Mumps is an infectious disease with a high morbidity in non-immunized children, caused by a respiratory-droplets transmitted virus, known as mumps virus. Usually, it causes swollen salivary glands under the ears, referred to as parotitis; other symptoms include fever, headache, muscle aches and tiredness. Especially in adults, mumps can also cause orchitis, mastitis, pancreatitis, encephalitis and meningitis. Despite the advent of an effective vaccine, mumps virus continues to circulate throughout the world and although the majority of mumps infections are subclinical in vaccinated individuals, severe complications still occur especially in underdeveloped countries. Moreover, in the last years, mumps outbreaks among vaccinated young adults have been also reported in different countries (Willocks et al., 2017; Donahue et al., 2020).

Mumps virus is a member of the Paramyxoviridae family, belonging to the genus Orthorubulavirus, related to parainfluenza and Newcastle disease viruses, containing non-segmented negative-strand RNA genome. Different glycoproteins spikes protrude away from the viral surface, playing key roles in the different steps of the infection process as host receptor recognition, binding and cleavage. Among the 8 gene-encoded viral glycoproteins, exhibiting hemagglutinin, neuraminidase and cell fusion activity, hemagglutinin-neuraminidase (HN) has attracted attention as potential target for the development of specific and potent inhibitors able to regulate the enzyme functions. This glycoprotein is a specific sialic-acid-binding lectin able to recognize host cell-surface glycans which serve as attachment points for the virus, allowing not only its entry into the host cells but also the spreading of the infection. HN protein, indeed, also participates in the internalization of viral particles activating the fusion protein and permitting the fusion of viral membranes to the host cell. Finally, HN prevents self-agglutination of viral particles and favors the release of virions from the infected cells since it acts as a sialidase, removing the sialic acid moiety from viral progeny (Lamb and Kolakofsky, 2001). Thus, the disruption of the interactions between virus-related HN protein and host glycans may prevent the virus attachment to host cells as well as its multiplication and release from the infected cells.

With the goal to provide the basis for the design and development of novel HN inhibitors, we here report the characterization of the molecular interaction of MuV-HN protein from SBL-1 strain with α-2,3-linked sialic acid containing ligands. In detail, we described the kinetic parameters that characterize the hydrolysis of the sialyllactosamine, 3’SLN, catalyzed by the HN protein; we identified the epitope mapping of the ligand when interacting with the protein and we determined the three-dimensional structure of the protein-glycan complex. To date, the crystal structure of the MuV-HN from SBL-1 strain has not been solved yet; thus, with the aim to achieve a 3D perspective of the molecular recognition mechanism, the homology model of the protein has been built. Finally, we synthesized a potential inhibitor of HN neuraminidase and elucidated by NMR, fluorescence analysis and computational studies its interaction with the viral protein.

Receptor motifs that can be recognized by MuV-HN (hemagglutinin-neuraminidase) protein from SBL-1 strain have been recently identified by glycan array screening (Kubota et al., 2019) as specific host-cell sialoglycans which terminate with neuraminic acid (Neu5Ac) moiety. Among them, Neu5Acα2-3Galβ1-4GlcNAc (3’SLN), a sialoglycan broadly expressed in and exposed on various host tissues, was reported as one of the strongest binders; thus, we decided to investigate, at molecular level, its recognition by MuV-HN SBL-1 strain by means of NMR spectroscopy, biophysical and computational approaches.

Given the enzymatic activity of MuV-HN SBL-1, upon addition of the sialoglycan substrate in solution, MuV-HN protein catalyzed the cleavage of the terminal Neu5Ac moiety, producing N-Acetyllactosammine (LacNAc, residues A’ and B’) and reducing sialic acid (red-Neu5Ac, K’) (Figure 1A). Therefore, the real time kinetics of 3’SLN hydrolysis by MuV-HN was followed by 1D 1H-NMR, as reported in Figure 1B, where the overlapped NMR spectra detected over time progression are shown. The progress of the hydrolysis was assessed by monitoring the decrease of the signals of the substrate (as B1 and K3_eq) and the concurrent appearance and following enhancement of the intensity of the products resonances (as B’1 and K’3_eq). Similarly to other viral sialidases, (Chan et al., 2012), MuV-HN acts as retaining glycosidase, with a net retention of substrates configuration. The H3_eq resonance of Neu5Ac was monitored to evaluate the variation of the substrate concentration and, by fitting the kinetic data through the Lambert W function (Goudar et al., 2004), the values of the enzyme K_M value of 5 μM and V_max of 6*10⁻³ mM/min were determined (Figure 1B). The low value of the K_M corresponds to high affinity for 3’SLN substrate.

Notably, the K_M evaluated for 3’SLN/MuV-HN from SBL-1 strain was lower with respect to that observed by our group for the same substrate in the interplay with MuV-HN from Hoshino strain. (K_M 12 μΜ) (Forgione et al., 2020), suggesting a higher enzymatic affinity of the protein from SBL-1 strain for 3’SLN.

Despite the neuraminidase activity of MuV-HN induces the cleavage of the sialic acid within NMR time scale, saturation transfer difference (STD) NMR allowed to identify the ligand binding epitope (Figure 1C) (Di Carluccio et al., 2021). A qualitative analysis of the STD effects demonstrated that the trisaccharide was entirely accommodated into the binding pocket of the protein. The moiety most contributing to the interaction was the neuraminic acid residue (K) which established several contacts with the protein, not only through its acetamide portion but also via the lateral chain; the highest STD effect, set at 100%, belonged indeed to the N-acetyl group; protons H7–H8–H9 also showed significant STD enhancements; finally, STD signals were detected for some protons of the Neu5Ac carbohydrate ring, as H4, H5 and H6. As already observed in the case of MuV-HN from Hoshino strain, also the other two residues (A and B) were involved in the interaction with the protein although to a lower extent. Furthermore, by evaluating the relative intensities of two N-acetyl groups of Neu5Ac K and GlcNAc A, it was clear that the contribution from the acetamide of the GlcNAc was not significant.

To get structural insights into sialoglycans recognition by MuV-HN (SBL-1 strain, genotype A), comparative homology modelling was carried out considering as template the crystal structure of MuV-HN head domain (Hoshino strain, genotype B), PDB ID: 5B2D, which possesses 95% of structural similarity to the target sequence (Figure 2A), in agreement with the low number of genomic mutations occurring among different MuV genotypes at HN encoding gene (Palacios et al., 2005). From the sequence analysis of the two different strains (Figures 2A,B and Supplementary Figure S11), the most relevant modifications of the amino acid sequences occurred at the known B-cell epitopes, particularly at 327–363; 375–403; 440–443; and 533 regions, as similarly detected when comparing other MuV-HN genotypes (Gouma et al., 2020).

The homology model of HN protein from SBL-1 strain was then built using Swiss Model server (Waterhouse et al., 2018), and, after refinement by means of Maestro (Schrödinger Suite, 2019), its quality was assessed through SAVES resource (https://saves.mbi.ucla.edu/). As result, the refined model, analyzed using ERRAT plot, exhibited an overall quality factor of 87.11. The Ramachandran plots generated by means of the PROCHECK tool (Supplementary Figure S11B) allowed to assess the stereochemical quality of the resulting structure (Laskowski et al., 1993), showing that 99% of the residues’ dihedrals belonged to the core/allowed regions, hence confirming the fitness of the model. Moreover, the 3D profile of the model structure obtained by VERIFY 3D, scoring the compatibility of the 3D structure model with respect to amino acid sequence, showed a scoring above 0.2 for above 80% of the amino acid residues, further confirming the accuracy of the model (Supplementary Figure S11C). Comparing our refined model with the target structure, a similar topology was noted, and the RMSD calculation in reference to the MuV-HN crystal structure of 0.05 Å further denoted its remarkable quality, in full agreement with the high sequence identity between the homologous proteins from different MuV strains (Supplementary Figure S11A). Indeed, as predicted by the sequence analysis, the changes in the amino acid sequence among the proteins from the two different strains did not affect neither the overall tridimensional structure of the sialic acid binding motif nor the accessibility of the protein to the glycans on the host cell surface. In detail, MuV-HN (SBL-1 strain) model showed a globular structure, characterized by six four-stranded antiparallel β−sheets (β1-β6) arranged in the so-called bladed beta propeller fold, that is characteristic of many viral neuraminidases (Colman et al., 1993) (Figure 2B). At the center of the domain, it was located the dual binding/sialic acid cleavage site of the protein, comprising the conserved sialidase active site residues, namely Arg180, Glu407, Arg422, Arg512, Tyr540, Glu561, and Asp204 (Kubota et al., 2016). Furthermore, the binding site comprised the Tyr369 and Val476, highly conserved among the MuV-HN genotypes, that revealed to be important determinants of Mumps recognition (Kubota et al., 2016; Forgione et al., 2020; Gouma et al., 2020).

As several studies highlighted the importance of the flexibility of specific loops in other viral sialidases (Winger et al., 2012; Kubota et al., 2019), the MuV-HN conformational behavior was explored by MD simulations. First, the MD simulation of the apo form of MuV-HN SBL-1 strain model was carried out using Amber18 (Case et al., 2018). As result, it was noted that MuV-HN showed a slight conformational rearrangement along the simulated time, with a small increase of the backbone RMSD which converged to 1.5 Å (Supplementary Figure S12A). The inter-strand loops RMSD was calculated (Supplementary Figure S12B) showing that the fluctuations from the initial geometry affected mostly β5 and β6 loops. No relevant conformational changes were instead detected when calculating the β1, β2 and β3 inter-strand loops RMSD along the trajectory (Supplementary Figure S12B). Furthermore, the atom-positional root-mean-square fluctuations (RMSFs) relative to the Cα-atoms was calculated along the trajectory, denoting smaller values in the secondary structure regions, further reinforcing the homology model quality (Supplementary Figure S12C). Interestingly, large fluctuations occurred in the loop regions, especially at the positions corresponding to the β4 and β6 inter-strand loops. In particular, the β4 showed high flexibility from the RMSF plot (Supplementary Figure S12B). Nonetheless, the highest flexibility belonged to the long loop connecting β4 and β5 sheets.

Our model was subsequently employed for docking calculation by Autodock 4.2 (Morris et al., 2009) program in the presence of 3’SLN as ligand. Based on the predicted binding energy and cluster population (Supplementary Table S1), the best pose of 3’SLN/MuV-HN complex was selected (Figures 2C,D) and it was submitted to 100 ns MD run using AMBER 18 (Supplementary Figures S13A–E). The stability of the complex has been assessed by the ligand RMSD calculated using the protein as reference; the results showed that the ligand remained stable around 1 Å for almost the entire simulation, thus indicating it was anchored to the receptor along the simulated time. Moreover, the notable structural similarity between the most populated MD clusters calculated with respect to the ligand RMSD along the trajectory further confirmed that the sialotrisaccharide didn’t exhibited conformational changes along the simulation (Supplementary Figures S13A,B).

In accordance with NMR data, computational analysis showed that all three sugar moieties were in contact with the receptor surface, as reported in Figure 2C. The interaction network of the selected complex showed many similarities with those observed for MuV-HN from Hoshino strain (Kubota et al., 2016; Kubota et al., 2019; Forgione et al., 2020). In detail, the sialic acid residue established strong interactions between its carboxylate moiety and three arginine residues, namely Arg180, Arg422 and Arg512. The N-Acetyl moiety formed the hydrogen bond with Glu407 and hydrophobic interactions with Ile181 and Val308. The GlcNAc ring was engaged in CH-pi interactions with Tyr369. Gal residue was in proximity to the receptor surface but no significant polar interactions were observed, as its hydroxyl groups were directed toward the solvent. Nonetheless, the Gal moiety was in proximity to the Val476 side chain thus entailing hydrophobic contacts with MuV-HN. Additionally, the analysis of the protein-ligand interactions showed that the main contacts were maintained along all the trajectory, including the electrostatic interactions of the Neu5Ac unit with the Arg triad, and the stacking interaction between Tyr369 and the GlcNAc moiety (Supplementary Figures S13C–E). This further suggested the importance of such residues to the molecular recognition of sialoglycans by Mumps virus hemagglutinin-neuraminidase receptor.

In order to verify if a particular loop stabilization occurred in the MuV-HN bound form, the RMSF and loop RMSD analysis of the 3’SLN/MuV-HN complex was carried out and compared to the free state. As a result (Supplementary Figures S14A,B), lower fluctuations upon binding of the inter-strands loops of β5 and β6 sheets were observed. A significant stabilization of the inter-sheet loops, namely β2-β3, β4-β5 and β5-β6 was also noted. Interestingly, the lower atomic fluctuations displayed by the loops of β4, β5 and β6 sheets matched with the MuV-HN region binding to the 3’SLN ligand, therefore implying that the stabilizing interactions with the sialotrisaccharide influenced the involved loops conformational flexibility.

Given that from NMR and computational analysis the entire trisaccharide seemed to be involved in the recognition and binding process, being the sialic acid the moiety anchoring the glycan to the binding pocket of the protein, we decided to synthesize a thioderivative of 3’SLN, namely thio-3’SL, as a potential HN inhibitor. The incorporation of a sulfur functional group on the anomeric position of the sialic acid should indeed affect the lability of the Sia-Gal glycosidic linkage without hampering the recognition of the sialoglycan by the enzyme.

The thio-3’SL (1) was prepared (see Scheme 1) starting from the thrichloroacetimidate 2, (Turnbull et al., 2000), and the methyl-2,3,6-tri-O-benzyl-β-D-glucopyranoside (3), (Yoneda et al., 2005), both prepared accordingly to literature methods. Glycosylation, catalyzed by BF₃∙Et₂O, gives the protected trisaccharide 4 with a 50% yield. Removal of benzyl protective groups by hydrogenation and of the acetylic groups with MeONa in MeOH was followed by hydrolysis of the methyl ester with aqueous NaOH to give the thiotrisaccharide 1 with a 75% yield on three steps.

The hydrolytic stability of the thio-3’SL and its ability to interact with the MuV-HN protein was hence investigated by means of NMR spectroscopy and fluorescence analysis. As expected, the thio-3’SL was not hydrolyzed by MuV-HN, as it was observed by monitoring over time the NMR spectra of MuV-HN/thio-3’SL mixture (data not shown). However, it was recognized by MuV-HN protein as suggested by fluorescence data. The binding interaction of thio-3’SL with MuV-HN from SBL-1 strain was indeed analyzed by fluorescence spectroscopy exploiting the presence of aromatic amino acids in the protein binding pocket. In detail, changes in the fluorescence intensity of tryptophane residues of MuV-HN upon the addition of increased amount of thio-3’SL were followed (Figure 3A left panel). The interaction was quantitatively evaluated by fitting the resulting binding curve achieved with the application of the quenching data elaboration described by Ribeiro et al. (Ribeiro et al., 2008). As a result, a binding constant (K_b) value of 0.44 ± 0.05 was obtained. (Figure 3A, right panel).

The molecular characterization of the complex has been then carried out by the combination of NMR and computational studies. The ligand interacting epitope has been indeed described by STD NMR experiments (Figure 3B). As expected, the strongest STD enhancements were observed for the sialic acid moiety and especially for its acetyl group and protons of Neu5Ac lateral chain (H7-H8-H9). Significant STD effects were further detected for protons H4-H5 and H6 of sialic acid ring. Lower STD signals (below 50%) were instead observed for protons belonging to galactose and glucose residues indicating that they pointed farer from the protein surface. The thio-3’SL structure was then modelled into the binding site of MuV-HN SBL-1. The docking calculations allowed to describe the relevant receptor-ligand binding interactions, highlighting that the protein accommodated the synthetic analogue and the 3’SLN with a similar binding mode. Indeed, the superimposition of the two complexes (Figure 4A) clearly showed a nearly identical orientation of the ligands into the receptor binding site. The interaction network was overall conserved; the Neu5Ac carboxylate engaged electrostatic interactions with the receptor Arg180, Arg422 and Arg512; Glu407, Glu264, and Tyr323 binding pocket residues were also interacting with sialic acid unit. Moreover, the Glc moiety of the trisaccharide formed hydrophobic interactions with Tyr369 and Val476 residues; this latter residue was also in close contact with Gal moiety.

Hemagglutinin-Neuraminidase activity is essential for the infection and propagation of viruses belonging to the Paramyxoviridae family, including parainfluenza (hPIV), the Newcastle disease and mumps viruses. Over the past years, different series of inhibitors have been developed toward viral neuraminidases, especially targeting hPIVs. However, the design and development of novel inhibitors and high-affinity ligands could prove worthwhile for a better understanding of virus tropism and pathogenesis and may help in the fight against viral infections allowing the advancement of new licensed anti-viral drugs.

Here we focused our attention on mumps virus, the leading cause of the mumps disease that could affect the central nervous system, causing meningitis. Mumps virus possesses different surface glycoproteins and among them, the HN represents a multitasking protein mediating both the early and late stages of viral infection, including the host-cell sialoglycans recognition, the trigger of virus and host-cell membranes fusion and finally the release of progeny virions from infected cells. MuV-HN thus represents an attractive target for the structure-based design and development of novel anti-viral drugs since interfering with these processes may affect the viral pathogenicity and infectivity.

Thus, we investigated the structural features of MuV-HN from SBL-1 strain when bound to host-cell sialoglycans with the aim to exploit them as basis for the guided development of tailored inhibitors. The structural characterization of the HN protein in the interaction with the 3’SLN has been performed by means of NMR and computational studies. In detail, the kinetic of 3’SLN hydrolysis catalyzed by MuV-HN from SBL-1 strain was characterized by progress curve analysis derived by NMR assay. Notably, the kinetic parameters obtained by the elaboration through the Lambert W function were in the same range of the values obtained for the same ligand interacting with MuV-HN from Hoshino strain, as shown in our previous work (Forgione et al., 2020). STD NMR allowed to describe the ligand interacting epitope stressing the importance of the sialic acid in the recognition and interaction process, as expected, but also demonstrating that the binding pocket of the protein was pliable enough to accommodate the three composing sugars of the 3’SLN.

The homology modelling of the protein was also carried out with the aim to achieve a 3D view of the protein-ligand complex. The crystal structure of MuV-HN from the Hoshino strain was used for building the model; that revealed a huge similarity between the 3D structures. It is worth to note that, from the sequence analysis of the two proteins, few variable positions were observed. Few of the mutations affected the important functional regions, as the known B-cell epitopes; the most relevant is the 354 (P→Q) mutation at the 327–363 region, that modifies the physiochemical features of the surface exposed area. This variation was already suggested to be significant for altered antibody recognition and neutralization by MuV-HN (Gouma et al., 2018; Gouma et al., 2020). Notably, the neuraminidase binding site was preserved, differently with respect to that observed when comparing the sequences of other MuV-HN genotypes (Gouma et al., 2018; Gouma et al., 2020). The refined 3D model of the protein has been used for docking analysis and MD simulations of the MuV-HN/3’SLN complex. Our results demonstrated that the main interactions involving the three sugars of the glycan receptor observed when bound to MuV-HN from Hoshino strain were completely conserved in the binding with MuV-HN from SBL-1 strain. We indeed observed that the Neu5Ac residue was deeply located in the receptor pocket where it established strong polar interactions with the canonical neuraminidases sialic acid recognizing residues, including the Arginine triad (Arg180, Arg422, Arg512). Nonetheless, the previously described interactions involving Tyr369, which were reported to strongly influence the binding affinity of MuV-HN from Hoshino strain, were also present in the case of MuV-HN from SBL-1 strain (Kubota et al., 2016). Interestingly, the MD simulation studies performed on the apo and bound forms of MuV-HN revealed the stabilization of particular inter- strand- and inter-sheet loops upon the ligand binding, thus suggesting that the receptor loop flexibility may play a relevant role in the molecular recognition of sialoglycans by mumps virus hemagglutinin neuraminidase and mechanism of catalysis as well. Indeed, in the bound state, lower fluctuations were observed for loop regions in intimate contact with the ligand bound to Muv-HN, on which some of the binding site residues were located, as Val476 on β4-β5 loop and Arg512 on β5 loop. Previous studies on the conformational behavior of other neuraminidases (Winger et al., 2012; Amaro et al., 2009; Amaro et al., 2011) reported a shift from open to closed loop conformation upon sialoglycans binding, thus, we also inspected this possibility by measuring the distance across the loops delimiting the binding site (Figure 5). Interestingly, the distance measured between the binding pocket walls showed an increased value for the apo-MuV-HN protein, thus suggesting that also for the mumps virus hemagglutinin neuraminidase the glycan receptor specificity may be tuned by loops flexibility.

Furthermore, we performed the synthesis of a thiotrisaccharide, thio-3’SL, as a potential inhibitor of the neuraminidase activity of the protein. The ability of thio-3’SL to interact with MuV-HN from SBL-1 strain was demonstrated by monitoring the intrinsic fluorescence of the tryptophane residues of the receptor and by NMR. Moreover, modelling analyses provided a picture of the protein-ligand complex, showing that the thio-3’SL established contacts with MuV-HN receptor which were comparable to those made from 3’SLN, consistently to the similar bound ligands’s topology. The main interactions mostly involved the Neu5Ac unit, that interacted with the characteristic receptor binding site amino acids Arg180, Arg422, Arg512, Glu407 and Tyr323. Accordingly, the third sugar exhibited stacking interactions with the Tyr369.

In conclusion, a combination of organic chemistry, NMR, fluorescence and computational studies could provide a 3D view of the MuV-HN from SBL-1 strain when interacting with the natural substrate, 3’SLN, and the synthetic thio-3’SL, showing a similar binding mode. Our analysis demonstrated that, in both cases, the main interactions involving the three sugars of the glycan receptor were also comparable to those observed when investigating MuV-HN from Hoshino strain (Kubota et al., 2016).

Our outcomes may help in the identification of new inhibitor scaffolds which could prove worthwhile in the fight against mumps virus.