Broad-Spectrum Small-Molecule Inhibitors of the SARS-CoV-2 Spike—ACE2 Protein–Protein Interaction from a Chemical Space of Privileged Protein Binders

Therapeutically useful small-molecule inhibitors (SMIs) of protein–protein interactions (PPIs) initiating the cell attachment and entry of viruses could provide novel alternative antivirals that act via mechanisms similar to that of neutralizing antibodies but retain the advantages of small-molecule drugs such as oral bioavailability and low likelihood of immunogenicity. From screening our library, which is focused around the chemical space of organic dyes to provide good protein binders, we have identified several promising SMIs of the SARS-CoV-2 spike—ACE2 interaction, which is needed for the attachment and cell entry of this coronavirus behind the COVID-19 pandemic. They included organic dyes, such as Congo red, direct violet 1, and Evans blue, which seem to be promiscuous PPI inhibitors, as well as novel drug-like compounds (e.g., DRI-C23041). Here, we show that in addition to the original SARS-CoV-2 strain, these SMIs also inhibit this PPI for variants of concern including delta (B.1.617.2) and omicron (B.1.1.529) as well as HCoV-NL63 with low- or even sub-micromolar activity. They also concentration-dependently inhibited SARS-CoV-2-S expressing pseudovirus entry into hACE2-expressing cells with low micromolar activity (IC50 < 10 μM) both for the original strain and the delta variant. DRI-C23041 showed good therapeutic (selectivity) index, i.e., separation between activity and cytotoxicity (TI > 100). Specificities and activities require further optimization; nevertheless, these results provide a promising starting point toward novel broad-spectrum small-molecule antivirals that act via blocking the interaction between the spike proteins of coronaviruses and their ACE2 receptor initiating cellular entry.


Introduction
COVID-19, the coronavirus disease caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is an ongoing pandemic that continues to spread worldwide [1,2]. As of mid-2022, there have been over 500 million confirmed cases and 6 million death cases reported by the World Health Organization (WHO). Since a gold-standard treatment is still lacking, it is urgent that new antiviral drugs are developed, especially oral ones that would allow more widespread use, including by those who are either not willing to be vaccinated or are unable to do so (e.g., due to some pre-existing medical condition). The sites of viral attachment and entry are of particular interest as possible therapeutic targets [3] since they are the first steps of the replication cycle and occur at a relatively accessible extracellular site [4]. SARS-CoV-2 uses the human angiotensin converting enzyme 2 (hACE2) as cell entry receptor, attaching itself via the receptor-binding domain (RBD) of its spike (S) protein [5][6][7][8]. SARS-CoV-2 RBD seems to have a higher ACE2-binding affinity than SARS-CoV(-1) due to some residue changes in the RBD stabilizing two virus-binding hotspots at the RBD-ACE2 interface [9]. The SARS-CoV-2 spike is a highly glycosylated homotrimer. It is post-translationally cleaved into S1 and S2 subunits, with S1 consisting of the amino-terminal domain and the RBD and S2 including the trimeric core and being responsible for membrane fusion [10,11]. The RBD switches between a stand-up position needed for binding to the receptor and a lie-down position used for evading the immune attack [5,12]. Since attachment of the spike (S) protein to its ACE2 receptor is a key step of viral infection, agents that block this, such as neutralizing antibodies or viral entry inhibitors, can be used to prevent infection [13][14][15].
Although vaccination has been successfully shown to control the outbreak of COVID-19, viral variants of increased fitness that facilitate the virus spread and transmission rate [16] and show resistance to the immunity induced by current vaccines against COVID- 19 have already arisen [17,18]. To date, five such variants have been defined as variants of concern (VoC) due to their increased transmissibility, higher disease severity, and resistance to neutralizing antibodies, including those elicited by the existing clinically approved vaccines [19][20][21][22][23]. In chronological order, they are: alpha (B.1.1.7), beta (B.1.351), gamma (P.1), delta (B.1.617.2), and omicron (B.1.1.529). The delta variant, which replaced alpha to become the predominant lineage around October 2020, is characterized by spike protein mutations T19R, ∆157-158, L452R, T478K, D614G, P681R, and D950N. These amino acid mutations play a crucial role in cell infection and may affect the immune responses directed against the main antigenic regions of receptor-binding proteins (452 and 478) and the deletion part of the N-terminal domain (157-158) [2]. The delta variant exhibited reduced sensitivity to antibodies versus the alpha variant [24]. After a single dose application, the Pfizer/BioNTech (BNT162b2) and AstraZeneca (ChAdOx1 nCoV-19) vaccines were less effective in people infected with the delta variant than in those infected with the alpha variant [25]. In November 2021, the omicron variant was identified; it had an alarmingly high number of mutations (>30) in its spike protein, including >15 in its RBD, which is the principal target of neutralizing antibodies, and it continues to spread around the world [26]. Because of the large number of spike mutations, the effectiveness of current COVID-19 vaccines and antibody therapies is likely compromised [27]. SARS-CoV-2 clearly has the potential to continue developing and become increasingly more infectious and less responsive to available therapies.
Development of effective broad-spectrum oral medications could have significant bearing on the coronavirus pandemic as such treatments could be started easily as soon as the first symptoms manifest. Remdesivir was the first small-molecule antiviral approved by the United States Food and Drug Administration (FDA) for the treatment of COVID-19 in October 2020, but it must be given intravenously [28]. Several attempts at repurposing (repositioning) approved drugs as possible small-molecule antiviral agents for SARS-CoV-2 have been pursued in the past two years; however, only very limited success has been achieved so far. For example, WHO's large Solidarity trial found that repositioned antiviral drugs-including hydroxychloroquine, lopinavir, remdesivir, and interferon-β1-had little or no effects in patients hospitalized with COVID-19 [29]. Thus, there is an ongoing need to develop novel drugs that can combat such infections, including those that might arise in the future [30]. Molnupiravir and nirmatrelvir, two small-molecule drugs that exert antiviral effects via inhibition of viral reproduction and protease activity, respectively, have received emergency use authorizations (EUAs) from the FDA in late 2021 as COVID-19 treatments. Molnupiravir is an oral prodrug of N4-hydroxycytidine, a ribonucleoside that exerts its antiviral action by introducing copying errors during viral RNA replication. Nirmatrelvir is an oral peptidomimetic inhibitor of M pro (the main SARS-CoV-2 protease) [31] that binds directly to the SARS-CoV-2 M pro active site [32] and is approved in combination with ritonavir being sold as Paxlovid.
In light of our past work on small-molecule inhibitors of protein-protein interactions (SMIs of PPIs) [33][34][35], we focused on SMIs of the CoV spike-hACE2 PPI as possible viral entry inhibitor antivirals [36][37][38]. Small molecules traditionally were not considered to be likely PPI modulators because protein surfaces tend to lack binding pockets that could make adequate binding possible for them. This changed during the past two decades as an increasing amount of evidence confirmed that SMIs can be effective against at least certain PPIs. SMIs which are or have been in preclinical development target more than 40 PPIs [39][40][41], and three such SMIs (venetoclax [42], lifitegrast [43], and fostemsavir [44]) were recently approved by the FDA for clinical use. Notably, enfuvirtide, maraviroc, and fostemsavir, which are all PPI inhibitors, are approved for clinical treatment against HIV-1, showing that such strategies can be successful for antiviral drug discovery. In particular, fostemsavir, which acts via blocking the gp120-CD4 interaction [44] and was approved for clinical use as an antiretroviral by the FDA in 2020 (Rukobia), provides strong support for the feasibility of the PPI SMI antiviral concept [15]. Compared to antibodies, SMIs could become antiviral therapies that are more patient-friendly (as they are more suitable for oral or inhaled formulations), less immunogenic (as they are not foreign proteins introduced into the body), more controllable (as they have shorter half-lives and better biodistributions), and possibly even more broadly active (i.e., less strain-and mutationsensitive) [15]. Oral drug delivery is the most convenient, cost effective, and safe route of administration [45], ensuring higher patient compliance and therefore being most suitable for broadly acceptable and long-term preventive use [46][47][48], which is also important in the control of viral diseases. SMIs could also be directly delivered into the respiratory system via inhaled or intranasal administration to obtain higher local concentration, which cannot be done with antibodies and could be relevant for COVID-19 treatments. Moreover, SMIs could be broadly active with multi-strain or even pan-CoV inhibitory activity, which is less likely with antibodies that tend to be highly specific [49,50].
As part of our immunopharmacology work, we started a search for SMIs of cosignaling PPIs as potential immunomodulatory agents, and we focused on the chemical space of organic dyes because they tend to have good affinity for proteins [51] and contain structural elements that are considered privileged structures for protein binding [52][53][54]. We hypothesized that this is a better starting point to identify SMIs of PPIs than most commonly available drug-like libraries, and indeed, we found several organic dyes and novel drug-like small-molecule compounds derived from them, designated as DRI-C series, that show promising inhibitory activity against immune checkpoint PPIs of interest such as CD40-CD40L [33][34][35]41]. More recently, following the outbreak of the COVID-19 pandemic, we have shown that some of these dyes and DRI-C compounds possess potential antiviral ability against the original SARS-CoV-2 strain via inhibiting the entry of the virus [36]. Interestingly, we also found that methylene blue, which is approved by the FDA for clinical use as treatment for methemoglobinemia, could be useful as a repurposed antiviral agent against COVID-19, including its delta variant [37,38]. In the present study, we further investigated the activity of some of our most promising drug-like novel compounds (DRI-C23041 and DRI-C24041) as possible inhibitors of CoV entry, including for VoCs such as delta and omicron, while also including some of the organic dyes that showed the most promising inhibitory activity for comparison.

Inhibition of SARS-CoV-2 Spike-hACE2 PPI (Original Strain)
Following up on our previous work [36], we reconfirmed here that two of our DRI-C compounds, DRI-C23041 (1) and DRI-C24041 (2) (Figure 1) inhibit the interaction between the SARS-CoV-2 spike RBD and hACE2 in classic concentration-responsive manner for the original strain in our ELISA-based assay, with DRI-C23041 showing particularly promising (sub-micromolar) activity (IC 50 = 0.66 µM), but DRI-C24041 also having low micromolar activity (IC 50 = 2.85 µM) ( Figure 2A). Thus, these compounds are indeed promising SMIs of the SARS-CoV-2-S-hACE2 PPI, which is essential for the attachment and entry of this coronavirus. We also reconfirmed the activity of three organic dyes identified earlier as having inhibitory activity and included here as comparators of possible interest: Congo red (5, CgRd), direct violet 1 (6, DV1), and Evans blue (7, EvBl) (IC 50 s of 1.90, 2.36, and 2.25 µM, respectively) ( Figure 2A). Two other organic dyes, sunset yellow FCF (8, SY(FD&C#6)) and naphthol blue black (9, NBlBk), were included as negative controls and indeed showed no activity in this assay (IC 50 > 500 µM). We also included two other DRI-C compounds of slightly different structure ( Figure 1) to confirm our structure-activity relationship (SAR) assumptions, and they indeed showed strongly diminished inhibitory activity against this PPI involving the RBD of the original strain: DRI-C41041 (4) and DRI-C2105041 (3), with IC 50 values of 26.09 and 263.00 µM, respectively. activity and included here as comparators of possible interest: Congo red (5, CgRd), direct violet 1 (6, DV1), and Evans blue (7, EvBl) (IC50s of 1.90, 2.36, and 2.25 μM, respectively) (Error! Reference source not found.A). Two other organic dyes, sunset yellow FCF (8, SY(FD&C#6)) and naphthol blue black (9, NBlBk), were included as negative controls and indeed showed no activity in this assay (IC50 > 500 μM). We also included two other DRI-C compounds of slightly different structure (Error! Reference source not found.) to confirm our structure-activity relationship (SAR) assumptions, and they indeed showed strongly diminished inhibitory activity against this PPI involving the RBD of the original strain: DRI-C41041 (4) and DRI-C2105041 (3), with IC50 values of 26.09 and 263.00 μM, respectively. Inhibitory activities were also assayed using not the RBD but the S1 fragment of the spike protein (Error! Reference source not found.B). The activity of DRI-C23041 and DRI-C24041, as well as of the dyes CgRd, DV1, and EvBl, remained within the same low- control SY (FD&C#6) showed no inhibition (IC50 > 500 μM). However, DRI-C41041, DRI-C2105041, and NBlBk showed some activity in this assay with IC50 values in the mid-micromolar range (8-20 μM; Error! Reference source not found.B). On one hand, this could be due to the somewhat weaker binding of S1 as used here to hACE2 than that of the RBD fragment (see Figure 2 in [36]). On the other hand, this may illustrate that slightly different structural requirements are needed to inhibit RBD and S1 binding-see also binding of the VoC PPIs below. and hACE2 in cell-free ELISA-type assays by the compounds tested here (1-9). Data (symbols) are mean ± SD from three experiments in duplicates. IC50 values obtained from fitting with standard sigmoid curves are shown in the legend confirming that several of the DRI-C compounds and organic dyes included here showed promising low micromolar or even sub-micromolar activity (IC50 < 1 μM).

Figure 2. Inhibition of SARS-CoV-2-S binding (original strain) to hACE2.
Concentration-response curves obtained for the inhibition of the PPI between SARS-CoV-2-S-RBD (A) or SARS-CoV-2-S1 (B) and hACE2 in cell-free ELISA-type assays by the compounds tested here (1-9). Data (symbols) are mean ± SD from three experiments in duplicates. IC 50 values obtained from fitting with standard sigmoid curves are shown in the legend confirming that several of the DRI-C compounds and organic dyes included here showed promising low micromolar or even sub-micromolar activity (IC 50 < 1 µM).
Inhibitory activities were also assayed using not the RBD but the S1 fragment of the spike protein ( Figure 2B). The activity of DRI-C23041 and DRI-C24041, as well as of the dyes CgRd, DV1, and EvBl, remained within the same low-micromolar range (0.2-3.0 µM; Figure 2B), and the negative control SY (FD&C#6) showed no inhibition (IC 50 > 500 µM). However, DRI-C41041, DRI-C2105041, and NBlBk showed some activity in this assay with IC 50 values in the mid-micromolar range (8-20 µM; Figure 2B). On one hand, this could be due to the somewhat weaker binding of S1 as used here to hACE2 than that of the RBD fragment (see Figure 2 in [36]). On the other hand, this may illustrate that slightly different structural requirements are needed to inhibit RBD and S1 binding-see also binding of the VoC PPIs below.

Inhibition of SARS-CoV-2 Spike-hACE2 PPI (Variants of Concern including Delta and Omicron)
Due to the increased importance of broad-spectrum activity, especially following the emergence of VoCs for SARS-CoV-2, we also tested these compounds to assess their inhibitory activity against the corresponding PPIs with mutant spike proteins including D614G, delta (B.1.617.2), and omicron (B.1.1.529). Just as for the original strain, we assessed activity using both the RBD and the S1 fragments (except for D614G, where this was not possible as the mutation is only in the S1 and not in the RBD region). Notably, both DRI-C23041 (1) and DRI-C24041 (2) maintained their low micromolar activity in all of these assays (IC 50 (5-7), which showed good activity against the original strain ( Figure 2B), retained their low-micromolar activity for these VoC PPIs as well (Figures 3-5). The negative control SY(FD&C#6) (8) remained consistently inactive in all assays (IC 50 > 500 µM).
In the meantime, just as for the original strain, DRI-C41041 (4), DRI-C2105041 (3), and NBlBk (9) showed little if any activity in the assays with RBDs, but some (mid-micromolar) activity in the S1 assays, indicating again somewhat more susceptibility to inhibition for the S1 fragment as compared to RBD only. Error! Reference source not found. (delta), and Error! Reference source not found. (omicron), indicating their potential for broad-spectrum activity. Similarly, the organic dyes CgRd, DV1, and EvBl (5-7), which showed good activity against the original strain (Error! Reference source not found.B), retained their low-micromolar activity for these VoC PPIs as well (Error! Reference source not found.-5). The negative control SY(FD&C#6) (8) remained consistently inactive in all assays (IC50 > 500 μM). In the meantime, just as for the original strain, DRI-C41041 (4), DRI-C2105041 (3), and NBlBk (9) showed little if any activity in the assays with RBDs, but some (mid-micromolar) activity in the S1 assays, indicating again somewhat more susceptibility to inhibition for the S1 fragment as compared to RBD only.

Inhibition of HCoV-NL63 S-hACE2 PPI
Finally, since there is one other known coronavirus that infects humans, causing a common cold that also uses hACE2 as its receptor for attachment and entry (HCoV-NL63), we also assessed the inhibitory activity of some of these compounds against the HCoV-NL63 S1-hACE2 PPI (Error! Reference source not found.). As the affinity of the HCoV-NL63 S1 to hACE2 is considerable less than that of SARS-CoV-2 S1 (46 vs. 16 nM in our setup [36]), larger protein amounts were needed for detectability; therefore, we assessed only a subset of compounds of interest. Notably, DRI-C23041 and DRI-C24041 lost some (2-5-fold) activity compared to the SARS-CoV assays, but still showed good inhibition (IC50 values of 4.16 and 6.54 μM, respectively), indicating possible pan-coronavirus inhibitory activity. The organic dyes CgRd, DV1, and EvBl, which we have shown before to be promiscuous and non-specific PPI inhibitors [36], maintained their low micromolar inhibitory activity (IC50 < 3.0 μM).

Inhibition of HCoV-NL63 S-hACE2 PPI
Finally, since there is one other known coronavirus that infects humans, causing a common cold that also uses hACE2 as its receptor for attachment and entry (HCoV-NL63), we also assessed the inhibitory activity of some of these compounds against the HCoV-NL63 S1-hACE2 PPI ( Figure 6). As the affinity of the HCoV-NL63 S1 to hACE2 is considerable less than that of SARS-CoV-2 S1 (46 vs. 16 nM in our setup [36]), larger protein amounts were needed for detectability; therefore, we assessed only a subset of compounds of interest. Notably, DRI-C23041 and DRI-C24041 lost some (2-5-fold) activity compared to the SARS-CoV assays, but still showed good inhibition (IC 50 values of 4.16 and 6.54 µM, respectively), indicating possible pan-coronavirus inhibitory activity. The organic dyes CgRd, DV1, and EvBl, which we have shown before to be promiscuous and non-specific PPI inhibitors [36], maintained their low micromolar inhibitory activity (IC 50 < 3.0 µM).

Inhibition of SARS-CoV-2 Pseudovirus Entry (Original Strain and Delta VoC)
Next, we confirmed inhibitory activities using a cell-based pseudovirus assay that allows the quantification of viral entry without having to use biosafety level 3 (BSL-3) or higher containment since it uses pseudoviruses that do not replicate in human cells. As before [36], this has been done with a baculovirus pseudotyped with SARS-CoV-2-S proteins and generated using BacMam-based tools. If the nuclei of pseudovirus-exposed host cells (here, ACE2-and red fluorescence expressing HEK293T) show green fluorescence, pseudovirus entry was completed. If entry is blocked, the cell nuclei remain dark. Here, we tested inhibitory activities using pseudoviruses expressing the SARS-CoV-2 spike protein corresponding to the original strain and the delta variant (Error! Reference source not found.).
In parallel with this, we also evaluated the cytotoxicity of these compounds under conditions similar to those used to evaluate their activity to assess their therapeutic (selectivity) index, TI (SI) = TC50/IC50. None of the DRI-C compounds tested showed significant effect on the viability of HEK293T cells, with DRI-C24041 being the most toxic (TC50 > 600 μM) and all others having TC50 > 1000 μM (Error! Reference source not found.). Accordingly, DRI-C23041 shows particularly promising separation between activity and toxicity, i.e., therapeutic index, as illustrated in Error! Reference source not found.A, which summarizes all its activity in a single graph. In agreement with their promiscuous PPI inhibitory activity, the organic dyes CgRd, DV1, and EvBl showed more pronounced cytotoxicity in the HEK293T cells as tested here (TC50: 130, 110, and 439 μM, respectively). Accordingly, there is much less separation between their inhibitory and toxic activities, as illustrated for DV1 in Error! Reference source not found.B.

Percent of Max Ligand Binding
Inhibition of HCoV-NL63 S1 -hACE2 Binding  (7) 2.60 Figure 6. Inhibition of HCoV-NL63 S binding to hACE2. Concentration-response curves obtained for the inhibition of the PPI between HCoV-NL63 S1 and hACE2 as done for SARS-CoV-2 in previous figures.

Inhibition of SARS-CoV-2 Pseudovirus Entry (Original Strain and Delta VoC)
Next, we confirmed inhibitory activities using a cell-based pseudovirus assay that allows the quantification of viral entry without having to use biosafety level 3 (BSL-3) or higher containment since it uses pseudoviruses that do not replicate in human cells. As before [36], this has been done with a baculovirus pseudotyped with SARS-CoV-2-S proteins and generated using BacMam-based tools. If the nuclei of pseudovirus-exposed host cells (here, ACE2-and red fluorescence expressing HEK293T) show green fluorescence, pseudovirus entry was completed. If entry is blocked, the cell nuclei remain dark. Here, we tested inhibitory activities using pseudoviruses expressing the SARS-CoV-2 spike protein corresponding to the original strain and the delta variant (Figure 7).
In agreement with their good PPI inhibitory activity, DRI-C23041 (1) and DRI-C24041 (2) showed good concentration-dependent inhibitory activity with similar IC 50  In parallel with this, we also evaluated the cytotoxicity of these compounds under conditions similar to those used to evaluate their activity to assess their therapeutic (selectivity) index, TI (SI) = TC 50 /IC 50 . None of the DRI-C compounds tested showed significant effect on the viability of HEK293T cells, with DRI-C24041 being the most toxic (TC 50 > 600 µM) and all others having TC 50 > 1000 µM (Figure 8). Accordingly, DRI-C23041 shows particularly promising separation between activity and toxicity, i.e., therapeutic index, as illustrated in Figure 9A, which summarizes all its activity in a single graph. In agreement with their promiscuous PPI inhibitory activity, the organic dyes CgRd, DV1, and EvBl showed more pronounced cytotoxicity in the HEK293T cells as tested here (TC 50 : 130, 110, and 439 µM, respectively). Accordingly, there is much less separation between their inhibitory and toxic activities, as illustrated for DV1 in Figure 9B.    individual experiments are shown for DRI-C23041, DRI-C24041, DV1, NBlBk, and SY (FD&C#6) in C (numbers within each panel indicating percent compared to corresponding controls in top row as 100%).

Discussion
Our results here reconfirmed that the DRI-C compounds we identified earlier, such as DRI-C23041 (1) as well as some organic dyes such as CgRd (5) and DV1 (6), have potential as SMIs of PPIs mediating the cell attachment and entry of SARS-CoV-2 and possibly even other coronaviruses. DRI-C23041 in particular showed promising separation between activity and non-specific cytotoxicity, having TI >> 100 ( Figure 9A), but even DRI-C24041 has TI ≈ 100 as calculated based on the ratio of the TC 50 and IC 50 values. The organic dyes, especially CgRd and DV1, also show good inhibitory activity; however, they seem to act as non-specific and promiscuous PPI inhibitors. Accordingly, they also have a quite narrow therapeutic index and, thus, low therapeutic promise ( Figure 9B). Nevertheless, this work confirms again that, as highlighted before, by being rich in privileged structures for protein binding, the chemical space of organic dyes provides a useful starting point in the search for effective SMI scaffolds for PPI inhibition in general [15,36,41]-more so than typical drug-like screening libraries. This is also corroborated by comparing these results obtained by us from screening a relatively limited number of compounds (<200) with those obtained by other efforts to identify SMIs of spike-hACE2 PPIs using various screening approaches that often involved multiple thousands of candidate compounds, which have been reviewed recently [15]. So far, SMIs that have antiviral activity with an IC 50 < 30 µM confirmed in a live viral or pseudoviral assay are limited to DRI-C23041 (1) [36] and some organic dyes, including Congo red (5), direct violet 1 (6), Evans blue (7), and methylene blue, from our work [36][37][38]; methylene blue [55][56][57][58] and Evans blue [59] also found by others; verteporfin [60]; and cannabigerolic and cannabidiolic acids [61]. It is notable, for example, that Evans blue was established as the best hit identified from a high-throughput screening of more than 3000 compounds, which were themselves preselected after in silico prescreening of~60,000 structures [59], whereas it ranked as not even the most promising hit in our screening of a much smaller library of slightly more than a hundred organic dyes. Results obtained there (K D = 2.2 µM for binding to SARS-CoV-2-S and IC 50 = 28.1 µM for SARS-CoV-2 infection inhibition in Vero E6 cells [59]) are quite consistent with those obtained here (Figures 2 and 7). Finally, regarding organic dyes as SMIs of PPIs, it is also of interest that methylene blue was found to inhibit the entry and replication of SARS-CoV-2, and has potential as a possible inexpensive, broad-spectrum antiviral for the prevention and treatment of COVID-19, possibly due to multiple mechanisms of action [37,38,[55][56][57][58], even if it is a promiscuous PPI inhibitor with a fairly narrow TI [38], since it is FDA approved for clinical use and is included in the WHO List of Essential Medicines.
Of course, structures identified from organic dye libraries will need considerable further structural optimization to eliminate their color, increase their specificity, and optimize their clinical potential [34,35]. SMIs of PPIs tend to be larger and more hydrophobic structures than what is considered typical for drug-like molecules, because such larger structures that typically contain multiple aromatic rings are usually needed for efficient protein binding and PPI inhibition [52][53][54]. Thus, they also tend to violate the so-called "rule-of-five" (Ro5) criteria widely applied to ensure adequate oral bioavailability and pharmacokinetic profile during lead candidate selection, since Ro5 criteria include, among others, the need to have a molecular weight less than 500 Daltons [62,63]. Nevertheless, several new drugs have been launched in the last decade or so that violate these selfimposed empirical rules and prove that oral bioavailability can be achieved in what is called the "beyond the Ro5" chemical space, even if not easily [64,65]. Thus, when looking at the potential of development and clinical translatability of identified PPI SMI hits, it is worth noting that it will most likely involve considerable optimization work anyway. For example, ABT-737, which was the first lead compound during the development process of venetoclax (ABT-199), was so far from suitable for formulation as a typical drug that it was described as having "the biophysical properties of brick dust" [66]. Similarly, it took an impressively tedious medicinal chemistry optimization process to make the original lead of the series into what ultimately became the clinical product fostemsavir (BMS-663068), as neatly summarized in [44]. As a first evaluation of the drug-likeness and clinical trans-latability of the present compounds, summaries of their physicochemical and absorption, distribution, metabolism, and excretion plus toxicity (ADMET) properties calculated by SwissADME [67] and ADMETlab2.0 [68] (accessed on 19 August 2022) are included in  Supplementary Information Table S1.
Finally, it should be mentioned that results here indicate that it seems feasible that SMI structures can be identified that have sufficiently broad-spectrum activity, i.e., they can inhibit several variants and mutants of SARS-CoV-2. In fact, it could be possible to find SMI that can also inhibit other coronaviruses, including not just SARS-CoV and MERS-CoV but even less dangerous ones such as the common cold-causing HCoVs (e.g., Figure 6). Ultimately, it is hoped that such inhibitory effects on coronaviral attachment can be translated into antiviral activity against the COVID-19-causing SARS-CoV-2 and maybe even other CoVs that bind to ACE2 such as SARS-CoV and the α-coronavirus HCoV-NL63.

Cytotoxicity Assay
HEK293T cells were seeded onto 96-wells and treated with five concentrations of test compounds (5,15,45,135, and 270 µM) for 48 h. Old media were removed and replaced with fresh ones for additional 48 h. Then, cells were treated with MTT solution at 37 • C for 2 h. Cell viability was determined by optical density at 570 nm using a microplate reader (SpectraMax iD3, Molecular Devices, San Jose, California, CA, USA).

Statistical Analysis and Data Fitting
All binding and pseudovirus entry inhibition assays were performed as duplicates or triplicates per plate, and all experiments were repeated at least three independent times (with the exception of the HCoV-NL63 S1-hACE2 assay, where, due to the larger protein amounts needed, only two repeats were done and only for a subset of compounds). As before [34,36,38], data were normalized as inhibition % and fitted with standard concentrationresponse models in GraphPad Prism (GraphPad, La Jolla, CA, USA) to determine EC 50 or IC 50 values (half-maximal effective or inhibitory concentrations).

Conclusions
In conclusion, even if specificities and activities of the compounds identified here still require further optimization, these results provide proof-of-principle evidence for the feasibility of SMIs targeting the coronavirus spike protein-hACE2 PPI that could lead to alternative, orally bioavailable therapies for the prevention and treatment of COVID-19 and maybe even other diseases caused by coronavirus infection.

Patents
Buchwald, P. Small molecule inhibitors of coronavirus attachment and entry, methods and uses thereof. International Patent Application, International Patent Application No. PCT/US21/52520, Filed 29 September 2021.

Conflicts of Interest:
The authors declare the following competing financial interest. The University of Miami has filed a patent on these compounds and their use for this application with P.B. as inventor. All other authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript: ACE2 angiotensin World Health Organization