Linoleic acid enhances ruminal carbohydrate metabolism and biohydrogenation to promote rumen epithelial development in calves

Abstract

Unlike mature ruminant, pre-weaned calves have limited ruminal biohydrogenation during early microbial establishment. Since linoleic acid (LA) is known to promote calf growth, elucidating whether dietary LA supplementation during this plastic phase can actively shape rumen biohydrogenation function, restructure the microbiota, and drive epithelial development represents a pivotal unanswered question. This study determined the optimal dietary supplementation dose of LA and elucidated its regulatory effects on the rumen microbiota, metabolites and gene expression in preweaning calves. LA supplementation reshaped the rumen microbial community of calves and enhanced microbial carbohydrate metabolism by increasing the abundance of carbohydrate-active enzymes, thereby promoting acetate synthesis. Hydrogen generated during carbohydrate fermentation further facilitated LA biohydrogenation. The metabolites generated from this process activate the microbial phosphatidylcholine synthesis pathway in the rumen, increasing PC concentrations in both rumen fluid and tissue. CYP3A24 was identified as a core rumen gene involved in the PC-CYP3A24 regulatory axis. PC promoted bovine rumen epithelial cells (BRECs) proliferation, barrier function, antioxidant capacity and inflammation suppression, while CYP3A24 silencing diminishes these effects. In conclusion, optimal LA supplementation regulates rumen metabolism, enhances PC synthesis via functional microbiota, and activates the PC-CYP3A24 axis, thereby improving rumen development and calf growth performance.

Introduction

Unsaturated fatty acids (UFAs) are indispensable components of ruminant nutrition and serve as structural building blocks for cell membranes, precursors of bioactive lipid mediators, and energy sources for growth and reproduction^1,2. Unlike monogastric animals, ruminants possess a unique rumen microbial ecosystem that profoundly alters dietary lipids before their absorption in the lower gastrointestinal tract. For young ruminants, including calves, the demand for UFAs is elevated because of the rapid proliferation of rumen epithelial cells, skeletal muscle development, and immune system maturation during the pre- and postweaning periods³. The rumen of neonatal calves is initially a sterile organ, and its colonization by microorganisms is a sequential and dynamic process that occurs primarily during the first few months of life⁴. This period is widely recognized as the “critical window” for rumen microbial establishment, where the composition and structure of the microbial community are highly plastic and sensitive to dietary interventions, host age, and environmental factors. The establishment of a stable and functional rumen microbiota is a prerequisite for the transition from milk-based to solid-feed-based nutrition, as it enables the fermentation of dietary fiber, the production of volatile fatty acids (VFAs), and the maturation of the rumen epithelium^5,6. For young ruminants, the successful development of the rumen microbial ecosystem directly affects their future rumen function, feed efficiency, and overall growth performance. A key feature of the rumen ecosystem, even in developing young ruminants, is its high hydrogenating capacity, which is primarily mediated by anaerobic microorganisms such as Butyrivibrio fibrisolvens and Megasphaera elsdenii^7,8. These microbes catalyze the biohydrogenation of dietary UFAs into SFAs through a series of enzymatic reactions, a process that is thought to protect rumen microbes from the potential toxic effects of UFAs and generate hydrogen as an electron sink for fermentation^9,10. However, this extensive biohydrogenation in the developing rumen of calves results in the loss of most dietary UFAs, including LA, thereby reducing their availability for host utilization.

Previous studies have demonstrated that dietary lipid supplementation can alter the rumen microbial composition in mature ruminants^11,12, but limited research has focused on the effects of LA on the rumen microbial ecosystem during the critical establishment period in calves. Given the essential role of LA in animal development and the unique microbial dynamics of the young calf rumen, it is hypothesized that LA supplementation may modulate the rumen microbial flora and their metabolic activities, potentially by selectively inhibiting or promoting specific microbial taxa involved in biohydrogenation, thereby influencing rumen development and growth performance.

Therefore, in this study, the effects of LA supplementation on the rumen microbial community structure and metabolic profile of calves were systematically investigated using high-throughput sequencing. We characterized the changes in the ruminal bacterial, archaeal, and fungal communities following LA addition and identified the key ruminal metabolites and metabolic pathways affected by LA. Furthermore, we established causal relationships among LA-induced alterations in the rumen microbiota, rumen metabolism, rumen development, and growth performance of calves.

Results

LA promoted rumen acetate metabolism and papilla development in calves

To investigate the effects of LA supplementation on the growth performance and rumen development of pre-weaning calves, we analyzed key growth indices, rumen fermentation parameters and rumen morphological traits of calves in different LA supplementation groups at 90 d of age (Fig. 1A). When calves were fed with LA until 90 d of age, the ADG and BW of calves in the LA II group were significantly higher than those in other treatment groups (P < 0.05; Fig. 1B, C), indicating that 0.21 × BW⁰^·⁷⁵ is the optimal supplementation dose of LA. Notably, the molar proportion of ruminal acetate in the LA group was significantly greater than that in the CON group (P < 0.05; Fig. 1D). Histological analysis further revealed that the height of the ruminal papilla of calves in the LA group was significantly greater than that in the CON group (P < 0.05; Fig. 1E, F). The above results suggested that LA supplementation promoted rumen morphological development by regulating the function of the rumen microbial community and altering the metabolism pattern of VFAs.

Fig. 1. LA improves the growth performance of calves and promotes rumen acetate metabolism and papilla development.

A Schematic diagram of the experimental design. B Average daily gain of calves in each group (n = 15). C BW of calves in each group (n = 15). D Rumen fermentation parameters (n = 10); E H&E staining of rumen tissue; F rumen papilla length. “*” indicates P < 0.05, “**” indicates P < 0.01; scale = 700 μm.

Characteristics of altered metabolic profiles in calf rumen fluid and rumen tissue under LA treatment

To further elucidate the metabolic regulatory mechanisms underlying LA-mediated rumen microbial and host phenotypic changes, we performed a nontargeted metabolomic analysis of rumen fluid and tissue to characterize LA-driven metabolic profile alterations and identify key differentially abundant metabolites. Principal component analysis (PCA) revealed that the metabolomes of the rumen fluid and rumen tissue in calves from the CON and LA groups were significantly separated (Fig. 2A, C). Using a variable importance in projection (VIP) > 1 and a |Fold change (FC)| > 2 as screening criteria, t test analysis revealed 600 and 585 significantly differentially abundant metabolites in the rumen fluid and rumen tissue, respectively (P < 0.05; Fig. 2B, D). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed that the differentially abundant metabolites in the rumen fluid were enriched mainly in pathways such as LA metabolism and UFA biosynthesis (Fig. 2E), whereas those in the rumen tissue were also significantly enriched in the LA metabolism pathway (Fig. 2F). Notably, three common differentially abundant metabolites were identified in the LA metabolic pathway between the two sample types: 1,2-dihexadecanoyl-sn-glycero-3-phosphocholine, PC (16:0/16:0), and LA (Fig. 2G). The abundances of 1,2-dihexadecanoyl-sn-glycero-3-phosphocholine and PC (16:0/16:0) in both the rumen fluid and the rumen tissue of the LA group were significantly greater than those in the CON group (P < 0.01; Fig. 2H, I). The change in LA abundance was tissue specific: the abundance of LA in the rumen fluid of the LA group was significantly lower than that in the CON group but significantly greater than that in the rumen tissue (P < 0.01; Fig. 2J). These findings indicate that LA hydrogenation in the rumen may occur.

Fig. 2. Variation characteristics of the metabolic profiles of calves' rumen fluid and rumen tissue under LA.

A Multivariate statistical PCA plot of the metabolome of the calves' rumen fluid. B Multivariate statistical PCA plot of the metabolome of calf rumen tissue. C Differential scatter plot of the rumen fluid metabolome. D Differential scatter plot of the rumen tissue metabolome. E KEGG enrichment bubble plot of differentially abundant metabolites in the rumen fluid. F KEGG enrichment bubble plot of differentially abundant metabolites in the rumen tissue. G Venn diagram of differentially abundant metabolites in the LA metabolic pathway of rumen fluid and rumen tissue. H Expression abundance of 1,2-dihexadecanoyl-sn-glycero-3-phosphocholine in the rumen fluid and rumen tissue. I Expression abundance of PC (16:0/16:0) in the rumen fluid and rumen tissue. J Expression abundance of LA in the rumen fluid and rumen tissue. (n = 10).

LA-mediated differential abundance of rumen bacteria and identification of core taxa

To characterize the effects of LA supplementation on the rumen microbial community structure and composition in preweaning calves, we performed high-depth metagenomic sequencing of the rumen fluid. A total of 1,405,903,516 raw sequences were obtained from rumen metagenomic sequencing, and the average sequencing depth of the samples was 70,296,175.8 ± 9,684,240.945 (mean ± standard error). After quality control and host gene filtering, 1,395,849,058 high-quality sequences were retained, with an average depth of 69,792,452.9 ± 9,632,956.57 (Table S1). Analysis of the microbial composition revealed that bacteria (91.66%), viruses (6.57%), archaea (1.72%), and eukaryotes (0.01%) were the main components (Fig. S1A and Table S2). At the domain level, the Wilcoxon rank-sum test revealed that the relative abundance of rumen bacteria in the LA group of calves (95.16%) was significantly greater than that in the CON group (P < 0.05; Table S2). Therefore, subsequent analyses focused on the bacterial community. The bacterial community was dominated by Bacteroidota (42.50%), Bacillota (41.35%), and Actinobacteriota (8.31%) (Fig. S1B and Table S3); the dominant taxa at the genus level included Prevotella (50.69%), Succinivibrio (7.30%), Xylanibacter (5.47%), Sharpea (5.63%), and Ruminococcus (4.62%) (Fig. S1C and Table S3); and the dominant bacterial species were Prevotella sp. (29.64%), Lachnospiraceae bacterium (27.35%), Bacteroidales bacterium (15.45%), and Bacteroidaceae bacterium (5.17%) (Fig. S1D and Table S3). Analysis of α - diversity showed that there were no differences in the Chao1 and ACE indices between the groups (P > 0.05; Table S4), but the Shannon index in the LA group was significantly lower (P < 0.05; Fig S1E and Table S4), suggesting that the community evenness differed between the two groups. PCA based on the Bray-Curtis distance further confirmed a significant separation between the LA and CON groups at the phylum, genus, and species levels (PERMANOVA, P < 0.05; Fig S1F–H).

Differential taxa of bacteria at the phylum, genus, and species levels were identified via linear discriminant analysis effect size (LEfSe) analysis, with LDA > 2 and P < 0.05 defined as the criteria for differentially abundant taxa. At the phylum level, Bacteroidota and Campylobacterota were significantly enriched in the LA group (LDA > 2, P < 0.01), whereas the abundances of Bacillota, Pseudomonadota, and Actinomycetota increased in the CON group (LDA > 2, P < 0.05; Fig. 3A). Analysis at the genus level revealed that 17 genera were significantly enriched in the LA group (LDA > 2, P < 0.05; Fig. 3B). At the species level, the abundances of 68 bacterial species increased significantly in the LA group (LDA > 2, P < 0.05; Fig. 3C). Furthermore, Spearman correlation analysis revealed that 19 bacterial species enriched in the LA group (such as Ruminococcus albus, Prevotellaceae bacterium, Schwartzia succinivorans, and Ruminococcus difficilis) were significantly positively correlated with host phenotypes (molar ratio of acetate, BW, and ADG) (R > 0.5, P < 0.05; Fig. 3D). Visualization with a Sankey diagram indicated that these 19 host phenotype-associated bacterial species belonged to 5 phyla and 7 genera (Fig. 3E). Heatmap analysis confirmed their specific enrichment in the LA group (P < 0.05); thus, they were defined as “host phenotype-associated microbiota (HPAM)”. Spearman correlation and co-occurrence network analyses confirmed that the core taxa within the HPAM were predominantly positively correlated with each other, suggesting that these functional bacteria act synergistically and cooperatively to promote rumen fermentation and rumen epithelial development (Fig. S2A, B).

Fig. 3. LA drives the differentiation of the rumen microbiota in calves and identifies the core microbiota associated with host phenotypes.

A Histogram of the LDA value distribution at the phylum level; B histogram of the LDA value distribution at the genus level. C Histogram of the distribution of LDA values at the species level. D Spearman correlation between host phenotype and rumen microorganisms. E Sankey diagram. (n = 10).

LA promotes rumen acetate production by increasing microbial carbohydrate metabolism

To elucidate the effects of LA supplementation on the functional potential of the rumen microbial community in preweaning calves, we conducted functional annotation analyses of high-quality metagenomic sequences against the KEGG and CAZymes databases. At KEGG Level-2, within the “Metabolism” pathway, subcategories including “Carbohydrate metabolism” were significantly enriched in the rumen microbiome of LA calves (LDA > 2, P < 0.05; Fig. 4A). Further exploration at the KEGG level 3 revealed significantly differential metabolic pathways. Specifically, pathways associated with “carbohydrate metabolism” (fructose and mannose metabolism, glycolysis/gluconeogenesis, starch and sucrose metabolism, pyruvate metabolism, butanoate metabolism, propanoate metabolism, glyoxylate and dicarboxylate metabolism, and the citrate cycle) were significantly enriched in the LA group (LDA > 2, P < 0.05; Figs. 4B and S3A). Analysis of key metabolic modules further revealed that the relative abundances of gluconeogenesis (oxaloacetate => fructose-6P, M00003), the citrate cycle (second carbon oxidation, 2-oxoglutarate => oxaloacetate, M00011), and glycogen degradation (glycogen => glucose-6P, M00855) significantly increased in the LA group (P < 0.05; Fig. S3B). To obtain a comprehensive understanding of carbohydrate metabolism, key pathways and modules were integrated for further analysis, which revealed that all these KEGG pathways were involved in the biosynthesis of acetate (Fig. 4C). Metabolomic analysis revealed that the relative abundances of key intermediate metabolites involved in acetate synthesis, α-D-glucose-1P and acetyl-CoA, were significantly greater in the LA group than in the CON group, whereas the relative abundance of glucose-6P was significantly lower (P < 0.05; Fig. 4C). We further analyzed the abundance of six major classes of carbohydrate-active enzymes (CAZy) in the calf rumen. Compared with the CON group, the LA group exhibited significantly increased relative abundances of all six CAZy classes, including glycoside hydrolases (GH), glycosyltransferases (GT), carbohydrate-binding modules (CBM), carbohydrate esterases (CE), auxiliary activities (AA), and polysaccharide lyases (PL) (P < 0.05; Fig. S3C). Subsequent analysis of key CAZy families revealed that the relative abundances of specific families, such as GT2, GT4, CBM50, GH13, GH23, GH3, GH43, GH1, GH5, and GH35, were significantly greater in the LA group than in the CON group (P < 0.05; Fig. S3D). Multinetwork analysis revealed that the 19 HPAM strains were significantly positively correlated with 7 carbohydrate metabolic pathways (e.g., glycolysis/gluconeogenesis and the TCA cycle) (R > 0.50, FDR-corrected P < 0.05), and these 7 pathways were, in turn, significantly positively correlated with host ADG, BW, and acetate concentration (R > 0.50, P < 0.05; Fig. S3E). These results indicated that LA supplementation could reshape the rumen microbiota of calves, increase carbohydrate metabolism, and thereby promote acetate production. These findings were further verified by an in vitro fermentation trial in which Ruminococcus albus was added (Fig. S4). Moreover, the fermentation of carbohydrates to acetate in the rumen is accompanied by the release of large amounts of hydrogen, which provides the essential reductive conditions required for the classical biohydrogenation of LA.

Fig. 4. LA promotes rumen acetate production by increasing microbial carbohydrate metabolism.

A Heatmap and LDA bar chart of the KEGG pathway level. B Heatmap of carbohydrate metabolism. C Metabolic pathways involved in the synthesis of acetate. D LA hydrogenation pathway. E PC biosynthesis pathway.

LA promotes rumen biohydrogenation and PC synthesis via the core functional microbiota

Integrated rumen metabolomic and metagenomic analyses were performed to investigate the biohydrogenation of LA in the calves' rumen. Compared with that in the CON group, the relative abundance of LA (c9, c12-18:2) significantly decreased in the LA group (P < 0.01; Fig. 4D). Notably, the relative abundance of trans-vaccenic acid (TVA, t11-18:1), an intermediate in LA biohydrogenation, significantly decreased, whereas the relative abundance of stearic acid (SA, 18:0), the final product of LA biohydrogenation, significantly increased (P < 0.01; Fig. 4D). These results indicate that the classical biohydrogenation pathway of LA was efficiently activated. LA supplementation significantly enriched the key functional bacteria involved in LA biohydrogenation, namely, Butyrivibrio fibrisolvens (which catalyzes the isomerization of LA to CLA and TVA) and Butyrivibrio proteoclasticus (which catalyzes the reduction of TVA to SA), whose relative abundances were significantly greater than those in the control group (P < 0.01; Fig. 4D). These findings confirm that LA supplementation drives the production of biohydrogenation products by modulating the abundance of the core functional microbiota. In the LA group, the relative abundances of 1-oleoyl-L-α-L, PA and the final product, PC, were significantly greater than those in the control group (P < 0.01; Fig. 4E), indicating that the supply of LA biohydrogenation products effectively promoted the overall activation of the PC biosynthesis pathway. LA supplementation also significantly enriched the core functional bacteria involved in PC synthesis, including Prevotella sp. (which participates in the conversion of DAG to PC) and Ruminococcus albus (which participates in PA synthesis), whose relative abundances significantly increased (P < 0.01; Fig. 4E). This provides a sufficient microbial source of precursors and an enzymatic basis for PC biosynthesis.

Transcriptional expression levels in the rumen tissue of LA-driven calves

To further elucidate the molecular mechanisms underlying LA-induced metabolic alterations, microbial changes, and host phenotypes, we analyzed the transcriptomes of the rumen tissue to identify differentially expressed genes (DEGs) and their regulatory networks. PCA revealed that the rumen tissue metabolomes of calves in the LA group and the CON group were significantly separated (Fig. 5A). On the basis of the criteria of FC > 10 and P < 0.05, a total of 822 DEGs were screened, among which 753 genes were significantly upregulated and 69 genes were significantly downregulated in the LA group (Fig. 5B). The association between DEGs and host phenotypes was evaluated by Spearman’s rank correlation analysis. A total of 555 genes were significantly correlated with host phenotypes (R > 0.05, P < 0.05; Table S5), which were defined as host phenotype-related DEGs (Table S5). Furthermore, Weighted gene coexpression network analysis (WGCNA) revealed that the turquoise module had the strongest correlation with host phenotypes (R > 0.05, P < 0.05; Fig. 5C). Among the 404 DEGs in this module, 390 genes were common to the 555 DEGs related to host phenotypes (BW, ADG and acetate) mentioned above (Fig. 5D). Protein-protein interaction (PPI) network analysis was performed on the 390 common genes, and the cytoNCA algorithm was used to screen hub genes. The results revealed that 9 genes, such as EVPL, had high betweenness centrality (BC) values (Fig. 7E and Table S6). These nine core genes were significantly upregulated in the rumen tissues of the LA group (P < 0.01; Fig. 5F, G), and GO functional enrichment analysis revealed that the DEGs were enriched mainly in cellular component terms such as cell periphery and plasma membrane (logFC > 10; P < 0.05; Fig. 5F). STRING analysis revealed PPI relationships among the nine core genes; multiple edges of different colors were observed between CYP3A24 and CYP2C18, indicating various types of interactions between these two proteins (e.g., coexpression and gene co-occurrence; Fig. 5H). Spearman’s rank correlation analysis further revealed that the nine genes (including EVPL) were significantly positively correlated with HPAM (R > 0.05, P < 0.05; Fig. 5I), and these genes were also significantly positively correlated with the metabolites 1,2-dihexadecanoyl-sn-glycero-3-phosphocholine, PC (16:0/16:0), and LA (R > 0.05, P < 0.05; Fig. 5J). In conclusion, CYP3A24, a core hub gene in the LA metabolic pathway, serves as the key downstream effector gene mediating the regulatory effects of PC. Accordingly, it was selected as the core target gene for subsequent mechanistic validation in this study.

Fig. 5. Effects of LA on the transcriptional level of calf rumen tissue.

A PCA plot of the rumen tissue samples. B Scatter plot of differentially expressed genes. C Heatmap of module–trait correlations. D Venn diagram of genes in the turquoise module and DEGs. E Hub genes were screened using the algorithm in Cytoscape (the larger the circle is in the counterclockwise direction, the higher the BC value; the yellow circles show the top 9 genes with the highest BC values). F Expression patterns of the hub genes and the top 10 GO analysis terms of the DEGs. G Core gene expression level. H Protein-protein interaction network. I Correlations between core genes and host-associated microorganisms. J Correlations between core genes and key metabolites. (n = 10).

Fig. 7. The PC-CYP3A24 axis enhances the barrier function and biological performance of BRECs.

A Experimental design. B Interference fragment screening. C mRNA expression of the CYP3A24 gene after PC treatment. D mRNA expression of tight junction proteins and mucins. E mRNA expression is related to intestinal cell renewal, differentiation, and proliferation. F mRNA expression is related to antioxidants. G mRNA expression is involved in the immune-inflammatory response. H EdU staining. I ROS detection.

Key targets of the PC-CYP3A24 axis in regulating calf rumen epithelial development

By jointly analyzing the transcriptome and metabolome of the rumen tissue, we identified the LA metabolic pathway as the coregulated pathway of the two omics analyses. We identified 13 DEGs (including PLA2G12A, PLA2G5, PLA2G3, CYP2C18, CYP2C19, CYP2J2, CYP3A24, PLB1, PLA2G4F, PLA2G4D, PLA2G4E, and PLAAT3) and 5 differentially abundant metabolites enriched in this pathway. Only PLA2G12A and 13-OxoODE were significantly downregulated in the rumen tissue (P < 0.01; Fig. 6A). The expression patterns of the DEGs in this pathway also significantly differed between the two sample groups (Fig. 6B). STRING analysis indicated that CYP3A24 interacts with other proteins (Fig. 6C). The heatmap shows that the expression patterns of the five differentially abundant metabolites significantly differed between the two groups (Fig. 6D). Spearman’s rank correlation analysis revealed that PLA2G12A was significantly negatively correlated with HPAM, whereas the remaining genes were significantly positively correlated with HPAM (Fig. S5A). Moreover, the differentially abundant metabolite 13-OxoODE was significantly negatively correlated with HPAM, whereas the other differentially abundant metabolites were significantly positively correlated with HPAM (Fig. S5B). Spearman’s rank correlation analysis further revealed a significant association between DEGs and differentially abundant metabolites in the pathway: 13-OxoODE was significantly positively correlated with PLA2G12A (R > 0.5, P < 0.05) but significantly negatively correlated with the remaining DEGs; phospholipid molecules (including SOPC, PC(16:0/16:0), and LA) exhibited the opposite pattern-they were significantly negatively correlated with PLA2G12A and significantly positively correlated with the other DEGs (Fig. S5C). On the basis of the metabolite-gene interaction rules revealed by the above Spearman’s rank correlation analysis, key regulatory nodes in the LA metabolic pathway were screened out, including the differentially abundant metabolites of PC (including PC (16:0/16:0) and 1,2-diarachidonoyl-sn-glycero-3-phosphocholine) and their downstream gene CYP3A24. The correlations among Ruminococcus albus, PC, and CYP3A24 were further analyzed. Changes in the abundance of the microorganism Ruminococcus albus were significantly positively correlated with both the gene expression of CYP3A24 and the abundance of the PC (16:0/16:0) (Fig. 6E–G).

Fig. 6. Key targets of the PC-CYP3A24 axis in the regulation of calf rumen epithelial development.

A Map of changes in genes and metabolites in the LA metabolic pathway (the origin represents differentially abundant metabolites, the squares represent differential genes, red indicates upregulation, and green indicates downregulation). B Expression patterns of DEGs in the LA metabolic pathway. C Protein interaction network of DEGs in the LA metabolism pathway. D Expression patterns of differentially abundant metabolites in the LA metabolism pathway. E Correlation between PC and CYP3A24 gene. F Correlation between PC and Ruminococcus albus. G Correlation beween CYP3A24 and Ruminococcus albus.

The PC-CYP3A24 axis enhances the barrier function and biological performance of BRECs

To validate the regulatory role of the PC-CYP3A24 axis in rumen epithelial barrier function and cellular homeostasis, we conducted in vitro functional assays on primary BRECs with PC treatment and CYP3A24 silencing (Fig. 7A). Primary BRECs were isolated via trypsin digestion. On day 7 postisolation, limited cell attachment was observed, followed by extensive attachment on day 9. Purification via differential centrifugation yielded F2-generation adherent cells, which were subsequently passaged to F4 for phenotypic characterization (Fig. S6A). Immunofluorescence analysis confirmed that more than 95% of the cells exhibited positive staining for CK-18, while no significant fluorescence was detected for IgG (Fig. S6B), verifying the successful isolation of BRECs. BRECs were treated with 0, 25, 50, 75, or 100 μM PC for 3, 6, 12, or 24 h, and cell viability was assessed via the CCK-8 assay at each time point. The results demonstrated that the optimal concentration and treatment duration for PC were 25 μM and 12 h, respectively, which yielded maximal efficacy (Fig. S6C). Three siRNA fragments targeting CYP3A24 were screened, with siCYP3A24-84 demonstrating the greatest interference efficacy (P < 0.01; Fig. 7B). Following the LA treatment of BRECs, significant upregulation of CYP3A24 gene expression was observed (P < 0.01; Fig. 7C).

PC exposure upregulated the mRNA expression of Occludin, hypoxia-inducible factor alpha (HIF-1α), zonula occludens-1 (ZO-1), and tight junction protein 1 (TJP1) in BRECs, whereas CYP3A24 silencing alleviated these effects (P < 0.05; Fig. 7E). Furthermore, PC treatment increased the expression levels of proliferating cell nuclear antigen (PCNA), catenin beta 1 (CTNNB1), monoclonal antibody Ki-67 (Ki-67), and transcription factor 4 (TCF4, P < 0.05; Fig. 7F) and increased proliferation, as indicated by an increased EdU⁺ cell ratio (Fig. 7H). In contrast, CYP3A24 silencing reversed the PC-induced upregulation of PCNA, CTNNB1, Ki-67, and TCF4 expression (Fig. 7H). PC increased the mRNA expression of antioxidant genes in BRECs (P < 0.05; Fig. 7F) and reduced intracellular reactive oxygen species (ROS) levels (Fig. 7I). However, compared with PC supplementation, CYP3A24 silencing downregulated the expression of these antioxidant genes (P < 0.05; Fig. 7F) and further decreased ROS levels in BRECs (Fig. 7I). Moreover, compared with PC supplementation, PC treatment downregulated the expression of proinflammatory cytokines (P < 0.05), whereas compared with PC supplementation, CYP3A24 silencing further promoted proinflammatory cytokine expression (P < 0.05; Fig. 7G). In summary, PC promoted tight junction formation, cell proliferation, and antioxidant capacity but suppressed inflammation in BRECs. However, silencing CYP3A24 reversed these effects, indicating that CYP3A24 is a key mediator underlying the functional effects of PC.

Discussion

LA, as an essential polyunsaturated fatty acid for ruminants, has been shown to participate in rumen microbial metabolism and host growth regulation. Previous studies have shown that the average daily intake of LA between 3 and 5 g/d during the first 30 d of life promoted the growth of preweaned Holstein calves, possibly by improving their immune status¹³. In this study, 60 Holstein bull calves under the same management and feeding conditions were selected to clarify the relationship between different dosages of LA supplementation and the host growth rate and to determine the optimal dosage of LA. By integrating multiomics technologies (metagenomics, metabolomics, and transcriptomics) and combining in vitro fermentation and cell culture experiments, we systematically revealed the regulation of the rumen microbiota in calves by LA hydrogenation and revealed that the PC-CYP3A24 axis promotes the development of rumen epithelial cells in calves.

The establishment of a functional rumen microbial community is the core basis for rumen fermentation and nutrient utilization in preweaning calves, and dietary interventions are the most important factors regulating the composition and structure of the rumen microbiota^14,15. Consistent with previous studies showing that dietary lipid supplementation can alter the rumen microbial community of ruminants¹⁶, our study revealed that LA supplementation significantly changed the rumen microbial composition of preweaning calves, with the relative abundance of rumen bacteria in the LA group being significantly greater than that in the control group, which indicated that LA supplementation could promote the proliferation of functional rumen bacteria and increase the microbial biomass of the rumen ecosystem. At the phylum level, Bacteroidota and Campylobacterota were significantly enriched in the LA group, while the abundances of Bacillota and Actinomycetota were relatively reduced; at the genus and species levels, a total of 17 genera and 68 species were significantly enriched in the LA group, among which 19 species were identified as HPAM because of their significant positive correlation with the calf ADG, BW and rumen acetate molar proportion. Notably, the HPAM identified in this study, including Ruminococcus albus, Prevotellaceae bacterium, Schwartzia succinivorans and Ruminococcus difficilis, are all core functional microbes in the rumen with multiple metabolic capabilities. Correlation and co-occurrence network analyses revealed that the 19 HPAM species identified in this study were dominated by significant positive correlations, indicating that these functional bacteria synergistically enhance carbohydrate metabolism and LA biohydrogenation, forming a core microbial module that cooperatively drives calf growth. These results further demonstrate that the significant reduction in the Shannon index following linoleic acid supplementation was caused by diet-induced microbial functional specialization and environmental filtering, rather than microbial growth inhibition. LA selectively enriched a core microbiota with high-efficiency carbohydrate degradation and biohydrogenation capabilities. Although this reduced community diversity, it markedly improved the metabolic functional efficiency of the rumen microbiota. Ruminococcus albus, as a classic cellulose-degrading and acetate-producing bacterium, is not only capable of degrading dietary fiber to produce fermentable sugars and VFAs^17,18 but also participates in the synthesis of microbe-derived PC in this study, which is a key microbe connecting carbohydrate metabolism and lipid metabolism in the rumen. Prevotellaceae bacterium, as the dominant genus in the rumen of calves, plays important roles in polysaccharide degradation and glycolysis¹⁹, and its enrichment further increased the carbohydrate metabolic capacity of the rumen microbiota in the LA group. The screening of HPAM in this study differs from that in previous studies that focused only on the taxonomic identification of rumen microbes but linked changes in microbial abundance to host phenotypic traits, which clearly revealed the functional relevance of the identified core microbes to calf growth and provided key microbial resources for the subsequent targeted regulation of the rumen microbiota to improve calf growth performance.

Rumen microbial carbohydrate metabolism is the main source of VFAs for calves, and acetate, as the most abundant VFA in the rumen, is not only the main energy substance for calves but also a key signaling molecule promoting rumen papilla development^20,21. Previous studies have shown that dietary additives can regulate rumen carbohydrate metabolism by altering microbial enzyme activity, but the regulatory effect of LA on rumen microbial carbohydrate metabolic pathways and the underlying molecular mechanisms have not been reported. Our study revealed that LA supplementation significantly enriched carbohydrate metabolism-related pathways in the rumen microbiome, including glycolysis/gluconeogenesis, starch and sucrose metabolism, pyruvate metabolism and the citrate cycle, and multinetwork analysis confirmed that these pathways were significantly positively correlated with HPAM abundance and host phenotypic traits, indicating that LA-induced microbial functional reprogramming is the core driving force for the increase in rumen acetate production. The functional annotation of CAZy further revealed the molecular basis of the enhanced carbohydrate metabolic capacity of the rumen microbiota in the LA group: the relative abundances of all six major CAZymes classes (GH, GT, CBM, CE, AA, and PL) and key families (GT2, GT4, CBM50, GH13, and GH3) significantly increased in the LA group. GH families (e.g., GH13 and GH3) are the core GHs for degrading starch, cellulose and hemicellulose, and their enrichment directly improved the efficiency of the rumen microbiota in the utilization of dietary carbohydrates^22,23; CBM families (e.g., CBM50) can increase the binding ability of enzymes to carbohydrate substrates, further improving the catalytic efficiency of hydrolases; and GT families (e.g., GT2 and GT4) are involved in microbial polysaccharide synthesis, which is beneficial to the proliferation and metabolic activity of microbes. In addition, the results of the metabolomic analysis of the rumen fluid further verified the activation of carbohydrate metabolism in the LA group: the relative abundances of key intermediate metabolites for acetate synthesis (α-D-glucose-1P and acetyl-CoA) significantly increased, whereas the relative abundance of glucose-6P significantly decreased, which indicated that the glycogen degradation and glycolysis pathways of the rumen microbiota were highly activated and that the metabolic flux was directed toward the synthesis of acetyl-CoA and further to acetate. In vitro fermentation experiments further confirmed that R. albus could significantly increase the molar proportion of rumen acetate, especially when starter was used as the substrate, which directly proved the key role of HPAM in promoting rumen acetate production. Moreover, the large amount of hydrogen produced during carbohydrate fermentation to acetate provided the essential reductive conditions for the classic biohydrogenation of LA in the rumen, forming a “carbohydrate metabolism–hydrogen production-LA biohydrogenation” positive feedback loop in the rumen ecosystem, which resulted in the synergistic metabolism of carbohydrates and lipids in the rumen.

The biohydrogenation of dietary UFAs by the rumen microbiota is a unique metabolic process of ruminants, and Butyrivibrio fibrisolvens and Butyrivibrio proteoclasticus are the core microbes that mediate LA biohydrogenation^24,25. Our study revealed that LA supplementation significantly activated the classic LA biohydrogenation pathway in the rumen: the relative abundance of LA in the rumen fluid of the LA group significantly decreased, the intermediate product TVA significantly decreased, and the final product SA significantly increased, which was consistent with the significant enrichment of Butyrivibrio fibrisolvens and Butyrivibrio proteoclasticus in the LA group. These results indicated that LA supplementation could enrich the core functional microbes of biohydrogenation, accelerate the conversion of LA to SA, and reduce the potential toxic effect of excessive free LA on the rumen microbiota²⁵, which is an important adaptive response of the rumen ecosystem to LA supplementation. Notably, the highly efficient complete biohydrogenation observed in the LA group of this study was mainly attributed to the significant enrichment of Butyrivibrio proteoclasticus, a key rate-limiting bacterium that mediates the conversion of TVA to SA²⁶. Meanwhile, LA promoted ruminal carbohydrate metabolism and generated abundant hydrogen, which provided sufficient reducing power for the final reduction step of biohydrogenation. Consequently, a large amount of SA was produced without obvious accumulation of TVA. A key new finding of this study is that the products of LA biohydrogenation can effectively promote the activation of the PC synthesis pathway in the rumen, and microbe-derived PC is the key metabolite connecting LA biohydrogenation and host growth regulation. The metabolomic results revealed that the relative abundances of both PC (16:0/16:0) and 1,2-dihexadecanoyl-sn-glycero-3-phosphocholine in the rumen fluid and in the rumen tissue of the LA group significantly increased, and the metagenomic results confirmed that LA supplementation enriched the core microbes involved in PC synthesis (Prevotella sp. and Ruminococcus albus). Ruminococcus albus participates in the synthesis of phosphatidic acid (PA), the key intermediate of PC, and Prevotella sp. mediates the conversion of diacylglycerol (DAG) to PC, resulting in the formation of a complete microbe-derived PC synthesis pathway in the rumen. PC is the main component of the microbial cell membrane and rumen epithelial cell membrane^27,28, and the increase in microbe-derived PC not only maintains the integrity and fluidity of the microbial cell membrane but also provides a large amount of exogenous PC for the rumen epithelium, which is the material basis for improving the function of BRECs. This study revealed for the first time the link between LA biohydrogenation and microbe-derived PC synthesis in the rumen, which complemented the understanding of the lipid metabolic function of the rumen microbiota in calves.

BRECs are the key barrier for material exchange and signal transduction between the rumen and the host, and their proliferation, barrier integrity and antioxidant capacity are directly related to rumen development and calf growth^29,30. Previous studies have shown that lipid metabolites can regulate the function of rumen epithelial cells³¹, but the specific molecular targets and regulatory pathways involved remain unclear. Our study revealed that LA supplementation significantly upregulated the expression of CYP3A24 in rumen tissue, and by WGCNA and PPI network analysis, CYP3A24 was identified as the core hub gene, with a significant positive correlation with PC abundance and HPAM abundance, indicating that a PC-CYP3A24 regulatory axis is mediated by LA. Besides CYP3A24, the other eight core hub genes also possess important biological functions. Among them, EVPL is mainly involved in the formation of cornified envelopes in epithelial cells and serves as a key structural gene for maintaining the structural integrity of epithelial tissue³². JUP, as a core protein of desmosomes and adherens junctions, regulates the proliferation and differentiation of epithelial cells³³. CDH1, a classic epithelial cadherin, primarily mediates tight intercellular adhesion and acts as a key gene in maintaining epithelial barrier function and cellular homeostasis³⁴. These results further indicate that dietary LA supplementation can effectively promote the development and maturation of ruminal epithelium in calves by regulating the expression of the above key genes. In vitro cell culture experiments further verified the regulatory role and mechanism of the PC-CYP3A24 axis on BRECs function: 25 μM PC significantly promoted the proliferation of BRECs, increased the expression of tight junction proteins (Occludin, ZO-1, and TJP1), increased the expression of antioxidant genes (glutathione peroxidase 4, GPX4, superoxide dismutase 2, SOD2, and nuclear factor erythroid 2-related factor 2, NFE2L2) and inhibited the expression of proinflammatory cytokines (toll-like receptor 4, TLR4, myeloid differentiation primary response 88, MyD88, and tumor necrosis factor, TNF), while silencing CYP3A24 completely reversed the above regulatory effects of PC, indicating that CYP3A24 is an essential downstream effector of the regulation of the function of BRECs by PC. CYP3A24 belongs to the cytochrome P450 enzyme family, which is widely involved in the metabolism of lipids and other endogenous substances^35,36. Our study revealed that CYP3A24 mediates the regulatory effect of PC on BRECs by activating the Wnt/β-catenin signaling pathway (upregulating PCNA, CTNNB1, and Ki-67) and inhibiting the TLR4/MyD88 inflammatory signaling pathway, which, for the first time, clarifies the biological function of CYP3A24 in BRECs and contributes to the understanding of the molecular mechanism of cytochrome P enzymes in ruminant nutritional metabolism. The activation of the PC-CYP3A24 axis by LA is the key link connecting rumen microbial metabolism and host rumen development: microbially derived PC is absorbed and utilized by BRECs and then upregulates the expression of CYP3A24, which further promotes the proliferation and barrier function of BRECs, improves the antioxidant capacity and reduces inflammation, thereby accelerating the development of rumen papillae, increasing the nutrient absorption area of the rumen, and ultimately improving the growth performance of calves. This regulatory axis enables precise cross-talk between the rumen microbiota and the host and is the core molecular mechanism through which LA promotes calf growth. Nevertheless, the mechanistic validation regarding how linoleic acid regulates ruminal epithelial development in calves was primarily based on multi-omics association analyses and in vitro cellular assays. Further in vivo experiments are still required to verify the causal relationships of the identified ruminal microbiota–metabolite–host gene regulatory axes.

To summarize, this study revealed that dietary LA supplementation at the optimal dose reshaped the rumen microbial community of preweaning calves, enriched HPAM with carbohydrate and lipid metabolic capabilities, enhanced rumen carbohydrate metabolism and acetate production, drove rumen LA biohydrogenation and microbe-derived PC synthesis, and further activated the PC-CYP3A24 axis in rumen tissue, which promoted the proliferation and barrier function of rumen epithelial cells, improved antioxidant capacity, inhibited inflammation, and ultimately accelerated rumen development and improved calf growth performance.

Methods

Ethics approval

All experimental procedures were approved by the Animal Experiment Committee of Ningxia University (approval number: no. NXU-A-2024-116). All animal experiments were conducted in accordance with the principles and regulations for the use of laboratory animals at Ningxia University and followed the ARRIVE guidelines (https://arriveguidelines.org/).

Animal management, diet, and experimental design

A total of 60 Holstein male calves with similar birth weights and 35 d of age were obtained from the Ningxia Farms and Land Reclamation Farm. All calves were individually housed in the cowshed until they were 90 d old.

Sixty healthy male dairy calves (35 ± 3 d, 65.44 ± 5.13 kg) were randomly assigned to four groups: the control group (CON, fed a basal diet, n = 15), the low-dose group (LA I, the basal diet supplemented with LA, the addition amount of LA = 0.15 × body × weight (BW)^0.75), the medium-dose group (LA II, the addition amount of LA = 0.21 × BW^0.75), and the high-dose group (LA III, the addition amount of LA = 0.49 × BW^0.75); the specific addition amount of LA is shown in Table S7. The experimental period ranged from 35 to 90 d after birth, with a 7-day adaptation period. The dosage of LA is dynamically adjusted according to the changes in the BW of calves. LA was obtained from Zebang Biotechnology Co., Ltd (Xi’an, China) and was ≥90% pure. An additional amount of LA was obtained from a previous study³⁷. The calves were allowed unlimited access to both food and water. All animals were housed under identical environmental, feeding, and management conditions throughout the experiment to minimize potential confounders. The nutritional content of the base diet is presented in Table S8. At 90 d of age, ten healthy calves were randomly selected from the CON and the best growth group (The group with the highest BW and ADG), fasted for 12 h, then humanely euthanized via electrical stunning and followed by exsanguination.

Sample collection and measurement

The BW of young calves at 35 and 90 d of age was recorded. The difference between the BW at 90 d old and the BW at 35 d old was divided by the age in days to obtain the ADG. Detailed documentation of the initial quantity of milk provided, the remaining amount, and the consumption of starter feed was maintained to calculate the average daily feed intake (ADFI). After slaughter using exsanguination, all the rumen contents were immediately mixed, the ruminal fluid sample was strained through four layers of sterile cheesecloth, and the collected ruminal contents were divided into three tubes and stored at −80 °C for subsequent microbial DNA extraction, metabolite extraction, and analysis of VFA concentration using gas chromatography. One portion of the rumen wall (3 × 3 cm) from the ventral sac was collected for morphology analysis using histomorphometry microscopy, as described previously. Rumen wall samples were fixed in 4% paraformaldehyde and processed for paraffin embedding and sectioning. Hematoxylin and eosin staining was used to measure the thickness of the ruminal epithelia and smooth muscle layer through Aperio ImageScope × 64 (Leica Biosystems, Germany). After being washed three times in cold phosphate-buffered saline, the ruminal epithelium and muscle layer tissue were collected and stored in liquid nitrogen for RNA extraction.

Determination of rumen fermentation parameters

Rumen fluid sampling occurred at 90 d posttrial initiation, with the initial 50 mL discarded to preclude oropharyngeal contamination. Subsequent filtration through quadruple-layer sterile gauze preceded immediate pH quantification via a calibrated meter (SMART SENSOR, Zhengzhou, China). To quantify the ammonia-nitrogen (NH₃-N) concentration in the rumen fluid, the colorimetric approach proposed by Broderick and Kang³⁸ was employed. To analyze the concentration of VFAs in the rumen, a gas chromatograph (GC-14B; Shimadzu, Japan) was used, and the detailed operational steps followed the protocol presented by Yang et al.³⁹.

Hematoxylin and eosin (H&E) staining

Rumen papilla morphology and light microscopy histomorphometric analysis were carried out as described previously⁴⁰. Briefly, the rumen tissues were embedded in paraffin, deparaffinized, stained with H&E, and sealed with neutral gum. Tissue morphology was observed using the TissueFAXS Plus system (Tissue Gnostics GmbH, Vienna, Austria). The intestinal villus height (VH) was measured using TissueFAXS Viewer (TissueGnostics GmbH, Vienna, Austria) in a double-blind manner, with each slide observed three times.

Metagenomic sequencing analysis

Total genomic DNA was individually extracted from the ruminal content samples using a HiPure Stool DNA Kit (D3141; Meiji, China) following the manufacturer’s protocol. The concentration and purity of the extracted DNA were assessed using a NanoDrop 2000 spectrophotometer. Additionally, DNA quality was evaluated using 1% agarose gel electrophoresis. The construction of paired-end libraries was carried out using the NEBNext® ULtra™ DNA Library Prep Kit (New England Biolabs, USA), where adapters containing sequencing primer hybridization sites were ligated to the blunt-end fragments. Paired-end sequencing was conducted on the NovaSeq X Plus platform (Life Technologies, USA) at GENE BENOVO (Shanghai, China).

FASTP (version 0.20.0) was used to filter the raw data from the Illumina platform. The filtering criteria were as follows: reads containing unknown nucleotides (N) with a proportion ≥10%, reads with a proportion of bases with a phred quality score ≤20 ≥ 50%, reads containing adapters, and reads with a length <50. The filtered clean reads were used for assembly analysis. MEGAHIT (version 1.2.9)⁴¹ was used to assemble the clean reads of each sample separately. MetaGeneMark (version 3.38)⁴² was used to predict genes in the assembled sequences of each sample. Bowtie2 (version 2.3.5.1)⁴³ was used to remap the reads to the assembled sequence set and count the number of reads. Pathoscope (v2.0.7)⁴⁴ was used to filter out genes with fewer than 2 read supports to obtain the final nonredundant gene set. Pathoscope software uses a Bayesian framework to examine the sequence and alignment quality of each read for reassignment. Afterward, for each sample, the gene abundance was calculated according to the following formula:

$$\mathrm{TPM}=\frac{N_i/L_i\times {10}^6}{{\epsilon }_{j=1}^n{N}_j/L_j}$$

TPM represents the relative abundance of genes in the sample. Li denotes the length of genei, Ni represents the number of genesi detected in the sample (for example, the number of aligned reads), and j indicates the number of genes. The nonredundant gene set was subsequently subjected to subsequent annotation analysis using DIAMOND software. The lowest common ancestor (LCA) algorithm⁴⁵ in MEGAN software (version 6.25.10)⁴⁶ was used to obtain species annotation and abundance information. The nonredundant gene set was aligned against the NR (2024.07.25), KEGG (Release 111), and CAZy (2024.07.14) databases using DIAMOND software (version 0.9.25, parameter e-value < 0.00001)⁴⁷. The protein sequence with the highest similarity was selected through alignment, after which the corresponding protein function annotation was obtained.

The difference in microbial alpha diversity between the two groups was evaluated utilizing the Wilcoxon rank-sum test, with a false discovery rate (FDR)-adjusted P value < 0.05 indicating statistical significance. To assess beta diversity, PCA was conducted employing the Bray-Curtis distance algorithm, with intergroup differences assessed through the Adonis method. Microbial domains were compared using the Wilcoxon rank-sum test, with an FDR-adjusted P value < 0.05 considered to indicate a significant difference. Microbial phyla, genera, and species were compared using LEfSe, with an LDA score >2 and a P value < 0.05 considered to indicate significant differences. The relative abundances of KEGG pathways, KEGG modules, KEGG enzymes, and CAZymes were also compared between the two groups using LEfSe, and significant differences were considered when the LDA score was >2 and P was <0.05. Sankey diagrams and correlation analyses were plotted using cloud tools on the OmicShare platform (https://www.omicshare.com/).

Analysis of the rumen fluid and rumen tissue metabolome

LC–MS/MS analyses of the rumen fluid metabolome and rumen tissue metabolome were conducted on a UHPLC (1290 Infinity LC, Agilent Technologies) coupled to a quadrupole time-of-flight (AB Sciex TripleTOF 6600) at Gene Denovo Biotechnology Co., Ltd (Guangzhou, China). The raw data were converted to MzML format using ProteoWizard, after which peak alignment, retention time correction, and peak area extraction were performed with the XCMS program. For the data extracted by XCMS, the following steps were conducted sequentially: first, metabolite structure identification and data preprocessing (including missing value filtering: removing ion peaks with a missing value rate > 50%; missing value imputation: K-nearest neighbor (KNN) imputation; and data filtering: filtering out features with a relative standard deviation (RSD) > 50%). Finally, data analysis was carried out.

The Ropls package was used to conduct orthogonal partial least squares discriminant analysis (OPLS-DA). To screen for differentially abundant metabolites, the VIP value from the OPLS-DA was integrated with the results of the t test and the P value obtained through univariate statistical analysis. Metabolites were considered significantly different when they met the criteria of having a VIP > 1 and a P < 0.05. Volcano plots were used to visualize differentially abundant metabolites by integrating VIP scores (VIP > 1), P < 0.05, and log2-transformed fold-change values from comparative analyses. Differentially abundant metabolites between the comparison groups were normalized (z-score) and clustered, and heatmaps were generated using the R package heatmap. A KEGG enrichment bubble diagram was constructed using R packages related to enrichment analysis, such as ClusterProfiler and EnrichPlot.

Analysis of the transcriptome of the rumen tissue

Total RNA was extracted with TRIzol (Invitrogen Life Technologies, Carlsbad, CA, USA), and its concentration and purity were measured using a NanoDrop (NanoDrop Technologies, Wilmington, DE, USA). mRNA was enriched using mRNA capture beads. After the magnetic beads were purified, the mRNA was fragmented at high temperature. Using the fragmented mRNA as a template, the first strand of cDNA was synthesized in a reverse transcriptase mixture system. While the second strand of cDNA was synthesized, end repair and A-tailing were completed. Subsequently, the adapters were ligated, and the target fragments were purified and selected using Hieff NGS® DNA Selection Beads. Afterward, PCR library amplification was performed, and finally, detection was carried out using an Illumina NovaSeq X Plus.

The raw reads obtained from sequencing were subsequently subjected to quality control using fastp⁴⁸. Low-quality data were filtered out to obtain clean reads. The clean reads were then aligned to the bovine ribosomal database using the short-read alignment tool Bowtie2⁴³. Reads that aligned to the ribosome were removed without allowing mismatches, and the remaining unmapped reads were used for subsequent transcriptome analysis. The paired-end sequencing reads were aligned to the reference genome using HISAT 2⁴⁹. In this study, the reference genome was GCF_002263795.2_ARS-UCD1.3. On the basis of the fragments per kilobase of transcript per million fragments mapped (FPKM) of genes in each sample, DEGs were identified using differential analysis software, and the differences were analyzed using DESeq2 software. |FC| ≥ 10 and FDR < 0.05 were used as the screening criteria for DEGs. Functional enrichment analysis was performed using the R/clusterProfiler and R/topGO packages. WGCNA was performed using the Omicsmart online platform (https://www.omicsmart.com/). A PPI analysis was performed with the STRING database. Core genes were identified according to their BC using the STRING database and Cytoscape. Venn diagrams and correlation analyses were plotted using cloud tools on the OmicShare platform (https://www.omicshare.com/).

In vitro fermentation experiment

Three Holstein dairy cows with rumen fistulas were selected as rumen fluid donors. Rumen fluid was collected before the morning feeding, filtered through four layers of gauze, thoroughly mixed, and stored in a thermos flask. Fermentation broth was collected using an in vitro fermentation system, with each bottle containing 80 mL of mixed buffer, 40 mL of rumen fluid, and 1 g of substrate. Substrates were divided into three categories: 1 g of milk replacer, 1 g of starter, and a mixture of 0.79 g of milk replacer and 0.21 g of starter. For each substrate, RAM (Ruminococcus albus medium) liquid medium was added as the control group (CON), and RAM medium inoculated with Ruminococcus albus (DMSZ, Germany) at a concentration of 10⁷ CFU/mL was used as the treatment group (RA). The headspace of each fermentation bottle was purged with carbon dioxide, followed by incubation at 39 °C for 48 h. After fermentation, the liquid was collected for VFA determination. For each fermentation sample from the different treatments, six replicates were established.

BRECs isolation

Healthy rumen tissue was collected from Holstein cattle, after which the surface digesta was rinsed off using physiological saline. The cleaned tissues were immediately immersed in PBS supplemented with 5× penicillin‒streptomycin, 5× amphotericin B, and 5× gentamicin, followed by rapid transport to the laboratory. After the serosa and muscular layers were removed by hand, the remaining rumen tissues were transferred to a biosafety cabinet for cell digestion. The tissues were subsequently washed 6 times with the aforementioned antibiotic-supplemented PBS, cut into 2 × 2 mm cubes and placed in 10 cm culture dishes. Differential time digestion was performed using 0.25% trypsin + 0.02% EDTA (sufficient to cover the tissue blocks) in a 37 °C incubator for 7 cycles (40, 30, 20, 10, 8, 5, and 5 min). After each digestion cycle, the supernatant was observed microscopically; fresh trypsin was used for each cycle, followed by rinsing with 1× antibiotic-supplemented PBS. When many small, oval/round cells with smooth edges appeared in the digestate, the supernatant and rinse solution were collected. Digestion was terminated by adding an equal volume of medium containing 20% fetal bovine serum. The cell suspension was filtered 3 times through a 200-mesh stainless steel sieve, and the filtrate was centrifuged at 1000 rpm for 5 min. The pellet was resuspended and washed twice with basic medium (1000 rpm, 5 min each), resuspended in complete medium and seeded into 10 cm dishes for adherence.

Immunofluorescence

BRECs were seeded into 6-well plates. After 24 h (approximately 70% confluence), the medium was discarded, and the cells were washed 3 times with PBS. The cells were fixed with 4% paraformaldehyde at room temperature for 20–30 min, followed by 3 washes with PBS (5 min each). After blocking with 5% BSA at room temperature for 2 h, the blocking solution was removed. The cells were subsequently washed with PBS and incubated with a CK-18 primary antibody (1:400 dilution) at 4 °C overnight; isotype-matched IgG was used as a control in the control group. Following 3 washes with PBS (10 min each), the cells were incubated with a Cy3-conjugated secondary antibody (1:100-500) at 37 °C for 1 h in the dark. After being washed with PBS, the cells were stained with ready-to-use DAPI for 5 min in the dark and observed and imaged under a fluorescence microscope.

CCK-8 assay

Cell proliferation was assessed using a CCK-8 kit (Solarbio, Beijing, China). For the CCK-8 assay, rumen cells were seeded into 96-well plates at 100 μL of medium per well, with 6 independent replicates per treatment group. PC (MCE, Japan) was added to the 96-well plates at final concentrations of 0, 25, 50, 75, and 100 μM, followed by continued incubation for 3, 6, 12, and 24 h. Before detection, 10 μL of CCK-8 reagent was added to each well. The cell culture plate was gently shaken to ensure uniform mixing and incubated in an incubator for 1 h, after which the absorbance at 450 nm was measured using a microplate reader.

Cell processing

Three small interfering RNA (siRNA) fragments targeting the CYP3A24 gene were designed and synthesized. BRECs were seeded into 6-well plates and transfected when the cell confluence reached approximately 80%, following the manufacturer’s instructions for Lipofectamine 3000 (Invitrogen, USA). Half an hour before transfection, the cell culture medium was replaced with serum-free OPTI-MEM. Solution A and solution B were mixed and incubated at room temperature for 20 min. The mixture was gently added dropwise to the cells in the 6-well plates and mixed thoroughly. The plates were incubated at 37 °C in a 5% CO₂ cell incubator for 6 h, after which the medium was replaced with serum-containing growth medium. The cells were harvested for subsequent assays 48 h posttransfection.

BRECs were seeded into 6-well plates. When the cell confluence reached 70-80%, the cells were treated with PC at a final concentration of 25 μM for 12 h. The medium was discarded, and the cells were washed 3 times with PBS. Subsequently, 1 mL of TRIzol reagent was added to lyse the cells, and the lysate was collected for RNA extraction. For the interference group, the cells were treated with PC for 12 h at 36 h post-CYP3A24 siRNA transfection, after which the cells were collected for subsequent assays.

Real-time qPCR

Total RNA extraction, reverse transcription, and real-time PCR were performed as previously described. The primer sequences for the genes are listed in Table S9. The expression levels of target genes related to BREC barrier integrity, Occludin, HIF-α, ZO-1 and TJP1; mucosal repair, including PCNA, CTNNB1, antigen identified by Ki-67, and TCF4; and antioxidant activity, including GPX4, kelch-like ECH-associated protein 1 (KEAP1), SOD2, and NFE2L2; and immune function TLR4, MyD88, reticuloendotheliosis viral oncogene homolog A (RELA) and TNF, were quantified using the 2^−ΔΔCt method, with the relative FC calculated against the CON group. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a reference gene.

EdU and ROS staining

Cell proliferation was detected using an EdU kit (Beyotime, Shanghai, China) following the aforementioned protocol⁵⁰. Intracellular ROS levels were measured with a ROS detection kit (Lablead, Beijing, China) as follows: DCFH-DA was diluted 1:1000 in serum-free medium to a final concentration of 10 μM; the treatment medium was aspirated, and an appropriate volume of the diluted DCFH-DA working solution was added to the cells. After being incubated at 37 °C in the dark for 30 min, the cells were washed 1–2 times with serum-free medium and imaged under an inverted fluorescence microscope.

Statistical analysis

All data were first tested for normality using the Shapiro‑Wilk test and for homogeneity of variance using Levene’s test. If the assumptions of normality and homogeneity were satisfied, parametric statistical methods (e.g., independent‑samples t‑test) were used. If the assumptions were violated, data were log‑transformed or non‑parametric tests (e.g., Wilcoxon rank-sum test) were applied instead. Statistical analyses of the rumen fermentation parameters were performed using the independent sample t test in SAS software, with a significance level of P < 0.05. For each experimental group, data are presented as the mean ± SEM. One-way ANOVA followed by t-test (two groups). The effect size and 95% confidence interval were calculated to quantify the magnitude of differences between groups, where applicable. All figures were constructed using GraphPad Prism software. Spearman’s rank correlation was used for all correlation analyses in this study (Spearman’s |r| > 0.50 and P < 0.05). Multiplex networks were visualized using Gephi software (version 0.10.1).

Author contributions

C.L.H. Writing—original draft, writing—review and editing, conceptualization, and prepared figures. X.H.M, X.X.L, and D.Z.W. data curation, software. S.W. and J.N.D. writing—review and editing; Y.F.M. conceptualization, framework design, manuscript finalization, quality control, funding acquisition, project administration, and supervision. All authors reviewed the manuscript.

Data availability

The rumen metagenomic sequencing data generated in this study have been deposited in the National Center for Biotechnology Information BioProject database under accession number PRJNA1464113 http://www.ncbi.nlm.nih.gov/sra/PRJNA1464113. The rumen transcriptomic sequencing data generated in this study have been deposited in the National Center for Biotechnology Information SRA database under accession number PRJNA1463994 https://www.ncbi.nlm.nih.gov/sra/PRJNA1463994.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41522-026-01063-y.