Structure of Human TMPRSS2 in Complex with SARS-CoV-2 Spike Glycoprotein and Implications for Potential Therapeutics.

The coronavirus disease 2019 (COVID-19) pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has infected more than 520 million people around the globe resulting in more than 6.2 million as of May 2022. Understanding the cell entry mechanism of SARS-CoV-2 and its entire repertoire is a high priority for developing improved therapeutics. The SARS-CoV-2 spike glycoprotein (S-protein) engages with host receptor ACE2 for adhesion and serine proteases furin and TMPRSS2 for proteolytic activation and subsequent entry. Recent studies have highlighted the molecular details of furin and S-protein interaction. However, the structural and molecular interplay between TMPRSS2 and S-protein remains enigmatic. Here, using biochemical, structural, computational, and molecular dynamics approaches, we investigated how TMPRSS2 recognizes and activates the S-protein to facilitate viral entry. First, we identified three potential TMPRSS2 cleavage sites in the S2 domain of S-protein (S2', T1, and T2) and reported the structure of TMPRSS2 with its individual catalytic triad. By employing computational modeling and structural analyses, we modeled the macromolecular structure of TMPRSS2 in complex with S-protein, which incited the mechanism of S-protein processing or cleavage for a new path of viral entry. On the basis of structure-guided drug screening, we also report the potential TMPRSS2 inhibitors and their structural interaction in blocking TMPRSS2 activity, which could impede the interaction with the spike protein. These findings reveal the role of TMPRSS2 in the activation of SARS-CoV-2 for its entry and insight into possible intervention strategies.

The coronavirus disease 2019 (COVID-19) pandemic, caused by the highly transmissible and virulent pathogen severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is a global health emergency[1,2] causing severe acute respiratory distress syndrome (ARDS) in humans. SARS-CoV-2 is an enveloped, single-stranded positive-sense RNA β-coronavirus[3] belonging to the family Coronaviridae, similar to SARS-CoV-1 and MERS-CoV-2.[4] To date, it has infected more than 520 million people resulting in more than 6.2 million deaths with varied incidences across the globe (www.covid19.who.int) as of May 2022. In different countries, various levels of virulence and pathogenicity have been identified, which could be due to the variance in viral strains and host factors, including host genetic makeup.[5−7] To develop improved therapeutics against SARS-CoV-2, understanding the molecular mechanism of viral entry and/or hijacking the host system and its interplay with the different and host proteins is imperative. The observed differences in the rates of infection across the world also raise an intriguing question of whether functionally relevant variants in ACE2, Furin, and TMPRSS2 contribute to differences in infection rates.[8,9]

Like other viruses, the entry of a coronavirus or SARS-CoV-2 into host cells is a vital determinant of viral infectivity and pathogenesis and also a significant challenge for host immune surveillance and combat strategies.[10,11] SARS-CoV-2 utilizes spike glycoprotein (S-protein) for host cell adhesion via host-receptor recognition followed by membrane fusion.[12,13] The N-terminal S1 subunit of S-protein contains the receptor binding domain (RBD) and binds to the peptidase binding domain on receptor angiotensin converting enzyme 2 (ACE2).[9,14] The cryo-electron microscopy (cryo-EM) structure of S-protein reveals the structural rearrangement of S-protein after binding to the receptor to allow fusion of the host cell and viral membranes.[13,15] It is also known that SARS-CoV-2 S-protein becomes activated through different host proteases.[16,17] Evidence suggests that coronavirus S-proteins are cleaved by host proteases such as furin at the S1/S2 cleavage site, exposing S2 to further processing by host serine proteases for subsequent viral entry.[16,18] Surface-expressed transmembrane protease serine 2 (TMPRSS2) is implicated in the activation of influenza A, influenza B, and coronaviruses, including SARS-CoV-2, to drive efficient pulmonary infection.[9,19] Inhibiting TMPRSS2’s proteolytic activity prevents efficient viral entry, making it a promising target for antiviral therapies.[20] In SARS-CoV-2, TMPRSS2 is known to attack the S2 prime (S2′) region and cleave the spike protein facilitating cell entry. In this regard, specific serine protease inhibitors such as Camostast and Nafmostast have been effectively used to block the TMPRSS2 as one of the therapeutic options against COVID-19 infection.[21] Recent clinical studies have also shown that mutations in host TMPRSS2 lower the infection rate in COIVD-19 patients.[22] However, the lack of structural and molecular studies concerning interaction of TMPRSS2 with S-protein limits our understanding of the key priming action of TMPRSS2 for the entry of the S-protein into the host cell. This warrants detailed structural and molecular studies to unravel the molecular mechanism and repertoire of the coronavirus hijacking the host system, which has potential therapeutic implications.

Here, using comprehensive structural, functional, and molecular modeling–dynamics studies, we report the structural basis of TMPRSS2-mediated activation of SARS-CoV-2 S-protein for its cellular entry. Along with known and functionally proven TMPRSS2 cleavage at the S2′ site, we report two additional potential TMPRSS2 cleavage sites in spike protein, which potentially contribute to immune evasion and cell infectivity. The structural studies also unveil the cleavage fragments of the S protein. Using computational approaches and structural interactions, we screened several clinically approved protease inhibitors (Chemostat, Upamostat, Nafmostat, and Bromhexine) that could potentially block TMPRSS2 activity and its further interaction with the S2 subunit of spike protein. The other two potential TMPRSS2 inhibitors are Ambroxol and Gabexate. These findings unravel one of the key strategies that SARS-CoV-2 adopts to infect humans while escaping immune surveillance, and the findings also provide insight into possible intervention strategies and development of other therapeutics.

First, to understand the structure and functional mechanism of TMPRSS2 and due to the lack of an available structure, we first constructed the soluble and surface-exposed functional domain structure of TMPRSS2 through de novo molecular modeling by employing two independent servers, SWISS-MODEL (https://swissmodel.expasy.org/) for homology modeling and I-TASSER for iterative threading-based prediction (www.zhang-lab.org), by selecting the best sequence- and topology-aligned structure [Protein Data Bank (PDB) entry 5CE1]. The modeled loops were reconfirmed and refined for the best fit to avoid steric hindrance clashes and ensure all residues are placed in Ramachandran favored positions using Coot (https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/). The model was validated using the C-score (confidence score) and TM score (structural similarity), demonstrating the most correct fold and confidence of the predicted structure. All amino acid residues were positioned according to their lowest-energy possible orientation in the final model, and it superimposes with 5CE1 (hepsin protease) at a 0.45 Å Ca root-mean-square deviation (RMSD) (Figure S1a–e). The overall structure of TMPRSS2 (amino acids 145–492) (Figure a) measures 42 Å in length and 24 Å in diameter, comprising an N-terminal (amino acids 145–243) activation domain and a C-terminal (amino acids 256–492) proteolytic or catalytic domain. The N-terminus of TMPRSS2 is oriented toward the transmembrane of the host cell, while the C-terminus is pointed outward and is arranged to receive its target peptides and/or proteins or spike protein. We mapped the catalytic site/pocket of TMPRSS2 using sequence and structural analysis of different serine proteases. Residues H296, S441, K432, W461, and Q438 were found to be highly conserved with other TTPs (type II transmembrane serine proteases) and act as prime residues for the catalytic activity. The N-terminus of TMPRSS2 (amino acids 1–145) belongs to the transmembrane region, which is not considered in this structural determination. We also compared our TMPRSS2 structure with recent AI-based computations from Alpha-fold,[23] which was found to be substantially comparable to our modeled structure as well as a recent unpublished crystal structure (PDB entry 7MEQ) with an RMSD of 0.32 Å (Figure e). The two minor observed differences are a change in the one-turn helix to a loop structure at Q317–G323 in the model and a small bending angle between the N- and C-terminal domains, which do not have a role in catalytic function.

Figure 1

Structures of TMPRSS2 and SARS-CoV-2 spike glycoprotein. (a) Surface and ribbon model showing the side view of the homology model structure of TMPRSS2. The substrate or catalytic binding region is labeled on top, and the unstructured and membrane regions are shown as dashed lines. (b) Surface and ribbon model showing the side view of SARS-CoV-2 spike glycoprotein. The S1 domain region is shown as a gray surface, and the internal S2 domain is shown as colored ribbons. Three protomers of the homotrimer are colored accordingly. (c) Domain arrangement of the S1 and S2 subunits of spike glycoprotein. Abbreviations: SP, signal peptide; NTD, N-terminal domain; RBM, receptor binding domain; FP, fusion peptide; HR1, heptad repeat 1; HR2, heptad repeat 2; TM, transmembrane domain; CP, cytoplasmic domain. (d) Cleavage of SARS-CoV-2 S-protein by TMPRSS2. The HEK293 cells were transfected with either an empty vector or the TMPRSS2 gene and incubated for 48 h in the absence or presence of the furin inhibitor MI-1851 (50 μM each). Cell lysates were subjected to SDS–PAGE and Western blot analysis using antibodies against the C-terminal Myc tag. Cleavage of the spike into S2, S2′, and S1 is shown. Figure adapted with permission from ref (29). (e) Overlay of our de novo validated structure of TMPRSS2 (blue) with Alpha-fold predicted (pink) and X-ray diffraction (green). Catalytic active sites and N- and C-termini are denoted.

With regard to the structure of the SARS-CoV-2 spike glycoprotein, the only two available cryo-EM structures (PDB entries 6VSB and 6VXX)[15,24] are incomplete and have several gaps in the built structure. Hence, we remodeled these structures to fill the missing loops using a previously published and validated model structure of full-length SARS-CoV-2 spike glycoprotein.[25] The loop regions connecting the S1 and S2 regions of the spike are highly flexible and hard to locate even in recent EM structures but play a key role in virion activation.[26] For enzyme catalytic activity, understanding the active site is critical. It has been shown that TMPRSS2 cleaved spike protein preferentially in the S2′ region (KPSKR815↓SFIED) in S-protien. With the aid of structure-guided serine protease cleavage sites, we mapped two additional potential TMPRSS2 binding or cleaving sites in S-protein (T1, amino acids 837–845, and T2, amino acids 976–986) located in the S2 domain region, as found in other serine protease target sites,[27,28] along with the putative sequence and structural analysis (Figure S2). The T1 and T2 binding sites on S2 spike protein were discovered in the structure of extended loop regions, where T2 is buried in S1 domains and likely exposed during either receptor binding or post-furin cleavage (PRRAR685↓SVAS). The T1 cleavage site on the S2 domain of S-protein, which is between the C-terminal region of the fusion peptide and the N-terminal region of heptad repeat region 1 (HR1) adjacent to the fusion peptide site, is rich in Lys and Arg residues. On the basis of cryo-EM and a molecular model of S-protein, the S1 and S2 domains and their inner structural arrangement are shown in Figure b. Similarly, the T2 cleavage site on the S2 domain overlaps with the C-terminal region of HR1 and contains similar residues (Figure c). It is also clear that most serine proteases target the Lys- and Arg-rich targets.

The SARS-CoV-2 spike protein endogenous cleavage assay demonstrates that human TMPRSS2 cleaves the spike protein at the S1 and S2′ regions even in the presence of MI-1851 (furin-specific inhibitor) (Figure d). The TMPRSS2 cleavage produces S1 and S2/S2′ fragments, but it has a strong preference for the S2′ region for cleavage.[29] Additionally, cleavage at the T1 site has also been reported. The absence of MI-1851, on the contrary, resulted in a greater cleavage of the spike, which was primarily driven by the endogenous furin. Furthermore, recent studies show that COVID-19 patients carrying WT-TMPRSS2 had an infection rate higher than that with the V160M mutation[22,30] and demonstrate the V160M mutation reduced TMPRSS2 processivity.[22] This suggests that WT-TMPRSS2 is one of the factors that contributes to enhanced virulence.

The functional and biophysical studies evidence that TMPRSS2 potentially recognizes and cleaves the spike protein at the S2 and S2′ sites (Figure d). To better understand the structural basis of the interaction of SARS-CoV-2 S-protein and TMPRSS2, molecular unbiased random docking and interaction studies were performed with the full-length and activated form (post-furin cleavage) of SARS-CoV-2 spike glycoprotein (S2 domain) and our modeled and validated TMPRSS2 (amino acids 145–492) as template structures. With these two individual structures, we docked the structures using two independent servers, Cluspro (https://cluspro.org/login.php) and HADDOCK 2.2 (www.bonvinlab.org/software/haddock2.2/), for further validation in the absence of water. The binding free energies were taken into consideration for selecting the best possible models. Further validation and refinement were completed by ensuring that the residues occupied Ramachandran favored positions using Coot (https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/). The final docked complex structures were then verified for the absence of any large conformational changes upon docking. The S-protein/TMPRSS2 complex structure is further validated and screened using different parameters (cluster size, mode of interaction, buried interphase, ΔG, etc.) obtained from the docking clusters (Figure S3). The prime interaction between S-protein and TMPRSS2 is mediated by classical protease cleavage binding mode (Figure ). The interaction encompasses a large buried interphase (2140 Å2), demonstrating the factual mode of interaction, and is mediated via electrostatic energy (−170 kcal), van der Waals energy (−63 kcal), and hydrogen bonding. The overall complex structure shows that TMPRSS2 binds to the S2′ site (amino acids 806–814; a loop region is extended outward and well-exposed) of the SARS-CoV-2 spike glycoprotein homotrimer. The active site amino acids in TMPRSS2 (Q276, E299, K300, P301, K340, K342, E389, K390, L419, S441, Q438, and W461) mediate major contacts for S-protein interaction (Figure a,b). Consistently, residues H296, S441, K432, and Q438 are highly conserved among several serine proteases.[31] This underlines the structural mechanism of human TMPRSS2 recognizing and cleaving the SARS-CoV-2 spike protein in the S2′ region.

Figure 2

Structural interaction between TMPRSS2 and SARS-CoV-2 spike glycoprotein. (a) S2′, the primary cleavage site of TMPRSS interaction in the S2 domain of S-protein in the S2′ region. (b) Enlarged view showing the detailed interaction between TMRPRSS2 and the S2′ site of spike protein. (c) Key residues and binding orientation topology at the interaction interface of the S2′ site. Color coding and labeling are the same in all of the figures. (d) T1, first model of TMPRSS interacting with the S2 domain of spike glycoprotein in the T1 region. (e) Enlarged view showing the detailed interaction between TMRPRSS2 and the T1 site in spike protein. (f) Key residues and binding orientation topology at the interaction interface of the T1 site.

The additional interaction and cleavage by the TMPRSS2 at the T1 and T2 sites of the S2 domain of spike glycoprotein occur via three possible modes (Figure S3b). The TMPRSS2 binding to the T1 site of spike protein (cluster 1 of the docked complex) represents another best possible mode of interaction (Figure d–f and Figure S4a–c) with a highest docking score of −101 with a large buried surface area of 1557 Å2. The interaction is mediated by electrostatic energy (−250 kcal), van der Waals energy (−45 kcal), and hydrogen bonding (Figure S3). The overall docked complex structure shows that TMPRSS2 binds to the T1 site (amino acids 835–850) of the SARS-CoV-2 spike glycoprotein homotrimer and adopts the typical protease binding mode. The TMPRSS2 catalytic pocket adopts a cup-like architecture and accommodates a large binding interface. With regard to interaction of TMPRSS2 with the T1 site (K835–I850) of activated spike protein (Figure d–f and Figure S4a–c), the extended and well-exposed loop region (G832–N856) of S-protein assumes the peptide binding mode with TMPRSS2. The entire “S-shaped” loop region of the T1 site of the S2 domain passes into the canyon-like crevice or cup-like structure of the TMPRSS2 catalytic pocket (Figure d–f), and the interaction is mediated via hydrogen bonding and van der Waals energy. Among them, S-protein residues K835, Y837, D839, C840, L841, D843, I844, R847, and R848 are the key interacting residues and are well positioned with respect to the catalytic pocket of TMPRSS2 for processing the spike protein.

On the contrary, we also notice the possible interaction of TMPRSS2 with the T2 site [amino acids 975–987, the linker region connecting α1/α2 and α3/α4 of the S2 domain of the spike protein (Figure S4d–f)]. This includes the interaction with the pin/tip regions of the activated/primed S2 domain trimer containing α-helices via hydrogen bonding, electrostatic energy (−129 kcal), and van der Waals forces (−69 kcal) with a larger buried surface area of 2282 Å2 (Figure S3). Residues D745 and T747 of the α1/α2 helix region and residues N977, D978, L981, R983, D985, and K986 of the α3/α4 helix region of spike glycoprotein (S2 domain) are the key interacting residues and well positioned and inserted into the catalytic pocket of TMPRSS2 (Figure S4d–f). The detailed position and alignment of the amino acids of TMPRSS2 and spike protein involved in the interaction provide a high degree of confidence in molecular binding (Figure S4d–f). This suggests that T1 and T2 sites are also potential targets for the TMPRSS2 protease.

To understand the real-time in-solution behavior of the SARS-CoV-2 S-protein/TMPRSS2 complex, we performed virtual biophysical experiments using molecular dynamics and simulations using DynOmics (http://dynomics.pitt.edu/) and LARMD (http://chemyang.ccnu.edu.cn/ccb/server/LARMD/). The time course molecular simulations for 10 ns of dynamics were recorded. As the main catalytic domain of TMPRSS2 is involved in only target spike protein interaction or drugs, which is stable and resembles the crystal structure, we studied the dynamics using only one conformation or ensemble of TMPRSS2. The structure of spike protein in complex with TMPRSS2 was solvated in an area of 125 Å3 with water molecules using the Desmond Builder (Schrödinger, LLC, New York, NY). All simulations were performed after energy minimization, and equilibration systems were neutralized with counterions. The simulations were performed for 10 ns using LARMD at a pressure of 1 atm and a temperature of 300 K. All analyses were carried out using LARMD analysis tools. The MD trajectories were analyzed to identify critical interactions that were formed, retained, and disrupted in the interface between spike protein and TMPRSS2. The DynOmics program was used to generate the changes in the B-factor, eigenvectors, and mode of interaction analysis using default settings. The B-factor profiles (thermal stability factor), RMSD, and domain separation analysis combined with simulation studies were performed and validated with Schrodinger molecular dynamics tertiaries.

Our extended biophysical, molecular dynamics, and simulation studies also principally suggest that the overall SARS-CoV-2 S-protein/TMPRSS2 complex is stable with regard to its interaction and dynamic motion (Figure and Figure S5). First, the inter-residue contact map of the SARS-CoV-2 spike/TMPRSS2 complex shows the clear and robust physical interaction between the molecules even in the dynamic or in-solution state (Figure a,b). The entire loop region of spike protein at S2 interacting with TMPRSS2 shows stability even under dynamic motion. The dynamic state of the complex was allowed to oscillate up to 4.5 Å (intermolecule lines in Figure a) (Movie S1). With respect to the B-factor, all residues in the SARS-CoV-2 S-protein/TMPRSS2 complex showed significantly lower B-factor values of <0.5 Å2 at the interaction interface (Figure c), which further supported the greater physical stability of the complex. In comparison, considerable B-factor values were observed in the disorder regions or loops. In addition, the SARS-CoV-2 S-protein/TMPRSS2 complex with a lower B-factor of the structure also demonstrates the higher thermodynamic stability. Hence, we next sought to check the domain separation possibilities of SARS-CoV-2 and TMPRSS2 through a biophysical study and time course eigenvectors (domain separation dynamics). As expected, very low eigenvectors were observed for the whole complex-forming region (eigenvector score of <0 ± 0.01), suggesting the high stability of the complex and the least likely physical separation possibilities (Figure d). On the contrary, some increase or higher eigenvectors were noticed for the S1 domain regions of SARS-CoV-2, which further evidence the separation of S1 domains (Figure d). Small differences in eigenvector levels of spike monomers could result from differences in interdomain interaction. Increased eigenvectors are directly linked with a higher likelihood of domain physical separation or movement from the rest of the complex. Consistent molecular dynamics results were also obtained with the post-furin cleaved spike protein[8] and with regard to TMPRSS2 recognition of spike (Figure S5). The time course simulation of the SARS-CoV-2 S-protein/TMPRSS2 complex for 10 ns was recorded, and the mobility scale (Movie S1) also shows that the interaction interphase is less mobile, thus providing a high degree of confidence of complex formation and physical stability. Furthermore, the mode shape oscillation profile shows that S1 domains of SARS-CoV-2 have different conformations but the S2 domains are more stable and exhibit minor changes in conformation, suggesting the stability of the S2 domains of the S-protein compared to the S1 domains.

Figure 3

(a) Intermolecular connectivity of the SARS-CoV-2 full-length trimer and TMPRSS2 complex during the real-time dynamic state and oscillation of atoms. Each monomer of the spike protein and TMPRSS2 are shown in different colors in the ball-and-stick model. (b) Ribbon model showing the mobility scale and stable complex-forming region. Highly stable and less stable residues are colored blue and red, respectively. (c) Molecular dynamics simulation studies showing the oscillation and B-factor (stability factor; the lower the value, the higher the stability) profiles of the S-protein/TMPRSS2 complex. The amino acid residue position is shown on the X-axis, and the degree of movement of amino acids as a B-factor is shown on the Y-axis. (d) Domain separation dynamics of the spike protein/TMPRSS2 complex. Low and studied eigenvectors for TMPRSS2 and moderate level for spike protein in most regions are noticeable, indicating the higher stability of the complex. (e) Mode shape profile of the spike trimer binding to TMPRSS showing the region that has a greater number of conformational shapes.

Currently, COVID-19 patients are undergoing treatment with a broad range of antiviral drugs such as remedesivir, arbidol, ritonavir, and other combinations of drugs, but the disease still warrants more drugs that aid in different functional aspects.[32−34] Drugs or inhibitors that could block the function of TMPRSS2 and structurally impede its interaction with the spike protein could be potential therapeutic targets. To explore the specificity and validate the precision of drugs that could fit in the catalytic pocket of TMPRSS2, we performed structure-guided screening of potential drugs via unbiassed random molecular docking, refinement, and dynamic studies with four potential TMPRSS2 protease inhibitors (Chemostat, Upamostat, Nafmostat, and Bromhexine hydrochloride) using two independent servers: (i) Cluspro (https://cluspro.org/login.php) with a refinement interface and HADDOCK2.2 (www.bonvinlab.org/software/haddock2.2/). Before that, both protein and protease inhibitors were prepared for these studies by ensuring the presence of all hydrogen atoms and water molecules at least 5 Å around the binding site or catalytic pocket using the Mastero package program.[35] The prediction results and drug binding location are assessed and corroborated on the basis of the solvent accessible surface area (SASA), C-score (confidence score), and Z-score (clash score) of the binding location and exposed residues of SARS-CoV-2 spike glycoprotein (Figures S6 and S7). We observed the specific binding location of all drugs directed to the catalytic pocket of TMPRSS2. Among the possible modes of small molecule binding, from the SwissDock server, it is evident that the three best possible solutions for the individual drug that could potentially bind to TMPRSS2 protein and the clusters C2 show higher redundancy of drug binding and correlate with the TMPRSS2 catalytic pocket (Figure S6). It is also interesting to note that all four drugs bind to TMPRSS2 with a greater affinity of approximately −9 kcal/mol with a higher docking score (Figures S6 and S7).

We next analyzed interactions of individual drugs with the TMPRSS2 catalytic pocket of active residues Q276, H296, E299, K300, P301, K340, K342, E389, K390, L419, S441, Q438, and W461 (Figure ). It was interesting to notice that all four inhibitors (Chemostat, Upamostat, Nafamostat, and Bromhexine hydrochloride) predominantly bind to the specific location of the TMPRSS2 catalytic pocket, which excluded residues Q276, Q317, and E395 (Figures and 4). Intriguingly, Nafamostat potentially could bind in two different modes and adjoining positions in the TMPRSS2 catalytic pocket. The inhibitor Chemostat prefers the TMPRSS2 catalytic pocket for interaction with Q276, H296, K342, W461, L419, Q389, and S441 (Figure a). Bromhexine hydrochloride also docks at the catalytic triad of the TMPRSS2 pocket and docks into the hydrophobic groove, forming strong interactions with H296, K342, W461, L419, Q389, and S436 (Figure b). Nafamostat shows promising binding in two different sites. Mode 1 has a higher affinity to bind and interact with the residues (H296, E299, K342, S441, Q438, W461, S436, and D435), whereas in binding mode 2, Nafamostat interacts mainly with the TMPRSS2 residues (H296, E299, S441, Q438, Q317, and E395) (Figure c,d). Similarly, Upamostat interacts with TMPRSS2 via Q389, E299, E389, and K399 in addition to the key triad residues (H296, K342, S441, and Q438) (Figure e). The detailed position and alignment of amino acids of TMPRSS2 and individual drugs involved in the interaction are shown in Figure .

Figure 4

Detailed structural view of the interaction between different potential protease inhibitors and the TMPRSS2 catalytic site (color coding of amino acids for TMPRSS2 only). The position and residue names are labeled accordingly, and the types of interactions between the individual drug and neighboring amino acids are marked as shown in the legend: (a) Camostat, (b) Bromhexine, (c and d) Nafamostat binding in two different modes, and (e) Upamostat.

In this study, we identified and structurally demonstrated the key regions of the SARS-CoV2 S-protein involved in interacting with and processing of human TMPRSS2 to establish the underlying structural mechanism in TMPRSS2-mediated S-protein processing for viral entry in human cells. On the basis of available biochemical and structural data for the target sites of other serine proteases[36,37] along with sequence analysis, we identified two prime TMPRSS2 recognition sites in the SARS-CoV-2 S-protein (T1, amino acids 837–845; T2, amino acids 976–986). The two identified sites span a part of HR1 of the six-helix bundle of S-protein near the fusion peptide. Recently, it has been reported that regions HR1 and HR2 aid in bringing the fusion peptide into the proximity of the transmembrane domain, thereby facilitating membrane fusion.[13,38,39] Docking results of the T1 site showed the maximum docking score and largest buried area, suggesting that T1 is a better substrate binding site. We also examined TMPRSS2 recognition sites located in the three-dimensional structure of S-protein. The first target site was observed in the position adjoining the furin site (amino acids 837–848), which was partially buried and shielded with S1 domains, and the entire loop region was poorly exposed to solvent. On the contrary, T2, the second TMPRSS2 recognition site (amino acids 976–986) in SARS-CoV-2 S-protein, was observed to be completely buried or hidden inside spike S1 domains with no access to the external solvent content. This raises an intriguing question. If both TMPRSS2 sites are buried and shielded inside the spike glycoprotein with S1 domains, how would TMPRSS2 interact with spike glycoprotein? Furthermore, the distance from the S1 domain to the TMPRSS2 binding site measures 105 Å, and the membrane-exposed structure of TMPRSS2 measures only 42 Å from the membrane. This makes it challenging for TMPRSS2 to recognize its target site on spike glycoprotein without the cleavage of furin protease.[8,40] It is convincing to speculate that furin protease acts first on the spike glycoprotein in the S1/S2 region and cleaves the spike protein into S1 (ACE2 and CD26 binding region) and S2 (trimerization domain), resulting in the complete exposure of the S2 domain and TMPRSS2 recognition sites as suggested previously.[8,18,40]

To assess the efficiency of binding of TMPRSS2 with identified cleavage sites T1/S2′ and T2 on the S2 subunit of S-protein, we tested the clinically proven protease inhibitors Camostat, Nafamostat, Upamostat, and Bromhexine hydrochloride (BHH). Camostat, used therapeutically for unrelated clinical conditions, was shown to inhibit influenza viral replication,[21,41] while Nafamostat is a potent inhibitor of MERS S-protein-mediated membrane fusion.[42] BHH is a Food and Drug Administration-approved mucolytic agent and a specific inhibitor of TMPRSS2.[43,44] Upamostat is another serine protease inhibitor under consideration for clinical trials.[45] All four drugs bind to the active site of TMPRSS2 with a high degree of precision and specificity (Figure and Figure S5). We demonstrate that Camostat, Upamostat, and BHH preferentially bind to a specific location in the catalytic pocket of TMPRSS2, while Nafamostat binds to three additional binding residues in two modes. It is noticeable that residues H296, E299, S441, Q438, and W461 of the TMPRSS2 catalytic pocket are highly conserved and interact with all four drugs. Furthermore, these potential TMPRSS2 inhibitors also share their interaction via several polar, charged, and hydrophobic interactions (Figure S5). The specific binding of these drugs with a high degree of precision confirms that they could potentially bind and impede the interaction between the S-protein and TMPRSS2. The interactions of the TMPRSS2 catalytic pocket with the linker region connecting α1/α2 and α3/α4 of the S2 domain involving residues D745 and T747 of the α1/α2 helix region, residues N977, D978, L981, R983, D985, and K986 of the α3/α4 helix region of T1, and residues K835, Y837, D839, C840, L841, D843, I844, R847, and R848 of T2 clearly demonstrate that the cleavage sites identified are the regions where TMPRSS2 cleaves the S2 domain of spike protein leading to membrane fusion and viral entry into the host cell. Therefore, this conserved epitope can be targeted for the development of vaccines and therapeutic drugs. On the basis of these findings, we propose a model of interaction of TMPRSS2 with the S2 subunit of spike protein of SARS-CoV-2 (Figure ).

Figure 5

Proposed model of the TMPRSS2 binding S2 unit of SARS-CoV-2 spike glycoprotein. The three protomers of spike protein are colored accordingly.

Besides, in addition to the recent findings demonstrating host genetic mutations in altered SARS-CoV-2 virulence,[22] we discovered a number of other genetic mutations in human TMPRSS2 in our whole-exome sequencing analysis of nearly 60 000 humans (Figure S8a) derived from next-generation sequencing data from the GTIxp portal and the GenomeAD V3.1 repository.[46] The exome sequence data were filtered to extract only putative loss-of-function (pLOF) and missense mutations that occurred in the whole exome of human TMPRSS2 encoded on chromosome 21. All missense and deleterious mutations/SNPs or genetic alleles or variants are tabulated for further structural and binding analyses. Among all of these alleles, only eight mutations were found to be located in the active site pocket of TMPRSS2. That could be involved in the interaction with the target SARS-CoV-2 spike protein or any drug candidate. Furthermore, all tissue expression profiling (www.GTExportal.org) of TMPRSS2 reveals its highest and very high level of expression in the prostate followed by the colon (Figure S8), which could explain why men are more susceptible to SARS-CoV-2 infection than women.