationship (SAR) analysis of peptides of varying lengths based on the HR2 domain of SARS-CoV-2 found that the degree of α-helicity is a key determinant of antiviral efficacy.²⁹ Through lead peptide screening, this study showed that IPB02, a peptide with cholesterol attached to the C-terminus of a 36-amino-acid HR2-derived sequence, exhibited potent antiviral effects against SARS-CoV-2 variants.²⁹ Furthermore, two variants of IPB02 (IPB02V2 and IPB02V3), modified by the addition of hydrophilic amino acids to the N-terminus, increased viral suppression by significant recovery of α-helicity.³² On the other hand, another study demonstrated that extending the N-terminus of a peptide derived from the SARS-CoV-2 HR2 domain by six amino acids (longHR2_42), outside the HR2 helical region, significantly enhanced viral inhibitory activity in an in vitro system using SARS-CoV-2 live viruses.³¹ Building on these strategies, the noveler HR2 lipopeptide IPB29^35,36 was developed by linking an optimized HR2-derived sequence to cholesterol through a rigid, helix-facilitating EAAAK linker, which increases α-helical content, stability and breadth of activity against coronaviruses, and has progressed to phase 2/3 clinical evaluation as an inhaled therapy for COVID-19. In parallel, the next-generation HR1-targeting lipopeptide P315V3,³⁷ incorporating an N-terminally extended HR1-binding segment and a PEG4-cholesterol modification, exhibits nanomolar inhibitory activity against diverse SARS-CoV-2 variants and other coronaviruses in vitro, provides robust prophylactic protection in mice following intranasal administration, and is currently being assessed in a phase 2 clinical trial as an intranasal pan-coronavirus fusion inhibitor. Peptide-based fusion inhibitors also display preventive activity. For instance, EK1C4 inhibited SARS-CoV-2 infection in mice when administered 30 min before SARS-CoV-2 inoculation.³⁸ Similarly, when administered 2 h before viral challenge, a SARS-CoV-2 HR2-derived peptide with bifunctional lipid modifications ([SARS_HRC-PEG₄]₂-chol) inhibited airborne viral infection in four of six infected ferrets.^39,40 Like EK1C4, IPB02s,²⁹ longHR2_42,³¹ IPB29, and P315V3, this peptide binds to the HR1 domain and antagonizes 6-HB formation, leading to inhibition of fusion. In addition to their promising antiviral properties, peptide fusion inhibitors are characterized by lower immunogenicity and lower manufacturing costs than biological products such as monoclonal antibodies,^38,41,42,43 further justifying their evaluation as potential pan-coronavirus therapeutics.

In this study, we sought to identify fusion inhibition peptides with properties that would increase their clinical utility, including: (1) high anti-SARS-CoV-2 activity, (2) high pan-coronavirus activity, (3) high stability, (4) low toxicity, (5) broad resistance to proteases, and (6) a potential for both therapeutic and preventive activities. We now describe a lipopeptide pan-coronavirus fusion inhibitor, CGM23, that potently inhibits infection by divergent human coronaviruses, including SARS-CoV-2 (wild-type (Wuhan-Hu-1), Alpha, Kappa, Delta, and Omicron (BA.1)), SARS-CoV, MERS-CoV, HCoV-NL43, and HCoV-229E. As a result of a combination of optimized peptide modifications, CGM23 displays both high helicity and stability as well as broad protease resistance. Because of their properties, we suggest that CGM23 or related peptides merit careful study as a potential defense against future outbreaks of zoonotic coronavirus infections in humans.

Results

Screening for lipidated peptides that potently inhibit cell-cell fusion mediated by SARS-CoV-2 spike-human ACE2 binding

To identify peptides that could exert potent inhibitory effects on viral fusion, we employed computational methods to rationally design such peptides and tested 159 candidates in quantitative cell-cell fusion assays and syncytia formation assays (Figure 1A). The peptides were classified into three categories based on the structure of the backbone: (1) peptides based on EK1 (CG-series, EK1, and EK1C4), (2) peptides based on SARS-CoV-2 HR2 (CGM-series), and (3) peptides with a hybrid of these two constructs as their backbone (CGMA-series) (Figures 1A and S1–S3). For the syncytia assay, 293T cells stably expressing GFP and SARS-CoV-2 wild-type (Wuhan-Hu-1) spike were co-cultured with Calu-6 cells stably expressing RFP and ACE2 in the presence and absence of each peptide. Fusion was quantified as the number of double-positive (GFP+ and RFP+) cells using fluorescence microscopy, and the peptides compared based on their half-maximal inhibitory concentration (IC₅₀) values. A scrambled EK1 peptide was employed as a negative control; as expected, this peptide exhibited no concentration-dependent reduction in syncytia formation (Figure 1A (Ⅰ)). In contrast, CGM23, which incorporates the SARS-CoV-2 HR2 backbone with 4-phenylbutanoic acid (PBA) at the N-terminus and palmitic acid at the C-terminus, induced a marked, concentration-dependent decrease in syncytia formation (Figure 1A(Ⅱ)). Among the 25 SARS-CoV-2 HR2-based peptides, 11 displayed IC₅₀ values less than 10 nM in the syncytia assay, while only 4 (including EK1C4) of the 131 EK1-based peptides exhibited this level of potency (Figures 1B and S1–S3). All three peptides with a hybrid EK1 and SARS-CoV-2 HR2 backbone had IC₅₀’s above 10 nM (Figure S3). The IC₅₀ of EK1C4 was 3.0 nM; peptides with moderately lower values included CGM19 (2.0 nM), CGM20 (2.1 nM), CGM21 (2.1 nM), CGM23 (1.9 nM), and CGM39 (3.0 nM), all containing lipid modifications at the C-terminus (Figure 1B). Since EK1C4 has been previously reported^30,39 and consistently showed potent fusion inhibitory activity in our assays, we used it as a benchmark to compare peptides used in this study. EK1C4, CGM19 (previously reported as SARS-CoV-2 HR2-chol³⁹), and CGM20 contain C-terminal cholesterol modifications, while CGM21, CGM23, and CGM39 contain palmitic acid modifications (Figure 1B). The peptide backbone of CGM19, CGM20, and CGM21 was identical to the SARS-CoV-2 HR2 sequence. In contrast, CGM23 features three N-terminal amino acid substitutions (Asp1Ser, Ser3Asp, and Gly4Gln) within the SARS-CoV-2 HR2 sequence and a 4-phenylbutanoic acid (PBA) cap. These substitutions are derived from the EK1C4 peptide. Additionally, CGM39 incorporates a hydrophobic staple motif into the structure of CGM23 and modifies the first amino acid substitution to include the D form of serine.

Figure 1.

Peptide design and evaluation

(A) Schematic representation of SARS-CoV-2 spike protein and amino acid sequences of HR2 peptides from SARS-CoV-2 and EK1 (HCoV-OC43 HR2 derived peptide). N, N-terminus; C, C-terminus; S1/S2, cleavage site at S1/S2 boundary; RBD, receptor-binding domain; HR1, heptad repeat 1; HR2, heptad repeat 2; HR2P, heptad repeat 2 peptide.

Syncytia assay. Left: GFP- and spike-expressing 293T cells were co-cultured with RFP- and ACE2-expressing Calu-6 cells for 16 h in the presence or absence of peptides. Center: double-positive cells (white arrows) indicative of syncytia formation are frequent in the presence of the scrambled EK1 peptide (Ⅰ) but not in the presence of CGM23 (Ⅱ) (100 nM). Right: quantification of syncytia formation in the presence of the scrambled EK1 peptide (top) or CGM23 (bottom) relative to mock treatment.

(B) Sequence, N- and C-terminal modifications and IC₅₀ of the 15 peptides with IC₅₀s < 10 nM in the syncytia assay. CoV-2, SARS-CoV-2; N-term, N-terminus; C-term, C-terminus. IC₅₀ data are means of samples from a representative experiment. Ac, acetylation; PPA, 4-phenylpropanoic acid; PBA, 4-phenylbutanonic acid.

(C) Correlation between IC₅₀ values in the pseudotyped SARS-CoV-2 spike virion assay and live SARS-CoV2 virus infection assay for the 15 peptides with IC₅₀ values below 10 nM in the syncytia assay. Statistical analysis was performed using Spearman’s rank test.

(D and E) Dose-dependent inhibitory activity of CGM23 and EK1C4 relative to CG167 (EK1 scrambled peptide with EK1C4 lipidation) in the pseudotyped SARS-CoV2 spike virion assay (D) and live SARS-CoV2 infection assay (E). IC_50, half-maximal inhibitory concentration.

To examine the impact of the 15 peptides with an IC₅₀ below 10 nM in the syncytia assay on viral entry, we performed infection experiments with pseudotyped SARS-CoV-2 spike virions and SARS-CoV-2 live virus (Figure 1C). The IC₅₀ values in the pseudotyped spike virion assay and the live virus infection assay were significantly correlated (Spearman’s r of 0.69, p = 0.005), and several peptides had IC₅₀ values lower than EK1C4’s (383.0 nM, 48.7 nM) in both assays, including CGM19 (31.6 nM, 5.0 nM), CGM20 (33.0 nM, 11.2 nM), CGM23 (115.0 nM, 1.3 nM), and CGM39 (170.0 nM, 13.0 nM) (Figures 1C–1E). Since CGM23 exhibited the strongest antiviral activity in the live virus infection assay, and a 37-fold higher potency than EK1C4 (Figure 1E), we evaluated its potential cytotoxic effects using an adenosine triphosphate (ATP) cell viability luciferase assay.⁴⁴ CGM23 induced no cytotoxicity except at 20 μM, the highest concentration tested (Figure S4C), confirming that its potent inhibition of infection at lower concentrations is not mediated by the disruption of target cells. CG167, a scrambled version of EK1 with the same cholesterol modification as EK1C4 (Figure S2) was employed as a negative control in these experiments.

CGM23 has antiviral activity against a range of different coronaviruses

To assess CGM23’s impact on the entry of other human coronaviruses, we generated pseudotyped virions expressing the spike proteins of SARS-CoV, MERS-CoV, and the HCoV-229E and HCoV-NL63 seasonal coronaviruses. The IC₅₀ of CGM23 was 2-fold lower than that of EK1C4 against HCoV-229E (96.4 nM vs. 213.3 nM) and 7-fold lower than that of EK1C4 against HCoV-NL63 (121.8 nM vs. 840.9 nM) (Figures 2A, 2B, S5A and S5B). The difference was more modest (1.3-fold lower IC₅₀ for CGM23 than EK1C4) against SARS-CoV pseudotyped virions (144.9 nM vs. 187.2 nM) and MERS-CoV pseudotyped virions (92.5 nM vs. 120.4 nM) (Figures 2C, 2D; S5A and S5B).

Figure 2.

Antiviral activity of CGM23 against coronaviruses other than SARS-CoV-2 and SARS-CoV-2 variants

(A–D) Inhibitory activity of CGM23 and EK1C4 in the pseudotyped spike virion assay against HCoV-229E (A) HCoV-NL63 (B), SARS-CoV (C), and MERS-CoV (D). CG167, EK1 scrambled peptide with EK1C4 lipidation.

(E and F) Inhibitory activity of EK1C4 (E) and CGM23 (F) against SARS-CoV-2 variants in the pseudotyped spike virion assay. IC_50, half-maximal inhibitory concentration. Ambig, ambiguous.

We next compared the activity of CGM23 and EK1C4 against a range of SARS-CoV-2 variants. The viral variants included Alpha, Delta, and Omicron, which harbor substitutions in the HR1domain relative to the original Wuhan strain, and Kappa, which harbors substitutions in the connector domain between HR1 and HR2 (Methods S1). CGM23 potently inhibited cell entry by all four variants at levels similar to WT and with lower IC₅₀ values than observed with EK1C4. For Omicron, the IC₅₀ value of CGM23 was 2.2-fold lower than that of EK1C4 in the pseudotyped virion assay (122.3 nM vs. 276.0 nM) (Figures 2E and 2F) and 9-fold lower in the cytopathic effect (CPE) assay using Omicron live virus (85.4 nM vs. 772.6 nM) (Figure S5C). Furthermore, CGM23 exhibited higher inhibitory activity than EK1C4 against multiple more recently emerging variants of SARS-CoV-2 (Figure S5D).

Omicron preferentially enters cells via fusion in the late endosome rather than fusion at the plasma membrane.⁴⁵ To assess potential route-specific differences in CGM23 effectiveness, we compared the ability of CGM23 to block the entry of SARS-CoV-2 pseudotyped virions into 293T cells that overexpress ACE2 (293T-ACE2), which promotes fusion in endosomes, versus 293T cells that overexpress both ACE2 and TMPRSS2 (293T-ACE2/TMPRSS2), which encourages entry at the cell surface.^46,47 There was no significant difference in the profiles of the dose-response curves and IC₅₀ values (293T-ACE2, 112.1 nM vs. 293T-ACE2/TMPRSS2, 124.8 nM) between the two cell lines (Figure S5E).

In these in vitro experiments with pseudotyped virions, dimethyl sulfoxide (DMSO) was used as a carrier solvent for these peptides. To potentially improve clinical utility, we assessed whether reformulating CGM23 in water affected its antiviral activity. We observed equivalent IC₅₀ values for CGM23 in water (101.3 nM) and DMSO (106.1 nM) (Figure S5F). These combined findings indicate that CGM23 exhibits potent and broad-spectrum anti-coronavirus activity, especially including the seasonal coronaviruses such as HCoV-NL43 and HCoV-229E (Figure S5B). In addition, the water-soluble properties of CGM23 were a favorable attribute for clinical development.

CGM23 displays greater helicity than EK1C4 and other related lipopeptides

Among the 159 peptides that were synthesized and tested during the course of this work, some were designed to test a specific concept while others sought to add diversity. In lieu of a detailed description of each modification, we focus here on the important modifications present in the lead peptide, CGM23.

Published X-ray structures reveal that the HR1 and HR2 domains of SARS-CoV and SARS-CoV-2 form a six-helix bundle (6-HB).^30,33 HR2 contains a conserved five-turn-helix that displays a hydrophobic face. This face binds to a groove between two helixes in the HR1 trimer. When free in solution, formation of the five-turn-helix in HR2 is energetically unfavorable. We therefore reasoned that any modification in the inhibitory peptides that encourages helix formation should improve anti-fusion potency.

To improve the antiviral efficacy of short peptides, we initially focused on modifying the N-terminus. Two modifications were introduced at the N-terminus of CGM23 (Figure S4A). First, CGM23 was capped with PBA to mimic Leu1166 of SARS-CoV-2. The published crystal structure of SARS CoV-2 6-HB (PDB: 6lxt³⁰), indicates that Leu1166 (n-2, in CGM23) is buried in a hydrophobic pocket between Phe970 and Ile973. Modeling indicated PBA could form similar hydrophobic interactions. PBA is also in position to interact with the Asn969Lys mutation found in all Omicron subvariants BA1, 2, 3, 4, and 5, which may contribute to the potency of CGM23 against BA.1 (Figures 2F and S5C). The N-terminal PBA cap also stabilizes CGM23 in an alpha helical conformation. Additionally, to enhance the peptide’s anti-pan-coronavirus activity, the N-terminal amino acid sequence was designed to align with that of EK1. Previous studies have shown that peptides derived from the HR2 sequences of SARS-CoV or SARS-CoV-2 exhibit relatively low antiviral activity against MERS-CoV.^33,39 Both SARS-CoV-2³⁰ and EK1³³ form a half helical turn in this location: DISG, 1168–1171, in SARS-CoV-2 and SLDQ, 1–4, in EK1. This half helical turn in EK1 appears to be stabilized by internal H-bonding, which was part of our rationale for employing this sequence in CGM23.

The C-terminus of CGM23 was lipidated with palmitic acid leading to an increase in antiviral activity and manufacturing convenience (Figure S4A). Finally, the CGM23 backbone was based on the HR2 sequence from SARS-CoV-2, whereas EK1’s backbone is based on the HR2 of seasonal coronavirus HCoV-OC43.³³ Close examination of the available X-ray structures indicated that the SARS-CoV-2 gave better fit to the conserved contacts with SARS-CoV-2 HR1. Experiments with chimeric peptides (e.g., CG71, Figure S2) indicate that the four hydrophobic substitutions in the middle helical segment are the most critical: Met16Ile, Ala22Val, Ile23Ala, and Tyr30Leu. These residues also play a crucial role in maintaining the pan-coronavirus potency of CGM23 (Figures S4B and S5B).

To investigate the effect of modifications in CGM23 on helicity, circular dichroism (CD) was employed. Based on a CD maximum near 190 nm and double minima at 208 and 222 nm, CGM23 showed an alpha helicity 1.2-fold greater than that of CGM20, which contains the SARS-CoV-2 HR2-derived backbone but lacks the N-terminal PBA modification and N-terminal amino acid substitutions (CGM23, 69.6% vs. CGM20, 60.4%) (Figure S6A). Furthermore, our computational models indicate that the PBA occupies the same hydrophobic pocket as Leu1162 in HR2, as seen in the crystal structure of 2019-nCoV S2 subunit, 6lxt³⁰ (Methods S2). Of note, EK-1 exhibited a relatively low helicity (38.7%) (Figure S6B). C-terminal lipidation with palmitic acid did not significantly improve helicity over the C4 lipidation introduced into EK1C4 (CGM23, 69.6% vs. CG127, 68.6%) (Figure S6C). Conversely, CGM27 containing bi-functional lipidation exhibited a 2.6-fold lower helicity than CGM23 (CGM23, 69.6% vs. CGM27, 26.6%) (Figure S6D).

CGM23 is more stable in plasma and lung secretions than EK1C4 and other lipopeptides

Next, we assessed CGM23 degradation by biologically relevant proteases present in human plasma and lung secretions. Among the various peptides evaluated, CGM23 demonstrated the greatest level of protease resistance. Peptides were incubated at 37 °C for 8 h, and the level of residual full-length peptide was determined by HPLC (high-performance liquid chromatography). Four EK1-derived peptides (EK1C4, CG114, CG115, and CG127) and four SARS-CoV-2 HR2-derived peptides (CGM23, CGM20, CGM26, and CGM27) were selected for investigation because of their distinctive structures and activities (Figure S1). When incubated in plasma, only 52% of EK1C4 survived as an intact peptide while approximately 80% of CGM23 did under identical conditions (Figure 3A). CGM23 was also more stable than EK1C4 in the presence of neutrophil elastase, a protease actively secreted by neutrophils under inflammatory conditions in the lung (87% residual CGM23 versus 43% EK1C4) (Figure 3B). CGM20, in which the backbone of EK1C4 was changed to SARS-CoV-2 HR2 sequence, showed a 1.7-fold increase in residuals, compared to EK1C4 (CGM20, 87% vs. EK1C4, 52%). CG127, which also has the same lipid modification as EK1C4, showed the same low stability as EK1C4 (49%). However, CGM23, in which the lipid modification of CGM127 was changed to palmitic acid, showed a 1.7-fold increase in stability (82%). CG115, with three amino acid substitutions in the backbone of EK1C4, and CGM26, with the PBA N-capping and bifunctional lipidation in the backbone of SARS-CoV-2 HR2, also showed high stability (CG115, 94%, CGM26, 91%). In the presence of cathepsin B,⁴⁸ CGM23 exhibited high stability, with twice as much peptide remaining than EK1C4 (100% vs. 55%) (Figure 3B). CGM23 was less stable in the presence of cathepsin D than B, but remained more stable than EK1C4 (52% vs. 41%) (Figure 3C). Cathepsin D is not an airway-specific protease and it differs from cathepsin B and K (aspartate proteases) as a cysteine protease. In the presence of cathepsin K,⁴⁹ CGM20 had lower residuals than EK1C4 (28% vs. 42%) (Figure 3D). CGM23 was 1.5 times more stable than EK1C4 (63% vs. 42%). Even in the presence of trypsin, a digestive enzyme located in the intestinal mucosa,⁵⁰ CGM23 showed a high degree of stability (88%) (Figure 3F). In summary, CGM23 exhibited overall heightened resistance to a wide array of biologically relevant proteases present in the pulmonary tract and plasma compared to the other peptides tested in vitro.

Figure 3.

CGM23 stability against plasma and enzymes and pharmacodynamics

(A–F), Stability of peptides in plasma (A) or in the presence of cathepsin D (B), cathepsin B (C), cathepsin K (D), neutrophil elastase (E), trypsin (F) in vitro. After incubation of peptides with plasma or enzymes for 8 h, the residual amount of peptides was determined by HPLC.

(G) Time course of CGM23 and EK1C4 concentration in plasma (LLOQ; 8.65 ng/ml) after intranasal administration to mice (CGM23, 0.865 mg/kg; EK1C4, 1 mg/kg). ∗p < 0.05. Statistical analysis was performed using unpaired two-tailed Student’s t test test. Black bars represent the mean concentrations. LLOQ, lower limit of quantification. Identical symbols indicate sequential samples from the same mouse.

Assessment of pharmokinetics of CGM23 in vivo

CGM23, dissolved in water, was administered intranasally at a dose of 0.865 mg/kg to CD-1 mice, a genetically diverse outbred strain (Figure 3G). Concentrations of CGM23 in plasma, lung homogenates, and bronchoalveolar lavage fluid (BALF) were measured over time. In the plasma of three mice, the concentration exceeded the lower limit of quantification (LLOQ: 8.65 ng/mL, 1.69 nM) within 30 min of intranasal administration and increased rapidly, peaking at 4 h at a maximum concentration (Cmax) of 78 ng/mL (15.2 nM). This concentration is 11 times higher than the in vitro IC₅₀ (1.3 nM), suggesting adequate uptake of the peptide for inhibition of virion fusion. The plasma concentration of CGM23 in the three mice fell below the LLOQ (1.69 nM) at 24 h, with an AUC of 425 ng h/mL and an estimated half-life of 7 h. CGM23 was still detectable 8 h after administration—beyond the calculated half-life—in lung homogenates (17.3 ng/g, 3.3 μM/g) and bronchoalveolar lavage fluid (BALF) (8.65 ng/mL, 1.69 nM). Given these concentrations relative to the in vitro IC₅₀ (0.9 nM), it is likely that CGM23 continues to exert antiviral activity within this time frame. In addition, to investigate the differences in the in vivo stability of CGM23 and EK1C4, EK1C4 was administered intranasally at a concentration of 1 mg/kg, a higher concentration than used with CGM23. Eight hours after administration, the concentrations of EK1C4 in plasma in all three mice had fallen the LLOQ (1.69 nM). Despite a lower dosing (0.865 mg/kg), CGM23 consistently achieved a higher plasma concentration than EK1C4 (p = 0.04) (Figure 3G). In sum, while CGM23 was metabolized within 24 h following intranasal administration to mice, it reached concentrations exceeding the IC₅₀ for inhibition of fusion rapidly and consistently in plasma, airways, and lung parenchyma. These findings suggest that intranasally administered CGM23 is expressed at levels capable of inhibiting virion fusion throughout the body including within the lungs with a T_1/2 of approximately 7 h.

CGM23 suppresses viral replication in vivo

To assess the in vivo antiviral activity of CGM23, 12.5 μg (0.865 mg/kg), the same dose used in the pharmacokinetics study, was administered intranasally to K18-hACE C57BL/6J mice together with live SARS-CoV-2 WA1 engineered to contain a nanoluciferase reporter gene. The impact on viral replication in the lungs was then evaluated 2 days after infection. The impact of CGM23 was compared to that of EK1C4 and CG167 (corresponding to a scrambled EK1 control peptide with the same C-terminal lipid). A non-infected and untreated cohort of mice served as additional controls. In lung homogenates, luminescent signals were detected in the EK1C4 and CG167 groups on days 2 post-inoculation, while no signals were observed in any of the mice in the CGM23 group (Figures 4A and 4B). Histopathological analysis revealed that spike protein signals were significantly reduced in the EK1C4 and CGM23 treatment groups compared to CG167. Notably, spike protein levels tended to be higher in EK1C4-treated mice than in those treated with CGM23 (Figure 4D). Expression of MAC-2, a marker of inflammatory macrophages localized around the airways, was elevated in the CG167 group at the early stage of infection, 2 days post-inoculation. In contrast, no significant differences were observed among the EK1C4 group, the CGM23 group, and the uninfected control group (Figure 4E). To evaluate the therapeutic efficacy of CGM23 against SARS-CoV-2 infection in a physiologically relevant model, wild-type mice were intranasally inoculated with mouse-adapted SARS-CoV-2,⁵¹ followed by intranasal administration of CGM23 or comparable amounts of EK1C4 8 h after virus innoculation. At 24 h post-infection, lungs were harvested, homogenized, and viral titers were quantified by plaque assay. Both EK1C4- and CGM23-treated groups exhibited reduced viral loads compared to the mock control. Notably, CGM23 treatment resulted in significantly lower viral loads than EK1C4, with undetectable virus in 3 out of 5 mice (Figure 4F). To assess the prophylactic efficacy of CGM23, the same model was employed. CGM23 was administered intranasally 30 min prior to viral challenge. Twenty-four hours after peptide administration, lungs were collected for plaque assay. Viral loads were significantly reduced in the EK1C4 group relative to the mock control. In the CGM23 group, viral titers were undetectable in all five mice (Figure 4G).

Figure 4.

In vivo prophylactic and therapeutic efficacy of CGM23 against SARS-CoV-2 live virus in mice

(A) The SARS-CoV-2 outgrowth assay. Lung homogenates were collected 2 days post viral infection (DPI) combined with intranasal administration of CGM23 and EK1C4 (12.5 μg, 0.865 mg/kg).

(B) Diluted lung homogenates were added to Calu-6-ACE2 cells and infection titers were measured by the luciferase assay 48 h later. CG167, EK1 scrambled peptide with EK1C4 lipidation. Data presented correspond to mean ± SD. ∗p < 0.05. Statistical analysis was performed using one-way ANOVA followed by Tukey’s post hoc test.

(C) Histopathological findings of mouse lungs at 2 days after virus inoculation. Images are shown clockwise starting from the top left: CG167 group, EK1C4 group, non-infected group, and CGM23 group. Lung sections were stained with anti-spike antibody (green) and anti-MAC-2 antibody (magenta) and DAPI (blue). Scale bars, 100μm.

(D and E) Quantitative analysis of lung histopathological findings for each group. Data shown represent means ± SD.

(F) Therapeutic treatment. CGM23 and EK1C4 were administered intranasally (12.5 μg, 0.865 mg/kg; 25 μg, 1.73 mg/kg) 8 h after SARS-CoV-2 inoculation, and lung homogenates were collected 24 h later for plaque assay analysis. Data presented correspond to mean ± SD. ∗p < 0.05, ∗∗∗p < 0.001. Statistical analysis was performed using one-way ANOVA followed by Tukey’s post hoc test.

(G) Prophylactic treatment. SARS-CoV-2 was administered intranasally 30 min after intranasal administration of CGM23 or EK1C4 (12.5 μg, 0.865 mg/kg; 25 μg, 1.73 mg/kg). Lung tissues were collected 24 h later for plaque assay analysis.

Discussion

To date, no fusion inhibitors against coronaviruses have been approved for clinical use in humans. Our systematic analysis of 159 lipopeptides based on the HR domains of SARS-CoV-2 and HCoV-OC43 (summarized in Figure S1) identified CGM23 as a potent inhibitor of infection by diverse human coronaviruses. CGM23 exhibits high stability and is resistant to degradation by human plasma and endogenous proteolytic enzymes. As a result, in vivo studies demonstrate that CGM23 administered 8 h after viral inoculation exhibits potent therapeutic effects during the acute phase of infection. Additionally, CGM23 shows a preventive effect when administered 30 min prior to viral challenge. These promising properties argue for further exploration of CGM23 as a preventive or therapeutic agent once it is further optimized for delivery into the lung and systemic circulation. An overall summary of our findings is provided in Table S1.

CGM23 is a SARS-CoV-2 HR2-derived peptide with N- and C-terminal modifications. Of note, an unmodified SARS-CoV-2 HR2-derived peptide, CGM1, lacks inhibitory activity in our syncytia assay (Figure S3), highlighting the functional importance of the N- and C-terminal modifications. Of our eleven SARS-CoV-2 HR2-derived peptides with the strongest cell-cell fusion inhibitory activity, eight were lipidated at the C-terminus (Figure 1C). Lipidation of EK1, a peptide derived from the HR2 of HCoV-OC43, had previously been shown to improve its binding affinity for the HR1 domain and inhibitory activity.³⁰ The enhancing effects of lipidation might be due to improved association of the peptides with membranes where fusion occurs.³⁹ The three highly active peptides without lipid modifications all shared modification on the N terminus with PBA (Figure 1C). The addition of PBA enhances peptide helicity, as shown in the CD spectroscopy experiments (Figure S6E), which could underlie the ability of these peptides to block fusion. In the live virus infection assays, CGM19 and CGM20, both of which are identical SARS-CoV-2 HR2-derived peptides with differing C-terminal cholesterol-based modifications, exhibit lower IC₅₀ values than EK1C4. Conversely, CGM21, a SARS-CoV-2 HR2-derived peptide with C-terminal palmitic acid modification displays a higher IC₅₀ value than CGM20. However, CGM23, where PBA and three hydrophilic substitutions are introduced at the N-terminus of CGM21 results in a modestly improved IC₅₀ value (0.9 nM vs. 11.2 nM). The increased stability of CGM23 structure produced by the N- and C-terminal modifications of the SARS-CoV-2 HR2-derived sequence is likely important for this improved antiviral activity.

CGM23 is able to block the growth of a broad-spectrum coronaviruses in vitro, including SARS-CoV, MERS-CoV, NL63, and 229E. Indeed, we predict it is a pan-coronavirus inhibitor based on the high conservation of the heptad repeats in the coronavirus family. Compared to EK1C4, CGM23 displays lower IC₅₀ values in pan-coronavirus pseudotyped spike virion assay, especially for the alpha coronaviruses HCoV-229E and HCoV-NL43 (Figures S5A and S5B). A previous report showed that two SARS-CoV-2 HR2-derived peptides linked by cholesterol exhibited weak inhibitory activity against MERS-CoV.³⁹ In contrast, CGM23 inhibits MERS-CoV at a lower IC₅₀ than EK1C4, confirming favorable pan-coronavirus properties. CGM23 also outperforms EK1C4 in fusion assays performed with SARS-CoV-2 variants, including Omicron BA.1. BA.1 contains three amino acid substitutions (Q954H, N969K, and L981F) in the HR1 domain (Methods S1), but the other four Omicron subvariants (BA.2, BA.3, BA.4, and BA.5) contain only the Q954H and N969K mutations.^52,53 CGM23 proved effective against all of the Omicron subvariants. Furthermore, the CGM23 displays comparable inhibitory activity for fusion occurring in endosomes, which is efficiently exploited by Omicron variants, and fusion occurring at the cell surface.⁵⁴ Together, these findings indicate that CGM23 exhibits favorable pan-coronavirus activities and maintains high activity against all historical and more contemporary SARS-CoV-2 variants.

A key characteristic of CGM23 is its high resistance to proteolytic degradation. In inhalation therapies, maintaining sufficient peptide concentrations in the airways and lung tissue over extended periods is essential to reduce the need for frequent administration. Furthermore, as observed with SARS-CoV and MERS-CoV infections,^54,55 plasma stability of therapeutic peptides is critical for combating systemic, multiorgan dissemination. Protease degradation assays based on HPLC analysis demonstrated that the superior stability of CGM23 stems not only from its high helicity but also from its structural features, including palmitate lipid modification and a backbone sequence derived from SARS-CoV-2. For instance, although CG127 shares N-terminal modification and backbone structure with CGM23, a comparative analysis of CG127 and CGM23—each possessing cholesterol and palmitate modifications, respectively—revealed that the palmitate modification in CGM23 confers increased stability against plasma proteases and trypsin (Figures 3A–3F). Additionally, comparison between peptides derived from the HR2 domains of HCoV-OC43 (EK1C4) and SARS-CoV-2 (CGM20), both containing the same cholesterol modification, showed that the SARS-CoV-2 HR2-derived backbone provides enhanced resistance to plasma, cathepsin B, cathepsin D, neutrophil elastase, and trypsin (Figures 3A–3F). Pharmacokinetic studies in mice following intranasal administration (Figure 3G) further demonstrated that the peptides were detectable in plasma within 30 min and remained detectable for at least 8 h post-administration. These results collectively indicate that CGM23 is stable in both intracellular and extracellular environments, which plays a critical role in its sustained presence and therapeutic potential.

An additional attractive property of CGM23 is its high level of helicity. Secondary structure analysis by CD spectroscopy demonstrated that the three N-terminal amino acid substitutions (Asp1Ser, Ser3Asp, and Gly4Gln) and the presence of the PBA cap improved helicity by 10% (CGM23 vs. CGM20) (Figure S6A). The three amino acid substitutions at the N-terminus of CGM23 are also present in the N-terminal amino acid sequence of EK1. Crystallographic analysis shows that a half-helix turn is present at the N-terminus of SARS-CoV and SARS-CoV-2.³³ Computer modeling suggests that these three hydrophobic amino acid substitutions form internal hydrogen bonds and reinforce this turn. The HR2 domain of HCoV-229E or HCoV-NL63 forms a longer helix to interact with 3-HR1 fusion core.^55,56 The pan-coronavirus activity of CGM23 might be enhanced by the half-helix at the CGM23 N-terminus, which compensates for the long HR2 domains of these alpha-coronaviruses. Of note, the cholesterol and palmitate modifications at the C-terminus had little effect on helicity (CGM23 vs. CG127). Conversely, the addition of dual C14 lipid modification markedly reduced helicity (CGM23 vs. CGM27) (Figures S6C and S6D). Comparing CGM23 with CGM8, which is CGM23 minus its lipidation, revealed a 5.5-fold difference in IC₅₀ in the syncytia assay (Figure S3). Furthermore, the four amino acid substitutions forming the helical structure, which interact with hydrophobic faces of coronaviruses, ensure a broad spectrum of antiviral activity for CGM23 (Figure S4B). These interactions may explain the 7-fold greater potency of CGM23 relative to EK1 against NL63. Thus, the unique structure of CGM23, which combines high helicity with the maintenance of a hydrophobic face, likely underlies its potent and broad-spectrum anti-fusogenic activity.

In conclusion, we have designed, synthesized, and tested peptides with various modifications to the backbone based on two HR2 domain sequences, and comprehensively analyzed their functions. These structure-function comparisons have led to the identification of CGM23 as a potent pan-Coronavirus inhibitor of viral fusion that is active against both beta- and alpha-coronaviruses such as HCoV-NL43 and HCoV-229E. The CGM23 peptide is water-soluble and appears resistant to the major proteases present in plasma and lung fluids. Taking all of our findings together, CGM23 merits further exploration as an effective preventive or a therapeutic agent in humans and could form an intervention to counter the next zoonotic transmission of a pathogenic Coronavirus into humans.

Limitations of the study

This study has several limitations. First, the effective concentration of inhaled peptides reaching the lower airways could not be precisely quantified, and peptide deposition within the lungs may be heterogeneous, potentially resulting in nonuniform exposure across different airway regions. Furthermore, although no overt toxicity was observed, the possibility of local airway irritation or inflammation following inhalation cannot be excluded. Differences in pulmonary anatomy and respiratory physiology between the animal model used in this study and humans may further limit direct extrapolation of dosing and efficacy. Finally, the long-term safety and immunogenicity associated with repeated peptide inhalation were not assessed and warrant further investigation.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Warner C. Greene (warner.greene@gladstone.ucsf.edu).

Materials availability

Materials generated in this study are available upon request to the lead contact.

Data and code availability

• •All data reported in this paper is available from the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to analyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank Dr. Anke Meyer-Franke, Dr. Nevan J. Krogan, Dr. David E. Gordon, and Dr. William DeGrado for the generous support of this study. This work was supported by the Roddenberry Foundation (Y.M., R.S., M.M., E.W.H., and W.C.G.), the NIH/NIAID AI162083 (M.F.S. and R.T.), and the James B. Pendleton Charitable Trust.

R.K.S. is supported by the NIH. Division of Intramural Research. This research was supported in part by the Intramural Research Program of the National Institutes of Health (NIH). The contributions of the NIH author were made as a part of their official duties as NIH federal employees, are in compliance with the agency policy requirements, and are considered works of the United States Government. However, the findings and conclusions presented in this paper are those of the author and do not necessarily reflect the views of the NIH or the U.S.

Author contributions

Y.M., R.S., M.M., R.T., and W.C.G. designed the study. Y.W., Y.A., and M.A. designed and synthesized the peptides. Y.M., R.S., M.M., E.W.H., R.K.S., T.Y.T., M.F.S., and H.J. conducted experiments and acquired data. Y.M., R.S., M.M., E.W.H., T.Y.T., H.J., M.F.S., R.T., M.A., and W.C.G. analyzed data. Y.M., R.S., M.M., M.A., Y.W., M.O., and W.C.G. wrote the manuscript. W.C.G., R.T., and M.O. provided funding. All authors assisted with editing the manuscript. W.C.G. and M.O. supervised the study.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
SARS-CoV-2 Spike antibody	GeneTex	Cat#GTX135360; RRID:AB_2887483
Anti-mouse/human Mac-2, purified (Clone M3/38)	Cedarlane Labs	Cat#CL8942AP; RRID:AB_10060357
Goat Serum, New Zealand origin	Gibco	Cat#16210064
Goat anti-Rabbit IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor Plus 488	Invitrogen	Cat#A32731; RRID:AB_2633280
Goat anti-Rat IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 647	Invitrogen	Cat#A-21247; RRID:AB_141778
Bacterial and virus strains
Stbl3 cells	Invitrogen	Cat#C737303
SARS-CoV-2 (Isolate USA-WA1/2020)	BEI	Cat#NR-52281
rVSVΔG-rLuc∗G	Condor et al.53	–
Mouse-adapted-SARS-CoV-2	This paper.	–
Chemicals, peptides, and recombinant proteins
Avicel	DuPont	Cat#RC-591
Cathepsin B human	Sigma-Aldrich	Cat#SRP6414
Cathepsin D human	Sigma-Aldrich	Cat#SRP6415
Cathepsin K human	Sigma-Aldrich	Cat#SRP6561
Trypsin	Sigma-Aldrich	Cat#SRP6311
Prolong Diamond Antifade Mountant	Invitrogen	Cat#P36961
DAPI	Thermo Scientific	Cat#62248
PEI MAX-Transfection Grade Linear Polyethylenimine Hydrochloride (MW 40,000)	Kyfora Bio	Cat#24765
Human Sputum Leucocyte Elastase	Elastin Products Company, Inc.	Cat#SE563
Critical commercial assays
Plasmid DNA Minipreps Kit	BioBasic	Cat#BS614
Viral ToxGlo Assay	Promega	Cat#G8941
Renilla Luciferase Assay System	Promega	Cat#E2810
Nano-Glo Luciferase Assay System	Promega	Cat#N1110
Xtra Maxi Plus kit for transfection-grade plasmid DNA	NucleoBond	Cat#740416
Experimental models: Cell lines
HEK293T	ATCC	Cat#CRL-3216; RRID:CVCL_0063
Calu-6	ATCC	Cat#HTB-56; RRID:CVCL_0236
Vero E6	ATCC	Cat#CRL-1586; RRID:CVCL_0574
Calu-6-ACE2	This paper.	–
VeroE6-ACE2-TMPRSS2	This paper.	–
293T-ACE2	This paper.	–
293T-ACE2-TMPRSS2	This paper.	–
Experimental models: Organisms/strains
Mouse: K18-hACE2 C57BL/6J: B6.Cg-Tg(K18-ACE2)2Primn/J	The Jackson Laboratory	Cat#034860; RRID:IMSR_JAX:034860
Mouse: C57BL/6J	The Jackson Laboratory	Cat#000664 RRID:IMSR JAX:000664
Recombinant DNA
SARS-CoV-2 S (Wuhan-Hu-1, Alpha, Kappa, Delta, BA.1)	This paper.	–
SARS-CoV S	This paper.	–
MERS-CoV S	This paper.	–
HCoV-NL63 S	This paper.	–
HCoV-229E S	This paper.	–
hACE2	OriGene	Cat#RC08442
plenti-Gluc-IRES-EGFP	Targeting Systems	Cat#GL-GFP
pLL-EF1α-rFLuc-T2A-mRFP-mPGK-Puro	System Biosciences	Cat#LL420PA-1
Software and algorithms
Prism10	GraphPad	https://www.graphpad.com/scientific-software/prism/
ImageJ	NIH	https://www.graphpad.com/scientific-software/prism/

Experimental model and study participant details

Study design and ethics

All research conducted in this study complies with all relevant ethical regulations. All experiments conducted with replication-competent viruses were performed in a certified biosafety level 3 (BSL3) laboratory and experiments were approved by the Institutional Biosafety Committee of the University of California, San Francisco and the Gladstone Institutes. All protocols concerning animal use were approved (AN203103-00E) by the Institutional Animal Care and Use Committees of the University of California, San Francisco and the Gladstone Institutes and conducted in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

• •Calu-6 and HEK293T cells were obtained from ATCC and were not further authenticated in our laboratory. All cell lines were routinely tested and confirmed to be mycoplasma-free.
• •For mouse sex, similar in vivo experiments were performed in both male and female mice, with age matched between groups.

Method details

Peptide design and synthesis

All the peptides were designed and synthesized in the laboratories of ChemPartner. Computational models guided the design of peptides. Models were built in MOE/CCG⁵⁷ based on the crystal structure 6lxt³⁰ (yellow in Figure S4B). The builder function in MOE was used to construct substitutions which were then minimized to relieve strain. Visual inspection was used to evaluate the models since calculated binding energies were not sufficiently accurate for design work. In the structural model, the PBA residue of CGM23 (Methods S2, orange) mimics the extended N-terminus of HR2 in SARS-CoV-2, 6lxt. The phenyl ring sits in a hydrophobic pocket formed by residues Phe970 and Ile973 (Methods S2, yellow) from adjacent HR1 helixes (Methods S2, green). In the crystal structure 6lxt, the SARS CoV-2 HR2 residue Leu1162 occupies the same pocket.

Solid phase peptide synthesis used Fmoc/tBu chemistry on a peptide synthesizer Syro II (Biotage). The general synthesis scale is 0.1 mM using Rink amide MBHA resin. The acylation reactions were carried out for 60 min with a 4-fold excess of activated amino acids, equimolar amounts of 2-(1H-benzotriazole-1-yl)-1,1,3,3tetramethyluronium hexafluorophosphate (HBTU), and a 6-fold molar excess of N, N-diisopropylethylamine (DIEA) in dimethylformamide (DMF). The dried peptidyl resins were cleaved using a trifluoroacetic acid (TFA) cocktail, 90% TFA, 5% Thioanisole, 2.5 % triisopropylsilane, and 2.5 % water, and 1% EDT for two hours at room temperature. The crude peptides were purified using preparative reverse-phase HPLC (Waters mass-directed auto purification system) with buffer (A) 0.1% TFA in water and (B) 0.1% TFA in acetonitrile, at a flow rate of 25 mL/min. Then, the purified peptides were characterized using analytical HPLC with waters X-bridge peptide BEH C18 column (4.6∗150 mm, 3.5 μm, 130 Å), flow rate 1 mL/min. The lipid, palmitic acid, was conjugated to the peptide on resin. Fmoc-Lys(alloc)-OH was used as the first amino acid to attach to Rink amide MBHA resin. After PBA on resin coupling at the N-terminus of the sequence, the protecting group, -Alloc at the side chain of lysine, was deprotected using Pd(PPh3)4. Then, palmitic acid was conjugated to lysine using HBTU/DIEA in two hours.

SARS-CoV-2 syncytia assay

All peptides were initially adjusted to 20 μM and then serially diluted 1:5 in DMEM before testing in the syncytia assay. 293T constitutively expressing SARS-CoV-2 Spike protein and GFP cells were plated in 96 well plates and mixed with the peptides. After incubation for 30 minutes at room temperature, Calu-6 cells constitutively expressing ACE2 and RFP cells were added to the mixtures and incubated overnight at 37 °C. The 96 well plates were then analyzed by automated microcopy (ArrayScan Instrument, Thermo Scientific) to determine the number of multinucleated large syncytia formed expressing both GFP and RFP fluorescence) for calculating the IC₅₀ values.

SARS-CoV-2 pseudotyped virion assay

For the preparation of virions, 293T cells were transfected with the Spike plasmid (Wuhan-Hu-1, GenBank accession number: MN908947.3), followed by inoculation with a previously generated working stock of rVSVΔG-rLuc∗G (G protein–deficient vesicular stomatitis virus containing an integrated Renilla luciferase reporter gene) to generate the pseudotyped rVSVΔG-rLuc∗SARS-CoV-2.⁵⁸ The SARS-CoV-2 Spike pseudotyped virions harvested from the supernatant of the 293T cells was assayed for titer and then aliquots mixed for 30 minutes with peptides. The mixtures were then used to infect Calu-6 cells constitutively expressing ACE2 plated in 96 well plates and incubated overnight. Levels of infection were measured by Renilla luciferase quantification (Renilla Luciferase Assay System, Promega). Other four coronavirus (SARS-CoV, MERS-CoV, HCoV-NL63, and HCoV-229E) pseudotyped virions were produced by the same method as the SARS-CoV-2 pseudotyped virion and used for testing pan-coronavirus activity of peptides. SARS-CoV-2 variant pseudotyped virions were also generated using Spike plasmids harboring mutations found in the Alpha variant (H69 deletion, V70 deletion, Y144 deletion, N501Y, A570D, D614G, P681H, T716I, S982A, and D1118H), the Kappa variant (L452R, E484Q, D614G, P681R, and Q1071H), the Delta variant (T19R, G142D, E156 deletion, F157 deletion, R158G, L452R, T478K, D614G, P681R, and D950N), and the Omicron variant (A67V, H69 deletion, V70 deletion, T95I, G142D, V143 deletion, Y144 deletion, Y145 deletion, N211I, L212 deletion, R241 insertion, G339D, S371I, S373P, S375F, K417N, N440K, G446S, T478K, E484A, Q493R, G496S, G496S, Q498R, N501Y, N505H, T547K, D614G, H655Y, N679K, P681H, N764K, D796Y, N856K, Q954H, N969K, and L981F) (Methods S1). The all types of virions were titrated based on the plaque assay and equalized the infectivity titers. A scrambled peptide with EK1C4 C-terminal lipidation was used as a negative control and EK1C4 was employed as a positive control in these studies.

ATP cell viability luciferase assay

All peptides were adjusted to 20μM and then serially diluted 1:5 in DMEM before testing in the cell viability assay. The culture medium with peptides was then applied on Calu-6 cells constitutively expressing ACE2 (Calu-6-ACE2) plated in 96 well plates and incubated overnight. Levels of cell viability were tested by cellular ATP measurement (Viral ToxGlo Assay, Promega).

In vitro live SARS-CoV-2 virus infection assay

The SARS-CoV-2 isolate USA-WA1/2020 was produced by cloning into a molecular clone expression plasmid. Live virus experiments were performed with Calu-6 epithelial cells (ATCC HTB-56) stably expressing human Angiotensin Converting Enzyme 2 (ACE2) (OriGene, RC08442) as target cells. Viral stocks were prepared using an infectious clone of SARS-CoV-2 WA1, expressing a nano-luciferase reporter.⁵⁹ For each experiment, 24 hours before administration of virus inoculum or virus/peptide cocktail, 5×10⁴ Calu-6-ACE2 cells were plated per well of a 96 well flat bottom tissue culture treated plate in 200 μL of complete DMEM. At the time of the experiment, test peptides were diluted in DMEM to six concentrations in triplicate as follows; 20 μM, 4 μM, 800 nM, 160 nM, 32 nM, and 6.4 nM. A virus inoculum of 1 × 10³ PFU was prepared in DMEM and was incubated with the peptides or media alone for 30 minutes at room temperature. Upon completion of the virus and peptide co-incubation, the Calu-6-ACE2 cells were washed one time with sterile 1×PBS then the peptide-virus cocktail or virus-media prep were added to the cells in triplicate. The plates were then incubated at 37°C and 5 % CO₂ for 24 hours. At 24 hours post infection the cells were washed once with 1×PBS then 100uls of fresh DMEM was add per well followed by 100 μLs of Nano-Glo Luciferase substrate/buffer cocktail (Promega, N1110). The plates were incubated for 5 minutes at 37°C and 5 % CO₂ then read on a GloMax Discover luminometer. Results were analyzed by GraphPad software version 10.4.2. All experiments were performed in the Gladstone ABSL3, adhering to BSL3 protocols.

Production of virus-like particles and luciferase-based transduction assay

Virus-like particles (VLPs) were generated by transient transfection of HEK293T cells (ATCC). Plasmids encoding SARS-CoV-2 nucleocapsid (CoV2-N, 10 μg), membrane protein with IRES-driven envelope (CoV2-M-IRES-E, 5 μg), spike protein (CoV-2 Spike, 24 ng), and a luciferase reporter construct (Luc-T20, 15 μg) were mixed in 2 mL Opti-MEM (Gibco).⁶⁰ Polyethylenimine (PEI; 90 μg) was added to the DNA mixture, and transfection complexes were incubated at room temperature for 20 minutes before being applied to HEK293T cells (1.2 × 10⁷ cells) seeded in T175 flasks containing DMEM. The culture medium was replaced 24 hours post-transfection. Supernatants containing VLPs were harvested 48 hours after transfection and clarified by filtration through a 0.45 μm syringe filter. For transduction assays, 100 μL of filtered VLP-containing supernatant was added to VeroE6-ACE2-TMPRSS2 cells seeded in 96-well plates and incubated overnight. The following day, supernatants were removed and cells were washed once with PBS (Corning). Cells were lysed by adding 20 μL of cell lysis buffer (Promega) per well and incubating at room temperature for 15 minutes with gentle rocking. Lysates were transferred to white 96-well plates, and 50 μL of reconstituted luciferase assay buffer was added to each well. Luminescence was measured immediately using a TECAN Infinite M plate reader and used as a readout of VLP transduction efficiency.

Peptide stability assay

Peptides adjusted to 100 μM with water were mixed with human plasma, cathepsin B (Sigma, SRP6414) 67 nM, cathepsin D (Sigma, SRP6415) 200 nM, cathepsin K (Sigma, SRP6561) 200 nM, neutrophil elastase (Elastin Products Company Inc., SE563) 100 nM, and trypsin (Sigma, SRP6311) 200 nM, respectively, and incubated at 37°C for 8 hours. After incubation, the remaining amount of peptide was determined by HPLC. Acetonitrile with 0.1 % TFA (mobile phase B) and water with 0.1 % TFA (mobile phase A) were purchased from Fisher Scientific. Samples were analyzed using an Agilent 1100 series liquid chromatograph, including a G1312A pump, G1329 Autosampler, G1300A ALS thermos tatted compartment and a G1315B Diode-array Detector (Agilent). The HPLC column used was an Agilent Pursuit 5, C18 (150×4.6 mm). Aliquots of 50-100 μL were prepared and kept at 4°C in HPLC-ready injection glass vials (VWR) to be run on the same day of collection. Stock solutions of peptides (1 mM) were made in water daily, diluted further as needed, and kept at 4 °C for same-day testing. The column effluent was monitored using a signal of 214 nm. The integrated area under the peak of each peptide run was used as the assay parameter for normalization of the signal.

Pharmacokinetics study

The pharmacokinetics study was completed at ChemPartner. Male CD-1 mice (Jihui Laboratory Animal Care Co., Ltd.) were treated with a single dose of CGM23 (0.865 mg/kg) following intranasal administration, and sampling at 5 minutes, 15 minutes, 30 minutes, 1 hours, 2 hours, 4 hours, 8 hours, and 24 hours post-dose, terminal collecting for lung and Bronchoalveolar lavage at 1 hours, 2 hours, 4 hours, and 8 hours (three mice per time point). Blood was collected via a facial vein in K2 EDTA polyethylene tubes. The blood sample was put on ice and centrifuged at 2,000 g for 5 min at 4 °C within 15 minutes to obtain a plasma sample. Plasma was removed to a new tube and stored at -70 °C until analysis. To get bronchoalveolar lavage (BAL) samples, 35 μL PBS / g (body weight) was slowly aspirated into the trachea using a flexible butterfly catheter via a 3-way tap before the fluid was gradually withdrawn. This procedure was repeated 3 times; all lavage fluids were mixed into a glass bottle. Then, an aliquot of 50 μL BAL was transferred to cell counting directly (storage on wet ice), and the rest of BAL was centrifuged to separate the cell pellet. BAL fluid (BALF), and the cell pellet and sup will be snap frozen and stored at -70 °C until analysis. When analysis, the cell pellet was thawed and reconstituted in 1,000 μL of water per 1x10⁶ cells. The lung was collected, rinsed with cold saline, dried on the filter paper, and directly frozen at -70°C in polyethylene tubes until analysis. CGM23 levels in mouse plasma, lung, and Bronchoalveolar lavage samples were measured using UPLC-MSMS (AB SCIEX, TRIPLE QUAD 6500, 5035183-K).

In vivo testing of peptides in mice

To evaluate the in vivo efficacy of candidate peptides, heterozygous K18-hACE2 C57BL/6J transgenic mice (strain: B6.Cg-Tg(K18-ACE2)2Primm/J) and C57BL/6J wild-type mice were housed in the Gladstone Animal Facility. At the time of experimentation, animals aged 5–8 weeks were randomly assigned to experimental groups and transferred to the ABSL-3 facility at the Gladstone Institutes. For all in vivo procedures, body weight and baseline body temperature were recorded. Mice were anesthetized via intraperitoneal injection with a solution containing ketamine (150 mg/kg) and xylazine (10 mg/kg). In studies using K18-hACE2 mice, anesthetized animals were intranasally inoculated with a mixture containing SARS-CoV-2 WA1 Nano-Luc virus⁶¹ (1 × 10³ PFU in 20 μL) and candidate peptide (12.5 μg), adjusted to a final volume of 40 μL with PBS. At 2 days post-inoculation, animals were euthanized by carbon dioxide asphyxiation followed by cervical dislocation. For experiments using wild-type mice, we constructed a mouse-adapted (MA)-SARS-CoV-2 (Spike: Q498Y/P499T) using pGLUE as described previously.⁶² MA-SARS-CoV-2 (1 × 10³ PFU in 40 μL) was administered intranasally under anesthesia. Peptides (12.5 μg or 25 μg) were administered intranasally 8 hours post-inoculation in a final volume of 40 μL. Mice were euthanized 24 hours after viral inoculation. In prophylactic experiments, wild-type mice received intranasal administration of the peptide (12.5 μg, 25 μg in 40 μL), followed 30 minutes later by intranasal inoculation with MA-SARS-CoV-2 (1 × 10³ PFU in 40 μL). Mice were euthanized 24 hours after peptide administration. For histological analysis, K18-hACE2 mice were perfused with 1× PBS followed by 4% paraformaldehyde. In animals not used for histology, lungs were harvested, coarsely minced, and homogenized in pre-filled zirconium bead tubes using a bead homogenizer. Lung homogenates from K18-hACE2 mice were also used to assess viral replication in Calu-6-ACE2 cells. Cells were seeded in 12-well plates at a density of 2 × 10⁵ cells per well and infected with a 10⁴-fold dilution of lung homogenate. After incubation at 37°C and 5% CO₂ for 2–4 hours with intermittent shaking, the homogenate was removed and replaced with fresh culture medium. After 48 hours of incubation, Nano-Luc luciferase activity was measured using Promega Nano-Glo reagent and quantified with a GloMax Discovery luminometer (Promega). Lung homogenates from wild-type mice were evaluated for infectious viral particle formation using a plaque assay. VeroE6-ACE2-TMPRSS2 cells were seeded into 24-well plates and incubated overnight at 37 °C with 5 % CO₂. The following day, cells were infected with 10^-1 to 10^-6 serial dilutions of lung homogenates in serum-free DMEM and incubated for 1 hour to allow viral adsorption. After adsorption, cells were overlaid with 2.5 % Avicel (DuPont, RC-591) prepared in complete DMEM. Plates were incubated for 72 hours, after which the overlay was removed. Cells were fixed with 10 % neutral buffered formalin for 1 hour and stained with crystal violet to visualize plaque-forming units (PFU).

Immunohistochemistry

The harvested mouse lungs were cut into 15 μm thick frozen sections with a cryostat (Leica CM 1900). and mounted to super frosted microscopic slides (VWR, 48311-703). The tissue sections were first washed in PBS and incubated in a blocking and permeabilization buffer [DPBS supplemented with 0.2 % Triton X-100, 3 % normal goat serum (Gibco, 16210074) and 5 % BSA] for 1 hour at room temperature. Sections were rinsed twice with PBS containing 0.1 % Triton X-100 before overnight incubation with anti-SARS-CoV-2 Spike rabbit polyclonal antibody (GeneTex, GTX135360) and anti-mouse MAC-2 monoclonal antibody (Cedarlane Labs, CL8942AP) (1:100 dilution) at 4 °C. All tissue sections were washed three times with DPBS containing 0.1 % Triton X-100 and incubated with goat anti-rabbit IgG secondary antibody, Alexa Fluor Plus 488 (Invitrogen, A32731) and goat anti-rat IgG secondary antibody, Alexa Fluor 647 (Invitrogen, A-21247) (1:1000 dilution) for one hour at room temperature. After a second wash, the tissue sections were stained with 4′,6-diamidino-2-phenylindole (DAPI, Thermo Scientific, 62248). Free floating sections were mounted on microscopic slides, covered with glass coverslip (VWR, 48404-452) and sealed with ProLong Diamond Antifade Mounting reagent (Invitrogen, P36961) and kept at 4 °C until imaging.

Confocal microscopy

Tissue sections were imaged using an Olympus FLUOVIEW FV3000RS confocal microscope, with a 60×oil UPLSAPO (NA = 1.35) objective and a FV31S-SW software (Olympus). Individual channels were captured, where a 405 nm laser was used for DAPI with collection through a 430/70 special detector (SD1), AlexaFluor488 excited using a 488nm laser with collection through a 500/40 special detector (SD2), and AlexaFluor568 excited with a 561 nm laser and collected through a 570/620 high sensitivity detector (HSD3), and AlexaFluor647 exited with a 650 nm laser and collected through a 650/750 high sensitivity detector (HSD4) (TruSpectral detector technology). Captured images were processed using Image J version 2.1.0/1.53c.

Circular dichroism (CD) spectroscopy

Peptide solutions were prepared at 50 μM in 10 mM phosphate buffer (pH 7.0). Peptide concentrations were determined by absorbance at 280nm after denaturation in 6 M guanidine HCl solution using 1280 M^-1 cm^-1 and 3936 M^-1 cm^-1 for extinction coefficient of tyrosine (Tyr) and naphthylalanine (Nal), respectively.^63,64 Circular dichroism studies were conducted at 25 °C on a JASCO J-810 spectropolarimeter equipped with a Peltier temperature control unit.

Quantification and statistical analysis

The IC₅₀ was determined by first normalizing the measured values using the average of the minimum and maximum values within the dataset (Normalized value = (x – $\overline{x}$_min)/($\overline{x}$_max – $\overline{x}$_min)). Curve fitting was performed on the normalized data using a three-parameter logistic (3PL) dose-response model with a fixed Hill slope of –1, Y = Bottom + (Top – Bottom)/(1 + (X/IC₅₀)⁻¹), and the IC₅₀ was calculated from the fitted curve with GraphPad Prism. Data analysis for the pharmacodynamic studies shown in Figure 3 was performed using an unpaired two-tailed Student’s t test, with three mice per group. ∗P < 0.05. For the peptide administration experiments shown in Figure 4, data were analyzed using one-way ANOVA followed by Tukey’s post hoc test, with three or five mice per group in the in vivo infection study. ∗P < 0.05, ∗∗∗P < 0.001. Correlations between IC₅₀ titers in the pseudotyped Spike virion assay and IC₅₀ titers in the live virus infection assay were analyzed using Spearman’s rank test. Analyses were performed with GraphPad Prism version 10.4.2. Heatmaps were created in R studio and GraphPad Prism version 10.4.2.

Published: March 11, 2026