Exploring the Effect of Structure-Based Scaffold Hopping on the Inhibition of Coxsackievirus A24v Transduction by Pentavalent N-Acetylneuraminic Acid Conjugates

Coxsackievirus A24 variant (CVA24v) is the primary causative agent of the highly contagious eye infection designated acute hemorrhagic conjunctivitis (AHC). It is solely responsible for two pandemics and several recurring outbreaks of the disease over the last decades, thus affecting millions of individuals throughout the world. To date, no antiviral agents or vaccines are available for combating this disease, and treatment is mainly supportive. CVA24v utilizes Neu5Ac-containing glycans as attachment receptors facilitating entry into host cells. We have previously reported that pentavalent Neu5Ac conjugates based on a glucose-scaffold inhibit CVA24v infection of human corneal epithelial cells. In this study, we report on the design and synthesis of scaffold-replaced pentavalent Neu5Ac conjugates and their effect on CVA24v cell transduction and the use of cryogenic electron microscopy (cryo-EM) to study the binding of these multivalent conjugates to CVA24v. The results presented here provide insights into the development of Neu5Ac-based inhibitors of CVA24v and, most significantly, the first application of cryo-EM to study the binding of a multivalent ligand to a lectin.


Introduction
Viral pandemics have over the course of human civilization, caused the suffering and deaths of millions of human beings [1]. The severe consequences of viral outbreaks and pandemics are currently on global display, with the ravaging covid-19 pandemic that have crippled societies causing serious and, potentially, long-lasting socio-economic effects at both a local level and on a global scale.
Coxsackievirus A24 variant (CVA24v), an antigenic variant of coxsackievirus A24 (CVA24), has over the last decades caused two pandemics and several recurring outbreaks of the highly contagious eye infection acute hemorrhagic conjunctivitis (AHC) with >10 million estimated cases of the disease [2,3]. AHC is characterized by an abrupt onset of ocular pain, swelling of the eyelids, watering, a foreign body sensation, and extensive subconjunctival hemorrhaging [4]. The infection is highly contagious with short incubation periods (24-48 h) that result in explosive outbreaks in communities, affecting up to 50% of the population [3][4][5]. AHC generally resolves spontaneously within 1-2 weeks, and although severe conditions are rare, the development of acute flaccid paralysis and even fatalities have been reported [5,6]. Treatment of AHC is supportive as neither vaccines nor antivirals are currently available to combat the infection.
To facilitate attachment and subsequent entry into cells, CVA24v engages glycancontaining receptors that terminate in 5-N-acetyl-neuraminic acid (Neu5Ac) and the intercellular adhesion molecule-1 (ICAM-1), with ICAM-1 serving as the primary receptor [7]. Binding to ICAM-1 is essential for productive replication of CVA24, including variant and non-variant strains, as binding initiates uncoating by triggering destabilizing rearrangements of the capsid proteins that eventually result in the formation of a capsid-mediated pore through which the viral genome is released into the cytoplasm of cells [7]. However, the virulence and pandemic potential of CVA24v seems to be linked to enhanced efficiency in binding Neu5Ac, which favors a tropism switch towards the eye [7]. This adaptation has further been suggested to promote CVA24v attachment to corneal and conjunctival cells [8] and thereby, resulting in enhancing the transmission of viral progeny [7]. CVA24v binds Neu5Ac glycans via binding sites located within viral protein 1 (VP1) that together cluster around the five-fold symmetry axis of the virion, in essence, forming pentameric Neu5Ac lectins [9].
Protein-carbohydrate interactions are of vital importance in a wide range of biological processes in health and disease. Yet, only a few carbohydrate-based drugs have reached the market. This may be due to challenges associated with overcoming the generally weak affinities of protein-carbohydrate interactions and the poor pharmacological properties of carbohydrates as a result of their high polarity and metabolic vulnerability [10]. In the case of CVA24v and many other viruses that employ Neu5Ac for host cell attachment, these issues can potentially be circumvented by design of multivalent ligands intended for a topical administration as the primary sites of replication are linked to the eyes and airways. Importantly, direct topical administration to the site of infection would circumvent the challenges posed by oral administration and systemic distribution of a carbohydrate drug. Indeed, the design of multivalent carbohydrate ligands is a proven strategy used to achieve high avidity binding to lectins (carbohydrate-binding proteins) via several mechanisms, including the chelate effect, subsite binding, steric stabilization, statistical rebinding, and receptor clustering [11]. Further, multivalent ligands tailored to lectin valency and topology have been demonstrated to result in rather exceptional enhancements in avidity [12,13]. Other promising approaches include repurposing existing topical eye medications to be used as broad-spectrum antivirals [14].
Bioisosteric replacements are broadly defined as the replacement of, e.g., functional groups or a scaffold part of a bioactive molecule with a substructure that is similar in size and properties. The strategy is widely used in compound optimization with the aim of improving target binding, drug-like properties, reduce metabolic vulnerability, or reduce toxicity. Replacement of the central scaffold of an active molecule with another scaffold has in recent years gained increased attention [15]. This is the result of, e.g., development of computational methods referred to as "scaffold hopping" [16] and the realization that the diversity of scaffolds in approved drugs are relatively low [17]. Whether the selection of a novel scaffold for an active compound is based on "scaffold hopping" methodology or on traditional structure-based reasoning [18], the general idea is to retain enough similarity for the compound to remain active, normally achieved by keeping key substituents for binding. This may result in altering the properties of the molecule or may move it into novel chemical space [15,16]. Scaffold-replacement has been demonstrated to result in improved binding affinity [19] and cell permeability [20], and reduced toxicity [21]. Further, decreased flexibility of the scaffold can increase the binding affinity of a molecule as the loss of conformational entropy is reduced upon binding of the ligand to its target [22].
To date, pentavalent Neu5Ac conjugates based on a glucose scaffold are the most efficient reported inhibitors of CVA24v attachment and infection of human corneal epithe-lial (HCE) cells [23]. Herein, we report on the synthesis of scaffold-replaced pentavalent Neu5Ac conjugates and, to the best of our knowledge, the first use of cryogenic electron microscopy (cryo-EM) to study the dynamic nature of a multivalent ligand binding to CVA24v.

Design
A total of 60 Neu5Ac binding sites decorate the CVA24v capsid. These are arranged into 12 relatively flat, local five-fold symmetries, each providing five shallow, surface exposed Neu5Ac binding sites [9]. We previously reported on the design and synthesis of pentavalent Neu5Ac conjugates with pseudoradial topology that inhibit CVA24v attachment to, and infection of, HCE cells [23]. These inhibitors were designed to chelate each of the pentameric lectins of CVA24v. Crystal structures of inhibitor-bound CVA24v confirmed compound binding, with observed electron density for the Neu5Ac units. However, negative-staining electron microscopy suggested that the major mode of binding and inhibition of these inhibitors likely resulted from cross-linking and aggregation of virus particles [23]. Thus, in order to further explore, and achieve, tailored chelation of the Neu5Ac binding sites, a modification of the strategy was needed.
We reasoned that replacement of the glucose scaffold of the pentavalent Neu5Ac conjugates potentially could result in direct contacts between the scaffold and the viral capsid and thus result in improved efficacy. We hypothesized that tailored pentavalent Neu5Ac conjugates with symmetric topology may facilitate detection of the remaining parts of the inhibitors as the number of unique binding modes are significantly reduced, and could thus result in more ordered binding events to the viral capsid. Theoretically, this should result in higher avidity binding due to reduced loss of entropy in the inhibitor-lectin interaction [24][25][26]. Further, although many different multivalent scaffold architectures have been developed, the effect of scaffold-replacements in the optimization of multivalent ligand structures is a rather underexplored topic as most reports have focused on ligand valency and density and on the refinement of spacer length and characteristics [24].
Penta-butynyl-corannulene [27] was selected as the scaffold based on its five-fold symmetry, equivalent to that of CVA24v, and due to the synthetic ease of accessing C5symmetric pentavalent Neu5Ac conjugates via the copper-catalyzed azide-alkyne cycloadditions (CuAAC). The scaffold may interact with amino acid residues of the capsid via CH-π, cation-π, or π-π interactions [28,29]. In addition, the aromatic bowl-shaped backbone of the corannulene scaffold imparts a dipole moment, through which interactions could be induced [28]. Furthermore, bowl-inversions of the scaffold occur readily [30,31], thus, either the concave or convex face of the scaffold can interact with the virus particle, which may be ideal due to the spherical form of CVA24v.
Quantum mechanical molecular modeling was used to probe potential interactions between the corannulene-scaffold and CVA24v. To reduce the computational demands of the quantum mechanical calculations, a penta-propyl substituted corannulene molecule was docked to a truncated region, consisting of 340 atoms, of one pentagon of CVA24v that included the five-fold symmetry axis of the virion ( Figure 1A). The docking started with a low-energy penta-propyl-substituted corannulene (Figure 1B) that was manually placed on top of the five-fold center of CVA24v. Subsequently, the complex was geometry optimized in Jaguar [32] using density functional theory (BLYP-D3 [33], 6-31G **), with the molecule allowed to move freely over the viral surface with fixed atomic positions. The optimization converged after 48 cycles according to the direct inversion of the iterative space method [34]. No atomic clashes were identified, and beneficial arene interactions were present between histidine side chains and the corannulene molecule ( Figure 1C). This docked structure was then used to study the length of the spacers required for chelation of the pentameric lectins ( Figure 1D) [23]. The docked structure of corannulene was connected to the sialic acids (X-ray coordinates), and the spacers were energy minimized. Spacers of increasing length were evaluated to minimize the strain where n = 1 led to a strained spacer and n = 5 was sufficiently long to avoid strain.

Synthesis of Scaffold-Replaced Pentavalent Neu5Ac Conjugates
Synthesis of the symmetric pentavalent Neu5Ac conjugates 12-14 was achieved in five steps from Neu5Ac donor 1 that was prepared from Neu5Ac as previously described (Scheme 1) [35]. The docking started with a low-energy penta-propyl-substituted corannulene ( Figure 1B) that was manually placed on top of the five-fold center of CVA24v. Subsequently, the complex was geometry optimized in Jaguar [32] using density functional theory (BLYP-D3 [33], 6-31G **), with the molecule allowed to move freely over the viral surface with fixed atomic positions. The optimization converged after 48 cycles according to the direct inversion of the iterative space method [34]. No atomic clashes were identified, and beneficial arene interactions were present between histidine side chains and the corannulene molecule ( Figure 1C). This docked structure was then used to study the length of the spacers required for chelation of the pentameric lectins ( Figure 1D) [23]. The docked structure of corannulene was connected to the sialic acids (X-ray coordinates), and the spacers were energy minimized. Spacers of increasing length were evaluated to minimize the strain where n = 1 led to a strained spacer and n = 5 was sufficiently long to avoid strain.

Corannulene-Based Pentavalent Neu5Ac Conjugates Attenuate CVA24v Transduction
To evaluate the biological effect of the scaffold-replaced Neu5Ac conjugates 12-14, the transduction of CVA24v at 37 • C was measured in HCE cells (Figure 2A), as previously described [23]. Ocular inspection of the cells by microscopy and cell counting ( Figure S1) indicated that the compounds are non-toxic. Compared with untreated virions, the transduction of CVA24v was significantly attenuated when preincubated with 12-14. All inhibitors were significantly more potent than Neu5Ac [23] and disialyllacto-N-tetraose [9]. Control compound ME0752 with a βglucose scaffold ( Figure 1B, compound 28 in reference [23]), attenuated transduction with >95% and 50% at 1.25 mM and 312.5 μM, respectively, but had no effect at lower concentrations. Compound 14, with the longest spacers, was the most efficient of the scaffoldreplaced compounds and reduced transduction by 80-85% and 25% at 1.25 mM and 312.5 μM, respectively. Compounds 13 and 12, containing successively shorter spacers, reduced transduction by 75% and 60%, respectively, but had no effect at 312.5 μM. A plausible explanation for this could be that the spacers of compounds 12 and 13 are not of sufficient length to allow simultaneous binding of the Neu5Ac units of the compounds to the pentameric lectins of CVA24v. It is also possible that the compounds cause cross-linking and aggregation as previously observed for glucose-based pentameric sialic acids, including ME0752 [23]. Nevertheless, the displayed inhibitory activity of compound 14 indicates that the glucose scaffold can be replaced with a corannulene scaffold, although for this particular system and set of compounds, it did not result in enhancing potency.

X-ray Crystallography Supports Binding of Inhibitors to CVA24v
The binding mode of 14 was structurally characterized by X-ray crystallography. The X-ray structure of 14-CVA24v was established to a resolution of 1.78 Å. Similar to the CVA24v structures with the glucose-based Neu5Ac inhibitors [23] (pdb code 6TSD), the 14-CVA24v structure showed clear electron density for the Neu5Ac units in all five binding sites, each located within viral protein 1 (VP1). The electron density of the spacer points outward to the solvent and shows that the spacer does not have specific contacts to the viral shell. This results in a high degree of flexibility in that region, resulting in a blurry map that does not allow the placement of more than the first three atoms of the spacers of compound 14. Thus, the remaining parts of 14 were not resolved, suggesting that it is either not ordered or that the inhibitor does not accommodate a single conformation. Further, it is possible that 14 does not bind in a chelating fashion similar to the Compared with untreated virions, the transduction of CVA24v was significantly attenuated when preincubated with 12-14. All inhibitors were significantly more potent than Neu5Ac [23] and disialyllacto-N-tetraose [9]. Control compound ME0752 with a β-glucose scaffold ( Figure 1B, compound 28 in reference [23]), attenuated transduction with >95% and 50% at 1.25 mM and 312.5 µM, respectively, but had no effect at lower concentrations. Compound 14, with the longest spacers, was the most efficient of the scaffold-replaced compounds and reduced transduction by 80-85% and 25% at 1.25 mM and 312.5 µM, respectively. Compounds 13 and 12, containing successively shorter spacers, reduced transduction by 75% and 60%, respectively, but had no effect at 312.5 µM. A plausible explanation for this could be that the spacers of compounds 12 and 13 are not of sufficient length to allow simultaneous binding of the Neu5Ac units of the compounds to the pentameric lectins of CVA24v. It is also possible that the compounds cause cross-linking and aggregation as previously observed for glucose-based pentameric sialic acids, including ME0752 [23]. Nevertheless, the displayed inhibitory activity of compound 14 indicates that the glucose scaffold can be replaced with a corannulene scaffold, although for this particular system and set of compounds, it did not result in enhancing potency.

X-ray Crystallography Supports Binding of Inhibitors to CVA24v
The binding mode of 14 was structurally characterized by X-ray crystallography. The X-ray structure of 14-CVA24v was established to a resolution of 1.78 Å. Similar to the CVA24v structures with the glucose-based Neu5Ac inhibitors [23] (pdb code 6TSD), the 14-CVA24v structure showed clear electron density for the Neu5Ac units in all five binding sites, each located within viral protein 1 (VP1). The electron density of the spacer points outward to the solvent and shows that the spacer does not have specific contacts to the viral shell. This results in a high degree of flexibility in that region, resulting in a blurry map that does not allow the placement of more than the first three atoms of the spacers of compound 14. Thus, the remaining parts of 14 were not resolved, suggesting that it is either not ordered or that the inhibitor does not accommodate a single conformation. Further, it is possible that 14 does not bind in a chelating fashion similar to the glucose-based pentavalent Neu5Ac conjugates that were observed to cross-link and aggregate CVA24v particles [23]. Thus, 14 may bind to Neu5Ac binding sites of separate CVA24v particles within the crystal lattice, resulting in the absence of observed density as the spacers likely adopt random orientations in the solvent. The interactions in the lectin binding sites of CVA24v with the Neu5Ac units of compound 14 were similar as previously described [23] ( Figure 3), with the main exception that the C8 hydroxyl rather than the C9 hydroxyl of Neu5Ac engages in intramolecular hydrogen bonding with the C2 carboxylate. glucose-based pentavalent Neu5Ac conjugates that were observed to cross-link and a gregate CVA24v particles [23]. Thus, 14 may bind to Neu5Ac binding sites of separa CVA24v particles within the crystal lattice, resulting in the absence of observed densi as the spacers likely adopt random orientations in the solvent. The interactions in the le tin binding sites of CVA24v with the Neu5Ac units of compound 14 were similar as pr viously described [23] (Figure 3), with the main exception that the C8 hydroxyl rather tha the C9 hydroxyl of Neu5Ac engages in intramolecular hydrogen bonding with the C2 ca boxylate.

Cryo-EM Supports Inhibitor Binding over the Five-Fold Axis of CVA24v
To understand how compound 14 binds to CVA24v in solution, cryo-EM analys was performed. To exclude the possibility of cross-linkage between the inhibitor and bin ing sites of neighboring capsids, we employed an on-grid binding strategy described pr viously [37,38]. This approach allowed us to determine the 14-CVA24v structure at a overall resolution of 2.3 Å ( Figure 4A). The spacer to the corannulene core is flexible, resulting in a fuzzy electron density. Therefore, neither the corannulene nor most parts of the spacer can be modeled into the crystal structure.

Cryo-EM Supports Inhibitor Binding over the Five-Fold Axis of CVA24v
To understand how compound 14 binds to CVA24v in solution, cryo-EM analysis was performed. To exclude the possibility of cross-linkage between the inhibitor and binding sites of neighboring capsids, we employed an on-grid binding strategy described previously [37,38]. This approach allowed us to determine the 14-CVA24v structure at an overall resolution of 2.3 Å ( Figure 4A). In the unsharpened map, which retains lower resolution features, a clear mushroomshaped density for compound 14 was visible at the capsid vertices ( Figure 4A). However, upon map sharpening, only the sialic acid component of the inhibitor was resolved at high resolution, with a binding mode virtually identical to that observed in the X-ray crystallography structure (Figures 3 and 4B,C). Because the core and spacer regions of compound 14 were poorly resolved, focused classification [39] was performed on a single five-fold symmetry axis of the CVA24v capsid to study alternative conformations of the inhibitor. In this approach, 60 symmetrically redundant orientations were assigned to each particle, and a mask was used to focus classification on a substructure within the virion. In the absence of symmetry averaging, this approach revealed a striking heterogeneity in the binding mode of compound 14 and demonstrated that the inhibitor density deviates from five-fold symmetry ( Figure 4D and Video S1). Moreover, in several classes, compound 14 is only localized to a subset of available sialic acid binding sites. This suggests that only a couple of Neu5Ac appendages of the compound are in contact with the capsid at any one time, and the others adopt random orientations in the solvent. Clearly, this indicates that 14 is not capable of fully chelating the Neu5Ac binding sites of the CVA24v lectins. Thus, it is plausible that a radial presentation of Neu5Ac ligands is not an appropriate design strategy for this particular system, possibly due to the overall topology of the lectin-receptor. The rising of the protein surface from the center of the five-fold symmetry axis of the virion in relation to the Neu5Ac binding sites gives the lectin-receptor an overall convex-concave-convex surface going from one Neu5Ac binding site across the center to another. Computational work studying the stability of complexes between divalent ligands and bivalent receptors with concave (i.e., depression or channel) or convex (i.e., rising or hill) surfaces between the binding sites, have shown that spacer-receptor interactions are particularly detrimental in convex receptors and result in vanishing avidity enhancements [40]. It is possible that similar factors play a role in this particular system as well. In the unsharpened map, which retains lower resolution features, a clear mushroomshaped density for compound 14 was visible at the capsid vertices ( Figure 4A). However, upon map sharpening, only the sialic acid component of the inhibitor was resolved at high resolution, with a binding mode virtually identical to that observed in the X-ray crystallography structure (Figures 3 and 4B,C). Because the core and spacer regions of compound 14 were poorly resolved, focused classification [39] was performed on a single five-fold symmetry axis of the CVA24v capsid to study alternative conformations of the inhibitor. In this approach, 60 symmetrically redundant orientations were assigned to each particle, and a mask was used to focus classification on a substructure within the virion. In the absence of symmetry averaging, this approach revealed a striking heterogeneity in the binding mode of compound 14 and demonstrated that the inhibitor density deviates from five-fold symmetry ( Figure 4D and Video S1). Moreover, in several classes, compound 14 is only localized to a subset of available sialic acid binding sites. This suggests that only a couple of Neu5Ac appendages of the compound are in contact with the capsid at any one time, and the others adopt random orientations in the solvent. Clearly, this indicates that 14 is not capable of fully chelating the Neu5Ac binding sites of the CVA24v lectins. Thus, it is plausible that a radial presentation of Neu5Ac ligands is not an appropriate design strategy for this particular system, possibly due to the overall topology of the lectin-receptor. The rising of the protein surface from the center of the five-fold symmetry axis of the virion in relation to the Neu5Ac binding sites gives the lectin-receptor an overall convex-concave-convex surface going from one Neu5Ac binding site across the center to another. Computational work studying the stability of complexes between divalent ligands and bivalent receptors with concave (i.e., depression or channel) or convex (i.e., rising or hill) surfaces between the binding sites, have shown that spacer-receptor interactions are particularly detrimental in convex receptors and result in vanishing avidity enhancements [40]. It is possible that similar factors play a role in this particular system as well.

Conclusions
In this study, we performed structure-based scaffold hopping on pentavalent Neu5Ac conjugates in an attempt to achieve a chelating binding to CVA24v, and evaluated the effect of the scaffold replaced compounds on CVA24v transduction. The rationale for the replacement of a glucose scaffold with corannulene was confirmed by quantum mechanical docking of the corannulene scaffold to CVA24v that indicated the presence of beneficial arene interactions to the protein surface of CVA24v. The effect of the scaffold-replaced pentavalent Neu5Ac compounds generated no beneficial effect on the inhibition of CVA24v transduction in relation to the glucose-based inhibitor ME0752. We used cryo-EM to study the binding of the corannulene-based compound 14 to CVA24v, which indicated that binding does not occur via the intended chelation of the binding sites. Thus, we find it reasonable to speculate that this is a major contributing factor to the small observed avidity gains achieved by both the glucose-and corannulene-based pentavalent Neu5Ac conjugates. Likely, a mixture of multivalent binding mechanisms are operating at any given time resulting in the inhibition of CVA24v transduction. This is valuable information for the future development of Neu5Ac-based inhibitors of CVA24v. Future studies could explore the effect of multivalent Neu5Ac conjugates with different topologies.

Computational Modelling
The molecular structure of the penta-propyl-corannulene was energy minimized using the semi-empirical method Austin model 1 (AM1) [41,42] followed by geometry optimization with density functional theory (DFT) in Jaguar (Schrödinger, New York, NY, USA) [32,43] using functional B3LYP [33,43,44] and basis set CC-PVTZ. This resulted in a low-energy structure with the expected bowl-shaped corannulene moiety ( Figure 1A). The docking of penta-propyl-corannulene to a truncated region of CVA24v, consisting of 340 atoms of 1 pentagon, started with the manual placement followed by a optimization of the interaction using DFT (BLYP-D3 [33,44,45], 6-31G **) in Jaguar [32,43]. During the geometry optimization, the penta-propyl-corannulene was allowed to move but the virus binding site, including protein atom in close proximity ( Figure 1A), had fixed atomic positions. The optimization converged after 48 cycles according to the direct inversion of the iterative subspace method [34]. No atomic clashes were identified, and there was an indication for a beneficial arene interaction between corannulene and the proximal histidine side chains in the binding site ( Figure 1B). Using this docked structure, spacers of different lengths to the sialic acid moiety were hypothesized and tested.

CVA24v Production and CVA24v Transduction
CVA24v of high purity ( Figure S2) was produced as described previously [8,9] and the CVA24v transduction assay was carried out according to established protocols [23].

Crystallization and Structure Determination
CVA24v was crystallized and derivatized as previously reported [9]. Briefly, compound 14 (17 mM) was soaked overnight at 4 • C before the crystals were harvested and flash-frozen in liquid nitrogen. Data were collected at beamline I03 at Diamond Light Source, Oxford, United Kingdom (UK). Data reduction was performed using XDS [46], followed by molecular replacement using rigid body refinement as implemented in REF-MAC5 [47]. Simulated annealing using Phenix [48] was applied to remove model bias. The final model resulted from cyclic real-space refinement using COOT [49] and reciprocal space refinement as implemented in REFMAC5. Strict NCS parameterization was applied in all later refinement steps (Table 1). The structure was validated using MOLPROBITY [50] before structure factor amplitudes and coordinates were deposited to the protein data bank (pdb entry 7OJ7). Figures were generated using PYMOL [51].

Cryo-Electron Microscopy
Cryo-EM grids were prepared by applying 3 µL of purified CV-A24v (9.6 mg/mL) to 400 mesh lacey grids coated in a 3 nm carbon film (Agar Scientific, Essex, UK). The sample was left to adsorb for 30 s, and then the excess sample was removed by manual blotting. On-grid binding of pentavalent Neu5Ac conjugate was then performed by applying 3 µL of a 5 mM solution of compound 14 in PBS to the pre-blotted, CV-A24v-coated grid, and leaving for 30 s before blotting and freezing using a Leica EM GP plunge freeze device (Leica Microsystems, Wetzlar, Germany). The Leica EM chamber temperature was set to 8 • C with 80% relative humidity and liquid ethane was used for sample vitrification. Grids were glow discharged at 20 mA for 30 s prior to application of the sample, using a PELCO easiGlow™ (Ted Pella, Redding, CA, USA). All data sets were collected on a Titan Krios (Thermo Fisher Scientific, Waltham, MA, USA) transmission electron microscope operating at 300 kV, at a nominal magnification of 165,000×. A total of 1443 exposures were recorded using EPU automated acquisition software on a Gatan K2 summit camera operating in super-resolution mode, using a total electron dose of 48.9 e−/Å 2 . Each movie had a total exposure of 10 s and contained 50 fractions. The final calibrated physical pixel size was 0.82 Å. Detailed information on data collection was shown in Table 2.  Following cryo-EM data collection, the RELION-3.0 pipeline [52] as used for image processing. Drift correction was first performed on micrograph stacks using MOTION-COR2 [53], and the contrast transfer function for each was estimated using Gctf [54]. A subset of virus particles was picked manually and subjected to 2D classification, with the resultant classes used as templates for automatic particle picking [55]. Particles were classified through multiple rounds of reference-free 2D classification, and particles in poor quality classes were removed after each round. Using the 'molmap' command in UCSF chimera [56], the CV-A24v crystal structure (PDB ID: 4Q4V [9]) was used to generate a 40 Å resolution starting model to use as a reference for 3D autorefinement, with icosahedral symmetry imposed. This reconstruction was postprocessed to mask and correct for the B-factor of the map before taking particles forward to contrast transfer function (CTF) refinement and Bayesian polishing [57]. The final reconstruction at 2.3 Å was computed after correcting for the curvature of the Ewald sphere [52]. The nominal resolution the map was determined according to the 'gold standard' Fourier shell correlation (FSC) criterion (FSC = 0.143) [55] and a negative B-factor of −49 Å 2 was applied during the final post-processing step. The local resolution estimation tool in RELION was used to generate a map filtered by local resolution.
To investigate the conformational heterogeneity of compound 14, a focused 3D classification approach was employed (as described previously [39]). Briefly, each particle contributing to the final icosahedral-symmetry-imposed reconstruction was assigned 60 orientations corresponding to its icosahedrally related views using the relion_symmetry_expand tool. SPIDER [58] was used to generate a spherical mask, which was manually placed over the map to isolate the fivefold symmetry axis of the capsid, and the symmetryexpanded particles were subjected to masked 3D classification without alignment using a regularization parameter ('T' number) of 20. During the no-alignments 3D classification, the resolution in the E-step was limited to 12 Å. Figures were generated using UCSF Chimera [56] and UCSF ChimeraX [59].

General Procedure for Synthesis of Deprotected Azido-Sialosides
An oven-dried round bottom flask was charged with magnetic stirring bar, powdered activated molecular sieves (4Å, 3.3 g), xanthate sialoside donor (1.0913 mmol, 1.0 eq) and placed under N 2 atmosphere. To this mixture was added CH 2 Cl 2 (50 mL), bromoalcohol (2, 3, or 4, respectively, 1.20 eq), and under dark conditions, a solution of silver trifluormethansulfonate (2.0 eq) in freshly distilled CH 3 CN (77 mL). The solution was cooled to −72 • C and stirred for 15 min, followed by dropwise addition of a solution of IBr (1.0 M in CH 2 Cl 2 , 1.40 eq). After complete addition of the IBr solution, the reaction was stirred for 2 h at −72 • C, and N,N-diisopropylethylamine (DIPEA) (6.0 eq) was added and stirred for another 20 min at −72 • C, and another 20 min before warming to rt. The mixture was filtered through a plug of celite and concentrated to dryness. The resulting mixture was purified by flash chromatography (CH 2 Cl 2 /CH 3 OH gradients), affording the protected sialosides as a mixture of alpha and beta anomers, in addition to the elimination product. The mixture (544 mg, 0.8312 mmol) was subsequently dissolved in DMF (25 mL) and treated with NaN 3 (6.0 eq) followed by tetra-n-butylammonium iodide (2.0 eq). The reaction was allowed to stir under nitrogen atmosphere for 22 h. The mixture was diluted in CH 2 Cl 2 , washed with water, HCl (1 M), and brine. The organic layer was dried of anhydrous Na 2 SO 4 , filtered and concentrated under reduced pressure affording crude product, which was used without additional purification. The crude was dissolved in a 0.03 M solution of sodium methoxide (a total of 10.0 equiv.). The reaction was allowed to stir at rt under nitrogen atmosphere for 24 h. The solution was neutralized by the addition of Dowex 50 × 8 H-form (pre-washed with MeOH), filtered, and concentrated to dryness. Mixture was purified using column chromatography (CH 2 Cl 2 /MeOH, from 6% MeOH to 10% MeOH), affording the deprotected azido sialosides 5-7 as pure α-anomers. See below for specific yields and analytical data.

General Procedure for CuAAC
An oven-dried round bottom flask equipped with magnetic stirring bar was charged with azido-sialoside (0.37 mmol, 11.5 eq). To this was added a solution of pentapropargylated corannulene core 8 (1.0 eq, 0.032 mmol) in tetrahydrofuran (THF, 7 mL). To the stirring solution was added CuSO 4 . 5H 2 O (1.59 eq) and sodium ascorbate (1.55 eq) in H 2 O (7 mL). The flask was equipped with rubber septa and the mixture was heated to 50 • C for 5 h and then the reaction was left to perform at room temp for 36 h. The THF was removed under reduced pressure, and the resulting mixture was injected on HPLC (MeCN/H 2 O 10% → 25% gradient in 25 min), affording the pentavalent methyl ester derivative after freeze-drying. See chemical synthesis for specific yields and analytical data.

General Procedure for Ester Hydrolysis
The pentavalent methyl ester derivate (0.01 mmol, 1.0 eq) was dissolved in CH 3 OH (1.35 mL), and to this stirring solution was added a 1M solution of LiOH (0.156 mL, 15.0 eq). The mixture was stirred for 48 h at room temperature in the dark. The mixture was neutralized (pH 7-8) with Dowex 50x8 H-form, filtered, and concentrated under reduced pressure. The resulting residue was diluted in water and freeze-dried to afford the pentavalent target compound.