Synthesis of Sulfo-Sialic Acid Analogues: Potent Neuraminidase Inhibitors in Regards to Anomeric Functionality

The design, synthesis and application of N-acetylneuraminic acid-derived compounds bearing anomeric sulfo functional groups are described. These novel compounds, which we refer to as sulfo-sialic acid analogues, include 2-decarboxy-2-deoxy-2-sulfo-N-acetylneuraminic acid and its 4-deoxy-3,4-dehydrogenated pseudoglycal. While 2-decarboxy-2-deoxy-2-sulfo-N-acetylneuraminic acid contains no further modifications of the 2-deoxy-pyranose ring, it is still a more potent inhibitor of avian-origin H5N1 neuraminidase (NA) and drug-resistant His275Tyr NA as compared to the oxocarbenium ion transition state analogue 2,3-dehydro-2-deoxy-N-acetylneuraminic acid. The sulfo-sialic acid analogues described in this report are also more potent inhibitors of influenza NA (up to 40-fold) and bacterial NA (up to 8.5-fold) relative to the corresponding anomeric phosphonic acids. These results confirm that this novel anomeric sulfo modification offers great potential to improve the potency of next-generation NA inhibitors including covalent inhibitors.

Based upon the higher acidity and electronegativity of sulfo relative to carboxy and phosphono functional groups, we predicted that sialic acid anomeric sulfonic acid analogues, which we refer to as sulfo-sialic acid analogues, should form the strongest electrostatic interactions with the conserved NA active site triarginyl cluster (Fig. 1). This strategy has been considered in an in silico analysis, which also indicated that the sulfo group should produce the strongest binding inhibitors 11 . Yet, the actual synthesis of sulfo-sialic acid analogues is quite challenging and has remained unreported. In fact, with exception to 2-amino and 2-acetamido anomeric sulfonates, no 2-deoxy sugar with an anomeric sulfo group has been reported to our knowledge 12,13 .
We first became interested in the synthesis of sulfo-sialic acid analogues during our research on covalent NA inhibitors, which have utilized an electronegative 3-fluoro group to destabilize formation of oxocarbenium ion transition states (Fig. 1A) 3 . In addition to enhancing electrostatic interactions with the triarginyl cluster, a strong electron withdrawing sulfo group might also destabilize oxocarbenium ion formation. In this report we reveal the first synthesis of sulfo-sialic acid analogues and their application as potent NA inhibitors.

Results and Discussion
Sulfo-sialic acid analogues were synthesized via oxidation of a mixture of acetylthio intermediates according to Fig. 2. Neu5Ac (1)-derived peracetylated octoses 2a and 2b, first reported by Potter and von Itzstein, were selected as the substrate for sulfur addition 14 . The decarboxylated Neu5Ac derivatives 2a and 2b were synthesized based on the method of Shie et al. with modifications 5 .
Nucleophilic substitution of the 2 anomeric position with HSAc was mediated by either BF 3 ·Et 2 O or TMSOTf, using CH 3 CN, CHCl 3 , or 1,2-dichloroethane (1,2-DCE) as solvent (Table 1). Although our original intention was to produce only 3, during NMR analysis of the acetylthio reaction products, a surprisingly large peak was always observed at 5.80 ppm, with coupling to another singlet at 6.27 ppm. These NMR peaks were later attributed to H-1, H-2 and H-3 of the 2,3-unsaturated pseudoglycal 4b, similar to compounds resulting from the Ferrier rearrangement of glycals in which the axial orientation of the anomeric acetylthio group is also preferred 15,16 . The axial orientation of the acetylthio group of 4b is indicated by a lack of NOE correlation between H-1 and H-5. The structure of 4b was further supported by high-resolution mass spectrometry (HRMS), and characterization of the oxidation products 6b and 8b. The dominance of the axial acetylthio orientation of this pseudoglycal could be explained by the anomeric effect. Less polar reaction conditions were used to produce a higher ratio of 4b to 3 (Table 1). More polar reaction conditions could be used to weaken the anomeric effect and increase the proportion of equatorial acetylthio compound 3a (Table 1).
Oxidation of the acetylthio functionality of 3 and 4 to the sulfo functionality was achieved using excess Oxone (potassium peroxymonosulfate, KHSO 5 · 0.5KHSO 4 · 0.5K 2 SO 4 ) in AcOH buffered with KOAc, based upon the method of Reddie [17][18][19] . Workup was somewhat difficult due to the presence of excess salt, so the procedure was modified with NH 4 OAc in place of KOAc for the oxidation of 3 to 5, which appeared to be quantitative. Yields for the oxidation of 4 to 6 were low due to concurrent oxidation of the pseudoglycal ring double bond by the excess Oxone. During HRMS analysis of polar Oxone oxidation products, negative ions m/z 452.0826 (6 + O), m/z 454.0985 (6 + O + 2H) and m/z 470.0949 (6 + 2O + 2H) were detected, indicating that 6 was further oxidized. Sulfonates 5a and 5b could also be converted to the corresponding methyl esters by reaction with trimethylsilyldiazomethane, a toxic reagent that should be handled with caution. Active sulfo-sialic acid analogues, 7a, 7b and 8b were obtained after deprotection of the precursors in the presence of NaOH.
The structures of sulfo-sialic acid analogues 5-8 were determined using HRMS, 1 H, GCOSY, NOESY and 13 C data. H-1, H-2 and H-5 are all shifted downfield in the axial anomeric sulfonates 5b and 7b relative to the equatorial isomers 5a and 7a, which is also observed in 2, 3, and the phosphonic acid analogues of 5 and 7 4, 5 . The H-1 to H-2 coupling constants of 5a, 5b, 7a and 7b match those of the corresponding phosphonic acids and SCientifiC REPORtS | 7: 8239 | DOI:10.1038/s41598-017-07836-y carboxylic acids as well 4,5,20 . A clear H-1 to H-5 NOE correlation is observed in the equatorial sulfo compound 7a. In contrast, the lack of any NOE correlation between H-1 and H-5 in 8b is consistent with the axial anomeric functional group orientations of 4b, 6b and 8b. A clear H-1 to H-2 NOE correlation in 8b further supports the equatorial orientation of H-1. Synthesis of psuedoglycals with anomeric sulfones has also been reported with axial anomeric stereoselectivity 21 . Isomerization of the anomeric position was never observed for any of the sulfo-sialic acid analogues.
Due to the high electronegativity of the sulfo group, the diagnostic H-1, H-2, H-5 and C-1 chemical shifts of 5 and 7 are shifted downfield relative to the corresponding carboxylic acids and phosphonic acids ( Table 2). This observance of a strong electron-withdrawing inductive effect of the anomeric sulfo group supports our hypothesis that sulfo-sialic acid analogues will be useful for the development of next-generation covalent NA inhibitors 3 .
In order to assess the relative effectiveness of sulfo-sialic acid analogues, the anomeric phosphonic acids corresponding to 7a (eqPO 3 H, ePANA) and 7b (axPO 3 22 . Inhibition of influenza and bacterial NA by sulfo-sialic acid analogues confirms that the sulfo group enables more potent NA inhibition relative to the analogous phosphonic acids (Table 3, Fig. 3). The NA inhibitory activity of axPO 3 H was difficult to quantify due to its low potency. For inhibition of influenza N2, the equatorial anomeric sulfonate 7a was 40-fold more  potent than eqPO 3 H. The axial anomeric sulfonates 7b and 8b were also more potent N2 inhibitors relative to eqPO 3 H. In the case of CpNA inhibition, 7a was 8.5-fold more potent than eqPO 3 H. The sialic acid anomeric sulfo modification is advantageous in terms of drug resistance because of the essential role of the triarginyl binding site for natural NA substrate binding. Substitution of Arg118, Arg292 and Arg371 is known to result in impaired viral fitness due to decrease in NA activity and substrate affinity 24 . Although the Arg292Lys substitution is well documented, we showed that it results in an 88-fold K m increase and a decrease in the enzymatic activity of H7N9 NA 25 . One of the most common forms of NA inhibitor resistance is caused by a His275Tyr substitution, which confers a high level of resistance to oseltamivir and moderate resistance to peramivir 24,26 . Consistent with the results for A/RI/5+/1957(H2N2) NA, 7a inhibits H5N1 N1 and its drug-resistant His275Tyr form at the same level (Table 3, Fig. 4), demonstrating the effectiveness of sulfo-sialic acid analogues against highly-prevalent drug-resistant NA and highly-pathogenic avian-origin NA.
Sialic acid anomeric phosphonic acid analogues were first reported by Walliman and Vasella in 1990 4 . The long delay in the synthesis of sulfo-sialic acid analogues might be explained by challenges with the current synthesis. Some difficulty includes the similar chromatographic mobility of the anomeric acetoxy and anomeric acetylthio intermediates, the precise conditions of the acetylthio oxidation, difficulty in sulfo protection, and high water solubility of the protected sulfo-sialic acid analogues.
The clinically used NA inhibitors zanamivir, oseltamivir, peramivir and laninamivir inhibit influenza NA at a low nanomolar level 1 . Despite having a lower potency, the sulfo-sialic acid analogue 7a contains no other structural improvements and is still a more potent inhibitor of influenza NA than the oxocarbenium ion transition state analogue Neu4Ac2en (N2 IC 50 : 19 μm) 27 . Therefore, our sulfo modification in combination with other modifications may offer potential to produce sub-nanomolar level NA inhibitors. In conclusion, this report proves the concept that sulfo-sialic acid analogues are the most potent NA inhibitors with regards to anomeric functionality. Further development of novel sulfo-sialic acid based inhibitors with increased potency and efficacy is currently underway.    Neuraminidase Inhibition. Inhibition of neuraminidase (NA) activity was quantified using a fluorogenic assay based upon the method by Potier et al. 22 with modifications. 4-methylumbelliferyl-α-d-Neu5Ac  (4-MU-Neu5Ac) substrate was utilized at a final concentration of 120 μm for inhibition of N2 and Streptococcus 6646K NA (Amsbio). For inhibition of Clostridium perfringens NanJ NA (Sigma) the 4-MU-Neu5Ac concentration was 60 μm when using 7a and 120 μm when using eqPO 3 H. For IC 50 determination, 5-8 inhibitor concentrations were used with 1-4 replicates for each condition. For N1 K i determination, 5 inhibitor concentrations and 3 substrate concentrations were used. NA and inhibitor were pre-incubated for 20 minutes at room temperature before starting each assay with addition of 4-MU-Neu5Ac. Fluorescence of liberated 4-methylumbelliferyl (Ex.

Methods
355-360 nm, Em. 460 nm) was detected at room temperature using a Thermo Scientific Varioskan ™ Flash or a Molecular Devices SpectraMax Paradigm microplate reader. IC 50 values and confidence intervals were calculated by analyzing inhibition data using the 'log(inhibitor) vs. response (three parameters)' function of Prism 7 software. N1 K i values were calculated using the Michaelis-Menten function of Prism 7, and N2 K i values were calculated using the Cheng-Prusoff equation 23 . A/Anhui/1/2005 (H5N1) NAs were purchased from Sino Biological. Influenza A/RI/5+/1957(H2N2) NA was produced according to our previously reported methods 3 .