C-2 Thiophenyl Tryptophan Trimers Inhibit Cellular Entry of SARS-CoV-2 through Interaction with the Viral Spike (S) Protein

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) causes COVID-19, by infecting cells via the interaction of its spike protein (S) with the primary cell receptor angiotensin-converting enzyme (ACE2). To search for inhibitors of this key step in viral infection, we screened an in-house library of multivalent tryptophan derivatives. Using VSV-S pseudoparticles, we identified compound 2 as a potent entry inhibitor lacking cellular toxicity. Chemical optimization of 2 rendered compounds 63 and 65, which also potently inhibited genuine SARS-CoV-2 cell entry. Thermofluor and microscale thermophoresis studies revealed their binding to S and to its isolated receptor binding domain (RBD), interfering with the interaction with ACE2. High-resolution cryoelectron microscopy structure of S, free or bound to 2, shed light on cell entry inhibition mechanisms by these compounds. Overall, this work identifies and characterizes a new class of SARS-CoV-2 entry inhibitors with clear potential for preventing and/or fighting COVID-19.


■ INTRODUCTION
The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), discovered in December 2019 as a new betacoranovirus, 1 is responsible of one of the largest pandemics the world has suffered in recent years, coronavirus disease 19 . 2 Prior to the emergence of SARS-CoV-2, six different coronaviruses were known to infect humans, most of which caused mild respiratory diseases. 3 The exceptions were two zoonotic betacoronaviruses, severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV), that caused high mortality in relevant epidemic outbreaks. 3,4 Unfortunately, no antiviral drugs against coronavirus infections were available at the time COVID-19 emerged, leaving the world defenseless against this new disease. In the search for nonbiological antivirals, special emphasis was placed on drug repurposing to accelerate the clinical implementation of effective drugs. As a result, remdesivir, a prodrug that targets viral RNA synthesis and that was initially developed against Ebola and Marburg viruses, 5 was approved for treatment of hospitalized COVID patients ( Figure  1). 6 More recently, two orally available drugs, molnupiravir, a prodrug of 4-hydroxycytidine, 7 and a combination of nirmatrelvir and ritonavir, named Paxlovid, 8 have received emergency authorization ( Figure 1). Despite these advances, concerns about the potential side effects of these drugs (particularly of molnupiravir), their interference with other drugs (in the case of Paxlovid), the potential appearance of resistant viral strains, and/or the possibility of insufficient therapeutic efficacy warrant the search for novel antiviral drug candidates.
The early stages of viral infection, namely, attachment to host cells and entry, represent attractive targets for antiviral therapy as their inhibition can block infection prior to the exponential growth phase of the virus. 9 In the case of coronaviruses, the spike protein (S), which protrudes from the viral envelope and gives viruses of this family their characteristic appearance, plays a crucial role in mediating viral entry and represents a key target for recognition by the host immune system. 10,11 In SARS-CoV-2, S is a transmembrane glycoprotein that forms homotrimers on the viral membrane and that is cleaved during virus maturation into two subunits, S1 and S2, that remain associated in the viral membrane-bound S trimer. 9,12 The receptor binding domain (RBD), located within S1, mediates the binding to the primary host receptor, the human angiotensin converting enzyme 2 (hACE2), while membrane fusion and viral entry are driven by the S2 domain. 12 Antibodies targeting the S protein are able to neutralize the virus, and several monoclonal antibodies have been used for COVID19 treatment, although reduced efficacy has been observed against new variants. 13,14 Moreover, antibody-based therapies are costly, 15 prohibiting their application in resource-poor settings. Hence, the use of compounds that can block viral entry represents an attractive antiviral strategy that could synergize with approved therapies to increase therapeutic efficacy, reduce the probability of drug-resistance, and cut down treatment costs.
Prior work from our group has shown multivalent functionalized tryptophan (Trp) derivatives to be potent inhibitors of different viruses, including human immunodeficiency virus (HIV), enterovirus 71 (EV-A71), and flavivirus infections, and to show low cytotoxicity. 16−22 Mechanistic studies demonstrated that these compounds interact with key elements of the viral surface (glycoprotein gp120 of HIV, 19 5-fold axis of the EV-A71 capsid 23 and domain III of the viral envelope glycoprotein dengue 2 virus 22 ) preventing virus attachment to the host cell membranes. In previous studies, the existence of carboxylates at the Trps was shown to be critical for antiviral activity. In addition, multivalency was proven to play an important role in the antiviral activity of this class of compounds, in which multiple tryptophan residues are exposed in such a way that their indole side chain and carboxylate groups map toward the periphery of the small dendrimer. It should be mentioned that there are other initiatives to target viral entry that also rely on multivalent inhibitors. 24,25 In this work, we screened our in-house Trp multivalent compounds for inhibition of SARS-CoV-2 entry using a highthroughput screening (HTS) assay based on pseudotyped vesicular stomatitis virus expressing the S protein of SARS-CoV-2 (VSV-S). The chemical space surrounding the top hit (defined as the compound of low cytotoxicity exhibiting the strongest antiviral activity) was explored via the synthesis of new analogues. These were screened in the VSV-S assay to identify improved compounds and validated against genuine SARS-CoV-2. Mechanistic studies confirmed that these compounds block viral entry. Thermofluor and microscale thermophoresis with pure proteins proved the binding of these compounds to the RBD in the spike of SARS-CoV-2 and corroborated their capacity to interfere with the binding of these proteins to hACE2. Finally, the structure of SARS-CoV-2 S bound to one the best inhibitors was obtained at near atomic resolution using cryoelectron microscopy (cryoEM), providing plausible structural mechanisms for the observed cellular entry-blocking antiviral activity.

High-Throughput Screening (HTS) Assay for Identification of SARS-CoV-2 Entry Inhibitors.
To assess whether our multivalent Trp derivatives could inhibit SARS-CoV-2, we implemented a HTS assay in Vero cells based on a pseudotyped VSV. Briefly, VSV lacking its own glycoprotein and encoding both GFP and luciferase was produced in cells that are engineered to express the ancestral Wuhan-Hu-1 SARS-CoV-2 S protein (VSV-S). 26 In the process of budding from the cell, VSV is coated with the S protein. This enables the viral particles to employ the S protein to enter cells via interaction with the ACE2 receptor in an analogous manner to SARS-CoV-2, and infection can be monitored by quantification of GFP fluorescence. Using this system, 50 multivalent Trp derivatives were tested at an initial concentration of 20 μM (data not shown). From this primary screening, tetramer 1 and trimer 2 ( Figure 2) showed significant antiviral activity in the absence of cytotoxicity ( Figure S1). Both compounds share a 4-NO 2thiophenyl ring attached to the C-2 position of the indole ring of each Trp residue. Thus, we screened related Trp derivatives with this functionalization and also identified compound 3 as being effective ( Figure 2 and Figure S1).
We next compared the antiviral activity and cytotoxicity of compounds 1−3 in the VSV-S assay in Vero cells. While the three compounds showed low cytotoxicity [concentration resulting in 50% cell death (CC 50 ) > 100 μM], the concentration that reduced virus infection by 50% (IC 50 ) was far better for trimer 2 (0.64 ± 0.47 μM) than for tetramer 1 and trimer 3 (respective IC 50 values, 21.43 ± 9.45 and 32.85 ± 4.48 μM). Thus, compound 2 with a Gly linked to the NH 2 at the central quaternary carbon (the focal point) was identified as a suitable hit and prompted us to synthesize trimers of general formula I ( Figure 2) bearing S-phenyl rings with different substituents (R 1 ) at the C-2 position of the indole ring for antiviral evaluation. Alternatives to the thioether (X) were also explored to attach the extra phenyl ring to the C-2 position of the indole. Moreover, the introduction of different chains (R 2 ) at the focal point was investigated with the aim of modifying the physicochemical properties of the compounds (i.e., lipophilicity).
Synthesis. Although compounds 1−3 were part of our inhouse collection, their synthesis has not been previously disclosed. The three compounds were synthesized using the Journal of Medicinal Chemistry pubs.acs.org/jmc Article divergent approach shown in Scheme 1, involving the reaction of a multivalent methyl-protected tryptophan derivative with pnitro-benzenesulfenyl chloride (pNPS-Cl) under acidic conditions, so that sulfenylation occurs selectively at the C-2 position of each indole ring. The synthesis of the tetrapodal derivative 1 was accomplished by reaction of the tetramer 4 21 with pNPS-Cl in the presence of acetic acid to afford intermediate 5 (71%). The subsequent saponification of the ester moieties (LiOH·H 2 O) gave the desired final compound 1 in 92% yield. In the case of trimer 2, the tripodal scaffold 7 was first prepared by reaction of the triacid 6 27 with OMe-protected Trp (H-TrpOMe·HCl) in the presence of HATU, as the coupling reagent, and N,N-diisopropylethylamine (DIPEA) as the base (Scheme 1). Then, trimer 7 was reacted with pNPS-Cl in the presence of acetic acid to afford the C-2 sulfenylated derivative 8. Subsequent saponification of the protecting ester moieties using LiOH·H 2 O, accompanied with the simultaneous removal of the Fmoc group, afforded the trimer 2 in 90% yield. The trimer 3 was obtained in 95% yield from the OMe-protected intermediate 9, previously described by our group 22 using a similar methyl ester saponification.
To synthesize compounds of general formula I following the synthetic strategy described in Scheme 1, a great variety of sulfenyl chlorides were needed. However, these chlorides are unstable and/or not commercially available. These limitations prompted us to explore a new synthetic strategy. Among the procedures described for the sulfenylation of indoles, a particularly appealing strategy was the "one-pot" tetrabutylammonium iodide (TBAI)-mediated sulfenylation using sulfonyl chlorides. 28 This procedure employs metal-free conditions and is compatible with benzenesulfonyl chlorides functionalized with electron-donating and electron-withdrawing groups at position 2, 3, or 4 of the phenyl ring. 28 However, reaction of the trimer 7 with 4-NO 2 PhSO 2 Cl in the presence of TBAI in DMF failed to yield the expected trimer 8. This unsuccessful result prompted us to study these reaction conditions on a simpler Figure 2. Trp derivatives 1−3, which were initially identified as having antiviral activity and low cytotoxicity in the VSV-S HTS assay, while I represents the general structure of the compounds here synthesized and tested.
substrate, FmocTrp(OMe) 29 (10) (Scheme 2). The election of Fmoc as a protecting group of the Trp was based on its stability under the acidic conditions generated in the sulfenylation reaction (due to the presence of IH) and its easy removal under basic conditions. However, reaction of 10 with 4-NO 2 PhSO 2 Cl also failed to provide the C-2 sulfenylated derivative 11, with most of the starting material remaining unaltered. According to the mechanism proposed by He et al., 28 the reaction of TBAI with the sulfonyl chlorides in DMF generates the corresponding disulfides together with I 2 , so that these are indeed the reactive species, as also reported by other groups. 30,31 Thus, we performed the reaction of FmocTrp(OMe) (10) with 4-NO 2diphenyl disulfide in the presence of I 2 in acetonitrile at 60°C for 4 h. In this way, the C-2 sulfenylated Trp compound 11 was obtained in 75% yield (Scheme 2).
Using this approach, different C-2 sulfenylated derivatives were synthesized (12−18), including those with a NO 2 group at positions 2 or 3 of the thiophenyl ring (compounds 12 and 13), or with other functional groups at position 4 (CF 3 , CN, F, COCH 3 , and SO 2 CH 3 , 14−18, respectively), with yields varying from 35 to 83%. The diphenyl disulfides required to obtain the intermediates 11−18 were commercially available in most cases, except for a few (R 1 = R 2 = H; R 3 = F, COCH 3 or SO 2 CH 3 ), which were synthesized from the corresponding benzenesulfonyl chlorides (19a−c) by reaction with TBAI in DMF at room temperature, as described for similar analogues 32 (for details see the Supporting Information). In our hands, this procedure to obtain these disulfides (20a−c) was simpler than previously described alternative methods. 33 Alternatives to the thioether used to link the 4-NO 2 -phenyl ring and C-2 position of the indole were also explored. A direct C−C bond was assayed through a metal-catalyzed (Pd II) crosscoupling reaction (Scheme 4). For this, a mixture containing H-Trp OMe (43) and 1-iodo-4-nitrobenzene in DMF was MWirradiated at 120°C for 30 min in the presence of 5 mol % Pd(OAc) 2 , AgBF 4 , and TFA, 35 to afford the C-2-arylated derivative 44. Subsequent HATU-mediated coupling of this Trp-derivative with the protected glycine scaffold 6 provided 45 in high yield. Finally, treatment of 45 with LiOH·H 2 O led to the acid 46.
We also envisioned the attachment of the 4-NO 2 -phenyl group to the C-2 position of the indole of Trp through a CO linker. To this end, Fmoc-Trp-OMe (10) reacted with 4-NO 2benzoyl chloride in the presence of SnCl 4 in anhydrous DCM at 0°C, as described for a similar indole derivative, 36 to afford the C-2 substituted derivative 47. Unexpectedly, reaction of 47 with piperidine led to the removal of the Fmoc group and concomitant cyclization and aromatization to afford the βcarboline 48. Previous synthesis of this compound involved a Pictet-Spengler condensation of the Trp derivative with 4-NO 2benzaldehyde followed by oxidation of the tetrahydro-βcarboline thus formed. 37 To avoid this cyclization, the acylation reaction was performed at the trimer 7 using 4-NO 2 -benzoyl chloride (4.5 eq) and SnCl 4 (9 eq). In this way, the acylated derivative 49 was obtained. Subsequent treatment with LiOH·H 2 O afforded the deprotected trimer 50.
As will be later discussed, compound 2 was still the compound providing the best antiviral activity in the VSV-S assay. Thus, the next set of modifications involved the introduction of fatty acid chains of different lengths (from 4 to 10 methylenes) at the NH 2 -group of the glycine moiety at the focal point in compound 2 to modulate the hydrophobicity of the resulting compounds (Scheme 5). Selective NHFmoc deprotection of the glycine intermediate 8 by treatment with piperidine afforded the free NH 2 derivative 51 (73% yield). Acylation reaction of 51 with aliphatic acyl chlorides with alkyl chains of different lengths in the presence of propylene oxide in dichloromethane afforded derivatives 52−55. Saponification of these methyl esters with LiOH·H 2 O gave the final compounds 56−59 (Scheme 5). As will be later discussed, these acyl derivatives maintained antiviral activity in the pseudotyped VSV-S antiviral assay. Thus, other functionalized chains at the focal point were also explored.
Reaction of 51 with monomethyl adipate in the presence of HATU and DIPEA afforded the ester derivative 60 in 53% yield (Scheme 6). Saponification of the four ester groups in 60 by treatment with LiOH·H 2 O afforded compound 61 in 60% yield. Similarly, reaction of 51 with Fmoc-9-amino-4,7-dioxanonanoic acid, also with HATU and DIPEA, provided the NHFmoc derivative 62 in 57% yield. Treatment of 62 with LiOH·H 2 O led to the saponification of the methyl ester moieties and concomitant removal of the Fmoc group to provide compound 63 in 91% yield.
Finally, we envisioned the synthesis of a dimer containing two units of compound 2 as a way to increase multivalency. Based on the antiviral data obtained with the compounds modified at the focal point (56−59, 61, and 63), a glycol linker was selected as a connecting unit. Thus, reaction of the Gly derivative 51 (2 equiv) with 4,7,10,13-tetraoxohexadecane-1,16-dioic acid (Scheme 7) (1 equiv) in the presence of HATU and DIPEA in DMF at 30°C for 48 h afforded the dimer 64 in 65% yield.
Saponification of the methyl esters by treatment with LiOH· H 2 O at rt overnight led to compound 65 in 81% yield.
In summary, we synthesized new Trp trimers substituted at position 2 of each indole with differently functionalized aryl rings (R 1 in general formula I). In most cases, the aryl ring is connected through a thioether to the C-2 position of the indole (X = S in I). Different chains have been incorporated at the NH 2 group of the Gly moiety at the focal point (R 2 in I), and this approach has been used to synthesize the dimer 65 connecting two units of compound 2 through a glycol spacer. Antiviral Evaluation. The compounds were evaluated using the VSV-S assay for their antiviral activity and cytotoxicity. Specifically, VSV-S was mixed with serial dilutions of the compound prior to infection of either Vero E6 or A549-Ace2-TMPRSS2 cells, and the concentration of compound resulting in 50% reduction of virus infection (IC 50 ) and 50% reduction in cell viability (CC 50 ) values were determined in parallel in the same wells following 16 h of culture (Table 1 and Supplementary Figure S1). Of note, none of the compounds reached 50% cytotoxicity at the highest concentration tested (100 μM), highlighting the low cellular toxicity of these compounds.
Overall, the antiviral activity of the compounds correlated well between Vero E6 and A549-ACE2-TMPRSS2 cells (Spearmen's rho = 0.56, p < 0.05; Figure S1). As already mentioned in the hit identification section, of the three initial compounds of our home library, compound 2, a trimer with a NHCOCH 2 NH 2 at the focal point, showed a more potent antiviral effect than the tetrameric compound 1 or the trimeric compound 3 that has a NO 2 group at the focal point. Thus, the skeleton of 2 was maintained in the next round of structure−activity relationship (SAR) studies. The data obtained with compounds 36−42 highlighted the importance of the substituent on the phenyl ring for the antiviral activity. Moving the 4-NO 2 group to positions 2 or 3 of the phenyl ring (36 and 37, respectively) led to significantly less potent compounds than those with the NO 2 group at position 4 (as in hit 2). Replacement of the 4-NO 2 group in 2 by other electron withdrawing groups (CF 3 in 38, CN in 39, or F in 40) also led to significantly less potent compounds. Of these derivatives, only 41, with a COCH 3 group at position 4, resembled compound 2 in having IC 50 values around 2 μM in the assays with both cell types. However, a similar compound with a SO 2 CH 3 group at position 4 (compound 42) had no antiviral activity. Interestingly, when the 4-NO 2 -phenyl ring is directly attached at the C-2 position of the indole (as in compound 46) or the 4-NO 2 phenyl is linked to the C-2 through a CO unit (compound 50), the antiviral activity is also lost.
Introduction of alkyl chains through acylation of the NH 2 of the Gly at the focal point (compounds 56−59) also diminished the antiviral activity in Vero E6 cells, although 56−58 showed IC 50 values around 10 μM in A549-ACE2-TMPRSS2 cells. Interestingly, 61, which carries a four-methylene alkyl chain and distal COOH group, showed IC 50 values around 2 μM in both cell lines, considerably better than the same analogue with a distal methyl group (compound 56). An acylated derivative with a glycol chain and terminal amino group (compound 63) had similar antiviral activity as the hit compound 2. Finally, the dimer 65 containing two units of 2 linked through a polyethylene glycol spacer had an IC 50 value of 0.28 ± 0.23 μM in Vero E6 cells. Thus, the data obtained point toward the importance of both the substituent at the aryl ring linked to the Trp and the linking moiety, so that the combination of thioether as the linker and the 4-NO 2 group at the aryl ring provides the best antiviral activity (compounds 2, 61, 63, and 65).
Mechanism of Action Studies in the Cell Culture. As compound 2 and some of its derivatives exhibited low IC 50 values in the VSV-S assay, we sought to define the mechanism by which these compounds inhibit infection. First, to examine if the compounds acted on the entry step, we compared the antiviral effect of adding the drug during the entry process or one-hour postinfection, when entry has been completed ( Figure 3A). For this assay, a high concentration (100 μM) of compound 2 was employed to ensure that even weak effects would be detected. While a strong reduction in infection (assessed as the degree of viral-produced GFP reporter expression) was observed when the compound was added together with the virus, no decrease in infection was observed when the addition of compound 2 was delayed until one-hour post virus addition, when the entry process was completed by removal of the viral inoculum ( Figure  3A). Since addition of the compound after infection resulted in a loss of antiviral activity, these data indicate that compound 2 specifically targets the entry process. Next, we evaluated the specificity of compound 2 for interfering with S-mediated entry.
For this, we generated a VSV pseudotyped with its native glycoprotein (VSV-G). As all steps in the infection process are identical between VSV-S and VSV-G, with the exception of the glycoprotein used for entry, these viruses can be used to assess the specificity of antivirals targeting S-mediated entry. Unlike for VSV-S, no antiviral activity was observed when VSV-G was preincubated with compound 2 prior to infection of cells ( Figure  3A), indicating that compound 2 specifically blocks S-mediated entry. Finally, we compared the effects of preincubation of the compound with the virus for 1 h prior to infection of cell or preincubation of the cells for 1 h, followed by addition of the virus ( Figure 3B). In both cases the compound was present when the virus was added to the cells. No significant difference in the antiviral effect at different concentrations of compound 2 was observed (IC 50 of 0.48 μM ± 0.14 vs 0.32 μM ± 0.13 when preincubated with the virus or with the cells, respectively). This suggests that the binding of compound 2 to its target is faster than the interaction of the virus with its receptor.
Validation of Antiviral Activity against SARS-CoV-2. To validate the results obtained with the VSV-S pseudotype assay, we next tested compounds 2, 63, and 65 against SARS-CoV-2 infection in two cell lines that support robust virus replication: Vero E6-TMPRSS2 and A549-ACE2 cells ( Table  2). For this, a SARS-CoV-2 virus carrying the D614G S mutation was preincubated with diluent alone (mock) or with the compounds at 10 μM prior to addition to the cells, and virus production was assayed after 24 h via limiting dilution. Remdesivir (10 μM) was used as a positive control. All compounds reduced virus production by >98% in A549-ACE2 cells and by >89% in VeroE6-TMPRSS2 cells, confirming the strong antiviral activity of these compounds against SARS-CoV-2.
Antiviral Activity of Selected Compounds against the Omicron BA.1 Variant. SARS-CoV-2 has undergone significant evolution since its emergence. To assess whether the identified compounds could inhibit the replication of newer SARS-CoV-2 variants, we next studied the ability of compounds 2, 63, and 65 to inhibit the replication of pseudotyped VSV carrying the S protein of the SARS-CoV-2 Omicron BA.1 virus (VSV-S Omicron ) compared to VSV carrying the S protein of the Wuhan-Hu-1 strain (VSV-S Wuhan-Hu-1 ) used for the initial screening (Table 3). Vero E6-TMPRSS2 cells were used in these experiments as they displayed higher susceptibility to infection by BA.1 pseudotyped VSV-S (data not shown). All three compounds showed antiviral activity against pseudotyped VSV-S Omicron at noncytotoxic concentrations. However, IC 50 values were 1−3 orders of magnitude higher for VSV-S Omicron than for VSV-S Wuhan-Hu-1 (Table 3). These data can be explained by the large number of mutations that the Omicron BA.1 variant accumulates in the S gene (35 mutations). Nevertheless, compound 65 exhibited moderate antiviral activity against this omicron variant, with an IC 50 value <10 μM in VSV-S Omicron assays.

Thermofluor Assays Indicate that Active Compounds Bind to the Receptor Binding Domain of the S Protein.
Having proven that the synthesized compounds inhibit SARS-CoV-2 infection by preventing viral entry, the next step was to determine if they interact with the S protein. With this aim, we performed a qualitative test using thermofluor assays. 38 This technique monitors changes in fluorescence of a protein-binding dye resulting from protein unfolding in solution due to a gradual increase of temperature. The binding of a ligand can modify the thermal stability of a protein, which changes the fluorescence profile obtained. The impact of such binding can be quantified by the increase or decrease of the temperature (T m ) at which the increase in fluorescence is 50% of the maximum fluorescence change. The target protein in our case was the purified recombinant RBD domain, produced in a baculovirus/insect cell system to guarantee its glycosylation (see the Experimental Section). Figure 4A shows the fluorescence profile for the RBD domain in the absence of the ligand, and the shift of the curve toward lower temperatures in the presence of a fixed concentration (100 μM) of compounds 2, 38, 41, 57, 61, and 65. All these compounds had been found to inhibit viral entry in the cellular assays (Table 1). In marked contrast, compound 42, which did not significantly inhibit viral entry, did not cause any change in the curve ( Figure 4A). Figure 4B illustrates the differences in T m values found in the presence of these compounds relative to the ligand-free protein, showing for all of them except 42, differences that are statistically significant.
Next, we tested the effect of variable concentrations of four of these compounds (2, 41, 61, and 65) on the binding ( Figure  4C). Our results showed that compounds 2 and 65, the two entry inhibitors that showed the lowest IC 50 in the cellular infection assays (IC 50 : 0.6 and 0.3 μM, respectively), caused RBD destabilization at lower concentrations than 41 and 61, which exhibited higher IC 50 in cellular infection assays (IC 50 : 3 and 2.1 μM, respectively).
In summary, the thermofluor assays indicate that the compounds that have a substantial effect on blocking viral entry are able to bind the RBD domain of the S protein.
Microscale Thermophoresis (MST) Provides Additional Proof of Binding of Active Compounds to the S

Protein, Revealing Competition with ACE2 Binding.
Further and more quantitative evidence of active compound binding to the S protein in its RBD was obtained using MST, a technique that titrates the influence of ligand concentration on the fluorescence change when heat is applied locally to capillary tubes hosting solutions of the fluorescent macromolecule and the ligand. 39 We used purified RBD and S proteins fluorescently labeled via the binding of a fluorophore to their respective polyHis-tags, titrating the effects of increasing concentrations of compounds 2 and 65 (the compounds that showed the highest potency in the cell culture, see above) as exemplified in Figure  5A. From the fluorescence traces obtained, the fraction of saturation of the protein by each concentration of the ligand can be estimated (see the legend of Figure 5).   Figure 5B illustrates the results of these binding assays, which fit single-site binding. K D values did not differ significantly for binding of 2 or 65 to the RBD, or for the binding of 2 to the complete S protein (mean ± SE values, 37.8 ± 4.6, 35.4 ± 5.6, and 30.6 ± 3.0 μM, respectively), while the K D value for the binding of 65 to S was somewhat lower (K D , 13.6 ± 1.7 μM; p < 0.05). Compound 42 was also tested and was found to be a very poor binder (K D ≥ 2.5 mM for binding to the RBD, Figure 5B), in line with its lack of substantial effects on VSV-S cell entry assays (Table 1) and on the thermal stability of the RBD ( Figure  4B). Thus, these results confirm that compounds 2 and 65 bind to the S protein in its RBD.
Next, the binding of the unlabeled recombinant catalytic domain of ACE2 to the fluorescent RBD domain was titrated in the absence and presence of 0.5 mM of 2 or 65. Our results showed that the concentrations of ACE2 needed for binding to the RBD domain were significantly (∼10-fold) increased by the presence of 0.5 mM of any of these two compounds ( Figure 5C). From this experiment, we concluded that compounds 2 and 65 compete with the cellular receptor ACE2 for the binding to the RBD domain. This competition was also observed when titrating the binding of ACE2 to the complete S protein, as exemplified for compound 65 in Figure 5D. These results help to explain how the synthesized compounds could act as entry inhibitors.
Detection of Compound 2 in the Structure of the Viral Spike Explains Viral Cell Entry Inhibition. Prior structural work at near-atomic resolution has shown S to be homotrimeric and to adopt distinct confirmations that can influence its ability to bind hACE2. Specifically, the RBD domains in the S trimer can adopt an "up" conformation or be buried inside S ("down" conformation). Only S trimers with one RBD in the up position (one RBD-up) can bind hACE2, which then promotes adjacent RBDs to adopt an open conformation that further increases the binding affinity for ACE2. 40,41 We obtained the cryo-EM structure of the SARS-CoV-2 S protein bound to compound 2, used as a prototype of present series of compounds with antiviral cell-entry inhibition. We used an S protein harboring the D614G mutation present in all major SARS-CoV-2 strains except the ancestral Wuhan-Hu-1 strain. From a dataset of 3500 movies ( Figure 6A−C), two main conformational populations were isolated by 3D EM data classification, without applying any internal symmetry restriction (see the Experimental Section for details). The most abundant conformation (74% of the particles; 3.4 Å resolution) had one RBD in up position and two in down position per S protein trimer ( Figure 6D, top panel). The remaining 26% of the particles had the three RBDs in down positions (4.3 Å resolution; Figure 6D, middle and down panels). Comparison of the present dataset with our recently published datasets, 42 which were prepared in an identical manner in the absence of any compound, showed that the 3-down class of the spike is highly enriched in the presence of compound 2 (24% of the particles versus <2% of the particles without drug). This finding is further supported by similar cryo-EM structures reported by others in which the conformation with the 3 RBDs down was represented in a small minority of the particles. 41,43 Hence, compound 2 appears to specifically induce a three RBD-down class that should inhibit binding to ACE2. Indeed, the structure of the spike in the one RBD-up conformation found here is very similar (root mean square, RMSD, of 1.27 Å for the superimposition over 3219 Cα atoms) to that of the same spike variant in the absence of compound 42 (PDB ID:7QDH). Detailed inspection did not reveal remarkable conformational differences nor additional nonprotein densities that could be attributable to the binding of compound 2. However, in the structure derived from the three RBD-down map, a clear density was found next to and between two RBDs ( Figure 6D, middle and down panels).
Although the local resolution of the map was not optimal in the site of this density, the approximate size of this extra density grossly fitted what would be expected for compound 2 ( Figure  6E). The limited resolution precludes more accurate compound 2 fitting into the density, or an unequivocal identification of residues interacting with this compound, although it is possible to confidently assign the closest residues to this drug ( Figure  6F). These residues are: N343 (with clear structural evidence of its glycosylation), L335, P337, G339, E340, D364, S366, and V367, all from subunit A, and S477, T478, F486, and N487 from the neighboring subunit C. Some of these residues are mutated in the Omicron BA.1 variant (G339D, S477N, and T478K) that may account for the reduced potency observed for the omicron variant (Table 3). Apart from the extra density assigned to compound 2, the spike in three RBD-down conformation is overall quite similar to other spikes having in their structures, all their RBDs in a down position. This is reflected in RMSDs of 2.02−2.04 Å for the superposition over the entire spike structures (2699 and 2916 Cα, respectively) in PDB IDs 7KRQ 44 and 7BNM. 41 The three subunits in the S protein form having its three RBDs in the "down" position are highly similar (RMSDs < 1 Å for mutual superpositions).
Biophysical in vitro assays showed that compound 2 inhibits the interaction of the isolated RBD domain with the ACE2 receptor (see Figure 5C above), suggesting direct competition of compound 2 with the ACE2 receptor for interaction with the RBD. This suggestion is substantiated structurally here, since ACE2 and compound 2 share the same RBD binding interface ( Figure 6G). This primary inhibitory mechanism could be enhanced if compound 2 physically holds two out of three RBDs of the spike in their down position, preventing the establishment of proper contacts of these RBDs with ACE2. Further enhancement of the inhibitory effect could be due to the possibility of an allosteric effect by which compound 2 binding could prevent the adoption of the up position in the third of these RBDs. This is supported by the facts that, in the spikes in 1up or 2-up conformations, the RBDs in up position/s interact with neighboring RBDs in the down position, 41−43 while in the present 3-down structure, the contact with compound 2 Figure 5. continued and F ∞ are the fluorescence in the absence, at a given concentration, and at infinite concentration of the ligand that is varied, respectively. F ∞ was estimated from the hyperbolic fitting. In the case of compound 42, the data were fitted to the minimal possible value of K D accepted by the fitting program (Graphpad Prism). Curves correspond to hyperbolic fitting (in semilog representation). Each point is the mean (±SE) for three different titrations. The K D values are the concentrations giving a half-maximum change. In panel B, given the lack of statistical differences in the K D values for 2 versus RBD and versus S, and of 65 versus RBD, a single hyperbola has been drawn fitting all the clumped points for these three data sets (K D , 34.6 μM). In panel C, a single hyperbola is shown for the clumped results for 2 and 65, given the lack of statistically significative differences between them (K D values of ACE2 for RBD, 55.1 ± 4 and 618 ± 86 nM in the absence or presence of the compounds, respectively). In panel D, K D values for the binding of ACE2 to S were 17.2 ± 2.2 and 388 ± 45 nM in the respective absence and presence of 0.5 mM 65. . Each subunit is shown in a different color (green, purple, and yellow). The top and middle panels represent side views of the spike with the molecular 3-fold axis of the trimer vertical, while in the bottom panel, the spike is seen from outside the virus, and the molecular threefold axis is perpendicular to the page. The nonprotein density observed (colored blue) in the middle and bottom panels appears to correspond to bound compound 2, present in the solution (see the Experimental Section), given the reasonably good gross fitting of this compound (in sticks) to the profile of the nonprotein density (panel E, in blue). (F) Zoom on this density profile to show nearby residues from two subunits. The Cα atoms of the indicated residues are localized with spheres in the backbone (yellow or green depending on the subunit) and are identified in single-letter amino acid notation. The N-glycosylation of N343 was visible on the map, and it is represented in sticks. (G) Superimposition of the structure of the backbone of the RBD of subunit C (colored yellow) of the spike in the 3-down conformation observed here, with that of the RBD (in gray) in the RBD-ACE2 complex (PDB 7A94 41 ), to illustrate that the density (blue) attributed here to compound 2 sits at the part of RBD that interacts with ACE2, strongly suggesting interference of compound 2 with the interaction of the spike with its receptor. (H) Partial view of the 1-up/2-down structure (in different shades of pink depending on the subunit) of the spike observed here, prevents an optimal location of the compound-contacting subunits to stabilize the third RBD in the up position ( Figure  6H).
In summary, our structures confirm the inference of the thermofluor and thermophoresis studies that compound 2 hampers RBD and S protein binding to the ACE2 receptor as a consequence of the binding of this compound to the S protein in its RBD portion, in a part of the protein that directly interacts with the ACE2 receptor. The cryoEM structures strongly suggest other more indirect effects, largely of allosteric nature, also resulting in competitive inhibition with respect to ACE2 binding. In this respect, compound 2, and by extension, other cell entry inhibitory compounds in our series, would act by a combination of various molecular mechanisms, all derived from the binding of the compound to the RBD domain in the trimeric S protein. The presence of compound 2 at the boundary of the RBD domain with the remainder of S ( Figure 6D, down panel) could potentially explain the increased affinity of compound 65 for S relative to compound 2 ( Figure 5B), the earlier being a dimeric variant of compound 2. Thus, the second unit of compound 2 contained in 65 may extend from the RBD domain to other parts of the same S subunit or even to other S subunits. Alternatively, the larger size of compound 65 might impose stronger allosteric restrictions on the S trimer simply by causing steric clashes. Further structural work using compound 65 will be required to try to differentiate between these possibilities, although a definitive answer may not be possible if the extra part of 65 (relative to compound 2) extends out freely in the S trimer and is not visible in the structures. In contrast with what is observed with compounds 2 and 65, compound 42 did not substantially inhibit cell entry in the VSV-S system (Table 1), it had no effect on the thermal stability of the RBD in thermofluor assays at 100 μM concentration ( Figure 5B) and exhibited extremely poor affinity for the RBD ( Figure 6B). All these effects can be attributed to the presence in the structure of 42 of the 4− SO 2 CH 3 thiophenyl substituent instead of the 4−NO 2 group present in compound 2. It is interesting that compound 41, bearing a 4−COCH 3 group as the thiophenyl substituent, inhibited cell entry of the VSV-S virus (Table 1) and thermally destabilized the RBD at 100 μM ( Figure 5B). Thus, the −SO 2 − moiety seems to be responsible for the lack of interaction of 42 with the RBD domain, inferring a paramount role of the substituents of the C-2 thiophenyl ring. Unfortunately, the low resolution of the structure of bound compound 2 precludes a detailed assessment of the ways in which the three 4-NO 2 thiophenyl substituents interact with the S protein.

■ CONCLUSIONS
In conclusion, we identify and explore, through the synthesis of a variety of structural analogues, a novel class of SARS-CoV-2 inhibitors. We further provide a detailed characterization of their antiviral activity and mechanism of action. The results indicate a direct role in the inhibition of viral entry by binding to the S protein mainly in the RBD domain, and in preventing the adoption by trimeric S of a conformation that enables efficient binding to the host receptor, ACE2. This new family of compounds opens a new avenue for the potential prophylactic (e.g., nasal spray) or combinatorial therapy of SARS-CoV-2 infection that requires further exploration. Compound 65 retained some antiviral activity against the omicron BA.1 variant despite the many amino acid changes in the S protein (15 changes in the RBD alone, relative to Wuhan RBD), supporting future attempts to develop more active compounds of this family for omicron viral variants. Their multivalent character might endow these compounds with particular resistance to viral escape. In this respect, it is also remarkable that different members of this family are active against a diverse group of viruses, such as HIV and picornaviruses, supporting their interest as useful prototypes for future medicinal chemistry studies. Monodimensional 1 H and 13 C spectra were obtained using standard conditions. 2D-inverse proton detected heteronuclear one-bond shift correlation spectra were obtained using the pulsed field gradient HSQC pulse sequence. Data were collected in a 2048 × 512 matrix with a spectral width of 3460 Hz in the proton domain and 22,500 Hz in the carbon domain and processed in a 2048 × 1024 matrix. The experiment was optimized for one bond heteronuclear coupling constant of 150 Hz. 2D inverse proton-detected heteronuclear long-range-shift correlation spectra were obtained using the pulsed field gradient HMBC pulse sequence. The HMBC experiment was acquired in the same conditions that HSQC experiment and optimized for long range coupling constants of 7 Hz. Compounds were also analyzed by HPLC/MS with a Waters e2695 LC (Waters, Milford, Massachusetts, USA), coupled to a Waters 2996 photodiode array detector and a Waters Micromass ZQ. The column used was a Waters SunFire C18 (2.1 × 50 mm, 3.5 μm), and the mobile phases were A: acetonitrile; B: H 2 O, together with a constant 5% of C (H 2 O with 2% formic acid) to assure 0.1% formic acid along the run. For high-resolution mass spectrometry (HRMS), an Agilent 6520 accurate mass quadrupole time-of-flight (QTOF) platform coupled with LC/MS and equipped with an electrospray interface (ESI) working in the positive-ion (ESI+) and negative-ion (ESI−) modes was used.
Separations on silica gel were performed by preparative centrifugal circular thin-layer chromatography (CCTLC) on a Chromatotron R (Kiesegel 60 PF 254 gipshaltig (Merck)), with layer thicknesses of 1 and 2 mm and flow rates of 4 or 8 mL/min, respectively.
For HPLC analysis, an Agilent Technologies 1120 Compact LC with a reverse-phase column ACE 5 C18-300 (4.6 mm × 150 mm, 3.5 μm) equipped with a PDA (photo diode array) detector was used. Figure 6. continued superimposed on the 3-down structure also observed here (in green and yellow). The RBDs of the A and C subunits of the two structures are superimposed, while subunit B of the three-down structure is not seen. In the two RBDs that are equally positioned in the two structures, there is some shift in the position of subunit A in the three-down structure (highlighted by the arrows) away from the position of the up RBD, expectedly destabilizing this up position. Given the location of the extra density equated here with compound 2 (symbolized with a blue ellipse), the binding of his compound could stabilize the down position of chain A, away from its optimal position for stabilizing the up position of chain B. Acetonitrile 0.05%TFA was used as mobile phase A, and water 0.05% of TFA was used as mobile phase B with a flow rate of 1 mL·min −1 , for 10 min, moving from 10 to 100% phase A. Final compounds had purities >95% based on HPLC.
General Procedure for the Simultaneous Deprotection of Fmoc and Methyl Ester Groups (General Procedure A). To a solution containing the corresponding methyl ester derivative (1.0 mmol) in THF (20 mL) at 0°C (ice bath), a solution of LiOH·H 2 O (2 equiv for each methyl ester group) in water (4 mL) was added, and the mixture was stirred at room temperature overnight. Then, 1 N hydrochloric acid aqueous solution was added to reach pH = 2, and volatiles were evaporated to dryness. The residue was dissolved in ethyl acetate (15 mL) and washed with H 2 O (3 × 10 mL). The organic layer was dried over anhydrous Na 2 SO 4 , filtered, and evaporated to dryness. The purification procedures are described individually.
General Procedure for the Coupling Reaction between the Carboxylic Acids and NH 2 Group of the Amino Acids (General Procedure B). To a solution containing the tripodal polyacid 6 27 (1.0 mmol), HATU (1.3 equiv of each carboxylic acid group), and the appropriate NH 2 free amino acid (1.1−1.2 equiv of each carboxylic acid group) in DMF (10 mL), DIPEA (2.3 equiv of each carboxylic acid group) was added. The resulting mixture was heated to 30°C for 24− 48 h. Then, it was quenched with a saturated solution of NH 4 Cl (5 mL) and volatiles were removed. The residue was dissolved in ethyl acetate (20 mL) and washed with water (10 mL). The organic layer was dried over Na 2 SO 4 , filtered, and evaporated to dryness, and the residue was purified as indicated for each compound.
General Procedure for the Sulfenylation Reaction (General Procedure C). Fmoc-Trp-OMe (10) 29 (1.0 mmol), the corresponding aromatic disulfide (0.7−1.2 mmol) and I 2 (0.6 mmol) were placed in a sealed tube and dissolved in acetonitrile (4 mL). The resulting mixture was heated at 60°C for 3−6 h. Then, it was diluted with ethyl acetate (20 mL) and washed with an aqueous solution of NaHSO 3 (10 mL). The organic layer was dried over Na 2 SO 4 , filtered, and evaporated to dryness, and the residue was purified by flash chromatography.
General Procedure for the Synthesis of Disulfides (General Procedure D). To a solution containing the corresponding benzenesulfonyl chloride (1.0 mmol) in anhydrous DMF (3 mL), a solution of TBAI (3.0 mmol) in anhydrous DMF (3 mL) was added dropwise. The resulting solution was stirred at rt for 24 h. Then, it was diluted with DCM (20 mL) and quenched with an aqueous solution of Na 2 S 2 O 3 (10 mL). The organic layer was washed with a saturated solution of NaHCO 3 (10 mL), dried over Na 2 SO 4 , filtered, and evaporated to dryness, and the residue was purified by flash chromatography.
General Procedure for Selective Fmoc Deprotection (General Procedure E). The appropriate Fmoc-protected compound (1 mmol) was dissolved in DCM (10 mL), and then piperidine (10 mmol) was added dropwise. The reaction was stirred at rt for 2−3 h. Volatiles were removed, and the residue was purified as indicated for each compound.

Methyl 2-Amino-3-(2-((4-fluorophenyl)thio)-1H-indol-3-yl)propanoate
For cryoEM studies, the S protein was produced in baculovirus/ insect cells exactly as described in our prior structural work on this protein, utilizing the form that incorporates the presently universal D614G mutation. 42 In this approach, allowance for intrinsic S protein flexibility and movements of its RBD domains is prioritized over yield of pure protein. 50 Thus, only two residues (K986 and V987) are replaced by proline while the furin cleavage site is also eliminated (change 682 RRAR 685 >A). The protein also includes C-terminal trimerization foldon and 9×His and Mic tags.
The N-terminal peptidase domain of human ACE2 (residues Ser19-Asp615) was produced as previously reported by our laboratory. 42 Purified proteins were concentrated by centrifugal ultrafiltration (Amicon Ultra Millipore devices holding membranes of 10, 30, or 100 kDa nominal cutoff, for, respectively, RBD, ACE2, and S proteins) and were stored at −80°C. They were quantified spectrophotometrically at 280 nm, using sequence-deduced (EXPASY Protparam tool, https:// www.expasy.org/) mass extinction coefficients of 13.7 for RBDs, 10.2 for S, and 21.8 for ACE2.
Thermal Shift Assays. Thermofluor assays 38 were performed in 20 μL of a solution of 17 μg/mL SARS-CoV-2 RBD in 10 mM Na Hepes pH 7.3, 150 mM NaCl, a 1:1000 dilution of the commercial preparation of SYPRO Orange from Invitrogen (Carlsbad, CA) and 5% dimethyl sulfoxide alone or as the solvent carrying into the mixture the compound of interest to a final concentration of 0.1 mM. The SYPRO Orange was the last component added, following a 10-min incubation at 21°C of the mixture without this fluorofore. Then, the mixtures were transferred to wells in a microwell plate, sealed with tape, and placed in a real-time PCR instrument (CFX Opus 96 Real-Time PCR System, Biorad, Hercules, CA, USA), which was used to monitor the increase in SYPRO Orange fluorescence (excitation at 470 nm; emission at 570 nm) with a temperature increase at a ramp of 1°C/min. Each assay with three replicate wells for each point was repeated at least two times on different days. Plots, curve fittings, and numerical calculations were performed with the program Graphpad Prism 7 (GraphPad Software, San Diego, CA, USA).
Microscale Thermophoresis (MST). Poly-His-tagged SARS-CoV-2 RBD or S proteins (see above) were labeled using His-Tag Labeling Kit RED-tris-NTA 2nd Generation (NanoTemper Technologies, Munchen, Germany) according to the manufacturer's instructions. Briefly, equal volumes of 200 nM (as protein chains) target protein and of 100 nM dye in phosphate buffered saline pH 7.1 (PBS; provided in the kit) were mixed and incubated 30 min at 21°C prior to 10-min centrifugation at 15000×g. The supernatant was used in the MST assays, by mixing with an equal volume of the other components. The final mixture contained 50 nM nominal concen-tration of labeled protein chains, PBS pH 7.1, 0.05% Tween-20, 5% dimethyl sulfoxide (DMSO), and the indicated concentrations of each compound tested (decreasing in 16 2-fold dilution steps from 5 mM). The 16 mixtures were incubated 15 min at 21°C before being used to fill the 16 corresponding Monolith Capillaries (catalog number MO-K022) composing a full run of MST measurements, carried out in a Monolith NT.115 apparatus (NanoTemper Technologies). In these measurements, the fluorescence profile is registered for several seconds before turning on the infrared laser, then for 21 s from the moment the infrared laser turns on, and finally for 4 s after the laser turns off (to corroborate the return of the fluorescence toward the initial values). The results were analyzed using a dedicated software (MO.Control software for affinity analysis; NanoTemper Technologies).
The binding of the catalytic domain of ACE2 to the labeled RBD and S proteins was tested in the same way, except for the use of decreasing concentrations of ACE2 (tag-less), in 2-fold dilution steps, from an initial concentration of 5 μM (as ACE2 chains). The titrations were done in the absence and in the presence of 0.5 mM compound 2 or 65.
Cryo-Electron Microscopy Sample Preparation. Cryo-EM Data Acquisition. The complex of S (see the section on Production of Proteins) with compound 2 (Na salt) was prepared by incubating 5 min at 25°C 0.5 mg/mL S and 0.3 mM compound 2 in 0.5 mM, Hepes pH 7.2, 150 mM NaCl. Then, the mixture was placed at 4°C, and 3 μL aliquots were placed on cryo-EM grids (QUANTIFOIL R 1.2/1.3 Au:300-mesh grids) that had been made hydrophilic by glow discharge (30 s using a Leica EM ACE600 device). The grids were then placed in the blotting chamber of a Leica EM GP2 at 10°C and 95% ambient humidity and immediately vitrified in liquid-N 2 -cooled liquid ethane.
Transmission cryoEM images were collected automatically (EPU Automated Data Acquisition Software for Single Particle Analysis; Thermo Fisher Scientific) on a FEI Talos electron microscope operated at 200 kV under low-dose conditions and images recorded on a FEI Falcon III detector operating in electron counting mode, at −1 to −2.5 μm defocus. A total of 3500 movies were recorded at a calibrated magnification of 120,000×, yielding a pixel size of 0.85 Å on the specimen. Each movie comprises 60 frames with an exposure rate of 0.55 e − /Å 2 per frame, with a total exposure time of 28 s and an accumulated exposure of 30 e − /Å 2 .
Image Processing. All image processing steps were performed using programs within Scipion. 51 Movies were motion-corrected and dose weighted with RELION-Motion Correction. 52 Aligned, nondoseweighted micrographs were then used to estimate the contrast transfer function (CTF) with GCTF. 53 662,994 particles were then automatically picked using Gautomatch 53 (https://www2.mrc-lmb.cam.ac.uk/ download/gautomatch-056/) and were 2D-classified with cryo-SPARC, 54 leading to the selection of 375,927 particles. The cryoSPARC initial model protocol was then used with a subset of the total particles to classify the particles into 4 classes. The best generated class (of around 50% of the particles) was uniformly refined in cryoSPARC using 3-fold-symmetry and then used as initial model for 3D heterogeneous refinement in cryoSPARC without applying symmetry to isolate two distinct particle classes, one (278,833 particles) for the spike trimer with 1 RBD domain up and 2 RBDs down, and another one (97,094 particles) with all three RBD domains down. We subsequently refined the maps obtained from these two groups of particles to 3.4 and 4.3 Å resolutions, respectively, based on the goldstandard criterion (FSC = 0.143). The resulting maps were sharpened with DeepEMhancer. 55 Model Building and Refinement. For model building of the spike in either the map in the 1-up or the 3-down conformation, we started with a deposited PDB file for the S protein in the 1-up conformation of the D614G variant (PDB: 7QDG). After fitting of 7QDG to the map using Chimera, 56 followed by rigid-body docking with Coot 57 of different regions of each monomer for keeping the secondary structure of the spike (residues 14-293; 294-319 plus 592-699; 320-330; 331-529 plus 530-591; 734-775; 944-962; 963-989; 990-1027), further model was built manually in Coot, followed by several rounds of refinement using REFMAC, 58 paying special attention to the contact region between different RBDs. In one of the subunits of the 3-down map, an extra density was found having a size compatible with that of compound 2 (see Results section). This process was repeated until acceptable refinement metrics were obtained.

■ ASSOCIATED CONTENT Data Availability Statement
Electron microscopy maps have been deposited in the Electron Microscopy Data Bank (EMDB, https://www.ebi.ac.uk/emdb/ ) with accession codes EMD-17576 and EMD-17578, for the 1up and 3-down forms, respectively of S:D614G. The corresponding atomic coordinates for these two forms are deposited in the PDB with accession codes 8P99 and 8P9Y, respectively.
Synthesis and spectroscopic data of disulfides 20a−c; ( Figure S1) antiviral dose−response profiles measured using the VSV-S assay in both Vero E6 and A549-Ace2-TMPRSS2; ( Figure S2) correlation between the IC 50 values for each compound observed in Vero E6 and A549-ACE2-TMPRSS2 cells obtained with the VSV-S antiviral assay; and 1 H, 13 C NMR spectra and HPLC chromatograms of selected compounds (PDF) Molecular formula strings of the novel synthesized compounds (CSV) ■ AUTHOR INFORMATION Corresponding Authors