Polydatin (purity ≥98%, 140 g, PO210425, Xi’an Guanjie BioTech Co. Ltd., Xi’an, China) and hawthorn flavonoids (purity ≥90%, no. AKH15-2; Linyi Aikang Pharmaceutical Co., Ltd., Shandong, China) were mixed to prepare a solution. simvastatin was purchased from Hangzhou MSD Pharmaceutical Co. Ltd. (20 mg, U010049; Zhangzhou, China). All the natural products using from a single manufacturing batch. The drug dosage standard for this experiment was in accordance with the First edition of Chinese Pharmacopoeia 2015. The clinical recommended daily oral doses for adults were 15 g knotcane and 12 g hawthorn. The dosage for mice was converted according to the equivalent dose ratio of adult (adult weight calculated by 60 kg) to the body surface area of mice and the content percentage of Chinese medicine components in Chinese medicinal materials. The high-fat diet (HFD) containing 21% saturated fat, 0.15% cholesterol and 1% choline was purchased from Keao Xieli Feed Co., Ltd. [SCXK (Beijing) 2019–0003].

Specific pathogen-free (SPF) male C57BL/6J mice (8-week-old, 19–21 g) and male apolipoprotein E knockout (ApoE^−/−) mice of the same age and genetic background were obtained from Vital River Laboratory Animals (Beijing, China). The mice were bred in a SPF laboratory and had free access to food and water. After 13 weeks, aortic atherosclerotic plaque formation was confirmed by microscopic observation after dissection, indicating that the atherosclerotic model was successfully prepared. Subsequently, all mice were randomly assigned to six groups (n = 10 in each group): (Bansilal et al., 2015): control: C57BL/6J mice received normal diet; (Joris and Gloor, 2019); model: ApoE^−/− mice received HFD; (Blagov et al., 2023); low dose of PH group (PHL): HFD supplemented with polydatin (50 mg/kg daily) and hawthorn flavonoids (25 mg/kg daily); (Vaccarezza and Galassi, 2023); medium dose of PH group (PHM): HFD supplemented with polydatin (100 mg/kg daily) and hawthorn flavonoids (50 mg/kg daily); (Zhou et al., 2020); high dose of PH group (PHH): HFD supplemented with polydatin (200 mg/kg daily) and hawthorn flavonoids (100 mg/kg daily); (Lau et al., 2017); simvastatin: HFD supplemented with simvastatin (5 mg/kg daily). The control and model groups received distilled water (0.2 mL daily). The experiment lasted for 24 weeks, and all mice were fasted overnight and euthanized using isoflurane inhalation at the end of the experiment. The experimental protocol was approved by Institutional Animal Care and Use Committee (IACUC), Guang’an men Hospital, China Academy of Chinese Medical Sciences. The protocol for in vivo studies was approved by the Ethics Committee of Guang’anmen Hospital, China Academy of Chinese Medical Sciences (IACUC-GAMH-2021-013).

After collection of blood samples, the circulatory system was rinsed with normal saline. The thoracic aorta was isolated, and adherent fat was removed. The heart and aorta were perfused with phosphate-buffered saline for 10 min and then with 3.5% paraformaldehyde for 5 min. The heart was then rapidly removed, fixed in 10% buffered neutral formalin for 24 h, and washed in tap water to remove the formalin. For enface staining of the plaque areas in the mice, the whole aorta was soaked with Oil Red O (ORO) solution and photographed against a black background. The aortic sinus specimens were fixed with 4% paraformaldehyde, dehydrated in layers, and embedded in paraffin wax. Continuous sections (5 μm) were taken from the aortic sinus to the aortic arch at intervals of 40 mm, and atherosclerotic aortic sinus lesions were evaluated using hematoxylin-eosin (H&E) staining. The remaining aortic sinus samples were buried at the optimal cutting temperature, and frozen sections were used for ORO staining to evaluate lipid deposition areas. The ratio used to make a working dilution of the red O dye was 3:2 (red O dye:distilled water). The ratio of isopropanol to distilled water was 3:2, which formed a 60% isopropanol solution. Subsequently, the aorta was removed from the 10% formalin solution, rinsed for 1–2 min with distilled water, dipped in 60% isopropanol for 1 min, stained with the oil red O working dilution for 10 min, and then the aorta was placed in 60% isopropanol. Aortas were differentiated until the vascular wall of the aortas was transparent, and the difference between the color and the plaque was clear. Finally, the aortas were stored in 10% formalin, after staining and taking photos. Three lesion areas were compared by using computer-supported morphometry (ImageJ software) at 30-μm intervals, and the average lesion size was estimated. The percentage of plaque in total vascular area represented the relative severity of AS.

The plasma concentrations of cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), very low-density lipoprotein cholesterol (VLDL-C), and high-density lipoprotein cholesterol (HDL-C) were measured using an automatic bioinformatic instrument (BackmanAU5821, United States).

Serum levels of interleukin-1-beta (IL-1β), interleukin-2 (IL-2), interleukin-6(IL-6), interleukin-17A (IL-17A), tumor necrosis factor alpha (TNF-α) and hs-CRP were measured using the Aimplex Mice Custom Premixed Analyte kit (Beijing Quantobio, China).

Paraffin-embedded fixed aortic arch tissues were sliced into approximately 10 μm sections. These sections were incubated with nuclear factor-кB (NF-κB) p65 Rabbit PolyClonal antibody (Proteintech; 10745-1-AP, 1:500); tumor necrosis factor alpha (TNF-α) antibody (Proteintech; 60291-1-Ig, 1:1,000); and Anti-FMO3 Antibody (abcam; ab126711, 1:5000) for overnight at 4°C. After being washed, slices were incubated with horseradish peroxidase (HRP)-conjugated anti-rabbit secondary antibody (Proteintech; SA00001-2, 1:300) for 1 h at 37°C. Hematoxylin was added to the slices for 30 s and the slices were counterstained after washing with PBS. The area of positive staining represented the levels of NF-κB and TNF-α in the aortic arch tissue, which was calculated and evaluated using the ImageJ software.

Protein concentrations in the aorta and liver tissues were quantified using a BCA kit (Thermo Scientific). Proteins were separated by 5% or 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred onto polyvinylidene fluoride membranes (Millipore, Burlington, MA). The membrane was blocked with 5% skimmed milk, for 1 h, and then incubated with primary antibody overnight at 4°C: TNF-α, NF-κB p65 and Anti-FMO3 antibody were same as the above (as shown in 2.6); β-actin (ABclonal, AC026, 1:1,000). The blots were incubated with HRP-conjugated anti-rabbit IgG for 1 h. After washing, the membranes were analyzed using chemiluminescence detection kit.

Total RNA was extracted from liver tissues using TRIzol reagent. The cDNA was synthesized using SuperScript II (TaKaRa Bio). RT-PCR for mRNA amplification was performed using a SYBR Green Master Mix kit. Details of the real-time PCR conditions are shown in Supplementary Table S1. The target mRNA level was normalized to the β-actin level. All primer sequences were designed using the PrimerBank software and are listed in Supplementary Table S2.

Serum TMAO and TMA levels were measured using the Waters ACQUITY ultra-high-performance liquid chromatography system (Waters Corporation, Milford, MA, United States). Samples were separated using a mobile phase gradient consisting of a mixture of 5 mM ammonium formate as solvent A and acetonitrile as solvent B. The analytes were eluted using the following gradient profile: 90% B, 10% A for 1 min; 3.5 min, 50% A, 50% B; and reconditioning at 90% B, 10% A for 8.6 min. The detailed contents of the mobile phase gradient and mass spectrometry conditions are shown in Supplementary Tables S3, S4. Data processing was completed using the TargetLynx software.

Bacterial DNA was extracted from 0.5 g fecal samples utilizing the FastDNA^® Spin Kit for SoilDNA (Qiagen, Limburg). DNA integrity was detected using 1% agarose gel electrophoresis, and the concentration and purity of DNA were detected using a NanoDrop 2000 microspectrophotometer (Thermo Fisher Scientific, United States). The forward primer 338F (5′-ACT​CCT​ACG​GGA​GGC​AGC​AG-3′) and reverse primer 806R (5′-GGACTACHVGGGTWTCTAAT-3′) were used to amplify the V3-V4 variable region of the bacterial 16S rRNA gene. Amplification was performed according to the standard protocols. PCR products were detected by 2% agarose gel electrophoresis and purified using an AxyPrep DNA Gel Extraction Kit (Invitrogen, Carlsbad, CA, United States). A NEXTFLEX Rapid DNA-Seq Kit (Bioo Scientific, United States) was used to prepare the sequencing library. The amplified products were collected and sequenced on an Illumina MiSeq PE300/NovaSeq PE250 platform (Illumina, San Diego, United States). All raw sequences have been submitted to the Sequences Read Archive database at the NCBI under accession number PRJNA1122934.

The quality control and splicing process of the raw sequencing sequences were based on the fastp software that was merged with FLASH. The UPARSE software was used to cluster operational taxonomic units (OTUs) with a 97% similarity cutoff. Each sequence was classified using the RDP classifier and compared to the Silva 16S rRNA database (V128) at a threshold of 70%. Alpha diversity and beta diversity analysis was used to show the distribution and similarity using principal coordinate analysis (PCoA). The differences at different taxonomic levels were evaluated using linear discriminant analysis effect size (LEfSe). A correlation heatmap was utilized to visualize the relationships between the different species and other parameters.

Quality control and optimization were performed using software fastp, and filtered reads were assembled (contigs ≥300 bp). In addition, open reading frames (ORFs) prediction and translation of contigs were performed using MetaGene to obtain the corresponding amino acid sequences. CD-HIT was used to cluster the predicted genetic sequences, and the non-redundant gene set was constructed. Diamond was used to align the amino acid sequences with the GENES database using BLASTP (version 2.2.28) and annotate the kyoto encyclopedia of genes and genomes (KEGG) databases (version 94.2). The amino acid sequences were aligned to the carbohydrate-active enzymes (CAZymes) database using HMMER (version 3.1b2). The abundance and differences in the corresponding species, CAZy, and KEGG functional categories were estimated using normalized RPKM values.

Statistical analyses were performed using GraphPad Prism 9.0. All data were presented as the mean ± standard deviation (SD), and multiple comparisons were analyzed using ANOVA. Statistical significance was set at P < 0.05.

Throughout the experiment, an obvious increase in body weight was observed in HFD-fed mice. After 11-week of intervention, the PH and simvastatin groups may have experienced a lower gain in body weight than that of the model group (Figure 1A). In addition to the possible anorexic effect that may be induced by HFD, supplementation with PH or simvastatin may also have exerted certain effects on the body weight and metabolic function of mice. Compared to the control group, a remarkably increasing trend of TC, TG, LDL-C, and VLDL-C was observed in HFD-fed mice (P < 0.001). Whereas mice supplemented with PH or simvastatin exhibited low levels of TC, TG, LDL-C, and VLDL-C, accompanied by high levels of HDL-C, reflecting the hyperlipidemia caused by the HFD; these levels were reversed by PH or simvastatin administration (Figures 1B–F). These results, together with the increase in body weight, highlight the synergistic effect of lipid metabolism as a potential modulator of atherosclerosis and the important role of PH on the treatment of dyslipidemia.

Atherosclerotic plaque size was measured to assess the pathological changes in atherosclerosis. The ratio of the plaque area to the entire aorta or aortic root in the model group increased after HFD feeding. The results of ORO staining revealed that the rising trend in atherosclerotic plaque area was inhibited, and even reduced in different doses of the PH or simvastatin groups as compared with the model group (Figures 2A, C). The aorta inner wall was smooth without any evidence of obvious plaques attached in the control group, whereas lipid deposition was more prominent with an amount of foam cells within the plaque and aortic vessel stenosis in the model group (Figures 2A, C). Specifically, quantitative analysis revealed that the proportion of aortic lesions to the total aortic vessel area in the PH group showed a descending trend compared to that of the model group (P < 0.0001) (Figure 2B). The therapeutic effects of PHM and PHH on atherosclerosis were significant. Consistently, the proportion of cross-sectional lesions of the aortic roots reduced in the PHM and PHH groups, compared to the model group (P < 0.01) (Figure 2D). The results of H&E staining also suggested that the atherosclerotic plaques and lesions in the Model group were higher than those in the Control group (Figure 2E). The lesion area in the aortas in the PH and simvastatin groups mice was significantly reduced (P < 0.001) (Figure 2F). In summary, PHM, PHH, and simvastatin had exerted significant effects on atherosclerosis and delayed plaque deterioration.

Besides lipid abnormalities, the progression of atherosclerosis is also influenced by systemic inflammation to a certain extent. Given that inflammatory cytokines have been recognized as an index of atherosclerosis in epidemiological studies, immunohistochemical staining of aortic root plaques was conducted. We observed that the expression of TNF-α and NF-кB was more prominent in mice fed an HFD than in mice fed a chow diet. Nevertheless, dietary supplementation with PH and simvastatin diminished the intensity of the staining, implying that a strong antagonistic action was exerted against the inflammatory reaction during the process of atherosclerotic plaque formation (Figures 3A, C). Quantitative analysis was utilized to assess the specific efficacy of PH and simvastatin. As shown in Figures 3B, D, the ratio of TNF-α and NF-кB positive area to the aortic sinus plaques in PH and simvastatin groups were dramatically decreased (P < 0.0001), especially obvious in PHH group. Thus, the therapeutic effect of PHH in down-regulating TNF-α and improving inflammation factor levels is superior to that of simvastatin (Figure 3B). Consistently, we observed that the downregulation of NF-κB protein in aortic plaque in PHM, PHH, and simvastatin group was more significant than that in model group (P < 0.0001) (Figure 3E), with the lowest protein expression in the PHH group (Figures 3F, G).

Furthermore, we measured the levels of other important inflammatory mediators to confirm the effects of PH on atherosclerosis. The results were consistent with the previously reported changes, demonstrating that the serum levels of hs-CRP, TNF-α, IL-1β, IL-2, IL-6, and IL-17A in model group mice dramatically increased in mice fed an HFD (Figures 3H–M). After PH and simvastatin administration, a dose-dependent decrease in the inflammatory cytokine levels was observed, with the lowest level in PHH group (P < 0.05) (Figure 3M). This suggested that inflammatory cytokines, represented by TNF-α and NF-кB, are closely related to the progression of atherosclerotic plaques. PH reduces inflammatory factor levels, thereby improving atherosclerotic plaque formation, with the most pronounced effects found in the PHH group.

To investigate the underlying mechanism of PH on atherosclerosis, atherosclerotic markers TMAO, and the TMAO metabolic pathway-related proteins were assessed. We observed that the HFD containing 1% choline induced an elevation of TMA and TMAO levels in the model group mice. Similar to the findings of previous studies (Li et al., 2021), choline accelerated atherosclerotic progression in the present study. We observed the attenuation of TMA and TMAO in the PHH group, compared to the model group (P < 0.001). Notably, TMA and TMAO levels were reversed by PH supplementation at high doses, indicating that the therapeutic effect of PHH was more prominent than that of the others (Figures 4A, B).

Hepatic FMO3 is a crucial enzyme involved in the conversion of TMA to TMAO. Thus, we focused on assessing the dynamic changes in FMO3 expression and activity. In line with previous findings, the results demonstrated that compared to the control group, the protein and mRNA levels of FMO3 in mice were upregulated in the model group (P < 0.01). PH downregulated the protein and mRNA expression of FMO3, which was more evident in terms of protein expression of FMO3. Similarly, a high dose of PH exhibited an unexpected effect on TMA/TMAO metabolism (Figures 4C–E). In conclusion, we speculated that activation of the TMA/FMO3/TMAO pathway is an important incentive for atherosclerosis, and PH inhibits TMAO generation through the involvement of hepatic FMO3.

To explore the microbial mechanism, 16S rRNA gene sequencing with a 97% similarity criterion for operational taxa was acquired from 36 samples. After sequence quality control and splicing, 3375 operational taxonomic units (OTUs) were identified. The rank-abundance curve declined smoothly and eventually flattened (Supplementary Figure S1A). The species cumulative and rarefaction curve revealed that the sequences per sample were sufficient and the sequencing depth was qualified for analysis later (Supplementary Figures S1B, C). Alpha (α) diversity analysis revealed that microbial diversity and abundance were reduced after an HFD intervention (Supplementary Figures S2, S3; Supplementary Tables S5, S6; Figures 5A–C); however, this trend was reversed or attenuated after PH administration (Figures 5D, E; Supplementary Figure S7). The gut microbiota characteristics were explored and analyzed at different levels, and 11 phyla, 14 classes, 32 orders, 50 families, 116 genera, and 201 species were identified. Moreover, the detail of distribution of different species were shown in Supplementary Figures S4–S6. At the phylum level, the proportion of Desulfobacterota markedly increased in the model group, while decreased in the PHM, PHH and simvastatin groups (Figure 5E).

Diversity analysis was performed to identify the alterations of gut microbial structure. The PCoA plot showed a remarkable separation among the four groups (control, model, PHH, and simvastatin) along the PC2 axis, which explained 13.96% of the variation in the microbial structure (Figure 6A). The segregated microbial structure among the three groups (control, PHH, and simvastatin) was consistently visualized using non-metric multidimensional scaling (NMDS) analysis (Figure 6B). The microbiota composition was clustered by partitioning around the medoids, and three enterotypes were identified. We identified that the proportion of Lactobacillus increased with the administration in PH or simvastatin (Figure 6C). Notably, the enterotypes in the PHH and simvastatin groups were mainly clustered into g_Lactobacillus and g_norank_f_Muribaculaceae_1 group, indicating that PHH and simvastatin reversed the HFD-induced enterotype switch (Supplementary Figure S8). Presumably, PHH and simvastatin protected the atherosclerotic mice against microbes. Accumulating evidence indicates that the HFD-induced decrease in gut microbial diversity can be restored to some extent by PH treatment. Bacteroidetes and Bacteroides were enriched in the control group, whereas Alistipes, f_Clostridiaceae, and f_Defluviitaleaceae were the predominant genera in the model group. PH administration significantly improved the proportion of p_ Actinobacteria, f_Atopobiaceae, and g_Coriobacteriaceae_UCG-002, whereas it decreased the proportions of f_Clostridiaceae and f_Defluviitaleaceae in the model group mice (Supplementary Figure S9). Finally, the multigroup difference analysis suggested that HFD increased the abundance of Antinobacteriota, Desulfobacterota, and Coriobacteriaceae_UCG-002 in the model group and reduced the abundance of Bacteroidetes, Bacteroides, and norank_f_Mribaculaceae, whereas PH treatment reversed these changes (Figures 6D–J).

We further conducted the metagenomic sequencing analysis to investigate the core functional capacity of different genes. Based on the LEfSe analysis, we observed that 37 differential KEGG pathways were identified in PHL and PHH groups, and ribosome, protein export, and RNA polymerase pathways were enriched in the model group. Most of the 37 differentially expressed functional pathways were highly enriched in the carbohydrate metabolism and membrane transport-related pathways. Among these, ABC transporters, carbon metabolism, glycerolipid metabolism, and glucagon signaling pathways were identified in the PHL group, and the biosynthesis of amino acids, starch, and sucrose metabolism were increased in the PHH group (Supplementary Figure S10A). Additionally, fatty acid metabolism, fatty acid and lipopolysaccharide biosynthesis were enriched in the control group, whereas the enrichment was reduced in the PHH group. A similar trend was observed in the multi-group comparison analysis (Figures 7B–I). Based on these results, we investigated the enrichment of CAZymes by LEfSe analysis. At the class level, glycosyl transferases (GTs) and carbohydrate esterases (CEs) were highly enriched in the PHL and PHH groups, whereas glycoside hydrolases (GHs) were mainly present in the simvastatin group (Figure 7A). The HFD-induced abundance of CAZymes was restored by PH administration, at the family level. Several CAZymes, including GT2_Glyco_tranf_2_3 and CE1, were enriched in the PHH group (Supplementary Figure S10B).

We constructed a heatmap to visualize the associations between the gut microbiota and bioinformatic parameters. Desulfobacterota demonstrated a significant positive correlation with lipid levels (TG, TC, and LDL-C) and inflammatory cytokines (hs-CRP, IL-2, IL-1β, and IL-17A). Additionally, Firmicutes displayed a similar trend with TG, LDL-C, IL-2, and IL-1β (Figure 8A). Notably, Firmicutes, Lachnoclostridium and Turicibacter demonstrated significant positive correlations with TMA, TMAO, lipid, and inflammatory cytokine levels. Desulfovibrio, Muribaculum, and Allobaculum also correlated with lipid levels and inflammatory cytokines. Bacteroides, Prevotellaceae_Ga6A1_group, and Eubacterium_fissicatena_group negatively correlated with TMA, TMAO, lipid levels, and inflammatory cytokines (Supplementary Figure S11A), thereby indicating that these microbial biomarkers may be involved in the development of atherosclerosis.

Associations between atherosclerotic variables, the KEGG functional pathways, and gut microbiota were analyzed. At Level 3, ABC transporters demonstrated positive correlations with lipid levels (TG, TC, LDL-C, and VLDL-C) and inflammatory cytokines (IL-2, IL-6, IL-17A, IL-1β, hs-CRP, and TNF-β). In particular, LDL-C demonstrated a positive correlation with the metabolic pathways, including carbon, starch, and sucrose metabolism (Figure 8B). At level 2, ko00520, ko00030, and ko00550 demonstrated similar associations with the lipid and inflammatory cytokine levels (Figure 8C). At module level, M00001 (Glycolysis), M00016 (Lysine biosynthesis), and M00051 (De novo pyrimidine biosynthesis) demonstrated positive correlations with lipid and IL-1β levels (Figure 8D). Uncultured_bacterium_g_Turicibacter, and uncultured_bacterium_g_Desulfovibrio exhibited positive correlations with TMAO, lipids, and inflammatory cytokine levels, whereas Bacteroides_acidifaciens exhibited negative correlations with TMAO, lipid, and inflammatory cytokine levels (Supplementary Figure S10B). In addition, we observed that metabolic pathways, including starch and sucrose metabolism, and ABC transporters, including M00001, M00051, and M00003, exhibited a positive correlation with s_Desulfovibrio_fairfieldensis and s_Allobaculum_stercoricanis, but a negative correlation with Bacteroides spp. and s_Lactobacillus_murinus (Supplementary Figures S12A, B).

Considering the pathways involved in the choline metabolism, TMAO, and associated gut microbiota, TMA-producing bacteria were selected for gene set construction. The gene set included CutC-related bacteria, which may have contributed to the therapeutic effects of PH. Critical KEGG pathways among different groups were identified using linear discriminant analysis. Glycolysis/gluconeogenesis, fluid shear stress, atherosclerosis, drug metabolism, and the HIF-1signaling pathway were highly enriched in the PHH group (Figure 9F; Supplementary Figure S13). Further multigroup comparison analysis revealed that the abundance of glycolysis/gluconeogenesis and HIF-1signaling pathway markedly increased after PHH treatment (Figures 9B–E). Thus, an analysis of intergroup differences in metabolic pathways was conducted based on these findings. The abundance of 6-phospho-beta-glucosidase [EC:3.2.1.86], which is involved in the glycolysis/gluconeogenesis metabolic pathway. The abundance of phosphoglycerate mutase [EC:5.4.2.11] and fructose-bisphosphate aldolase [EC:4.1.2.13] were increased in the model group. In addition, other related enzymes, such as 6-phosphofructokinase [EC:2.7.1.90] and fructose-bisphosphatase [EC:3.1.3.11], may also exerted an important effect (Figure 9G). We hypothesized that phosphoglycerate mutase and fructose-bisphosphate aldolase were the main contributors of glycolysis/gluconeogenesis pathway. Moreover, the gene ontology (GO) functional annotation results suggested that the biological processes (BP) of each group were mainly enriched in metabolic and biosynthetic processes (Figure 9A).

Correlation analyses between metabolic pathways and TMA/TMAO revealed that ko00240, ABC transporters, and auxiliary activities positively correlated with TMA, suggesting that TMA-producing bacteria were mainly involved in the progression of atherosclerosis via metabolic pathways and related enzymes (Figures 9H–J). Moreover, we identified the top 20 species that contributed to the differences in pathways at level 3. Metabolic functions are involved in different pathways, including glycolysis/gluconeogenesis, microbial metabolism, biosynthesis of secondary metabolites, and carbon metabolism. Desulfovibrio, Prevotella, Muribaculum, Clostridium, unclassified_f_Lachnospiraceae, unclassified_o_Bacteroidales, unclassified_f_Muribaculaceae, and unclassified_f_Coriobacteriaceae were contributed to these metabolic pathways (Figures 9K, L).

Finally, we further investigated the association between the TMA-producing bacteria and TMA/TMAO levels. Interestingly, Lachnospiraceae_bacterium, Muribaculaceae_bacterium_Isolate-104_(HZI), Clostridiales_bacterium, bacterium_0.1xD8-71, Helicobacter_ganmani, Desulfovibrio_sp., Clostridium_sp._CAG:417, and Clostridium_sp._CAG:451 exhibited positive correlations with TMA and TMAO. However, Coriobacteriaceae_bacterium, Ileibacterium_valens, Dubosiella_newyorkensis, Olsenella_sp._KGMB02461, unclassified_g_Parabacteroides, and Lachnoclostridium_sp. An181 expression exhibited negative correlations with TMA and TMAO (Figures 10M, N). Specifically, the abundance of s_Clostridium_sp._CAG:557, s_Clostridium_cuniculi, and s_Clostridium_disporicum increased in the model group, while the HFD-induced increase in the abundance of Clostridium spp. Was partially restored by PH administration. In addition, similar changes were observed in certain gut microbial species, such as s_Lachnoclostridium_sp._An118, s_Muribaculaceae_bacterium_Isolate-104_(HZI), and s_unclassidied_g_Desulfobibrio (Figures 10A–L). We speculated that CutC-related bacteria, such as Lachnospiraceae, Muribaculaceae, Clostridium, and Desulfovibrio prompted the progression of atherosclerosis via glycolipid metabolic pathways.