Molecular mechanisms behind anti SARS-CoV-2 action of lactoferrin
Mattia Miotto, Lorenzo Di Rienzo, Leonardo Bò, Alberto Boffi, Giancarlo Ruocco, Edoardo Milanetti
MMolecular mechanisms behind anti SARS-CoV-2 action of lactoferrin
Mattia Miotto,
1, 2
Lorenzo Di Rienzo, Leonardo B`o, Alberto Boffi,
3, 2
Giancarlo Ruocco,
2, 1 and Edoardo Milanetti
1, 2 Department of Physics, University of Rome ‘La Sapienza’, Piazzale Aldo Moro, 5, I00185, Rome, Italy Fondazione Istituto Italiano di Tecnologia (IIT), Center for Life Nano Science, Viale Regina Elena 291, I00161 Roma, Italy Department of Biochemical Sciences ‘A. Rossi Fanelli’ Sapienza University, Piazzale Aldo Moro 5, 00185, Rome, Italy
Despite the huge effort to contain the infection, the novel SARS-CoV-2 coronavirus has rapidlybecome pandemics, mainly due to its extremely high human-to-human transmission capability, anda surprisingly high viral charge of symptom-less people. While the seek of a vaccine is still ongoing,promising results have been obtained with antiviral compounds. In particular, lactoferrin is found tohave beneficial effects both in preventing and soothing the infection. Here, we explore the possiblemolecular mechanisms with which lactoferrin interferes with SARS-CoV-2 cell invasion, preventingattachment and/or entry of the virus. To this aim, we search for possible interactions lactoferrinmay have with virus structural proteins and host receptors. Representing the molecular iso-electronsurface of proteins in terms of 2D-Zernike descriptors, we (i) identified putative regions on thelactoferrin surface able to bind sialic acid receptors on the host cell membrane, sheltering the cellfrom the virus attachment; (ii) showed that no significant shape complementarity is present betweenlactoferrin and the ACE2 receptor, while (iii) two high complementarity regions are found on theN- and C-terminal domains of the SARS-CoV-2 spike protein, hinting at a possible competitionbetween lactoferrin and ACE2 for the binding to the spike protein.
I. INTRODUCTION
Lactoferrin (Lf) is a versatile glycoprotein, which playsa key role in many biological functions [1]. In this work,we focus on the Lf as a crucial player in natural immunity,since it has been proposed to play a strong antiviral activ-ity against a wide range of RNA and DNA viruses [2–6].Lf is composed of a single chain of about 700 residuesfolded into two symmetrical lobes. Each lobe possesses ametal-binding site, able to bind iron but also other ionslike Cu , Zn and M n [7, 8].This protein is present in saliva, tears, seminal fluid,white blood cells, and milk of mammals [9]. From itsdiscovery in 1939 [10, 11] lactoferrin has been identifiedas the most important iron-binding protein in milk. Be-sides, in recent years, lactoferrin has been found involvedin a multitude of biological processes. In fact, despitethe name, the iron cargo capacity of Lf is not the promi-nent activity exerted by this molecule. Instead, it per-forms antioxidant, anti-inflammatory, and anticancer ac-tivities [8, 12], together with a broad antimicrobial actionagainst bacteria and fungi. The latter activity, in partic-ular, is due to Lf’s ability to reversibly bind two atomsof iron with high affinity in the presence of bicarbon-ate. The iron-free form of Lf, apo-lactoferrin (apoLf),deprives bacteria of iron, thus inhibiting their metabolicactivities in vivo.Besides all the aforementioned activities, Lf has beendemonstrated to prevent infection of a wide range of di-verse viral species [13].Many viruses make use of glycans, such as sialic acid(SIA) or glycosaminoglycan, like heparan sulfate (HF) asattachment factors. See, for example, [14] for details onglycans. When the contact between the virus particle andthese receptors is established, they roll toward their spe- cific viral receptor and subsequently enter the host cell,for instance by fusing with the host cell membrane [15].While the interaction between Lf and HF has beenobserved [16], studies on its interaction with sialic acidderivatives are still missing. On the other hand, Lf hasbeen reported to interact with virus structural proteins,S, M, and E [17].In general, depending on the specifics of the virus,lactoferrin prevents infection of the target cell by either(i) interfering with the attachment factor or (ii) by bind-ing to host cell molecules that the virus uses as a receptoror co-receptor (competition) or (iii) by direct binding tovirus particles, as described for herpesvirus [18], polio-and rotavirus [19, 20], and possibly human immunod-eficiency virus [21]. See, for example, [22] for a moredetailed discussion.While we are writing this article, a novel virus, firstobserved in the autumn of 2019, has rapidly becomepandemic. This virus, called SARS-CoV-2, belongs tothe coronavirus family and causes severe acute respira-tory syndrome [23, 24], somewhat similar to those causedby two other coronaviruses, SARS-CoV and MERS-CoV, which crossed species in 2002-2004 [25, 26] and2012 [27]. In fact, SARS-CoV-2, similarly to SARS-CoV and MERS-CoV, attacks the lower respiratory sys-tem, thus provoking viral pneumonia. However, this in-fection can also lead to effects on the gastrointestinalsystem, heart, kidney, liver, and central nervous system[24, 28, 29].As for SARS-CoV [30–32], recent in vivo experimentsconfirmed that also SARS-CoV-2 cell entry is medi-ated by high-affinity interactions between the receptor-binding domain (RBD) of the virus S glycoproteinand the human-host Angiotensin-converting enzyme 2(ACE2) receptor [33]. The spike protein is located on the a r X i v : . [ q - b i o . B M ] J u l virus envelope and promotes the host attachment and fu-sion between the viral and cellular membrane. [34, 35].Structural studies determined the structures of such pro-tein both in free form and bound to ACE2 [36]. Furtherstudies investigate the possible interaction of SARS-CoV-2 to sialic acids [37–40] or heparan surfate receptors [39],both considered involved in SARS-CoV-2 as well as inother coronavirus infections [14, 16, 41–43].While the use of Lf as an antiviral against previouscoronaviruses infections has been poorly investigated, ex-cept for [16], where evidence of an effect in the attach-ment process is shown, new studies are proving the an-tiviral effect of Lactoferrin against the novel SARS-CoV-2 infection. In particular, administration of a liposomalformulation of Lf to a significant sample of Covid-19 pos-itive patients, has been shown to provide an immediatebeneficial effect [44].Here, we computationally investigate the possiblemolecular mechanisms behind the observed antiviral ac-tion of lactoferrin. In particular, we make use of a re-cently developed computational protocol based on the2D Zernike Polynomials, able to rapidly characterize theshape conformation of given protein regions [37]. In thisframework using a simple pairwise distance, it is possibleto evaluate the similarity between 2 protein pockets orthe shape complementarity between the binding regionsof 2 interacting proteins.In order to assess whether lactoferrin could influencethe attachment factors, we investigated the ability of Ltto bind to sialic acid (SIA) or heparan sulfate (HS) re-ceptors [14], both considered involved in SARS-CoV-2infection [37–40] as well as to other coronavirus infec-tions [16, 41–43].We moreover checked for a possible direct interactionbetween LF and ACE2 receptor, which could inhibit theinterplay between spike and ACE2 binding necessary tothe virus infection.Finally, we investigate the possible interaction betweenLf and the 3 proteins present on the SARS-CoV-2 mem-brane, i.e. the spike (S), membrane (M), and envelope(E) proteins. II. RESULTS
The entry of the virus inside the host cells requires theoccurrence of a sequence of molecular interactions. Sialo-side (SIA) and/or heparan sulfate (HF) chains mediatethe attachment of the virion to the cell surface. Oncein the proximity of the cellular receptor (ACE2), SARS-CoV-2 spike protein binds to the receptor and initiatesthe internalization process. Figure 1a shows a sketch ofthe mechanism. The observed antiviral action of Lf mayconsist of interference in one or more of those steps. Wethus investigate in the next three sections the possibilityof direct binding between lactoferrin and SIA, ACE2, andSARS-CoV-2 Spike protein, respectively (see Figure 1b).
A. Interaction with SIA
Possible binding regions for sialic acid, the termi-nal molecule of the SIA receptors on human lactoferrinare investigated on the basis of the procedure describedin [37], i.e. we select the portion of the molecular surfaceof the MERS-CoV spike protein in interaction with sialicacid, experimentally solved in [45] (Figure 2a), and wesearch for similar patches on the Lf molecular surface.Within the same strategy, we search for similar patcheson the surface of human lactoferrin.In the 2D Zernike framework, the geometrical shape ofa protein surface patch is compactly summarized in a setof ordered numerical descriptors, whose number - 121 inour case - modulate the detail of the description. Deal-ing with ordered numerical descriptors, the comparisonbetween different protein patch can be performed with aEuclidean distance.Figure 2b and c show the four most geometrically sim-ilar patches identified in the Apo and Holo form of Lf.An a posteriori check of the electrostatic potential on thepatches, allows us to select only some of the possible solu-tions identified based on the shape comparison analysis,i.e. the ones having also a similar electrostatic surfacewith the SIA binding site on the MERS-CoV spike.The region on Lf surface identified as the most similarto the MERS region interacting with sialic acid, both inshape and in electrostatics, is the one centered on VAL346.
B. Interaction with ACE2
To check whether the action of lactoferrin can be as-cribed to a competition with the virion spike proteinsin binding directly the ACE2 receptor (Figure 1), weperformed a blind search of the molecular surfaces ofboth ACE2 and human Lf to identify possible bindingregions having a meaningful shape complementarity. Un-der this hypothesis, if the interaction between the Lf andthe ACE2 receptor occurs, Lf could hinder the molec-ular binding between the spike protein of SARS-CoV-2and the corresponding ACE2 receptor. Figure 3 showsthe molecular surface of the ACE2 receptor colored ac-cording to its propensity to bind regions of the Lf protein.The redder the region, the greater the shape complemen-tarity between that region and another one found on thesurface of the putative molecular partner, i.e. holo lacto-ferrin. As one can see from Figure 3, a complementaryregion is indeed present, however, it is located far fromthe binding site of the spike (grey in the figure) and in apart of ACE2 that looks toward the membrane. To bet-ter visualize the result, we have represented two points ofview of the binding between spike and ACE2, one rotated180 o with respect to the other. On the other hand, wecan see that the ACE2 region interacting with the SARS-CoV-2 spike protein has no low shape complementaritywith Lf regions. Virion
HS-PG SIA-PG ACE2
Lf (apo)Lf (holo)S, ACE2 proteinM,N, E proteinLactoferrin (Lf) a) b )b )b ) FIG. 1:
SARS-CoV-2 attachment and entry to host cell in physiological condition and possible actions oflactoferrin. a)
Sketch of SARS-CoV-2 initial interactions with the host cell. Sialoside (SIA) and heparan sulfate (HS) glycanchains present on glycoproteins (PG) of the cell membrane are thought to facilitate the attachment of the virion to the cellsurface. This favors the establishment of an interaction between the virus spike protein and the ACE2 receptor, which startsthe internalization of the virus in the host cell. b) Human lactoferrin has been found to play an antiviral action against SARS-CoV-2 infection although it is not clear whether this action consists in (1) competition for the binding with glycan chains,and/or (2) competition for binding ACE2 receptor and/or (3) direct interaction with one of the proteins in the virion envelope,i.e. with S, M or E proteins.
C. Interaction with virion membrane proteins
A third possible mechanism at the basis of the observedantiviral activity of Lf could be ascribed to a direct inter-action with the membrane proteins present on the virionenvelope. In particular, SARS-CoV-2 presents three dif-ferent kinds of proteins on its membrane, i.e. S, M, andE proteins [46, 47]. While the 3D structure of the S pro-tein has been determined - even if some loop regions inthe S1 sub-unit are not solved - unfortunately, no struc-tures are available for the E and M proteins. Thus, themolecular surfaces of Lf were compared with those of thethree proteins in order to check whether an interactionwith lactoferrin is possible. For this analysis, we adoptedthe same computational procedure used in the previousparagraph. For both S, M, and E proteins, we sample their wholemolecular surface and compared all the possible patcheswith those of Lf. In this way, all molecular surfaces, bothmembrane proteins, and Lf ones are colored according tothe corresponding binding propensity.E and M presented a possible region of interactionslocated in the intra-membrane region (data not shown).The most robust and relevant result of this analysis re-gards the compatibility between the spike and lactoferrin.According to our findings, Lf presents two regions of highcomplementarity with one portion of the C-terminal do-main of the spike S1 subunit and another located in theN-terminal part.To test the reliability of the found signals, we per-formed a molecular dynamics simulation of the spiketrimer (see Methods for details) and sampled 5 config-urations for each of the three chains at equilibrium (see
GLN 165 a) b)c)
ALA 150
LEU 503 ALA 539
PHE 1307 VAL 1346 LEU 1663 ALA 1539
FIG. 2:
Putative sialic-binding regions on human lactoferrin. a)
Cartoon representation of the sialic-acid bindingregion of MERS-CoV (PDB id: 6Q04) and its representation in the Zernike disk (left) and local Coulombic surface (right). b) Four most complementar regions on holo human lactoferrin (PDB id: 1LFG) obtained comparing the lactoferrin molecularsurface with the sialic acid binding region of MERS-CoV. c) Same as in b) but for the apo human lactoferrin (PDB id: 1CB6).Molecular surfaces are colored according to the Coulombic potential.
Figure 4a).In particular, for each chain, we performed a PrincipalComponent Analysis (PCA) on the frames of the dynam-ics and thus projected them on the plane identified by thetwo principal components. Upon clustering these pointswe obtain 5 subgroups. For each subgroup, we extract thecentroidal configuration. Since distant points in the PCAplane correspond to the different 3D structure, pickingone point from each identified cluster assured that theselected configurations have high structural differencesbetween them in the explored configuration space (seeFigure 4b).Remarkably, repeating the blind search for comple-mentarity regions on the 15 surfaces of the extractedspike monomers, we found conservation of the signal overthe conformational noise. Figure 4c) shows the identifiedregions and their conservation in the different sampledframes.Among the found regions, the one involving the spikeC-terminal domain is the one with higher shape comple-mentarity. In particular, according to our method, hu-man lactoferrin could interact with the spike protein withthe surface region centered in residue ALA 539. Alter-natively, the spike protein may interact with the Lf pro- tein using a molecular patch centered in the residue PHE490. It must be pointed out that the complementarityachieved by these patches is comparable with those ofexperimentally solved complexes. Indeed, analyzing theshape complementarity of over 4600 x-ray protein-proteincomplexes (see Methods for details), we have the distri-bution shown in Figure 4d, where the lower the distancethe higher the complementarity of the binding region.Red dotted line shows the complementarity found be-tween the spike and Lt. As it is evident the proposedpatch is characterized by values of shape complementar-ity typical of experimental complexes.Finally, to further support this result with an inde-pendent and external methodology to our approach, weperformed a completely blind molecular docking analy-sis between the spike protein and the Lf protein. Tothis end, Zdock server was used as a state of the art ofmolecular docking software [48]. As per default settings,only the first 10 docking poses have been selected and thepredicted contacts analyzed. In particular, five out of tenposes show bindings involving the spike region involve theregion we found with our protocol when the residues aredefined in contact if their C-alpha atoms have a distanceless than 8 ˚ A . FIG. 3:
Analysis of the binding between lactoferrinand the ACE2 receptor.
Molecular representation of theexperimentally solved complex of the ACE2 receptor withSARS-CoV-2 spike protein. The ACE2 molecular surface iscolored according to its binding propensity to bind lactofer-rin regions. Dark red indicates high binding propensity whilewhite means no interaction.
III. DISCUSSION
In the last decade, the Zernike formalism has beenwidely applied for the characterization of molecular sur-faces [49–53].Very recently, we developed a new representation,based on the 2D Zernike polynomials, which allows anextremely efficient, fast, and completely unsuperviseddescription of the local geometrical shape, allowing foreasy comparison between different regions of molecules.Through this compact description, it is possible both toanalyze the similarity between 2 different regions - sug-gesting, for example, a similar ligand for binding regions- and to study the complementarity between interactingsurfaces [37].Here, we used our novel method to shed some lighton the molecular mechanisms ruling the observed antivi-ral action human lactoferrin exerts against SARS-CoV-2infection [44].In particular, we focused on the early stages of theinfection, i.e. the attachment and entry of the virus to thehost cell, when lactoferrin can interfere with the virus-host interaction without the need to be internalized inthe cell. We thus tried to establish whether lactoferrincould compete with the virus in binding to sialic acid, thesticky end of sialoside chains, which has been suggestedto mediate the attachment of SARS-CoV-2 to the hostcell. Interestingly, comparing the binding region of sialicacid in the MERS-CoV coronavirus with patches on thelactoferrin surface, we found possible spots on both apoand holo forms of Lf, which could compete in forminglow affinity but high avidity interactions.We then proceeded to test the hypothesis of an inter- action between lactoferrin and the primary SARS-CoV-2 protein receptor, ACE2. A blind search for comple-mentarity regions highlighted a hot-spot in a region thatin physiological conditions is oriented toward the mem-brane, while no significative complementarity is presentin the ACE2 region involved in the interaction with thevirus spike protein. At last, we analyzed the three mem-brane proteins on the virus envelope, i.e. the E, M, and Sones. Similarly to ACE-2, both E and M presented possi-ble interacting regions in portions of the surface, that areburied in the virion membrane under normal conditions(data not shown). On the other hand, the spike proteinshowed two main hot spots, one in the N-terminal do-main of the S1 subunit and another in the C-terminalone. Those two regions are robust to molecular noise, asthe signal endures using different configurations sampledfrom a molecular dynamics simulation and each of thethree chains of the trimer. Notably, the most comple-mentary region is the one in the C-terminal region, theone involved in the spike-ACE2 interaction. Thus ourfinding suggests a possible competition between ACE2and lactoferrin for the binding of the SARS-CoV-2 spike,which may explain the observed antiviral action.
IV. MATERIALS AND METHODSA. Datasets
The protein, whose structures are analyzed in this pa-per are: • Human lactoferrin, in the apo (PDB id: 1CB6) andholo (PDB id: 1LFG) forms. • ACE2, in its apo state (PDB id: 1R42). • SARS-CoV-2 S protein, modeled using I-Tasser [54]. • SARS-CoV-2 M protein, modeled using I-Tasser [54]. • SARS-CoV-2 E protein, modeled using I-Tasser [54].To set a reference for the measured complementari-ties, a dataset of protein-protein complexes experimen-tally solved in x-ray crystallography is taken from [55].We only selected pair interactions regarding chains withmore than 50 residues. The Protein-Protein dataset istherefore composed of 4605 complexes. For each com-plex, the binding region is identified as the portions ofthe two protein molecular surfaces distant less than 3 ˚ A . B. Computation of molecular surfaces
For each protein of the dataset (x-ray structure in PDBformat [56]), we use DMS [57] to compute the solvent-accessible surface, using a density of 5 points per ˚ A and trimer chains RMSD residuesresidues f r a m e s f r a m e s a)b) c)d) FIG. 4:
Possible interaction between SARS-CoV-2 spike protein and human lactoferrin. a)
Root mean squaredisplacement as a function of time of the SARS-CoV-2 spike trimer as provided by the molecular dynamics simulation. b) Clustering analysis of the trimer’s A chain in the plane of the two principal components of a PCA analysis over the MDconfigurations. Five regions of major variability of the chain are identified. c) Binding propensity computed from the Zernikedescriptors between the 15 most variable conformations of the chains of the SARS-CoV-2 Spike protein and human lactoferrin.Each residue of the proteins is colored from white to red according to its increasing shape complementarity with the partner. d) Comparison between the complementarity score (red dashed line) of the best binding site (Zernike disks) and the distributionof complementarity scores belonging to 4600 binding regions of experimental complexes. a water probe radius of 1.4 ˚ A . The unit normals vector,for each point of the surface, was calculated using theflag − n . C. Patch definition and complementarityevaluation
A molecular surface is represented by a set of points inthe three-dimensional space. We define a surface patch,as the group of points that fall within a sphere of radius R s = 6˚ A , centered on one point of the surface. Once thepatch is selected, • we fit a plane that passes through the points andreorient the patch in such a way to have the z-axisperpendicular to the plane and going through thecenter of the plane. • we define the angle θ as the largest angle betweenthe perpendicular axis and a secant connecting agiven point C on the z-axis to any point of thepatch. C is then set in order that θ = 45 ◦ . r is thedistance between C and a surface point. • build a square grid and associate each pixel withthe mean r of the points inside it. This 2D func-tion can be expanded on the basis of the Zernikepolynomials (see next section).Once a patch is represented in term of its Zernike de-scriptors, the similarity between that patch and anotherone can be simply measured as the Euclidean distancebetween the invariant vectors. The relative orientationof the patches before the projection in the unitary cir-cle must be considered. In fact, if we search for similarregions we must compare patches that have the same ori-entation once projected in the 2D plane, i.e. the solvent-exposed part of the surface must be oriented in the samedirection for both patches, for example as the positivez-axis. If instead, we want to assess the complementaritybetween two patches, we must orient the patches con-trariwise, i.e. one patch with the solvent-exposed parttoward the positive z-axis (‘up’) and the other towardthe negative z-axis (‘down’). D. 2D Zernike polynomials and invariants
Each function of two variables, f ( r, φ ) (polar coordi-nates) defined inside the region r < f ( r, φ ) = ∞ (cid:88) n =0 m = n (cid:88) m =0 c nm Z nm (1)with c nm = ( n + 1) π (cid:104) Z nm | f (cid:105) == ( n + 1) π (cid:90) drr (cid:90) π dφZ ∗ nm ( r, φ ) f ( r, φ ) . (2)being the expansion coefficients, while the complexfunctions, Z nm ( r, φ ) are the Zernike polynomials. Eachpolynomial is composed by a radial and an angular part, Z nm = R nm ( r ) e imφ . (3)where the radial part for any n and m , is given by R nm ( r ) = n − m (cid:88) k =0 ( − k ( n − k )! k ! (cid:0) n + k − k (cid:1) ! (cid:0) n − k − k (cid:1) ! r n − k (4)Since for each couple of polynomials, the following re-lation holds (cid:104) Z nm | Z n (cid:48) m (cid:48) (cid:105) = π ( n + 1) δ nn (cid:48) δ mm (cid:48) (5)the complete set of polynomials forms a basis andknowing the set of complex coefficients, { c nm } allowsfor a univocal reconstruction of the original image (witha resolution that depends on the order of expansion, N = max ( n )). Since the modulus of each coefficient ( z nm = | c nm | ) does not depend on the phase, i.e. it isinvariant for rotations around the origin of the unitarycircle, the shape similarity between two patches can beassessed by comparing the Zernike invariants of their as-sociated 2D projections. In particular, we measured thesimilarity between patch i and j as the Euclidean dis-tance between the invariant vectors, i.e. d ij = (cid:118)(cid:117)(cid:117)(cid:116) M =121 (cid:88) k =1 ( z ki − z kj ) (6) E. Molecular dynamics simulations
The starting structure of the SARS-CoV-2 spiketrimeric complex was taken from the model structure pro-posed by the I-Tasser server [54]. All steps of the simula-tion were performed using Gromacs 2019.3 [58]. Topolo-gies of the system were built using the CHARMM-27force field [59]. The protein was placed in a dodecahedricsimulative box, with periodic boundary conditions, filledwith 131793 TIP3P water molecules [60]. We checkedthat each atom of the trimer was at least at a distance of1.1 nm from the box borders. The addition of 3 sodiumcounterions rendered the systems electroneutral. The fi-nal system, consisting of 448572 atoms, was first mini-mized with 2064 steps of steepest descent. Relaxationof water molecules and thermalization of the system inNVT and NPT environments were run each for 0.1 nsat 2 fs time-step. The temperature was kept constant at300 K with v-rescale algorithm[61]; the final pressure wasfixed at 1 bar with the Parrinello-Rahman algorithm[62]which guarantees a water density of 1004 kg / m , close tothe experimental value. LINCS algorithm[63] was usedto constraint h-bonds.Finally, the systems were simulated with a 2 fs time-step for 140 ns in periodic boundary conditions, us-ing a cut-off of 12 ˚ A for the evaluation of short-rangenon-bonded interactions and the Particle Mesh Ewaldmethod [64] for the long-range electrostatic interactions. [1] E. M. Redwan, V. N. Uversky, E. M. El-Fakharany,and H. Al-Mehdar, Comptes rendus biologies , 581(2014).[2] B.-L. Waarts, O. J. Aneke, J. M. Smit, K. Kimata,R. Bittman, D. K. Meijer, and J. Wilschut, Virology ,284 (2005).[3] J. H. Andersen, S. A. Osbakk, L. H. Vorland, T. Traavik,and T. J. Gutteberg, Antiviral Research , 141 (2001).[4] M. Marchetti, E. Trybala, F. Superti, M. Johansson, andT. Bergstr¨om, Virology , 405 (2004).[5] P. Drobni, J. N¨aslund, and M. Evander, Antiviral re-search , 63 (2004).[6] P. Puddu, P. Borghi, S. Gessani, P. Valenti, F. Belardelli,and L. Seganti, The International Journal of Biochem- istry & Cell Biology , 1055 (1998).[7] E. Baker and H. Baker, Cellular and Molecular Life Sci-ences , 2531 (2005).[8] F. Giansanti, G. Panella, L. Leboffe, and G. Antonini,Pharmaceuticals , 61 (2016).[9] B. Niaz, F. Saeed, A. Ahmed, M. Imran, A. A. Maan,M. K. I. Khan, T. Tufail, F. M. Anjum, S. Hussain, andH. A. R. Suleria, International Journal of Food Properties , 1626 (2019).[10] M. Sorensen, S. Sorensen, et al., Compte rendu desTravaux du Laboratoire de Carlsberg, Ser. Chim. , 55(1940).[11] M. L. Groves, Journal of the American Chemical Society , 3345 (1960). [12] D. Caccavo, N. M. Pellegrino, M. Altamura, A. Rigon,L. Amati, A. Amoroso, and E. Jirillo, Journal of Endo-toxin Research , 403 (2002).[13] B. van der Strate, L. Beljaars, G. Molema, M. Harmsen,and D. Meijer, Antiviral Research , 225 (2001).[14] A. Langford-Smith, A. J. Day, P. N. Bishop, and S. J.Clark, Frontiers in Immunology (2015).[15] R. Raman, K. Tharakaraman, V. Sasisekharan, andR. Sasisekharan, Current Opinion in Structural Biology , 153 (2016).[16] J. Lang, N. Yang, J. Deng, K. Liu, P. Yang, G. Zhang,and C. Jiang, PLoS ONE , e23710 (2011).[17] M. Yi, S. Kaneko, D. Y. Yu, and S. Murakami, J. Virol. , 5997 (1997).[18] M. C. Harmsen, P. J. Swart, M.-P. d. Bethune,R. Pauwels, E. D. Clercq, T. B. The, and D. K. F. Meijer,Journal of Infectious Diseases , 380 (1995).[19] K. McCann, A. Lee, J. Wan, H. Roginski, and M. Coven-try, Journal of Applied Microbiology , 1026 (2003).[20] F. Superti, R. Siciliano, B. Rega, F. Giansanti, P. Valenti,and G. Antonini, Biochimica et Biophysica Acta (BBA)-General Subjects , 107 (2001).[21] P. Puddu, P. Borghi, S. Gessani, P. Valenti, F. Belardelli,and L. Seganti, The International Journal of Biochem-istry & Cell Biology , 1055 (1998).[22] F. Berlutti, F. Pantanella, T. Natalizi, A. Frioni, R. Pae-sano, A. Polimeni, and P. Valenti, Molecules , 6992(2011).[23] C. Huang, Y. Wang, X. Li, L. Ren, J. Zhao, Y. Hu,L. Zhang, G. Fan, J. Xu, X. Gu, et al., The Lancet ,497 (2020).[24] N. Zhu, D. Zhang, W. Wang, X. Li, B. Yang, J. Song,X. Zhao, B. Huang, W. Shi, R. Lu, et al., New EnglandJournal of Medicine (2020).[25] C. Drosten, S. G¨unther, W. Preiser, S. Van Der Werf,H.-R. Brodt, S. Becker, H. Rabenau, M. Panning,L. Kolesnikova, R. A. Fouchier, et al., New England jour-nal of medicine , 1967 (2003).[26] T. G. Ksiazek, D. Erdman, C. S. Goldsmith, S. R. Zaki,T. Peret, S. Emery, S. Tong, C. Urbani, J. A. Comer,W. Lim, et al., New England journal of medicine ,1953 (2003).[27] A. M. Zaki, S. Van Boheemen, T. M. Bestebroer, A. D.Osterhaus, and R. A. Fouchier, New England Journal ofMedicine , 1814 (2012).[28] E. Prompetchara, C. Ketloy, and T. Palaga, Asian PacificJ. allergy Immunol (2020).[29] S. Su, G. Wong, W. Shi, J. Liu, A. C. Lai, J. Zhou,W. Liu, Y. Bi, and G. F. Gao, Trends in microbiology , 490 (2016).[30] F. Li, W. Li, M. Farzan, and S. C. Harrison, Science ,1864 (2005).[31] F. Li, Journal of virology , 6984 (2008).[32] W. Li, C. Zhang, J. Sui, J. H. Kuhn, M. J. Moore, S. Luo,S.-K. Wong, I.-C. Huang, K. Xu, N. Vasilieva, et al., TheEMBO journal , 1634 (2005).[33] P. Zhou, X.-L. Yang, X.-G. Wang, B. Hu, L. Zhang,W. Zhang, H.-R. Si, Y. Zhu, B. Li, C.-L. Huang, et al.,Nature pp. 1–4 (2020).[34] R. L. Graham and R. S. Baric, Journal of virology ,3134 (2010).[35] L. Kuo, G.-J. Godeke, M. J. Raamsman, P. S. Masters,and P. J. Rottier, Journal of virology , 1393 (2000).[36] R. Yan, Y. Zhang, Y. Li, L. Xia, Y. Guo, and Q. Zhou, Science (2020).[37] E. Milanetti, M. Miotto, L. Di Rienzo, M. Monti,G. Gosti, and G. Ruocco, arXiv:2003.11107 pp. arXiv–2003 (2020).[38] A. Vandelli, M. Monti, E. Milanetti, R. D. Ponti, andG. G. Tartaglia, arXiv preprint arXiv:2003.13655 (2020).[39] L. Liu, P. Chopra, X. Li, M. A. Wolfert, S. M. Tompkins,and G.-J. Boons, bioRxiv (2020).[40] B. Robson, Computers in Biology and Medicine p. 103849(2020).[41] C. Schwegmann-Weßels and G. Herrler, Glycoconjugatejournal , 51 (2006).[42] M. A. Tortorici, A. C. Walls, Y. Lang, C. Wang, Z. Li,D. Koerhuis, G.-J. Boons, B.-J. Bosch, F. A. Rey, R. J.de Groot, et al., Nature structural & molecular biology , 481 (2019).[43] R. J. Hulswit, Y. Lang, M. J. Bakkers, W. Li, Z. Li,A. Schouten, B. Ophorst, F. J. van Kuppeveld, G.-J.Boons, B.-J. Bosch, et al., Proceedings of the NationalAcademy of Sciences , 2681 (2019).[44] G. Serrano, I. Kochergina, A. Albors, E. Diaz, M. Oroval,G. Hueso, and J. M. Serrano, International Journal OfResearch In Health Sciences (2020).[45] Y.-J. Park, A. C. Walls, Z. Wang, M. M. Sauer, W. Li,M. A. Tortorici, B.-J. Bosch, F. DiMaio, and D. Veesler,Nature Structural & Molecular Biology , 1151 (2019).[46] I. Seah, X. Su, and G. Lingam, Eye , 1155 (2020).[47] M. Bianchi, D. Benvenuto, M. Giovanetti, S. Angeletti,M. Ciccozzi, and S. Pascarella, BioMed Research Inter-national , 1 (2020).[48] R. Chen, L. Li, and Z. Weng, Proteins: Structure, Func-tion, and Genetics , 80 (2003).[49] L. Di Rienzo, E. Milanetti, R. Lepore, P. P. Olimpieri,and A. Tramontano, Scientific reports , 1 (2017).[50] S. Daberdaku and C. Ferrari, Bioinformatics , 1870(2019).[51] D. Kihara, L. Sael, R. Chikhi, and J. Esquivel-Rodriguez,Current Protein and Peptide Science , 520 (2011).[52] L. Di Rienzo, E. Milanetti, J. Alba, and M. DAbramo,Journal of Chemical Information and Modeling , 1390(2020).[53] V. Venkatraman, Y. D. Yang, L. Sael, and D. Kihara,BMC bioinformatics , 407 (2009).[54] A. Roy, A. Kucukural, and Y. Zhang, Nature Protocols , 725 (2010).[55] P. Gainza, F. Sverrisson, F. Monti, E. Rodol`a,D. Boscaini, M. Bronstein, and B. Correia, Nature Meth-ods , 184 (2020).[56] H. M. Berman, P. E. Bourne, J. Westbrook, andC. Zardecki, in Protein Structure (CRC Press, 2003), pp.394–410.[57] F. M. Richards, Annual review of biophysics and bioengi-neering , 151 (1977).[58] D. V. D. Spoel, E. Lindahl, B. Hess, G. Groenhof, A. E.Mark, and H. J. C. Berendsen, Journal of ComputationalChemistry , 1701 (2005).[59] B. R. Brooks, C. L. Brooks, A. D. Mackerell, L. Nilsson,R. J. Petrella, B. Roux, Y. Won, G. Archontis, C. Bartels,S. Boresch, et al., Journal of Computational Chemistry , 1545 (2009).[60] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W.Impey, and M. L. Klein, The Journal of Chemical Physics , 926 (1983).[61] G. Bussi, D. Donadio, and M. Parrinello, The Journal of Chemical Physics , 014101 (2007).[62] M. Parrinello and A. Rahman, Physical Review Letters , 1196 (1980).[63] B. Hess, H. Bekker, H. J. C. Berendsen, and J. G. E. M.Fraaije, Journal of Computational Chemistry , 1463 (1997).[64] T. E. I. Cheatham, J. L. Miller, T. Fox, T. A. Darden,and P. A. Kollman, Journal of the American ChemicalSociety117