A sustainable functional food ingredient from Saussurea involucrata cell culture alleviates atopic dermatitis-like inflammation via arachidonic acid metabolic remodeling

Abstract

Nutritional interventions may complement established approaches for managing atopic dermatitis (AD). Here, we evaluated Saussurea involucrata cell culture (SIC), a sustainable functional-food material, in a DNCB-induced mouse model of AD-like inflammation. Chemical profiling established the phytochemical composition of SIC, and both topical and oral interventions improved macroscopic and histopathological outcomes. Under the tested conditions, the oral groups showed more consistent changes in systemic indices, providing the basis for subsequent mechanistic analyses. Integrated transcriptomic and metabolomic analyses, targeted lipid-mediator profiling, and qRT-PCR identified coordinated changes in the arachidonic acid (AA) metabolic network. Exploratory docking and molecular dynamics provided descriptive computational support for target-dependent retention or reorientation of selected ligand-protein configurations, while tissue-level immunohistochemistry supported corresponding changes in selected AA pathway-associated, type 2 inflammatory, and epidermal-junction proteins. Together, these findings associate oral SIC supplementation with attenuation of AD-like inflammation and AA metabolic remodeling.

Introduction

Considered the most widespread chronic inflammatory skin disease worldwide, atopic dermatitis (AD) affects approximately 15–25% of pediatric populations and 3–7% of adults, creating profound physical discomfort and socioeconomic burdens for patients and healthcare systems alike^1–3. While conventionally viewed as a localized cutaneous disorder, accumulating evidence has redefined AD as a systemic immune-metabolic disease. The pathogenesis encompasses not only localized epidermal barrier defects but also profound systemic immune dysregulation, largely driven by an overactive type-2 inflammatory cascade and abnormal lipid mediator networks, which leads to heterogeneous manifestations and a variable disease course^4–6.

Currently, conventional pharmacological interventions, such as topical corticosteroids and systemic immunosuppressants, remain the primary therapeutic modalities. However, their prolonged utilization is frequently restricted by severe long-term adverse effects and toxicity^7,8. Driven by the emerging concept of the “gut-skin axis“^9,10, there is a surging interest in exploring natural and bioactive compound-rich dietary interventions capable of modulating systemic immunity from the inside out. Nutritional strategies using functional foods or botanical supplements offer a promising alternative to manage chronic skin inflammation while overcoming the limitations associated with long-term pharmacological treatments.

Saussurea involucrata (Kar. et Kir.) Sch.-Bip, renowned in traditional systems as a medicine-food homologous material, represents a precious natural resource^11,12. Its scarcity stems from extreme high-altitude habitats, prolonged growth cycles, and challenging cultivation. Recent advancements in food biotechnology have led to the development of Saussurea involucrata cell cultures (SIC) as a sustainable, eco-friendly, and highly standardized alternative to wild-harvested material, obtained through sterile in vitro propagation^13–15. Officially recognized as a novel food ingredient, SIC shows great potential to be incorporated into functional beverages and dietary supplements. These characteristics make SIC particularly attractive as a functional food ingredient, as its standardized quality, sustainable production, and regulatory approval facilitate future dietary applications and translational development compared with conventionally harvested botanical materials.

Phytochemical analyses confirm that SIC contains multiple bioactive compounds, exhibiting profiles comparable to wild plants¹². Studies demonstrate that SIC possesses anti-inflammatory, lipid-regulating, antioxidant, and immunomodulatory activities^14,16,17. Previous research indicates that SIC flavonoids inhibit cyclooxygenase activity, thereby suppressing the production of pro-inflammatory mediators^18,19. Furthermore, SIC effectively modulates immune parameters, such as macrophage and lymphocyte functions, which help maintain physiological homeostasis^20–22. Given that the systemic immune-metabolic dysregulation of AD is critically driven by the aberrant remodeling of bioactive lipids, particularly the hyperactivation of the arachidonic acid (AA) metabolic cascade, targeting the upstream enzymatic switches of this pathway represents a crucial therapeutic node. Because the predominant phenolic and phenylpropanoid constituents of SIC have been reported to possess anti-inflammatory properties and the potential to modulate lipid metabolism, we hypothesized that they may influence arachidonic acid metabolic networks relevant to AD pathogenesis. Although these mechanisms are relevant to AD pathogenesis, the comprehensive ameliorative potential of SIC as a specific dietary intervention for AD, as well as its underlying systemic metabolic mechanisms, remains to be systematically elucidated.

We therefore characterized the phytochemical composition of SIC and evaluated its effects in a DNCB-induced mouse model of AD-like inflammation. Oral dietary supplementation and topical application were assessed in separate cohorts at the phenotypic and histopathological levels. Because the oral regimen produced more consistent changes under the tested conditions, skin samples from this cohort were analyzed by transcriptomics, untargeted metabolomics, targeted lipid-mediator profiling, qRT-PCR, and tissue-level immunohistochemistry. Molecular docking and MD were used as hypothesis-generating analyses of selected compounds and pathway-associated proteins. This design examined molecular changes associated with oral SIC supplementation.

Results

Chemical profiling and quantitative analysis of key bioactive markers in SIC

UPLC-MS/MS profiling characterized 87 compounds in SIC (Fig. 1A, B). The largest classes were phenolic acids (31 compounds), flavonoids (22 compounds), and lignans and coumarins (16 compounds), with smaller numbers of terpenoids, organic acids, and other constituents (Fig. 1C). Of the 87 compounds, 51 were confirmed using authentic standards; the remaining assignments were based on chromatographic and mass-spectrometric annotation (Table 1).

Fig. 1. Determination of the chemical constituents in SIC extract.

A Total ion chromatogram (TIC) acquired in positive ion mode. B TIC in negative ion mode. C Classification and relative distribution of non-volatile metabolites identified by LC-MS. D Classification and relative distribution of volatile components identified by GC-MS. E HPLC chromatograms of three major compounds in SIC extract.

Table 1.

UPLC-ESI-MS/MS data of the compositions of SIC

No.	Inform	Identification	Type	tR(min)	Formula	Q1 (Da)	Q3 (Da)
1	[M-H]-	Sucrose*	Others	0.8	C12H22O11	341.11	119.03
2	[M-H]-	Fumarate*	Organic acids	1.1	C4H4O4	115	71.01
3	[M-H]-	Quinate*	Organic acids	1.1	C7H12O6	191.06	85.03
4	[M-H]-	Succinate*	Organic acids	1.3	C4H6O4	117.02	73.03
5	[M-H]-	(R)-3-Hydroxybutanoate*	Organic acids	1.4	C4H8O3	103.04	59.01
6	[M-H]-	Gallate*	Phenolic acids	1.8	C7H6O5	169.0143	125.0244
7	[M-H]-	1-Caffeoylquinic acid*	Phenolic acids	2.2	C16H18O9	353.09	191.06
8	[M-H]-	3,4-Dihydroxybenzoate*	Phenolic acids	2.5	C7H6O4	153.02	109.03
9	[M + H]+	Skimmin (7-Hydroxycoumarin-7-O-glucoside)	Lignans and Coumarins	2.6	C15H16O8	325.09	163.04
10	[M-H]-	isoscopolin	Others	2.7	C16H18O10	369.08	191.05
11	[M-H]-	Chlorogenate*	Phenolic acids	2.7	C16H18O9	353.0878	191.0553
12	[M-H]-	Syringin*	Others	2.8	C17H24O9	371.1348	209.081
13	[M-H]-	1-O-rhamnose-3-O-Caffeoyl Quinic Acid	Phenolic acids	2.8	C22H28O13	499.15	191.06
14	[M-H]-	3’-p-Coumaroyl-sucrose	Phenolic acids	2.8	C21H28O13	487.15	145.03
15	[M-H]-	methyl 4-O-galloylchlorogenate	Phenolic acids	2.8	C24H24O13	519.11	165.02
16	[M + H]+	Neochlorogenic acid*	Phenolic acids	2.9	C16H18O9	355.1	163.04
17	[M + H]+	Scopolin*	Lignans and Coumarins	3	C16H18O9	355.1	193.05
18	[M-H]-	( + )-Catechin*	Flavonoids	3	C15H14O6	289.07	245.08
19	[M-H]-	4-Hydroxybenzoate*	Phenolic acids	3	C7H6O3	137.0244	93.0346
20	[M-H]-	4-p-Coumaroylquinic acid*	Phenolic acids	3.1	C16H18O8	337.09	173.05
21	[M + H]+	Berchemol 4’-Glucoside	Lignans and Coumarins	3.2	C26H34O12	539.21	359.15
22	[M + H]+	(-)-Epicatechin*	Flavonoids	3.3	C15H14O6	291.09	139.04
23	[M-H]-	Vanillate*	Phenolic acids	3.3	C8H8O4	167.03	108.02
24	[M-H]-	3,4-Dihydroxybenzaldehyde	Others	3.3	C7H6O3	137.02	93.04
25	[M + H]+	1-O-Feruloylquinic acid	Phenolic acids	3.4	C17H20O9	369.12	177.05
26	[M-H]-	7,8-Dihydroxycoumarin*	Lignans and Coumarins	3.4	C9H6O4	177.02	121.03
27	[M-H]-	Eleutheroside E*	Lignans and Coumarins	3.4	C34H46O18	741.26	579.21
28	[M-H]-	Caffeate*	Phenolic acids	3.4	C9H8O4	179.03	135.05
29	[M + H]+	3,5-di-O-caffeoylquinic acid glucoside	Others	3.5	C31H34O17	679.19	163.04
30	[M + H]+	3-O-Feruloylquinic acid*	Phenolic acids	3.5	C17H20O9	369.12	177.06
31	[M-H]-	Lariciresinol 4-O-glucoside*	Lignans and Coumarins	3.5	C26H34O11	521.2	329.14
32	[M + H]+	Ethyl chlorogenate	Phenolic acids	3.6	C18H22O9	383.13	163.04
33	[M-H]-	Secoisolariciresinol 4-O-glucoside	Lignans and Coumarins	3.6	C26H36O11	523.22	361.14
34	[M-H]-	Matairesinol-4,4’-O-diglucoside	Lignans and Coumarins	3.7	C32H42O16	681.24	357.13
35	[M-H]-	Chlorogenic acid methyl ester*	Phenolic acids	3.7	C17H20O9	367.1	190.9
36	[M + H]+	Rutin*	Flavonoids	3.8	C27H30O16	611.16	303.05
37	[M + H]+	Desacetoxymatricarin	Terpenoids	3.8	C15H18O3	247.13	187.11
38	[M-H]-	Eriodictyol-7-O-glucuronide*	Flavonoids	3.8	C21H20O12	463.09	287.06
39	[M + H]+	Scolymoside*	Flavonoids	3.87	C27H30O15	595.1658	287.0584
40	[M + H]+	Zaluzanin C	Terpenoids	3.9	C15H18O3	247.13	131.08
41	[M + H]+	Luteolin 7-O-beta-D-glucoside*	Flavonoids	3.95	C21H20O11	449.1078	287.0622
42	[M + H]+	Isoquercitrin*	Flavonoids	4	C21H20O12	465.1	303.05
43	[M + H]+	1,3,4-O-tricaffeoylquinic acid	Others	4	C34H30O15	679.17	163.04
44	[M + H]+	Eriodictyol-7-O-glucoside*	Flavonoids	4	C21H22O11	451.12	289.07
45	[M + H]+	1,3-Dicaffeoylquinic acid*	Phenolic acids	4	C25H24O12	517.1334	163.0394
46	[M-H]-	Ferulate*	Phenolic acids	4	C10H10O4	193.05	134.04
47	[M-H]-	( + )-Syringaresinol O-beta-D-glucoside*	Lignans and Coumarins	4	C28H36O13	579.21	417.16
48	[M-H]-	Isochlorogenic acid C	Phenolic acids	4	C25H24O12	515.12	353.09
49	[M + H]+	Isorhamnetin 3-O-glucoside	Flavonoids	4.1	C22H22O12	479.12	317.07
50	[M-H]-	Phenyllactate	Phenolic acids	4.1	C9H10O3	165.06	103.06
51	[M + H]+	Scopoletin	Lignans and Coumarins	4.2	C10H8O4	193.0495	133.0255
52	[M-H]-	Kaempferol-7-O-glucoside*	Flavonoids	4.2	C21H20O11	447.09	285.04
53	[M-H]-	Umbelliferone	Lignans and Coumarins	4.2	C9H6O3	161.02	133.03
54	[M-H]-	syringaresinol-4-O-glucoside	Lignans and Coumarins	4.2	C28H36O13	579.21	417.16
55	[M-H]-	Isochlorogenic acid b	Phenolic acids	4.2	C25H24O12	515.12	353.09
56	[M + H]+	Taxifolin*	Flavonoids	4.21	C15H12O7	305.0645	231.0647
57	[M + H]+	Apigenin 7-O-beta-D-glucoside*	Flavonoids	4.3	C21H20O10	433.11	271.07
58	[M + H]+	Diosmetin-7-O-glucoside*	Flavonoids	4.3	C22H22O11	463.12	301.07
59	[M-H]-	Chrysoeriol-7-O-glucoside*	Flavonoids	4.3	C22H22O11	461.11	299.06
60	[M + H]+	Myricetin*	Flavonoids	4.4	C15H10O8	319.04	153.02
61	[M + H]+	Homoplantaginin*	Flavonoids	4.4	C22H22O11	463.13	301.07
62	[M-H]-	Dihydrocoumarin*	Lignans and Coumarins	4.4	C9H8O2	147.05	119.05
63	[M-H]-	3-O-Caffeoyl-4-O-p-Coumaroylquinic acid	Phenolic acids	4.4	C25H24O11	499.12	337.09
64	[M + H]+	Benzoate*	Phenolic acids	4.5	C7H6O2	123.04	79.05
65	[M-H]-	4-Hydroxycoumarin*	Lignans and Coumarins	4.5	C9H6O3	161.02	117.03
66	[M-H]-	3-O-Caffeoyl-5-O-p-Coumaroylquinic acid	Phenolic acids	4.5	C25H24O11	499.12	337.09
67	[M + H]+	1,3,4-Tri-O-Caffeoyl Quinic Acid	Phenolic acids	4.7	C34H30O15	679.17	163.04
68	[M-H]-	Methyl caffeate	Phenolic acids	4.7	C10H10O4	193.05	135.05
69	[M-H]-	3,4-Di-O-p-Coumaroylquinic acid	Phenolic acids	4.8	C25H24O10	483.13	163.04
70	[M + H]+	1,3-Dicoumaroylquinic acid	Phenolic acids	4.9	C25H24O10	485.15	147.04
71	[M-H]-	3,5-Di-O-p-Coumaroylquinic acid	Phenolic acids	4.9	C25H24O10	483.13	163.04
72	[M-H]-	3,4,5-Tricaffeoylquinic acid*	Phenolic acids	4.9	C34H30O15	677.15	515.12
73	[M + H]+	3,5-Di-O-caffeoyl-4-O-coumaroylquinic acid	Phenolic acids	5.1	C34H30O14	663.17	163.04
74	[M + H]+	Quercetin	Flavonoids	5.2	C15H10O7	303.05	229.05
75	[M + H]+	Luteolin*	Flavonoids	5.2	C15H10O6	287.06	153.02
76	[M-H]-	Eriodictyol*	Flavonoids	5.2	C15H12O6	287.05	135.04
77	[M-H]-	( + )-Syringaresinol*	Lignans and Coumarins	5.3	C22H26O8	417.16	402.1
78	[M + H]+	Apigenin*	Flavonoids	5.7	C15H10O5	271.06	153.02
79	[M-H]-	Kaempferol*	Flavonoids	5.7	C15H10O6	285.04	151
80	[M + H]+	Hispidulin*	Flavonoids	5.8	C16H12O6	301.07	286.05
81	[M-H]-	Reynosin	Terpenoids	6	C15H20O3	247.13	203.14
82	[M + H]+	Arctigenin	Lignans and Coumarins	6.18	C21H24O6	373.1662	137.0603
83	[M + H]+	beta-Sitosterol*	Steroids	7.75	C29H50O	415.3934	119.197
84	[M + H]+	Dihydroisoalantolactone	Terpenoids	9.1	C15H22O2	235.17	133.1
85	[M + H]+	1-Linoleoylglycerol	Lipids	10.7	C21H38O4	355.28	263.24
86	[M + H]+	Oleamide	Lipids	11.2	C18H35NO	282.28	247.24
87	[M-H]-	Hexadecanoic acid*	Lipids	11.7	C16H32O2	255.23	237.22

HS-SPME-GC-MS detected 158 volatile compounds with estimated OAVs ≥ 1 (Table S4). Because the analysis was semi-quantitative, these compounds are described as potential odor-active contributors rather than definitively quantified key odorants. Esters (30 compounds), aldehydes (21), terpenoids (20), and alcohols (20) were the predominant classes (Fig. 1D).

To establish a clear material basis for the dietary intervention, three key bioactive markers in SIC (chlorogenic acid, syringin, and 1,5-dicaffeoylquinic acid) were selected for absolute quantification via UPLC. The rationale for selecting these specific compounds is primarily because they represent the predominant phenolic and phenylpropanoid classes in the extract, ensuring high natural abundance. Prior to sample analysis, the analytical method was rigorously validated. As detailed in Table 2, the established UPLC method exhibited excellent linearity (R² ≥ 0.9999), high precision and stability (RSD < 0.5%), and reliable recovery rates (97.23–101.65%), thereby confirming the high accuracy and reproducibility of the quantification. The quantitative results demonstrated that these three compounds accounted for 1.527%, 1.113%, and 4.668% of the total extract mass, respectively (Fig. 1E).

Table 2.

Methodological validation results for the quantitative analysis of three key bioactive markers in SIC

Compound	Linear range(μg/ml)	Calibration curve	R2	Precision (RSD, %)	Repeatability (RSD, %)	Stability (RSD, %)	Recovery (%)
Chlorogenic acid	0.0872-0.872	y = 4 × 106×– 1428.4	1	0.226	0.464	0.458	97.234 ± 0.635
1,5-Dicaffeoylquinic acid	0.3198-3.198	y = 5 × 106×– 12507	0.9999	0.127	0.466	0.202	101.651 ± 0.965
Syringin	0.0834-0.834	y = 107×–1078.8	0.9999	0.125	0.197	0.132	101.508 ± 0.555

SIC interventions alleviate macroscopic AD-like features in mice

The normal control mice remained in good condition, whereas DNCB-treated model mice developed visible erythema, crusting, and scaling. Both topical SIC and oral SIC supplementation reduced dermatitis scores relative to the model groups (Fig. 2B). Ear-swelling measurements showed corresponding improvements in both intervention cohorts (Fig. 2C). Under the tested conditions, oral supplementation produced stronger phenotypic improvement, suggesting systemic modulation of the inflammatory response.

Fig. 2. Ameliorative effects of SIC on DNCB-induced atopic dermatitis in mice.

A Schematic diagram of the experimental design and the establishment of the atopic dermatitis model. B Representative images of dorsal skin morphology and corresponding dermatitis scores for mice in each group. C Ear swelling rate and (D) spleen index across different treatment groups. Complete blood count analysis showing the levels of (E) leukocytes, (F) eosinophils, (G) monocytes, and (H) neutrophils. (Data represent the mean ± SD. ^#p < 0.05, ^##p < 0.01, ^###p < 0.001 vs control^{, *}p < 0.05, ^**p < 0.01, ^***p < 0.001 vs model, n = 6).

SIC-associated changes in histopathological features and hematological indices

The spleen index was used as a macroscopic indicator of systemic immune activation. Most SIC-treated groups showed lower spleen indices than the model group, although the reduction in the high-dose topical group did not reach statistical significance. All oral SIC groups showed significant reductions in the spleen index (Fig. 2D). The high-dose oral group also showed values within the range observed for the positive-reference group. Overall, oral SIC supplementation was consistently associated with a lower spleen index across the tested doses.

Peripheral blood white blood cell, neutrophil, monocyte, and eosinophil counts were examined as hematological indicators of systemic inflammation. Topical SIC treatment reduced several of these indices, although the magnitude and statistical significance of the changes varied among cell types and doses. Oral SIC supplementation significantly reduced all four measured indices across the tested doses (Fig. 2E–H). The consistent changes in the spleen index and peripheral-blood parameters provided the basis for focusing subsequent mechanistic analyses on oral SIC supplementation.

Dorsal and auricular skin sections were examined by H&E and toluidine blue staining to assess DNCB-induced histopathological changes. The model group showed disorganization of the squamous epithelium, epithelial-cell degeneration and necrosis, ulceration, congestion, edema, and disrupted collagen architecture. These changes were accompanied by extensive inflammatory-cell infiltration in the epidermis, dermis, and subcutaneous tissue (Fig. 3A, B). Toluidine blue staining further showed increased mast-cell accumulation in the dermal and subcutaneous layers (Fig. 3C, D). These histopathological alterations were attenuated after oral SIC supplementation, with improvements broadly resembling those observed in the positive-reference group. Topical SIC treatment also reduced several features of tissue injury and inflammation. Together, these findings indicate that SIC treatment alleviated local histopathological alterations in the DNCB-induced model and support further evaluation of oral SIC as a functional food intervention.

Fig. 3. Histological evaluation of skin tissue.

Representative images of histological skin (A) and ear (B) sections stained with H&E. Toluidine blue staining results of the dorsal skin (C) and ear (D) in each group. (Data represent the mean ± SD. ^#p < 0.05, ^##p < 0.01, ^###p < 0.001 vs control, ^*p < 0.05, ^**p < 0.01, ^***p < 0.001 vs model, n = 3).

Untargeted metabolomics reveals the modulation of metabolic profiles by SIC

Motivated by the stronger phenotypic alleviation observed with oral supplementation under the tested conditions, untargeted metabolomic profiling of skin tissues was conducted to elucidate the mechanistic basis of SIC from a metabolic perspective using 6 biological replicates per group. Initial Principal Component Analysis (PCA) revealed distinct clustering among the three groups, highlighting significant metabolic perturbations in the AD model (Fig. 4A, B). Importantly, all quality control (QC) samples clustered tightly in the center of the unsupervised PCA score plots (Figure. S1), demonstrating excellent analytical stability and sequence reproducibility throughout the LC-MS run. To enhance the separation of improved variables and minimize noise, Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) was subsequently applied (Fig. 4C–F). The model’s reliability was validated via permutation tests, yielding high predictability and reliability indices (Strong polarity: R²Y = 0.954-0.969, Q² = 0.833-0.871; Weak polarity: R²Y = 0.996-0.998, Q² = 0.960-0.974). Furthermore, the statistical validity of the OPLS-DA models was confirmed by the corresponding permutation regression lines, where all calculated Q² intercepts successfully plotted below zero (Figure. S2), rigorously demonstrating excellent robustness and the absence of model overfitting.

Fig. 4. SIC alleviates the skin metabolic profile in DNCB-induced AD mice.

A, B The PCA plot of skin in strong and weak polarity fractions. C–F The OPLS-DA plot of skin in strong and weak polarity fraction. G, H Comparative S-plots (control vs. model, model vs. SIC) for skin samples in the strong and weak polarity fractions. I Volcano plot analysis of skin in control vs model and model vs SIC. J Heatmap analysis of skin in control vs model and model vs SIC. K, L KEGG functional enrichment analysis of control vs model and model vs SIC. (n = 6).

S-plots and volcano plots were used to prioritize metabolites according to the stated VIP, fold-change, and significance criteria (Fig. 4G–I). The control-versus-model comparison yielded 366 differential metabolites (233 decreased and 133 increased), whereas the model-versus-SIC comparison yielded 557 (439 decreased and 118 increased). Heatmap and KEGG analyses summarized the resulting patterns (Fig. 4J–L), including changes in purine, linoleic acid, glutathione, and PPAR-related pathways. AA metabolism was subsequently examined as one pathway shared across the omics datasets (Tables S5 and S6).

Transcriptomic analysis identifies key signaling pathways regulated by SIC intervention

Skin transcriptomes from the control, model and oral SIC groups separated by PCA (Fig. 5A). Differential-expression analysis identified 3465 genes in the control-versus-model comparison and 1042 genes in the model-versus-SIC comparison, with 356 genes shared between comparisons (Fig. 5B). The corresponding volcano plots contained 1673 decreased and 1792 increased genes for control versus model and 325 decreased and 717 increased genes for model versus SIC (Fig. 5C, D). Hierarchical clustering analysis based on DEG profiles further substantiated the therapeutic efficacy of SIC against AD (Fig. 5E, F). Subsequent GO enrichment analysis indicated that these DEGs were predominantly associated with biological processes such as muscle system processes and skin development, as well as molecular functions including myofibril components and actin binding (Fig. 5G, H). KEGG analysis revealed arachidonic acid metabolism, adipocytokine signaling pathway and glutathione metabolism (Fig. 5I, J).

Fig. 5. Transcriptomics for SIC treatment of AD.

A The PCA plot of skin. B The Venn plot of DEGs. C-D: DEGs volcano map of control vs model and model vs SIC. E, F Heatmap of the differentially expressed genes. G, H GO functional enrichment analysis of model vs SIC. I, J KEGG functional enrichment analysis of control vs model and model vs SIC. (n = 3).

Integrated multi-omics analysis identifies associations with the arachidonic acid network

To decipher the systemic regulatory network associated with the SIC intervention, an integrated analysis was performed on matched transcriptomic and metabolomic datasets. Pearson correlation analysis identified strong positive associations (R ≥ 0.8, p < 0.05) between numerous gene-metabolite pairs (Fig. 6A, B). These relationships were further visualized using hierarchical clustering heatmaps, which not only depicted the intensity of inter- and intra-group correlations but also confirmed distinct separation among experimental groups (Fig. 6C, D).

Fig. 6. Integration of transcriptomics and metabolomics reveals that SIC is associated with the modulation of the AA metabolic network and the amelioration of atopic dermatitis.

A, B Nine-quadrant diagrams showing the correlation between differentially expressed genes (DEGs) and differentially abundant metabolites (DAMs) in the Control vs. Model and Model vs. SIC comparisons. C, D Hierarchical clustering heatmaps of significant DEGs and DAMs across all experimental groups. E, F Joint KEGG pathway enrichment analysis based on the integrated multi-omics data. G Gene Set Enrichment Analysis (GSEA) plot highlighting the regulation of the AA metabolism pathway. H Interaction network of key genes and metabolites involved in the AA metabolic pathway.

Joint KEGG pathway enrichment revealed that differentially expressed entities from both omics layers converged on several key pathways. These included glycine, serine, and threonine metabolism, oxytocin signaling, bile secretion, and most notably, arachidonic acid (AA) metabolism (Fig. 6E, F). Crucially, Gene Set Enrichment Analysis (GSEA) demonstrated that the AA metabolic pathway, which was aberrantly activated in the model group, underwent significant suppression following SIC intervention (Fig. 6G). This suggests that the dietary intervention of SIC exerts its ameliorative effects in close association with the modulation of lipid metabolic homeostasis and the alleviation of inflammation-associated AA network alterations. Finally, a compound-target-pathway network was constructed (Fig. 6H) to systematically map the potential specific interactions between the quantified bioactive markers and key nodes within the AA pathway.

qRT-PCR validation and exploratory molecular docking and molecular-dynamics analyses of AA-pathway targets

To validate the transcriptomic findings, four genes showing marked changes in the arachidonic acid (AA) pathway were selected for qRT-PCR analysis. The expression levels of Pla2g4a, Alox5, Ptges, and Ptgr1 were higher in the model group than in the control group. Oral SIC supplementation reduced the expression of all four genes relative to the model group (Fig. 7A–D).

Fig. 7. qRT-PCR validation, molecular docking, and molecular-dynamics analyses of selected AA-pathway targets.

A–D Relative expression levels of Pla2g4a, Alox5, Ptges and Ptgr1, respectively, measured by qRT-PCR in skin tissues. (Data represent the mean ± SD. ^#p < 0.05, ^##p < 0.01, ^###p < 0.001 vs control^{, *}p < 0.05, ^**p < 0.01, ^***p < 0.001 vs model, n = 3). E–G Molecular docking binding poses and interaction-site maps of syringin, chlorogenic acid, and 1,5-dicaffeoylquinic acid with ALOX5, PTGR1, and PLA2G4A, respectively. H Ligand self-fit RMSD, representing internal ligand conformational change. I Ligand pose RMSD after protein-backbone alignment represents displacement relative to the initial docking pose. J Mean instantaneous protein–ligand hydrogen-bond counts. Rows in H–J represent ALOX5, PTGR1, and PLA2G4A, and columns represent syringin, chlorogenic acid, and 1,5-dicaffeoylquinic acid.

Molecular docking was then performed using syringin, chlorogenic acid and 1,5-dicaffeoylquinic acid against PLA2G4A, ALOX5, PTGES, and PTGR1. The docking results generated candidate binding configurations for the three representative SIC compounds (Table 3). These docking scores were used to generate interaction hypotheses rather than to demonstrate direct binding. Representative binding poses and interaction-site maps for ALOX5, PTGR1, and PLA2G4A are shown in Fig. 7E–G. Because the PTGES docking profile was comparatively weaker and less consistent across the tested compounds, PTGES was not advanced to molecular-dynamics analysis. Accordingly, no protein-level or enzyme-activity conclusion was drawn for PTGES.

Table 3.

Binding affinity scores (kcal/mol) of key active compounds with core targets in the arachidonic acid metabolic network

Ligand	ALOX5	PLA2G4	PTGES	PTGR1
Chlorogenic acid	−9.9	−9.9	−7.8	−9.1
1,5-Dicaffeoylquinic acid	−7.0	−9.5	−6.6	−7.5
Syringin	−6.3	−5.5	−5.3	−7

Molecular-dynamics simulations were performed for nine systems comprising ALOX5, PTGR1, or PLA2G4A complexed with syringin, chlorogenic acid, or 1,5-dicaffeoylquinic acid. The summary matrices in Fig. 7H–J show ligand self-fit RMSD, ligand pose RMSD, and mean protein-ligand hydrogen-bond counts, respectively. For the ALOX5 systems, mean ligand pose RMSD values were 0.130, 0.131, and 0.545 nm for syringin, chlorogenic acid, and 1,5-dicaffeoylquinic acid, respectively, whereas the corresponding mean hydrogen-bond counts were approximately 3.0. For PTGR1, mean pose RMSD values ranged from 0.294 to 0.348 nm, and mean hydrogen-bond counts ranged from 1.68 to 3.31. For PLA2G4A, mean ligand self-fit RMSD values were 0.145, 0.104, and 0.273 nm, mean pose RMSD values were 0.422, 0.431, and 0.461 nm, and mean hydrogen-bond counts were 1.972, 2.672, and 2.483 for syringin, chlorogenic acid, and 1,5-dicaffeoylquinic acid, respectively. These values provide descriptive information on ligand conformation, pose behavior, and polar contacts under the specified simulation conditions. Complete trajectory profiles are provided in Supplementary Figure. S3B–D and in Supplementary Table S7.

Targeted quantification of AA-associated lipid mediators

Targeted lipid-mediator profiling was performed to determine whether the transcriptional changes were accompanied by changes in AA-associated metabolites. Compared with the control group, the model group showed increased levels of AA and multiple metabolites associated with the LOX, COX, and CYP branches. Oral SIC supplementation reduced the levels of AA and several downstream mediators, including PGA2, PGD2, PGE2, PGB2, leukotriene B4, 20-HETE, and 5-HETE (Fig. 8A). These changes were consistent with treatment-associated remodeling of the AA-associated lipid-mediator profile.

Fig. 8. SIC alters the key metabolites of skin samples in DNCB mouse.

A The levels of key metabolites in the arachidonic acid metabolism pathway were determined by UPLC–MS/MS. (Data represent the mean ± SD. ^#p < 0.05, ^##p < 0.01, ^###p < 0.001 vs control^{, *}p < 0.05, ^**p < 0.01, ^***p < 0.001 vs model, n = 6).

Tissue-level immunohistochemistry supports molecular changes associated with oral SIC supplementation

IHC was performed to determine whether the molecular changes identified in the oral-treatment cohort were accompanied by corresponding changes in local protein immunoreactivity. Compared with the control group, the model group showed higher PLA2G4A, ALOX5, and PTGR1 immunoreactivity, whereas both oral SIC groups showed lower immunoreactivity for these proteins than the model group (Fig. 9A–C). GATA3-positive nuclei were also increased in the model group and reduced after oral SIC supplementation (Fig. 9D). ZO-1 immunoreactivity was lower in the model group than in the control group and increased in both oral SIC groups (Fig. 9E). These results indicate treatment-associated changes in selected AA-pathway-associated, type 2 inflammatory and epidermal-junction proteins, but they do not establish enzymatic activity or direct compound–protein engagement. Supplementary IHC analysis showed higher IL-4, IL-13, IL-31, and TNF-α immunoreactivity in the model group than in the control group. The high-dose oral group showed significant reductions in all four markers, whereas the low-dose oral group showed numerical decreases that did not reach statistical significance (Supplementary Fig. S4A–D). These findings support an association between oral SIC supplementation and local changes in inflammatory cytokine immunoreactivity.

Fig. 9. Immunohistochemical analysis of AA-pathway-associated, type 2 inflammatory and epidermal-junction proteins after oral SIC supplementation.

A–E Representative skin sections and quantitative analysis of PLA2G4A, ALOX5, PTGR1, GATA3 and ZO-1. (Data represent the mean ± SD. ^#p < 0.05, ^##p < 0.01, ^###p < 0.001 vs control^{, *}p < 0.05, ^**p < 0.01, ^***p < 0.001 vs model, n = 3).

Discussion

SIC is officially approved as a novel food ingredient by the National Health Commission of the People’s Republic of China. It is also recognized as a new cosmetic ingredient by the National Medical Products Administration of the People’s Republic of China. This study evaluated a standardized SIC food material containing 87 characterized constituents and 158 potential odor-active volatile compounds. Chlorogenic acid, syringin, and 1,5-dicaffeoylquinic acid were quantified as representative compositional markers. The selection of these three key markers is justified by their high abundance, good quantifiability, as well as significant biological and regulatory relevance. Extensive literature highlights their potent anti-inflammatory and immunomodulatory properties, particularly their capacity to alleviate skin inflammation and regulate lipid metabolism^24–26. Furthermore, these compounds serve as characteristic phytochemical markers. Specifically, Chlorogenic acid and 1,5-dicaffeoylquinic acid are widely recognized as classic quantitative standards for evaluating the quality of Saussurea involucrata²⁷. Furthermore, standard pharmacokinetic evidence confirms that these three representative markers possess favorable intestinal permeability and achieve robust systemic bioavailability post-oral ingestion, supporting the rationale for oral SIC dietary intervention^28–30.

Despite recent advances, AD management remains constrained by a trade-off between efficacy and safety³¹. Conventional immunosuppressants pose risks of long-term toxicity. Meanwhile, targeted biologics may not fully address the heterogeneous pathology of the disease³². These limitations underscore a critical demand for innovative dietary interventions or functional foods. Such natural alternatives are required to manage the multifaceted inflammation of AD without the severe adverse effects of conventional treatments.

Collectively, our findings support a potential protective effect of SIC against AD-like inflammation under the tested experimental conditions. These effects were manifested by reduced ear swelling³³ and improved histopathology. The infiltration of eosinophils, neutrophils, and mast cells was also significantly suppressed. The pathogenesis of AD is highly complex^2,34. Therefore, such broad anti-inflammatory effects across multiple cell lineages argue against a single-target pharmacological blockade. This observation prompted the investigation of upstream regulatory networks modulated by SIC³⁵.

Integration of the transcriptomic and metabolomic datasets identified AA metabolism as a prominent pathway associated with oral SIC supplementation. qRT-PCR, targeted lipid-mediator profiling, and IHC extended this pathway-level association to selected transcripts, metabolites, and local protein immunoreactivity. Docking and molecular dynamics supplied additional hypothesis-generating structural context for selected compound-protein configurations. The combined evidence is therefore most consistent with network-level AA-associated remodeling rather than a single-target pharmacological mechanism^36,37. The proposed model integrating these measured molecular and tissue-level changes is summarized in Fig. 10.

Fig. 10. Proposed model of SIC-associated changes in DNCB-induced AD-like inflammation.

The left panel summarizes DNCB-associated skin-barrier disruption, arachidonic acid metabolism, and inflammation- and pruritus-associated signaling. The right panel summarizes changes associated with oral SIC, including pathway-associated gene expression, AA-related metabolites, tissue inflammatory markers, and histopathological features. Solid arrows indicate findings measured in this study, whereas dashed arrows indicate proposed or literature-linked mechanisms.

Across the molecular measurements, lower Pla2g4a transcript abundance was accompanied by lower PLA2G4A immunoreactivity and AA abundance. ALOX5 and PTGR1 showed directionally similar changes at the transcript and tissue-immunoreactivity levels, while several downstream lipid mediators were also reduced. This convergence supports an association between oral SIC supplementation and remodeling of an AA-associated network^38–40. It does not demonstrate substrate deprivation, enzyme inhibition, or direct regulation of these proteins by individual SIC constituents.

The MD analysis provided a trajectory-based sensitivity check on docking-derived starting poses. Several systems retained relatively small internal ligand RMSD while showing greater pose RMSD, consistent with reorientation of an intact ligand rather than simple conformational instability. This pattern is most appropriately interpreted as model-dependent configurational behavior. There is no MM/GBSA or other binding free-energy calculation supplied, and the values therefore cannot establish binding, comparative affinity, or target preference.

The evidence depth differed among pathway nodes. PTGES was retained in the docking and qRT-PCR analyses, but its docking profile was comparatively weaker and less consistent across the tested compounds. PTGES was therefore not advanced to MD, and no IHC, protein-abundance, or enzyme-activity measurement was performed for this target. The available data do not support a direct-target or protein-level conclusion for PTGES.

The IHC findings added local protein-level support to the molecular interpretation. Reduced PLA2G4A, ALOX5 and PTGR1 immunoreactivity, together with lower GATA3 and inflammatory-cytokine staining, was consistent with treatment-associated changes in AA-pathway-associated and type 2 inflammatory markers. Increased ZO-1 immunoreactivity indicated altered staining of an epidermal-junction protein. However, IHC is semi-quantitative and does not establish protein activity, direct compound–protein interaction, or functional skin-barrier recovery.

The targeted lipid profile provides biochemical context for the inflammatory phenotype. Primarily, the reduction in eosinophil and neutrophil infiltration stems from the suppression of the LOX pathway. Metabolomic profiling revealed a marked reduction in 5-oxoETE and LTB4⁴¹. These metabolites are potent chemoattractants for eosinophils and neutrophils, respectively^42,43. The attenuation of these critical chemotactic signals impeded the recruitment of inflammatory cells from the circulation to cutaneous tissues⁴⁴. Consequently, the local inflammatory load was significantly attenuated^45,46. Concurrently, the reduction in ear swelling and mast-cell accumulation was associated with changes in COX-pathway metabolites. SIC lowered PGE2 and selected metabolites related to cysteinyl leukotriene synthesis, consistent with modulation of arachidonic acid-derived mediator profiles^47,48.. Through this reduction, the dietary intervention effectively resolved edema induced by vascular exudation^49,50. These associations may help explain the broad inflammatory changes observed after SIC supplementation, but they do not establish causal pathway suppression or a direct antipruritic effect.

These metabolomic changes also provide limited biochemical context for pruritus-associated signaling. Reduced PGD2 and LTB4, mediators linked to mast-cell and leukocyte signaling, may be relevant^51,52. Together with the lower Alox5 expression, the reduced LOX-associated metabolite profile is consistent with attenuation of leukotriene-related signaling. Because basophil-derived leukotrienes can activate sensory neurons and promote acute itch⁵³, this pattern is compatible with reduced basophil-neuron signaling, although that axis was not directly tested.

Reduced 20-HETE, a CYP-derived metabolite implicated in mechanical itch through TRPV1 activation on MrgprA3+ sensory neurons⁵⁴, provides a second pruritus-associated biochemical signal. However, individual CYP isoforms, neuronal activation, and scratching behavior were not measured. The present findings therefore indicate coordinated changes in metabolites associated with mast-cell, basophil, and CYP-related itch pathways, but do not establish a specific molecular mechanism or an antipruritic effect.

On a systemic level, the restoration of the spleen index suggests a remission of general immune inflammation. Atopic dermatitis is widely recognized as a systemic disorder driven by a strong type 2 immune response, rather than a mere localized skin condition⁵⁵. This systemic pathogenesis may explain why oral SIC supplementation showed stronger efficacy than topical application under the currently tested conditions. Mechanistically, topical delivery provides direct intervention at the epidermal lesion. However, its physiological impact is largely constrained by limited transdermal absorption and localized bioavailability. Notably, a relatively high topical concentration of SIC was employed because it is a whole-cell matrix rather than a purified singular monomer, requiring a higher baseline to ensure trace bioactive constituents effectively cross the disrupted epidermal barrier. Crucially, our histopathological and macroscopic observations revealed no aggravation of skin irritation or lesions in the topical group, confirming the localized safety and biocompatibility of the preparation. In contrast, oral administration facilitates efficient gastrointestinal digestion and widespread systemic distribution via the circulatory system. This comprehensive nutritional reach allows the bioactive constituents of SIC to access the cutaneous microenvironment through the dermal vascular bed. Furthermore, it enables the direct modulation of peripheral immune organs and circulating leukocytes. PGE2 acts as a critical cofactor that facilitates Th2 polarization while suppressing Th1 differentiation^56,57. Therefore, the sustained downregulation of systemic PGE2 aids in rectifying the canonical Th2-skewed immune imbalance characteristic of AD. Ultimately, this systemic regulation reverses the compensatory hyperplasia of lymphoid organs. Nevertheless, it must be acknowledged that a direct pharmacokinetic or molecular superiority of the oral route over the topical route cannot be definitively concluded here, given the inherent differences in dosing scales, vehicle formulations, and specific positive controls between the two cohorts. Future studies incorporating rigorous cross-cohort dose-exposure matching are warranted to systematically delineate the pathway-specific bioavailability of these delivery routes.

It is also noteworthy that the prostaglandin reductase gene ptgr1 and its metabolically inactivated product, 13,14-dihydro-15-ketoPGE2⁵⁸, were downregulated parallel to the subsidence of inflammation. This pattern is distinct from the compensatory overexpression typically observed in the AD model group⁵⁹. This finding implies that the protective mechanism of SIC is not driven by enhanced metabolic clearance of inflammatory mediators. Rather, the dietary intervention restricts upstream substrate availability. Through this upstream modulation, SIC facilitates a systemic transition from a stress state characterized by “high inflammation and high metabolism” back to normal physiological homeostasis.

Despite these promising findings, several limitations should be acknowledged. The DNCB-induced BALB/c mouse model cannot reproduce the complex genetic background and chronic relapsing course of human AD. The multi-omics, qRT-PCR and IHC findings are associative, and direct enzymatic interaction, ligand–target engagement and pathway necessity were not tested using biochemical, genetic or rescue experiments. The MD results were derived from one supplied trajectory per system and were analyzed descriptively without independent repeats or binding free-energy estimation. IHC was restricted to the oral-treatment cohort and therefore cannot establish molecular equivalence or difference between oral and topical SIC administration. Moreover, increased ZO-1 immunoreactivity indicates a change in epidermal-junction protein staining but does not demonstrate functional barrier restoration, which would require dedicated measurements such as transepidermal water loss. Moreover, gut microbiota, microbial metabolites, and the gut-skin axis were not assessed; consequently, the contribution of intestinal microbial remodeling to the oral response remains unknown. Future studies combining microbiome profiling, microbial-metabolite measurements, and exposure assessment will be needed to clarify this possibility. Clinical studies and more direct mechanistic experiments are required before the findings can be translated to long-term dietary management of human chronic skin inflammation.

Oral SIC supplementation alleviated macroscopic and histopathological features of DNCB-induced AD-like inflammation and altered peripheral-blood indices in mice. Transcriptomic, metabolomic, and targeted lipid-mediator analyses identified coordinated changes in skin AA metabolism, while qRT-PCR and tissue-level IHC supported corresponding changes in selected pathway-associated and AD-related markers in the oral-treatment cohort. Docking and single-trajectory MD supplied only hypothesis-generating structural context and did not establish direct target engagement or comparative affinity. These data support further evaluation of standardized SIC as a sustainable functional-food ingredient; pathway causality, functional skin-barrier restoration, and clinical relevance require additional validation.

Methods

Preparation of SIC extract

The SIC was supplied by Dalian Practical Biotechnology Co., Ltd., Dalian, Liaoning, China. Briefly, the preparation process was as follows: Explants selected from Saussurea involucrata tissues were induced to dedifferentiate into callus. These calli served as seed cultures and were subcultured under controlled conditions to generate granular cell aggregates. The resulting biomass was harvested, dried, and crushed, followed by micronization to obtain an ultrafine powder for experimental use.

Chemical profiling of SIC via LC-MS and GC-MS

Samples were analyzed using a UPLC system coupled to ESI-QQQ-MS/MS. Separation was performed on an Agilent SB-C18 column (100 × 2.1 mm, 1.8 µm) maintained at 40 °C. The binary mobile phase comprised 0.1% formic acid in water (phase A) and acetonitrile containing 0.1% formic acid (phase B). The gradient increased from 5% to 95% B over 0–9 min, was held at 95% B for 1 min, returned to 5% B over 1.1 min and was re-equilibrated at 5% B until 14 min. The flow rate was 0.35 mL/min and the injection volume was 2 μL. The effluent was introduced into an ESI-triple-quadrupole-linear ion trap (QTRAP) mass spectrometer. The ESI parameters were as follows: source temperature, 500 °C; ion-spray voltage, 5500 V in positive-ion mode and −4500 V in negative-ion mode; GSI, GSII and CUR, 50, 60, and 25 psi, respectively; and CAD gas, high. Scheduled MRM experiments were performed with nitrogen collision gas set to medium. The DP and CE values were optimized for each MRM transition. Data acquisition was divided into retention-time windows so that each transition was monitored during elution of its corresponding analyte.

Gas chromatography was performed using an Agilent 8890 instrument with the injector maintained at 250 °C. Compounds were separated on a DB-5MS capillary column (30 m × 0.25 mm, 0.25 μm). Detailed operating conditions and the calculation of odor activity values (OAVs) are provided in the Supplementary Information.

Quantitative analysis of key bioactive markers via UPLC

Three key bioactive markers in SIC were quantified to establish the material basis of the dietary intervention. Separation was performed on a Phenomenex Gemini C18 column (4.6 × 250 mm, 5 μm) using methanol (solvent A) and 0.1% formic acid in water (solvent B). The gradient was 20% A at 0–15 min, 20–36% A at 15–25 min, 36-37% A at 25–60 min, 37–70% A at 60–85 min, 70–100% A at 85–90 min and 100% A at 90–100 min. The injection volume was 10 μL, the flow rate was 0.9 mL/min, the column temperature was 30 °C and detection was performed at 265 nm.

Animals and dietary intervention

SPF-grade male BALB/c mice (20–22 g) were purchased from Beijing HFK Bioscience Co., Ltd. (License No. SYXK (Jing) 2023-0077) and housed under controlled conditions with a 12-h light-dark cycle and a temperature of 22 °C ± 3 °C. The mice were fed a standardized diet and had access to water throughout the experimental period. The study was approved by the Animal Ethics Review Committee of the Medical Experimental Center, China Academy of Chinese Medical Sciences. To establish the AD-like model and compare intervention strategies, 80 mice were randomly assigned to two distinct experimental cohorts (Fig. 2A, B). The topical intervention cohort comprised five groups (n = 8 per group): normal control, AD model, positive reference (0.1% hydrocortisone butyrate, TY), low-dose topical SIC (10%, TL), and high-dose topical SIC (20%, TH). The dietary supplementation (oral) cohort comprised five groups (n = 8 per group): normal control, AD model, positive reference (0.1% dexamethasone, OY), low-dose oral SIC (0.3 g/kg, OL), and high-dose oral SIC (0.9 g/kg, OH). Notably, the oral doses of 0.3 and 0.9 g/kg in mice translate to a Human Equivalent Dose of approximately 1.46 and 4.38 g/day for a 60 kg adult, respectively. These doses fall well within the feasible range for daily dietary supplementation. For AD modeling, the dorsal fur of all mice was removed one day prior to sensitization. On days 1 and 2, a 2% DNCB solution (dissolved in a 3:1 mixture of acetone and olive oil) was topically applied to the dorsal skin (100 µL) and the right ear (30 µL) of all groups except the normal controls. Following a 3-day resting period, a 1% DNCB solution was applied to the same areas every two days from day 6 to day 18. This was followed by daily DNCB applications until day 25. Successful modeling was confirmed by the emergence of AD-like symptoms, including erythema, crusting, and scaling.

SIC interventions began on day 26 and continued for one week. In the topical cohort, control and model animals received petroleum jelly on the dorsal skin (40 mg) and ears (10 mg), whereas the SIC groups received the corresponding ointments twice daily. In the oral cohort, control and model animals received distilled water (0.1 mL/10 g body weight) by gavage, and SIC groups received the corresponding aqueous suspensions once daily. After the final intervention, blood was collected, and animals were euthanized for the collection of skin, spleen, and ear tissues. The spleen index was calculated as fresh spleen mass divided by body mass. Ear edema was calculated as (right-ear mass - left-ear mass)/left-ear mass. Because the OL group showed the more pronounced phenotypic response within its cohort, its skin samples were selected for transcriptomic and metabolomic analyses. For IHC, skin samples from the control, model, OL, and OH groups were examined. These selections were not intended as a dose- or exposure-matched molecular comparison between oral and topical administration.

Evaluation of AD-like symptoms and phenotypic observation

Skin lesions were photographed throughout the intervention period. Their severity was assessed using a dermatitis scoring system. Three investigators independently scored the lesions every two days, and their scores were averaged to reduce subjective variation. The scoring criteria are provided in Supplementary Table S1.

Histopathological assessment and hematological analysis

EDTA-anticoagulated whole blood was analyzed using an automated hematology analyzer to determine the absolute white blood cell (WBC), eosinophil (EO), monocyte (MONO), and neutrophil (NEUT) counts.

Skin tissues were fixed in 4% paraformaldehyde, dehydrated through a graded ethanol series, and embedded in paraffin. Sections (4 μm thick) were stained with hematoxylin and eosin (H&E) to evaluate tissue architecture and epidermal thickness. Adjacent sections were stained with toluidine blue to visualize mast cells. For quantitative histopathological analysis, skin sections from three mice per group were examined under consistent imaging conditions. Mast cells were counted in eight fields per section at 400× magnification using Image-Pro Plus 6.0.

Immunohistochemistry and image analysis

For tissue-level analysis, paraffin-embedded skin sections from the control, model, OL, and OH groups were examined by IHC (n = 3 mice per group). Sections were deparaffinized in xylene, rehydrated through graded ethanol, and subjected to heat-mediated antigen retrieval in EDTA buffer (pH 9.0) at 80 °C for 20 min. After cooling and washing in PBS (pH 7.4), endogenous peroxidase activity was blocked with 3% H2O2 for 25 min, and nonspecific binding was blocked with 3% bovine serum albumin for 30 min at room temperature. Sections were incubated with primary antibodies at 4 °C overnight, followed by HRP-conjugated goat anti-rabbit secondary antibody (SeraCare, 5220-0336; 1:500) for 50 min at room temperature. Immunoreactivity was visualized with DAB, and sections were counterstained with hematoxylin, dehydrated, and mounted. Primary-antibody suppliers, catalogue numbers, and dilutions are provided in Supplementary Table S2.

Five non-overlapping fields were selected from comparable epidermal and superficial dermal regions for each mouse. Images were analyzed in ImageJ/Fiji using color deconvolution and identical thresholding settings across groups. DAB-positive area was quantified for PLA2G4A, ALOX5, PTGR1, IL-4, IL-13, IL-31, TNF-α, and ZO-1; GATA3 was quantified as the proportion of GATA3-positive nuclei among total nuclei. Fields containing folds, tears, major tissue loss, or non-comparable anatomy were excluded before analysis. The five field-level measurements were averaged to obtain one value per mouse before group-level analysis.

Non-targeted metabolomics

A sample of skin tissue (50 mg) was homogenized and ultrasonicated after being mixed with 500 µL of pre-cooled 75% methanol aqueous solution. Subsequently, 300 µL of dichloromethane was added, and the mixture was vortexed for 3 min. After centrifugation at 12,000 rpm for 10 min at 4 °C, the upper layer (200 µL, strong polarity fraction) and lower layer (150 µL, weak polarity fraction) were collected separately. The solvent in each fraction was evaporated using a rotary concentrator. The strong and weak polarity fractions were then redissolved using 150 µL of acetonitrile/water (4:1) and 100 µL of acetonitrile/water (2:3), respectively.

The chromatographic separation conditions were individually optimized for two different fractions. In the case of the strongly polar fraction, separation was achieved on a Waters ACQUITY UPLC BEH Amide column (2.1 × 100 mm, 1.7 μm) with the column temperature set to 25 °C. The mobile phase system consisted of 0.1% formic acid in water (solvent A) and acetonitrile containing 0.05% formic acid (solvent B). A gradient program beginning with a high proportion of organic solvent was applied: 97% B was maintained from 0 to 1 minute, then decreased to 92% B by 5 min, further reduced to 80% B at 10 minutes, to 60% B at 15 min, and finally to 50% B at 16 minutes, after which the column was re-equilibrated to initial conditions. The mobile phase was delivered at a flow rate of 0.2 mL/min, and the injection volume was 8 μL.

The weakly polar fraction was analyzed under different conditions, employing a Waters ACQUITY UPLC HSS T3 column (2.1 × 100 mm, 1.8 μm) operated at 35 °C. An isocratic mobile phase composition was avoided by implementing a gradient: starting at 5% B and increasing to 55% B during the first 2 minutes, then continuing to 95% B by 10 minutes, where it was held for 5 min. Both solvent A (water) and solvent B (acetonitrile) contained 0.05% formic acid. A flow rate of 0.3 mL/min was applied, and the injection volume was 3 μL. To ensure sample stability, the autosampler temperature was consistently maintained at 7 °C across both analytical procedures.

Mass spectrometric detection was carried out using an electrospray ionization (ESI) source operating in switching mode, allowing for both positive and negative ion acquisition within the same analytical run. Nitrogen served as the nebulizer gas, while ultra-pure helium was used for collision. Common parameters included a capillary temperature of 350 °C and auxiliary gas flow at 10 arb. Sheath gas flow varied: 60 arb for positive mode and 35 arb for negative mode (with S-Lens RF at 60%). Data were acquired in FT mode (resolution: 30,000) across a mass range of m/z 50–1500.

All LC-MS data were processed using Compound Discoverer 3.3 software. Multivariate statistical analyses were performed to identify significant potential biomarkers based on variable importance values (VIP), p-values from t-tests, and fold change (FC) values. Multiple testing corrections using the Benjamini-Hochberg procedure were subsequently applied to control the False Discovery Rate (FDR). The strict identification criteria were established as VIP > 1, raw p < 0.05, and FDR-adjusted q-value < 0.05.

Pathway enrichment analysis for metabolites was undertaken in MetaboAnalyst 6.0 (https://www.metaboanalyst.ca/), employing a standard pathway significance criterion based on raw p < 0.05 to capture comprehensive functional metabolic networks.

Transcriptomic sequencing and data assessment

RNA extraction, library construction, and sequencing services were provided by Metware Biotechnology Co., Ltd. (Wuhan, China), employing the Illumina NovaSeq 6000 platform. After the reference genome and annotation data were acquired, sequence alignment was performed using HISAT to map clean reads to the indexed genome. Gene abundance was determined via featureCounts, followed by the calculation of FPKM values to normalize for gene length. Differential expression analysis was conducted using DESeq2, screening for genes meeting the thresholds of |log2 Fold Change | ≥ 1.0 and FDR-adjusted q-value < 0.05. Furthermore, functional interpretation through GO and KEGG pathway enrichment was achieved using the clusterProfiler R package.

Molecular docking

The three-dimensional structures of syringin (CID 5316860), chlorogenic acid (CID 1794427), and 1,5-dicaffeoylquinic acid (CID 122685) were obtained from PubChem, assigned protonation states at physiological pH (7.0 ± 0.2), and energy-minimized. Protein structures for ALOX5 (PDB 3O8Y), PTGR1 (PDB 9D6W), PLA2G4A (PDB 1CJY), and PTGES (PDB 5T36) were obtained from the RCSB Protein Data Bank. Structures were prepared in AutoDockTools 1.5.7. Non-catalytic water molecules and unrelated heteroatoms were removed, whereas cofactors or structural ions required by the original screening model were retained. Hydrogen atoms and Gasteiger partial charges were added. Docking sites were centered on co-crystallized inhibitors or functionally annotated catalytic residues. AutoDock Vina 1.1.2 was run with an exhaustiveness of 32 and a 30 × 30 × 30 Å grid centered on each active pocket. Binding poses were visualized in PyMOL 2.6.0.

Molecular dynamics simulations

Molecular dynamics (MD) simulations were performed using the Amber/AmberTools platform to examine the dynamic behavior of docking-derived protein–ligand configurations. Nine complexes comprising ALOX5, PTGR1, or PLA2G4A with syringin, chlorogenic acid, or 1,5-dicaffeoylquinic acid were simulated. Initial conformations were generated by an independent AutoDock Vina docking step, and the rank-1 poses were used for system construction. Protein parameters were assigned using the ff14SB force field, whereas ligand parameters were generated with GAFF2 and AM1-BCC partial charges. The ligand net charges were set to 0 for syringin and -1 for both chlorogenic acid and 1,5-dicaffeoylquinic acid. Each complex was solvated using the explicit TIP3P water model in a periodic simulation box and subjected to staged energy minimization, heating, and equilibration before a nominal 100-ns production simulation. Long-range electrostatic interactions were treated using the particle-mesh Ewald method. Trajectories were processed using cpptraj with periodic-boundary reimaging and protein-backbone alignment. Protein and complex RMSD, ligand self-fit RMSD, ligand pose RMSD, protein RMSF, hydrogen-bond counts, radius of gyration, and protein-ligand center-of-mass distances were calculated. The ligand self-fit RMSD described internal conformational change, whereas the ligand pose RMSD described displacement relative to the initial docking pose after protein alignment.

Quantitative real-time polymerase chain reaction

Total RNA isolation from skin tissues was achieved using the E. Z. N. A. Total RNA Kit (Omega Bio-Tek, GA, USA). Subsequently, cDNA generation was performed with the ReverTra Ace® qPCR RT Master Mix (TOYOBO, Osaka, Japan), which includes a genomic DNA elimination step. Expression levels of target genes were assessed via quantitative real-time PCR (qRT-PCR) on an ABI Prism 7900 HT system (Thermo Fisher Scientific, MA, USA). Reactions were performed using SYBR Green Master Mix supplied by TransGen Biotech (Beijing, China). Data analysis was conducted according to the comparative Ct (2^−ΔΔCt) method, with β-actin serving as the reference gene for normalization. The Supporting Information contains the complete sequences of all primers used in this investigation (Table S3).

Targeted metabolomic profiling

Quantification of metabolites was conducted on a QTRAP® 6500 + LC-MS/MS platform (AB SCIEX, CA, USA). Separation of skin sample components was achieved using a Waters ACQUITY UPLC HSS T3 column (2.1 × 100 mm, 1.8 μm). The method was adapted from our previous study²³. Detailed instrument parameters are provided in the Supplementary Information.

Statistical analysis

All experimental results are expressed as mean values with corresponding standard deviations. Statistical evaluation was carried out utilizing GraphPad Prism software (version 9.5). Group comparisons were assessed via one-way analysis of variance (ANOVA), where probability values below 0.05 were deemed to reflect statistically significant differences.

Author contributions

Xinwei Yan: Writing - original draft, Visualization, Methodology, Investigation, Data curation. Yin Chen: Writing - review & editing, Formal analysis. Ping Yu: Investigation, Visualization. Zitong Lu: Writing - review & editing. Jiahui Peng: Supervision. Shuyi Feng: Investigation. Shuangrong Gao: Formal analysis. Jingzhe Li: Supervision. Baolin Bian: Project administration. Yu Liu: Resources. Wenya Gao: Writing - review & editing. Nan Si: Writing - review & editing, Funding acquisition, Conceptualization.

Data availability

The RNA-sequencing data generated in this study are publicly available in the NCBI BioProject database under accession number PRJNA1505169, which includes nine linked BioSample and SRA records. Other datasets generated or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this manuscript, the authors used Gemini to improve readability and language. The authors subsequently reviewed and edited the content and take full responsibility for the published work.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41538-026-01072-3.