View all hidden authors and organizations

Edited by Malcolm A. Martin, National Institute of Allergy and Infectious Diseases, Bethesda, Maryland, approved on August 18, 2021 (reviewed on May 10, 2021)

MM3122 represents an advanced drug candidate for the clinical development of a new antiviral drug for COVID-19. In addition to being new drugs, these selective TMRSS2 inhibitors can also be used as valuable chemical probes to help elucidate the pathogenesis of the virus. As TMPRSS2 as a viral protein processing protease plays a key role in the pathogenesis of other coronaviruses (SARS-CoV, MERS-CoV) and influenza viruses, MM3122 and such TMPRSS2 inhibitors are expected to be new drugs not only for the treatment of SARS-CoV -2 infection, but may also represent a broad-spectrum antiviral drug.

The host cell serine protease TMPRSS2 is an attractive therapeutic target for COVID-19 drug discovery. This protease activates the Spike protein of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) and other coronaviruses, and is essential for the spread of the virus in the lungs. Using the combination of rational structure-based drug design (SBDD) and TMPRSS2 substrate-specific screening, we discovered covalent small-molecule ketobenzothiazole (kbt) TMPRSS2 inhibitors, which are structurally different from the existing known Inhibitors Camostat and Nafamostat, and have significantly increased activity. The lead compound MM3122 (4) has an IC50 (half maximum inhibitory concentration) of 340 pM for the recombinant full-length TMPRSS2 protein, and an EC50 (half maximum effective concentration) of 430 pM for blocking host cells from entering the Calu-3 human lung. The epithelial cells of the VSV-SARS-CoV-2 chimeric virus, with an EC50 of 74 nM, can inhibit the cytopathic effect induced by the SARS-CoV-2 virus in Calu-3 cells. In addition, MM3122 prevents the entry of Middle East Respiratory Syndrome Coronavirus (MERS-CoV) cells with an EC50 of 870 pM. MM3122 has excellent metabolic stability, safety and pharmacokinetics in mice, with a plasma half-life of 8.6 hours and a lung tissue half-life of 7.5 hours. It is suitable for in vivo efficacy evaluation and is a promising drug candidate for the treatment of COVID-19 .

Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) is a new and highly spreading coronavirus that has caused the ongoing COVID-19 pandemic. As of April 12, 2021, the world has There are 136 million cases and nearly 3 million deaths. https://coronavirus.jhu.edu/map.html). Although the FDA has recently approved three vaccines, there are no clinically approved small molecule drugs that can be used to treat the disease except Redcivir, and the effectiveness of vaccines against immune escape mutations may be reduced. A variety of treatment strategies have been proposed (1-2), including viruses and host proteins, but clinical applications have not been fully verified. One class of protein targets that have shown promising results are proteolytic enzymes, including viral proteases (1, 3⇓ –5), papain-like proteases (PLpro), and 3C-like or "major proteases" (3CL or MPro), And several host protein enzymes are involved in virus entry, replication and impact on the immune system, leading to life-threatening symptoms of COVID-19 infection (4⇓ –6). The latter includes various members of the cathepsin family of cysteine ​​proteases, including cathepsin L, furin and serine protease factor Xa, plasmin, elastase, tryptase, TMPRSS2 and TMPRSS4.

The entry of coronaviruses (SARS CoV-2, SARS-CoV and Middle East Respiratory Syndrome Coronavirus [MERS]) is mediated by the viral spike protein, which must be cleaved by the host protease to trigger membrane fusion and enter the host cell. Binding host cell receptor angiotensin-converting enzyme-2 (ACE2) (7⇓ ⇓ –10). This is mediated by the initial cleavage of the S1/S2 junction of the spikes, which is believed to occur during the processing of the producer cell, and is then treated by serine proteases or late cathepsin proteases on the cell surface at the S2' site Cut endosomes or endolysosomes (9, 10). Whether serine protease or cathepsin protease is used for S2' cleavage depends on the cell type. Although entry into Calu-3 (human lung epithelium) or HAE (mainly human respiratory epithelium) cells is independent of cathepsin, entry into Vero cells (African green monkey kidney epithelium) that do not express the required serine protease is completely dependent on cathepsin . 7, 9⇓ –11).

TMPRSS2 (12) is a type II transmembrane serine protease (TTSP) (13), which has been shown to be essential for host cell virus entry and SARS-CoV-2 transmission (7, 8, 14⇓ –16), and Such as SARS-CoV (17, 18), MERS-CoV (19) and influenza A virus (20⇓ ⇓ ⇓ ⇓ ⇓ ⇓ –27). Spike protein needs to be proteolytically processed/primed by TMPRSS2 to mediate its entry into lung cells; therefore, small molecule inhibitors of this target offer great hope as new therapies for COVID-19 and other coronavirus diseases (7, 8) . As recently reported, the expression level of TMPRSS2 determines the entry pathway for SARS-CoV-2 to enter the cell (28). In cells that express little or no TMPRSS2, the cell enters through the endosomal pathway, and the cleavage of the spike protein is carried out by cathepsin L. It has been demonstrated that the lung epithelial Calu-3 cells expressing TMPRSS2 are highly resistant to SARS-CoV-2 infection. The irreversible serine protease inhibitors Camostat (7) and Nafamostat (29) effectively prevent host cell entry and SARS-CoV-2 replication in Calu-3 cells through a TMPRSS2-dependent mechanism (14, 15).

Here, we report the discovery of a class of substrate-based TMPRSS2 ketobenzothiazole (kbt) inhibitors, which have potent antiviral activity against SARS-CoV-2, which is significantly improved compared to Camostat and Nafamostat. Several compounds were found to be strong inhibitors of virus entry and replication, and their EC50 (half the maximum effective concentration) value exceeds the potency of Camostat and Nafamostat, and has no cytotoxicity. The newly developed compound MM3122 (4) has excellent pharmacokinetics (PK) and safety in mice, so it is a promising lead candidate for the treatment of COVID-19.

It is well known that several TTSPs, including TMPRSS2, not only play a role in infectious diseases (12), but also play a role in cancer progression and metastasis (30⇓ –32), which is thought to be mainly through their activation of hepatocyte growth factor ( HGF), the only ligand for MET receptor tyrosine kinase. This is achieved by proteolytically processing the inactive single-chain precursor pro-HGF into a double-stranded active form. TMPRSS2 shares pro-HGF as a protein substrate with other HGF-activated serine proteases, HGFA (HGF-activator), hepsin and Matriptase (33, 34). Like TMPRSS2, other TTSPs (such as Matriptase and hepsin) also have a typical serine protease domain, which serves as the C-terminal domain of the protein and is anchored to the cell membrane through the N-terminal type II signal peptide to display their enzymes outside the cell membrane. active. cell. We have previously reported the discovery and anti-cancer properties of the peptidomimetic ketothiazole (kt) and kbt inhibitors of HGFA, matriptase and hepsin (35⇓ ⇓ ⇓ –39), called synthetic HGF activation inhibitors or sHAIs. Since TMPRSS2 and HGF-activated protease have overlapping endogenous substrate-specific characteristics, we hypothesized that our substrate-based sHAI would also inhibit TMPRSS2.

Based on our assumption that sHAI will inhibit TMPRSS2, we first selected two sHAI lead compounds, ZFH7116 (1) (37, 40) and VD2173 (2) (Figure 1) for preliminary antiviral testing, and confirmed that they effectively inhibit TMPRSS2 by SARS- The spike protein of CoV-2 (SARS-2 S; Figure 2A) drives host cell-dependent viruses to enter Calu-3 lung epithelial cells with EC50 of 307 and 104 nM, respectively. The irreversible inhibitor Camostat also detected significant inhibition of SARS-2 S-driven entry, consistent with published data (7, 29). It is worth noting that neither of these two compounds showed activity in TMPRSS2-negative Vero cells, nor did they enter Calu-3 cells through pseudotyping with VSV glycoprotein (VSVpp VSV-G) (SI Appendix, Figure S6) This indicates that the inhibition of SARS-CoV-2 spike drive into Calu-3 cells is due to the blockade of TMPRSS2.

The initial structure of the TMPRSS2 inhibitor was discovered by screening the existing HGFA, matriptase and hepsin serine protease inhibitors.

ZFH7116 (1) and VD2173 (2) use (A) VSV-SARS-CoV-2-Spike protein pseudotype virus and (B) VSV-SARS- inhibit SARS-CoV-2 cells from entering Calu-3 lung epithelial cells CoV- 2-Spike protein chimeric virus. EC50 is calculated based on the average of three independent experiments. Statistically compare the relative infection at each concentration (*P <0.05, ** P <0.01, *** P <0.001, Student's t test). Camostat is used as a positive control.

Using a replication-competent chimeric VSV expressing SARS-CoV-2 Spike [showing Spike-dependent entry that mimics real SARS-CoV-2 (41)], we also proved that 1 and 2 prevent the virus from entering Calu-3 cells. Dose dependent. Compared with the pseudotyped virus (Figure 2A), the EC50 values ​​of the two drugs against the chimeric virus VSV-SARS-CoV-2 (Figure 2B) were 459 and 197 nM, respectively, and for VSV-eGFP (VSV G-dependent Sex item), or against any virus in TMPRSS2-negative Vero cells (SI Appendix, Figure S7). This confirms our preliminary results and allows the establishment of a system for screening these inhibitors for antiviral activity using VSV-SARS-CoV-2.

By overexpressing TMPRSS2 in the human cell line HEK-293T, we tested the inhibitory activity of 1 and 2 on the proteolytic activity of TMPRSS2 in a cell-based enzyme-based fluorescence assay. HEK-293T is often used in experiments due to its high transfection ability. . As mentioned earlier, the fluorescent peptide reporter substrate Boc-QAR-AMC can be used to accurately measure the expression of TMPRSS2 in cell culture (42, 43). Compound 1 inhibited cell-based TMPRSS2 enzyme activity in a concentration-dependent manner between 10 μM and 10 nM, with IC50 (half maximum inhibitory concentration) of 314 nM (SI appendix, Figure S4). Compound 2 is a more potent inhibitor of TMPRSS2 proteolytic activity with IC50 of 57 nM. These data confirm that 1 and 2 regulate their functions by effectively inhibiting TMPRSS2.

We analyzed the selectivity of 1 and 2 to a set of 43 serine and cysteine ​​proteases (SI appendix, Tables S1 and S2) to compare with published selectivity data for Camostat and Nafamostat proteases (44). We found that 1 is a potent inhibitor of Matriptase-2, plasma kallikrein, proteinase K, trypsin, tryptase b2 and G1, but also factor Xa, factor XIIa, kallikrein 5 and 14 and cysteine S, a moderate inhibitor of the acid protease cathepsin, also shows some activity against cathepsin L. 1 Overlaps the selectivity curve of Camostat, except that they do not show any activity against cathepsin. In addition, Camostat more effectively inhibits factor XIIa, plasma kallikrein, matriptase-2 and plasmin, while also increasing the activities of trypsin, tryptase and urokinase. Cyclic peptide 2 also inhibits urokinase and factor XIIa more effectively than 1, but not cathepsin S or L, but a moderate inhibitor of cathepsin B. The results of these different selective characteristics can explain that the different active viral entry of VSV pseudotypes and chimeras leads to the most significant urokinase potency, of which 1 is significantly weaker.

We published (38) Camostat and Nafamostat (Figure 3B) are inhibitors of Matriptase and hepsin, and the analogue Nafamostat is the most active SARS-CoV-2 cell entry inhibitor published (8). Although we previously used this chemical series as HGFA, matriptase and hepsin inhibitors (38), the best series we have developed are molecules based on small peptides such as 1, 2 and 3 (Ac-SKLR-kbt; Figure 3B) It exhibits an IC50 of low nanomolar to picomolar. These compounds contain kbt warheads (35⇓ –37, 39) that capture serine, which can react covalently with proteases, but it is important to react in a reversible manner, unlike Camostat or Nafamostat, whose inhibitory effect is irreversible. Therefore, we use kbt inhibitors for lead identification studies to find more effective and selective TMPRSS2 inhibitors.

(A) A molecular docking model using Glides/Schrödinger to combine compound 3 (yellow) and Nafamostat (grey) with the TMPRSS2 homology model. (B) The structure of compound 3, Camostat and Nafamostat.

We used X-ray eutectic structure data to optimize the rational design of HGFA, matriptase and hepsin inhibitors, which have higher potency and selectivity (35⇓ ⇓ –38). As shown in Figure 3, here we use the homology model of human TMPRSS2 (45) to computationally model our existing inhibitor (using Glide in Schrodinger), where we docked compound 3 (yellow; Figure 3A) And Nafamostat (grey). For Nafamostat, the naphthyl moiety penetrates deep into the S2 pocket, and the biphenylguanidine in the S1 pocket binds to the conserved S1 Asp189. The P4 Ser of 3 forms two H bonds with Asp162 and Asn163 outside the S4 pocket, and the Ser backbone carbonyl group is also within the H bond distance of P2 Lys98. P3 Lys is close to S4 Glu134 to generate electrostatic interaction, while P2 Leu is located in the S2 pocket, and benzothiazole fills the S1' area. The reactive ester of Nafamostat and the ketone of kbt are adjacent to the Ser195-His57-Asp102 catalytic triad. Our model of 3 and Nafamostat combined with TMPRSS2 (Figure 3A) showed a good fit for the S1 to S4 and S1' pockets of TMPRSS2. For 3, it seems that P4 Ser is forming a double H bond with Asp162 and Asn163 residues in TMPRSS2. This region is occupied by Gln in hepsin and Matriptase, but by Asp in HGFA. Interestingly, the extra Asn163 residue is unique to TMPRSS2. This indicates that the free amine in P4 can provide selectivity superior to Matriptase and hepsin. The structure also shows that Lys87 is located in the S2 pocket, indicating that P2 Glu or Asp may be ideal for TMPRSS2 inhibitors and leads to potential choices for HGFA (P2 Ser), matriptase (P2 Phe), and hepsin (P2 Asn) Sex. The Lys87 is also close to the side chain of P4 Ser, indicating that Asp or Glu at this position may have a third electrostatic interaction with Lys87. In addition, the S2 pocket is large and seems to be able to accommodate larger side chains such as Phe, Tyr or Trp.

To enhance our rational design of more effective and selective TMPRSS2 inhibitors, we analyzed existing data on the peptide substrate preferences of TMPRSS2, HGFA, matriptase, and hepsin. When comparing the substrate combinatorial library (PS-SCL) data of TMPRSS2 (Figure 4) (30) with the position scans of Matriptase (47), hepsin (48) and HGFA (36), it is clear that there is a preference for their substrates. There is a significant overlap. These data indicate that TMPRSS2 can tolerate many different P2 side chains, but prefer Phe and Ala/Thr, such as Matriptase (Figure 4), and they can also tolerate large and small populations (preferring Ser/Ala), but hepsin and HGFA prefers columns. For P3, TMPRSS2 preferred Gln/Glu and Met, while Lys/Gln preferred hepsin and Matriptase, and Lys/Arg preferred HGFA (Figure 4). The most obvious difference is in the P4 position, where HGFA, matriptase and hepsin all like basic residues, such as Lys/Arg, while TMPRSS2 likes Ile/Gly and Pro, which are common attributes with hepsin.

For TMPRSS2, matriptase, hepsin and HGFA substrate specific combination PS-SCL. Using protease substrate terminology, the cleavage (hydrolysis) site of the protease protein substrate is located between the N-terminal P1 and P1' positions of the peptide (P) substrate, and S1 and S1' refer to the subsite (S)1 Protease that binds to 1'P1 and P1' amino acid side chains (46). Red indicates the overlap of the amino acid specificity of TMPRSS2 with other proteases. The one-letter name of the amino acid is shown.

To enhance our compound design, we obtained further information about the specificity of TMPRSS2 extended substrates using multiple substrate analysis by mass spectrometry (MSP-MS) (49). In the MSP-MS analysis, the physiochemically diverse library of 228 tetrapeptides was incubated with human recombinant TMPRSS2 in the activity buffer for several hours. At different time points (15, 60, and 240 minutes), aliquots of the reaction mixture were extracted, quenched with 8 M guanidine hydrochloride, and analyzed by tandem MS to monitor the cleavage products produced by the protease. As a TTSP with trypsin folding, TMPRSS2 is known for its high affinity for substrates containing arginine residues at the P1 position (30). Our MSP-MS analysis confirmed this outstanding P1 specificity, which showed that the P1 position on the cleavage product generated by TMPRSS2 was almost occupied by Arg or Lys (Figure 5A). Peptide sequencing by LC-MS/MS was able to identify the 25 most preferred TMPRSS2 substrates in our peptide library (Figure 5B). The IceLogo frequency chart (Figure 5C) shows the substrate-specific characteristics of TMPRSS2 expanded at pH 7.5, which reveals a preference for hydrophobic amino acids at the P2 and P1' positions flanking the cleavage site. The analysis shows that the preferred cleavage of the N-terminal tetrapeptide is PLFR, with other P4 amino acids H and M and P3 amino acids G, Y, V, and Q, but only F at P2, which is amazing. To a large extent, these data summarize what has been clarified in the PS-SCL study (Figure 4). Based on the careful analysis of the computational modeling work of these and PS-SCL results, we selected other existing compounds in our HGFA/matriptase hepsin inhibitor library for testing and synthesis (SI Appendix, Scheme S1-S3). Specifically designed for TMPRSS2: Ac-GQFR-kbt (4), Ac-PQFR-kbt (5), Ac-QFR-kbt (6) and Ac-IQFR-kbt (7).

(A) The heat map shows the overall amino acid frequency of each P4-P4' position of a set of positive peptides cleaved by human TMPRSS2. It is observed that positive enrichment is red, and the zero setting is yellow. (B) The total spectral count determined by LC-MS/MS highlights the peptide sequence as a high turnover rate substrate for TMPRSS2. (C) IceLogo describes the extended substrate specificity of human TMPRSS2 based on 205 cleavage events detected by MSP-MS analysis.

The analogs based on initial hit compounds 1 and 2 and substrate specificity data are Ac-SKLR kbt (3), Ac-WFR-kbt (8), Ac-SKFR-kt (9), Ac-KQFR-kt (10 ), Ac-SQLR-kt (11), Ac-FLFR-kbt (12), Ac-dWFR-kbt (13), dWFR-kbt (14), dWFR-kbt-CO2H (15), Ac-WLFR-kbt (16), Ac-KQLR-kbt (17), Ac-LLR-kt (18), Cyclo(DMK)R-kbt (19), Cyclo(DQK)R-kbt (20) and Cyclo(allylGLY)R- kbt (21). According to predictions from PS-SCL studies, many of these compounds contain Phe (F) residues at the P2 position, while others contain Gln (Q) at the P3 position. When tested in the VSV pseudotype assay, all showed improved activity relative to 1 (Table 1), of which the best compound was 13. Compound 13 is a tripeptide that contains the unnatural amino acid D-Trp at the P3 position and Phe at the P2 position. It is predicted to be the first choice from the analysis of PS-SCL and MSP-MS. Compound 17 contains P3 Gln, which is also predicted from PS-SCL, and its activity is the second with an EC50 of 78 nM. In addition, 16, which also has P2 Phe, shows an excellent potency of 150 nM. The corresponding cyclic amide analogs 2, 19 and 20 retain good activity (EC50 of 119 and 138 nM) and are used to inhibit VSV pseudotyped virus entry, but the allyl phenyl ether analog 21 shows a significant decrease in activity (EC50 565 nM).

Structure and inhibition data of protease activity and VSV pseudotype and chimera SARS-CoV-2 virus entering human Calu-3 lung epithelial cells

The remaining compounds were tested in the VSV-SARS-CoV-2 chimera test. What is exciting is that the most promising results (Table 1) come from our new rationally designed TMPRSS2 inhibitor, 4 to 7. These compounds are designed by adding preferred side chains at all P1 to P4 (as determined by the PS-SCL data above) to design inhibitor positions. This group of existing compounds all show good efficacy. The best 12 of them have an EC50 of 101 nM in preventing viruses from entering Calu-3 cells, which is six times higher than the second-best analogue. In pseudotyping analysis, 17 Excitingly, three new rationally designed analogues, 4 to 6, showed significantly increased potency (Figure 6), with a sub-nanomolar EC50, while the fourth new compound, 7, still Very effective, with an EC50 of 3.6 nM. This not only increased the activity of 4 (EC50 0.43 nM) by 230 times compared to the previous best compound 12, but also increased by 80 times compared to Camostat. Among other compounds, tripeptide 15 (EC50 105 nM) showed similar potency to 12, but the potency observed from the analogs of 14, 13 was lower, which was the best compound tested in the pseudotyping assay. Also showing excellent efficacy is 10, with an EC50 of 262 nM.

Reasonably designed structure of TMPRSS2 covalently reversible kbt inhibitor (4 to 7).

In order to demonstrate the activity of a wide range of coronaviruses, we tested Camostat, 2 and our best new compounds 4 to 7 (Figure 6) based on the results of VSV chimera, which are active in blocking the entry of MERS-CoV chimeric virus into host cells . As shown in Table 2, compared with SARS-CoV-2, all compounds inhibited the entry of MERS-CoV virus with almost the same potency (Table 1). Although the dependence of MERS-CoV on TMPRSS2 has been previously demonstrated, this provides strong evidence that our compound has clinically broad-spectrum potential against other coronaviruses. To confirm that this activity is indeed related to TMPRSS2, we next set out to produce recombinant TMPRSS2 protein to test the enzyme inhibitory activity of our compound.

The activity of the lead compound on the entry of VSV pseudotype MERS virus into human Calu-3 lung epithelial cells

The recombinantly expressed TMPRSS2 protease domain (SI appendix, Figure S8A) was purified by Ni-nitrilotriacetic acid (NTA) column (SI appendix, Figure S8C), and then purified by sodium dodecyl sulfate polyacrylamide gel electrophoresis ( SDS/PAGE) and Western blot verification (SI appendix, Figure S8 D and E), showed the absence of other proteins. The catalytic activity was tested by combining 150 nM purified TMPRSS2 protease domain with (MCA)-K-KARSAFA-K-(DnP), an 8-mer peptide fluorescence resonance energy transfer (FRET) based on the synthesis of the following peptides Substrate From our MSP-MS analysis, TMPRSS2 is the most effective first choice (Figure 5). Michaelis-Menten kinetics showed a Km of 0.6 ± 0.05 μM (SI appendix, Figure S8B). Relative to the full-length TMPRSS2 (see below), the kcat (catalytic rate constant) value achieved by the protease domain (0.07/s) is 1,000 times lower than that observed for constructs carrying the LDLR class A domain and scavenger receptor The N-terminal of the cysteine ​​(SRCR)-containing module is connected to the catalytic domain (kcat = 94.54/s). Further research is needed to understand these differences and the interactions between the catalytic domain, LDLR and SRCR domains, which will affect the extended specificity and potential catalytic activity of TMPRSS2.

Using the active recombinant full-length TMPRSS2 as described above, together with Boc-QAR-AMC as the fluorescent substrate (44), we found that the Km (Mie's constant) was 85.6 μM when the enzyme concentration of 3 nM was used (SI appendix, Figure 2) . S2). In a standard kinetic analysis using Nafamostat and Camostat as controls, we used this substrate to test the inhibition of TMPRSS2 proteolytic enzyme activity by all inhibitors, where we measured the compound within 1 hour after incubating the compound with the enzyme for 30 minutes IC50. Camostat and Nafamostat have IC50 values ​​of 1.5 and 0.14 nM, which are similar but slightly improved compared to previously reported values ​​of 6.2 and 0.27 nM, respectively (44). We found that the IC50 data we generated for TMPRSS2 inhibition are closely related to the VSV pseudotype and chimera assay data, with some exceptions. Many compounds are significantly more potent than Camostat, some of which are equivalent to Nafamostat (Table 1). The new rationally designed lead compounds 4 to 7 (Figure 6) all show very effective sub-nanomolar IC50 values, of which 7 is the best with an IC50 of 250 pM, which is similar to Nafamostat. However, it should be noted that nalfalimstat is an irreversible inhibitor, while kbt inhibitors are reversible, so it is difficult to directly compare the IC50 values ​​between two series with different inhibitory mechanisms. Although the initial hit compound 1 still showed good potency against TMPRSS2 (IC50 of 74 nM), it was significantly weaker than the expected rationally designed TMPRSS2 inhibitor. However, another initial hit compound 2 is a more potent inhibitor of TMPRSS2 activity than 1, with an IC50 of 2.6 nM, which is only about 10 times lower than the activity of 4 to 7. It is worth noting that the P3 to P1 tripeptide 6 without the P4 side chain is almost as active as any tetrapeptide 4, 5 or 7 with P4 residues. However, this compound 6 and other tripeptides, including 8 and 13, are highly inhibited by plasma protease factor Xa, which is undesirable for further drug discovery due to the patient's potential bleeding side effects.

The least active compound is the aryl ether cyclic peptide 21 with an IC50 of 197 nM, however, this is still considerable. Presumably, the larger and more constrained aryl ether ring system is not ideal for binding to the active site of TMPRSS2 by bridging the S2-S4 pocket. It is important to note that although TMPRSS2 activity is effective in all inhibitors, most of the tested compounds still have relative activity against the other proteases HGFA, matriptase, and hepsin.

The most selective TMPRSS2 analogs are 16 with IC50 of 6.3 nM, which is 40 times more active than HGFA and twice that of Matriptase, while 12 has 60 times higher selectivity than HGFA and twice the selectivity for Matriptase . Like other compounds in this series of peptidyl kbt inhibitors, based on our experience, we have found that it is challenging to obtain selectivity to hepsin, but one of our best examples is the initial hit 2, which is relative to TMPRSS2. The activity on TMPRSS2 is four times higher than hepsin and the activity on matriptase is almost 10 times higher. The consequences of inhibiting these other serine proteases, especially other HGF-activated proteases, in the case of COVID-19 are not yet clear, but it may be beneficial rather than harmful to treatment.

The 21 compounds tested have a clear structure-activity relationship (SAR). For example, it seems that TMPRSS2 prefers the large group that extends beyond the C-terminal part of the inhibitor kbt S1' because 1 has higher activity than the unsubstituted kbt analog 3. Strengthening this hypothesis, compounds with smaller kt warheads in P1' positions as seen in 9 to 11 are also ineffective relative to analogs with larger kbt. In addition, the P3 position is important for activities that do not seem to require basic groups such as Lys (K) (9, 3, 1), and this seems to be the same for the P4 position seen in compound 17. Finally, in the PS-SCL data as modeled and predicted, Phe is clearly the first choice in the P2 position. Therefore, it is likely that various new analogs can be developed through other aromatic side chains (such as Trp and Tyr), which will also produce inhibitors with high TMPRSS2 activity and potentially higher selectivity.

Since the homology model of TMPRSS2 used in this study was only used for the catalytic domain of TMPRSS2, the catalytic domain and full-length TMPRSS2 were tested against candidate compounds. The inhibitory IC50 values ​​of the recombinant protease domain (SI appendix, Figure S1) and the full-length TMPRSS2 showed significantly different values. Compared with the commercial full-length picomolar to two-digit nanomolar IC50 value, the estimated IC50 value exceeds 100 μM TMPRSS2 (Table 1 and SI appendix, Figure S3). From our current research, it is not clear what is the direct cause of this significant difference, but we suspect that the protein domain/module of TMPRSS2 located at the N-terminus of the catalytic domain plays an important role in ligand interaction and subsequent catalysis.

After confirming and quantifying the inhibitory effects of all compounds on the TMPRSS2 enzyme of recombinant proteins and their selectivity to HGFA, matriptase, hepsin, thrombin and factor Xa, we evaluated the most promising lead compounds 2 and 4 to 7 in inhibiting cells The activity of the pathological effect was detected by the CellTiter-Glo (Promega) cell viability assay on the lung epithelial Calu-3 cells to detect wild-type SARS-CoV-2 (50). We used Remdesivir and Nafamostat (n = 1) as positive controls. Although all compounds showed excellent activity in this assay (Figure 7 and Table 1), the most effective compounds were rationally designed TMPRSS2 inhibitors 4 and 5 (Figure 6), with EC50 of 74 and 52 nM, respectively. >20 times better than Remdesivir and more active than Nafamostat (note that this is based on n = 1). Importantly, all five kbt inhibitors showed no signs of toxicity to Calu-3 cells (up to 50 μM, almost 1,000 times the IC50 of 5), while Remdesivir and Nafamostat were at the highest concentration tested (50 μM) The following shows toxicity.

(A) The CellTiter-Glo (Promega) assay was used to inhibit the cytopathic effect (viral toxicity) of wild-type SARS-CoV-2 Calu-3 with a lead TMPRSS2 inhibitor. (B) The viability of Calu-3 cells in the absence of virus (compound toxicity).

The in vitro stability of compound 2 (VD2173) and lead compounds 4 to 7 in mouse and human plasma were tested (Table 2). All compounds have excellent stability in mouse and human plasma, with a half-life of more than 289 minutes, but the half-life of 7 compounds in mouse plasma is 154 minutes, and the half-life of 2 compounds is 222 minutes. Based on potency, selectivity, and in vitro properties, compound 4 (MM3122) and compound 2 (VD2173) were selected as the main candidates, and their in vivo PK was tested in mice. Compounds 5 to 7 were not tested for PK because of their low antiviral activity, while the factor Xa activity of 5 and 6 was relatively high. As shown in Figure 8, the half-life of VD2173 (2) in mice is 2.5 hours area under the curve (AUC) and plasma exposure exceeds 24 hours. Excitingly, the rationally designed TMPRSS2 lead compound MM3122 (4) also has an excellent PK with a half-life of 8.6 hours in plasma. The lung exposure of VD2173 and MM3122 was then tested within 24 hours after intraperitoneal (IP) administration. It was found that both of these compounds reached high levels in the lungs and had excellent AUC, but MM3122 was even better, with a half-life of 7.5 hours, while VD2173 had a half-life of 4.2 hours. The efficacy of MM3122 and VD2173 in TMPRSS2 enzyme assays and virus assays is much higher (500 times), making MM3122 an ideal lead candidate for COVID-19 in vivo efficacy studies, and future reports will further optimize communication.

PK of mice after administration of MM3122 (4) and VD2173 (2) at 16.7 mg/kg IP. The change in the concentration of the compound in plasma and lung tissue over time is related to the IC50 of the compound in the SARS-CoV-2 VSV mixed cell into the test. The half-life (t1/2) of MM3122 (4) is 8.6 hours in plasma and 7.5 hours in lung, and the half-lives of VD2173 (2) are 2.5 hours and 4.2 hours, respectively. The dotted line indicates the IC50 of the compound in the VSVSARS-CoV-2 cell entry test.

We further tested the safety of the lead compound MM3122 in mice. MM3122 was administered to mice daily for 7 days by IP injection at three different single dose levels of 20, 50 and 100 mg/kg. No adverse reactions were observed in any treatment group. Compared with the control group, no weight loss or changes in harvested organs (liver, spleen, and kidney) were observed (SI Appendix, Figure S5).

As mentioned above, TMPRSS2 has been shown to be essential for SARS-CoV-2 host cell virus entry and replication. Based on molecular docking studies using the published TMPRSS2 homology model (45) and substrate-specific data from PS-SCL, we hypothesize that our existing set of peptidyl kbt inhibitors of HGFA, matriptase and hepsin will also inhibit TMPRSS2. In fact, we proved that these compounds not only effectively inhibit the TMPRSS2 enzymatic activity of recombinant proteins and cell surface proteins, but also inhibit host cell entry and the toxic entry driven by the Spike protein of SARS-CoV-2 into Calu-3 expressing ACE2/TMPRSS2 Lung epithelial cells. After further optimization, we have identified a variety of effective inhibitors of TMPRSS2. Four of them are rationally designed analogues of TMPRSS2 in terms of enzyme determination and blocking the entry of VSV-SARS-CoV-2 chimera into human Clau-3 epithelial lung cells. Shows sub-nanomolar activity. In addition, some of these compounds show excellent efficacy against another prominent coronavirus, MERS. We further confirmed the effective antiviral activity of the five lead compounds against the wild-type SARS CoV-2 virus, thereby identifying the most promising lead compound MM3122 (4). We have clearly determined that this TMPRSS2 inhibitor MM3122 is more effective than Remdesivir, Camostat and Nafamostat. Importantly, compared with Remdesivir and Nafamostat, this compound is also not toxic to Calu-3 cells, which is toxic at higher concentrations. MM3122 has excellent metabolic stability in mouse and human plasma, and has excellent PK and safety in mice. The cyclic peptide VD2173 (2) and the rationally designed TMPRSS2 inhibitor MM3122 both have an impressive half-life of >8 hours in mice. In addition to excellent TMPRSS2 inhibition, PK and antiviral activity, VD2173 also has anticancer effects in animal models of lung cancer. Judging from the protease selectivity data and the potency of MM3122 and other new compounds, it is expected that these compounds will also show corresponding anti-cancer activity, especially for HGF-driven prostate cancer in which TMPRSS2 is known to play a key role (30, 51) .

Recent studies have identified other serine proteases of the TMPRSS family that can mediate the entry of SARS-CoV-2, including TMPRSS13 (293T-ACE2 and Calu-3 cells) (9, 52, 53) and TMPRSS4 (intestinal-like cells) (16) . Although the dependence of entry into Calu-3 cells on TMPRSS13 is much lower than that of TMPRSS2 (9, 28), the mechanism contribution and/or redundancy of these proteases to SARS-CoV-2 entry is still unclear, and their relative inhibition is not yet clear. The scope of this research. Future work in animal models can address the biological and tissue-specific relevance of protease inhibition, and is best suited to answer these questions.

MM3122 represents an advanced drug candidate for clinical development. It is a new type of antiviral drug that can be used for COVID-19 and against infections caused by other coronaviruses (such as MERS). We have shown the equivalent activity of MM3122 in it. We are currently optimizing the TMPRSS2 selectivity and antiviral efficacy of MM3122 and such inhibitors, and will test MM3122 and other optimized leads in appropriate animal models of COVID-19 when appropriate. In addition to being new drugs, selective TMRSS2 inhibitors can also be used as valuable chemical probes to help elucidate viral pathogenesis, including host cell-virus interactions, Spike protein processing, and ACE2 receptor binding. Since TMPRSS2 plays a key role in the pathogenesis of other coronaviruses (SARS-CoV, MERS-CoV) and influenza viruses (25, 54⇓ ⇓ ⇓ –58), MM3122 and these new inhibitors, it may not only play a key role as a viral protein processing protease Effective against COVID-19, and effective against most or all infections caused by coronavirus and influenza viruses. Therefore, these small molecule inhibitors of TMPRSS2 are not only expected to become new drugs for the treatment of SARS-CoV-2 infection, but may also represent broad-spectrum antiviral drugs.

It has been widely reported that periplasmic secretory bacterial expression of serine proteases can produce correctly folded high-quality enzymes because its reducing environment allows disulfide formation and fewer proteases are present compared to the cytoplasm (59, 60). Therefore, we adopted a similar strategy for the TMPRSS2 protease domain. Using the standard Gibson assembly procedure (61), the human TMPRSS2 extracellular domain (residues 106 to 492) was cloned into the pET28a vector with the pelB leader sequence followed by the N-terminal 6x His tag (SI appendix, Figure S8A). The Gibson assembly reaction mixture was used to transform E. coli DH5a cells, the resulting plated colonies were selected and Sanger DNA sequencing was used to verify the correct assembly of the cloned product. The correctly cloned plasmid was used to transform E. coli BL21(DE3) cells, and then used for subsequent expression. Pick a single colony of BL21(DE3) cells and inoculate it into 50 mL luria broth (LB) containing 2% glucose and 50 mg/mL kanamycin culture, and grow it overnight at 37 °C and 220 rpm . Upgraded liquid medium (LB 0.1% glucose, 50 mg kanamycin) inoculate the overnight culture with an initial optical density (OD600) of 0.05 at 600 nm, and induce expression with a final concentration of isopropylthio-β-half The lactoside (IPTG) is 1 mM, and the OD 600 is 0.7. The expression is carried out at 16°C for approximately 72 hours. The cells were collected by centrifugation, and a periplasmic extract was prepared by osmotic shock. In short, resuspend the cell pellet in TES buffer [200 mM Tris·HCl pH 8, 0.5 mM (oxadionitrile) tetraacetic acid, 500 mM sucrose] and incubate at 4°C for 1 hour, then add Incubate in cold water at 4°C for 45 minutes. Centrifugation (10,000 × g for 30 minutes) at 4°C to separate the periplasmic extract, and add imidazole (10 mM final concentration) and MgCl2 (100 mL/L expression culture volume) combined with Ni-NTA resin in batches overnight (each Liter expression culture volume 2 mL slurry). Use 10 column volumes of wash buffer 1 (50 mM Tris, 250 mM NaCl, pH 7.6) and then 10 column volumes of wash buffer 2 (50 mM Tris, pH 7.6, 250 mM NaCl, 20 mM imidazole). Purification) and elution with elution buffer (50 mM Tris, pH 7.6, 250 mM NaCl, 4 mM benzamidine, 1 mM CaCl2, 500 mM imidazole) (SI appendix, Figure S8C). SDS/PAGE was performed using fractions of the eluate, and those bands showing the correct size of the TMPRSS2 protease domain were combined. The combined fractions were buffer exchanged with 25 mM citrate pH 6, 250 mM NaCl, 4 mM benzamidine, 1 mM CaCl 2 to remove excess imidazole and prevent autolysis. Concentrate in the same buffer in the ultrafiltration device. Use 25 mM citrate pH 6, 250 mM NaCl, 4 mM benzamidine, 1 mM CaCl 2 and 10% glycerol and Superdex 200 to increase 10/300 GL, and remove potential aggregates and degradation products by size exclusion chromatography. The fractions containing the protease domain of TMPRSS2 were combined by SDS/PAGE (SI appendix, Figure S8D). A final Western blot against the protease domain (TMPRSS2 monoclonal antibody [M05], clone 2F4) was performed to verify that the collected sample was the TMPRSS2 protease domain. The combined samples were quickly frozen for subsequent analysis (SI appendix, Figure S8E).

Recombinant human TMPRSS2 was purchased from Cusabio Technology (CSB-YP023924HU) and analyzed in 25 mM Tris·HCl, 150 mM NaCl, 5 mM CaCl2, 0.01% Triton X-100, pH 8.0, and a final concentration of 3 nM. In order to calculate KM, Boc-QAR-AMC (Vivitude, MQR-3135-v) was serially diluted in dimethyl sulfoxide (DMSO), and then each diluent was further diluted in the assay buffer to make the final The substrate concentration ranges from 0.514 μM to 200 μM, and the final DMSO concentration is 0.5%. Measured in a total volume of 30 μL in three replicate wells of a black 384-well plate (Nunc 262260), and measured the initial velocity on Biotek at 35 second intervals for 7 minutes, excitation wavelength is 360 nm, emission wavelength is 460 nm HTX plate reader. KM and Vmax are calculated from the Michaelis-Menten graph using GraphPad Prism. For inhibitor studies, compounds were serially diluted three-fold in DMSO and then pre-incubated with TMPRSS2 in assay buffer for 30 minutes at room temperature. The reaction started after Boc-QAR-AMC was added, and the fluorescence was measured for 90 minutes at 190 second intervals. The final concentrations of enzyme and substrate are 3 nM and 86.6 μM, respectively, and the inhibitor concentration ranges from 2 μM to 0.15 pM. The assay was performed in quadruplicate plates, and the IC50 was calculated from the dose response curve using GraphPad Prism. The IC50 values ​​of HGFA, matriptase, hepsin, factor Xa and thrombin are determined using our published assay (37⇓ –39).

Human recombinant TMPRSS2 (100 nM) expressed in yeast (see Expression of Human Recombinant TMPRSS2 in Pichia pastoris) was incubated with a physiochemically diverse library of 228 myristide peptides at a final concentration of 500 nM. At different time intervals (15, 60, and 240 minutes), aliquots were taken and then quenched with 1 equivalent volume of 8 M guanidine hydrochloride. Desalt the sample using a C18 tip (Rainin), and perform liquid chromatography (LC) using a quadrupole Orbitrap mass spectrometer (LTQ Orbitrap) and a 10,000-psi nanoACQUITY ultra-high performance liquid chromatography system (Waters) for peptides- MS/MS peptide sequencing analysis reversed-phase LC separation. The peptides were separated on a Thermo ES901 C18 column (75 μm inner diameter, 50 cm length) coupled with an EASY-SprayTM ion source, and the flow rate was 300 nL/min from 2 to 50% buffer B (acetonitrile, 0.5% formic acid). ). Survey scans were recorded in the range of 325 m/z to 1,500 m/z, and up to three strongest precursor ions (MS1 features with charge ≥ 2) were selected for collision-induced dissociation. The data was acquired using Xcalibur software and processed as described previously (62). In short, use MSConvert to process the raw MS data to generate a peak list. Then the peak list was searched in Protein Prospector v.6.2.2 (63) against a database containing sequences from the 228 fourteen peptide library. The search uses a precursor ion mass accuracy tolerance of 20 ppm and a fragment ion mass accuracy tolerance of 0.8 Da. Variable modifications include the conversion of N-terminal pyroglutamate from glutamine or glutamic acid and the oxidation of tryptophan, proline and tyrosine. Then use MSP-xtractor software (https://pharm.ucsf.edu/craik/research/extractor) to process the search, which extracts the peptide cleavage sites and spectral counts of the corresponding cleavage products. Spectral counting is used for the relative quantification of peptide cleavage products.

According to the previously described protocol (30), human recombinant TMPRSS2 was expressed and purified from yeast with slight modifications. In short, according to the manufacturer, the SRCR and serine protease domain regions of TMPRSS2-N249G were cloned into the pPICZα-B construct (EasySelect Pichia Expression Kit; Invitrogen), transformed into P. pastoris strain × 33, and then contained The concentration of Zeocin is selected on the plate. A single colony was grown overnight in 10 mL of buffer medium containing glycerol (BMGY) at 30 °C and 230 rpm. The overnight culture is used to inoculate 1 L of BMGY and grow to OD600 2 to 6. The cells were pelleted and resuspended in 100 mL of buffered medium containing methanol. Cells were induced with 5% methanol every 24 hours for 72 to 96 hours. The secreted TMPRSS2 was precipitated with 70% ammonium sulfate overnight at 4 °C, precipitated at 27,000 × g for 45 minutes, then resuspended in 50 mM Tris pH 8, 0.5 M NaCl, 0.01% CHAPS, and dissolved at 4 °C 2 hours. Then the solubilized protein was purified on a gravity column containing soybean trypsin inhibitor immobilized agarose (Pierce). The TMPRSS2 was concentrated and purified by size exclusion chromatography using 50 mM 83 potassium phosphate buffer, pH 6, 150 mM NaCl, 1% glycerol to prevent self-proteolysis.

The human TMPRSS2 open reading frame containing the PLX304 plasmid from ORFeome Collaboration (Dana-Farber Cancer Institute, Broad Institute of Harvard, and Massachusetts Institute of Technology [HsCD00435929]) was obtained from DNASU Plasmid Repository, and the control PLX304 vector was from Addgene. HEK-293T cells Grown in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% Fetal Bovine Serum (FBS) and seeded in a black 96-well plate (75,000 cells per well). On the second day, in 100 μL OptiMEM per well, the cells were transfected overnight with the control plasmid (PLX) or TMPRSS2 (PLX-TMPRSS2) with TransIT LT-1 Transfection Reagent (Mirus Bio). 24 hours after transfection, the medium was replaced with 80 μL phosphate buffered saline (PBS). Add inhibitors or PBS individually to the wells of the five specified concentrations and incubate at 25°C for 15 minutes. Then add the fluorescent substrate Boc-QAR-AMC (R&D Biosystems) to each well to a final concentration of 100 μM. Fluorescence (excitation 365 nm, emission 410 nm) was dynamically measured with a GloMax plate reader (Promega) at 37°C every 15 minutes for a total time of 150 minutes.

VSV-eGFP is a recombinant VSV expressing the GFP reporter gene (depending on the entry of VSV glycoprotein G), which has been described previously (64). VSV-SARS-CoV-2 is a replication-competent infectious VSV chimera. It uses SARS-CoV-2 Spike (S) protein instead of VSV G to enter the virus and express eGFP. It has also been described before (41, 65 ). VSV-MERS is created in the same way as VSV-SARS-CoV-2, except that the MERS spike (HCoV-EMC/2012 strain) with 21 amino acids deleted at the end is inserted instead of VSV G.

Inoculate human Calu-3 lung epithelial cells or Vero cells (African green monkey kidney) in 96-well black plates in DMEM containing 10% FBS for 24 hours (37 °C and 5% CO2). On the second day, cells were pretreated with inhibitors or vehicle control (DMSO) in 50 μL of serum-free DMEM for 2 hours, and then infected with multiple infections (MOI) with VSV-SARS-CoV-2, VSV-MERS or VSV-eGFP Is 0.5. 7 hours after infection (single round of infection), the cells were fixed in 2% formaldehyde, and the nuclei were stained with 10 μg/mL Hoechst 33342 (Invitrogen) for 30 minutes at room temperature. Wash the cells once and store them in PBS after fixation, and use the InCell 2000 analyzer (GE Healthcare) in the DAPI and fluorescein isothiocyanate (FITC) channels for automatic microscopy (10x objective lens, 9 fields per well) , Covering the entire well). Use the multi-target analysis module of the InCell Analyzer 1000 workstation software (GE Healthcare) to analyze the image to determine the percentage of GFP-positive cells in each well (top hat segmentation). The percentage of GFP-positive cells under experimental conditions was normalized to control (DMSO-treated) cells and expressed as relative infectivity. GraphPad Prism (version 8.4.2) is used to calculate the EC50 of each drug. Statistics (comparison of VSV-SARS-CoV-2 and VSV-eGFP with each drug; Student's t-test) were performed in Microsoft Excel. Three biological replicates were performed.

We analyzed the pseudotyped entry into the TMPRSS2-positive human lung cell line Calu-3 (7 , 29). VSV-G was used as a control because it does not rely on TMPRSS2 to enter the host cell. In addition to Calu-3 cells, we further used Vero cells (African green monkey, kidney) as a control, because these cells do not express TMPRSS2, so any reduction in SARS-2-S-driven entry is related to any non-specific side effect or cell toxicity. Each compound was tested in a separate pseudotyped batch in three separate experiments. We treated the cells (96-well format) with different concentrations of inhibitors or solvents (DMSO) diluted in medium (50 μL per well) at 37 °C and 5% CO2 for 2 hours, and then on the top and at 37 Incubate for 16 hours at °C and 5% CO2. Next, we measured the virus-encoded firefly luciferase activity (an indicator of pseudotyped entry into target cells) in the cell lysate. The data was normalized against the control (DMSO-treated cells = 100% pseudotyped entry) and plotted with GraphPad Prism (version 8.3.0) to calculate the effective concentration EC50 and perform statistics (compared with the corresponding control; use Dunnett's for two-way ANOVA post-test) .

We inoculated Calu-3 cells on a white flat-bottom 96-well plate with 10,000 cells per well, added DMEM, and added 10% heat-inactivated FBS, 100 U/mL penicillin-streptomycin (Life Technologies, Inc., 15140 -163), and buffered with 10 mM Hepes, and cultured in a humidified 5% CO2 incubator at 37 °C. After 16 hours, the medium was removed and replaced with 50 μL per well of the same medium, but containing 2% heat-inactivated FBS (D-2) instead of 10%. In D-2, the compound was serially diluted three times in a nine-point series and added to a 96-well plate in a volume of 25 μL, so that the final concentration range on the assay plate was 50 μM to 7.6 nM for 1 hour. Then, we infect the cells (with compounds still on them) with 25 μL of SARS-CoV-2 per well at an MOI of 0.1 plaque forming unit (PFU) per cell, with the final total volume of each well being 100 μL. We incubate the plate for 72 hours as described above, then add 25 μL of CellTiter-Glo reagent (prepared according to the manufacturer's instructions; Promega, G7573) to each well, then shake for 5 minutes and incubate at room temperature for 20 minutes before detecting luminescence. Use Biotek Synergy H1 plate reader. The same assay was used to determine the cytotoxicity of the compound, but instead of adding virus, we added 25 μL of D-2. These experiments were performed 3 times. We performed nonlinear regression on the nonlinear regression of log (inhibitor) and response, with variable slope, and performed nonlinear regression on the four parameters of the compiled results from the three experiments. Similarly, the IC50 is calculated repeatedly based on all three biology.

All animal procedures were performed in accordance with the guidelines and approvals of the Institutional Animal Care and Use Committee of the University of Washington. Maintain the animal in a controlled temperature (22 °C ± 0.5 °C) and light (12 L:12 D) environment. Provide standard laboratory food and water at will.

NOD-scid IL2Rgnull (NSG) male and female mice (three in each group) were injected with 0, 20, 50, or 100 mg/kg MM3122 (in 1% to 5% DMSO/95% saline; IP) daily for 7 sky. Daily assessment of drug toxicity for animal health (fur shrinkage, eye secretions, weight loss, dehydration, lethargy, hypothermia, abnormal tissue growth), and weight measurement every 2 to 3 days. At the end of the experiment, the animals were sacrificed and their organs were examined. Collect the liver, spleen and two kidneys and weigh them.

The PK of MM3122 was tested in mice (single IP administration, actual dose 16.7 mg/kg). Each of the eight animal groups contained three animals (male, CD-1), and plasma and plasma were collected at eight different sampling time points (0.08, 0.25, 0.5, 1, 2, 4, 8 and 24 hours) after administration. Collect lung tissue. The PK of VD2173 was tested in mice (single IP administration, actual dose 16.7 mg/kg). Each of the three animal groups contained three animals (male, CD-1), and plasma and lung tissue were collected from each group of animals at three different sampling time points (0.25, 8 and 24 hours) after administration. Use standard LC/MS/MS techniques for biological analysis of plasma and lung tissue extracts.

Figure S1 of the SI appendix provides single-point inhibition data using recombinant protease domains 4 to 6 with FRET substrates. Tables S1 and S2 in the SI appendix provide protease selectivity data for 1 (ZFH7116) and 2 (VD2173). The synthesis and NMR and high performance liquid chromatography mass spectrometry (HPLC-MS) spectra of new compounds 2, 4 to 7 and 19 to 21 are provided in Schemes S1-S3 in the SI appendix. The Km curve of Boc-QAR-AMC using the full-length TMPRSS2 is provided in Figure S2 of the SI Appendix. The IC50 inhibition curve of the full-length TMPRSS2/Boc-QAR-AMC is provided in Figure S3 of the SI Appendix. Figure S4 of the SI Appendix provides the cell-based enzymatic activity of compounds 1 and 2 in HEK-293 cells. The acute toxicity of MM3122 (4) is provided in Figure S5 of the SI Appendix. Figure S6 in the SI appendix provides the activity of 1 and 2 and Camostat using pseudotyped Vero cells, and figure S7 in the SI appendix provides the activity of the chimeric VSV-SARS-CoV-2 virus. The expression and purification of the protease domain of TMPRSS2 are provided in Figure S8 of the SI Appendix.

All research data are included in the article and SI appendix.

We thank David Griggs and Scott Campbell of St. Louis University for their pharmacokinetic studies on MM3122. We would also like to thank Michael Winter of the University of California, San Francisco for his assistance in the initial MSP-MS and PS-SCL analysis. This work was funded by Washington University's Siteman Cancer Center (Grants 16-FY18-02 and SCC P30CA091842) and the Barnes Jewish Hospital Foundation in St. Louis (JWJ) (Grant BJHF 4984). This work was also funded by NIH awards R43 CA243941 (JWJ and LK), R43 CA224832 (JWJ and LK), U19 AI142784 (CLS), P50AI150476 (CSC) and U19 AI070235 (MERging), and the Eosic Research Disease Foundation ( MER). Additional support was provided by a rapid grant from Emergent Ventures at the Mercatus 9 Center at George Mason University (CSC). SP thanks Bundesministerium für Bildung und Forschung (Germany: Federal Ministry of Education and Research; Bonn, Germany) for funding; RAPID Alliance, awarded 01KI1723D and 01KI2006D; RENACO Award 01KI20328A; SARS_S1S2 Award 01KI20396; Charité-Universitätsmedizinä Berlin (med Universitätsmedizinä Berlin) Consortium, 01KX2021 prize); Lower Saxony (authorization 14-76103-184); and German Research Foundation (granted PO 716/11-1 and PO 716/14-1). The work using live SARS-CoV-2 was funded by Burroughs Wellcome Fund Investigators in the Pathogenic of Infectious Disease Award (17008, CLS).

↵1M.M., VCD and MAT made the same contribution to this work.

Author contributions: MAT, DhC, ALL, CLS, MER, SP, SPJW, AJO, CSC and JWJ design research; MM, VCD, MAT, DhC, ALL, DP, AEMB, MH, JV, PK, NPA, AMK, CET Conducted research with JWJ; PWR and ML contributed new reagents/analysis tools; MM, VCD, MAT, DhC, ALL, AEMB, MH, JV, NPA, LK, CLS, MER, SP, SPJW, AJO, CSC and JWJ analyzed the data; MM, VCD, MAT, DhC, ALL, AJO, CSC and JWJ wrote this paper.

Competitive interest statement: These compounds are included in two patent application documents of the University of Washington, of which JWJ and VCD are the inventors. JWJ and LK own shares in ProteXase Therapeutics, and the company has been licensed for two patent applications.

This article is directly contributed by PNAS.

This article contains online support information at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2108728118/-/DCSupplemental.

This open access article is distributed under Creative Commons Attribution 4.0 (CC BY).

Thank you for your interest in advertising on PNAS.

Note: We only ask you to provide your email address so that the people you recommend the page to know that you want them to see it, and that it is not spam. We do not capture any email addresses.

Feedback privacy/legal

Copyright © 2021 National Academy of Sciences. Online ISSN 1091-6490. PNAS is a partner of CHORUS, COPE, CrossRef, ORCID and Research4Life.