Novel sites for Cathepsin L cleavage in SARS-CoV-2 spike guide treatment strategies

The spike (S) protein of SARS coronavirus 2 (SARS-CoV-2) is an ideal target for the development of specic vaccines or drugs. However, treatments targeting viruses with mutant S proteins that have recently emerged in many countries are limited. Cleavage of the S protein by host proteases is essential for viral infection. Here, we discovered two novel sites (CS-1 and CS-2) in the S protein for cleavage by the protease Cathepsin L (CTSL). Both sites are highly conserved among all SARS-CoV-2 variants of concern. Cryo-electron microscopy structural studies revealed that CTSL cleavage increases the dynamics of the receptor binding domain of S and induces novel conformations. In our pseudovirus (PsV) infection experiment, alteration of the cleavage site signicantly reduced the infection eciency, and CTSL inhibitors markedly inhibited infection with PsVs of both the wild-type and emerged SARS-CoV-2 variants. Furthermore, six highly ecient CTSL inhibitors were found to effectively inhibit live virus infection in human cells in vitro, and two of these were further conrmed to prevent live virus infection in human ACE2 transgenic mice in vivo. Our work suggested that the CTSL cleavage sites in SARS-CoV-2 S are emerging new but effective targets for the development of mutation-resistant vaccines and drugs. (c) Luciferase gene expression by ERE and ESR1 bound and activated the ERE promoter to upregulate luciferase expression. (d and e) Effector Huh7 cells were cotransfected with plasmids expressing ERE-luciferase and different S proteins as indicated, and another plate of target Huh7 cells was transfected with plasmid expressing ESR1. After 24 h, the effector cells were detached and added to the target cells for 30-60 min. Then, the supernatant was removed and treated with PBS or CTSL (8 (cid:0) g/mL) for 20 min. The reaction was stopped by adding 500 (cid:0) L of medium, and culture was continued for another 24 h to allow cell-cell fusion. When a target cell and effector cell fused to form a syncytium, ESR1 bound and activated the ERE promoter to upregulate luciferase expression. Luciferase activity was then measured as a proxy for the fusion rate. The data were normalized to the WT-PBS group (n=3). Statistical signicance was assessed between the indicated group and the WT-PBS group by two-way ANOVA with Dunnett’s post hoc test. f, Images of syncytium formation induced by CS mutant SARS-CoV-2 S proteins. Huh7 cells were transfected with plasmids to express the WT, CS-1M, CS-2M or CS-1M+2M S protein. Cells were treated in the absence (PBS, pH=5.8) or presence of CTSL (4 (cid:0) g/mL, pH=5.8). Images were acquired after an additional 10-16 h of incubation in medium (scale bars, 50 μm). The black arrowheads indicate syncytia. Representative data from three independent experiments are shown. g, Overexpression or knockdown of the CTSL gene dose-dependently promoted or inhibited, respectively, infection with WT (Wuhan-1) and three mutant SARS-CoV-2 furin FM-ARAA, a luciferase assay as the relative activity (n=3). signicance by promoted syncytium formation induced cells signicance by one-way with post cells with scramble or S


Introduction
The coronavirus disease 2019 (COVID-19) pandemic, caused by the novel SARS coronavirus 2 (SARS-CoV-2), has resulted in a global public health crisis. Most treatment strategies for SARS-CoV-2 infection are focused on vaccines or antiviral drugs targeting the viral spike (S) protein 1,2 . However, the evolution of RNA viruses is driven by their high mutation rates, allowing their evasion of host immune attacks 3,4 .
Despite multiple ongoing efforts to treat SARS-CoV-2 infections, a new wave of infections caused by S mutant strains has emerged in many countries and has exhibited high morbidity [3][4][5][6] . Therefore, the development of antiviral drugs or vaccines targeting the S protein may encounter the problem of resistant mutations. Indeed, three vaccines (Janssen, Novavax, and AstraZeneca) tested against the B.1.351 variant exhibited reduced effects on preventing symptomatic infection, with effectiveness estimates of 57%, 49% and even a statistically insigni cant percentage, respectively 7 . The serum neutralizing activity against the B.1.351 variant among vaccinated persons was reduced by 6.5-and 8.6-fold for the BNT162b2 (P zer) and mRNA-1273 (Moderna) mRNA vaccines, respectively 8 . Thus, there is an urgent need to develop broad-spectrum treatment strategies for SARS-CoV-2 infection.
During viral infection, the S protein requires activation by host cell proteases, such as furin, TMPRSS2 and cathepsin L (CTSL) [9][10][11] . CTSL is a member of the lysosomal cysteine protease family, whose major function is proteolysis of antigens produced by pathogens. Recently, we found that in SARS-CoV-2infected patients, the circulating level of CTSL in blood samples is highly correlated with the severity and course of COVID-19. We also found that CTSL promotes SARS-CoV-2 pseudovirus (PsV) infection by cleaving the S protein and enhancing viral entry into cells 9 . However, the precise cleavage site and mechanism by which CTSL activates the S protein, as well as the effectiveness of CTSL inhibitors in preventing or treating infections with live SARS-CoV-2, remain unknown. In this study, we identi ed two of the S protein are mutated in these variants. However, the CTSL cleavage sites are highly resistant to viral mutational escape (Fig. 1f), suggesting that these CTSL cleavage sites may be essential for the SARS-CoV-2 life cycle.
It has been reported that SARS-CoV-2 may originate from bat coronaviruses 16 , such as RaTG13; therefore, we performed sequence alignment of residues around these sites in coronaviruses found in bats and pangolins. As shown in Supplementary Fig. 2a, CS-2 is highly conserved among most SARS-CoV-and SARS-CoV-2-related CoVs, while CS-1 is conserved only among SARS-CoV-2-related CoVs. This result suggests that the two CTSL cleavage sites may have different roles during evolution in various animal reservoirs. CS-2 seems to be the most important for the CoV life cycle, while CS-1 may have gradually evolved from SARS-CoV-1 to SARS-CoV-2 to improve viral infection. Moreover, we compared the conservation of the two CTSL cleavage sites among seven known human CoVs and found that these CTSL cleavage sites exist only in SARS-CoV-1 and SARS-CoV-2 ( Supplementary Fig. 2b). This nding was consistent with previous studies showing that CTSL is involved in SARS-CoV-1 infection 17 and suggested that CTSL plays a unique role during the process of both SARS-CoV-1 and SARS-CoV-2 infection. CTSL cleavage activates SARS-CoV-2 S with increased dynamics To investigate how CTSL cleavage affects the structure and conformation of SARS-CoV-2 S, we performed cryo-EM studies on both the CTSL-treated and untreated S protein. We found three structural populations of S in the untreated control group (Supplementary Fig. 3) and six populations in the treated group ( Supplementary Fig. 4). The resolutions of these structural populations ranged from 4.6 Å to 3.1 Å according to the gold standard Fourier shell correlation coe cient at 0.143 (Supplementary Figs. 3 and 4). The three structural populations in the untreated control group (Fig. 2a) included one inactive state (closed) and two active states (1R-weak and 1R-open), which have been discovered and reported before 12,18 . In the closed inactive state, the RBD is buried and not accessible for ACE2 binding, while in the activated states, one RBD becomes upright and exposed (1R-open) or becomes dynamic (1R-weak) and ready for ACE2 binding 19 . In the CTSL-treated group (Fig. 2a), in addition to three populations (closed, 1R-weak and 1R-open) similar to those found in the control group ( Supplementary Fig. 5a-c), we found three novel populations: the 1N-attached state, with presumable CTSL binding (Fig. 2b); the active state 2R-weak, with two RBDs missing ( Supplementary Fig. 5d); and the ready-for-fusion state 1N-2R-weak, with two RBDs and one NTD missing (Fig. 2c-e). All six populations exhibit shapes generally similar to that of the prefusion S protein. We were aware of that the furin-treated S also kept similar shape with that of the prefusion S 18 . Notably, in cryo-EM, the structure maps of the closed, 1N-attached, 1R-open and 1N-2R-weak states exhibited elongation effects resulting from the preferred orientations of raw particles ( Supplementary Fig. 6); however, the functional domains of the S1 subunit, such as the NTD, CTD and RBD, could be clearly distinguished and assigned ( Fig. 2 and Supplementary Fig. 6). In addition, we observed no signi cant differences in the density maps at the cleavage site regions between the untreated and treated groups ( Supplementary Fig. 7), similar to the ndings for the furin cleavage site, as previously reported 18 .
Next, we sought to analyze the proportions of these populations before and after CTSL treatment. Without CTSL treatment, the proportion of the inactive-state S protein population was approximately 43% in our study, approximately the same as that in previous in vivo and in vitro results (Supplementary Table   1) 14,20 . Notably, a previous study reported that in the presence of the ACE2 receptor, the proportion of the inactive state population signi cantly decreased to 11% 19 . In our study, the proportion of the inactive state population (closed) was also signi cantly reduced from 43-11% after cleavage by CTSL ( Fig. 2a and f). In addition, with CTSL treatment, the proportions of the 1N-attached, 1R-open, 1R-weak, 2R-weak and 1N-2R-weak populations were 11%, 17%, 24%, 16% and 21%, respectively (Fig. 2a). The proportion of the 1R-weak population remained the same as that in the untreated group, while the proportion of the 1Ropen population was reduced from 34-17% (Fig. 2a). In addition, by categorizing the 1R-open, 1R-weak, 2R-weak and 1N-2R-weak populations as the active state populations, the overall percentage of activation was 78% in the CTSL-treated group but 58% in the untreated group ( Fig. 2a and f).
The unique population of 1N-attached emerged in the 3D classi cation step using the closed state as a reference and exhibited a shape obviously different from a regular threefold symmetry ( Supplementary  Fig. 4). The extra density surrounds CS-1 (R246-A262 loop region) in one NTD, which can be suitably tted with the crystal structure of CTSL (PDB entry 3OF8) ( Fig. 2b and Supplementary Fig. 5f). Thus, this enzyme-substrate complex population could represent an intermediate state during CTSL cleavage and be ready for activation.
Regarding the novel population 1N-2R-weak, with many domains that exhibited signi cant changes and offsets compared with the structure in the untreated group ( Fig. 2c-e), we postulated that this state is not only an activated state but also a new conformation with the S2 subunit exposed and ready for fusion with the host cell membrane (Fig. 2f). In the structure of this population, a major part of the RBD and the top half of the NTD in one protomer of S (promoter 1) is disordered, while half of the RBD in another protomer (promoter 2) is disordered ( Fig. 2c and d). In addition to domain disordering, we also observed signi cant domain movement in this population compared to the inactive state population ( Fig. 2e and Supplementary Fig. 5e). In protomer 1, the remaining part of the RBD moved 8 Å away, while the CTD and the remaining parts of the NTD and CTD moved 7 Å and 15 Å away, respectively. In promoter 2, the remaining part of the RBD moved 14 Å away from its location in the closed con guration. In the third protomer (protomer 3), the intact RBD moved 5 Å away from its original position in the inactive state. These observations imply that this novel population has greatly decreased stability compared with that of the S protein in its prefusion state and is thus ready for fusion with the host membrane (Fig. 2f).
Overall, these results suggested that CTSL cleavage induces signi cantly increased dynamics of the RBDs and NTDs with an increased proportion of activated S, which is accessible for ACE2 binding and then becomes ready for membrane fusion, resulting in improved potential for infection (Supplementary Movie 1).

CTSL cleavage sites are essential for SARS-CoV-2 infection and cell fusion
To verify the observations from our structural studies, we investigated the functions of CTSL in the viral infection process. We generated several mutants of PsV according to the two cleavage sites, including a CS-1 mutant (CS-1M), CS-2 mutant (CS-2M), and combination mutant (CS-1M + 2M) (Fig. 3a). As the proper control, we generated three PsV mutants proven to escape furin cleavage ( Supplementary Fig. 8a): furin mutant (FM)-delta (deletion of residues 682-685), FM-ARAA ( 682 ARAA 685 ) and FM-GSAS ( 682 GSAS 685 ). Then, we compared the infectivity of the wild-type (WT) and mutant PsVs in four human and monkey cell lines, which we have previously validated to have the highest susceptibility to SARS-CoV-2 infection 3,21 . Mutations in the furin cleavage site did not reduce the infectivity of either the PsV or live SARS-CoV-2 22,23 , suggesting that furin cleavage may not be essential for viral infection. However, once the CTSL cleavage sites were mutated to glycine, these PsVs lost most of their infection ability in all four cell lines. When both CTSL cleavage sites were mutated, the viral infectivity was reduced 100 ~ 700-fold compared with that of WT PsV (Fig. 3b), showing that the two novel CTSL cleavage sites CS-1 and CS-2 are essential for SARS-CoV-2 infection. Moreover, CS-2 seems to play a more important role than CS-1 in SARS-CoV-2 entry. Similar results were observed with different PsV concentrations ( Supplementary Fig.  8b).
The fusogenic activity of SARS-CoV-2 S is a characteristic of SARS-CoV-2 infection, and the presence of syncytia in COVID-19 patient lung tissue indicates that this activity plays a role in the pathological process 24 . Syncytium formation was quanti ed with an ESR1-ERE transactivation system (Fig. 3c-e and Supplementary Fig. 9) and observed by bright eld microscopy (Fig. 3f). Compared with the control group, the CS-1M group exhibited signi cantly reduced syncytium formation ( Fig. 3e and f), syncytium formation was inhibited even more severely in the CS-2M group, and almost no syncytia were observed in the CS-1M + 2M group ( Fig. 3e and f). More importantly, the inhibitory effects of these mutations were not rescued by the addition of CTSL into the system ( Fig. 3e and f), con rming that the promotive effect of CTSL on syncytium formation, as previously shown 9 , was due to S protein cleavage at CS-1 and CS-2. Taken together, our ndings proved that CTSL is essential for SARS-CoV-2 infection by cleaving the viral S protein at CS-1 and CS-2.

CTSL cleavage promotes SARS-CoV-2 infection independent of furin
As the CS-2 region is very close to the furin cleavage site, we next compared the function of these two proteases using different FM PsVs. PsV infection assays were performed in the human Huh7 cell line 3,21,25 under CTSL overexpression or knockdown conditions. Overexpression of the CTSL gene markedly increased but knockdown of the CTSL gene signi cantly reduced the infection e ciency of SARS-CoV-2 PsV (Fig. 3g), similar to our previous nding 9 . This result indicated that CTSL plays a crucial role in the SARS-CoV-2 infection process. Furthermore, overexpression or knockdown of the CTSL gene also signi cantly affected the infection levels of the three FM PsVs in a dose-dependent manner (Fig. 3g), similar to its effects on WT PsV. This result suggested that CTSL-enhanced SARS-CoV-2 viral entry is independent of furin cleavage.
The furin cleavage site has been reported to be required for SARS-CoV-2 S-driven cell-cell fusion 10 . Here, we used the same cell fusion system (Fig. 3d) to evaluate the effect of CTSL on FM S protein-mediated syncytium formation. FM-ARAA S protein induced a low level of syncytium formation similar to that in the scramble group ( Fig. 3h and 3i). However, after CTSL was added to the mutant S group, many syncytia began to form, and this rescue effect exhibited dose dependency with the concentration of CTSL ( Fig. 3h  and 3i). This result suggested that CTSL cleavage is another important factor that induces SARS-CoV-2 S-driven cell-cell fusion.

CTSL inhibitors prevent SARS-CoV-2 infection
Since CTSL plays a key role in mediating SARS-CoV-2 infection, we selected six compounds (K777,  cathepsin  Then, we further examined the antiviral effects of the six compounds against the recently emerging variants of SARS-CoV-2 using the PsV system (Fig. 4b).
As we expected, all six compounds also e ciently inhibited infection with all mutant PsVs. We found that E64d showed a higher inhibitory e ciency against P.1, while MDL-28170 and oxocarbazate were more effective against B.1.429. Thus, these data proved that CTSL inhibitors are resistant to mutational escape of SARS-CoV-2, including B.1.351, against which the effects of most current vaccines are reduced 7,8 .
Finally, we investigated the in vivo antiviral effects of CTSL inhibitors in human ACE2 (hACE2) transgenic mice, which are susceptible to SARS-CoV-2 infection 26 . We selected two compounds, E64d and Z-FY-CHO, which have been validated to be safe for in vivo studies 9,27 . E64d and Z-FY-CHO were administered in prophylactic and therapeutic regimens, and mice were treated at appropriate doses according to previous studies 9, 27 . Each mouse was infected with 10 6 PFU SARS-CoV-2 at 0 days post infection (dpi) by intranasal instillation. Tissue samples were collected at 4 dpi ( Fig. 5a), when the viral load peaked, and exhibited obvious histopathological changes. Both compounds signi cantly decreased the number of viral RNA copies in lung tissues by ~ 1 to 4 log10, with undetectable (below the lower limit of detection, LOD) viral loads in lung tissues from 2 mice treated with Z-FY-CHO and 4 mice treated with E64d ( Fig. 5b). Both drugs also signi cantly decreased the viral load in the nasal turbinate (Fig. 5b).
Histological analysis of the lungs showed that vehicle-treated mice exhibited moderate pathological changes, as evidenced by the large areas of alveolar septal thickening, in ammatory cell in ltration and bronchiolar epithelial cell degeneration. In contrast, the lungs of E64d-and Z-FY-CHO-treated mice exhibited improved morphology and less in ltration ( Fig. 5c and 5d). Taken together, these results indicated that CTSL inhibitors e ciently block SARS-CoV-2 infection both in vitro and in vivo and are resistant to viral mutational escape.

Discussion
In this study, we found that CTSL cleaves SARS-CoV-2 S at two novel sites in two loops in the S1 subunit, named CS-1 and CS-2. CS-1 is located in the NTD, while CS-2 is located in the RBD and is near the S1/S2 site. Both sites are highly conserved among all SARS-CoV-2 variants, including recently emerging variants. Our cryo-EM structural studies suggested that CTSL cleavage can greatly enhance S protein dynamics, with an increased proportion of activated state populations as well as novel activation conformations. Our viral infection assay using the PsV system proved that CTSL cleavage, especially cleavage at CS-2, is an essential step in SARS-CoV-2 infection.
Combining these results with those of our structural analysis, we proposed a possible model for the CTSL-mediated SARS-CoV-2 infection process ( Fig. 6 and Supplementary Movie 1). When viral S encounters active CTSL, CTSL can bind to CS-1 and CS-2 in the loops on the S protein to form the S-CTSL complex. CTSL cleavage at CS-1 greatly induces the 2R-weak conformation and increases the dynamics of the RBD and NTD to increase the ACE2 binding activity of S. Then, CTSL cleaves S at CS-2, which induces the 1N-2R-weak conformation, and the S1 and S2 subunits are then ready to separate to expose S2 for membrane fusion and complete viral infection.We proved that SARS-CoV-2 utilizes human CTSL to infect host cells independent of furin. In addition, many TMPRSS2-de cient cells, including Vero cells, have been found to be highly susceptible to SARS-CoV-2 infection, suggesting that TMPRSS2 is not indispensable for infection 28 . Therefore, it seems that instead of furin and TMPRSS2 inhibitors, CTSLspeci c inhibitors are a better choice to prevent SARS-CoV-2 infection. Indeed, CTSL inhibitors reduced infection with live SARS-CoV-2 by ~ 10 3− to 10 4 -fold in Vero E6 cells and nearly completely prevented infection with SARS-CoV-2 variant PsVs. Moreover, CTSL inhibitors effectively prevented SARS-CoV-2 infection in hACE2 transgenic mice.
We identi ed the precise CTSL cleavage loops in the S protein. This nding provides the structural basis for the development of new drugs and vaccines. In addition to CTSL-speci c inhibitors, drugs that can speci cally protect the cleavage loops from CTSL-mediated activation or competitively inhibit the binding of CTSL with the cleavage loops are new strategies for COVID-19 treatment. Moreover, vaccines targeting CS-1 and CS-2 to counteract viral mutations are another approach.
In conclusion, the high mutation rate of SARS-CoV-2 often leads to viral escape from neutralizing antibodies or vaccines. An understanding of the precise mechanisms underlying SARS-CoV-2 infection is urgently needed to develop speci c antiviral drugs or vaccines with increased resistance to viral mutations. Our study shows that the new CTSL cleavage sites in the SARS-CoV-2 S protein are crucial new targets for the development of drugs and vaccines to treat infections with SARS-CoV-2 variants.

Methods
Protein expression and puri cation The S protein was puri ed as previous reported 14 by Novoprotein company. Brie y, the prefusion S ectodomain, the mammalian codon-optimized gene coding SARS-CoV-2 (Wuhan-Hu-1 strain, GenBank ID: MN908947.3) S glycoprotein ectodomain (residues M1-Q1208) with proline substitutions at K986 and V987, a "GSAS" substitution at the furin cleavage site (R682 to R685), a C-terminal thrombin tag, a T4 britin trimerization motif and a 6xHisTag was synthesized and cloned. To express 4-Gly-mutant S protein, residues W258, T259, V635 and Y636 were substituted into glycine. These expression vectors were used to transiently transfect to Expi293 cells and puri ed by using cOmpleteHis-Tag Puri cation Resin to capture the target protein. The protein was then subjected to additional puri cation by sizeexclusion chromatography using Superdex 200 10/300 GL column (GE Healthcare) in PBS. CTSL treatment Recombinant CTSL (Novoprotein) was used to cleave SARS-CoV-2 S protein. Puri ed SARS-CoV-2 S protein was incubated with CTSL (2-8 mg/mL as indicated and 8 mg/ml for cryo-EM sample preparation) in the presence of 100 mM NaAC, 1 mM EDTA and 1 mM DTT, pH 5.5 for 24 h at 25℃. CTSL was preactivated in 30°C for 1 min before use. The resulting protein was then subjected to SDS-PAGE and stained by Coomassie blue or silver staining methods as indicated. For cryo-EM sample preparation, the treated S protein was then subjected to additional puri cation by size-exclusion chromatography using

Image processing
Image processing steps were performed using RELION 32,33 and cryoSPARC 34 . The output super resolution movies were rst subjected to motion correction using MotionCor2 35 , with binning level of 2 in Fourier space, and dose weighting was also performed during the process. Contrast transfer function (CTF) parameters estimation were performed using Gctf 36 . Gautomatch (https://www.mrclmb.cam.ac.uk/kzhang/Gautomatch/) was used to pick 364k particles from 1249 micrographs in untreated S dataset, and 1500k particles from 5025 micrographs in CTSL treated S dataset. Then good particles were extracted and sorted by two rounds of 2D classi cation. After 2D classi cation, one round of 3D classi cation was performed to investigate different populations of Ss. Three different populations were found for the untreated dataset and six for the CTSL treated dataset. For the S-1N-attach population, another round of 3D classi cation was performed to further sort particles. To prevent any model bias, previously reported cryo-EM structure of SARS-CoV-2 S (PDB entry 6VXX) lowpass ltered to 60 Å was used as initial reference and no additional symmetry was imposed throughout the whole image processing pipeline. Then the well-aligned classes with clear secondary structure features were selected separately for subsequent image processing, including auto-re nement, Bayesian polishing, CTF re nement and no-alignment 3D classi cation. All these steps were performed in RELION. For the nal round of re nement, the non-uniform (NU) re nement in cryoSPARC was performed to re ne the nal maps to a higher resolution (see also Figures S3 and S4).

Experimental mice
The study used human ACE2 transgenic mice, a mouse model expressing human ACE2 (hACE2) generated by using CRISPR/Cas9 knock-in technology as previously reported 26 . The hACE2 mice used in this manuscript were 17-weeks-old female C57BL/6 mice, with the body weight between 22 g. All animal protocols were approved by the Ethical Review Committee at the Institute of Zoology, Capital Medical University, China.

SARS-2-S plasmids and site-directed mutagenesis
SARS-2-S plasmid was constructed by inserted the codon-optimized S gene of SARS-CoV-2 (GenBank: MN_908947) into pcDNA3.1as previously described 3 . For the site-directed mutagenesis, 15 to 20 nucleotides before and after the target mutation site were selected as forward primers, while the reverse complementary sequences were selected as reverse primers. Following site-directed mutagenesis PCR, the template chain was digested using DpnI restriction endonuclease (NEB, USA). The PCR product was transformed into E. coli DH5a competent cells; single clones were selected and then sequenced. Production and quanti cation of PsVs PsVs incorporated with S protein from either SARS-CoV-1, SARS-CoV-2, or mutants were constructed using a procedure described by us previously. For this VSV-based PsV system, the backbone was provided by VSV-G pseudotyped virus (G*ΔG-VSV) that packages expression cassettes for re y luciferase instead of VSV-G in the VSV genome. For quanti cation of PsV, viral RNA was extracted by reverse-TGCTCTTCCACTCCATCCTCTTGG. All PsVs were normalized to the same amount as previously described 3 .

Western blot analysis
Seven millilitres of SARS-CoV-2 PsVs with a titre of 1.86 × 10 5 TCID 50 /ml were pelleted through a 25% sucrose cushion by ultra-centrifugation at 100,000× g for 3 h. The layer of supernatant was discarded.
The pellet was collected and subjected to western-blot analysis as previously described, and detected by anti-S1 mice serum and VSV-M antibody (KeraFast, Cat. No. EB0011). PsV infection assay Before infection, the 96-well plates were seeded with cells adjusted to 2 × 10 5 cells/ml. Then, 100 mL of the normalized PsV with indicated dilution fold was added to wells in 96-well cell culture plate. After 24 h incubation at 37 ℃, the activities of re y luciferase were measured on cell lysates using luciferase substrate (Perkinelmer, BRITELITE PLUS 100 ml KIT, Cat. No. 6066761) following the manufacturer's instructions. Luciferase activity was measured using a luminometer (Promega).

Overexpression and knockdown of CTSL
Overexpression and knockdown of CTSL gene in huh7 cells were described by us previously 9 . Brie y, for CTSL knockdown, Huh7 cells were plated in 48-well plates, and transfected with indicated concentrations of siRNAs against homo CTSL mRNA (si-CTSL) or 50 nM negative control siRNA (scramble) using Lipofectamine 3000 reagent (Thermo Fisher Scienti c). For CTSL overexpression, Huh7 cells were plated in 48-well plates and transfected with indicated concentrations of human CTSL expression plasmid (GenBank: NM_001912, Vigenebio) or 0.2 mg control plasmid. The overexpression and knockdown e ciencies were validated in both mRNA and protein levels by us previously 9 . Twenty-four hours post transfection, the medium was replaced with fresh medium. Then the cells were infected with SARS-CoV-2 or mutant PsV (1.3 × 10 4 TCID 50 /ml) and cultured for another 24 h before re y luciferase activity analysis. siRNA sequences were provided in Supplementary Table 5.

Syncytium-formation assay
Syncytium-formation assay was performed as previously described by us 9 . Brie y, Huh7 cells were seeded in 24-well plates and transfected with SARS-2-S or mutant protein expression plasmids using Lipofectamine 3000 reagent (Thermo Fisher Scienti c). The transfection solutions were changed to standard culture medium 4-6 h post transfection and cells incubated for an additional 6-8 h. Next, cells were treated in the absence (PBS, pH = 5.8) or presence of 2 or 4 mg/ml CTSL (Novoprotein) (in PBS, pH = 5.8) for 20 min at 37°C. Then, the solutions were changed to a standard culture medium, and the cells further incubated for 10-16 h. The pictures were captured under bright-eld microscopy (Olympus).

Quanti cation of syncytium-formation
Huh7 target-cells were seeded in 48-well plate at 10-20 % con uency and transfected with ESR1 expression plasmid. Huh7 effector-cells were seeded in a 6-well dish at 50-70 % con uency and transfected with ERE-Luciferase reporter plasmid as well as expression plasmid for WT or mutant SARS-2-S or empty plasmid (1:2 ratio). At 24 h post-transfection, effector-cells were detached by resuspending and added into the target-cells at 1:1 ratio. Where indicated, the medium was changed by CTSL, trypsin or PBS 2 h after mixture and reaction was stopped after 20 min by adding 500 mL culture medium.

Analysis of CTSL activity
The CTSL inhibition by drugs was measured using the CTSL substrate Ac-FR-AFC (Abcam, Cat. No. ab157769). CTSL will cleave the synthetic substrate to release free AFC (amino-4-tri uoromethyl coumarin), which can be measured in a spectrophotometer at Em = 505 nm. Puri ed CTSL (Novoprotein) was incubated with CTSL substrate in the absence (vehicle, control) or presence of 10 mM compounds as indicated in 37 ℃ for 1-2 h, the CTSL activity was measured by a spectrophotometer (Promega) at

SARS-CoV-2 infection assay in vivo
Prophylactic treatment was given on -2, -1, 0, 1, 2, and 3 dpi, while therapeutic regimen was given on 0, 1, 2, and 3 dpi. The mice were infected at 0 dpi with 10 6 PFU of SARS-CoV-2 by intranasal instillation. Lung and nasal turbinate tissues were collected at 4 dpi. The viral RNAs copies from tissues were measured by RT-qPCR. Tissue pathology of infected mice was examined by H&E staining and immunohistochemical assays as described previously 43,44 .

Statistical analysis
All statistical analyses were conducted with the software GraphPad Prism version 7.0 Data are presented as means ± s.e.m. Statistical signi cance was determined using unpaired Student's t-test or the Mann-Whitney U-test if n < 9, or one-way ANOVA as appropriate.  CTSL cleavage site 2 (CS-2). Furin cleaves at the S1/S2 site. b, Overall structure of the SARS-2 S ectodomain (PDB entry: 6VXX). The CTSL cleavage sites, CS-1 and CS-2, are colored red, while the furin cleavage site at S1/S2 is colored orange. The three protomers of the S trimer are colored pink, purple and green. c, SDS-PAGE analysis of CTSL-mediated cleavage of the S protein. The puri ed SARS-CoV-2 S protein ectodomain was incubated with different concentrations of CTSL (2-8 μg/mL). The N-terminal sequencing results for band 1 (60 kDa) and band 2 (100 kDa) are also shown. d, 4-Gly-mutant SARS-CoV-2 S cannot be cleaved by CTSL into the 60 kDa (band 1) and 100 kDa (band 2) fragments. WT S protein (1 μg) and mutant S protein (500 ng) were incubated with CTSL (8 μg/mL). All samples were subjected to SDS-PAGE, and bands were detected by silver staining. e, The P1 and P2 residues in CS-1 and CS-2 were mutated to glycine, and the mutant SARS-2-S glycoprotein was named 4-Gly-mutant S. f, Amino acid sequence alignment of residues around CS-1 and CS-2 in SARS-CoV-2 variants. P1 and P2 residues that are the same as the WT residues are highlighted in red. The symbol * indicates amino acid residues that are conserved among all tested sequences. analysis of syncytium formation induced by CS mutant SARS-CoV-2 S proteins (see also Supplementary  Fig. 11). (c) Luciferase gene expression was driven by the ERE promoter, and ESR1 (activator) bound and activated the ERE promoter to upregulate luciferase expression. (d and e) Effector Huh7 cells were cotransfected with plasmids expressing ERE-luciferase and different S proteins as indicated, and another plate of target Huh7 cells was transfected with plasmid expressing ESR1. After 24 h, the effector cells were detached and added to the target cells for 30-60 min. Then, the supernatant was removed and treated with PBS or CTSL (8 g/mL) for 20 min. The reaction was stopped by adding 500 L of medium, and culture was continued for another 24 h to allow cell-cell fusion. When a target cell and effector cell fused to form a syncytium, ESR1 bound and activated the ERE promoter to upregulate luciferase expression. Luciferase activity was then measured as a proxy for the fusion rate. The data were PsV infectivity in Huh7 cells was measured by a luciferase assay and is shown as the relative luciferase activity (n=3). Statistical signi cance was assessed by one-way ANOVA with Tukey's post hoc test. h, CTSL promoted syncytium formation induced by the FM-ARAA mutant SARS-CoV-2 S protein. Effector cells were cotransfected with ERE-luciferase plasmids and either FM-ARAA S or scramble vectors (Control). Target cells were transfected with ESR1 expression plasmid. After the effector cells and target cells were mixed, the supernatant was removed and treated with PBS or CTSL (8 g/mL and 16 g/mL).
Luciferase activity was then measured and normalized to that in the control group (n=4). Statistical signi cance was assessed by one-way ANOVA with Tukey's post hoc test. i, Huh7 cells were transfected with scramble vector or FM-ARAA S protein expression plasmid. Cells were treated in the absence (PBS) or presence of CTSL (2 or 4 g/mL). Images were acquired after an additional 10-16 h of incubation in medium (scale bars, 50 μm). The black arrowheads indicate syncytia. Representative data from four independent experiments are shown. The data are presented as the mean ± s.e.m. values. * P < 0.05, ** P < 0.01, *** P < 0.001, *** P < 0.0001.  CTSL inhibitors prevent SARS-CoV-2 infection in vivo. a, E64d and Z-FY-CHO were administered intraperitoneally at -2 ~ 3 dpi as prophylactic treatment, and mice were challenged with 106 PFU at 0 dpi; the two drugs were administered therapeutically at 0~3 dpi. Tissue samples were collected at 4 dpi. b, Viral RNA copies in mouse lung and nasal turbinate tissues (n=5 mice/group). The dotted line indicates the lower limit of detection (LOD). Statistical signi cance was assessed between the indicated group and control group by one-way ANOVA with Tukey's post hoc test. c, Representative images from histological analysis of lungs from SARS-CoV-2-infected hACE2 mice at 4 dpi. Magni ed views of the boxed regions in each image are shown below the corresponding image. The black arrows indicate in ammatory cell in ltration, the black arrowheads indicate bronchiolar epithelial cell degeneration, and the red arrows indicate alveolar septal thickening. d, Semiquantitative histological scoring of each lung tissue was performed by grading the severity of bronchiolar epithelial cell damage (0-10), alveolar damage (0-10) and in ammatory cell in ltration in blood vessels and bronchioles (0-10) and summing these scores to calculate the total score. Normal = 0, indeterminate = 1-2, mild = 3-4, moderate = 5-7, severe = 8-10. (n=3) Statistical signi cance was assessed between the indicated group and control group by one-way ANOVA with Tukey's post hoc test. Proposed mechanism by which CTSL promotes SARS-CoV-2 infection. SARS-CoV-2 S contains S1 and S2 subunits. CS-1 is located in the NTD of the S1 subunit, and CS-2 is located near the S1/S2 site. CTSL cleaves the SARS-CoV-2 S protein: (1) By binding with the S protein on the surface of SARS-CoV-2, CTSL cleaves S at CS-1 to increase the dynamics of the RBD and make it accessible to ACE2 for binding. (2) CTSL further cleaves S at CS-2 to activate the S2 subunit for membrane fusion. (3) The virus fuses with the cell membrane, and the viral genetic material is released into the host cell.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.