Evaluation of a Set of C9 N-acyl Neu5Ac2en Mimetics as Viral Sialidase Selective Inhibitors

Identification of selective influenza viral sialidase inhibitors is highly desirable in order to minimize or avoid the adverse effects due to the possible inhibition of endogenous human sialidases. We recently reported the evaluation of C9 N-acyl Neu5Ac2en mimetics as probes for human sialidases. Herein, we describe the in vitro activity of the same set of C9 N-acyl Neu5Ac2en mimetics against sialidases expressed by influenza virus A/PR/8/34 (H1N1), A/Memphis/1/72 (H3N2), and A/Duck/313/78 (H5N3) strains. Compound 8 is identified as a promising starting point for the development of viral sialidase selective inhibitors. Multiple sequence alignment and molecular docking techniques are also performed to explore the plausible interaction of compound 8 with viral sialidases.


Introduction
Influenza is a perceivably benign condition that develops in approximately 20% of the world's population and kills 0.25 to 0.5 million people every year worldwide, according to the WHO [1]. Influenza can cause a high level of mortality, particularly in children, elderly, or those with chronic underlying conditions of lung, heart, kidney, and so forth [2]. There have been three influenza pandemics in the 20th century, and this has lead to millions of deaths with the appearance of a new strain of the virus in each pandemic [3,4]. Since June 11, 2009, a new strain of swine-origin influenza A virus subtype H1N1 has been declared as the first global influenza pandemic of the 21st century. As of July 4, 2010, over 18311 deaths in more than 214 countries have been confirmed (http://www.who.int/csr/ don/2010 07 09/en/index.html). Influenza viruses belong to the Orthomyxoviridae family and are divided into three types, namely, A, B, and C. Influenza A virus, in particular, represents a significant health risk to the public due to both its ability to spread rapidly among humans and being associated with major epidemic outbreaks [5].
In late 1960s, DANA 2 (5-acetamido-2,6-anhydro-3,5dideoxy-D-glycero-D-galacto-non-2-enonic acid, Neu5Ac2en), a transition-state analogue, is identified as an effective first inhibitor of sialidase enzymes [23]. Using the advantage of structure-based drug design method, two potent sialidase inhibitors, Zanamivir 3 (Relenza, GlaxoSmithKline) and an ester prodrug oseltamivir 4 (Tamiflu, Gilead/Roche), were designed and developed for the prevention and the treatment of influenza and were complementing the widespread use of influenza vaccines [24,25] (Figure 1). Both sialidase (NA) enzyme inhibition and X-ray crystallography studies of these inhibitors also suggest that the strategy of designing an inhibitor of NA that binds to the highly conserved active site of the NA achieves the desired goal of activity against all influenza NA subtypes, N1-N9, and influenza B viruses [26,27]. However, with use of these licensed drugs, several mortalities, severe allergic reactions, and neuropsychiatric events have been reported, particularly in Japan [28]. Also, the FDA has issued a warning label for Tamiflu after reports of serious psychiatric side effects in patients receiving the drug, especially in children [29]. Some of the observed adverse effects have been speculated as a reason of nonselective endogenous human sialidase inhibition by these drugs, although no statistically significant relationship has been established till present [30,31].
Four types of human sialidases are known and have been classified based on their subcellular localization, namely, the intralysosomal sialidase (NEU1), the cytosolic sialidase (NEU2), the plasma membrane-associated sialidase (NEU3), and the lysosomal or mitochondrial membraneassociated sialidase (NEU4) [32]. These isoforms differ in their substrate specificities, enzymatic properties, and physiological functions. Human sialidases are involved in a wide variety of biological processes through modulating the sialoglycoconjugates [33]. They are implicated in various cellular events such as cell metabolism, cell differentiation, cell growth, and apoptosis including immune functions [34]. Human sialidases, although differ from the viral sialidases in their primary structure and enzymatic properties, show striking similarities in the tertiary structural aspects and active-site architecture [35]. Therefore, active site-directed nonspecific inhibitors of viral sialidase could also potentially inhibit human sialidase isoforms. For instance, inhibition of NEU3 in normal subjects may contribute to the onset of neuropsychiatric symptoms since NEU3 is proved to be an important molecule in the neuronal differentiation [30]. The above mentioned facts suggests that there is  (4).
Glu264 Figure 2: Schematic representation of DANA and its four constituent groups required for binding and conserved residues of NEU1 active site (purple) in comparison with viral sialidase active site (green). a significant need for viral sialidase selective inhibitors with other improved properties (including increased efficacy and reduced sensitivity to resistance) relative to the currently marketed drugs. The structural analysis of various sialidase-DANA complexes suggests that four main functional groups that bind to the enzyme active site are 2-carboxylate, 4-hydroxyl, 5-Nacetyl, and 6-glycerol moiety ( Figure 2) [36]. More importantly, 2-carboxylate (-C 1 O 1A O 1B H) group is absolutely necessary for the inhibitory activity. It is well established that International Journal of Medicinal Chemistry 3 the Arg triad in the immediate vicinity of the carboxylate group plays a key role in orienting and stabilizing various inhibitors [37,38]. In our previous work, we reported a series of amide-linked C9 modified DANA analogues as human sialidase inhibitors, and structure-based methods were used to investigate the observed activity [39]. Results showed that some of the C9 amide-linked hydrophobic analogues of DANA are selective for human lysosomal sialidase (NEU1). The mixed structure and sequence alignment of viral sialidase with human sialidases revealed that NEU1 has a higher sequence similarity (34%) as compared to other isoforms (25-27%), and the active site architecture of NEU1 is more closely related to viral sialidase than other isoforms though striking variations are present [36]. In particular, near C9 of DANA, the conserved E277 of viral sialidase that forms important H-bonds with O8 and O9 hydroxyl groups of DANA is also conserved in NEU1 with a chemically equivalent residue (Asp263), while other isoforms are substituted with nonpolar amino acid residues ( Figure 2). Considering this, we planned to evaluate certain C9 DANA amides, (I), against viral sialidases and compared their selectivity over human sialidases. Herein, we report the viral sialidase inhibitory activities and complete experimental details of synthesis and screening of C9 amidelinked hydrophobic analogues.

Sialidase Inhibitory Activity. A set of C9 N-acyl
Neu5Ac2en derivatives were synthesized according to the procedures previously described. The viral sialidase inhibitory activities of compounds (5)(6)(7)(8)(9)(10)(11)(12)(13)(14) were evaluated against three different strains of influenza A virus (H1N1, H3N2, and H5N3) using DANA 2 as a reference compound. Results expressed as 50% inhibitory concentration (IC 50 ) values are presented in Table 1 together with the previously reported human sialidase inhibitory activities of compounds (for 2, 5, and 8) [39]. Some compounds demonstrate moderate to good inhibitory activities against viral sialidases with IC 50 of lower than 100 μM. Moreover, most of the compounds show more activity against sialidase of H1N1 strain over other tested viral strains. The substitution of the small groups by higher homologues at C9 amide position results in a substantial loss of inhibitory activity and might reflect a spatial restriction in the active site of viral sialidases. In particular, the bulky groups of compounds (6, 7, and 14) severely diminish the activity, and this effect becomes more evident with other viral sialidases of H3N2 and H5N3. Indeed, compounds 5 (22 μM for H1N1, 34 μM for H3N2, and 97 μM for H5N3) and 8 (9 μM for H1N1, 16 μM for H3N2, and 98 μM for H5N3) show IC 50 values close to DANA (1 μM for H1N1, 9 μM for H3N2, and 11 μM for H5N3) and inhibit the viral sialidases with higher affinity as compared to other compounds in this series. Interestingly, the viral/human sialidase selectivity profiles of compounds 5 (IC 50 58 μM for NEU1, >1000 μM for NEU2, >1000 μM for NEU3, and 580 μM for NEU4) and 8 (680 μM for NEU1, >1000 μM for NEU2, >1000 μM for NEU3, and 825 μM for NEU4) are better than those of DANA (143 μM for NEU1, 43 μM for NEU2, 61 μM for NEU3, and 74 μM for NEU4). These data indicate that compounds 5 and 8 could be further explored for the design of viral sialidase selective inhibitors.

Molecular Modeling.
A mixed structural-sequence analysis and molecular docking were performed to explain why there are notable differences in binding affinities and selectivities. The crystal structure of H1N1 (A/Brevig Mission/1/18) (PDB code, 3B7E) was selected as the representative for determining the information regarding spatial disposition discriminating residues [40]. The multiple-sequence alignments of the tested viral strains (H1N1 [AcqP03468], H3N2 [AcqP03475], and H5N3 [AcqA6YJ51]) and 3B7E were performed ( Figure 3). The best compound 8 was chosen to be docked into the DANA's binding site of H1N1 (3B7E), and then obtained protein-ligand complex was energetically minimized. We can infer from Figures 3 and 4 that the putative active site residues interacting with compound 8 are highly conserved in all three viral sialidases, except a striking difference at Ser246 (3B7E) and its neighboring residues. This difference is positioned in the vicinity of C-9 of DANA's scaffold and could be a reason for differential interaction of the present series of compounds with their binding regions. Taking a close look at the binding mode of C9 substituent of compound 8 in the active site of H1N1, it can be seen that one out of two conserved H-bond interactions of DANA with Glu (276 in H1N1) is broken and a new H-bond is formed with Ser246 (H1N1). In the case of H3N2 and H5N3, this residue is substituted with a hydrophobic residue Ala that can cause a loss of H-bond and a subsequent drop in the inhibitory activity as compared to H1N1. The cyclopropyl ring of compound 8 interacts with a fleck of hydrophobic surface mainly defined by the side chains of Asn211, Ile222, Arg224, Ser246, and Pro227. As seen in Table 1, the small structural differences at C9 substituents that likely interact with this hydrophobic surface had significant differences in the inhibitory activity. The above mentioned observations indicate that cyclopropyl group can form optimal interactions, and even a slight increase in size of the substituent can probably cause repulsive interactions with this hydrophobic surface. Although the difference is not significant, the slightly increased activity of compound 8 over compound 5 could be due to the conformational rigidity of the cyclopropyl group that is also proposed for its decreased activity against human sialidases as compared to DANA.
The results indicate that the cyclopropyl substituent at C9 shows an optimal inhibitory activity for the current series of compounds and also exhibits a better selectivity profile than the reference compound DANA. Multiple sequence alignment and molecular docking studies give some possible explanations on the interaction of compound 8 with the viral sialidases. We strongly believe that the information derived from present study could be potentially utilized for the design of selective viral sialidase inhibitors over human sialidases and such selective inhibitors are of interest in the context of the future development of anti-influenza agents with minimum or less adverse effects in influenza patients.

The Set of Investigated Compounds.
A total of 10 compounds (5-Acetamido-9-acylamido-2,3,5,9-tetra deoxy-Dglycero-D-galacto-non-2-eno-pyranosonic acids) (5)(6)(7)(8)(9)(10)(11)(12)(13)(14) selected for viral sialidase screening in this study were prepared according to the procedure previously described [39].  compound 8 and DANA were manually docked into the active site of sialidase structure separately. The obtained complexes were then energetically minimized with 500-1000 iterations of 'in situ ligand minimization algorithm' using the SMART MINIMIZER program. A distance constraint was applied between carboxylic group of inhibitors and conserved Arg triad in the active site. The interactions between compound 8 and active site amino acid residues were analyzed and also compared with DANA's binding mode to investigate the relative differences in interactions for understanding the affinity differences.