Parsley root of P. crispum (Mill.) Fuss (syn. Petroselinum sativum Hoffm., Apiaceae) was purchased from a local herbal market in Shanghai, China, where the plants were traditionally cultivated on an organically composed natural substrate. The plant materials were taxonomically authenticated by Prof. Guixin Chou, Institute of Chinese Materia Medica, Shanghai University of Traditional Chinese Medicine, Shanghai, China.

After air-drying, the roots were ground into a fine powder to obtain the crude extract. To obtain the aqueous extract of parsley (AP), each plant powder (100 g) was macerated in 900 mL of distilled water for 1 h at 95 °C. The HEP was obtained by reflux extracting plant powder using a mixture of ethanol and distilled water (50:50 V/V) in a ratio of 100 g powder to 1,000 mL of hydro-ethanolic solvent for 1 h at 90 °C. Ethanolic extracts of parsley (EP) were obtained by reflux extraction of 100 g plant powder using 1,000 mL of 95% alcohol for 1 h at 80 °C. In all cases, the extracts were filtered (Whatman No. 1), concentrated by rotary evaporation under reduced pressure, and subsequently freeze-dried to constant weight. The extraction yields were as follows: 16.35 g (AP), 12.35 g (HEP), and 7.90 g (EP), corresponding to drug–extract ratios (DER) of approximately 6.12:1, 8.10:1, and 12.66:1, respectively. Dried extracts were stored in airtight containers at 4 °C until use.

Extracts were chemically characterized to ensure batch-to-batch consistency. Flavonoid constituents of HEP extract were initially profiled using Liquid Chromatography-Mass Spectrometry (LC-MS) as previously described (Obmann et al., 2011), enabling the identification of key bioactive compounds. Quantitative analysis of the principal marker flavonoids, isorhamnetin, quercetin, and luteolin, was subsequently carried out using High-Performance Liquid Chromatography (HPLC), with reference standards for each compound. Details of the extraction procedure, chromatographic conditions, and representative results were provided in the Supplementary Methods and Supplementary Tables S1 and S2. The phytochemical evaluation and documentation of the extract followed the GA best-practice and ConPhyMP (Consensus on the Phytochemical Characterization of Medicinal Plant Extracts) reporting guidelines for botanical products (Heinrich et al., 2022), and the corresponding completed GA checklist tables are included as Supplementary Material.

Recombinant human TNF-α and IFN-γ were from Peprotech (Rocky Hill). Tert-Butyl hydroperoxide (t-BHP) and other chemicals were purchased from Sigma-Aldrich. Primary antibodies for Nrf2 (cat. no. 12721), NF-κB P65 (cat. no. 8242), IκBα (cat. no. 4814), p-JAK1 (cat. no. 74129), JAK1 (cat. no. 3344), p-STAT1 (cat. no. 9167), STAT1 (cat. no. 14994), and β-actin (cat. no. 4970) were obtained from Cell Signaling Technology. The primary antibodies against anti-Lamin B (cat. no. ab16048) were purchased from Abcam. Horseradish peroxidase-conjugated secondary antibodies were purchased from Jackson ImmunoResearch Laboratories.

Total polyphenol content (TPC) was determined using the Folin-Ciocalteu method (Pyeon et al., 2021). Briefly, the extract was dissolved in deionized water. Dissolved extract solution (1 mL) was mixed with 9 mL of distilled deionized water (dd H₂O) and treated with 1 mL of Folin-Ciocalteu reagent (Sigma-Aldrich). After reacting at room temperature for 5 min, this solution was mixed with 10 mL of 7% sodium carbonate (Na₂CO₃) and 4 mL of dd H₂O. The mixture stood at room temperature for 90 min, and the absorbance was measured at 750 nm by a fluorescence microplate reader (BioTek Instruments). Gallic acid (GA, Sigma-Aldrich) was used as a standard. The data were expressed as mg GA equivalents (GAE)/g of lyophilized extract powder.

Total flavonoid content (TFC) was determined by the aluminum chloride (AlCl₃) colorimetric assay (Pyeon et al., 2021). The extract was dissolved in deionized water. Dissolved extract solution (1 mL) was mixed with 4 mL of dd H₂O and treated with 0.3 mL of 5% sodium nitrate (NaNO₂). After 6 min at room temperature, 0.3 mL of 5% AlCl₃, 2 mL of 1 M sodium hydroxide (NaOH), and 2.4 mL dd H₂O were added. Additionally, the absorbance was then measured at 510 nm by a fluorescence microplate reader. Rutin (Sigma-Aldrich) was used as a standard. The data were expressed as mg rutin equivalents (CE)/g lyophilized extract powder.

Antioxidant activity of the extracts was evaluated using the DPPH radical scavenging assay and Ferric Reducing Antioxidant Power (FRAP) assay.

HaCaT cells were purchased from The Cell Bank of Chinese Academy of Sciences. Cells were cultured in Dulbecco’s modified Eagle medium (DMEM) (Sigma-Aldrich), supplemented with 10% (v/v) fetal bovine serum (FBS) (Gibco) and 1% penicillin/streptomycin (Gibco) at 37 °C and in a humidified 5% CO₂ incubator.

Cell viability was measured with a Cell Counting Kit-8 (CCK-8) assay (APExBio) according to the manufacturer’s instructions. The HEP solution was prepared by dissolving lyophilized HEP in DMSO to generate a 50 mg/mL stock solution. Prior to cell treatment, this DMSO stock was diluted with culture medium to achieve final treatment concentrations. HaCaT cells were seeded in 96-well plates at a density of 1 × 10⁴ cells per well and incubated for 24 h. Cells were treated for 6 h with different concentrations of the hydro-ethanolic extract of parsley and then treated with 0.4 mM t-BHP for 2 h. After that, the cells were cultured with fresh medium supplemented with the same concentrations of HEP for another 24 h. 15 μL of CCK-8 reagent (5 mg/mL) was then added to cells and incubated for 2 h at 37 °C. The optical density at a wavelength of 450 nm (OD₄₅₀) was measured using a microplate reader (BioTek Epoch). Each measurement was performed in triplicate. Cell viability was expressed relative to that of the vehicle control group.

Intracellular ROS levels in t-BHP-stimulated HaCaT cells were estimated by DCFH-DA fluorescence staining method. HaCaT cells were seeded in 6-well plates at a density of 1 × 10⁵ cells/well for 24 h. Cells were pretreated with different concentrations of the HEP for 6 h and then stimulated with t-BHP for 2 h. Subsequently, cells were further incubated with 10 μM DCFH-DA (diluted in fresh DMEM) for 30 min at 37 °C. DCFH-DA labeled intracellular ROS were observed under an Olympus fluorescent microscope. Fluorescence intensity of the DCFH-stained cells for each condition was quantified by Olympus Softimage solution software (Olympus Imaging America), and data were represented as fold change relative to untreated control values.

Total protein was extracted from HaCaT cells using cell lysis buffer (Beyotime Biotechnology, Shanghai, China). Nuclear and cytoplasmic protein fractions were extracted from HaCaT cells using an extraction reagent kit (Beyotime Biotechnology, Shanghai, China), according to the manufacturer’s instructions. The protein concentration was determined by a modified BCA protein assay (Beyotime Biotechnology, Shanghai, China). About 20 μg protein was subjected to electrophoresis on an SDS–polyacrylamide gel, and transferred onto a polyvinylidene difluoride (PVDF) membrane. The membranes were incubated with a blocking solution (5% BSA dissolved in Tris-buffered saline with 0.05% TWEEN-20 (TBST) for 30 min at room temperature and then incubated overnight with a primary antibody (dilution, 1:1,000 in TBST) at 4 °C. After washing thrice with TBST, the blots were incubated with a horseradish peroxidase-conjugated secondary antibody (West Grove) (dilution, 1:2,000) for 2 h at room temperature and washed again thrice with TBST. The blots were visualized using a SuperPico ECL Chemiluminescence Kit (Nanjing Vazyme Biotech). Densitometry was performed using the ImageJ software.

Total RNA was extracted from HaCaT cells using an RNA isolation kit (Yeasen Biotechnology Co., Ltd., Shanghai, China) according to the manufacturer’s instructions. For each sample, 1 µg total RNA was reverse transcribed into cDNA using PrimeScript RT reagent Kit (Accurate Biotechnology Co., Ltd., Hunan, China). qPCR was performed on the quantum Studio 6 real-time PCR system (Thermo Fisher Scientific, Inc., Waltham, MA, United States) using SYBR Green Master Mix (Accurate Biotechnology Co., Ltd., Hunan, China). The sequences of used primers are presented in Table 1. The relative messenger RNA (mRNA) level of each target gene was quantified using the 2^−ΔΔCT method and normalized to that of β-actin.

All in vivo studies were approved by Institutional Animal Care and Use Committee of Shanghai University (project approval number ESSHU2024-003). After 1 week of acclimation, mice were randomly divided into the following four groups: (1) Control (Ctrl): Untreated mice receiving vehicle only; (2) 2,4-dinitrofluorobenzene (DNFB): Mice treated with DNFB to induce AD-like dermatitis; (3) DNFB+ dexamethasone (DEX): DNFB-treated mice receiving DEX (positive control); (4) DNFB+HEP: DNFB-treated mice receiving HEP. The hapten-induced AD model was established using a two-phase sensitization protocol (Yang et al., 2023). On day 0, mice were sensitized by topical application of 25 μL of 1% DNFB (Sigma, St. Louis, MO, United States) dissolved in acetone/olive oil (4:1 v/v) to the shaved abdominal skin (Figure 8A). On day 5, mice received challenge applications of 5 μL of 0.4% DNFB solution to both inner and outer ear surfaces to elicit localized inflammatory responses (Ding et al., 2020). Given the rapid anti-inflammatory action of DEX during peak inflammation, HEP was administered prophylactically on Day 3 post-induction (1% DNFB), whereas DEX intervention commenced at the inflammatory peak on Day 5. Briefly, HEP group: Mice received daily pretreatment with HEP (50 mg/kg) for two consecutive days prior to 0.4% DNFB challenge. DEX group: Mice received daily subcutaneous injections of DEX (2 mg/kg/day) from day 5. All treatment groups except controls received a final 0.4% DNFB application on day 5 to maintain inflammatory stimulation. Disease severity was evaluated using a standardized scoring system applied by blinded observers (Shi et al., 2019). The severities of erythema (hemorrhage), edema, erosion, and dryness in individual mice were scored as 0 (absent), 1 (mild), 2 (moderate), and 3 (severe) in a blinded manner. Ear thickness was measured at 0, 8, 24, and 48 h after DEX administration with a digital caliper. Ear thickness was calculated by four groups compared with Ctrl group. The tissue samples were collected on day 7, and used for subsequent experiments.

Mouse ear skin tissues were collected and fixed with 4% paraformaldehyde (PFA) in PBS overnight before paraffin embedding, and sectioned. Tissue sections (4 μm) were stained with hematoxylin and eosin (H&E) and toluidine blue (TB, Biosharp). Ear epidermal thickness was taken using ImageJ software.

The expression of CD3 and CD68 was detected using immunohistochemistry staining as described previously (Tang et al., 2024; Liu et al., 2024). The tissues sections underwent heat-induced epitope retrieval at 95 °C for 20 min. After blocking with 3% BSA, the slides were stained with primary antibodies: CD3 (Abcam: ab16669), CD68 (Cell Signaling Technology: #97778) overnight at 4 °C. Subsequently, sections were incubated with HRP-conjugated secondary antibody (Goat anti-rabbit IgG) for 1 h at RT. DAB substrate kit was used to detect the signal. Images were taken using the Olympus Microscope IX71 (Olympus, Tokyo, Japan). Five visions were counted per tissue section used for measurement. The mean intensity and cell quantity of skin tissue was quantified from photographs using ImageJ software with the IHC Toolbox.

All data were expressed as a mean ± standard deviation (SD) from three independent experiments. Comparisons among groups were carried out using one-way analysis of variance followed by Dunnett’s post hoc test using GraphPad Prism 6.0 software (Graphpad Software). Statistical significance was set at p < 0.05.

The analysis of TPC and TFC revealed significant variations among AP, EP and HEP. Notably, HEP demonstrated the highest concentration of both polyphenols (43.87 ± 0.09 mg GA/g extract) and flavonoids (57.76 ± 0.30 mg rutin/g extract) (Table 2), indicating superior extraction efficiency of bioactive constituents. LC-MS profiling was employed to identify the major flavonoid constituents of HEP extract (Supplementary Methods). HPLC was used to quantify key marker compounds, including isorhamnetin, quercetin, and luteolin. The results from HPLC corroborated the LC-MS findings, confirming comparable levels of these flavonoids across analytical methods (Supplementary Tables S1 and S2). These findings suggest that the hydro-ethanolic extraction method is more effective in extracting TPC and TFC, the key bioactive metabolites from parsley.

The antioxidant capacity of parsley extracts was evaluated using DPPH and FRAP assays. Among the extracts, HEP showed the highest DPPH scavenging activity with an IC₅₀ of 113.467 μg/mL, while AP and EP had higher IC₅₀ (221.800 μg/mL and 239.067 μg/mL, respectively) (Figure 1A; Table 3), indicating a decreased potency in scavenging DPPH radicals for the latter two extracts. Similarly, the FRAP assay demonstrated that HEP had the greatest ferric reducing antioxidant power with an EC₅₀ of 96.42 μg/mL, compared to AP (186.50 μg/mL) and EP (164.97 μg/mL) (Figure 1B; Table 3). These findings highlight the superior antioxidant potency of HEP, which correlates with its higher TPC and TFC content in parsley extracts.

Given the superior potential of HEP, our following experiments mainly concentrated on investigating its activities. The possible cytotoxic effects of HEP on HaCaT cells were first evaluated via CCK-8 assay. The results revealed that HEP at concentrations of 25 μg/mL or lower did not exhibit any effect on the cell viability over a 24-h period (Figure 2A), excluding the cytotoxicity on HaCaT cells. Therefore, further experiments were performed at concentrations of 1, 5, and 10 µg/mL of HEP. t-BHP is an organic peroxide that can be metabolized by cytochrome P450 in cells to generate peroxyl and alkoxyl radicals or detoxified to tert-butanol, both of which can inflict oxidative damage to cells. t-BHP is a commonly used trigger to induce oxidative stress in vitro and in vivo (Taffe et al., 1987; Bellomo, Thor, and Orrenius, 1984; Kučera et al., 2014; Pinho, Oliveira, and Cunha-Oliveira, 2025). To assess the protective effect of HEP on t-BHP-induced oxidative damage, HaCaT cells were pretreated with HEP (1, 5, and 10 μg/mL) for 6 h and then exposed to 0.4 mM t-BHP. The CCK-8 assay demonstrated that t-BHP stimulation inhibited cell viability, and HEP pretreatment alleviated its inhibition (Figure 2B), suggesting that HEP protected cells from t-BHP-induced oxidative damage.

Excessive ROS contributes to the skin inflammation and disruption of skin barrier (Woodby et al., 2020; Sies and Jones, 2020). Since HEP alleviated the t-BHP-induced oxidative damage, we therefore tested whether HEP alleviated the t-BHP-mediated ROS production. DCFH-DA is a fluorescent indicator. Upon oxidation by ROS, it converts into DCF and emits green fluorescence. Hence, fluorescence intensity in cells can be used to indicate the ROS level. HaCaT cells were pretreated with indicated concentrations of HEP (1–10 μg/mL), following with t-BHP treatment. Compared to the untreated control cells, t-BHP stimulation alone triggered a remarkable increase in fluorescence intensity. However, this effect was dramatically alleviated by HEP-pretreatment in a concentration-dependent manner (Figures 3A,B), indicative of the decreased ROS level. Hence, HEP exhibited antioxidant activity in t-BHP stimulated HaCaT cells.

Nrf2 signaling pathway has an important role in the regulation of oxidative responses (Liu et al., 2022). Therefore, we wondered whether HEP exerted antioxidant activity via activating Nrf2. Nuclear-translocated Nrf2, the activated form, was investigated using Western blotting. As shown in Figure 4A, HEP treatment prior to t-BHP stimulation increased nuclear and total Nrf2 protein levels compared to the t-BHP-only group, indicating that HEP treatment resulted in Nrf2 accumulation and nuclear translocation (Figure 4A).

Nrf2 in the nucleus promotes the transcription of multiple antioxidant genes, including superoxide dismutase (SOD) and catalase (CAT). To consolidate the transcription activity, we analyzed the expression of its target genes via qPCR. HEP treatment increased mRNA expression levels of SOD and CAT in t-BHP-stimulated HaCaT cells (Figure 4B). These findings support the notion that HEP activates Nrf2 signaling pathway.

To evaluate the anti-inflammatory activity of HEP, the inhibitory effect of HEP on the expression of TNF-α/IFN-γ-stimulated pro-inflammatory cytokines in HaCaT cells were assessed using qPCR. As shown in Figure 5, TNF-α/IFN-γ stimulation significantly increased the mRNA expression levels of IL-33, IL-6, and IL-8 in HaCaT cells, compared to the untreated control group, whereas HEP treatment markedly suppressed the mRNA expression levels of these proinflammatory cytokines in TNF-α/IFN-γ-stimulated HaCaT cells. These findings indicate that HEP possesses anti-inflammatory property.

JAK/STAT and NF-κB signaling pathways contribute to TNF-α/IFN-γ-mediated inflammatory responses. Given that HEP inhibited the expression of TNF-α/IFN-γ-induced inflammatory cytokines, we examined its effect on JAK/STAT and NF-κB pathways. Western blot analysis revealed that stimulation of HaCaT cells with TNF-α/IFN-γ markedly induced the phosphorylation of JAK1 and STAT1, and this effect was inhibited by HEP pretreatment (Figure 6A). In addition, TNF-α/IFN-γ stimulation induced degradation of the inhibitor of NF-κB (IκBα) and nuclear translocation of NF-κB subunit p65 (NF-κB p65), whereas HEP treatment inhibited the TNF-α/IFN-γ-induced degradation of IκBα and also prevented the nuclear localization of NF-κB p65 (Figure 6B). These results collectively revealed that HEP inhibited the activation of JAK/STAT and NF-κB signaling pathways in TNF-α/IFN-γ-stimulated HaCaT cells, thereby accounting for its suppressive impact on inflammatory reactions.

To determine the protective potential of HEP on inflammation-induced skin barrier dysfunction, the effect of HEP on the expression of skin barrier-related proteins in TNF-α/IFN-γ-stimulated HaCaT cells was assessed using qPCR. TNF-α/IFN-γ stimulation downregulated mRNA expression levels of genes related to skin barrier formation, including Claudin-1 and Filaggrin (FLG), compared to the untreated control group, whereas HEP treatment markedly increased both mRNA expressions in TNF-α/IFN-γ-stimulated HaCaT cells (Figure 7).

To evaluate the therapeutic potential of HEP in AD, we utilized a well-established DNFB-induced AD model that recapitulates key features of human AD (Figure 8A) (Li et al., 2013). Following DNFB challenge, characteristic inflammatory responses developed progressively, with edema and erythema manifesting at 8, 24, and 48 h post-application. Comprehensive clinical assessment using standardized dermatitis scoring revealed that both the positive control dexamethasone (DEX) and HEP treatment significantly reduced overall dermatitis severity compared to DNFB-treated controls (Figure 8B). H&E staining demonstrated that DNFB-induced lesions were characterized by pronounced epidermal hyperplasia and extensive dermal immune cell infiltration, hallmarks of inflammatory skin disease. In contrast, both DEX- and HEP-treated mice exhibited marked amelioration of these pathological changes, with substantially reduced inflammatory cell accumulation and normalization of epidermal thickness (Figure 8C).

To further characterize the inflammatory response, we performed immunohistochemical analysis targeting key immune cell populations, including macrophages (CD68), T lymphocytes (CD3), and mast cells (toluidine blue, TB). DNFB-induced skin lesions exhibited significantly increased CD68⁺ macrophages, confirming robust innate immune activation, while both DEX and HEP treatment substantially decreased CD68⁺ macrophage number, indicating the reduced macrophage infiltration and inflammatory responses. Similarly, CD3⁺ T-cell accumulation and TB-positive mast cell numbers were markedly decreased in both treatment groups compared to DNFB controls (Figure 8D). Collectively, these findings indicate that HEP exhibits potent anti-inflammatory properties in vivo, effectively attenuating multiple pathological features of AD-like skin inflammation through suppression of both innate and adaptive immune responses.