Based on the gut–heart axis: Polygonum capitatum improves atherosclerosis by modulating gut microbiota and TMAO, supporting MCPIP1/p53-associated endothelial protection

Abstract

Polygonum capitatum (PC), known as “Tou Hua Liao” (Chinese name), is an essential source of Hmong medicinal plants, which has been used for treating various human diseases. This study examined whether PC has lipid-lowering and anti-atherosclerotic effects and explored the underlying mechanisms. We focused on PC’s influence on gut microbiota–derived metabolites. First, we analyzed animal-derived serum containing PC components and the botanical compounds of PC by UPLC-MS/MS to identify potential bioactive constituents. Second, we treated high-fat diet–fed hamsters with PC to determine whether the treatment improved plasma lipids and attenuated atherosclerosis progression. We then assessed PC’s effects on the gut microbiota by 16S rDNA sequencing and performed fecal microbiota transplantation in hamster models. Finally, we used human umbilical vein endothelial cells (HUVECs) to probe molecular mechanisms by which PC might inhibit oxidative stress and apoptosis. In a diet-induced atherosclerotic hamster model, PC treatment reduced atherosclerosis by decreasing lipid accumulation, oxidative stress, and apoptosis, and it restored gut microbiota balance while markedly lowering the abundance of TMAO-producing bacteria. PC also exerted antioxidant and anti-apoptotic effects and inhibited endothelial apoptosis via an MCPIP1-dependent mechanism. Together, these results indicate that PC suppresses atherosclerosis through two likely pathways: reduction of gut microbiota–derived TMAO production and inhibition of oxidative stress–driven endothelial apoptosis. Network pharmacology analysis of PC-specific blood-absorbed components supports these findings.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13020-026-01355-7.

Introduction

Atherosclerosis (AS), the primary pathological basis of cardiovascular disease [1], remains the leading cause of major cardiovascular events and mortality among residents in both urban and suburban China, attracting widespread attention [2]. The disease involves multiple mechanisms, including vascular dysfunction, inflammatory responses, oxidative stress, and lipid metabolism disorders [3]. Endothelial cells, as fundamental components maintaining the integrity and permeability of the vascular intima, play a crucial role in resisting AS development. However, prolonged inflammatory stimulation and oxidative stress can lead to endothelial injury and dysfunction [4]. When blood flow continuously interacts with damaged endothelium, lipid infiltration and atherosclerotic plaque formation are promoted [5]. This process represents both the initiating event and a hallmark in the emergence and progression of AS.

The gut–heart axis is a complex, bidirectional communication network that links the gut microbiota and its metabolites to cardiovascular disease. Trimethylamine N-oxide (TMAO), a metabolite produced in a microbiota-dependent manner, occupies a central position in this network and has been identified as an independent risk predictor for atherosclerosis [6]. Studies show that a high-fat, high-choline diet elevates circulating TMAO by altering the gut microbiota, which disrupts the endothelial oxidative–antioxidative balance [7] and promotes reactive oxygen species (ROS) accumulation [8]. This ROS accumulation induces endothelial expression of interleukin-1β (IL-1β), monocyte chemoattractant protein-1 (MCP-1), and tumor necrosis factor-α (TNF-α) [9, 10], thereby forming a cascade involving inflammation, oxidative stress, and apoptosis [11, 12]. These processes culminate in endothelial dysfunction, a major contributor to atherosclerosis [13]. However, the mechanisms by which TMAO modulates key downstream apoptotic nodes—such as dynamic changes in the tumor protein p53 (p53) pathway—remain incompletely understood, and comprehensive, systematic evidence is lacking.

While TMAO has been linked to endothelial oxidative stress and apoptosis, whether it functions as a key mediator or a central regulatory hub in these processes remains unclear. Noticeably, monocyte chemoattractant protein-induced protein 1 (MCPIP1), induced by MCP-1 and other mediators, is located within endothelial cells [14, 15] and plays a crucial role in regulating angiogenic balance and cardiovascular homeostasis [16, 17]. Previous research demonstrated that MCPIP1 negatively regulates the expression of TNF-α and IL-1β, thereby exerting anti-inflammatory effects [18]. However, these inflammatory factors also significantly contribute to apoptosis. Currently, the precise regulatory mechanisms involved in TMAO-mediated apoptosis of vascular endothelial cells remain incompletely understood. Hence, novel therapeutic approaches aimed at reducing endothelial cell damage are of significant importance for the prevention and treatment of AS.

In recent years, Statins remain the primary therapeutic option for AS, effectively reducing Low-Density Lipoprotein Cholesterol(LDL-C) levels and inhibiting disease progression. However, statin therapy also carries potential risks, including muscle toxicity, liver dysfunction [19], and disturbances in glucose metabolism [20]. Traditional Chinese medicine has demonstrated potential suitability for treating AS. Specifically, Polygonum capitatum (PC), a prominent representative of Miao ethnomedicine from Guizhou, China, has received considerable attention due to its unique pharmacological properties. Studies have indicated that PC is abundant in flavonoids, phenolic acids, organic acids, alcohols, aldehydes, esters, and other bioactive components [21]. Its preventive and therapeutic efficacy is associated with anti-inflammatory, antioxidant, and antibacterial activities [22]. Notably, flavonoids and phenolic acids extracted from the entire plant exhibit excellent reactive oxygen species-scavenging capabilities [23]. Flavonoid components also significantly improve lipid metabolism in AS rat models and effectively inhibit the release of pro-inflammatory factors such as TNF-α and IL-1β [24]. Furthermore, combining drug-containing serum metabolomics with liquid chromatography-based phytochemical analysis facilitates more precise predictions of herbal drug targets and pathways through network pharmacology. The golden hamster model used in this study exhibits distinctive variations in lipid metabolism and reproductive cycles not observed in other small animal models. Its metabolic and cardiovascular systems closely resemble those of humans [25], and both golden hamsters and humans specifically express CETP in the liver [26]. Aortic plaques formed under high-fat conditions in golden hamsters exhibit characteristics similar to those in human lesions [27].

Currently, the complex interplay between gut microbiota remodeling, TMAO metabolic regulation, and MCPIP1-mediated endothelial protection remains incompletely characterized. It is also unclear whether endothelial MCPIP1 modulates TMAO-induced oxidative stress and apoptosis, and the role of MCPIP1 in p53-related apoptotic signaling requires further validation. Components rich in flavonoids and phenolic acids identified in PC extract show anti-inflammatory, antioxidant, and lipid-modulating properties, which position PC as a candidate for AS intervention. However, a systematic evaluation of PC’s effects on AS and the endothelial molecular mechanisms involved is lacking. Therefore, this study assessed the anti-atherosclerotic effects of PC in a Syrian golden hamster AS model. We combined microbiota and TMAO-related phenotyping with mechanistic validation and omics/network analyses. Finally, using Surface Plasmon Resonance(SPR) and Co-Immunoprecipitation (Co-IP), we tested a potential TMAO–MCPIP1–p53 axis and probed MCPIP1/p53-related signaling as a mediator of PC’s effects.

Methods and materials

Preparation of PC and medicinal serum

Polygonum capitatum (PC) was soaked in 600 µL of methanol supplemented with 4 ppm 2-Amino-3-(2-chloro-phenyl)-propionic acid. In a tissue grinder, the sample was homogenized at 55 Hz for 60 s in presence of steel balls, followed by 15-min ultrasonication at room temperature, and a subsequent 10-min centrifugation at 4 ℃ and 12,000 rpm. Finally, the supernatant was filtered through a 0.22-μm filter and loaded to UPLC-MS/MS.

Thaw the Medicinal serum at 4 ℃, vortex the sample for 1 min after thawing, and mix evenly; Accurately transfer an appropriate amount of sample into a 2 mL centrifuge tube; Add 300 µL of 2-chloro-l-phenylalanine (4 ppm) solution prepared with 80% methanol water and vortex for 1 min; Centrifuge for 10 min at 12,000 rpm and 4 °C, filter the supernatant by 0.22 μm membrane and transfer into the detection bottle for UPLC-MS/MS detection. The samples were processed by PANOMIX Biomedical Technology, Suzhou.

Liquid chromatography and mass spectrum conditions

Liquid chromatography analysis was performed on a Vanquish UHPLC system (Thermo Fisher Scientific, USA) using an ACQUITY UPLC^® HSS T3 column (2.1 × 100 mm, 1.8 µm; Waters, USA) maintained at 40 °C. The flow rate was 0.3 mL/min with an injection volume of 5 μL. For ESI(+) mode, the mobile phase consisted of 0.1% formic acid in water (A2) and 0.1% formic acid in acetonitrile (B2); for ESI(-) mode, 5 mM ammonium formate (A3) and acetonitrile (B3) were used. The gradient elution program for both modes was as follows: 0–1 min, 8% B; 1–8 min, 8%–98% B; 8–10 min, 98% B; 10–10.1 min, 98%–8% B; 10.1–12 min, 8% B.

Mass spectrometry was conducted on an Orbitrap Exploris 120 instrument (Thermo Fisher Scientific, USA) with an ESI source operating in full MS-ddMS² mode. Key settings included: sheath gas pressure, 40 arb; aux gas flow, 10 arb; spray voltage, ± 3.50 kV (ESI +)/−2.50 kV (ESI-); capillary temperature, 325 °C; MS¹ scan range, m/z 100–1000 at 60,000 FWHM; up to 4 dependent MS² scans per cycle at 15,000 FWHM; normalized collision energy, 30%; dynamic exclusion set to automatic. Liquid chromatography and Mass spectrum was conducted by PANOMIX Biomedical Technology, Suzhou.

Network pharmacology prediction

Bioactive components of PC were identified by integrating UPLC-MS/MS data with the TCMSP repository (Traditional Chinese Medicine Systems Pharmacology, http://bionet.ncpsb.org.cn/batman-tcm/index.php) using pharmacokinetic thresholds (DL ≥ 0.18, OB ≥ 30%). Targets listed in the "Related targets" column for qualifying components were extracted. For components not present in TCMSP, 2D structures were obtained from PubChem (https://pubchem.ncbi.nlm.nih.gov/) and screened in the ADME database (high gastrointestinal absorption and drug-likeness > 2) (http://www.swissadme.ch/). Qualified bioactive components were compiled, and their potential targets were predicted with SwissTargetPrediction (http://swisstargetprediction.ch/) by inputting SMILES notations, restricting predictions to Homo sapiens and using Probability > 0 as the threshold. Finally, bioactive components and their corresponding targets were deduplicated and tabulated.

Potential atherosclerosis targets were identified by searching GeneCards (https://www.genecards.org/) and OMIM (https://www.omim.org/) with the query term "atherosclerosis." From GeneCards, targets with relative scores above the median were retained, and putative false positives were excluded using OMIM. Endothelial cell apoptosis targets were then intersected with predicted PC targets. The overlapping genes were analyzed in STRING (https://string-db.org); protein–protein interaction networks with confidence score ≥ 0.7 were generated and imported into Cytoscape 3.9.1, where core proteins were identified using the CytoNCA plugin based on degree, closeness, and betweenness centralities (DC/CC/BC). Gene ontology (GO) analysis was performed in DAVID 6.8 (https://david.ncifcrf.gov) (p < 0.05) to examine the biological processes, cellular components, and molecular functions of the overlapping human targets. Pathway analysis was conducted using the Kyoto Encyclopedia of Genes and Genomes (KEGG).

Molecular docking

Chemical structures of PC bioactive constituents were acquired from PubChem and TCMSP databases, and protein structures of core targets were retrieved from the PDB (https://www.rcsb.org/). Molecular docking of bioactive compound and protein targets were computed by AutoDock Vina 1.1.2, as ligand-receptor interactions. Predicted structures were visualized in PyMOL. Meanwhile, AlphaFold3 (https://alphafoldserver.com/) was employed to predict joint structures. Protein sequences or SMILES strings were uploaded to the AlphaFold3.

Drug preparation

The PC was obtained from Relinqing Capsules (No. Z52020319, Hongkang Pharmaceutical). For cell culture applications, PC was dissolved in complete medium to prepare stock solution of 20 mg/mL, and stored at 4 °C. For animal administration, PC was dissolved in ultrapure water at pre-defined concentrations and administered immediately.

Animal model and drug intervention

This manuscript did not involve human participants, and all the animal experiments were approved by the Ethics Committee of Guizhou University of Traditional Chinese Medicine (No.20250501001, Guiyang, China). Golden hamsters (Male, 120–140 g) were purchased from Liaoning Changsheng Biotechnology Co., LTD., and maintained in the Animal Research Institute of Guizhou University of Traditional Chinese Medicine. After a one week adaptive feeding, hamsters were randomly divided into 6 groups and fed with a high-lipid and high-choline diet (cholesterol 2%, lard 24.5%, choline chloride 3%, basal feed 70.5%) for 8 weeks, except for Control group. Subsequently, PC at different dosages were fed to hamster AS model for another 8 weeks. In detail, Control group was given regular feed; Model group was given a high-fat, high-cholesterol diet, without PC treatment; PC-L low dose group was given PC at 6.75 mg/100 g; PC-M median dose group was given PC at 13.5 mg/100 g; PC-H high dose group was given PC at 27 mg/100 g; Atorvastatin was applied as positive control. Animals were anesthetized by intraperitoneal injection of citalazine (10 mg/kg) at week 16.

16S rDNA sequencing

Fecal microbial DNA was extracted using the Fecal Genome DNA Extraction Kit (AU46111-96, BioTeke, China) and quantified with Qubit (Invitrogen, USA). The V3-V4 region of the 16S rRNA gene was amplified with primers 341F/805R, and the resulting libraries were purified with AMPure XT Beads (Beckman Coulter Genomics, USA) and quality-checked using Kapa library quantitative kits (Kapa Biosciences, USA) and an Agilent 2100 Bioanalyzer (Agilent, USA) prior to sequencing on the Illumina NovaSeq 6000 platform (PE250). After demultiplexing, primer removal with cutadapt (v1.9), and merging of paired-end reads with FLASH (v1.2.8), raw sequences were quality-filtered using fqtrim (v0.94) to remove low-quality reads (quality score < 20), short sequences (< 100 bp), and reads with excessive ambiguous bases. Chimeras were filtered with Vsearch (v2.3.4), and amplicon sequence variants (ASVs) were generated using DADA2. Taxonomic annotation was performed in QIIME2 using the SILVA and NT-16S databases. Alpha and beta diversity analyses were conducted in QIIME2, and differential abundance at the genus level was assessed using the Kruskal–Wallis test (P < 0.05). LEfSe analysis was applied to identify biomarkers (LDA ≥ 3.0, P < 0.05). All analyses were performed using R (v3.4.4).

Silencing MCPIP1

MCPIP1-siRNA and NC-siRNA (negative control) were synthesized by Shanghai GenePharma, as shown in Table 1. HUVECs cells were seeded at density of approximately 2 × 10⁵ per well, in 6-well cell culture plate. MCPIP1 siRNA was transfected at 70–80% confluency and analyzed by Western blot 24 h post transfection. Validated siRNAs were adopted to construct adenoviral shRNA expression vector (hMCPIP1-shRNA). Recombinant adenovirus were produced via the transient-transfection of HEK293 cells. Viral particles were collected after multiple propagation and purified by gradient centrifugation (AD-hMCPIP1-shRNA).

Table 1.

Sequence of MCPIP1 SIRNA

siRNA name	5ʹ → 3ʹ
MCPIP1-HOMO-1001	GUGUCCCUAUGGAAGGAAATT UUUCCUUCCAUAGGGACACTT
MCPIP1-HOMO-1262	GGGUCUAACACAGACCUAUTT AUAGGUCUGUGUUAGACCCTT
MCPIP1-HOMO-1747	CCAGCGUGUAUACUAAGCUTT AGCUUAGUAUACACGCUGGTT
MCPIP1-HOMO-1848	GAGAUCCUCUCCUACAAGUTT ACUUGUAGGAGAGGAUCUCTT

Vascular endothelial cell apoptosis

Vascular endothelial cell apoptosis was induced by treating HUVECs with 300 μL TMAO (HY-116084, MedChemExpress) for 24 h[28]. For the siRNA-MCPIP1 group, cells were first treated with virus for 24 h, pretreated with PC for 30 min, and then co-incubated with TMAO for an additional 24 h. After interventions, cells were collected, and proteins were extracted. CCK8 Cell Proliferation Assay Kit (HYCCK8-500 T, HYCEZMBIO) was employed to examine cell viability.

ELISA

The ELISA kits were obtained from Jingmei Bioengineering. For the detection of IL-1β, MCP-1, ROS, Malondialdehyde (MDA), Superoxide Dismutase (SOD) or TMAO Follow the kit instructions, the signal were developed and measured by microplate reader (1510-01061C, Thermo Fisher Scientific), at the 450-nm Optical Density (OD).

Western blot

Hamster aorta of 20 mg were resected and cut into small pieces, then extracted by High-Efficiency RIPA Buffer (R0010, Solarbio) by homogenization in a tissue grinder. For complete lysing, the sample were incubated on ice for 30 min, and then quantified by BCA Protein Assay (PC0020, Solarbio). Seperated by SDS-PAGE (10% PAGE red gel ultra-fast formulation kit, G2043-50 T, Servicebio), the proteins were transferred onto PVDF membrane. The membranes were then sequentially blocked with Blocking solution (QuickBlock^™, P0233, Beyotime), probed using primary antibodies and further exposed to secondary antibody. Primary antibodies are consisted of MCPIP1(25009-1-AP, Proteintech Group), p53(60283-2-Ig, Proteintech Group), Mouse Double Minute 2 homolog (MDM2)(A23388, Abclonal), Inducible Nitric Oxide Synthase (iNOS) (AF0199, Affinity), Cysteine-aspartic acid protease-3 (Caspase-3) (AF7022, Affinity), 47 kDa phagocyte oxidase subunit (p47phox)(DF6282,Affinity), NADPH Oxidase 2 (NOX2)(A19701,Abclonal), Endothelial Nitric Oxide Synthase (eNOS)(27120-1-AP, Affinity). Protein were detected by ECL substrate reagent (KF8003,affinity) and visualized with ChemiDoc XRS imager (Bio-Rad).

Flow cytometry

Cells were collected and incubated with DHE diluted in 1 mL serum-free medium for 1000 times, at 37 °C for 20 min with gentle agitation at 3-min intervals (Dihydroethidium, S0063, Beyotime). After three washes with serum-free medium, reactive oxygen species were analyzed by flow cytometry (cytoFLEX, Beckman Coulter). Cells were collected and stained with 5 μl of 7-AAD and 5 μl Annexin V-APC sequentially (Annexin V-APC/7-AAD Detection Kit, KGA1026, KeyGen Biotech). Apoptotic Cell populations were then analyzed by flow cytometer.

Serum TC, TG, LDL-C and HDL-C test

The assay kits were obtained from Nanjing Jiancheng Biological Engineering, to assess serum Total Cholesterol (TC) (A110-1-1), Triglyceride (TG) (A111-1-1), LDL-C (A112-1-1), andHigh-Density Lipoprotein Cholesterol (HDL-C) (A113-1-1). Follow the kit instructions, working fluid, ultrapure water, standard substance, and serum sample of 2.5 μl were added to each well in sequential, and incubate at 37℃. The colorimetric signal was measured by microplate reader.

Pathological tissue section examination

Aortic tissues were routinely dehydrated, cleared, and embedded in paraffin or embedding medium. The embedded tissues were then sectioned and stained with Hematoxylin and Eosin (H&E) or Oil Red O, respectively, followed by observation and imaging under an optical microscope.

Immunofluorescence staining

Following collection, fixation, embedding, and sectioning, samples were treated with 3% methanol–hydrogen peroxide for 20 min at room temperature, washed with tap water, and subjected to antigen retrieval and waxing. Sections were blocked with normal goat serum (20 min, RT), incubated with primary antibody MCPIP1 antibody (25009-1-AP, Proteintech Group) and p53 antibody(60283-2-Ig, Proteintech Group), overnight at 4 °C, then rinsed with PBS. Afterward, fluorophore-conjugated secondary antibody (IgG) was applied and incubated at 37 °C for 20 min protected from light. Finally, sections were washed in PBS, air-dried, and mounted with neutral resin.

IHC staining

Paraffin-embedded sections were dewaxed and rehydrated, then subjected to antigen retrieval in a microwave oven filled with citric acid pH 6.0 (G1202, Servicebio). Endogenous peroxidase was blocked by incubating tissue sections in 3% hydrogen peroxide, in dark for 25 min. the area of interest were circled with immunohistochemical pen, then the tissue section was blocked with 3% BSA (G5001, Servicebio) for 30 min. MCPIP1 antibody (25009-1-AP, Proteintech Group) and p53 antibody (60283-2-Ig,Proteintech Group), Flavin-Containing Monooxygenase 3 (FMO3)(BS7313,Bioworld)was incubated overnight. Afterwards, the tissue section was incubated with secondary antibody, washed, dried, and developed using freshly prepared DAB as chromogen. IHC sections were counterstained with hematoxylin. In the end, the sections were mounted with neutral gum and examined under Microscopy.

TUNEL assay

Paraffin-embedded tissue sections were dewaxed and rehydrated, then permeabilized with 1% Triton X-100 supplemented with Proteinase K, for 15 min at room temperature. The Terminal deoxynucleotidyl transferase (TdT) dUTP Nick-End Labeling (TUNEL) reaction was performed by applying 50 ~ 100 μL TdT enzyme reaction solution on tissue section, followed by 1-h incubation at 37 °C in a humidified dark chamber. After washing, 50 ~ 100 μL Streptavidin-Fluorescein solution was added and incubated for 60 min (TUNEL kit, KGA1406-100, KeyGen Biotech). Following another wash, nuclei were counterstained with 50 ~ 100 μL DAPI solution (KGA1807-100, KeyGen Biotech) for 5 min at room temperature in the dark. Finally, sections were mounted with anti-fade medium and examined using Olympus BX63 fluorescence microscope.

SPR

MCPIP1 protein was immobilized on a CM5 sensor chip via amine coupling for interaction analysis. The chip was installed with the labeled side facing upwards, and channel 7 was activated at a flow rate of 10 μL/min using an EDC/NHS mixture in 1 × PBS-P + buffer (pH 7.4). The ligand protein was diluted to 20 μg/mL in sodium acetate and immobilized on the same channel at 10 μL/min, followed by blocking with ethanolamine to complete the surface preparation. For analyte binding experiments, the running buffer was exchanged for 1 × PBS-P + containing 5% (v/v) DMSO to match the sample matrix, and a solvent correction calibration was performed. Each analyte was serially diluted in a 96-well plate. These samples were injected over the functionalized chip surface in order of increasing concentration at a flow rate of 30 μL/min for 360 s. Following each injection, the chip surface was regenerated with a 10 mM glycine hydrochloride solution (pH 2.0) for 5 min to remove bound analyte before the next cycle. The resulting sensorgram data were globally fitted to a 1:1 Langmuir binding model using Biacore Insight evaluation software. This analysis yielded the association rate constant (Ka), the dissociation rate constant (Kd), and the equilibrium dissociation constant (KD), which quantifies binding affinity.

Co-IP

This study employed Co-IP to validate protein–protein interactions. The coding sequences of the target proteins were codon-optimized for human expression, and the corresponding plasmids were synthesized. Following transformation and extraction, plasmid quality was verified. Proteins were expressed using the Expi293F mammalian system. Cells were transfected with the plasmids using PEI Max when the viable cell density reached 2.5–3 × 10⁶/mL with > 96% viability. Culture was continued with feed medium addition.The supernatant was collected as the cell lysate. For co-immunoprecipitation, HA-tagged affinity magnetic beads (Magneti-G anti-HA beads) (Boyi,Changzhou) and isotype control magnetic beads (Magnetic-G Control IgG Beads)(Boyi,Changzhou) were pretreated with 1 × IP Wash Buffer. The cell lysate was incubated with the treated beads at 4 °C for 60 min. After incubation, bead–protein complexes were washed thoroughly with 1 × IP Wash Buffer. PBS and 2 × SDS loading buffer were then added to the protein-bound beads and heated at 37 °C for 5–10 min to elute the proteins. Following magnetic separation, the eluate supernatant was used for SDS–PAGE, and target proteins were detected by Western blot with Anti-Flag and Anti-HA antibodies to verify expression and interaction. (See Supplementary Table 4 for specific gene synthesis).

Fecal microbiota suspension donor hamsters

Donor hamsters were randomly divided into a Model group and a PC group. Hamsters in the Model group were fed a high-fat, high-choline diet. The PC group received the same dose as the PC-H group in the in vivo experiment via gavage administration and was also fed a high-fat, high-choline diet. After 8 weeks, fresh fecal samples were collected into sterile tubes and prepared fresh for immediate use.

Gnotobiotic hamsters were established by administration of a four-antibiotic cocktail.

Hamsters were randomly assigned to a gnotobiotic model group (GF group) and fed a high-fat, high-choline diet for 8 weeks, then randomly subdivided into a Model Fecal Microbiota Transplantation group (GFM) and a PC Fecal Microbiota Transplantation group (GFT). Both gnotobiotic model groups received daily gavage administration of an antibiotic cocktail in sterile water (200 μL per animal containing Ampicillin 1 g/L + Neomycin sulfate 1 g/L + Metronidazole 1 g/L + Vancomycin 500 mg/L) for 2 consecutive weeks, with sterile water replaced daily while providing sterilized standard feed and bedding. On Day 15, fresh fecal samples were collected under aseptic conditions into sterile cryovials, followed by random selection of hamsters from each group for 16S rRNA gene sequencing to verify successful gnotobiotic model establishment.

Fecal microbiota transplantation (FMT) procedure

Fresh contamination-free fecal samples from hamsters in each group were transferred to a laminar flow hood; sterile centrifuge tubes, gauze, mesh filters, and beakers were prepared. For every 200 mg of fecal sample, 2 ml of sterile PBS/DTT solution was added, followed by vortex mixing for 2 min in a sealed container. The mixture was sequentially filtered through a 100 μm sterile nylon mesh three times to obtain a 100 mg/mL fecal microbiota suspension, which was aliquoted into 50 mL sterile Eppendorf tubes and promptly transplanted into gnotobiotic rodents. This procedure was repeated weekly for 4 consecutive weeks.

DIA proteomics

Equal amounts of protein from each sample were subjected to tryptic digestion. The samples were diluted to a final volume of 200 μL and reduced with 10 mM dithiothreitol (DTT) at 37 °C for 60 min. Alkylation was performed using iodoacetamide (IAM) at a final concentration of 50 mM in the dark at room temperature for 45 min. Trypsin (Promega) was added at a 1:50 (enzyme-to-protein) ratio and incubated overnight at 37 °C. After digestion, desalting was carried out using C18 tips activated with 100% methanol and equilibrated with 0.1% trifluoroacetic acid (TFA). The samples were loaded onto the tips, washed with 0.1% TFA to remove impurities, and eluted with 90% acetonitrile. The eluate was collected and lyophilized. The processed peptides were analyzed using an Orbitrap Astral mass spectrometer equipped with a nanoflow HPLC column (QL-HPLC-100*15) and a Thermo Scientific^™ NanoSpray Flex^™ ion source. Data were acquired in data-independent acquisition (DIA) mode with positive ionization over a mass range of m/z 380–980. The full MS resolution was set to 24,000 (at m/z 200), and the MS2 accumulation time was 3 ms. The capillary voltage was set to 1.9 kV, with a total cycle time of 0.6 s. Raw mass spectrometry data were generated for subsequent analysis.The samples were processed by PANOMIX Biomedical Technology, Suzhou.

Statistical analysis

All data in the present work are represented as means ± standard deviations(± s). GraphPad Prism 9 was utilized for statistical analysis and data visualization. Shapiro–Wilk test was used to examine data normality. Levene’s test was used to assess the quality of variance. One-way ANOVA was adopted for homogeneous data, and Brown-Forsythe and Welch ANOVA tests were adopted for heterogeneous data. A threshold of p < 0.05 was deemed as statistically significant. Values of *p < 0.05, **p < 0.01, ***p < 0.001 and were considered significant between different groups. All the experiments were repeated for at least three times. The Ropls package in R language for multivariate statistical analysis. The methods include:Principal Component Analysis, PCA;Partial Least Squares-Discriminant Analysis, PLS-DA;Orthogonal Partial Least Squares Discriminant Analysis, OPLS-DA. But, Prior to the analysis, the data need to be subjected to appropriate weight transformation, namely standardization (Scaling) processing.

Results

Metabolomic profiling of PC-containing serum

Total ion chromatograms of serum components absorbed from PC were obtained using UPLC-MS/MS in both positive and negative ion modes (Supplementary Fig. A-B). Multivariate analyses PCA (R²X = 0.636), PLS-DA (R²Y = 0.659, Q² = 0.963), and OPLS-DA (R²Y = 0.633, Q² = 0.923)revealed distinct patterns: intragroup clusters were well defined, and intergroup separation was clear, indicating the results are reliable (Fig. 1A; Supplementary Fig. C-D). Volcano plots (Fig. 1B) and pie charts (Fig. 1C) illustrated the upregulated and downregulated serum metabolites and their distribution across chemical classes. A total of 605 metabolites were identified, mainly comprising terpenoids, alkaloids, flavonoids, phenols, phenylpropanoids, polyketides, and quinones. Terpenoids were the most abundant (98 compounds, 20%), followed by alkaloids (58 compounds, 12%) and flavonoids (48 compounds, 10%). By comparing retention times between blank serum and drug-containing serum, we performed an initial deduplication that yielded 160 metabolites. Using Xcalibur software, we examined relative molecular masses, secondary ion fragments, and elemental compositions in both ion modes. Differential metabolites were screened by significance (p-value < 0.05) and model contribution (VIP value > 1), resulting in 40 distinct metabolites (Table 2). Functional pathway enrichment and topological analyses of these differential metabolites were performed with MetaboAnalyst, and enriched pathways were mapped using KEGG Mapper (Fig. 1D).

Fig.1.

Metabolomic analysis of drug-containing serum metabolites using a mass spectrometry platform. A OPLS-DA score plot. B Volcano plot showing upregulated and downregulated differential metabolites. C Pie chart illustrating the classification of serum metabolites. D KEGG pathway visualization of differential metabolites analyzed by MetaboAnalyst. E–H Common targets of differential metabolites were identified through database retrieval, followed by GO, KEGG, and PPI analyses, resulting in the construction of compound–target–disease networks

Disease association and target enrichment analysis of differential metabolites

We used metabolite-driven network pharmacology to analyze the drug-containing serum metabolite results. Metabolic targets were predicted with tools such as Swiss Target Prediction, and candidate targets were filtered using DisGeNET, GeneCards, and OMIM to link metabolite components with disease-associated targets. We then performed KEGG (Fig. 1E) and GO (Fig. 1F)enrichment analyses and constructed a PPI (Fig. 1G)network for these targets to generate metabolic–target–disease network relationships (Fig. 1H). The analyses revealed associations with apoptosis and identified pathways related to atherosclerosis, oxidative stress, and cell death. Predicted key targets included p53, MDM2, Casp3, IL-1β, and MCP-1 (CCL2). These results indicate that differential metabolites produced by PC after in vivo metabolism may influence major disease-related pathways and support AS as the principal disease context for the network pharmacology and subsequent mechanistic investigations.

The chemical components of PC were determined using UPLC-MS/MS

The composition was profiled in both positive and negative ion modes by UPLC-MS/MS (Fig. 2A). By cross-referencing public databases, the PANOMIX internal standard library, the Chinese herbal medicine compound library, and prior literature on PC, 19,637 major characteristic peaks were detected in the positive ion mode, yielding 1084 identified metabolites. In the negative ion mode, 17,273 major characteristic peaks were observed, leading to identification of 748 metabolites. Together, these analyses produced 1832 identified components, which were mainly flavonoids, terpenoids, alkaloids, ketones, aldehydes, and acids (Fig. 2C). Applying pharmacokinetic screening thresholds (OB ≥ 30%, DL ≥ 0.18) resulted in 104 candidate active components (Table 3).

Fig.2.

Prediction of AS-related diseases by combining UPLC-MS/MS component analysis with network pharmacology. A Total ion current chromatograms from UPLC–MS analysis in negative and positive ion modes. B Predicted key targets of PC in AS. C Pie chart of main chemical components. D Venn diagram showing the overlap of targets between PC and AS. E Top 20 enriched CC. F Top 20 enriched BP. G Top 20 enriched MF. H Enriched KEGG pathways

Network pharmacology prediction based on the compounds detected in the liquid phase

We selected 104 candidate active ingredients for correlation analysis with atherosclerosis. High-confidence AS-related protein targets were retrieved from GeneCards (median "correlation score" > 8.51), OMIM, and DisGeNET (Score > 0.5). After merging and deduplication, we cross-validated these with PC-related targets and identified 138 potential targets (Fig. 2D). Overlapping protein targets were then analyzed in STRING (confidence ≥ 0.7); after removing irrelevant nodes, protein interaction (PPI) network was constructed. CytoNCA analysis of this network identified nodes with degree (DD), betweenness centrality (BC), and closeness centrality (CC) values above the median. We screened the resulting subnetworks and, from the second subnetwork, identified core proteins including STAT3, AKT1, TP53, IL-1β, Caspase-3, and MDM2 (Fig. 2B). Gene Ontology enrichment yielded 646 entries, comprising 257 biological processes, 203 cellular components, and 186 molecular functions. The top 20 most significant GO-BP-MF terms in each category are shown in (Fig. 2E–G). Protein phosphorylation (including peptide serine phosphorylation and peptide tyrosine phosphorylation) emerged as the principal biological process through which PC affects endothelial cell apoptosis. The associated proteins localize mainly to membrane regions and protein kinase complexes, and they exhibit serine, threonine, and tyrosine kinase activities. Inflammatory responses and apoptosis-related processes are also significantly enriched, indicating that PC may drive atherosclerosis by modulating endothelial cell apoptosis. KEGG enrichment analysis identified 144 pathways, among which the lipid and atherosclerosis pathways and the apoptosis pathway are closely linked to PC targets (Fig. 2H).

Pharmacodynamic effects of Polygonum capitatum on atherosclerosis hamster

Based on two sets of network pharmacology predictions from the drug-containing serum metabolome and the traditional Chinese medicine liquid phase, PC demonstrated lipid-lowering, antioxidant, cytoprotective, and anti-apoptotic activities. PC administration produced dose-dependent improvements in serum lipid profiles by lowering TG, TC, and LDL and by raising HDL in AS hamsters (Fig. 3A). H&E staining showed that, relative to the model group, different doses of PC limited plaque expansion, reduced foam cell numbers, and ameliorated elastic fiber disarray (Fig. 3C). Oil Red O staining confirmed decreased lipid deposition (Fig. 3D). TUNEL analysis indicated reduced endothelial-layer apoptosis after PC treatment (Fig. 3E, K). Together, these results indicate that PC curtailed the progression of atherosclerotic plaques induced by a high-fat, high-choline diet by improving hyperlipidemia and inhibiting endothelial apoptosis.

Fig.3.

Pharmacological effects of PC in high-fat, high-choline diet–induced hamsters. A Serum lipid profiles, including TG, TC, LDL-C, and HDL-C. B Serum TMAO levels. C Pathological changes in the aorta assessed by H&E staining. D Pathological changes in the aorta assessed by Oil Red O staining. E Quantification of apoptotic regions detected by TUNEL fluorescence staining. F–H Serum oxidative stress indicators, including ROS, SOD, and MDA. I, J Serum apoptosis-related factors MCP-1 and IL-1β. K The expression of TUNEL is shown as a bar chart. All the data significance was analyzed by ANOVA by Graph Pad Prism Software, Data are expressed as the mean ± standard deviation of three or more independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

To evaluate whether PC mitigates atherosclerosis via apoptosis and oxidative stress pathways, we measured related indicators in the predicted models. In hamsters fed a high-lipid, high-choline diet, serum TMAO levels were elevated compared with non-diet controls (Fig. 3B), accompanied by increased ROS and MDA levels and decreased SOD activity (Fig. 3F–H). Consistent with observations in endothelial cells, large amounts of the chemokine MCP-1 and the cytokine IL-1β were secreted into the serum of AS hamsters (Fig. 3I, J), These results indicate that rising TMAO levels coincide with heightened oxidative stress and apoptosis. Administration of Polygonum capitatum at three different dosages restored abnormal levels of ROS, MDA, SOD, MCP-1, and IL-1β in a dose-dependent manner.

Extraordinary elevations of Abnormally elevated levels of oxidative stress markers (p47phox, NOX2, iNOS) and apoptosis markers (p53 and Caspase-3) in the AS hamster model, which were restored to levels comparable to non-diet controls after high-dose PC treatment. Conversely, eNOS, MDM2 expression was reduced in AS hamsters and was normalized by low-dose PC administration (Fig. 4A–I). Interestingly, MCPIP1 expression wasdecreased in AS hamsters but returned to normal or even higher levels in the aortic tissue of PC-treated animals, But, p53 exhibited the opposite trend (Fig. 4J, L, M). In the liver, immunohistochemistry showed that FMO3 (Fig. 4K, N), the key enzyme in TMAO metabolism, was reduced following PC intervention. ELISA results corroborated that lower hepatic FMO3 limits TMAO conversion (Fig. 4O). Together, these findings suggest that elevated TMAO may suppress MCPIP1 expression, thereby promoting oxidative stress and endothelial apoptosis in AS hamsters, while PC mitigates AS progression by lowering blood lipids and reducing oxidative stress–induced endothelial cell apoptosis.

Fig.4.

PC restores MCPIP1 expression and attenuates TMAO-induced oxidative stress and apoptosis in hamster aortic tissue. A–I Expression of p47phox, NOX2, eNOS, iNOS, MCPIP1, p53, MDM2, and cleaved Caspase-3 proteins in AS hamsters. Western blot signals were quantified and normalized to β-Actin, with results shown as bar charts. Data are presented as mean ± standard deviation from three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001. J Localization of MCPIP1 and p53 in aortic tissue determined by in situ fluorescence analysis. K Localization of FMO3 in liver tissue determined by in situ IHC. L, M The expression of MCPIP1 and p53 is shown as a bar chart. N The expression of FMO3 is shown as a bar chart. Data are presented as mean ± standard deviation from three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001.O Expression of TMAO in the liver. All the data significance was analyzed by ANOVA by Graph Pad Prism Software, Data are presented as mean ± standard deviation from six independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

Polygonum capitatum modulate gut microbiota in AS hamsters

It is known that atherosclerosis is associated with distinct fecal microbiome signatures. To explore the intestinal microbiota differences of golden hamsters administered with or without PC, the fecal samples were collected from high-lipid and high-choline diet-induced hamsters, and subjected to 16S rDNA sequencing. As shown in the(Fig. 5A), Chao1 and observed_species indices increase in PC treated hamsters, exhibiting slightly dose dependency. In the context of PC intervention, microbial richness was restored (Fig. 5B). Principal Component Analysis (PCA) demonstrated significant differences among the non-diet control, non-treated model, and PC treated animals (Fig. 5C). Assessing microbiome communities similarity with unweighted unifrac, and jaccard algorithms, the species composition of intestinal microflora differed significantly across animal groups (Fig. 5D). In conclusion, PC demonstrated beneficial effects in regulating the decrease in gut microbiota richness and diversity caused by AS.

Fig.5.

Increased fecal microbial diversity in golden hamster treated with PC. A The Rarefraction curves of α diversity analysis. B Violin plot of α diversity analysis, C Principal Component Analysis (PCA) of β diversity analysis, D Community similarity analysis,The dots in the figure represent large discrete values with special markings. PCG, Atorvastatin intervention; PC_H, high dose PC treated diet-induced hamster; PC_M, median dose PC treated diet-induced hamster; PC_L, low dose PC treated diet-induced hamster; Model, diet-induced hamster without PC treatment; control, non-diet hamster. Multivariate statistical analysis was performed using principal component analysis (PCA) from the Ropls package in R language

In (Fig. 6A–C), at phylum level, comparing to the control hamsters, the Firmicutes and Desulfobacterota are relatively decreased in diet-induced hamster model, while Bacteroidota were elevated. In the context of PC intervention, relative abundances of the microbiota mentioned above returned to nearly normal. Notably, the high-dose PC group exhibited a particularly significant elevation in Verrucomicrobiota. At family level, Muribaculaceae, Lachnospiraceae, Ruminococcaceae, Oscillospiraceae, and Desulfovibrionaceae were altered in high-lipid high-choline induced hamster. Their relative abundances were reversed by intervention. From genus perspective, Lachnospiraceae_NK4A136_group, Lactobacillus, and Ruminococcus were rescued by PC intervention. Especially, in the high-dose group, Akkermansia were significantly elevated. Notably, after PC treatment, the composition of microbial genera were improved, suggesting that PC could modulate the gut microbiota in AS hamsters.

Fig.6.

Phylogenetic profiles of intestinal flora. A–C Species composition of fecal microbiota at different taxonomic levels: phylum, family, and genus. D Bubble plot of Indicator Species Analysis. PCG, atorvastatin intervention; PC_H, high-dose PC-treated diet-induced hamster; PC_M, medium-dose PC-treated diet-induced hamster; PC_L, low-dose PC-treated diet-induced hamster; Model, diet-induced hamster without PC treatment; Control, non-diet hamster. E Relative abundance levels of specific differential taxa. F Changes in F/B ratio. G Increased microbial communities at the genus level in non-diet, untreated, and high-dose PC-treated hamsters analyzed by LEfSe. H Functional prediction of gut microbial genes using PICRUSt2. I Cluster correlation analysis heat map. All the data significance was analyzed by ANOVA by Graph Pad Prism Software, Data are presented as mean ± standard deviation from six independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

To further investigate the intestinal microbiota regulated by PC, effect size measurement (LEfSe) analysis was performed with linear discriminant analysis (LDA) to identify predominant and differentially enriched bacterial taxa among groups. Indicator species analysis (ISA) further identified significant biomarkers, highlighting taxa enriched in the PC-treated group. Akkermansia, which was elevated in high-dose PC-treated hamsters (Fig. 6D), was identified as a high-confidence indicator species (p < 0.01). Statistical analysis of differential taxa combined with correlation analysis showed that Verrucomicrobiota and Akkermansia were significantly negatively correlated with TMAO levels, whereas the Firmicutes/Bacteroidota (F/B) ratio was positively correlated with TMAO levels (Fig. 3B), suggesting that PC intervention modulates gut microbiota to suppress harmful metabolite accumulation (Fig. 6E–G). Functional prediction of microbial genes using PICRUSt2 revealed that the phosphotransferase system pathway was reduced in the PC-H group compared with the model group, whereas fatty acid biosynthesis, pantothenate and CoA biosynthesis, and histidine metabolism pathways were increased (Fig. 6H).The Intergroup Correlation Analysis Heatmap Further Validates the Beneficial Impact of Increased Akkermansia muciniphila Abundance on Specific Atherosclerosis-Related Parameters (Fig. 6I).

Gut microbiota transplantation slightly reduced the indexes of endothelial cell apoptosis

Gnotobiotic hamsters were established through quadruple-antibiotic intervention in the model group (Fig. 7A). Analysis of gut microbiota structure in post-intervention gnotobiotic hamsters showed reduced microbial richness (α-diversity) and community dissimilarity (β-diversity) compared with both the model group (pre-transplantation) and the transplantation group (post-transplantation) (Fig. 7B). Heatmap analysis further revealed distinct microbial abundance patterns between gnotobiotic hamsters and the other two groups, with substantially lower abundance in gnotobiotic specimens (Fig. 7C). These results demonstrate that antibiotic gavage effectively eliminated most gut microbiota in hamsters, thereby providing a suitable model for subsequent fecal microbiota transplantation experiments.

Fig.7.

Successful establishment of gnotobiotic and FMT-treated groups. A α-Diversity analysis of gnotobiotic rodents. B β-Diversity analysis of gnotobiotic rodents. C Heatmap analysis of microbial abundance comparing gnotobiotic with pre-and post-FMT groups. D α-Diversity analysis of FMT-treated rodents. E β-Diversity analysis of FMT-treated rodents. Utilize Anosim (Analysis of Similarities) for similarity analysis, based on unweighted unifrac, weighted unifrac, Jaccard, and Bray–Curtis algorithms. All the data significance was analyzed by ANOVA by Graph Pad Prism Software, Data are presented as mean ± standard deviation from six independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

The GFM group received fecal microbiota transplantation (FMT) from the Model group (high-fat, high-choline diet), whereas the GFT group received FMT from the PC-treated group. Comparative analysis showed that both GFM and GFT groups had significantly lower observed species and Chao1 indices compared with the Control and Model groups, while Simpson and Shannon diversity indices remained comparable across all groups (Fig. 7D). In addition, evaluation of microbial community similarity using weighted UniFrac and Bray–Curtis analyses revealed statistically significant differences in intestinal microbial composition among the animal groups (Fig. 7E).

Differential analysis revealed that, compared with GFM, the GFT group exhibited higher phylum-level abundances of Firmicutes, Desulfobacterota, and Patescibacteria, but lower abundances of Bacteroidota, Spirochaetota, and Actinobacteriota (Fig. 8A–D). At the class level, GFT showed reduced Bacteroidia, Bacilli, and Coriobacteriia, while displaying elevated Desulfovibrionia, Actinobacteria, Clostridia, and Saccharimonadia. At the family level, GFT demonstrated decreased Erysipelotrichaceae but increased beneficial taxa, including Lactobacillaceae, Lachnospiraceae, and Bifidobacteriaceae. At the genus level, the GFT group harbored significantly higher abundances of Lactobacillus, Limosilactobacillus, and Bifidobacterium compared with both GFM and Model groups. The Intergroup Correlation Analysis Heatmap Further Validates the Beneficial Impact of Increased Lactobacillaceae, Lachnospiraceae, and Bifidobacteriaceae on Specific Atherosclerosis-Related Parameters (Fig. 8E, I). Notably, GFM exhibited positive correlations between TMAO-metabolizing genera (Allobaculum, Bacteroides, Ruminococcus, Prevotella) and TMAO production, whereas GFT reversed these abundance patterns. Statistical analysis of differential taxa (Fig. 8F, G), combined with correlation analysis of TMAO levels, demonstrated that gut microbiota transplanted from PC-treated donors suppressed TMAO production (Fig. 8H). Heatmap analysis of microbial composition (Fig. 8I) further confirmed substantial structural divergence in FMT-treated groups relative to Control and Model groups.

Fig.8.

Phylogenetic profiles of intestinal flora. A–D Species composition of fecal microbiota at different taxonomic levels: phylum, class, family, and genus. E Cluster correlation analysis heat map. F Relative abundance levels of specific differential taxa. G, H Changes in the Firmicutes/Bacteroidota (F/B) ratio, serum and liver TMAO levels. I Heatmap showing microbial abundance distribution after transplantation in each group. All the data significance was analyzed by ANOVA by Graph Pad Prism Software, Data are presented as mean ± standard deviation from three or more independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

Building upon these findings, ELISA assays were performed on serum samples from all four hamster groups. GFT intervention improved blood lipid profiles (Fig. 9A) and reduced TMAO levels compared with both GFM and Model groups, consistent with the observed microbiota shifts (Fig. 8H). In addition, ROS and MDA concentrations were decreased, whereas SOD activity was elevated in the GFT group (Fig. 9B–D). Pro-apoptotic inflammatory cytokines, including IL-1β and MCP-1, were downregulated (Fig. 9E, F). Western blot analysis further showed decreased expression of apoptosis-related proteins p53 and Caspase-3, while MDM2 levels remained unchanged (Fig. 9G–J). Immunohistochemistry (IHC) additionally confirmed altered expression of MCPIP1 and p53 in GFT tissues (Fig. 9K–M). However, hepatic TMAO levels showed no significant change, which may reflect regulation of metabolites by the microbial community itself rather than direct effects on the liver. Collectively, these results indicate that microbiota transplantation modestly suppresses progression of apoptosis.

Fig.9.

Levels of various indicators in FMT-treated groups. A Blood lipid levels. B ROS levels in hamster serum. C, D SOD and MDA levels in hamster serum. E, F IL-1β and MCP-1 levels in serum. G–J Expression of apoptosis-related proteins, including p53, MDM2, and Caspase-3. K–M Expression levels of MCPIP1 and p53 in aortic tissue of the GFT group. All the data significance was analyzed by ANOVA by Graph Pad Prism Software, Data are presented as mean ± standard deviation from three or more independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

Joint proteomics identified the stimulate effect of TMAO on endothelial cell apoptosis

As PC was shown to alter gut microbiota in AS model animals, we hypothesized that trimethylamine-N-oxide (TMAO) is modulated in PC-treated hamsters. TMAO, a key gut microbiota–derived metabolite produced from dietary choline, is strongly associated with atherosclerosis and represents an important regulator of the gut–heart axis. To investigate the effects of TMAO on endothelial protein expression, proteomic analysis was performed in endothelial cells following TMAO induction. A total of 110 differentially expressed proteins were identified in the TMAO group compared with the control group (Fig. 10A, B), with a volcano plot illustrating representative expression changes (Fig. 10C). The most significantly enriched GO terms for each category are presented in (Fig. 10D), with biological processes notably enriched in apoptosis, suggesting that endothelial cell apoptosis may be a major mechanism contributing to AS. KEGG pathway analysis (P < 0.05) based on differential protein expression highlighted strong associations of TMAO with atherosclerosis and the p53 signaling pathway (Fig. 10E). Furthermore, GSEA-based KEGG enrichment of quantitative proteins confirmed enrichment in atherosclerosis and apoptosis pathways (Fig. 10F). Collectively, these results indicate that TMAO promotes AS progression by upregulating p53 pathway–related proteins to induce endothelial apoptosis, consistent with findings from serum metabolomics and network pharmacology analyses.

Fig.10.

Proteomic analysis of TMAO-induced endothelial cells. A Venn diagram showing overlapping protein expression among Con, TMAO, and Si-MCPIP1 + TMAO groups. B PCA plot of principal component distribution. C Volcano plot of differentially expressed proteins between TMAO and Con groups. D Top 20 enriched GO terms for biological process BP,CC, and MF. E KEGG pathways identified based on differentially expressed proteins. F KEGG pathway enrichment analysis of quantitative proteins using GSEA

MCPIP1, a regulator of angiogenic balance and cardiovascular homeostasis. The molecular interaction between MCPIP1 and TMAO was modeled using AlphaFold3 (Fig. 11A), with an interaction confidence score exceeding 0.7, indicating high-quality predictions since the combined ipTM and pTM scores were well above the recommended threshold of 0.5 [29], The predicted model suggested that TMAO and MCPIP1 may bind. Guided by this prediction, After purification of MCPIP1 (Fig. 11B), we performed SPR analysis using the analyte at the dissolved concentration, and the kinetic fitting supported a physical interaction between TMAO and MCPIP1(Fig. 11C), although the functional consequences of this binding remain unclear. Co-immunoprecipitation further confirmed interaction, as Mcpip1-HA enrichment coincided with enrichment of P53-Flag(Fig. 11D). We evaluated the effective dose range of PC by measuring dose-dependent effects on endothelial cell viability (Fig. 11E). Of the si-RNAs tested, siRNA-1848 produced the most efficient MCPIP1 knockdown (Fig. 11F). TMAO treatment increased ROS and MDA levels and decreased SOD activity in endothelial cells (Fig. 11G–I); PC rescued these abnormalities by modulating MCPIP1. Likewise, TMAO elevated inflammatory factors MCP-1 and IL-1β, and PC restored their levels to baseline in the presence of MCPIP1 (Fig. 11J, K).

Fig.11.

The role of MCPIP1 in oxidative stress and apoptosis was investigated by linking TMAO-MCPIP1-P53 and using Si-RNA. A Docking structure of MCPIP1 and TMAO, predicted by AlphaFold 3. B Results of MCPIP1 protein purification. C Analysis of Physical Binding Level of TMAO and MCPIP1 by SPR Technology. D Co-IP validation of the interaction between MCPIP1 and p53. E CCK8 endothelial cell viability caused by dose-gradient PC.F Transfection efficiency of MCPIP1 siRNAs were determined by reduction of MCPIP1 protein abundance. G, H released SOD and MDA in cell culture supernatant. I ROS in cell culture supernatant. J, K Secreted IL-1β and MCP-1 in cell culture supernatant. All the data significance was analyzed by ANOVA by Graph Pad Prism Software, Data are presented as mean ± standard deviation from three or more independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

Polygonum capitatum suppresses endotheliocyte apoptosis via MCPIP1

Under TMAO-induced oxidative stress and inflammation, PC inhibits p47phox activity, prevents the increase in NOX2, preserves eNOS expression, and reduces iNOS levels. PC also suppresses oxidative stress, lowers ROS, and reduces p53 and the executioner caspase-3. In contrast, MDM2, an inhibitor of p53, shows an upward trend in response to PC (Fig. 12A–I). All of these regulatory effects of PC depended on appropriate MCPIP1 expression. Consistent with this, TMAO was found to promote oxidative stress and endothelial apoptosis, whereas PC alleviated oxidative and inflammatory stress and suppressed apoptotic signaling. Flow cytometry analysis showed that treatment with TMAO alone increased the proportion of early apoptotic endothelial cells from 2.42% to 12.71%, whereas co-treatment with PC reduced this proportion to 4.14%. The protective effect of PC was lost — even when PC was co-treated in the knockdown group, the levels of ROS and p53 did not show significant downregulation, remaining only similar to those in the Model group, and early apoptotic cells remained at 7.11% despite PC co-treatment (Fig. 13J, K). These results indicate a key anti-apoptotic role for MCPIP1 in TMAO-induced endothelial apoptosis and suggest that PC may exert its antioxidant and anti-apoptotic effects via activation of MCPIP1.

Fig.12.

Apoptosis rate of endothelial cells and expression of apoptosis-related proteins in MCPIP1-depleted endothelial cells following PC and TMAO treatment. A–I Expression of p47phox, NOX2, eNOS, iNOS, MCPIP1, p53, MDM2, and cleaved Caspase-3 proteins. Western blot signals were quantified and normalized to β-Actin, with results shown as bar charts. Data are presented as mean ± standard deviation from three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001. J, K Flow cytometry analysis of endothelial cell apoptosis. The APC-A channel represents Annexin V–stained cells undergoing apoptosis, while the 7-AAD-A channel represents 7-AAD–stained dead cells. All the data significance was analyzed by ANOVA by Graph Pad Prism Software, Data are presented as mean ± standard deviation from three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

Fig.13.

Metabolite with AS targets. A Binding energy between selected components and target proteins

Metabolite-target molecular docking prediction

Based on prior studies, We utilized molecular docking to predict the binding affinity of metabolites with proteins, thereby assisting in validating the conclusions of the aforementioned molecular biology experiments. we screened the top metabolites by degree value and evaluated selected PC metabolites using molecular docking (Ganoderic acid G, Hydroxyhydroquinone, Methyl 3,4,5-trimethoxycinnamate, Neoabietic acid, Nonadecanoic acid Podofilox, and Atorvastatin),The binding affinities for all receptor–ligand pairs were below −6 kcal/mol, indicating strong molecular interactions and supporting the likelihood that these metabolites bind core protein targets related to atherosclerosis. We found that atorvastatin demonstrated strong binding affinity for all target proteins, and in experiments, we noted its effect on proteins like p53 marginally outperformed that of PC. Among these, Ganoderic acid G, Neoabietic acid, and Podofilox showed the strongest binding affinities, and their binding patterns were visualized using PyMOL (Fig. 13A).

Discussion

This study used drug-containing serum metabolomics to identify differential metabolites enriched in oxidative stress and apoptosis pathways, suggesting a mechanism for the anti-atherosclerotic effect. UPLC-MS/MS then characterized the herbal chemical profile and provided candidate compounds for network pharmacology analysis. Two independent network pharmacology workflows converged on apoptosis pathway enrichment and identified targets including p53, MDM2, and caspase-3. Consistent with previous reports, podofilox in PC may affect apoptosis by regulating the p53 pathway [30], while ganoderic acid may downregulate endothelial Caspase-3 via antioxidant and lipid-modulating effects, thereby reducing oxidative stress-related apoptosis [31]. These results illustrate PC multi-component, multi-target, multi-pathway character and implicate targets including TP53, IL-1β, caspase-3, and MDM2 in the regulation of oxidative stress and apoptosis. In AS model hamsters, PC administration remodeled the gut microbiota, reduced TMAO, and concurrently improved blood lipids, ROS/oxidative stress markers, and apoptosis. Overall, the data suggest that PC may attenuate AS by altering gut microbiota and metabolites (for example, TMAO) to mitigate associated oxidative stress and apoptosis, although the precise causal mechanisms require further study.

The gut microbiota, a central component of the gut–heart axis, can accelerate AS progression when its composition becomes dysregulated. Such dysbiosis favors the systemic dissemination or local accumulation of harmful metabolites and thereby activates pathogenic pathways and targets [32, 33]. Accordingly, we first examined diet-induced microbial perturbations and the modulatory effects of PC. Evidence indicates that diet-driven hyperlipidemia and gut microbiota shifts in AS patients contribute substantially to disease pathogenesis, primarily through altered metabolites such as Lipopolysaccharide(LPS), TMAO, and fatty acids, which are key predictive risk factors for AS. Our study observed that, compared with the control group, AS hamsters induced by a high-fat and high-choline diet exhibited a significant reduction in the abundance of Firmicutes and Desulfobacterota, whereas Bacteroidota was elevated. Additionally, unclassified Prevotella_9 was significantly enriched; Prevotella abundance correlated positively with serum TMAO levels and may be involved in diverse metabolic processes [34]. Following PC treatment, Atherosclerotic hamsters exhibited a trend toward partial recovery in gut microbiota abundance and diversity. Specifically, abundances of Firmicutes and Desulfobacterota increased, while Bacteroidota and Actinobacteriota decreased, and Verrucomicrobiota notably increased in the PC-H group. It has been reported that Verrucomicrobiota is associated with intestinal mucosal integrity and immune-inflammatory regulation. Its enrichment in the treatment group suggests that the microbiota alterations following PC intervention may be linked to enhanced barrier protection and an anti-inflammatory state. [35]. At the genus level,The abundance of Akkermansia was significantly increased in the high-dose group. Akkermansia, a representative genus of Verrucomicrobiota, has been shown to attenuate TMAO-induced AS, promote Short-chain fatty acids secretion, and negatively regulate inflammatory responses, thus influencing the onset and progression of obesity, diabetes, and cardiometabolic diseases [36, 37]. In addition, Clostridium can lower circulating TMAO by converting choline into alternative metabolites [9]. Given that PC treatment was associated with gut microbiota alterations and reduced TMAO levels, we further used FMT to evaluate the transferability of the microbiota and its role in the phenotypic improvements induced by PC, while also aiming to distinguish between microbiota-mediated and host-directed pathways. The FMT results showed that microbiota from PC-treated donors partially shifted the gut microbial composition of pseudo-germ-free hamsters toward the donor's profile, consistent with decreased TMAO levels and reduced markers of oxidative stress/endothelial injury. Changes in the relative abundance of groups such as Lactobacillaceae, Oscillospiraceae, and Bifidobacteriaceae might be involved in this process, though their specific functions require further verification. However, microbial richness and diversity remained lower after FMT compared to the control group, suggesting that while FMT partially restored the microbiota after antibiotic depletion, some low-abundance bacteria might not have colonized fully or recovered adequately. Due to the resolution and functional inference limits of 16S rRNA sequencing, this study primarily captured overall structural changes and could not pinpoint alterations in specific functional bacterial strains. Nevertheless, integrating prior work with our correlations, these results remain biologically meaningful: PC may reshape gut microbiota composition and related metabolic functions, thereby influencing glucose and lipid metabolism and reducing TMAO. We also found that absorbed PC components acted on hepatic FMO3, which together with microbiota changes contributed to lower TMAO, although the relative contributions of these two mechanisms require further experimental clarification. As TMAO declined, markers of oxidative stress and apoptosis in the animals also decreased.

TMAO is recognized as a central metabolite linking the gut–heart axis. Prior work has shown that elevated circulating TMAO associates with vascular endothelial oxidative stress, activation of apoptotic signaling, impaired endothelial barrier function, and progression of atherosclerosis. Accordingly, we investigated whether a “TMAO-related oxidative stress–apoptosis axis” underlies PC-mediated protection. High TMAO levels have been reported to promote NOX activation and ROS accumulation, which deplete tetrahydrobiopterin (BH4) and destabilize eNOS. This endothelial dysfunction, together with increased iNOS expression and NO release, ultimately triggers apoptosis [38, 39]. In our study, increased TMAO coincided with upregulated p47phox expression in hamsters. Because p47phox is a principal regulatory subunit of the NOX2 complex, this observation is consistent with augmented NOX2-derived ROS generation [40, 41]. The resulting ROS further destabilize eNOS, induce iNOS upregulation, and promote apoptosis. Furthermore, existing studies suggest that TMAO may impair mitochondrial oxidative phosphorylation. This is associated with altered expression of respiratory chain complex II subunits (SDHB) and increased mitochondrial ROS production [42–44], potentially leading to mitochondrial dysfunction [45]. Excessive ROS can oxidatively modify and regulate redox-sensitive signaling proteins (p53, NF-κB), thereby influencing cellular stress responses and transcription programs linked to inflammation and apoptosis [46, 47]. A portion of this apoptotic signaling is mediated by the p53 pathway, which activates the caspase cascade to execute cell death [48]. Thus, in this context, ROS acts as an upstream regulator of p53 activation, with p53 functioning as a key downstream mediator of apoptosis [49, 50]. The existing literature on the "oxidative stress–apoptosis axis" supports the changes in ROS and apoptotic markers that we observed in our high-fat, high-choline diet–induced hamster model and in cellular models. The reversal of these markers after PC treatment suggests that PC may protect tissues by attenuating TMAO-associated oxidative stress and apoptotic signaling. However, the specific mechanistic intermediates in this pathway remain to be identified and require further validation.

MCPIP1 has drawn attention for its roles in controlling inflammation and apoptosis. Primarily known for its RNase activity and its involvement in transcriptional regulation, MCPIP1 may serve as a key mediator linking metabolites to apoptotic pathways. However, direct evidence that the metabolite TMAO binds MCPIP1 and thereby modulates p53-related apoptotic signaling is limited, and the mechanisms remain unclear. Likewise, it is not yet established whether PC can prevent endothelial cell apoptosis by modulating MCPIP1-related pathways. MCPIP1 directly interacts with TRAF-associated substrates, promoting degradation of cytokine mRNA transcripts, thereby negatively regulating the activities of IκB and JNK kinases and suppressing NF-κB signaling, ultimately exerting anti-inflammatory effects [51, 52]. MCPIP1 has also been reported to exacerbate TNF-α-mediated apoptosis via downregulation of the NF-κB/cFLIP signaling axis [53], whereas metformin can prevent LPS-induced acute kidney injury by upregulating the MCPIP1/SIRT1 axis [54, 55]. Owing to its multifaceted molecular functions, we investigated the specific contribution of MCPIP1 to endothelial cell apoptosis, evaluating whether it mediates TMAO-triggered oxidative stress and p53-related apoptotic signaling. We used SPR and Co-IP assays to assess the potential for direct physical binding between TMAO and MCPIP1. Furthermore, by examining the protein–protein interaction between MCPIP1 and p53, we aimed to establish the connectivity of the TMAO–MCPIP1–p53 axis. Subsequently, we used AD-hMCPIP1-shRNA to knock down MCPIP1 expression in endothelial cells. Under this knockdown condition, TMAO treatment led to increased ROS levels, exacerbated oxidative stress markers, and a corresponding rise in the apoptosis rate. When PC was administered in the AD-hMCPIP1-shRNA + TMAO setting, MCPIP1 expression showed an upward trend compared to the knockdown group, suggesting that PC may influence MCPIP1 expression or protein stability. Meanwhile, PC administration lowered NOX2-linked ROS, oxidative stress, and apoptosis. This indicates a potential role for MCPIP1 in PC's protective effects; nevertheless, further studies are needed to verify the causal link.

Taken together, our findings indicate that PC may modulate the pathological cascade of “gut microbiota dysbiosis—metabolic perturbation—dysregulation of key signaling proteins” via mechanisms linked to the gut–heart axis. Concurrently, PC treatment was associated with reductions in blood lipid and TMAO levels and prevented MCPIP1 downregulation, thereby attenuating activation of signaling pathways involved in oxidative stress and apoptosis. Moreover, PC consistently upregulated MCPIP1 and ameliorated oxidative stress and apoptosis, implicating MCPIP1 in both TMAO-induced endothelial apoptosis and the protective effects of PC. These molecular and cellular effects coincided with improved AS-related phenotypes; however, definitive causal relationships require further mechanistic validation.

The pathogenesis of AS involves complex pathological processes and metabolic pathways across multiple organs. The mechanism discussed in this study—oxidative stress-induced apoptosis leading to vascular endothelial injury—represents only one part of this broader picture. A more comprehensive understanding of AS requires further integration and validation of other relevant pathways. First, subsequent animal studies could establish an MCPIP1-deficient model in vascular tissues to further elucidate its role in AS. Concurrently, since the specific active components within PC have not been screened, future work should isolate and purify candidate unique compounds from PC and assess their binding affinity to MCPIP1 using SPR. Additionally, molecular docking should be combined with molecular dynamics simulations to enhance the reliability of predictions. Second, regarding TMAO, the gut—liver axis is crucial for its generation and clearance. Although this study observed correlated changes in gut microbiota structure and TMAO levels, a more systematic analysis of alterations in TMAO metabolic pathways is still lacking. Third, concerning the oxidative stress and apoptosis axis, this study only covered partial signaling pathways. Other relevant pathways suggested by proteomics and changes in mitochondrial ultrastructure require further validation and in—depth discussion in follow—up studies. This study also has methodological constraints, including the limited resolution of 16S rRNA sequencing and the restricted dynamic range of the proteomics approach. Additionally, the analysis of microbiota alterations following drug treatment or transplantation lacked integration with metagenomic and metabolomic (targeted/untargeted) data to identify specific strains and critical metabolites, and did not involve re—colonization experiments with isolated PC—modulated bacteria. Thus, the evidence supporting key functional molecules and causal relationships remains incomplete. Meanwhile, the FMT experiment in this study primarily compared the effects of transplants from model donors versus PC—treated donors. The focus was to assess whether PC—associated microbiota could be partially transferred and reconstituted in recipients under the same high—fat, high—choline diet. However, the absence of an FMT control from healthy donors makes it difficult to distinguish whether the PC—modified microbiota restored a state closer to health or induced a new, drug—related homeostasis [54]. To this end, future studies should integrate multi-omics strategies—such as network pharmacology, metagenomics, microbiomics, metabolomics, and transcriptomics—to obtain more comprehensive mechanistic evidence and further elucidate the related regulatory mechanisms of the gut—heart axis.

Conclusions

This study comprehensively investigated the potential active components of PC and their therapeutic mechanisms in AS. The findings demonstrated that PC not only effectively suppressed oxidative stress and apoptosis in endothelial cells but also improved gut microbiota composition in an AS hamster model. Notably, PC significantly alleviated apoptosis and oxidative stress by modulating the ROS/p53 signaling pathway in an MCPIP1-dependent manner. These novel findings provide crucial theoretical evidence for the clinical treatment of AS and offer valuable insights for the development of related therapeutic strategies.

Abbreviations

OMIM

Mendelian Inheritance in Man

AS

Atherosclerosis

PDB

Protein data bank

BP

Biological processes

CC

Cellular components

CTD

Comparative toxicogenomics database

PPI

Protein–protein interaction

PC

Polygonum capitatum

ELISA

Enzyme linked immunosorbent assay

GO

Gene ontology

HDL-C

High density lipoprotein cholesterol

HE

Hematoxylin and eosin

TCH/T-CHO

Total cholesterol

TCMSP

Traditional Chinese medicine systems pharmacology database

KEGG

Kyoto Encyclopedia of Genes and Genomes

TG

Triglyceride

LDL-C

Low density lipoprotein cholesterol

TTD

Therapeutic target database

MF

Molecular functions

SOD

Superoxide Dismutase

MDA

Malondialdehyde

Author contributions

Yunpei Wang: Writing-review & editing, Writing-original draft, Visualization, Software, Resources, Project administration, Methodology. Weiyi Tian: Investigation, Formal analysis, Data curation, Conceptualization. Zi Ye: Supervision, Software, Methodology. Yuanzhu Liao: Investigation, Formal analysis, Conceptualization. Chunhua Huang: Project administration, Investigation, Data curation. Dake Qi: Validation, Software, Resources, Project administration. YuHui Wang: Data curation, Resources,Validation. YaJie Chen: Resources,Software, Data curation. Yixia Zhou: Writing – review & editing, Supervision, Methodology, Funding acquisition, Conceptualization.

Funding

This work was supported by National Natural Science Foundation of China [grant numbers 82160099]; Guizhou High-level Innovative Talents Project [grant number Qian-ke-he basic talents-GCC[2023]085]; Science and Technology Plan Project of Guizhou Province [grant number Qiankehe support [2022] generally 263]; Guizhou Provincial Postgraduate Education Innovation Program project funding [grant number 2024YJSKYJJ376].

Data availability

Raw data is available from the corresponding author upon request. All data were generated inhouse, and no paper mill was used. All authors agree to be accountable for all aspects of work in ensuring integrity and accuracy.

Declarations

Ethics approval and consent to participate

This manuscript did not involve human participants, and all the animal experiments were approved by the Ethics Committee of Guizhou University of Traditional Chinese Medicine (No.20250501001, Guiyang, China). Animal source: Liaoning Changsheng Biotechnology Co. LTD., Animal production license number: SCXK(Liao) 2020–0001. Experimental animal use license number: syxk(Qian) 2021–0005.

Competing interests

The authors declare no competing interests.