Daily feed offered was recorded for each sheep, and residual feed from the pre-vious day was collected and weighed before the morning feeding. Body weights were measured on days 0 and 30 of the formal trial period prior to morning feeding.

Calculation Formulas: Average Daily Gain (ADG) = (Final Body Weight – Initial Body Weight)/Trial Duration (days); Dry Matter Intake (DMI) = (Feed Offered – Residual Feed) × Dietary Dry Matter Content (%); Feed-to-Gain Ratio (F:G) = DMI/ADG.

The digestion and metabolism trial was conducted from days 25 to 29 of the experiment. Approximately 200 g of forage samples were collected using the five-point sampling method. During the trial period, residual feed from each experimental sheep was collected daily, weighed, air-dried, and then ground through a 40-mesh sieve. The total fecal collection method was employed by fitting custom-made fecal collection bags to the tail of each sheep. Fecal samples were collected every 4 h from all sheep, with fresh weight recorded continuously for 3 days. The daily collected samples were thoroughly mixed, and 10% of the total was preserved. For total urine collection, custom-made urine collection devices were attached to the urination site of the sheep prior to the digestion trial. Urine was collected every 4 h from all animals. The urine samples from each sheep during the trial period were homogenized, with 10% of the total volume preserved using 10% concentrated sulfuric acid for nitrogen fixation.

The moisture and crude protein contents in the feed and feces were determined according to the Chinese National Standards GB/T 6435-2014 and GB/T 6432-2018, respectively. The neutral detergent fiber and acid detergent fiber were analyzed using a fully automated fiber analyzer (ANKOM A2000i). Specific results are presented in Appendix 1.

Rumen fluid samples were collected via ruminal fistula at 0 h (pre-feeding), 2 h, 4 h, 6 h, 8 h, and 10 h post-morning feeding on days 0, 10, 20, and 30 of the trial. The samples were immediately filtered through four layers of gauze and analyzed using a pH meter. Additional rumen fluid samples (~200 mL) were collected on day 0 (pre-feeding) and day 30 (at 0 h, 2 h, 4 h, 6 h, 8 h, and 10 h post-feeding), filtered similarly, and processed for further analysis. Volatile fatty acids (VFA) were quantified using a Shimadzu GC-2010 gas chromatograph equipped with a Stabilwax column (30 m × 0.25 mm × 0.25 μm), with 4-methylvaleric acid as the internal standard. Ammonia nitrogen (NH₃-N) concentrations in rumen fluid collected on days 0 and 30 were determined according to the method described in reference (12).

On day 30 of the experiment, rumen fluid was collected via ruminal fistula, snap-frozen in liquid nitrogen, and subsequently subjected to 16S rDNA high-throughput sequencing. The specific methods are described in Appendix 2.

On day 30 of the experiment, rumen fluid was collected via ruminal fistula, snap-frozen in liquid nitrogen, and subsequently subjected to metabolomic analysis. For methodological details, please refer to Appendix 3.

Following AOC supplementation, the ruminal pH dynamics in sheep showed an initial decrease followed by a gradual increase (Figure 1A). The ruminal pH values in the trial group were higher than those in the control group at 2 h, 4 h, and 6 h post-feeding, indicating a trend toward increased ruminal pH with AOC supplementation.

On day 0, there were no significant differences in volatile fatty acid (VFA) or NH₃–N concentrations between the two groups (p > 0.05, Figure 1B). By day 30, the acetate and total VFA concentrations in the trial group were significantly higher than those in the control group (p < 0.05), while the NH₃–N level was 8.59% higher than that of the control, though this difference was not statistically significant (p > 0.05).

The postprandial ruminal VFA concentrations at different time points are shown in Figure 2. The acetate and total VFA concentrations in the trial group were significantly elevated at multiple time points (acetate: 0 h, 4 h, 8 h; total VFA: 0 h, 2 h, 4 h, 8 h) (p < 0.05). The acetate-to-propionate ratio was also significantly higher in the trial group (p < 0.05). Concentrations of other VFAs (propionate, butyrate, isobutyrate, valerate, isovalerate) showed an increasing trend, but the differences were not significant.

The alpha diversity of ruminal microbiota is illustrated in Figure 3. The sequencing coverage of all samples exceeded 99%, indicating sufficient sequencing depth (Figure 3B). The rarefaction curves plateaued, confirming adequate sequencing volume (Figure 3A). Venn diagram analysis revealed that the number of unique OTUs in the trial group (1,363) was greater than that in the control group (1,311), suggesting a potential increase in species richness (Figure 3C). However, no significant differences were observed in the alpha diversity indices (Chao1, Shannon, Simpson) between groups (p > 0.05), indicating that overall microbial diversity and evenness remained unchanged.

At the genus level, detectable taxa in the rumen microbiota included Succiniclasticum, Methanobrevibacter, Prevotella, Quinella, Veillonellaceae_UCG-001, Lachnospiraceae_NK3A20_group, Lachnospiraceae_XPB1014_group, Selenomonas, Rikenellaceae_RC9_gut_group, Solobacterium, with other genera present at low abundance. Compared to the control group, the trial group exhibited: Significantly increased relative abundance of Succiniclasticum (p < 0.05). Non-significant increasing trends (p > 0.05) in Lachnospiraceae_NK3A20_group, Lachnospiraceae_XPB1014_group, and Rikenellaceae_RC9_gut_group. Nonsignificant decreasing trends (p > 0.05) in Methanobrevibacter, Prevotella, Veillonellaceae_UCG-001, Selenomonas, and Solobacterium.

At the phylum level, bacterial phyla detected in the rumen content of sheep included Firmicutes, Bacteroidota, Euryarchaeota, Actinobacteriota, Synergistota, Spirochaetota, Proteobacteria, Patescibacteria, Desulfobacterota, and Cyanobacteria, among others. Compared to the control group, the trial group exhibited a significantly lower relative abundance of Patescibacteria (p < 0.05). Changes in the relative abundances of the other phyla were not statistically significant.

At the genus level, detectable genera in the rumen microbiota included Succiniclasticum, Methanobrevibacter, Prevotella, and others. Compared to the control group, the trial group showed a significantly increased relative abundance of Succiniclasticum (p < 0.05). Non-significant trends of increase or decrease were observed for other genera such as those within Lachnospiraceae.

The LEfSe analysis of bacterial communities in sheep rumen contents revealed significant dominant taxa in the trial group (Figure 3). The LDA scores indicated that the trial group exhibited notable enrichment in the following taxa at specific classification levels: Bacteroidales_BS11_gut_group, Desulfovibrionales, Desulfovibrionaceae, Desulfovibrionia, Clostridium_sensu_stricto_13, and Desulfovibrio. Additionally, the trial group showed dominance in bacterium_ND2018, Saccharimonadia, Candidatus_Saccharimonas, Saccharimonadales, Saccharimonadaceae, Patescibacteria, Clostridia_UCG_014 (class), Clostridium__methylpentosum_group, RC-11rumen_bacterium_RC_11, and Ruminococcus_sp.

Tax4Fun functional prediction indicated that metabolic pathways represented the most abundant functional category, followed by other categories including genetic information processing and environmental information processing. However, no significant differences were observed between the experimental and control groups in these functional categories (p > 0.05).

The quality control (QC) chart for the metabolite sample quality analysis is presented in Figure 4. The correlations of QC samples exceeded 0.99 in both positive and negative ion modes. These results indicate high stability of the metabolite detection process and superior data quality, fulfilling the prerequisites for subsequent research data analysis.

The PCA results (Figure 4.) of the total samples revealed that in the positive ion mode, the contribution rates of PC1, PC2, and PC3 were 23.45, 21.04, and 11.49%, respectively. In the negative ion mode, the contribution rates of PC1, PC2, and PC3 were 29.23, 18.46, and 12.21%, respectively.

As shown in Figure 4, the PLS-DA model clearly distinguished the metabolomic profiles between the two groups, with high model validity (R²Y and Q² values; intercept of the Q² regression line < 0), indicating that the model is reliable and not overfitted.

The classification of rumen metabolites in positive (A) and negative (B) ion modes is shown in Figure 5. In positive ion mode, 10 categories were identified, dominated by lipids and lipid-like molecules (34.59%), organic acids and derivatives (23.53%), and heterocyclic compounds (16.71%). In negative ion mode, 8 categories were identified, primarily lipids and lipid-like molecules (38.87%), organic acids and derivatives (18.94%), and heterocyclic compounds (10.63%).

Figure 5 shows the volcano plots of differential metabolites in positive (C) and negative (D) ion modes. Differential metabolites were screened based on the criteria of VIP > 1.0, FC > 1.2 or FC < 1/1.2, and p-value < 0.05. A total of 673 metabolites were identified in positive ion mode, among which 82 showed significant differences (58 up-regulated and 23 down-regulated). In negative ion mode, 342 metabolites were identified, with 45 being significantly different (12 up-regulated and 32 down-regulated).

The top 20 upregulated and downregulated differential metabolites were ranked by FC. As shown in Table 3, under positive ion mode: Upregulated metabolites were predominantly saponins, amino acids and derivatives, esters, sugars, heterocyclic compounds, phenolic compounds, and neurotransmitters. Downregulated metabolites included amino acids and derivatives, heterocyclic compounds, hormones and their metabolites, antibiotics, aromatic amines, fatty acids and derivatives, and non-steroidal anti-inflammatory drugs (NSAIDs).

From Table 4, under negative ion mode, the upregulated metabolites were predominantly coumarins and derivatives, fatty acids and derivatives, heterocyclic compounds, esters, amino acids and derivatives, and nucleotides and their metabolites. Conversely, the downregulated metabolites primarily included sugars and derivatives, amino acids and derivatives, phenolic compounds, organic acids, and nucleotide derivatives.

The cluster analysis of differential metabolites is presented in Figure 6. Color intensity correlates with metabolite abundance, where darker hues indicate higher expression levels and lighter hues represent lower expression levels. The results demonstrated that dietary supplementation with AOC significantly altered rumen microbial metabolites in sheep.

The correlation analysis plots (A/C: positive ion mode; B/D: negative ion mode) and chord diagrams of differential metabolites are shown in Figure 6. In the positive ion mode, metabolites with significant correlations included methyl 3-indolyacetate, methyl indole-3-acetate, loxoprofen, pleuromutilin, phosphoethanolamine, D-(−)-glutamine, and 4-ethoxy-N-[(4-methoxyphenyl)methylene]aniline. In the negative ion mode, significantly correlated metabolites were primarily N-acetyl-L-methionine, ochratoxin A, thymidine, pseudouridine, all-cis-4,7,10,13,16-docosapentaenoic acid, tréhalose, and methadone-d9.

Based on the metabolite classification information provided by the KEGG database, statistical visualization of annotated differential metabolites was performed Figure 6. In positive ion mode, the differential metabolites were widely involved in multiple biological layers, including Organismal Systems (e.g., endocrine and digestive systems), Metabolism (e.g., lipid and amino acid metabolism), Genetic Information Processing, and Environmental Information Processing. Among these, Global and Overview Maps accounted for the highest proportion (27.78%), followed by Amino acid metabolism (13.89%) and Lipid metabolism (11.11%), indicating that these broad metabolic processes were significantly influenced. In negative ion mode, the differential metabolites were more concentrated in the domain of Metabolism, particularly in Amino acid metabolism (23.08%), Nucleotide metabolism (10.26%), and Carbohydrate metabolism (10.26%).

KEGG pathway enrichment analysis was performed to identify the primary biological functions of differential metabolites. Based on the enrichment results, bubble plots of the enriched KEGG pathways were generated Figure 6. In the positive ion mode, differential metabolites were predominantly enriched in metabolic pathways, steroid hormone biosynthesis, phenylalanine metabolism, ovarian steroidogenesis, and aminoacyl-tRNA biosynthesis pathways among the top 20 KEGG pathways. Under negative ion mode, significant enrichment was observed in phenylalanine metabolism, pyrimidine metabolism, citrate cycle (tricarboxylic acid cycle, TCA cycle), butanoate metabolism, arginine biosynthesis, and alanine, aspartate, and glutamate metabolism pathways. These findings highlight critical roles in steroidogenesis, amino acid and nucleotide metabolism, and energy homeostasis, aligning with the observed metabolic shifts in the trial group.

Pearson correlation analysis was performed between differential bacterial genera at the genus level and differential metabolites (Figure 7). In positive ion mode (Figure A), Candidatus Saccharimona showed a significant positive correlation with anti-inflammatory agents such as loxoprofen. Clostridium_sensu_stricto_13 was positively correlated with the neurotransmitter acetylcholine, but negatively correlated with metabolites including methyl indole-3-acetate. In negative ion mode, Clostridium_sensu_stricto_13 demonstrated positive correlations with multiple fatty acids (e.g., eicosatetraenoic acid and docosapentaenoic acid), and a negative correlation with 4-methylphenol.