The FPR was obtained from BOYA Biotechnology Co., Ltd (Leizhou, Guangdong, China). The raw pineapple peel was squeezed to maintain 78%−80% initial moisture, then evenly sprayed with a mixed lactic acid bacterium (Lactobacillus plantarum GIM1.191) and yeast (Saccharomyces cerevisiae GIM2.133) liquid. Finally, the FPR mixture was pressed into polyethylene bags (50 kg each) and fermented anaerobically for 20 days.

The experiment was conducted in a beef cattle company in Yunfu, Guangdong, China. A total of 30 healthy Simmental bulls (20 months old, 546 ± 44 kg weight) were used in a completely randomized design for a 3-day adaptation period and a 30-day experimental period. They were randomly divided into three groups in an open sawdust-bedded cowshed: the CON group (no FPR or control, fed basic diet), T25 (25% FPR replaced SPR), and T50 (50% FPR replaced SPR). All bulls were fed a total mixed ration (TMR) at 10:00 and 16:00, and water was provided ad libitum. To meet nutritional requirements, the TMR was based on SPR and rice straw as the main forage components and corn flour as the major concentrate component, according to NRC standards (NRC, 2016). The ingredients and nutrient composition of the three diets are shown in Table 1. The remaining feed was collected and recorded daily at 8:30.

The samples of FPR and FPR (days 0 and 20) were analyzed for dry matter (DM), crude protein (CP), ether extract (EE), and ash according to the AOAC International guidelines (AOAC, 1990). The neutral detergent fiber (NDF) and acid detergent fiber (ADF) contents were determined using the method reported by Van Soest et al. (1991). These contents were determined from water extract, while wet FPR (20 g) was transferred to a glass bottle filled with 180 ml of deionized water, sealed, mixed, and stored at 4°C overnight (Fang et al., 2016). Then, the water extract was passed through filter paper, and the filtrate pH was measured using a glass-electrode pH meter (Horiba D-21, Horiba, Tokyo, Japan). The FPR had low DM, CP, NDP, ADF, and ash of 21.15, 6.66, 63.46, 33.03, and 4.33% of DM basis, respectively (Table 2). Additionally, it had high starch of 3.2%.

On days 1, 12, and 24, the feed offered to the bulls was sampled and used for nutrient analysis and chemical analysis. The methods of nutrient determination, including CP, NDF, and ADF, were consistent with the method described in the “Nutritional compositions” section. Chemical analysis of the calcium (Ca) and phosphorus (P) contents was performed using inductively coupled plasma spectroscopy (Chemists and Horwitz, 1990).

The cattle were weighed before the morning feeding on days 1 and 30, and the average daily gain and feed weight ratio were calculated. Eight bulls were randomly selected from each group for the slaughter test. They were fasted for 12 h, and water was withheld for 3 h before slaughter. After being stunned, the cattle were slaughtered according to a general process, including hanging upside down, slaughtering, bloodletting, skinning, removing head and tail, and eviscerating. The live weight was recorded before slaughter and carcass weight after slaughter. Then the samples of the longissimus thoracis (LT) were excised between the 12th and 13th rib. After measuring the eye muscle area, pH, and flesh color of LT, cut into two uniform pieces vertically. One piece was put on ice for 24 h to measure drip loss, centrifugal water loss rate, pH and shear force, and the other was frozen in dry ice for the later determination of nutritional indicators.

The outline of the LT cross-section was delineated with sulfuric acid paper, and the eye muscle area was calculated with ADOBE PDF (version 1.2, San Jose Co., Ltd., CA, United States) after scanning. At 45 min, 24 h, and 48 h after slaughter, the pH was determined using a pH meter (FE28-Standard, METTLER-TOLEDO Co., Ltd., Shanghai, China) in the cut surface of the LT. A colorimeter (NR10QC, 3nh Co., Ltd., Shenzhen, China) was used to measure meat color on the surface of the LT about 45 min after slaughter. The meat samples were cut into 1 cm thick pieces, wrapped in plastic bags, heated to 70°C, and taken out to cool. The meat was cut into strips with a cross-section of 1 cm × 1 cm along the fiber direction, and then the shear force was measured using a tenderness meter (TA. XTPlus, SANHAO Co., Ltd., Suzhou, China). The meat was cut into long strips (5 cm × 2 cm × 2 cm) and weighted as N1. After being packed into plastic bags and hanging suspended for 24 h in a 4°C freezer, the weight was scored as N2. Drip loss% = (N2/N1) × 100%. Then, after centrifugation (1,500 × g, 30 min at 4°C), the centrifugal water loss rate was calculated. Analyzing DM, CP, EE, and ash of meat according to the AOAC method and amino acids using a full-automatic amino acid analyzer (LA8080, Hitachi Co., Ltd., Tokyo, Japan).

Rumen fluid samples were collected from all the bulls on the last day by a rumen tube before the morning feeding. To avoid the contamination of oral saliva, the first 20 ml of rumen fluid was discarded. Approximately 150 ml of rumen fluid sample from each bull was collected and then strained through four layers of cheesecloth. The filtrate was dispensed into 50-ml centrifuge tubes (REF430829, Corning Life Science Co., Ltd.) and 2-ml storage tubes. The samples in the storage tubes were put into liquid nitrogen and then transferred to a −80°C laboratory refrigerator for future use.

The pH of rumen fluid in centrifuge tubes was immediately measured by a pH meter (FE28-Standard, METTLER-TOLEDO Co., Ltd., Shanghai, China). Then, the samples were centrifuged at 5,000 × g for 15 min (BR4I, Thermo Co., Ltd., NY, United States) to collect the supernatant. The supernatant was divided into three 15-ml centrifuge tubes. Two tubes were used to measure volatile fatty acid (VFA) (acetic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, and valeric acid) content and ammonia nitrogen (NH₃-N) concentration using a gas chromatograph (SP-3420, BEIFENGRUILI Co., Ltd., Beijing, China) and ELIASA (ST-360, KEHUA Co., Ltd., Shenzhen, China), respectively, according to Erwin et al. (1961) and Broderick and Kang (1980), and were stored at −20°C, and the remaining tube was stored at −80°C as a spare.

The total genomic DNA was extracted from rumen fluid samples using the modified cetyltrimethylammonium bromide/sodium dodecyl sulfate method (Zhang et al., 2021). The DNA samples were tested for integrity using 1% agarose gel electrophoresis, and their concentration was determined using a Qubit fluorometer (Nanodrop2000/2000C, Thermo Co., Ltd., NY, United States). Then, DNA was diluted to 1 ng/μl using sterile water according to the concentration. The V1–V9 regions of the 16S ribosomal DNA (rDNA) genes were amplified by polymerase chain reaction using the TransStart® FastPfu DNA Polymerase Kit (TransGen Biotech Co., Ltd., Beijing, China). In detail, the amplification was performed with the universal primers (forward primer, 27F: AGAGTTTGATCCTGGCTCAG; reverse primer, 149R: GNTACCTTGTTACGACTT). Sequencing libraries were generated using the SMRTbellTM Template Preparation Kit (Pacific Bioscience, CA, United States) on the PacBio Sequel sequencer.

Single-end reads were assigned to samples based on their unique barcode in the adaptor sequence. Quality filtering of the raw reads was performed to obtain high-quality clean reads according to the PacBio SMRT Portal Provisioning Agreement. The reads were compared with the reference database using the UCHIME algorithm (http//www.drive5.com/usearch/manual/uchime_algo.html) to detect chimeric sequences (Haas et al., 2011; Quast et al., 2012), and clean reads were finally obtained using the Uparse software (Uparse v7.0.1001) Edgar RCUPARSE, 2013. Sequences with ≥97% similarity were assigned to the same operational taxonomic units (OTUs). For each representative OTU, the Silva Database (https://www.arb-silva.de/) was used to annotate taxonomic information based on the Mothur algorithm (Quast et al., 2012). Alpha diversity was applied to analyze the complexity of species diversity within groups, including the observed species, Chao1, Shannon, and ACE indices. Beta diversity analysis was used to evaluate differences between groups using nonmetric multidimensional scaling (NMDS). All these indices were calculated using the quantitative insights into microbial ecology (QIIME) pipeline (Version 1.7.0).

The data were analyzed using the INFLUENCE Statement and GLM Model of SAS (version 9.4; SAS Institute Inc., Cary, NC, United States). A CONTRAST Statement was used to analyze the effects of each index between treatment and control. The test results were presented as the mean and standard error of the mean (SEM), with P < 0.05, indicating a significant difference, and P < 0.01, indicating a highly significant difference. Growth performance, meat quality indicators, rumen fermentation parameters, and relative abundance of rumen flora were analyzed using the analytical model I: Yi = μ + Pi + εi, where Yi is the dependent variable value of the bull in different treatments, μ is the overall mean, Pi is the dietary treatment effect, and εi is the random error. The KENWARDROGERS method is used to perform DOF correction.

The FPR had low DM, CP, NDP, ADF, and ash of 21.15, 6.66, 63.46, 33.03, and 4.33% of DM basis, respectively (Table 2). Additionally, it had high starch of 3.2.%.

The daily dry matter intake (DMI) of the control and treatments were approximately similar (Figure 1). The average daily weight gain of the T25 and T50 groups, respectively, increased by 0.17 and 0.29 kg, and DMI/weight gain significantly (P < 0.05) decreased (Table 3). According to the purchase and sale prices, the benefit of fattening each bull per day improved from ¥3.52 (CON) to ¥ 8.86 (T25) and ¥7.21 (T50; Table 4).

Fermented pineapple residue did not adversely affect the slaughter performance and beef sensory quality (Table 5). The crude fat indicators were significantly higher (P < 0.05) in T50 than in CON (Table 6). The content of cysteine, glycine, histidine, phenylalanine, proline, and tyrosine in treatments was raised (P < 0.05; Table 7), indicating that FPR can improve the amino acid composition of meat.

The FPR increases (P < 0.05) the rumen fluid pH while CON was 7.05, T25 was 7.18, and T50 was 7.26 (Table 8). The concentrations of isobutyric acid and isovaleric acid significantly (P < 0.05) decreased in T25 (0.83 mmol/L) and T50 (0.70 mmol/L) while CON was 0.94, whereas isovaleric acid descent (P < 0.05) in T25 (0.88 mmol/L) and T50 (0.62 mmol/L) while CON was 1.92. Butyrate raised (P < 0.05) while CON was 7.60%, T25 was 8.87%, and T50 was 9.65%. Thus, FPR had a regulating effect on the fluctuation range of the rumen fermentation parameters.

The V1–V9 regions of the 16S rDNA were enriched, and 356,162 raw reads were collected using high-throughput analysis. After quality control, each sample produced 11,339 valid sequences with a read length of 1,447 nucleotides. Venn diagram analysis yielded 4,119 unique OTU candidates with 97% sequence similarity, and 1,331 candidates shared across all samples were defined as core OTUs (Figure 2). The core OTUs were ~32.31% of the total candidates, whereas 456, 447, and 480 OTUs were identified as unique in the CON, T25, and T50 groups, respectively. A total of 21 phyla, 26 kingdoms, 46 orders, 66 families, 102 genera, and 97 species were found using the OTU annotations. The main bacterial phyla were Firmicutes, Bacteroidetes, and Tenericutes (48.62, 38.19, and 5.65%, respectively; Figure 3). At the species level (Figure 4), Rumen_bacterium_YS3 (1.32%) was the most common species. Unclassified bacteria accounted for 93.45% of the OTUs, while the identified secondary strains accounted for 97.03%.

Observed species, Chao1, Shannon, Simpson, ACE, and PD_whole_tree were used to evaluate the microbial diversity after FPR treatment (Figure 5). The addition of FPR had no significant effect on the above-mentioned indexes, but the diversity and richness tended to decrease as the proportion of FPR increased. The diversity and richness of the T50 group were the lowest. The rumen flora of the groups was roughly distributed in the same area (Figure 6A). The sample distances were more concentrated within each group, presenting three different colonies as a whole; this indicated that FPR affected the main bacterial groups in the rumen. The locations of the sample points in each group were not completely separated (Figure 6B), and the area of intersection of sample colonies in each group was the smallest for CON and T50, which indicated that 25% FPR affected more than 50% but not vigorously. The differences between and within bacterial groups showed obvious discrimination under the nonlinear structure (Figure 6C); the samples were clustered more centrally within each treatment, and the groups were well distinguished.

A total of 25 different bacterial strains were statistically detected between the groups, with nine species in CON, two in T25, and 14 in T50 (Figure 7A). The phylogenetic tree showed multiple clades (Figure 7B). The evolutionary routes of the three treatment groups were mixed with each other, indicating they had similar evolutionary directions. This showed that the environment created in the rumen was not the same when the proportion of FPR was different, so the rumen microbes followed different evolutionary directions.

To analyze the correlations of FPR and rumen microbiota, Pearson correlation analysis was performed and then found that four phyla, seven genera, and three species were related to the rumen fermentation parameters (Figure 8). At the genus level, unidentified_Rikenellaceae and Psedobutyrivibrio were positively (P < 0.05) correlated to pH, whereas Succiniclasticum was significantly (P < 0.05) negatively correlated. Fibrobacter was negatively (P < 0.05) correlated to NH3-N while Pirellula was positive. Alcaligenes was positively (P < 0.05) correlated to lactic acid and butyric acid. Marvinbryantia was correlated to butyric acid positively (P < 0.05). At the species level, Lachnospiraceae_bacterium_RM and Butyrivibrio_fibrisolvens were negatively (P < 0.05) correlated to pH. Butyrivibrio_fibrisolvens was positively (P < 0.05) correlated to NH₃-N. Alcaligenes_faecalis was significantly correlated to lactic acid and butyric acid (P < 0.05).