The research procedures were approved by the West Virginia University’s Institutional Animal Care and Use Committee (IACUC Protocol Number: 2206054350). A group of 59 Angus bulls (average body weight (BW) = 428 ± 18.8 kg; 350 ± 13.4 d of age) were fed a high-forage total mixed ration (TMR; primarily consisting of corn silage, hay, cracked corn, and a ration-balancing supplement; see Table 1 ) for 60 days in two pens (pen 1: n = 31; pen 2: n = 28). Each pen was equipped with two GrowSafe8000 intake nodes (GrowSafe Systems Ltd., Airdrie, Alberta, Canada) to measure individual feed intake and In-Pen Weighing Positions (IPW, Vytelle LLC), positioned at a water trough in each pen to measure the body weight (BW) of individual animals several times daily (Wells et al., 2021). The use of IPW to measure BW has enabled the measurement of feed efficiency with sufficient accuracy over a test period of 59 days (MacNeil et al., 2021; Wells et al., 2021; Taiwo et al., 2022a). Following this test period, growth performance and feed intake data were collected. The phenotypic RFI values of the bulls were determined as described previously by Taiwo et al., 2022a. Briefly, the daily BW was regressed on time to calculate the beginning BW, mid-test BW, and average daily gain (ADG). Thereafter, ADG and metabolic mid-test BW (mid-test BW0.75) were regressed against individual daily DMI, and RFI was calculated as the difference between the predicted value of the regression and the actual measured value using the following equation; Y =β₀ + β₁X₁ + β₂X₂ + ϵ, where Y is the dry matter intake (DMI; kg/d), β₀ is the regression intercept, β₁ and β₂ are partial regression coefficients, X₁ represents the metabolic mid-test BW (MMTW = mid test BW^0.75; kg), and X₂ is the average daily gain (ADG; kg/d) (Durunna et al., 2011; Taiwo et al., 2022a). Genetic evaluation of the bulls was performed by Vytelle (Vytelle Insight Beef Genetics) using data collected by Vytelle SENSE systems and Vytelle INSIGHT analytics services, which include at least three generations of pedigrees to determine the RFI-EPD values. At the conclusion of the trial and subsequent calculations, the bulls were ranked based on their Residual Feed Intake Expected Progeny Difference (RFI-EPD) coefficients. The bulls with the most negative RFI-EPD values (NegRFI; genetically efficient; n = 10) and the most positive RFI-EPD values (PosRFI; genetically inefficient; n = 10) were identified for further evaluation.

Rumen fluid samples were collected prior to morning feeding on day 60 of the testing period from all the bulls using an orally administered stomach tube connected to a vacuum pump (Ruminator; profs-products.com). The first 200 mL of rumen fluid was discarded to prevent saliva contamination. Then, 200 mL of fluid was collected from each animal and placed into 50-mL polypropylene conical bottom tubes. The samples were immediately placed on ice after collection and subsequently stored at −80°C until DNA extraction and sequencing were performed.

Before DNA extraction, rumen fluid samples from Angus bulls identified as NegRFI and PosRFI were thawed at room temperature. Microbial DNA was extracted from 500 µL of rumen fluid samples using the Qiagen DNeasy Powersoil Pro DNA Isolation Kit following the manufacturer’s instructions (Qiagen; catalog number: 47014, Germantown, MD, USA). Using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), total DNA purity was measured, with an A260/A280 ratio ranging from 1.8 to 2.0. To prevent contamination, Kimtech wipes (Fisher scientific; catalog number 06-666, Pittsburg, PA, USA) were utilized throughout spectrophotometry quantification steps prior to each sample measurement. DNA was stored at −80°C until sequencing.

All subsequent steps were performed at Novogene Bioinformatics Technology (UC Davis Sequencing Center, CA; Batch ID: X202SC23094803-Z01-F001; Contract ID: H202SC23094803). For library preparation and construction, quality control (QC) of DNA samples was performed using agarose gel electrophoresis to assess DNA degradation and potential contamination. The concentration of DNA in the library was measured using a Qubit^® 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA) with a Qubit^® dsDNA Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). The purity of samples ranged from 1.8 to 2.0, with concentrations greater than 1 µg used to construct the library. Following library construction, the DNA was measured using the Qubit Fluorometric Quantification system (Thermo Fisher Scientific, Waltham, MA, USA). Genomic DNA of each sample was randomly sheared into short fragments (350 bp) for sequencing and processed using the Illumina NovaSeq 6000 PE150bp sequencing platform according to its effective concentration and expected data volume, as described by Li et al. (2022).

Preprocessing of raw data from the Illumina sequencing platform was performed to obtain clean data for subsequent analysis. Clean data were obtained by removing low-quality bases (default quality ≥ 38) exceeding the default length of 40 bp, reads containing N bases exceeding the default length of 10 bp, and reads with overlaps with adapters exceeding the default length of 15 bp. Subsequently, clean data were BLASTed against the host database (B. Taurus ARS-UCD2.0)., using Bowtie2 software (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml) by default, followed by MEGAHIT software for assembly analysis of clean data. For Scaftigs without N assembly, resulting Scaffolds from the N junction were broken as previously reported (Qin et al., 2010; Li et al., 2015).

DIAMOND software (https://github.com/bbuchfink/diamond/; Buchfink et al., 2015) was used to align unigenes to the sequences of bacteria, fungi, archaea, and viruses from the non-redundant National Center for Biotechnology Information (NCBI) database. To prevent multiple alignment results and ensure species annotation and abundance of the sequences, a Lossless Compression Algorithm (LCA) was applied in the taxonomic software MEGAN. To form leveled taxonomic abundance tables, a relative abundance overview, and an abundance clustering heatmap, principal coordinate analysis (PCA) was performed. To identify differences between groups, analysis of similarities (ANOSIM), Metastat, and linear discriminant analysis (LDA, default score of 4) effect size (LEfSe) were conducted.

To evaluate the rumen microbial function, DIAMOND software (https://github.com/bbuchfink/diamond/) was used to BLAST Unigenes alignment with the functional database according to the default parameter settings of BLAST. The functional databases include KEGG (http://www.kegg.jp/kegg/) and CAZy (http://www.cazy.org/). Additionally, LEfSe analyses (LDA ≤ 2) of the functional differences between the two groups were conducted.

The growth performance data (DMI, average daily gain (ADG), initial and final BW) and RFI values of the PosRFI and NegRFI Angus bulls were analyzed using the GLIMMIX procedure of SAS version 9.4 (SAS Institute Inc., Cary, NC). Animals were included as a random effect nested within RFI-EPD. The RFI-EPD status was included as a fixed effect, and initial body weight values were included as a covariate for the final body weight. Significant effects were reported at P ≤ 0.05.

The results of the growth performance of the NegRFI and PosRFI bulls are shown in Table 2 . The average RFI-EPD values of the NegRFI and PosRFI bulls were -0.39 kg/d and 0.29 kg/d, respectively. The initial and final body weights and ADG were similar (P > 0.05) between the groups. However, the DMI was lower (P = 0.0002) in the NegRFI bulls (7.74 kg/d) compared to the PosRFI bulls (9.90 kg/d).

A total of 132, 526, 840 raw reads were obtained from the 20 rumen fluid samples of the Angus bulls, with an average of 6,626,342.00 ± 815,282.6 raw reads per sample. After quality control and removal, 131,849, 910 clean reads, with 6,592,455 ± 803,592 reads per sample, were retained ( Supplementary Table 1 ). Using Megahit, a total number of 6,993,174 scaftigs with an average of 349,659 ± 72,350 scaftigs per sample were obtained by interrupting scaffolds at the N-site, resulting in an average max length of 170,278bp ( Supplementary Table 2 ).

The results revealed numerically greater gene richness in NegRFI bulls compared to PosRFI bulls ( Figure 1 ). ANOSIM results revealed a small but noticeable difference (P = 0.052; R = 0.097) between the rumen microbial communities of the two groups ( Figure 2 ). However, the results of the Metastats analysis identified a total of 35 differentially abundant taxa at the species level. The relative abundance of 29 species such as Fibrobacter UWB5, UWT2, UWB1, Trepomena C6AB and Trepomena JC4 were greater in NegRFI bulls, whereas 6 species such as Frischella japonica, Prevotella veroralis, and Helicobacter cetorum were more abundant in PosRFI bulls ( Supplementary Figure 1 ). Abundance heatmaps showing the distribution of the 35 dominant taxa at the phylum and species levels are reported in Supplementary Figures 2 , 3 . Figure 3 shows the principal coordinate analysis (PCA) plot based on the differentially abundant (P < 0.05) species. Results of the LEfSe analysis showed the class Fibrobacteria (LDA = 5.1) and genus Fibrobacter (LDA = 4.8) as the most differentially enriched taxa in NegRFI bulls compared to PosRFI bulls ( Figure 4 ).

Utilizing the first-level KEGG orthology database, 46 core pathways belonging to 6 functional categories were identified within the rumen microbiome of the bulls. Predominant genes involved in carbohydrate metabolism, amino acid metabolism, translation, replication and repair, and folding, sorting and degradation were identified as the top 5 most abundant functions ( Figure 5 ). No differences in KEGG functional pathways were detected between the two groups.

To identify the relative abundance of carbohydrate-active enzymes (CAZy), predicted genes were annotated to the CAZy database containing the following 6 classes: glycoside hydrolases (GH), glycosyltransferases (GT), polysaccharide lyases (PL), carbohydrate-binding modules (CBM), carbohydrate esterases (CE), and auxiliary activities (AA). The relative distribution of all annotated CAZy genes is shown in Figure 6 . The LEfSe analysis of CAZy genes revealed that the relative abundance of GH5, CBM86, CBM35, GH43, and CBM6 (LDA ≥ 2.0; P ≤ 0.05) was greater in NegRFI bulls, whereas the relative abundance of GH13 and GT2 was greater in PosRFI bulls (LDA ≥ 2.0; P ≤ 0.05) ( Figure 7 ).

While the beef cattle production industry provides food and nutrition globally, feed efficiency is economically important for creating sustainable practices (de Ondarza and Tricarico, 2017; Stewart et al., 2018). Within ruminants, the influence of the rumen microbiome is crucial in facilitating feed intake, metabolism of carbohydrates, and energy utilization through the production of VFAs and nitrogen (Wang et al., 2013; Shabat et al., 2016; Wang et al., 2020). In this study, the quantitative gene abundance was greater in bulls identified as NegRFI. While little has been reported regarding a tendency of different gene richness in the rumen of cattle selected for feed efficiency, a study by Lima et al. (2019) reported a substantial link between rumen microbial abundances and feed intake, associating efficient cattle with greater gene abundance in the rumen while consuming less feed. Our research indicates an association between ruminal gene richness and genetic feed efficiency.

In cattle, the rumen serves as a station for diet fermentation, as it harbors a diverse microbial community that contributes to the breakdown of complex plant materials into absorbable nutrients, mediated by several mechanisms. These processes all contribute to cattle metabolic homeostasis (Sakata and Tamate, 1979; Lima et al., 2019). Relative to gene abundance, the expression of genes related to feed metabolism within the rumen microbiome significantly influences nutrient conversion and energy acquisition, ultimately impacting overall metabolic efficiency and production outcomes in cattle. The results of this study indicate greater gene richness in NegRFI bulls, which could suggest a greater ability to metabolize feed, requiring quantitatively less feed to exhibit the same production gain as their less efficient counterparts. However, conclusive results surrounding microbial diversity and gene richness in feed-efficient cattle are far from finalized. In a study conducted by Shabat et al. (2016), cattle identified as feed efficient revealed a less diverse and genetically rich microbial community. Nevertheless, while these cattle were reported as such, they were accompanied by a significantly higher dominance of certain taxa involved in energy metabolism and VFA production at the genus and species level, which complements our results. While the results surrounding microbial community and gene richness are still widely variable, feed-efficient cattle consistently possess dominant taxa that contribute to energy metabolism and production (Myer et al., 2015; Shabat et al., 2016; Auffret et al., 2020).

Cattle possess a diverse microbial community in their rumen, characterized by dominant phylogenetic taxa such as Firmicutes, Fibrobacterota, Bacteroidetes, and Proteobacteria, which are crucial for efficient digestion and nutrient metabolism (Jami and Mizrahi, 2012; Myer et al., 2015). While a core microbiome exists in the rumen, research has exhibited altered rumen microbial profiles in cattle selected for feed efficiency (Li et al., 2019b; Idowu et al., 2023). In our study, the microbial communities between PosRFI and NegRFI bulls were different and can be explained by the greater gene abundance in NegRFI bulls. Additionally, the relative abundance of 35 species belonging to taxa such as Fibrobacter, Lentisphaerae, Bacteroidetes, and Treponema were found to be different between the two groups of bulls.

The relative abundance of the genus Treponema and its species T. C6A8, T. JC4, and T. porcinum were greater in bulls identified as NegRFI, compared to PosRFI. Many Treponema species have been reported within the rumen of cattle and play a crucial role in the complex microbial ecosystem of the rumen; furthermore contributing to the efficient breakdown of fibrous plant materials and the production of energy-rich VFAs for the host animal (Radolf, 1996; Nordhoff et al., 2005; Rosewarne et al., 2012). The relative abundance of Bacteroidetes bacterium sp. Adurb.BinA104, Candidatus Endolissoclinum, and Lentisphaerae bacterium sp. GWF2528 were greater in NegRFI bulls compared to PosRFI bulls. These bacteria are classified as symbiotic proteobacteria and their phylogeny has previously been identified within the rumen of cattle divergent in feed efficiency (Kwan et al., 2012; Myer et al., 2015; Reis et al., 2023). Symbiotic proteobacteria contribute to ruminal fermentation through hydrolysis of complex polysaccharides, contributing to the formation and fermentation of biofilms, and have been previously reported to be involved in nucleotide, carbohydrate, and nitrogen metabolism (Hart et al., 2018; Zhang et al., 2021; Hernández et al., 2022). Schwartzia sp. succinovorans is a gram-negative bacterium that can ferment succinate to propionate in the rumen and was found in greater abundance in NegRFI bulls compared to PosRFI bulls (van Gylswyk et al., 1997). In ruminants, the production of VFAs is crucial in anaerobic digestion processes, serving as substrates for microbial metabolism (Bergman, 1990). Propionate and succinate use acetate: succinate transferases, which feed into the tricarboxylic acid (TCA) cycle, thereby potentially enhancing energy generation and microbial growth efficiency in the rumen (Kwong et al., 2017).

The relative abundance of species such as Fibrobacter UWB5, F. UWT2, and F. UWB were greater in NegRFI bulls compared to PosRFI bulls. Additionally, the genus Fibrobacter and the class Fibrobacteria were identified as the most enriched taxa in NegRFI Angus bulls. Fibrobacteria, particularly represented by the genus Fibrobacter and its species F. UWB5, F. UWT2, and F. UWB, are major cellulose-degrading bacteria. They break down plant material, including cellulose and hemicellulose, through the secretion of fibrolytic enzymes (Suen et al., 2011; Cammack et al., 2018). This enzymatic activity facilitates the breakdown of complex polysaccharides into simpler compounds, enhancing the accessibility of nutrients for both the microbial community and the host animal (Comtet-Marre et al., 2017; Jewell et al., 2013). A study by Neumann and Suen (2018) highlighted the importance of herbivore-associated Fibrobacter spp. within the gut microbiome of ruminant livestock, noting that species F. UWB5, F. UWT2, and F. UWB are associated with lignocellulose degradation.

Lignocellulose, an important building block in plant cell walls, is among the most abundant natural feedstocks in agriculture (Vázquez-Vuelvas et al., 2021). Cattle heavily rely on microbial digestion of plant lignocellulosic roughages, which make up major parts of their diets and consist mostly of lignin and three polysaccharides: cellulose, hemicellulose, and pectin (Naraian and Gautam, 2018; Gharechahi et al., 2023). The breakdown and conversion of complex plant material into polysaccharides within the rumen are suggested to support up to 70% of a ruminant animal’s daily energy requirements (Flint et al., 2008). This suggests that NegRFI bulls are able to break down and utilize feed more efficiently for increased energy production and metabolism, which likely explains their better feed efficiency.

The increased abundance of fiber-degrading species in the rumen of NegRFI bulls suggests an enhanced capacity to break down and utilize fibrous components of high-forage diets, compared to PosRFI bulls. This improved fiber degradation likely supports the ability of these bulls to grow at similar rates while consuming less feed compared to NegRFI bulls. Improved fiber degradative species in cattle selected for feed efficiency have previously been reported (McGovern et al., 2018; Elolimy et al., 2018; Liu et al., 2022). McGovern et al. (2018) employed 16S rRNA gene sequencing to further explore the relationship between the cellulolytic rumen microbiome and feed efficiency in bulls. Their analysis revealed a higher abundance of operational taxonomic units (OTUs) associated with fiber-degrading phyla in efficient RFI bulls, suggesting that the rumen bacterial community was adapted for enhanced fiber degradation, which may influence the feed efficiency phenotype. In a separate study, Elolimy et al. (2018) reported a higher abundance of cellulose-degrading bacteria in the rumen of low RFI beef steers, suggesting their role in improving feed digestibility and energy production in feed-efficient steers. These findings collectively indicate the importance of ruminal fiber-degrading communities in enhancing feed efficiency in cattle, suggesting that selecting for these microbial traits may contribute to the development of more feed-efficient herds fed high-forage diets.

Carbohydrate-active enzymes are essential for breaking down and utilizing complex carbohydrates in cattle feed (Abbott et al., 2018). In this study, we observed differences in the levels of certain enzymes, including glycoside hydrolases, glycoside transferases, and carbohydrate-binding modules, between two groups of Angus bulls. These enzymes are crucial components of the rumen microbiome and have been extensively studied for their biochemical and enzymatic properties (Lombard et al., 2014; Amin et al., 2021). Glycoside hydrolases, also known as glycosidases or glycosyl hydrolases, break glycosidic bonds in complex carbohydrates and play roles in antibacterial defense and pathogenesis in living organisms (Davies and Henrissat, 1995; Sjögren and Collin, 2014).

In our study, we found that the levels of GH5 and GH43 were higher in NegRFI bulls, while GH13 was more abundant in PosRFI bulls. GH5 proteins are one of the largest families and are commonly found in various environments (Henrissat et al., 1989; Dai et al., 2012). GH43 enzymes are known for their ability to break down biomass and are important in degrading plant cell walls (Cantarel et al., 2009; Mewis et al., 2016). Both GH5 and GH43 are involved in breaking down mannan bonds, a type of hemicellulose found in plant cell walls, into usable sugars (Aspeborg et al., 2012; Wang et al., 2013). This breakdown increases fermentable substrates available to rumen microorganisms, potentially increasing VFA production and microbial protein availability (Wang et al., 2019; Hua et al., 2022). The higher levels of GH5 enzymes in NegRFI bulls suggest increased energy substrates and nutrient availability, enhancing ruminal fermentation efficiency and nutrient utilization.

The GH13 family, also known as the α-amylase family, is responsible for rapidly converting starch into glucose and maltose (Bourne and Henrissat, 2001; Stam et al., 2006). Interestingly, despite being fed the same diet as NegRFI bulls, PosRFI bulls had higher levels of GH13, indicating differences in carbohydrate utilization mechanisms between the groups (Lin et al., 2019; Zhao et al., 2022). The relative abundance of GT2 was higher in PosRFI bulls. These enzymes are involved in the biosynthesis of structural molecules from sugar moieties, affecting the overall utilization of carbohydrates for energy substrates and production (Keenleyside et al., 2001).

The relative abundance of CBM86, CBM35, and CBM6 was higher in NegRFI bulls compared to PosRFI bulls. Carbohydrate-binding modules influence carbohydrate binding, degradation, and utilization by facilitating GH binding to carbohydrate structures (Boraston et al., 2004; Ficko-Blean and Boraston, 2006). The greater relative abundance of these enzymes in NegRFI bulls suggests increased enzymatic recognition and hydrolysis, potentially providing additional energy substrates to support better nutrient utilization.