The outer surface of coronaviruses contains a critical transmembrane spike glycoprotein that is essential for entry of viral particles into host cells. This viral glycoprotein possesses a trimeric structure, which gives the virus its typical crown-like halo (Figure 1
A). This outer protein contains domains and structural motifs that are essential for binding to host cells and for viral fusion. Two major subunits (S1 and S2) need to be processed by host cell proteases to enable conformational rearrangement of the C-terminal domain and exposure of the epitopes that allow the virus to enter and subsequently egress from host cells (Figure 1
]. Hence, recent studies suggested that impairing the spike glycoprotein processing represents a viable therapeutic strategy [3
]. There are three proteolytic cleavage sites (S1, S2, and S2′; Figure 1
B) in the spike glycoprotein. The sequence of these sites can determine whether the virus can cross species, for example from bats or camels to humans [5
]. The cleavage site (S2) of sequence ATY↓MS (the arrow indicates the cleavage site) is likely cleaved by cathepsin L (Figure 1
]. Because this site is conserved among coronaviruses, its cleavage cannot explain differences in pathogenicity among them [3
On the contrary, and unlike less virulent coronavirus strains, the SARS-CoV2 glycoprotein presents the S1 cleavage of sequence SPR
; consensus residues are depicted in bold characters), which represents a consensus furin recognition motif [3
]. Furin and related proprotein convertases (PC2, PC1/3, PC4, PACE4, PC5/6, and PC7) are specialized serine endoproteases, which cleave R-X-(R/K/X)-R↓(S)(V/A/L) multibasic motifs [9
]. The highly pathogenic MERS-CoV coronavirus also contains a putative furin cleavage S1 site [2
] (Table 1
). On the contrary, less pathogenic strains such as the SARS coronavirus (SARS-CoV) and the bat coronavirus strains (Bat-RaTG13, Bat-ZXC21, or Bat-ZC45) possess the S1 sequence S(L/I)LR
T, which cannot be readily cleaved by furin. For these sites, the membrane trypsin-like serine protease, TMPRSS2, has been identified as a possible major priming protease [8
]. This observation suggests that furin may be essential for the viral entry and/or egress in highly pathogenic strains [2
However, earlier studies indicated that furin was dispensable for the MERS-CoV entry, while TMPSSR2 was necessary and sufficient for viral entry [13
]. A more recent study with SARS-CoV2 corroborated these findings [4
]. In particular, pharmacological inhibition of TMPRSS2 by camostat mesylate, a covalent inhibitor, attenuated the entry of SARS-CoV2 surrogate viral particles into human cells, albeit only partially and at relevantly high (10–50 μM) concentrations [4
]. Here, we closely analyzed the coronavirus cleavage sequences to delineate possible parameters that may confer increased virulence and pathogenicity to SARS-CoV2. In contrast to other coronaviruses, the S1 SARS-CoV2 site presents the peculiar property of being a substrate for both furin and TMPRSS2. Furin is involved in numerous pathogenic processes including not only viral propagation but also bacterial toxin activation. In anthrax toxin, for example, furin plays an essential role in the cleavage of anthrax protective antigen (PA), representing a necessary step for the entry of anthrax toxin into macrophages. Anthrax toxin PA protein (83 kDa) is secreted by the bacterium, binds to anthrax toxin receptor (ATR) on host cell membranes and is then cleaved by cell-surface furin to generate a cell-associated 63 kDa PA and a free 20 kDa PA [14
]. The cell-associated PA molecule heptamerizes, forming a membrane channel that allows entry of lethal factor (LF) toxin into the host-cell cytoplasm, resulting in shock and eventual death. If furin is absent or inactive, the toxin fails to assemble and therefore is not lethal. Interestingly, the S1 SARS-CoV2 site is homologous to the processing site of the anthrax toxin PA protein, which can also be processed by TMPRSS2-like proteases and furin (Table 1
). Moreover, similar to LF/PA anthrax toxin, SARS-CoV infects macrophages, as well as the airway epithelium [26
]. Hence, we monitored the ability of compound 1
to protect RAW macrophages from LF/PA-induced cell death. The agent protected RAW macrophages in a dose-dependent manner with EC50
in the low micromolar range indicating a robust cellular inhibition of various possible PA-activating serine proteases, including furin and, potentially, also TMPRSS2. Finally, using an in vivo anthrax toxemia model, we probed whether systemic administration of serine protease inhibitors is potentially a viable therapeutic strategy.
Furin and related PCs (PC2, PC1/3, PC4, PACE4, PC5/6, and PC7) are specialized serine endoproteases that cleave the multibasic motifs R
]. In addition to its normal cellular functions, furin is also implicated in many pathogenic states. Thus, furin cleaves to maturity membrane fusion proteins of viruses and pro-toxins of a variety of bacteria, including anthrax and botulinum toxins, influenza, measles, flaviviruses and many others [9
]. Acquisition of furin-like priming sequences correlates with increased virulence and pathogenicity. For example, the acquisition of a furin cleavage site in the priming site of the viral protein hemagglutinin (HA), necessary for influenza virus entry, is associated with the increased pathogenicity of the avian influenza viruses [33
]. Perhaps more interestingly, such evolution of the influenza virus to contain furin-like sequences can be induced by repeated passages in cell culture or through animals [34
Furin-like sequences that may contribute to increased virulence have also been identified in the coronavirus spike glycoproteins (Table 1
]. The complex mechanism of viral fusion in coronaviruses is not fully understood, but it likely comprises a first cleavage of the S1 site that allows the S2 subunit to more easily dissociate from the S1 subunit (Figure 1
). The S2 subunit contains a fusion peptide, an internal fusion peptide, two heptad-repeat domains, and a transmembrane domain (Figure 1
). The spike protein S1 attaches the virion to the cell membrane by interacting with its host receptor, thus initiating the infection. This occurs most likely by binding to the ACE2 receptor causing internalization of the virus into the endosomes of the host cell. Proteolysis by serine proteases of the S1 site (or by cathepsin L, in the adjacent S2 cleavage site) [8
] may unmask the fusion peptide and activate membrane fusion within the endosomes. This step seems to require an additional cleavage at the S2′ site to unmask the internal fusion peptide in the S2 viral fusion protein.
However, while the S2 site is conserved among various coronavirus strains, the S1 site in SARS-CoV2 contains a furin cleavage site (Table 1
). The S1 sequence is located in an exposed unstructured loop in the structure of the SARS-CoV2 spike protein (Figure 1
A,B). As a result, no electron density was observed in this loop region in the recently reported Cryo-EM structure [1
]. Most intriguingly, the unusual SARS-CoV2 S1 site may have also acquired an increased cleavage propensity for TMPRSS2 (Table 1
) (Supplementary Materials
). This may explain why TMPRSS2 appeared more important than furin for the entry of the surrogate SARS-CoV2 viral particles in cell [4
]. However, while TMPRSS2 is abundant in the respiratory tract, furin is more ubiquitously found in many other organs (https://www.proteinatlas.org/ENSG00000140564-FURIN/tissue
); hence, acquisition of a furin cleavage site most likely increases the tropism and the pathogenicity of the strain. Furthermore, because furin is localized in the trans-Golgi network and cycles between the trans-Golgi and the cell surface, furin cleavage in the spike protein may occur also during viral egress from the infected cells. As a result, pre-primed viral particles may be more ready to enter and infect other cell types and/or to spread among hosts.
The common mechanisms of cell trafficking mediated by furin cleavage by both viral fusion proteins and bacterial toxins is striking. For example, anthrax toxin, similar to SARS-CoV2, requires processing of the PA sequence RKKRST (Table 1
) to chaperone the internalization of the LF toxin into macrophages. Intriguingly, much like the S1 of SARS-CoV2, the PA cleavage site also contains both furin-like and TMPRSS2-like proteases recognition sites (Table 1
), and it also invades macrophages, making a potentially good model system to study inhibition of priming in vivo. Intact toxins, like viral proteins, are incapable of accomplishing these processes in absence of proper priming by the host proteases. Hence, while it must be emphasized that the detailed molecular mechanisms of the spike glycoprotein-mediated viral fusion [8
] and of the PA-mediated LF entry are fairly different [14
]; both end processes depend on the activity of these priming enzymes. Nonetheless, further viral replication experiments with live SARS-CoV2 will be required for a full understanding of the potential of this approach. In recent years, several reports emerged describing improved furin inhibitors [36
]. Here, we evaluate the potential of systemic administration of a furin protease inhibitor to prevent priming, using the serine protease dependent anthrax toxin as a model system.
When tested in cell, the pan-active compound 1
was efficacious in protecting RAW macrophages from anthrax toxin (Figure 2
), suggesting that the prototype agent possesses favorable pharmacological properties for in vivo studies. Because the toxicity in vivo of LF/PA toxin intimately depends on PA cleavage by serine proteases [31
], this model is ideal for evaluating the inhibition of priming in vivo. Hence, Balb/C mice receiving a mixture of LF and PA (100 µg via the tail vein) were injected with either a single dose of 3 mg/kg (I.P.) of compound 1
, or two doses spaced by 2 h, or with vehicle control. According to our pharmacokinetics studies, these doses should reach blood levels of the drug sufficient to inhibit furin effectively (Figure 2
B). Based on previous studies [31
], mice treated with such a lethal dose of toxin die by roughly 48 h post treatment depending on the LF and PA lots and the age and strain of mice and their weight. A potent direct LF inhibitor given at 30 mg/kg I.P. 3 times a day was reported to prevent death of mice at 48 h, while no survivors were present in the control group [41
]. However, no information was provided in the literature on the fate of treated mice after 48 h (no time to death was reported). In our experiments, all mice in the untreated group died by time = 33 h, in close agreement with the published studies, while a remarkable and significant increase of both median survival time (MST) and time to death (TTD) was observed in both groups treated with compound 1
B) even at the single dose of 3 mg/kg.
These data clearly suggest that at least for anthrax toxin and likely for other pathogens including SARS-CoV2, furin-targeting pan protease inhibitors could be used as antiviral agents or be deployed prophylactically in emergency medicine in case of pandemic outbreaks in patients that are suspected or at risk of viral infection. Systemic administration of the inhibitors in the treated group was tolerated by mice at 3 mg/kg doses, but a maximum tolerated dose of about 10–15 mg/kg was observed in separate toxicity studies, suggesting that more targeted delivery strategies may improve the observed therapeutic window. Similar to other antiviral drugs such as Zanamivir, this could be perhaps simply accomplished by devising proper inhalable formulations, which should be facilitated by the high aqueous solubility of the agents (>1 mM).
Recently, camostat mesylate, a covalent TMPRSS2 inhibitor already clinically approved for other indications in Japan (Figure 2
D), has been proposed given that it partially blocked viral entry in surrogate cellular assays [4
]. However, we found that camostat mesylate did not appreciably inhibit furin (Figure 2
E), and while it may attenuate entry at relatively high concentrations (10–50 µM) [4
], in our opinion it would do little to prevent furin-mediated egress of partially primed (at S1 site), hence more virulent, SARS-CoV2 viral particles.
Hence, while we await for the results of the efficacy of camostat in a very recently initiated human clinical trial with COVID19 patients (https://clinicaltrials.gov/ct2/show/NCT04321096
), this report wishes to incentivize once again private and public efforts to consider developing new pan-serine protease inhibitors, perhaps taking advantage of several agents already being reported in pre-clinical studies [36
], into emergency therapeutics to combat the new coronavirus SARS-CoV2 and to ward off future similar pandemics that are likely to occur when pathogens acquire the further optimized furin cleavage sites within their priming entry mechanisms.
These development efforts are particularly significant especially for coronaviruses as no viable treatments or vaccines are currently available, and at the same time other future furin-like cross-species transmission in coronaviruses seems likely. Mutations of the cleavage site in either S1 or S2′ of coronavirus strains’ spike glycoprotein can be correlated with pathogenicity, increased tropism, and crossing zoonotic barriers. Unfortunately, one could envision several mutations in SARS-CoV2 (or any other coronavirus strain) that could transform these sequences into more efficient furin and/or dual furin and TMPRSS2 cleavable sites, hence increasing their pathogenicity, virulence, and potential for spread.