A dimeric proteomimetic prevents SARS-CoV-2 infection by dimerizing the spike protein

Protein tertiary structure mimetics are valuable tools to target large protein–protein interaction interfaces. Here, we demonstrate a strategy for designing dimeric helix-hairpin motifs from a previously reported three-helix-bundle miniprotein that targets the receptor-binding domain (RBD) of severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2). Through truncation of the third helix and optimization of the interhelical loop residues of the miniprotein, we developed a thermostable dimeric helix-hairpin. The dimeric four-helix bundle competes with the human angiotensin-converting enzyme 2 (ACE2) in binding to RBD with 2:2 stoichiometry. Cryogenic-electron microscopy revealed the formation of dimeric spike ectodomain trimer by the four-helix bundle, where all the three RBDs from either spike protein are attached head-to-head in an open conformation, revealing a novel mechanism for virus neutralization. The proteomimetic protects hamsters from high dose viral challenge with replicative SARS-CoV-2 viruses, demonstrating the promise of this class of peptides that inhibit protein–protein interaction through target dimerization.

C onformationally stabilized protein secondary structure mimics of α-helices, β-sheets and loops have been extensively used as protein-protein interaction (PPI) inhibitors 1 . However, the efficient mimicry of relatively large and flat PPIs necessitates the use of more complex topologies that could be recapitulated by protein tertiary structure mimetics 2 . Helix-turn-helix motifs or helix-hairpins, wherein a pair of antiparallel amphipathic helices are connected by a covalent loop, are particularly well-suited for this purpose 3 . The design of synthetically amenable and conformationally stable monomeric helix-hairpin peptides was pioneered by Wells and coworkers, who showed the structural and functional mimicry of an IgG-binding three-helix-bundle Z-domain 4 . Their approach involved truncating the third helix that does not engage in target binding, incorporating charged residues at the exoface and stabilizing the helix-hairpin motif by a disulfide bridge. Subsequent work by Gellman and others successfully used this strategy to derive potent binders against human epidermal growth factor receptor 2 (HER2) 5 and vascular endothelial growth factor (VEGF) 6 . In addition to disulfide bridges, monomeric helix-hairpin mimics targeting intracellular PPIs have been stabilized by substituting an interhelix salt bridge with a lactam bridge by Grossmann, Waldmann and coworkers, and covalent crosslinkers by Arora et al. 7,8 .
Dimeric helix-hairpin proteins such as Rop occur in nature 9 . However, seminal work of DeGrado 10 and Baltzer 11 toward the de novo design of four-helix-bundle proteins and helix-loop-helix motifs, respectively, established the principles for the design of dimeric helix-hairpins. To drive the formation of dimeric helix-hairpins, appropriately sculpting the heptad repeats in either helix with hydrophobic/hydrophilic residues is as important as the proper choice of loop residues 10,12 . However, unlike the β-turn in β-hairpins, which have been extensively studied and designed 13 , little attention has been paid to the design of interhelical loops.
Thus, we sought to design dimeric helix-hairpin proteomimetics through loop optimization that could mimic the topology necessary for targeting complex PPIs and potentially induce dimerization of the target. The interaction between the receptor-binding domain (RBD) of severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) and the human angiotensin-converting enzyme 2 (ACE2) served as a fertile ground for testing our hypothesis.
SARS-CoV-2, which emerged in late 2019, caused a global COVID-19 outbreak 14 by infecting roughly 500 million humans and killing more than 6 million humans worldwide. SARS-CoV-2 infection is mediated by RBD of the spike glycoprotein on the viral surface through binding to the cell-surface receptor ACE2 (ref. 15 ). After the attachment of the virus to the ACE2 receptor, the spike protein is cleaved by the transmembrane serine protease 2 (TMPRSS2). This facilitates conformational transitions that induce membrane fusion and subsequent release of the viral RNA genome into the host cell cytoplasm 16 17 . The X-ray and cryogenic-electron microscopy (cryo-EM) structures of the SARS-CoV-2 spike protein bound to ACE2 revealed that the N-terminal helices of the ACE2 receptor mainly engage with the RBD of SARS-CoV-2 (ref. 18 ). Therefore, rational structure-guided design of α-helical peptides that mimic the ACE2 receptor interaction has been a sought-after strategy to develop pre-exposure prophylactics [19][20][21][22][23] . Nonetheless, most ACE2 derived single-helical peptide mimics are not potent enough to compete against the high-affinity SARS-CoV-2-ACE2 interaction (K D of roughly 14 nM) 24 , and thus impractical for clinical development 21,25,26 . However, the recent de novo design approaches have We noted the absence of a SIH-1-RBD complex at the highest achievable concentration of SIH-1 (12 μM). We confirmed the observation through binding-induced thermal shift change of RBD using nanoscale differential scanning fluorimetry (nanoDSF), wherein we could detect a small population of SIH-1-RBD complex (Extended Data Fig. 2). Thus, this suggested that the native loop of SIH-1 is inefficient in inducing the hairpin dimerization, resulting in solvent exposure of the hydrophobic patch and, consequently, its poor solubility.
Next, we systematically substituted the loop residues connecting the two helices to enforce a compact helix-hairpin. We initially substituted Ala 22 at the N terminus of helix 2 with the helix-promoting aminoisobutyric acid, which modestly increased the thermal stability of SIH-2 (Fig. 1d), although its solubility remained low (Supplementary Table 1). SIH-2 is a better competitor of ACE2 than SIH-1, however, still with low efficiency (Supplementary Fig. 4). Based on the dihedral angles (ϕ, ψ) of the residues -Leu 19 -Gly-His 21determined from the structure of LCB1, the loop can be categorized into the γ-α L -β conformation put forth by Efimov 3,32 . To introduce a conformationally constrained loop, we substituted the flexible Gly 20 in SIH-2 with D-Ala to enforce a rigid left-handed α-helical (α L ) conformation 33 . The sharp increase in helicity of SIH-3 (62.3%) against SIH-2 (45.9%) is indicative of the favorable influence of α L conformation adopted by D-Ala on the helix-hairpin. The enhanced helicity leads to improved knob-into-hole interaction of the hydrophobic residues at the a and d sites from helix 1 and helix 2 to form a hydrophobic seam. This subsequently leads to dimerization of SIH-3 by forming a hydrophobic core, which is evident from the increased solubility, marked increase in thermal stability (T M ), and its predominantly dimeric state (Supplementary Figs. 3 and 5).
To further improve the helicity of SIH-3, His 21 adopting the β-conformation was substituted with Asn (SIH-4) and Asp (SIH-5) to introduce an N-cap interaction with the N terminus of helix 2 (ref. 34 ). The high helical content, thermostability and solubility of these loop-engineered variants result from a stable and compact helix-hairpin structure (Fig. 1d). Consequently, SIH-3, SIH-4 and SIH-5 efficiently out-compete ACE2 in a dose-dependent manner, with SIH-5 being the most potent ( Fig. 2a and Supplementary Fig. 4).
To evaluate the oligomeric state of SIH-5, we performed sedimentation-velocity and sedimentation-equilibrium analytical ultracentrifugation, which indicated that SIH-5 predominantly exists in a dimeric state (Supplementary Figs. 5 and 6). The dimeric state of SIH-5 led us to assess its use as a scaffold for targeting SARS-CoV-2 with mutations in the RBD and other PPIs. Thus, the structural and functional tolerance of SIH-5 on amino acid substitutions was determined. We incorporated isosteric unnatural amino acids at six different sites on SIH-5 that included residues at the RBD binding interface, buried residues between helices 1 and 2, and solvent-exposed residues ( Supplementary Fig. 2).
Binding of helix-hairpin to RBD increase its stability. All the tested substitutions retained the stability of the helix-hairpins and efficiently inhibited the RBD-ACE2 interaction (Supplementary Figs. 3 and 4). However, in particular, three analogs SIH-7 (Tyr 10 3,4difPhe), SIH-11 (Phe 35 Cha) and SIH-13 (Tyr 10 3,4difPhe ;Phe 35 Cha), despite their occurrence in varying oligomeric states ( Supplementary Fig. 5), were potent competitors of ACE2. The comparable potency of SIH-5, SIH-7, SIH-11 and SIH-13 to LCB1 in inhibiting RBD binding to ACE2, led us to evaluate their binding affinity to RBD by SPR 24 . All the peptides bound to RBD with binding affinities in the range of 0.9-3 nM ( Supplementary Fig. 7). For SIH-5 we could not determine the binding affinity since the sensorgram at the highest concentration (160 nM) showed a rise in signal post 200 s dissociation time (Fig. 2b). Thus, we re-analyzed the data considering 200 s as the dissociation time, which returned a binding affinity of roughly 900 pM. LCB1 bound to RBD with a binding affinity of 42 pM (Fig. 2b). Thus, the helix-hairpins showed more than 25-fold loss in binding affinity to RBD relative to LCB1. However, the comparable efficacy of the peptides and LCB1 in preventing the binding of RBD to ACE2 in the competition assay ( Fig. 1d) was perplexing. Thus, we monitored the binding-induced thermal shift change of RBD in the presence of inhibitors by nanoDSF ( Fig. 2c) 24 , where the change in T M reports on the tight binding and stability of the inhibitor-RBD complex. On binding, LCB1 increased the melting temperature of RBD by 16 °C, whereas the SIH peptides increased the melting temperature of RBD maximally by 11 °C, confirming their weaker affinity to RBD. Although LCB1 binds tightly to RBD, two distinct thermal unfolding events corresponding to the free RBD (52.9 °C) and LCB1-RBD complex (69.2 °C) were observed. Curiously, a single thermal unfolding corresponding to the SIH-RBD complex was observed for SIH-5 and SIH-13 ( Supplementary Fig. 8), indicating the absence of free RBD under the experimental condition.
Helix-hairpins induce dimerization of RBD. Next, we determined the binding stoichiometry of the SIH peptides to the RBD through SEC-MALS ( Fig. 2d and Extended Data Fig. 3). The RBD eluted as a single peak with a molecular weight of roughly 33.1 kDa. Incubation of LCB1 (6.8 kDa) with RBD resulted in a single stable complex (39.2 kDa), indicating a binding stoichiometry of 1:1 (ref. 20 ). Incubation of SIH-5 and SIH-13 (4.6 kDa) with RBD resulted in a complex that eluted earlier than RBD with a molecular weight of roughly 75 kDa, suggesting a 2:2 binding stoichiometry. Thus, the formation of stable SIH-5-RBD and SIH-13-RBD dimer results in a single thermal unfolding as observed in nanoDSF ( Supplementary Fig. 8    distinct complexes with RBD, where the early eluting peak (75 kDa) correspond to the 2:2 complex and the later peak (40 kDa) correspond to a 1:1 (SIH-3-RBD) binding stoichiometry. Therefore, a stable dimeric helix-hairpin (SIH-5) can simultaneously engage two RBDs forming a stable complex. Despite the higher oligomeric state of SIH-13 ( Supplementary Fig. 5), the formation of a 2:2 complex in the presence of RBD suggests a strong dimeric interface in SIH-13, but a relatively weak dimer-dimer association in the oligomeric state, which is disrupted on binding to the RBD. A similar phenomenon is observed for SIH-7 and SIH-11 (Supplementary Fig. 5 and Extended Data Fig. 3).

Cryo-EM confirms the dimerization of spike protein by SIH-5.
To investigate the structure of dimeric SIH-5 bound to RBD and its consequence on the spike protein, we resorted to electron microscopy at physiological pH where 68% of the spike protein are present in an 1-RBD up conformation 35 . Room temperature negative staining transmission electron microscopy (TEM) followed by reference-free two-dimensional (2D) class averages were performed to visualize the SIH-5-spike protein complex. Negative staining indicated the absence of aggregation or distortion of the spike protein in the presence of SIH-5 ( Supplementary Fig. 9). Nonetheless, the TEM images showed the occurrence of elongated spike protein particles, and the 2D reference-free class averages confirmed the association of two spike proteins (Fig. 2e). To obtain finer detail of the dimeric complex, we visualized the entire assembly at cryogenic conditions using single-particle cryo-EM. The three-dimensional (3D) classification indicated that most particles adopt a dimeric state ( Fig. 2f and Supplementary Figs. 10 and 11). The structure solved to a global 5.4 Å resolution at 0.143 Fourier shell correlation ( Supplementary Fig. 12), showed a dimeric spike protein ectodomain trimer with all three RBDs from either spike protein oriented  To identify the interacting residues between SIH-5 and spike protein, we improved the resolution of the single spike protein trimer bound to SIH-5, which was solved at a resolution range of 3.5-4.5 Å using a local resolution calculation and a global resolution of 4.47 Å at 0.143 Fourier shell correlation ( Supplementary Fig. 13). The side chains of the S2 region in spike protein are visible in high-resolution maps, which allowed us to calculate the atomic model from the cryo-EM map ( Supplementary Fig. 14). The atomic model successfully docked into the dimeric spike protein's cryo-EM map, wherein the two spike protein models fit appropriately. The SIH-5 fit with high accuracy within the additional density present between the RBDs of either spike protein (Fig. 3b,c and Supplementary Video 1). The fitting of SIH-5 revealed an antiparallel four-helix bundle formed by the two helix-hairpin protomers, where both the loops are at one end and the N and C termini at the other (Fig. 3e). The total interacting surface area of SIH-5 with RBD is 944 Å 2 . However, at this resolution, we could identify only a few interacting partners (Fig. 3d). Of note, Gln 7 , Aib 22 and Glu 23 of SIH-5 interact via H-bonding interaction with Glu 484 and Asn 501 of the RBD that are known to undergo mutations 17 . The comparison of SIH-5-RBD spike protein complex with LCB1-RBD revealed that SIH-5 binds to RBD in the same fashion as LCB1, however, with a displacement in the loop (Tyr 473 -Tyr 489 ), which moves toward the SIH-5 (Extended Data Fig. 5). Last, we noted that the dimer interface of SIH-5 is stabilized by a crucial electrostatic interaction between Asp 1 and Lys 38 from either protomer and several hydrophobic interactions (Fig. 3e). The data represent the average and standard deviation of replicates (n = 2). All pseudoviral assays were repeated three times. c, Kinetics of the pseudoviral infection in the presence of LCB1 and SIH-5. SIH-5 displayed significant enhancement in virus neutralization from 0 to 30 min preincubation compared to LCB1. The data represent the average and standard deviation of replicates (n = 2). All pseudoviral assays were repeated three times. d, DLS intensity distribution for pseudovirus (control) and pseudovirus incubated with LCB1, SIH-5 and SIH-11. A single representative intensity distribution out of three independent DLS measurements for each condition is provided.

Homodimerization amplifies the potency of helix-hairpins.
To evaluate the beneficial role of dimeric helix-hairpins in target engagement, we sought to develop monomeric helix-hairpins. Since the hydrophobic residues at the g sites are crucial for homodimerization, we substituted the Ile12 and Leu31 in SIH-5 with Ala and Glu to obtain SIH-14 and SIH-15, respectively (Figs. 1c and 3e). As the dimer interface transition from a hydrophobic to the polar surface, we observe a gradual reduction in helicity and θ 222 /θ 208 of the helix-hairpins ( Fig. 1d and Extended Data Fig. 6), resulting in monomeric SIH-15 (Extended Data Fig. 7). The partial destabilization of SIH-5 homodimer on alanine substitution results in 1:1 and 2:2 binding stoichiometries of SIH-14-RBD (Extended Data Fig. 6), while the complete destabilization by glutamate substitution results in a 1:1 complex of SIH-15-RBD (Fig. 2d). The loss of a stable homodimeric structure weakens the binding affinity of SIH-14 (3 nM) and SIH-15 (1.7 nM) to RBD against SIH-5 (0.9 nM). Despite the comparable binding affinities of SIH-15 and SIH-13 (1.6 nM) to RBD, we note a twofold lower potency of monomeric SIH-15 in inhibiting the RBD binding to ACE2 (Fig. 1d). Finally, on subjecting to thermal stress, the 1:1 SIH-15-RBD complex shows a distinct peak of RBD as opposed to the 2:2 SIH-5-RBD complex (Fig. 2c). Thus, the homodimerization of helix-hairpins results in bivalent inhibitors that form high-affinity and thermally stable 2:2 ligand-target complexes.
Virus neutralization by helix-hairpins. We next used surrogate virus to evaluate the ability of the helix-hairpins to block the interaction of SARS-CoV-2 spike with the cellular receptor ACE2 and thereby inhibit the viral entry ( Fig. 4 and Extended Data Fig. 7). In the standard pseudovirus neutralization assay, SIH-5 and SIH-11 were very potent with half-maximum inhibitory concentration (IC 50 ) of 326 and 337 pM, respectively, approaching the potency of LCB1 (86 pM). Although the SIH-5 variants with noncanonical substitution in the helix, SIH-7, SIH-11 and SIH-13 formed a stable 2:2 complex with RBD, none could outperform SIH-5 in neutralizing the virus. The oligomeric state of these helix-hairpins sheds light on their relative potency of pseudovirus neutralization ( Fig. 4a and Extended Data Fig. 7). The comparable binding affinities to RBD The spike protein is in the prefusion state on the surface of free virions. It undergoes molecular breathing 36 that exposes the RBDs over time to enable interaction with ACE2. To assess the differential mode of action of LCB1 and the hairpin peptides, we preincubated the virus with the most potent peptides for different time points and determined the inhibition of viral entry (Fig. 4c). The entry inhibition by SIH-5 and LCB1, where the compounds were not preincubated with the virus (0 min) or preincubated for 10 min before the infection, correlated well with their respective binding affinities to RBD. At these time points, LCB1 was roughly 60-75-fold more potent than SIH-5 in neutralizing the virus. However, SIH-5 was equipotent to LCB1 in neutralizing the virus when preincubated for 30 min. An identical trend in neutralizing the virus was observed for SIH-11 (Extended Data Fig. 7). The significant enhancement (>450-fold) in the neutralization potency of SIH-5 at 30 against 0 min, as opposed to LCB1 (>10-fold), is indicative of the large structural reorganization for the dimerization of the spike proteins on the surface of virions by SIH-5. Since LCB1 neutralizes the virus through binding to the RBD without inducing spike dimerization, it shows more rapid neutralization kinetics than SIH-5.
The dimerization of the spike protein on the surface of virions would result in higher-order complexes/viral aggregates in the solution. Therefore, we incubated SIH-5 and SIH-11 with the pseudovirus and determined the size of viral particles by dynamic light scattering (DLS) (Fig. 4d). Since LCB1 binds to the spike protein in 1:1 stoichiometry, it served as a control to distinguish the mode of action of dimeric helix-hairpins. The viral particles had an average diameter of 155.5 ± 0.5 nm, agreeing well with the previous reports 37 . On incubating with LCB1, we observed a modest increase in the diameter of the viral particles. However, a stark increase in the size of viral particles, suggestive of aggregation 37 was observed with SIH-5 and SIH-11, wherein SIH-11 (Supplementary Table 3) formed a discrete higher-order population. Furthermore, the sharp peak shape of the SIH-5 and SIH-11 treated pseudovirus, as opposed to a broad distribution, suggests a stoichiometric basis for viral aggregation.
Prophylactic activity of SIH-5 in hamsters. Encouraged by the potency of SIH-5 and SIH-11 in neutralizing the virus, we evaluated the cytotoxicity of these peptides on Vero-E6 cells ( Supplementary  Fig. 16). At the highest concentration of the peptides tested (20 μM), SIH-5 and SIH-11 showed >75% cell viability at 48 h. A selectivity index (calculated by dividing the 50% cytotoxic concentration with IC 50 ) of >50,000 was observed for both the peptides, suggesting their potential as antivirals. This led us to evaluate SIH-5 in preventing SARS-CoV-2 infection in vivo. We used the well-established Syrian hamster model 38,39 and performed a single-dose prophylactic experiment. An intranasal dose of 2.5 mg kg −1 of SIH-5 was administered to the hamsters 8 h before the viral challenge with 5 × 10 6 plaque-forming units (PFU) of SARS-CoV-2 US strain (USA-WA1/2020) (Supplementary Fig. 17). SIH-5 was efficient in preventing weight loss in hamsters from the second day postinfection ( Fig. 5a) and accompanied a significantly reduced viral load in the lungs 4 days postinfection (Fig. 5b). We also monitored the clinical signs of disease progression in the three groups ( Supplementary  Fig. 18 and Table 4). Histopathology of the lung tissues at 4 days postinfection showed that the virus-challenged hamsters had a larger percentage area of lungs affected than the prophylactic and the unchallenged groups (Fig. 5c). The virus-challenged group showed moderate to severe pulmonary inflammation, bronchitis, bronchial and alveolar necrosis, and infiltration of lung tissues by immune cells (neutrophils, macrophages and lymphocytes). Although these phenotypes were observed in the prophylactic group, there was a significant reduction in severity and incidence (Fig. 5d).
Loop and g site optimization results in hyperstable hairpin. To demonstrate the broad applicability of the design strategy to develop thermostable dimeric helix-hairpins, we chose an antiparallel coiled coil formed by helices α6 and α7 from the N-terminal domain of the transcription factor STAT-4 (signal transducers and activators of transcription), which mediate several PPIs 40 . Because our design allowed for the stabilization of helix-hairpins with less than three heptad repeats in each helix, we truncated STAT-4 to an identical length as SIH-1 (Fig. 6a,b). The 38-residue wild-type sequence (BGF-1) showed a modest helical signature and low thermal stability (T M = 48.7 °C) ( Fig. 6c and Supplementary Fig. 19). However, by introducing -D-Ala-Asp-Aib-at the interhelical loop and substituting the polar Gln 12 and Glu 19 at the g sites with hydrophobic Leu, we obtained BGF-2 that showed a significantly enhanced helicity and θ 222 /θ 208 , indicative of dimerization ( Fig. 6d and Supplementary  Fig. 20). BGF-2 showed exceptional thermal stability (T M > 95 °C), which presumably results from the formation of a hydrophobic core achieved through the loop and g-site engineering.

Discussion
In the pursuit of stable proteomimetics to target PPIs, we explored dimeric helix-hairpin peptides that present two identical faces for interacting with the target of interest. The de novo designed miniprotein LCB1, which bears structural similarity to the Z-domain of staphylococcal protein A derived three-helix-bundle affibody molecules 41 , was used for the investigation. The three critical steps used in the design of dimeric helix-hairpins are (1) truncation of the third helix; (2) widening of the narrow hydrophobic channel carved by the a and d residues through incorporating hydrophobic residue at the g sites and (3) optimization of the turn residues to enforce a rigid conformation inducing the proximity of helices 1 and 2. The design of helix-hairpins with fewer than three heptad repeats is challenging. Our findings indicate that interhelical loop re-engineering coupled with g site substitution to obtain a 3-3-1 hydrophobic repeat in the helices results in stable dimeric helix-hairpins. We note that subtle alteration in the amphipathicity of the helix-hairpin on substitution with hydrophobic amino acids reorganizes the oligomeric state of the dimeric helix-hairpin. However, when subjected to target binding, the high affinity between the helix-hairpin and the target protein dissociates the transient oligomeric states of the ligand into a dimeric ligand-protein complex that is stable under thermal stress.
The inhibition of PPI through target dimerization 42 , allowed us to prevent the binding of SARS-CoV-2 onto cell-surface ACE2 by dimerizing the spike protein. The dimerization of the spike protein on the surface of virions leads to noncovalent crosslinking of the individual spike proteins and enhanced blocking of the RBD through avidity effects. This significantly contributes to virus neutralization as observed for dimeric SIH-5 against monomeric SIH-15. It is worth noting that the head-to-head dimerization of spike trimer has also been identified in the SARS-CoV-2 kappa variant of concern (B.1.617.1). However, unlike in the SIH-5 induced spike trimer dimerization that inhibits the viral entry into cells, the labile dimer in the kappa variant has been suggested to increase the viral fitness against neutralizing antibodies by masking the antibody-accessible region of RBD 43 . Thus, the spike trimer dimerization is a unique mechanism to tune the viral fitness 44 . Despite its weaker affinity to RBD, the new binding mode of SIH-5 allowed us to achieve comparable potency in neutralizing the virus as the high-affinity RBD binder LCB1. The in vivo efficacy of the dimeric helix-hairpin devoid of toxicity suggests the potential for their development as new therapeutics. We see two different routes for the use of this class of molecules, one in which SIH-5 is used as a scaffold, and by substituting the solvent-exposed residues engaged in RBD binding, specific interaction with the target of interest is achieved. While the other involves a stepwise optimization of the turn residues and g sites of the helix-hairpin extracted from the PPI interfaces 45 or three-helix-bundle affibody molecules 4-6,41 to develop dimeric helix-hairpins. The exceptional stability of these targetligand-ligand-target complexes reminiscent of irreversible covalent inhibitors would encourage the exploration of this unique class of proteomimetics to target PPIs while achieving the favorable pharmacology displayed by targeted covalent inhibitors 46 .

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41589-022-01060-0.
LCB1 was synthesized on 2Cl-TCP resin (1.3 mmol g −1 ) using standard Fmoc-based chemistry. The global deprotection of protected protein was carried out with 95% TFA, 2.5% triisopropylsilane and 2.5% water for 2 h. Deprotected crude protein was precipitated in chilled ether. The white solid was pelleted by centrifugation and dissolved in 6 M Guanidine HCl in phosphate buffer for purification. Protein was purified by RP-HPLC (solvent A 0.1% TFA in water, solvent B 0.1% TFA in acetonitrile, gradient 25-85% B solvent over 30 min). The purity was checked by RP-HPLC and the identity was confirmed by electrospray ionization-mass spectrometry.
Solubility measurement. Solubility of SIH-1 to SIH-5 was determined in sodium-phosphate buffer (pH 7.4). Approximately 2 mg of each peptide was added to 500 µl of buffer in 1.5-ml Eppendorf tubes. The tubes were shaken at 25 °C for 2 h on a 900 r.p.m. vortex shaker. The samples were centrifuged at 12,000 r.p.m. for 15 min to remove the precipitate and the supernatant was analyzed by ultravioletvisible light (UV-vis) spectrometry.

Circular dichroism spectroscopy.
Far-UV circular dichroism spectra of the peptides were recorded on a JASCO-715 spectropolarimeter using 0.1 cm path-length cuvette at 50 μM concentration (SIH-1 and SIH-2 were measured at 15 μM concentration) in 20 mM sodium-phosphate buffer (pH 7.4). Scans were carried out at 20 and 95 °C and again at 20 °C after fast refolding (10 min) over the range of 190-260 nm with 0.5-nm increments and a 2-nm bandwidth. Thermal unfolding experiments were monitored at 222 nm over a temperature range of 4-98 °C with 0.5 °C interval in 20 mM sodium-phosphate buffer (pH 7.4). The samples were equilibrated for 2 min at each temperature. Temperature-dependent circular dichroism data were fit to a two-state unfolding model to obtain the melting temperature (T M ).
Molar extinction coefficients were calculated for the peptides based on the number of Trp and Tyr residues in their sequences. In the case of SIH-6 and SIH-8, the molar extinction coefficient for unnatural amino acids l-2-napthylalanine and l-1-napthylalanine were used 3,936 M −1 cm −1 as reported earlier 47 . Mean residue ellipticity ((θ) deg cm 2 dmol −1 ) and percentage helicities were calculated using equations (1) and (2), respectively.
[θ] = θ observed /(10 × l × c × n) (1) where θ observed is the ellipticity measured in millidegrees, l is the path length of the cuvette (cm), c is the concentration in moles per liter and n is the number of residues in the peptide.

Expression and purification of mammalian expressed RBD (mRBD) from
Expi293F. Proteins were expressed and purified as described previously 24 Tag removal. The His-tag was removed by subjecting proteins to digestion with PreScission HRV-3C protease (Protein, HRV-3C = 50:1) in PBS (pH 7.4) buffer and incubating at 4 °C for 16 h. The untagged protein was separated from the remaining tag protein and protease by immobilized metal affinity chromatography using Ni Sepharose 6 Fast flow resin (GE Healthcare). The unbound tag-free protein was collected, and protein concentration was determined by absorbance (A 280 ) using NanoDrop 2000c with the theoretical molar extinction coefficient calculated using the ProtParam tool (ExPASy).

SPR-direct binding kinetics determination of immobilized RBD to synthetic peptides.
Direct binding kinetics to immobilized mRBD were assayed by titration with synthetic ACE2 competing peptides (160-10 nM). Briefly, amine coupling GLM sensor chip was activated with sulfo-NHS, EDC (Sigma) reaction. mRBD was covalently coupled to roughly 3,500 RU using 10 mM sodium acetate buffer Analytical ultracentrifugation. Sedimentation-velocity analytical ultracentrifugation. The sedimentation-velocity analytical ultracentrifugation data for SIH-3, SIH-5, SIH-7, SIH-11, SIH-13, SIH-14 and SIH-15 were collected between the concentration range of 20-50 µM, and the data for BGF-1 and BGF-2 were collected at 100 μM concentration using an optima XL-I analytical ultracentrifuge equipped with absorbance optics with an An-50Ti 8 place rotor (Beckman Inc.). Peptides were dissolved in the buffer (20 mM Phosphate buffer pH 7.4). Sedimentation-velocity experiments were carried out at 40,000 or 42,000 r.p.m. at 20 °C using two-channel charcoal-filled centerpieces with Sapphire glass windows. Samples were loaded onto the two-sector centerpiece (400 µl of reference cells and 390 µl of sample cells). The velocity data were collected by scanning samples at a wavelength of 280 nm with a spacing of 0.003 cm and an average of three scans per step. The standard partial specific volumes of proteins were taken into a calculation that had a value of 0.73 cm 3 g −1 at 20 °C. Buffer density (roughly 1.00 g cm −3 ) and viscosity (roughly 0.01002 poise) were calculated using the program SEDNTERP and these values were kept fixed for all the peptides. Assembly states of peptides were analyzed by direct curve fitting of sedimentation boundaries using Sedfit 48 . Fit to data was selected based on the root mean square deviations (r.m.s.d.) less than 0.008. The sedimentation coefficients were normalized to 20 °C in water, S 20,w under standard conditions.

Sedimentation-equilibrium analytical ultracentrifugation.
Sedimentation-equilibrium experiments were carried out at 25,000, 33,000 and 42,000 r.p.m. at 20 °C using six channel charcoal-filled centerpieces with Sapphire windows. The equilibrium data of SIH-5 were collected by scanning samples at 280 nm at a spacing of 0.003 cm with an average of five scans per step. The equilibrium data were edited using WinREEDIT (J. Lary, National Analytical Ultracentrifuge Center) program, and edited data sets were then analyzed using nonlinear least squares using the program WINNONLIN (D. Yphantis, University of Connecticut; M. Johnson, University of Virginia and J. Lary, National Analytical Ultracentrifuge Center).
nanoDSF thermal melt studies. Equilibrium thermal unfolding of mRBD (−10x His-tag) protein with and without synthetic peptide was carried out using a nanoDSF (Prometheus NT.48) as described previously 39 . The mRBD was incubated with different synthetic helix-hairpin peptides (1:1 stoichiometry) for at least 10 min at room temperature before analysis. Two independent measurements were carried out in duplicate with 10 μM of mRBD protein in the presence or absence of equimolar concentration of peptide using a temperature range of 15-95 °C at 100% LED power and initial discovery scan counts (350 nm) ranging between 5,000 and 10,000.

SEC-MALS.
SEC-MALS of mRBD and the peptide-mRBD complex were performed using PBS (pH 7.4) buffer equilibrated analytical Superdex-200 10/300GL gel filtration column (GE Healthcare) on a Shimadzu HPLC. The mRBD and synthetic peptides were complexed in equimolar amounts (30:30 µM), 60 min at 4 °C before the assay. However, due to the low solubility of SIH-1 in phosphate buffer, the mRBD was incubated with SIH-1 at 12 µM concentration to obtain equimolar amounts (12:12 µM). Gel filtration resolved protein peaks were subjected to in-line refractive index (Waters Corp.) and MALS (mini DAWN TREOS, Wyatt Technology Corp.) detection for molar mass determination. The acquired data from UV, MALS and refractive index (RI) were analyzed using ASTRA 6.1 software (Wyatt Technology).

Negative staining-TEM (NS-TEM) sample preparation and data acquisition.
Freshly prepared spike protein and SIH-5 samples were incubated in the ratio 1:3 (spike protein:SIH-5) for 20 min at 25 °C. The reaction mixture was analyzed for the presence of uniformly distributed homogeneous spike protein-SIH-5 complexes using conventional NS-TEM as previously described 35 . In brief, carbon-coated Cu grids were hydrophilized for 30 s using a GloQube glow discharge system before the sample application. 3.5 μl of protein-peptide sample (roughly 0.1 mg ml −1 ) was applied on the glow discharged grids and incubated at room temperature for 1.5 min. The surplus sample was gently blotted off using Whatman filter paper, and negative staining was performed for 30 s with 1% fresh uranyl acetate solution. The stained grids were observed, and data were acquired at room temperature using a Talos L120C transmission electron microscope (Thermo Fisher Scientific) equipped with a Ceta (4,000 × 4,000 K) camera. Images were recorded with calibrated pixel sizes of 2.38 Å per pixel.
NS-TEM data processing. The negative staining electron micrographs were assessed using the EMAN v.2.1 software package. Spike protein-SIH-5 peptide complexes were picked manually as well as automatically. Coordinates of the 1,500 selected particles were extracted using e2boxer.py in EMAN v.2.1 software. Reference-free 2D classification was performed using simple_prime2D of the SIMPLE v.2.0 software package.
Cryo-EM sample preparation and data collection. Quantifoil R1.2/1.3 (Electron Microscopy Sciences) 300 mesh gold grids were glow discharged for 130 s at 20 mA in GloQube glow discharge system (Gatan Inc.) before sample application. Here, 3 μl of freshly prepared spike protein-SIH-5 sample was applied onto the glow discharged holey grids and incubated at room temperature for 10 s. The excess sample was blotted for 8 s at 100% humidity, followed by rapid plunging into liquid ethane using an FEI Vitrobot Mark IV plunger. Cryo-EM data were captured in Thermo Scientific 200 kV Talos Arctica transmission electron microscope (FEI) equipped with K2 Summit Direct Electron Detector (Gatan Inc.). Data were acquired using automated data collection software LatitudeS (Gatan Inc.) at a magnification of ×42,200 and pixel size 1.17 Å at specimen level. 80 e − /Å 2 total electron dose over defocus ranging from −0.75 to −2.5 μm was subjected to the sample. Data were recorded for a total of 20 frames for 8 s.
Cryo-EM data processing. Cryo-EM raw movie files were first imported into RELION v.3.1, and beam-induced motion correction was performed using MotionCor2 software. The motion-corrected images were then imported into cisTEM software 49 , and only data up to 8 Å resolution were screened and selected for further processing. A total of 4,794 of the best micrographs were processed using RELION v.3.1 software. The contrast transfer function of these micrographs was estimated using RELION implementation of CTFFIND v.4.1.13. Following an initial round of manual picking, reference-free 2D classification was performed to use as a reference for automated particle picking. A total of 1,057,624 particle coordinates were extracted at a box size of 600 pixels binned by 2, yielding a pixel size of 2.34 Å. The dataset was initially screened robustly by performing three rounds of 2D classification. The best classes from 2D classification were selected, and 376,308 particles were split into three classes for 3D classification (without imposing any symmetry, C1 symmetry). Class 2 (101,296 particles) was selected for further refinement. The selected particle set for Class 2 was extracted at a box size of 600 pixels yielding a pixel size of 1.17 Å. Further auto-refinement was performed using a soft mask generated against dimeric spike molecules. The refined map sharpened with a B factor of −120 was resolved at 6.3 Å resolution. The particles were further subjected to contrast transfer function refinement with correction for beam-tilt fitting, anisotropic magnification correction and per-particle defocus. This was followed by Bayesian polishing. The 'shiny' particles obtained were subjected to another round of auto-refinement with a soft mask followed by map sharpening with a B factor of −120 and resolved at 5.4 Å resolution. Local resolution estimation was performed using unfiltered auto-refined maps with ResMap.
To understand the structural details of spike protein and SIH-5 interaction, masked auto-refinement was performed based on a single spike protein. For masked refinement, a soft mask was applied on a single spike protein with a 3-RBD up SIH-5 bound state (101,296 particles), at a pixel size of 1.17 Å with a box size of 256 pixels. The auto-refined map was sharpened and resolved at 4.8 Å resolution. The particles were further subjected to contrast transfer function refinement with correction for beam-tilt fitting, anisotropic magnification correction and per-particle defocus. This was followed by Bayesian polishing. The 'shiny' particles obtained were subjected to another round of auto-refinement with a soft mask followed by map sharpening with a B factor of −120 and resolved at 4.47 Å resolution. For atomic model building, the 3D auto-refined map was sharpened using PHENIX. Local resolution estimation was performed using unfiltered auto-refine maps with ResMap.
Atomic model building and structural analysis. SIH-5 bound 3-RBD up single S protein map resolved at 4.47 Å was docked with the Protein Data Bank (PDB) 7KMS (ref. 50 ) (for spike trimer), and rigid body fitting was performed in UCSF Chimera. PDB 7KMS was taken as the initial model to perform phenix.dock_ in_map in PHENIX followed by refinement with respect to the cryo-EM map using phenix.real_space_refine. Then, the homology model for SIH-5 was fitted manually in UCSF Chimera. The whole model combining spike and peptide was docked in the cryo-EM map using phenix:dock_in_map followed by refinement with respect to the cryo-EM map using phenix:real_space_refinement. The refined atomic model and the cryo-EM density map were imported to COOT, and the side chain fit was corrected by manual fitting. After every round of fitting in COOT, the model was refined in PHENIX using phenix.real_space_refine. The final atomic model used for assessment was analyzed using MolProbity and validated using phenix:validation_tool in PHENIX.EM ringer score 51 was calculated using PHENIX. All structural assessments and representations were done using UCSF Chimera and UCSF Chimera X.
Entry-inhibition assays were done three times in two replicates by using synthetic peptides. The pseudovirus was incubated with serially diluted peptide in a total volume of 100 μl for 1 h at 37 °C as a standard assay format. The 293T-hACE2 cells were then trypsinized, and 1 × 10 4 cells per well were added to make up the final volume of 200 μl per well. The plates were further incubated for 48 h in a humidified incubator at 37 °C with 5% CO 2 . The time-course entry-inhibition assay was performed by incubating the virus with peptides for 0, 10 and 30 min. The 293T-hACE2 cells were plated in a 96-well plate a day before, containing 100 μl of growth media per well. The virus and peptide mixture was added onto the cells after the incubation period, and the plate was further incubated for 48 h: 0 min was indicated when no incubation of the virus with peptide was done. After 48 h of incubation of cells, 140 μl of cell culture media was removed and 50 μl of nano-Glo luciferase substrate (Promega Inc.) was added. After lysing the cells for 2-3 min, 80 μl of lysate was transferred to white plates and luminescence was measured using a Cytation-5 multi-mode reader (BioTek Inc.) The luciferase activity measured as Relative luminescence units from SARS-CoV-2 pseudovirus in the absence of peptide was considered as 100% infection and was used as a reference to calculate the percentage inhibition in the presence of peptides. The inhibitory concentration (IC 50 ) at which 50% inhibition of viral entry was seen was calculated.

Pseudovirus preparation and DLS measurements.
Exponentially growing HEK293T cells were transfected at the cell density of 50% by using SARS-CoV-2 spike and HIV-1-NL4.3 expression constructs. After 12 h of transfection, the media was removed, cells were washed with plain DMEM and fresh FBS-free DMEM containing penicillin-streptomycin was added. After 30 h of further incubation, virus supernatant was harvested, spun to remove cell debris and filtered through a 0.22 µM filter to obtain a clean virus supernatant. Pseudovirus production was examined by infectivity in HEK293T-hACE2 cells.
The freshly prepared virus was incubated at 37 °C with 200 nM of LCB1/ SIH-5/SIH-11 for 1 h. After incubation, the virus-peptide suspension was used for virus size determination. The average particle size distribution of virion-peptide complexes and their zeta potential was determined by a Malvern Zetasizer Nano ZS DLS system. All DLS data were collected and analyzed using Malvern Zetasizer software (v.8.01). The average diameters of pseudovirus or pseudovirus-peptide complexes were calculated as the mean size of particle population ± s.d. from three independent experiments.
Cell viability assay. The cell viability assay was performed by using the MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay. Briefly, Vero-E6 or A549 cells were plated at 20,000 per well in 100 μl of growth medium (DMEM with 5% FBS and penicillin-streptomycin) in 96-well microtiter plates for 16-24 h at 37 °C in a humidified chamber with 5% CO 2 . The next day, 50 μl of media was removed from each well and 50 μl of the peptide was added in a dose-dependent manner. 10% DMSO (final concentration v/v) was used as a positive control for the induction of cytotoxicity. The plate was further incubated for 48 h (37 °C, 5% CO 2 ), and 10 μl of MTT solution was added to each well. The plate was incubated for another 4 h in the humidified CO 2 incubator, after which 100 μl of dissolving solution (10% SDS in 0.01 M HCl) was added to each well. The plate was then incubated in a CO 2 incubator overnight. The next day, the plate was read for absorbance at 570 nm. The percentage viability of cells was recorded in comparison to the untreated cells.
Hamster experiments. Regarding ethics, all animal experiments were performed in the Virus BSL-3 Laboratory at the Centre for Infectious Disease Research, Indian Institute of Science, Bangalore, India, following CPCSEA (The Committee for the Purpose of Control and Supervision of Experiments on Animals) and ARRIVE guidelines 52 . The work plans were reviewed and approved by the Indian Institute of Science, Institute Animal Ethical Committee. Required number of Syrian golden hamsters (Mesorectums auratus) of both sexes (80-100 gm of weight) were procured from CPCSEA registered, Biogen Laboratory Animal Facility (Bangalore, India). The hamsters were housed in individually ventilated cages, maintained at 23 ± 1 °C temperature and 50 ± 5% relative humidity.
Prophylactic treatment of peptide. After acclimatization of 7 d in individually ventilated cages at the IISc Virus BSL-3 Laboratory, the hamsters were divided into three groups: (1) control-PBS treated (n = 3), (2) challenged with 5 × 10 6 PFU dose of SARS-CoV-2 (n = 5) and (3) SIH-5 treated (day 0 at −8 h) and challenged (n = 5). After taking the basal weight of all groups, hamsters were anesthetized with xylazine (10 mg kg −1 body weight) and ketamine (150 g kg −1 per body weight) cocktail intraperitoneally. Hamsters were treated intranasally (i.n.) either with a dose of 2.5 mg kg −1 SIH-5 in 100 μl of PBS or with PBS for the control group. Hamsters were challenged, 8 h later, with 100 μl of i.n. administered SARS-CoV-2 US strain (USA-WA1/2020 obtained from BEI resources) at 5 × 10 6 PFU per hamster, by sedating/ anesthetizing the hamsters with xylazine (10 mg kg −1 body weight) and ketamine (150 g kg −1 body weight) cocktail intraperitoneally. In addition, the control group of three hamsters were dosed similarly with PBS. The health of hamsters, body temperatures, body weights and clinical signs were monitored daily by an expert veterinarian. Clinical sign scoring systems were developed following the previous studies 19 with some modifications. In the present experiment considering 14 clinical signs, we measured average clinical scores based on the points system as follows, lethargy (1 point), rough coat (1 point On the fourth day after the challenge, all the hamsters were humanely euthanized by an overdose of xylazine through intraperitoneal injection. The left lobe of the lung was harvested and fixed in 4% paraformaldehyde for histopathological examination of the lungs. The right lobes were frozen at −80 °C for determining the virus copy number by quantitative PCR with reverse transcription (RT-qPCR).
Histopathological examination. After necropsy, the left lobe of the lung of each hamster was fixed in 4% of paraformaldehyde, and the next day, these were processed, embedded in paraffin and cut into 4 µm sections by microtome for hematoxylin and eosin staining. The lung sections were microscopically examined and evaluated for different pathological scores by a veterinary immunologist. Four different histopathological scores were assigned as follows: (1) percentage of the infected part of lung tissues considering the consolidation of lung, (2) lung inflammation scores, considering the severity of alveolar and bronchial inflammation, (3) immune cell influx score, considering the infiltration of lung tissue with the numbers of neutrophils, macrophages and lymphocytes and (4) edema score, considering the alveolar and perivascular edema. The scores and parameters were graded as absent (0), minimal (1), mild (2), moderate (3) or severe (4) 53 .
RNA extractions and RT-qPCR to quantitate subgenomic virus copy in lungs. The freeze-thawed right lower lobe of each hamster lung was homogenized in 1 ml of RNAiso Plus Reagent (Takara), and total RNA was isolated as per the manufacturer's protocol. The quantity and quality (260/280 ratios) of RNA extracted was measured by NanoDrop. The extracted RNA was diluted to 27 ng μl −1 in nuclease-free water. The viral subgenomic copy number was quantified using 100 ng of RNA per well for 10 μl of the reaction mixture using AgPath-ID One-Step RT-PCR kit (AM1005, Applied Biosystems). The following primers and probes were used 2019-nCoV_N1-Fwd-5′-GACCCCAAAATCAGCGAAAT-3′; 2019-nCoV_N1-Rev-5′-TCTGGTTACTGCCAGTTGAATCTG-3′; 2019-nCoV_ N1 Probe (6-FAM/BHQ-1) ACCCCGCATTACGTTTGGTGGACC (Sigma Aldrich) for targeting SARS-CoV-2 N-1 gene. The subgenomic virus copy number per 100 ng of RNA was estimated by generating a standard curve from serial dilution of SARS-CoV-2 genomic RNA standard.
Statistical analysis. Data are described as mean ± s.d. unless otherwise stated. The P value for hamster weight change between virus challenged and SIH-5 treated was analyzed by a two-way analysis of variance test using the GraphPad Prism software.
The P values for viral load in the lungs and histopathological scores were analyzed by a two-tailed Mann-Whitney test using the GraphPad Prism software (v.8.0.1).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The cryo-EM maps and atomic models have been deposited in the Electron Microscopy Data Bank and the PDB with accession codes: EMD-32388, EMD-33042 and PDB entry ID 7X7N. All other data are present as source data provided with this paper or in the Supplementary Information.