Preparation of BSG protein hydrolysates
The BSG protein hydrolysates used in the current study were prepared following alkali extraction, acid precipitation and enzyme hydrolysis as previously reported [11, 13]. In brief, BSG was milled through a 0.5-mm screen (Wiley Mill; standard model 4; Arthur Thomas Co., Philadelphia, PA, USA) solubilized in 0.1 M NaOH at 20% (w/v) rate, followed by continuous stirring at 350 rpm at 50oC for 2 h for protein extraction. The supernatant was then collected by centrifuging at 8,000 × g for 15 min at 20oC, passed through an ultrafiltration membrane to obtain the protein fraction > 1 kDa which was further precipitated by acid (adjusting pH to 3.5), centrifuged at 8,000 × g for 15 min at 20oC and then freeze dried. The obtained protein was dispersed in deionized water to reach a 5% (w/v) solution, hydrolyzed by 1% alcalase and flavourzyme at their optimum pH and temperatures (pH 8.0, 55oC and pH 6.6, 50oC, respectively). At the end of hydrolysis, all of the hydrolysate solutions were adjusted to pH 7.0, heated at 95oC for 5 min to deactivate the enzyme and centrifuged at 8,000 × g for 30 min to separate the solubilized peptides and amino acids from the non-soluble substrates. The obtained hydrolysates with alcalase and flavourzyme were referred to AlcH and FlaH, respectively. The protein content and sample degree of protein hydrolysis were determined as described [11, 13] and results were shown in Table 1. The antioxidant properties, including 1,1-diphenyl-2-picryl hydrazyl (DPPH) free radical scavenging activity, ferrous ion chelating activity and reducing power were measured using chemical methods (Supplemental material S1).
Experimental design and treatments
The experiment was a completely randomized design with four treatments assigned to sixteen fermentation vessels in two units of RUSITEC apparatus. Treatments were: 1) basal diet without hydrolysates or antibiotics (Control), 2) basal diet + 1% AlcH (AlcH), 3) basal diet + 1% FlaH (FlaH), and 4) basal diet + 33 mg monensin + 11 mg tylosin/kg diet DM (positive control; ANT). The basal diet contained 10% barley silage, 87% dry-rolled barley grain, and 3% vitamin and mineral supplement (DM basis; Table 2), The diet was prepared as total mixed ration, and grounded through a 4-mm sieve (Arthur Thomas Co., Philadelphia, PA, USA). Approximately 10 g (DM) of diet was weighed into nylon bags (10 × 20 cm; pore size of 50 μm, Ankom Technology Corp., Macedon, NY, USA). The protein hydrolysates and antibiotics were added to bags at the desired amount (DM basis) and manually mixed. The experiment was 15 d in duration, with 8 d for adaptation followed by 7 d for sampling and data collection.
Three ruminally fistulated Aberdeen Angus cross cows (average, 768 ± 95.1 kg BW) offered a high grain diet containing 8.2% barley silage, 89.2% dry rolled barley grain, and 2.6% vitamin and mineral supplement (DM basis) were used as rumen inoculum donor. Two hours after morning feeding, solid and liquid contents were collected from four locations within the rumen of each cow via rumen cannula. The contents were immediately filtered through four layers of cheesecloth, pooled (4 L per cow) and pH recorded before adding into the fermenters. The experimental protocols were reviewed and approved by the Lethbridge Research and Development Centre Animal Care Committee and the cows were handled in accordance with the guidelines of the Canadian Council on Animal Care .
The RUSITEC procedure was carried out using two units of RUSITEC apparatus, equipped with eight 920-mL anaerobic fermenters of each unit, as described previously [23, 24]. In brief, to initiate the fermentation each fermenter was filled with 200 mL of McDougall’s buffer  and 700 mL of prepared rumen inoculum. Then, one bag containing 20 g of prepared solid rumen digesta and one bag containing 10 g of experimental diet (DM basis) were placed in each fermenter at 0900 h on d 0. Fermenters were placed in a circulating water bath at 39°C for the duration of the incubation period. After 24 h of incubation, the bag containing solid rumen digesta was replaced with a bag containing experimental diet. Thereafter, a feed bag was replaced daily, so that each bag remained in each fermenter for 48 h. During the daily feed-bag exchange, the fermenter was flushed with N2 gas to maintain anaerobic condition in the fermenters. The artificial saliva was continuously infused into fermenters using a peristaltic pump (Model ISM 932D, Ismatec, Index Health and science GmbH, Wertheim, Germany) at a dilution rate of 2.9%/h. Effluent and fermentation gasses from each fermenter were collected, respectively, into a 2 L Erlenmeyer flask and a reusable 2 L gas-tight collection bag (CurityR; Conviden Ltd., Mansfield, MA, USA) and the volume recorded at the feed bag exchange.
Disappearance of DM, OM, CP, acid detergent fiber (ADF), neutral detergent fiber (NDF) and starch was measured with 48-h incubated feed bags from d 9 to 13 of the sampling period. Bags were withdrawn from fermenters and washed manually under running cold water until the water was clear and dried at 55°C for 48 h (AOAC, 2005; method 930.15) to determine DM disappearance. Thereafter, the bag residues were pooled over 5 d by fermenter and ground through a 1-mm sieve for DM, OM, NDF and ADF analysis. A portion of ground sample was further ground using a ball mill (Mixer Mill MM2000; Retsch, Haan, Germany) for total N and starch analysis. Disappearance of DM, OM, CP, NDF, ADF and starch was calculated as the differences between the amount of input and the amount remaining of each nutrient in the residues.
Gas production and dissolved gases
Fermentation gasses were measured every 24 h using a gas meter (Model DM3A, Alexander-Wright, London, England, UK) from d 9 to 15 of the sampling period. A sample (20 mL) was obtained from each bag once daily for determining gas profiles. On d 14 and 15, fermentation liquid was sampled in duplicate (35 mL each) from each fermenter to determine the concentration of dissolved H2 (dH2) and dissolved CH4 (dCH4). The sampling procedure was carried out using two syringes as reported . In brief, a 50 mL syringe containing 35 mL of fermentation liquids was connected to a 20 mL syringe filled with 5 mL of N2. The N2 was injected from the small syringe into the large syringe via a T tube and valve and the apparatus was vigorously shaken by hand. The entire gas phase was then transferred from the large syringe into the small one to determine the gas volume. Finally, the small syringe was removed from the T tube and 6 mL of gas was sampled for both gas and dissolved gas analyses using a Varian 4900 Gas Chromatograph (Agilent Technologies Canada Inc., Mississauga, ON, Canada).
The pH of fermentation fluid of each fermenter was measured daily using a pH meter (Orion model 260A, Fisher Scientific, Toronto, ON, Canada) at the time of feed-bag exchange. From d 9 to 13 of sampling period, effluent (5 mL) was sampled and preserved with 1 mL of 25% metaphosphoric acid for VFA analysis, with another 5 mL of effluent preserved with 1 mL of H2SO4 (1% vol/vol) for NH3-N analysis. All samples were well mixed and frozen at -20°C until analyzed. The production (mmol/d) of total VFA, individual VFA and NH3-N were determined based on daily effluent volume.
Microbial protein synthesis
Bacteria in the fermenters were labeled using 15N. From d 7 to 15, 0.3g/L (NH4)2SO4 in McDougall’s buﬀer was replaced with 0.3g/L 15N-enriched (NH4)2SO4 (Sigma Chemical Co., St. Louis, MO, USA; minimum 15N enrichment 10.01 atom%). From d 9 to 15, the effluent was preserved by adding 3 mL of a sodium azide solution (20%; wt/vol) to each effluent flask. On d 14 and 15, daily volume of effluent from each fermenter was recorded and 35 mL was sampled and centrifuged at 20,000 × g, 4°C for 30 min to isolate liquid-associated bacteria (LAB). The obtained pellets were washed with deionized water and centrifuged three times (20,000 × g, 30 min, 4°C), suspended in distilled water, freezing and lyophilisation for determination of N and 15N.
Feed particle-associated (FPA) and feed particle-bound (FPB) bacterial fractions were measured from 48-h feed residues. After 48-h of incubation, feed bags were squeezed to expel excess liquids, placed individually in a plastic bag with 20 mL of McDougall’s buffer and processed for 1 min using a Stomacher 400 Laboratory Blender (Seward Medical Ltd, London, UK). Then, the liquid from each feed bag was squeezed into a 50 mL centrifuge tube, the feed residues were washed twice with 10 mL of McDougall’s buffer in each wash, and the washed buffer pooled with the squeezed liquid to obtain the FPA bacterial fraction. The washed feed residues were considered as the FPB bacterial fraction. The obtained FPA samples were centrifuged at 500 × g, 4°C for 10 min to remove large feed particles, the supernatant centrifuged (20,000 × g, 30 min, 4 °C) to isolate bacterial pellets and further processed as described previously. Washed feed residues (FPB fraction) were dried at 55°C for 48 h, weighed for DM determination, and ball ground (MM400; Retsch Inc., Newtown, PA, USA) for N and 15N analysis. Total microbial protein synthesis was estimated as the sum of LAB, FPA and FPB.
The microbial communities of FPA samples were assessed through high-throughput sequencing. Total DNA was extracted from FPA samples (30 mg) using a QIAamp Fast DNA stool mini kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions. The quality and quantity of extracted DNA was measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), then the extracted DNA was stored at -20°C until sequencing. The V4 hypervariable region of the archaeal and bacterial 16S rRNA gene was amplified using the modified 515-F and 806-R primers, with the PCR conditions and sequencing steps carried out as previously described by . Briefly, the 16S rRNA gene amplicons were generated using a two-step PCR, and then the amplicons were subjected to Illumina paired-end library preparation, cluster generation, and sequenced on an Illumina MiSeq instrument (Illumina, Inc., San Diego, CA, USA).
The obtained 16S rRNA gene sequencing raw data were processed using QIME2  and the R-package DADA2 (Version 1.4) denoise method as described. In brief, after removing primer sequences and truncating both the forward and reverse reads at 225 bp, quality control was done for the reads using the QIME2, with chimeric sequences identified and removed. Then, the richness (number of OTUs) and diversity (Shannon index) were calculated and non-metric multidimensional scaling (NMDS) was performed based on Bray-Curtis similarity distances using R packages vegan (Version 2.4.4; ) and phyloseq (Version 1.20.0; ). Fold change of ruminal bacterial at genus level with a threshold of 5% was analyzed using R-package Deseq2 .
The chemical analysis of the feeds and feed residues were conducted in duplicate, and repeated when the CV for the replicate analysis was more than 5%. Analytical DM was measured by oven drying at 135°C for 2 h (AOAC, 2005; method 930.15)  and ash content was determined by combustion of samples at 550°C for 5 h (AOAC, 2005; method 942.05), with OM content calculated as the difference between 100 and the ash content. The concentration of NDF (ash-free) was determined using a VELP Fiber Digestion System (VELP Scientifica, Burlington, ON, Canada) using the method of Van Soest et al. , with heat stable α-amylase (Termamyl 120 L, Novo Nordisk Biochem, Franklinton, NC, USA) and sodium sulfite included; while ADF was determined according to AOAC (2005; method 973.18). Total N of feed and residue samples and 15N of LAB, FPA and FPB samples were analyzed using combustion analyzer (NA 2100, Carlo Erba Instruments, Milan, Italy), with CP calculated as total N × 6.25. Starch was determined by enzymatic hydrolysis of α-linked glucose polymers as reported previously . Concentration of VFA and NH3-N in the effluent was determined using a gas chromatograph (model 5890, Hewlett-Packard Lab, Palo Alto, CA, USA).
Data were analyzed in a completely randomized design using the MIXED procedure of SAS (Version 16.0.0, SAS Inst. Inc., Cary, NC, USA), with treatment considered as a fixed effect, day of sampling as repeated measures, and the fermenter and RUSITEC apparatus as random effects. For the repeated measures, various covariance structures were tested with the final structure chosen based on the lowest Akaike’s information criteria value. Results are reported as least squares means, which were compared using the Tukey correction for multiple comparisons. Significance among treatments was declared at P ≤ 0.05 and a trend at 0.05 < P ≤ 0.10 unless otherwise stated.