The experiments were conducted in the Microorganism and Biochemical Pharmacy Innovation Lab at College of Pharmacy, Changchun University of Chinese Medicine, during March to October, 2021.

Dexamethasone (Dex) and ovalbumin (OVA) were obtained from Suicheng Pharmaceutical Co., Ltd. (Zhengzhou, China) and Shanghai Yuanye Bio-Technology Co., Ltd. (Shanghai, China), respectively. Chelidonium majus samples were provided by Hebei Renxin Pharmaceutical Co., Ltd. (Hebei, China). The botanical identity of the samples was confirmed by Prof. Shumin Wang of Changchun University of Chinese Medicine. The honey-processing procedure for Chelidonium majus was conducted following the general rules of the Chinese Pharmacopeia 2020 Edition [43]. Refined honey was subjected to dilution using boiling water at a boiling water/refined honey ratio of 1:1.9 (w/w).

After stirring of obtained mixed sample (20.7:100, diluted honey/raw C. majus, w/w) till total moistening, the sample was put in a preheated pan, followed by mild fire frying to dryness. Before being stored in a desiccator, the honey-processed C. majus (HC) was cooled in the plate. The HC was ground into a powder and sieved through a 60 mesh filter. The precisely weighed powder (400 g) was subjected to reflux extraction with 80% methanol for two 1 hr cycles. After conflating, the filter liquor was concentrated under reduced pressure, freeze-dried into powder and finally storage at –80 °C.

The HC extract was dissolved in a solution of methanol containing 2-amino-3-(2-chloro-phenyl)-propionic acid (4 ppm) and was then passed through a 0.22 µm membrane. Composition identification was performed on a Vanquish UHPLC System (Thermo Fisher Scientific, Waltham, MA, USA) coupled with a Q-Exactive Focus mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) equipped with an electrospray ionization (ESI) source [44]. The detailed instrument parameters for UPLC-Q-Exactive MS analysis are provided in Supplementary Table 1.

Composition was initially confirmed based on accurate molecular weight (MW error $\le$30 ppm) and subsequently obtained by annotating mzClound, PubChem, LipidMaps and Human Metabolome Database (HMDB) based on MS/MS fragmentation patterns [45, 46].

Specific pathogen-free (SPF) male Sprague-Dawley (SD) rats (220 $\pm$ 20 g) were obtained from Changchun Yisi Laboratory Animal Technology Co., Ltd. (Changchun, China). The rats were housed in plastic cages under 22 $\pm$ 2 °C, 50 $\pm$ 5% humidity and 12/12 hrs light/dark cycle conditions. All animal experimentation procedures adhered to the guidelines and were approved by the Animal Ethics Committee of Changchun University of Chinese Medicine (protocol No.: 2020345; approval date: October 11, 2020).

Forty-eight rats were randomized as 6 groups (n = 8): Control, Model (asthma), Dexamethasone (Dex, positive control) and low, medium and high-dose HC (HCL, HCM and HCH) groups. An asthma model was established after 7 days of routine feeding. The sensitizing solution was prepared by dissolving 100 mg OVA and 100 mg aluminum hydroxide in 1 mL of normal saline. Rats in each group (except control) received intraperitoneal injections of sensitization solution on days 1 and 8. Control rats were given injections of 1 mL of normal saline in an identical way. Following the last sensitization (day 15), each group was continuously gavaged for 7 days. The Control and Model groups received normal saline, the Dex group received dexamethasone (0.5 mg/kg) and the HC groups received their respective doses (0.81, 1.21 or 1.62 mg/kg) according to the regulations of Chinese Pharmacopoeia 2020 Edition [43]. The gavage volume was 10 mL/kg/day. Thirty minutes after each administration, the rats were challenged with 1% (w/v) OVA solution in a plexiglass chamber for 30 min once daily. Likewise, normal saline was given to control rats.

Following the final atomization inhalation, rats were fasted for 12 hrs. The next day, they were deeply anesthetized under intraperitoneal sodic pentobarbital anesthesia (100 mg/kg). After confirming the absence of a pedal reflex, the rats were euthanized via exsanguination or thoracotomy. A portion of lung tissue was collected for pathological analysis. After fixation with 4% paraformaldehyde, the lung tissues were dehydrated in ethylic alcohol, embedded in paraffin and sliced (4 µm). Hematoxylin and Eosin (H&E) staining was performed with sections for assessing inflammation. Periodic acid-Schiff (PAS) staining was performed to evaluate goblet cell hyperplasia. A scoring system (0–4 scale) was adopted to evaluate the severity of peribronchial and perivascular inflammation, while goblet cell hyperplasia was graded (0–4 scale) based on the number of PAS-positive cells in each airway [47, 48]. Masson staining was employed for assessing collagen fiber deposition in airways, which stained blue. ImageJ software version 1.54 (NIH, Bethesda, MD, USA) was utilized to quantify the percentage of collagen fiber area [49]. A double-blind approach was employed for the evaluation of inflammation.

Serum samples were collected and the supernatant was subjected to centrifugation (3000 rpm, 10 min, 4 °C). The concentrations of inflammatory mediators (IL-10, IL-13 and IL-1$\beta$) were quantified according to the instructions of ELISA kits. Commercial kits for measuring rat IL-10 (MM-0176M2), IL-13 (MM-089802) and IL-1$\beta$ (MM-0040M2) were offered by Jiangsu Meimian Industrial Co., Ltd., Yancheng, China.

Serum samples from the Control, Model and HCH groups were subjected to metabolomic analysis utilizing a Waters Acquity UPLC I-Class PLUS (Waters, Medford, MA, USA) with a Waters Xevo G2-XS QTOF mass spectrometer (Waters, Medford, MA, USA) (Supplementary Table 2) [50].

Raw data were processed with Progenesis QI software version 3.0 (Nonlinear Dynamics, Wilmslow, UK) for peak extraction, alignment, or additional data implementation, with mass deviation and theoretical fragment identification within 100 ppm.

Identified metabolites were subsequently searched in established databases (KEGG, LipidMaps and HMDB) to retrieve pathway and classification information. Differential multiples between groups were calculated and statistically evaluated utilizing Student’s t-tests. To visualize overall separation between groups, Principal Component Analysis (PCA) was employed. Orthogonal projection to latent structures-discriminant analysis (OPLS-DA) modeling was completed using R language package ropls. The model reliability was assessed through 200-fold permutations and model variable importance projection (VIP) was determined through multiple cross-validation. Differentially accumulated metabolites (DAMs) were identified (fold change, FC $>$1, VIP $>$1, p 0.05). Subsequently, pathway enrichment analysis of identified metabolic biomarkers (impact $>$0, –logP $>$1) was utilized using MetaboAnalyst 5.0. The significance of KEGG pathway enrichment was determined using a hypergeometric test.

This work collected lung tissues from Control (n = 3), Model (n = 3) and HCH groups (n = 3) for RNA sequencing. The quality and quantity of the isolated RNA were analyzed with a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). The RNA integrity was evaluated utilizing an Agilent Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA, USA) with an RNA Nano 6000 Assay Kit. The cDNA libraries were constructed and their quality was verified using an Agilent Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA, USA). Raw reads were further processed using BMKCloud platform [51].

Differential gene expression analysis was performed using the DESeq2 R package with the criteria of p 0.05 and $|$log2(FC)$|$ $>$1. Gene functions were annotated using the KEGG Ortholog (KO) database. A protein-protein interaction (PPI) network was constructed for differentially expressed genes (DEGs) commonly identified in the Model-vs-Control and HCH-vs-Model comparisons based on the STRING database (interaction score $>$0.7), followed by importing in Cytoscape 3.9.1 software (Cytoscape Consortium, San Diego, CA, USA) for network visualization. The CytoNCA plugin was employed to count node topological parameters within the PPI network for identifying hub genes.

To validate the mRNA expression levels of hub genes, RT-qPCR was employed. Total RNA was isolated from lung tissues utilizing TRIzol reagent (Cat. No. 15596026, Thermo Fisher Scientific, Waltham, MA, USA). The mRNA levels were quantified using SYBR Green Premix Ex TaqII (Cat. No. RR820A, TaKaRa, Dalian, China). The cycling conditions were 95 °C for 30 sec, then 40 cycles of 95 °C for 15 sec and 60 °C for 30 sec. The relative levels of gene expression were calculated by reference to the $\beta$-actin level in each sample. Each sample was analyzed with triplicate biological replicates and triplicate technical replicates per biological replicate. The 2^{-Δ⁢Δ⁢Ct} method was employed to calculate the relative expression levels of genes, and primer sequences are provided in Table 1.

To elucidate the mechanism by which HCH alleviates OVA-induced asthma, 35 metabolic biomarkers and 273 DEGs were coanalyzed using the Cytoscape 3.9.1 software (Cytoscape Consortium, San Diego, CA, USA) plug-in Metscape 3.1.3 (University of Michigan, Ann Arbor, MI, USA) to generate a compound-reaction-enzyme-gene network. Spearman’s rank correlation coefficients were calculated between 35 metabolic biomarkers and 273 DEGs to identify key genes and metabolites with significant correlations. For validation purposes, RT-qPCR was conducted to verify mRNA levels of 8 key genes and primer sequences were shown in Table 1.

Data were represented as Mean $\pm$ Standard Error of the Mean (SEM). GraphPad Prism 8 (GraphPad Software Inc., San Diego, CA, USA) was utilized to analyze pharmacodynamic profiles. One-way ANOVA was employed to assess inter-group differences (p 0.05).

The UPLC-Q-Exactive MS analysis was performed to identify the chemical constituents present in the HC extract. This analysis revealed 28 compounds (Supplementary Fig. 1A,B). Under positive ion mode, 21 compounds were identified, including chelidonic acid, p-coumaric acid, caffeic acid, magnocurarine, magnoflorine, chelamine, protopine, chelidonine, tetrahydrocoptisine, coptisine, homochelidonine, berberine, quercetin, chelerythrine, chelilutine, kaempferol, isorhamnetin, oxysanguinarine, dihydrochelerythrine, dihydrosanguinafine and corysamine.

In negative ion mode, 7 compounds were identified, including citric acid, gallic acid, ferulic acid, genistein, luteolin, calycosin and formononetin (Table 2). Further analysis of these identified compounds using the Traditional Chinese Medicine Systems Pharmacology Database and Platform (TCMSP) and relevant literature exploration suggested that all 28 compounds were considered major contributors to the therapeutic effects of HC extract.

At the start of the experiment, there were no notable differences in body weight and relative lung weight among rats in all six groups. Compared to Control group, OVA challenge caused a notable decrease (p 0.01) in body weight whereas a remarkable increase (p 0.001) in relative lung weight in Model group. Administration of Dex, medium- and high-dose HC (HCM and HCH) remarkably elevated body weight (p 0.05) whereas suppressed (p 0.01, p 0.05 and p 0.01) relative lung weight compared to Model group (Fig. 1a,b).

Fig. 1.

Effect of HC administration on body weight and relative lung weight in OVA-induced asthmatic rats. (a) Body weight and (b) Relative lung weight of groups after administration. Values are shown as Mean $\pm$ SEM (n = 8), ^#⁢#p 0.01, ^#⁢#⁢#p 0.01 vs Control group and *p 0.05, **p 0.01 vs Model group.

Histological examination of bronchial tissues revealed no abnormalities in the Control group. In contrast, the Model group displayed marked infiltration of inflammatory cells in bronchi and perivascular. Administration of Dex, HCM and HCH obviously reduced the severity of inflammatory cell infiltration compared to Model group (Fig. 2a). Based on PAS staining, there was few goblet cells in Control group. On the contrary, the Model group displayed significant proliferation of goblet cells on airway epithelium. The Dex, HCL, HCM and HCH groups obviously attenuated goblet cell proliferation compared to the model group (Fig. 2b).

Fig. 2.

HC alleviates pulmonary inflammation and goblet cell proliferation. (a) Hematoxylin and Eosin (H&E) staining of lung sections ($\times$200) and inflammation scores and (b) Periodic acid-Schiff (PAS) staining of pulmonary sections ($\times$200) and the PAS score. Scale bar = 100 µm. (a) Yellow arrows indicate goblet cell proliferation and (b) Green arrows indicate inflammatory cell infiltration. Data are represented as Mean $\pm$ SEM (n = 3), ^#⁢#p 0.01 vs Control group and *p 0.05, **p 0.01 vs Model group.

In Model group, IL-13 and IL-1$\beta$ levels were notably elevated, while IL-10 level was lower in comparison to Control group. Inversely, compared to Model group, treatment with Dex, HCM and HCH reduced IL-13 and IL-1$\beta$ levels and notably increased IL-10 levels. The HCL group also significantly reduced IL-1$\beta$ and increased IL-10 levels (p 0.05) compared to Model group (Fig. 3a–c). These findings demonstrated that medium- and high-dose HC exerted anti-inflammatory effects, with high-dose HC potentially exhibiting a more pronounced effect.

Fig. 3.

Serum levels of inflammatory markers. (a) IL-10, (b) IL-13 and (c) IL-1$\beta$ in asthmatic rats treated with HC. Data are represented as Mean $\pm$ SEM (n = 8), ^#p 0.05, ^#⁢#p 0.01 vs Control group and *p 0.05, **p 0.01, ***p 0.001 vs Model group.

To assess the stability of the UPLC-MS method, three quality control (QC) samples, prepared by pooling equal aliquots of experimental samples from all groups, were run before initiating the experiment and then once every six samples throughout the analysis. Furthermore, polar and nonpolar tissue samples underwent analysis using negative and positive ion modes, respectively (Supplementary Fig. 2A–F). Following data calibration and normalization, PCA analysis was performed using datasets from all samples. The PCA score plots (Supplementary Fig. 2G,H) demonstrate tight clustering of QC samples in the center for both ion modes, suggesting good reproducibility and stability of the developed method.

The PCA score plots (Fig. 4a) revealed a basic separation between Control and Model groups. Furthermore, HCH group clustered closer to Control group compared to Model group. This suggested distinct metabolite profiles between Control and Model groups, with HCH group exhibiting more similarity to Control group. OPLS-DA yielded R2Y and Q2 values of 0.990 and 0.902 for Control-vs-Model (Fig. 4b), as well as 0.962 and 0.859 for HCH-vs-Model (Fig. 4d), consistently showed clear discrimination. Permutation tests confirmed the reliability of models, with R2 and Q2 values well below the original values (Fig. 4c,e). Based on the multivariate analysis of OPLS-DA (FC $>$1, VIP $>$1 and p 0.05), 46 candidate serum metabolites were detected as potential biomarkers. These metabolites were primarily associated with lipid, organic acid, fatty acid and amino acid. Compared to Control group, 15 metabolites were up-regulated, while 31 were down-regulated in Model group. The HCH group exhibited 16 down-regulated and 19 up-regulated metabolites compared to the model group (Supplementary Table 3). The metabolite heatmap was shown in Fig. 4f.

Fig. 4.

Metabolomic profiling among the Control, Model and HCH groups. (a) Principal component analysis (PCA) score plot showcasing the overall separation of metabolic profiles. (b–e) Orthogonal projections to latent structures discriminant analysis (OPLS-DA) score plots depicting model validation and group separation. (f) Heatmap visualization of metabolic biomarkers among groups (n = 6) and (g,h) Metabolic pathway analysis of Model-vs-Control group, Model-vs-Control and HCH-vs-Model groups. (g,h) Size of the bubble corresponds to the pathway impact score and the color intensity represents the –log(p-value).

Metabolic pathway analysis of the identified metabolites revealed distinct alterations across experimental groups. The Model group, compared to Control, exhibited remarkable enrichment in twelve key metabolic pathways, primarily related to lipid metabolism (glycerophospholipid, arachidonic acid, linoleic acid and sphingolipid), amino acid metabolism (arginine, proline, tyrosine, phenylalanine, tryptophan and beta-alanine) and energy metabolism (citrate cycle, pyruvate metabolism and purine metabolism) (Fig. 4g). Treatment with HCH partially reversed these, with 11 major pathways showing significant modulation, notably those associated with amino acid and energy metabolism (Fig. 4h).

To sum up, asthma induction disturbed these metabolic biomarkers as well as associated metabolic pathways. The alteration was mainly associated with lipid, amino acid and fatty acids metabolic disorders. After HCH intervention, the disorder reversed to some extent.

The RNA-seq was performed on lung tissues from Control, Model and HCH groups to identify major DEGs and functional pathways involved in HCH action against OVA-induced allergic asthma. In Model-vs-Control group comparison, 754 DEGs were detected, with 454 up-regulated and 300 down-regulated. The HCH group exhibited 273 DEGs compared to Model group, of which 167 up-regulated and 106 down-regulated (Fig. 5a). The 20 most significantly enriched pathways of KEGG enrichment analysis are presented in Fig. 5b,c. In Model-vs-Control group, the most enriched pathways were primarily related to inflammation pathways (cytokine-cytokine receptor interaction, chemokine signaling pathway, TNF signaling pathway and IL-17 signaling pathway), immune pathways (hematopoietic cell lineage, antigen processing and presentation, intestinal immune network for IgA production) and asthma. In Model-vs-Control and HCH-vs-Model groups, the enrichment pathways were mostly associated with airway mucus secretion (Hedgehog pathway), inflammation pathways (PI3K-Akt signaling pathway, MAPK signaling pathway, chemokine pathway, TNF signaling pathway and IL-17 signaling pathway) and airway remodeling (ECM-receptor interaction).

Fig. 5.

Transcriptomic analysis revealing differentially expressed genes (DEGs), key pathways and hub genes modulated by HCH in the OVA-induced asthma model. (a) Heatmap showing DEGs in Control, Model and HCH groups. (b,c) Bubble chart showing Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of DEGs in Model-vs-Control, HCH-vs-Model groups and (d) Protein-protein interaction (PPI) networks for DEGs in Model-vs-Control and HCH-vs-Model comparisons.

According to the degree centrality, the PPI network analysis identified 6 genes with the highest degree values, including Il6 (degree = 82), Il10 (degree = 62), Ccl2 (degree = 52), Cxcl2 (degree = 34), Ptgs2 (degree = 26) and Cxcl6 (degree = 26), in the Model-vs-Control and HCH-vs-Model groups (Fig. 5d).

To validate the functions of HC in asthma by those potential hub genes, RT-qPCR was performed to assess Il6, Il10, Ccl2, Cxcl2, Ptgs2 and Cxcl6 mRNA expression (Fig. 6). The OVA-induced asthma resulted in a significant up-regulation of Il6, Ccl2, Cxcl2, Ptgs2 and Cxcl6 relative to Control group. Administration with HCH significantly down-regulated Il6, Ccl2, Cxcl2 and Cxcl6 expressions (p 0.05). However, the Ptgs2 expression did not change obviously and Il10 was not detected.

Fig. 6.

Validation of hub gene expression by RT-qPCR. Relative mRNA expression levels of (a) Il6, (b) Ccl2, (c) Cxcl2, (d) Ptgs2, and (e) Cxcl6 in the lung tissues of Control, Model, and HCH-treated rats. Results are represented by Mean $\pm$ SEM (n = 3), ^#p 0.05, ^#⁢#p 0.01 vs Control group and *p 0.05 vs Model group.

To elucidate potential connections between DEGs and metabolic biomarkers, a gene-metabolite network was constructed using Metscape software (Supplementary Fig. 3). This analysis revealed seven key targets (P4ha1, Pde3a, Pde5a, Adcy9, Pck2, Pla2g16 and Entpd8), corresponding to eleven differential metabolites (L-proline, L-arginine, L-malic acid, inosine, xanthosine, hypoxanthine, LysoPC(20:0), LysoPC(20:1(11Z)), arachidonic acid, PGE2 and PGF2a), involved in six important metabolic pathways (purine metabolism; arachidonic acid metabolism; urea cycle and metabolism of arginine, proline, glutamate, aspartate and asparagine; glycerophospholipid metabolism; glycolysis and gluconeogenesis; TCA cycle) in Model-vs-Control and HCH-vs-Model groups.

Spearman rank correlation was performed to assess the potential co-relationship between DEGs and differential metabolites. The analysis revealed potential correlations between DEGs and metabolites. Further analysis identified statistically significant correlations between some metabolites and DEGs (Fig. 7a). P4ha1 was positively interrelated with L-proline and L-arginine (p 0.05). Pck2 showed a positive correlation with L-malic acid (p 0.01). Pde7b was negatively interrelated with ADP (p 0.05). Adcy9 exhibited negative correlations with inosine, xanthosine and hypoxanthine (p 0.05). Entpd8 showed positive correlations with inosine and xanthosine (p 0.01 and p 0.05). Pde3a was negatively interrelated with inosine and xanthosine (p 0.05 and p 0.01). Pla2g16 exhibited negative correlations with LysoPC(20:0) and LysoPC(20:1(11Z)) (p 0.05), while showing positive correlations with arachidonic acid, PGE2 and PGF2a (p 0.05). Ptgs2 displayed a positive correlation with arachidonic acid (p 0.05). The Pde3a was negatively interrelated with inosine and xanthosine (p 0.05 and p 0.01). The Pde5a was negatively interrelated with hypoxanthine and xanthosine (p 0.05). The Rrm2b displayed negative correlations with inosine and xanthosine (p 0.01 and p 0.05).

Fig. 7.

Correlation analysis of metabolites, differentially expressed genes (DEGs) and validation of hub gene expression. (a) Significantly correlated metabolic biomarkers and DEGs and (b) RT-qPCR validation of 10 genes. Data are expressed as Mean $\pm$ SEM (n = 3), ^#p 0.05, ^#⁢#p 0.01 vs Control group and *p 0.05, **p 0.01 vs Model group.

Furthermore, to evaluate the results of the network analysis, RT-qPCR was performed to verify the mRNA expression levels of those 10 genes in lung tissues (Fig. 7b). Compared with Control group, the expression of P4ha1, Pde3a, Pde5a, Pde7b and Adcy9 were notably reduced (p 0.05). Conversely, the expression levels of Ptgs2, Pck2, Pla2g16 and Entpd8 were markedly elevated in Model group. The HCH group notably promoted the expression of P4ha1, Pde3a, Pde5a and Adcy9 and obviously inhibited the expression levels of Pck2, Pla2g16 and Entpd8 compared to the Model group. However, the mRNA levels of Ptgs2 and Pde7b did not significantly improve. The changes of key metabolites and DEGs within five metabolic pathways identified in the analysis are shown in Fig. 8. It emphasizes the interconnectedness of metabolic and transcriptional responses and underscores HC’s modulatory effects on these pathways.

Fig. 8.

Changes of metabolites (blue squares) and DEGs (red circles) within five metabolic pathways based on correlation analysis. Red arrows indicate increased/decreased levels in Model-vs-Control group and blue arrows represent increased/decreased levels in HCH-vs-Model group.