Napier grass and anthocyanin-rich black cane were grown and harvested at the goat and sheep research farm of Suranaree University of Technology (SUT), Nakhon Ratchasima, Thailand (14°52'49.1“N, 102°00'14.9”E, 243 m above sea level). The two grasses were grown between August 2017 and January 2018, during the monsoon season. After harvesting, the Napier grass and anthocyanin-rich black cane materials were chopped (2–3 cm using a crop cutter) and ensiled without wilting (8, 15). Both silages fermented well after 90 d of ensiling, as indicated by sensory evaluation and a pH value measured using a portable pH meter (Eutech Pc 700, Italia) when the drum containing the silage was opened. The chemical composition and fermentation characteristics of Napier grass silage (NS) and anthocyanin-rich black cane (AS) are provided in Supplementary Table S1.

Thirty-six healthy crossbred Thai-native Anglo-Nubian male goats with a body weight (BW) of 14.42 kg (± 0.6 SD) were selected among 500 goats reared by all collaborating farmers in Thailand. The goats were divided into two groups, housed individually, and randomly assigned to one of two experimental diets (Table 1): (1) Napier grass silage-based diet (a diet containing 50% NS, n = 18), or (2) anthocyanin-rich black cane silage-based diet (a diet containing 50% AS, n = 18). The remaining ingredients were identical to those used in a basal total mixed ration (TMR) diet prepared in accordance with the NRC requirements (16). An adjustment period of 14 d allowed the goats to become acclimated to routine feeding and to allow time for proper diet adjustment prior to the experiment. During the adjustment period, the animals were gradually fed experimental diets. Following the 14-d period, the animals were fed the experimental diets on a daily basis for 90 d. The animals were offered on average ~350 g DM of TMR/day, which was supplied in two equal portions at 0700 and 1,600. Throughout the experiment, the goats had free access to fresh water and a trace mineral salt block. On average, the animals consumed ~95% of the amount of feed supplied. Daily feed refusals were recorded and used to calculate DM intake (DMI). The BW of the goats was recorded to determine their growth performance throughout the routine feeding.

Samples of dietary ingredients, two experimental diets, and feed refusals were collected every 2 wk, dried at 55°C in an air oven, and ground in a Wiley Mill using (Retsch SM 100 mill; Retsch Gmbh, Haan, Germany) through a 1-mm screen. The dried samples were stored for chemical and nutrient content analysis. The contents of neutral detergent fiber (NDF) (with heat stable α-amylase), acid detergent fiber (ADF), and acid detergent lignin (ADL) were determined according to Van Soest et al. (17), with the use of a fully automated system (FibertecTM 8000, FOSS, Hilleroed, Denmark). The hemicellulose content of the feed was calculated as NDF minus ADF, whereas the cellulose content of the feed was calculated as ADF minus ADL. The nitrogen content of two experimental diets and diet refusals were analyzed with a Kjeltec™ 8400 fully automated Kjeldahl analyser (FOSS, Hilleroed, Denmark); 6.25 was used as the conversion factor to obtain crude protein (CP) values. Furthermore, the second subsample was extracted at 50°C for 24 h with 0.01 N hydrochloric acid (HCl) dissolved in an 80% methanol solution, and the supernatant was collected and transferred into a 50-ml volumetric flask for HPLC determination of anthocyanin composition (15, 18, 19). The chromatographic separation was carried out on a reversed-phase Zorbax SB-C18 column (3.5 μm particle size, i.d. 4.6 mm × 250 mm, Agilent Technologies, Santa Clara, CA, USA) for 65 min at 28°C. The injection volume was kept constant at 20 μl. The mobile phase was composed of acetonitrile of HPLC grade and 10% acetic acid (1:9). Chromatographic separation was accomplished at a flow rate of 0.8 ml/min using a binary gradient of (A) 10% acetic acid, 5% acetonitrile, and 1% phosphoric acid, and (B) acetonitrile. The absorption of the compounds that were tested and measured was evaluated using a photodiode array UV detector set to 520 nm. The Agilent OpenLAB CDS 1.8.1 system manager was used to make sure that the data was correct and to analyze chromatographic data. When there were differences in the nutritional values of the diet between the measured and initial values, the dietary formula was adjusted according to the measured nutrient values.

Blood samples were taken 2 h after morning feeding at 09:00 on the last feeding wk through jugular venipuncture into a single 10-ml heparin-containing vacuum tube. The samples were placed on ice for transfer to the laboratory. Subsamples (5 ml) were centrifuged at 5,000 × g for 20 min at 4°C (Sorvall Legend XT/XF Centrifuge Series, Thermo Fisher Scientific, Waltham, MA, USA), and collected plasma were frozen at −20°C until analysis. Plasma concentrations of urea, total protein, glucose, insulin, triglycerides, alanine transaminase, and aspartate aminotransferase were determined using an automated enzymatic colorimetric method on a Cobas Integra 400 Instrument (Roche Diagnostics, Mannheim, Germany), quadruplicate, as described by Xue et al. (20) and Niu et al. (21). Other subsamples were centrifugated at 3,000 × g for 15 min at 4°C. Plasma was then used to determine plasma antioxidant (TAC, TBARS, SOD, CAT, GSH-Px and GSH-Rx) using an automated enzymatic colorimetric method on a Microplate (96 wells, UV plate), quadruplicate, equipped into microreader (Varioskan-LUX multimode microplate reader, Thermo Scientific, USA) as explained in prior investigation (5).

All goats (n = 36) were killed at the end of the experiment. Then, samples (~500 ml, mixture of liquid and solid) were collected from the dorsal, central, and ventral regions of the rumen to form a composite sample, and then strained through four layers of cheesecloth to collect rumen fluid. Then, the pH of the rumen fluid was immediately determined using a handheld pH meter (Eutech Pc 700, Italia). The strained rumen fluid was subsequently put in a sterilized thermos flask and directly transported to the laboratory. Upon arrival at the laboratory, the filtered rumen fluid was divided in two aliquots. The first aliquot of the filtrates (5 ml) was treated with 0.5 ml of 50% (v/v) HCl, 0.5 ml of a metaphosphoric acid solution (187.5 g/L) and a formic acid (250 ml) solution and stored at −18°C pending the chemical analysis of ammonia-N and volatile fatty acid (VFA) concentrations. After thawing, the concentration of ammonia N in the filtrates was determined using a micro-Kjeldahl method [Kjeltec 8100, Hillerd, Denmark, AOAC (22, 23)]. The concentration of VFAs in the filtrates was quantified using a gas chromatography (Agilent 6890 GC, Agilent Technologies, Wilmington, DE, USA) with a 30 m × 0.25 mm × 0.25 μm column (DB-FFAP) and peak detection was compared and calculated as described by Purba et al. (24). The analyses for ammonia N and VFAs were carried out in quadruplicate, and the mean data was used for statistical analysis.

The second aliquot of the filtrates (5 ml) was homogenized for microbiological detection and stored at −80°C until the analysis the relative abundances of selected rumen microbes. After thawing, DNA was extracted from the homogenized filtrates by means of a quantitative real-time PCR (qPCR) reaction. Total genomic DNA was extracted from 1 ml homogenized filtrates using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) following the RBB+C method (25). The DNA yield was determined using a NanoDrop NanoVue spectrophotometer (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA) set to a 260:280 absorbance ratio. To prevent the DNA from degradation, the DNA yield was eluted with appropriate dilutions (volume of nuclease-free water) and stored at −20°C until further analysis. The relative abundances of selected primers in genomic DNA extracted from rumen fluids were quantified using a QuantiTect SYBR Green RT–PCR Kit (full master mix; Qiagen) fitted with the selected primer set and a Roche Lightcycler 480-II (Roche Applied Science, Basel, Switzerland), following the recently reported amplification and qPCR settings (26). References for the relative abundances of total bacteria, Ruminococcus albus, Ruminococcus flavefaciens, Fibrobacter succinogens, Butyrivibrio fibrisolvens, Megasphaera elsdenii, Streptococus bovis, Methanogen, and Protozoa were obtained from Vivantis Technologies Sdn Bhd (Selangor Darul Ehsan, Malaysia). The primers used for qPCR were forward primer 5′-CGGCAACGAGCGCAACCC-3′ and reverse primer 5′-CCATTGTAGCACGTGTGTAGCC-3′, forward primer 5′-TCTGGAAACGGATGGTA-3′ and reverse primer 5′-CCTTTAAGACAGGAGTTTACAA-3′, forward primer 5′-GTTCGGAATTACTGGGCGTAAA-3′ and reverse primer 5′-CGCCTGCCCCTGAACTATC-3′, forward primer 5′-TTCGGTGGATCDCARAGRGC-3′ and reverse primer 5′-GBARGTCGWAWCCGTAGAATC-3′, and forward primer 5′-CTTGCCCCTCYAATCGTWCT-3′ and reverse primer 5′-GCTTTCGWTGGTAGTGTATT-3′, for total bacteria, R. flavefaciens, F. succinogens, methanogen, and protozoa, respectively (27); forward primer 5′-ACACACCGCCCGTCACA-3′ and reverse primer 5′-TCCTTACGGTTGGGTCACAGA-3′, forward primer 5′-GACCGAAACTGCGATGCTAGA-3′ and reverse primer 5′-CGCCTCAGCGTCAGTTGTC-3′, for B. fibrisolvens and M. elsdenii, respectively (28); forward primer 5′-CCCTAA AAGCAG TCTTAGTTCG-3′ and reverse primer 5′-CCTCCTTGCGGTTAGAACA-3′ for R. albus (29); forward primer 5′-GCCA GGCTATTTAGGTGACACTATAG-3′ and reverse primer 5′-GGGT AATACGACTCACTATAGGG-3′, for S. bovis (30). Prior to beginning qPCR tests, a standard curve was established using a 6 fold serial dilution of pooled DNA. To ensure reproducibility, qPCR tests were performed in quadruplicate for each selected species or group of bacteria using both standards and genomic DNA samples. The Ct data were transformed into normalized relative numbers using the LightCycler 480 software version 1.2.9.11 (Roche Applied Science, Basel, Switzerland), which accounted for PCR efficiency. The values for the 16S rRNA gene of a particular species or group of microorganisms are defined as a relative percentage of total bacteria.

At the culmination of the routine feeding trail, goats were fasted for 24 h before being loaded (at 0500 h) and transported half a mile to a university slaughterhouse. Goats were spared as much as possible. Briefly, large ramps with steep slopes (11°) were employed for the loading and unloading of goats, and non-slip flooring was installed on the floors of the transport vehicles. Once the transport vehicles arrived at the slaughterhouse, effective scheduling processes were developed to guarantee rapid unloading of goats. To assist the smooth movement of goats, pre-slaughter handling methods were deployed. The goats were slaughtered standing up, with their heads fastened and their necks exposed for the throat incision. The goat knifes had a long, razor-sharp, and unbroken blade. These appropriate knives were prepared to cause a rapid outpouring of blood by cutting both jugular veins and both carotid arteries.

Prior to slaughter, the final BW of each goat was determined to establish the dressing percentage. The hot carcass weight was determined on the day of slaughter, and the carcasses were then refrigerated at 4°C for 48 h after slaughter. The weights of cold carcasses and the weights of shrinking carcasses were determined. Between the 12^th and 13^th ribs, carcasses were cut up to evaluate carcass quality features and yield grade metrics. Carcass quality characteristics included 12^th-rib fat thickness, longissimus muscle (LM) area, intramuscular fat, and final pH values. Carcass color was determined by measuring the L^*, a^*, and b^* color values on the cut lean surface and carcass external fat along the lateral side of the carcass with a portable Chroma Meter (CR-300, Minolta Corporation, Osaka, Japan). Quadruplicate color readings were done in the L^* (0 = black, 100 = white), a^* (negative values = green, positive values = red), and b^* (negative values = blue, positive values = yellow) color space (CIELAB); big chunks of connective tissue and intramuscular fat were excluded. Color saturation was estimated in accordance with the operational manual (31).

The Warner-Bratzler shear force (WBSF) values were calculated using the AMSA (1995) recommendations with slight modifications (32). Following the collection of carcass data, strip loins were removed from the left side of each carcass, vacuum-packed, and stored for 14 d at 4°C. Following the aging period, 2.54-cm-thick steaks were prepared, vacuum bagged, and frozen (−20°C) for further examination. WBSF steaks were defrosted at 4°C for 24 h and grilled to an internal temperature of 70°C on an electric grill (Kashiwa KW-308, Nakhon Pathom, Thailand) equipped with thermocouples inserted about in the geometric center of the steak. Cooking loss, reported as a percentage of weight loss, was obtained by dividing the initial weight (before cooking) by the end weight (after cooking). Using a steel hollow-core equipment, at least six 1.27-cm-diameter pieces were excised from each steak parallel to the muscle fiber orientation (32). The pieces were sheared perpendicular to the muscle fiber orientation using a shear device Texture analyzer for Warner-Bratzler Meat Shear (TA-TX2 Texture Analyzer, Stable Micro Systems, UK). A crosshead speed of 200 mm/min was used, as well as a 5 kN load cell calibrated to 113 read over the range 0x100 N. Six pieces were evaluated for their peak shear forces and expressed in kilograms.

External fat and connective tissue were removed from meat samples to determine ash, protein, moisture, and ether extractable lipid. In brief, CP content was estimated from nitrogen content (%N × 6.25) using a KjeltecTM 8400 fully automated Kjeldahl analyser (FOSS, Hilleroed, Denmark); ash content was determined by heating the steak sample using a furnaces (Carbolite AF1100, Parsons Lane, England) at 550°C for 15 h; moisture content was evaluated by weight loss after freeze-drying at −55°C for 5 d; and lipid content was determined using a Soxtec 2050 [Foss, Höganäs, Sweden, AOAC (22)]; petroleum ether 35–60°C used for extractable solvent. The analyses were carried out in quadruplicate, and the mean data was used for statistical analysis.

Data on animal performance, blood biochemical indices, rumen fermentation, microbial community, and carcass characteristics were statistically analyzed using the SAS 9.4 general linear model procedure (33) using the model: Y_ij = μ+ τ_i + ε_ij, where Y_ij is the response variable, μ is the overall mean, τ_i is experimental diet (i = NS or AS), and ε_ij is the residual error. In this model, diet was considered as a fixed effect, while the animal was considered as a random effect. The Shapiro–Wilk test was used to normalize all data. All of the data was analyzed using the Student's t-test. The least-squares means were reported, and P <0.05 was declared significant.

The chemical composition of the experimental diets is shown in Table 1. The experimental diets were isocaloric and isonitrogenous. Moreover, the contents of hemicellulose, cellulose, and lignin were similar between the two diets. The anthocyanin content of the AS vs. NS was found to be almost 6 times higher. Within the anthocyanin fraction, the differences between AS and NS were most pronounced for the Malvidin-3-O-glucoside and Cyanidin fractions, i.e., 5.7 and 6.2 times higher in AS compared to NS (Table 1).

The DMI, ADG, and the ratio between ADG and DMI of the crossbred Thai-native Anglo-Nubian male goats are summarized in Table 2. The AS diet had the same effect on animal growth performance as the NS it replaced (P > 0.05) after 90 d of routine feeding.

The variables of rumen fermentation are summarized in Table 3. The pH values of rumen fluids were similar between the two groups. The rumen ammonia N concentration was found to be 5.2% lower (P = 0.002) when the goats were fed AS. In contrast, the total VFA concentration was ~ 6.8% higher (P = 0.001) when AS was fed to the goats. The feeding of AS, instead of NS, caused higher proportions of acetate (P = 0.005) and lower proportions of propionate (P = 0.026). Consequently, the acetate to propionate ratio was found to be 13.5% higher (P = 0.003) after the feeding of AS. The proportions of butyrate were not affected by the feeding of AS (P = 0.401).

Both, the abundances of total bacteria and total protozoa (P = 0.432) in rumen fluid of the goats were similar between the two diets, i.e., P = 0.619 and P = 0.432, respectively. Overall, total protozoa and total bacteria of two groups were found to have log₁₀ copies of 16 rRNA genes ranging from 6.22 to 6.31 and 9.35 to 9.46, respectively. The relative abundances of R. flavefaciens, F. succinogens, B. fibrisolvens, M. elsdenii, and S. bovis (Figure 1), expressed as percentage of the 16S rRNA gene copy number of the total bacteria (%), were similar between the two groups (P ≥ 0.118). The relative abundance of methanogen was found to be lower (P = 0.001) when AS was fed while the relative abundance of R. albus was found to be higher (P = 0.007) after the feeding of AS.

The AS diet did not affect (P ≥ 0.110) on plasma concentrations of glucose, albumin, cholesterol, insulin, triglycerides, HDL, LDL, VLDL, or plasma urea nitrogen (Table 4). Similarly, there were no significant differences in the concentrations of IgG, alanine transaminase or aspartate aminotransferase between the two groups (P ≥ 0.324). The concentration of TBARS, however, was almost 9% lower (P = 0.024) when AS was fed to the goats (Table 4). In contrast, the feeding of AS instead of NS caused higher plasma concentrations of TAC (P = 0.002) and also higher activities of SOD, CAT, GSH-Px, and GSH-Rx (P ≤ 0.009).

Both the hot and cold carcass weights at slaughter (Table 5) were similar between the two diets (P ≥ 0.194). Consequently, the dressing percentage also was similar between diets (P = 0.217). Likewise, the 12^th rib fat thickness, LMA, and pH were not affected by AS (P ≥ 0.157). There were no significant differences between the two groups in the indices related to the external fat color or 12^th rib lean color (P ≥ 0.235).

As shown in Table 6, the cooked rate, drip loss, moisture, protein, and ash percentages of steaks were similar between the two experimental diets (P ≥ 0.111). However, the WBSF value of tenderness was ~ 13.8% lower (P = 0.035) when AS was fed. In contrast, the values on the percentage of intramuscular fat in steak samples were found to be ~ 28% higher (P = 0.015) after the feeding of AS.