Cattle selected for this experiment were provided by a commercial beef cattle farm (Jiangxi Agriculture University Test Base, Xinyu, China). The glycyrrhetinic acid was provided by Beijing Centre Technology (Beijing, China). Twenty-four Simmental cattle (initial mean ± SE: 588 ± 54.6 kg of body weight, and 15 ± 1 month old) were randomly divided into two treatments (n = 12 per treatment): control treatment and GA treatment. All the cattle were fed individually, and the control treatment cattle were fed the standard diet, while those in the GA treatment group were fed the standard diet +0.1% glycyrrhetinic acid (the dosage of glycyrrhetinic acid added was determined from previous rumen in vitro fermentation experiments). Cattle received the feed two times a day at 8:00 a.m. and 4:00 p.m., and clean, fresh water was provided as required. All treatments lasted 80 days, and the first 7 days were allocated for adaptation. The specific formulation and nutritional components of the standard diet utilized throughout the study can be found in Table 1.

At both the commencement and conclusion of the experiment, each animal was weighed individually in the morning before they were given their feed. To determine the average daily gain (ADG), the total weight gain for each individual cattle was divided by the number of days the trial lasted. Additionally, the average daily dry matter intake (ADMI) was assessed by taking the total amount of diet provided to each cattle, subtracting the leftover feed, and then dividing this figure by the total duration of the trial in days. The residue was collected and weighed before morning feeding. The feed conversion rate was ADMI divided by ADG. On the concluding day of the experiment, blood samples were collected from the caudal vein of each cattle just before their morning feeding. After standing for 30 min, the blood samples underwent centrifugation at a speed of 3,500 revolutions per min for 15 min. Following this, the serum was isolated and preserved at −20°C for subsequent analysis. The levels of serum total protein (TP), albumin (ALB), globulin (GLB), total cholesterol (TC), triglycerides (TG), high-density lipoprotein (HDL), low-density lipoprotein (LDL), alanine aminotransferase (ALT), aspartate aminotransferase (AST), lactate dehydrogenase (LDH), and alkaline phosphatase (ALP), along with interleukin-4 (IL4) and interleukin-6 (IL6), were detected by automatic biochemical analyzer (3,100 Hitachi, China).

The diet offered and refused was measured and sampled daily. Between 78 and 80 days, the total amount of feces excreted by each cattle was recorded, and a fecal sample (1/10 of the wet weight) was collected and mixed with sulfuric acid solution (100 mL/L). TMR, refusals, and fecal samples were subjected to drying in an oven set at a temperature of 65°C for a duration of 48 h. After the drying process, the samples were carefully ground to achieve a uniform consistency that enabled them to pass through a 1-mm sieve for further analysis.

Before feeding on the 80th day, rumen fluid was collected (150 mL) utilizing an oral stomach tube, following the protocol established by Shen et al. (16). Following collection, the rumen fluid was immediately flash-frozen in liquid nitrogen to preserve its integrity and biochemical composition. The frozen samples were then securely stored at a temperature of −80°C for later analysis. In addition to the preservation of the samples, a portion of the rumen fluid was utilized to measure volatile fatty acids (VFAs), using the technique described by Hu et al. (17). In summary, ortho-phosphoric acid (25% w/v) was introduced to the rumen fluid sample prior to filtration and centrifugation (12,000 g for 10 min at 4°C), after which VFAs were examined using gas chromatography (GC-2014 Shimadzu, Japan).

The analysis of diet, refusals, and fecal samples focused on several parameters, including acid-insoluble ash (AIA), dry matter (DM), crude protein (CP), neutral detergent fiber (NDF), acid detergent fiber (ADF), and ether extract (EE). The AIA served a crucial role as an internal marker, facilitating the estimation of total fecal output, which is essential for determining apparent nutrient digestibility coefficients. To ensure the reliability of the AIA measurement, the procedure was conducted using the 2 N HCl method in triplicate, thereby enhancing the validity of the results obtained (18). In addition to the analyses involving AIA, all the remaining sample examinations were performed in duplicate to ensure the accuracy and reproducibility of data. The laboratory dry matter content was determined by subjecting the samples to a drying process at 105°C in a forced-air oven for a duration of 24 h. The ash content was evaluated to determine organic matter (OM) by incinerating the samples at 550°C for a period of 4 h (19). Neutral detergent fiber and acid detergent fiber were detected using the Van Soest method by the ANKOM²⁰⁰ fiber analyzer (A200i Ankom, United States), with the addition of sodium sulfite and α-amylase for the NDF procedure. Crude protein was determined using the Kjeldahl method by the FOSS Kjeltec 8,400 Nitrogen Analyzer (8,400 FOSS, Denmark). The ether extract was evaluated using Soxhlet extraction (20). Apparent total tract digestibility CP, NDF, ADF, and EE were determined from the following equation: 100–100 × [(C_AIA-feed/C_AIA-feces) × (C_{nutrient-feces}/C_{nutrient-feed})], where C_AIA-feed represents the concentration of AIA in the feed, C_AIA-feces represents the concentration of AIA in the feces, C_{nutrient-feces} represents the concentration of nutrient in the feces, and C_{nutrient-feed} represents the concentration of nutrient in the feed.

E.N.Z.A Stool DNA Kit method were used to extract total DNA from microbial sources in rumen fluid. The isolation process was carried out in strict adherence to the instructions provided by the manufacturer. The quantity and purity of isolated DNA were measured on an ND-1000 spectrophotometer (NanoDrop, United States). The amplification of the eukaryotic ribosomal RNA gene’s 16S rDNA V3–V4 region was achieved using polymerase chain reaction (PCR), incorporating an initial step at 95°C for 3 min, followed by 30 cycles consisting of 30 s at 95°C,30 s at 55°C, and 45 s at 72°C, culminating with a final extension at 72°C for 10 min (eventually stopped by the user). Specific primers were used in this process, namely 338F (5’-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5’-GGACTACHVGGGTWTCTAAT-3′). The PCR reactions were executed in triplicate to ensure the reliability and reproducibility of the results. Each 20 μL of the reaction mixture was carefully prepared, incorporating 4 μL of 5× FastPfu buffer, 2 μL of 2.5 mM dNTPs, and equal volumes of both the forward and reverse primers, each at a concentration of 5 μM. Additionally, the mixture contained 0.4 μL of FastPfu polymerase, 0.2 μL of BSA, and 10 ng of template DNA, which served as the starting material for the amplification process.

PCR product purification used an AxyPrep DNA Gel Extraction Kit (Axygen, United States), following the guidelines provided by the manufacturer. Additionally, quantification was performed with a Quantus™ Fluorometer (Promega, United States). Purified amplicons were pooled in equimolar and paired-end sequenced on an Illumina NovaSeq PE250 platform (Illumina, United States) according to the standard protocols by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). The initial sequence reads were submitted to the NCBI database (Accession Number: PRJNA884686).

The raw sequencing reads of the 16S rRNA gene underwent demultiplexing, and the quality of the sequences was assessed and filtered using fastp version 0.20.0 (21), and the filtered reads were merged using FLASH version 1.2.7 (22). Operational taxonomic units (OTUs) were grouped based on a similarity threshold of 97% (23, 24) utilizing UPARSE version 7.1 (23), with the identification and removal of chimeric sequences. Each representative sequence of the OTUs was classified taxonomically through RDP Classifier version 2.2 (25), referencing the 16S rRNA database (e.g., Silva v138) and applying a confidence threshold of 0.7.

The analysis of the experimental data was conducted using SPSS 26.0 (SPSS, United States), where independent sample t-tests were used to assess the significance of growth performance, nutrient apparent digestibility, and serum biochemical parameters. The final results are presented as mean values, with differences regarded as showing a tendency when 0.05 < p < 0.10 and considered statistically significant at a p-value of ≤0.05.

Table 2 illustrates the impact of GA on the growth performance of beef cattle. When compared to the control (Con) treatment, the inclusion of GA significantly enhanced the average daily gain (p < 0.05). The average daily dry matter intake in GA treatment indicates a potential to be greater than in the Con treatment (p < 0.10).

The effects of GA on the serum biochemical parameters of beef cattle are shown in Table 3. The concentrations of TP, ALB, GLB, TC, TG, HDL, LDL, AST, ALT, LDH, ALP, and UREA in the serum of the Con group were not significantly different from the GA group. The IL4 concentration of the GA group was significantly higher than that in the Con treatment (p < 0.05), and the IL6 concentration tended to increase (p < 0.10).

As shown in Table 4, the apparent digestibility of crude protein, neutral detergent fiber, and acid detergent fiber in the feed of the GA treatment was significantly higher when compared to the Con treatment (p < 0.05).

The rumen pH, isovaleric acid portion, and the ratio of acetic acid to propionic acid in the GA treatment were significantly lower than those observed in the Con treatment (p < 0.05, Table 5). In contrast to the Con treatment, the addition of GA significantly enhanced the total VFAs, the concentration of microbial proteins, and the content of propionic acid (p < 0.05).

The findings from sequencing analysis are illustrated in Figure 1A. Good coverage for the detected OTUs was 97.92 ± 0.09, and the rarefaction curves displayed distinct asymptotes, suggesting that the community had been nearly completely sampled. Additionally, the unweighted principal coordinates analysis (PCoA) demonstrated a clear separation between the two groups at the OTU level, with statistical significance reported (p < 0.01). The similarities in the bacterial community between samples were compared by ANOSIM on Bray–Curtis and revealed significant differences in rumen microbiota structure among different rumen fluid samples (Figure 1B). There were 2009 OTUs identified in GA and control treatments. There were 1954 OTUs and 1901 OTUs in the GA treatment and control treatment, respectively. A total of 1846 OTUs were identified in both treatments, comprising 91.89% of all OTUs. The Sobs index was notably greater in the GA treatment compared to the control treatment (p < 0.05), while the Shannon and Chao1 indices showed a trend toward being higher in the GA treatment relative to the control (Shannon p < 0.10; Chao1 p < 0.10). The GA treatment had a lower Simpson index and a higher Ace index, but neither was significantly different between the two treatments (Figure 1C).

In the analysis of rumen fluid samples, a total of 21 distinct bacterial phyla were identified, and those with relative abundance >1% included Bacteroidota, Firmicutes, Patescibacteria, Spirochaetota, Actinobacteriota, Fibrobacterota, and Proteobacteria (Figure 2). Bacteroidota and Firmicutes had the highest relative abundance in both treatments. The GA treatment exhibited a significantly lower relative abundance of Bacteroidota compared to the CON group (p < 0.01), whereas the relative abundance of Firmicutes was significantly greater (p < 0.01, Table 6). In the analysis of rumen fluid samples, a total of 21 distinct bacterial phyla were identified. The relative abundances of Patescibacteria, Actinobacteriota, and Proteobacteria increased by 30.92, 40.74, and 14.00%, respectively, compared with the Con group, but they were not significantly different. Additionally, the Bacteroidetes and Firmicutes ratio in GA treatment was significantly lower than that in Con treatment (p < 0.01).

A total of 241 bacterial genera were identified, 32 of which had a relative abundance >1% (Figure 3), including norank_f__Bacteroidales_RF16_group, norank_f__F082, norank_f__Muribaculaceae, norank_f__Bacteroidales_BS11_gut_ group, norank_f__Prevotellaceae, norank_f__Bacteroidales_UCG-001 and norank_f__p-251-o5 in the Bacteroidota phylum, norank_f__UCG-011, norank_f__Eubacterium_coprostanoligenes_ group, norank_f__Ruminococcaceae, norank_f__UCG-010, UCG-005, norank_f__norank_o__Clostridia_UCG-014 and UCG-004 in Firmicutes, norank_f__norank_o__Absconditabacteriales_SR1 in Patescibacteria, Treponema in Spirochaetota, and Fibrobacter in Fibrobacterota. The relative abundance of Prevotella (p < 0.05), NK4A214_ group (p < 0.05), norank_f__UCG-011 (p < 0.05), Prevotellaceae_UCG-003 (p < 0.05), Christensenellaceae_R-7_treatment (p < 0.01), Prevotellaceae_UCG-001 (p < 0.05), norank_f__Bacteroidales_UCG-001 (p < 0.05), Pseudobutyrivibrio (p < 0.01), and Butyrivibrio (p < 0.01) significantly differed between the two groups, and the relative abundance of norank_f__Bacteroidales_RF16_ group (p < 0.10) and Lachnospiraceae_NK3A20_ group (p < 0.05) tended to differ (Table 7).

To enhance our understanding of the critical function of microbiota in two distinct groups, the PICRUSt program was used to forecast our high-throughput sequencing data based on 16S rRNA. Additionally, the analysis was conducted with reference to the Cluster of Orthologous Groups (COG) database. It is worth mentioning that the GA treatment exhibited significant enrichment in areas such as carbohydrate transport and metabolism, amino acid transport and metabolism, and energy production and conversion (p < 0.05; Figure 4).