The RNA-dependent RNA polymerase of influenza viruses, which is responsible for replication and transcription of the single-stranded viral RNA genome, is heterotrimeric and composed of three subunits, namely PA, PB1 and PB2. Transcription of viral mRNA involves a unique process called ‘cap-snatching’, wherein the PB2 subunit binds host cell pre mRNAs via their cap structure, followed by cleavage at 10-13 nucleotides by an endonuclease activity of the PA subunit [9,10,11, 13]. Thereafter, these short capped oligomers are used to prime mRNA synthesis through the PB1 subunit [19]. Based on recent findings, it has been demonstrated that the PA gene of the 2009 pandemic H1N1 (pH1N1) virus is involved in increased virulence to both avian and mammalian hosts [24, 28]. PA mutations have also been shown to be linked to the adaptation of pH1N1 virus to new host species [21]. Several studies have investigated non-synonymous adaptive mutations [7, 28] located within the PA gene that are associated with enhanced polymerase activity in the mutant strains. In a recent study, Seyer et al. studied the adaptive potential of the influenza A virus subtype H1N1 variant isolate A/Hamburg/04/09 (HH/04) by sequential passaging in mouse lungs [22]. They found three synergistically acting mutations, which were defined as pathogenicity determinants, comprising two mutations in the hemagglutinin (HA[D222G] and HA[K163E]), and one in the PA (PA[F35L]), whereby the HA(D222G) mutation was shown to determine receptor binding specificity, and the PA protein F35L mutation increased the polymerase activity. They observed a 2.5-fold higher polymerase activity upon reconstitution of the polymerase complex with a PA containing the F35L mutation. Their study also showed a replication advantage and thus enhanced virulence conferred by the F35L mutation in the PA (F35L)-infected mice.

In the present study, we attempted to investigate computationally the effect of the PA-F35L mutation on the nature of specific binding aspects of mononucleotide uridine 5’-monophosphate (UMP), a component of the RNA endonuclease substrate. Several recent studies have demonstrated that MD simulations can provide key information about conformational changes of functional sites and are vital for analysis of the flexibility of ligand-binding cavities and, specifically, loop regions, both of which are known to govern the biological function of the viral protein [1, 2, 4, 6, 12, 17, 18, 20].

To select the wild-type and mutant-type viral strains, the PA gene sequences as of December 2016 (n = 8093) of pH1N1 viruses of the period 2009─January 2016 were downloaded from the Influenza Virus Database of NCBI and aligned using MEGA 7 [25]. Strain A/England/01220639/2009 was selected as the wild-type strain, and A/Mexico/200102/2014 was selected as the mutant strain (Fig. S1). The structures of the endonuclease domain of PA with bound UMP of wild-type and mutant-type strains were derived by homology modeling using the SWISS-MODEL server [5] (http://swissmodel.expasy.org/interactive#structure) and the X-ray crystallographic structure of the N-terminal domain of the PA subunit (PA-Nter) of A/California/04/2009/pH1N1 (PDB ID: 4AWH) as the template. Validation of the modeled structures was carried out using the PROCHECK tool from the Structure Analysis and Verification Server (SAVES) [https://services.mbi.ucla.edu/SAVES/]. The 3D structures of the protein complexes were visualized using BIOVIA Discovery Studio Visualizer (www.accelrys.com/products). The interaction energy of the complexes of PA with the respective ligand was calculated using Protein Interfaces, Surfaces and Assemblies (PISA), a web-based tool (http://www.ebi.ac.uk/pdbe/prot_int/pistart.html). The tool PISA evaluates the change in the standard Gibbs free energy of dissociation, ∆Gdiss= -∆Gint - T∆S, where ∆Gint is the binding energy, T is the temperature, and ∆S is the entropic cost of dissociation. The complex is driven towards the dissociated state when entropic cost prevails (∆Gdiss < 0). Hence, complexes with positive ∆Gdiss are considered chemically stable [15, 16]. MD simulations of the wild- and mutant-type complexes with the ligand UMP (PAN1 and PAN2, respectively) were performed using YASARA package version 13.6.16. The simulation system was solvated using explicit TIP3P water molecules in a periodic simulation cell with boundaries extending 20 Å from the surface of the complex, and neutrality was achieved by addition of counter ions (Cl- and Na+) of physiological strength (0.15 M). Simulated annealing was carried out until convergence was reached, i.e., during each 200 steps its energy enhanced by less than 0.05 KJ/mol per atom [14]. Minimization was followed by an equilibration procedure using the NVT ensemble to 310K. The resulting minimized and equilibrated models were used for MD simulations. The simulations were carried out at constant pressure (NPT Ensemble), and the temperature was controlled using a Berendsen Thermostat by rescaling atom velocities [3]. The production simulation was carried out for 20 ns. Snapshots of the trajectories at 25-ps time intervals were taken, trajectory analysis was performed in YASARA, and the Grace plotting tool (http://plasma-gate.weizmann.ac.il/Grace/) was used for plotting the RMSD values of the Cα atoms. Fluctuations ≤ 0.5 Å were considered to imply the stability of the complexes.

The 3D models for PA-Nter (wild and mutant) with bound UMP (PAN1 and PAN2, respectively) were retrieved from the SWISS-MODEL Server in complex with the ligand UMP and two metal ions (Mn2+), as was in the template 4AWH.PDB, and validated using SAVES. The template (4AWH.PDB) shows 100% and 99.49% sequence identity to the query sequence of the wild-type and mutant strain, respectively. Overall, 89% of the residues fell into the allowed region, ~9.3% residues in the additionally allowed region, ~0.5% in the generously allowed region, and ~0.5% in the disallowed region. Interaction energy analysis showed an interaction energy value of 8.2 kcal/mol for the PAN1 complex and 8.3 kcal/mol for PAN2 complex. Upon protein-ligand interaction analysis of both complexes (Fig. S2), it was observed that in both cases the intermolecular contacts were the same as observed in the crystal structure complex, 4AWH.PDB (Fig. 1).

Fig. 1
figure 1

Endonuclease domain of influenza virus A/H1N1 based on crystal structure 4AWH.PDB (a) 3D view of the PA N-terminus, indicating active site residues (labeled in green), metal ions Mn2+ (ball representation) and ligand UMP (in yellow) along with the mutational position Phe35 (labeled in white). The close-up view of UMP in the endonuclease pocket is shown at the right. (b) Two-dimensional interaction plot of ligand-interacting residues and metal ions Mn2+using the Discovery Studio Visualizer

Full size image

The root mean square deviation (RMSD) plot of the backbone Cα atoms as a function of simulation time for PAN1 and PAN2 complexes (Fig. 2) revealed a stable conformation of ligand UMP in the endonuclease domain after ~10 ns for PAN1, while stability was observed at ~12 ns in the case of PAN2. The final stable conformation of the PAN1 complex at 20 ns showed the interaction energy value to be 4.7 kcal/mol, while the value for the PAN2 complex was 4.8 kcal/mol. The stable conformation of the PAN1 complex showed pi-pi stacking of the uridine base with Tyr24 and a hydrophobic interaction of the base with Lys34. On the other hand, the stable conformation of the PAN2 complex at 20 ns showed an H-bond interaction between the base and Lys34, pi-pi stacking of the base with Tyr24, and an additional H-bond interaction between the base and Tyr24. Further, both complexes showed an attractive charge interaction with Lys134. Other residues such as Ile38, Ala37, His41, Glu119, Glu80, Asp108, Tyr130, Val122, and Arg124 were involved in van der Waals interactions in the case of the PAN1 complex. In the case of the PAN2 complex, residues Ile38, Ala37, His41, Glu119, Glu80, Asp108 and Tyr130 formed van der Waals interactions. In both the PAN1 and PAN2 complexes, the two Mn2+ metal ions formed interactions with the phosphate group of ligand UMP (Fig. 2).

Fig. 2
figure 2

RMSD plot of the MD simulations carried out over a period of 20 ns for ligand UMP (left) and two-dimensional (2D) analysis of interaction of the endonuclease domain with ligand UMP upon MD simulation using the Discovery Studio Visualizer at the 20 ns simulation time point (right). (a) PAN1 complex. (b) PAN2 complex

Full size image

The PA-Nter possesses an α/β architecture consisting of five β-strands surrounded by seven α-helices and encompasses the active site of the endonuclease [9, 26]. The binding site in the endonuclease domain is known to exert some selectivity on nucleotide binding [23, 27]. In the crystal structure complex 4AWH, the active site comprises four metal-binding residues (His41, Glu80, Asp108 and Glu119) and one catalytic residue (Lys134), which are conserved amongst all influenza A and B virus strains (Fig. 1). Isothermal titration calorimetry (ITC) and studies of crystal structure complexes of PA-Nter with bound nucleotides have shown that two Mn2+ ions bind more tightly than one Mg2+ ion in the active site. Other important residues of the active site include Tyr24 and Lys34 [8, 10, 13]. Here, we have investigated the interaction of PA-Nter with only one ligand (viz. UMP) although the crystal structure of PA-Nter with AMP was available in PDB for influenza virus A/H5N1 (3HW5.PDB). This was because it was observed that rUMP in the endonuclease domain of pH1N1 (PDB entry: 4AWH) showed a different conformation [13] than that of the published structure [27] of the endonuclease domain of H5N1 with the ligand UMP (PDB entry 3HW3). The ribose and base positions were relatively different in 3HW3.PDB and unable to interact with Lys34 or Tyr24 [13].

Our results showed that contacts with the active site residues Tyr24 and Lys34 play a significant role for proper positioning of the ligand in the endonuclease domain cavity. We noted that, in the case of the mutant strain (PAN2), in addition to the pi-pi stacking interaction of Tyr24 with the base, the H-bond interaction was also favorable with Lys34. Compared to the crystal structure, an additional H-bond was made by Tyr24 with the base. This may imply that the mutant site Leu35 being in the close vicinity of these residues may have an impact on the bonding pattern of Lys34 and Tyr24 (Fig. 3a and b). The additional H-bond interactions in the case of the PAN2 complex may indicate a better binding affinity for the mutant strain PA with UMP. The positive values of interaction energies in terms of the Gibbs free energy of dissociation in both PAN1 and PAN2 reflect a stable system (Table S1). Structural comparison of both complexes (PAN1 and PAN2) showed a shift in the orientation of the α3 helix, where the mutation site (F35L) is located (Fig. 3b and c). This also results in a displacement of the loop connecting the α3 and α2 helices and a subsequent bending of the α2 helix such that Tyr24 orients itself closer to the uridine base, forming the additional H-bond. Notably, the α3 and α2 helices of the crystal structure and PAN1 complex (wild type) superimposed well (Fig. 3a). The displacement of the α3 helix might be due to the linear nature of the side chain of leucine as opposed to the aromatic nature of the side chain of phenylalanine at residue position 35 in the wild-type cases.

Fig. 3
figure 3

Superimposition of (a) PAN1 and crystal structure (4AWH.PDB), (b) PAN1 and crystal structure (4AWH.PDB), and (c) PAN1 and PAN2. The displacement of the α3 helix in the case of the mutant endonuclease domain (PAN2) is illustrated in b and c

Full size image

Our study, which demonstrates improved binding of UMP in the PAN2 complex, is suggestive of increased fitness of the mutant in terms of cleavage at a UMP site of the host pre-mRNA, which could enhance the polymerase activity. This is in correlation with the reverse genetics studies done by Seyer et al. [22]. MD simulation studies can thus complement experimental studies of novel mutations in influenza virus strains. A search of the GenBank database for pH1N1 strains possessing this mutation revealed only a few strains from human hosts (seven strains from 2009, two strains from 2010, a single strain from 2011 and 2014, and two strains from 2016 out of a total of 8,093 strains as of January 2016). Further studies related to the binding affinity of the other linear nucleotide AMP [27] in the endonuclease domain of PA of pandemic influenza A virus could provide additional clarity.

In summary, the binding affinity for PA-Nter endonuclease ligand (nucleoside monophosphate) related to cleavage of host cell pre-mRNAs in pH1N1 viruses was studied. We attempted to throw some light on the molecular mechanism involved in PA-protein-based RNA–enzyme interactions, which might be further be useful for the design of endonuclease-specific inhibitors to combat influenza virus infections.