key: cord-256156-mywhe6w9
authors: Clausen, Thomas Mandel; Sandoval, Daniel R.; Spliid, Charlotte B.; Pihl, Jessica; Perrett, Hailee R.; Painter, Chelsea D.; Narayanan, Anoop; Majowicz, Sydney A.; Kwong, Elizabeth M.; McVicar, Rachael N.; Thacker, Bryan E.; Glass, Charles A.; Yang, Zhang; Torres, Jonathan L.; Golden, Gregory J.; Bartels, Phillip L.; Porell, Ryan; Garretson, Aaron F.; Laubach, Logan; Feldman, Jared; Yin, Xin; Pu, Yuan; Hauser, Blake; Caradonna, Timothy M.; Kellman, Benjamin P.; Martino, Cameron; Gordts, Philip L.S.M.; Chanda, Sumit K.; Schmidt, Aaron G.; Godula, Kamil; Leibel, Sandra L.; Jose, Joyce; Corbett, Kevin D.; Ward, Andrew B.; Carlin, Aaron F.; Esko, Jeffrey D.
title: SARS-CoV-2 Infection Depends on Cellular Heparan Sulfate and ACE2
date: 2020-09-14
journal: Cell
DOI: 10.1016/j.cell.2020.09.033
sha: 
doc_id: 256156
cord_uid: mywhe6w9

We show that SARS-CoV-2 spike protein interacts with both cellular heparan sulfate and angiotensin converting enzyme 2 (ACE2) through its Receptor Binding Domain (RBD). Docking studies suggest a heparin/heparan sulfate-binding site adjacent to the ACE2 binding site. Both ACE2 and heparin can bind independently to spike protein in vitro and a ternary complex can be generated using heparin as a scaffold. Electron micrographs of spike protein suggests that heparin enhances the open conformation of the RBD that binds ACE2. On cells, spike protein binding depends on both heparan sulfate and ACE2. Unfractionated heparin, non-anticoagulant heparin, heparin lyases, and lung heparan sulfate potently block spike protein binding and/or infection by pseudotyped virus and authentic SARS-CoV-2 virus. We suggest a model in which viral attachment and infection involves heparan sulfate-dependent enhancement of binding to ACE2. Manipulation of heparan sulfate or inhibition of viral adhesion by exogenous heparin presents new therapeutic opportunities.

The COVID-19 pandemic, caused by the novel respiratory coronavirus 2 (SARS-CoV-2), has swept across the world, resulting in serious clinical morbidities and mortality, as well as widespread disruption to all aspects of society. As of September 1, 2020, the virus has spread to 215 countries, causing more than 25.4 million confirmed infections and at least 851,000 deaths (World Health Organization). Current isolation/social distancing strategies seek to flatten the infection curve to avoid overwhelming hospitals and to give the medical establishment and pharmaceutical companies time to develop and test antiviral drugs and vaccines. Currently, only one antiviral agent, Remdesivir, has been approved for adult COVID-19 patients (Beigel et al., 2020) and vaccines may be 12-18 months away. Understanding the mechanism for SARS-CoV-2 infection and its mechanism of infection could reveal other targets to interfere with viral infection and spread.

The glycocalyx is a complex mixture of glycans and glycoconjugates surrounding all cells.

Given its location, viruses and other infectious organisms, must pass through the glycocalyx to engage receptors thought to mediate viral entry into host cells. Many viral pathogens have evolved to utilize glycans as attachment factors, which facilitates the initial interaction with host cells, including influenza virus, Herpes simplex virus, human immunodeficiency virus, and different coronaviruses (SARS-CoV-1 and MERS-CoV) (Cagno et al., 2019; Koehler et al., 2020; Stencel-Baerenwald et al., 2014) . Several viruses interact with sialic acids, which are located on the ends of glycans found in glycolipids and glycoproteins. Other viruses interact with heparan sulfate (HS) (Milewska et al., 2014) , a highly negatively charged linear polysaccharide that is attached to a small set of membrane or extracellular matrix proteoglycans (Lindahl et al., 2015) . In general, glycan-binding domains on membrane proteins of the virion envelope mediate initial attachment of virions to glycan receptors. Attachment in this way can lead to the engagement of protein receptors on the host plasma membrane that facilitate membrane fusion or engulfment and internalization of the virion. J o u r n a l P r e -p r o o f 5 Like other macromolecules, HS can be divided into subunits, which are operationally defined as disaccharides based on the ability of bacterial enzymes or nitrous acid to cleave the chain into disaccharide units (Esko and Selleck, 2002) . The basic disaccharide subunit consists of α1-4 linked D-glucuronic acid (GlcA) and α1-4 linked N-acetyl-D-glucosamine (GlcNAc), which undergo various modifications by sulfation and epimerization as the copolymer assembles on a limited number of membrane and extracellular matrix proteins (only 17 heparan sulfate proteoglycans are known) (Lindahl et al., 2015) . The variable length of the modified domains and their pattern of sulfation create unique motifs to which HS-binding proteins interact (Xu and Esko, 2014) . Different tissues and cell types vary in the structure of HS, and HS structure can vary between individuals and with age (de Agostini et al., 2008; Feyzi et al., 1998; Han et al., 2020; Ledin et al., 2004; Vongchan et al., 2005; Warda et al., 2006; Wei et al., 2011) . These differences in HS composition may contribute to the tissue tropism and/or host susceptibility to infection by viruses and other pathogens.

In this report, we show that the ectodomain of the SARS-CoV-2 spike (S) protein interacts with cell surface HS through the Receptor Binding Domain (RBD) in the S1 subunit. Binding of heparin to SARS-CoV-2 S protein shifts the structure to favor the RBD open conformation that binds ACE2. Spike binding to cells requires engagement of both cellular HS and ACE2, suggesting that HS acts as a coreceptor priming the spike for ACE2 interaction. Therapeutic unfractionated heparin (UFH), non-anticoagulant heparin and HS derived from human lung and other tissues blocks binding. UFH and heparin lyases also block infection of cells by S protein pseudotyped virus and authentic SARS-CoV-2. These findings identify cellular HS as a necessary co-factor for SARS-CoV-2 infection and emphasizes the potential for targeting S protein-HS interactions to attenuate virus infection. The trimeric S proteins from SARS-CoV-1 and SARS-CoV-2 viruses are thought to engage human ACE2 with one or more RBD in an "open" active conformation (Fig. 1A ) (Kirchdoerfer et al., 2018; Walls et al., 2020; Wrapp et al., 2020) . Adjacent to the ACE2 binding site and exposed in the RBD lies a group of positively-charged amino acid residues that represents a potential site that could interact with heparin or heparan sulfate ( Fig. 1A and Suppl. Fig. S1 ). We calculated an electrostatic potential map of the RBD (from PDB ID 6M17 (Yan et al., 2020) ), which revealed an extended electropositive surface with dimensions and turns/loops consistent with a heparin-binding site (Fig. 1B) (Xu and Esko, 2014) . Docking studies using a tetrasaccharide (dp4) fragment derived from heparin demonstrated preferred interactions with this electropositive surface, which based on its dimensions could accommodate a chain of up to 20 monosaccharides ( Fig. 1B and 1C ). Evaluation of heparin-protein contacts and energy contributions using the Molecular Operating Environment (MOE) software suggested strong interactions with the positively charged amino acids R346, R355, K444, R466 and possibly R509 (Figs. 1A, 1D, and 1E) . Other amino acids, notably F347, S349, N354, G447, Y449, and Y451, could coordinate the oligosaccharide through hydrogen bonds and hydrophobic interactions. Notably, the putative binding surface for oligosaccharides is adjacent to, but separate from the ACE2 binding site, suggesting that a single RBD could simultaneously bind both cell surface HS and the ACE2 protein receptor. The putative HS binding site is partially obstructed in the "closed" inactive RBD conformation, while fully exposed in the open state (Suppl. Fig. S1 ).

The amino acid sequence of S protein RBD of SARS-CoV-2 S is 73% identical to the RBD of SARS-CoV-1 S (Fig. 1F) , and these domains are highly similar in structure with an overall Cα r.m.s.d. of 0.929 Å (Fig. 1G) . However, an electrostatic potential map of the SARS-CoV-1 S J o u r n a l P r e -p r o o f 7 RBD does not show an electropositive surface like that observed in SARS-CoV-2 (Fig. 1H ).

Most of the positively charged residues comprising this surface are conserved between the two proteins, with the exception of SARS-CoV-2 K444 which is a threonine in SARS-CoV-1 (Fig.   1F ). Additionally, the other amino acid residues predicted to coordinate with the oligosaccharide are conserved with the exception of Asn354 in SARS-CoV-2, which is a negatively charged glutamate residue in SARS-CoV-1. SARS-CoV-1 has been shown to interact with cellular HS in addition to its entry receptors ACE2 and transmembrane protease, serine 2 (TMPRSS2) (Lang et al., 2011) . Our analysis suggests that the putative heparin-binding site in SARS-CoV-2 S may mediate an enhanced interaction with heparin or HS compared to SARS-CoV-1, and that this change evolved through as few as two amino acid substitutions, Thr Lys444 and Glu Asn354.

To test experimentally if the SARS-CoV-2 S protein interacts with heparin/HS, recombinant ectodomain and RBD proteins were prepared and characterized. Initial studies encountered difficulty in stabilizing the S ectodomain protein, a problem that was resolved by raising the concentration of NaCl to 0.3 M in HEPES buffer. Under these conditions, the protein could be stored at room temperature, 4 o C or at -80 o C for at least two weeks. SDS-PAGE showed that each protein was ~98% pure ( J o u r n a l P r e -p r o o f 8 Recombinant S ectodomain and RBD proteins were applied to a column of heparin-Sepharose. Elution with a gradient of sodium chloride showed that the RBD eluted at ~0.3 M NaCl, with a shoulder that eluted with higher salt (Fig. 2A) . Recombinant S ectodomain also bound to heparin-Sepharose, but it eluted across a broader concentration of NaCl. The elution profiles suggest that the preparations contained a population of molecules that bind to heparin, but that some heterogeneity in affinity for heparin occurs, which may reflect differences in glycosylation, oligomerization or the number of binding sites in the open conformation.

The RBD protein from SARS-CoV-2 also bound in a saturable manner to heparin-BSA immobilized on a plate (Fig. 2B ). The RBD domain from SARS-CoV-1 showed significantly reduced binding to heparin-BSA and a higher K D value (640 nM [95% C.I.; 282 -1852 nM] for SARS-CoV-1 RBD vs. 150 nM [95% C.I. 123 -173 nM]) for SARS-CoV-2 RBD), in accordance with the difference in electropositive potential in the proposed HS binding regions (Fig. 1H) . A monomeric form of SARS-CoV-2 S ectodomain protein also bound in a saturable manner to heparin immobilized on a plate (Suppl. Fig. S3A ). The trimeric protein bound to heparin-BSA with an apparent K D value of 3.8 nM [95% C.I. 3.1 -4.6 nM] (Fig. 2C ). Binding of recombinant S ectodomain, mutated to lock the RBDs into a closed (Mut2) or that favors an open (Mut7) conformation, showed that the heparin binding site in the RBD domain is accessible in both conformations (Fig. 2D ). However, the K D value for Mut7 is lower (4.6 nM [95% C.I. 3.8 -5.5 nM] vs. 9.9 nM [95% C.I. 8.7 -11.3 nM] for Mut2), which is in line with the partial obstruction of the site in the closed conformation (Suppl. Fig. S1 ). As expected, only trimer with an open RBD conformation bound to ACE2 (Fig. 2E ).

In contrast to spike protein, ACE2 did not bind to heparin-BSA (Fig. 2C) . ACE2 also had no effect on binding of S protein to heparin-BSA at all concentrations that were tested (Fig. 2C , inset). Biotinylated ACE2 bound to immobilized S protein (Suppl. Fig. S3B ) and a ternary complex of heparin, ACE2 and S protein could be demonstrated by titration of S protein bound to immobilized heparin-BSA with ACE2 (Fig. 2F ). Binding of ACE2 under these conditions J o u r n a l P r e -p r o o f 9 increased in proportion to the amount of S protein bound to the heparin-BSA. Collectively, these findings show that (i) spike protein can engage both heparin and ACE2 simultaneously and (ii) that the heparin binding site is somewhat occluded in the closed conformation, but it can still bind heparin albeit with reduced affinity.

The simultaneous binding of ACE2 to spike protein and heparin suggested the possibility that heparin binding might affect the conformation of the RBD, possibly increasing the open conformation that can bind ACE2. To explore this possibility, spike protein was mixed with ACE2 (6-fold molar ratio) with or without dp20 oligosaccharides derived from heparin (9-fold molar ratio). The samples were then stained and analyzed by transmission electron microscopy, and the images were deconvoluted and sorted into 3D reconstructions to determine the number of trimers with 0, 1, 2, or 3 bound ACE2 (Fig. 2G -H and Suppl. Fig. S3C-D) . The different populations were counted and the percentage of particles belonging to each 3D class was calculated. Two time points were evaluated after mixing ACE2 and trimeric S: at 15 min 29,600 and 31,300 particles were analyzed in the absence or presence of dp20 oligosaccharides, respectively; at 60 min, 17,000 and 21,000 particles were analyzed in absence or presence of dp20 oligosaccharides, respectively. At both time points, the presence of dp20 increased the total amount of ACE2 protein bound to spike . After 15 minutes in the absence of dp20 very few of the trimers had conformations with 1 or 2 bound ACE2 (5% each), whereas the inclusion of dp20 oligosaccharides greatly increased the proportion of trimers bearing one (37%) or two (21%) ACE2, with a proportional drop in the unbound conformers from 90% in the absence of heparin to 42% in its presence (Fig. 2G ). Extending the incubation to 60 minutes resulted in a mixture of trimers containing 1 (45%), 2 (11%) and 3 ACE2 (13%) in the absence of heparin. Inclusion of dp20 further increased the proportion of bound spike trimers bearing 2 (19%), and 3 (27%) ACE2 (Fig. 2H) . The imaging studies suggest that, under these J o u r n a l P r e -p r o o f 10 experimental conditions, heparin may stabilize the ACE2 interaction, increasing the proportion of spike bound to ACE2 as well as the occupancy of individual spikes.

The SARS-CoV-2 spike protein depends on cellular heparan sulfate for cell binding.

To extend these studies to HS on the surface of cells, S ectodomain protein was added to human H1299 cells, an adenocarcinoma cell line derived from Type 2 alveolar cells (Fig. 3A ).

Spike ectodomains bound to H1299 cells, with half-maximal binding achieved at ~75 nM.

Treatment of the cells with a mixture of heparin lyases (HSase), which degrades cell surface HS, dramatically reduced binding ( Fig 3A) . The S ectodomain also bound to human A549 cells, another Type 2 alveolar adenocarcinoma line, as well as human hepatoma Hep3B cells (Fig.   3B ). Removal of HS by enzymatic treatment dramatically reduced binding in both of these cell lines as well (Fig. 3B ). Recombinant RBD protein also bound to all three cell lines dependent on HS (Fig. 3C) . A melanoma cell line, A375, was tested independently and also showed HS dependent binding ( Fig 3D) . The extent of binding across the four cell lines varied ~4-fold. This variation was not due to differences in HS expression as illustrated by staining of cell surface HS with mAb 10E4, which recognizes a common epitope in HS ( We also measured binding of the S ectodomain and RBD proteins to a library of mutant Hep3B cells, carrying CRISPR/Cas9 induced mutations in biosynthetic enzymes essential for synthesizing HS (Anower et al., 2019) . Inactivation of EXT1, a subunit of the copolymerase required for synthesis of the backbone of HS, abolished binding to a greater extent than enzymatic removal of the chains with HSases ( Fig. 3F and Suppl. Fig. S4 ), suggesting that the HSase treatment may underestimate the dependence on HS. Targeting NDST1, a GlcNAc N-J o u r n a l P r e -p r o o f 11 deacetylase-N-sulfotransferase that N-deacetylates and N-sulfates N-acetylglucosamine residues, and HS6ST1 and HS6ST2, which introduces sulfate groups in the C6 position of glucosamine residues, significantly reduced binding (Figs. 3F and Suppl. Fig. S4 ). Although experiments with other sulfotransferases have not yet been done, the data suggests that the pattern of sulfation of HS affects binding to S and RBD.

To further examine how variation in HS structure affects binding, we isolated HS from human kidney, liver, lung and tonsil. The samples were depolymerized into disaccharides by treatment with HSases, and the disaccharides were then analyzed by LC-MS (Experimental Methods). The disaccharide analysis showed that lung HS has a larger proportion of Ndeacetylated and N-sulfated glucosamine residues (grey bars) and more 2-O-sulfated uronic acids (green bars) than HS preparations from the other tissues (Fig. 4A ). The different HS preparations also varied in their ability to block binding of RBD to H1299 cells (Fig. 4B ).

Interestingly, HS isolated from lung was more potent compared to kidney and liver HS, consistent with the greater degree of sulfation of HS from this organ (Suppl. Table 1 ). HS from tonsil was as potent as HS from lung, but the overall extent of sulfation was not as great, supporting the notion that the patterning of the sulfated domains in the chains may affect binding.

Unfractionated heparin is derived from porcine mucosa and possesses potent anticoagulant activity due to the presence of a pentasaccharide sequence containing a crucial 3-O-sulfated Nsulfoglucosamine unit, which confers high affinity binding to antithrombin. Heparin is also very highly sulfated compared to HS with an average negative charge of -3.4 per disaccharide (the overall negative charge density of typical HS is -1.8 to -2.2 per disaccharide). MST cells, which were derived from a murine mastocytoma, make heparin-like HS that lacks the key 3-O-sulfate group and anticoagulant activity (Gasimli et al., 2014; Montgomery et al., 1992) . The 12 anticoagulant properties of heparin can also be removed by periodate oxidation, which oxidizes the vicinal hydroxyl groups in the uronic acids, resulting in what is called "split-glycol" heparin (Casu et al., 2004) . All of these agents significantly inhibited binding of the S protein to H1299 and A549 cells ( Fig. 4C and 4D ) yielding IC 50 values in the range of 0.01-0.12 µg/ml (Suppl .   Table 1 ). Interestingly, the lack of 3-O-sulfation, crucial for the anticoagulant activity of heparin, had little effect on its inhibition of S binding. In contrast, CHO cell HS (containing 0.8 sulfates per disaccharide) only weakly inhibited binding (IC 50 values of 18 and 139 µg/ml for A549 and H1299, respectively) (Suppl. Table 1 ). These data suggest that inhibition by heparinoids is most likely charge dependent and independent of anticoagulant activity per se.

The experiments shown in Fig. 2G -H indicate that binding of heparin to spike protein can increase binding to ACE2. To explore if HS, ACE2 and spike interact at the cell surface, we investigated the impact of ACE2 expression on S protein cell binding. Initial attempts were made to measure ACE2 levels by Western blotting or flow cytometry with different mAbs and polyclonal antibodies, but a reliable signal was not obtained in any of the cell lines tested (A375, A549, H1299, and Hep3B). Nevertheless, expression of ACE2 mRNA was observed by RT-qPCR (Suppl. Fig. 5A ). Transfection of A375 cells with ACE2 cDNA resulted in robust expression of ACE2 (Fig. 5A) , resulting in an increase in S ectodomain protein binding by ~4fold (Fig. 5B) . Interestingly, the enhanced binding was HS-dependent, as illustrated by the loss of binding of S protein after HSase-treatment (Fig. 5B ). CRISPR/Cas9 mediated deletion of the B4GALT7 gene, which is required for glycosaminoglycan assembly (Suppl. Fig. S5B ), also reduced binding of spike protein (Fig. 5B ) despite the overexpression of ACE2 (Fig. 5A ). To explore the impact of diminished ACE2 expression, we examined spike protein binding to A549 cells and in two CRISPR/Cas9 gene targeted clones C3 and C6 bearing biallelic mutations in ACE2 (Suppl. Fig. S5C ). Binding of S ectodomain protein was greatly reduced in the ACE2 -/-J o u r n a l P r e -p r o o f 13 clones and the residual binding was sensitive to HSases (Fig. 5C ). These findings show that binding of spike protein on cells requires both HS and ACE2, consistent with the formation of a ternary complex (Figs. 2F-H).

Assays using purified components provide biochemical insights into binding, but they do not recapitulate the multivalent presentation of the S protein as it occurs on the virion membrane.

Thus, to extend these studies, pseudotyped vesicular stomatitis virus (VSV) was engineered to express the full-length SARS-CoV-2 S protein and GFP or luciferase to monitor infection. Vero E6 cells are commonly used in the study of SARS-CoV-2 infection, due to their high susceptibility to infection. Spike protein binding to Vero cells also depends on cellular HS as binding was sensitive to HSases, heparin and split-glycol heparin (Fig. 6A ). Interestingly, HSase treatment reduced binding to a lesser extent than the level of reduction observed in A549, Heparin very potently reduced infection more than ~4-fold at 0.5 µg/mL and higher concentrations (Fig. 6G) . In contrast, studies of SARS-CoV-1 S protein pseudotype virus showed that HSase-treatment actually increased SARS-CoV-1 infection by more than 2-fold, suggesting that HS might interfere with binding of SARS-CoV-1 in this cell line (Fig. 6H ).

Infection of H1299 and A549 cells by SARS-CoV-2 S pseudotype virus was too low to obtain J o u r n a l P r e -p r o o f 14 accurate measurements, but infection of Hep3B cells could be readily measured (Fig. 6I ).

HSase and mutations in EXT1 and NDST1 dramatically reduced infection 6-to 7-fold.

Inactivation of the 6-O-sulfotransferases had only a mild effect unlike its strong effect on S protein binding (Fig. 3F) , possibly due to the high valency conferred by multiple copies of S protein on the pseudovirus envelope. Hep3B cells were not susceptible to infection by SARS-CoV-1 S protein pseudotyped virus, but was infected by MERS-CoV S protein pseudotyped virus and infection was independent of HS (Suppl. Fig. S6 ).

Studies of pseudovirus were then extended to authentic SARS-CoV-2 virus infection using strain USA-WA1/2020. Infection of Vero E6 cells was monitored by double staining of the cells with antibodies against the SARS-CoV-2 nucleocapsid (N) and S proteins ( heparin inhibition (maroon and blue symbols). To rule out that the treatments caused a decrease in ACE2 expression or a reduction in cell viability, Vero cells were treated with heparin lyases and 100 µg/mL UFH, and ACE2 expression was measured by western blotting and cell viability by CellTiter-Blue® (Suppl. Fig. S7A -B) . No effect on ACE2 expression or cell viability was observed. These findings further emphasize the potential for using unfractionated heparin or other non-anticoagulant heparinoids to prevent viral attachment.

J o u r n a l P r e -p r o o f 15 These findings were then extended to Hep3B cells and mutants altered in HS biosynthesis using a viral plaque assay. Virus was added to wildtype, NDST1 -/and HS6ST1/2 -/cells for 2 hr, the virus was removed, and after 2 days incubation a serial dilution of the conditioned culture medium was added to monolayers of Vero E6 cells. The number of plaques were then quantitated by staining and visualization. As a control, culture medium from infected Vero E6 cells was tested, which showed robust viral titers. Hep3B cells also supported viral replication, but to a lesser extent than Vero cells. Inactivation of NDST1 in Hep3B cells abolished virus production, whereas inactivation of HS6ST1/2 -/reduced infection more mildly, ~3-fold (Fig. 7D) . HSase and UFH reduced infection more than 5-fold, but it had no effect on cell viability (Suppl. 

In this report, we provide compelling evidence that HS is a necessary host attachment factor that promotes SARS-CoV-2 infection of various target cells. The receptor binding domain of the SARS-CoV-2 S protein binds to heparin/HS, most likely through a docking site composed of positively charged amino acid residues aligned in a subdomain of the RBD that is separate from the site involved in ACE2 binding (Fig. 1) . Competition studies, enzymatic removal of HS, and genetic studies confirm that the S protein, whether presented as a recombinant protein (Figs. 2-J o u r n a l P r e -p r o o f 16 5), in a pseudovirus (Fig. 6) , or in authentic SARS-CoV-2 virions (Fig. 7) , binds to cell surface HS in a cooperative manner with ACE2 receptors. Mechanistically, binding of heparin/HS to spike trimers enhances binding to ACE2, likely increasing multivalent interactions with the target cell. This data provides crucial insights into the pathogenic mechanism of SARS-CoV-2 infection and suggests HS-spike protein complexes as a novel therapeutic target to prevent infection.

The glycocalyx is the first point of contact for all pathogens that infect animal cells, and thus it is not surprising that many viruses exploit glycans, such as HS, as attachment factors. For example, the initial interaction of herpes simplex virus with cells involves binding to HS chains on one or more HS proteoglycans (Shieh et al., 1992; WuDunn and Spear, 1989 ) through the interactions with the viral glycoproteins gB and gC. Viral entry requires the interaction of a specific structure in HS with a third viral glycoprotein, gD (Shukla et al., 1999) , working in concert with membrane proteins related to TNF/NGF receptors (Montgomery et al., 1996) .

Similarly, the human immunodeficiency virus binds to HS by way of the V3 loop of the viral glycoprotein gp120 (Roderiquez et al., 1995) , but infection requires the chemokine receptor CCR5 (Deng et al., 1996; Dragic et al., 1996) . Other coronaviruses also utilize HS, for example NL63 (HCoV-NL63) binds HS via the viral S protein in addition to ACE2 (Lang et al., 2011; Milewska et al., 2018; Milewska et al., 2014; Naskalska et al., 2019) . In these examples, initial tethering of virions to the host cell plasma membrane appears to be mediated by HS, but infection requires transfer to a proteinaceous receptor. The data presented here shows that SARS-CoV-2 requires HS in addition to ACE2. We imagine a model in which cell surface HS acts as a "collector" of the virus and a mediator of the RBD-ACE2 interaction, making viral infection more efficient. HS varies in structure across cell types and tissues, as well as with gender and age (de Agostini et al., 2008; Feyzi et al., 1998; Ledin et al., 2004; Vongchan et al., 2005; Warda et al., 2006; Wei et al., 2011) . Variation in competition by HS from different tissues supports this conclusion and raises the possibility that HS contributes to the tissue tropism and J o u r n a l P r e -p r o o f 17 the susceptibility of different patient populations, in addition to levels of expression of ACE2 .

Coronaviruses can utilize a diverse set of glycoconjugates as attachment factors. Human coronavirus OC43 (HCoV-OC43) and bovine coronavirus (BCoV) bind to 5-N-acetyl-9-Oacetylneuraminic acid (Hulswit et al., 2019; Tortorici et al., 2019) , middle east respiratory syndrome virus (MERS-CoV) binds 5-N-acetyl-neuraminic acid (Park et al., 2019) , and guinea fowl coronavirus binds biantennary di-N-acetyllactosamine or sialic acid capped glycans (Bouwman et al., 2019) . Whether SARS-CoV-2 S protein binds to sialic acid remains unclear.

Mapping the binding site for sialic acids in other coronavirus S proteins has proved elusive, but modeling studies suggest a location distinct from the HS binding site shown in Fig. 1 (Park et al., 2019; Tortorici et al., 2019) . The S protein in murine coronavirus contains both a hemagglutinin domain for binding and an esterase domain that cleaves sialic acids that aids in the liberation of bound virions (Rinninger et al., 2006; Smits et al., 2005) . Whether SARS-CoV-2 S protein, another viral envelope protein, or a host protein contributes to HS-degrading activity to aid in the release of newly made virions is unknown.

The repertoire of proteins in organisms that bind to HS make up the so called "HS interactome" and consists of a variety of different HS-binding proteins (HSBPs) (Xu and Esko, 2014) . Unlike lectins that have a common fold that helps define the glycan binding site, HSBPs do not exhibit a conserved motif that allows accurate predictions of binding sites based on primary sequence. Instead, the capacity to bind heparin appears to have emerged through convergent evolution by juxtaposition of several positively charged amino acid residues arranged to accommodate the negatively charged sulfate and carboxyl groups present in the polysaccharide, and hydrophobic and H-bonding interactions stabilize the association. The RBD domains from the SARS-CoV-1 and SARS-CoV-2 S proteins are highly similar in structure (Fig.   1G ), but the electropositive surface in SARS-CoV-1 S RBD is not as pronounced in SARS-CoV-2 S RBD (Fig. 1H ). In accordance with this observation, recombinant RBD protein from SARS-J o u r n a l P r e -p r o o f 18 CoV-2 showed significantly higher binding to heparin-BSA, compared to RBD from SARS-CoV-1 (Fig. 2B) . A priori we predicted that the evolution of the HS binding site in the SARS-CoV-2 S protein might have occurred by the addition of arginine and lysine residues to its ancestor, SARS-CoV-1. Instead, we observed that four of the six predicted positively charged residues that make up the heparin-binding site are present in SARS-CoV-1 as well as most of the other amino acid residues predicted to interact with heparin ( Fig. 1) . SARS-CoV-1 has been shown to interact with cellular HS in addition to its entry receptors ACE2 and transmembrane protease, serine 2 (TMPRSS2) (Lang et al., 2011) . Our analysis suggests that the putative heparinbinding site in SARS-CoV-2 S may mediate an enhanced interaction with heparin compared to SARS-CoV-1, and that this change evolved through as few as two amino acid substitutions, Thr444Lys and Glu354Asn. Further studies are underway to define the amino acid residues in the combining site for heparin/HS to test this hypothesis.

The ability of heparin and HS to compete for binding of the SARS-CoV-2 S protein to cell surface HS and the inhibitory activity of heparin towards infection of pseudovirus and authentic SARS-CoV-2 illustrates the therapeutic potential of agents that target the virus-HS interaction to control infection and transmission of SARS-CoV-2. There is precedent for targeting proteinglycan interactions as therapeutic agents. For example, Tamiflu targets influenza neuraminidase, thus reducing viral transmission, and sialylated human milk oligosaccharides can block sialic acid-dependent rotavirus attachment and subsequent infection in infants (Hester et al., 2013; von Itzstein, 2007) . COVID-19 patients typically suffer from thrombotic complications ranging from vascular micro-thromboses, venous thromboembolic disease and stroke and often receive unfractionated heparin or low molecular weight heparin (Thachil, 2020) .

The findings presented here and elsewhere suggest that both of these agents can block viral infection (Courtney Mycroft-West, 2020; Kim et al., 2020; Liu et al., 2020; Mycroft-West et al., 2020; Tandon et al., 2020; Wu et al., 2020) . Effective anticoagulation is achieved with plasma levels of heparin of 0.3-0.7 units/ml. This concentration is equivalent to 1.6-4 µg/ml heparin (assuming that the activity of UFH is 180 Units/mg). Although this is sufficient to block spike protein binding to cells (Fig. 4) , it would not be expected to prevent viral infection, but it should attenuate infection depending on the viral load (Fig. 7) . The anticoagulant activity of heparin, which is typically absent in HS, is not critical for its antiviral activity based on the observation that MST derived heparin and split-glycol heparin is nearly as potent as therapeutic heparin ( Figs. 4 and 6) . Additional studies are needed to address the potential overlap in the dose response profiles for heparin as an anticoagulant and antiviral agent and the utility of nonanticoagulant heparins. Antibodies directed to heparan sulfate or the binding site in the RBD might also prove useful for attenuating infection.

In conclusion, this work revealed HS as a novel attachment factor for SARS-CoV-2 and suggests the possibility of using HS mimetics, HS degrading lyases, and metabolic inhibitors of HS biosynthesis for the development of therapy to combat COVID-19. 

Further information and request for resources should be directed to the Lead Contact, Thomas Mandel Clausen (tmandelclausen@health.ucsd.edu)

All developed SARS-CoV-2 expression plasmids produced in this study can be made available upon request to the Lead Contact.

J o u r n a l P r e -p r o o f 28 This study did not generate any unique datasets or code.

Cell Lines NCI-H1299, A549, Hep3B, A375 and Vero E6 cells were from the American Type Culture Collection (ATCC). NCI-H1299 and A549 cells were grown in RPMI medium, whereas the other lines were grown in DMEM. Hep3B cells carrying mutations in HS biosynthetic enzymes were previously derived from the parent Hep3B line as described (Anower et al., 2019) . All cell media were supplemented with 10% (v/v) FBS, 100 IU/ml of penicillin and 100 µg/ml of streptomycin sulfate, and the cells were grown under an atmosphere of 5% CO 2 and 95% air. Cells were passaged at ~80% confluence and seeded as explained for the individual assays. Protein was produced in ExpiCHO or HEK293-6E cells that were acquired from Thermo Fisher and grown according to the manufacturer's specifications.

Human bronchial epithelial cells were acquired from Lonza. They were cultured in PneumaCult-Ex Plus Medium or to PneumaCult-ALI Medium according to the manufacturer's instructions (StemCell Technologies). Specific details on the culture methods are described in the Methods section.

The collection of human tissue in this study abided by the Helsinki Principles and the 

An electrostatic potential map of the SARS-CoV-2 spike protein RBD domain was generated from a crystal structure (PDB:6M17) and visualized using Pymol (version 2.0.6 by Schrödinger).

A dp4 fully sulfated heparin fragment was docked to the SARS-CoV-2 spike protein RBD using the ClusPro protein docking server (https://cluspro.org/login.php) (Kozakov et al., 2013; Kozakov et al., 2017; Vajda et al., 2017) . Heparin-protein contacts and energy contributions were evaluated using the Molecular operating environment (MOE) software (Chemical Computing Group).

Recombinant SARS-CoV-2 spike protein, encoding residues 1-1138 (Wuhan-Hu-1;

GenBank: MN908947.3) with proline substitutions at amino acids positions 986 and 987, a "GSAS" substitution at the furin cleavage site (amino acids 682-682), TwinStrepTag and His 8x , was produced in ExpiCHO cells by transfection of 6 x10 6 cells/ml at 37 ºC with 0.8 µg/ml of plasmid DNA using the ExpiCHO expression system transfection kit in ExpiCHO Expression Medium (ThermoFisher). One day later the cells were refed, then incubated at 32 ºC for 11 days. The conditioned medium was mixed with cOmplete EDTA-free Protease Inhibitor (Roche). 

Samples of the recombinant trimeric spike protein ectodomain were diluted to 0.03 mg/ml in 1X TBS pH 7.4. Carbon coated copper mesh grids were glow discharged and 3 µL of the diluted sample was placed on a grid for 30 sec then blotted off. Uniform stain was achieved by depositing 3 µL of uranyl formate (2%) on the grid for 55 sec and then blotted off. Grids were transferred to a Thermo Fisher Morgagni operating at 80 kV. Images at 56,000 magnification J o u r n a l P r e -p r o o f 31 were acquired using a MegaView 2K camera via the RADIUS software. A dataset of 138 micrographs at 52,000x magnification and -1.5 µm defocus was collected on a FEI Tecnai Spirit (120keV) with a FEI Eagle 4k by 4k CCD camera. The pixel size was 2.06 Å per pixel and the dose was 25 e − /Å 2 . The Leginon (Suloway et al., 2005) software was used to automate the data collection and the raw micrographs were stored in the Appion (Lander et al., 2009) database.

Particles on the micrographs were picked using DogPicker , stack with a box size of 200 pixels, and 2D classified with RELION 3.0 (Scheres, 2012) .

Secreted human ACE2 was transiently produced in suspension HEK293-6E cells. A plasmid encoding residues 1−615 of ACE2 with a C-terminal HRV-3C protease cleavage site, a

TwinStrepTag and an His 8x Tag was a gift from Jason S. McLellan, University of Texas at

Austin. Briefly, 100 mL of HEK293-6E cells were seeded at a cell density of 0.5 × 10 6 cells/ml 24 hr before transfection with polyethyleneimine (PEI). For transfection, 100 µg of the ACE2 plasmid and 300 µg of PEI (1:3 ratio) were incubated for 15 min at room temperature.

Transfected cells were cultured for 48 hr and fed with 100 mL fresh media for additional 48 hr before harvest. Secreted ACE2 were purified from culture medium by Ni-NTA affinity chromatography (Qiagen). Filtered media was mixed 3:1 (v/v) in 4X binding buffer (100 mM Tris-HCl, pH 8,0, 1,2 M NaCl) and loaded on to a self-packed column, pre-equilibrated with washing buffer (25 mM Tris-HCl, pH 8, 0.3 M NaCl, 20 mM imidazole). Bound protein was washed with buffer and eluted with 0.2 M imidazole in washing buffer. The protein containing fractions were identified by SDS-PAGE.

J o u r n a l P r e -p r o o f SARS-CoV-2 spike protein in dPBS was applied to a 1-ml HiTrap heparin-Sepharose column (GE Healthcare). The column was washed with 5 ml of dPBS and bound protein was eluted with a gradient of NaCl from 150 mM to 1 M in dPBS.

For binding studies, recombinant spike protein and ACE2 was conjugated with EZ-Link TM Sulfo-NHS-Biotin (1:3 molar ratio; Thermo Fisher) in Dulbecco's PBS at room temperature for 30 min. Glycine (0.1 M) was added to quench the reaction and the buffer was exchanged for PBS using a Zeba spin column (Thermo Fisher).

Heparin ( and incubated with S protein (100 nM). ACE2 binding was measured to bound spike protein as described above.

Mixtures of stabilized (Mut7) spike protein, 6x molar excess soluble ACE2 ectodomain, with or without 9x molar excess an icosasaccharide (dp20) fragment derived from heparin were incubated at 4°C for 15 min or 1 hr. Samples were diluted to 0.02 mg/mL with respect to spike protein in 1X PBS pH 7.4. Carbon coated copper mesh grids were glow discharged at 20 mA for 30 s and 3 µL sample was applied for 20 s and blotted off. Grids were washed five times in 10 µL 1X TBS pH 7.4 for 15 sec then stained and blotted twice with 3 µL 2% uranyl formate for 15 sec. Grids were imaged with an FEI Tecnai Spirit (120 keV) or FEI Tecnai F20 (200 keV) with an FEI Eagle CCD (4k) camera. Data were collected on the FEI Tecnai F20 at 62,000x magnification, -1.5 µm defocus with a pixel size of 1.77 Å per pixel. These datasets employed a box size of 256 and comprised 167 to 331 micrographs. Data were collected on the FEI Tecnai Spirit as described above. Data collection on both microscopes was automated through Leginon (Suloway et al., 2005) . Stored in the Appion (Lander et al., 2009 ) database, and particles were picked with DoG Picker . Particles were 2D classified with RELION 3.0 J o u r n a l P r e -p r o o f 34 (Scheres, 2012) . Trimeric 2D classes were selected for iterative 3D classification with RELION 3.0. Classifications were performed until 3D classes demonstrated ACE2 occupancy throughout the relevant threshold-level of the spike protein as visualized using ChimeraX (Goddard et al., 2018) . Particle counts of final 3D classes were obtained with RELION 3.0 (Scheres, 2012) and

the percentages of particles bound to 0, 1, 2, or 3 ACE2 were calculated and visualized in GraphPad Prism 8.

Cells at 50-80% confluence were lifted with PBS containing 10 mM EDTA (Gibco) and 

Fresh human tissue was washed in PBS, frozen, and lyophilized. The dried tissue was crushed into a fine powder, weighed, resuspended in PBS containing 1 mg/mL Pronase with 90% ethanol (Esko, 1993) .

For HS quantification and disaccharide analysis, purified HS was digested with a mixture of heparin lyases I-III (2 mU each) for 2 hr at 37 °C in 40 mM ammonium acetate buffer containing 

The ACE2 expression plasmid (Addgene, plasmid #1786) (Li et al., 2003) qPCR mRNA was extracted from the cells using TRIzol (Invitrogen) and chloroform and purified using the RNeasy Kit (Qiagen). cDNA was synthesized from the mRNA using random primers and the SuperScript III First-Strand Synthesis System (Invitrogen). SYBR Green Master Mix (Applied Biosystems) was used for qPCR following the manufacturer's instructions, and the expression of TBP was used to normalize the expression of ACE2 between the samples. The qPCR primers used were as follows: ACE2 (human) forward: 5' -CGAAGCCGAAGACCTGTTCTA -3' and reverse: 5' -GGGCAAGTGTGGACTGTTCC -3';

and TBP (human) forward: 5' -AACTTCGCTTCCGCTGGCCC -3' and reverse: 5' -GAGGGGAGGCCAAGCCCTGA -3'.

To generate the Cas9 lentiviral expression plasmid, 2.5 x 10 6 HEK293T cells were seeded to a 10-cm diameter plate in DMEM supplemented with 10% FBS. The following day, the cells J o u r n a l P r e -p r o o f 38 were co-transfected with the psPAX2 packaging plasmid (Addgene, plasmid #12260), pMD2.g envelope plasmid (Addgene, plasmid #12259), and lenti-Cas9 plasmid (Addgene, plasmid #52962) (Sanjana et al., 2014) in DMEM supplemented with Fugene6 (30µL in 600µL DMEM).

Media containing the lentivirus was collected and used to infect A549 WT and A375 WT cells,

which were subsequently cultured with 5 µg/mL and 2 µg/mL blasticidin, respectively, to select for stably transduced cells. A single guide RNA (sgRNA) targeting ACE2 (5'-TGGATACATTTGGGCAAGTG -3') and one targeting B4GALT7 (5'-TGACCTGCTCCCTCTCAACG-3') was cloned into the lentiGuide-Puro plasmid (Addgene plasmid #52963) following published procedure (Sanjana et al., 2014) . The lentiviral sgRNA construct was generated in HEK293T cells, using the same protocol as for the Cas9 expression plasmid, and used to infect A549-Cas9 and A375-Cas9 cells to generate CRISPR knockout mutant cell lines. After infection, the cells were cultured with 2 µg/mL puromycin to select for cells with stably integrated lentivirus. After 7 d, the cells were serially diluted into 96-well plates.

Single colonies where expanded and DNA was extracted using the DNeasy blood and tissue DNA isolation kit (Qiagen). Proper editing was verified by sequencing (Genewiz Inc.) and gene analysis using the online ICE tool from Synthego (Suppl. Fig. 5 ).

Vesicular Stomatitis Virus (VSV) pseudotyped with spike proteins of SARS-CoV-2 were generated according to a published protocol (Whitt, 2010) . Briefly, HEK293T, transfected to express full length SARS-CoV-2 spike proteins, were inoculated with VSV-G pseudotyped ∆Gluciferase or GFP VSV (Kerafast, MA). After 2 hr at 37°C, the inoculum was removed and cells were refed with DMEM supplemented with 10% FBS, 50 U/mL penicillin, 50 µg/mL streptomycin, and VSV-G antibody (I1, mouse hybridoma supernatant from CRL-2700; ATCC).

Pseudotyped particles were collected 20 hr post-inoculation, centrifuged at 1,320 × g to remove cell debris and stored at −80°C until use. Briefly, 100 µL of luciferin lysis solution was added to the cells and incubated for 5 min at room temperature. The solution was transferred to a black 96-well plate and luminescence was detected using an EnSpire multimodal plate reader (Perkin Elmer). Data analysis and statistical analysis was performed in Prism 8. Fluor 594 labeling kits (Invitrogen), respectively. Zombie UV™ was used to gate for live cells in the analysis. Cells were then analyzed using an MA900 Cell Sorter (Sony).

for 4 days. Fresh medium, 100 µl in the apical chamber and 500 µl in the basal chamber, was added daily. At day 7, the medium in the apical chambers was removed, and the basal chambers were changed every 2-3 days with apical washes with PBS every week for 28 days.

The apical side of the HBEC ALI culture was gently washed three times with 200 µl of phosphate buffered saline without divalent cations (PBS-/-). Heparinase was added to the apical side for half an hour prior to infection. An MOI of 0.5 of authentic SARS-CoV-2 live virus (USA-WA1/2020 (BEI Resources, #NR-52281)) in 100 µl total volume of PBS was added to the apical chamber with either DMSO, Heparinase (2.5mU/ml heparin lyase II, and 5mU/ml heparin lyase III (IBEX)) or 100ug/ml of Unfractionated Heparin. Cells were incubated at 37C and 5% CO2 for 4 hours. Unbound virus was removed, the apical surface was washed and the compounds were re-added to the apical chamber. Cells were incubated for another 20 hours at 37C and 5% CO2.

After inoculation, cells were washed once with PBS-/-and 100 µl TrypLE (ThermoFisher) was added to the apical chamber then incubated for 10 min in the incubator. Cells were gently pipetted up and down and transferred into a sterile 15 ml conical tube containing neutralizing medium of DMEM + 3% FBS. TrypLE was added again for 3 rounds of 10 minutes for a total of 30 min to clear transwell membrane. Cells were spun down and resuspended in PBS with Zombie UV viability dye for 15 min in room temp. Cells were washed once with FACS buffer then fixed in 4% PFA for 30 min at room temp. PFA was washed off and cells were resuspended in PBS. Zombie UV™ was used to gate for live cells in the analysis. Infection was analyzed by flow cytometry as explained above.

Cell viability was assessed using the CellTiter-Blue® assay (Promega). Briefly, Vero cells were seeded into a 96 well plate. The cells were treated with HSase mix (2.5 mU/ml HSase II, and 5 mU/ml HSase III; IBEX) or 100 µg/mL UFH for 16 hrs. The viability of the cells using CellTiter-Blue® was measured according to the manufacturers protocol. Briefly, the J o u r n a l P r e -p r o o f 42 CellTiter-Blue® reagent was added directly to the cell culture and the cells were incubated overnight. Fluorescence was read at excitation 560nm and emission 590nm, using an EnSpire multimodal plate reader (Perkin Elmer). Data analysis was performed in Prism. The human

Bronchial epithelial cells were grown at an air-liquid interface as explained above. Cell viability after treatment with HSase mix (2.5 mU/ml HSase II, and 5 mU/ml HSase III; IBEX) or 100 µg/mL UFH for 16 hrs was measured by adding CellTiter-Blue® reagent directly to the transwell inserts and developed as explained above.

All statistical analyses were performed in Prism 8 (Graphpad). All experiments were performed in triplicate and repeated as indicated in the figure legends. Data was analyzed statistically using unpaired T-tests when two groups were being compared or by one-way ANOVA without post-hoc correction for multiple comparisons. IC 50 values and confidence intervals were determined using non-linear regression using the inhibitor vs. response least squares fit algorithm. The error bars in the figures refer to mean plus standard deviation (SD) values. The specific statistical tests used are listed in the figure legends and in the methods section.

Experiments were evaluated by statistical significance according to the following scheme; ns: p > 0.05, *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001, ****: p ≤ 0.0001.

After 48 hr, cell culture supernatants were collected and stored at -80°C. Virus titers were determined by plaque assays on Vero E6 monolayers

Greiner bio-one, #662160) and rocked for 1 hr at room temperature. The cells were subsequently overlaid with MEM containing 1% cellulose

The plaques were visualized by fixation of the cells with a mixture of 10% formaldehyde and 2% methanol (v/v in water) for 2 hr. The monolayer was washed once with PBS and stained with 0.1% Crystal Violet (Millipore Sigma # V5265) prepared in 20% ethanol

The Pennsylvania State University, following the guidelines approved by the Institutional Biosafety Committees. Human bronchial epithelial cell air-liquid interface generation and infection Human Bronchial Epithelial Cells (HBECs, Lonza) were cultured in T75 flasks in

Plus Medium according to manufacturer instructions (StemCell Technologies)

To generate air-liquid interface (ALI) cultures, HBECs were plated on collagen I-coated 24 well transwell inserts with a 0.4-micron pore size (Costar, Corning) at 5x10 4 cells/ml. Cells were maintained for 3-4 days in PneumaCult-Ex Plus Medium until confluence, then changed to PneumaCult-ALI Medium

Triglyceride-rich lipoprotein binding and uptake by heparan sulfate proteoglycan receptors in a CRISPR/Cas9 library of Hep3B mutants

Remdesivir for the Treatment of Covid-19 -Preliminary Report

Guinea Fowl Coronavirus Diversity Has Phenotypic Consequences for Glycan and Tissue Binding

Heparan Sulfate Proteoglycans and Viral Attachment: True Receptors or Adaptation Bias? Viruses 11

Undersulfated and glycol-split heparins endowed with antiangiogenic activity

The 2019 coronavirus (SARS-CoV-2) surface protein (Spike) S1 Receptor Binding Domain undergoes conformational change upon heparin binding

Identification of a major co-receptor for primary isolates of HIV-1

HIV-1 entry into CD4+ cells is mediated by the chemokine receptor CC-CKR-5

Special considerations for proteoglycans and glycosaminoglycans and their purification

Order out of chaos: Assembly of ligand binding sites in heparan sulfate

Age-dependent modulation of heparan sulfate structure and function

Bioengineering murine mastocytoma cells to produce anticoagulant heparin

UCSF ChimeraX: Meeting modern challenges in visualization and analysis

Structural analysis of urinary glycosaminoglycans from healthy human subjects

Human milk oligosaccharides inhibit rotavirus infectivity in vitro and in acutely infected piglets

Human coronaviruses OC43 and HKU1 bind to 9-O-acetylated sialic acids via a conserved receptor-binding site in spike protein domain A

Loss of Bcl-6-Expressing T Follicular Helper Cells and Germinal Centers in COVID-19

Stabilized coronavirus spikes are resistant to conformational changes induced by receptor recognition or proteolysis

Initial Step of Virus Entry: Virion Binding to Cell-Surface Glycans

How good is automated protein docking?

The ClusPro web server for protein-protein docking

Appion: an integrated, database-driven pipeline to facilitate EM image processing

Inhibition of SARS Pseudovirus Cell Entry by Lactoferrin Binding to Heparan Sulfate Proteoglycans

Evolutionary differences in glycosaminoglycan fine structure detected by quantitative glycan reductive isotope labeling

Heparan sulfate structure in mice with genetically modified heparan sulfate production

Assessing ACE2 expression patterns in lung tissues in the pathogenesis of COVID-19

Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus

Proteoglycans and Sulfated Glycosaminoglycans

SARS-CoV-2 spike protein binds heparan sulfate in a length-and sequence-dependent manner

Entry of Human Coronavirus NL63 into the Cell

Human coronavirus NL63 utilizes heparan sulfate proteoglycans for attachment to target cells

Stable heparin-producing cell lines derived from the Furth murine mastocytoma

Herpes simplex virus-1 entry into cells mediated by a novel member of the TNF/NGF receptor family

Heparin inhibits cellular invasion by SARS-CoV-2: structural dependence of the interaction of the surface protein (spike) S1 receptor binding domain with heparin

Membrane Protein of Human Coronavirus NL63 Is Responsible for Interaction with the Adhesion Receptor

Structures of MERS-CoV spike glycoprotein in complex with sialoside attachment receptors

Localisation and distribution of O-acetylated N-acetylneuraminic acids, the endogenous substrates of the hemagglutinin-esterases of murine coronaviruses, in mouse tissue

Mediation of human immunodeficiency virus type 1 binding by interaction of cell surface heparan sulfate proteoglycans with the V3 region of envelope gp120-gp41

Improved vectors and genome-wide libraries for CRISPR screening

RELION: implementation of a Bayesian approach to cryo-EM structure determination

Cell surface receptors for herpes simplex virus are heparan sulfate proteoglycans

A novel role for 3-O-sulfated heparan sulfate in herpes simplex virus 1 entry

Nidovirus sialate-O-acetylesterases: evolution and substrate specificity of coronaviral and toroviral receptor-destroying enzymes

The sweet spot: defining virus-sialic acid interactions

Automated molecular microscopy: the new Leginon system

Effective Inhibition of SARS-CoV-2 Entry by Heparin and Enoxaparin Derivatives. bioRxiv

The versatile heparin in COVID-19

Structural basis for human coronavirus attachment to sialic acid receptors

The war against influenza: discovery and development of sialidase inhibitors

Structural characterization of human liver heparan sulfate

DoG Picker and TiltPicker: software tools to facilitate particle selection in single particle electron microscopy

Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Isolation and characterization of heparan sulfate from various murine tissues

Site-specific glycan analysis of the SARS-CoV-2 spike

A comprehensive compositional analysis of heparin/heparan sulfate-derived disaccharides from human serum

Generation of VSV pseudotypes using recombinant DeltaG-VSV for studies on virus entry, identification of entry inhibitors, and immune responses to vaccines

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

Vaccines and Therapies in Development for SARS-CoV-2 Infections

Initial interaction of herpes simplex virus with cells is binding to heparan sulfate

Demystifying heparan sulfate-protein interactions

Structural basis for the recognition of SARS-CoV-2 by full-length human ACE2

CoV-2 spike protein interacts with heparan sulfate and ACE2 through the RBD • Heparan sulfate promotes Spike-ACE2 interaction • SARS-CoV-2 infection is co-dependent on heparan sulfate and ACE2

• Heparin and non-anticoagulant derivatives block SARS-CoV-2 binding and infection In brief

provide evidence that heparin sulfate is a necessary co-factor for SARS-CoV-2 infection. They show that heparin sulfate interacts with the Receptor Binding Domain of the SARS-CoV-2 spike glycoprotein

We thank Scott Selleck (The Pennsylvania State University), Eugene Yeo (UC San Diego), John Guatelli (UC San Diego), Mark Fuster (UC San Diego) and Stephen Schoenberger (La Jolla Institute for Immunology) for many helpful discussions, and Annamaria Naggi and Giangiacomo Torri from the Ronzoni Institute for generously providing split-glycol heparin. This