Chemical compounds of QF were obtained from the databases of traditional Chinese medicine systems pharmacology (TCMSP, http://tcmspw.com/tcmsp.php) and bioinformatics analysis tool for molecular mechanism of traditional Chinese medicine (BATMAN-TCM, http://bionet.ncpsb.org/batman-tcm) (Ru et al., 2014; Liu et al., 2016). Active compounds were screened with oral bioavailability (OB)≥30% and drug-likeness (DL)≥0.18. Protein targets of the active components in QF were converted to official symbols by checking with reviewed Homo sapiens gene list in the Protein Knowledgebase (UniProtKB, http://www.uniprot.org/) (Tao et al., 2020). The disease-related targets were collected by inputting keywords “respiratory syncytial virus pneumonia,” “RSV pneumonia” and “RSV-induced lung inflammation” into online platforms including GeneCards (http://www. genecards. org/) and online Mendelian inheritance in man (OMIM; http://www.omim.org/) (Stelzer et al., 2016; Amberger and Hamosh, 2017). The common target genes were selected out as QF targets in the treatment of RSV-LI. The core targets were further analyzed by STRING (https://string-db.org/cgi/input.pl) to construct Protein–Protein Interactions (PPI) network (Szklarczyk et al., 2017). The enrichment analysis in the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) were performed to demonstrate the main biological functions and signal transductions involved by core targets. The top 20 KEGG pathways (p < 0.05) were displayed in a histogram using the MetaScape Bioinformatics Resource database (http://metascape.org/) (Zhou et al., 2019). GO enrichment was conducted by ClueGO, a plug-in of Cytoscape (v3.8.2) software (http://www.cytoscape.org/) to visualize biological process (BP), cellular component (CC) and molecular function (MF) enrichments of core targets.

QF was made up of the following ingredients: Ephedra sinica Stapf. (batch No. 190807, 5g), Prunus armeniaca L. (batch No.191321, 12g), Gypsum Fibrosum (batch No. 191201, 40g), Morus alba L. (batch No. 190407, 12g), Lepidium virginicum L. (batch No. 190312, 10g), Angelica decursiva (batch No. 190129, 12g), Bistorta officinalis Delarbre (batch No. 191227, 15g), Reynoutria japonica Houtt. (batch No. 191115, 15g), Bombyx mori Linnaeus. (batch No. 191104, 8g) and Salvia miltiorrhiza Bunge (batch No. 191226, 8g). The ten products were purchased from Tongling Hetian Chinese Medicine herbal tablets Co., Ltd. Anhui, China. QF was prepared in accordance with standard operating procedures (Wang et al., 2003). In brief, the ten drugs were mixed and decocted twice in filtered water for 1 hour at 100°C, and the filtrate was concentrated to a final concentration of 3.65 g/ml using a rotary evaporator (YRE-2000A, China). The concentrating agent was kept at −20°C for future use. The fingerprinting analysis of QF (30 ug/ml) was performed in both negative and positive ionization modes using a Linear Ion Trap Quadrupole-Orbitrap Mass Spectrometer (LTQ-Obitrap MS; Thermo Fisher Scientific, USA).

Hep2 cells (human laryngeal carcinoma cells; China Center for Type Culture Collection, Wuhan, China) were cultured with DMEM medium (10% FBS, 100 U/ml penicillin, 0.1 mg/ml streptomycin) to provide condition for RSV strain A2 (China Center for Type Culture Collection, Wuhan, China). The viral supernatant was collected after 4–5 days and a plaque assay was performed to ensure that the virulence reached 1*10^6 PFU/ml. The viral solution was stored in liquid nitrogen after being centrifuged for 10 min at 3,000 rpm at 4°C. The animal experiment in this study was approved by the ethics committee of the Laboratory Animal Center at Nanjing University of Chinese Medicine (ethical # 201912A009). The in vitro and in vivo experiments were conducted out in Yangzhou University’s biosafety level-2 laboratory. A total of 18 female BALB/c mice (aged 6–8 weeks, 18 ± 2 g) were purchased from the Laboratory Animal Department of Shanghai Family Planning Research Institute (permission number: 202116007) and randomly assigned to three groups: a control group (C), an RSV group (M) and a QF group (n = 6 per group). The RSV and QF groups received 80 μl RSV suspension intranasally under inhalation anesthesia on the first day of modeling, whereas the control group received equal DMEM medium. QF (27.6 g/kg/day) was delivered intragastrically for four consecutive days, equivalent to the clinical human dosage as assessed by host weight (Zhu et al., 2021). On the fifth day after intervention, mice were sacrificed. The body weight, energy level, diet, drinking water, fur color, and activity of the mice in each group were recorded daily.

After the mice were euthanized, middle lobe of right lung was fixed in 4% neutral buffered paraformaldehyde for 12 h dehydration before being embedded in paraffin wax and stained with hematoxylin and eosin (H&E). Tissue lesions and inflammatory cell infiltrates were measured and captured under a light microscope at high magnification (200×) (OLYMPUS, Japan). Lung tissue was divided into 20 mg aliquots for lipidomic analysis and the remaining was stored at −80°C for the biochemical indexes including PI3K, p-PI3K, AKT, p-AKT, mTOR, p-mTOR, Lipin-1, Beclin-1, Atg5, LC3B, VPS34, IL-1β, IL-6, and TNF-α.

ABF converter (http://www.reifycs.com/AbfConverter) was used to convert raw data from the Xcalibur 2.2 program (Thermo Fisher Scientific, USA) to ABF formats. The files were then analyzed using the MS-DIAL (v4.10) software tool. The parameters of MS-DIAL were described in the previous study (Yang et al., 2014). For lipid identification, the public LipidBlast library was used with accurate mass and MS/MS matching. Adduct ions were set as + H, +NH4 and +Na for the positive ion mode. MetaboAnalyst 5.0 (http://www.metaboanalyst.ca) was used to further normalize multidimensional data following median and pareto scaling. The difference in lipidomic profiles between the control, model, and QF groups was represented in the principal component analysis (PCA) model. False discovery rate (FDR <0.05, Kruskal-Wallis test) and fold change (FC > 1.2 or <0.833, ratio of median) were used to screen for significant lipids. The number of screened lipids in each of the two groups was represented by a Venn diagram. Furthermore, heatmaps revealed the pattern of key lipids that were expressed in opposite directions between the two groups. In the boxplots, the FDR values were used to compare the relative concentrations of diacylglycerol (DAG).

Protein was extracted from lung tissues on ice for 30 min using RIPA lysis buffer with proteinase and phosphatase inhibitors (100:1:1). The tissue lysate was then centrifuged for 30 min at 13,000 rpm and 4°C. Using bovine serum albumin as a reference, protein content was determined using a BCA Protein Assay Kit (Thermo Fisher Scientific, USA). 30 μg total proteins of lung tissue lysate were run on 10% tris-glycine SDS-PAGE gel (Biosharp), 80 V for 0.5 h and 110 V for 1 h. Gels were transferred to a polyvinylidene difluoride (PVDF) membrane by Trans-Blot (Bio-Rad, USA), and blocked with 5% BSA for 1 h. The bands were incubated overnight at 4°C with primary antibodies: PI3K (1:1,000, CST, USA), p-PI3K (1:1,000, CST, USA), AKT (1:1,000, CST, USA), p-AKT (1:2000, CST, USA), mTOR (1:1,000, CST, USA), p-mTOR (1:1,000, CST, USA), Beclin-1 (1:1,000, CST, USA), Atg5 (1:1,000, CST, USA), LC3B (1:1,000, CST, USA), VPS34 (1:500, Proteintech, China), β-actin (1:20000, Proteintech, China). Proteins on the band were observed using a ChemiDoc XRS Gel Imaging System (Bio-Rad, USA) after 1 h at room temperature incubation with the appropriate secondary antibodies (Fcmacs, China). ImageJ software was used to examine the protein expression of each group (v1.8.0).

Lung tissue paraffin slides were deparaffinized in xylene, rehydrated using an ethanol gradient, antigen-retrieved in a 10 mM sodium citrate buffer (pH 6), blocked with 3% H2O2, and permeabilized with 1% BSA. The sections were then blocked with Lipin-1 antibody (1:500, Abcam, USA) overnight at 4°C. Afterwards, the secondary antibody (1:250, MaxVision, China) was added and incubated at room temperature for 20 min. After DAB staining, the slices were stained with hematoxylin and eosin for 10 min, dehydrated, and sealed in neutral resin. Lipin-1 protein expression levels were observed in different groups using a light microscope at high magnification (200×).

GraphPad Prism software (v8.0.2) was used to analyze and visualize all of the biochemical data. The standard deviation is shown by standard error of the mean (SEM). One-way analysis of variance (ANOVA) with the Student-Newman-Keuls post hoc test was used for multi-group comparisons, with p < 0.05 considered significant.

Detailed information on formula composition of QF was shown in Table 1. The chemical profile of QF was originally investigated using UPLC-ESI/LTQ-Orbitrap-MS. Based on the reference compounds, a total of six components from 10 crude materials were identified qualitatively. The typical total ion chromatograms (TICs) are shown in Supplementary Figure S1. QF’s primary bioactive components, including Catechin (C15H14O6), Luteolin (C15H10O6), Quercetin (C15H10O7), Naringenin (C15H12O5), Kaempferol (C15H10O6), and Tanshinone IIA (C19H18O3), were processed under quality control (Supplementary Table S1). The detail parameters of the quality control were listed in supplementary file.

To develop a BALB/c mouse model of RSV-LI, 80 μl RSV suspension (1*10⁶ PFU/ml) intranasally under inhalation anesthesia on the first day of modeling (Figure 2A). Body weight (BW) changes revealed that RSV infection led to a significant weight loss within 24 h of nasal drip, with the mean value reaching the lowest level on the third day of modeling compared to controls (*p<0.01). Such weight loss, however, was successfully avoided in mice given QF therapy (*p>0.05 versus control, Figure 2B). The expression of RSV F and G proteins in lung tissues represented viral load. These two indicators’ transcription levels were evidently increased in the model group (*p<0.0001, versus control, Figure 2C), but they were significantly reduced in the QF group (^# p<0.0001, versus model; Figure 2C). Figure 2D showed HE staining of pulmonary pathologies in the three groups. In RSV-LI mice model, there was clear evidence of alveolar wall damage and inflammatory cell infiltration surrounding the bronchi, which was significantly alleviated after QF treatment.

A total of 148 active compounds were identified in QF after integrating information from databases in TCMSP and BATMAN-TCM Specifically, 23 of them were found in Ephedra sinica Stapf (Ma Huang), 19 in Prunus armeniaca L. (Xingren), 12 in Lepidium virginicum L. (Ting Lizi), 27 in Morus alba L. (Sang Baipi), 8 in Angelica decursiva (Zihua Qianhu), 5 in Bistorta officinalis Delarbre (Quan Shen), 10 in Reynoutria japonica Houtt. (Hu Zhang), 1 in Bombyx mori Linnaeus. (Jiang Can), and 69 in Salvia miltiorrhiza Bunge (Dan Shen). Quercetin (MOL000098), Luteolin (MOL000006), Kaempferol (MOL000422), Naringenin (MOL004328), Tanshinone IIA (MOL007154), Beta-sitosterol (MOL000358), Cryptotanshinone (MOL007088), Ellagic acid (MOL001002), Isorhamnetin (MOL000354) and Dihydrotanshinlactone (MOL007100) were the top ten active chemicals for QF in the treatment of RSV-LI (Table 2). By referring to target mappings in TCMSP and translating protein names to official symbols on the UniProt database, 275 compound-related targets were identified. Furthermore, 859 disease-related targets were discovered in the GeneCards and OMIM databases. Finally, 101 symbols were identified as important targets for QF against RSV-LI by evaluating the shared parts in the two segments (Table 3). The network in Figure 3 constructed with 101 main targets and their matching 117 active molecules demonstrated the unique therapeutic benefits of QF working on RSV-LI. Based on interactions, active components primarily worked on targets such as Prostaglandin G/H synthase 2 (PGST2), Beta-2 adrenergic receptor (ADRB2), Muscarinic acetylcholine receptor M1 (CHRM1), Androgen receptor (AR), Nitric oxide synthase (NOS2), Peroxisome proliferator-activated receptor gamma (PPARG), and Phosphatidylinositol 4,5-bisphosphate 3-kinase catalytic subunit gamma isoform (PIK3CG).

The PPI network illustrated the relationship between core targets (Figure 4A). In the network, proteins including RAC-alpha serine/threonine-protein kinase (AKT1), Cellular tumor antigen p53 (TP53), Mitogen-activated protein kinase 3 (MAPK3), Vascular endothelial growth factor A (VEGFA), Tumor necrosis factor (TNF), Interleukin-6 (IL6), Transcription factor AP-1 (JUN), Caspase-3 (CASP3), Mitogen-activated protein kinase 8 (MAPK8), Signal transducer and activator of transcription 3 (STAT3), Prostaglandin G/H synthase 2 (PTGS2), Mitogen-activated protein kinase 1 (MAPK1), Pro-epidermal growth factor (EGF), Epidermal growth factor receptor (EGFR) and Myc proto-oncogene protein (MYC) were identified as hub nodes with a higher degree value.

The signaling pathways were shown using KEGG and GO analysis to better study the biofunctions of key targets involved in the QF against RSV-LI process. The top 20 KEGG pathways, as shown in Figure 4B, were mostly related to immunological function, inflammatory response, and cell stress. Because of the lowest ^a p value, the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT) signaling pathway was rated top. The downstream of PI3K/AKT signal transduction including mammalian target of rapamycin (mTOR) and forkhead box (FoxO) were also shown in the list. Furthermore, ClueGO software was used to cluster fifteen significant GO terms in order to visualize the biological process (BP), cellular component (CC), and molecular function (MF) of core targets. With enrichment ratios greater than 10%, the apoptotic process, protein serine/threonine kinase activity, lipid response, and cellar response to chemical stress were found to be the most significant (Figure 4C).

Untargeted lipidomics were applied based on our previous study that revealed decreased levels of lipids including triglyceride (TG) and glycerophosphates in RSV-infected mice (Shan et al., 2018), as well as the GO enrichment results of network pharmacology (Figure 3C) that connected to a lipid response. The principal component analysis (PCA) model in Figure 5A demonstrated full separations of lipidomic profiles across the control (C), model (M), and treatment (QF) groups. There were 281 differential lipids detected in the M vs. C group, 158 in the QF vs. M group, and 134 in both comparisons (Figure 5B). The heatmaps in Figures 5C,D were established using 97 lipids that were expressed differently in the M vs. C and QF vs. M groups. Diacylglycerol (DAG), N-acylethanolamine (NAE), and ceramide (Cer) were higher expressed in lung tissues of RSV-infected mice (versus control, FDR<0.05, FC>1.2), whereas triglyceride (TG) and glycerophospholipids including phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylinositol (PI), phosphatidylglycerol (PG), phosphatidylserine (PS) were lower expressed. (versus control, FDR<0.05, FC<0.833). After QF therapy, all of these lipid metabolites recover to normal levels at various degrees.

In this investigation, however, there was no discernible change in the relative amounts of lysophospholipids and oxidized phospholipids (Figures 5C,D). Supplementary Table S3 provided detailed information of lipids in heatmaps. Above all, we found that QF controlled the lipidomic abnormalities produced by RSV infection in the lungs of mice, as evidenced by decreased DAG, Cer, NAE, and increased production of TG, PC, PE, PG, PS, and PI.

DAG was reported as the second messenger triggering signaling cascades in the previous study (Sim et al., 2020). We discovered that relative lung concentrations of DAG 30:0, DAG 30:5, DAG 32:0, DAG 16:0_18:0, DAG 17:0_17:0, DAG 34:1, DAG 36:0, DAG 36:1 and DAG 40:2 were elevated in model group, whereas QF suppressed lung levels of the DAGs in mice with RSV-induced LI (Figure 5E). As a result, we hypothesized that QF might have important therapeutic effects through modulating DAG synthesis. Lipin-1, a key enzyme in DAG production, was shown to be elevated in the lung tissue of mice with RSV-LI, and it was reduced following QF treatment (Figures 6B,C). In addition, lipin-1 serves as the downstream of the PI3K/AKT/mTOR signaling pathway (Peterson et al., 2011). The increased levels of phosphorylated versions of the proteins PI3K, AKT, and mTOR in the lung tissues of mice in the QF group were all reduced in our investigation (versus model, ^## p<0.01; Figure 6A). In RSV-LI mouse models, the data showed that QF could inhibit DAG synthesis via modulating the PI3K/AKT/mTOR signaling pathway and its related protein levels of lipin-1.

Previous study discovered that Lipin-1-mediated DAG synthesis might activate excessive autophagy via stimulating type III phosphatidylinositol kinase (VPS34) signaling (Zhang et al., 2014). In our study, VPS34 was triggered by RSV-LI and suppressed after QF treatment (versus control, *p<0.01; versus model, ^# p<0.05, Figure 7B). The top-ranked list of KEGG analysis in Figure 3B included autophagy, a highly conserved cellular recycling mechanism that is intimately connected to viral replication and immune response during RSV infection (Li et al., 2018; Oh et al., 2020). To see if such a pathogenic process was activated by RSV-LI, we detected the expression of autophagy-related proteins including Beclin-1, Atg5, and LC3B. Figure 7A showed that protein levels of Beclin-1, Atg5, and LC3B II were all increased in the model group (versus control, *p<0.05, **p<0.01, ***p<0.001), but significantly decreased in the QF group (versus model, ^# p<0.05, ^## p<0.01). Considering the close relationship between excessive autophagy and inflammation, our results demonstrated that QF effectively reduced high levels of proinflammatory cytokines, such as IL-1β, TNF-α and IL-6(versus control, *p<0.05, **p<0.01, ***p<0.001; versus model, ^# p<0.05, ^### p<0.001, Figures 7C–E). As a consequence, we conclude that QF can suppress DAG-induced excessive autophagy and inflammation in lung tissues of RSV-LI.