Screen Anti-influenza Lead Compounds That Target the PAC Subunit of H5N1 Viral RNA Polymerase

The avian influenza (H5N1) viral RNA polymerase protein PAC was used as a target to screen nine chlorogenic acid derivatives for their polymerase inhibitor activity. Among them, seven compounds were PAC ligands, and four inhibited influenza RNA polymerase activity. These results aid in the design of anti-influenza agents based on caffeoylquinic acid.


Introduction
Influenza virus, which has high rates of morbidity and mortality, is one of the major causes of viral respiratory infections. This virus mutates frequently, spreads rapidly, and occasionally transfers from animals to humans (e.g., avian influenza A, H5N1) [1]. The increasing geographic distribution of this epizootic virus has aroused serious concerns about the therapeutic methods currently available to curb a potential pandemic of this disease. To treat influenza, two classes of anti-influenza agents, M2 ion channel blockers and neuraminidase (NA) inhibitors, have been used [2,3]. However, the emergence of drug-resistant viruses has limited the effectiveness of these drugs [4][5][6]. Therefore, alternative anti-influenza agents with a low risk of drug resistance are urgently needed.
In the search for a new generation of anti-influenza drugs, the RNA polymerase has been a valuable target to confront the disease [7][8][9]. This heterotrimer, which contains PB1, PB2 and PA subunits, is responsible for viral RNA (vRNA) replication and transcription. Structural analysis of the PA subunit revealed several potential active sites. Mutations of certain residues in these sites could significantly reduce the polymerase activity or disrupt vRNA replication [10][11][12][13][14][15][16]. Moreover, the structure of the PA subunit is conserved across type A, B and C influenza viruses. This conservation indicates that the antiinfluenza agents targeting the PA subunit may be effective against most influenza strains and less susceptible to drug resistance [10][11][12]. Actually, some works have been done in the anti-influenza molecule screening targeting at PA [17][18][19]. In the pursuit of novel antiinfluenza agents, we present a screen against the carboxyl-terminal domain of PA (termed PA C , residues 257-716).

Results and Discussion
In previous work, we have discovered chlorogenic acid (CA) from Flos Lonicera Japonica and Eucommia Ulmoides Oliv extracts as a PA C ligand and potential anti-influenza active compound [17]. Herein, nine CA derivatives, the principal active components of several anti-influenza traditional Chinese medicines (e.g., Flos Lonicera Japonica [20], Stemona Japonica [21]), were tested (Table 1,  Table S1). To guide in the discovery of PA C ligands, flexible docking simulations were utilized to evaluate the binding affinities between the compounds and the target. The docking was carried out using the AutoDock (v.4.01) software package with active sites established according to the structural analysis of PA C [9]. The results were shown in Table S2 and Figure S9, S10, S11, S12, S13, S14, S15, S16, and S17, and binding affinities were summarized in Table 2. The binding affinities were scaled by pK d values, where pK d is the negative logarithm of the dissociation constant of the binding complex. Larger pK d values indicate stronger binding affinities. The docking results suggested that candidate compounds a-g are potential binders of PA C .
The binding between the candidate compounds and PA C was determined experimentally by analyzing the transverse relaxation change of the small molecule upon the addition of protein in a relaxation-edited NMR ( Figure 1, Figure S1, S2, S3, S4, S5, S6, S7, and S8). As shown in Figure 1, the application of the Carr-Purcell-Meiboom-Gill (CPMG) spin-lock can suppress or even eliminate the fast-relaxing signals [22,23] from the PA C ligand. By using this method, seven of the nine compounds were confirmed to be PA C ligands (candidates a, b, c, d, e, f, and g).
Thus, the number and position of caffeoyls on quinic acid is an important factor in PA C binding. On the other hand, although the docking simulation predicted that candidate b would bind tightly to the target, experimentally this compound behaved oppositely in affinity evaluation by NMR and SPR. This may be due to steric hindrance of the bulky trans-disubstitution of caffeoyl at the 1-and 3-position of the quinic acid. Other disubstitutedcaffeoyl quinic acids bind to PA C with high affinities because the caffeoyls are either cis-disubstituted or both in equatorial positions.
In order to address specificity of the binding, the interaction between compound a and Human Serum Albumin (HSA) was evaluated by relaxation-edited NMR. As shown in Supporting Information ( Figure S30), compound a did not specifically interact with HSA.
The PA C subunit is essential in influenza's RNA polymerase activity, so one may expect that PA C ligands should inhibit RNA polymerase activity as well. The effects of these compounds on polymerase activity were evaluated by an ApG primer extension assay [24]. The polymerase can use ApG as a primer to synthesize cRNA from vRNA promoters; therefore, the length of cRNA can be used to judge polymerase activity. As shown in Figure 2, the candidate compounds a, b, c, and d inhibited the synthesis of cRNA while candidate f showed slight inhibition. The inhibition rates of compound a, b, c, d, and f on polymerase activity were 73%, 54%, 81%, 67%, and 26%, respectively. Notably, candidate e enhanced the polymerase activity at a rate of 42%. Although these compounds were assayed at relatively high concentrations (5 mM), the discovery of the inhibition effectiveness on polymerase activity of these CA derivatives suggests that CA could be a lead structure for potential anti-influenza drugs. These results clearly show that disubstituted caffeoylquinic acids with low steric hindrance are more likely to be effective inhibitors against polymerase because they bind strongly to PA C . It was also observed that some PA C ligands did not exhibit activity in the ApG assay. This may well be due to their non-specific or weak binding to PA C .
In summary, the avian influenza viral RNA polymerase protein PA C , a conserved key target in the design of a new generation of anti-influenza agents, was used to screen lead plant-derived antiinfluenza compounds. Seven compounds were identified as PA C ligands. Among them, four compounds inhibited polymerase activity. Therefore, PA C made a useful target to screen for antiinfluenza agents. Based on the structure-activity relationship of CA derivatives as polymerase inhibitors, the position and number, and maybe also the synergistic effect, of the caffeoyls in quinic acid played important roles in the inhibition potential of polymerase activity. These results provided an important step in caffeoylquinic acid structure-based design of anti-influenza agents.

Protein expression and purification
Methods for the preparation of PA C protein were previously described [9]. Briefly, residues 257-716 of the PA subunit of avian H5N1 influenza A virus (A/goose/Guangdong/1/96) were cloned into a pGEX-6p vector (GE Healthcare) and transformed into Escherichia coli strain BL21. Cells were cultured in LB medium at 37uC with 100 mg/L of Ampicillin. When the OD600 reached 0.6-0.8, the culture was induced with 0.5 mM isopropyl-thio-Dglactosidase (IPTG) at 16uC. After 20 hours of incubation, the cells expressing PA C were harvested and combined by centrifugation at 5000 rpm for 10 min. Recombinant protein was purified with a glutathione affinity column (GE Healthcare). Glutathione S-transferase (GST) was cleaved with PreScission protease (GE Healthcare), and the protein complex was further purified by Q sepharose FF ion exchange chromatography and Superdex-200 gel filtration chromatography (GE Healthcare). Methods for the preparation of the RNA polymerase were previously described [25]. Briefly, the RNA polymerase (PA, PB1, and PB2) complex was expressed in hi5 insect cell and purified by Ni-affinity column, ion-exchange column, and gel exclusion chromatography.

The affinity analysis based on virtual docking
To predict the binding affinities between candidate compounds and PA C , the simulated flexible docking of ligands was carried out using the AutoDock (v.4.01) software package. The structure of PA C was retrieved from the Protein Data Bank (PDB entries: 3CM8) and modified for visual docking. First, in the PDB file,  Candidates f and g were identified to be PA C ligands, but the binding affinities were too weak to be evaluated by NMR methods. doi:10.1371/journal.pone.0035234.t002 water molecules were removed, polar hydrogen atoms were added, and non-polar hydrogen atoms were merged using the Hydrogen module in the AutoDock Tools (ADT). Then, Kollman united atom partial charges were assigned. The grid map of the docking simulation was established in a 61661661 cube centered on the target active sites referred to in a previous report [10]. The targets are defined as site 1: center of K328, K539, R566 and K574; site 2: center of K539, R566, K574 and N696; site 3: center of E410, K461, E524 and K536; site 4: center of F411, M595, L666, W706, F710, V636 and L640; and site 5: center of 620 and 621. There is a spacing of 0.375 Å between the grid points. When the ligand was docked to the PA C target, the Lamarckian genetic algorithm was used to optimize the conformation of the ligand in the binding pocket. The parameters were set to the following: the size of the population was 150; the number of energy evaluations was set to 1.0610 8 as the run terminates; for clustering the conformations, the root mean square deviation tolerance was set to 2.0; fifty independent docking runs were carried out for every ligand; and other parameters were set to default. The binding affinities of the candidate compounds to the targets were summarized in Table S2, and the average pK i of the five active sites was summarized in Table 2.

NMR experiments
NMR experiments were performed on a Bruker AVANCE 600 spectrometer equipped with a 5 mm BBI probe capable of delivering z-field gradients, using TOPSPIN software (Bruker, version 2.1) was used in experimental manipulating and data processing. All experiments were carried out at 298.2 K. The relaxation-edited NMR experiments utilized a [D/pre-saturation-90 x -(D-180 y -D) n -acquire] pulse sequence, in which the CPMG sequence was used for the spin-lock. For all relaxation-edited experiments the following variables were used: pre-saturation water suppression was applied in pre-acquisition delay (D = 3 s), P 90 was measured and set up for each sample, D = 1.5 ms, and 26n6D = total spin-lock time. The spectra were collected with 64 k of data points and 32 scans. Transverse relaxation times were measured by a pseudo-2D experiment using the CPMG sequence. Presaturation was applied in a pre-acquisition delay for water suppression, and 32 k and 16 data points were set for the F2 and F1 dimension, respectively. In each experiment, the following variables were used: D = 3 s (pre-acquisition delay); P 90 was measured and set up for each sample; D = 1.5 ms; the list of spin-lock loop (n) was set to 0, 10, 20, 30, 40, 50, 80, 100, 150, 200, 250, 300, 350, 400, 500, 600; and 26n6D = total spin-lock time. The spectra were collected with 32 scans, and each T 2 was calculated using TOPSPIN (Bruker, version 2.1) software by simulating the peak attenuation curve in different spin-lock times.

The affinity analysis based on T 2 simulation
The method to evaluate K d by T 2 simulation was previously reported [26]. When a ligand (L) and a protein (P) form a complex (LP), there is a dissociation equilibrium LP = L+P in the solution. The dissociation constant, K d , of this equilibrium is a key factor to describe the binding strength of the ligand and protein. It can be evaluated by the transverse relaxation time (T 2 ) of the ligand in the presence of the protein. The expression describing the observed effective transverse relaxation rate R 2obs (1/T 2obs ) as a function of molar ratio of protein to ligand (C P /C L ) is: Since the total concentration of PA C , C P , and the ligand, C L , in the experiments were known, it is possible to obtain the T 2F and T 2B from the plot of T 2obs versus C P /C L . Extrapolation of the curve to C P /C L = 0 should give T 2F and to infinite give T 2B . Then, by simulating the plot of R 2obs versus C P /C L using the above equation, K d can be simultaneously obtained. To evaluate the binding affinities between the candidate compounds and PA C , a series of samples containing small molecules and PA C with different concentration ratios (shown in Figure S9, S10, S11, S12, and S13) were prepared. These T 2 of the ligands were determined by the CPMG method using NMR. As we can see from Figure S9, S10, S11, S12, and S13, the observed T 2 (T 2obs ) of the ligand decreases when the concentration ratio of protein to ligand (C P /C L ) increases and the variation trends according to exponential decay. By simulating the plot of R 2obs (R 2obs = 1/T 2obs ) versus C P /C L using equation (6), the K d value can be obtained. The results of the simulation were summarized in Table 2.

The affinity analysis based on Surface Plasmon Resonance (SPR)
An affinity analysis of the interaction of candidate compounds a-g and PA C was carried out using an SPR spectrometer (Biacore 3000, GE Healthcare Bio-Sciences, Sweden). PA C at a concentration of 10 mg/mL in 10 mM NaAC, pH 5.5, was used to couple this protein to a CM5 sensor chip. To determine the affinity of small molecules to PA C , increasing concentrations (labeled in Figure S14, S15, S16, S17, S18, S19, and S20) of the small molecules diluted in running buffer (130 mM PBS, pH 7.5) were injected over the sensor chip for 60 s (association phase), which was followed by dissociation for 180 s and recording of the spectra. All of the experiments were performed at 25uC, and the flow rate was 30 mL/min. The surface at the end of each experiment was regenerated using 20 mM NaOH at 30 mL/min for 10 s to remove any bound analyte. The data were analyzed using the BIAevaluation software (4.1 version) to calculate the affinity constant. The association and dissociation kinetics plots of the small molecules to PA C were displayed in Figure S14, S15, S16, S17, S18, S19, and S20, and the results were summarized in Table 2.

Polymerase activity test
The ApG primer extension assay was previously described [24]. We performed 5 mL reactions with 2.5 mL 3P and 0.7 mM model vRNA promoter (an equimolar mixture of the 59-end vRNA 59-AGUAGAAACAAGGCC-39 and 39-end vRNA 59-GGCCUG-CUUUUGCU-39) in the presence of 5 mM MgCl 2 , 5 mM drug, 5 mM dithiothreitol, 1 mM ATP, 0.5 mM UTP, 0.5 mM CTP, 0.1 mM[a-32 P]GTP (3,000 Ci/mmol), and 2 U/mL RNasin (Promega). Where indicated, 0.5 mM ApG (Sigma) was added to the reaction. The reaction system can be described as: 0.25 mL 0.1 M MgCl 2 , 0.25 mL 0.1 M DTT, 0.5 mL 10 mM ApG, 0.25 mL RNasin, 0.25 mL 206NTP, 0.25 mL 206promoter, 0.25 mL [a-32 P]GTP, 2.5 mL polymerase, 0.5 mL depc H 2 O or 0.5 mL small molecule. Reactions were incubated at 30uC for 1 h. The loading buffer, 5 mL 26formamide/bromophenolblue/EDTA, and the mixture were heated at 95uC for 2 min. Analysis was performed by running the samples on an 18% PAGE with 16Trisborate-EDTA and 8 M urea followed by autoradiography. Figure S1 Binder screening by relaxation-edited NMR. Spectra of 3,4-dicaffeoylquinic acid (compound a) in the absence (plots a, b) and presence (plots c, d) of PA C . The CPMG spin-lock time of each experiment was labeled beside the spectrum. The concentration of the small molecule and PA C was 1.0610 23 mol/ L and 7.0610 26 mol/L, respectively. The water peak located at d 4.8 and 1 mM of TSP was added to the sample as a reference (d 0). The ligand peaks that attenuated when applying CPMG spinlock in the presence of PA C were marked with ''*'' in plot d. (TIF) Figure S2 Binder screening by relaxation-edited NMR. Spectra of 4,5-dicaffeoylquinic acid (compound c) in the absence (plots a, b) and presence (plots c, d) of PA C . The CPMG spin-lock time of each experiment was labeled beside the spectrum. The concentration of the small molecule and PA C was 1.0610 23 mol/ L and 7.0610 26 mol/L, respectively. The water peak located at d 4.8 and 1 mM of TSP was added to the sample as a reference (d 0). The ligand peaks that attenuated when applying CPMG spinlock in the presence of PA C were marked with ''*'' in plot d.