Fermented soybean meal affects the ruminal fermentation and the abundance of selected bacterial species in Holstein calves: a multilevel analysis

The effect of soybean meal (SBM) replacement with fermented SBM (FSBM) on ruminal fermentation and bacterial abundance in Holstein calves was investigated in this study. Thirty nine calves were randomized to: (1) control: 27% SBM + 0% FSBM (FSBM0, n = 13); (2) 18% SBM + 9% FSBM (FSBM9, n = 13); and (3) 13.5% SBM + 13.5% FSBM (FSBM13, n = 13). SBM contained a greater amount of large peptides containing 3 to 10 amino acids (AAs), while FSBM had a greater amount of ammonia nitrogen (NH3–N), free AAs, and small peptides containing 2 to 3 AAs. The calves fed FSBM13 had the lowest acetic acid, NH3–N, and the ratio of acetate to propionate, with the greatest concentration of caproic acid, valeric acid and isovaleric acid in ruminal fluid. Compared to those fed FSBM9 or FSBM13, the calves fed FSBM0 had the greatest proportion of Butyrivibrio fibrisolvens and Ruminococcus albus in rumen fluid. However, the ruminal abundance of Prevotella ruminicola in calves fed FSBM13 was greater than in calves fed FSBM0. Network analysis showed that the abundance of the Ruminococcus albus was associated with large peptides, and butyric acid was correlated with small peptide. Taken together, our findings suggest that FSBM may have the potential to boost calf performance by changing the fermentation products and the relative abundance of some members of the bacterial community in the rumen.

www.nature.com/scientificreports/ The chemical composition of the feed, such as peptides and amino acids (AAs), has been shown to alter in vitro and in vivo fermentation and the bacterial community in the rumen [10][11][12] . Carro and Miller 11 reported that feeding peptides resulted in increased digestibility of fiber, VFA production, microbial nitrogen flow and microbial protein synthesis efficiency in calves. Fermentation is an efficient and cost-effective process that improves the quality of feed through microbial activity or microbial enzymatic action 13 . Soybean meal (SBM) proteins are extensively hydrolyzed to AAs and low molecular weight peptides due to microbial enzymatic activity and biochemical changes during fermentation with Bacillus subtilis 14,15 . Previous studies have shown that the fermentation of soya bean with Bacillus subtilis increases the proportion of A fraction; non-protein nitrogen is defined as the fraction A, which is metabolically closer to the soluble protein and consists of small peptides, free amino acids and NH 3 16,17 . In addition, the amount of rumen undegraded protein (RUP) may increase in fermented soybean meal (FSBM) due to intensive heat treatment before fermentation (i.e., cooking) 18 . Protein fractions, such as RUP, have been reported to affect the performance and digestibility of calves 18,19 . Therefore, feeding fermented protein sources may affect the bacterial community, the production of VFAs and the growth of calves.
Previous studies have concentrated on the effects of FSBM on pig performance [20][21][22][23] , although few have studied the impact of FSBM on the fermentation of rumen and the bacterial community in dairy calves [24][25][26] . Our previous work has shown that feeding FSBM improved calf performance, increased starter intake and reduced weaning stress during abrupt weaning 24 . Weaning stress has been known to reduce feed intake and calf performance during the immediate post-weaning period [24][25][26] . Therefore, we hypothesized that fermentation may improve the protein fractions and peptide composition of SBM and affect ruminal fermentation and the abundance of some cellulolytic, fibrolytic and proteolytic bacteria in Holstein calves after abrupt weaning. Furthermore, we employed a multilevel approach, such as the hierarchical clustering analysis (HCA) and the principal component analysis (PCA), to determine the potential relationships between rumen fermentation and bacterial abundance.

Results
Protein fractions and peptide composition. As shown in Table 1, the protein content of SBM increased after fermentation from 43.3 to 48.2%. FSBM had greater levels of small peptides and free AAs, although it had lower levels of large peptides than SBM. The ammonia nitrogen (NH 3 -N), fractions A, B 3 , and C increased, although the fractions B 1 and B 2 decreased in FSBM compared to SBM. In comparison, RUP increased and rumen degraded protein (RDP) decreased in FSBM as compared with SBM. FSBM13 had greater amounts of crude protein, RUP, and fractions A, B3 and C, but had lower quantities of starch than FSBM0 and FSBM9.

Concentrations of VFAs and NH 3 -N in the rumen fluid of Holstein calves. As there was a linear
increase in the FSBM level (0, 9%, and 13.5%), a linear contrast analysis was performed ( Table 2). A linear reduction in acetic acid concentration was found with an increase in the FSBM level (P = 0.04); calves fed FSBM13 had the lowest acetic acid concentration compared to those fed FSBM9 or FSBM0. Treatments did not affect the concentration of propionic acid (Table 2). This, in turn, resulted in a greater ratio of acetate to propionate in calves fed with FSBM0 ( Table 2).
The ruminal concentrations of NH 3 -N (P < 0.01), caproic acid (P < 0.01), valeric acid (P = 0.005) and isovaleric acid (P < 0.01) increased linearly in response to FSBM such that the FSBM13-fed calves had the greatest concentrations of NH 3 -N, caproic acid, valeric acid and isovaleric acid in rumen fluid (Table 2). Treatments did not affect the concentration of butyric acid and isobutyric acid in ruminal fluid ( Table 2).

Detection of ruminal bacteria by polymerase chain reaction (PCR).
The results of the PCR products gel electrophoresis were shown in Supplementary Fig. 1. On the basis of the PCR assay, samples derived from fresh rumen fluid of cows were found to be positive for Ruminococcus flavefaciens, Fibrobacter succinogenes, Ruminococcus albus, Butyrivibrio fibrisolvens, and Prevotella ruminicola ( Supplementary Fig. 1).
Next, PCR gel electrophoresis was conducted using experimental samples (calves). The PCR assay showed that the experimental samples (ruminal fluid collected from the claves) were found to be negative for Ruminococcus flavefaciens and Fibrobacter succinogenes. Samples were found to be positive for Ruminococcus albus, Butyrivibrio fibrisolvens and Prevotella ruminicola. qPCR for the determination of bacterial abundance in the rumen fluid of Holstein calves. The qPCR assay detected signals for all bacteria in the rumen fluid of cows (positive control). In relation to the total 16S rDNA gene copies per ml rumen fluid, the relative abundance of Ruminococcus albus (P = 0.04, linear) and Butyrivibrio fibrisolvens (linear, P = 0.0001, and quadratic, P = 0.0008) decreased with an increase in FSBM (Table 2). Compared with FSBM9-or FSBM13-fed calves, calves fed FSBM0 had the greatest abundance of Ruminococcus albus and Butyrivibrio fibrisolvens in relation to the total 16S rDNA gene copies per ml rumen fluid ( Table 2). The relative abundance of Prevotella ruminicola (in relation to the total 16S rDNA gene copies per ml rumen fluid) increased linearly (P < 0.05) in response to an increase in FSBM. The calves fed FSBM13 had the greatest abundance of Prevotella ruminicola (in relation to the total 16S rDNA gene copies per ml rumen fluid) compared with calves fed FSBM0 ( Table 2). The qPCR did not detect any signal for Ruminococcus flavefaciens and Fibrobacter succinogenes at Ct < 40; this threshold was set on the basis of the negative control Ct. The specificity of qPCR was shown in Supplementary Fig. 2.
The PCA analysis revealed that the first four principal axis factors accounted for a sufficient amount of total variance (69.5%, Supplementary Table 1). The PCA plot showed that the proportion of Ruminococcus albus was positively associated with large peptides (directional vectors at < 45°; Fig. 2b). The proportion of Ruminococcus albus was negatively associated with Prevotella ruminicola, small peptides, butyric acid and isovaleric acid (directional vectors approaching 180°, Fig. 2b). In addition, small peptides were positively associated with ruminal Prevotella ruminicola, valeric acid, isovaleric acid, caproic acid and NH 3 -N (directional vectors at < 45°). However, ruminal NH 3 -N, valeric acid, and caproic acid were negatively associated with Butyrivibrio fibrisolvens (directional vectors approaching 180°). PCA showed a negative association between Butyrivibrio fibrisolvens and butyrate, its main fermentation product (directional vectors approaching 180°, Fig. 2b).
In the case of a 60% cut-off point, a correlation between acetic acid and Butyrivibrio fibrisolvens was found in a network analysis (darker line shown in Fig. 3c). This 60% cut-off point was selected as all parameters were Table 1. Protein fractions, peptide composition, and amino acid profile of soybean meal (SBM) and fermented SBM (FSBM). Data were presented as mean ± SD; DM: dry matter. a ) Large peptide: consisting of 3-10 amino acids. b ) Small peptide: consisting of 2-3 amino acids. www.nature.com/scientificreports/ disconnected from this level. Butyric acid was found to be associated with small peptides, while large peptides were linked with Ruminococcus albus (darker line shown in Fig. 3c). Caproic acid was also found to be associated Calves were abruptly weaned at 59 days of age but remained in the experiment until day 67 when rumen fluid was collected (n = 13 calves per treatment). a ) FSBM0 (control, 27% SBM + 0% FSBM); FSBM9 (18% SBM + 9% FSBM); and FSBM13 (13.5% SBM + 13.5% FSBM). b ) The data was analyzed by the SAS software using the oneway ANOVA, followed by multiple comparisons tests, the Tukey's test. The data was presented as Least Squares Means (LSM) ± Standard Error of the Mean (SEM). Different superscripts (a, b) in a row indicate statistically significant differences at P < 0.05. * indicates P < 0.10: a tendency towards statistical significance between FSBM9 and FSBM13. Orthogonal polynomial contrasts were used to determine the linear or quadratic effect of FSBM using SAS software (SAS Institute Inc., Cary, NC). c ) The relative abundance of bacteria was measured using qPCR as a proportion of the total bacterial 16S rDNA. www.nature.com/scientificreports/ with ruminal NH 3 -N. The network analysis confirmed the HCA and PCA findings, thus providing a better picture of the interactions between dietary peptides and ruminal NH 3 -N, VFAs and bacterial abundance.

Discussion
In this study, the fermentation of SBM increased the total protein content and modified the protein fractions.
The fermentation of SBM has been reported to increase the protein content of SBM from 43 to 49% 15,20,21 . The fermented SBM provides approximately 10% more protein than the traditional SBM 18 . Potentially, an increase in the production of microbial proteins during fermentation may explain the greater protein content of FSBM 15,27 . It was also found that free AAs, small peptides, and fractions A (non protein nitrogen) increased in FSBM, in agreement with other studies 15,16,22,[25][26][27] . This may be the result of proteolytic activity of microorganisms (such as Bacillus spp.) and the release of proteolytic enzyme during fermentation 16,18,28 . Weng and Chen 28 found that the concentration of small peptides in soybean increased after 24 h of fermentation. In the present study, the  www.nature.com/scientificreports/ fermentation time for SBM was 48 h, which may explain the greater percentage of fraction A (which consists of free AAs and small peptides 16,17 ) in FSBM. Zhang et al. 29 found that small peptides had a positive effect on nitrogen retention and the use of protein in goats. In addition, Cao et al. 30 reported that, despite digesting greater amounts of nitrogen than those fed SBM, soybean peptide-fed sheep had smaller losses of nitrogen from the feces and urine. However, Wang et al. 31 demonstrated a decrease in NDF and ADF digestibility due to infusion of soybean small peptides. It should be remembered that, due to greater amounts of small peptides and RUP, the feeding of FSBM13 may compromise the digestibility of the fiber and the concentration of acetate in calves. In response to RUP, Griswold et al. 32 observed a decrease in total concentrations of VFAs, acetate and butyrate in continuous culture. It seems that by improving the metabolism of nitrogen, the use of FSBM with greater small peptides has led to an increase in the growth performance of FSBM13-fed calves reported in our previous study 24 .
Importantly, it should be noted that FSBM was prepared by incubation of SBM with Bacillus subtilis, which has been shown to have probiotic effects. Hence, Bacillus subtilis may enter the rumen and alter the microbial composition, fermentation and digestive processes of the rumen 33,34 . For example, Zhang et al. 33 found that the probiotic bacteria (a mixture of Lactobacillus plantarum and Bacillus subtilis) influenced the rumen bacterial community and reduced the numbers of cellulolytic bacteria such as Ruminococcus albus. Therefore, further research with killed colonies (i.e., sterilized FSBM) is required to distinguish the role of peptide size and Bacillus subtilis (as a probiotic bacterium) in rumen physiology. In addition, an investigation into the presence of these bacteria in the rumen (i.e., qPCR) may help to explain the possible probiotic effect of Bacillus subtilis.
As described above, fermentation increased small peptides and free AAs in FSBM, resulting in an increased fraction A, consisting mostly of free AAs and small peptides 35,36 . In addition, thermal administration during fermentation tends to be capable of transforming fraction B2 into fractions B 3 (slowly degradable true protein) and C (undegradable true protein) 35 . Toti et al. 37 reported that heat treatment decreases fraction B1, but increases B3 and C fractions in canola meal.
The present results revealed a decline in the ruminal concentration of acetic acid in calves fed FSBM13 or FSBM9. Similarly, several studies have reported that feeding peptides or SBM small peptides reduces the concentrations of acetic acid in the rumen 31,38 . The decreased concentration of acetic acid in calves fed FSBM13 seems to be partially due to a decrease in the abundance of Ruminococcus albus (one of the fiber digestive bacteria) 39,40 and Butyrivibrio fibrisolvens in the rumen 41 . In particular, network, Spearman correlation, PCA, and HCA analyses showed that acetic acid was associated with Butyrivibrio fibrisolvens, suggesting that feeding FSBM13 can result in a decrease in both acetic acid and Butyrivibrio fibrisolvens abundance in calves. It is important to note that during the entire study and at the time of the collection of rumen fluid (day 67), calves were fed experimental starters with 21-23% NDF (almost medium-NDF levels) and 10% chopped alfalfa hay. The use of straw bedding by calves during the experiment was found to be very rare. It should also be noted, as a limitation of this study, that the stomach tube was used in this study to collect rumen fluid. This method mainly collects the liquid parts of rumen, not the solid particles of rumen. Many bacterial species, particularly cellulolytic bacteria, are strongly attached to solid particles 39,40,42 . These limitations might explain the low abundance of Butyrivibrio fibrisolvens (a fibrolytic bacterium) and Ruminococcus albus (a cellulolytic bacterium). Moreover, no signals were found in this study for Ruminococcus flavefaciens and Fibrobacter succinogenes in experimental samples (i.e., calves) using qPCR. However, positive signals for these bacteria were detected in the fresh rumen fluid of cows. Shinkai and Kobayashi 42 did not detect one group of Fibrobacter succinogenes in the rumen content of sheep by qPCR test. They reported that the fluorescence signals from rumen fluid were mainly weak, and the abundance of Fibrobacter succinogenes in rumen fluid was considerably smaller than in rumen-incubated hay samples 42 . It appears that a large number of cellulolytic bacteria were not obtained due to the poor efficiency of the stomach tubes used in this study. As a result, the present results did not cover the full spectrum of the rumen bacterial population; further research is required to identify the effect of FSBM on the rumen bacterial communities in calves.
The increase in FSBM resulted in greater concentrations of valeric acid, isovaleric acid and caproic acid. Both Ruminococcus albus and Butyrivibrio fibrisolvens species are the main consumers of branched-chain VFAs (BCVFAs, e.g. isovaleric acid) 43 . Therefore, the decreased abundance of these bacteria might explain greater concentrations of isovaleric acid in FSBM13-fed calves. In fact, BCVFAs are primarily derived from ruminal digestion of dietary proteins 44 . Hence, the greater concentrations of isovaleric acid and valeric acid in FSBM13-calves can be explained by the greater amount of crude protein and small peptides in FSBM than in SBM. Moreover, the Spearman correlation demonstrated a positive, but not significant, correlation between small peptides and ruminal concentrations of isovaleric acid and valeric acid. In addition, there was a positive correlation between Butyrivibrio fibrisolvens and isobutyrate, although there was no significant relationship between cellulolytics and isobutyrate. Previous studies have shown that BCVFAs improve the apparent digestibility of dry matter, microbial growth, microbial function, and enzyme activity in ruminants 45,46 . We also recently reported an improvement in the performance of FSBM13-fed calves compared to SBM-fed calves 24 . These findings suggest that the feeding of FSBM may alter the rumen conditions for the growth of the calves.
Data revealed that FSBM13 increased Prevotella ruminicola but decreased the abundance of Ruminococcus albus and Butyrivibrio fibrisolvens. Wang et al. 31 reported that the infusion of soybean small peptides had no effect on abundance of Prevotella ruminicola and Ruminococcus albus in cows. The data revealed a negative correlation between Butyrivibrio fibrisolvens abundance and small peptides, while Wang et al. 31 reported an increase in Butyrivibrio fibrisolvens abundance in cows in response to the infusion of soybean small peptides. This implies that other factors of SBM/FSBM or fermentation (i.e., Bacillus subtilis) may have an effect on the composition of the bacteria in the rumen fluid, in addition to small peptides.
Generally, the primary fermentation products of Ruminococcus albus, Butyrivibrio fibrisolvens and Prevotella ruminicola are acetate, butyrate and propionate, respectively 38 www.nature.com/scientificreports/ shown by the fact that these two strains together account for only a small portion of the total bacterial population, several other species may have been responsible for the bulk of the increase in acetic acid as their main fermentation product. Interestingly, this was discovered by the PCA, Butyrivibrio fibrisolvens was adversely associated with butyrate, its main fermentation product 47 , but the cause remains uncertain. Nevertheless, Butyrivibrio fibrisolvens have been found to be substantially associated with butyrate in the subacute rumen acidosis model 48 . The Spearman analysis also revealed that Butyrivibrio fibrisolvens had a negative association with the concentrations of NH 3 -N in rumen. Sales et al. 49 reported that, the proteolytic activity of Butyrivibrio fibrisolvens increased as a reaction to NH 3 -N, while the proteolytic activity of Prevotella albensis decreased. They proposed that each species would follow its own dietary strategy to conform to the environmental conditions of the rumen ecosystem 49 . The inverse association between Butyrivibrio fibrisolvens and butyrate or between Butyrivibrio fibrisolvens and NH 3 -N may be part of the regulatory feedback within the rumen microbial ecosystem; more research is required to validate this hypothesis. It is important to note that the Spearman correlations can only indicate whether there is an association between two variables, and this does not mean a casual relationship 50 .
The PCA showed that large peptides had a strong association with the abundance of Ruminococcus albus and Butyrivibrio fibrisolvens. Regarding variations in FSBM levels, fermentation and bacterial abundance in rumen may vary depending on the size of the peptide. As described above, the potential role of Bacillus subtilis in rumen physiology needs to be investigated in order to differentiate the effects of peptide size and Bacillus subtilis on rumen bacterial community.
Calves fed FSBM13 had a greater proportion of Prevotella ruminicola in the rumen fluid. This proteolytic bacterium effectively consumes NH 3 -N, peptides and, to a lesser degree, uses AAs to produce microbial proteins and adenosine triphosphate (ATP) 51,52 . The PCA showed that Prevotella ruminicola was positively associated with the concentration of NH 3 -N in rumen. Prevotella ruminicola has been reported to grow faster and to a greater final dry weight in the presence of soybean protein than casein 53 . Hence, feeding FSBM13 containing small peptides and free AAs might contribute to increased growth of proteolytic Prevotella ruminicola to reduce the high NH 3 -N levels 52 .
In summary, proteolytic Prevotella ruminicola responded positively to FSBM, while other cellulolytic/fibrolytic bacteria, such as Ruminococcus albus and Butyrivibrio fibrisolvens, reacted negatively. On the basis of the present findings and our previous research 24 , it can be concluded that feeding FSBM will improve the growth performance and starter intake of Holstein calves by changing the fermentation and abundance of some bacteria in the rumen. More studies with both liquid and solid phases of the rumen content and a broader variety of bacteria are needed to provide a deeper understanding of the effects of FSBM on the rumen ecosystem. Animals, experimental design and diets. Thirty-nine Holstein calves with an average initial body weight of 39.5 ± 3.3 kg were assigned to 3 treatments (7 male and 6 female per treatment) from birth until 67 days of age. The calves were separated from their dams immediately after birth, weighed, and moved to individual hutches (1 × 1.5 m) bedded with straw, which was renewed every 24 h. The calves were housed in individual hutches during the entire study (67 days). The use of straw bedding by calves was also checked and was found to be very rare during the experiment.

Methodology
Calves received 2.5 L of colostrum in the first 1.5 h of life, followed by 5 L of colostrum in the first 12 h of life. By the third day, colostrum and transitional milk were given to calves. From day 4, whole milk (5 L per d, temperature: 39 ± 2 °C) was fed to calves using a plastic bucket at 0,040 and 1,600 h. The average milk composition used during the trial was 3.30 ± 0.13% fat, 2.85 ± 0.01% crude protein, 4.93 ± 0.06% lactose, and 11.00 ± 0.52% total solids (mean ± SD). During the entire study, the calves had free access to water and the ad libitum starter intake (which included different quantities of SBM or FSBM) was accomplished by providing a quantity that resulted in a residue of 10% of the feed offered after 24 h. The feed refusal for each individual calf was collected at 0,800 h. From the first day of life, a starter was provided mixed with chopped alfalfa hay until day 67 of the experiment. During the entire study (67 days) and at the time of the collection of rumen fluid (day 67), calves were fed experimental starters. All calves were suddenly weaned by removing the milk meal at 8 weeks of age, but remained in individual hutches until the end of the experiment (day 67).
Fermentation of SBM. The FSBM was developed by a commercial company (Mehre Bisotun Co., Isfahan, Iran) as described by others 27 . The method used to prepare the FSBM was detailed in our previous work 24 . In brief, the dried SBM (90% dry matter) was immersed in distilled water (385 g water/1,000 g SBM) for 60 min, taking the moisture content of SBM to about 35%. The hydrated SBM was then cooked in a steam tank at 65 °C for 1 h and allowed to cool to room temperature. The cooled SBM was then inoculated with Bacillus subtilis GR-101 (4 log cfu/g of SBM), mixed and fermented in a bed-packed incubator for 48 h. After fermentation, the  . The first 20 ml of rumen contents was discarded to avoid contamination with saliva. In addition, samples with abnormal viscous appearance were deemed to have been contaminated with saliva and discarded. Rumen fluid samples were immediately filtered through four layers of cheese cloth and divided into two parts. The first portion (10 ml) was mixed with 2 ml of 250 g/l (w/v) metaphosphoric acid and preserved at − 20 °C for VFAs analysis. The second part of the rumen fluid (10 ml) was immediately centrifuged at 9,000 × g for 10 min. The supernatants were discarded and the resulting pellets were resuspended into a microtube containing 1 ml of the Table 3. Ingredients and chemical composition (% of dry matter, unless otherwise stated) of experimental starter feeds. a ) FSBM0 = 27% soybean meal + 0% fermented soybean meal, FSBM9 = 18% soybean meal + 9% fermented soybean meal, FSBM13 = 13.5% soybean meal + 13.5% fermented soybean meal. b ) Contained per kilogram of supplement: 250,000 IU of vitamin A, 50,000 IU of vitamin D, 1,500 IU of vitamin E, 2.25 g of Mn, 120 g of Ca, 7.7 g of Zn, 20 g of P, 20.5 g of Mg, 186 g of Na, 1.25 g of Fe, 3 g of S, 14 mg of Co, 1.25 g of Cu, 56 mg of I, and 10 mg of Se. c ) Calculated according to NRC (2001). d ) NFC = 100-% neutral detergent fiber-% crude protein-% ether extract-% ash; Data are presented as mean ± SD. www.nature.com/scientificreports/ RNAlater solution (Life Technologies, Gaithersburg, MD, USA) and stored at -80 °C for DNA extraction and qPCR. The concentrations of VFAs were measured using a capillary column (SGE DBX5, length: 30 m, inside diameter: 0.25 mm; with a 25 micron film; Varian CP-WAX 52 CB, Varian Inc., Palo Alto, CA) on a Varian 3,900 gas chromatograph equipped with a flame ionization detector. The temperature of the injector and detector was 230 °C and 240 °C, respectively. Nitrogen was used as a carrier gas with a flow rate of 1.8 ml min −1 .
The concentration of ruminal NH 3 was measured by a spectrophotometer (Ultramicroplate Reader, ELx808, Bio-Tek Instruments, Inc., Winooski, VT, USA) using a colorimetric method based on the reaction of NH 3 with hypochlorite and then with phenol 54 . Cultivation and isolation of anaerobic bacteria. In order to provide microbial DNA, anaerobic bacteria were cultivated and isolated using fresh rumen fluid from cows. The microbial DNA was used to prepare the qPCR standard curve and to test the integrity of the cDNA. Also, in the event that these samples (fresh rumen fluid from cows) were found to be positive for the target bacteria (using PCR gel electrophoresis), these samples were used as a positive control for the qPCR.
Fresh rumen fluid (50 ml per each replication) was collected from the ruminally cannulated lactating Holstein cows for microbial cultivation. Rumen fluid samples were placed in a plastic bottle previously flushed with CO 2 using an oxygen absorber-CO 2 generator (GasPak, AnaeroPack sachet, Mitsubishi Gas Chemical America, Inc.) to protect it from air 55 . Fresh rumen fluid samples were immediately transported to the laboratory within 30 min using an insulated container and cultivated under anaerobic conditions. Hungate's roll-tube anaerobic technique 56 , modified by Bryant and Burkey 57 , was used for the cultivation of anaerobic bacteria. The anaerobic culture medium used was based on the rumen fluid-glucose-cellobiose-agar (RGCA) composition as follows: 1,000 ml of culture medium contains 399.5 ml of fresh rumen fluid, 216.5 ml of solution IV (including 2.0 g of glucose, 2.0 g of cellobiose, 14.98 g of agar), 149.8 ml of mineral solution I (including 0.3% K 2 HPO 4 ), 149.8 ml of mineral solution II (including 0.6% (NH 4 ) 2 SO 4 , 0.6% NaCl, 0.3% K 2 HPO 4 , 0.06% MgSO 4 , 0.06% CaCl 2 ), 66.7 ml of 6% Na 2 CO 3 solution, 16.7 ml of 3% cysteine hydrochloride solution, and solution III (1 ml of 0.1% resazurin). Cysteine hydrochloride was used to reduce the culture media prior to inoculation 55 . In addition, adequate concentrations of Na 2 CO 3 were added to the culture media in order to maintain the pH of culture at 5.6-5.8. This mixture was put in a dilution bottle and sterilized with an autoclave at 120˚C for 20 min. Next, the solution was allowed to cool about 45˚C and then bubbled with O 2 -free CO 2 gas until resazurin indicator changed from pink to colorless. After solidifying the culture media using agar (2% wt/vol), RGCA (4 ml) was put in a roll tube, autoclaved at 120˚C for 5 min, and held in a water bath at 45˚C for 30 min. Upon inoculation, the tubes were gently rolled in cold water to form a thin layer of rumen fluid-agar around the inside.
Microbial DNA extraction and real-time polymerase chain reaction. Microbial DNA was extracted from rumen samples (pellets derived from the experimental rumen fluid of the calves) and from total bacterial colonies (derived from the fresh rumen fluid collected from cows to provide positive control). The extraction was fully mixed (2 mg) with 500 μl of cetyltrimethyl ammonium bromide (CTAB) lysis buffer (composed of 2% CTAB, 100 mmol Tris-HCL, 25 mM EDTA and 2 mol NaCl) and incubated at 65 °C for 15 min according to the CTAB method 58,59 . The tubes were then centrifuged at 14,000 × g for 5 min at 4 °C, and the resulting supernatant was transferred to 2 ml microcentrifuge tube containing 300 μl of chloroform: isoamyl alcohol (24:1). The mixture was centrifuged at 14,000 × g for 5 min at 4 °C, then precipitated with 200 μl of ethanol 70%, centrifuged at 14,000 × g for 5 min at 4 °C and allowed to air-dry at room temperature. The resulting pellets were collected and dissolved in a Tris-EDTA buffer (100 μl) to dissolve DNA; the extracted DNA was stored at − 80 °C for further use 58,59 . In the event that samples (from cows) were found to be positive for the target bacteria using PCR product gel electrophoresis, microbial DNA was extracted from these samples and used as positive control for the qPCR assay.
In this study, three cellulolytic bacteria, including Ruminococcus albus, Ruminococcus flavefaciens, Fibrobacter succinogenes, and two proteolytic bacteria, including Butyrivibrio fibrisolvens and Prevotella ruminicola, were qPCR-tested species. The abundance of bacteria in rumen fluid was measured using qPCR as a proportion of the total bacterial 16S rDNA based on the following formula 60-63 : The Cycle threshold (Ct) was the number of cycles needed for the fluorescent signal to exceed the threshold. The method used to compare the relative expression data in real-time PCR was the delta-delta Ct (2 -ΔΔCt ) method, relative quantification without efficiency correction 64 ; since this method assumes that the target DNA is optimally doubled during each real-time PCR cycle 64 .
The primers used for qPCR and their efficiency are shown in Table 4. The efficiency of the qPCR reaction was sufficient, on average 99.68 ± 2.71 (mean ± SD, Table 4). The results were omitted where Ct values of the samples were greater than Ct of the corresponding negative control (i.e., > 40). In the case of Ruminococcus flavefaciens and Fibrobacter succinogenes, the Ct values of qPCR (for all samples derived from rumen fluid of experimental calves) were greater than Ct of the corresponding negative control. All real-time qRT-PCR reactions were triplicated and conducted on a StepOne Plus™ real time PCR device (Applied Biosystems, Life Technologies, Massachusetts, USA). The reaction mixture (10 μl) consisted of 5 μl Master mix, 1 μl of each primer (10 pmol/ μl), 0.2 μl of 50 × ROX Reference Dye (Invitrogen, Madrid, Spain ×), 1.8 μl ddH 2 O and 1 μl of extracted DNA.
The microbial DNA used for standard curves was directly extracted from the cultivated colonies of cow rumen fluid. In brief, the standards for each real time PCR test were developed using the regular PCR. Then, the PCR product was purified using a QIAquick PCR purification kit (Qiagen, Valencia, CA, USA Quality control of DNA extracts and polymerase chain reactions. In order to test the integrity of cDNA, PCR products gel electrophoresis was conducted on a number of fresh rumen fluid samples obtained from lactating cows ( Supplementary Fig. 1). Next, PCR gel electrophoresis was conducted on experimental samples derived from experimental calves. The integrity of microbial DNA was tested using gel electrophoresis with a 1% (w/v) agarose gel. The purity and quantity of microbial DNA was evaluated using a spectrophotometer, Eppendorf BioPhotometer D30 (Eppendorf AG, Hamburg, Germany). The 260/280 ratio of 1.8-2.0 confirmed the purity of DNA. Tris acetate EDTA buffer (0.5X) was used for the preparation of the agarose buffer. PCR was performed in 15 μl aliquots consisting of 7.5 μl of PCR Master Mix (Takara Bio Inc., Japan), 2.0 μl of each primer (10 pmol/μl, forward and reverse, Table 4), 6.0 µl of nuclease-free water, and 1.0 µl of DNA template (75 ng/µl). PCR was carried out using the PCR system (Master Cycler, Eppendorf, Germany) under the following conditions (37 cycles): initial denaturation at 94 °C for 120 s, final denaturation at 94 °C for 25 s, annealing at 60 °C for 20 s, extension at 72 °C for 20 s, and final extension72°C for 7 min. In addition, a subset of negative samples (without DNA template but with primers) was used in conjunction with the test samples to determine the specificity of qPCR. Then, a limit (i.e., < 40 Ct) was set on the basis of the Ct observed for the negative sample at each run. Positive controls included a DNA template that was synthesized from fresh rumen fluid samples (from cows); these samples were cultivated on anaerobic culture medium and displayed a band using PCR gel electrophoresis.
Chemical analysis. The ingredients and the chemical compositions of the calf starters are shown in Table 3.
Nutrient analysis of the starter feed was carried out according to AOAC (2016) 65 : dry matter (DM, method 925.40), crude protein (CP, method 2001.11), ether extract (EE, method 920.39), ash (method 942.05), starch (method 996.11), neutral detergent fiber (NDF, using 100 μl/heat resistant alpha-amylase sample, sodium sulfite was used during ND extraction to reduce nitrogenous contamination), and acid detergent fiber (ADF) using an Ankom Fiber Analyzer System (Ankom Technology, Macedon, NY) 66 . Non-fiber carbohydrate (NFC) was calculated as the following formula: For the determination of large-peptide N (LPep N) and small-peptide plus AA N (SPep + AA N), 30 mg of feed samples were weighed into a glass-stoppered flask and mixed with tungstic acid (TA) and trichloroacetic acid (TCA) as described by Licitra et al. 17 . Then, TA-soluble N (TA N) and TCA-soluble N (TCA N) were used to calculate LPep N and SPep + AA N according to the following formula: For ammonia extraction, 20 g samples were put into glass-stoppered flasks, mixed with approximately 60 ml of 2 M KCl, stirred for 30 min, and then filtered through Whatman filter paper (no. 54). NH 3 -N was measured by phenol-hypochlorite assay in the extract ether 67 . For the determination of peptides, five separate samples (i.e., experimental starters, SBM and FSBM) were used.  www.nature.com/scientificreports/ Table 1 shows the protein content, the CNCPS protein fractions and the peptide and AAs profiles of SBM and FSBM (n = 5, five different feed samples). The values of rumen undegradable protein (RUP) and rumen degradable protein (RDP) were calculated using the equation below (NRC, 2001) 68 : where k d is the digestion rate of fraction B, and k p is the passage rate of undigested feed.
The AAs content of the feed samples was measured using ion-exchange HPLC with post-column ninhydrin derivatization and fluorescence detection. To this aim, SBM and FSBM samples (n = 5, five separate feed samples) were first hydrolyzed with HCl (containing 0.1% phenol) for 24 h at 110 °C. The resulting chromatograms were incorporated using specialized software (LKB, Biochrom 20; AminoAcid Analyzer, Biotronik GmbH, Maintal, Germany). Cysteine and methionine were analyzed as cysteic acid and methionine sulfone, respectively, by oxidation with performic acid for 16 h at 0 °C. Hydrobromic acid was used after oxidative treatment to consume excess/residual peroxide 69 . For the determination of AAs, five separate samples (i.e., SBM and FSBM) were used.