Establishment of an at-line nanofractionation-based screening platform for rapid identification of influenza PA_N/PA_N I38T inhibitors from Artemisiae Argyi Folium

Abstract

The N-terminal domain of influenza viral polymerase (PA_N), a highly conserved region with critical catalytic function related to viral RNA replication and transcription, is considered as a very promising anti-influenza drug target. There is an urgent need for highly efficient and rapid screening methods to identify potential PA_N inhibitors (PA_NIs) from complex matrices. In this work, a novel high-throughput screening (HTS) platform was established by coupling high performance liquid chromatography and high-resolution mass spectrometry (HPLC-HRMS) with a fluorescence resonance energy transfer (FRET)-based endonuclease activity assay through an at-line nanofractionation (ANF) system. The proposed screening platform could rapidly identify potential PA_NIs from plant extracts with good sensitivity (baloxavir at the half maximal inhibitory concentration (IC₅₀) could be detected) and reliability (Z’ factor of 0.77). This platform was then successfully applied to the screening of potential inhibitors against PA_N/PA_N I38T from an aqueous extract of Artemisiae Argyi Folium and 17 potential PA_NIs were identified. Among them, three compounds (cynarine, isochlorogenic acid B, and isochlorogenic acid C) showed comparable inhibitory activity against PA_N and even better activity against PA_N I38T, compared to baloxavir. This study not only established a novel high-throughput ANF-based PA_NIs screening platform, but also proved the feasibility to discover PA_NIs from complex traditional Chinese medicines (TCMs), which has a great potential in future anti-influenza drug discovery.

Graphical abstract

Highlights

• •An at-line nanofractionation screening platform was established for PA_N/PA_N I38T inhibitors for the first time.
• •This proposed platform presented good sensitivity and reliability.
• •17 potential PA_N/PA_N I38T inhibitors were identified from Artemisiae Argyi Folium.

1. Introduction

Seasonal influenza, a highly contagious acute respiratory disease caused by influenza viruses, is responsible for up to 650,000 deaths annually and poses a growing threat to public health [1]. While vaccination remains the primary preventive measure, antiviral drugs, particularly neuraminidase (NA) inhibitors, play an extremely important role in the war against influenza virus infections. However, the rapid evolution of influenza viruses can lead to mutations that result in reduced susceptibility and antiviral resistance. For example, the influenza viruses with the H275Y mutation in the NA proteins have shown significantly decreased efficacy of oseltamivir, peramivir, and other NA inhibitors [2]. The continuing evolution of influenza virus underscores the urgent need for new antiviral strategies, especially those focusing on different viral targets with highly conserved structural domains.

The N-terminal domain of polymerase acidic protein (PA_N), a core region of the influenza viral polymerase, has a cation-dependent endonuclease active-site pocket for viral RNA transcription and replication [3]. Several catalytic residues within this pocket are highly conserved among different influenza strains, and therefore it is considered as a promising target for novel anti-influenza drug development [4,5]. Baloxavir marboxil (BXM), the only U.S. Food and Drug Administration (FDA) approved PA_N inhibitors (PA_NIs) for clinical use, can effectively inhibit the transcription of influenza messenger RNA (mRNA), exhibiting potent efficacy against influenza A and B viruses with just a single dose. However, recent BXM clinical trials have reported serious side effects and poor effects in specific populations [6]. Therefore, substantial efforts have been devoted to discovering novel PA_NIs via pharmacophore modelling and virtual screening. For instance, the drug candidate L-735882 developed by Merck has shown good PA_N inhibitory activity, with a half maximal inhibitory concentration (IC₅₀) of approximately 1 μM [3].

Natural products exhibit remarkable chemical diversity and have continuously played a crucial role in drug discovery and development. Many plant-derived compounds have shown anti-influenza activity, such as terpenoids, quinones, phenylpropanoids, polyphenols, flavonoids, and alkaloids [7]. Among them, some flavonoids and polyphenols, particularly those enriched in traditional medicinal herbs, have been reported for their effective inhibitory activity against PA_N. These compounds interact with the catalytic sites of endonuclease through the tight bounds between their polyhydroxy functionalities and metal ions [8]. However, despite their potential, a comprehensive screening of PA_NIs from natural herbs has never been conducted. To rapidly screen PA_NIs from plant extracts still faces significant technical challenges due to the structural complexity of natural compounds and the variety of bioactivities in herbs. While bioassay-guided fractionation (BGF) is widely used for separating, purifying, and identifying bioactive compounds from herbs, this method is a time-consuming procedure, costly, and often leads to the loss of low-abundance bioactive ingredients. These limitations emphasize the urgent need for a rapid, high-throughput, and sensitive screening method to discover novel PA_NIs from herbs.

To overcome these challenges, we established a rapid and high-throughput screening (HTS) platform for potential PA_NIs in complex plant extracts by coupling high performance liquid chromatography and high-resolution mass spectrometry (HPLC-HRMS) with a fluorescence resonance energy transfer (FRET) based endonuclease activity assay through an at-line nanofractionation (ANF) system. The ANF system seamlessly links the analytical and pharmacological workflows, leading to a synchronization among chromatographic, mass spectrometric, and bioactivity profiles. This approach allows for the rapid identification of bioactive compounds in a single injection through the synchronized structural and bioactivity information [9]. Due to its high efficiency, high throughput, low sample consumption, and high compatibility, the ANF-based screening platforms have been successfully applied to the screening of inhibitors against NA [10], monoamine oxidases [11], and so on [12,13]. To establish a reliable PA_NIs screening platform, a robust and compatible PA_N inhibitory activity evaluation module is essential with the characteristics of high sensitivity, high stability, satisfactory speed, and ease of operation. Kowalinski et al. [14] reported a FRET-based endonuclease bioassay using an RNA oligonucleotide-based fluoroprobe. This method can rapidly recognize inhibitors at nanomolar concentrations in a fluorescence manner, offering a feasible solution for the rapid and sensitive evaluation of PA_N inhibitory activity [14,15].

In this work, the ANF and FRET-based endonuclease bioassay conditions were first optimized to maintain LC resolution and achieve satisfactory bioassay sensitivity. The validity, sensitivity, and reproducibility of the established ANF-based PA_NIs screening platform were then systematically evaluated using model compounds. Additionally, the versatility of this platform was investigated by applying it to PA_N I38T, a clinically relevant PA_N mutant with I38T substitution showing reduced susceptibility to baloxavir. Finally, the proposed ANF-based PA_NIs screening platform was applied to identify potential inhibitors against PA_N/PA_N I38T from the aqueous extract of Artemisiae Argyi Folium, a traditional Chinese herb used to treat dyspepsia, arthroncus, and anaphylactic diseases. Molecular docking experiments were further conducted to elucidate the interaction mechanism between the identified active compounds and PA_N/PA_N I38T.

2. Experimental

2.1. Chemicals and materials

Escherichia coli (E. coli) BL21 (DE3) was purchased from Thermo Fisher Scientific Inc. (Waltham, MA, USA). pGEX-6P-1-PA_N plasmid was supplied by Tsingke Biotechnology Co., Ltd. (Beijing, China). NaCl, glycine, MgCl₂, MnCl₂, l-glutathione (reduced), dimethyl sulfoxide (DMSO), KCl, Na₂HPO₄, KH₂PO₄, galantamine hydrobromide, Tris-HCl, NaOH, HCl, and formic acid (FA) were purchased from Aladdin Chemistry (Shanghai, China). 3-[(3-Cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS), glycerin, d,l-1,4-dithiothreitol (DTT), guanidine hydrochloride, phenylmethanesulfonyl fluoride (PMSF), isopropyl β-d-thiogalactoside (IPTG), Triton X-100, 10% NP-40, bovine serum albumin (BSA), ampicillin sodium salt (AMP), yeast powder, tryptone, and sodium dodecyl sulfate (SDS) were acquired from Sigma-Aldrich (Steinheim, Germany). Vicenin II, neochlorogenic acid, chlorogenic acid, cryptochlorogenic acid, schaftoside, cynarine, protocatechuic aldehyde, isochlorogenic acid B, isochlorogenic acid A, isochlorogenic acid C, favipiravir, baloxavir, esculetin, and caffeic acid were all bought from Chengdu Push Bio-technology (Chengdu, China). 12% SDS-polyacrylamide gel electrophoresis (PAGE) gel super quick preparation kit was supplied by Beyotime Biotechnology (Shanghai, China). 6-Carboxyfluorescein (6-FAM) 5′-TGGCAATATCAGCTCCACA-3′ black hole quencher 1 (BHQ1) and 6-FAM 5′-TGGCAATATCAGCTCCACA were purchased from Sangon Biotech (Shanghai, China). Glutathione Sepharose 4B was obtained from Cytiva (Marlborough, MA, USA). HPLC-grade methanol (MeOH) and acetonitrile (ACN) were purchased from Merck (Darmstadt, Germany). Deionized water was purified using a Milli-Q water purification system (Millipore, Molsheim, France). The 384-well plates were bought from Corning Inc. (New York City, NY, USA). Lyophilized aqueous extracts of Cassiae Semen, Pulsatillae Radix, Dioscoreae Rhizoma, Gynostemmatis Pentaphylli Herba seu Radix, Siraitiae Fructus, Dendrobii Caulis, Citri Reticulatae Pericarpium, Artemisiae Argyi Folium, Salviae Miltiorrhizae Radix et Rhizoma, Eucommiae Cortex, Nelumbinis Folium, Epimedii Folium, and Forsythiae Fructus were kindly supplied by Guangdong Yifang Pharmaceutical Co., Ltd. (Foshan, China).

2.2. Expression and purification of endonucleases

The DNA sequence encoding the PA_N from the influenza virus A/goose/Guangdong/1/1996 (H₅N₁) genome was synthesized by Tsingke Biotechnology Co., Ltd., and inserted between the BamHI and XhoI sites of a plasmid pGEX-6p-1. The recombinant PA_N with glutathione S-transferase (GST) affinity tag was expressed in E. coli BL21 (DE3) cells. The transformed bacteria, inoculated in Luria-Bertani (LB) medium containing 100 μg/mL ampicillin (Amp), was induced with 0.8 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) at 16 °C for 14 h. After induction, the cells were harvested and resuspended in buffer A (140 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄) and lysed with a high-pressure cell cracker (Union-Biotech (Shanghai) Co., Ltd., Shanghai, China) at the pressure of 640 bar. The target protein was purified by Glutathione Sepharose affinity chromatography (Cytiva) on an NGC™ Quest 10 Plus system (Bio-Rad Laboratories, Hercules, CA, USA). Unbound proteins were washed out with buffer A, while the purified PA_N was eluted using buffer B (10 mM GSH and 50 mM Tris-HCl at pH 8.0). The mutant protein PA_N I38T was cloned from the PA_N through site-directed mutagenesis, and then it was expressed and purified following the same procedure as described above [16].

2.3. Establishment and validation of the ANF-based PA_NIs screening platform

2.3.1. Liquid chromatography (LC) separation and eluate collection

The complete ANF-based PA_NIs screening platform combines LC-MS analysis and an endonuclease bioassay. A 20 μL of sample was injected into a Prominence LC-20A Modular HPLC system (Shimadzu, Kyoto, Japan) equipped with a photo-diode array (PDA) detector, and then separated on an RD-C₁₈ column (4.6 mm i.d. × 250 mm, 3 μm; Zhongpu Science Technology, Fuzhou, China). The mobile phases consisted of water with 0.1% FA (phase A) and MeOH with 0.1% FA (phase B). A binary gradient program was applied for sample separation, as detailed in the corresponding figure caption. The ultraviolet (UV) detection wavelength was set to 280 nm. After the LC separation, the eluate from the PDA detector was split at a 1:2 ratio. Two-thirds of the eluate was sequentially collected through a CTC PAL autoinjector (CTC Analytics AG, Zwingen, Switzerland) into 384-well plates at a designed fractionation frequency. The remaining eluate was directed into the MS system. After the eluate collection, the 384-well plates were dried in a vacuum oven at 35 °C to remove the mobile phase, and then used for the endonuclease bioassay.

2.3.2. MS analysis

MS analysis was performed on an AB SCIEX ZenoTOF 7600 system (AB SCIEX, Framingham, MA, USA). In negative mode, the electrospray ionization (ESI) parameters were set as follows: ion source temperature, 500 °C; pressure of ion source gas, 55 psi; pressure of curtain gas, 25 psi; ion spray voltage, −4500 V; time-of-flight (TOF)-MS scan range was set from 50 to 1500 Da with an ion accumulation time of 0.25 s ion, a declustering potential of −80 V, and a collision energy of −10 V; and TOF-MS/MS scan range was set from 50 to 1000 Da with a collision energy of −45 V (±20 V) and an ion accumulation time of 0.1 s. In positive mode, the ESI parameters were set as follows: ion source temperature, 500 °C; pressure of ion source gas, 55 psi; pressure of curtain gas, 25 psi; ion spray voltage, 5500 V; TOF-MS scan range was set from 50 to 1500 Da with an ion accumulation time of 0.25 s, a declustering potential of −80 V, and a collision energy of 10 V; and TOF-MS/MS scan range was set from 50 to 1000 Da with a collision energy of 45 V (±20 V) and an ion accumulation time of 0.1 s. SCIEX OS software was used for MS data analysis.

2.3.3. FRET-based endonuclease bioassay

A FRET-based bioassay using a TaqMan-like probe (sequence of 6-FAM 5′-TGGCAATATCAGCTCCACA-3′ BHQ1) as the enzymatic substrate was used in this study. Endonuclease and the probe were first dissolved in Tris-HCl (50 mM, 0.5% NP-40, pH 7.5) and Tris-HCl (10 mM, 0.5% NP-40, pH 8.0), respectively. Afterwards, 2 μL of DMSO (100%), 14 μL of reaction buffer (50 mM Tris-HCl, 50 mM NaCl, 1 mM DTT, 5 mM MgCl₂, 0.5 mM MnCl₂, and 1 mM CHAPS; pH 7.5), 2 μL of endonuclease (1:5-fold dilution), and 2 μL probe (250 nM) were sequentially added to each well of the dried 384-well plates using electric multi-channel pipettes (Thermo Fisher Scientific Inc., Ermelo, the Netherlands) under light-proof conditions. The final reaction volume was 20 μL per well. Finally, the 384-well plates were put into PerkinElmer EnVision® 2105 Multimode Plate Reader (Waltham, MA, USA), and the fluorescence intensity of each well was continuously recorded for 50 min at an excitation wavelength of 485 nm and an emission wavelength of 535 nm. The activity of endonuclease in the wells was calculated as the slope of the kinetic curve (fluorescence intensity vs. time), resulting from continuous measurements over 5–15 min. The time of each nanofraction collected was determined from the event logs of the above-mentioned CTC PAL autoinjector. The inhibitory bioactive profile was plotted using the slope values (Y-axis) and the time of each nanofraction collected (X-axis). The final integrated PA_NIs screening results were constructed by integrating the bioactive profile with HPLC chromatogram or total ion chromatogram (TIC) using GraphPad Prism 9.5 software (GraphPad Software, San Diego, CA, USA).

For IC₅₀ measurements, 2 μL of test compounds or natural product extracts at gradient concentrations, 14 μL of reaction buffer, 2 μL of endonuclease (1:5-fold dilution), and 2 μL probe (250 nM) were sequentially added to each well in the 384-well plate with a total final reaction volume of 20 μL, and its fluorescence intensity was measured. The inhibition rate was calculated according to the following equation:

$$\text{Inhibition}(\%)=\frac{S_i-S_n}{S_i-S_b}\times 100\%$$

where S_i and S_n represent the fitted slopes of the kinetic curve in the presence and absence of the sample, respectively. S_b is the fitted slope of the kinetic curve in the absence of endonuclease and the sample. IC₅₀ values were estimated by GraphPad Prism 9.5 software.

2.3.4. Validation

To assess the validity and reproducibility of the established ANF-based PA_NIs screening platform, a model mixture consisting of 1 mM baloxavir (positive control, polymerase acidic protein (PA) inhibitor), 1 mM galantamine hydrobromide (negative control, acetylcholinesterase inhibitor), and 1 mM favipiravir (negative control, polymerase basic protein inhibitor) was prepared and analyzed. To evaluate the sensitivity of the screening platform, the concentration of baloxavir was adjusted to 0.1, 1, 5, 10, 25, and 50 μM, and then was injected and analyzed by the screening platform. All experiments were performed in triplicate to ensure reproducibility.

2.4. HTS of potential PA_NIs from plant extracts

2.4.1. IC₅₀ measurement of plant extracts

13 lyophilized aqueous plant extracts were initially dissolved in 100% DMSO to prepare stock solutions at a concentration of 5 mg/mL. These stock solutions were then serially diluted (1:1) in DMSO to yield a series of concentrations (500, 250, 125, 62.5, 31.2, 15.6, 7.81, 3.90, and 1.95 μg/mL) on 384-well plates. Then the endonuclease inhibitory activity of each well was measured to determine the IC₅₀ values of the plant extracts.

2.4.2. HTS of the aqueous extract of Artemisiae Argyi Folium

The aqueous extract of Artemisiae Argyi Folium was dissolved in 10% MeOH at a concentration of 7.5 mg/mL and subsequently filtered through 0.22-μm nylon filter membranes prior to HPLC analysis. Then ANF-based PA_NIs screening was conducted following the procedure as described in Section 2.3, to generate the corresponding bioactive profile. All experiments were performed in triplicate.

2.4.3. Molecular docking analysis

The X-ray crystal structures of PA_N (Protein Data Bank (PDB) ID: 6FS6) and PA_N I38T (PDB ID: 6FS7) were obtained from Research Collaboratory for Structural Bioinformatics (RCSB) PDB database. The two-dimensional (2D) structures of all tested compounds were initially drawn using ChemDraw 21.0.0 (Revvity, Waltham, MA, USA) and then converted to 3D structures using Chem3D software (Revvity). The Open Babel GUI (available from http://openbabel.org/docs/) was used to convert ligands from “mol2” format to “PDB” format. After adding hydrogen, calculating and adding charges in AutoDock Tools 1.5.7 (available from http://vina.scripps.edu/), the small molecular candidates and protein targets were saved as PDBQT files. Molecular docking calculations were performed using AutoDock Vina 1.2.5 (available from http://vina.scripps.edu/) with two grid boxes (PA_N: 40 number of points in 3D with a 58.555, 122.372, and 3.681 (X, Y, and Z) center and spacing 0.510 Å; PA_N I38T: 40 number of points in 3D with a −30.947, 35.066, and −35.929 (X, Y, and Z) center and spacing 0.510 Å). The visualization of protein-ligand complexes and prediction of their interactions were achieved using Pymol 3.1.0 (Schrödinger, New York, NY, USA) and Discovery Studio 2019 (BIOVIA, San Diego, CA, USA), respectively.

3. Results and discussion

3.1. Establishment and optimization of the ANF-based PA_NIs screening platform

3.1.1. Optimization of endonuclease activity assay conditions

Two endonucleases (PA_N and its variant I38T) were used in this study. PA_N was employed for the optimization and validation of the established screening platform, while PA_N I38T was used for evaluating the versatility of this platform, and subsequently for constructing parallel endonucleases bioassays to distinguish the bioactive ingredients against both wild-type and mutant.

The endonuclease bioactive profile was constructed by using the slope values of linear growth region (5–15 min) in the measured growth kinetic curve of each well. The slope values and its robustness are highly related to the baseline of bioactive profile, which in turn affects the sensitivity of the whole system. In order to obtain a satisfactory slope value, the concentrations of PA_N and probe, and the buffer composition were systematically optimized for the FRET-based endonuclease bioassay [11,15,17].

As shown in Fig. 1A, the growth rate of the kinetic curve increased with decreasing the dilution ratio of PA_N. When the enzyme dilution ratio was 1:1.25, the growth rate was the highest, but the fluorescence response of the system gradually decreased after 20 min, which may be related to the possible re-accumulation of fluorescence quenching groups due to the increased concentration of cleaved oligonucleotide probe per unit of time and volume. When the enzyme dilution ratio was 1:5, no decreasing tendency of the fluorescence response was observed, and the growth rate was still high enough and stable within 20 min, which could be used to determine the inhibitory activity of compounds. Similarly, as can be seen in Fig. 1B, at the optimal enzyme concentration, the kinetic curve of the fluorescence response measured at a substrate concentration of 25 nM showed a good linear growth trend (good slope value) over 5–15 min. Therefore, an enzyme concentration of 1:5-fold dilution and a substrate concentration of 25 nM were finally chosen for further assessment of enzyme activity.

Fig. 1.

Growth kinetic curves of endonuclease activity assay measured at different concentrations of (A) the N-terminal domain of influenza viral polymerase (PA_N) and (B) probe.

In our previous work, we noticed that the buffer composition (i.e., pH, salt, and even the surfactants) affects enzyme stability and solubility of test compounds, further influencing bioassay performance [18,19]. To optimize the assay conditions, the effects of different surfactants, pH levels, and types of buffer systems on PA_N activity were carefully investigated. Two buffer systems (phosphate-buffered saline (PBS) and Tris-HCl) were first evaluated. Figs. S1A and B shows that PA_N exhibited significantly enhanced activity in Tris-HCl buffer than in PBS across a pH range of 6.5–7.5. This could be attributed to phosphate ions in PBS, which can bind to Mg²⁺, an essential cofactor for PA_N to exert its bioactivity. Furthermore, the enzymatic activity of PA_N showed a pH-dependent enhancement in Tris-HCl buffer, with the maximal reaction kinetics, as indicated by the slope value of 1500 achieved at pH 7.5. As for surfactants, Triton X-100 and NP-40 were separately added into the bioassay system. In Figs. S1C and D, compared with the blank control, the growth rate of the kinetic curve was significantly increased when proper concentration of Triton X-100 or NP-40 was added to the buffer system. Notably, a satisfactory slope value of 1500 with the lowest relative standard deviation (RSD) value of 1.33% was obtained, when 0.5% NP-40 was added in the buffer. Therefore, 50 mM Tris-HCl (pH 7.5) containing 0.5% NP-40 was chosen as the optimal buffer composition for the PA_N bioassay.

Z′ factor has been broadly used as a quality control metric in HTS assays [20]. In this work, baloxavir (500 μM), galantamine hydrobromide (500 μM), and DMSO were used as the positive control (n = 94), negative control (n = 94), and blank control (n = 24), respectively. Generally, 0.5 ≤ Z′ < 1 indicates that the signal distribution of positive samples is well separated from that of negative samples in the measurement system [21]. Therefore, the result in Fig. 2 shows that the optimized FRET-based PA_N bioassay method is suitable for HTS of PA_NIs.

Fig. 2.

Z′ factor of the fluorescence resonance energy transfer (FRET)-based the N-terminal domain of influenza viral polymerase (PA_N) bioassay.

3.1.2. Optimization of fractionation frequency and reaction volume

Theoretically, a higher fractionation frequency can usually enhance the resolution of bioactive profiles by increasing both the quantity of collected fractions and associated bioassay data points, while it also means less LC eluate in each collecting well, leading to a lower bioactive response for individual inhibitory peaks in bioactive profile. Therefore, four different fractionation frequencies (7, 9, 11, and 13 s) were investigated using a mixed standard solution containing a positive compound (baloxavir, PA_NI), and two negative compounds (galantamine hydrobromide, acetylcholinesterase inhibitor; favipiravir, polymerase basic protein 1 (PB1) inhibitor). As shown in Fig. 3A, no significant enhancement in peak area or peak height of bioactive profile was observed using different fractionation frequencies. It has been reported that a lower reaction volume can not only reduce the dilution effects and consumption of enzyme and substrate, but also improve the reaction rate and sensitivity of low-abundance enzyme assay [22]. Fig. 3B shows that a reaction volume of 20 μL led to a stronger inhibitory peak for baloxavir compared to 50 μL (the absolute height of inhibitory peak increases from 41% to 98%). Therefore, a fractionation frequency of 9 s/well and a reaction volume of 20 μL were selected for PA_NIs screening from natural products.

Fig. 3.

Optimization of (A) fractionation frequency (bioactivity profiles (top) and the corresponding high performance liquid chromatography (HPLC) chromatograms (bottom)) and (B) reaction volume ((bioactivity profiles (top) and the corresponding HPLC chromatogram (bottom)) of the at-line nanofractionation (ANF)-based the N-terminal domain of influenza viral polymerase (PA_N) inhibitors (PA_NIs) screening platform. Conditions: RD-C₁₈ column (4.6 mm i.d. × 250 mm, 3 μm); mobile phase A, H₂O:formic acid (FA) (99.9:0.1, v/v); mobile phase B, methanol (MeOH): FA (99.9:0.1, v/v); gradient program: 0 min, 60% B, 12 min, 100% B, and 15 min, 60% B; flow rate, 0.5 mL/min; ultraviolet (UV) wavelength, 295 nm.

3.2. Validation of the ANF-based PA_NIs screening platform

To validate the established screening platform, its sensitivity, repeatability, and versatility were checked using the above mixed standard solution.

As shown in Fig. 4A, an evident inhibitory peak for baloxavir was observed at 8.7 min, with consistent area and height across three injections. No inhibitory peaks were observed from 4 to 7 min, where the negative compounds were located, proving that this platform could stably and effectively distinguish potential PA_NIs from inactive compounds in complex samples. Moreover, the absolute peak height of the inhibitory peaks of baloxavir continually increased with the rising concentrations of mixed control solutions, reaching 38.43, 94.98, and 114.38 at baloxavir concentrations of 50, 100, and 250 μM, respectively (Fig. 4B), demonstrating a positive correlation. To evaluate the sensitivity of the platform, a series of baloxavir concentrations from 0.1 to 50 μM were injected and analyzed. As shown in Fig. S2, when the concentration of baloxavir was decreased to 10 μM, the inhibitory peaks of baloxavir still can be detected with the peak height three-fold higher than the signal-to-noise ratio (SNR), indicating that this platform has satisfactory sensitivity for recognizing PA_NIs.

Fig. 4.

Validation of the at-line nanofractionation (ANF)-based the N-terminal domain of influenza viral polymerase (PA_N) inhibitors (PA_NIs) screening platform with PA_N: (A) repeatability (bioactivity profiles (top) and the corresponding high performance liquid chromatography (HPLC) chromatograms (bottom)) and (B) sensitivity (bioactivity profiles (top) and the corresponding HPLC chromatogram (bottom)). Conditions: 65% B isocratic elution; other chromatographic conditions are the same as those in Fig. 3.

By replacing PA_N with PA_N I38T, its sensitivity and repeatability were also evaluated. Obviously, the PA_N I38T screening platform almost presented the same performance for inhibitor screening as PA_N, according to the results in Fig. S3, except for a slight decrease in the inhibitory peak height of baloxavir at the concentration of 50 μM compared to that in the PA_N screening platform, which can be explained by the lower susceptibility of baloxavir toward PA_N I38T. Those results also demonstrate good versatility of the current ANF-based PA_NIs screening platform, providing a potential method for parallel evaluation of different endonucleases bioassays. Therefore, the proposed screening platform has good repeatability and effective potential to screen PA_NIs from complex systems.

3.3. Screening of the aqueous extract of Artemisiae Argyi Folium

13 herbals with reported anti-influenza activity were collected according to literature, and then extracted by water. Their PA_N inhibitory rates at different concentrations were further measured by in-vitro PA_N bioassay and shown in Fig. 5A. The heatmap shows that the aqueous extract of Artemisiae Argyi Folium has the highest inhibitory activity, and the measured IC₅₀ value is 6.15 ± 0.49 μg/mL (Fig. 5B). Therefore, it was selected for the following screening of PA_NIs.

Fig. 5.

Evaluation of the inhibitory effect of several traditional Chinese medicines (TCMs) extracts to the N-terminal domain of influenza viral polymerase (PA_N). (A) Heatmap of the PA_N inhibition rate of the aqueous extracts of 13 herbals at different concentrations. (B) A dose-response curve of Artemisiae Argyi Folium on PA_N activity. IC₅₀: half maximal inhibitory concentration.

The developed ANF-based PA_NIs screening platform was then applied to identify PA_N/PA_N I38T inhibitors from the aqueous extract of Artemisiae Argyi Folium. A total of 22 inhibitory peaks were observed (Fig. S4). As shown in Fig. 6, by comparing the screening results of PA_N and PA_N I38T, 13 inhibitory peaks were found in both PA_N and PA_N I38T bioactive profiles, with particularly nine inhibitory peaks only present in PA_N I38T, indicating that these compounds have different inhibitory activities against PA_N and PA_N I38T. To identify the active compounds in Artemisiae Argyi Folium, the screening results were analyzed with the relevant literature summarizing the MS information of Artemisiae Argyi Folium components [[23], [24], [25], [26], [27]], in combination with chromatographic retention time, inhibitory peak position, and accurate MS data, collected by the AB SCIEX ZenoTOF 7600 Mass Spectrometer. According to LC-MS data in Table 1, the specific structures of 17 compounds were successfully identified. However, one compound was not identified due to its positional isomerism, and four compounds showed almost no corresponding matches in existing literature.

Fig. 6.

Screening results of the N-terminal domain of influenza viral polymerase (PA_N)/PA_N I38T inhibitors on the aqueous extract of Artemisiae Argyi Folium. Conditions: RD-C₁₈ column (4.6 mm i.d. × 250 mm, 3 μm); gradient program: 0 min, 10% B, 15 min, 30% B, 50 min, 43% B, 75 min, 60% B, 85 min, 70% B, 95 min, 100% B, and 105 min, 100% B. Other chromatographic conditions are the same as those in Fig. 3. Peaks 1−22 indicate candidate inhibitors identified in this study. Peaks marked by black dashed lines are common inhibitors of both PA_N and PA_N I38T, while those marked by orange dashed lines selectively inhibit PA_N I38T. TIC: total ion chromatogram; UV: ultraviolet.

Table 1.

Liquid chromatography-mass spectrometry (LC-MS) data of the potential the N-terminal domain of influenza viral polymerase (PA_N) inhibitors (PA_NIs) screened out from the aqueous extract of Artemisiae Argyi Folium.

No.	tR (min)	Ion mode	m/z	MS/MS	Chemical formula	Compound
1	22.526	[M−H]−	353.0858	191.0543 (100), 179.0340, 135.0440, and 134.0360	C16H18O9	Neochlorogenic acid
2	26.244	[M−H]−	137.0238	137.0230 (100), 136.0149, 108.0210, 109.0284, and 92.0256	C7H6O3	Protocatechualdehyde
3	28.349	[M−H]−	353.0859	163.0366 (100), 135.0418, and 89.0324	C16H18O9	Chlorogenic acid
4	29.724	[M−H]−	353.0860	135.0439 (100), 173.0438, 191.0543, 179.0337, and 93.0337	C16H18O9	Cryptochlorogenic acid
5	31.154	[M−H]−	177.0173	177.0181 (100), 78.9582, 133.0281, 105.0341, and 89.0394	C9H6O4	Esculetin
6	31.822	[M−H]−	515.1121	191.0541 (100), 179.0333, 353.0839, and 135.0436	C25H24O12	Cynarine
7	32.709	[M−H]−	179.0340	135.0446 (100), 134.0361, and 89.0388	C9H8O4	Caffeic acid
8	41.993	[M−H]−	563.1357	563.1354 (100), 353.0630, 383.0728, 443.0951, 473.1029, and 503.1146	C26H28O14	Schaftoside isomer
9	43.703	[M−H]−	563.1334	563.1333 (100), 353.0628, 383.0721, 443.0932, 473.1054, and 297.0729	C26H28O14	Schaftoside
10	45.109	[M−H]−	463.0842	151.0021 (100), 135.0433, 287.0530, 463.0823, 175.0226, and 85.0287	Unknown	Unknown
11	48.455	[M+H]+	565.1523	379.0774 (100), 349.0683, 529.1317, 325.0666, 295.0567, 397.0900, and 547.1407	Unknown	Unknown
12	49.910	[M−H]−	563.1337	563.1351 (100), 353.0628, 383.0722, 473.1024, and 443.0928	Unknown	Unknown
13	50.961	[M−H]−	515.1144	173.0440 (100), 179.0337, 353.0843, 191.0541, 135.0438, and 515.1141	C25H24O12	Isochlorogenic acid B
14	51.509	[M−H]−	563.1332	563.1332 (100), 353.0619, 443.0928, 383.0739, 473.1042, and 297.0735	C26H28O14	Isoschaftoside
15	52.114	[M−H]−	515.1154	191.0541 (100), 179.0330, 353.0843, and 135.0437	C25H24O12	Isochlorogenic acid A
16	53.709	[M−H]−	431.0937	311.0518 (100), 283.0575, 341.0637, and 431.0933	C21H20O10	Isovitexin
17	56.388	[M−H]−	463.0834	300.0249 (100), 463.0821, 301.0571, and 136.9859	C21H20O12	Hyperoside
18	57.850	[M−H]−	463.0835	300.0245 (100), 271.0209, 463.0832, 301.0324, 255.0278, and 243.0276	C21H20O12	Isoquercetin
19	62.253	[M−H]−	515.1152	173.0443 (100), 179.0335, 353.0843, 191.0545, 135.0441, and 515.1148	C25H24O12	Isochlorogenic acid C
20	67.049	[M−H]−	549.1926	387.1621 (100), 161.0231, 549.1907, 59.0137, 207.1002, 133.0279, and 89.0235	C27H34O12	Tracheloside
21	76.488	[M−H]−	677.1430	173.0437 (100), 179.0335, 353.0841, 515.1146, 191.0541, and 135.0440	C34H30O15	3,4,5-Tricaffeoy-lquinic acid
22	80.013	[M+H]+	509.8837	114.0900 (100), 209.1630, 500.8767, 228.1575, 322.2463, and 341.2402	Unknown	Unknown

t_R: retention time.

As shown in Fig. 7, the identified potential PA_NIs from Artemisiae Argyi Folium include eight caffeoylquinic acids (CQAs) (compounds 1, 3, 4, 6, 13, 15, 19, and 21), five flavonoids (compounds 9, 14, 16, 17, and 18), one phenolic acid (compound 7), one coumarin (compound 5), one lignan glycoside (compound 20), and one polyphenol (compound 2). Coincidentally, it has been reported that CQAs have anti-influenza virus activity at low micromolar concentrations [28]. Their ability to inhibit the invasion and replication of the influenza virus has also been demonstrated in an earlier study [29]. This suggests that CQAs have the potential to produce anti-influenza virus effects by targeting PA_N. Furthermore, it was shown that flavonoids are good influenza PA_NIs, with their high affinity and inhibitory potency attributed to surface complementarity and strong interaction between hydroxyls on the B-ring and metal ions [8]. Particularly, flavonoids with hydroxyl substitutions at positions 5 and 7 on the A-ring and positions 3′ or 4’ on the B-ring in the basic skeleton tend to exhibit excellent anti-influenza virus activity [30]. This structural characteristic is consistent with the five flavonoids (schaftoside, isoschaftoside, hyperoside, isoquercetin, and isovitexin) screened in this study. Additionally, the catechol structure in other compounds may function as a metal-binding pharmacophore affecting PA_N activity.

Fig. 7.

Structures of the identified the N-terminal domain of influenza viral polymerase (PA_N) inhibitors (PA_NIs) from the aqueous extract of Artemisiae Argyi Folium.

To verify the accuracy of the above results, seven commercially available compounds were tested for their in-vitro endonuclease activity against both PA_N and PA_N I38T. According to the inhibitory activity information in Table 2, most of the screened CQAs exhibited similarly low IC₅₀ values against both endonucleases, particularly cynarine, isochlorogenic acids A, isochlorogenic acids B, and isochlorogenic acids C. Among them, cynarine, isochlorogenic acids B, and isochlorogenic acids C showed higher inhibitory activity against PA_N I38T compared to baloxavir. However, chlorogenic acid and caffeic acid exhibited low inhibitory activity against endonucleases. These compounds can be considered as false-positive due to their high abundance in the extract, which could lead to prominent inhibitory peaks in the bioactive profile. In summary, the established screening platform demonstrates significant potential not only for identifying PA_NIs from complex herbal extracts, but also for differentiating the inhibitory effects of each compound on different endonucleases (PA_N and PA_N I38T).

Table 2.

Half maximal inhibitory concentration (IC₅₀) values of screened the potential the N-terminal domain of influenza viral polymerase (PA_N) inhibitors (PA_NIs) on the endonuclease bioassay (n = 3).

Peak No.	Compound	IC50 (PAN) (μM)	IC50 (PAN I38T) (μM)
3	Chlorogenic acid	>100	>100
5	Esculetin	59.64 ± 3.36	61.36 ± 2.43
6	Cynarine	23.69 ± 18.70	25.91 ± 15.39
7	Caffeic acid	>100	>100
13	Isochlorogenic acid B	29.32 ± 1.65	26.51 ± 1.88
15	Isochlorogenic acid A	56.77 ± 18.47	61.81 ± 33.76
19	Isochlorogenic acid C	22.85 ± 2.39	24.44 ± 1.62
Positive control	Baloxavir	9.11 ± 0.41	12.88 ± 0.23

3.4. Molecular docking analysis

To understand the molecular interactions between the screened compounds with PA endonuclease, molecular docking analysis was performed. The results revealed that esculetin (compound 5) could bind to the active pocket of PA endonuclease by chelating with two Mn²⁺ ions, forming hydrogen bonds with Ala37, Val122, and Lys134, and interacting with His41 and Asp108 through π−cation and π−anion interaction (Fig. S5A). When binding to PA_N I38T (Fig. S5B), esculetin could also produce multiple hydrogen bond interactions with protein, resulting in a similar binding energy (PA_N: −5.945 kJ/mol; PA_N I38T: −5.798 kJ/mol).

Compared with esculetin, cynarine (compound 6) can effectively occupy more active sites within the target protein due to its larger molecular structure (Fig. S6). It forms multiple hydrogen bonds (e.g., Glu80, Val122, Arg124, and Tyr 130) and π−interactions with key amino acid residues in both PA_N and PA_N I38T. Although it forms a coordination bond with only one Mn²⁺ ion, cynarine still shows a low binding energy (PA_N: −7.896 kJ/mol; PA_N I38T: −7.833 kJ/mol).

Isochlorogenic acid B (compound 13), isochlorogenic acid A (compound 15), and isochlorogenic acid C (compound 19) could rely on strong hydrogen bonds with Tyr24, Ala37, Glu80, His41, Val122, and Arg124, as well as various π−interactions with Lys34, Ala37, and His41 to form stable bonds (Figs. S7−S9). Among them, isochlorogenic acid C was able to fit two binding pockets with a better binding posture and from hydrophobic contact with Lys34 and Ala37. As a result, it demonstrates a low binding energy comparable to that of baloxavir (PA_N: −8.985 kJ/mol; PA_N I38T: −8.727 kJ/mol), indicating its potential as a PA endonuclease inhibitor.

As shown in Table S1, the binding affinities of the tested compounds with PA_N endonuclease were closely associated with their structural frameworks and interaction patterns. Isochlorogenic acid C exhibited the strongest affinity due to its extended conformation, featuring a tricyclic scaffold and multiple ester linkages that enable deep insertion into the catalytic pockets. This compound formed extensive hydrogen bonds with residues, i.e., Glu80, Tyr24, Ser194, and Glu195, and engaged in π–π stacking with His41. Isochlorogenic acids A and B, as well as cynarine, also possessed dual aromatic rings and flexible ester bridges, allowing the formation of hydrogen bonds and π–cation or π–π interactions, though minor steric hindrance near Tyr130 slightly reduced their binding efficiency. Esculetin, despite having several hydroxyl groups, showed limited engagement with deep pocket due to its compact structure, although moderate π and hydrogen bond interactions with His41 and Glu80 were observed. In docking studies with PA_N I38T endonuclease, similar structure-activity trends were observed. Isochlorogenic acids C and A again displayed superior binding, forming strong hydrogen bonding networks with Arg124, Glu80, Lys34, and Tyr24, while maintaining π–π stacking with His41. These ligands demonstrated excellent spatial complementarity and avoided steric hindrance with residues like Val122 and Ala37. Cynarine showed moderate adaptability through its flexible scaffold but exhibited affinity due to fewer interactive moieties. Overall, compounds with extended, flexible scaffolds, rich polar groups and dual aromatic systems, such as isochlorogenic acid C and A, exhibited superior binding to both endonucleases by establishing multiple hydrogen bonds and aromatic interactions with key residues of endonuclease. Their ability to adapt conformationally and form multi-point interactions highlights the importance of structural adaptability and interaction density in achieving potent enzyme inhibition.

4. Conclusions

Herein, a highly efficient screening platform was successfully established for the rapid identification of influenza virus PA_N and PA_N I38T inhibitors from complex plant extracts. After systematical optimization and validation, this platform was applied to the screening of PA_N and PA_N I38T inhibitors from the aqueous extract of Artemisiae Argyi Folium, with 17 potential active compounds successfully identified. Among them, five compounds demonstrated high inhibitory activity against both PA_N and PA_N I38T, exhibiting comparable efficacy to baloxavir and thereby emerging as promising candidates for the therapeutics of drug-resistant influenza. In addition, molecular docking analysis confirmed that these five compounds formed strong interactions with the central active site of PA endonuclease and the four hydrophobic pockets to varying degrees, suggesting their potential as lead structures for further systematic structural modification. Given the continuous evolution of influenza viruses and the increasing challenge of antiviral resistance, the proposed platform provides a rapid, high-throughput, and sensitive strategy for PA_NIs screening from complex natural matrices. This strategy offers a powerful tool for the accelerated discovery of broad-spectrum anti-influenza drugs, paving the way for future antiviral drug development.

CRediT authorship contribution statement

Yuexiang Chang: Writing – original draft, Validation, Methodology, Data curation. Hao Tian: Software, Methodology, Investigation. Jia-Huan Qu: Supervision, Investigation. Jiaming Yuan: Software, Formal analysis. Rongkai Gu: Software, Formal analysis. Tingting Zhang: Writing – review & editing, Supervision, Methodology. Jincai Wang: Writing – review & editing, Supervision, Methodology, Funding acquisition. Zhengjin Jiang: Writing – review & editing, Supervision, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. As an editorial board member, Zhengjin Jiang recused himself from all review processes related to this article to ensure the fairness and objectivity of the review.

Appendix A. Supplementary data

The following is the Supplementary data to this article: