In vivo studies involving growing sheep were conducted to investigate the impact of replacing concentrate with OBW on nutrient intake, utilization, enteric CH₄ emissions, and the compositional and functional abundances of rumen microbiota, adhering to the ethical guidelines set forth by the institute for animal handling. Approval was secured from the Institute Animal Ethics Committee under reference number NIANP/IAEC/1/2024/6 for conducting experiments on sheep, and all procedures were executed in strict adherence to the established guidelines of the IAEC.

Twenty-one Bannur male sheep, approximately 5 months old, having an average body weight of 17.7 ± 0.117 kg, were purchased from the Mandya district of Karnataka, India, and transported to the Experimental Livestock Farm of the National Institute of Animal Nutrition and Physiology in Bangalore (12°56′55″N 77°36′25″E). The growing sheep were randomly assigned into three groups of seven individuals each in a completely randomized design and accommodated in a well-ventilated shed arranged in a tail-to-tail configuration. Each sheep was assigned to an individual pen equipped for feeding and watering continuously over a 24-h period. The animals received deworming treatment with Ivermectin at a dosage of 200 mcg per kg of body weight. The growing sheep were fed according to the ICAR feeding standard (ICAR, 2013), ensuring their dry matter, energy, and protein needs were fulfilled. The diet consisted of equal parts finger millet (Eleusine coracana) straw and a concentrate. The concentrate was formulated by combining the ground components, including maize grain (320 g/kg), soybean meal (130 g/kg), groundnut cake (120 g/kg), wheat bran (400 g/kg), mineral mixture (20 g/kg), and salt (10 g/kg). In the control group (C), the sheep were offered a basal diet that consisted of equal parts of finger millet straw and concentrate, without OBW. In test group I (T₁), the animals were offered a diet consisting of 50 parts finger millet straw, 30 parts concentrate, and the remaining 20 parts OBW. In test group II (T₂), the animals offered 50 parts of finger millet straw, 20 parts of concentrate, and the remaining 30 parts comprised OBW. The concentrate without OBW (group C ) or with OBW (groups T₁ and T₂) was offered (Figure 1) in the morning at 09.00 h, whereas the finger millet straw was offered separately at 17.00 h. Daily the leftover of concentrate and finger millet straw was recorded and the intake was calculated. Based on the intake of individual animal and to ensure equal ratio of finger millet straw and concentrate, an additional 10% allowance was added to the daily offering of each so that the individual animal has ad libitum access to the feed. All the animals under different groups were adapted to the basal diet and local environment for 21 days before the onset of actual experiment.

The fresh OBW was collected each day from the SVP Farm in in Thattaguppe, Bangalore, India, and transported to the institute in the early morning hours. Three animal studies were carried out over 90 days to find out how replacing concentrate with OBW (w/w) at two different levels affected nutrient utilization, CH4 levels, and growth performance.

Every day, fresh OBW samples from the brewery were tested for aflatoxin B1 using an indirect competitive ELISA kit (PerkinElmer, Cat. No. SC0021) according to the instructions provided by the manufacturer. A total of 5 g of fresh OBW sample were placed in a 100 mL flask, followed by the addition of 25 mL of 60% methanol, and then vortexed for 5 min. Subsequently, the solution was placed into a tube and subjected to centrifugation at 4,000 × g for duration of 5 min. The supernatant underwent filtration using Whatman filter paper 1. Then, 1 mL of the filtered liquid was mixed with 4 mL of deionized water for 5 s, and 50 μL of this mixture was set aside for the ELISA test. The aflatoxin contamination in OBW was examined over a period of 3 days during aerobic storage in a well-ventilated covered shed.

The growing sheep underwent a 7-days metabolic trial after a 60-day preliminary feeding. Throughout the metabolic trial, the daily allowance of concentrate without (C) or with OBW (T₁ and T₂) as well as finger millet straw given to the individual sheep were recorded and the representative samples of feed offered were collected and dried in a hot air oven at 100°C for 24 h for the chemical composition analysis including energy content. The concentrate refusal for the individual sheep was recorded daily at 16.30 h, whereas the refusal of finger millet straw for the sheep individually was recorded next day at 08.30 h. The representative samples of refusals were collected and dried in a hot air oven at 100°C for 72 h. Based on the actual intake of concentrate and finger millet straw, the actual ratio of concentrate-roughage was calculated for the individual animal. The dung voided was daily collected throughout the metabolic trial in the fecal bags tied to the individual sheep, while the total urine was collected in a tightly screwed glass bottle placed under the metabolic cages and connected through a tube to the urine bags tied to the sheep. The total dung voided was weighed next day at 08.30 h for the individual sheep and the aliquots for dry matter (DM), N and energy estimation were collected, preserved and processed in accordance to Krishna and Ranjhan (1980). Daily, urine volume was measured and freeze dried for GE and N estimation (Morris et al., 2021). On the completion of metabolic trial, the dried samples were ground in a Cyclotec mill (CT293, FOSS, India) and stored for further analysis. The proximate components in the feed, refusals, and dung samples were analyzed in accordance with standard procedures. The crude protein (CP) was determined through the analysis of nitrogen content using the automatic nitrogen analyzer (VAPODEST 450, Gerhardt, Cäsariusstraße, Germany), followed by multiplication by 6.25. The ash content was determined through the incineration of a two-gram sample in a muffle furnace at 550°C for 4 h. In contrast, the organic matter (OM) was derived by calculating the difference between the dried weight of the sample and the ash content. The ether extract (EE) in the samples was determined according to AOAC (2005) utilizing a Soxtherm instrument (Gerhardt, Germany). The determination of fiber fractions in the samples, including neutral detergent fiber (NDF) and acid detergent fiber (ADF), was carried out following the methodology of Van Soest et al. (1991) using an automatic fiber analyzer (Fibertherm FT12, Gerhardt, Cäsariusstraße, Germany).

Along with the digestibility study, the CH4 measurement was carried out in growing sheep using the sulfur hexafluoride (SF6) tracer technique (Berndt et al., 2014). In summary, the brass permeation tubes equipped with a Swagelok nut served as a source for the SF6 gas. The permeation tubes were filled with SF6 (99.9%) in liquid nitrogen and calibrated at 39°C by measuring tube weight weekly over a period of 12 weeks. The release rate (mg/day) was calculated using linear regression, taking into account the change in tube weight throughout the 12-week calibration period. Tubes with a R² value of 0.99 or more were chosen for use in the experiment. Ten days before the CH4 measurement, the calibrated permeation tubes were placed in the rumen of sheep. The mean release rate of SF6 from the permeation tubes was 2.87 ± 0.128 mg/day. The vacuumed PVC canisters (>95 kPa) were fastened individually to collect the breath samples from sheep over a full feeding cycle of 24 h. Each day, the vacuumed PVC canisters were tied at 09:15 and removed next day from the animal after collecting the breath samples for the complete feeding cycle. Throughout the experiment, we obtained a minimum of five successful breath samples from each individual sheep. The N₂-diluted breath gas samples were withdrawn using an airtight Hamilton glass syringe to analyze the sample in the gas chromatograph (GC 2010 plus, Shimadzu, Japan) equipped with a flame ionization detector (FID) and an electron capture detector (ECD) for the analysis of CH4 and SF6 gases, respectively. The similar conditions for gas chromatography were followed as described previously (Malik et al., 2021; Thirumalaisamy et al., 2022). The concentrations of CH4 (ppm) and SF6 (ppt) were determined following the methodology outlined by Lassey et al. (2014), with minor modifications for local elevation and atmospheric pressure.

G_S indicated the levels of CH4 (ppm) or SF6 (ppt) under atmospheric pressure of 90 kPa and at an elevation of 920 m. τ_f (kPa) indicated the final vacuum in the canister following N₂ dilution, τ_s (kPa) denoted the post-sampling vacuum in the canister, while τ_e represented the vacuum in the evacuated canister. G_A referred to the concentration of CH4 (ppm) or SF6 (ppt) in the samples analyzed by the gas chromatograph.

Daily enteric CH4 emissions was calculated using the equation of Moate et al. (2014) and expressed as gram per day.

R_CH4 indicated the CH4 emission (g/day); R_SF6 denoted the SF6 release rate from the tubes (mg/day); [CH4]_M – [CH44]_BG referred to the CH4 concentration in the sample and background; [SF6]_M – [SF6]_BG represented the SF6 concentration in the sample and background; MW_CH4 and MW_SF6 signified the molecular mass of CH4 and SF6, respectively.

The DM and OM intake, as well as their digestibility in the individual sheep during the experiment, were taken into consideration and the CH₄ emissions was expressed as g/100 g DM intake, g/100 g OM intake, g/100 g digestible DM intake and g/100 g digestible OM intake.

The gross energy (GE) of feed, refusals, fecal and urine samples was determined using a bomb calorimeter (RSB 7, Rajdhani Scientific Instrument, New Delhi, India). In brief, around 0.5 g of the ground dried sample was weighed and converted into pellets prior to being placed into the bomb crucible. The GE in urine sample was determined by drying 4 g of sample in a bomb capsule at 60°C (Morris et al., 2021). The GE was determined through the incineration of a sample in a sealed oxygen-rich environment, with the temperature raised during the combustion being recorded to ascertain the GE content. The GE was quantified in megajoules per kilogram of dry matter (MJ/kg DM). Partitioning of energy was determined according to Da Fonseca et al. (2019). In brief, the GE intake was calculated by multiplying DM intake with the GE of feed, whereas the energy lost through feces (FE) and urine (UE) was determined by taking the energy content of feces and urine into account and multiplying with the total quantity of dung and total volume of urine, respectively. Energy loss in CH4 was estimated by taking daily enteric CH4 emission and its energy value i.e., 55 MJ/L (World Nuclear Association, 2020) into account. The digestible energy (DE) was determined by subtracting the FE loss from the GE intake and expressed as MJ/day. Further, the metabolizable energy (ME) was calculated by subtracting the total UE and the energy lost through daily enteric CH4 emissions. The ME intake was expressed in MJ/day.

At the completion of the trial, rumen fluid samples were collected 3 h after feeding from five sheep in each group selected randomly with the help of a stomach tube. About 45 mL of the rumen fluid, which included both fluid and solid components, was collected and equally divided into three subsets of 15 mL each. The first subset was set aside to extract metagenomic RNA, while the second and third subsets were used to measure ammonia-volatile fatty acid (VFA) and count protozoa, respectively. In the first subset for RNA isolation, the rumen digesta, which had both liquid and solid parts, was used without filtering it. The liquid part from the second subset, which was separated by spinning at 11,300 × g and 4°C for 15 min, was kept for determining ammonia-nitrogen (N) and VFA. To preserve the supernatant for determining ammonia-N, 2–3 drops of saturated HgCl₂ were added, and for estimating VFA, 25% metaphosphoric acid was mixed and stored at −20°C until processing. The third subset of the digesta was used for the enumeration of ruminal protozoa.

The preserved supernatant samples were thawed at room temperature for the determination of ammonia-N and VFA. The concentrations of individual VFAs were estimated according to the methodology outlined by Filípek and Dvorák (2009) using a gas chromatograph (Agilent 7890B, Santa Clara, USA) with the functional conditions described previously (Malik et al., 2021; Thirumalaisamy et al., 2022). The concentration of ammonia-N in the supernatant was measured following the method of Conway (1957). In brief, 1 mL of boric acid was placed into the inner chamber, while an equivalent volume of Na₂CO3 was added to the outer chamber of the Conway disk. Subsequently, 1 mL of supernatant was put in the outer chamber opposite from the Na₂CO3, and the disk was covered with a lid and allowed to remain undisturbed at room temperature for 2 h, after which titration was performed using 0.01 N H₂SO4. The concentration of ammonia-N was determined using the following formula and expressed in mg/dL.

The third subset of the rumen fluid collected from sheep was used for protozoal enumeration by combining equal volumes of the rumen fluid with formal saline. The mixed samples were allowed to remain undisturbed at room temperature overnight. The morphological characterization of protozoa was carried out following the methods outlined by Hungate (1966), while enumeration was performed using a phase-contrast microscope (Nikon Eclipse, Japan) at a 10 × objective, in accordance with the procedures of Kamra and Agarwal (2003).

Throughout the experimental period, the body weight of each sheep was recorded on a weekly basis. Feed intake was recorded daily, and the feed allowance was adjusted weekly according to body weight. The average daily gain (ADG, g/day) for individual sheep across all groups was determined by calculating the difference between final and initial body weights over a 90-day period, and to obtain the ADG, the result was divided by the total number of days. The ADG was expressed as g per day. The Feed conversion ratio (FCR) was calculated considering total feed consumed (kg, DM basis) by the individual sheep and dividing by the body weight gain (kg) of the individual animal over the entire period. The mean FCR in different groups was calculated based on the FCR of all seven sheep within the group.

To isolate RNA, the rumen digesta was first pelleted at a low speed of 10,000 × g for 5 min, then resuspended in 100 μL of RNAlater stabilizing solution (GCC Biotech, India) and preserved until further processing (Comtet-Marre et al., 2017). The RNA was isolated from 0.25 g preserved digesta sample using the RNeasy PowerMicrobiome Kit (Qiagen). All procedures were carefully followed in accordance with the manufacturer's instructions. The purity and concentration of RNA were determined using the Qubit 4.0 fluorometer (Invitrogen, USA) with the Qubit RNA BR Assay Kit (ThermoFisher Scientific, USA). RNA samples were sent for metatranscriptome sequencing using the NovaSeq 6000 (Illumina Inc., USA) to an external facility (Eurofin Genomics India Pvt. Ltd, Bangalore, India). The paired-end sequencing libraries were generated employing the Illumina TruSeq RNA library preparation kit, and the ligated products undergo size selection using AMPureXP beads. The product was amplified using the index primer according to the instructions provided in the kit protocol. The libraries enriched through PCR were subjected to purification using AMPureXP beads and were subsequently analyzed on the Agilent TapeStation 4200. The paired-end libraries were prepared for loading onto the Illumina platform to facilitate cluster generation and sequencing.

The raw metatranscriptome data was used for exploring the compositional abundances and functional analysis. The rumen metatranscriptome data containing total RNA obtained after quality filtration and removal of host contamination, were used for assessing the compositional abundances. In brief, the raw metatranscriptome data were processed to remove adapter contamination, filter low-quality reads, and correct misrepresented bases using Fastp v 0.23.2 (Chen, 2023). Subsequently, removal of the host contamination from the reads was performed using Hisat2 v2.2.1 (Kim et al., 2019), and the host-contaminated reads were mapped against the sheep genome Oar_v4.0 (RefSeq assembly accession: GCF_000298735.2), the unmapped read files were considered as cleaned reads. The processed cleaned paired and single-end reads were uploaded at BV-BRC version 3.30.19 for the microbial taxonomic affiliation by K-mer using Kraken2 database (Wood et al., 2019). The Kraken2 output files were exported to the Pavian v 0.8.4 (Breitwieser and Salzberg, 2020) to obtain microbial abundances at different taxonomic levels. The metatranscriptome data at different taxonomic levels obtained from the Pavian were normalized by Total Sum Scaling (TSS) and count filters at a cutoff value of minimum four reads with 80% prevalence in MicrobiomeAnalyst 2.0 (Lu et al., 2023).

For functional analysis, from the clean reads, the rRNA was removed using SortMeRNA (v4.3.6, Kopylova et al., 2012). The contigs were built by using rRNA depleted reads in MEGAHIT (v1.2.9; Li et al., 2015) at a length cutoff of 300 bp. The contigs were then used to predict the potential protein coding regions in Prodigal (v2.6.3, Hyatt et al., 2010). Thereafter, the non-redundant open reading frame (ORF) Catalog was prepared by clustering the amino acid sequences at an identity threshold of 95% using CD-HIT (v4.8.1, Li and Godzik, 2006). The non-redundant amino acid sequences were assigned to the KEGG orthologs (KO) and clusters of orthologous genes (COG) in GHOSTX (Kanehisa et al., 2017) and COG classifier (v2.0.0, Shimoyama, 2022; Galperin et al., 2021), respectively. The coverage of predicted unigene ORFs was determined using RNASeq tool in the CLC genomics workbench (v21.0.2; Qiagen, USA) and the transcripts per million (TPM) were added to the corresponding KOs and COG.

Data was evaluated for the outliers using the identify outliers feature in GraphPad Prism version 10.4.0. The outliers were identified by using the animal traits data such as intake, digestibility, growth, fermentation and energy partitioning were evaluated for gaussian distribution through the Kolmogorov–Smirnov test at a 5% significance level, using GraphPad Prism version 10.4.0 (GraphPad Software, San Diego, CA, USA). The analysis of the data was carried out using a One-way ANOVA within the specified model.

Where Y_ij represents the j^th observation (j = 1, 2,… 0.0.7) on the i^th group (i = 1, 2, 3). μ was the common effect of the experiment, τ_i represents the i^th group effect, and ∑_ij represents the random error due to the j^th observation of the i^th group. The mean of each group was compared with the mean of every other group. The significance between the groups was ascertained using Tukey post-hoc analysis at a family-wise alpha threshold and confidence level of 0.05 in GraphPad prism version 10.4.0.

The significance among the groups was established using non-parametric Kruskal–Wallis rank sum tests, followed by the Dunn post-hoc test (Miles et al., 2022) in R Studio. The alpha diversity of the metatranscriptome data was determined using the Shannon index, while the beta diversity was estimated using the Bray-Curtis dissimilarity index.

The differential expression in the functionality of rumen microbiota between the groups was checked by DESeq2 (Love et al., 2014) and the significance was ascertained at a FDR value cutoff of 0.05.

Chemical composition (g/kg) of feed constituents such as finger millet straw, concentrate, OBW and diet is presented in Table 1.

Results from the digestibility trial indicated a significantly (P < 0.05) lower DM intake (g/day) in group T₂ as compared to C and T₁ (Table 2). Despite the similar DM intake, there was a significant difference (P < 0.05) in the digestible DM intake (g/day) between C and T₁ groups. The actual ratio of concentrate-roughage in DM intake deviated from the ratio of offering (50:50) and it was recorded 65:35, 63:37, and 63:37 in C, T₁ and T₂ groups, respectively. The organic matter intake (g/day) and digestible organic matter intake (g/day) in the growing sheep was significantly (P < 0.05) lower in T₂ as compared to C and T₁ groups. The CP intake (g/day) was significantly (P < 0.05) different between the group T₁ and T₂, whereas the CP intake between C and T₁, as well as C and T₂, was comparable. In contrary, the digestible CP intake (g/day) was significantly (P < 0.05) higher in T₂ as compared to C, while there was no significant (P > 0.05) difference between group C and T₁. The NDF intake was significantly (P < 0.05) higher in group T₁ as compared to group C, whereas the ADF intake in T₁ was significantly (P < 0.05) than that was in group C and T₂. The adjustment of NDF and ADF intake data to the corresponding digestibility revealed the significantly (P < 0.05) lower intake in group T₂ (Figure 1).

Data showed a significant (P < 0.05) decrease in daily enteric CH4 emissions (g/day) in group T₁ and T₂ as compared to group C. The replacement of concentrate with OBW at 20 and 30% of diet resulted into a significant (P < 0.05) decrease of 13.4 and 15.4% in daily enteric CH4 emissions in T₁ and T₂ groups, respectively (Figure 2a). These findings indicated that the replacement of concentrate with OBW beyond 20% of diet did not lead to any further significant (P > 0.05) decrease in daily enteric CH4 emissions (g/day). Adjustment of CH4 emissions data to per 100 g of DM intake in growing sheep indicated a significant (P < 0.05) decrease in CH4 emissions in T₁ as compared to group C and T₂ (Figure 2b). However, the CH4 emissions between C and T₂ did not reveal any significant (P > 0.05) difference (4.29 vs. 4.14 g/100 g DM intake). Adjustment of CH₄ emission data to per 100 g of OM intake revealed a remarkable (P < 0.05) reduction of about 13.2% in CH4 emissions was recorded in T₁ as compared to group C (Figure 2c). On the contrary, the CH4 emissions between group C and T₂ were comparable (4.60 vs. 4.45 g/100 g OM intake).

Results from the study demonstrated significantly (P < 0.05) lower CH4 emissions in group C and T₁ as compared to T₂ when the emissions among the groups was compared based on per 100 g of digestible DM intake (Figure 3a) and per 100 g of digestible OM intake (Figure 3b). A reduction of about 18 and 16% in CH4 emissions was observed in group C and T₁ as compared to T₂.

The GE intake (MJ/day) was higher (P < 0.05) in C and T₁ groups as compared to T₂ (Table 3). Similarly, the intake of DE was also significantly (P < 0.05) higher in C and T₁ groups as compared to T₂. The loss of energy in feces was significantly (P < 0.05) higher in group T₁ as compared to C, whereas the fecal energy loss between C and T₂ and T₁ and T₂ was comparable. However, the fecal energy loss (% of GE) was significantly (P < 0.05) higher in group T₂ as compared to the other two groups. The fecal energy loss (% of GE) was minimum in group C (Table 3), however, this group had the maximum loss of GE in the form of CH4 (14.8 vs. 12.5 vs. 13.9 %). On the other hand, the UE (MJ/day) loss was lowest (P < 0.05) in group C as compared to groups T₁ and T₂. Overall, the ME intake (MJ/day) was similar in group C and T₁, whereas sheep in T₂ had the lowest (P < 0.05) ME intake.

The ammonia-N (mg/dL) levels did not change (P > 0.05) when concentrate was replaced with two amounts of OBW in groups T₁ and T₂, and the levels were similar across the groups (Table 4). However, the total volatile fatty acid (TVFA) concentration (mmol) was significantly (P < 0.05) lower in group T₂ compared to C and T₁, whereas the TVFA concentration between C and T₁ was not different (P > 0.05). Similarly, the acetate and propionate concentrations (mmol) were also recorded as the lowest in group T₂. The concentration of other VFAs except isovalerate was not different (P > 0.05) among the groups (Table 4). These findings indicated that the replacement of concentrate with the OBW at the highest level, i.e., 30%, affected the VFA production adversely, whereas the replacement at the 20% level did not affect VFA production significantly (P > 0.05) compared to group C.

The total protozoa (cells × 10⁷/mL) were significantly (P < 0.05) decreased with the replacement of concentrate with OBW at the 30% level in T₂ as compared to group C. However, the total protozoal numbers between C–T₁ and T₁-T₂ were comparable (Figure 4a). Similarly, the numbers of Entodinimorphs (cells × 10⁷/mL) were decreased (P < 0.05) in T₂ as compared to the control, but there was no difference in the Entodinimorphs numbers between the other groups (Figure 4b). The Holotrichs were significantly (P < 0.05) lower in group T₂ as compared to groups C and T₁ (Figure 4c).

The initial and final body weight of the sheep showed no significant difference (P > 0.05) among the groups (Table 5). Similarly, the ADG (g) was also not significant (P > 0.05) between the C and test groups T₁ and T₂. There was a significant (P < 0.05) difference in the daily DMI (g/day) and total DMI for 90 days between the group T₁ and T₂. The FCR in C, T₁ and T₂ groups was 7.61, 6.87, and 6.77; however, there was no significant (P > 0.05) difference in the FCR among the groups.

The alpha diversity measured with the Shannon index showed no significant difference (P > 0.05, Figure 5a), suggesting that the species richness and evenness across the groups were relatively consistent. Beta diversity (Figure 5b) analyzed the differences in microbial composition between the groups and showed no significant differences (P > 0.05). In this study, a total of 432 million transcriptomic raw reads were generated, with an average of 33.2 million per sample (Supplementary File 1). Approximately 4.32% of reads were excluded during quality filtration due to poor quality, short length (< 150 bp), and adapter contamination. Additionally, 0.145% of the metatranscriptomics reads were excluded because of host contamination. Subsequently, the high-quality paired and single-end metatranscriptomics reads were analyzed for taxonomical abundances.

In the study, across the groups, a total of 22 microbial phyla were identified (Supplementary File 1), where Bacteroidota, Bacillota, and Pseudomonadota were the largest microbial phyla, constituting 90% of the ruminal microbiota (Figure 6a). Among these, the Bacteroidota was the most abundant phylum, constituting more than 40% of the total rumen microbiota. The compositional abundance of Bacteroidota was significantly (P < 0.05) decreased with the graded replacement of concentrate with OBW in the T₁ and T₂ groups. On the contrary, the compositional abundance of the third largest phylum, Pseudomonadota, was increased (P < 0.05) in groups T₁ and T₂ as compared to group C, whereas the abundance of Bacillota was comparable (P > 0.05) among the groups. Similarly, the abundances of other phyla, including Euryarchaeota, were not different among the groups.

At the order level, a total of 101 microbial orders were identified, where microbes affiliated with Bacteroidales constituted ~50% of the ruminal microbiota (Figure 6b). The Eubacteriales was the second largest order of the ruminal microbes; however, their abundance was approximately 3–4 times less than the Bacteroidales. The compositional abundance of Coriobacteriales in the T₂ group was significantly (P > 0.05) greater than the control group, whereas there was no difference in the abundance between groups C and T₁. The abundance of Synechococcales was increased (P < 0.05) with the graded replacement of concentrate with OBW in groups T₁ and T₂. Methanogens affiliated with the order Methanobacteriales constituted the largest phylum of the archaea; however, their abundances were not different (P > 0.05) among the groups.

A total of 344 microbial genera were present in the rumen. Among these genera, Prevotella was the largest, constituting approximately 50% of the total rumen microbiota. Fibrobacter and Pseudomonas were second and third most abundant microbial genera. However, the abundances of the three largest genera were similar among the groups (Figure 6c). Among the archaea, Methanobrevibacter was the most abundant genus, which constituted 2.5%−3.0% of the total microbiota (Figure 6c); nevertheless, their abundances were also not different (P > 0.05) among the groups. Replacing concentrate with OBW significantly increased the abundances of Olsenella, Pseudobutyrivibrio, Marvinbryantia and Parafannyhessea. Conversely, the OBW in the test groups led to a decrease in the abundances of Eubacterium and Thermoclostridium.

After removal of host contamination and rRNA, on an average 9,642,117 reads of mRNA per sample were retained from each metatranscriptome library (Supplementary File 1). After de novo assembly of mRNA reads, it generated an average of 69,789 contigs (N50 length 862 ± 59 bp). On an average 56.3% ± 2.0% and 30.4% ± 1.2% of mRNA was mapped to the COG and KEGG reference annotated genes.

About 33.5%, 29.3% and 30.6% of non-redundant genes of the total identified COG were assigned to the function cellular processes and signaling in groups C, T₁, and T₂, respectively (Figure 7). In cellular processes and signaling category of COG functions, the cell motility and intracellular trafficking, secretion and vascular transport were significantly different between the groups. The cell motility was significantly decreased in T₁ (P = 0.04) and T₂ (P = 0.01) groups as compared to C. Similarly, there was a decrease in intracellular trafficking, secretion and vascular transport in T₁ (P = 0.02) as compared to C, whereas no significant difference (P = 0.07) was recorded between C and T₂. The translation, ribosomal structure and biogenesis function under information storage and processing category of COG was different between the groups (Figure 7). A significant decrease (P = 0.04) was recorded between the groups C and T₁. Similarly, there was a significant difference (P = 0.005) in the carbohydrate transport and metabolism function between the C and T₁ groups. Other COG functions were not found significantly different between the groups.

KEGG methane metabolism pathway (Figure 8) analysis revealed a significant difference in the EC1.5.98.1-K00319 (methylenetetrahydromethanopterin dehydrogenase), EC1.12.98.1-K00441 (coenzyme F420 hydrogenase), EC1.8.98.6, EC1.8.98.5, EC1.8.98.4, EC1.8.7.3-K03388 (heterodisulfide reductase subunit A2) and EC2.8.4.1-K00401 (methyl coenzyme M reductase sub unit beta). The expression of methylenetetrahydromethanopterin dehydrogenase was significantly decreased (P = 0.03) in T₂ as compared to C. Similarly, the expression of coenzyme F420 hydrogenase was also significantly lower (P = 0.004) in T₂ as compared to C. However, the expression of heterodisulfide reductase subunit A2 was significantly lower in both the T₁ and T₂ groups as compared to C. The expression of methyl coenzyme M reductase sub unit beta was lower (P = 0.03) in T₂ as compared to C.

KEGG nitrogen metabolism pathway (Figure 9) analysis revealed a significant difference in the EC1.1.7.5.2 and EC1.7.99 assigned to K00370 (nitrate reductase) in the present study were expressed significantly lower in T₁ (P = 0.01) and T₂ (P = 0.03) groups as compared to C. Another EC which was significantly different in the present study was 1.4.1.4 assigned to K00262 (glutamate dehydrogenase) was significantly lower in both the test groups T₁ (P = 0.04) and T₂ (P = 0.04) as compared to C.