Content uploaded by David Alan Winkler
Author content
All content in this area was uploaded by David Alan Winkler on May 21, 2020
Content may be subject to copyright.
Content uploaded by David Alan Winkler
Author content
All content in this area was uploaded by David Alan Winkler on May 15, 2020
Content may be subject to copyright.
In silico comparison of spike protein-ACE2 binding affinities across species;
significance for the possible origin of the SARS-CoV-2 virus
Sakshi Piplani1,2, Puneet Kumar Singh2, David A. Winkler3-6*, Nikolai Petrovsky1,2*
1 College of Medicine and Public Health, Flinders University, Bedford Park 5046, Australia
2 Vaxine Pty Ltd, 11 Walkley Avenue, Warradale 5046, Australia
3 La Trobe University, Kingsbury Drive, Bundoora 3042, Australia
4 Monash Institute of Pharmaceutical Sciences, Monash University, Parkville 3052, Australia
5 School of Pharmacy, University of Nottingham, Nottingham NG7 2RD. UK
6 CSIRO Data61, Pullenvale 4069, Australia
*Joint senior authors
Abstract
The devastating impact of the COVID-19 pandemic caused by SARS–coronavirus 2 (SARS-
CoV-2) has raised important questions on the origins of this virus, the mechanisms of any
zoonotic transfer from exotic animals to humans, whether companion animals or those used for
commercial purposes can act as reservoirs for infection, and the reasons for the large variations
in SARS-CoV-2 susceptibilities across animal species. Traditional lab-based methods will
ultimately answer many of these questions but take considerable time. Increasingly powerful in
silico modeling methods provide the opportunity to rapidly generate information on newly
emerged pathogens to aid countermeasure development and also to predict potential future
behaviors. We used an in silico structural homology modeling approach to characterize the
SARS-CoV-2 spike protein which predicted its high affinity binding to the human ACE2
receptor. Next we sought to gain insights into the possible origins and transmission path by
which SARS-CoV-2 might have crossed to humans by constructing models of the ACE2
receptors of relevant species, and then calculating the binding energy of SARS-CoV-2 spike
protein to each of these. Notably, SARS-CoV-2 spike protein had the highest overall binding
energy for human ACE2, greater than all the other tested species including bat, the postulated
source of the virus. This indicates that SARS-CoV-2 is a highly adapted human pathogen. Of the
species studied, the next highest binding affinity after human was pangolin, which is most likely
explained by a process of convergent evolution. Binding of SARS-CoV-2 for dog and cat ACE2
was similar to affinity for bat ACE2, all being lower than for human ACE2, and is consistent
with only occasional observations of infections of these domestic animals. Snake ACE2 had low
affinity for spike protein, making it highly improbable that snakes acted as an intermediate
vector. Overall, the data indicates that SARS-CoV-2 is uniquely adapted to infect humans,
raising important questions as to whether it arose in nature by a rare chance event or whether its
origins might lie elsewhere.
Introduction
The devastating impact of COVID-19 infections caused by SARS–coronavirus 2 (SARS-CoV-2)
has stimulated unprecedented international activity to discover effective vaccines and drugs for
this and other pathogenic coronaviruses.1-16 It has also raised important questions on the
mechanisms of zoonotic transfer of viruses from animals to humans, questions as to whether
companion animals or those used for commercial purposes can act as reservoirs for infection,
and the reasons for the large variations in SARS-CoV-2 susceptibility across animal species.17-19
Understanding how viruses move between species may help us prevent or minimize these
pathways in the future. Elucidating the molecular basis for the different susceptibilities of species
may also shed light on the differences in susceptibilities in different sub-groups of humans.
Very recently, Shi et al. published the results of experiments to determine the susceptibility to
SARS-CoV-2 of ferrets, cats, dogs, and other domesticated animals.20 They showed that SARS-
CoV-2 virus replicates poorly in dogs, pigs, chickens, and ducks, but ferrets and cats are
permissive to infection. Other studies have reported the susceptibility of other animal species to
SARS-CoV-2.17,20,21 Susceptible species such as macaques, hamsters and ferrets are used as
animal models of SARS-CoV-2 infection.22-24 In the absence of purified, isolated ACE2 from all
the relevant animal species that could be used to measure the molecular affinities to spike protein
experimentally, computational methods offer considerable promise for determining the rank
order of affinities across species, as a method to impute which species may be permissive to
SARS-CoV-2.
Here we show how computational chemistry methods from structure-based drug design can be
used to determine the relative binding affinities of the SARS-CoV-2 spike protein for its
receptor, angiotensin converting enzyme (ACE)-2, a critical initiating event for SARS-CoV-2
infection, across multiple common and exotic animal species.25-27 The aim of these studies was to
better understand the species-specific nature of this interaction and see if this could help
elucidate the origin of SARS-CoV-2 and the mechanisms for its zoonotic transmission.
Materials and Methods
Homology modelling of structures
To construct the three-dimensional structure of the SARS-CoV-2 spike protein, the sequence was
retrieved from NCBI Genbank Database (accession number YP_009724390.1). A PSI-BLAST
search against the PDB Database for template selection was performed and the X-ray structure of
SARS coronavirus spike template (refcode 6ACC) was selected with 76.4% sequence similarity
to SARS-CoV-2 spike protein. The protein sequences of the ACE2 proteins for different species
is summarized in Table 11 and full sequence alignment in Supplementary Figure 1. The
phylogenetic tree for ACE2 proteins from selected animal species is illustrated in Supplementary
Figure 2. The 3D-structures were built using Modeller 9.21 (https://salilab.org/modeller/).28 The
quality of the generated models was evaluated using the GA341 score and DOPE scores, and the
models assessed using SWISS-MODEL structure assessment server
(https://swissmodel.expasy.org/assess).29
Protein Structure preparation
The X-ray crystal structures of human ACE2 (recode 3SCI) and human SARS spike protein
(refcode 5XLR) were retrieved from Protein Data Bank. Protein preparation and removal of non-
essential and non-bridging water molecules for docking studies were performed using the UCSF
Chimera package (https://www.cgl.ucsf.edu/chimera/).30
Molecular Docking
These modelled structures were docked against SARS-CoV-2 spike protein structure using the
HDOCK server (http://hdock.phys.hust.edu.cn/).31,32 Molecular docking was performed on the
homology modelled SARS-CoV-2 spike protein with human and animal ACE2 proteins. The
SARS-CoV spike protein was also docked with human ACE2 protein to obtain the docking pose
for binding energy calculations. The docking poses were ranked using an energy-based scoring
function. The docked structures were analyzed using UCSF Chimera software.
Molecular Dynamics Simulation
The final models were optimized using the AMBER99SB-ILDN force field in Gromacs2020
(http://www.gromacs.org/).33 Docked complexes (SARS-CoV-2 spike with human ACE2, human
SARS-CoV spike with human ACE2, SARS-CoV-2 spike with bat ACE2 etc) were used as
starting geometries for MD simulations. Simulations were carried out using the GPU accelerated
version of the program with the AMBER99SB-ILDN force field I periodic boundary conditions
on an Oracle Cloud Server. Docked complexes were immersed in a truncated octahedron box of
TIP3P water molecules. The solvated box was further neutralized with Na+ or Cl− counter ions
using the tleap program. Particle Mesh Ewald (PME) was employed to calculate the long-range
electrostatic interactions. The cutoff distance for the long-range van der Waals (VDW) energy
term was 12.0 Å. The whole system was minimized without any restraint. The above steps
applied 2500 cycles of steepest descent minimization followed by 5000 cycles of conjugate
gradient minimization. After system optimization, the MD simulations was initiated by gradually
heating each system in the NVT ensemble from 0 to 300 K for 50 ps using a Langevin thermostat
with a coupling coefficient of 1.0/ps and with a force constant of 2.0 kcal/mol·Å2 on the
complex. Finally, a production run of 100 ns of MD simulation was performed under a constant
temperature of 300 K in the NPT ensemble with periodic boundary conditions for each system.
During the MD procedure, the SHAKE algorithm was applied for the constraint of all covalent
bonds involving hydrogen atoms. The time step was set to 2 fs. The structural stability of the
complex was monitored by the RMSD and RMSF values of the backbone atoms of the entire
protein. Finally, the free energies of binding were calculated for all the simulated docked
structures. Calculations were also performed for up to 500 ns on human ACE2 to ensure that
100ns is sufficiently long for convergence. Duplicate production runs starting with different
random seeds were also run to allow estimates of binding energy uncertainties to be determined
for the strongest binding ACE2 structures.
MM‐PBSA binding free energy study
The binding free energies of the protein‐protein complexes were evaluated in two ways. The
traditional method is to calculate the energies of solvated SARS-CoV-2 spike and ACE2 proteins
and that of the bound complex proteins and derive the binding energy by subtraction.
ΔG (binding, aq) = ΔG (complex, aq) – (ΔG (spike, aq) + ΔG (ACE2, aq) (1)
We also calculated binding energies using the molecular mechanics Poisson Boltzmann surface
area (MM-PBSA) tool in GROMACS that is derived from the nonbonded interaction energies of
the complex.34,35 The method is also widely used method for binding free energy calculations.
The binding free energies of the protein complexes were analyzed during equilibrium phase from
the output files of 100 ns MD simulations. The g_mmpbsa tool in GROMACS was used after
molecular dynamics simulations, the output files obtained were used to post-process binding free
energies by the single-trajectory MM-PBSA method. Specifically, for a non-covalent binding
interaction in the aqueous phase the binding free energy, ΔG (bind,aq), is: –
ΔG (bind,aqu) = ΔG (bind,vac) + ΔG (bind,solv) (2)
where ΔG (bind,vac) is the binding free energy in vacuum, and ΔG(bind,solv) is the solvation
free energy change upon binding: –
ΔG (bind,solv) = ΔG (R:L, solv) - ΔG (R,solv) - ΔG (L,solv) (3)
where ΔG (R:L,solv), ΔG (R,solv) and ΔG (L,solv) are solvation free energies of complex,
receptor and ligand, respectively.
Free energy decomposition analyses were also performed by MM-PBSA decomposition to get a
detailed insight into the interactions between the ligand and each residue in the binding site. The
binding interaction of each ligand–residue pair includes three terms: the van der Waals
contribution, the electrostatic contribution, and the solvation contribution.
Interaction energy
Another estimate of the strength of the interaction between protein-protein complex can be
obtained from the non-bonded interaction energy between the complex. GROMACS has the ability
to decompose the short-range nonbonded energies between any number of defined groups. To
compute the interaction energies as a part of our analysis, we reran the trajectory files obtained
during simulation to recompute energies using -rerun command. The interaction energy is the
combination of short range Coulombic interaction energy (Coul-SR:Protein-Protein) and the
short-range Lennard-Jones energy (LJ-SR:Protein-Protein (see Table 3).
While this paper was being prepared, a paper by Guterres and Im described a substantial
improvement in protein-ligand docking results using high-throughput MD simulations.36 They
employed docking using AutoDock Vina, followed by MD simulation using CHARMM. The
parameters they advocated were very similar to those used in our study. Proteins were solvated in
a box of TIP3P water molecules extending 10 Å beyond the proteins and the particle-mesh
Ewald method was used for electrostatic interactions. Nonbonded interactions over 10 and 12 Å
were truncated. Their systems were minimized for 5000 steps using the steepest descent method
followed by 1 ns equilibration with an NVT setting. For each protein-ligand complex, they ran 3
× 100 ns production runs from the same initial structure using different initial velocity random
seeds and an integration step size of 2 fs.
Results
The ancestry of SARS-CoV-2 traces back to the human, civet and bat SARS-CoV strains, which
all use the same ACE2 proteins for cellular entry.37-39 The similarities and variations in
sequences for both the SARS-CoV-2 spike protein and the SARS spike protein were determined
from sequences retrieved from NCBI GenBank Databank and aligned using CLUSTALW. The
spike protein receptor binding domain (RBD) region showed a 72% identity between the two
viruses (Figure 1).
Figure 1. SARS and SARS-CoV-2 receptor binding domain sequence alignment showing
conserved regions in black and non-conserved residues in red and blue.
Spike Protein-ACE2 interaction
The three-dimensional structure of SARS-CoV-2 Spike protein (Figure 2) used was PDB refcode
6VXX. ACE2 is the receptor for SARS-CoV-2 and is needed for its cellular entry. The low
affinity binding of SARS spike protein for mouse ACE2 has been postulated as the reason mice
are largely non-permissive to SARS infection. Similarly, we postulated that variation in ACE2
between species may determine the affinity of binding of SARS-CoV-2 Spike protein and hence
determine which species are permissive to SARS-CoV-2 infection. The sequence alignment of
the selected ACE2 species is shown in Supplementary Figure 1 and the phylogenetic tree
showing relatedness of ACE2 proteins across selected animal species is illustrated in
Supplementary Figure 2. ACE2 modelled structures were generated for each relevant species
and the ones with the lowest DOPE (Discrete Optimized Protein Energy) score were chosen and
refined by energy minimization and used for further analysis.
Figure 2. 3D structure of SARS-CoV-2 spike protein (PDB refcode 6VXX)
The Ramachandran scores of the modeled ACE2 structures for selected species are summarized
in Table 1 with the actual predicted ACE2 structures and Ramachandran plots for each selected
species shown in Supplementary Figure 3.
Table 1. Ramachandran scores for ACE2 modelled structures for selected species
Species
MolProbity Score
Ramachandran Score
(favoured region)
Rhinolophus sinicus (bat)
2.45
97.8%
Mus musculus (mouse)
2.49
98.2%
Mustela putorius furo (ferret)
2.5
98.3%
Mesocricetus auratus (hamster)
2.59
98.7%
Felis catus (cat)
2.42
98.5%
Paguma larvata (civet)
2.52
97.3%
Macaca fascicularis (monkey)
2.59
98.2%
Manis javanica (pangolin)
2.42
98.3%
Ophiophagus hannah (king cobra)
2.93
96.0%
Canis lupus familiaris (dog)
2.9
97.1%
Equus caballus (horse)
2.29
96.3%
Panthera tigris altaica (tiger)
2.25
96.6%
Bos Taurus (cattle)
2.70
96.2%
The three-dimensional structures of ACE2 receptor of selected species were created using
Modeller9.23. The templates used to model the various ACE2 structures were 1R42 (human
ACE2), 3CSI (glutathione transferase) and 3D0G (spike protein receptor-binding domain from
the 2002-2003 SARS coronavirus human strain) (Table 2). Similarity of the template and query
is important while building the model, the query sequence of Macaca fascicularis (monkey)
accession number A0A2K5X283 was found to be 96.9% similar to the structure of human ACE2
(PDB id: 1R42) whereas Ophiophagus hannah (king cobra) had a much lower similarity of
61.42% to its template, 1R42 (Table 2). All structures were subjected to energy minimization
and those with the lowest DOPE score energy minimized structures were selected for further
study (Supplementary Figure 3).
Table 2. Template structures used to model selected ACE2 species and similarity scores of each
ACE2 sequence to the selected template used
Species
Accession No.
Template
Similarity
Rhinolophus sinicus (Bat)
AGZ48803.1
3CSI
79.73%
Mus musculus (Mouse)
Q8R0I0
1R42
84.27%
Mustela putorius furo (Ferret)
Q2WG88
1R42
83.44%
Mesocricetus auratus (Hamster)
A0A1U7QTA1
1R42
87.58%
Felis catus (Cat)
Q56H28
1R42
85.93%
Canis luparis (Dog)
J9P7Y2
1R42
84.93%
Paguma larvata (Ccivet)
Q56NL1
3D0G
86.77%
Macaca fascicularis (Monkey)
A0A2K5X283
1R42
96.91%
Manis javanica (Pangolin)
XP_017505752.1
1R42
85.57%
Ophiophagus hannah (King cobra)
V8NIH2
1R42
61.42%
Equus caballus (Horse)
F6V9L3
6MI7
85.91%
Panthera tigris altaica (Tiger)
XP_007090142.1
6MI7
85.91%
Bos taurus (Cow)
NP_001019673.2
6MI7
80.30%
Structure Quality Assessment
The modelled structures were further assessed for quality control using Ramachandran Plot and
molprobity scores in SWISSModel structure assessment. The Ramachandran Plot checks the
stereochemical quality of a protein by analyzing residue-by-residue geometry and overall
structure geometry, is also a way to visualize energetically allowed regions for backbone
dihedral angles ψ against φ of amino acid residues in protein structure. The Ramachandran score
of SARS-CoV-2 spike protein was 90% in the binding region and molprobity scores that provide
an evaluation of model quality at both the global and local level was 3.17. Further, the ACE2
modelled structures from selected species were also assessed using Swiss model structure
assessment. Ramachandran score or the percentage of amino acid residues falling into the
energetically favored region for all species ranged from 96-99% (Table 1). These Ramachandran
and Molprobity scores show that all the built structures were of good quality and were suitable
for use in further studies. The Ramachandran graphs are presented in Supplementary Figure 3.
Molecular Docking analysis
The Receptor Binding Domain (RBD) of SARS-CoV-2 spike protein was docked against ACE2
receptor of various species using HDOCK server. The interacting residues of ACE2 and SARS-
CoV-2 spike protein are depicted in Table 3. We found certain key amino acids in the receptor
binding motif (RBM) that were in accordance with previous studies40. Certain amino acids
including PHE28, ASN330, ASP355 and ARG357 were conserved in ACE2 of most of the
selected species and were observed to take part in the interaction with spike protein. TYR41,
LYS353, ALA386 and ARG393 also interacted with spike protein residues and were highly
conserved across all species except bat, mouse, ferret and pangolin, respectively. Spike protein
interacting residues with Ophiophagus hannah (King cobra) were least common amongst all the
ACE2 species included in the study, consistent with its low sequence similarity to human ACE2.
Table 3. Protein interacting residues in selected ACE2 species when docked with SARS-CoV-2 spike protein. Residues common to spike
interacting residues in human ACE2 are labelled red.
Species
Accession
Number
Position
19
24
27
28
30
31
34
37
38
41
42
79
83
330
353
393
Homo sapiens (Human)
Q9BYF1
S
Q
T
F
D
K
H
E
D
Y
Q
L
Y
N
K
R
Macaca fascicularis (Monkey)
A0A2K5X283
S
Q
T
F
D
K
H
E
D
Y
Q
L
Y
N
K
R
Rhinolophus sinicus (Bat)
U5WHY8
S
E
M
F
D
K
T
E
D
H
Q
L
Y
N
K
R
Mustela putorius furo (Ferret)
Q2WG88
D
L
T
F
E
K
T
E
E
Y
Q
-
Y
N
K
R
Manis javanica (Pangolin)
XP_017505752.1
-
E
T
F
E
K
S
E
E
Y
Q
I
Y
N
K
R
Ophiophagus hannah (Snake)
ETE61880.1
Q
V
K
F
E
Q
A
-
D
Y
N
N
F
N
L
R
Canis luparis (Dog)
J9P7Y2
-
L
T
F
E
K
Y
E
E
Y
Q
L
Y
N
K
R
Mesocricetus auratus (Hamster)
A0A1U7QTA1
S
Q
T
F
D
L
Q
E
D
Y
Q
L
Y
N
K
R
Panthera tigris (Tiger)
XP_007090142.1
S
L
T
F
D
K
H
E
E
Y
Q
L
Y
K
K
R
Bos Taur u s (Cow)
NP_001019673.2
S
Q
T
F
E
K
H
E
D
Y
Q
M
Y
N
K
R
Mus musculus (Mouse)
Q8R0I0
S
N
T
F
N
N
Q
E
D
Y
Q
Y
F
N
K
R
Paguma larvata (Civet)
Q56NL1
S
L
T
F
E
K
Y
E
Q
Y
Q
L
Y
N
K
R
Felis catus (Cat)
Q56H28
S
L
T
F
E
K
H
E
E
Y
Q
L
Y
N
K
R
Equus ferus caballus(Horse)
F6V9L3
S
L
T
F
D
K
S
E
E
H
Q
L
Y
N
K
R
Molecular dynamics simulations
The molecular dynamics simulation of complexes of SARS-CoV-2 spike protein and ACE2 receptors of various species were performed for
100ns. All complexes became stable during simulation with RMSD fluctuations converging to a range of 0.5 to 0.8 nm from the original
position. The calculated binding energies for the interactions of SARS-CoV-2 with ACE2 from the species studied are presented in Table 4. The
MMPBSA binding energies are summarized for comparison. The energies calculated by equation 1 and those from the MMPBSA algorithm
correlate strongly (r2=0.92). The binding energies (Table 2) correlate poorly with sequence similarity (r2=0.27). Also shown is the observed and
predicted SARS-CoV-2 susceptibilities of the species from analysis of phylogenetic clustering and sequence alignment with currently known
ACE2s utilized by SARS-CoV-2, as described by Qiu et al.41 (Supplementary Figure 2) Also listed is the interaction energies of the spike and
ACE2 proteins for each species calculated by Wu et al. using an automatic docking method, ICM-Pro42. Table 4 also includes observational in
vivo data on SARS-CoV-2 infectivity and disease symptoms in the species where this has been reported.
Table 4. Binding energies of SARS-CoV-2 spike protein to ACE2 of selected species and potential species susceptibilities from other studies
Species
Binding
energy
kcal/mol
MMPBSA e nerg y
kcal/mol
COVID infectivity
n/a = not assessed
Homo sapiens (human)
-52.8
-57.6
Permissive, high infectivity, severe disease in
5-10%,
Manis javanica (pangolin)
-52.0
-56.5
Permissive
Canis luparis (dog)
-50.8
-49.5
Permissive, low infectivity, no overt disease
Macaca fascicularis (monkey)
-50.4
-50.8
Permissive, medium infectivity, lung disease
Mesocricetus auratus (hamster)
-49.7
-50.0
Permissive, high infectivity, lung disease
Mustela putorius furo (ferret)
-48.6
-49.2
Permissive, medium infectivity, mild disease
Felis catus (cat)
-47.6
-48.9
Permissive, high infectivity, lung disease
Panthera tigris (tiger)
-47.3
-42.5
Permissive, overt respiratory symptoms
Rhinolophus sinicus (bat)
-46.9
-49.6
n/a
Paguma larvata (civet)
-45.1
-46.1
n/a
Equus ferus caballus (horse)
-44.1
-49.2
Permissive, low infectivity, no overt disease
Bos taurus (cow)
-43.6
-42.5
n/a
Ophiophagus hannah (snake)
-39.5
-52.5
n/a
Mus musculus (mouse)
-38.8
-39.4
n/a
Discussion
Knowing which species may be permissive to SARS-CoV-2 is very important, both in respect of
identifying intermediate hosts and potential source of the original virus, as well as helping to
identify suitable species for use as infection models to allow testing of COVID-19 drugs and
vaccines. Directly testing broad ranges of species for virus susceptibility is difficult and time
consuming and in the cases of very rare species, may not even be practicable. A less direct method
to try and obtain the same information might involve measurement of binding affinity of the SARS-
CoV-2 spike protein to the target ACE2 receptor from different species. This can be done, for
example, through use of cell lines transfected with ACE2 receptors from each individual species,
but this would again be time consuming and not practicable. A third approach, adopted here, is to
use fast, efficient in silico structural modelling and docking algorithms, using available genomic
and structural biology data, to generate relevant ACE2 structural models and use molecular
dynamics to calculate the binding energies. Notably, this approach surprisingly revealed that the
binding energy between SARS-CoV-2 spike protein and ACE2 was highest for humans out of all
species tested, suggesting that SARS-CoV-2 spike protein is uniquely evolved to bind and infect
cells expressing human ACE2. This finding is particularly surprising as, typically, a virus would be
expected to have highest affinity for the receptor in its original host species, e.g. bat, with a lower
initial binding affinity for the receptor of any new host, e.g. humans. However, in this case, the
affinity of SARS-CoV-2 is higher for humans than for the putative original host species, bats, or for
any potential intermediary host species.
The calculated binding energies of SAR-Cov2 spike protein with ACE2 from selected species are in
general agreement with the relatively limited amount of published information. SAR-Cov2 binding
affinity to monkey ACE2 was lower than for human ACE2. Young cynomolgus macaques when
infected expressed viral RNA in nasal swabs but did not develop overt clinical symptoms whereas
aged animals showed higher viral RNA loads and some weight loss and rapid respiration associated
with moderate interstitial pneumonia and virus replication in upper and lower respiratory tract.23,48
Syrian hamsters and ferrets are highly susceptible to SARS-CoV infection and are used as animal
models of the disease. Mahdy published a preprint reviewing SARS-CoV-2 infection in animals
and suggested that SARS-CoV-2 might recognize ACE2 from a variety of animal species, including
palm civet, the intermediate host for SARS-CoV. 45
Although bats carry many coronaviruses including SARS-CoV, a relative of SARS-CoV-2, direct
evidence for existence of SARS-CoV-2 in bats has not been found. As highlighted by our data, the
binding strength of SARS-CoV-2 for bat ACE2 is considerably lower than for human ACE2,
suggesting that even if SARS-CoV-2 did originally arise from a bat precursor it must later have
adapted its spike protein to optimise its binding to human ACE2. There is no current explanation
for how, when or where this might have happened. Instances of direct human infection by
coronaviruses or other bat viruses is rare with transmission typically involving an intermediate host.
For example, lyssaviruses such as Hendra are periodically transmitted from bats to horses and then
to humans who contact the infected horse. Similarly, SARS-CoV was shown to be transmitted from
bats to civet cats and from them to humans. To date, a virus identical to SARS-CoV-2 has not been
identified in bats or any other non-human species, making its origins unclear. To date, the most
closely related coronavirus to SARS-CoV-2, is the bat coronavirus, BatCoV RaTG1, which has
96% whole-genome identity to SARS-CoV-2.50. The fact that SARS-CoV-2 has also not been
found in any likely intermediate host raises questions of the origins of the original SARS-CoV-2
virus that infected human case zero in late 2019. Wuhan, the epicentre of the outbreak, hosts
China’s only BSL4 facility and is the site of considerable bat coronavirus research. Identification of
an intermediate animal host in which SARS-CoV-2 might have adapted to a human ACE2
permissive form would go a long way to alleviating concerns that SARS-CoV-2 is not a natural
virus. Lam et al.44 made confused public claims of finding SARS-CoV-2 in Malayan pangolins,
suggesting that pangolins were an intermediate vector for SARS-CoV-2. However, further sequence
analysis of these claims by Zhang et al. established that Pangolin-CoV was a very different
coronavirus that had modest at best ~90% sequence similarity to SARS-CoV-2. While Pangolin-
CoV spike RBD shared some similarities to SARS-CoV-2, its spike protein did not share the furin
cleavage site that was a prominent feature of SARS-CoV-2 spike protein.51 Hence, any similarity of
Pangolin-CoV to SARS-CoV-2 was restricted to the residues in the RBD and RBM. Overall,
Pangolin-CoV is only a distant relative of SARS-CoV-2.
Based on our data, the similarity of Pangolin-CoV to SARS-CoV-2 in the spike RBM could be a
case of convergent evolution of the two viruses, whereby the close similarity of the structure of the
pangolin ACE2 spike binding domain (SBD) to the same region of human ACE2, drove the
convergent evolution of the spike protein RBD of both viruses, allowing them to bind to pangolin
and human ACE2, respectively. Such close similarity of pangolin and human ACE2 SBD could
make it easy for any pangolin CoV to cross from pangolins to humans. Nevertheless, with no virus
matching the SARS-CoV-2 sequence identified in pangolins this makes it less likely that SARS-
CoV-2 has a pangolin origin. However, this does call for more intensive survey of coronaviruses in
pangolin populations, should such viruses pose future human pandemic threats. This finding also
supports the need for strict enforcement of a worldwide ban on any trafficking of pangolins to
reduce risks of pangolin coronaviruses crossing to humans and becoming a trigger for a future
coronavirus pandemic. However, there continues to be a lack of evidence to indicate that SARS-
CoV-2 is a pangolin-derived virus that first crossed from pangolins to humans in late 2019.
Early in the COVID-19 outbreak it was suggested that snakes may be an intermediate vector. ACE2
of turtle and snake has very low homology to human ACE2 and SARS-CoV-2 was shown to not
bind to reptile ACE2, making snakes unlikely as possible hosts for SARS-CoV-2. 46 This is
consistent with our own data predicting moderate to low binding of SARS-CoV-2 to snake ACE2.
Rodents varied widely in their permissiveness to SARS-CoV with mice being resistant and
hamsters permissive. Similarly there is inefficient virus replication of SARS-CoV-2 in mice
making them unsuitable as models to test SARS or COVID-19 vaccines or drugs. 47 This reflects
the low predicted binding energy of SARS-CoV-2 spike protein for mouse ACE2, as demonstrated
by our model data. Mice only became permissive for SARS infection when made transgenic for
human ACE2, with the same likely to be true for SARS-CoV-2. By contrast, hamsters were highly
permissive for SARS-CoV infection. Our model predicted that hamster should be permissive to
SARS-CoV-2 infection based on the predicted strong binding of SARS-Cov-2 spike protein for
hamster ACE2 (Table 4). Consistent with our model data, Syrian hamsters have been shown to
exhibit clinical and histopathological responses to SARS-CoV-2 that closely mimic human upper
and lower respiratory tract infections, with high virus shedding and ability to transmit to naïve
contact animals.24 Ferrets were another permissive model of SARS CoV infection, and our
modelling data indicated that SARS-CoV-2 has a similar binding energy to ferret, as it does to
hamster, ACE2 (Table 4). Consistent with our model data, ferrets have been shown to be permissive
to infection with SARS-CoV-2, with high virus titre in the upper respiratory tract, virus shedding,
infected ferrets showing acute bronchiolitis but without severe disease or death, and active
transmission to naïve ferrets through direct contact.20,22
Cat and tiger ACE2 were shown by our model to have similar binding affinity for SARS-CoV-2
spike protein and both these species have been shown to be permissive for SARS-CoV-2 infection.
Similarly our data suggests that SARS-CoV-2 binds with moderate affinity to dog ACE2 suggesting
that dogs may be susceptible to infection by SARS-CoV-2. In the case of companion animals that
live in close proximity to humans, Shen at al. state that SARS-CoV-2 can be efficiently transmitted
in cats and dogs while Shi et al. found that SARS-CoV-2 replicates poorly in dogs, pigs, chickens,
and ducks, but that ferrets and cats were permissive to infection. Temmam et al. tested 9 cats and 12
dogs living in close contact with their owners, two of whom tested positive for SARS-CoV-2 and
11 of 18 others showed clinical signs of COVID-19 but no antibodies against SARS-CoV-2 were
detectable in their blood using an immunoprecipitation assay. Goumeniu et al. published an
editorial querying the role of dogs in the Lombardy COVID-19 outbreak and recommended use of
computational docking experiments to provide evidence for or against infection of dogs.43 Hence,
our model data suggesting that dog ACE2 might be permissive for SARS-CoV-2 binding and
infection is therefore consistent with anecdotal reports of dogs being infected with SARS-CoV-2.
Hence other genetic factors could underlie the apparent lack of susceptibility of dogs to COVID-19
clinical infection.
It is known that gain of function (GOF) mutations occur in viruses that can lead to pandemics. GOF
means viruses gain a new property e.g. in influenza virus GOF has been associated with the
acquisition of a new function, such as mammalian transmissibility, increased virulence for humans,
or evasion of existing host immunity.49 The conditioning of viruses to humans as pandemics
progress is well recognized. However, the SARS-CoV-2 structures and sequences that we employed
were from viruses collected very early in the pandemic. It is therefore not clear how SARS-CoV-2
could have developed such a high affinity for human ACE2, notably higher than for those of
putative zoonotic sources for SARS-CoV-2, unless it has been previously selected on human ACE2
or an ACE2 of another species bearing a closely homologous spike protein binding domain.
Interestingly, pangolin ACE2 bears some similarities in its SBD to human ACE2. This marries with
the fact that Pangolin-CoV shares a highly similar RBD to SARS-CoV-2, although their remaining
sequence has only 90% similarity. This could be consistent with a process of convergent evolution
whereby human and pangolin coronaviruses infecting via ACE2, have come to the same solution in
respect of evolving an optimal spike RBD for binding of either human or pangolin ACE2,
respectively. Our data does indicate that humans might be permissive to pangolin CoVs that use
ACE2 for cell entry, a fact that needs to be borne in mind in respect of future potential coronavirus
pandemic sources. However, this does not mean that pangolin ACE2 was the receptor on which the
SARS-CoV-2 spike protein RBD was initially selected, with the strength of binding to pangolin
ACE2 lower than binding to human ACE2. This makes it unlikely that pangolins are the missing
intermediate host. If SARS-CoV-2 spike was selected on pangolin ACE2, then given the higher
affinity of SARS-CoV-2 for human ACE2 than for bat ACE2, SARS-CoV-2 would have to have
circulated in pangolins for a long period of time for this evolution and selection to occur and to date
there is no evidence of a SARS-CoV-2 like virus circulating in pangolins.
Another possibility would be a short term evolutionary step where a pangolin was recently co-
infected with a bat ancestor to SARS-CoV-2 at the same time as it was infected by a pangolin CoV
allowing a recombination event to occur whereby the spike RBD of the pangolin virus was inserted
into the bat CoV, thereby conferring the bat CoV with high binding for both pangolin and human
ACE2. Such recombination events are known to occur with other RNA viruses and can explain
creation of some pandemic influenza strains49. Nevertheless, such events are by necessity rare as
they require coinfection of the one host at exactly the same time. Most importantly, if such a
recombination event had occurred in pangolins it might have been expected to have similarly
triggered an epidemic spread of the new highly permissive SARS-CoV-2 like virus among pangolin
populations, such as we now see occurring across the human population. Currently there is no
evidence of such a pangolin SARS-CoV-2 like outbreak, making this whole scenario less likely.
Indeed, pangolins might be protected from SARS-CoV-2 infection due to the existence of cross-
protective spike RBD neutralising antibodies induced by exposure to pangolin CoV, given the RBD
similarity of these two viruses. Another possibility which still cannot be excluded is that SARS-
CoV-2 was created by a recombination event that occurred inadvertently or consciously in a
laboratory handling coronaviruses, with the new virus then accidentally released into the local
human population.
Given the seriousness of the ongoing SARS-CoV-2 pandemic, it is imperative that all efforts be
made to identify the original source of the SARS-CoV-2 virus. In particular, it will be important to
establish whether COVID-19 is due to a completely natural chance occurrence where a presumed
bat virus was transmitted to humans via an intermediate animal host or whether COVID-19 has
alternative origins. This information will be of paramount importance to help prevent any similar
human coronavirus outbreak in the future.
Acknowledgements
We would like to thank Harinda Rajapaksha for assistance to optimise Gromacs for this project. We
would like to thank Oracle Corporation for providing their Cloud computing resources through the
Oracle for Research program for the modelling studies described herein. In particular, we wish to
thank Peter Winn, Dennis Ward, and Alison Derbenwick Miller for facilitating this access. S.P.,
P.K.S and N.P. are supported by the National Institute of Allergy and Infectious Disease of the
National Institutes of Health under Contracts HHSN272201400053C and HHSN272201800044C.
This publication’s contents are solely the responsibility of the authors and do not necessarily
represent the official views of their affiliated institutions, funding bodies or Oracle Corporation.
References
1 Zhang, T. et al. Clinical trials for the treatment of Coronavirus disease 2019 (COVID-19): A
rapid response to urgent need. Sci China Life Sci, doi:10.1007/s11427-020-1660-2 (2020).
2 Zhang, J. et al. Progress and Prospects on Vaccine Development against SARS-CoV-2.
Vaccines (Basel) 8, doi:10.3390/vaccines8020153 (2020).
3 Yavuz, S. & Unal, S. Antiviral Treatment of Covid-19. Turk J Med Sci, doi:10.3906/sag-
2004-145 (2020).
4 Whitworth, J. COVID-19: a fast evolving pandemic. Trans R Soc Trop Med Hyg 114, 241-
248, doi:10.1093/trstmh/traa025 (2020).
5 Thanh Le, T. et al. The COVID-19 vaccine development landscape. Nat Rev Drug Discov,
doi:10.1038/d41573-020-00073-5 (2020).
6 Sohrabi, C. et al. World Health Organization declares global emergency: A review of the
2019 novel coronavirus (COVID-19). Int J Surg 76, 71-76, doi:10.1016/j.ijsu.2020.02.034
(2020).
7 Schlagenhauf, P., Grobusch, M. P., Maier, J. D. & Gautret, P. Repurposing antimalarials and
other drugs for COVID-19. Travel Med Infect Dis, 101658,
doi:10.1016/j.tmaid.2020.101658 (2020).
8 Sanders, J. M., Monogue, M. L., Jodlowski, T. Z. & Cutrell, J. B. Pharmacologic
Treatments for Coronavirus Disease 2019 (COVID-19): A Review. JAMA,
doi:10.1001/jama.2020.6019 (2020).
9 Rosales-Mendoza, S., Marquez-Escobar, V. A., Gonzalez-Ortega, O., Nieto-Gomez, R. &
Arevalo-Villalobos, J. I. What Does Plant-Based Vaccine Technology Offer to the Fight
against COVID-19? Vaccines (Basel) 8, doi:10.3390/vaccines8020183 (2020).
10 Rosa, S. G. V. & Santos, W. C. Clinical trials on drug repositioning for COVID-19
treatment. Rev Panam Salud Pub 44, e40, doi:10.26633/RPSP.2020.40 (2020).
11 Olsen, M., Cook, S. E., Huang, V., Pedersen, N. & Murphy, B. G. Perspectives: Potential
Therapeutic Options for SARS-CoV-2 Patients Based on Feline Infectious Peritonitis
Strategies: Central Nervous System Invasion and Drug Coverage. Int J Antimicrob Agents,
105964, doi:10.1016/j.ijantimicag.2020.105964 (2020).
12 Mendes, A. Research towards treating COVID-19. Br J Community Nurs 25, 204-205,
doi:10.12968/bjcn.2020.25.4.204 (2020).
13 Lu, S. Timely development of vaccines against SARS-CoV-2. Emerg Microbes Infect 9,
542-544, doi:10.1080/22221751.2020.1737580 (2020).
14 Jiang, S. Don't rush to deploy COVID-19 vaccines and drugs without sufficient safety
guarantees. Nature 579, 321, doi:10.1038/d41586-020-00751-9 (2020).
15 Ciotti, M. et al. COVID-19 Outbreak: An Overview. Chemother, 1-9,
doi:10.1159/000507423 (2020).
16 Berkley, S. COVID-19 needs a big science approach. Science, doi:10.1126/science.abb8654
(2020).
17 Almendros, A. & Gascoigne, E. Can companion animals become infected with COVID-19?
Vet Rec 186, 419-420, doi:10.1136/vr.m1322 (2020).
18 Almendros, A. Can companion animals become infected with COVID-19? Vet Rec 186,
388-389, doi:10.1136/vr.m1194 (2020).
19 Ahmad, T. et al. COVID-19: Zoonotic aspects. Travel Med Infect Dis, 101607,
doi:10.1016/j.tmaid.2020.101607 (2020).
20 Shi, J. et al. Susceptibility of ferrets, cats, dogs, and other domesticated animals to SARS-
coronavirus 2. Science, doi:10.1126/science.abb7015 (2020).
21 Temmam, S. et al. Absence of SARS-CoV-2 infection in cats and dogs in close contact with
a cluster of COVID-19 patients in a veterinary campus. bioRxiv (2020) doi:
10.1101/2020.04.07.029090..
22 Kim, Y. I. et al. Infection and Rapid Transmission of SARS-CoV-2 in Ferrets. Cell Host
Microbe, doi:10.1016/j.chom.2020.03.023 (2020).
23 Yu, P. et al. Age-related rhesus macaque models of COVID-19. Animal Model Exp Med. ,
1-5, doi:10.1002/ame2.12108 (2020).
24 Chan, J. F. et al. Simulation of the clinical and pathological manifestations of Coronavirus
Disease 2019 (COVID-19) in golden Syrian hamster model: implications for disease
pathogenesis and transmissibility. Clin Infect Dis, doi:10.1093/cid/ciaa325 (2020).
25 Yan, R. et al. Structural basis for the recognition of SARS-CoV-2 by full-length human
ACE2. Science 367, 1444-1448, doi:10.1126/science.abb2762 (2020).
26 Touyz, R. M., Li, H. & Delles, C. ACE2 the Janus-faced protein - from cardiovascular
protection to severe acute respiratory syndrome-coronavirus and COVID-19. Clin Sci
(Lond) 134, 747-750, doi:10.1042/CS20200363 (2020).
27 Shang, J. et al. Structural basis of receptor recognition by SARS-CoV-2. Nature,
doi:10.1038/s41586-020-2179-y (2020).
28 Sali, A. & Blundell, T. L. Comparative protein modelling by satisfaction of spatial
restraints. J Mol Biol 234, 779-815, doi:10.1006/jmbi.1993.1626 (1993).
29 Benkert, P., Biasini, M. & Schwede, T. Toward the estimation of the absolute quality of
individual protein structure models. Bioinform 27, 343-350,
doi:10.1093/bioinformatics/btq662 (2011).
30 Pettersen, E. F. et al. UCSF Chimera--a visualization system for exploratory research and
analysis. J Comput Chem 25, 1605-1612, doi:10.1002/jcc.20084 (2004).
31 Yan, Y., Zhang, D., Zhou, P., Li, B. & Huang, S. Y. HDOCK: a web server for protein-
protein and protein-DNA/RNA docking based on a hybrid strategy. Nucleic Acids Res 45,
W365-W373, doi:10.1093/nar/gkx407 (2017).
32 Yan, Y., Tao, H., He, J. & Huang, S. Y. The HDOCK server for integrated protein-protein
docking. Nat Protoc, doi:10.1038/s41596-020-0312-x (2020).
33 Abraham, M. J. et al. GROMACS; High performance molecular simulations through multi-
level parallelism from laptops to supercomputers. SoftwareX 1-2, 19-25 (2015).
34 Baker, N. A., Sept, D., Joseph, S., Holst, M. J. & McCammon, J. A. Electrostatics of
nanosystems: application to microtubules and the ribosome. Proc Natl Acad Sci USA 98,
10037-10041, doi:10.1073/pnas.181342398 (2001).
35 Kumari, R., Kumar, R., Open Source Drug Discovery, C. & Lynn, A. g_mmpbsa--a
GROMACS tool for high-throughput MM-PBSA calculations. J Chem Inf Model 54, 1951-
1962, doi:10.1021/ci500020m (2014).
36 Guterres, H. & Im, W. Improving Protein-Ligand Docking Results with High-Throughput
Molecular Dynamics Simulations. J Chem Inf Model, doi:10.1021/acs.jcim.0c00057 (2020).
37 Ge, X. Y. et al. Isolation and characterization of a bat SARS-like coronavirus that uses the
ACE2 receptor. Nature 503, 535-538, doi:10.1038/nature12711 (2013).
38 Li, W. et al. Angiotensin-converting enzyme 2 is a functional receptor for the SARS
coronavirus. Nature 426, 450-454, doi:10.1038/nature02145 (2003).
39 Wan, Y., Shang, J., Graham, R., Baric, R. S. & Li, F. Receptor Recognition by the Novel
Coronavirus from Wuhan: an Analysis Based on Decade-Long Structural Studies of SARS
Coronavirus. J Virol 94, doi:10.1128/JVI.00127-20 (2020).
40 Yuan, M. et al. A highly conserved cryptic epitope in the receptor-binding domains of
SARS-CoV-2 and SARS-CoV. Science, doi:10.1126/science.abb7269 (2020).
41 Qiu, Y. et al. Predicting the angiotensin converting enzyme 2 (ACE2) utilizing capability as
the receptor of SARS-CoV-2. Microbes Infect, doi:10.1016/j.micinf.2020.03.003 (2020).
42 Wu, C. et al. In Silico Analysis of Intermediate Hosts and Susceptible Animals of SARS-
CoV-2. ChemRxiv (2020) doi: 10.26434/chemrxiv.12057996.v1.
43 Goumenou, M., Spandidos, D. A. & Tsatsakis, A. [Editorial] Possibility of transmission
through dogs being a contributing factor to the extreme Covid19 outbreak in North Italy.
Mol Med Rep 21, 2293-2295, doi:10.3892/mmr.2020.11037 (2020).
44 Lam, T. T. et al. Identifying SARS-CoV-2 related coronaviruses in Malayan pangolins.
Nature, doi:10.1038/s41586-020-2169-0 (2020).
45 Mahdy, M. An Overview of SARS‐CoV‐2 and Animal Infection. Preprints (2020) doi:
10.20944/preprints202004.0192.v1.
46 Luan, J., Jin, X., Lu, Y. & Zhang, L. SARS-CoV-2 spike protein favors ACE2 from
Bovidae and Cricetidae. J Med Virol, doi:10.1002/jmv.25817 (2020).
47 Bao, L. et al. The pathogenicity of 2019 novel coronavirus in hACE2 transgenic mice.
bioRxiv (2020) doi: 2020/2002.2007.939389.
48 Rockx, B. et al. Comparative pathogenesis of COVID-19, MERS And SARS in a non-
human primate model. bioRxiv (2020) doi: 2020/2003.2017.995639.
49 Casadevall, A. & Imperiale, M. J. Risks and benefits of gain-of-function experiments with
pathogens of pandemic potential, such as influenza virus: a call for a science-based
discussion. mBio 5, e01730-01714, doi:10.1128/mBio.01730-14 (2014).
50 Zhou, P., Yang, X.-L., Wang, X.-G., Hu, B., Zhang, L., Zhang, W., Si, H.-R.,Zhu, Y., Li,
B., Huang, C.-L., et al. (2020). A pneumonia outbreak associated with a new coronavirus of
probable bat origin. Nature. Published online February 3,
2020.https://doi.org/10.1038/s41586-020-2012-7.
51 Zhang, T.; Wu, Q.; Zhang, Z. 2020, Current Biology 30, 1346–1351, April 6, 2020
SUPPLEMENTARY INFORMATION
Supplementary Figure 1. Sequence alignment of ACE2 amino acid sequence from selected species. The
SAR-Cov-2 spike protein binding region is highlighted in red.
Supplementary Figure 2. Phylogenetic tree showing relatedness of sequences of ACE2 proteins from selected species.
BAT ACE2
CAT ACE2
CIVET ACE2
DOG ACE2
FERRET ACE2
HAMSTER ACE2
MONKEY ACE2
MOUSE ACE2
PANGOLIN ACE2
SNAKE ACE2
HORSE ACE2
CATTLE ACE2
TIGER ACE2
Supplementary Figure 3. Modelled ACE2 structures for selected species, with Ramachandran plots and quality metrics
Supplementary Figure 4. Predicted and confirmed utilization of ACE2 receptors by the SARS-Cov-2 spike protein based on
sequence homology from Qui et al.41
