All animal works were approved by the Animal Care and Use Committee of Jiangxi Agricultural University (No. JXAU 2011-006).

The experimental cattle used in this study were sourced from a Chinese indigenous breed named Xiangzhong black cattle. Twenty-one uncastrated Xiangzhong black cattle (initial body weight of 385.08±15.20 kg) at the age of 1.5 years were randomly allocated into three groups with seven individuals per group. Each group was assigned to one of the following dietary treatments: i) basal diet+0 mg NA/kg concentrate (CL), ii) basal diet + 800 mg NA /kg concentrate (NA1), iii) basal diet+1,200 mg NA/kg concentrate (NA2). The cattle were housed in individual stalls (1.7×3.5 m) and fed a diet consisting of 60% concentrate and 40% rice straw twice daily at 07:00 and 16:00, respectively, adopting ad libitum feeding manner. Water was provided ad libitum. All cattle were in good health and had not received any antibiotic treatment for two months before slaughter. The composition and nutrient levels of the basal concentrate diet are presented in Table 1. NA was purchased from Jiangsu Nantong Biotechnology Co., Ltd. (Purity >97%; China). For NA treatments, the mineral premix was initially blended with NA and subsequently added to the concentrate. After a formal experiment period of 120 days, all animals were slaughtered following the Chinese standard procedures at a commercial slaughterhouse. Rumen samples were collected from multiple sites of the rumen, equally mixed, and strained with cheese cloth to obtain rumen fluid that are used for analyses. Three out of seven rumen samples from each group were randomly chosen for 16S rRNA gene sequencing. All samples for 16S rRNA gene sequencing were promptly dipped in liquid nitrogen and transferred to a −80°C refrigerator until utilization.

Microbial DNA samples were extracted from rumen samples within one month after sample collection using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, North Rhine-Westphalia, Germany) according to the manufacturer’s instructions. The concentration and purity of DNA samples were determined using Nanodrop-1000 and 1% agarose gel electrophoresis. The V4 region of bacterial 16S rRNA gene was amplified using the fusion primers 515F (5′-GTGCCAG CMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGG TWTCTAAT-3′) [14]. Subsequently, the polymerase chain reaction products were normalized to equimolar quantities and subjected to sequencing on an Illumina MiSeq platform (Illumina, San Diego, CA, USA) using a pair-end 250 bp strategy.

Raw sequencing data were filtered to exclude the adaptor and low-quality sequences to obtain clean reads. Clean sequence reads with overlapped fragments longer than 10 bp and zero mismatches were assembled into tags by using the FLASH program (v.1.2.11) [15]. The USEARCH software (v.7.0.1090) [16] was employed for the operational taxonomic unit (OTU) clustering at the threshold of 97% sequence identity. Alpha-diversity metrics including observed species, Shannon, Simpson, Chao, Ace, and Good’s coverage were calculated using Mothur (v.1.39.5) [17]. The beta-diversity analysis of rumen microbiota including principal component analysis (PCA), principal co-ordinates analysis (PCoA), non-metric multidimensional scaling analysis (NMDS) and analysis of similarities (ANOSIM) were conducted using QIIME (v1.80) [18]. The figures were plotted with ggplot2 in the package (v 3.5.2). The potential functional capacity of the rumen microbial community was predicted using PICRUSt 2 based on 16S rRNA marker gene sequences. Normalization, metagenome prediction and function categorization based on Kyoto encyclopedia of genes and genomes (KEGG) pathways and metaCyc Metabolic Pathways were performed by PICRUSt 2 according to the standard procedures [19]. linear discriminate analysis effect size (LEfSe) was used to identify the bacterial taxa and KEGG pathways exhibiting differential abundances among the three groups.

The rumen pH was immediately measured after collection using the Mettler Toledo Delta 320 pH meter (Mettler Toledo, Greifensee, Switzerland). The concentrations of VFAs including acetate, propionate, and butyrate in the rumen were determined by high-performance liquid chromatography (HPLC) using an LC-10A system (Tokyo, Japan). Specifically, a Shodex RSpak KC-811S-DVB gel C column (8.0 mm×30 cm) was applied under an oven temperature of 30°C with 3 mmol/L of HClO4 as the mobile phase at a flow rate of 1.0 mL/min. A total of 5-μL sample supernatant was injected for each measurement and data were collected using the SPD-M10AVP detector. Total VFA (TVFA) is defined as the sum of acetate, propionate, and butyrate concentrations. Ammonia nitrogen (NH₃-N) concentration was determined by a spectrophotometer (Model 721/721-100; Shanghai, China) using ammonium chloride solution as standard.

Fresh fecal samples (250 g per animal), feed, and dried rice straw were collected daily during the last two weeks of the feeding experiment. Subsequently, all fecal samples were mixed, and 10% hydrochloric acid added to fix nitrogen. All samples were subjected to oven drying at 65°C for 72 h until constant weight. The nutrient contents in both the feed and the mixed feces were determined using AOAC methods (AOAC, 2001). Briefly, The dry matter (DM) and organic matter (OM) contents were determined according to the method described by Luo et al [6]. Crude protein (CP) was measured using the Kjeldahl nitrogen method and calculated as N×6.25. Crude fat (CF) was extracted with diethyl ether extraction using a Soxhlet apparatus. The acid detergent fiber (ADF) and neutral detergent fiber (NDF) contents were determined using the Van Soest methods [20]. Acid insoluble ash (AIA) was employed as an internal marker to calculate the apparent digestibility of nutrients as previously described by Luo et al [6]. The apparent digestibility of nutrients was calculated using the formula: apparent digestibility of a specific nutrient (%) = [1 − (nutrient content in feces × hydrochloric AIA content in feed) /(nutrient content in feed × hydrochloric AIA content in feces)] × 100.

One-way analysis of variance analysis was performed to compare rumen fermentation parameters and apparent digestibility of nutrients using SPSS 19.00 software (SPSS Inc., Chicago, IL, USA). Multiple comparisons were conducted for identifying significant differences among the three groups. The results were expressed as mean values and the standard error of the mean. The significance thresholds were set at p≤0.01, p≤0.05, and 0.05<p≤0.10, respectively. The correlations between the concentrations of VFAs in the rumen and the abundances of OTUs were evaluated by the Spearman rank correlation analysis. The significance level was set at the false discover rate <0.05. The heatmap was plotted with the pheatmap package in R (v 3.3.1).

A total of 781,946 clean reads were obtained for all tested samples. To avoid the influence of sequencing depth on microbial composition, the sequencing depth for all tested samples was normalized to at least 15,803 tags (31,606 reads) per sample (the least number for individuals). The Good’s coverage based on the observed OTUs reached 98% (Table 2), and the rarefaction curves also achieved saturation (Supplementary Figure S1). These results indicated that the sequencing depth was sufficient.

To investigate the effect of NA on rumen microbiota composition, the alpha-diversity and beta-diversity were compared among three groups. The results showed that the OTU number, diversity indices (Sobs and Ace), and richness index (Chao) in the NA1 group were significantly lower than those in CL and NA2 groups (p<0.05) (Table 2). In the PCA and PCoA, the samples from each of the CL, NA1, and NA2 groups were clustered into their corresponding group, except for one sample from the control group, which was close to the NA groups (Figure 1). The results also demonstrated that the inter-sample distances within each group were remarkably close, enabling their distinct separation from other groups, and suggested the significant effect of NA on rumen microbial compositions. The cattle from the two NA groups had a higher similarity of rumen microbial compositions than the cattle from the control group.

Based on the 97% sequence identity, a total of 2,160 OTUs were obtained from all nine tested samples. Among them, 1,172 OTUs were shared in samples across all three groups. The number of OTUs that were specific to the CL, NA1, and NA2 groups was 249, 126, and 131, respectively (Figure 2). These 2,160 OTUs were annotated to a total of 21 bacterial phyla. After the phyla with low relative abundance were filtered, 10 bacterial phyla accounting for more than 99.0% of total abundance were left. The predominant phyla in bovine rumen were Bacteroidetes, Firmicutes, Spirochaetes, and Proteobacteria, which accounted for 31.89% to 41.22%, 18.78% to 22.24%, 8.20% to 13.75%, and 7.70% to 12.36% of relative abundances, respectively, followed by Fibrobacteres, Tenericutes, Cyanobacteria, Verrucomicrobia and Lentisphaerae (Figure 3a; Supplementary S2a). The relative abundance of Firmicutes was similar in all three experimental cattle groups (CL, 20.25%; NA1, 22.24%; and NA2, 18.78%) (Figure 3a; Supplementary Figure S2a). However, the NA1 (13.75%) and NA2 (12.02%) groups had a significantly higher relative abundance of Spirochaetes compared to the CL group (8.20%). Additionally, compared with the CL group, there was also a significant increase in the relative abundance of Tenericutes (CL vs NA1: 2.57% vs 4.93%) and Proteobacteria (CL vs NA1: 7.70% vs 12.36%) in the NA1 group as well (Figure 3a; Supplementary Figure S2a). However, the relative abundances of Bacteroidetes (CL vs NA1 vs NA2: 41.22% vs 31.89% vs 37.22%) and Lentisphaerae (CL vs NA1 vs NA2: 1.30% vs 0.34% vs 048%) were decreased in the NA1 and NA2 groups compared to the CL group. The relative abundances of other bacterial phyla were not significantly altered by the intervention of dietary NA. At the genus level, Prevotella, Treponema, Ruminococcus, and Fibrobacter were identified as the predominant genera in the rumen (Figure 3b; Supplementary Figure S2b). The relative abundances of Prevotella, Treponema, and Ruminococcus were higher in both NA1 and NA2 groups than that in the CL group. Conversely, the abundance of Bacteroidia was lower in both NA1 and NA2 groups than that in the CL group (Figure 3b; Supplementary Figure S2b).

LEfSe analysis was further employed to identify the differential bacterial taxa among the three groups. At the significant threshold of linear discriminant analysis (LDA) ≥2 and p<0.05, a total of 78 bacterial taxa showed different enrichments among three experimental cattle groups (Figure 4; Supplementary Table S1). Most numbers of 47 bacterial taxa were enriched in the CL group. Among these 47 differential bacterial taxa, eight taxa belonged to Erysipelotrichi in which the taxa from Erysipelotrichi to the genus RFN20, and to RFN20 OTUs showed enrichment in the CL group. A total of 15 bacterial taxa belonging to Clostridiales were also enriched in the CL group. Among them, five taxa were classified into Ruminococcaceae, three taxa into Mogibacteriaceae, and two taxa into Lachnospiraceae. Five taxa belonging to Sphaerochaetales (three taxa were annotated to Sphaerochaeta) and six taxa classified into Pedosphaerales were also enriched in the control group. A total of 11 bacterial taxa had higher abundances in the NA1 group (800 mg/kg NA), including two taxa belonging to Roseburia faecis. There were 20 taxa showing enrichments in the NA2 group, including seven taxa belonging to Prevotella, three taxa classified into Paraprevotellaceae, three taxa belonging to Bifidobacteriaceae, and two OTUs annotated to Fibrobacter succinogenes (LDA value >2.0 and p<0.05) (Figure 4; Supplementary Table S1).

Potential functional capacities of rumen microbiota were predicted by PICRUSt 2 with 16S rRNA marker gene sequences for all nine experimental samples based on KEGG pathways and metaCyc metabolic pathways. Metabolic pathways, biosynthesis of secondary metabolites, biosynthesis of amino acids, microbial metabolism in diverse environments, and carbon metabolism were the KEGG pathways whose abundances were ranked in the top five in all nine samples. Pentose phosphate pathway (non-oxidative branch), gondoate biosynthesis (anaerobic), adenosine ribonucleotides de novo biosynthesis, cis-vaccenate biosynthesis, and superpathway of pyrimidine nucleobases salvage were the top five metaCyc metabolic pathways according to their abundances in the tested samples. LefSe analysis was used to identify the differential pathways among the three groups. At the significance threshold of LDA>2.0 and p<0.05, the KEGG pathway of biosynthesis of secondary metabolites was enriched in the NA2 group, the pathways of Necroptosis and Amoebiasis had higher abundances in the NA1 group, and Monobactam biosynthesis showed the enrichment in the CL group (LDA> 3.0, p<0.05) (Figure 5a). For the metaCyc metabolic pathways, superpathway of thiamin diphosphate biosynthesis I, superpathway of thiamin diphosphate biosynthesis II, and thiazole biosynthesis I (Escherichia coli) were significantly enriched in the NA2 group (LDA value >2.0 and p<0.05). However, we did not identify any pathways enriched in the CL and NA1 groups (Figure 5b).

To investigate the impact of supplemental NA on rumen fermentation and digestion, the rumen fermentation parameters and apparent digestibility of nutrients were measured for all 27 experimental cattle. As shown in Table 3, NA supplementation showed certain effects on rumen fermentation in uncastrated bulls. The concentrations of acetate, TVFA, and the acetate/propionate ratio (A/P) in the rumen were significantly increased in both NA1 and NA2 groups (p<0.05). In addition, the concentration of butyrate was also significantly increased in the NA2 group (p<0.05). When comparing the proportion of individual VFA, it was observed that the proportion of acetate was significantly increased in both the NA1 group and NA2 groups. However, there was a significant decrease in the proportion of propionate in both NA treatment groups. However, dietary NA had no significant effects on rumen pH, NH₃-N, propionate concentration, and butyrate proportion among the three groups (p>0.05).

As shown in Table 4, the apparent digestibility of ADF exhibited a significant increase in both NA1 and NA2 groups (p<0.01). Conversely, the apparent digestibility of DM and OM demonstrated a decrease in the NA2 group. No statistically significant differences were observed in the apparent digestibility of CP, CF, and NDF among all three groups (p> 0.05).

To evaluate the correlation between rumen fermentation parameters and the rumen microbial compositions, a correlation analysis was conducted between the concentrations of rumen VFAs and the abundances of 50 OTUs with high abundance in tested samples (Figure 6; Supplementary Table S2). l r l >0.5 and p<0.05 was set as the threshold of significant correlations, and l r l >0.5 was treated as the tendency of correlation. A total of 13 significant correlations and 35 tendencies were obtained (Figure 6; Supplementary Table S2). In details, acetate concentration exhibited a strong association with six OTUs (l r l >0.5), among which the OTU annotated to BF311 displayed negative correlation with acetate concentration (r = −0.53), while the OTU442 annotated to Ruminococcus (r = 0.63), OTU539, OTU1800, and OTU516 to Prevotella (r = 0.65, 0.67, and 0.67, p<0.05), and the OTU1187 to Anaeroplasma (r = 0.67, p<0.05) showed a positive correlation with acetate concentration. Propionate concentrations in the rumen showed significant correlations with the abundances of seven OTUs (l r l >0.5), among them, the OTU1484 annotated to Bacteroidales (r = −0.63), OTU724 to Treponema (r = −0.60), OTU1986 to Paraprevotellaceae (r = −0.60), and OTU757 to Prevotella (r = −0.87, p<0.05) exhibited negative correlations with propionate levels of rumen fluid, while the OTU436 annotated to RFP12 (r = 0.52), OTU655 to bacteria (r = 0.60), and OTU1634 to BS11 (r = 0.72, p<0.05) displayed positive associations with propionate concentration. Butyrate concentrations in the rumen were negatively correlated with the OTU509 annotated to Ruminococcus (r = −0.53) and OTU2275 to Treponema (r = −0.56). However, the OTU919 (Bacteria) was positively associated with the concentration of butyrate. The concentration of TVFA exhibited a negative correlation with the OTU1616 annotated to BF311 (r = −0.60). However, the OTUs annotated to Ruminococcus (r = 0.60), Prevotella (r = 0.68, p<0.05), and Anaeroplasma (r = 0.71, p<0.05) showed positive correlations with TVFA. The greatest number of 16 OTUs showed positive correlations with the NH₃-N concentrations (r>0.50). These OTUs were annotated to Prevotella, Ruminococcus, Fibrobacter (r = 0.72, p<0.05), and Treponema (r = 0.83, p<0.05), while negative correlations were observed on the OTU1616 annotated to BF311 (r = −0.53) and OTU834 to Bacteroidales (r = −0.68, p<0.05). Rumen pH exhibited a positive correlation with the OTU1616 annotated to BF311 (r = −0.58). A total of five OTUs were negatively associated with rumen pH values. Among them, four OTUs belonged to Treponema (r = −0.57 to −0.68, p<0.05), and the other one OTU to Prevotella (r = 0.87, p<0.05).