Structural basis of TMPRSS11D specificity and autocleavage activation

Abstract

Transmembrane Protease, Serine-2 (TMPRSS2) and TMPRSS11D are human proteases that enable SARS-CoV-2 and Influenza A/B virus infections, but their biochemical mechanisms for facilitating viral cell entry remain unclear. We show these proteases spontaneously and efficiently cleave their own zymogen activation motifs, activating their broader protease activity on cellular substrates. We determine TMPRSS11D co-crystal structures with a native and an engineered activation motif, revealing insights into its autocleavage activation and distinct substrate binding cleft features. Leveraging this structural data, we develop nanomolar potency peptidomimetic inhibitors of TMPRSS11D and TMPRSS2. We show that a broad serine protease inhibitor that underwent clinical trials for TMPRSS2-targeted COVID-19 therapy, nafamostat mesylate, was rapidly cleaved by TMPRSS11D and converted to low activity derivatives. In this work, we develop mechanistic insights into human protease viral tropism and highlight both the strengths and limitations of existing human serine protease inhibitors, informing future drug discovery efforts targeting these proteases.

Fig. 1. The zymogen activation motif of TTSPs are cleaved by trypsin-like serine proteases and their residue composition is distinct to each TTSP.

a Schematic of an inactive (zymogen) TTSP at the cell surface. The catalytic Serine Protease (SP) domain is connected to the non-catalytic (stem) domains through a disulfide bond (S-S) and the zymogen activation motif peptide bond, shown as a pink line. The zymogen motif peptide bond is cleaved (indicated with scissors) to form (b) the matured TTSP that has enzymatic activity and can cleave protein and/or peptide substrates. c Multiple sequence alignment of the zymogen activation motif of all human TTSPs. TTSPs are colored by TTSP subfamily; hepsin/TMPRSS-black; HAT/DESC-blue; matriptase-magenta; corin-orange. The length of the zymogen activation motif is indicated in parentheses for each TTSP. Scissors and a dashed black line indicate where TTSPs are cleaved during protease zymogen activation.

Fig. 2. Active soluble TMPRSS11D is accessible by replacing its zymogen activation motif with a DDDDK sequence and its activity is blocked by peptidomimetic and small molecule inhibitors.

a Schematics of soluble human TMPRSS11D protein constructs that span the TMPRSS11D ectodomain. The two protein domains include the Sea urchin, Enteropeptidase and Agrin (SEA) and Serine Protease (SP) domains which are connected by a disulfide bond (S-S) and the R186-I187 peptide bond. Mutations targeting the R186-I187 cleavage site for each protein construct are indicated. b TMPRSS11D protein test expression studies from baculovirus-infected Sf9 insect cells. The indicated TMPRSS11D protein was purified from media through IMAC purification and protein content evaluated by SDS-PAGE. Samples were thermally denatured and reduced (4x Laemmeli buffer containing 5 mM β-mercaptoethanol, 95 °C, 5 min) prior to gel separation. c Purified, active dasTMPRSS11D protein. SDS-PAGE samples were thermally denatured and reduced (+) or were not heated and not reduced (−) in advance of gel separation. d Purified, activated eTMPRSS11D S368A protein. All protein gel images (b–d) are representative of n ≥ 3 independent biological experiments. e Chemical structure of an arginine ketobenzothiazole (kbt) peptidomimetic inhibitor. A ligand Position 1 (P1) arginine is shown, and N-terminal amino acid residues at P2 and P3 are shown in simplified format. f dasTMPRSS11D (left; 15 nM enzyme) and dasTMPRSS2 (right; 1.5 nM enzyme) half-maximal inhibitory concentration (IC₅₀) plots for the indicated small molecule and peptidomimetic inhibitors. Assays contained a final concentration of 100 µM Boc-QAR-AMC substrate and relative protease activities were determined across the first 60 s of the reaction after substrate addition. Inhibitors were pre-incubated with dasTMPRSS11D or dasTMPRSS2 for 10 minutes prior to the start of the assay. Data are shown as mean values +/− SD for experiments performed in technical duplicate across 4 independent biological replicates (total n = 8). Peptidomimetic 1 (PM-1): Ac-Glu-Gln-Arg-kbt. Peptidomimetic 2 (PM-2): Ac-Gln-Ser-Arg-kbt.

Fig. 3. Nafamostat rapidly acylates dasTMPRSS11D, then hydrolyzes to restore protease activity.

a The putative nafamostat covalent inhibition mechanism for TMPRSS11D. b Peptidase activity progress curves of dasTMPRSS11D (3 nM) with nafamostat at the indicated inhibitor concentrations added simultaneously with Boc-QAR-AMC substrate (100 µM final). c Peptidase activity progress curves of dasTMPRSS11D (8 nM) pre-incubated (10 min) with the indicated concentrations of nafamostat before being transferred to wells containing Boc-QAR-AMC substrate (100 µM final). Data for (b, c) are shown as mean values for experiments performed in technical duplicate (n = 2) and are consistent with results obtained across n = 4 independent biological replicates d Melting temperature shifts (ΔT_ms) of dasTMPRSS11D protein in the presence of the indicated concentrations of nafamostat (teal datapoints) or 6-amidino-2-naphthol (violet datapoints) ligands. Each assay contained 2 µg dasTMPRSS11D, 5X SYPRO orange dye, and 50 mM Tris pH 8.0 with 200 mM NaCl. Data are shown as mean values for experiments performed in technical triplicate (n = 3), with consistent data observed across n = 3 independent biological replicates. The ΔT_m data were curve-fitted for one-site EC₅₀ in GraphPad Prism.

Fig. 4. TMPRSS11D crystallizes by using intermolecular contacts with its own zymogen motif peptide occupying the substrate binding cleft.

a Schematic of the TMPRSS11D crystal lattice containing the TMPRSS11D serine protease (SP) domain and its cleaved zymogen activation motif (magenta squiggle). b Cartoon representation of the dasTMPRSS11D crystal structure (PDB 8VIS). The cleaved zymogen activation motif (magenta sticks) of Molecule A interacts with the substrate binding cleft of Molecule B. c Zoomed-in view of the dasTMPRSS11D active site occupied by the DDDDK¹⁸⁶−CO₂^- peptide. The TMPRSS11D catalytic S368 residue and the TMPRSS2 Subsite 1 (S1 residue) D362 are shown as salmon sticks. Additional TMPRSS11D subsites are denoted in gray text. d Zoomed-in view of the eTMPRSS11D S368A active site occupied by the LSEQR¹⁸⁶-CO₂^- peptide (PDB 9DPF). e Comparison of the cleaved zymogen activation motifs (attached through a disulfide bond) of the indicated TTSPs, with the SP domain of TMPRSS11D shown as a green surface. The terminal residues of the zymogen motifs of dasTMPRSS11D, eTMPRSS11D S368A, and human TMPRSS15 are shown as sticks. f Side view of the zymogen activation motifs from (e). g Side view of the back side of the TMPRSS11D SP domain. The stem domain of TMPRSS2 (salmon cartoon; PDB 7MEQ) is shown and is covalently attached to the SP domain of TMPRSS2 through an interdomain disulfide bond.

Fig. 5. Structural comparison of the S1’, S1, S2 and S3 binding sites of TMPRSS11D, TMPRSS11E, TMPRSS2 and TMPRSS13.

a Surface representation of TMPRSS11D, with Subsite 1’ (S1’; green), S1 (red), S2 (purple), and S3 (blue) subsites colored. The TMPRSS11D catalytic triad, S368-H227-D272, are shown as sticks. b Comparison of the amino acids at each protease subsite for every TTSP. TTSPs with experimental crystal structures are indicated in black text. Protease amino acids predicted to form salt bridges are indicated in red (electronegative) or blue (electropositive) whereas hydrophobic amino acids are indicated in gray text. Amino acids in green are predicted to participate in H-bonding with ligands. c S1 of TMPRSS11D shown in stick representation. The R186 ligand residue (from PDB 9DPF) is shown as green sticks. H-bonds between the ligand and the TMPRSS11D S1 are shown as dashed orange lines. TMPRSS11D residues- purple TMPRSS11E (PDB 2OQ5)-teal, TMPRSS2 (PDB 8V04)-yellow, TMPRSS13 (PDB 6KD5)-pink. d S1’ of TMPRSS11D. The carboxylate of R¹⁸⁶-CO₂^- is shown as green sticks and points towards S1’ (indicated by a green squiggle). e, f S2 and S3 of TMPRSS11D. g dasTMPRSS11D (15 nM enzyme) IC₅₀ plot for ketobenzothiazole (kbt)-containing peptidomimetics 3-5. Assays contained a final concentration of 100 µM Boc-QAR-AMC substrate and relative protease activity was determined across the first 60 s of the reaction after substrate addition. Data are shown as mean values +/−SD for experiments performed in technical duplicate across 4 independent biological replicates (total n = 8). PM-3:Ac-Glu-Glu-Arg-kbt, PM-4:Ac-Glu-D-Glu-Arg-kbt, PM-5:Ac-Glu-Orn-Arg-kbt.

Fig. 6. A proposed model of TMPRSS11D autocleavage activation and shedding from the cell surface.

a AlphaFold 2.0 model of full-length human TMPRSS11D (AF-O60235-F1). The transmembrane (TM) domain (green), the Sea urchin, Enteropeptidase and Agrin (SEA) domain (teal), and the serine protease (SP) domain (purple) are shown in cartoon representation. The TMPRSS11D zymogen activation motif is shown in magenta sticks. b Superposed model of the AlphaFold 2.0 TMPRSS11D protein structure (magenta sticks) and the cleaved TMPRSS11D zymogen activation motif (PDB 9DPF; yellow sticks). The TMPRSS11D cleavage site spanning the R186-I187 peptide bond is denoted with a scissor graphic. c Cartoon representation of the TMPRSS11D cleavage event in the SEA domain leading to protease shedding from the cell surface. The SP domains of the AlphaFold 2.0 TMPRSS11D structure (purple) and the dasTMPRSS11D crystal structure (cartoon) were superposed. The suspected SEA domain cleavage site is indicated with a black bar. d Graphical representation of the shed TMPRSS11D SP domain.