Forty-eight Hu sheep, 150 days of age with an average body weight (BW) of 31.40±0.55 kg (mean±standard deviation), were used for a 58-day feeding experiment. The sheep were assigned to two groups in a completely randomized design, with six pens (four sheep per pen) per group. Two diets with identical nutritional compositions were formulated: TMR and PTMR (Table 1). The TMR diet was prepared by chopping dried peanut vines into 3–4 cm pieces and thoroughly mixing them with concentrate feed at a 1:1 ratio (wt/wt). The PTMR diet, containing the same 50:50 ratio of chopped peanut vines to concentrate feed, was processed into pellets using a pelleting machine (9KLP-300; Henan Hengfu Machinery Equipment, Xinxiang, Henan, China) with a pellet diameter of 6 mm. A single batch of feed was prepared for the entire experimental period.

Hu sheep were fed twice daily (at 07:00 and 17:00) on an ad libitum basis during the experimental period. Daily feed residues were weighed, and the feed supply was adjusted to maintain a 5% daily residue. Each pen was equipped with a separate drinking bowl to provide unlimited access to water.

Each sample was dried in a laboratory oven at 60°C for 24 h to a constant weight, then ground using a mill and sieved through a 1 mm mesh for further analysis. DM, crude ash (Ash), calcium (Ca), and phosphorus (P) contents were determined according to the AOAC method [10]. Crude protein (CP) content was determined by Kjeldahl analysis using a Kjeltec 2300 analyzer (FOSS Analytical AB, Hoganas, Sweden). NDF and ADF contents were analyzed according to Van Soest et al [11] using an Ankom Fiber Analyzer (Ankom Technology, Fairport, NY, USA).

The daily TMR from each pen was weighed and sampled every day. Before the morning feeding on days 0, 28, and 56 of the feeding experiment, all sheep were weighed. Average daily gain (ADG), dry matter intake (DMI), and feed conversion ratio (F/G) were calculated using the aforementioned measurements.

Ruminal fluid samples were collected from each sheep two hours after morning feeding on day 56 of the experiment as previously described [12]. Samples were aliquoted into cryopreservation tubes, immediately frozen in liquid nitrogen, and transported to the laboratory for storage at −80°C.

Ruminal fluid pH was measured immediately after sample collection using a Russell CD700 pH meter. Two samples from each pen were randomly selected for ruminal volatile fatty acids (VFA) analysis. The samples were centrifuged at 10,000×g for 15 minutes at 4°C to pellet particulate matter. The supernatant was then filtered through a 0.45-μm membrane for clarification. Samples were treated with 25% (w/v) metaphosphoric acid to stabilize VFA at a final concentration of 2% (w/v). Acid-treated samples were centrifuged again under the same conditions (10,000×g, 10 minutes, 4°C) to remove protein precipitates. Samples were stored at −20°C and thawed for 30-fold dilution with deionized water immediately before analysis. VFA analysis was performed using an ICS-3000 ion chromatography system with an IonPac AS11-HC column (4×250 mm). Chromatographic separation used a 0.5 mL/min flow rate of 30 mM KOH eluent with the column temperature held at 35°C. Calibration curves were prepared from mixed standard solutions containing acetic, propionic, and butyric acids at 50, 100, 150, and 200 mg/L. A 20 μL aliquot of each diluted sample was injected and quantification was performed using external standards. Ruminal NH₃-N concentration was determined using UV spectrophotometry.

Seven samples per group were randomly selected from the aforementioned 12 samples per group for 16S rRNA gene sequencing. Rumen fluid samples were processed for DNA extraction and subsequent microbial community analysis. Total bacterial DNA was extracted using a QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. The 16S rRNA gene was then amplified with the universal primer pair (5′-ACTCC TACGGGAG GCAGCA-3′ and 5′-GGACTACHVGGGTWTCTAAT-3′) spanning the V3-V4 hypervariable region [13]. The integrity of the PCR product was verified and it was purified, followed by measurement of its concentration. The PCR products were then pooled in equal proportions based on their DNA concentrations. Sequencing of the 16S rRNA gene amplicons was performed using the Illumina HiSeq 2500 system (Illumina, San Diego, CA, USA) [14].

Raw sequencing data were processed through the following steps: 1) denoising, 2) paired-end read merging with FLASH v1.2.11, 3) quality filtering using Trimmomatic v0.33, and 4) chimera removal with UCHIME v8.1. Filtered reads were clustered into operational taxonomic units (OTUs) at 97% sequence similarity with USEARCH v10.0, against the Greengenes database (release 13_8). Taxonomic annotation of representative OTU sequences was performed using the RDP Classifier with a confidence threshold of 0.8 [15]. Abundance was normalized based on the least sequenced sample. Alpha diversity indices and rarefaction curves were generated in MOTHUR v1.3.0. Beta diversity was visualized through principal coordinates analysis (PCoA) based on weighted UniFrac distance matrices. Functional potential was predicted from 16S rRNA gene data using PICRUSt2, with KEGG pathways and COG categories annotated for metabolic and functional profiling respectively.

Statistical analyses were performed using SAS v9.2 software (SAS Institute, Cary, NC, USA). Daily intake measurements were aggregated into monthly means. BW, DMI and ADG data were analyzed using a repeated-measures mixed model with a randomized complete design. The model incorporated fixed effects of treatment, time (days on-feed), and treatment× time interaction, with pen as a random effect and sheep within a pen. Independent-samples t-tests were performed in IBM SPSS Statistics 26 (IBM, Armonk, NY, USA) to compare ruminal VFA concentrations and alpha diversity indices (ACE, Chao1, Shannon, Simpson), and significance was declared when p<0.05. Bacterial community differences were assessed by linear discriminant analysis (LDA) effect size (LEfSe), with a significance threshold of LDA score>3.0. Spearman’s rank correlation coefficient was used for correlation testing.

The effects of feeding PTMR diets on sheep intake and growth performance were presented in Table 2. The PTMR group had significantly higher BW (p<0.05) and showed a significant effect on ADG (p<0.05). However, the PTMR group had no significant effect on DMI (p = 0.115) or F/G (p = 0.866). Significant time-treatment interaction effects were observed for DMI (p<0.05), with the PTMR group exhibiting increased intake over time. Longitudinal analysis revealed significant temporal variations in BW, ADG, DMI, and F/G between PTMR and TMR groups throughout the experiment (p<0.05).

No significant differences were observed in ruminal fermentation parameters between TMR and PTMR groups, including pH (p = 0.429), acetate (p = 0.270), propionate (p = 0.321), butyrate (p = 0.302), acetate/propionate (p = 0.985), total VFA concentrations (p = 0.444), or NH₃-N (p = 0.210) (Table 3).

16S rRNA gene sequencing yielded 47,800±1,688 high-quality reads per sample (mean±standard error of the mean). The TMR group exhibited significantly higher alpha diversity indices (Sob, Ace, Bootstrap, Shannon, and Chao) than the PTMR group (p<0.01), including Shannon, Chao1, ACE, and Simpson indices. A Venn diagram showed 1,623 shared OTUs between groups, with 514 OTUs unique to TMR and 146 OTUs specific to PTMR (Figure 1).

The relative abundance of Bacteroidetes in both TMR and PTMR was greater than 59%, while the relative abundance of Firmicutes was greater than 27%, accounting for more than 70% of the total abundance (Figure 2A). In addition, Desulfobacterota had a significantly greater relative abundance in TMR than in PTMR (p<0.05). The relative abundance of Desulfobacterota in TMR accounted for 3.12%, while in PTMR it was 1%. At the genus level, Prevotella was the dominant genus in the rumen of both the TMR and PTMR groups (Figure 2B). A total of 45 different species were identified between the two groups, with significant differences (p<0.05). Taking the top fifteen different species as an example, the relative abundance of Rikenellaceae_RC9_gut_group, norank_f__Bacteroidales_RF16_group, NK4A214_group, Fretibacterium, Streptococcus, norank_f__p-251-o5, UCG-002, norank_f__UCG-011, norank_f__Selenomonadaceae, norank_f__Bacteroidales_BS11_gut_group, Anaerovibrio, Family_XIII_AD3011_group, CAG-352, and Anaerobiospirillum was significantly higher in TMR than in PTMR (p<0.05) (Figure 2D). Additionally, the relative abundance of Succinivibrionaceae_UCG-001 was significantly higher in PTMR than in TMR (p<0.05). LEfSe analysis of samples from the two groups (Figure 2F) revealed that there were 26 differential biomarkers (LDA score>3) for TMR and PTMR.

Functional profiles predicted from 16S rRNA gene data revealed 39 KEGG pathways and 23 COG categories. Among the 23 COG categories, “Energy production and conversion” was the sole functional class showing significant intergroup differences between TMR and PTMR (Figure 3A). Similarly, of the 39 KEGG pathways, “Signal transduction” was the only one that exhibited significant differences between TMR and PTMR (p<0.05). Metabolic pathways predominated in functional classifications, with carbohydrate metabolism being the most predominant subcategory, followed sequentially by amino acid metabolism and energy metabolism. None of these metabolic subcategories showed significant intergroup differences (Figure 3B).

As shown in Figure 4, the correlation heatmap was constructed based on rumen bacteria, specifically the top 15 genus level bacteria that exhibited significant differences according to p-values. The result revealed that VFA were significantly correlated with three genus level bacteria, of which two were positively correlated and one was negatively correlated (p<0.05). Furthermore, Anaerovibrio showed significant negative correlations with acetic acid, butyrate, propionate and total VFA (p<0.05). Moreover, Rikenellaceae_RC9_gut_group showed significant positive correlations with Acetate/Propionate and significant negative correlations with ADG (p<0.05). In addition, butyrate was positively correlated with norank_f__Bacteroidales_RF16_group of bacteria.

Dietary physical form significantly influences ruminant growth performance. Our results demonstrated that Hu sheep receiving PTMR exhibited higher ADG compared to those fed conventional TMR. This observation aligns with previous reports that pelleted diets enhanced growth performance in small ruminants [16,17]. Notably, while the DMI of the PTMR group was not significantly different from that of the TMR group overall, a significant time-treatment interaction was observed. The difference may be attributed to rumen adaptation to PTMR diets, as Zhong et al reported higher DMI in PTMR groups compared with our findings [18]. The PTMR group achieved higher ADG, suggesting that pelleting improves growth performance through mechanisms potentially related to enhanced rumen escape of nutrients. High density and sinking velocity may promote higher rumen escape rates in PTMR-fed sheep [19]. The enhanced ADG observed in pelleted diet studies may be linked to shifting nutrient digestion patterns toward the hindgut, a mechanism hypothesized to improve metabolic efficiency. It is noteworthy that PTMR-fed sheep showed only a non-significant DMI trend alongside a significant ADG increase, suggesting that pelleting subtly enhances growth performance through mechanisms beyond the current study’s scope. Further research measuring digestibility and rumen dynamics is needed to clarify these mechanisms.

The pH and VFA concentrations in rumen fluid serve as critical indicators for evaluating ruminal fermentation, reflecting both the stability of the rumen environment and fermentation efficiency. In the present study, the PTMR exhibited no significant effects on rumen pH or VFA concentrations between the experimental groups, which aligns with the findings reported by Karimizadeh et al [20]. Similarly, Li et al [21] observed unchanged rumen VFA concentrations but a decreased rumen pH in lambs fed PTMR for 3 hours, attributing this pH reduction to the accelerated feeding rate associated with PTMR. These results suggest that the degradation and fermentation processes of PTMR in the rumen reach a stable state compared to conventional TMR, although granulation may simultaneously increase the rumen passage rate [22]. In this study, the rapid passage rate of PTMR might have shifted the primary site of fermentation from the rumen to the hindgut, where undigested substrates could undergo further microbial metabolism. This mechanism could explain the absence of significant changes in rumen VFA concentrations despite the observed improvement in average ADG. In contrast, Zhang et al [23], reported significant differences in acetic, propionic, and butyric acid concentrations even in post-slaughter measurements after 12 hours of PTMR feeding, hypothesizing that pelletization enhances feed intake rate and stimulates microbial fermentation activity. These divergent findings highlight the need for more systematic and dynamic measurements of rumen pH and VFA profiles in future studies to comprehensively elucidate the effects of granular TMR on rumen fermentation kinetics.

The regulation of metabolic functions of the body through the interaction between the host and bacteria has been extensively studied. Dietary habits are among the crucial factors that contribute to bacterial diversity and community composition, ultimately influencing body health [24]. Nutritional intervention can influence estrogen levels, modulate the gut microbiota, and facilitate increased dietary microbiota diversity [25]. The present study sequenced the rumen bacteria in Hu sheep and found that the three most abundant bacterial phyla were Bacteroidetes, Firmicutes, and Proteobacteria, which is consistent with the findings reported by Zhang et al [23]. In ruminants, Firmicutes and Bacteroidetes play a crucial role in fiber degradation and the digestion of complex carbohydrates [26]. The most abundant bacteria at the genus level included Prevotella, Rikenellaceae RC9 gut group, Quinella, norank f Bacteroidales RF16 group, Prevotellaceae UCG 001, Ruminococcus, Prevotellaceae UCG 003 and Succiniclasticum group. These bacterial genera remained unaffected by the physical forms investigated in this study. The stability of these abundant bacterial genera may indicate the presence of a core microbiome [27]. In the present study, we observed that the relative abundance of the Rikenellaceae RC9 gut group at the genus level was significantly higher in TMR compared to PTMR, which aligns with the findings reported by Bo Trabi et al [28]. Conte et al found that the Rikenellaceae RC9 gut group is associated with lipid metabolism in the bovine rumen [29]. Moreover, the relative abundance of norank f Bacteroidales RF16 group, NK4A214 group, Fretibacterium, Streptococcus, UCG-002, and other bacteria was significantly higher in TMR compared to the PTMR group at the genus level. This may be attributed to the longer retention time of TMR in the rumen, allowing for more extensive fermentation by fiber-degrading bacteria such as Rikenellaceae RC9, as well as the increased activity of other bacterial genera (e.g., norank_f_Bacteroidales_RF16_group, NK4A214_group, Fretibacterium, Streptococcus, and UCG-002) involved in nutrient metabolism processes, including carbohydrate and fatty acid metabolism [30,31]. In contrast, Succinivibrionaceae UCG-001 was the only bacterial genus in the PTMR group that exhibited a significantly higher relative abundance compared to the TMR group. An increase in the abundance of Succinivibrionaceae may significantly enhance the fermentation environment in the rumen, thus improving the digestibility and utilization of feed [32]. This correlation may be attributed to alterations in the physical properties of starch in TMR resulting from granulation treatment, which could promote the growth of starch-degrading bacteria such as Succinivibrionaceae. Although the rapid passage rate of PTMR in the rumen generally reduces the retention time for bacterial fermentation, Succinivibrionaceae, as a starch-degrading bacterium, may have a competitive advantage in utilizing the rapidly available starch substrates in PTMR, leading to its increased relative abundance [33]. Additionally, Zhang et al [23] found that PTMR feeding decreased cecal microbial abundance (e.g., Lachnospiraceae), indicating its modulatory role in hindgut metabolic processes. These findings suggest that granulation not only reduces the retention time of feed in the rumen but also selectively enriches bacterial taxa that are adapted to the altered dietary conditions, thereby influencing both rumen fermentation and nutrient utilization. Further in-depth research is warranted to elucidate the precise mechanisms underlying these observations.

Previous studies have reported changes in the response of microbial community composition and function to dietary alterations [34]. Consequently, we speculated that the bacterial abundance and functional profiles might respond to PTMR treatment. The prediction of bacterial gene functions further revealed distinct differences between Hu sheep fed TMR and those fed PTMR. Within the COG gene family, a significant decrease in energy production and conversion was observed in PTMR-fed animals. Relevant studies have reported a strong correlation between energy utilization and production performance in bacteria when subjected to changes in dietary nutrient composition [35]. In the present study, no significant alteration in rumen VFA concentration was observed in the PTMR group, whereas a notable difference was detected in the prediction of rumen bacterial functions. We hypothesize that this discrepancy may arise from alterations in energy supply due to the short-term retention of feed in the rumen following PTMR treatment. The rapid passage rate of PTMR likely reduced the extent of rumen fermentation, shifting the site of nutrient digestion and energy production to the hindgut. Such alterations may, to some extent, be attributed to changes in dietary energy density and substrate availability following pelletization.

Furthermore, we performed a prediction of bacterial functions based on the KEGG database, and the results revealed that signaling pathways were significantly downregulated in the PTMR group. This finding suggests that pelletization may influence the regulatory mechanisms of rumen bacteria, potentially affecting their ability to sense and respond to dietary changes. Despite the scarcity of relevant literature on the signaling role of rumen bacteria, the crosstalk among rumen bacteria and between rumen bacteria and the host constitutes one of the crucial regulatory mechanisms in ruminants [36].

Correlation analysis between rumen bacteria and the metabolites VFA, ADG, and DMI revealed significant associations among these variables. Prior studies have demonstrated a significant and positive correlation between norank_f_Bacteroidales_RF16 group and both propionate and acetate [37]. Another correlation analysis indicated that Anaerovibrio exhibited a significant negative correlation with acetic acid, propionic acid, butyric acid, as well as total VFA. This finding contrasts with previous studies, which have reported a positive correlation between Anaerovibrio and VFA production due to its role in lipid hydrolysis [38]. However, Chen et al also reported contrasting findings, showing that Anaerovibrio was positively correlated with the VFA [39]. In the present study, we hypothesize that under PTMR feeding conditions, the rapid passage rate of granular feed may have limited the extent of lipolysis by Anaerovibrio, thereby reducing its contribution to VFA production. Additionally, the altered physical properties of lipids in pelletized feed could further restrict substrate accessibility for Anaerovibrio, leading to the observed negative correlation. These results highlight the complex interplay between dietary form, bacterial activity, and metabolic outcomes in the rumen.