Guanidinoacetic acid enhances Tibetan sheep–meat quality through the gut–muscle axis: insights from metabolome–microbiome integration Authors: Rengeerli Sa, Fengshuo Zhang, Xianhua Zhang, Yu Zhang, Shengzhen Hou, Linsheng Gui Published: 2026 ABSTRACT AbstractThis study evaluated whether dietary guanidinoacetic acid (GAA) was associated with meat quality, jejunal microbiota and muscle metabolomic profiles in Tibetan sheep. A total of 120 healthy, 2-month-old, weaned male Tibetan sheep (17.33 ± 0.21 kg) were allocated to four groups: CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. The feeding trial lasted 90 days after a 10-day adaptation. Compared with CON, 0.12% GAA was associated with higher relative abundances of selected jejunal taxa (Eubacterium_nodatum_group, Family_XIII_AD3011_group and Bacillus) and increased acetate and propionate concentrations. Untargeted metabolomics identified significantly altered muscle metabolites associated with glycerophospholipid metabolism, including PC 37:5, PC 38:4 and LPA 14:0. The HG group also showed higher glutathione peroxidase activity and total antioxidant capacity, a higher a* value and lower shear force. Correlation analyses indicated associations among jejunal microbiota, muscle metabolites and meat-quality traits. These findings provide further evidence supporting the potential use of dietary GAA as a feed additive for improving selected meat-quality traits in Tibetan sheep.Keywords: Ruminant nutrition, Jejunal microbiota, Short-chain fatty acids, Glycerophospholipid metabolism, Antioxidant capacity, Feed additive Introduction Tibetan sheep (Ovis aries) are an important livestock species in Qinghai Province. This breed contributes substantially to the local economy and income of herders owing to its adaptive traits such as cold tolerance and disease resistance (Lu et al., 2025). Sheep production in Qinghai Province has traditionally relied on extensive grazing systems that depend entirely on natural rangeland resources. However, recent years have seen intensified grazing pressure, which has caused forage resource depletion. Consequently, sheep growth performance and meat quality may be compromised, while grazing pressure may further degrade the local environment. Simultaneously, the ecological environment has degraded (Fan et al., 2021). To address these challenges, intensive housing systems have been progressively implemented to enhance Tibetan sheep productivity. Although the transition from extensive grazing to intensive feeding systems has improved production efficiency, it has raised significant concerns about meat quality, nutritional value and safety (Zhang et al., 2022). As market demand for Tibetan sheep meat continues to expand, consumer expectations for high-quality meat have correspondingly increased. Concurrently, rising feed costs highlight the need for strategies that optimise feed conversion efficiency while maintaining or improving meat quality (Jin et al., 2024). Accordingly, feed additives may offer a practical strategy. Additives that modulate intestinal microbiota and their metabolic functions may offer novel strategies for improving meat quality and production efficiency in Tibetan sheep farming (Wang et al., 2024). Guanidinoacetic acid (GAA), recognised as a nutritional feed additive, is a metabolic intermediate biosynthesised from arginine and glycine and serves as the sole endogenous precursor for creatine synthesis in vertebrates (Portocarero & Braun, 2021). Creatine contributes to rapid energy buffering and cellular energy metabolism in animals. Compared with creatine, GAA exhibits practical advantages such as lower production costs, higher bioavailability and better chemical stability. Thus, GAA is a widely studied alternative to direct creatine supplementation (Ostojic & Jorga, 2023). In broilers, dietary supplementation with 0.04% GAA has been reported to modulate creatine metabolism by increasing creatine kinase activity and serum arginine concentrations, thereby improving growth performance and meat quality parameters (Cao et al., 2024). In fattening pigs, GAA supplementation increased pH at 45 min, reduced water loss and shear force, and improved meat-quality traits (Li et al., 2018). In ruminants, dietary incorporation of GAA at 0.8 and 1.6 g kg−1 enhanced growth performance, modulated ruminal fermentation parameters and haematological indices and improved digestibility and nitrogen metabolism efficiency in Angus steers (Yi et al., 2024). In Kazakh male sheep, phosphocreatine (PCr) and adenosine diphosphate concentrations in the longissimus dorsi muscle increased linearly with GAA supplementation, accompanied by higher CK activity (He et al., 2018). These studies indicate that GAA supplementation has been associated with improved animal growth performance and meat quality. However, research on the application of GAA in Tibetan sheep production systems remains limited. In the present study, we used rumen-protected GAA, which is designed to reduce ruminal degradation and increase post-ruminal availability. Given the important role of the jejunum to examine potential associations among GAA supplementation, jejunal microbial changes and meat-quality traits. The jejunum represents an important anatomical site for nutrient digestion and absorption in ruminants. Simultaneously, it serves as a site involved in for microbial community establishment and immune function modulation (Liu, Lu, et al., 2025; Liu, Zhang, et al., 2025). The abundance and composition of jejunal microbiota have been associated with feed digestibility, intestinal barrier function and meat-quality traits (Liu et al., 2024). The gut-muscle axis has been proposed as a bidirectional communication pathway; microbially derived metabolites may act as signalling molecules and influence skeletal muscle metabolism and phenotype. However, direct evidence in ruminants remains limited (Mancin et al., 2023). The intestine plays a crucial role in maintaining animal health and performance through regulation of metabolic and endocrine functions (Han et al., 2023). The species composition of jejunal microbiota can be accurately identified using 16S rRNA amplicon sequencing, and its functional potential can be predicted through comparative analysis of microbial community data (Ma et al., 2023). Metabolite profiles of the longissimus dorsi muscle can be analysed using untargeted metabolomics, enabling a comprehensive assessment of muscle quality characteristics and nutritional value (Zhang et al., 2025). Therefore, metabolomics provides a valuable tool for examining metabolic changes associated with GAA supplementation and meat quality (Ueda et al., 2024). To address these knowledge gaps, we integrated 16S rRNA sequencing and untargeted metabolomics to examine whether different dietary doses of GAA were associated with changes in jejunal microbial communities, muscle metabolomic profiles and meat-quality traits in Tibetan sheep. We further assessed correlations among jejunal taxa, muscle metabolites and meat-quality traits to identify potential microbiota-metabolite-phenotype associations related to GAA supplementation. Materials and methods Animals and housing A total of 120 2-month-old weaned male Tibetan sheep with similar initial body weights (17.33 ± 0.21 kg) were selected from a well-managed large-scale commercial Tibetan sheep farm in Gonghe County, Hainan Prefecture, Qinghai Province. All sheep were clinically healthy and had completed the farm's routine vaccination programme. The 100-day experiment comprised a 10-day adaptation period followed by a 90-day feeding trial. Each replicate pen housed five sheep in a 4.0 m × 5.0 m indoor pen connected to a 5.0 m × 6.0 m outdoor area. Animals had free access to both indoor and outdoor areas, allowing adequate space for movement and natural behaviours. Sheep were fed twice daily at 08:00 and 17:00 h, with ad libitum access to feed and water. Animal health and welfare status were monitored at each feeding by observing appetite, behaviour, locomotion, and general condition. Experimental design and randomisation The 120 eligible sheep were assigned to 24 pens, with five sheep per pen. The pens were then randomly allocated to four dietary treatments using a computer-generated random-number sequence, resulting in six replicate pens per treatment. The treatments were CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. Initial body weight was assessed before the dietary intervention to confirm baseline comparability among treatments. No additional blocking or stratification was applied. The pen was defined a priori as the experimental unit for treatment allocation and statistical analysis. The composition and nutrient levels of the concentrate diet are presented in Table S1. Sample collection and processing At the conclusion of the experiment, one sheep was selected from each replicate pen for slaughter and sample collection using a within-pen random selection procedure, resulting in a total of 24 slaughtered sheep and n = 6 independent biological samples per treatment. Briefly, the five sheep in each pen were individually identified by ear-tag number and assigned numbers from 1 to 5. A random number between 1 and 5 was then generated using a computer-based random-number function, and the sheep corresponding to that number was selected for slaughter. No selected animals were replaced or excluded after selection. Thus, each analysed biological sample represented a distinct replicate pen. For phenotypic measurements, each biological sample was assessed in three technical replicates, and the mean value of the three technical replicates was used for statistical analysis. Technical replicates were not treated as independent biological replicates. Sample size was based on the available pen-level design; no formal a priori power calculation was performed (Ma et al., 2024; Ma et al., 2025). Selecting one sheep per pen yielded six independent biological replicates per treatment for slaughter, microbiome and metabolomic analyses. Subsequently, longissimus dorsi muscle and jejunal content samples were collected. The jejunal contents and muscle tissues were aliquoted into cryogenic storage tubes, transported to the laboratory in liquid nitrogen, and stored at −80 °C for subsequent 16S rRNA amplicon sequencing and untargeted metabolomic analysis. Blinding was partially implemented using coded group information. Personnel and assessors were aware that samples belonged to different treatment groups but were not informed of the specific group identities (i.e., control, low-, medium- or high-GAA) during carcass and meat-quality measurements, antioxidant assays, 16S rRNA sequencing and metabolomic analyses. Data analysts were also blinded to the treatment identities during statistical analysis. Treatment identities and GAA inclusion levels were decoded only after data acquisition, processing and statistical analysis had been completed. No slaughtered animals or collected biological samples were excluded from the final analyses, and no data points were removed as outliers. For microbiome and metabolomic datasets, features were filtered solely according to the quality-control criteria described below. Carcass traits and meat quality parameter determination Carcass and meat-quality traits of Tibetan sheep were evaluated at the end of the 90-day feeding trial. After slaughter, carcass weight was determined as the weight of bone-in meat after removing the head, hooves and viscera. Rib meat thickness (RMT) was measured between the 12th and 13th ribs. Abdominal wall thickness was measured using vernier calipers at the 12th rib level on the ventral abdominal wall. Meat colour was measured on the longissimus dorsi muscle at the 12th–13th rib region after slaughter. Before measurement, the freshly cut muscle surface was allowed to bloom in air for 30 min (Chen et al., 2022; Devapriya et al., 2021; Güngör et al., 2022). Measurements were performed using a PHSJ-4 A colorimeter. The colorimeter was calibrated with black and white standard plates before measurement. L* (lightness), a* (redness) and b* (yellowness) values were recorded. Three readings were taken at different positions on each sample, and the mean value was used for statistical analysis. Muscle pH was measured in the longissimus dorsi muscle at the 12th–13th rib region at 45 min and 24 h post-mortem using a calibrated pH meter (PHSJ-6L, Shanghai, China). The pH meter was calibrated with standard buffer solutions of pH 4.00 and 7.00 before measurement. Three readings were taken from pre-marked positions on each sample, and the mean value was used for statistical analysis. Subsequently, thawing loss, cooking loss, water loss and shear force were determined according to Honikel (Honikel, 1998). Specifically, for thawing loss, frozen meat samples were weighed before thawing at 4 °C for 12 h (W1, g). After thawing, samples were gently blotted dry with filter paper and weighed again (W2, g). Thawing loss was calculated as follows: thawing loss (%) = [(W1 − W2)/W1] × 100. For cooking loss, thawed meat samples were blotted dry and weighed before cooking (M1, g). Samples were then sealed in cooking bags and heated in a water bath at 85 °C for 20 min. After cooking, samples were cooled to room temperature, blotted dry and weighed again (M2, g). Cooking loss was calculated as follows: cooking loss (%) = [(M1 − M2)/M1] × 100. Water loss was measured using a MAEC-18 hydraulic testing instrument (Nanjing, China). Meat cubes (1 cm × 1 cm × 1 cm) were weighed before compression (m1, g), subjected to a compressive force of 350 N, and weighed again after compression and removal of surface exudate (m₂, g). Water loss was calculated as follows: water loss (%) = [(m₁ − m₂)/m₁] × 100. Shear force was determined using a MAQC-12 shear force analyser (Nanjing, China). For this measurement, meat samples were first heated in a water bath at 80 °C until the core temperature reached 70 °C. Meat cubes (1 cm × 1 cm × 1 cm) were selected, with muscle fibres oriented perpendicular to the cutting blade during measurement. Antioxidant activity determination in longissimus dorsi muscle Initially, 1.0 g of longissimus dorsi muscle tissue was weighed and transferred to a centrifuge tube. Subsequently, 0.9 mL of phosphate-buffered saline was added, and the mixture was homogenised using a tissue homogeniser (Tiss-Basic48, Jingxin Company, Shanghai, China) to obtain a turbid suspension. The homogenate was centrifuged at 3000×g for 20 min, and the supernatant was stored at −20 °C. Antioxidant parameters in the muscle tissue were evaluated using commercial sheep-specific antioxidant assay kits according to the manufacturer's instructions. These kits were used to quantify antioxidant enzyme activities or oxidative stress-related indices: sheep glutathione peroxidase (GSH-Px) kit (Catalog No. YJ774934), sheep superoxide dismutase (SOD) kit (Catalog No. YJ061711), sheep total antioxidant capacity (T-AOC) kit (Catalog No. YJ022376), sheep catalase (CAT) kit (Catalog No. YJ775000), and sheep malondialdehyde (MDA) kit (Catalog No. YJ774922). All kits were manufactured by Shanghai Yuanju (Shanghai, China). Fatty acid (FA) composition analysis in longissimus dorsi muscle FA composition in longissimus dorsi muscle was quantified by Beijing Novogene Co., Ltd. (Beijing, China) using an ExionLC™ AD UHPLC system coupled to a QTRAP® 6500+ mass spectrometer (AB SCIEX Corp., Boston, MA, USA) with isotope-labelled internal standards (ISs). Samples were ground in liquid nitrogen and extracted with isopropanol/acetonitrile (1:1, v/v), followed by centrifugation. For derivatisation, 50 μL of the supernatant was mixed with 150 μL of 3-nitrophenylhydrazine (3-NPH) reagent containing EDC·HCl and pyridine in acetonitrile/water and incubated at 40 °C for 40 min. The derivatised samples were diluted with isopropanol/acetonitrile (1:1, v/v), and 95 μL of the diluted supernatant was mixed with 5 μL of the mixed IS solution before analysis. Calibration standards containing the 30 selected FAs were prepared in a fatty acid-free matrix. The complete analyte list is provided in Table S2B. QC samples and solvent blanks were included to monitor instrument stability and background signals. Chromatographic separation was performed on a Waters ACQUITY UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm) maintained at 40 °C. mobile phases A and B were 0.1% formic acid in acetonitrile/water (1:1, v/v) and isopropanol/acetonitrile (1:1, v/v), respectively. The flow rate was 0.30 mL min−1, and the injection volume was 2 μL. The complete gradient program is provided in Table S2A. The mass spectrometer was operated with an electrospray ionisation source in negative-ion multiple-reaction monitoring (MRM) mode. The source parameters were as follows: ion spray voltage, −4500 V; curtain gas, 35 psi; ion source temperature, 550 °C; and ion source gases 1 and 2, 60 psi. The analyte-specific precursor-to-product ion transitions, retention times, declustering potentials, and collision energies are provided in Table S2B. Calibration curves were constructed by plotting the nominal standard concentration on the x-axis and the analyte-to-IS peak-area ratio on the y-axis. Calibration curves with correlation coefficients of r > 0.99 were accepted. FA concentrations were calculated from the corresponding calibration equations, corrected for the dilution factor and sample weight, and subsequently converted to mg/100 g fresh meat. Method-validation results for accuracy, intra- and inter-day precision, and stability are provided in Table S3. Short-chain fatty acid (SCFA) determination in jejunal contents Jejunal-content aliquots (100 μL) were mixed with 400 μL of 80% methanol in water (80:20, v/v), homogenised, and centrifuged at 12,000×g and 4 °C for 10 min. A 50 μL aliquot of the supernatant was mixed with 150 μL of derivatisation reagent and incubated at 40 °C for 40 min. The derivatised sample was diluted 100-fold with 80% methanol, and 95 μL of the diluted supernatant was mixed with 5 μL of a mixed isotope-labelled internal standard (IS) solution before UHPLC–MS/MS analysis. Individual standards for the six SCFAs quantified in this study were accurately weighed and combined to prepare a mixed calibration stock solution, which was serially diluted with methanol to prepare the working calibration solutions. Isotope-labelled IS solutions were prepared at appropriate concentrations and mixed thoroughly. All stock and working solutions for calibration standards, ISs, and QCs were stored at −20 °C. SCFAs were quantified using an ultrahigh-performance liquid chromatography-tandem mass spectrometry system (Vanquish™ Flex UHPLC coupled to a TSQ Altis™ mass spectrometer; Thermo Fisher Scientific, Waltham, MA, USA). Separation was performed on a Waters ACQUITY UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm; Waters Corp., Milford, MA, USA) maintained at 40 °C. The mobile phases were 10 mM ammonium acetate in water (solvent A) and acetonitrile/isopropanol (1:1, v/v) (solvent B), both of LC-MS grade. The flow rate was 0.30 mL min−1, and the injection volume was 2 μL. The gradient program was as follows: 25% B from 0.0 to 2.5 min; increased to 30% B at 3.0 min, 35% B at 3.5 min, 38% B at 4.0 min, 40% B at 4.5 min, 45% B at 5.0 min, 50% B at 5.5 min, 55% B at 6.5 min, 58% B at 7.0 min, and 70% B at 7.5 min; increased to 100% B at 7.8 min and maintained until 10.0 min; returned to 25% B at 10.1 min; and maintained at 25% B until 12.0 min for column re-equilibration. The complete gradient program is also provided in Table S4A. The mass spectrometer was operated with an electrospray ionisation source in negative-ion multiple-reaction monitoring (MRM) mode. The source parameters were as follows: spray voltage, −4500 V; sheath gas, 35 psi; ion-source temperature, 550 °C; auxiliary gas, 50 psi; and collision gas, 55 psi. The precursor-to-product ion transitions, retention times, declustering potentials, and collision energies for the six SCFAs and their corresponding isotope-labelled ISs are provided in Table S4B. SCFA concentrations were calculated using isotope-normalised calibration curves, with the analyte-to-IS concentration ratio plotted on the x-axis and the analyte-to-IS peak-area ratio plotted on the y-axis. Calibration curves with correlation coefficients of r > 0.99 were accepted. Sample concentrations were calculated from the corresponding regression equations and corrected for the 100-fold dilution factor. Method-validation results, including intra-day and inter-day precision, recovery, and stability, are presented in Table S4C. Jejunal microbiota analysis DNA was extracted from jejunal content samples using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. The samples were thawed, mixed with InhibitEX Buffer, vortexed and incubated at 70 °C for 5 min. After centrifugation, the supernatant was transferred to a new tube and proteinase K was added. DNA was further purified using a spin-column system and finally eluted with finally eluted with Buffer ATE. The bacterial community was characterised by amplification of the V4 region of the 16S rRNA gene using primers 515F and 806R. The polymerase chain reaction (PCR) mixture contained 15 μL of Phusion High-Fidelity PCR Master Mix, 0.2 μM of each primer and approximately 10 ng of genomic DNA template. Thermal cycling was performed under the following conditions: initial denaturation at 98 °C for 1 min; 30 cycles of 98 °C for 10 s, 50 °C for 30 s and 72 °C for 30 s, followed by a final extension at 72 °C for 5 min. PCR products were purified using magnetic beads and quantified using Qubit fluorometry and quantitative PCR. Sequencing libraries were prepared by adding sample-specific indices, pooled at equimolar concentrations and sequenced by Novogene Bioinformatics Technology Co., Ltd. (Beijing, China) on an Illumina NovaSeq 6000 platform, generating 300-bp paired-end reads. Raw paired-end reads were assigned to samples according to unique barcodes, and barcode and primer sequences were removed. Sequence processing was performed using the QIIME2 pipeline. Reads were quality filtered, denoised, merged and checked for chimeric sequences using DADA2, generating amplicon sequence variants (ASVs), a feature table, and representative ASV sequences. Representative ASV sequences were taxonomically annotated against the SILVA 138.1 database (https://www.arb-silva.de/documentation/release-1381). Downstream analyses, including diversity analyses and taxonomic composition analyses, were performed based on the ASV abundance table (Tables S5–S7). Untargeted metabolomic analysis Muscle tissue samples (100 mg) were homogenised, and metabolites were extracted using 80% methanol in water. After vortexing, ice-bath incubation and centrifugation, the extracts were diluted to a final methanol concentration of 53% and centrifuged again, and the supernatants were collected and analysed via LC-MS. QC samples were prepared by pooling equal volumes from each experimental sample. The pooled QC samples were analysed before, during, and after the analytical sequence to monitor analytical reproducibility and instrument stability. Blank samples were prepared by replacing the sample with 53% methanol in water. Untargeted metabolomic profiling was performed using a Vanquish™ UHPLC system coupled to a Q Exactive™ HF Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA). Chromatographic separation was conducted on a Hypersil Gold C18 column (100 × 2.1 mm, 1.9 μm; Thermo Fisher Scientific) maintained at 40 °C with a flow rate of 0.2 mL min−1. Mobile phase A consisted of 0.1% formic acid in water, and mobile phase B consisted of methanol. The mass spectrometer was operated in both positive and negative electrospray ionisation modes with a resolution of 60,000 (at m/z 200) over a scan range of m/z 100–1500. Data-dependent MS/MS acquisition was employed for structural identification. Data preprocessing was performed using Compound Discoverer 3.3 software (Thermo Fisher Scientific), including peak alignment, feature extraction, peak-area integration, and metabolite identification. Missing peak-area values were automatically filled using the automatic filling function implemented in Compound Discoverer 3.3 during data preprocessing. For metabolite identification, a mass tolerance of 5 ppm was applied, and identification confidence was assessed using the reverse dot score from the mzCloud database, with only metabolites achieving a score ≥ 70 retained for further analysis. QC clustering was assessed by principal component analysis incorporating both biological samples and pooled QC samples. Instrument stability was further evaluated by calculating Pearson correlation coefficients among the pooled QC samples based on metabolite relative peak areas. The pairwise Pearson correlation coefficients ranged from 0.997 to 0.998 in the positive-ion mode and from 0.995 to 0.996 in the negative-ion mode, indicating high analytical reproducibility and satisfactory instrument stability (Fig. S1). Peak area correction was conducted using the first QC sample, with parameters set as follows: mass deviation ≤ 5 ppm, signal intensity deviation ≤ 30%, and signal-to-noise ratio ≥ 3 with a minimum signal intensity of 100,000. Background ions were removed using blank samples, followed by total-peak-area normalisation and exclusion of compounds with a coefficient of variation (CV) > 30% (Tables S8–S10). Statistical analysis The pen was defined as the experimental unit for all statistical analyses. For slaughter-related phenotypic, microbiome, and metabolomic endpoints, one sheep was randomly selected from each replicate pen; thus, each observation represented a distinct pen, yielding n = 6 independent biological replicates per treatment. Technical replicates were averaged prior to statistical analysis and were not included as independent observations in the statistical model. Statistical analyses were performed using IBM SPSS Statistics, version 25.0 (IBM Corp., Armonk, NY, USA). Before parametric testing, normality and homogeneity of variance were assessed using the Shapiro-Wilk and Levene's tests, respectively. One-way analysis of variance (ANOVA) was applied to analyse carcass traits, carcass traits, meat-quality indices, antioxidant indices and FA concentrations and SCFAs. For data meeting parametric assumptions, Duncan's multiple range test was used for post hoc comparisons of the means. In cases where assumptions were violated, the non-parametric Kruskal-Wallis test followed by Dunn's post hoc test with Benjamini-Hochberg correction was used. Given that standard errors were similar across treatment groups, pooled standard error of the mean (SEM) derived from the residual mean square of the ANOVA was reported for each parameter. All data are presented as means with pooled SEM, and the p-values from ANOVA, linear contrasts, and quadratic contrasts are reported for each parameter, with p < 0.05 considered statistically significant. Correlation analysis among jejunal microbiota, muscle metabolites, and phenotypes was performed using Spearman's rank correlation, while Mantel tests were used to assess the associations between the selected microorganisms or metabolites and phenotypic variables. Low-abundance variables were filtered (metabolite detection rate ≥ 80%; microbial genera relative abundance ≥ 0.1%). Heatmaps were generated using ChiPlot (https://www.chiplot.online/). Results Results of carcass traits and meat quality parameters As shown in Table 1, rib meat thickness was significantly higher in the LG, MG and HG groups than in the CON group (p = 0.049). The a* value was significantly higher in the HG group than in the other groups (p = 0.013). Shear force was significantly lower in the HG group than in the CON group (p = 0.042). In addition, L*, b* and pH at 45 min showed significant linear increases with dietary GAA level (p = 0.032, p = 0.044 and p = 0.023, respectively), whereas thawing loss showed a quadratic response (p = 0.042). Table 1Effects of dietary guanidinoacetic acid (GAA) supplementation at different concentrations on carcass traits and meat quality indices in Tibetan sheep.ItemsGroupsSEMp-valueCONLGMGHGANOVALinearQuadraticCarcass weight (kg)17.017.217.717.40.210.7290.4310.582Rib meat thickness (mm)11.8b14.7a14.1a14.1a0.450.0490.1170.093Abdominal wall thickness (mm)2.72.52.52.30.080.9470.0500.874L*33.633.636.335.80.500.0880.0320.778a*20b18b21b26a1.130.0130.0650.107b*14.214.615.917.70.650.2100.0440.581pH45 min7.027.037.057.130.020.0780.0230.215pH24 h6.136.086.086.060.010.1920.0610.457Shear force (N)105a100ab100ab80b5.470.0420.1340.502Water loss rate (%)3330302910.4270.1610.573Thawing loss (%)332410.1910.8970.042Cooking loss (%)3839393720.4220.5400.131Note: Values are presented as means and pooled SEM. Means within the same row with different superscript letters (a, b) differ significantly (p < 0.05). CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. ANOVA represents the overall treatment effect, whereas linear and quadratic p-values represent orthogonal polynomial contrasts. Effects of dietary guanidinoacetic acid (GAA) supplementation at different concentrations on carcass traits and meat quality indices in Tibetan sheep. Note: Values are presented as means and pooled SEM. Means within the same row with different superscript letters (a, b) differ significantly (p < 0.05). CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. ANOVA represents the overall treatment effect, whereas linear and quadratic p-values represent orthogonal polynomial contrasts. Muscle antioxidant activity analysis As shown in Table 2, GSH-Px activity was significantly higher in the HG group than in the CON, LG and MG groups (p = 0.019). T-AOC was also significantly higher in the HG group than in the other groups (p = 0.002). In addition, GSH-Px, MDA and T-AOC showed significant linear responses to increasing dietary GAA levels (p = 0.030, p = 0.049 and p = 0.001, respectively), and T-AOC also showed a significant quadratic response (p = 0.003). Table 2Effects of dietary guanidinoacetic acid (GAA) supplementation at different concentrations on antioxidant properties of the longissimus dorsi muscle in Tibetan sheep.ItemsGroupsSEMp-valueCONLGMGHGANOVALinearQuadraticSOD17.816.418.119.00.400.1300.1410.132GSH-Px12.2b12.9b11.5b15.9a0.580.0190.0300.067MDA27.726.425.025.30.490.1920.0490.419CAT14.514.314.913.70.610.9260.7750.708T-AOC12.4b14.6b12.5b19.6a0.690.0020.0010.003Note: Values are presented as means and pooled SEM. Means within the same row with different superscript letters (a, b) differ significantly (p < 0.05). CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. ANOVA represents the overall treatment effect, whereas linear and quadratic p-values represent orthogonal polynomial contrasts. T-AOC (μmol/mL), total antioxidant capacity; CAT (U/L), catalase; SOD (U/mL), superoxide dismutase; MDA (nmol/mL), malondialdehyde; GSH-Px (U/mL), glutathione peroxidase. Effects of dietary guanidinoacetic acid (GAA) supplementation at different concentrations on antioxidant properties of the longissimus dorsi muscle in Tibetan sheep. Note: Values are presented as means and pooled SEM. Means within the same row with different superscript letters (a, b) differ significantly (p < 0.05). CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. ANOVA represents the overall treatment effect, whereas linear and quadratic p-values represent orthogonal polynomial contrasts. T-AOC (μmol/mL), total antioxidant capacity; CAT (U/L), catalase; SOD (U/mL), superoxide dismutase; MDA (nmol/mL), malondialdehyde; GSH-Px (U/mL), glutathione peroxidase. Muscle FA analysis As shown in Table 3, the concentrations of C8:0 and C12:0 were significantly higher in the HG group than in the CON, LG, and MG groups (p = 0.047 and p = 0.032, respectively). For MUFAs, C18:1n12 was significantly higher in the CON and HG groups than in the LG and MG groups (p = 0.047), whereas C20:1n9 was significantly higher in the MG and HG groups than in the CON and LG groups (p = 0.048). For PUFAs, C20:3n6 was significantly higher in the MG and HG groups than in the CON and LG groups (p = 0.047), while C20:5n3 was significantly higher in all GAA-supplemented groups than in the CON group (p = 0.021). In addition, C20:1n9, C18:3n3, C20:5n3, total PUFA, total UFA, and total n-3 PUFA increased linearly with increasing dietary GAA levels (p = 0.021, 0.015, 0.003, 0.041, 0.042, and 0.032, respectively). No significant quadratic responses were observed (p > 0.05). Table 3Effects of dietary guanidinoacetic acid (GAA) supplementation at different concentrations on fatty acid composition of the longissimus dorsi muscle in Tibetan sheep (mg/100 g fresh meat).ItemsGroupsSEMp-valueCONLGMGHGANOVALinearQuadraticSFAC8:065b64b71b87a8.130.0470.3760.661C10:017621122619215.120.7380.6860.329C11:0191925251.480.2400.0740.975C12:0656b741b856b1031a72.840.0320.0870.752C13:0514050383.260.4960.3120.900C14:01818171117701435140.000.8220.4560.724C15:043948849044823.990.8640.9110.422C16:012,54213,86714,42512,796570.190.6740.8150.261C17:063666276069154.620.9030.6450.712C18:010,85912,38912,66511,067595.140.6860.8790.256C20:02062392232139.290.4840.7550.465ΣSFA27,46630,43331,56027,99713790.7420.8450.309MUFAC14:1n513114216017917.740.8340.3840.923C15:1n591114151.390.3960.1140.787C16:1n710049011073125691.920.6390.3170.483C17:1n751351366073548.130.2820.0780.685C18:1n12132a109b90b154a18.280.0470.8030.292C18:1n7535962584.150.9310.7160.610C20:1n977b82b125a151a9.630.0480.0210.412ΣMUFA1919181621832548178.890.5350.2090.543PUFAC18:2n64896508259106031379.610.6970.2800.969C18:3n633135639042230.120.7840.3310.962C18:3n310,07711,04612,78714,784902.110.0810.0150.765C20:3n6731b794b902a926a69.110.0470.3380.903C20:4n613514219317717.630.6570.3170.764C20:2n657471278787255.880.3020.0740.807C20:3n3242538302.580.2080.1490.350C20:5n31452b2091a2344a3069a230.260.0210.0030.881C22:2n659173283391925.80.2720.0640.814C22:5n314023822819321.550.4120.4550.158C22:5n61618169119692135148.730.6420.2310.887C22:6n361175877470561.320.8310.6400.451ΣPUFA21,17923,66827,15630,2651593.540.1970.0410.916ΣUFA23,09825,48429,33932,8131716.970.1990.0420.863Σn-3 PUFA12,30414,16016,17118,7811072.580.1600.0320.844Σn-6 PUFA8875950810,98511,484666.070.5280.1670.962Note: Values are presented as means and pooled SEM. Means within the same row with different superscript letters (a, b) differ significantly (p < 0.05). CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. ANOVA represents the overall treatment effect, whereas linear and quadratic p-values represent orthogonal polynomial contrasts. Abbreviations: SFA, saturated fatty acids; MUFA, monounsaturated fatty acids; PUFA, polyunsaturated fatty acids; ΣSFA, sum of saturated fatty acids; ΣMUFA, sum of monounsaturated fatty acids; ΣPUFA, sum of polyunsaturated fatty acids; ΣUFA, sum of monounsaturated and polyunsaturated fatty acids; Σn-3 PUFA, sum of n-3 polyunsaturated fatty acids; Σn-6 PUFA, sum of n-6 polyunsaturated fatty acids. Effects of dietary guanidinoacetic acid (GAA) supplementation at different concentrations on fatty acid composition of the longissimus dorsi muscle in Tibetan sheep (mg/100 g fresh meat). Note: Values are presented as means and pooled SEM. Means within the same row with different superscript letters (a, b) differ significantly (p < 0.05). CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. ANOVA represents the overall treatment effect, whereas linear and quadratic p-values represent orthogonal polynomial contrasts. Abbreviations: SFA, saturated fatty acids; MUFA, monounsaturated fatty acids; PUFA, polyunsaturated fatty acids; ΣSFA, sum of saturated fatty acids; ΣMUFA, sum of monounsaturated fatty acids; ΣPUFA, sum of polyunsaturated fatty acids; ΣUFA, sum of monounsaturated and polyunsaturated fatty acids; Σn-3 PUFA, sum of n-3 polyunsaturated fatty acids; Σn-6 PUFA, sum of n-6 polyunsaturated fatty acids. Jejunal SCFA analysis As shown in Table 4, acetic acid concentration was significantly higher in the HG group than in the CON, LG and MG groups (p = 0.045). Propionic acid concentration was also significantly higher in the HG group than in the CON and MG groups, while the LG group showed an intermediate value (p = 0.028). No significant differences were observed in isobutyric acid, butyric acid, isovaleric acid or valeric acid among groups (p > 0.05). Table 4Effects of dietary guanidinoacetic acid (GAA) supplementation at different concentrations on short-chain fatty acids in the jejunum of Tibetan sheep (μg/mL).ItemsGroupsSEMp-valueCONLGMGHGANOVALinearQuadraticAcetic acid37b41b39b73a9.900.0450.2910.493Propionic acid1.9b2.8ab1.8b4.7a0.580.0280.1650.397Isobutyric acid0.470.440.480.650.050.4520.2140.328Butyric acid0.500.420.370.560.040.4680.7490.162Isovaleric acid0.170.160.150.240.020.2570.1800.165Valeric acid0.070.090.060.090.010.5870.8910.676Note: Values are presented as means and pooled SEM. Means within the same row with different superscript letters (a, b) differ significantly (p < 0.05). CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. ANOVA represents the overall treatment effect, whereas linear and quadratic p-values represent orthogonal polynomial contrasts. Effects of dietary guanidinoacetic acid (GAA) supplementation at different concentrations on short-chain fatty acids in the jejunum of Tibetan sheep (μg/mL). Note: Values are presented as means and pooled SEM. Means within the same row with different superscript letters (a, b) differ significantly (p < 0.05). CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. ANOVA represents the overall treatment effect, whereas linear and quadratic p-values represent orthogonal polynomial contrasts. Jejunal microbiota composition analysis Based on 16S rRNA gene sequencing, rank-abundance curves indicated comparable ASV richness and community evenness among the four dietary treatment groups (Fig. 1A). A total of 4760 amplicon sequence variants (ASVs) were detected across the four dietary treatment groups. Among these, 1036, 998, 892, and 620 ASVs were unique to the CON, LG, MG, and HG groups, respectively, whereas 395 ASVs were shared among all four groups (Fig. 1B). Alpha diversity indices did not differ significantly among the groups, including Chao1, Shannon, Pielou's evenness, and Simpson indices (p > 0.05) (Fig. 1C–F). Bray-Curtis PCoA and PERMANOVA showed no significant treatment-related difference in jejunal microbial community composition (pseudo-F = 0.967, R2 = 0.127, p = 0.513) (Fig. 1G and Table S11). Relative-abundance profiling identified 22 phyla, 106 families, and 215 genera. UPGMA clustering based on the 20 most abundant phyla and the 30 most abundant families and genera, together with Circos plots, summarised the dominant taxonomic patterns across treatments (Fig. 1H-M). (A) Rank–abundance curves. (B) Venn diagram. α-diversity indices: Chao1 index (C); Shannon index (D); Simpson index (E); Pielou index (F). (G) Principal coordinate analysis (PCoA) plot. The unweighted pair group method with arithmetic mean (UPGMA) clustering shows the microbial community composition of the four groups at the phylum level (top 20) (H), family level (top 30) (I), and genus level (top 30) (J). The Circos map shows the microbial distribution and percentages of the four groups at the phylum (K), family (L), and genus (M) levels. CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. LEfSe analysis (Kruskal–Wallis test, p < 0.05; LDA score > 2.5) identified exploratory treatment-associated taxonomic signatures among the four dietary groups (Fig. 2A–B). The HG group was characterised by higher LDA scores for Anaerovoracaceae, Eubacterium_nodatum_group, Family_XIII_AD3011_group, Syntrophococcus, Family_XIII_UCG-001, Bacillaceae, and Bacillus. DNF00809 was identified as an LG-associated signature, whereas Muribaculaceae was identified as an MG-associated signature. In contrast, Eubacterium_coprostanoligenes_group, Oscillospirales, Erysipelotrichaceae_UCG-009, and Alphaproteobacteria showed higher LDA scores in the CON group. In separate univariate comparisons of the relative abundances of LEfSe-selected genera, significant group differences were observed for Bacillus, Family_XIII_UCG-001, Family_XIII_AD3011_group, Eubacterium_nodatum_group, and DNF00809 (p < 0.05; Fig. 2C). All five genera showed their highest mean relative abundances in the HG group and were significantly more abundant in HG than in CON (p < 0.05). Notably, although DNF00809 was identified as an LG-associated signature by LEfSe, its mean relative abundance was highest in the HG group in the separate univariate comparison. These univariate comparisons were based on relative-abundance data and are presented as supportive descriptive analyses rather than confirmatory compositional differential-abundance tests. Tax4Fun analysis suggested a higher predicted representation of functions related to amino acid metabolism, pyrimidine metabolism, DNA repair and recombination, purine metabolism, tRNA biogenesis, aminoacyl-tRNA biosynthesis, and ribosome-associated processes in the HG group (Fig. 2D–E). These results represent functions inferred from 16S rRNA gene profiles and do not constitute direct measurements of microbial functional activity. (A) Dendrogram showing treatment-associated taxa with higher LDA scores in jejunal samples from the four treatment groups, with groups coloured as follows: CON (red), HG (green), LG (blue), and MG (purple). (B) Linear discriminant analysis (LDA) scores highlighting microbial taxa in the jejunum; markers with LDA scores >2.5 are shown. (C) Asterisks (*) indicate significant differences between groups (*p < 0.05). (D) Bar plot of the relative abundances of the top predicted microbial functions and (E) heatmap showing the clustering of predicted functional profiles among the four groups, as inferred by Tax4Fun analysis. CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Untargeted metabolomic analysis Based on 16S rRNA gene sequencing, GAA supplementation did not significantly alter the overall jejunal microbial community structure, as determined by PERMANOVA (p = 0.513). To characterise GAA-associated changes in muscle metabolism, untargeted metabolomics analysis was performed via UHPLC–Orbitrap high-resolution MS. Three-dimensional principal component analysis (Fig. S2A) showed separation between CON and GAA-treated (LG, MG and HG) groups in positive and negative ionisation modes. In the positive-ion mode, PC1 and PC2 explained 24.39% and 11.09% of the total variance, whereas in the negative-ion mode, PC1 and PC2 explained 21.11% and 13.00%, respectively. Two-dimensional partial least squares discriminant analysis (PLS-DA) further showed separation of the metabolic profiles among groups (R2Y > 0.8; Q2Y > 0.4) (Fig. S2B). Model performance was evaluated using sevenfold cross-validation. The high R2Y values indicated good model fit, whereas Q2Y values >0.4 indicated acceptable predictive performance. Permutation testing (200 iterations) confirmed that all Q2 regression lines had negative y-axis intercepts and that R2 values were consistently higher than Q2 values (Fig. S2C), indicating no evident model overfitting. These findings indicate that GAA supplementation was associated with changes in the muscle metabolic profile; however, the present data do not establish a causal microbiota-host mechanism. Metabolomic profiling of muscle samples from CON, LG, MG and HG groups was performed using LC-MS. A total of 4908 metabolites were identified across all samples. The identified metabolites were classified according to their chemical categories, and their distributions were summarised using pie charts. The major chemical classes included lipids and lipid-like molecules, organic acids and derivatives, organoheterocyclic compounds and nucleosides, nucleotides and analogues (Fig. 3A). Of the 4908 identified metabolites, 1636 showed distinct abundance trends across increasing dietary GAA concentrations. Mfuzz clustering grouped these 1636 metabolites into 10 clusters, of which Cluster 10 (263 metabolites), Cluster 6 (203 metabolites), and Cluster 9 (184 metabolites) were the three largest (Fig. 3B). These cluster-specific abundance patterns showed heterogeneous dose-associated metabolic responses to GAA supplementation. Volcano plots were used to visualise the distributions of differentially abundant metabolites in the LG vs CON, MG vs CON, and HG vs CON comparisons (Fig. 3C). Based on |log₂(fold change)| > 1 and p < 0.05, 113, 222, and 264 differentially abundant metabolites were identified in the LG vs CON, MG vs CON, and HG vs CON comparisons, respectively (Table S10). Specifically, the LG vs CON comparison included 33 increased and 80 decreased metabolites. The MG vs CON comparison included 60 increased and 162 decreased metabolites, whereas the HG vs CON comparison included 64 increased and 200 decreased metabolites. Thus, the 113, 222, and 264 differentially abundant metabolites represent comparison-specific subsets of the 4908 metabolites identified across the complete dataset. Fold-change values were log₂-transformed, and metabolites were ranked according to their log₂(fold-change) values. The top 20 increased and top 20 decreased metabolites in each comparison were visualised using stem plots (Fig. 3D). Ten phenotype-related metabolites were selected for further analysis. PC 26:3 was increased in both the MG vs CON and HG vs CON comparisons. GMP was increased in LG relative to CON. In the MG vs CON comparison, prostaglandin H₂ (PGH₂) was decreased, whereas PC 21:1 and PC 40:6 were increased. In the HG vs CON comparison, PC 37:5, PC 40:5, and PC 38:4 were decreased, whereas 1-acylglycerol-3-phosphate (LPA) 14:0, LPC 20:3-sn-1 was increased. (A) Pie chart showing the categorisation of all metabolites. (B) Trend analysis plot showing clusters of metabolites. (C) Volcano plot showing the distribution of differentially abundant metabolites for the three comparison groups. Red circles represent significantly upregulated metabolites, blue squares represent significantly downregulated metabolites, and grey circles represent non-significant metabolites. (D) Lollipop plots for the top 20 increased/decreased metabolites. In each plot, the colour of the matchsticks distinguishes between upregulated (blue) and downregulated (red) metabolites, and the length of the matchsticks directly reflects the magnitude of the log₂(fold change) value (i.e., longer sticks indicate greater fold differences). The size of the dots represents the VIP value. CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Weighted correlation network analysis, implemented using the WGCNA framework, was applied to identify modules of metabolites with similar abundance profiles. The scale-free topology fit index increased markedly between soft-thresholding powers 1 and 3 and changed only slightly thereafter. Meanwhile, mean connectivity decreased rapidly with increasing soft-thresholding power and subsequently stabilised (Fig. 4A). Considering both the scale-free topology fit and mean connectivity, a soft-thresholding power of 3 was selected for network construction. Hierarchical clustering and intermodule correlation analysis identified seven metabolite modules, each represented by a distinct colour (Fig. 4B–C and Table S12). Metabolites within the same module exhibited similar abundance patterns across the samples. Module eigenmetabolite profiles, defined as the first principal component of the metabolite abundance profiles within each module, illustrated module-level variation among samples (Fig. 4D–J). Module–trait correlation analysis identified significant associations between specific metabolite modules and jejunal or meat-quality traits (Fig. 4K–N). The yellow module was positively correlated with MDA and negatively correlated with T-AOC, C12:0 and propionate. The blue module was positively correlated with CAT and C20:1n9. The brown module was positively correlated with CAT and RMT. The red module was positively correlated with propionate and negatively correlated wieh SOD. The black module was positively correlated with L*. The green module was negatively correlated with cooking loss. The turquoise module was negatively correlated with C12:0. PC 37:5 was assigned to the yellow module. PC 40:5, PC 26:3, LPC 20:3-SN1, PC 21:1 and PC 40:6 were assigned to the turquoise module. PC 38:4 and GMP were assigned to the black module. LPA 14:0 and PGH₂ were assigned to the green and brown modules, respectively. Weighted correlation network analysis using the WGCNA framework. (A) Soft-threshold power selection, showing the relationship between soft-threshold power and average connectivity. (B) Hierarchical clustering tree of metabolites. (C) Module membership plot showing a total of seven modules. Distribution of individual samples across modules: (D) black module, (E) turquise module, (F) yellow module, (G) blue module, (H) brown module, (I) green module, and (J) red module. (K) Heatmap of correlations between modules and muscle antioxidant activity. (L) Heatmap of correlations between modules and carcass traits and meat quality parameters. (M) Heatmap of correlations between modules and short-chain fatty acids in the jejunum. (N) Heatmap of correlations between modules and fatty acids in muscle. These heatmaps illustrate the correlations between metabolite modules and target phenotypic traits. CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Through integrated analysis of differentially abundant metabolites detected in positive and negative ionisation modes, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed, and the top 20 enriched pathways were summarised for each comparison (Fig. S3A–C). Pathway analysis showed that purine metabolism and taste transduction were among the top-ranked pathways in the LG vs CON comparison. In the MG vs CON and HG vs CON comparisons, glycerophospholipid metabolism and arachidonic acid metabolism were among the top-ranked pathways. Based on the integrated analysis of the log2(fold-change) lollipop plots and metabolite modules identified using the WGCNA framework, ten candidate metabolites associated with phenotypic traits were identified: GMP, PGH2, PC(21:1), PC(40:6), PC(26:3), PC(37:5), PC(40:5), PC(38:4), LPA(14:0) and LPC(20:3-SN1). These metabolites showed associations with muscle FA composition, antioxidant indices, meat-quality traits, and jejunal SCFA concentrations. These candidate metabolites were associated with GAA-related phenotypic variation. Glycerophospholipid metabolism was among the enriched pathways in both the MG vs CON and HG vs CON comparisons and included several of the candidate lipid metabolites. Therefore, this pathway is interpreted as a candidate pathway potentially associated with the observed metabolic responses. Correlation analysis of jejunal microbiota, muscle metabolites and meat quality To explore the hypothesis that GAA-associated differences in jejunal microbial profiles may be related to muscle metabolic profiles and meat-quality traits, associations among these variables were examined. Thus, the relationships among different microorganisms, SCFAs, metabolites, eating-quality traits, antioxidant capacity and FA profiles in Tibetan sheep samples were investigated. Correlation heatmaps display Spearman's rank correlations, whereas Mantel tests were used to evaluate associations between the selected microorganisms or metabolites and phenotypic traits. All Mantel P values reported in this section are presented solely to describe exploratory association patterns among the selected variables. Correlation analysis between microbiota and phenotypic traits (Fig. 5A–C) showed positive associations of T-AOC with Eubacterium_nodatum_group and Anaerovoracaceae (Mantel's P < 0.001), as well as with Bacillaceae, Bacillus, and Family_XIII_AD3011_group (Mantel's P < 0.05). Bacillus and Bacillaceae were positively associated with FA C8:0 (Mantel's P < 0.001), GSH-Px, and a* values (Mantel's P < 0.01). C8:0 was positively associated with Family_XIII_AD3011_group and Eubacterium_nodatum_group (Mantel's P < 0.05). Anaerovoracaceae was positively associated with propionate, C8:0, and L* (Mantel's P < 0.01). Correlation analysis between metabolites and phenotypic traits (Fig. 5D–F) showed that PC 37:5 was positively associated with T-AOC (Mantel's P < 0.001), MDA and propionate (Mantel's P < 0.05). PC 40:5 was positively associated with b* (Mantel's P < 0.001), GSH-Px, and a* (Mantel's P < 0.01). Furthermore, FA C12:0 was positively associated with LPC 20:3-SN1, PC 40:6, and PC 21:1 (Mantel's P < 0.05). The Sankey diagram illustrating correlations between jejunal microbiota and muscle metabolites (Fig. 5G) was based on Spearman's rank correlation and showed positive correlations of Eubacterium_nodatum_group, Bacillus,and Family_XIII_AD3011_group with PC 37:5, LPA 14:0, and PC 38:4, respectively. Spearman's rank correlation and Mantel test analyses among jejunal microbiota, muscle metabolites, and phenotypic traits. (A–C) Associations between differential jejunal microorganisms and (A) Longissimus dorsi muscle carcass traits and meat-quality indices, (B) Longissimus dorsi muscle antioxidant indices, and (C) Longissimus dorsi muscle fatty acids and jejunal short-chain fatty acids. (D–F) Associations between differential muscle metabolites and (D) Longissimus dorsi muscle carcass traits and meat-quality indices, (E) Longissimus dorsi muscle antioxidant indices, and (F) Longissimus dorsi muscle fatty acids and jejunal short-chain fatty acids. In panels A-F, the heatmaps show Spearman's rank correlations among the phenotypic variables, whereas the connecting lines indicate Mantel-test associations between the selected microorganisms or metabolites and the corresponding phenotypic variables. Line colours indicate Mantel P values: red, P ≤ 0.001; orange, P ≤ 0.01; green, P ≤ 0.05; grey, P > 0.05. (G) Sankey diagram showing Spearman's rank correlations between differential metabolites (left) and differential microorganisms (right); red lines indicate positive correlations and blue lines indicate negative correlations. Discussion By integrating the jejunal microbiome, muscle metabolome, and phenotypic data, we constructed an exploratory association model summarising the potential relationships among GAA supplementation, jejunal microbial signatures, muscle metabolic profiles, antioxidant status, and meat-quality traits in Tibetan sheep (Fig. 6). Compared with the CON group, supplementation with 0.12% GAA was associated with treatment-related taxonomic signatures involving the Eubacterium nodatum group, Family XIII AD3011 group, and Bacillus. These microbial signatures showed statistical associations with selected muscle metabolites, including PC 37:5, PC 38:4, and LPA 14:0. GAA supplementation was also associated with differences in the measured antioxidant indices and meat-quality traits, including the a* value, shear force, and rib meat thickness. Glycerophospholipid metabolism is therefore presented as a candidate pathway associated with the observed muscle metabolic and phenotypic patterns rather than as a demonstrated central regulatory hub. Exploratory integrative model of the associations among GAA supplementation, jejunal microbial signatures, muscle metabolic profiles, antioxidant indices, and meat-quality traits in Tibetan sheep. Left: Longissimus dorsi muscle metabolome; right: jejunal microbiome. Red indicates metabolites/microbes significantly altered by 0.12% GAA vs. CON (p < 0.05). Arrows summarise literature-based putative relationships and do not establish causal direction. CON, basal diet (control group); LG, basal diet supplemented with 0.08% GAA; MG, basal diet supplemented with 0.10% GAA; HG, basal diet supplemented with 0.12% GAA. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Post-mortem pH dynamics and meat quality Meat pH serves as a critical quality parameter that significantly influences other meat quality attributes (Knight et al., 2019). Terlouw et al. emphasised that muscle pH and temperature declined relatively rapidly during the early post-mortem period (Terlouw et al., 2021) because post-mortem changes encompass two fundamental phases: rigour mortis and meat maturation (Lebret & Čandek-Potokar, 2022). Although no significant group differences were observed in either pH45 min or pH24 h, the HG group exhibited a numerically higher pH45min than the other groups. A muscle pH > 7.0 at 45 min post-mortem is uncommon and has previously been associated with severe pre-slaughter stress and altered post-mortem glycolysis (Nelis et al., 2022). However, pre-slaughter stress cannot be inferred from the present results because stress biomarkers and muscle glycogen concentrations were not measured. At 24 h post-mortem, the HG group showed a numerically lower pH than the other groups. Despite the elevated initial pH, the pH declined to approximately 6.0 after 24 h. This value was below the pH threshold commonly associated with dark, firm, and dry meat, which generally remains above 6.4 at 24 h post-mortem (Ijaz et al., 2020). Therefore, the ultimate pH values did not indicate a typical DFD condition. Several factors may plausibly contribute to this distinct post-mortem pH pattern: Because stress indicators, glycogen and lactate were not measured, the physiological basis of this pH pattern remains uncertain and requires targeted validation (Miranda-de la Lama et al., 2012). However, stress indicators, muscle glycogen concentrations, and altitude-related physiological traits were not measured, these explanations remain hypothetical, and the present data do not establish that the observed pH pattern directly caused the meat-quality outcomes. Overall, further studies incorporating stress biomarkers, muscle glycogen and lactate concentrations, and post-mortem glycolytic measurements are needed to clarify the physiological basis of this pH pattern. Sensory quality and tenderness Appearance attributes (meat colour) and intrinsic quality parameters (cooking loss, thawing loss, water loss, SF and pH) are commonly used to evaluate meat quality (An et al., 2021). Meat colour considerably influences consumer purchasing behaviour, serving as the primary visual indicator of quality (Eom et al., 2024). Meat colour is typically characterised using CIE L*a*b* coordinates (Joseph et al., 2012). Higher a* values indicate greater redness because fresh meat contains high concentrations of oxymyoglobin, displaying a visually appealing bright red appearance and stimulating consumer appetite (Xu et al., 2024). Significant differences in a* values were observed among the four treatment groups, with HG exhibiting notably elevated a* values. This difference may be related to post-mortem biochemical and structural changes affecting myoglobin status and muscle colour stability (Holman & Hopkins, 2021). Correlation analysis showed positive associations between a* values and Bacillus, Bacillaceae, and PC 40:5. However, these correlations do not establish that Bacillus-derived phospholipase activity altered PC 40:5 abundance or directly influenced meat colour. The observed associations may reflect parallel responses to GAA supplementation and require mechanistic validation in the future (Lee et al., 2009). Low SF values, a critical indicator of sheep tenderness, indicate superior tenderness and palatability (Meenongyai et al., 2025). Sheep cuts with thick rib sections demonstrate enhanced tenderness and juiciness owing to uniform intramuscular fat (IMF) distribution (Ncube et al., 2025). The present results suggest that dietary GAA supplementation was associated with improvements in selected meat-quality traits, including tenderness-related indices, with the HG group showing the most favourable overall profile. GAA acts as a creatine precursor, enhancing muscle energy metabolism and meat quality (DeGroot et al., 2018). Previous studies have suggested that GAA supplementation delays post-mortem pH decline, reducing protein denaturation and drip loss while enhancing meat texture and tenderness (Kim et al., 2014), which is broadly consistent with the present observations. Antioxidant capacity enhancement The health of animals is closely associated with their antioxidant capacity. T-AOC, GSH-Px, CAT, SOD and MDA are key biomarkers for assessing antioxidant status (Gan et al., 2025). Improved antioxidant status may contribute to the preservation of meat quality by limiting oxidative damage. T-AOC is a comprehensive index reflecting overall enzymatic and non-enzymatic antioxidant defence, while GSH-Px helps limit lipid peroxidation-induced cellular damage by regulating free radical generation and scavenging (Zhao et al., 2025). GAA supplementation has been associated with changes in the antioxidant status of animals. In Jinjiang bulls, GAA enhanced SOD and GSH-Px activities (Li et al., 2021). In broilers, GAA increased T-AOC while reducing reactive oxygen species (ROS) and MDA levels (Zhao et al., 2021). In swine, GAA increased muscle GSH-Px and SOD activities while reducing MDA content (Wang et al., 2012). In the present study, significant differences were observed in antioxidant parameters among Tibetan sheep receiving different GAA concentrations. As the GAA concentration increased GSH-Px activity and T-AOC increased significantly, while MDA showed a significant linear decrease, with the 0.12% GAA supplementation group showing the highest GSH-Px and T-AOC value. GAA is synthesised from arginine and glycine and acts as a direct precursor of creatine, because creatine has been shown to scavenge superoxide anions, GAA-creatine metabolism may provide a plausible biochemical context for the observed antioxidant responses (Lawler et al., 2002). Correlation analysis further showed positive associations of T-AOC with the Eubacterium nodatum group, Anaerovoracaceae, and PC 37:5. Previous studies have linked microbial SCFA-related activity with the maintenance of intestinal barrier integrity and the regulation of systemic inflammatory and oxidative responses (Xiao et al., 2020). Creatine-related metabolism has also been associated with cellular energy homeostasis and mitochondrial redox regulation, providing a possible metabolic context for the observed association between Anaerovoracaceae and antioxidant indices (Beaumont et al., 2020). Similarly, PC 37:5 may be related to phospholipid-derived signalling processes potentially involving redox-sensitive pathways, including Nrf2/ARE-associated antioxidant regulation (Che et al., 2018). Collectively, these literature-based mechanisms provide plausible biological explanations for the observed correlations; however, the present data do not demonstrate direct mechanistic roles for the individual taxa or metabolites, and these hypotheses require further experimental validation. FA composition and nutritional value With the growing health consciousness among consumers, the demand for meat products with favourable nutritional profiles is increasing. Muscle FAs play crucial roles in determining the nutritional value, flavour and oxidative stability of meat (Cui et al., 2022). FA composition has gained increasing attention (Jeong et al., 2024). High dietary intake of certain SFAs has been associated with an increased risk of coronary heart disease by raising total cholesterol and low-density lipoprotein levels (Zong et al., 2016). Together with the observed changes in individual FAs, this finding suggests that GAA supplementation may contribute to a more favourable nutritional FA profile in Tibetan sheep meat. However, certain medium-chain SFAs play essential roles in energy metabolism and exhibit antimicrobial, antioxidant and antiviral functions (Hinton & Ingram, 2006). In the present study, C8:0 (caprylic acid) and C12:0 (lauric acid) contents in Tibetan sheep muscle increased significantly with GAA supplementation. These changes may be related to the effects of GAA on creatine synthesis, energy metabolism, and antioxidant status. C8:0 showed significant positive correlations with Bacillus and Bacillaceae. Previous studies have suggested that Bacillus supplementation may influence meat FA profiles, nutritional quality, and flavour (Wood et al., 2004). Tang et al. reported that Bacillus supplementation increased meat pH₂₄ h and may be associated with the relationship between C8:0 and post-mortem muscle pH (Tang et al., 2021). Herein, GAA supplementation maintained pH24h values of 6.06–6.08 within the observed range. Regarding MUFAs, different GAA concentrations significantly affected the contents of C18:1n12 and C20:1n9. As a crucial component of cell membrane phospholipids, C18:1n12 accumulation might be associated with IMF deposition, and may therefore be related to meat tenderness and juiciness (Song et al., 2022). Although the metabolic functions of C18:1n12 remain unclear, its accumulation in ruminant muscle might reflect altered gastrointestinal microbial biohydrogenation. The observed change in C20:1n9 may reflect alterations in long-chain FA elongation or β-oxidation following GAA supplementation. Elevated C20:1n9 content has previously been associated with improved membrane stability, reduced cooking loss and greater water-holding capacity (Cardoso et al., 2019). For PUFAs, the changes in C20:3n6 and C20:5n3 (EPA) were particularly notable. As an intermediate metabolite in n-6 series PUFA metabolism, the change in C20:3n6 content may indicate an effect of GAA supplementation on n-6 PUFA metabolism, potentially involving Δ-5 desaturase-related processes (Chen et al., 2024). Arachidonic acid (C20:4n6) is involved in growth and muscle development, immune function and nervous system health (DiNicolantonio & O'Keefe, 2018). Meat rich in arachidonic acid may also have favourable flavour and texture (Takahashi, 2018). EPA (C20:5n3) is an anti-inflammatory n-3 PUFA. The observed change in EPA content may be related to altered n-3 PUFA metabolism following GAA supplementation, potentially involving changes in mitochondrial and lipid metabolic processes (Wu et al., 2025). This finding aligns with a previous study reporting that GAA supplementation in calves significantly enhanced muscle EPA deposition, which may improve the nutritional value of meat (Giraldi et al., 2024). Feeding n-3 PUFA-rich diets during fattening increases meat n-3 PUFA content and quality and reduces production costs in rabbits (Cavani et al., 2009). The HG group had the highest mean total n-3 PUFA content. Collectively, these changes in selected medium-chain SFAs, MUFAs, and n-3 PUFAs suggest that dietary GAA supplementation may improve the nutritional characteristics of the FA profile in Tibetan sheep muscle. Gut microbiota and SCFAs The gut microbiome plays an essential role in host health by modulating nutrient metabolism, immune responses and disease susceptibility (Wang et al., 2020). GAA supplementation has been reported to influence gut microbial composition (Dong et al., 2024). Consistent with these findings, alpha-diversity indices did not differ significantly among groups, although some values were numerically higher in the HG group, suggesting that jejunal microbial richness and evenness are positively associated with dietary GAA supplementation levels. SCFAs, including acetate, propionate and butyrate, are generated through gastrointestinal microbial fermentation of carbohydrates and proteins. These SCFAs can support intestinal epithelial cell proliferation, enhancing absorptive capacity and feed nutrient utilisation efficiency by facilitating nutrient transport and absorption (Dai et al., 2022; McNeil, 1984). SCFA concentrations vary in response to dietary composition (Tilocca et al., 2017). In the present study, acetate and propionate concentrations differed significantly among treatment groups, with HG exhibiting the highest levels. The higher acetate and propionate concentrations in the HG group may reflect changes in microbial fermentation activity and substrate utilisation associated with GAA supplementation. GAA supplementation has also been associated with changes in acetate and propionate production that may contribute to intestinal homeostasis (Alaa et al., 2024). Overall, the observed SCFA changes provide a possible functional context for the effects of GAA supplementation, although the underlying microbial metabolic processes require further investigation. Gut microbiota structure and function Several GAA-associated jejunal genera were identified, including Bacillus, Family_XIII_UCG-001, Family_XIII_AD3011_group and the [Eubacterium] nodatum group, all within the phylum Firmicutes. Firmicutes constitute a major component of the intestinal microbiota, and members of this phylum are involved in nutrient metabolism, mucosal homeostasis, and resistance to pathogen colonisation (Pushpanathan et al., 2019). In the present study, the relative abundances of Bacillus, Family_XIII_UCG-001, Family_XIII_AD3011_group, and the Eubacterium nodatum group differed significantly among the dietary groups, with the highest mean values observed in the HG group. Dietary Bacillus subtilis supplementation has been reported to improve meat quality and flavour by altering amino acid and FA profiles while improving carcass characteristics in broiler chickens (Tang et al., 2021). Breast muscle pH after 24 h post-mortem was also increased following Bacillus subtilis supplementation in that study, providing a possible context for the association between Bacillus and meat-quality traits observed here. Family_XIII_UCG-001 and Family_XIII_AD3011_group belong to the Family_XIII lineage, members of which have been associated with cellulose and hemicellulose fermentation and the production of SCFAs, particularly acetate and butyrate. These microbial activities may be relevant to host energy metabolism, lipid deposition, and muscle-quality characteristics (Meehan & Beiko, 2014). Family_XIII_UCG-001 was reported to be more abundant in the caecal microbiota of male chickens (Yang et al., 2024). The Eubacterium nodatum group has previously been associated with intestinal barrier function, microbial homeostasis, and oxidative status (Zhou, Liu, Qin, et al., 2025). GAA supplementation has also been reported to increase total SCFA content and microbial abundance in yak rumen (Liu et al., 2019). Tax4Fun analysis suggested differences in predicted jejunal microbial functions among the dietary groups, with the HG group showing higher predicted representation of pathways related to protein translation and nucleotide metabolism, including aminoacyl-tRNA biosynthesis, purine metabolism, and pyrimidine metabolism. Purine metabolism has previously been associated with meat flavour-related characteristics in Tibetan sheep (Zang et al., 2022). Overall, 0.12% GAA was associated with the most pronounced microbial and SCFA differences; their functional implications require validation. Metabolomics and glycerophospholipid metabolism Metabolomics has provided an important analytical approach in nutrition and food science (Gibbons et al., 2015). Permutation test results indicated that the PLS-DA models showed acceptable performance without evident overfitting and supported the discrimination of metabolic profiles among the treatment groups. The WGCNA framework can be applied to metabolite-abundance data to identify metabolite modules associated with complex traits (Li et al., 2025). We identified seven metabolite modules. Based on WGCNA-derived metabolite modules and KEGG enrichment pathways, we identified candidate metabolites associated with muscle FA composition, antioxidant indices, meat-quality traits, and jejunal SCFA concentrations. Among these, PC 21:1, PC 40:6, PC 26:3, PC 37:5, PC 40:5 and PC 38:4 were phosphatidylcholine (PC) species and were prominent among the candidate metabolites associated with meat-quality traits. PC is the most abundant phospholipid in mammalian tissues, constituting a substantial proportion of cellular membrane lipids and contributing to membrane structure and lipid-related meat-quality traits (Li et al., 2025). Alterations in PC abundance have been associated with IMF content, with specific phospholipid molecules being associated with lipid-droplet formation and IMF deposition (Ben M'barek et al., 2017). PC undergoes hydrolysis during meat cooking, resulting in the formation of volatile flavour compounds such as aldehydes and ketones, which may contribute to meat flavour (Li et al., 2020). In the present study, several phospholipid species showed GAA-associated abundance changes, and glycerophospholipid metabolism was among the enriched pathways. Glycerophospholipids are polar lipids ubiquitously present in all tissues, serving as essential components of cellular membranes and participating in physiological and biochemical functions, including cell proliferation, differentiation, secretion and volatile compound production (Cheng et al., 2024). Previous research has linked glycerophospholipid metabolism with DNA methylation-related changes in PLA2G4F expression, lipid-metabolic enzymes, and PPARγ signalling, which may influence glycerophospholipid synthesis and muscle lipid deposition (Zhu et al., 2024). This possible relationship is supported by a previous study in beef cattle reporting changes in hepatic PC metabolites and lipid metabolism-related genes, including LPCAT2, following GAA supplementation (Yi et al., 2025). GAA supplementation has also been reported to affect lipid deposition together with changes in lipoprotein lipase expression in ducks (Wu et al., 2024). Together, these studies suggest that GAA may influence phospholipid-related metabolic processes across different animal species. Glycerophospholipid metabolism has also been associated with meat-flavour formation through the generation of volatile compounds derived from unsaturated FAs (Xu et al., 2025). Glycerophospholipids are also known as glycerophosphatidyl lipids (Dennis et al., 2011). Their metabolism generates monoacylated phospholipids and phosphatidic acid (PA), an important intermediate in glycerophospholipid synthesis. Previous studies have reported that PA can interact with mTORC1, a signalling pathway associated with adipogenesis, FA synthesis, IMF deposition, and related meat-quality traits (Turi et al., 2023). Taken together, the metabolite changes and pathway-enrichment results suggest that glycerophospholipid metabolism may be involved in the response to dietary GAA supplementation. However, the activities or expression of PLA2G4F, LPCAT2, lipoprotein lipase, PAP, CPT, PPARγ, mTORC1, and related signalling components were not directly measured; therefore, these literature-supported pathways are presented as possible explanations. Key metabolites and meat quality regulation PC 37:5 was assigned to the yellow module and showed changes in abundance following GAA supplementation. Previous studies suggest that methyl-donor availability, including S-adenosylmethionine, may influence PC synthesis, providing a possible explanation for the observed change in PC 37:5 (McBreairty et al., 2015). PC 40:5, PC 26:3, LPC 20:3-SN1, PC 21:1 and PC 40:6 were assigned to the turquoise module and were negatively correlated with C12:0. FA C12:0 can serve as a substrate for PC and LPC synthesis (Ghaffari et al., 2024). One possible explanation is that GAA-related changes in the creatine-phosphocreatine energy system may influence ATP utilisation and lipid metabolism, thereby contributing to the inverse association between C12:0 and long-chain phospholipids such as PC 40:6 and PC 40:5. However, this proposed pathway requires validation in future studies through direct measurement of muscle creatine and PCr levels, CK activity, key β-oxidation enzyme activity and dynamic changes in specific phospholipids (Adriano et al., 2022). Nevertheless, earlier studies have primarily attributed the effects of GAA to enhanced creatine-PCr energy metabolism (Zhang. et al., 2022). The associations of long-chain polyunsaturated PC species, including PC 40:6, PC 40:5, and PC 38:4, with jejunal SCFAs suggest a possible relationship between GAA supplementation, microbial fermentation products, and muscle phospholipid metabolism in ruminants. LPA 14:0 was assigned to the green module and was negatively correlated with cooking loss. LPA 14:0 abundance tended to increase with GAA supplementation, and its negative correlation with cooking loss suggests that it may be associated with muscle water-holding properties. From a molecular mechanistic perspective, LPA 14:0, a specific ligand for G protein-coupled receptors (LPAR1–6) has been linked to Rho/ROCK signalling in other systems; this mechanism was not measured here (Laface et al., 2024). In the black module, GMP abundance showed a significant positive correlation with L* values. GMP possesses antimicrobial properties that may limit the growth of spoilage bacteria on meat surfaces, delay meat colour darkening caused by microbial metabolism and maintain meat colour stability (Gu et al., 2025). Further pathway enrichment analysis associated GMP with the cGMP-PKG signalling pathway. The cGMP-PKG signalling pathway participates in energy metabolism pathways in Chaka sheep, increasing energy efficiency and meat quality (Guo et al., 2022). In the brown module, PGH₂ abundance showed a significant positive correlation with CAT activity and RMT. Previous studies suggest that PGH₂-derived lipid mediators may be associated with antioxidant regulation through pathways involving peroxisome proliferator-activated receptor γ (PPARγ) (Hao et al., 2024). PPARγ is a major regulator of adipogenesis (Chen et al., 2023). PGH₂-derived mediators have been linked to PPARγ-related adipogenic signalling in other models; this pathway was not measured here (García-Alonso et al., 2013). PGH2 was mapped to the arachidonic acid metabolism pathway. The arachidonic acid metabolism network participates in the transduction of multiple signalling molecules during dry-ageing processes of meat, effectively enhancing textural characteristics and flavour compound formation in Yanbian cattle muscle (Sun et al., 2024). Overall, dietary GAA supplementation was associated with changes in metabolites related to energy metabolism, phospholipid metabolism, arachidonic acid metabolism, antioxidant status, and meat-quality traits. These findings provide possible explanations for the favourable meat-quality responses observed with GAA supplementation, particularly at 0.12%. Microbiome–metabolite association network In ruminants, gut microbiota might interact with the host and be associated with muscle metabolite profiles (Wang et al., 2021). In the present study, the 0.12% GAA group showed the most pronounced changes in selected jejunal microbial taxa and muscle metabolites. Through integrated correlation analysis of the microbiome and metabolome, we identified significant associations between specific jejunal microorganisms and selected muscle metabolites. Significant positive correlations were observed between (i) Eubacterium_nodatum_group and PC 37:5, (ii) Family_XIII_AD3011_group and PC 38:4 and (iii) Bacillus and LPA 14:0. PC 37:5 and PC 38:4, important membrane phospholipids within the PC family, are components of the glycerophospholipid metabolism pathway. In this pathway, glycerol-3-phosphate is sequentially acylated to form LPA and phosphatidic acid (PA), which provide precursors for the synthesis of downstream glycerophospholipids (Wang et al., 2021). PA can subsequently be converted to diacylglycerol (DAG), after which phosphocholine transfer contributes to PC synthesis through the Kennedy pathway (Takeuchi & Reue, 2009). Previous studies have shown that microbially derived SCFAs, including butyrate, can act as energy substrates and signalling molecules and may influence host lipid metabolism through G protein-coupled receptor-related pathways (Kalkan et al., 2025; Tian et al., 2021). This provides a possible biological context for the observed association between the Eubacterium nodatum group and PC 37:5. Within glycerophospholipid metabolism, PA may be dephosphorylated by phosphatidic acid phosphatase to form DAG (Liu, Lu, et al., 2025). DAG serves as a precursor for synthesising glycerophospholipids such as PC and PE. Microbial metabolites have previously been associated with lipid catabolism and DAG-related metabolic processes (Sando & Chertihin, 1996). Changes in DAG metabolism may therefore be relevant to glycerophospholipid synthesis. The association between Family_XIII_AD3011_group and PC 38:4 may likewise reflect a relationship between microbial fermentation-related activity and host lipid metabolism (Zhou, Liu, Zhao, et al., 2025). Taken together, these associations may reflect links between microbial fermentation-related activity and muscle glycerophospholipid metabolism. However, the causal direction and the involvement of specific enzymes, including PAP and CPT, remain to be established through further investigation. Limitations and future directions Several limitations of this study should be acknowledged. Although correlation analyses suggest potential associations among gut microbiota, metabolites, and phenotypic traits, our study does not establish causality. Additionally, we did not directly measure classical biochemical markers of GAA metabolism, including GAA, creatine, PCr, CK activity, methylation status, or energy metabolism markers. The relatively limited sample size may also affect the generalisability of our findings. To address these limitations and validate our findings, future studies should conduct large-scale validation across diverse Tibetan sheep populations and production conditions, assess long-term safety and cost-benefit, and integrate classical metabolic markers with gut microbiota and metabolomic analyses. Moreover, mechanistic investigations-such as targeted metabolite verification, microbial isolation and intervention experiments, faecal microbiota transplantation, and isotope tracing or metabolic flux analysis-are required to test causality within the proposed gut-muscle axis and elucidate the specific biochemical pathways through which microbial activities influence muscle phospholipid composition. Conclusion This study investigated the associations among dietary GAA supplementation, jejunal microbial profiles, muscle metabolism, and meat-quality traits in Tibetan sheep. Integrated multi-omics analysis showed that the 0.12% GAA group showed the most favourable responses in several carcass and meat-quality traits, as evidenced by increased meat-colour a* values and rib meat thickness, reduced shear force, improved antioxidant indices, and changes in the muscle FA profile. These responses were accompanied by changes in selected jejunal microbial taxa and muscle metabolites. Glycerophospholipid metabolism was among the pathways associated with the observed metabolic and meat-quality differences. Exploratory association analysis identified relationships between the Eubacterium nodatum group, Family_XIII_AD3011_group and Bacillus and the metabolites PC 37:5, PC 38:4 and LPA 14:0. Collectively, these findings suggest possible links among GAA supplementation, jejunal microbial signatures, muscle phospholipid metabolism, and meat-quality traits. However, these associations do not establish a causal gut-muscle mechanism. From a practical perspective, dietary supplementation with 0.12% GAA may be considered a nutritional strategy for improving selected carcass and meat-quality traits in Tibetan sheep, although further validation in larger populations and under different production conditions is required. CRediT authorship contribution statement Rengeerli Sa: Writing – original draft, Formal analysis, Data curation, Conceptualization. Fengshuo Zhang: Methodology, Data curation. Xianhua Zhang: Investigation. Yu Zhang: Visualization, Validation. Shengzhen Hou: Project administration, Funding acquisition. Linsheng Gui: Writing – review & editing, Project administration, Funding acquisition, Conceptualization. Animal ethics statement Ethics approval and consent to participate Our study was carried out in compliance with the ARRIVE guidelines (AVMA Guidelines for the Euthanasia of Animals: 2020 Edition). All experimental animal procedures complied with the laboratory animal management and welfare regulations approved by the Ethics Committee of Qinghai University, Xining, Qinghai, China. All animal experiments were approved by the Animal Care Committee of Qinghai University (approval number, QUA-2020-0710). Funding This work was supported by the Shuaicai Scientists Project of Qinghai Province [grant number 2022NK169-1]. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.fochms.2026.100440. Supplementary data Supplementary material 1 Supplementary material 2 Data availability The 16S rRNA gene sequencing data are available in the NCBI Sequence Read Archive under BioProject accession PRJNA1273294. The metabolomics data are available in NCBI SRA under accession PRJCA041283. References Associated Data Supplementary Materials Supplementary material 1 Supplementary material 2 Data Availability Statement The 16S rRNA gene sequencing data are available in the NCBI Sequence Read Archive under BioProject accession PRJNA1273294. The metabolomics data are available in NCBI SRA under accession PRJCA041283.