This research was performed under the approval of the Institutional Animal Care and Use Committee of Shanxi Agricultural University (sxnd202028). A total of 18 Hu ewes (18 months, 50 kg ± 1 kg) were randomly selected, equally assigned into 3 groups (n = 6) and fed corresponding diet containing different concentration of taurine (purity ≥ 99%, Shanghai Jin Sui Biotechnology Co., Ltd.): control group (0% taurine), 0.1% TAU group (0.1% taurine) and 0.2% TAU group (0.2% taurine). The basal diets were formulated to be isonitrogenous and isoenergetic, taurine was administered orally by gavage, and ewes had unrestricted access to both feed and water. Estrus synchronization was performed to ensure all ewes were inseminated simultaneously using semen from Dorper rams.

Feed nutritional content was evaluated following standard procedures (AOAC, 2023), and the basal diet’s complete nutrient profile is presented in Table 1. Samples were freeze-dried for 72 h and ground to pass a 1 μm sieve. Analyses included dry matter (DM) by drying at 105 °C for 2 h (method 930.15); crude protein (CP) via the Kjeldahl method (method 980.10), calculated as N × 6.25; ether extract (EE) by Soxhlet extraction (method 920.39); and ash by combustion at 550 °C for 2 h (method 942.05). Neutral detergent fiber (NDF) and acid detergent fiber (ADF) were determined following Van Soest et al. (1991) using heat-stable α-amylase and sodium sulfite.

The following sampling and analyses were conducted with six biological replicates per group, including ewe and lamb serum taurine concentration, lamb body weight, lamb intestinal gene expression (qPCR), and lamb serum immune and antioxidant parameters. Blood was sampled from the ewe’s jugular vein during early, mid, and late gestation, as well as from offspring for serum collection. Aliquots of serum were prepared and frozen at −80 °C for future analyses. To minimize potential sex-related variability, only male lambs were selected for downstream analyses. Hu sheep is a prolific breed with a high incidence of twin and multiple births, which ensured the availability of sufficient male lambs per treatment group without introducing selection bias. Male lambs were weighed, euthanized, and sampled at 15 days after birth. Prior to euthanasia, lambs were deeply anesthetized by intravenous injection of sodium pentobarbital (30 mg/kg body weight) via the jugular vein to ensure complete loss of consciousness and absence of pain. Once deep anesthesia was confirmed by the absence of corneal reflex and response to toe pinch, euthanasia was performed by exsanguination. After euthanasia, intestinal tissues were collected from multiple segments for histological evaluation, including the duodenum, jejunum, ileum and colon. For microbiome profiling and gene expression analyses, samples were intentionally collected from the jejunum only. This segment was selected because it represents a functionally central region for nutrient absorption and early-life intestinal barrier and immune regulation, and is directly involved in taurine uptake via specific intestinal transport systems, while maintaining analytical focus and statistical robustness (Anderson et al., 2009). Jejunum segments (approximately 3–4 mm) were fixed in 4% paraformaldehyde solution (Biosharp, BL539A, Hefei, China) for paraffin embedding and histological analysis. Additional jejunal tissues and digesta were immediately stored at −80 °C for subsequent molecular analyses.

For histomorphometric analysis, jejunal tissues were collected from six lambs per group. After fixation and paraffin embedding, 2–3 tissue sections per animal were prepared and stained with hematoxylin and eosin (HE). Within each section, at least 10 complete, well-oriented villus-crypt units were randomly selected and measured. The villus height and crypt depth for each animal were calculated as the mean of all measured units. Dehydration of the fixed tissues was performed through increasing concentrations of ethanol (75, 85, 95, and 100%) with 2 h at each step, cleared sequentially in ½ ethanol:½ xylene, xylene I, and xylene II (30 min each), and then infiltrated with ½ xylene:½ paraffin, low-melting paraffin (58 °C), and high-melting paraffin (60 °C) for 3 h each. After embedding, the tissues were sliced to a thickness of 5 μm by means of a microtome (RM2265, Leica, Wetzlar, Germany) and mounted on glass slides. Prior to staining, the sections were dewaxed in xylene I and II (10 min each), rehydrated through a descending ethanol series (100, 95, 85, 70, and 50%, 2 min each), and stained with HE by the instructions (Solarbio, G1120, Beijing, China). After staining, the sections underwent dehydration, xylene clearing, and neutral balsam mounting. Images were captured under microscope (DMi8, Leica, Wetzlar, Germany).

5-μm-thick tissue sections were dewaxed in xylene I and II (10 min each), followed by rehydration through a stepwise ethanol series (100, 95, 85, 70, and 50%), with 2 min at each concentration. For each animal, 2–3 tissue sections were analyzed, and within each section, at least 10 representative microscopic fields were randomly selected for evaluation of goblet cells. After rehydration, according to the instructions (Solarbio, G1285, Beijing, China), AB-PAS staining was conducted according to standard procedures. Tissue sections were first immersed in Alcian blue solution for 15 min, distilled water was used to rinse the samples. They were then treated with 0.5% periodic acid for 10 min, rinsed again, and subsequently exposed to Schiff reagent for 15–20 min. After coloration, the sections were washed in running water for 5–10 min. After section mounting and drying, images were captured microscopically (DMi8, Leica, Wetzlar, Germany).

A sheep-specific ELISA kit (MEIMIAN, MM-8115401, Jiangsu, China) was applied to measure taurine levels in serum. In addition, the concentration of bile acids (Nanjing Jiancheng, E003-2-1, Nanjing, China) in the serum was determined. The jejunal tissue was weighed and ground into powder. Then, saline was added at 1:9 (w/v), and commercial assay kits (Nanjing Jiancheng, Nanjing, China) were used to determine the content of glutathione (GSH, A006-2-1), superoxide dismutase (SOD, A001-3-2), and malondialdehyde (MDA, A003-1-2) in the samples.

TRIzol reagent (Invitrogen, 15596026CN, US) was used to extract total RNA from jejunal tissue as per the supplier’s instructions. After that, a reverse transcription kit (Takara, RR037A) was employed to synthesize cDNA from RNA, qRT-PCR was conducted using qPCR reagents (Mei5bio, MF013-01, Beijing, China). Relative mRNA expression was calculated using the 2^–ΔΔCt method, with β-actin as the reference gene. Briefly, the threshold cycle (Ct) values of target genes were first normalized to the Ct of β-actin to obtain ΔCt values. The ΔCt of each sample was then compared to that of the control group to calculate ΔΔCt, and relative gene expression was determined as 2^−ΔΔCt, representing the fold change relative to the control (Livak and Schmittgen, 2001). Primer information is detailed in Table 2. All primers were designed based on gene sequences available in the NCBI database and synthesized by Sangon Biotech Co., Ltd. (Shanghai, China).

Jejunal total protein was isolated using RIPA buffer (Bosterbio, AR0102, Wuhan, China). Following SDS-PAGE separation, equal protein amounts were transferred to nitrocellulose membranes. After a 1-h block with 5% non-fat milk, membranes were treated with primary and secondary antibodies (LI-COR, 926–32,211, 1:20,000). The antibodies information were: anti-ZO-1 (Affinity, AF5145, 1:1000; Jiangsu, China), anti-Claudin-1 (Affinity, AF0127, 1:1000), anti-Occludin (Affinity, DF7504, 1:1000), anti-IL-6 (Affinity, DF6087 1:1000), anti-IL-1β (Affinity, AF5103, 1:1000), anti-IL-10 (Affinity, DF6894, 1:1000), anti-P65 (ABclonal, A19653, 1:5000; Wuhan, China), anti-P-P65 (ABclonal AP1294, 1:10,000), anti-β-actin (Bioss, bsm-33036 m, 1:10,000). Protein signals were captured and quantified via an infrared dual-color imaging system (LI-COR Biosciences, Lincoln, NE, USA). For each lane, a rectangular region of interest (ROI) of identical size was applied around the target band, and the integrated intensity was extracted following background subtraction. Signal intensities were normalized to β-actin for each lane to control for loading variability. All six biological replicates per group were included in the quantification and statistical analysis. For visual clarity in the figures, representative blots from three replicates per group are shown; the reported quantitative results, however, are based on data from all replicates.

The microbial diversity in the jejunum was assessed via 16S rRNA gene sequencing. Total microbial DNA was isolated from jejunal content samples utilizing the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany) following the protocol. DNA integrity and concentration were assessed via electrophoresis on a 1.0% agarose gel and spectrophotometry (NanoDrop2000, Thermo Scientific). All DNA samples were placed at −80 °C prior to downstream analysis. A preliminary PCR was conducted using randomly selected jejunal content samples to ensure successful amplification with minimal cycle numbers and appropriate product concentration. Once pretests were validated, a formal PCR was performed using Pro Taq polymerase and the V3–V4 region was amplified using primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) (Guo et al., 2022). Each 20 μL reaction contained 2 × Pro Taq buffer, primers (0.4 μM final concentration each), template DNA (10 ng), and nuclease-free water. Purified amplicons were pooled in equal proportions and sequencing adapters were ligated to construct sequencing libraries. Sequencing was carried out using the Illumina HiSeq PE250 platform (Illumina, San Diego, USA) targeting the V3-V4 region. Following demultiplexing, raw paired-end reads were processed for quality control, filtering, and merging. The optimized sequences were denoised using the DADA2 algorithm (Callahan et al., 2016), yielding amplicon sequence variants (ASVs) with representative sequences and abundance tables.

Genomic DNA was isolated from the jejunal content using the Mag-Bind® Soil DNA Kit (Omega Bio-tek, USA), a protocol suited for high-complexity samples. DNA integrity and purity were assessed using 1% agarose gel electrophoresis. Fragmentation was performed using a Covaris M220 sonicator (Gene Company Limited, China) to yield ~350 bp fragments. Paired-end libraries were constructed, following standard procedures including adapter ligation, self-ligated fragments were removed via magnetic bead-mediated size selection, PCR amplification for template enrichment, and bead purification to obtain the final libraries. Sequencing libraries were run on a NovaSeq 6,000 system at Shanghai Majorbio Bio-Pharm Technology Co. using bridge-PCR cluster generation and sequencing-by-synthesis chemistry. Raw reads were filtered with fastp v0.20.0: reads < 50 bp, average Q < 20, or containing ambiguous bases were discarded and adapters trimmed (Chen et al., 2018). Host-derived sequences were then filtered out by aligning the clean reads against the sheep reference genome¹using BWA v0.7.17 (Li and Durbin, 2009); reads showing high similarity to the host genome were discarded. Using MEGAHIT v1.1.2, high-quality reads were assembled de novo; contigs ≥ 300 bp were retained (Li et al., 2015). Coding sequences were predicted with Prodigal, and those ≥ 100 bp translated into proteins (Hyatt et al., 2010). Protein sequences were clustered with CD-HIT (90% identity and coverage) to generate a non-redundant gene catalogue, keeping the longest member of each cluster as representative (Fu et al., 2012). Gene representation profiles were generated by mapping clean reads to the non-redundant gene catalogue using SOAPaligner v2.21 at ≥ 95% sequence identity (Li et al., 2008). Taxonomic annotation employed DIAMOND searches against the NCBI NR database (E-value ≤ 1 × 10⁻¹⁰⁰), followed by the lowest-common-ancestor algorithm to derive taxonomic profiles. Genes from bacterial taxa showing significant differential features among groups (LEfSe, LDA > 3, p < 0.05) were extracted and functionally annotated to infer microbial functional potential against the KEGG database² using DIAMOND, with an E-value threshold of ≤ 1 × 10⁻¹⁰⁰. KEGG pathway enrichment analysis was subsequently performed using the Majorbio Cloud Platform (Shanghai, China), with pathways identified as significantly altered based on a reporter score threshold of | ≥ 1.65| (Hong et al., 2024).

All sequencing and bioinformatics analyses in this study were conducted by Shanghai Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). Alpha diversity indices were computed via the Mothur platform to evaluate microbial diversity within samples (Schloss et al., 2009). To assess differences in microbial community structure among groups, Principal Coordinates Analysis (PCoA) was carried out using Bray-Curtis distance metrics (Lozupone et al., 2011). Group-level variation in community composition was statistically assessed using Permutational Multivariate Analysis of Variance (PERMANOVA) (Anderson, 2001). To identify taxonomic biomarkers that differed among treatment groups, the Linear Discriminant Analysis Effect Size (LEfSe) method was applied with a threshold of LDA > 2 and p < 0.05 (Segata et al., 2011). Spearman analysis was further conducted to examine the associations between representative microbial features and the measured indicators, and the results were visualized using heatmap software tools.

Data were expressed as means ± standard error (SEM), and intergroup differences were determined using a one-way analysis of variance (ANOVA), followed by Tukey’s post hoc test for multiple comparisons, with significance assessed using GraphPad Prism version 10.0. Differences were considered significant at p < 0.05 and extremely significant when p < 0.01.

Although serum taurine levels in ewes were unchanged during early gestation, dietary taurine supplementation increased serum taurine concentrations in both middle and late gestation (Figures 1A–C). Similarly, the bile acid levels were increased in taurine fed ewes and their offspring lambs (Supplementary Figures S1A–D). Moreover, the serum taurine abundances in offspring at 15 days of age were elevated in a dose-dependent manner (Figure 1D). Notably, compared with the control group, the offspring’s BW in the taurine-fed groups was elevated (Figure 1E).

Lambs in taurine-treated groups displayed increased villus length and villus-to-crypt ratio, along with decreased crypt depth compared to the control group in a dose-dependent manner (Figures 2A–D). AB-PAS staining results suggested an increased number of goblet cells in the intestinal epithelium of offspring from those taurine-fed ewes (Figure 2E). As expected, the MUC2 mRNA expression was markedly elevated in treatment lambs (Figure 2F). Furthermore, the abundances of tight junction proteins and mucin-related factors, including ZO-1, Claudin-1, and Occludin in treatment groups were elevated at the mRNA level, with consistent trends observed at the protein level (Figures 2G–I). In addition, the expression of Notch2 was decreased by maternal dietary taurine supplementation.

The mRNA abundance of IL-6 and IL-1β, were reduced in intestines of taurine-treated group compared to the control lambs (Figure 3A), and these trends were supported by the Western blot results (Figures 3B,C). Protein abundance of IL-10 was increased in treatment groups, but the difference was not statistically significant (Figure 3D). Meanwhile, both NF-κB and p-NF-κB contents were decreased in intestines of lambs from taurine-supplemented ewes (Figure 3E). Moreover, the SOD and GSH contents in intestine were elevated, while the MDA levels were decreased in treatment group (Figures 3F–H).

To evaluate the effects of maternal taurine intervention on the gut microbial composition of offspring, 16S rRNA gene sequencing was performed. After quality control and denoising with DADA2, ASV tables were generated and used for downstream analyses.

A total of 530 representative ASVs were detected across all samples, with 289, 377, and 287 ASVs detected in the CON, 0.1% TAU, and 0.2% TAU groups, respectively. Among these, 153 core ASVs were shared across all three groups. These shared ASVs, which were assigned to 153 genera, were used to generate the Venn diagram at the genus level (Figure 4B). The majority of sequences in each group were accounted for by these shared ASVs, indicating a relatively stable core microbiota, while taurine-related effects were primarily reflected in taxon-specific differences rather than in the number of shared ASVs. PCoA analysis at the genus level did not show a clear separation among the three groups, with PC1 and PC2 explaining 17.49 and 9.94% of the total variation (Figure 4C), indicating that maternal taurine supplementation did not induce a pronounced shift in overall microbial community structure. The α-diversity analysis showed that the Shannon index was significantly increased in the 0.1% TAU group compared with the control group (n = 6 per group) (Figure 4A). No statistically significant difference was observed in the 0.2% TAU group. Other α-diversity indices were also calculated but did not show statistically significant differences (Supplementary Figures S2A–D).

LEfSe analysis was performed to identify discriminative bacterial taxa among the three groups based on both statistical significance and effect size (Figure 4D) (LDA > 2, p < 0.05). A total of 13 genera were identified as group-specific biomarkers. The control group was characterized by Peptostreptococcus and Gemella, genera that have been associated with intestinal inflammation or pathogenic potential. In the 0.1% TAU group, several genera, including Desulfovibrio, Butyricicoccus, and Terrisporobacter, were identified as discriminative features. Notably, Butyricicoccus has been reported to produce butyrate and contribute to intestinal barrier function (Xiao et al., 2023; Kwon et al., 2024). The 0.2% TAU group was characterized by Coprococcus, Ruminococcus_gauvreauii_group, and Lachnospiraceae_FE2018_group, taxa previously associated with antioxidant capacity, anti-inflammatory responses, and intestinal homeostasis.

Spearman correlation analysis was performed between the LEfSe-identified differential genera and the aforementioned phenotypic indices. Several genera enriched in the treatment groups, particularly those in the 0.2% TAU group, exhibited significant correlations with gut health-related indicators (Figure 4E). Although correlations in the 0.1% TAU group did not reach statistical significance, several phenotypic indicators already showed clear improvement, suggesting a trend toward beneficial associations (given the modest sample size, some correlations may not reach statistical significance despite observable trends, n = 6 per group). In contrast, genera enriched in the control group were generally positively associated with unfavorable phenotypes and negatively associated with intestinal barrier and antioxidant indices.

Venn diagram analysis identified a total of 429 core microbial features shared among the three groups (Figure 5A). The number of group-specific microbial features observed in the 0.2% TAU group was higher than that in the control group. While this difference may be partially influenced by technical factors such as sequencing depth or assembly sensitivity, the consistent presence of these group-specific features across replicates indicates they may be of interest for further study alongside dominant taxonomic features. To assess overall differences in microbial community structure among the groups, non-metric multidimensional scaling (NMDS) was performed to visualize β-diversity. NMDS analysis suggested modest differences in microbial community structure among the three groups (Figure 5B).

At the genus level, Limosilactobacillus, Enterocytozoon, Lactobacillus, Chlamydia, Escherichia, Clostridium, Sarcina, Mogibacterium, Eubacterium, Blautia were the dominant genera (Figure 5C). Similar distribution patterns were observed at the species level (Figure 5D). Several genera, including Eubacterium, Blautia, Ruminococcus, and Prevotella showed higher representation in the treatment groups. LEfSe analysis further identified bacterial species discriminating among groups (Figures 5E,F) (LDA > 3, p < 0.05), several of which have been previously associated with gut health in prior studies. It is worth noting that, alongside Ruminococcus_sp., the 0.1% TAU group also harbored detectable levels of certain taxa, including Escherichia coli and Salmonella enterica, which are occasionally associated with opportunistic infections under specific conditions, potentially indicating an altered gut microbial configuration at this dosage. In contrast, the 0.2% TAU group maintained a more stable profile of treatment-responsive beneficial taxa.

Using Spearman correlation analysis, we focused on Eubacterium_sp., Ruminococcus_sp. and Blautia_sp., taxa that showed higher representation in the treatment groups, and examined their relationships with the aforementioned phenotypic indicators. In the 0.2% TAU group, significant positive correlations were observed with intestinal barrier and antioxidant-related indicators such as MUC2, Claudin-1, GSH, and ZO-1, and significant negative correlations with inflammation-related markers such as Notch2 and IL-6 (Figures 5G,H). In the 0.1% TAU group, the overall correlation patterns were directionally consistent with those in the 0.2% TAU group. Notch2 and IL-1β also showed negative associations with Ruminococcus sp., and other phenotypic indicators followed similar trends, although none reached statistical significance. At the functional level, KEGG pathway analysis revealed distinct microbial metabolic profiles among the three groups (Figures 5I,J). The top 20 differentially represented pathways ranked by reporter score are shown in the figure. In the 0.1% TAU group, pathways with positive reporter scores were primarily associated with microbial adaptability and core metabolic processes, including two-component system, lysine degradation, nitrotoluene degradation, lipopolysaccharide biosynthesis, biofilm formation by Escherichia coli, bacterial chemotaxis, and flagellar assembly. In contrast, the 0.2% TAU group exhibited a broader set of biosynthetic pathways with positive reporter scores relative to the control group, such as the biosynthesis of amino acids, aminoacyl-tRNA biosynthesis, lysine biosynthesis, and phenylalanine, tyrosine, and tryptophan biosynthesis. These results suggest that moderate taurine supplementation was associated with distinct shifts in microbial functional potential related to adaptability and metabolic capacity.