Structural and functional characterization of neuraminidase-like molecule N10 derived from bat influenza A virus

Overall Structure.

The ectodomain of N10 from influenza virus A/little yellow-shouldered bat/Guatemala/153/2009 (H17N10) was cloned and expressed using a baculovirus expression system based on a previously reported method (18, 19, 23) with slight modifications. The crystal structure of N10 was solved at a resolution of 2.20 Å and exhibited an overall structure similar to that of canonical influenza virus NAs. N10 subunits were assembled into a box-shaped tetramer, with each monomer containing a propeller-like arrangement of six-bladed β sheets (Fig. 1A). Comprehensive structural alignment among N10, influenza B NAs (with NA from influenza B Beijing/1/87 as the representative), and all available canonical influenza A NA subtypes revealed rmsds for Cα atoms of one N10 monomer relative to the other NAs ranging from 1.463 to 2.757 Å (Table S1). The average rmsd was much higher than that between two canonical influenza A NAs (<1 Å), but only moderately higher than that between canonical influenza A NAs and influenza B NA. N10 is most structurally similar to N1, particularly 09N1, and least related to influenza B NA (Table S1). Therefore, the structural alignment is consistent with the recent phylogenetic analysis indicating that N10 is divergent from all other influenza NAs (7).

Unexpectedly, N10 was found to not contain four antiparallel β strands in each blade. As illustrated by superimposition with the N2 structure from A/Tokyo/3/67 (H2N2) virus (PDB ID code 1NN2), N10 lacks a β strand in blade 6, which is buried in the protein core, and instead contains a loop structure (Fig. 1B). Two additional unique NA structural features were also found by comparison with available influenza NA structures (excluding the drug escape mutants); N9 derived from tern [Protein Data Bank (PDB) ID code 7NN9] has only three β strands in blade 6, whereas the duck influenza N6 (PDB ID code 1V0Z) lacks one β strand in both blade 4 and blade 6.

A detailed structural analysis confirmed the presence in N10 of the highly conserved N-glycosylation site N146 (N2 numbering is used throughout the text according to the sequence alignment shown in Fig. S1) shared by all known influenza A NA structures. Interestingly, two sites, N259 and N269, were also identified in blade 4 of N10. The N-linked glycans of N259 are located at the bottom of the N10 head and protrude toward the virion membrane, whereas in N269, the glycans are near the top of the head and point away from the center of the putative enzymatic active site (Fig. 1C). Regarding the calcium-binding site, each N10 monomer has the conserved high-affinity site important for stability in all known NAs (19). This site in N10 is formed with slight differences from other NA structures, in which the calcium ion is coordinated by the main chain carbonyl oxygens of N293, D297, G345, and G347; the carbonyl group of the D324 side chain; and a water molecule (Fig. 1D).

Lack of Sialidase Activity.

Influenza A virus NAs can be grouped based on their primary sequences (group 1 and group 2) and featured by variant structural properties (open or closed state) of the 150-loop (24). As demonstrated by an in vitro fluorescence-based activity assay, typical group 2 N2, atypical group 1 09N1, and typical group 1 N5 displayed high sialidase activity and similar K_m values (20.3–36.1 μM) as reported previously (25, 26). In contrast, N10 exhibited no NA activity at a concentration of 1 μM using the standard NA substrate methylumbelliferyl-N-acetylneuraminic acid (MUNANA) (Fig. 2). Even at a concentration of 10 μM, no N10 activity was detected, demonstrating that N10 lacks sialidase activity and may have an unknown function distinct from canonical influenza NAs.

Alterations of Conserved NA Active Site Architecture.

We next investigated the structural basis for the inability of N10 to bind and hydrolyze the sialidase substrate with a terminally linked sialic acid. As illustrated in Figs. 1E and 3 by the N2-Neu5Ac complex (PDB ID code 2BAT), a classical reference structure, the charged pocket accommodating sialic acid in typical influenza A NA, can be divided into two distinct regions. First, there is an important “edge” region consisting of a positively charged arginine triad (R118, R292, and R371), as well as R152 and R224. The arginine triad forms critical high-energy salt bridges with the negatively charged C1 carboxylate of sialic acid, whereas R152 forms a hydrogen bond with the sialic acid N-acetyl group. Second, there is a negatively charged “platform” region composed of E119, E227, E276, and E277, located below the bound sialic acid. E119 and E276 form hydrogen bonds to noncharged portions of the sialic acid, and E276 is thought to play an important role in the catalytic mechanism via interaction with Tyr406 (27). All of these residues, including influenza B, are highly conserved in N1–N9. In contrast, N10 lacks the conserved arginine triad responsible for binding of the sialic acid carboxylate group (Fig. 1F), and the corresponding negatively charged platform region is replaced by a positively charged region in the N10 structure. Why the platform region is negatively charged rather than positively charged or neutral remains unclear, and should be addressed in the near future.

In addition, N10 has a more open pseudoactive site cavity, conferring a lower capability, if any, to tightly bind sialic acid (Fig. 1 E and F). More specifically, the key catalytic site residues and the framework residues responsible for stabilizing the active site are highly conserved (with the exception of D/N for residue 198) among N1–N9, as well as influenza B virus NA, but is highly divergent in N10 (Fig. 3A). Surprisingly, N10 contains only three of the eight canonical influenza NA catalytic site residues. The salt bridge between R224 and E276 is maintained in N10. The conformation of E276 is nearly the same as that observed in the NA-oseltamivir complex structures and might be sufficiently flexible to hydrogen-bond with the glycerol group of sialic acid, as reported previously (11, 16). R118 is shifted by 1.20 Å (Cα atom), and the position of its side chain is oriented toward the bottom of the pseudoactive site. Furthermore, residue 371 is found 2.35 Å (Cα atom) away from the center of the referenced NA active site, and both residues 292 and 371 are replaced with shorter polar residues, T and Q, respectively. In addition, residues 151 and 152 are shifted even further away from the pseudoactive site (11.72 and 11.63 Å, respectively, in terms of Cα atoms compared with the N2-Neu5Ac complex). These alterations lead to the loss of the crucial interactions essential for the recognition of the sialidase substrate. Most importantly, the proposed key nucleophilic residue Y406 is replaced by a positively charged R406 in N10 (27), which is unlikely to function in a similar manner, especially with its side chain oriented in the opposite direction relative to the canonical Y406 (Fig. 3B). The deficiency of a key catalytic residue further supports the idea that N10 is not a canonical sialidase.

As stated earlier, only 3 of the 11 framework residues of canonical influenza NAs could be identified in N10. The conformations of both E425 and S179 in N10 resemble those in the N2-Neu5Ac complex; however, the conserved E119, E227, W178, and I222 residues, which also form contacts with sialic acid, are replaced in N10 by R119, N227, R178, and P222, respectively. This both further weakens the sialic acid-binding capability of N10 and greatly contributes to replacement of the highly conserved acidic pocket with a predominantly basic site in N10. Further analysis of the remaining five residues revealed an interesting pattern in which framework residues are altered in coordination with the substitution of key catalytic site residues. In the canonical N2-Neu5Ac complex, E277, N294, H274, and D198 help stabilize the conformation of Y406, R292, E276, and R152, respectively, and R156 is important for stabilizing the conformation of E119 and the 150-loop. Nevertheless, the N10 D198 is shifted 3.6 Å (Cα atom) outward from the center compared with D198 in the N2-Neu5Ac complex. Moreover, the remaining four residues are replaced by Y277, L294, N274, and E156, resulting in significantly different chemical properties (Fig. 3C). Overall, the effects caused by these alterations explain why N10 has a unique pseudoactive site architecture.

Special Features of the 150-Loop.

Numerous studies have confirmed that the 150-loop is one of the most flexible parts of the influenza NA structure (28–30), and there are striking structural differences among variant influenza NA subtypes at this site (17, 18, 20). No 150-cavity in N10 could be identified compared with the uncomplexed typical group 1 VN04N1 (NA from Vietnam 2004 H5N1; PDB ID code 2HTY) and typical group 2 N2 structures (Fig. 4 A–C).

Detailed comparison of these three structures revealed an abnormally extended conformation of the N10 150-loop, which is located far beyond the VN04N1 active site. The 430-loop of N10 is also moved outward relative to the center of the referenced active sites so as to maintain coordination with the 150-loop, in accordance with other known NAs (18, 28). Compared with the van der Waals interaction between P431 and I149 in 09N1 (PDB ID code 3NSS), L144 forms a more stable hydrophobic interaction with L433 in N10 (Fig. 4D).

In view of a tetrameric N10, the 150-loop of each monomer is positioned much closer to the respective neighboring monomer compared with that in other influenza NAs (Fig. 4E). Interestingly, it is directly involved in the intermolecular interaction at the interface of two NA monomers, which has not been reported previously. There are strong polar contacts between the 150-loop of one N10 molecule and the 100-loop (residues 105–109) of the other. In particular, the S151 main chain nitrogen hydrogen-bonds with the P105 carbonyl, and the N146 carbonyl group forms a hydrogen bond with the E109 peptide bond nitrogen. In addition, two water molecules mediate the interaction of the S151 side chain hydroxyl group with the G107 carbonyl oxygen, whereas the E109 side chain exhibits an extended state to hydrogen-bond with the N146 peptide-bound nitrogen (Fig. 4F). We also noted the consistent presence of a helix in place of the N10 100-loop in all other active influenza NA structures. The unique secondary structure of the N10 100-loop may be attributed primarily to the presence of two proline residues at positions 105 and 108.

Cloning, Expression, and Purification of Influenza Virus NAs.

The preparation procedures for recombinant 09N1, N2, and N5 were as described in our previous reports (18, 20, 24). Recombinant N10 protein was prepared using our established baculovirus expression system. The cDNA encoding the ectodomain (residues 83–460, based on N2 numbering) of influenza A/little yellow-shouldered bat/Guatemala/153/2009(H17N10) NA-like protein N10 was cloned into the pFastBac1 baculovirus transfer vector (Invitrogen). A GP67 signal peptide was added at the N terminus to facilitate secretion of the recombinant protein, followed by a His tag, a tetramerizing sequence, and a thrombin cleavage site (18). Recombinant pFastBac1 plasmid was used to transform DH10BacTM Escherichia coli (Invitrogen). The recombinant baculovirus was obtained following the manufacturer’s protocol, and Hi5 cell suspension cultures were infected with high-titer recombinant baculovirus. After growth of the infected Hi5 suspension cultures for 2 d, centrifuged media were applied to a 5-mL HisTrap FF column (GE Healthcare), which was washed with 20 mM imidazole. Then NA was eluted using 300 mM imidazole. After dialysis and thrombin digestion (3 U/mg NA; BD Biosciences) overnight at 4 °C, gel filtration chromatography was performed with a Superdex 200 10/300 GL column (GE Healthcare) with 20 mM Tris⋅HCl and 50 mM NaCl (pH 8.0). High-purity NA fractions were pooled and concentrated using a membrane concentrator with a molecular weight cutoff of 10 kDa (Millipore).

Enzymatic Activity Assay.

The activities of purified 09N1, N2, N5, and N10 were tested using MUNANA (J&K Scientific) as a fluorogenic substrate (26). The appropriate protein and substrate concentrations were chosen after several rounds of preliminary tests. Protein (10 μL) was mixed with 10 μL of buffer containing 33 mM MES and 4 mM CaCl₂ (pH 6.0) in each well of a 96-well plate, after which the plate was incubated at 37 °C. The final concentrations were 10 nM for 09N1, 10 nM for N2, 10 nM for N5, 1 μM for N10, and 10 μM for BSA. Serial dilutions (0–500 μM) of preheated MUNANA (30 μL) were then added. The fluorescence intensity of the released product was measured every 30 s for 1 h at 37 °C on a microplate reader (SpectraMax M5; Molecular Devices), with excitation and emission wavelengths of 355 nm and 460 nm, respectively. All assays were performed three or more times, and the K_m and V_m values for active NAs were calculated using GraphPad Prism.

Crystallization, Data Collection, and Structure Determination.

N10 crystals were obtained using the sitting-drop vapor diffusion method. N10 protein [1 μL of 10 mg/mL protein in 20 mM Tris and 50 mM NaCl (pH 8.0)] was mixed with 1 μL of reservoir solution [30% (wt/vol) PEG2000 monomethyl ether (MME), 0.1 M sodium acetate (pH 4.6), and 0.2 M ammonium sulfate]. N10 crystals were cryoprotected in mother liquor with the addition of 20% (vol/vol) glycerol before being flash-cooled at 100 K. Diffraction data for N10 were collected at Shanghai Synchrotron Radiation Facility beamline BL17U. The collected intensities were indexed, integrated, corrected for absorption, and scaled and merged using HKL-2000 (38). Data collection statistics are summarized in Table S3. The structure of N10 was solved by molecular replacement using Phaser (39) from the CCP4 program suite (40), with the structure of 09N1 (PDB ID code 3NSS) as the search model. The initial model was refined by rigid body refinement using REFMAC5 (41), and extensive model building was performed using COOT (42). Further rounds of refinement were performed using the phenix.refine program implemented in the PHENIX package (43) with energy minimization, isotropic ADP refinement, and bulk solvent modeling. Final statistics for the N10 structure are presented in Table S3. The stereochemical quality of the final model was assessed with PROCHECK (44).