Angiotensin-converting enzyme 2 (ACE2) has been shown to be the functional receptor for severe acute respiratory syndrome coronavirus (SARS-CoV) (
18,
24), the etiological agent of an acute infectious disease that spreads mainly via the respiratory route. Although ACE2 is present in the vascular endothelial cells of all organs, SARS-CoV is highly pathogenic only in the lungs (
12). Furthermore, while ACE2 expression in the lung has been shown for both type I and type II pneumocytes (
12), cell tropism of SARS-CoV does not strictly correlate with ACE2 expression, suggesting that other factors are required to explain the pathogenesis of this disease (
8,
32). One such factor is the critical role played by host cellular proteases in the process of viral entry into cells. For example, a variety of proteases such as trypsin, tryptase Clara, mini-plasmin, human airway trypsin-like protease (HAT), and TMPRSS2 (transmembrane protease, serine 2) are known to cleave the glycoprotein hemagglutinin (HA) of influenza A viruses, a prerequisite for the fusion between viral and host cell membranes and viral cell entry. Cleavage of HA is critical for viral infection, with the tissue distribution of proteases determining cell tropism of virus strains (
16). There are two major mechanisms responsible for proteolytic activation of viral glycoproteins. For many viruses, such as the human immunodeficiency virus (HIV) and Nipah virus, cellular proteases (e.g., furin or cathepsin) cleave the glycoprotein during biogenesis, separating receptor binding and fusion subunits and converting the precursor glycoprotein to its fusion-competent state (
36). Alternatively, for other viruses such as Ebola and SARS-CoV, cleavage of the viral glycoprotein by endosomal proteases induces conformational changes during viral entry following receptor binding and/or endocytosis (
6,
20,
28,
31).
Three proteases—trypsin, cathepsin L, and elastase—have been previously reported to activate the spike (S) protein of SARS-CoV (
3,
13,
20,
30,
31). In the absence of proteases at the cell surface, SARS-CoV enters cells by an endosomal pathway, and S protein is activated for fusion by cathepsin L in the endosome (
14,
30,
37). Conversely, in the presence of proteases at the cell surface, such as trypsin and elastase, viral S proteins attached to the receptor at the host cell surface are activated, inducing envelope-plasma membrane fusion and direct entry of SARS-CoV into cells (
21). Viral replication in the latter case has been shown to be 100 times higher than replication via the endosomal pathway (
21), suggesting that the higher infectivity of SARS-CoV in the lungs could be due to an enhancement of direct viral cell entry mediated by proteases. In this study we test the possibility that TMPRSS2 is an activator of SARS-CoV entry into host cells. TMPRSS2 is highly expressed at epithelial cells in human lungs (
9,
26) and activates influenza A virus and metapneumovirus in culture cells (
4,
5,
7,
29). Here, we present data indicating that the distribution of TMPRSS2 correlates with SARS-CoV infection in the lung and that this protease can efficiently activate SARS-CoV S protein to induce virus-cell membrane fusion at the cell surface.
DISCUSSION
In the present study we showed the distribution of ACE2 in mild and severe inflammatory lesions of cynomolgus lungs infected by SARS-CoV. We found that type II pneumocytes, which are frequently observed in regenerated tissues during inflammation and which express high levels of ACE2, were refractory to SARS-CoV infection, whereas type I pneumocytes, which do not express detectable levels of ACE2, were readily infected by SARS-CoV. This inconsistency may be explained by the downregulation of ACE2 in SARS-CoV-infected cells, as previously reported (
11,
17). Here, we showed that the localization of TMPRSS2-expressing cells in normal lung tissues, rather than ACE2-expressing cells, is closely tied to SARS-CoV infection in mild lesions, indicating that TMPRSS2 may determine viral tropism at an early stage of SARS-CoV infection. However, TMPRSS2 expression occurred in cells adjacent to virus-infected cells, not in infected cells, suggesting that TMPRSS2 may also be downregulated in infected cells, as observed for ACE2. We have previously reported that SARS-CoV infectivity was enhanced in culture cells by the addition of exogenous elastase, which enabled virus entry via the cell surface (
21). Elastase is a major protease produced by neutrophils during inflammation and may be relevant to the high level of SARS-CoV replication in lungs that results in severe pneumonia in the mouse model (
1). We have also reported that in the late stage of SARS-CoV infection in the cynomolgus lung, which remarkably occurs with severe inflammation in the lower lobe, SARS-CoV is distributed in both type I and type II pneumocytes (
23), indicating that elastase may trigger viral entry via the surface of these cells and determine SARS-CoV pathogenicity in this late stage of infection.
Activation of viral glycoprotein by TMPRSS2 has been previously reported for influenza A virus and metapneumovirus (
4,
5,
7,
29). The most distinctive difference between these viruses and SARS-CoV is the stage during virus replication in which viral glycoproteins are cleaved by proteases. In influenza A virus and metapneumovirus, the protease makes a simple cut in the glycoprotein during maturation, in a manner similar to that made by furin. In contrast, SARS-CoV S protein is cleaved by the protease following receptor-induced conformational changes. The protease cleavage site in S protein, located nearer the C-terminal region than predicted for a cleavage site, is thought to be exposed only after receptor binding (
2,
35). In support of this model, we recently reported that the S protein of mouse hepatitis virus type 2 (MHV-2), which is highly similar to the S protein of SARS-CoV (
27), requires two-step conformational changes mediated by sequential receptor binding and proteolysis to be activated for fusion (
20). Such a mechanism allows for tight temporal control over fusion by protecting the activating cleavage site from premature proteolysis yet allowing efficient cleavage upon binding to the receptor on target cells.
Previous studies have clearly demonstrated that the SARS-CoV S protein requires proteolytic cleavage for S-mediated cell-cell or virus-cell fusion (
3,
13,
21,
30,
31). Recently, the cleavage of S protein by the airway transmembrane protease, TMPRSS11a, was also reported (
15). Treatment with a purified soluble form of TMPRSS11a following receptor binding to the pseudotyped SARS-CoV S protein strongly enhanced viral infection. Moreover, proteolytic cleavage by TMPRSS11a was observed at the same position as cleavage by trypsin in the purified soluble form of S protein (
15). While the reported concentration of the previously mentioned proteases required to induce membrane fusion is extremely high (
3,
15,
21 ), the amount of TMPRSS2 expressed in Vero-TMPRSS2 cells is thought to be low (
29). Still, our results showed the formation of massive syncytia when S protein was expressed (Fig.
2), suggesting that TMPRSS2 may be more efficient than previously characterized proteases. The inhibition of cell-cell fusion in TMPRSS2-expressing cells in the presence of a protease inhibitor (Fig.
2) strongly suggests that TMPRSS2 proteolytically affects S protein. Because we could not detect a considerable amount of cleaved S protein in cells expressing both S and TMPRSS2, even when massive syncytia were observed in these cells, we hypothesized that proteolytic activation of S protein by TMPRSS2 occurs only during cell entry, following receptor binding. The results shown in Fig.
5 support this hypothesis and clearly show that the spatial orientation of TMPRSS2 in relation to S protein is a key mechanism underling this phenomenon: TMPRSS2 must be expressed in the opposing cell membrane to activate S protein and induce cell-cell fusion. We speculate that the encounter of receptor-bound S protein with TMPRSS2 at the right time and in the correct spatial orientation at the cell surface results in efficient cleavage of S protein and subsequent membrane fusion. Therefore, only a small amount of S protein needs to be cleaved to enable viral or cell-cell membrane fusion. These cleavage products would be difficult to detect by Western blot analysis.
Additionally, we have previously reported that HR2 peptide inhibits SARS-CoV entry into Vero E6 cells in the presence, but not the absence, of trypsin. Likewise, inhibition of SARS-CoV cell entry does not occur when the virus is allowed to use the normal endosomal, cathepsin L-dependent, entry pathway (
34). Here, we show (Fig.
3) that HR2 peptide efficiently inhibits viral entry into TMPRSS2-expressing cells. This result provides further evidence that TMPRSS2 is efficient for viral entry and is localized at the cell surface, exposing the HR2 peptide binding site before endocytosis can occur. Thus, even if HR2 peptide does not efficiently work in commonly used tissue cell cultures, it may be a suitable candidate antiviral for the inhibition of SARS-CoV infection of lung cells that express membrane-associated proteases.