The SPF experimental KM mice were all sourced from the Laboratory Animal Center of Guangxi Medical University (Certificate of Conformity: scxk (Gui) 2020–0003). All experiments were approved and supported by the Animal Research Ethics Committee of Guangxi University (Approval No. GXU-2022-326). All animal experiments were carried out in accordance with the “Guidelines for the Care and Use of Laboratory Animals” published by the National Institutes of Health.

There is a close connection between different extraction factors and the content of the target components to be extracted. First, several portions of Pulsatilla chinensis medicinal materials (50 g each) were weighed in parallel. The quality of the medicinal materials was tested and confirmed to meet the requirements; that is, the content of Anemoside B4 in the medicinal materials was ≥ 4.6%. In accordance with HPLC-SPD 20A (Shimadzu (Shanghai) Global Laboratory Consumables Co., Ltd., Shanghai, China), the following variables were investigated: extraction time (0.5 h, 1 h, 1.5 h, 2 h, 3 h), extraction time (once, twice, three times, four times, five times), ethanol volume (50, 60, 70, 80, 90%), solid–liquid ratio (1:5, 1:10, 1:15, 1:20, 1:25), extraction temperature (50°C, 60°C, 70°C, 80°C, 90°C), and particle size of the medicinal materials (whole plant, decoction pieces, 30 mesh, 50 mesh, 80 mesh). Finally, the extracted liquid was filtered and then concentrated under reduced pressure to 50 mL. After that, the content was detected by HPLC.

The three factors with the most significant influence among the above variables were screened out for the response surface methodology (RSM) experiment. The percentage of Anemoside B4 in the extract (Y) was taken as the response value of the response surface optimization curve. Design Expert 10 was used to design the experiment. According to the 17 schemes simulated by the three factors and three levels designed in the experiment, a response surface optimization curve analysis experiment was carried out to obtain the optimal extraction process and prepare the PCE for subsequent studies.

Three S. aureus model strains, including CMCC (B) 26003 (National Center for Medical Culture Collections, Beijing, China), ATCC 25904, and USA 400 (feed from the Department of Veterinary Pharmacology, Jilin University, Jilin, China), and one S. aureus clinical isolate, GXU 2017 (isolated, identified and preserved in Guangxi Key Laboratory of Animal Breeding Disease Control and Prevention, Nanning, China), were used in this in vitro study, cultured at a constant temperature of 37°C for 12 h and configured into bacterial suspensions for backup.

Forty 7-week-old male KM mice and eighty 7-week-old female KM mice were housed in a single mouse cage at a male-to-female ratio of 1:2 and were allowed to mate freely to obtain pregnant female mice. Considering animal welfare, the experiment was carried out on the 15th day after the female mice gave birth (by which time the suckling mice had already acquired the ability to survive independently) (11).

The 80 female mice were divided into two physiological cycles, namely, the lactation period and the weaning period. For the lactation period, the maternal mice were separated from the suckling mice 3 h before the experiment (19), whereas for the weaning period, the maternal mice were separated from the suckling mice 5 days before the experiment. There were 40 female mice in each physiological cycle, and each cycle was randomly divided into a control group (n = 8) and four gradient S. aureus model groups (n = 8, with concentrations of 1 × 10⁵, 1 × 10⁶, 1 × 10⁷, and 1 × 10⁸ CFU/mL). Referencing the previous method and optimizing it (11), female mice were anesthetized after a 12-h fasting period. The breasts and the surrounding skin were disinfected with 75% alcohol. The mammary ducts of the fourth pair of nipples were exposed. A flat-tip microsyringe was used to inject 50 μL of S. aureus suspension into the mammary ducts of the mice in the model group, while the same volume of normal saline was injected into those in the Control group. Gentle massage was performed to promote the uniform distribution of the injected liquid. After 24 h, blood and mammary gland tissues were collected and stored at −80°C.

Sixty 5-week-old KM mice (with an equal number of males and females) were provided with free access to water and were fed standard feed according to the standard of 10–15% body weight. Six gradient groups were randomly set up according to the dose of PCE (n = 10, with an equal number of females, and the experiment was carried out in separate cages), namely, 0 (Control), 1,250, 2,500, 5,000, 10,000 and 20,000 mg/kg. After a single intragastric administration, the clinical symptoms, food intake, water intake and average weight gain of the mice were continuously observed for 7 days. After the experiment, the mice were sacrificed, and pathological changes in the internal organs were observed. The aim of this study was to investigate the oral safety of PCE.

Sixty lactating female mice after giving birth were randomly divided into six groups, namely, the Control group (n = 10, continuously gavaged with sterile water for 7 days and then injected with 0.9% NaCl into the milk ducts on the 7th day), the S. aureus group (n = 10, continuously gavaged with sterile water for 7 days and then injected with 1 × 10⁷ CFU/mL S. aureus suspension into the milk ducts on the 7th day), the dexamethasone (DEX) control group (n = 10, continuously gavaged with 5 mg/kg of DEX for 7 days and then injected with 1 × 10⁷ CFU/mL S. aureus suspension into the milk ducts on the 7th day) (20, 21), and the PCE 100, 200, 400 mg/kg dose groups (n = 10, continuously gavaged with different doses of PCE for 7 days and then injected with 1 × 10⁷ CFU/mL S. aureus suspension into the milk ducts on the 7th day). The doses of PCE were determined with reference to previous reports and preliminary experiments (22). Blood and mammary gland tissues were collected and stored at −80°C.

Aseptic procedures were performed with a super clean bench. After the mice were anaesthetized, blood was collected from the eyeballs. The first 3 drops of blood were placed into an anticoagulant tube and diluted 5 times the volume of sterile normal saline. Then, 100 μL of the diluted sample was spread on nutrient agar medium. The mixture was incubated at a constant temperature of 37°C for 24 h, and the bacterial load in the blood was calculated.

Ten milligrams of mammary gland tissue was homogenized with 100 times the volume of sterilized normal saline. One hundred microlitres of the tissue homogenate was spread on nutrient agar medium. The mixture was incubated at a constant temperature of 37°C for 24 h, and the bacterial load in the mammary gland tissue was calculated.

The protocol was performed in accordance with the requirements of the instruction manual. Enzyme-linked immunosorbent assay (ELISA) kits were used to evaluate the levels of interleukin-6 (IL-6), interleukin-1β (IL-1β) and tumor necrosis factor alpha (TNF-α) in blood and mammary gland tissues (Shanghai Enzyme Linked Biotechnology Co., Ltd., Shanghai, China).

In accordance with the requirements of the kit instruction manual, serum and mammary gland homogenates were collected to detect the expression level of myeloperoxidase (MPO) (NanJing JianCheng Bioengineering Institute, Nanjing, China).

The isolated mammary gland tissues were rinsed with normal saline to reduce the stagnation of milk and then placed in 10% formalin fixative for 24 h. The tissues were dehydrated with gradient ethanol and then embedded in paraffin blocks. The paraffin blocks containing the tissues were cut into slices with a thickness of 3 μm and incubated at 65°C for 4.5 h. The samples were subsequently dewaxed and rehydrated. Finally, the tissues were stained with dyes, and the slides were mounted. The prepared hematoxylin–eosin (HE)-stained pathological sections were used for histological observation and scoring of the mammary gland tissues (19).

The contents of the mouse caecum were collected in sterile freezing tubes, snap-frozen in liquid nitrogen and stored, and total DNA was extracted from the contents of the caecum using a DNA extraction kit (AU46111-96, BioTeke, China). Total DNA was amplified by PCR via the primers 341F/805R (341F: 5′-CCTACGGGNGGCWGCAG-3′; 805R: 5′-GACTACHVGGTATCTAATCC-3′; GACTACHVGGTATCTAATCC-3′). PCR amplification of total DNA was performed via the primers 341F/805R (341F: 5′-CCTACGGNGGCWGCAG-3′; 805R: 5′-CACTACHVGGTATCTAATCC-3′), and the resulting PCR products were purified, quantified and analyzed sequentially. Sequencing services were provided by LC-Bio Technology Co., Ltd., Hangzhou, China.

Each experiment was replicated in triplicate and the data derived therefrom were collected and calculated with Excel. Unless otherwise explicitly stated, all data were tested for normality distribution using GraphPad Prism 10.1.2 (GraphPad Software Inc., La Jolla, CA). In the case of normally distributed data, one-way ANOVA and Tukey’s multiple comparison test were employed to analyze and compare the significance of the experimental results, followed by statistical analysis and graphing. A p-value of less than 0.05 was regarded as statistically significant.

Taking the content of Anemoside B4 in the PCE as the standard, through a climbing test, it was found that the extraction time (Figure 1B), number of extractions (Figure 1C), and volume of ethanol (Figure 1D) had the most significant impacts on the target components in the PCE. Seventeen RSM tests were carried out on the above three factors and the content (%) of Anemoside B4 (Tables 1, 2).

Multiple regression analysis yielded a fitting equation (Y = 29.62 + 1.07A + 2.08B + 1.86C + 0.38AB – 0.92 AC + 0.91 BC – 0.79A²–2.59B²–1.45C²). In the analysis of variance, the model was extremely significant (p < 0.01), and the lack of fit was not significant (p > 0.05) (Table 3), indicating that the results were reasonable and could be used for the analysis and prediction of the PCE content under multiple factors. As seen from the F values (Table 3), the number of extractions and the volume of ethanol had relatively large effects on the content of Anemoside B4 in the PCE. In the plot of residuals versus predicted values, the data points showed a random distribution (Figure 1G), and in the normal probability plot, all the points were distributed on a straight line (Figure 1H), which could accurately reflect the authenticity of the experiment. The RSM is a three-dimensional spatial surface graph composed of three factors, namely, extraction time, the number of extractions, and the volume of ethanol. The steeper the slope is, the more significant the impact (Figures 1J–L). The contour plots all showed irregular ellipses (Figures 1M–O), so the interactions among the three factors were significant.

On the basis of the results of the climbing test and RSM, the optimal extraction process for determining the PCE was determined. The content of Anemoside B4 in the PCE extract was quantified by HPLC as 31.15 (Table 4). One kilogram of Pulsatilla chinensis medicinal material was added, 10 times the amount of 75% ethanol was added, hot reflux extraction was performed at 80°C for 70 min, suction filtration was carried out, and the filtrate was temporarily preserved. The filtrate was extracted three times, combined, and then concentrated under reduced pressure to 1 L. After vacuum drying (at 70°C and 0.095 Mpa), it was processed into powder to obtain 255 g of PCE. Then, 1 g of PCE was equivalent to 4 g of the original medicinal material, and the content of Anemoside B4 in the PCE powder was measured to be 65.17%.

The diameter of the inhibition zone of the PCE against different S. aureus strains was investigated by the Oxford cup method (Figure 2A). When the concentration of S. aureus was 1 × 10⁷ CFU/mL, the antibacterial effect was better, and the edge of the inhibition zone was clear. Therefore, a concentration of 1 × 10⁷ CFU/mL was selected for subsequent studies. For the determination of the MIC and the MBC, the MIC values of PCE against S. aureus (1 × 10⁷ CFU/mL) CMCC (B) 260,003, ATCC 25904, USA 400 and GXU 2017 were 31.25, 31.25, 15.625 and 62.5 mg/mL, respectively, and the MBC values were 31.25, 31.25, 31.25 and 62.5 mg/mL, respectively. The results revealed that when PCE was cocultured with S. aureus, it effectively inhibited the growth of different strains of S. aureus within 8 h (Figures 2B–E). The above results indicated that PCE exhibited favorable anti-S. aureus effects in vitro.

To investigate the protective effect of PCE on mastitis, we examined mice in two physiological cycles, namely, the lactation and weaning periods, and established a mastitis model on the basis of their sensitivity to invasion by different concentrations of S. aureus (GXU2017) (Figure 3). Compared with those in the weaning period, the bacterial loads in both mammary gland tissues (Figure 3B) and blood (Figure 3C), which exhibited a concentration-dependent pattern, were greater in the lactating period than in the weaning period, and the bacterial load in mammary glands was much greater than that in blood. Compared with those in the control group, the expression levels of IL-6 (Figures 3D,E) and MPO (Figures 3F,G) in the serum and mammary gland tissues of lactating mice were significantly greater (p < 0.01). In contrast, the expression of these indicators in weaning mice was significant only under high-concentration bacterial infection. Histological analysis demonstrated that (Figure Compared with those in the weaning period), the mammary gland tissue of lactating mice presented larger-volume acini, thicker acinar walls and a more complete honeycomb structure, more pronounced inflammatory cell infiltration, acinar wall hyperplasia and more severe pathological alterations. In conclusion, an experimental surrogate model of dairy cow mastitis was established by infecting lactating mice with 1 × 10⁷ CFU/mL S. aureus for 24 h.

Before a drug is taken orally, its safety needs to be determined. Therefore, we conducted a 7-day oral safety evaluation. When the oral concentration of PCE was less than 20,000 mg/kg, there was no significant difference (p > 0.05) in water intake (Figure 4A) or weight gain (Figure 4B) between the test groups and the control group. No death or poisoning phenomenon occurred, and the mice were in a good mental state with a normal appetite and thirst. No obvious pathological changes were found in the internal organs during necropsy. In summary, PCE has excellent oral safety in vivo.

To evaluate the effect of PCE in the mouse mastitis model, S. aureus was inoculated into the mammary ducts after continuous intragastric administration of PCE for 7 days (Figure 5A). Studies have shown that pretreatment with PCE (100, 200, or 400 mg/kg) can significantly reduce the excessive expression of bacteria (Figures 5B,C) and inflammatory factors (IL-6, IL-1β, TNF-α and MPO) (Figures 5D–K) in mammary gland tissues and blood (p < 0.01). Histological analysis revealed that the invasion of S. aureus into the mammary gland damaged the tissue structure of the mammary gland and induced a large number of infiltrating inflammatory cells and colloid secretions in the alveolar cavity (Figures 5L,M). In mice pretreated with PCE, the tissue structure of the mammary gland was relatively intact and clear, and only a small number of inflammatory cells adhered to the alveolar walls and around the lobular ducts. Overall, the 200 mg/kg dose had the greatest effect. These results indicate that oral administration of PCE can significantly reduce the bacterial load and inflammatory response in blood and mammary gland tissues induced by S. aureus and alleviate damage to mammary gland tissues.

We investigated the microbiome of the caecal contents of the mice. Principal coordinate analysis (PCoA) revealed that the microbiota structures of the CON group and the PCE group were relatively similar, while there were significant differences in the microbiota structures between the DEX group and the Model group compared with those of the CON group (Figure 6B). Therefore, it is unreasonable to simply determine whether it is beneficial or harmful to health on the basis solely of the increase or decrease in the number of amplicon sequence variants (ASVs). The number of ASVs in the gut microbiota represented only a quantitative change, as demonstrated by the greater number of ASVs in the DEX group and the Model group (Figure 6A). The species richness at the phylum level is shown in Figure 6C. Firmicutes and Bacteroidota constituted the main framework of species at the phylum level (accounting for ≥80%). Firmicutes are mainly involved in energy acquisition, intestinal homeostasis, nutrient absorption, and immune regulation, while Bacteroidota are mainly involved in food digestion, nutrient synthesis, intestinal immunity, and body metabolism. The ratio of the two is considered a key indicator of intestinal health status. They work together to maintain the balance of the intestinal microecology. An imbalance of either one may lead to intestinal dysfunction, which in turn can trigger a variety of health problems (23). A significant difference analysis at the phylum level revealed that, compared with those in the CON group, the relative abundances of Proteobacteria, Deferribacterota and Planctomycetota in the intestines of the Model and DEX groups were significantly greater (Figure 6D). Moreover, the PCE group presented a framework similar to that of the control group at the phylum level. Interestingly, Planctomycetota appeared only in the intestines of the mice in the Model group and the DEX group but not in those of the other two groups. The proportional and chord diagrams (Figure 6E) more clearly demonstrated the distribution proportions of the dominant microbiota in each group at the phylum level. In addition, through linear discriminant analysis effect size (LEfSe) analysis, significant differences in the species of the gut microbiota among the different treatment groups of mice were detected (Figures 6F,G). Notably, both the DEX group (purple) and the Model group (green) presented a large number of species with significant differences, and these differences may affect the health of the animals. A correlation clustering-labeled heatmap was used to evaluate the correlations between the gut microbiota and multiple phenotypes in mammary gland tissues (Figure 6H) and blood (Figure 6I). The abundances of Actinobacteriota and Bacteroidota were more negatively correlated with the relevant phenotypes (p < 0.05). Therefore, these microbiota may have inhibited the occurrence of mastitis to some extent. However, the abundances of Firmicutes, Fusobacteriota and Proteobacteria were more positively correlated with the relevant phenotypes (p < 0.05), indicating that they promoted the occurrence of mastitis. These differential results are expected to become biomarkers for the clinical diagnosis of mastitis.