Keywords heparin - coronavirus - SARS-CoV-2 - COVID-19 - spike - S1 - RBD - circular dichroism
- surface plasmon resonance - molecular modelling
Introduction
The evolution of multicellular animals required the emergence of networks of interacting
extracellular molecules to provide sophisticated communication between cells, underpinning
the transition of the unit of natural selection from the individual cell to the entire
organism. It is likely that the interactions of the glycosaminoglycans (GAGs) of the
peri- and extracellular matrix with proteins were amongst the earliest such molecular
networks. For example, choanoflagellates, considered to be extant unicellular descendants
of the last common unicellular ancestor of multicellular animals, produce GAGs[1 ]
[2 ]
[3 ] and interactions between their GAGs and a protein from bacterial prey regulate their
reproduction.[1 ]
[3 ] In mammals, the GAG heparan sulfate (HS) has the lion's share of protein partners;
over 830 extracellular regulatory proteins that bind HS have now been identified.[2 ]
[4 ]
[5 ]
[6 ] Numerous pathogens, including parasites, bacteria and viruses exploit the dependence
of multicellular organisms on protein–GAG interactions by producing proteins that
also bind GAGs.[6 ]
[7 ] The pathogen protein–host GAG interactions are integral to infection, as they are
documented to be involved in decoying host cell communication, pathogen adhesion and
cell entry.[8 ]
[9 ]
The binding specificity of the interactions of proteins and GAGs, such as HS, depends
largely on the complementary spatial disposition of basic groups on the protein and
of sulfate and carboxyl groups on the polysaccharide.[4 ]
[6 ]
[7 ]
[10 ] HS biosynthesis provides a large repertoire of sulfated saccharide sequences and,
with sulfation affecting the three-dimensional structure of sugars, a degree of specificity
is achieved. This enables cells to differentially bind and respond to extracellular
regulatory proteins. The need to accommodate hundreds of protein partners and the
redundancy inherent in HS binding sites arising from its biosynthesis makes it difficult
to envisage how an organism could enter an evolutionary race with a pathogen by altering
protein-binding structures in the polysaccharide. In this light, HS represents an
Achilles' heel of organisms with respect to pathogen invasion.
It is, therefore, perhaps not surprising that the list of pathogens exploiting HS
is extensive. This includes viruses. Importantly, there appears to be an overlap between
the site(s) on viral proteins required for interaction with cell receptors and cell
entry, and the site(s) that bind HS. Thus, there is extensive evidence to show that
heparin, some of its derivatives and HS inhibit infectivity of a broad spectrum of
viruses, including Coronaviridae and the severe acute respiratory syndrome (SARS)-associated coronavirus isolate HSR1,[11 ] flaviviruses,[12 ]
[13 ] herpes,[14 ] influenza[15 ] and human immunodeficiency virus.[16 ]
[17 ] Not only does heparin inhibit infectivity of coronavirus isolate HSR1,[11 ] but there is considerable sequence homology between the receptor-binding domain
(RBD) of this Coronaviridae and the spike (S1) protein RBD (S1 RBD) of SARS coronavirus-2 (SARS-Cov-2), with
basic patches that are likely to interact with heparin overlapping the angiotensin-converting
enzyme 2 (ACE2) binding site.
Therefore, we investigated the hypothesis that heparin would both bind the S1 RBD
of SARS-CoV-2 and that this interaction would inhibit viral infectivity, using a combination
of SARS-CoV-2 plaque assays in Vero cells, surface plasmon resonance (SPR) to establish
binding and circular dichroism (CD) spectroscopy to monitor structural changes in
the protein following binding. The data demonstrate that heparin does indeed inhibit
SARS-CoV-2 infection of Vero cells, and more effectively than for SARS-CoV. Moreover,
the S1 RBD binds heparin, as well as the clinical low-molecular-weight heparin, enoxaparin.
These interactions elicit a conformational change in the protein; a hexasaccharide
being the shortest effective oligosaccharide. The availability, low cost and clinical
grade of heparin, HS and derivatives, as well as the fact that they are safe to administer
to patients in critical care, make these polysaccharides attractive first-line therapeutic
candidates for viral diseases such as COVID-19 and could, perhaps, contribute to a
prophylactic control strategy.
Methods
Vero Cell Culture and Assays
African green monkey Vero kidney epithelial cells were purchased from ECACC and maintained
at 50 to 75% confluence in DMEM (Gibco, United Kingdom), supplemented with 10% (v/v)
foetal calf serum (SLS, United Kingdom), 20 mM L-glutamine (Gibco, United Kingdom),
100 U mL−1 penicillin-G (Gibco, United Kingdom) and 100 U mL−1 streptomycin sulfate (Gibco, United Kingdom), with 5% CO2 at 37°C.
For cellular invasion assays, Vero cells were plated at 2.5 × 105 cells per well in 24-well plates and incubated with EMEM (Sigma-Aldrich, Italy) supplemented
with 10% (v/v) foetal calf serum (complete medium). Twenty four hours later, cells
were incubated with porcine mucosal heparin (6.25–200 μg mL−1 ; Celsus Laboratories, Cincinnati, United States) in 300 µL of complete medium, 1 hour
prior to infection and then incubated with virus solution containing 50 plaque forming
units (PFUs) of either Italy/UniSR1/2020 isolate (GISAID accession ID: EPI_ISL_413489[18 ] or, SARS-CoV HSR-1 (EID 2004). After incubation for 1 hour at 37°C, supernatants
were discarded and 500 µL of 1% (w/v) methylcellulose (Sigma-Aldrich, Italy) overlay
(in complete medium) were added to each well. After 3 days, cells were fixed using
a 6% (v/v) formaldehyde:phosphate-buffered saline (PBS) solution and stained with
1% (w/v) crystal violet (Sigma-Aldrich, Italy) in 70% (v/v) methanol (Sigma-Aldrich,
Italy). The plaques were counted using a stereoscopic microscope (SMZ-1500, Nikon).
For RBD-binding assays, Vero cells were plated into 96-well cell culture plates at
1,000 cells per well in 100 µL of in DMEM supplemented with 10% (v/v) foetal calf
serum, 20 mM L-glutamine, 100 U mL−1 penicillin-G and 100 U mL−1 streptomycin sulfate, and were allowed to adhere overnight. Medium was aspirated
prior to washing thrice with 200 µL of 1× PBS (CMF-PBS; Ca2+ and Mg2+ free; Lonza, United Kingdom). Cells were fixed with 100 µL of 10% neutral buffered
formalin (Thermo Fisher, United Kingdom) for 10 minutes at room temperature. Wells
were washed thrice with 200 µL of CMF-PBS. Then 100 µL CMF- PBS was added to each
well and plates were stored at 4°C until use. Prior to binding assays, wells were
blocked with 200 µL CMF-PBS + 1% (w/v) bovine serum albumin (BSA; Sigma-Aldrich, United
Kingdom) for 1 hour at room temperature, washed thrice with 200 µL CMF-PBS with 0.1%
(v/v) Tween-20 (PBST, Sigma-Aldrich, United Kingdom) followed by two washes with 200
µL of CMF-PBS. SARS-CoV-2 S1 RBD (50 µg mL−1 ) with or without heparin (100 µg mL−1 ) were added to each well with 25 µL PBST + 0.1% (w/v) BSA prior to incubation for
1 hour at room temperature, while rocking. Post incubation, wells were washed thrice
with 200 µL PBST and twice with 200 µL CMF-PBS before incubation (with agitation)
for 1 hour at room temperature, under dark conditions with an Alexa Fluor 488 anti-his
tag antibody (clone J095G46, Biolegend, United Kingdom) 1:5,000 (v/v) in 25 µL PBST
with 0.1% (w/v) BSA per well. Wells were washed thrice with 200 µL PBST and twice
with 200 µL CMF-PBS. Fluorescence emission was measured using a Tecan Infinite M200Pro
plate reader (λex. = 495 nm, λem. = 520 nm) with results presented as the mean ± the standard deviation (n = 3).
Expression and Purification of Recombinant SARS-CoV-2 S1 RBD
Residues 330 − 583 of the SARS-CoV-2 spike protein (GenBank: MN908947) were cloned
upstream of an N-terminal 6xHis-tag in the pRSETA expression vector and transformed
into SHuffle T7 Express Competent E. coli (NEB, United Kingdom). Protein expression was performed in MagicMedia E. coli Expression Media (Invitrogen, United Kingdom) at 30°C for 24 hours, 250 rpm. The
bacterial pellet was suspended in 5 mL of lysis buffer (BugBuster Protein Extraction
Reagent, Merck Millipore, United Kingdom; containing DNAse) and incubated at room
temperature for 30 minutes. The protein was purified from inclusion bodies using immobilized
metal affinity chromatography under denaturing conditions. On-column protein refolding
was performed by applying a gradient with decreasing concentrations of the denaturing
agent (6 to 0 M urea). After extensive washing, protein was eluted using 20 mM NaH2 PO4 , pH 8.0, 300 mM NaCl, 500 mM imidazole. Fractions were pooled and buffer-exchanged
to PBS (140 mM NaCl, 5 mM NaH2 PO4 , 5 mM Na2 HPO4 , pH 7.4; Lonza, United Kingdom) using a Sephadex G-25 column (GE Healthcare, United
Kingdom). Recombinant protein was stored at −20°C until required.
Preparation of Chemically Modified Heparin Derivatives
All chemically modified heparin polysaccharides ([Table 1 ]) were synthesised from parental unfractionated porcine mucosal heparin (M
w = 12 kDa; Celsus Laboratories, Cincinnati, United States) as previously described.[15 ]
[19 ] The veracity of all chemical modifications was ascertained using 1 H and 13 C NMR, with chemical shifts compared with TSP (Sigma-Aldrich, United Kingdom) as an
external reference standard.
Table 1
Repeating disaccharide structures of heparin derivatives
R1
R2
R3
Heparin 1
SO3
−
SO3
−
SO3
−
Heparin 2
SO3
−
COCH3
SO3
−
Heparin 3
H
SO3
−
SO3
−
Heparin 4
SO3
−
SO3
−
H
Heparin 5
H
COCH3
SO3
−
Heparin 6
SO3
−
COCH3
H
Heparin 7
H
SO3
−
H
Heparin 8
H
COCH3
H
Heparin 9[a ]
SO3
−
SO3
−
SO3
−
a Additionally, chemically O-sulfated at positions C-3 of glucosamine and C-3 of the
uronic acid.
Secondary Structure Determination of SARS-CoV-2 S1 RBD by Circular Dichroism Spectroscopy
The CD spectrum of the SARS-CoV-2 S1 RBD in PBS was recorded using a J-1500 Jasco
CD spectrometer (Jasco, United Kingdom), Spectral Manager II software (JASCO, United
Kingdom) and a 0.2-mm path length, quartz cuvette (Hellma, United States) scanning
at 100 nm min−1 with 1 nm resolution throughout the range of 190 to 260 nm. All spectra obtained
were the mean of five independent scans, following instrument calibration with camphorsulfonic
acid. SARS-CoV-2 S1 RBD was buffer-exchanged (prior to spectral analysis) using a
5 kDa Vivaspin centrifugal filter (Sartorius, Germany) at 12,000 g thrice and the
CD spectra were collected using 21 μL of a 0.6 mg mL−1 solution in PBS, pH 7.4. Spectra of heparin (porcine mucosal heparin), its derivatives
and oligosaccharides were collected in the same buffer at approximately comparable
concentrations, since these are polydisperse materials. Collected data were analysed
with Spectral Manager II software prior to processing with GraphPad Prism 7, using
second-order polynomial smoothing through 11 neighbours. Secondary structural prediction
was calculated using the BeStSel analysis server.[20 ] To ensure that the CD spectral change of SARS-CoV-2 S1 RBD in the presence of porcine
mucosal heparin did not arise from the addition of the heparin alone, which is known
to possess a CD spectrum at high concentrations,[21 ]
[22 ] a difference spectrum was analysed. The theoretical CD spectrum, which resulted
from the arithmetic addition of the CD spectrum of the SARS-CoV-2 S1 RBD and that
of the heparin, differed from the observed experimental CD spectrum of SARS-CoV-2 S1
RBD mixed with heparin. This demonstrates that the change in the CD spectrum arose
from a conformational change following binding to porcine mucosal heparin.
Surface Plasmon Resonance Determination of SARS-CoV-2 S1 RBD Binding to Unfractionated
Heparin
Human fibroblast growth factor 2 (FGF2) was produced as described by Duchesne et al.[23 ] Porcine mucosal heparin was biotinylated at the reducing end using hydroxylamine
biotin (Thermo Fisher, United Kingdom) as described by Thakar et al.[24 ] Heparin (20 µL of 50 mg mL−1 ) was reacted with 20 µL of hydroxylamine biotin in 40 µL of 300 mM aniline (Sigma-Aldrich,
United Kingdom) and 40 µL, 200 mM acetate pH 6 for 48 hours at 37°C. Free biotin was
removed by gel-filtration chromatography on Sephadex G25 (GE Life Sciences, United
Kingdom).
A P4SPR, multi-channel SPR instrument (Affinté Instruments; Montréal, Canada) was
employed. The functionalisation of the gold sensor surface was a modification of a
described method.[25 ] Briefly, a gold sensor chip that was plasma-cleaned prior to derivatisation: a self-assembled
monolayer of polyethylene glycol methyl ether (mPEG) thiol and biotin mPEG was formed
by incubating the chip in a 1 mM solution of these reagents at a 99:1 molar ratio
in ethanol for 24 hours. The chip was rinsed with ethanol and placed in the instrument.
PBS was used as the running buffer for the three sensing channels and a fourth background
channel at 500 µL min−1 , using an Ismatec pump. Twenty micrograms of streptavidin (Sigma, United Kingdom;
1 mL in PBS) was injected over the four channels. Subsequently, biotin–heparin (1 mL)
was injected over the three measurement channels, the fourth serving as a control.
Binding experiments were performed using PBS with 0.02% (v/v) Tween 20 as the running
buffer. The ligand was injected over the three sensing channels after dilution to
the concentration indicated in the figure legends, at 500 µL min−1 . Sensor surfaces with bound FGF2 were regenerated using a wash of 2 M NaCl (Fisher
Scientific, United Kingdom); however, this was found to be ineffectual for SARS-CoV-2 S1
RBD. Partial regeneration of the surface was achieved with 20 mM HCl (VWR, United
Kingdom), although only 0.25% (w/v) SDS (VWR, United Kingdom) was found to remove
the bound SARS-CoV-2 S1 RBD protein. After regeneration with 0.25% (w/v/) SDS, fluidics
and surfaces were washed with 20 mM HCl to ensure all traces of detergent were removed.
Background binding to the underlying streptavidin bound to the mPEG/biotin mPEG self-assembled
monolayer was determined by ligand injection over the control channel. Responses are
reported as the change in plasmon resonance wavelength (nm) and for the three control
channels represent their average response. Reliable kinetic parameters could not be
calculated since the data for interactions involve several potential binding sites
in the ligand and mass transport artefacts associated with flow systems meant that
the data did not fit a single site model.
Prediction of SARS-CoV-2 S1 Heparin-Binding Domain
The methods employed to predict heparin-binding sequences within the SARS-CoV-2 S1
RBD are described in Rudd et al.[6 ]. A brief description of the method follows. In combination, Ori et al[2 ] and Nunes et al[5 ] identified 786 heparin-binding proteins. In this study, 776 heparin-binding proteins,
from the previously identified list, were decomposed into amino acid sequences containing
no less than three residues with at least one basic amino acid. To simplify the problem,
the heparin-binding proteins were fragmented into sequences containing the following
combinations of amino acids: BX, BXA, BXS, BXP, BXAS, BXAP and BXPS (where, B = basic,
X = hydrophobic, A = acidic, P = polar and S = special). These sequences were compared
using a metric based on the Levenshtein distance, a measure of the similarity between
strings of characters, which is based on the minimum number of insertions, deletions
or substitutions that need to be applied to the characters to alter one sequence so
that it matches the other. Sequences with a similarity score of greater than 0.7 (70%
similarity) were considered highly conserved. These sets of highly conserved basic
amino acid containing sequences found in heparin-binding proteins were then used to
identify possible heparin-binding sequences within SARS-CoV-2 S1, utilising the same
Levenshtein similarity cut-off of 0.7.
Molecular Docking and Molecular Dynamics Simulations of Heparin-SARS-CoV-2 RBD Domain
The in silico docking of heparin binding to the spike RBD based on the crystal structure
6LZG (residues 333–527)[26 ] was performed using the ClusPro server version 2.0 that includes an advanced option
for heparin tetrasaccharide docking. At the time of this work, another crystal structure
was released (PDB code: 6M0J).[27 ] Both RBD structures (PDB: 6LZG and 6M0J) are superimposed with a root mean square
deviation of 0.6Å and have similar interactions with ACE2. Therefore, for this work,
molecular modelling was performed on the spike RBD chain (PDB code: 6LZG) after removing
the ACE2 domain. In alignment with the protein sequence used in this experimental
work, we also performed modelling of heparin oligosaccharides on the structure (PDB:
6ZGG; chain B) containing both RBD and SD1 domains (residues 330 −583).
ClusPro is a fast and crude approach to predict heparin-binding sites on the protein
surface using generic tetrasaccharides; however, it cannot capture the flexibility
and structural variability of longer oligosaccharides (greater than tetrasaccharides)
with sulfated and acetylated regions that might be involved in protein binding. Therefore,
docking calculations of heparin hexasaccharide and octasaccharide ligands (containing
the trisulfated disaccharide sequence; Heparin 1) to spike RBD were performed using
the GlycoTorch Vina program.[28 ] These model ligands were built using the Glycam server. The protein and ligands
were then converted to appropriate file formats for docking using the GlycoTorch tools.
The binding sites on the protein surface were defined based on the output from the
ClusPro heparin docking analysis. The docked poses obtained from docking studies on
sites I and III were subjected to the molecular dynamics (MD) simulations.
To investigate the conformational changes in the protein–heparin complex in the presence
of explicit solvent, classical MD simulations using the AMBER simulation package were
performed on the heparin hexasaccharide–RBD complex (site III) and heparin octasaccharide–RBD
(site I) complex. AMBER ff99SB*-ildn[29 ] and GLYCAM (version 06j1)[30 ] force fields were used for protein and heparin, respectively. The systems consisting
of the protein–heparin complex were solvated using TIP3P solvent[31 ] in an octahedral box by keeping a minimum distance of 12.0 Å between each face of
the box and the complex in both cases. Sodium and chloride ions were added to obtain
0.15 M ionic strength, as well as to neutralise the charges. The AMBER16 pmemd.cuda[32 ] module was used to perform MD simulations. The system preparation consisted of four
steps. In the first step the solvent and the ions were relaxed in a 2,000-step steepest
descent energy minimisation with the protein–heparin position restrained using a force
constant of 10 kcal/mol·Å2 . A further step of energy minimisation was performed with the position restraints
removed (500 steps of steepest descent and 1,500 steps of conjugate gradient minimisation).
After minimisation, the system was heated to T = 298 K in a 100 ps Langevin dynamics simulation under NVT conditions with restraints
on protein–heparin using a force constant of 10 kcal/mol·Å2 , followed by a 2 nanosecond NPT simulation using a Berendsen barostat and a force
constant of 2 kcal/mol·Å2 on protein–heparin. A further 2 nanosecond NPT unrestrained simulation was performed
for the density equilibration of the system. The production runs were performed for
500 nanoseconds using a Langevin dynamics integrator under constant pressure (p = 1 bar) and temperature (T = 298 K), with a time step of 2 fs. Long-range electrostatic calculations were performed
using the PME method. Heavy atom–hydrogen bonds were constrained using SHAKE. A cut-off
of 12.0 Å was applied to van der Waals forces. Periodic boundary conditions were applied
throughout.
Visualisation of the molecular models was performed using UCSF Chimera.[33 ] Molecular surfaces coloured by electrostatic potential were generated using APBS
and PDB2PQR[34 ] web servers with a protein interior dielectric constant of 2, a solvent dielectric
constant of 78.54. To obtain more detailed information about the contribution of binding
residues, the MM/GBSA module of AMBER 16.0[32 ] was used to compute the per-residue energy decompositions for every individual frame,
adding 1 to 4 energy terms to internal energy terms; the energies for every residue
were averaged over the 50,000 frames of production dynamics. The polar component of
the desolvation energy was determined via Onufriev's GB (igb = 5)[35 ] and the nonpolar contribution was estimated based on the solvent-accessible surface
area using the LCPO method[36 ] implemented in Amber. MM/GBSA calculations were performed using mbondi2 radii and
default settings for the nonpolar decomposition scheme, surface tension, and cavity
offset, whereas the ionic strength was changed from 0.0 to 0.150 M to be consistent
with MD simulations in explicit solvent. The outer dielectric and solute dielectric
constants were set to 80 and 1, respectively. More details of the MM/PB(GB)SA methods
can be found in the literature.[37 ]
Results
SARS-CoV-2 Viral Plaque-Forming Assays and RBD Binding
The titer of infectious virions was measured using standard plaque-forming assays
on Vero cells with and without a heparin pre-treatment (from 200 μg mL−1 , 1 hour prior to infection) with both a historical SARS-CoV isolate HSR-1 and the
recent SARS-CoV-2 isolate, Italy/UniSR1/2020. Significant decreases were observed
in the number of PFU upon heparin treatment for both SARS-CoV and SARS-CoV-2, with
the latter demonstrating significantly higher levels of inhibition (80%) than for
SARS-CoV ([Fig. 1 ]). Vero cells have the advantage that they are defined and well-characterised and,
when infected by viruses, do not secrete α- or β-interferon. Heparin treatment was
also able to inhibit the SARS-CoV-2 RBD binding to the cell surface of Vero cells
([Fig. 1B ]) The observed inhibition may be due to direct competition of the exogenous heparin
for host cell surface GAGs and/or for the ACE2 receptor and demonstrates that addition
of exogenous heparin can disrupt the interaction of the virus with the cell. It is
noteworthy that the heparin concentrations used cover both prophylactic and therapeutic
heparin nebulisation protocols,[38 ] highlighting that this in vitro regime correlates with heparin use clinically, which
is not the case for some of the other drugs tested thus far against COVID, e.g. ivermectin.[39 ]
Fig. 1 The heparin-mediated inhibition of SARS-CoV-2 viral invasion of Vero cells. (A ) The effect of unfractionated porcine mucosal heparin added 1 hour before the infection
of Vero cells with 50 PFU of SARS-CoV-2 or SARS-CoV. The results are expressed as
the number of PFU per well and represent the mean ± SD of quadruplicate cultures.
The p -value was calculated using the Mann–Whitney U test, ∗
P ≤ 0.05; ∗∗P ≤ 0.01; ∗∗∗P ≤ 0.001. (B ) The effect of unfractionated porcine mucosal heparin (100 μg mL−1 ) on RBD binding to Vero cells. Nil represents no treatment; no secondary represents
no secondary antibody control; au, arbitrary units of fluorescence. PFU, plaque-forming
unit; RBD, receptor-binding domain; SD, standard deviation.
Surface Plasmon Resonance Binding Studies
Fibroblast growth factor 2, a well-characterised heparin-binding protein, was used
to confirm the successful functionalisation of the three sensing channels with biotin–heparin.
Injection of 1 mL 100 nM FGF2 over the sensing channels elicited a significant response
([Fig. 2A ], injection between the blue and red arrows). However, 100 nM FGF2 elicited no response
in the control channel, functionalised solely with streptavidin ([Fig. 2A ]). The bound FGF2 was removed by a wash with 2 M NaCl, as previously for the IASys
optical biosensor.[40 ]
Fig. 2 Interaction of FGF2 and 800 nM SARS-CoV-2 S1 RBD with immobilised heparin. Reducing
end biotinylated heparin was immobilised on a streptavidin-functionalised P4SPR sensor
surface [no biotin-heparin (−) control]. PBS running buffer flow rate was 500 µL min−1 . The data for the three sensing channels are reported as an average response (−).
The start of protein injections is indicated by black arrows and the return of the surface to the running buffer (PBST) by red arrows . (A ) Injection of 100 nM FGF2. (B ) Injection of 800 nM SARS-CoV-2 S1 RBD protein. PBS, phosphate buffered saline; PBST,
phosphate buffered saline with Tween-20; RBD, receptor-binding domain.
When 800 nM SARS-CoV-2 S1 RBD was injected over the three sensing channels, there
was an increase in the signal of binding ([Fig. 2B ] injection starts at the blue arrow). At the end of the injection, the system returned
to the running buffer (PBST, red arrow). There was no dissociation. This is common
in surface measurement, particularly in flow systems, due to the extensive boundary
layer[41 ] of liquid enabling rebinding of the analyte and often causes a substantial underestimation
of the dissociation rate, k
off , which can be remedied by the addition of soluble ligand.[42 ] The injection of 800 nM SARS-CoV-2 S1 RBD over the control channel, functionalised
with just streptavidin, showed a small increase in response ([Fig. 2B ]) of the order of 10% of that seen in the measurement channel. Repeated measurements
indicate that this is the maximum level of background binding. These data demonstrate
that 800 nM SARS-CoV-2 S1 RBD binds specifically to heparin immobilised through its
reducing end and does not interact extensively with the underlying streptavidin/ethylene
glycol surface. It should be noted that when biotin–heparin is anchored to the streptavidin
layer, as in the measurement channels, such background binding will be reduced, since
less of the underlying surface is exposed. This is illustrated in the competition
experiments, in which soluble heparin was able to completely abrogate the binding
of SARS-CoV-2 S1 RBD to the surface ([Fig. 3A ]).
Fig. 3 Competition for SARS-CoV-2 S1 RBD binding to immobilised heparin with model heparin-derived
oligo- and polysaccharides. 800 nM SARS-CoV-2 S1 RBD was injected onto the surface
in the presence or absence of the indicated concentration of heparin-derived oligo-
and polysaccharides. Since the SARS-CoV-2 S1 RBD does not dissociate appreciably when
the system returns to PBST, this is when the response was measured, to avoid any confounding
effects of differences of refractive index between samples. A control measurement
(800 nM SARS-CoV-2 S1 RBD alone) was performed before each competition and used to
calculate the percentage of maximum binding. This ensured that small changes over
time in the responsiveness of the surface did not confound the analysis. (A ) Competition for 800 nM SARS-CoV-2 S1 RBD binding to immobilised heparin by heparin
and by enoxaparin. (B ) Competition for 800 nM SARS-CoV-2 S1 RBD binding to immobilised heparin by 0.17 mg
mL−1 heparin-derived dp 8 and dp 10, with the corresponding value for heparin from panel
(A) shown to aid comparison. (C ) Competition for 800 nM SARS-CoV-2 S1 RBD binding to immobilised heparin by a panel
of orthogonally chemically desulfated heparins ([Table 1 ]). PBST, phosphate buffered saline with Tween-20; RBD, receptor-binding domain.
In a first set of competition experiments, 800 nM SARS-CoV-2 S1 RBD was mixed with
heparin or enoxaparin at the indicated concentration and injected over the sensing
channels. The response when the surface returned to the running buffer, PBST, was
then measured. This is appropriate, since there is no appreciable dissociation of
the heparin-bound SARS-CoV-2 S1 RBD when the surface returns to PBST ([Fig. 2B ]) and it ensures that any differences in refractive index of the samples do not interfere
with the measurement. At 1.7 µg mL−1 heparin, a small reduction in the binding of SARS-CoV-2 S1 RBD was observed ([Fig. 3A ]) and, as the concentration of heparin increased, the binding of SARS-CoV-2 S1 RBD
decreased in a dose-dependent manner. SARS-CoV-2 S1 RBD binding was completely abrogated
by 1.7 mg mL−1 heparin. The low-molecular-weight heparin, enoxaparin, was also found to inhibit
the binding of SARS-CoV-2 S1 RBD, but on a weight basis it was less potent than heparin
([Fig. 3A ]). Thus, a small inhibition of binding was observed at 17 µg/mL enoxaparin and the
maximal inhibition observed with this polysaccharide was 70% at 1.7 mg mL−1 . These data indicate that heparin is approximately 30-fold more potent an inhibitor
of the interaction of SARS-CoV-2 S1 RBD with immobilised heparin than enoxaparin.
We then examined whether short oligosaccharides could inhibit SARS-CoV-2 S1 RBD binding.
A heparin-derived octasaccharide (dp 8) was without effect, but a decasaccharide (dp10)
at 0.17 mg mL−1 showed a modest inhibition of SARS-CoV-2 S1 RBD binding to immobilised heparin ([Fig. 3B ]). Whether these data reflect a true requirement for a longer structure for SARS-CoV-2 S1
RBD binding or instead relate to selective reduction in the presence of particular
binding structures in the polysaccharide as a consequence of the manufacturing processes
employed (β-elimination for enoxaparin and partial enzymatic heparin degradation for
oligosaccharides) remains to be determined.
The ability of a library of heparins that had been selectively modified ([Table 1 ]) to inhibit SARS-CoV-2 S1 RBD binding was determined. This showed different levels
of inhibitory activity, depending on their pattern of sulfation. Of the singly desulfated
heparins, heparin 2 ([Table 1 ]), which is de-N- sulfated/N -acetylated, showed inhibitory activity at 0.05 and 0.17 mg mL−1 , whereas heparin 3 and heparin 4 ([Table 1 ]), which are de-O -sulfated at C-2 and C-6, respectively, had no detectable inhibitory activity ([Fig. 3C ]). This suggests that SARS-CoV-2 S1 RBD has a preference for regions of saccharides
that are 2-O - and 6-O -sulfated. Interestingly, the doubly desulfated heparins (heparin 5, heparin 6, heparin
7; [Table 1 ]) possessed inhibitory activity, albeit lower than the native heparin ([Fig. 3C ]). Since the sulfation of the polysaccharide has a marked effect on its conformation,[43 ] these data suggest that the SARS-CoV-2 S1 RBD may have a preference for a particular
spatial arrangement of charged groups. Completely desulfated heparin (heparin 8) had
no inhibitory activity, indicating that ionic interactions with sulfate groups make
an important contribution to the interaction of the SARS-CoV-2 S1 RBD with the polysaccharide.
Over-sulfated heparin (heparin 9) inhibited most strongly of the heparin derivatives,
but not as effectively as native heparin. This highlights the likely importance of
polysaccharide conformation for SARS-CoV-2 S1 RBD binding, since sulfation on all
available hydroxyls renders the polysaccharide more rigid and restricts its conformational
flexibility.[43 ]
[44 ]
Secondary Structure Determination of SARS-CoV-2 S1 RBD Protein by Circular Dichroism
Spectroscopy
Circular dichroism spectroscopy in the far-ultraviolet region (190–260 nm) detects
conformational changes in protein secondary structure that occur in solution and can
also infer binding by an added ligand. The SPR data showed that there is binding between
heparins and SARS-CoV-2 S1 RBD, whereas the CD spectra would indicate whether this
binding interaction was accompanied by changes in secondary structure in the protein.
Such structural changes can be quantified using deconvolution of the CD spectra.[45 ] SARS-CoV-2 S1 RBD underwent conformational change in the presence of heparin ([Figs. 4 ] and [5 ]), consisting of an increase in α-helix content of 1.5% and a decrease in global
β-sheet of 2.1%. The observed changes demonstrate that when the SARS-CoV-2 S1 RBD
interacts with heparin in aqueous conditions of physiological relevance, it undergoes
conformational change. A chemically modified heparin derivative with the predominant
repeating disaccharide structure, IdoA-GlcNAc,6S (heparin 5), was able to induce closely
comparable secondary structural changes in the SARS-CoV-2 S1 RBD as heparin ([Fig. 5A–C ]; [Supplementary Fig. S1 ] [available in the online version]). Analysis by CD spectroscopy of the role of chain
length for heparin-derived oligosaccharides ([Fig. 6A–D ]) revealed that a hexasaccharide fraction was able to induce similar conformational
changes to heparin even though small heparin oligosaccharides were not effective inhibitors
in the SPR experiments. Together, the data show that although there is an apparent
dependence on size and charge for the binding of SARS-CoV-2 S1 RBD with polysaccharides,
the structure–function relationship seems to be more complex. It is also important
to note that heparin and its derivatives are produced by methods designed to enrich
anticoagulant properties, making it difficult to propose antiviral structure–function
studies without proper selection based on the actual antiviral properties of heparin.
Fig. 4 The conformational change of the SARS-CoV-2 S1 RBD observed in the presence of heparin
by CD spectroscopy. (A ) Circular dichroism spectra (190–260 nm) of SARS-CoV-2 S1 RBD alone (black solid line ) and with heparin (red solid line ) in PBS, pH 7.4. The red, dotted line represents the sum of the two individual spectra and the fact that this is distinct
from the spectrum of the RBD with heparin (red solid line ) indicates that a conformational change and, therefore, binding have occurred. The
dotted vertical line indicates 193 nm. (B ) Details of the same spectra expanded between 200 and 240 nm. Vertical dotted lines indicate 222 and 208 nm. PBS, phosphate buffered saline; RBD, receptor-binding domain.
Fig. 5 The conformational change of the SARS-CoV-2 S1 RBD observed in the presence of a
chemically modified heparin derivative by CD spectroscopy. (A ) Circular dichroism spectra (190–260 nm) of SARS-CoV-2 S1 RBD alone (black solid line ) with heparin (red solid line ) and with a chemically modified derivative, heparin 5 ([Table 1 ]), with the predominant repeating disaccharide structure; –IdoA2OH-GlcNAc6S–(blue solid line ) in PBS, pH 7.4. The vertical dotted line indicates 193 nm (B ). The same spectra expanded between 200 and 240 nm. Vertical dotted lines indicate 222 and 208 nm. (C ) Secondary structure content analysed using BeStSel (20) for SARS-Cov-2 S1 RBD [analysis
using BeStSel was performed on smoothed CD data from (A) between 190 and 260 nm].
PBS, phosphate buffered saline; RBD, receptor-binding domain.
Fig. 6 The conformational change of the SARS-CoV-2 S1 RBD observed in the presence of size-defined
heparin oligosaccharides by CD spectroscopy. Circular dichroism spectra between 200
and 240 nm of SARS CoV-2 S1 RBD in PBS, pH 7.4, alone (black solid line ), with heparin (red solid line ), and PMH-derived, size-defined oligosaccharides (blue solid line ): (A ) Tetrasaccharide, (B ) hexasaccharide, (C ) octasaccharide, and (D ) decasaccharide. Vertical dotted lines indicate 222 and 208 nm. PBS, phosphate buffered saline; RBD, receptor-binding domain.
Heparin-Binding Site Analysis
Interactions with basic amino acids are known to dominate the binding between proteins
and heparin. With that in mind, primary sequence analysis of the expressed protein
domain and analysis of the modelled SARS-CoV-2 S1 RBD structure ([Fig. 7 ]) were conducted, which indicated that there are several potential heparin-binding
sites and, importantly, these patches of basic amino acids are exposed on the protein
surface.
Fig. 7 SARS-CoV-2 S1 RBD protein model. Basic amino acids that are solvent accessible on
the surface are indicated (dark blue ); these form extensive patches. Sequences with the highest normalised count ([Tables 2 ] and [3 ]) are highlighted. R346 is also shown as it indicates a potential heparin-binding
gain of function mutation (T346R) from the Bat-RaTG13. RBD, receptor-binding domain.
Analysis of the RBD sequence for potential heparin-binding sites employing a metric
based on the Levenshtein distance (a measure of the similarity between two sequences)
found that the basic amino acid sequences within SARS-CoV-2 S1 RBD were similar to
278 sequences found in 309 heparin-binding proteins. The predicted heparin-binding
basic amino acid sequences are shown in [Tables 2 ] and [3 ]. Basic amino acids with a normalised count greater than 0.5 were found at: RKR 355–356
(0.941), LVK 533–535 (1.00), KK 557–558 (0.544) and R 557 (0.688). There were also
several possible secondary sites of interaction (a normalised count greater than 0.3):
R 346 (0.368), R 403 (0.386), K 417 (0.309) and H 519 (0.305). One consequence of
the ability of SARS-CoV-2 S1 RBD to interact with HS is that it may provide a route
for adhering to cell surfaces, enabling invasion. Conversely, this region also interacts
with the orthodox receptor of the spike protein ACE2 (SARS-CoV-2 S1 RBD 436–529),
suggesting that heparin and its derivatives interfere with the interactions between
the virus, via these residues. Furthermore, most of the identified sequences, with
three exceptions [TKLN (385–389), GKIADYNYKLP (416–427) and PYRVVVL (507–514)], are
exposed on the protein surface and available for binding.
Table 2
Sequence analysis of SARS-CoV-2 S1 RBD (330–583)
AA No.
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
AA
P
N
I
T
N
L
C
P
F
G
E
V
F
N
A
T
R
F
A
S
V
Y
A
W
N
R
K
R
I
S
N
HBP
0.022
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.368
0.368
0.368
0.000
0.000
0.000
0.000
0.000
0.000
0.941
0.941
0.941
0.941
0.000
0.000
AA No.
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
AA
C
V
A
D
Y
S
V
L
Y
N
S
A
S
F
S
T
F
K
C
Y
G
V
S
P
T
K
L
N
D
L
C
HBP
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.081
0.081
0.081
0.081
0.000
0.000
0.000
AA No.
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
AA
F
T
N
V
Y
A
D
S
F
V
I
R
G
D
E
V
R
Q
I
A
P
G
Q
T
G
K
I
A
D
Y
N
HBP
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.386
0.386
0.386
0.386
0.004
0.169
0.169
0.173
0.173
0.004
0.004
0.004
0.000
0.000
0.000
0.000
0.088
0.309
0.309
0.309
0.007
0.015
0.007
AA No.
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
AA
Y
K
L
P
D
D
F
T
G
C
V
I
A
W
N
S
N
N
L
D
S
K
V
G
G
N
Y
N
Y
L
Y
HBP
0.235
0.235
0.235
0.066
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.063
0.114
0.114
0.051
0.051
0.000
0.000
0.000
0.000
0.000
0.000
AA No.
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
AA
R
L
F
R
K
S
N
L
K
P
F
E
R
D
I
S
T
E
I
Y
Q
A
G
S
T
P
C
N
G
V
E
HBP
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.110
0.110
0.110
0.114
0.004
0.004
0.004
0.004
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
AA No.
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
AA
G
F
N
C
Y
F
P
L
Q
S
Y
G
F
Q
P
T
N
G
V
G
Y
Q
P
Y
R
V
V
V
L
S
F
HBP
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.015
0.029
0.029
0.029
0.029
0.029
0.029
0.000
0.000
AA No.
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
AA
E
L
L
H
A
P
A
T
V
C
G
P
K
K
S
T
N
L
V
K
N
K
C
V
N
F
N
F
N
G
L
HBP
0.000
0.305
0.305
0.305
0.305
0.004
0.004
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
1.000
1.000
1.000
0.000
0.015
0.015
0.015
0.000
0.000
0.000
0.000
0.000
0.000
0.000
AA No.
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
AA
T
G
T
G
V
L
T
E
S
N
K
K
F
L
P
F
Q
Q
F
G
R
D
I
A
D
T
T
D
A
V
R
HBP
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.544
0.544
0.544
0.544
0.000
0.000
0.000
0.000
0.026
0.026
0.033
0.007
0.007
0.007
0.007
0.000
0.000
0.011
0.688
0.688
0.688
AA No.
578
579
580
581
582
583
AA
D
P
Q
T
L
E
HBP
0.011
0.000
0.000
0.000
0.000
0.000
Abbreviations: AA No; amino acid number. AA; amino acid identity. HBP; relative frequency
of sequence among heparin-binding proteins; RBD, receptor-binding domain.
Note: This table shows the amino acid sequence, number, and normalized count for similar
sequences in the SARS-CoV-2 RBD as found in a library of 776 heparin-binding proteins.
The higher the value, the more often the short heparin-binding sequence was identified
among the set of heparin-binding proteins. Basic amino acids within regions of high
similarity are identified; arginine (blue), lysine (red), and histidine (green).
Table 3
Predicted heparin-binding domains in SARS-CoV-2 S1 RBD (330–583)
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
Human – SARS-CoV-2
P
N
I
T
N
L
C
P
F
G
E
V
F
N
A
T
R
F
A
S
V
Y
A
W
N
R
K
R
I
S
Bat – RaTG13
P
N
I
T
N
L
C
P
F
G
E
V
F
N
A
T
T
F
A
S
V
Y
A
W
N
R
K
R
I
S
Human – SARS-CoV
P
N
I
T
N
L
C
P
F
G
E
V
F
N
A
T
T
F
P
S
V
Y
A
W
E
R
K
R
I
S
Bat – SARS-CoV-related
P
N
I
T
N
L
C
P
F
G
E
V
F
N
A
T
K
F
P
S
V
Y
A
W
E
R
K
K
I
S
Bat – SARS-CoV-related
P
N
I
T
N
L
C
P
F
G
E
V
F
N
A
T
T
F
P
S
V
Y
A
W
E
R
K
R
I
S
0.022
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.368
0.368
0.368
0
0
0
0
0
0
0.941
0.941
0.941
0.941
0
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
Human – SARS-CoV-2
N
C
V
A
D
Y
S
V
L
Y
N
S
A
S
F
S
T
F
K
C
Y
G
V
S
P
T
K
L
N
D
Bat – RaTG13
N
C
V
A
D
Y
S
V
L
Y
N
S
T
S
F
S
T
F
K
C
Y
G
V
S
P
T
K
L
N
D
Human – SARS-CoV
N
C
V
A
D
Y
S
V
L
Y
N
S
T
S
F
S
T
F
K
C
Y
G
V
S
A
T
K
L
N
D
Bat – SARS-CoV-related
N
C
V
A
D
Y
S
V
L
Y
N
S
T
F
F
S
T
F
K
C
Y
G
V
S
A
T
K
L
N
D
Bat – SARS-CoV-related
N
C
V
A
D
Y
S
V
L
Y
N
S
T
S
F
S
T
F
K
C
Y
G
V
S
A
T
K
L
N
D
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.081
0.081
0.081
0.081
0
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
Human – SARS-CoV-2
L
C
F
T
N
V
Y
A
D
S
F
V
I
R
G
D
E
V
R
Q
I
A
P
G
Q
T
G
K
I
A
Bat – RaTG13
L
C
F
T
N
V
Y
A
D
S
F
V
I
T
G
D
E
V
R
Q
I
A
P
G
Q
T
G
K
I
A
Human – SARS-CoV
L
C
F
S
N
V
Y
A
D
S
F
V
V
K
G
D
D
V
R
Q
I
A
P
G
Q
T
G
V
I
A
Bat – SARS-CoV-related
L
C
F
S
N
V
Y
A
D
S
F
V
V
K
G
D
D
V
R
Q
I
A
P
G
Q
T
G
V
I
A
Bat – SARS-CoV-related
L
C
F
S
N
V
Y
A
D
S
F
V
V
K
G
D
D
V
R
Q
I
A
P
G
Q
T
G
V
I
A
0
0
0
0
0
0
0
0
0
0
0.386
0.386
0.386
0.386
0.004
0.169
0.169
0.173
0.173
0.004
0.004
0.004
0
0
0
0
0.088
0.309
0.309
0.309
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
Human – SARS-CoV-2
D
Y
N
Y
K
L
P
D
D
F
T
G
C
V
I
A
W
N
S
N
N
L
D
S
K
V
G
G
N
Y
Bat – RaTG13
D
Y
N
Y
K
L
P
D
D
F
T
G
C
V
I
A
W
N
S
K
H
I
D
A
K
E
G
G
N
F
Human – SARS-CoV
D
Y
N
Y
K
L
P
D
D
F
M
G
C
V
L
A
W
N
T
R
N
I
D
A
T
S
T
G
N
Y
Bat – SARS-CoV-related
D
Y
N
Y
K
L
P
D
D
F
M
G
C
V
L
A
W
N
T
R
N
I
D
A
T
S
T
G
N
Y
Bat – SARS-CoV-related
D
Y
N
Y
K
L
P
D
D
F
L
G
C
V
L
A
W
N
T
N
S
K
D
S
S
T
S
G
N
Y
0.007
0.015
0.007
0.235
0.235
0.235
0.066
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.063
0.114
0.114
0.051
0.051
0
0
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
Human – SARS-CoV-2
N
Y
L
Y
R
L
F
R
K
S
N
L
K
P
F
E
R
D
I
S
T
E
I
Y
Q
A
G
S
T
P
Bat – RaTG13
N
Y
L
Y
R
L
F
R
K
A
N
L
K
P
F
E
R
D
I
S
T
E
I
Y
Q
A
G
S
K
P
Human – SARS-CoV
N
Y
K
Y
R
S
L
R
H
G
K
L
R
P
F
E
R
D
I
S
N
V
P
F
S
P
D
G
K
P
Bat – SARS-CoV-related
N
Y
K
Y
R
Y
L
R
H
G
K
L
R
P
F
E
R
D
I
S
N
V
P
F
S
P
D
G
K
P
Bat – SARS-CoV-related
N
Y
L
Y
R
W
V
R
R
S
K
L
N
P
Y
E
R
D
L
S
N
D
I
Y
S
P
G
G
Q
S
0
0
0
0
0
0
0
0
0
0
0
0.110
0.110
0.110
0.114
0.004
0.004
0.004
0.004
0
0
0
0
0
0
0
0
0
0
0
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
Human – SARS-CoV-2
C
N
G
V
E
G
F
N
C
Y
F
P
L
Q
S
Y
G
F
Q
P
T
N
G
V
G
Y
Q
P
Y
R
Bat – RaTG13
C
N
G
Q
T
G
L
N
C
Y
Y
P
L
Y
R
Y
G
F
Y
P
T
D
G
V
G
H
Q
P
Y
R
Human – SARS-CoV
C
T
P
P
–
A
F
N
C
Y
W
P
L
N
D
Y
G
F
F
T
T
N
G
I
G
Y
Q
P
Y
R
Bat – SARS-CoV-related
C
T
P
P
–
A
L
N
C
Y
W
P
L
N
D
Y
G
F
Y
T
T
T
G
I
G
Y
Q
P
Y
R
Bat – SARS-CoV-related
C
S
A
I
–
G
P
N
C
Y
N
P
L
R
P
Y
G
F
F
T
T
A
G
V
G
H
Q
P
Y
R
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.015
0.029
0.029
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
Human – SARS-CoV-2
V
V
V
L
S
F
E
L
L
H
A
P
A
T
V
C
G
P
K
K
S
T
N
L
V
K
N
K
C
V
Bat – RaTG13
V
V
V
L
S
F
E
L
L
N
A
P
A
T
V
C
G
P
K
K
S
T
N
L
V
K
N
K
C
V
Human – SARS-CoV
V
V
V
L
S
F
E
L
L
N
A
P
A
T
V
C
G
P
K
L
S
T
D
L
I
K
N
Q
C
V
Bat – SARS-CoV-related
V
V
V
L
S
F
E
L
L
N
A
P
A
T
V
C
G
P
K
L
S
T
D
L
I
K
N
Q
C
V
Bat – SARS-CoV-related
V
V
V
L
S
F
E
L
L
N
A
P
A
T
V
C
G
P
K
L
S
T
D
L
I
K
N
Q
C
V
0.029
0.029
0.029
0.029
0
0
0
0.305
0.305
0.305
0.305
0.004
0.004
0
0
0
0
0
0
0
0
0
0
1.000
1.000
1.000
0.000
0.015
0.015
0.015
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
Human – SARS-CoV-2
N
F
N
F
N
G
L
T
G
T
G
V
L
T
E
S
N
K
K
F
L
P
F
Q
Q
F
G
R
D
I
Bat – RaTG13
N
F
N
F
N
G
L
T
G
T
G
V
L
T
E
S
N
K
K
F
L
P
F
Q
Q
F
G
R
D
I
Human – SARS-CoV
N
F
N
F
N
G
L
T
G
T
G
V
L
T
P
S
S
K
R
F
Q
P
F
Q
Q
F
G
R
D
V
Bat – SARS-CoV-related
N
F
N
F
N
G
L
T
G
T
G
V
L
T
P
S
S
K
R
F
Q
P
F
Q
Q
F
G
R
D
V
Bat – SARS-CoV-related
N
F
N
F
N
G
L
T
G
T
G
V
L
T
S
S
S
K
R
F
Q
P
F
Q
Q
F
G
R
D
V
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.544
0.544
0.544
0.544
0
0
0
0
0.026
0.026
0.033
0.007
0.007
570
571
572
573
574
575
576
577
578
579
580
581
582
583
Human – SARS-CoV-2
A
D
T
T
D
A
V
R
D
P
Q
T
L
E
Bat – RaTG13
A
D
T
T
D
A
V
R
D
P
Q
T
L
E
Human – SARS-CoV
S
D
F
T
D
S
V
R
D
P
K
T
S
E
Bat – SARS-CoV-related
S
D
F
T
D
S
V
R
D
P
K
T
S
E
Bat – SARS-CoV-related
S
D
F
T
D
S
V
R
D
P
K
T
S
E
0.007
0.007
0
0
0.011
0.688
0.688
0.688
0.011
0
0
0
0
0
Abbreviation: SARS-CoV, severe acute respiratory syndrome coronavirus.
Note: This table shows the amino acid sequence, number, and normalized count for similar
short, heparin-binding sequences in the SARS-CoV-2 RBD found in a library of 776 heparin
binding proteins. In addition to SARS-CoV-2, the table also contains the aligned sequences
from similar viruses, SARS-CoV, and related bat viruses. The predicted heparin-binding
regions are shown in yellow and variations within those regions between SARS-CoV-2
and the related viruses are highlighted in red.
Studying SARS-CoV-2 spike protein structure and behaviour in solution is an important
step in the development of effective therapeutics against SARS-CoV-2. Here, the ability
of the SARS-CoV-2 S1 RBD to bind pharmaceutical heparin, studied using spectroscopic
techniques, showed that SARS-CoV-2 S1 RBD binds to heparin and that, upon binding,
a significant conformational change is induced. Moreover, moieties of basic amino
acid residues that are common constituents of heparin-binding domains, and accessible
to solvent, are present on the SARS-CoV-2 S1 RBD surface, forming a series of continuous
patches ([Fig. 7 ]), suitable for heparin binding.
Comparison of the RBD amino acid sequence (residues 330–583) with an extensive library
of sequences from known heparin-binding proteins ([Tables 2 ] and [3 ]) based on the Levenshtein distance, which provides a measure of the degree of similarity between the sequences, suggests
that wider areas of the RBD surface may be available ([Fig. 7 ]) for binding to host cell surface GAGs, which can be disrupted by heparin ([Fig. 1B ]). These may relate to the propensity for the virus to select particular species,
individuals, age groups and cell types, since the GAG composition is known to vary
with these different parameters. There is also evidence of a potential heparin-binding
gain of function mutation (T346R) from Bat-RaTG13 ([Fig. 7 ]).
Molecular Modelling of SARS-CoV-2 RBD and Heparin Oligosaccharides
To evaluate the basic residue clusters obtained from the bioinformatics analyses,
we used systematic docking calculations that searched the entire protein surface of
SARS-CoV-2 RBD, using models of heparin tetrasaccharide as probes. ClusPro results
are in agreement with the results shown in [Tables 2 ] and [3 ], indicating the presence of four basic heparin-binding clusters, here onwards referred
to as sites I, II and III. Docking of multiple tetrasaccharides indicates the presence
of a binding site consisting of residues K356, R357, R355, R466, R346, K444 and R509
(site I). The distances between various residues indicate that site I could bind a
heparin oligosaccharide longer than dp6 ([Supplementary Fig. S1 ], available in the online version). Site II comprises the region around basic residues
R457, K458 and K462 and could accommodate a heparin tetrasaccharide ([Supplementary Fig. S1 ], available in the online version). Site III is adjacent to the ACE2 interface and
comprises residues R408, R403 and K417 ([Supplementary Fig. S2 ], available in the online version). In the absence of glycosylation and other domains
of S1, ClusPro analyses of the truncated structure (PDB:6ZGG; chain B) identified
residues LVK 533–535 and KK 557–558 of subdomain 1 (SD1) as heparin tetrasaccharide
binding sites (data not shown), which agrees with the sequence analysis above.
The results for the docking of a heparin octasaccharide, using GlycoTorch Vina, revealed
that the entire oligosaccharide has favourable affinity for site I (docking score = −5.4
kcal/mol) whereas the same octasaccharide showed unfavourable preference for site
III. Docking of a heparin hexasaccharide indicates that site III ([Supplementary Fig. S4 ], available in the online version) can accommodate heparin fragments equal or shorter
than a hexasaccharide. The obtained docking results confirmed that the binding of
heparin fragments to sites I, II and III is length-dependent.
To explore the key binding residues that may play an important role in binding to
the SARS-CoV-2 RBD region, the intermolecular interactions between the heparin oligosaccharides
and the protein residues were quantified from the MD trajectories using the pairwise
per residue decomposition energy module in AMBER. Amino acid residues found to be
of importance for interacting with a heparin octasaccharide were T345, R346, N354,
R355, K356, R357, I358, S359, L441, D442, K444, N448, Y449, N450, R466 and R509 ([Fig. 8A ]). The strength of these interactions varied from −1.0 to −17.5 kcal/mol. The interactions
between a heparin hexasaccharide and site III of RBD was also examined, demonstrating
a favourable interaction with residues R403, R408, N409, T415, G416, K417, Y421, N501
and Y505 ([Fig. 8B ]) and a strength of interactions in the range of −1.0 to −10 kcal/mol. The interactions
of heparin with residues D405-E406 are unfavourable due to the negative charge.
Fig. 8 Snapshots taken from MD simulations of SARS-CoV-RBD in the presence of either a heparin
octasaccharide (A) or a heparin hexasaccharide (B ). The heparin oligosaccharides are shown as sticks whereas amino acids of the RBD
are shown as spheres. The residues are coloured as per elements. Hydrogen atoms are
not shown for clarity. The regions subjected to conformational changes in the protein
during the simulations are highlighted in yellow ribbon. MD, molecular dynamics; RBD,
receptor-binding domain.
Conformational change in the SARS-CoV-2 spike RBD using 500 nanosecond MD simulations
was investigated using the DSSP (Define Secondary Structure of Proteins) program implemented
in AmberTools. Changes in the secondary structure elements of the RBD in the presence
of heparin oligosaccharides trajectory are shown in [Supplementary Fig. S3 ] (available in the online version). The β-sheets maintain the most stable regions
of secondary structure, whereas the α- and 310 -helices, turns and loops connecting the β-sheets, undergo conformational change in
both systems. The C-terminal of the RBD domain is connected to the SD1 domain through
a long flexible loop (residues A520–K535). This will further contribute to a conformational
change in the system upon binding of oligosaccharides.
Heparin-interacting residues Y449, K417, Y505, N501 and G502 of the RBD are also hotspots
for ACE2 receptor binding (known as the receptor-binding motif). This complements
the experimental data showing that heparin binding induces a conformational change
and inhibits the binding of ACE2.
Discussion and Conclusion
Discussion and Conclusion
The rapid spread of SARS-CoV-2 represents a significant challenge to global health
authorities. Moreover, it is likely that humanity will face future epidemics. Therefore,
strategies are required for treatments that might reduce the burden of disease. Basing
therapeutics on fundamental aspects of the biology of pathogen–host interactions that
will be reasonably constant at the molecular level would provide the means to develop
treatments in the face of new, rapidly spreading pathogens, such as SARS-CoV-2. The
interaction of a sizable number of pathogens, including the Coronaviridae with peri- and extracellular HS, represents such an opportunity.
Glycosaminoglycans such as HS are present on almost all animal cells and this class
of carbohydrates is central to the strategy employed by numerous pathogens,[46 ]
[47 ] including the Coronaviridae ,[48 ] to attach to host cells. It is likely that the host has a very limited ability to
alter HS structures employed by the pathogen. This is due to at least two factors.
First, the large number of protein partners of HS,[2 ]
[5 ]
[7 ] each showing some level of selectivity for different saccharide structures. Second,
binding specificity is selective for patterns of charged groups in a particular geometry,
which are attainable by different saccharide sequences.[49 ] The consequence is a high degree of degeneracy and selectivity of proteins for HS,
but never absolute specificity, i.e., one protein, one saccharide sequence.[7 ]
[44 ] However, on the pathogen's part, there are also restrictions. In the case of viruses,
their limited repertoire of extracellular proteins may result in overlapping binding
sites. The likely overlap of the ACE2 site in SARS-Cov-2 S1 RBD with its predicted
heparin (likewise HS) binding site being one example. Thus, HS presents a reasonably
constant molecular target for a pathogen, which in turn offers unique therapeutic
possibilities, as illustrated in the present work.
Our initial observation of an interaction between SARS-CoV-2 S1 RBD and heparin[50 ] has now been confirmed by subsequent work.[51 ]
[52 ] Here, we show that this interaction is associated with an inhibition by heparin
of cell invasion by SARS-CoV-2, in a dose-dependent manner at concentrations from
6.25 to 200 μg mL−1 , up to 80% in Vero cells, concentrations within the range experienced during anticoagulation
therapy. Furthermore, host cell surface GAGs are likely to act as SARS-CoV-2 binding
sites, as the addition of exogenous heparin disrupted RBD binding to the cell surface.
These data suggest that heparin, or one of its derivatives, may be effective in controlling
infection by SARS-CoV-2. Moreover, infections by the Coronaviridae including SARS-CoV-2 are often accompanied by marked coagulopathy, which is the usual
clinical target of heparin.[53 ] There are also data to suggest that heparin can modulate excessive inflammatory
responses,[54 ]
[55 ] and these are another hallmark of SARS-CoV-2 infections.
Protective effects of heparin in COVID-19 patients have been described[53 ] as research begins to reveal the mechanism behind increased clotting during the
disease. Although promising, questions concerning the timing of treatment, the appropriate
dosage and the selection of anticoagulation therapy remain unanswered.[56 ] In addition, there is the possibility of targeting with heparin directly the luminal
surface of the respiratory tract both to reduce infectivity on this surface and thromboses
in the air sacs, through the delivery of nebulised heparin. This is a strategy behind
two clinical trials.[38 ]
The present data provide direct evidence for the pursuit of the CHARTER and COVID-19
HOPE trials.[38 ] Moreover, together with the many prior studies on the interactions of pathogens
with host GAGs, they suggest that this apparent Achilles' heel of the host may be
exploited therapeutically in a range of infectious diseases using heparin and its
derivatives, drugs that are readily available and cheap world-wide, and which are
well tolerated even in extremely ill patients.
What is known about this topic?
The coronavirus SARS-CoV-2 is the causative agent of COVID-19.
The pandemic COVID-19 disease is of global significance and an international health
emergency.
The SARS-CoV-2 spike protein is responsible for host cell attachment and invasion.
The SARS-CoV-2 spike S1 receptor-binding domain (RBD) is known to bind to human angiotensin
I converting enzyme 2 (ACE2) and this is believed to facilitate host cell invasion.
Heparin has been shown previously by the authors to inhibit viral invasion by the
historical S-associated coronavirus HSR strain [responsible for infections in the
original 2002–2004 severe acute respiratory syndrome (SARS) outbreak], although the
potency of inhibition for this coronavirus was not believed to be of therapeutic significance.
What does this paper add?
This manuscript demonstrates for the first time that:
Heparin inhibits cellular invasion of SARS-CoV-2 coronavirus in live virus assays
and this inhibition is significantly more potent than that of historical coronavirus
strains.
The inhibition of live virus by heparin occurs at concentrations that would enable
therapeutic use, unlike those observed for historical SARS-causing coronavirus strains.
Heparin blocks the binding of recombinantly expressed SARS-CoV-2 spike S1 RBD protein
to pathologically relevant Vero cells.
Heparin binds to the SARS-CoV-2 spike S1 RBD and this manuscript highlights the structural
dependencies with regard to heparin.
The interaction of heparin with SARS-CoV-2 spike S1 RBD induces a conformational change
in the SARS-CoV-2 spike S1 RBD protein.