key: cord-0788093-pcexw3nj
authors: Wu, Canrong; Zheng, Mengzhu; Yang, Yueying; Gu, Xiaoxia; Yang, Kaiyin; Li, Mingxue; Liu, Yang; Zhang, Qingzhe; Zhang, Peng; Wang, Yali; Wang, Qiqi; Xu, Yang; Zhou, Yirong; Zhang, Yonghui; Chen, Lixia; Li, Hua
title: Furin, a potential therapeutic target for COVID-19
date: 2020-10-05
journal: iScience
DOI: 10.1016/j.isci.2020.101642
sha: 7e31c56e8d0d07cfc6b0c5e263e4cdb860973186
doc_id: 788093
cord_uid: pcexw3nj

COVID-19 has broken out since the end of December 2019 and is still spreading rapidly, which has been listed as an international concerning public health emergency. We found the Spike protein of SARS-CoV-2 contains a furin cleavage site, which did not exist in any other betacoronavirus subtype B. Based on a series of analysis, we speculate that the presence of a redundant furin cut site in its Spike protein is responsible for SARS-CoV-2’s stronger infectious than other coronaviruses, which leads to higher membrane fusion efficiency. Subsequently, a library of 4,000 compounds including approved drugs and natural products were screened against furin through structure-based virtual screening and then assayed for their inhibitory effects on furin activity. Among them, an anti-parasitic drug, Diminazene, showed the highest inhibition effects on furin with an IC50 of 5.42 ± 0.11 μM, which might be used for the treatment of COVID-19.

with good effects and high specificity has been found so far.

The epidemiological observations showed the infectious capacity of SARS-CoV-2 is stronger than SARS-CoV, so there are likely to be other mechanisms to make the infection of SARS-CoV-2 easier. We suppose the main possibilities as follows, first, SARS-CoV-2 RBD combining with ACE2 may have other conformations; second, the SARS-CoV-2 Spike protein can also bind to other receptors besides ACE2; third, Spike is more easily cleaved by host proteases and easily fused with host cell membrane.

We compared the Spike proteins from three sources, SARS-CoV-2, SARS-CoV and Bat-CoVRaTG13, and found that the SARS-CoV-2 virus sequence had redundant PRRA sequences. Through a series of analyses, this study proposes that one of the important reasons for the high infectivity of SARS-CoV-2 is a redundant furin cleavage site in its Spike protein. And through structure-based virtual ligand screening and in vitro enzyme-based assay, the anti-parasitic drug Diminazene was found to show competitive inhibition on furin, with IC 50 of 5.42 ± 0.11 μM.

By sequence alignment of Spike protein sequence of SARS-CoV-2 with its highly homologous sequences, it was found that the Spike cleavage site of SARS-CoV-2 possessed 4 redundant amino acids-PRRA, and these were not found in those of high homology coronavirus, thus forming a furin-like restriction site as RRAR ( Figure S1 and S2). Through prediction in ProP 1.0 Server, it is true that the sequence was easily digested by furin ( Figure S3 ). In order to explore the evolution of this sequence, we used the BLASTp method to find 1,000 homologous Spike sequences with homology from 100% to 31%, which are all from β-CoVs. Multiple sequence alignments were performed on these thousands of Spike sequences. One sequence was selected from each highly homologous class (homology greater than 98.5%) for further sequence alignment, and about 155 sequences were finally selected. A homologous multiple sequence alignment was performed on these 155 sequences, and then a phylogenetic tree was constructed ( Figure 1 ). As shown in the phylogenetic tree, the J o u r n a l P r e -p r o o f Spike of SARS-CoV-2 exhibited the closest linkage to those of Bat-SL-CoV and SARS-CoV, and far from those of MERS-CoV, HCoV-HKU1 and HCoV-OC43. In general, most of the Spike proteins in α-CoVs do not contain a furin cleavage site, but it is very popular in γ-CoVs Spike protein, and with or without furin cleavage site are common in β-CoVs (Millet and Whittaker, 2015) . We systematically analyzed the four subtypes of β-CoVs and found that only SARS-CoV-2 in the subtype B β-CoVs contains the furin cleavage site, and most of the subtype A β-coronavirus contains the furin restriction site.

We performed furin digestion site prediction on the sequence of each type of coronavirus Spike through online software. It was found that all Spike with a SARS-CoV-2 Spike sequence homology greater than 40% did not possess a furin cleavage site ( Figure 1 , Table 1 ), including Bat-CoV RaTG13 and SARS-CoV (with sequence identity as 97.4% and 78.6%, respectively). The furin cleavage site "RRAR" of SARS-CoV-2 is unique in its family, rendering by its unique insert of "PRRA". The furin cleavage site of SARS-CoV-2 is unlikely to have evolved from MERS, HCoV-HKU1, and so on. From the currently available sequences in databases, it is difficult for us to find the source. Perhaps there are still many evolutionary intermediate sequences waiting to be discovered.

By analysis of the SARS-CoV-2 Spike protein sequence, we found that its most features are similar to SARS-CoV. It has an N-terminal signal peptide and is divided into two parts, S1 and S2. Among them, S1 contains N-terminal domain and receptor binding region. And S2 is mainly responsible for membrane fusion. The C-terminal region of S2 is S2', containing a fusion peptide, heptad repeat 1, heptad repeat 2, and a transmembrane domain (Figure 2A ). There are two cleavage sites between S1 and S2', named CS1 and CS2. However, there are some differences in this two cleavage sites.

Unlike SARS-CoV, SARS-CoV-2 contains polybasic amino acids (RRAR) at the CS1 digestion site, and trypsin digestion efficiency will be significantly improved here (Belouzard et al., 2009 ). More importantly, as mentioned above, this site can be recognized and cleaved by the furin enzyme. The cleavage of Spike protein promotes J o u r n a l P r e -p r o o f structural rearrangements of RBD for the adaptation to receptor, thus increasing the affinity (Walls et al., 2019) . More importantly, the digestion of Spike is indispensable for membrane fusion of S2 part (Kirchdoerfer et al., 2016) . In this case, the cleavage efficiency of the SARS-CoV-2 Spike protein cleavage is significantly higher than that of SARS-CoV, and the SARS-CoV-2 Spike protein could be cleaved during the process of biosynthesis, which has been verified by a recent research (Walls et al., 2020) ( Figure   2B ). The receptor affinity and membrane fusion efficiency of SARS-CoV-2 would be significantly enhanced when compared to that of SARS-CoV. The membrane fusion of SARS-CoV-2 Spike protein is more likely to occur on the host cell plasma membrane.

This may explain the strong infectious capacity of SARS-CoV-2. So, the development of furin inhibitors may be a promising approach to block its transmissibility.

In our previous studies (Wu et al., 2020) , both SARS and SARS-CoV-2 Spike RBD structures have been docked with human ACE2 to calculate their binding free energy.

In that time, the complex structure of SARS-CoV-2 RBD with ACE2 was not available.

Its energy was calculated based on the homology model generated from SARS_RBD-ACE2 complex. The binding free energy between the SARS-CoV-2 Spike RBD and human ACE2 was -33.72 KCal mol -1 , and that between SARS-CoV Spike RBD and ACE2 was -49.22 KCal mol -1 . This means the binding affinity between SARS-CoV-2 Spike and ACE2 is weaker than that of SARS Spike. During this manuscript was prepared, the structure of SARS-CoV-2 Spike RBD-ACE2 complex was disclosed . Based on this new real structure of SARS-CoV-2 Spike RBD-ACE2 complex, we re-did the calculation and found that the binding free energy between SARS-CoV-2 Spike RBD and ACE2 was -50.13 KCal mol -1 . ( Figure S4 ). This means the binding affinity between SARS-CoV-2 Spike and ACE2 is slightly stronger than that of SARS Spike. By inspecting the crystal structure of SARS-CoV-2 RBD-ACE2 complex and SARS RBD-ACE2 complex, one can find that one key loop of SARS-CoV-2 RBD in the complex interface had very different conformation compared to that of SARS RBD and previous modeled SARS-CoV-2 RBD ( Figure S5 ).

In order to further explore the possible mechanism of furin cleaving SARS-CoV-2 J o u r n a l P r e -p r o o f Spike, we perform protein-protein docking for furin and Spike. We already built a homology model of SARS-CoV-2 Spike in our recently published paper (Wu et al., 2020) . SARS-CoV-2 Spike structure was built by using the SARS-CoV Spike structure as the temple (PDB code: 5X58) (Yuan et al., 2017) . In order to verify the accuracy of homologous modeling, we aligned the computational structure of the SARS-CoV-2 Spike that modeled from the SARS-CoV spike with its Cryo-EM structure (6VXX) just solved and released during this manuscript was being revised and submitted (Walls et al., 2020) . The computational model of the SARS-CoV-2 Spike showed a Cα RMSD of 1.571 Å on the overall structure compared to the SARS-CoV-2 Spike Cyro-EM structure, demonstrating a very subtle difference.

By superimposing the SARS-CoV Spike with the SARS-CoV-2 Spike, we can find that the major conformation differences between two structures are RBD domain, Arg685/677 loop region (furin/trypsin/TMPRSS2 cut site) and S2 loop region just after fusion peptide ( Figure 3A ). The trypsin/TMPRSS2 cut sites of both SARS-CoV and SARS-CoV-2 Spikes were disordered and missing from the original Cryo-EM structures possibly due to its flexibility and without electro density, we built this region by modeling software. The "PRRA" inserting in this region of SARS-CoV-2 apparently generates the more flexible loop region and accessible cut site for protease. We performed protein-protein docking by setting SARS-CoV-2 Spike furin cleavage loop as the receptor, and furin active pocket as the ligand. The protein-protein docking results showed that furin acidic/negative active pocket can be well fitted onto the SARS-CoV-2 Spike basic/positive S1/S2 protease cleavage loop with low energy . This implies that the extra "PRRAR" loop of SARS-CoV-2 Spike renders it more fragile to the protease. And this may allow this site to be cut during the maturation, efficiently enhancing the infection efficiency.

By the online ProP 1.0 Server software prediction, we proposed that furin can cleave the RRAR sequence, but the cleavage efficiency of this sequence has not been determined. To measure the cleavage efficiency, we performed a set of experiments J o u r n a l P r e -p r o o f to determine the Michaelis-Menten constant (Km), limiting rate (Vmax) of furin using fluorogenic peptide substrates, Boc-RRAR-AMC (Table 2, Figure 7A ). Kinetic parameters were also determined by using the second AMC substrate containing the authentic cleavage motif Arg-Val-Arg-Arg (Table 2, Figure 7A ). Boc-RRAR-AMC has a similar Km value toward furin compared to Boc-RVRR-AMC. Using this substrate, a 2-fold higher hydrolysis efficiency was observed as demonstrated by the limiting rate, indicating that the "RRAR" motif is easily recognized by furin and quickly hydrolyzed .

Structure-based virtual ligand screening method was used to screen potential furin protein inhibitors through ICM 3.7.3 modeling software (MolSoft LLC, San Diego, CA) from a ZINC Drug Database (2924 compounds), a small in-house database of natural products (including reported common antiviral components from traditional Chinese medicine) and derivatives (1066 compounds), and an antiviral compounds library contains 78 known antiviral drugs and reported antiviral compounds.

Compounds with lower calculated binding energies (being expressed with scores and mfscores) are considered to have higher binding affinities with the target protein.

The screening results for the ZINC Drug Database (Table S1) Here, we show one example of screen hits, Diminazene, which was predicted to bind in the active site of furin with low binding free energy. In the generated docking model, Diminazene was well fitted into the binding pocket of the substrate and adopted similar conformation as substrate analogous inhibitor MI-52 in PDB model 5JXH (Dahms et al., 2016) , and occupied two arms' position of MI-52 ( Figure 4A ).

Asp154, Asp258 and Asp306 were predicted to form three hydrogen bonds with imine groups of compounds ( Figure 4B ). It looks like that Diminazene mimic at least J o u r n a l P r e -p r o o f two arginines. Weak hydrophobic interaction of His194, Leu227, the backbone of Trp254 and Asn295 with the compound may further stabilize its conformation.

Another example was anticancer drug Imatinib. It was also predicted to bind in the active site of furin. In the generated docking model, Imatinib was fitted well in the binding pocket, and occupied the top two arms' position of MI-52 ( Figure 5A ). Two hydrogen bonds were predicted to form between the compound with Glu236 and Gly255. Weak hydrophobic interaction between Val231, Pro256, Trp254 and Gly294 and the compound was found ( Figure 5B ).

For the natural products (Table S2) (Table S4) .

Among them, the antiparasitic drug Diminazene showed an inhibition ratio over 95%, and other compounds like Aminopterin, Methotrexate and Silybin showed inhibition ratio over 70%. Among them, Diminazene showed significantly dose-dependent inhibition on furin protease, with IC 50 value of 5.42 ± 0.11 μM ( Figure 7B ). Enzyme kinetics has also been measured to explore its mode of inhibition. As shown in Figure   7C and 7D, Diminazene displayed a competitive inhibition mechanism characterized by dose-dependent increase in Km and little effects on Vmax (Table 3 ).

Our previous study (Wu et al., 2020) analyzed the amino acid composition of the RBD domain of the ACE2 receptor of SARS-CoV-2 and Bat-CoVRaTG13. We found that several key amino acids determining binding were mutated in SARS-CoV-2, which are more similar to that of SARS-CoV. The calculation results showed that in the same conformation as the SARS-CoV protein, the binding free energy of SARS-CoV-2 and ACE2 receptors was a little higher, but this result cannot fully explain the epidemiologically high contagion, so we speculate (1) other mechanisms that enhance infectivity. During this manuscript was prepared, the Cryo-EM structure of SARS-CoV-2 Spike was solved (Walls et al., 2020) . Comparing the structure of SARS-CoV-2 with the Spike structure of SARS-CoV, combined with biophysical detection, they found that SARS-CoV-2 binds more strongly to cellular ACE2 receptors (Walls et al., 2020) . Furthermore, the just disclosed crystal structure of SARS-CoV-2 RBD-ACE2 complex showed a distinct conformational change in the key loop of complex binding interface. And the binding free energy calculation indicated a slightly stronger binding for SARS-CoV-2 RBD compared to that for SARS RBD. These results confirm our supposition that the conformational change of the RBD domain of SARS-CoV-2 leads to stronger binding. However, stronger receptor binding still can't fully explain the more infectious problem of SARS-CoV-2.

So we put forward these hypotheses: (1) Boc-RRAR-AMC can be effectively recognized and cleaved by furin, and its hydrolysis efficiency was higher than Boc-RVRR-AMC, a known substrate of furin.

We systematically analyzed the four subtypes of β-CoV and found that SARS-CoV-2 was the only one in the subtype B β-CoV which contains the furin cleavage site, while most of the subtype A contains the furin restriction site (Figure 1 ). We aligned 1000

Spike sequences and found that all Spikes with sequence homology greater than 40% of SARS-CoV-2 Spike did not have a furin cleavage site, but its possible evolutionary J o u r n a l P r e -p r o o f source cannot be found currently, and more novel viruses are needed to be discovered. Very recently, a new close related virus RmYN02 containing PAA at the CS1 of the S protein has been reported, but it also had no furin cleavage sites (Zhou et al., 2020) .

The "PRRA" insert and subsequent arginine (R) constitute a RRAR sequence that could be recognized and cleaved by furin-like proteases, which may be the reason why SARS-CoV-2 infection is stronger than SARS-CoV. What's more, we performed a homologous alignment and phylogenetic analysis of the SARS-CoV-2 sequence, and found that "PRRA" insert did not appear at any other close relatives of SARS-CoV-2,

indicating that this insert was completely novel in this genus virus. The existence of such a motif may allow Spikes to be cut into S1 and S2 by furin-like proteases before maturity, which provides S1 with the flexibility to change the conformation to better fit the host receptor. According to studies of Simmons G et al., overexpression of furin can increase the activity of SARS-CoV Spike, but it will not cause Spike to be cleaved (Simmons et al., 2011) . This is consistent with our prediction. (Simmons et al., 2005) . In addition, SARS-CoV Spike can be activated by TMPRSS2 cleavage on the host cell surface (Glowacka et al., 2011) . As we can see in Figure 2B , the Spike protein of SARS-CoV-2

can be cleaved at multiple stages, which greatly increases the efficiency of fusion.

Markus H et al. demonstrated that the CS1 of the S protein of SARS-CoV-2 was easily cleaved in the host cell, and they mentioned that TMPRSS2 play an important role in the cleavage of CS2 (Hoffmann et al., 2020) . It is likely that the virus will fuse with the cell in the cell plasma membrane and release the genome into cells. In addition, the receptor affinity of the cleaved Spike is also greatly enhanced (Parka et al., 2016) .

According to our study, furin-like proteases may be potential drug targets for anti-SARS-CoV-2 treatment. At present, some peptide inhibitors have been Materials and the information used for the experiments are available upon reasonable request.

All data used in the study are included in this publication. The present research did not use any new codes

All methods can be found in the accompanying Transparent Methods supplemental file. 

Bat-to-human: Spike features determining 'host jump' of coronaviruses SARS-CoV, MERS-CoV, and beyond

Middle east respiratory syndrome coronavirus Spike protein is not activated directly by cellular furin during viral entry into target cells

Host cell proteases: Critical determinants of coronavirus tropism and pathogenesis

A general and fast scoring function for protein ligand interactions: a simplified potential approach

Proteolytic processing of middle east respiratory syndrome coronavirus Spikes expands virus tropism

Different host cell proteases activate the SARS-coronavirus Spike-protein for cell-cell and virus-cell fusion

Inhibitors of cathepsin L prevent severe acute respiratory syndrome coronavirus entry

Characterization of cellular furin content as a potential factor determining the susceptibility of cultured human and animal cells to coronavirus infectious bronchitis virus infection

Structure, function and antigenicity of the SARS-CoV-2 spike glycoprotein

Structural Analysis of Rabies Virus Glycoprotein Reveals pH-Dependent Conformational Changes and Interactions with a Neutralizing Antibody

Proteolytic activation of the Spike protein at a novel RRRR/S motif is implicated in furin-dependent entry, syncytium formation, and infectivity of coronavirus infectious bronchitis virus in cultured cells

Cryo-EM structures of MERS-CoV and SARS-CoV Spike glycoproteins reveal the dynamic receptor binding domains

A novel bat coronavirus reveals natural insertions at the S1/S2

cleavage site of the Spike protein and a possible recombinant origin of HCoV-19. bioRxiv

Protease inhibitors targeting coronavirus and filovirus entry

The unique "RRAR" motif in SARS-CoV-2-S is easily recognized and hydrolyzed by furin

A reason for high infectivity of SARS-CoV-2 is the furin cleavage site in Spike

We are very grateful to the Protein Data Bank, DTU Health Tech, ZINC database and researchers for making data freely available. We acknowledge support from NationalMega-project for Innovative Drugs (grant number