TMPRSS2 (transmembrane serine proteinase 2) is a multidomain type II transmembrane serine protease that cleaves the surface glycoprotein HA (haemagglutinin) of influenza viruses with a monobasic cleavage site, which is a prerequisite for virus fusion and propagation. Furthermore, it activates the fusion protein F of the human metapneumovirus and the spike protein S of the SARS-CoV (severe acute respiratory syndrome coronavirus). Increased TMPRSS2 expression was also described in several tumour entities. Therefore TMPRSS2 emerged as a potential target for drug design. The catalytic domain of TMPRSS2 was expressed in Escherichia coli and used for an inhibitor screen with previously synthesized inhibitors of various trypsin-like serine proteases. Two inhibitor types were identified which inhibit TMPRSS2 in the nanomolar range. The first series comprises substrate analogue inhibitors containing a 4-amidinobenzylamide moiety at the P1 position, whereby some of these analogues possess inhibition constants of approximately 20 nM. An improved potency was found for a second type derived from sulfonylated 3-amindinophenylalanylamide derivatives. The most potent derivative of this series inhibits TMPRSS2 with a Ki value of 0.9 nM and showed an efficient blockage of influenza virus propagation in human airway epithelial cells. On the basis of the inhibitor studies, a series of new fluorogenic substrates containing a D-arginine residue at the P3 position was synthesized, some of them were efficiently cleaved by TMPRSS2.
The gene of TMPRSS2 (transmembrane serine proteinase 2; also known as epitheliasin) was first identified in 1997 on human chromosome 21 by systematic exon-trapping experiments. The full-length cDNA encodes a predicted protein of 492 amino acids , which is anchored to the plasma membrane and belongs to the family of TTSPs (type II transmembrane serine proteases). TTSPs are characterized by a short intracellular N-terminal domain, a transmembrane domain and a large extracellular domain containing a variable stem region and a C-terminal serine protease domain of the chymotrypsin S1 fold [2,3]. The TTSPs can be divided into four subfamilies on the basis of the phylogenetic analysis of the serine protease domain and the composition of the stem region , including the HAT (human airway-trypsin-like serine protease)/DESC, the matriptase, the hepsin/TMPRSS (transmembrane protease serine) and corin subfamilies.
TMPRSS2 belongs to the hepsin/TMPRSS subfamily, which contains an additional six proteolytically active TTSPs. Hepsin is an effective activator of the pro-hepatocyte growth factor in vitro and was the first protease which was recognized as a multidomain TTSP. The cloning of enteropeptidase cDNA, a long-known and well-studied protease in the digestive tract, revealed that it has a very similar structure and also belongs to the TTSPs. However, the physiological role of the remaining members of this subfamily, TMPRSS2–5 and MSPL (mosaic serine protease large-form; also known as TMPRSS13), is still unknown .
The multidomain stem region of TMPRSS2 contains an LDLRA [LDL (low-density lipoprotein) receptor A] domain, which forms a binding site for LDL and calcium, an SRCR (scavenger-receptor cysteine-rich) domain-group A and a serine protease domain with trypsin-like substrate specificity . In humans, TMPRSS2 is predominantly expressed in prostate, pancreatic and colon tissues, but also in lung, liver and kidney tissues [5,6]. Its expression is increased in prostate cancer cells and regulated by androgens [6,7]. Nevertheless, the physiological role of the protein is still unknown . TMPRSS2-knockout mice develop normally and survive to adulthood with no abnormalities in organ histology or function .
Two protein species of TMPRSS2 with apparent molecular masses of 70 and 32 kDa are described in literature. In cells and tissues, autocatalytic cleavage occurs between Arg255 and Ile256, providing the active protease . The still membrane-bound protease can be partially shed and liberated into the extracellular space . The protease domain contains a catalytic triad consisting of residues His296, Asp345 and Ser441  corresponding to His57, Asp102 and Ser195, according to the chymotrypsinogen numbering, which is used in the following paragraphs.
It has been demonstrated that, in addition to other proteases, TMPRSS2 can activate the G-protein-coupled PAR-2 (protease-activated receptor-2) in vitro, which is up-regulated in many tumour cells, including prostate tumours. Activation of PAR-2 causes increased levels of MMP (matrix metalloprotease)-2 and MMP-9, which are key proteases that enable tumour cells to metastasize . Host cell proteases have been shown to play an important role in activating surface glycoproteins of various viruses, which is a prerequisite for virus-cell fusion and replication. TMPRSS2 can cleave the precursor of the surface glycoprotein HA (haemagglutinin) of various influenza viruses at a monobasic cleavage site, including human viruses of subtypes H1, H2 and H3, which are responsible for millions of infected people worldwide during seasonal outbreaks and occasional pandemics . This cleavage is essential for the ability of HA to mediate fusion between viral and endosomal membranes following receptor-mediated endocytosis of the virus [11–13]. Owing to its crucial role in influenza virus replication, TMPRSS2 emerged as a potential target for the treatment of influenza infections. In analogy, the fusion protein F of the HMPV (human metapneumovirus) is synthesized as a single surface glycoprotein and requires cleavage by host cell proteases to enable virus propagation . This virus causes severe bronchiolitis and pneumonia predominantly in young children. In a previous study, it was demonstrated that TMPRSS2 is an efficient activator of the HMPV F protein . Furthermore, it was shown that the spike protein S of the SARS-CoV (severe acute respiratory syndrome coronavirus) can be activated by TMPRSS2 in addition to cathepsin L [16,17]. The cleavage of the S protein by TMPRSS2 seems to facilitate the direct entry of the SARS-CoV from the cell surface independent of cathepsin L activity in the endosomal pathway . In a recent study, it was described that the inhibition of TMPRSS2 by the non-specific serine protease inhibitor camostat caused a 10-fold reduction in infection of Calu-3 cells by SARS-CoV .
On the basis of this information, TMPRSS2 could be a potential target for drug design. However, in contrast with many other well-characterized trypsin-like serine protease, there is only very limited information available on the substrate specificity and inhibition of TMPRSS2 . The spread of the influenza virus after HA activation in TMPRSS2-expressing MDCK (Madin–Darby canine kidney) cells could be inhibited by the ovomucoid trypsin inhibitor and Pefabloc® SC [4-(2-aminoethyl)-benzenesulfonylfluoride], whereas aprotinin was less effective at the same concentration of 50 μM . Wilson et al.  described the inhibition of TMPRSS2 by the irreversible inhibitor Cbz (carboxybenzyl)-phosphono-Lys(OPh)2. However, most of these analogues have limited potential for further drug development, although an aerosol formulation of aprotinin has been approved in Russia for local respiratory application in mild-to-moderate influenza .
For further characterization of TMPRSS2 we have cloned and expressed its catalytic domain in Escherichia coli. The activated protein was used for screening the substrate specificity of TMPRSS2 with a set of chromogenic and fluorogenic substrates and to identify the first synthetic inhibitors of this protease with a reversible competitive binding mode. The most potent inhibitor was used to inhibit the propagation of human pandemic H3N2 and the 2009 H1N1 influenza strains in cell culture. On the basis of the results from enzyme kinetic studies, several new fluorogenic substrates were synthesized, some of which were efficiently cleaved by TMPRSS2. The results are described in the present paper.
Standard reagents for synthesis, protein expression and enzyme kinetics were purchased from Sigma–Aldrich, Fluka, Acros Organics and VWR. Amino acid derivatives were obtained from Bachem and Novabiochem®. The substrates used were purchased from Pentapharm and Sigma, or were synthesized in our group starting from commercially available H-Arg-pNa (p-nitroanilide)·2HCl (Orpegen) or H-Arg-AMC (7-amino-4-methylcoumarin)·2HCl (Pentapharm). Matriptase, which was used only as a reference in the SDS/PAGE analysis of TMPRSS2, was available from previous studies .
Cloning of the catalytic domain of TMPRSS2
The nucleotide sequence of the serine protease domain was amplified from the plasmid pCAGGS-TMPRSS2  by PCR using 5′-GGATATCATATGAAACATCACCATCACCATCACATCGTGGGCGGTGAGAG-3′ and 5′-GGATATGAATTCTTAGCCGTTTGCCTTCATTTG-3′ as sense and antisense primers respectively. The primers were chosen to introduce a sequence coding for Met-Lys-(His)6 at the 5′-end of the cDNA encoding the protease domain. The approximately 750-bp long amplification product was purified and subcloned into a pET24(b) vector (Novagen, Merck Bioscience) for expression into E. coli.
Expression, purification, refolding and activation of the catalytic domain of TMPRSS2
Expression, purification, refolding and activation were performed according to a strategy described previously for the preparation of the catalytic domain of the related TTSP matriptase . The expression vector, encoding the protease domain, was transformed into E. coli BL21 (DE3) CodonPlus competent cells. The cells were incubated in LB (Luria–Bertani) medium containing 30 μg/ml kanamycin at 37°C for 3 h and 220 rev./min. The expression of the catalytic domain was induced by the addition of 1 mM IPTG (isopropyl β-D-thiogalactopyranoside) at D600=1 and the incubation was continued for 10 h at 5°C. The cells were harvested and suspended in buffer (50 mM Tris/HCl and 0.9% NaCl, pH 7.5) and lysed via ultrasound. After DNA depletion with Benzonase (25 units/g of cell pellet, Novagen), the inclusion bodies were washed and denatured in denaturation buffer (8 M urea, 10 mM Tris and 100 mM sodium phosphate, pH 8.0). The denatured protein was freed from insoluble constituents by centrifugation and filtration (0.2 μm) and the His-tagged TMPRSS2 was purified by metal chelate chromatography (Ni2+-nitrilotriacetate agarose, Qiagen). TMPRSS2-containing fractions were pooled and renatured by rapid dilution in 50-fold volume refolding buffer (50 mM Tris, pH 7.5, 0.5 M L-arginine, 20 mM CaCl2, 1 mM EDTA, 100 mM NaCl, 0.05% Brij 58, 0.05 mM GSSG and 0.5 mM GSH). After 3 days of incubation at 8°C, the refolding solution was concentrated by tangential filtration (Vivaflow 200, 10 kDa cut-off, Sartorius) and the buffer was exchanged to activation buffer (50 mM Tris, pH 7.5, 1 M NaCl and 0.05% Brij 58).
The refolded TMPRSS2 was activated by removal of the N-terminal Met-Lys-(His)6 sequence, because a free isoleucine residue in position 16 at the N-terminus of the protease domain is required for activity. This was obtained by incubating the protease for 5 h with 2.5 m-units/ml of activated DAPase (Qiagen) at room temperature (~20°C). The activated protease was separated from non-activated protease and His-tagged DAPase by metal chelate chromatography and is later designated as active TMPRSS2.
A polyclonal antiserum against TMPRSS2 was obtained from GENOVAC. Rabbit antisera against H3 and H1 were derived from rabbits immunized with influenza viruses A/Aichi/2/68 (H3N2) and A/Hamburg/5/09 (H1N1) respectively, carried out under the appropriate ethical approval granted by the regional council of Hessen to the Institute of Virology in Marburg. Species-specific HRP (horseradish peroxidase)-conjugated secondary antibodies were purchased from Dako.
SDS/PAGE, Western blot analysis and Coomassie Brilliant Blue staining
Proteins were denatured in reducing SDS sample buffer by heating to 95°C for 5 min and subsequently subjected to SDS/PAGE (12% gel). Then, the proteins were stained by using 0.1% Coomassie Brilliant Blue or were transferred onto a PVDF membrane (GE Healthcare). Proteins were detected by incubation with TMPRSS2-specific and HA-recognizing antibodies respectively, and species-specific HRP-conjugated secondary antibodies, followed by subsequent incubation with Pierce ECL (enhanced chemiluminescence) peroxidase substrate (Thermo Sientific) and exposure of the PVDF membrane to autoradiography films (CEA, Germany) .
All measurements were performed at room temperature in 50 mM Tris/HCl buffer (pH 8.0; containing 154 mM NaCl). All substrate stock solutions (2 mM) were prepared in ultrapure water containing 10% DMSO and further diluted by water to the appropriate concentrations.
Measurements with chromogenic pNa substrates
The cleavage of the pNa substrates was measured at 405 nm using a microplate IEMS Reader MF 1401 (Labsystems). The initial screening was performed with a single substrate concentration of 200 μM in the assay. For the five best substrates, the enzyme kinetic parameters Km and Vmax were determined from two independent experiments.
Measurements with fluorogenic AMC substrates
The measurements were performed using a Safire2 fluorescence plate reader (Tecan; λEx=380 nm and λEm=460 nm). The Km and Vmax values were calculated as the average of two independent measurements.
The Ki determinations were performed according to the method of Dixon  using the fluorogenic substrate H-D-cyclohexylalanine-Pro-Arg-AMC (200, 100 and 50 μM). The Ki values were calculated as the average of two independent measurements.
Synthesis of new AMC substrates
Analytical HPLC measurements were performed on a Shimadzu LC-10A system [column: Nucleodur C18, 5 μm, 100 Å (1 Å =0.1 nm), 4.6 mm×250 mm; Machery-Nagel] with a linear gradient of ACN (acetonitrile) (solvent B) in water (solvent A) both containing 0.1% TFA (trifluoroacetic acid) (10–50% increase in solvent B at 40 min, detection at 220 nm) at a flow rate of 1 ml/min. The final substrates were purified to more than 97% purity (detection at 220 nm) by preparative HPLC (pumps: Varian PrepStar Model 218 gradient system; detector: ProStar Model 320, fraction collector: Varian Model 701) using a C8 column (Nucleodur, 5 μm, 100 Å, 32 mm×250 mm; Macherey-Nagel) and a linear gradient of ACN containing 0.1% TFA at a flow rate of 20 ml/min. All substrates were finally obtained as TFA salts after freeze-drying. Their molecular mass was determined using a QTrap 2000 ESI spectrometer (Applied Biosystems), as described previously [25,26].
The synthesis of the AMC substrates was performed by standard methods as described previously for the design of substrate analogue inhibitors of various trypsin-like serine proteases [26–28] and is shown for the most suitable substrate CH3-SO2-D-Arg-Pro-Arg-AMC (115) in 1. Briefly, mesyl chloride was introduced to H-D-Arg[Pbf]-OH [26,29], and provided intermediate 123 (MS calculated 504.17, found 503.35 [M−H]−, HPLC tR=35.3 min). Boc-Pro-OSu was coupled with commercially available H-Arg-AMC·2HCl, followed by cleavage of the Boc-protection group to give intermediate 124 (MS calculated 428.48, found 429.23 [M+H]+, HPLC tR=15.2 min). Both intermediates were coupled by PyBOP [benzotriazol-1-yl-oxytripyrrolidinophosphonium hexafluorophosphate], followed by cleavage of the Pbf group and final purification by preparative HPLC to provide substrate 115 (for analytical data, see Table 1).
Synthesis of substrate 115
|Compound .||Substrate .||MS calcd [M+H]+ .||MS found [M+2H]++/2 .||tR HPLC (min) .|
|Compound .||Substrate .||MS calcd [M+H]+ .||MS found [M+2H]++/2 .||tR HPLC (min) .|
Cells and viruses
Calu-3 (cultured human bronchial epithelial cells, HTB-55; A.T.C.C.) were cultured in DMEM (Dulbecco's modified Eagle's medium)/Ham's F12 (1:1 dilution) (Gibco) supplemented with 10% FBS (fetal bovine serum) and 1% penicillin, streptomycin and glutamine each, with fresh culture medium replenished every 2–3 days. MDCK cells were cultured in DMEM supplemented with 10% FBS, 1% penicillin, streptomycin and glutamine each. All cell growth and incubations occurred under humified air conditions at 37°C and 5% CO2. The influenza viruses used in the study were A/Hamburg/5/2009 (H1N1) (provided by Mikhail Matrosovich, Institute of Virology, Philipps University Marburg) and A/Aichi/2/68 (H3N2). Influenza A viruses were propagated in MDCK cells in infection medium (DMEM supplemented with 0.1% BSA, 1% glutamine, penicillin and streptomycin each) containing 1 μg/ml TPCK (tosylphenylalanylchloromethane)-treated trypsin (Sigma) .
Infection and multicycle viral replication in inhibitor-treated cells
All infection experiments were performed using infection medium. For analysis of influenza virus multicycle replication in Calu-3 cells in the presence of inhibitors, cells were seeded in six-well plates and grown to confluence. The cells were then inoculated with virus at a low MOI (multiplicity of infection) of 0.0001 for 1 h in the absence of inhibitors, washed with PBS and replenished with fresh infection medium containing inhibitors at the indicated concentrations. The cells were incubated for 72 h and at 24, 48 and 72 h post-infection, supernatants were collected and viral titres were determined as pfu (plaque-forming units) by plaque assay as described previously . Briefly, MDCK cells grown in 24-well plates were inoculated with 10-fold serial dilutions of each virus sample for 1 h. The inoculum was then removed and replaced by Avicel overlay containing 1 μg/ml TPCK-treated trypsin. The cells were incubated for 48 h and subsequently immunostained using virus-specific antibodies, HRP-conjugated secondary antibodies and the peroxidase substrate TrueBlue® (KPL).
To analyse cleavage of HA of progeny virus, Calu-3 cells were infected at a 100-fold higher MOI of 0.01 and incubated for 24 h (for A/Aichi/H3N2) or 48 h (for A/Hamburg/H1N1). Virus-containing cell supernatants were cleared from cell debris by low-speed centrifugation (4100 g, 5 min) and then pelleted by ultracentrifugation (Beckman Coulter rotor SW 41 Ti, 30000 rev./min, 2 h, 4°C). Pellets were resuspended in reducing SDS sample buffer, heated at 95°C for 5 min and subjected to SDS/PAGE and Western blot analysis using antibodies against H1 or H3 as described above.
To determine the viability of inhibitor-treated cells, a quantitative colorimetric MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide] assay (Sigma) was used. Calu-3 cells grown in 96-well plates were treated with PBS or the different inhibitors at the indicated concentrations in infection medium (total volume 100 μl per well) for 48 h at 37°C. Then, 20 μl of MTT stock solution (2 mg/ml in PBS) was added to each well and the cells were further incubated for 2–3 h at 37°C, until purple formazan crystals were visible. Finally, the MTT-containing medium was removed and the formazan dissolved in 200 μl of DMSO, after which the absorbance was measured at 562 nm on a microplate ELISA reader.
The catalytic domain of the serine protease of TMPRSS2 was cloned and expressed in E. coli. The protein was obtained in the form of inclusion bodies, which were separated from cell fragments by washing with buffer followed by centrifugation. To obtain native and correctly folded protein, the inclusion bodies had to be denatured in urea buffer, cleaned via affinity chromatography and renatured using the rapid-dilution method. After concentration, the emerged protein was activated by incubation with DAPase, a dipeptidylpeptidase, which in a stepwise manner removes the N-terminal Met-Lys-(His)6-tag. This provides the catalytic domain of TMPRSS2 with an N-terminal isoleucine residue (Ile16), which is a prerequisite for its enzymatic activity. The time-dependent activation of TMPRSS2 by DAPase could be monitored by cleavage of the chromogenic substrate CH3-SO2-D-Arg-Gly-Arg-pNa. Samples of the incubation solution were analysed every 30 min after starting the activation. After 5 h no further increase in activity was detected. Finally, DAPase was removed by metal chromatography.
The activated TMPRSS2 was analysed by SDS/PAGE followed by Coomassie Brilliant Blue staining or Western blotting with specific antibodies against its catalytic domain, whereby matriptase was used as a control. However, the amount of active TMPRSS2 obtained was not sufficient to detect the protein by Coomassie Brilliant Blue staining, whereas a 10-fold higher amount of purified matriptase was detectable by Coomassie Brilliant Blue (Figure 1a). In contrast, the activated TMPRSS2 was detected as a single band of approximately 28 kDa (lanes 1 and 2) by immunoblotting, whereas, as expected, matriptase could not be visualized by TMPRSS2-specific antibodies (Figure 1b). Owing to the very limited amount of protein, no further purification was possible and therefore only the crude preparation of active TMPRSS2 could be used for the substrate and inhibitor screening.
Characterization of the activated crude TMPRSS2 preparation, matriptase was used as reference
Only very few data describing the substrate specificity of TMPRSS2 are available. It was reported that the fluorogenic trypsin substrates Cbz-Gly-Gly-Arg-AMC  and Boc-Leu-Gly-Arg-AMC  were cleaved by TMPRSS2; although no Km or kcat values were provided. Therefore an initial screen with 16 chromogenic pNa substrates of various trypsin-like serine proteases at a fixed substrate concentration of 200 μM in the assay was performed, whereby the highest cleavage rate observed for H-D-cyclohexylalanine-Gly-Arg-pNa (1) was set to 100% (Table 2). These data reveal that a glycine residue in the P2 position and hydrophobic P3 amino acids in the D-configuration are well accepted. Substrates 1 and 2 possess identical Vmax values and show only marginal differences in Km. Both substrates contain an N-terminally unprotected P3 residue, whereby a Mes (methylsulfonyl) or methoxycarbonyl group is also well tolerated in the P4 position (3 and 4).
|Compound .||Substrate .||Cleavage rate (%) .||Vmax (M·s−1) .||Km (μM) .||Vmax/Km (s−1) .|
|Compound .||Substrate .||Cleavage rate (%) .||Vmax (M·s−1) .||Km (μM) .||Vmax/Km (s−1) .|
On the basis of the initial screen, the Michaelis–Menten curves were only measured for substrates 1–5, the Km and Vmax values obtained are summarized in Table 2. The quality of a substrate is normally described by the term kcat/Km, whereby the calculation of the kcat value requires the exact knowledge of the enzyme concentration used normally obtained from active-site titration [31,32]. However, owing to the very low enzyme content in the used TMPRSS2 preparation, we were not able to perform an active-site titration using 4-methylumbelliferyl p-guanidinobenzoate . It should be noted that active-site titrations using bovine thrombin as a reference were routinely possible with a 40 nM enzyme concentration in the assay under identical conditions, which were also used for the titration experiments with TMPRSS2. This suggests that the TMPRSS2 concentration in the enzyme stock solution was not sufficient to reach approximately 40 nM in the titration assay. Therefore only Vmax/Km values in the unit 1/s could be calculated. This term is proportional to the expression kcat/Km and therefore provides an identical ranking within a series of substrates, which were measured under identical conditions. The most efficiently cleaved substrates are analogues 3 and 5, both possessing a Mes group and glycine as P4 and P2 residues respectively.
Owing to the limited amount of TMPRSS2 available, additional fluorogenic substrates were tested, which allowed reliable measurements at 20-fold reduced enzyme concentrations (Table 3). The highest Vmax/Km value was determined for the substrate H-D-cyclohexylalanine-Pro-Arg-AMC (17), which was therefore used for the subsequent inhibitor screening.
|Compound .||Substrate .||Cleavage rate (%) .||Vmax (M·s−1) .||Km (μM) .||Vmax/Km (s−1) .|
|Compound .||Substrate .||Cleavage rate (%) .||Vmax (M·s−1) .||Km (μM) .||Vmax/Km (s−1) .|
In previous publications, several groups have reported the inhibition of various trypsin-like serine proteases by substrate analogue structures containing a 4-amidinobenzylamide [27,34,35] as a P1 mimetic or by arylsulfonylated amides of 3-amidinophenylalanine [23,36,37]. Although no crystal structure of TMPRSS2 is so far available, the catalytic domains of all trypsin-like serine proteases share a similar overall folding pattern. Therefore we assumed that such compounds could also be useful for TMPRSS2 inhibition.
Substrate analogue inhibitors
A first screen was performed with substrate analogue 4-amidinobenzylamide derivatives, which were described previously as inhibitors of HAT , matriptase  and thrombin . All of the compounds summarized in Table 4 contain a constant Bzls group in the P4 position and a proline residue as the P2 residue, whereby they differ only in the D-configured P3 amino acid.
The Ki values reveal a preference of TMPRSS2 for basic or hydrophobic P3 residues, whereby Ki values of approximately 20 nM were found for the most potent derivatives (24–27). It seems that TMPRSS2 accepts the more flexible D-homophenylalanine residue slightly better than D-phenylalanine, which was similarly observed for substrate analogue Factor Xa inhibitors in previous studies . A poor potency with Ki>1 μM was determined for the glycine and D-alanine inhibitors 38 and 39 or the D-aspartic acid analogue 40.
Proline is a preferred P2 residue of several trypsin-like serine proteases, especially for thrombin. However, its incorporation often provides inhibitors with limited selectivity. Therefore the influence of different P2 residues was investigated in a second series maintaining the well-accepted D-arginine as the constant P3 residue. Results shown in Table 5 indicate that alanine (41) and arginine (42) residues are suitable replacements for proline (24), whereby all other residues are less suited in the P2 position and provide inhibitors with Ki>50 nM.
Additional inhibitors were derived from compounds 25 and 33 (Table 4), whereby their t-butyl ester group in the P3 side chain of D-glutamic acid or D-aspartic acid was replaced by a series of piperazides or by morpholide (Table 6) . However, all of these analogues have reduced potency against TMPRSS2. The strongest inhibition within this series was observed for piperazide 54, which exhibits a 2.5-fold lower affinity compared with analogue 25 containing D-aspartic acid β-t-butyl ester as the P3 residue.
Several trypsin-like serine proteases, such as thrombin , Factor Xa , uPA (urokinase-type plasminogen activator) , Factor VIIa  plasmin and plasma kallikrein , prefer a Bzls group in the P4 position of substrate analogue inhibitors. The comparison of the inhibition constants of compounds 68–70 (Table 7) with their Bzls analogues (Table 4) reveals that the Bzls group also contributes to an improved affinity for TMPRSS2. Interestingly, the strongest potency with a Ki value of 20 nM was found for compound 66 containing an unprotected D-homotyrosine residue in the P3 position.
Some trypsin-like serine proteases accept chloro-substituted benzylamides as the P1 anchor . For instance, an excellent potency was found for substrate analogue inhibitors of thrombin  and Factor Xa  containing a 2-(aminomethyl)-5-chlorobenzylamide. However, the previously published inhibitors 71–75 (Table 8) [26,39] have reduced affinity compared with their 4-amidinobenzylamide analogues shown in Table 4. The most active compound (71) of this series contains a D-arginine residue in the P3 position.
Sulfonylated amides of 3-amidinophenylalanine
A first measurement within this series was performed with compounds 76–79 (Figure 2), which were originally described as matriptase  and thrombin inhibitors . The Ki value of 8 nM reveals a strong inhibition of TMPRSS2 by the matriptase inhibitor 76, whereby the thrombin inhibitors 77–79 possess reduced potency. These results encouraged us for an extended screen with related matriptase inhibitors. A potential drawback of inhibitor 76 could be its tribasic and strongly hydrophilic character, which is likely to limit its bioavailability. Therefore additional analogues of compound 76 were selected, which possess only two or one basic group within the molecule. In analogues 80–91 (Table 9) the C-terminal aminoethyl group on the piperidide moiety was replaced by uncharged urea structures of different length, whereby the N-terminal sulfonyl group was maintained . Several analogues have a comparable potency as found for compound 76 and possess inhibition constants of <20 nM.
Inhibition of TMPRSS2 by previously described 3-amidinophenylalanine-derived matriptase and thrombin inhibitors [
In the next series, the N-terminal β-alanylamide moiety of inhibitor 76 was replaced by substituted phenyl groups, whereby two different C-terminal piperidide residues were used (Table 10) [49,50]. In general, all dibasic inhibitors possess a 3–6-fold higher affinity against TMPRSS2 than their monobasic analogues. The strongest potency was found for compounds 92 and 94, which inhibit TMPRSS2 with Ki values of 0.9 and 1.0 nM respectively.
Design of new fluorogenic substrates
Although H-D-cyclohexylalanine-Pro-Arg-AMC (17) was identified as the best fluorogenic TMPRSS2 substrate in the initial screen and therefore used for the subsequent inhibitor studies, it possesses only a moderate Km value. However, the screening with substrate analogue inhibitors revealed a preference of TMPRSS2 for basic P3 residues in the D-configuration. This encouraged us to design a new series of fluorogenic AMC substrates containing a protected or free D-arginine residue in that position, whereby the well accepted and non-racemizable proline or glycine residues were used as P2 groups (Table 12).
All of these new substrates possess significantly reduced Km and in most cases also improved Vmax values compared with the reference 17 (Vmax=1.43·10−10 M · s−1 and Km=43 μM). The best AMC derivatives 115 and 116 contain a Mes group in the P4 position, whereby only negligible differences were observed between the use of glycine or proline residues as the P2 group. Both substrates have approximately 60-fold improved Vmax/Km values and their D-arginine side chain also contributes to an excellent solubility in water.
Suppression of influenza virus replication in human airway epithelial cells by inhibition of HA cleavage by TMPRSS2
The influenza virus surface glycoprotein HA is synthesized as a precursor protein HA0 and requires cleavage by an appropriate host cell protease into the subunits HA1 and HA2 to mediate fusion of the viral envelope and the endosomal membrane in order to release the virus genome into the cell after endocytosis . We identified TMPRSS2 as an HA-activating protease of influenza viruses with a monobasic cleavage site in the human airway epithelium . In the present study, Calu-3 bronchial epithelial cells were used to investigate the effect of TMPRSS2 inhibitors on influenza virus activation and multicycle replication in airway epithelial cells, because TMPRSS2 was shown to be responsible for activation of human influenza viruses in these cells .
First, we investigated whether treatment of Calu-3 cells with the potent inhibitors 92, 93, 113 and 114, possessing Ki values ≤5 nM, affects cell viability. Cells were treated with 10 or 50 μM of the selected inhibitors for 48 h and, subsequently, cell viability was measured using a quantitative MTT assay. Untreated cells were used as the control. In the presence of 10 μM inhibitor concentration, no reduction in cell viability was observed (results not shown). The presence of 50 μM inhibitors 92 and 114 did not affect cell viability, whereas a reduction of 20% was observed in cells treated with compounds 93 and 113 (Figure 3).
Quantitative viability assay of Calu-3 cells in the presence of 50 μM of the indicated inhibitors over a period of 48 h
On the basis of the results of the MTT assay inhibitor, 92 was selected for further studies, since it was the most potent TMPRSS2 inhibitor and had no cytotoxic effect. To investigate the efficacy of inhibitor 92 on influenza virus activation and propagation, Calu-3 cells were infected with human pandemic influenza viruses A/Aichi/2/68 (H3N2) or A/Hamburg/5/09 (H1N1) at a low MOI of 0.0001 and incubated in the presence of 10 or 50 μM compound 92 for 72 h. As a control, cells were infected in the absence of inhibitor 92. At the time points indicated virus titres were determined by plaque assay and virus-containing cell supernatants were subjected to SDS/PAGE and Western blotting to analyse HA cleavage. As shown in Figure 4(a), virus propagation was suppressed in a dose-dependent manner by inhibitor 92. Treatment of cells with 10 μM inhibitor caused a 10-fold and 100-fold reduction in virus titres of A/Hamburg H1N1 and A/Aichi H3N2 respectively, at 24 h and a 10-fold reduction at 48 h post-infection. Treatment with 50 μM inhibitor strongly suppressed virus propagation with a reduction in viral titres of 100–1000-fold at 24 h, 1000–10000-fold at 48 h and 100-fold at 72 h post-infection (Figure 4a). Cleavage of HA of progeny virus released from the cells was analysed by Western blotting at 72 h post-infection. As shown in Figure 4(b), the virus containing HA1 and HA2 was released from untreated cells and only a slight reduction in HA0 cleavage was detected in virus released from cells treated with 10 μM inhibitor. However, treatment of cells with 50 μM analogue 92 drastically reduced virus propagation and only low amounts of H3N2 containing non-cleaved HA0 were detected (Figure 4b, right panel). The amount of A/Hamburg H1N1 released from cells treated with 50 μM inhibitor was too low to be detected by immunoblotting (Figure 4b, left panel). Our data demonstrate that inhibition of virus activation by inhibition of TMPRSS2 activity using compound 92 strongly suppresses influenza virus propagation in Calu-3 airway epithelial cells. Therefore this analogue or related potent TMPRSS2 inhibitors could be promising candidates for the treatment of influenza virus infections.
Inhibition of influenza virus propagation by inhibitor 92
Although the TMPRSS2 gene was identified back in 1997, there has been little work carried out to further characterize this TTSP. Meanwhile, first reports have shown that proteolytically active TMPRSS2 could be involved in various viral diseases [11,13–17,52]. It was also suggested that TMPRSS2 might be involved in cancer progression via PAR-2 activation , although PAR-2 can be additionally activated by several other trypsin-like serine proteases in vitro such as trypsin, various tissue kallikreins, the coagulation Factors Xa and VIIa, mast cell tryptase, acrosin or matriptase . PAR-2 is also involved in inflammation and myocardial infarction; however, its relevant activator in vivo is still unknown. These reports encouraged us to screen our previously described [26,38,39,49,50] inhibitors of various trypsin-like serine proteases against TMPRSS2.
The expression of the catalytic domain of TMPRSS2 in E. coli according to a previous strategy for the preparation of matriptase  and its subsequent activation provided only a limited amount of active enzyme. Despite several attempts and modifications of the protocol, we could not identify a reliable refolding system, which provided larger quantities of active TMPRSS2 after the activation step with DAPase. In some cases, the protein also precipitated during the final activation with DAPase. Therefore the preparation of active TMPRSS2 requires further optimization to obtain sufficient amounts for structure determination. Nevertheless, the activated crude enzyme preparation could be used for the first substrate and inhibitor screening.
The initial substrate screening revealed a preference for analogues with hydrophobic P3 residues in the D-configuration, providing H-D-cyclohexylalanine-Pro-Arg-AMC as a suitable tool for the subsequent inhibition studies. During the course of investigation, the relatively high potency of substrate analogue inhibitors containing basic P3 residues suggested the synthesis of a new series of fluorogenic substrates containing D-arginine in that position. This provided not only excellently soluble compounds, but also several analogues with strongly improved Km and Vmax values resulting in a significantly higher net catalytic efficiency. The use of these new substrates will enable measurements with reduced enzyme and substrate concentrations in future.
Measurements with available substrate analogue inhibitors containing a 4-amidinobenzylamide as P1 residue provided several compounds with submicromolar inhibitory potency against TMPRSS2 possessing inhibition constants of approximately 20 nM. Although known examples exist for relatively specific substrate analogue inhibitors of, e.g. thrombin , uPA  or Factor Xa , they often suffer from limited selectivity against closely related trypsin-like serine proteases. For example, compound 24 causes a much stronger Factor Xa, thrombin and plasma kallikrein inhibition with Ki values of 2.4, 3.5 and 8.3 nM respectively . In general, most compounds with proline in the P2 position are relatively potent thrombin inhibitors [38,39]. In addition, a very strong Factor Xa inhibition (Ki=1.4 nM) was described previously for compound 41 .
The removal of the P4 Bzls group reduced the inhibitory potency against TMPRSS2 (Table 7); however, this drop in affinity was less pronounced as found for other proteases. Especially in the case of uPA  and Factor Xa , an approximately 1000-fold improved potency was found for the Bzls derivatives compared with their direct analogues with an unprotected P3 residue. Crystal structure analysis of substrate analogue inhibitor complexes with thrombin , uPA  or Factor Xa  revealed that this P4 Bzls moiety is most likely to contribute to the stabilization of a horseshoe-like inhibitor conformation by making strong van der Waals forces with the phenyl ring of the P1 4-amidinobenzylamide. In these crystal structures the Bzls residue is accommodated in a characteristic shallow subpocket above the disulfide bridge between Cys220 and Cys191. This disulfide bridge is highly conserved in all trypsin-like serine proteases and located close to the S1 binding site. Owing to the lack of any crystal structure, we can only speculate that this subpocket is less defined in the case of TMPRSS2, which explains the reduced contribution of the P4 Bzls group for its inhibition.
Thrombin and Factor Xa are two serine proteases, which prefer a 2-(aminomethyl)-5-chlorobenzylamide over 4-amidinobenzylamide as P1 residue [39,46,47]. X-ray studies with thrombin have demonstrated that the chloro atom points directly on to the π-system of Tyr228, making an attractive Cl–π interaction . This Tyr228 residue is highly conserved in nearly all trypsin-like serine proteases including TMPRSS2. However, in the case of TMPRSS2, a 2–8-fold reduced potency was found for the chlorobenzylamides shown in Table 8 compared with their benzamidine analogues. A reduced potency of substrate analogue inhibitors with chloro-aromatic P1 residues was observed also for uPA, plasmin, matriptase and trypsin (F. Sielaff and T. Steinmetzer, unpublished work). Whereas thrombin and Factor Xa possess an alanine residue in position 190 at the bottom of the S1 pocket, the proteases mentioned above, including TMPRSS2, belong to the Ser190 subfamily of trypsin-like serine proteases . Many crystal structures with proteases of the Ser190 subfamily reveal that the Ser190 side chain is hydrogen-bonded to a conserved water molecule directly located above Tyr228, which has to be displaced to allow the binding of chloro-substituted aromatic P1 residues. The lack of a comparable hydrogen bond with this water from the Ala190 side chain led to the assumption that the displacement of the more loosely bound water molecule is preferred with proteases of the Ala190 subfamily. However, we have to admit that several other proteases of the Ala190 subfamily also exist, such as matriptase-2 , HAT and plasma kallikrein (F. Sielaff and T. Steinmetzer, unpublished work), which are only poorly inhibited by this type of substrate analogue chlorobenzylamide inhibitors. Therefore additional reasons should exist as to why such chloroaromatic P1 groups are only suitable for inhibitors of certain trypsin-like serine proteases.
A significantly stronger inhibitory potency was found among the arylsulfonylated amides of 3-amidinophenylalanine, which were originally described as matriptase inhibitors . Within a first series containing a β-alanylamide substitution on the phenyl-3-sulfonyl group and a substituted uncharged urea structure at the piperidide moiety two analogues (86 and 89)  possess Ki values <10 nM (Table 9). The compounds of a second series are modified in their N-terminal sulfonyl and C-terminal piperidide residues. Analogues 92 and 94 have inhibition constants ≤1 nM, and are the most potent TMPRSS2 inhibitors identified so far. Both analogues contain a 2′,4′-dichloro- or dimethoxy-substituted biphenyl-3-sulfonyl group in combination with a C-terminal 2-aminoethyl-piperidide. The single chloro-substituted biphenylsulfonyl derivatives 99 and 100 have reduced potency. In all examples summarized in Table 10, the dibasic inhibitors containing the C-terminal 2-aminoethylpiperidide have enhanced potency compared with their monobasic 4-piperidylbutanoyl-methylamide analogues, although compounds 93 and 95 still inhibit TMPRSS2 with Ki values ≤5 nM. A very similar potency was also found for the monobasic benzamidine derivatives 113 and 114, which were obtained by combination of the N-terminal biphenyl-sulfonyl group with urea-substituted piperidides (Table 11). In principle, it should also be possible to design non-basic TMPRSS2 inhibitors with improved bioavailability by using one of the well-known prodrug strategies for benzamidine drugs . For example, a conversion into a hydroxyamidine or acylation was successfully used for the development of the orally available thrombin inhibitors ximelagatran and dabigatran.
The strongest TMPRSS2 inhibitor 92 significantly reduces the replication of the tested H1N1 and H3N2 influenza virus strains in Calu-3 cells. In this cell line, TMPRSS2 was identified as the major HA-cleaving protease . HAT was described as an additional HA-activating enzyme in the human respiratory tract. However, this protease is poorly inhibited by compound 92 with a Ki value of 1.7 μM and its expression was not detected in Calu-3 cells .
The results of the kinetic measurements revealed that the Ki values against TMPRSS2 strongly correlate with the inhibition constants against matriptase. For a total of 67 compounds, we have determined Ki values for both enzymes, which are plotted as pKi values in Figure 5. However, owing to the lack of any TMPRSS2 crystal structure, we cannot explain this very similar inhibition behaviour of both enzymes. Interestingly, both proteases also possess very similar P4–P4′ autocatalytic cleavage sites, which are RQSR↓IVGG and RQAR↓VVGG for TMPRSS2 and matriptase respectively (UniProt accession codes O15393 and Q9Y5Y6). This could indicate some kind of redundancy in physiological relevant substrate processing between the two proteases.
Correlation of the determined p
Ki values for the inhibition of TMPRSS2 and matriptase ( n=67 inhibitors)
In summary, the present study is the first dealing with the identification of synthetic inhibitors of TMPRSS2. We have to admit that none of the compounds identified by our screening is a highly specific TMPRSS2 inhibitor. Besides their matriptase affinity, some analogues also show a considerable potency against thrombin or Factor Xa, as described in previous studies [23,28,36,41]. It should be noted that our most potent compound 92 (Ki=0.9 nM for TMPRSS2) inhibits thrombin and Factor Xa with inhibition constants of 20 and 50 nM respectively. Although this might be a drawback, some inhibitors could be useful tools for further investigation of the physiological or pathophysiological function of TMPRSS2, e.g. for the inhibition of influenza virus propagation.
Dulbecco's modified Eagle's medium
fetal bovine serum
human airway-trypsin-like serine protease
Madin–Darby canine kidney
multiplicity of infection
severe acute respiratory syndrome coronavirus
transmembrane protease serine
transmembrane serine proteinase 2
type II transmembrane serine protease
urokinase-type plasminogen activator
Daniela Meyer prepared the TMPRSS2, performed the kinetic measurements and synthesized the new fluorogenic substrates. Frank Sielaff, Maya Hammami and Torsten Steinmetzer synthesized the inhibitors used. Eva Böttcher-Friebertshäuser performed the virological experiments and cell viability assay. Daniela Meyer and Torsten Steinmetzer wrote the paper, and Eva Böttcher-Friebertshäuser and Wolfgang Garten contributed to data analysis and discussion. Torsten Steinmetzer supervised the project.
M. Hammami was supported by a grant from the Yousef Jameel Scholarship Fund.