Exploring divalent conjugates of 5-N-acetyl-neuraminic acid as inhibitors of coxsackievirus A24 variant (CVA24v) transduction

Coxsackievirus A24 variant (CVA24v) is responsible for several outbreaks and two pandemics of the highly contagious eye infection acute hemorrhagic conjunctivitis (AHC). Currently, neither prevention (vaccines) nor treatments (antivirals) are available for combating this disease. CVA24v attaches to cells by binding Neu5Ac-containing glycans on the surface of cells which facilitates entry. Previously, we have demonstrated that pentavalent Neu5Ac conjugates attenuate CVA24v infection of human corneal epithelial (HCE) cells. In this study, we report on the structure-based design of three classes of divalent Neu5Ac conjugates, with varying spacer lengths, and their effect on CVA24v transduction in HCE cells. In relative terms, the most efficient class of divalent Neu5Ac conjugates are more efficient than the pentavalent Neu5Ac conjugates previously reported.


Introduction
Acute haemorrhagic conjunctivitis (AHC) is a highly contagious eye infection. 1 It is predominantly caused by two members of the Picornaviridae family, coxsackievirus A24 variant (CVA24v), an antigenic variant of the CVA24 strain, and enterovirus 70 (EV70). 2 The disease may also be caused by human adenoviruses. However, since its emergence in 1970, CVA24v has been reported as the principal etiological agent. 3,4 AHC is characterized by a sudden onset of ocular pain, foreign body sensation, excessive lacrimation, periorbital swelling and subconjunctival hemorrhages. 5 The disease spreads rapidly within communities, affecting up to 50% of the population. Such outbreaks can exhaust local healthcare resources of affected regions and severely disrupt the economy. In otherwise healthy individuals, AHC is generally self-limiting, resolving in 1-2 weeks. However, the infections are also associated with visual impairments, symptoms in the respiratory tract, and in rare cases cause acute accid paralysis and fatalities. 1 Over the last decades CVA24v has caused two pandemics, numerous recurring outbreaks, and caused >10 million cases of AHC worldwide. 1 Neither antiviral agents nor vaccines are currently available for treating or preventing the disease as is the case for most viruses causing diseases in humans. 6 Members of the Picornaviridae engage a range of different cellular receptors facilitating attachment and entry. These include coxsackievirus and adenovirus receptor (CAR), decay accelerating factor (DAF, CD55), low density lipoprotein receptor (LDL-R), human P-selectin glycoprotein ligand-1 (PSGL-1), heparan sulfate, integrins, intercellular adhesion molecule-1 (ICAM-1), and glycan-containing receptors terminating in 5-N-acetyl-neuraminic acid (Neu5Ac). [7][8][9][10][11] CVA24v is reported to engage receptors with terminal a2,3and a2,6linked Neu5Ac, with some preference for the a2,6-linkage. 9,10,12 In addition, binding to the ICAM-1 receptor is essential for productive replication of CVA24 and the AHC-causing CVA24v. 13 However, the enhanced ability of CVA24v to bind Neu5Ac has been suggested to contribute to its notable virulence and pandemic potential. Thus, Neu5Ac-based derivatives have potential to prevent attachment of virions to cells, limiting the infection and subsequent spread.
Protein-carbohydrate interactions play essential roles in a vast array of biological processes. However, only a few carbohydrate and carbohydrate-based drugs have reached the market. This is perhaps due to the challenges associated with overcoming the poor pharmacological properties of carbohydrates imparted by their high polarity and metabolic vulnerability. 14 In addition, protein-carbohydrate interactions are generally of low affinity. 15 These downsides are not of direct consequence for the development of carbohydrate-based therapeutics in the case of CVA24v as its primary sites of replication in humans are linked to the eyes and airways. 10 Thus, by design

Results and discussion
Design CVA24v attaches to cells via shallow, surface exposed, Neu5Ac pockets. 12 The CVA24v capsid is decorated by a total of sixty individual Neu5Ac binding sites which are further arranged into twelve local ve-fold symmetries (i.e. twelve regular pentagons). Thus, at its most basic form the template for the design of multivalent Neu5Ac-inhibitors can be based on a single pentagon with its ve Neu5Ac binding sites (Fig. 1A). We have previously reported on the design and synthesis of pentavalent Neu5Ac conjugates of radial 18 and pseudoradial 17 topology (Fig. 1B), that inhibit CVA24v infection of HCE cells. X-ray crystallography conrmed inhibitor binding, with observed electron density for the Neu5Ac units while spacers and the respective scaffold were not detected. The major mode of inhibition of these inhibitors likely resulted from a high local concentration of ligands, and inhibitor caused aggregation of viral particles as indicated by negative staining electron microscopy (EM) 17 and cryogenic electron microscopy (cryo-EM), 18 rather than intended fully chelated binding sites. Building on these observations we hypothesized that multivalent Neu5Ac inhibitors based on tailored linear topology, could engage in additional contacts with the virion through spacer fragments thus generating more potent inhibitors.
As a starting point to probe this strategy, divalent Neu5Ac compounds posed as attractive tools as they have potential to bind to the virion by simultaneously occupying two Neu5Ac binding sites. The spacers connecting both Neu5Ac entities can engage in additional binding events with the protein backbone of the virion (Fig. 1C). Despite the inherent synthetic challenges, a linear strategy has potential to be expanded into linear trivalent, tetravalent, or eventually macrocyclic pentavalent structures (Fig. 1D).
Upon examination of the Neu5Ac binding sites of CVA24v of Neu5Ac bound (PDB code 4Q4Z) and inhibitor bound (PDB code 6TSD) structures, it was evident that the closest points of contacts between two neighbouring Neu5Ac residues were from the C9 position of one Neu5Ac residue to the C2 or C4 position of a second Neu5Ac residue (Fig. 2). The hydroxyls attached to C2 and C9 serve as promising connection points as they are pointing towards the solvent, thereby minimizing possible steric clashes with the virion, and are not key contributors to Neu5Ac binding.
Both these hydroxyl groups are also directed towards a shallow canyon that interconnects the ve Neu5Ac binding sites of each pentagon, thus opening up the possibility of the spacer binding to amino acid residues of the canyon. Further, both C2 and C9 Neu5Ac alkynyl or azido building blocks are readily accessible and can be conjugated under mild conditions (e.g. copper azide alkyne cycloaddition) to provide access to C2-C9 linked divalent Neu5Ac compounds. Contrarily, the C4 hydroxyl is engaged in two water mediated hydrogen bonds to the protein backbone of the virion and projects towards the protein.
However, the canyon that interconnects the Neu5Ac binding sites is directly accessible from the C4 hydroxyl via a narrow passage making it highly interesting as a point of connection as interactions to, or steric clashes with, the protein backbone of the virion are forced. Further, 4-O-alkynyl Neu5Ac building blocks are accessible via etherication of appropriately protected building blocks 19 and aer deprotection conjugated to a C9 modied azido Neu5Ac compound to yield C4-C9 linked divalent Neu5Ac compounds. Another, more straightforward method, of assembling divalent Neu5Ac compounds is preparation of a common Neu5Ac building block, e.g. alpha-azido Neu5Ac, 20 and subsequently conjugate this to a di-alkynyl spacer fragment to yield C2-C2 triazole linked divalent Neu5Ac compounds. The spacer fragments of such divalent C2-C2 linked Neu5Ac compounds are however less likely to bind to the canyon as alpha anomeric Neu5Ac substituents are pointing towards the solvent. Nevertheless, we decided to probe all three strategies.

C2-C9 linked divalent Neu5Ac compounds reduce CVA24v transduction
The antiviral effect of the three synthesized classes of divalent Neu5Ac compounds were evaluated by measuring CVA24v transduction using HCE cells. 17 Ocular inspection of cells treated with divalent compounds did not reveal any cytotoxicity at any of the tested concentrations. Compared with untreated virions (control), the transduction was signicantly attenuated when preincubated with compounds 16-21 at all of the measured compounds concentrations (Fig. 3A). This is in contrast to Neu5Ac, that failed to reduce CVA24v transduction at 10 mM (not shown) 12,17 and the natural occurring divalent Neu5Ac glycan disialyllacto-N-tetraose (DSLNT) that reduced CVA24v transduction by 65% at 5 mM (  published data. 17 At 312.5 and 19.5 mM compounds 17-21 were equally effective and reduced transduction by 60-70% and 30-35%, respectively, while ME0752 reduced transduction by 85% and 50%, respectively. At 78.2 mM compounds 18-21 were $1.5 times more efficient than 16 and 17 and reduced transduction by 60%, while ME0752 reduced transduction by 75%. Taken together, a relationship between spacer length and potency could not be demonstrated due to small differences in efficacy between the compounds. This is in agreement with previous reports that have demonstrated that the length of exible spacers in divalent ligands have minimal inuence on avidity. 25,26 Nevertheless, compounds 18 and 21 seemed to be among the top performers at each of the tested concentrations. Further, in terms of absolute values ME0752 was the most potent inhibitor. However, in terms of relative inhibitor potency, the ratio of reduced transduction produced by an inhibitor divided by the number of Neu5Ac units present in the inhibitor or the number of heavy atoms in the molecule, 16-21 were more efficient at reducing CVA24v transduction. Contrarily, the C2-C2 linked compounds 40 and 41 were only effective at reducing CVA24v transduction by 60% at 5 mM (Fig. 3B). The C4-C9 linked compounds (compounds 32-34) did not attenuate CVA24v transduction at any of the tested concentrations (Fig. 3B), indicating the C4 hydroxyl as a key contributor in the binding and recognition of Neu5Ac by CVA24v or alternatively that steric factors imposed by the spacer prohibit binding. Thus, only the C2-C9 linked Neu5Ac compounds were considered for further evaluation.
X-ray crystallography support binding of the C2-C9 divalent Neu5Ac compounds to CVA24v To investigate the binding of compounds 16-21, we derivatized CVA24v by soaking CVA24v crystals in crystallization solution containing the corresponding compound (16 mM) as described previously. 12 The following structure determination resulted in data sets with resolutions below or around 2Å and showed that all compounds, beside 17, bind by their Neu5Ac entity to the characterized binding site. 12 The linker regions of all compounds seem not to interact with the capsid protein and are therefore not visible in the resolved X-ray crystal structures (Fig. 4). A plausible reason is the C2-C9 linkage of these dimeric compounds. The C2 and C9 hydroxyls of the parent Neu5Ac molecule are pointing towards the solvent. Thus, we nd it likely that the spacer unit that links the C2 position of one Neu5Ac entity to the C9 position of a second Neu5Ac entity is oating in the solvent in a non-ordered fashion rather than folding back towards the viral surface to interact with the canyon on top of the VP1 capsid protein.
C2-C9 divalent Neu5Ac compounds do not affect particle stability of CVA24v Capsid binding antivirals have been suggested to act by stabilization of the virus particle, 27 preventing or delaying receptormediated conformational changes required for uncoating. 28 In analogy, capsid-binding molecules could also mimic the receptor-mediated interaction thus triggering conformational changes resulting in early uncoating and inactivation of the virus. To study the effect of the inhibitors on the stability of CVA24v, we performed the Particle Stability Thermal Release Assay (PaSTRy). 29 The C2-C9 divalent inhibitors 18 and 21 were selected for testing together with the pentavalent Neu5Ac conjugate ME0752 (28 in ref. 17). The melting temperature of the CVA24v capsid, as measured by release of viral RNA (T m RNA), was calculated to be 50.96 C, in good agreement with previous observations 17 and in the same range of other known Enteroviruses. 28 In line with previous observations, ME0752 had a mild stabilizing effect and shied the release of viral RNA to 52.96 C (Fig. 5A and D), which is 1.0 C more than previously reported. 17 Treatment with 18 and 21 shied the release of viral RNA to 50.46 C and 51.46 C (Fig. 5B-D), respectively. Thus, the divalent Neu5Ac compounds does not seem to affect the thermal stability of the CVA24v particle indicating the mechanism of inhibition is not related to CVA24v uncoating.

Conclusion
In the present study we employed a structure-based approach to design three classes of novel divalent Neu5Ac conjugates with the aim to achieve chelation binding of the Neu5Ac binding sites of CVA24v. We presented efficient synthesis of each class of compounds, using "click" chemistry in the key chemical transformation connecting the two Neu5Ac entities with exible spacers. Compounds with a C2-C9 linkage between the two Neu5Ac entities proved more efficacious blocking CVA24v transduction than C4-C9 and C2-C2 linked compounds. In terms of relative efficiency, the C2-C9 linked compounds were better than the previously reported pentavalent Neu5Ac conjugate ME0752. X-ray crystallography conrmed the binding of the C2-C9 linked divalent Neu5Ac compounds, however electron densities were only detected for the Neu5Ac entities. This suggests that the spacers of the compounds do not interact with the protein backbone of CVA24v or that the spacers are of sufficient exibility to generate largely non-ordered binding events. The effect of designing and synthesizing precise and rigid spacers were previously demonstrated for divalent galactoside ligands targeting LecA. 30 A similar strategy could be applied to optimize binding of these divalent Neu5Ac ligands to CVA24v. However, based on the results presented here and previous results with pentavalent Neu5Ac attachment inhibitors 17,18 we conclude that further optimization of divalent Neu5Ac compounds will be challenging.

Experimental
Particle stability thermal release assay (PaSTRy) CVA24v (1 mg) and compounds (1 mM and 100 mM) were incubated for 30 min at room temperature in a total volume of 20 ml sample buffer (10 mM HEPES pH 8.0, 200 mM NaCl) before adding dyes. PaSTRy assay was performed as previously described. 29 SYPRO red (stock 5000Â, Invitrogen) and SYT09 (stock 50 mM, Thermo Fisher Scientic) were diluted 100Â in Milli-Q water freshly before each experiment. The dyes were added to the CVA24v + compound samples to total volume of 50 ml sample buffer and the nal concentrations of the dyes were 3Â of SYPRO red and 5 mM of SYT09. Samples and dyes were added in a microamp optical 96-well reaction plate (Applied Biosystems, California, USA) and ran a real-time PCR system (StepOnePlus, Applied Biosystems). The melting curve was set to increase 1 C every 15 s (log uorescence every 1 C increased), ranging from 25 C to 99 C.

CVA24v transduction assay
One day prior infection, HCE cells (2 Â 10 4 per well) were seeded in a black 96-well plate with transparent bottom. Next day, Neu5Ac conjugates were added to give four-fold dilution series with 5 mM as the highest concentration and plates were incubated at 37 C with 10 CVA24v per cell (approximately 4 Â 10 4 cells per well) for 1 hour. HCE cells in the black 96-well plate were washed twice with BB2 then incubated with 50 ml of CVA24v (10 CVA24v per cell) + compound mixture per well. Aer 1 hour incubation at 37 C, cells were washed to remove nonbound virions and incubated in HCE growth medium for 16-18 hours. Aer xation with 99.5% ice-cold methanol, mouse monoclonal antibodies against enterovirus VP1 (DakoCytomation, Glostrup, Denmark) were diluted 1 : 200 in PBS and 50 ml was added per well. Aer incubation for 1 hour at room temperature, the cells were washed again and incubated with 50 ml Alexa uor 488-labeled donkey anti-mouse immunoglobulin G (Thermo sher scientic) (diluted 1 : 400 in PBS) per well at room temperature. One hour later, the cells were washed again, and the numbers of infected cells were quantied using a Trophos system.

CVA24v crystallography
CVA24v was produced and described previously. 12 To incorporate 16-21, the compounds where diluted in the crystallization solution and pipetted to crystals. The nal concentration of each compound was 16 mM. The crystals were incubated for 1 hour at 4 C before harvesting and ash-freezing in liquid nitrogen. Data was recorded at beamline I03 at Diamond Light Source Ltd (Didcot, UK). All data sets were reduced by XDS 31 and scaled to the native data set that was recorded previously. Phasing was performed by applying the phases of the native structure (PDB code 4Q4W) followed by a simulated annealing approach as implemented in PHENIX. 32 Renement of the model was done in a cyclic procedure by reciprocal space renement as implemented in REFMAC5 33 including strict NCS renement and real space renement using COOT 34 The nal models were validated using MOLPROBITY. 35 Figures were generated with PYMOL. 36 Data statistics and renement statistics are summarized in Table 1. As all C2-C9 linked compounds essentially show the same complex formation, we picked compound 18 as a reference and deposited the coordinates together with the structure factor amplitudes to the protein data bank (PDB code 7QB5).

General chemical procedures
1 H nuclear magnetic resonance (NMR) and 13 C NMR spectra were recorded with a Bruker DRX-400 spectrometer at 400 MHz and 100 MHz respectively, or with a Bruker DRX-600 spectrometer at 600 MHz and 150 MHz respectively. NMR experiments were conducted at 298 K in D 2 O (residual solvent peak ¼ 4.79 ppm (d H )), CD 3 OD (residual solvent peak ¼ 3.31 ppm (d H ) and 49.00 ppm (d C )), or CDCl 3 (residual solvent peak ¼ 7.26 ppm (d H ) and 77.16 ppm (d C )). Liquid chromatography mass spectrometry (LC-MS) was performed by detecting positive/negative ion (electrospray ionization, ESI) on Agilent 1290 innity II-6130 Quadrupole using H 2 O/CH 3 CN (0.1% formic acid) as the eluent system or on Agilent 1290 innity-6150 Quadrupole using YMC Triart C18 (1.9 mm, 20 Â 50 mm column) and H 2 O/CH 3 CN (0.1% formic acid) as the eluent system. High resolution mass spectrometry (HRMS) was performed using a Bruker MicroTOF II time of ight (TOF) mass spectrometer with ESI + ; Tune Mix ESI solution was used for the calibration. Semi-preparative high performance liquid chromatography (HPLC) was performed on a Gilson system HPLC, using a YMC-Actus Triart C18, 12 nm, S-5 mm, 250 Â 20.0 mm with a ow rate 20 mL min À1 , detection at 214 nm and eluent system A: aqueous 0.005% formic acid, and B: CH 3 CN 0.005% formic acid. Column chromatography was performed on silica gel (Merck, 60Å, 70-230 mesh ASTM). Thin layer chromatography (TLC) were performed on Silica gel 60 F 254 (Merck) with detection under ultraviolet (UV) light or development with 5% H 2 SO 4 in ethanol (EtOH) and heat. Automated ash column chromatography was performed using a Biotage® Isolera One system and purchased pre-packed silica gel cartridges (Bio-tage® SNAP Cartridge, KP-Sil). Freeze drying was performed by freezing the diluted CH 3 CN/water solutions in dry ice-acetone bath and then employing a Scanvac CoolSafe freeze dryer connected to an Edwards 28 rotary vane oil pump. Organic solvents were dried using a SG Water Glass Contour Solvent Systems except CH 3 CN (freshly distilled from CaH 2 ) and methanol (MeOH) that were dried over molecular sieves 3Å. All commercial reagents were used as received. All target compounds were $95% pure according to HPLC UV-traces. Statistics were calculated using GraphPad Prism 7 (GraphPad Soware, Inc, La Jolla, CA). Microwave irradiation reactions were performed using a Biotage® Initiator microwave synthesizer; temperatures were monitored by an internal infrared (IR) probe; stirring was mediated magnetically and the reaction were carried out in sealed vessels. Automated ash column chromatography was performed using a Biotage® Isolera One system and purchased pre-packed silica gel cartridges (Bio-tage® SNAP Cartridge, KP-Sil).
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 ask was equipped with rubber septum and the mixture heated to 50 C for 5 h and then the reaction was le to perform at room temp for 36 h. The THF was removed under reduced pressure and the resulting mixture was puried by HPLC (MeCN/H 2 O 10% / 25% gradient in 25 minutes) affording the pentavalent methyl ester derivative aer freeze-drying. See the ESI † for specic yields and analytical data.