The study was conducted at the Angus beef cattle ecological farm of Meizhou Guangshunhai Food Co., Ltd. Twenty-seven Black Angus steers, aged approximately 22 months (±1) and with an average body weight of 520 ± 40 kg, were selected for the experiment. These Angus steers were randomly allocated into three treatment groups, each consisting of nine steers, in accordance with a single-factor completely randomized design. A pre-feeding period of 14 days was followed by a 90-day experimental phase. Throughout the experimental phase, each group received a total mixed ration (TMR) tailored to the specified dietary net energy for maintenance (Ne_mf): 6.657 MJ/kg for the LE group, 7.323 MJ/kg for the ME group, and 7.990 MJ/kg for the HE group. TMR diets were formulated according to the Chinese Feed Composition and Nutritional Value Tables (31st edition) (Xiong et al., 2020). Feeding was conducted twice daily at 9:00 a.m. and 4:00 p.m., and water was available ad libitum. Uniform management was applied to all experimental cattle. Details regarding the composition and the nutritional content of the dietary ingredients are shown in Table 1.

Throughout the main trial, fresh samples of the TMR mixed diets were collected on days 1, 20, 40, 60, 80, and 90. Leftover feed was collected the following morning before feeding and was then dried in a 65°C oven to constant weight. Subsequently, the dried feed samples were ground, sieved through a 40-mesh sieve, and stored in polyethylene bags. Stomach fluid samples were collected from the cattle on days 1, 45, and 90 prior to the morning feeding. During each sampling, approximately 150 mL of stomach fluid was collected. After collection, stomach fluid pH was immediately measured using a pH meter (FE28-Standard, METTLER-TOLEDO Corporation), and samples were then transferred to 2 mL cryotubes for storage in liquid nitrogen. The remaining stomach fluid was filtered through four layers of sterile gauze, aliquoted into 20 mL centrifuge tubes, and centrifuged at 5,000 rpm for 15 min. Subsequently, the supernatant was stored at −20°C for further analysis of stomach fluid fermentation parameters and 16S DNA sequencing. On day 91 of the trial, three cattle from each of the LE, ME, and HE groups were randomly selected for slaughter. Before slaughter, the cattle underwent a 16-h fasting period and a 3-h water restriction period. After slaughter, samples were collected from the longissimus muscle between the left 12th and 13th rib intercostal spaces. Measurements taken included the eye muscle area, meat color, shear force (after 24 h of cold storage), centrifugal moisture loss, and drip loss. Portions of the meat samples were sealed in bags and stored at −20°C for subsequent analysis of meat nutritional indicators.

Throughout the main trial, at a 5-day interval and specifically at 8:30 a.m., feed remnants from each group were collected, weighed, and meticulously recorded. These remnants were then subjected to a 48-h drying process in a 65°C oven, after which they were reweighed to determine their final weight. Consumption was calculated as the difference between the total feed quantity provided the previous day and the weight of the remaining feed the following day. Furthermore, on days 1, 45, and 90 of the trial, prior to morning feeding, cattle were individually weighed using electronic scales to ensure precision within a margin of error of less than 0.5 kg, and these measurements were accurately recorded for analysis.

The nutrient composition of the feed, including crude protein (CP), dry matter (DM), and ether extract (EE), was determined using the methods of the Association of Official Analytical Chemists (AOAC, 1990). The acid detergent fiber (ADF) and neutral detergent fiber (NDF) content in the feed were analyzed using the method described by Van Soest et al. (1991), with a fiber analysis instrument. Starch, calcium (Ca), and phosphorus (P) levels were assessed using near-infrared spectroscopy (NIRS). The net energy content of the feed was calculated using the methodology outlined in the “Institute of Animal Husbandry et al. (2004).” For meat quality assessment, the eye muscle area was outlined on a grid sulfuric acid paper between the 12th and 13th rib cuts, and then, the area was calculated. After slaughter, meat samples were stored under low light conditions for 45 min, followed by color analysis using a calibrated colorimeter (NR10QC, 3nh Company). Five distinct points on each sample were selected for color measurement, and the average value was recorded. Shear force was measured by taking meat samples with a core sampler, sealing them in bags, and heating in a 75°C water bath until the center of the meat reached 70°C. Following cooling, surface moisture was blotted dry with filter paper. Meat strips, 1 × 1 cm in cross-section, were cut perpendicular to the muscle fibers using a texture analyzer (tenderness meter). For each meat sample, 10 measurements were taken and the average was calculated. Drip loss and centrifuged water loss in meat samples were evaluated following the 2015 “Objective Evaluation Method for Meat Quality” from the Agricultural Industry Standards of the People’s Republic of China. Frozen meat samples were pulverized and then sieved through an 18-mesh screen. The CP and EE content in meat samples were analyzed using AOAC (1990) methods. The analysis of amino acid and fatty acid profiles in the meat samples was conducted by Shanghai Tianxiang Quality Technology Service Co., Ltd.

For the analysis of the rumen fluid, volatile fatty acid (VFA) concentrations were determined using gas chromatography, following the method described by Erwin et al. (1961), with an Agilent 6890B gas chromatograph (Leiden Scientific Instruments Co., Ltd., Suzhou City, Jiangsu Province, China) equipped with an HP-INNOWAX capillary column (30.0 m × 320 μm × 0.5 μm, Catalog No: 19091 N-213). Ammonia nitrogen (NH₃-N) concentration in the rumen fluid was measured using a colorimetric method employing a Biotron Synergy H1 enzyme-linked analyzer (Guangzhou Huqiao Instrument Technology Co., Ltd., Guangzhou City, Guangdong Province, China).

Genomic DNA was extracted from rumen fluid samples using magnetic bead-based extraction, followed by an evaluation of DNA purity and concentration. Upon confirming that the standards were met, the samples were diluted with sterile water to achieve a concentration of 1 ng/μl. Primers specific to the 16S V3-V4 region (515F and 806R) were used for amplification. The PCR mixture, consisting of 15 μL Phusion® High-Fidelity PCR Master Mix (New England Biolabs), 0.2 μM primers, and 10 ng genomic DNA template, was subjected to an initial denaturation at 98°C for 1 min, 30 cycles of denaturation at 98°C for 10 s, annealing at 50°C for 30 s, extension at 72°C for 30 s, and a final extension at 72°C for 5 min. Gel electrophoresis was used to analyze the PCR products, followed by enzymatic quantification of the purified products. After quantification, samples were pooled in equimolar ratios and were subjected to electrophoresis for the confirmation of target band presence. Library construction was carried out using the NEB Next® Ultra™ II FS DNA PCR-free Library Prep Kit (NEB/E7430L). After library preparation, quantification was conducted using Qubit and quantitative PCR (Q-PCR) methods. Once qualified, the libraries underwent paired-end (PE) 250 sequencing on a NovaSeq 6,000 platform. Following sequencing, sample data were demultiplexed based on barcodes and PCR primer sequences. Barcodes and primer sequences were removed using FLASH software (Version 1.2.11) (Magoč and Salzberg, 2011). Raw tags for each sample were then assembled into original tag data (raw tags). Raw tags were subjected to rigorous filtering and quality control using fastp software (Version 0.23.1) (Bokulich et al., 2013), resulting in high-quality tag data (clean tags). Following processing, clean tags were further analyzed to eliminate chimeric sequences. Tag sequences were aligned with species annotation databases (Silva for 16S/18S, Unite for ITS) to detect and remove chimeric sequences, yielding the final effective data (effective tags) (Edgar et al., 2011). Effective tags underwent denoising, employing the DADA2 module or deblur within QIIME2 software (version QIIME2-202006, defaulting to DADA2), to obtain final amplicon sequence variants (ASVs) and a feature table (Wang et al., 2021). Taxonomic annotation of ASVs was performed using QIIME2 software and Silva databases 138.1 for 16S/18S and Unite v8.2 for ITS. Rapid multiple sequence alignment of all ASV sequences was conducted using QIIME2 software for phylogenetic tree construction. Data from all samples were normalized based on the sample with the lowest data volume, followed by alpha and beta diversity analyses using the normalized data.

Experimental data were initially processed and formatted using Microsoft Excel. The statistical analysis was performed using the one-way analysis of variance (ANOVA) in SPSS 2.0, with post-hoc comparisons conducted via LSD and Duncan’s multiple range tests. The p-value was used to determine statistical significance among the three groups. Experimental data are presented in tables as mean ± standard error of the mean (SEM), with significance denoted by p < 0.05.

Figure 1 illustrates that, during the experimental period, the HE group showed a 1.1% decrease in the average daily matter intake (ADMI) compared to the ME group (p < 0.05) and a 4.4% decrease compared to the LE group (p < 0.05). The ME group exhibited a 3.4% decrease in ADMI relative to the LE group (p < 0.05). Notably, ADMI in all groups demonstrated an increasing trend throughout the experimental period. Table 2 reveals that, in terms of overall daily weight gain (ADG), the HE group showed a 25.6% increase compared to the LE group (p < 0.05), and the ME group exhibited a 30.5% increase (p < 0.05). Regarding the overall feed-to-weight ratio, the LE, ME, and HE groups recorded ratios of 14.78, 11.32, and 11.17, respectively. The HE and ME groups exhibited reductions in feed-to-weight ratio of 24.4 and 23.4%, respectively, compared to the LE group. In the 0–45 day period, the feed-to-weight ratios for the LE, ME, and HE groups were 15.92, 13.35, and 10.30, respectively. The HE and ME groups showed reductions in the feed-to-weight ratio of 35.3 and 16.1%, respectively, compared to the LE group. During the 45–90 day period, the feed-to-weight ratios for the LE, ME, and HE groups were 14.78, 10.42, and 13.24, respectively. The HE and ME groups exhibited feed-to-weight ratio reductions of 10.4 and 29.5%, respectively, compared to the LE group.

Across different experimental groups with varying dietary energy levels, there were no significant differences in parameters such as the eye muscle area, shear force, centrifugal dehydration rate, drip loss rate, meat color, and crude protein. However, the Longissimus dorsi muscle of the HE group exhibited a significantly higher crude fat content compared to the ME (p < 0.05) and the LE groups (p < 0.05). According to the analysis results, dietary energy levels had no significant impact on sensory evaluation indicators of beef, as shown in Table 3. Table 4 shows that varying dietary energy levels did not significantly impact the amino acid levels in the Longissimus dorsi muscle of Angus steers. However, dietary energy levels significantly affected the total fatty acid content in the muscle. The muscle of the HE group had a significantly higher total fatty acid content compared to both the ME and LE groups (p < 0.05). For unsaturated fatty acids, the HE group’s linoleic acid (octadecadienoic acid) content was significantly higher than in the ME and LE groups (p < 0.05). Additionally, pentadecanoic acid and arachidonic acid were exclusively found in the muscle of the HE group.

Table 5 shows the fermentation parameters of the rumen fluid for each experimental group. Dietary energy levels were observed to have no significant effect on the rumen fluid pH and NH₃-N content but significantly impacted various volatile fatty acids. In the LE group, the concentration of acetic acid in the rumen fluid was significantly higher compared to the ME group (p < 0.05) and the HE group (p < 0.05). The concentration of propionic acid in the HE group’s rumen fluid was significantly higher than in both the LE and ME groups (p < 0.05). Valeric acid concentration in the HE group was significantly higher than in the ME group (p < 0.05), and isovaleric acid concentration was notably higher compared to both the LE and ME groups (p < 0.05). The total volatile fatty acid concentration was significantly higher in the HE group compared to the ME group (p < 0.05). Regarding the acetic acid/propionic acid ratio, the LE group exhibited significantly higher values than both the ME and HE groups (p < 0.05). As dietary energy levels increased, there was a decrease in the molar ratio of acetic acid (p < 0.01) and an increase in the molar ratio of propionic acid (p < 0.01).

According to Figure 2, sequencing of the V3 and V4 regions in this experiment resulted in 2,091,965 raw data reads, with 1,747,504 high-quality sequences obtained after quality control. The sequences were clustered into 8,353 operational taxonomic units (OTUs) using a 97% similarity threshold. The clustering diagram revealed that the LE group possessed the highest number of unique OTUs, indicating a more diverse rumen microbiota compared to the ME and HE groups. Notably, the HE group demonstrated the lowest diversity in rumen microbiota.

Table 6 indicates that the LE group displayed significantly higher values in the chao1, observed_features, Shannon, and Simpson metrics compared to the HE group (p < 0.01). Additionally, the LE group exceeded the ME group in chao1, observed_features, and Shannon metrics (p < 0.01). Conversely, the ME group outperformed the HE group in all four metrics, although these differences were not statistically significant. Therefore, it is evident that the LE group possesses the highest diversity in rumen fluid microbiota.

In Figure 3, the ME and HE groups demonstrate similar distribution patterns, characterized by closely clustered samples within each group. The LE group, while showing some overlap with the ME and HE groups, predominantly occupies a distinct region. This suggests that increased dietary energy levels have, to some extent, altered the structure of the rumen microbiota. According to Tables 7, 8, a smaller Observe-Delta value signifies less variation within groups, whereas a larger Expect-Delta value indicates more pronounced differences between groups. A positive A value signifies that differences between groups exceed those within groups, while a negative A value suggests the opposite, with greater differences within groups than between groups. It is evident that significant differences exist between the LE group and both the ME and HE groups, while the differences between the ME and HE groups are not significant. This finding corroborates the previous observation of higher rumen microbiota diversity in the LE group.

The statistical analysis identified the top 20 bacterial relative abundances at both the phylum and genus levels. Comparisons were conducted among the different experimental groups to identify differences. Table 9 shows that the relative abundance of Proteobacteria in the HE group is significantly higher than in the LE group (p < 0.05) but not significantly different from the ME group. Relative abundances of Bacteroidota (p < 0.05), Spirochaetota (p < 0.05), and Fibrobacterota (p < 0.01) are significantly higher in the LE group than in the HE group. Table 10 indicates that the relative abundance of the Succinivibrio genus in the HE group is significantly higher than in both the LE and ME groups (p < 0.05). In the LE group, bacterial relative abundances of the Prevotella and Ruminococcus genera were significantly higher than in the HE group (p < 0.01). Bacterial relative abundances of the UCG-005, CAG-352, UCG-002, UCG-010, and Kroppenstedtia genera were all significantly higher in comparison to those in the HE group (p < 0.05).

Figure 4A reveals 18 distinct taxonomies among the experimental groups, encompassing 1 species, 5 genera, 6 families, 2 orders, 2 classes, and 2 phyla. Of these, 14 taxonomies were found in the LE group, while 2 were identified in each of the ME and HE groups. At the phylum level, a significant increase was observed in the relative abundance of Proteobacteria in the HE group and Bacteroidota in the LE group (p < 0.05). At the genus level, the LE group showed a significant increase in the abundance of Eubacterium_coprostanoligenes_group, Prevotella, and Muribaculaceae and a notable decrease in Rikenellaceae_RC9_gut_group (p < 0.05). In the evolutionary tree, the LE, ME, and HE groups exhibit distinct evolutionary pathways, indicating that varying dietary energy levels lead to different evolutionary directions in the rumen microbiota (Figure 4B).

Pearson correlation analysis was performed to analyze the correlation between fermentation parameter rates (FPR) and rumen microbiota, revealing associations between eight phyla and nine genera with rumen fermentation parameters (Figure 5). At the phylum level, Proteobacteria showed a positive correlation with isovaleric acid (p < 0.05), whereas Bacteroidota and Patescibacteria both exhibited negative correlations with isovaleric acid (p < 0.05). Spirochaetota and Desulfobacterota were negatively correlated with isobutyric acid (p < 0.05). Euryarchaeota was positively correlated with acetic acid and butyric acid (p < 0.05), while Fusobacteria negatively correlated with these acids (p < 0.05). Fibrobacterota showed a positive correlation with acetic acid (p < 0 0.05). At the genus level, Pseudomonas was positively correlated with isovaleric acid (p < 0.05). Conversely, the Eubacterium_coprostanoligenes_group and UCG-002 genera were negatively correlated with acetic acid and butyric acid (p < 0.01 for Eubacterium_coprostanoligenes_group; p < 0.05 for UCG-002). Similarly, Rikenellaceae_RC9_gut_group, Christensenellaceae_R-7_group, Saccharofermentans, Muribaculaceae, and F082 genera were all negatively correlated with isovaleric acid (p < 0.05 for Rikenellaceae_RC9_gut_group; p < 0.01 for others). The NK4A214_group was negatively correlated with both propionic acid (p < 0.01) and isovaleric acid (p < 0.05).