TMPRSS2 Protease Inhibitors May Prolong But Heparins Accelerate SARS-CoV-2 Clearance

TMPRSS2 Protease Inhibitors May Prolong But Heparins Accelerate SARS-CoV-2 Clearance 1 2 Shu Yuan, Si-Cong Jiang, Zhong-Wei Zhang, Zi-Lin Li, Chang-Quan Wang, Ming Yuan, Yang-Er 3 Chen, Qi Tao, Ting Lan, Xiao-Yan Tang, Guang-Deng Chen, and Jian Zeng 4 5 College of Resources, Sichuan Agricultural University, Chengdu 611130, China 6 Chengdu KangHong Pharmaceutical Group Comp. Ltd., Chengdu 610036, China 7 Department of Cardiovascular Surgery, Xijing Hospital, Medical University of the Air Force, Xi’an 710032, 8 China 9 College of Life Science, Sichuan Agricultural University, Ya’an 625014, China 10 Co-first author 11 Correspondence: roundtree318@hotmail.com (S.Y.) 12 13 SUMMARY 14 The 2019 novel SARS-like coronavirus (SARS-CoV-2) entry depends on the host membrane serine protease 15 TMPRSS2, which can be blocked by some clinically-proven drugs. Here we analyzed spatial relevance 16 between glycosylation sequons and antibody epitopes and found that, different from SARS-CoV S, most 17 high-surface-accessible epitopes of SARS-CoV-2 S are blocked by the glycosylation, and the optimal 18 epitope with the highest surface accessibility is covered by the S1 cap. TMPRSS2 inhibitor treatments may 19 prevent unmasking of this epitope and therefore prolong virus clearance and may induce 20 antibody-dependent enhancement. Interestingly, a heparin-binding sequence immediately upstream of 21 the S1/S2 cleavage site has been found in SARS-CoV-2 S but not in SARS-CoV S. Binding of SARS-CoV-2 with 22 heparins may lead to exposure of S686, which then facilitates the S1/S2 cleavage, induces exposure of the 23 optimal epitope, and therefore increases the antibody titres. A combination of heparin and vaccine (or 24 convalescent serum) treatments thus is recommended. 25 26 Graphical Abstract 27


In Brief 1
Most strong epitopes of SARS-CoV-2 S are blocked by the glycosylation, and the optimal epitope with the 2 highest surface accessibility is covered by the S1 subunit. Heparin facilitates the S1/S2 cleavage. Therefore, 3 TMPRSS2 inhibitors may prolong but heparins may accelerate SARS-CoV-2 clearance. 4 5 1

RESULTS AND DISCUSSION 2
Positive Electrostatic Potential of SARS-CoV-2 S Protein May Explain Its High Affinity to ACE2. 3 4

Figure 1. Electrostatic Potential of SARS-CoV S, SARS-CoV-2 S and Human ACE2 5
The red-to-blue color on the molecular surface indicates the electrosta c poten al (red: −1.8; blue: 1.8). 6 The S1/S2 cleavage sites are marked with the dark purple color. The receptor-binding motifs (RBM) are 7 marked with the pale lavender color. 8 9 The predominant state of the trimer has one of the three receptor-binding domains (RBDs) rotated up in a 10 receptor-accessible conformation. Biophysical and structural evidences indicated that ACE2 bound to the 11 SARS-CoV-2 S ectodomain with about 15 nM affinity, which is 10 to 20-fold higher than ACE2 binding to 12 SARS-CoV S (Yan et al., 2020). Here we calculated electrostatic potential of SARS-CoV S protein, 13 SARS-CoV-2 S protein and human ACE2 ( Figure 1). Interestingly, both SARS-CoV S and ACE2 protein 14 surfaces are uniformly negatively-charged, and therefore they repel each other. However, a large part of 15 SARS-CoV-2 S protein surface is electrically neutral but its receptor-binding motif (RBM) is positive-charged, 16 and therefore SARS-CoV-2 S and ACE2 attract each other. The S1/S2 cleavage sites are distributed in the 17 middle of both SARS-CoV S and SARS-CoV-2 S proteins, implying that TMPRSS2-mediated S1/S2 cleavage 18 may not influence ACE2 binding. 19 20

SARS-CoV-2 S Fusion Core Peptides Are More Hydrophobic than SARS-CoV S 21
A study of the X-ray crystal structure revealed that the six-helical fusion core in the SARS-CoV-2 S protein 22 S2 subunit is formed by interaction between two heptad repeat domains HR1 and HR2 . 23 The three HR1 domains (894-966 of SARS-CoV S protein or 912-984 of SARS-CoV-2 S protein) form a 24 parallel trimeric coiled-coil center, around which three HR2 domains (1145-1195 of SARS-CoV S protein or 25 1163-1213 of SARS-CoV-2 S protein) are entwined in an antiparallel manner . The 26 interaction between these two domains is predominantly a hydrophobic force. Each pair of two adjacent 27 HR1 helices forms a deep hydrophobic groove, providing the binding site for hydrophobic residues of the 1 HR2 domain. The hydrophobic appearance (electrically neutral surface) plays an important role in the 2 membrane fusion process . 3 4 Figure 2. Electrostatic Potential of SARS-CoV and SARS-CoV-2 S1 and S2 Subunits 5 The Viral spike (S) protein could be divided into S1 and S2 subunits upon the cleavage by TMPRSS2. The  6 red-to-blue color on the molecular surface indicates the electrosta c poten al (red: −1.8; blue: 1.8). The 7 S1/S2 cleavage sites are marked with the dark purple color. The heptad repeat domain HR1 on one of the 8 three monomers is marked with orange (invisible segment covered by the S1 cap), green (the fusion core) 9 and yellow colors (visible segment without a cover of S1 cap). In the SARS-CoV fusion core, only three aa 10 distribute on an electrically-neutral area (marked with the pale green color); while the others distribute on 11 the hydrophilic area. Different from SARS-CoV, the SARS-CoV-2 fusion core is much more hydrophobic that 12 only three aa distribute on an electrically-negative area (marked with the brown color) and the others 13 distribute on the electrically-neutral area. 1 2 The SARS-CoV fusion core is composed of 19 amino acids (aa; 911-929 of SARS-CoV S); while the 3 SARS-CoV-2 fusion core is also composed of 19 aa (929-947 of SARS-CoV-2 S; Figure 2). Interestingly, a 4 majority of SARS-CoV fusion core peptide surface is negatively-charged, which could be converted into 5 positive-charged after the TMPRSS2 cleavage, indicating an electrical charge redistribution. Among the 19 6 aa, only three aa distribute on an electrically-neutral area; while the others distribute on the hydrophilic 7 area. Different from SARS-CoV, the SARS-CoV-2 fusion core is much more hydrophobic that only three aa 8 distribute on an electrically-negative area and the others distribute on the electrically-neutral area. More 9 hydrophobic appearance of SARS-CoV-2 fusion core may be another reason for its higher infectivity 10 compared to SARS-CoV. Interestingly, C-terminus of HR1 domain in either SARS-CoV S (930-966 aa) or 11 SARS-CoV-2 S (948-984 aa) is covered by the S1 subunit, which could be unmasked upon proteolysis by 12 TMPRSS2, also confirming the role of TMPRSS2 in the conformational changes required for the membrane 13 fusion process. 14 15 The Optimal Epitope with the Highest Surface Accessibility Is Covered by the S1 Cap 16 Being exposed on the viral surface, S proteins are a major target for host antibodies and are referred to as 17 viral antigens; these antigens are therefore targets for vaccine development (Zheng and Song, 2020).

18
However, viral envelope proteins are often modified by the attachment of complex glycans. The 19 glycosylation of these surface antigens helps the pathogen evade recognition by the host immune system 20 by cloaking the protein surface from detection by antibodies, and can influence the ability of the host to 21 raise an effective adaptive immune response or even be exploited by the virus to enhance infectivity 22 (Baum and Cobb, 2017; Pereira et al., 2018).

23
In this study, we computed sequence-based antibody epitopes on spike proteins of SARS-CoV and 24 SARS-CoV-2 (Tables S1 and S2). As the surface accessibility of epitope is the most important determinant 25 to the interaction between antibody and antigen, the possible antibody epitopes were filtered with the 26 surface accessible scores by using the default threshold value of 1.0 (Emini et al., 1985). Then the epitope 27 candidates were re-scored by using BepiPred-2.0 bioinformatic tool with the default threshold value of 28 0.50 (Jespersen et al., 2017). 27 epitopes were found on SARS-CoV S protein, among which 10 epitopes 29 had been ruled out due to the low epitope scores. And 30 epitopes were identified on SARS-CoV-2 S, 30 among which 9 epitopes had been ruled out due to the low epitope scores (Tables S1 and S2). In SARS-CoV 31 RBD region (306-527) and SARS-CoV-2 RBD region (319-541) respectively, 4 epitopes and 6 epitopes were 32 screened out finally. Our epitope prediction has been proved by two clinical studies. In one study, 399 33 human monoclonal antibodies (mAbs) have been sorted in 10 SARS-CoV-2 patients, but only 35 34 S-protein-specific mAbs were acquired, among which, 4 mAbs recognize RBD (Chi et al., 2020). Another 35 study indentified the S230 antibody, which was isolated from memory B cells of a SARS-CoV-infected 36 individual and potently neutralized a broad spectrum of SARS-CoV isolates of human and animal origins 37 (Rockx et al., 2008). The S230 epitope is centered around L443 on S protein and Y408, Y442, F460 and 38 Y475 participate binding to this antibody (Rockx et al., 2008), which matches to a 14 aa epitope candidate 39 (431-444) screened out in this study with a high surface accessibility (SA) score of 3.149 (Table S1). 40 1 Figure 3.

Distribution of Glycosylation Sequons and Antibody Epitopes on SARS-CoV S and SARS-CoV-2 S 2
The Viral spike (S) protein could be divided into S1 and S2 subunits after the cleavage by TMPRSS2. The sequons are marked with the green color. The putatively-optimal epitope (755-761) of SARS-CoV with the 7 highest SA score of 4.431 is located on the cutting surface of S2 subunit, which would be uncovered only 8 after TMPRSS2 cleavage. And the putatively-optimal epitope (773-779) of SARS-CoV-2 with the highest SA 9 score of 4.868 is also located on the cutting surface of S2 subunit, whose binding requires removal of the 1 S1 cap. Due to the coverage limitation in the Swiss model, glycosylation sequons and epitopes in 2 1120-1255 aa of SARS-CoV S or in 1147-1273 aa of SARS-CoV-2 S are not shown in the figure. To present 3 the sites more clearly, only one of the three monomers is labeled. with these data, spatial relevance between glycosylation sequons and antibody epitopes were further 7 analyzed ( Figure 3). Grant et al. (2020) demonstrated that most SARS-CoV and SARS-CoV-2 epitopes are 8 shielded by glycans, and only areas of the protein surface at the apex of the S1 domain appear to be 9 accessible to known antibodies (Vankadari and Wilce, 2020). A visual examination of the structures from 10 molecular dynamics simulation also confirmed that the most exposed epitopes comprise the ACE2 11 receptor site RBD, specifically at the apex region of the RBM domain (Grant et al., 2020). Similar results 12 were also obtained in this study. On SARS-CoV S, only three strong epitopes with SA scores >3.0 have been 13 identified. One epitope (431-444; matching to the S230 epitope as mentioned above) recognizes RBD and 14 is not surrounded by glycosylation sequons. Another epitope (1238-1243) is located in the C-terminal 15 transmembrane domain ( Figure S1) and therefore should not be accessible to any antibody. The 16 putatively-optimal epitope (755-761) with the highest SA score of 4.431 is located on the cutting surface of 17 S2 subunit, which could be uncovered only after TMPRSS2 cleavage ( Figure 3). Besides, the epitope 18 540-548 is also not surrounded by glycosylation sequons, however its relatively low SA score (2.396) may 19 suggest a low neutralizing ability ( Figure 3 and Figure S2). 20 Unfortunately, no strong epitopes (SA scores >3.0) is available that recognizes SARS-CoV-2 RBD. This 21 finding is consistent with the fact that only low level of binding of SARS-CoV-2 S to polyclonal rabbit 22 anti-SARS S1 antibodies T62 was detected (Ou et al., 2020). Two strong epitopes are located on 23 SARS-CoV-2 S1 (674-685) and S2 (808-817) subunit surfaces respectively. However both of them are 24 accompanied with glycosylation sequons. Although these two epitopes have large surface areas, their 25 accompanying glycosylation sequons are located on raised areas, and therefore may form the steric 26 hindrance ( Figure 3). There is also a strong epitope (1256-1261) located in the C-terminal transmembrane 27 domain ( Figure S1). And the putatively-optimal epitope (773-779) with the highest SA score of 4.868 is 28 also located on the cutting surface of S2 subunit, whose binding requires removal of the S1 cap ( Figure 3). 29 Notably, a remarkable alterations in the antigenicity was observed in SARS-CoV-2 that no strong 30 RBD-targeting epitopes is available and almost all high-surface-accessible epitopes are blocked by the 31 glycosylation, including the 4A8 epitope sorted recently (matching to a 10 aa epitope 144-153 indentified 32 in this study; Chi et al., 2020). These results might explain why the sera from convalescent SARS-CoV-2 33 patients exhibited a much weaker neutralizing antibody response compared to SARS-CoV (Hoffmann et al., 34 2020).

35
These results also imply that developing of monoclonal antibodies may not be an idea strategy to treat 36 SARS-CoV-2 infections. Alternatively, recombinant virus vector vaccines, DNA vaccines or inactivated virus 37 vaccines may induce strong cellular immunity rather than humoral immunity that produces antibodies Considering that almost all high-surface-accessible epitopes of SARS-CoV-2 are blocked by the glycan 43 shield, people may deduce that the virus should not be cleaned up by the immune system. But that is not 44 the truth. The epitope on the cutting surface usually have no time to bind with the corresponding 1 antibody, since the membrane fusion occurs immediately following the S1/S2 cleavage. However, free 2 TMPRSS2 makes the antibody binding possible. TMPRSS2 is a secreted protease that is highly expressed in (C-terminal) may detach from the membrane and be released (secreted) to the extracellular space. As a 10 result, a small part of SARS-CoV-2 S proteins may be cleavaged by free TMPRSS2 before they bind the 11 receptor ACE2 and then the epitope on the cutting surface may have a time to induce a neutralizing SARS-CoV-2 entry into primary human lung cells. However as analyzed above, the optimal epitope with 19 the highest surface accessibility is covered by the S1 cap and thus TMPRSS2 inhibitors may prevent 20 unmasking of this epitope and prolong virus clearance subsequently. Nevertheless, the delay in virus 21 clearance caused by TMPRSS2 inhibitors may not occur in SARS-CoV infections, because that the 22 neutralizing antibody S230 would play a crucial role in the virus clearance (Rockx et al., 2008). 23 Antibody-dependent enhancement (ADE) of viral entry has been a major concern for epidemiology, 24 vaccine development, and antibody-based drug therapy ( with SARS-CoV-2, one patient (1/4) still developed severe diseases. These clinical data imply that higher 4 TMPRSS2 levels in prostate cancer patients did not increase their illness duration, but decreased the 5 mortality rate significantly; inhibition to TMPRSS2 (as androgen-deprivation therapy) may not improve the 6 outcomes. 7 Nevertheless, only 4 of 5273 (0.076%) prostate cancer patients receiving androgen-deprivation therapy 8 were infected with SARS-CoV-2; while 114 of 37,161 (0.307%) prostate cancer patients without 9 androgen-deprivation therapy were infected with SARS-CoV-2. The infection rate decreased by 75.1% after 10 the androgen-deprivation therapy. Camostat, nafamostat, or other TMPRSS2 inhibitors (e.g. bromhexine 11 as recommended by Stopsack et al., 2020) may be used as prophylactic drugs to reduce the risk of 12 infection, because that TMPRSS2 inhibitors may decrease the initial viral load during the incubation period. 13 However they may be inefficient for the patients who already develop symptoms, or even have a 14 detrimental effect on the virus clearance.

Figure 4. Distribution of a Heparin-Binding Sequence Immediately Upstream of the S1/S2 Cleavage Site 19 on SARS-CoV-2 S But Not on SARS-CoV S 20
A heparin-binding sequence immediately upstream of the S1/S2 cleavage site has been found in 21 SARS-CoV-2 S but not in SARS-CoV S. The heparin-binding sequence is marked with the red color. Both 22 R667 and S668 in SARS-CoV S cleavage site are exposed on the protein surface (marked with the dark 23 purple color). Contrastingly, although R685 in SARS-CoV-2 S cleavage site is exposed on the protein surface 24 (marked with the dark purple color), S686 in SARS-CoV-2 S is embedded under the protein surface (marked 25 with the light purple color), which may be exposed above the protein surface via a conformational change 26 induced by the heparin binding. To present the sites more clearly, only one of the three monomers is 27 labeled.

29
Heparin is a mucopolysaccharide sulfuric acid ester that is found especially in the liver and lungs. Heparin SARS-CoV-2 S, and interestingly found that only SARS-CoV-2 S has a HS-binding sequence (681-686 PRRARS) 9 immediately upstream of the S1/S2 cleavage site (R685-S686; Figure 4). Another intriguing difference 10 between SARS-CoV S and SARS-CoV-2 S is that both R667 and S668 in SARS-CoV S cleavage site are 11 exposed on the protein surface, but S686 in SARS-CoV-2 S is embedded under the protein surface ( Figure  12 4). These findings imply that heparin binding may be required for SARS-CoV-2 S1/S2 cleavage, but not for 13 SARS-CoV S1/S2 cleavage. Binding of SARS-CoV-2 with membrane-bound heparins may lead to exposure of 14 S686 by a conformational change, which then facilitates the S1/S2 cleavage and the subsequent 15 membrane fusion (virus entry). While if SARS-CoV-2 S binds free heparins in the interstitial fluids or in the 16 blood, the enhanced S1/S2 cleavage may induce more exposure of the optimal epitope 773-779, which 17 therefore accelerates SARS-CoV-2 clearance ( Figure 5). One copy of the HS-binding motif adjacent to the The enhancement to antigenicity by free heparins has been confirmed by a serological assay (Perera et  22 al., 2020). They observed a 1.0-1.5 log 10 reduction in TCID 50 (median tissue culture infective dose) when 23 the SARS-CoV-2 was diluted in the heparin medium compared with the control medium. They also carried 24 out titrations of three sera (from COVID-19 patients) with known micro-neutralisation antibody titres of spatial relevance between the heparin-binding sequence and the S1/S2 cleavage site on SARS-CoV-2 S, a 40 much lower IC 90 specific to SARS-CoV-2 could be expected. 41 The therapeutic effects of heparins on SARS-CoV-2 infections have been confirmed clinically.

42
Anticoagulant therapy with low molecular weight heparin (LMWH) has been suggested to treat COVID-19, Moreover, heparin also showed a good therapeutic effect to acute respiratory distress syndrome (ARDS), 2 which is a common complication of viral pneumonia (Thompson et al., 2017). Here we added an important 3 information that heparin may also inhibit SARS-CoV-2 entry by both enhancing neutralizing antibody titres 4 and preventing viral adhesion on the cell surface. Thus, LMWH anticoagulant therapy may also work for 5 the non-severe patients. On the other hand, COVID-19 has a prominent feature, that is, a large amount of 6 mucus (oedema and plasma exudation) could be found in the small airway, and it may eventually block the 7 airway, which may be an important reason for the high mortality after later mechanical ventilation and 8 high-flow oxygen inhalation (Barton et al., 2020). Therefore, nebulized heparin, oxygen supply or other 9 inhalation therapies should be given at the early stages of COVID-19 (Yuan et al., 2020b). 10 11

Figure 5. Drugs against SARS-CoV-2 Entry and Their Effects on the Virus Clearance 12
Coronavirus rolls onto the cell membrane by binding to cell-surface cholesterols and heparan sulfate 13 proteoglycans (HSPGs) and scans for the specific entry receptor ACE2, which leads to subsequent cell entry. 14 Camostat, nafamostat or bromhexine inhibits the plasma membrane protease TMPRSS2, which is 15 responsible for the proteolysis of viral S proteins in the post-receptor-binding stage. Methyl-β-cyclodextrin 16 and heparin inhibit virus binding with cholesterols and HSPGs respectively. Chloroquine neutralizes acidic 17 pH in the endosome, which is necessary for viral nucleocapsid release into the cytoplasm. PIKfyve 1 inhibitors apilimod and YM201636, TPC2 inhibitor tetrandrine and cathepsin L inhibitors E64D and SID 2 26681509 prevent the virus entry. On the other hand, in the interstitial fluids or in the blood, free heparin 3 binding may lead to exposure of the S1/S2 cleavage site by a conformational change. Then the enhanced 4 S1/S2 cleavage by free TMPRSS2 may induce more exposure of the optimal epitope 773-779, which 5 therefore accelerates neutralizing-antibody-mediated SARS-CoV-2 clearance. Contrastingly, TMPRSS2 6 inhibitors prevent unmasking of the optimal epitope and thus hamper neutralizing antibody activities, 7 prolonging the virus clearance. Although TMPRSS2 inhibitors may prevent macrophage death caused by 8 the SARS-CoV-2 entry, they increase the likelihood of viral attachment to the macrophage surface, which 9 induces proinflammatory responses and antibody-dependent enhancement (ADE). efficiently. More effective TMPRSS2 inhibitors still need to be developed. 26 Contrastingly, nebulized heparin is inhaled directly into the lung, so it can reach a local high 27 concentration in alveolar cells. Although the alveolar concentration cannot be easily estimated (1 mg/mL 28 LMWH is usually used for the ultrasonic atomization, which is equal to about 67 μM), it may be higher 29 than 10 μM that can induce a 1.0-1.5 log 10 reduction in TCID 50 of SARS-CoV-2 (Perera et al., 2020). 30 Nevertheless, given that heparin may cause thrombocytopenia and thrombosis, more clinical trials are still 31 required to determine the optimal dosage and therapeutic time.

32
Besides TMPRSS2 inhibitors and heparins, other drugs that inhibit coronavirus entry are summarized 33 and listed in Table 1 and Figure 5. The cellular alkalizers also repress virus entry through neutralizing acidic 34 pH in the early endosomes, which is necessary for viral nucleocapsid release into the cytoplasm.

35
Chloroquine and its derivative hydroxychloroquine are such alkalizers and are used clinically as 36 antimalarial medicines. In-vitro experiments confirmed that chloroquine is highly effective in the control of  showed that IC 50 of nafamostat was about 10 times lower than that of camostat against MERS-CoV. Thus, it 5 could be deduced that IC 90 of nafamostat against either SARS-CoV or SARS-CoV-2 may be 0.5 μM. † 10 μM 6 heparins induced a 1.0-1.5 log 10 reduction in TCID 50 of SARS-CoV-2. Thus, it could be deduced that IC 90 of 7 heparin against SARS-CoV-2 may be < 10 μM. N.A., Not Available. Nevertheless, a recent study demonstrated that phosphatidylinositol 3-phosphate 5-kinase (PIKfyve), two 14 pore channel subtype 2 (TPC2), and cathepsin L are critical for SARS-CoV-2 entry (Ou et al., 2020), and 15 PIKfyve inhibitors apilimod and YM201636, TPC2 inhibitor tetrandrine and cathepsin L inhibitors E64D and 16 SID 26681509 prevent the virus entry (Table 1; Ou et al., 2020). However, none of them are FDA-approved 17 drug and may have many side effects. In a nutshell, among all FDA-approved drugs against SARS-CoV-2 18 entry putatively, camostat, nafamostat or bromhexine may be candidate prophylactic drugs, nebulized 19 heparin may be a promising therapeutic drug, and validity and safety of (hydroxy)chloroquine require 20 further clinical investigations. 21 22