TMPRSS2 and ADAM17 Cleave ACE2 Differentially and Only Proteolysis by TMPRSS2 Augments Entry Driven by the Severe Acute Respiratory Syndrome Coronavirus Spike Protein

293T and Cos-7 cells were maintained in Dulbecco's modified Eagle medium (DMEM; Gibco, Invitrogen) supplemented with 10% heat-inactivated fetal calf serum (FCS; Biochrom) and 1% penicillin/streptomycin sulfate (Cytogen). The cells were grown in a humidified atmosphere at 37°C and 5% CO₂.

Expression plasmids for SARS-S and vesicular stomatitis virus G protein (VSV-G) as well as the plasmid used for production of the Fc-tagged S1 subunit of SARS-S have been described previously (39–41). Plasmids encoding the human transmembrane proteases TMPRSS2, TMPRSS3, TMPRSS4, TMPRSS6, and HAT, as well as murine, porcine, and avian TMPRSS2, murine TMPRSS4, and murine HAT, were also described earlier (26, 29, 42, 43). The sequences encoding the enzymatic inactive proteases TMPRSS2 and HAT with an N-terminal myc tag as well as the ACE2 mutants were generated by overlap-extension PCR and inserted into the plasmids pCAGGS (44) and pcDNA3.1 zeo, respectively. The integrity of all PCR-amplified sequences was confirmed by automated sequence analysis. Finally, for generation of lentiviral pseudotypes, the vector pNL-Luc-E⁻R⁻ was employed (45).

For the detection of ACE2 cleavage by TTSPs, 293T cells were cotransfected with an expression plasmid encoding ACE2 and either plasmids encoding the specified proteases or empty plasmid. The medium was replaced with fresh DMEM at 6 to 8 h posttransfection. At 48 h posttransfection, the cells were washed with phosphate-buffered saline (PBS) and lysed in radioimmunoprecipitation assay (RIPA) buffer (0.1% sodium dodecyl sulfate [SDS], 1% Triton X-100, 1% sodium deoxycholate, 50 mM Tris-HCl [pH 7.3], 150 mM NaCl). Subsequently, 2× SDS loading buffer was added and the lysates were incubated at 95°C. Thereafter, the lysates were separated by SDS-PAGE and blotted onto nitrocellulose membranes, and ACE2 was detected by staining with either a mouse anti-ACE2 antibody directed against the ectodomain (R&D Systems) at a dilution of 1:2,000 or with a rabbit anti-ACE2 serum specific for the C terminus (Abgent) at a dilution of 1:500, followed by incubation with horseradish peroxidase (HRP)-coupled anti-mouse or rabbit secondary antibodies (Dianova). As loading control, expression of β-actin was detected by employing an anti-β-actin antibody (Sigma). For the analysis of ADAM17-mediated ACE2 shedding, 10 μM phorbol 12-myristate 13-acetate (PMA; Sigma) was added for 16 h to cultures of ACE2-transfected 293T cells. Subsequently, the supernatants were harvested and cleared from cell debris by centrifugation. The proteins present in cleared supernatants were precipitated as described previously (36). In brief, trichloroacetic acid (TCA) was added to supernatants, the mixtures were incubated for 30 min at 4°C, and precipitated proteins were pelleted through centrifugation at 14,000 rpm for 5 min. Subsequently, the pellet was washed twice with ice-cold acetone and then resuspended in alkaline loading buffer (50 mM Tris [pH 8], 2% SDS, 100 mM dithiothreitol [DTT], 10% glycerol, bromophenol blue) and incubated at 95°C before analysis by SDS-PAGE and immunoblotting as described above. The program ImageJ (46) was employed for signal quantification.

The SARS-S-mediated host cell entry was analyzed by employing a lentiviral vector system, as described previously (40). In brief, lentiviral pseudotypes were produced by cotransfecting 293T cells with the HIV-1-derived vector pNL-Luc-E⁻R⁻ and expression plasmids for SARS-S, VSV-G, or empty plasmid (pcDNA). At 6 to 8 h posttransfection, the culture medium was replaced with fresh medium, and at 48 h posttransfection, the supernatants were harvested, passed through 0.45-μm-pore-size filters, aliquoted, and stored at −80°C. To analyze the impact of TMPRSS2 and HAT on SARS-S-driven transduction, plasmids encoding the receptor ACE2 or ACE2 mutants were cotransfected with plasmids encoding the indicated proteases or empty plasmid. One day prior to infection, the target cells were seeded in 96-well plates at 30,000 cells/well. Subsequently, pseudotypes were added and the cells were incubated for 6 h at 37°C. Thereafter, the medium was replaced with fresh medium. Finally, the luciferase activities in cell lysates were determined at 72 h postransduction using a commercially available kit (Promega).

The cathespin B/L inhibitor MDL28170 (Calbiochem), the ADAM17 inhibitor TAPI-1 (Calbiochem), or PMA (Sigma) was diluted in solvent as recommended by the manufacturers. For the analysis of entry inhibition by MDL28170, target cells expressing ACE2 or coexpressing ACE2 and protease were incubated with inhibitor (10 µM final concentration) or solvent alone for 60 min before pseudotypes were added. After incubation at 37°C for 8 h, the supernatants were removed and fresh culture medium without inhibitor was added. Transduction efficiency was determined at 72 h after the addition of pseudotypes as described above. In order to determine the impact of TAPI-1 and PMA on SARS-S-dependent transduction, target cells were incubated with pseudotypes at 4°C for 1 h followed by removal of unbound particles by washing. Subsequently, the indicated concentrations of TAPI-1 or PMA (see the legend to Fig. 8) were added and the cultures incubated at 37°C for 8 h. Thereafter, the culture supernatants were removed and replaced by fresh culture medium without TAPI-1 or PMA. Transduction efficiency was determined at 72 h after the addition of pseudotypes.

For analysis of ACE2 surface expression, plasmids encoding the ACE2 wild type (wt) or the specified ACE2 mutants were transiently transfected into 293T cells. At 48 h posttransfection, the cells were detached and washed with fluorescence-activated cell sorter (FACS) buffer (1× PBS, 5% FCS, 2 mM EDTA) and stained with a goat anti-ACE2 antiserum (R&D Systems). After binding of the primary antibodies for 45 min at 4°C, cells were washed three times with FACS buffer and incubated for 45 min at 4°C with Cy5-coupled anti-goat secondary antibody (Dianova). After three washes with FACS buffer, the cells were fixed with 2% paraformaldehyde (PFA) and analyzed in a Becton, Dickinson LSR II flow cytometer. FCS Express software (De Novo Software) was employed for data analysis.

In order to prevent retention of ACE2 cleavage fragments at the cell surface, the ACE2 and protease-coexpressing cells were pelleted and incubated with high-salt buffer (0.5 M, 1 M, and 1.5 M NaCl) on ice. Subsequently, the cells were pelleted again and the supernatants were collected. The proteins present in cleared supernatants were precipitated as described previously (36), and the presence of ACE2 in cell pellets and supernatants was analyzed by Western blotting as specified above.

For the analysis of cleavage of isolated ACE2, 1 μg of recombinant ACE2 (R&D Systems) was incubated with 0.2 μg of recombinant HAT (R&D Systems) in assay buffer (50 mM Tris, 0.05% [wt/vol], and Brij 35 [pH 9.5]) for 2 h at 37°C in a total volume of 25 μl. Subsequently, the reactions were stopped by the addition of SDS loading buffer, and the reaction products were analyzed by 12.5% SDS-PAGE and Western blotting. For the analysis of cleavage sites by mass spectrometry (MS), ACE2 cleavage products were separated on precast NuPAGE bis-Tris 4 to 12% gradient gels using a morpholinepropanesulfonic acid buffer system according to the manufacturer (Invitrogen). After colloidal Coomassie staining, gel bands were excised, and one part of the band was subjected to in-gel digestion with trypsin, while endoproteinase Asp-N was used for the other part. In-gel digestion and mass spectrometric analysis of the proteolytic peptides by matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) MS on an Ultraflex MALDI-TOF/TOF mass spectrometer (Bruker Daltonics) were performed as described previously (43, 47). Extracted peptides were dried, redissolved in 0.5% trifluoroacetic acid/0.1% octyl-glucopyranoside, and spotted onto an AnchorChip target (Bruker Daltonics) precoated with α-cyano-4-hydroxycinnamic acid. Preparation was according to either the standard thin-layer affinity method (47) or an adapted method to improve coverage of small hydrophilic peptides. For this purpose, 0.5 μl was spotted onto the matrix surface, let dry, and washed once with ammonium dihydrogen phosphate (10 mM in 0.1% trifluoroacetic acid [TFA]).

293T cells were seeded on coated coverslips, transfected with plasmids encoding ACE2 and TMPRSS2 or empty vector, and incubated with the Fc-tagged S1 subunit of SARS-S for 1 h. Incubation was performed at 4°C to allow binding but not uptake or at 37°C to allow both processes. Subsequently, the cells were prepared for confocal microscopy. For this, the cells were washed three times with PBS followed by 20 min of fixation with 4% PFA. To stop the PFA fixation, the cells were treated with 50 mM NH₄C for 10 min. Subsequently, the cells were washed three times with PBS and then permeabilized by treatment with 0.2% Triton X-100 for 15 min. Thereafter, the cells were washed again three times with PBS and then blocked with 3% bovine serum albumin (BSA). Bound SARS-S1-Fc fusion protein was detected by using a fluorescein isothiocyanate (FITC)-coupled, anti-human Ig-specific secondary antibody (Dianova), while ACE2 expression was detected by staining with a mouse anti-ACE2 antibody, followed by staining with Red-X-coupled anti-mouse secondary antibody (Dianova). Finally, the cells were washed again, the coverslips were embedded in mounting medium, and staining was analyzed by confocal microscopy employing a LSM5 Pa confocal microscope (Carl Zeiss). The FACS-based analysis of SARS-S1-Fc binding to cells was conducted as described previously (41).

Statistical significance was calculated using two-tailed Student's t test for independent samples.

We first examined if ACE2 cleavage by TTSPs is detectable in transiently transfected 293T cells, which we routinely employ to assess SARS-S-driven cell-cell and virus-cell fusion. Indeed, coexpression of ACE2 and TMPRSS2 or HAT of human and animal origin resulted in ACE2 cleavage with a C-terminal ACE2 fragment of 13 kDa being readily detectable in cell lysates (Fig. 1A and B, middle panels). ACE2 processing into a 13-kDa C-terminal fragment was also observed for hepsin, while TMPRSS3, TMPRSS4, and TMPRSS6 did not facilitate ACE2 proteolysis (Fig. 1A and B, middle panels), despite efficient expression in transiently transfected cells (not shown). Production of the 13-kDa ACE2 fragment was dependent on the enzymatic activities of TMPRSS2 and HAT, since the fragment was not generated in cells expressing enzymatically inactive mutants of these proteases (Fig. 1C). Finally, titration experiments showed that the efficiency of ACE2 cleavage was dependent on the protease expression level. Thus, low protease expression resulted in the generation of additional cleavage products of approximately 15 to 17 kDa (Fig. 1C), which likely represent incompletely processed C-terminal fragments of ACE2. In sum, our results demonstrate that TMPRSS2 and HAT remove a short C-terminal fragment from ACE2, in keeping with previous findings (28). In addition, a new ACE2-processing TTSP, hepsin, was identified, which cleaves ACE2 in a manner similar to that observed for TMPRSS2 and HAT.

The identification of the TMPRSS2 and HAT cleavage sites in ACE2 is a prerequisite to understanding how ACE2 cleavage by these proteases impacts SARS-S-driven entry. We employed recombinant ACE2 and HAT, both of which are commercially available, to address this question. Incubation of ACE2 with HAT but not reaction buffer alone produced an 80-kDa fragment (Fig. 2A), indicating that ACE2 proteolysis by HAT in cells can be reproduced in vitro with recombinant proteins. To identify the cleavage site, we performed a mass spectrometric peptide mapping analysis of intact ACE2 and the 80-kDa fragment using two different proteases with a complementary cleavage specificity, trypsin and endoproteinase Asp-N. In contrast to the peptide maps of intact ACE2, no peptides matching the C-terminal sequence of amino acids 603 to 733 were detectable in the trypsin/Asp-N digests of the 80-kDa fragment (Fig. 2B). This finding, together with the observation that the Asp-N cleavage product 598-DQSIKVRISLKSALG-612 is present in the Asp-N digest of intact ACE2 but not in that of the 80-kDa fragment (Fig. 2C), indicated that the HAT cleavage site resides in this 15-amino-acid sequence. When we screened the Asp-N digests for the corresponding candidate peptide species, we found the Asp-N cleavage product 598-DQSIKVR-604 to be present in the Asp-N digest of the 80-kDa fragment but not in that of intact ACE2 (Fig. 2D), indicating that the cleavage occurred at R604, in agreement with the preference of HAT for cleavage C-terminally of Arg (48). These results were confirmed by mass spectrometric sequencing of peptides 598-DQSIKVRISLKSALG-612 and 598-DQSIKVR-604 (data not shown). Thus, mass spectrometric analysis revealed that the 80-kDa fragment was generated upon HAT-dependent proteolysis of recombinant ACE2 at residue R604, corresponding to R621 of cellular ACE2. In order to elucidate if cellular ACE2 is also cleaved by HAT at this site or at an arginine or lysine residue located in close proximity, we generated K619A, R621A, K619A R621A, and K625A mutants. However, all ACE2 mutants were cleaved with an efficiency similar to that of the ACE2 wt in cells coexpressing HAT or TMPRSS2 (Fig. 2E). These findings indicate that the cleavage site in recombinant ACE2 might be located at R604, while the corresponding residue in cellular ACE2 and R621, as well as nearby basic residues, are dispensable for cleavage.

Since work with recombinant proteins did not allow the identification of the ACE2 site cleaved by HAT or TMPRSS2 in cells, we next employed mutagenesis to identify the cleavage site(s) of these proteases. Sequence analysis revealed the presence of five clusters of arginine and lysine residues located between R619 and the transmembrane domain of ACE2 (residues 741 to 761), which could be recognized by TTSPs (Fig. 3A). In order to assess the importance of these clusters for ACE2 cleavage by TMPRSS2 and HAT, we mutated the clustered arginine and lysine residues to alanine, resulting in mutants C0 to C4 (Fig. 3A). In addition, we combined a mutation of cluster 0, which harbors the residue (R621) cleaved in the context of recombinant proteins, with mutations of the remaining clusters, giving yield to mutants C0 + C1 to C0 + C4 (Fig. 3A). Expression of all ACE2 mutants in cell lysates was readily detected by Western blotting, although expression of mutants C2 and C0 + C2 was reduced compared to expression of the ACE wt (Fig. 3B and C, top panels). Analysis of ACE2 cleavage revealed that mutation of lysine and arginine residues in clusters 0 to 3 did not interfere with ACE2 processing by TMPRSS2 and HAT (Fig. 3B and C, bottom panels). In contrast, mutation of the arginine and lysine residues within cluster 4 markedly reduced cleavage by both TMPRSS2 and HAT (Fig. 3B, bottom panel), and cleavage was further diminished when a mutation of cluster 4 was combined with a mutation of cluster C0 (mutant C0 + C4) (Fig. 3C, bottom panel, and Fig. 3D). These results identify the arginine and lysine residues within cluster 4 (amino acids 697 to 716) as essential for ACE2 proteolysis by TMPRSS2 and HAT and suggest that cluster 0 (amino acids 619 to 625) contributes to efficient cleavage of cluster 4.

We next used our ACE2 mutants to investigate the role of ACE2 proteolysis in the TMPRSS2- and HAT-dependent enhancement of SARS-S-driven entry. As a prerequisite to these studies, we first assessed surface expression and receptor function of the ACE2 mutants in the absence of protease expression. Surface expression of most ACE2 mutants was robust, although mutation of one and particularly two arginine and lysine clusters invariably caused a modest decrease in ACE2 expression levels (Fig. 4A). A more notable reduction in ACE2 surface expression was measured upon mutation of cluster 2 (mutants C2 and C0 + C2) (Fig. 4A), in keeping with the reduced expression of these mutants in cell lysates (Fig. 3B and C).

The receptor activity of the ACE2 mutants mirrored their expression profile: most ACE2 mutants analyzed, including the protease-resistant mutants C4 and C0 + C4, supported transduction by SARS-S-bearing pseudotypes with an efficiency similar to that of the ACE2 wt (Fig. 4B). In contrast, a 5- and 50-fold-reduced transduction efficiency was measured for mutants C2 and C0 + C2, respectively (Fig. 4B).

In order to investigate protease-dependent enhancement of infectivity, the ACE2 wt or ACE2 mutants C4 and C0 + C4 were coexpressed with TMPRSS2 or HAT, and the efficiency of SARS-S-mediated transduction was examined. Coexpression of the ACE2 wt with TMPRSS2 and to a lesser degree with HAT augmented SARS-S- but not VSV-G-driven entry (Fig. 4C), as expected (28). In addition, augmentation of entry was dependent on the enzymatic activity of the proteases (Fig. 4C), again in keeping with published results (28). In contrast, expression of TMPRSS2 and HAT did not increase SARS-S-mediated entry into cells expressing the cleavage-resistant ACE2 mutants C4 and C0 + C4 (Fig. 4C), demonstrating that ACE2 proteolysis is essential for the protease-dependent augmentation of SARS-S-driven cellular entry.

The expression of TMPRSS2 in viral target cells can increase entry efficiency (Fig. 4C) (28) and allows for cathepsin L-independent SARS-S activation (26–28). The first phenotype, increased infection, is due to ACE2 cleavage, as demonstrated above. The second phenotype, cathepsin L independence, has so far been linked to SARS-S cleavage but could also depend on ACE2 proteolysis. To address this question, we asked whether TMPRSS2 expression also facilitates cathepsin L-independent entry into target cells expressing the cleavage-resistant ACE2 mutant C4. For this, the ACE2 wt or ACE2 mutant C4 were expressed in 293T cells, and the cells were treated with PBS or cathepsin B/L inhibitor MDL28170 and transduced by pseudotypes bearing VSV-G or SARS-S. All cells were readily transduced by pseudotypes bearing VSV-G (Fig. 5). Transduction facilitated by SARS-S was profoundly augmented upon expression of the ACE2 wt, and a further increase was observed upon coexpression of TMPRSS2, in accord with our previous observations (Fig. 4C). Transduction of cells expressing ACE2 alone was markedly reduced upon pretreatment with MDL28170 and transduction efficiency was rescued upon coexpression of TMPRSS2, in agreement with published data (26–28). Notably, the same effects were observed for cells expressing ACE2 mutant C4 in conjunction with TMPRSS2 (Fig. 5), demonstrating that ACE2 cleavage is not required for TMPRSS2-mediated, cathepsin L-independent entry into target cells.

The results obtained so far demonstrated that ACE2 cleavage is essential for augmentation of SARS-S-dependent entry by TMPRSS2 and HAT expression. In order to obtain insights into the underlying mechanism, we first investigated whether expression of TMPRSS2 facilitates SARS-S binding to cells. The soluble S1 subunit of SARS-S fused to human Fc (SARS-S1-Fc) bound efficiently to ACE2-expressing cells, while binding to TMPRSS2-positive cells was within the background range (Fig. 6A), and comparable results were obtained upon analysis of SARS-S1-Fc binding by immunofluorescence (data not shown). These results suggest that increased SARS-S binding does not account for the ability of TMPRSS2 to augment SARS-S-driven entry. Next, we assessed whether protease expression facilitates SARS-S uptake into target cells. For this, we incubated cells expressing ACE2 or coexpressing ACE2 and TMPRSS2 with SARS-S1-Fc and examined the cellular localization of ACE2 and SARS-S1-Fc by confocal microscopy. When 293T cells were incubated with SARS-S1-Fc at 4°C, bound SARS-S1-Fc and ACE2 colocalized at or close to the cell surface, and no difference was observed upon coexpression of TMPRSS2 (Fig. 6B). Similarly, both SARS-S1-Fc and ACE2 colocalized at the cell surface when cells were incubated at 37°C (Fig. 6B). Interestingly, a slightly different result was obtained with cells coexpressing ACE2 and TMPRSS2. Again, colocalization of SARS-S1-Fc and ACE2 at the cell surface was detectable, but a substantial part of the SARS-S1 signal was now localized inside the cell, just beneath the plasma membrane, consistent with more efficient cellular uptake of SARS-S under these conditions. In contrast, an increase in SARS-S1-Fc uptake was not observed upon coexpression of TMPRSS2 and the cleavage-resistant ACE2 mutant C4 (data not shown). Thus, TMPRSS2 expression increases uptake of SARS-S1-Fc and potentially authentic SARS-CoV, which might account for increased SARS-S-driven entry into TMPRSS2-positive cells.

It has been proposed that SARS-S binding to ACE2 induces ACE2 shedding by ADAM17, which in turn increases cellular uptake of SARS-CoV (34, 35). In the light of these findings, we thought to clarify whether TMPRSS2 and ADAM17 modulate SARS-CoV entry via similar mechanisms. For this, we first asked whether TMPRSS2, like ADAM17, facilitates ACE2 shedding. PMA is known to induce ACE2 shedding in an ADAM17-dependent fashion (49) and was used as positive control. Indeed, PMA treatment of ACE2-expressing cells induced ACE2 release into the supernatant in a concentration-dependent fashion (Fig. 7A). In contrast, efficient ACE2 release from cells coexpressing ACE2 and TMPRSS2 was only observed upon treatment with the highest concentration of PMA examined (Fig. 7A). Similarly, constitutive shedding of ACE2, which is known to be partially ADAM17 dependent (49, 50), was suppressed by TMPRSS2 expression (Fig. 7A, lanes without PMA). Finally, the failure of TMPRSS2-expressing cells to release ACE2 into the supernatant was not due to retention of cleaved ACE2 at the cell surface, as determined by high-salt washes, and suppression of ACE2 shedding was dependent on the enzymatic activity of TMPRSS2 (Fig. 7B). These observations indicate that TMPRSS2 does not facilitate ACE2 shedding and even interferes with ACE2 shedding by ADAM17.

The failure of TMPRSS2 to shed ACE2 suggests that TMPRSS2 and ADAM17 cleave ACE2 at different sites. To explore this possibility, we investigated the ADAM17 cleavage site employing our set of ACE2 mutants. The analysis of cell lysates for expression (Fig. 8, top panel) and TMPRSS2-mediated cleavage (Fig. 8, middle panel) of the ACE2 wt and mutants yielded results comparable to the ones we had obtained previously (Fig. 3), with a short C-terminal cleavage product being readily detectable for all ACE2 variants tested except mutant C4. In contrast, no C-terminal cleavage product was observed for cellular lysates upon PMA treatment of cells expressing the ACE2 wt and C0 to C3 (Fig. 8, middle panel), in keeping with the published observation that the cleavage product is instable and not readily detectable by immunoblotting (34, 51). A weak signal was consistently observed for mutant C4, suggesting that the amino acid changes introduced into this mutant stabilized the cleavage product. Finally, a prominent ACE2 signal was detected for supernatants of PMA-treated cells expressing the ACE2 wt and all mutants except C2 (Fig. 8, bottom panel), indicating that PMA treatment had induced ACE2 shedding and that shedding was inhibited by the changes introduced into mutant C2. Thus, arginine and lysine residues within amino acids 652 to 659 are critical for ADAM17-dependent ACE2 shedding.

Having demonstrated that different determinants in ACE2 control cleavage by TMPRSS2/HAT and ADAM17, we sought to further investigate the role of ACE2 processing by ADAM17 in SARS-S-driven entry. Our previous work indicated that ADAM17 activity is not required for SARS-S-mediated transduction or for spread of authentic SARS-CoV (36), findings that contrast with those of previous studies (34, 35). However, the activation of ADAM17 by PMA and potentially also by SARS-S is a fast, transient process (52), which might have been missed in our previous study, since cells were incubated for prolonged times with an excess of free virus in the presence and absence of ADAM17 inhibitor (36). To address this possibility, we investigated the role of ADAM17 in SARS-S-driven entry under conditions of synchronized infection in the absence of cell-free virus. For this, we incubated ACE2 wt-expressing 293T cells with SARS-S-bearing pseudotypes at 4°C, conditions which allow particle binding but prevent uptake. Subsequently, unbound particles were removed by washing and the cells were incubated with PMA or TAPI-1, which activates and represses ADAM17 activity, respectively. Thereafter, the compounds were removed and the transduction efficiency was quantified. Neither PMA nor TAPI-1 treatment appreciably modulated SARS-S-driven transduction under these conditions (Fig. 8B), indicating that ADAM17 activity is not required for S-protein-mediated transduction, at least in the experimental setup chosen here.