Digestive Ability, Physiological Characteristics, and Rumen Bacterial Community of Holstein Finishing Steers in Response to Three Nutrient Density Diets as Fattening Phases Advanced

The aim of this study is to track the dynamic alterations in nutrient intake and digestion, rumen fermentation and plasma metabolic characteristics, and rumen bacterial community of Holstein finishing steers in response to three nutrient density diets as fattening phases advanced. A total of eighteen Holstein steers were randomly allocated into three nutrient density groups and steers in each group were fed under a three-phase fattening strategy, with nutrient density increased in each group when fattening phase advanced. Results showed that both fattening phase and dietary nutrient density significantly influenced the nutrient digestion, most of the rumen fermentation parameters, and part of bacteria at phylum and genus levels. Individually, dietary nutrient density affected the concentrations of plasma alanine aminotransferase and urea N, bacterial richness and evenness. All determined nutrient intake and plasma biochemical parameters, except for alanine aminotransferase and triglyceride, differed among fattening phases. Spearman correlation analysis revealed strong correlations between fiber intake and bacterial richness and evenness, rumen fermentation characteristics and certain bacteria. Moreover, Patescibacteria abundance was positively correlated with ambient temperature and plasma total protein. These results indicate that rumen fermentation and nutrient digestion were influenced by both dietary nutrient density and fattening phase, and these influences were regulated by certain rumen bacterial community and ruminal bacteria may be affected simultaneously by ambient temperature. This study may provide insights into diet optimization and potentially adaptive mechanism of rumen bacterial community in response to fattening phases and gradually climatic change.


Introduction
The rumen is of peculiarity to ruminants and harbors complex microbiota, including bacteria, protozoa, archaea, and fungi, which together play pivotal roles in transforming plant material into exploitable nutrients. Among these ruminal microorganisms, bacteria are the most diverse population and have been investigated intensively in recent years [1][2][3]. Many studies have revealed the correlations 1 H = high energy and high protein diet group; C = moderate energy and moderate protein diet group; L = low energy and low protein diet group. The ingredients and nutrient composition were in accordance with our previous study ). 2  Feed intake and refusal were recorded daily, and the temperature and humidity were collected a quarter of an hour during the 334-day trail. Plasma samples were collected from the coccygeal vein into evacuated tubes with heparin sodium two hours before morning feeding on last day of each month, and then centrifuged at 4000 rpm/min for 20 min to obtain the plasma. Plasma samples were aliquoted into 1.5-mL centrifuge tubes and stored at −80 • C before determination by an automatic biochemical analyzer (Hitachi 7020; Hitachi Co., Tokyo, Japan).
Ruminal content was obtained three hours after morning feeding every month as described by Paz et al. [14], wherein both solid and liquid fractions were collected using esophageal tubing. Fecal samples were collected on three consecutive days every month, with sampling interval of six hours on each day. Feces from three days in each month of each steer were equally mixed into one sample for chemical analysis.

DNA Extraction and Volatile Fatty Acids (VFA) Determination
One hundred and twenty-six ruminal samples (42 samples from each fattening phase, with 4, 5, and 5 in H, C, and L diets, respectively) were extracted using a DNA Kit (OMEGA, Omega Bio-Tek, Norcross, GA, USA) as described by Qiu et al. [8]. Purity and quality of genomic DNA were checked on 1% agarose gels and quantified using a spectrophotometer (NanoDrop 2000 Technologies Inc., Wilmington, DE, USA).
Ruminal pH was immediately determined by a portable pH meter (Testo 205, Testo AG, Schwarzwald, Germany) after collecting digesta from rumen. Ruminal contents were centrifuged at 20,000× g for 20 min at 4 • C, and the supernatants were the original samples for subsequent VFA and ammonia nitrogen (NH 3 -N) determination. VFA measurement was according to our previous report [9]. Briefly, 800 µL separated ruminal fluid was mixed evenly with 200 µL internal standard solution (filling 25 g metaphosphoric acid and 217 µL 2-ethylbutyric acid with sterile water to 100 mL) and the determination was by means of a gas chromatograph (GC-2014 Shimadzu Corporation, Kyoto, Japan) with a 30-m capillary column (Rtx-Wax, 0.25 mm ID × 0.25 µm film, Restek, Evry, France) and nitrogen as the carrier gas. The oven program was well optimized as follows: initial 110 • C for 30 s, up to 120 • C at a constant rate of 10 • C/min, 120 • C hold for 4 min, and gradually rise to 150 • C in 3 min. The split ratio and flow rate were controlled at 20:1 and 2.5 mL/min, respectively. NH 3 -N was determined according to Broderick and Kang [15] using the phenol-sodium hypochlorite colorimetry.

PCR Amplification, Sequencing, and Analysis
The primers and PCR reaction system used in the current study were according to our previous report [8]. Briefly, the V3 to V4 was selected as the target hypervariable region, which was amplified with the following primers: 338F (5 -ACTCCTACGGGAGGCAGCAG-3 ) and 806R (5 -GGACTACHVGGGTWTCTAAT-3 ). A well-constructed 25-µL reaction system was performed as 12.5 µL of KAPA 2G Robust Hot Start Ready Mix, 2 µL equal amount of forward and reverse primers (5 µM), 5 µL of ruminal DNA, and 5.5 µL of sterile double distilled water. The well-mixed reaction was amplified under the following program: 95 • C for 5 min; followed by 32 cycles of 95 • C for 45 s, 55 • C for 50 s, and 72 • C for 45 s; and 72 • C for 10 min. Each sample was performed in triplicate, and PCR products were evaluated on 2% agarose gels and purified using an DNA Gel Extraction kit (AxyPrep, Axygen Biosciences, Union City, CA, USA). Paired-end sequences were obtained via the Illumina MiSeq platform (San Diego, CA, USA).
The raw data were analyzed using QIIME (version 1.9.1, https://qiime.org/). Paired end reads were merged by FLASH (version 1.2.11, http://ccb.jhu.edu/software/FLASH/). The length of the sequences was controlled at minimum of 260 bp and maximum of 500 bp. Sequences that contain any ambiguous base or chimera, with quality score of below 20 or mismatching to primer sequences were out of consideration. Filtered sequences were clustered into operational taxonomic units (OTUs) at a similarity level of 0.97 using UPARSE method (USEARCH v11.0.667, http://www.drive5.com/usearch/). OTUs across all samples were rarefied to the lowest sample depth (38,708 reads) based on the pseudo-random number generator of QIIME. Alpha diversity metrics were calculated using Mothur (version 1.39.5, Patrick Schloss, Ann Arbor, USA) [16]. The Ribosomal Database Project (RDP) classifier (http://sourceforge.net/projects/rdp-classifier/) was selected to taxonomy classifications by assigning against the SILVA bacterial alignment database version 132 [17]. Spearman correlation coefficients (r) and FDR corrected values (q) were calculated using the Psych packages version 1.8.12 to show correlations between ambient temperature, nutrient intake and apparent digestibility, plasma biochemical parameters, rumen fermentation characteristics, and bacterial composition. The graph was finished by means of GraphPad Prism (version 8.0.2, GraphPad Software, Inc., San Diego, CA, USA).
The raw sequences involved in this study were deposited in the Sequence Read Archive (SRA) of NCBI, and the SRA accession number is PRJNA 573499. It should be noted especially that 30 ruminal samples from phase 1 had been taken to explore the sampling frequency regarding rumen bacterial community composition in our previous study [8], but the analysis methods and purposes were different from the current study.

Chemical Analysis and Apparent Digestibility Calculation
The dry matter (DM, method 934.01), crude protein (CP, method 990.03), ether extract (EE, method 920.39) and ash (method 942.05) of feeds, orts and feces were analyzed according to the methods of the Association of Official Analytical Chemists [18]. Neutral detergent fiber (NDF) and acid detergent fiber (ADF) were measured according to Van Soest et al. [19], with the former assayed with alpha amylase. Acid-insoluble ash was selected as a natural marker to estimate the total feces, and the apparent digestibility of nutrient was calculated using the method described by Niu et al. [12].

Statistical Analysis
The average daily temperature-humidity index (THI) was calculated according to Vitali et al. (2009), which is expressed as (1.8 × T + 32) -(0.55 -0.55 × H/100) × [(1.8 × T + 32) -58], where T denotes temperature ( • C) and H indicates relative humidity (%); THI greater than 74 implied that cattle suffered from heat stress [20]. All three samples from the same steer in each fattening phase were averaged as one value to show a better presentation, because our previous study [8] showed that ruminal bacteria were in dynamic variation before an adaptation period of three months. Therefore, each fattening phase had 18 average values for each parameter, except for the bacterial community. All averaged data were analyzed as a repeated measures design in Mixed Models procedure of SPSS (version 21, IBM Corporation, Armonk, New York, United States) with first-order autoregressive covariance structure after comparing compound symmetry, first-order ante dependence and unstructured covariance structure based on the minimum Akaike information criterion (AIC) values. The statistical model used for this analysis is expressed as follows: y ijt = µ + α i + γ t + c j(i) + (αγ) it + e ijt , where y ijt is the parameter measured at fattening phase t on the jth cattle allotted to the ith diet, µ is the overall mean, α i is the ith fixed diet effect, γ t is the fixed tth fattening phase effect, c j(i) is the random effect of the jth cattle within the ith diet, (αγ) it is the fixed interaction effect between diet and fattening phase, and e ijt is the random error in the jth cattle allotted to the ith diet in fattening phase t. Multiple comparison was conducted using Tukey tests and significance was declared at 0.05 (p < 0.05).

Temperature-Humidity Index, Nutrient Intake, and Apparent Digestibility
The average semimonthly temperature and humidity are presented in Figure S1. Days with THI over 74 were 0, 4, and 69 in phase 1, phase 2 and phase 3, respectively. Additionally, the proportion of days with THI over 74 to all days (69/154) in last phase was 44.81%. The nutrient intake and apparent digestibility of nutrient are shown in Table 2. All determined nutrient intake differed significantly (p < 0.001) in three fattening phases, with the highest in phase 2, followed by phase 1 and phase 3. The intakes of CP and EE increased as dietary nutrient density increased, whereas intakes of NDF and ADF decreased as dietary nutrient density increased (p < 0.05), with no differences (p > 0.10) in intakes of DM and OM due to dietary treatments. No interaction effects (p > 0.10) were found in nutrient intakes between fattening phase and dietary treatment.
Both fattening phase and dietary treatment influenced the apparent digestibility of measured nutrients. The apparent digestibility of DM, OM, CP, and EE were higher (p < 0.001) in phase 3 as compared to phase 1 and phase 2, whereas apparent digestibility of NDF and ADF were higher (p < 0.001) in phase 1 when compared with phase 2 and phase 3. The apparent digestibility of DM, OM, CP, and EE were higher (p < 0.05) in steers fed H diet than steers fed C and L diets, and opposite results were observed in the apparent digestibility of NDF and ADF. Interaction effects were observed in the apparent digestibility of ADF and EE between fattening phase and dietary treatment.

Plasma Metabolic Parameters
The plasma metabolic parameters are summarized in Table 3. Fattening phase significantly influenced (p < 0.05) the detected metabolic parameters, except for alanine aminotransferase (ALT) and triglyceride. The concentration of ALT was higher (p = 0.005) in steers fed L diet than that in steers fed H and C diets, and opposite result was observed in concentration of urea N (p < 0.001). Interaction effects were observed in concentrations of glucose, triglyceride, and urea N between fattening phase and dietary treatment.

Rumen Fermentation Characteristics
The rumen fermentation characteristics are shown in Table 4. Rumen pH, VFA, and total VFA (TVFA) significantly differed (p < 0.05) in three fattening phases, whereas no differences (p > 0.05) were observed in concentration of NH 3 -N and ratio of acetate to propionate (A/P). Dietary nutrient density altered (p < 0.05) most of the rumen fermentation profiles, except for concentrations of NH 3 -N and TVFA. Interaction effects, except for concentrations of NH 3 -N and acetate, were observed between fattening phase and dietary treatment.

Rumen Bacterial Community
A total of 16, 340, 654 sequencing reads were obtained after filtering low-quality data, with 99.99% of high-quality sequences between the lengths of 380 bp and 440 bp. After clustering high-quality sequences, an average of 1446 OTUs was obtained for each sample. The converged and slow-growing Shannon-Wiener curves ( Figure S2) indicated that the sequencing depth in the current study was adequate to evaluate the vast composition of ruminal bacteria.
The estimation of alpha diversity is listed in Table 5. No differences (p > 0.05) in richness and evenness were observed among fattening phases. Chao 1, observed species, PD whole tree, and Shannon index increased as dietary nutrient density decreased. An interaction was observed in Shannon index between fattening phase and dietary treatment.
As is shown in Table 6, ten phyla were observed with relative abundance over 0.5%, and Bacteroidetes and Firmicutes dominated the most two. Fattening phase affected (p < 0.05) abundances of Bacteroidetes, Firmicutes, Proteobacteria, Fibrobacteres, and Patescibacteria, whereas dietary treatment had significant effects (p < 0.05) on eight out of ten phyla, except for Bacteroidetes and Actinobacteria. Besides, no interactions (p > 0.05) were observed between fattening phase and dietary treatment regarding phyla with relative abundance greater than 0.5%.
The taxonomic analysis with average abundance over 0.5% at the genus level is shown in Table 7. Twenty-six genera were observed with relative abundance greater than 0.5%, with Prevotella predominated 32.78%, 29.29%, and 26.86% in H, C, and L diets, respectively. Eight and ten out of twelve with relative abundance over 1% were affected (p < 0.05) by fattening phase and dietary treatment, respectively. Of these, no interactions (p > 0.05) were observed between fattening phase and dietary treatment. However, interactions were found in Veillonellaceae UCG-001, Prevotellaceae NK3B31 group, and Saccharofermentans between fattening phase and dietary treatment, whose relative abundances were below 1%.

Correlation Analysis
Correlations between ambient temperature, nutrient intake and apparent digestibility, plasma metabolic and rumen fermentation characteristics, and rumen bacterial composition are shown in Figure 1. Ambient temperature was positively associated with the phyla Firmicutes and Patescibacteria (r > 0.7 and q < 0.01), and it also correlated negatively with the phylum Bacteroidetes (r = −0.770 and q < 0.001). The estimated alpha diversity (both richness and evenness) was negatively associated with propionate (r < −0.7 and q < 0.01) and positively associated with intakes of NDF and ADF and A/P (r > 0.7 and q < 0.01); richness (Chao 1 and observed species) correlated negatively with isovalerate, valerate, and total branched-chain VFA (BCVFA) (r < −0.7 and q < 0.01).
The intakes of NDF and ADF were positively correlated with phyla Kiritimatiellaeota, Cyanobacteria, and Fibrobacteres (r > 0.7 and q < 0.01). The apparent digestibility of NDF and ADF and rumen pH were negatively associated with the genus Succiniclasticum (r < −0.7 and q < 0.01), whereas the relative abundance of Succiniclasticum was positively correlated with propionate, isovalerate, valerate, and total BCVFA (r > 0.7 and q < 0.01). The plasma total protein correlated positively with the phylum Patescibacteria (r = 0.734 and q < 0.01). The concentration of propionate was negatively correlated with the genus Rikenellaceae RC9 gut group (r = −0.865 and q < 0.01). The A/P was positively associated with the relative abundance of Kiritimatiellaeota (r = 0.837 and q < 0.01).

Correlation Analysis
Correlations between ambient temperature, nutrient intake, and apparent digestibility, plasma metabolic and rumen fermentation characteristics, and rumen bacterial composition are shown in Figure 1. Ambient temperature was positively associated with the phyla Firmicutes and Patescibacteria (r > 0.7 and q < 0.01), and it also correlated negatively with the phylum Bacteroidetes (r = −0.770 and q < 0.001). The estimated alpha diversity (both richness and evenness) was negatively associated with propionate (r < −0.7 and q < 0.01) and positively associated with intakes of NDF and ADF and A/P (r > 0.7 and q < 0.01); richness (Chao 1 and observed species) correlated negatively with isovalerate, valerate, and total branched-chain VFA (BCVFA) (r < −0.7 and q < 0.01).
The intakes of NDF and ADF were positively correlated with phyla Kiritimatiellaeota, Cyanobacteria, and Fibrobacteres (r > 0.7 and q < 0.01). The apparent digestibility of NDF and ADF and rumen pH were negatively associated with the genus Succiniclasticum (r < −0.7 and q < 0.01), whereas the relative abundance of Succiniclasticum was positively correlated with propionate, isovalerate, valerate, and total BCVFA (r > 0.7 and q < 0.01). The plasma total protein correlated positively with the phylum Patescibacteria (r = 0.734 and q < 0.01). The concentration of propionate was negatively correlated with the genus Rikenellaceae RC9 gut group (r = −0.865 and q < 0.01). The A/P was positively associated with the relative abundance of Kiritimatiellaeota (r = 0.837 and q < 0.01). Figure 1. Correlations between ambient temperature, physiological and digestive characteristics, and bacterial abundances of Holstein finishing steers fed three nutrient density diets. Only significant correlations (r > 0.70 or r < −0.70 and q-value < 0.05) and bacteria abundances greater than 0.5% are shown. The colors indicate positive (red, closer to 1) or negative (green, closer to −1) correlations between investigated characteristics and bacterial abundances. A/P, acetate to propionate ratio; TBCVFA, total branched-chain volatile fatty acids. Correlations between ambient temperature, physiological and digestive characteristics, and bacterial abundances of Holstein finishing steers fed three nutrient density diets. Only significant correlations (r > 0.70 or r < −0.70 and q-value < 0.05) and bacteria abundances greater than 0.5% are shown. The colors indicate positive (red, closer to 1) or negative (green, closer to −1) correlations between investigated characteristics and bacterial abundances. A/P, acetate to propionate ratio; TBCVFA, total branched-chain volatile fatty acids.

Digestive and Physiological Characteristics
It is generally accepted that nutrient intake increases as animals grow, so as to meet the requirement of maintenance and weight gain [13]. However, nutrient intake is affected by both feed intake and dietary nutrient density. In this study, nutrient intake decreased in phase 3 because of low feed intake because steers suffered from heat stress at over 40% of time in phase 3. Besides, nutrient intake increased as dietary nutrient density improved as a result of similar feed intake in each phase ( Table 2). Steers in phase 3 had higher apparent digestibility of DM, OM, CP, and EE, which is similar to the reports of Yadav et al. [11], and these differences could be due to longer retention time of digesta in phase 3 because previous study revealed digesta retention time increased as ambient temperature increased when temperature was below 35 • C [11]. However, this effect was quite different in the apparent digestibility of NDF and ADF. Previous study reported that the activity of cellulolytic bacteria was partly inhibited by low ruminal pH and further reduced the digestibility of NDF and ADF [21]. In this study, ruminal pH values were lower in phase 3 as compared to that in phase 1 and phase 2, thereby a lower digestibility of NDF and ADF was observed. Steers fed H diet improved the apparent digestibility of DM, OM, and EE, which could be attributed to the high starch content in H diet, a digestible nutrient for ruminants. The higher apparent digestibility of CP in H diet was similar to the result of Menezes et al. [22], who suggested a positive association between apparent digestibility of CP and protein level. Dietary nutrient density decreased the apparent digestibility of NDF and ADF, which is in accordance with the results of Tjardes et al. [23], who suggested that high-fiber diet increased the apparent total-tract digestibility of both NDF and ADF. These differences could be explained by the longer rumen retention time and more cellulolytic bacteria induced by higher content of fiber [21].
Previous study [24] reported that most blood biochemical characteristics changed with ages. In this study, albumin, cholesterol, glucose, total protein and urea N differed among fattening phases because of generally physiological reasons, which are consistent with the reports of Mohri et al. [24]. The urea N is the main end-product of protein hydrolysis and amino acid metabolism [25], and its concentration has shown to be positively associated with intake of CP [12]. Therefore, it was reasonable that plasma urea N increased as dietary nutrient density increased as a result of increased CP intake.
The ruminal pH is a vital ruminal characteristic because of its direct influence on microbial growth, and the adaptable pH for microbiota ranges from 5.5 to 7.0 [26]. In the current study, the ruminal pH values were in the tolerable range and decreased as dietary nutrient density improved and fattening phase advanced. High-density diet decreased the saliva production and chewing activities [27], and subsequently diminished the buffering capacity of suddenly increased VFA because of the high intake of fermentable carbohydrates. In addition, these differences among diets were more obvious in the last fattening phase, indicating that long-term stimulation would enlarge the effect of diets on ruminal pH, as seen by the interaction effect between fattening phase and diet regarding ruminal pH values. Generally, structural carbohydrates are mainly fermented into acetate and nonstructural carbohydrates produce more propionate [13], which was also mirrored in L and H diets in the present study, respectively. The A/P increased as dietary nutrient density decreased in the current study, which is in accordance with the fact that the A/P increased as the forage to concentrate ratio increased [28]. BCVFA usually derives from degradation of crude protein and has been used as an indicator of ruminal protein fermentation [29]. Higher concentrations of isobutyrate and isovalerate, as well as total BCVFA, were observed in H and C diets because of higher CP intake in these two diets. Our results showed high concentration of butyrate in H diet, which was similar to the report of Xia et al. [28] wherein bulls fed diet with high concentrate to forage ratio yielded high concentration of butyrate. Previous study [30] reported an elevated valerate concentration as the proportion of dietary concentrate increased. In this study, similar results in valerate concentration were obtained in terms of both diet and fattening phase, because higher dietary nutrient density, as well as later fattening phases, contained more fermentable nonstructural carbohydrates.

Rumen Bacterial Community Profiles
Alpha diversity metrics are used to estimate the species richness and evenness in a certain sample or a single community [31]. In this study, rumen samples from steers fed C and L diets showed higher diversity as compared to H diet, which is in line with many reports [3,6,7] in which grain-based diet decreased alpha diversity. These differences may be explained by the well-established theory that ruminal pH has a large impact on the rumen bacterial diversity [32].
Bacteroidetes and Firmicutes dominated about 90% of the bacterial composition, with 62.88% and 27.15%, respectively. The higher abundance of Firmicutes in steers fed H and C diets was due to lower ruminal pH in these two diets because previous study reported that lower ruminal pH increased the proportion of Firmicutes in bacterial composition [32]. Proteobacteria plays an important role in the rumen metabolism despite of the relative low abundance, and was frequently observed in starch-rich diet [3,33]. However, present study showed the opposite result wherein high abundance of Proteobacteria was observed in fiber-rich diet. It is tempting to assume that certain species in the phylum Proteobacteria may also actively take part in the digestion of fiber, but further studies are needed to confirm this assumption and certain species. Kiritimatiellaeota, a novel phylum previously assigned to the Verrucomicrobia subdivision 5, occupied niches mainly characterized by anaerobic environment [34]. The relative abundance of Kiritimatiellaeota increased as dietary nutrient density decreased, indicating that this phylum may be involved in fiber degradation. Fibrobacteres, as well as Fibrobacter, is well-known for its vital role in degrading cellulose, and they are commonly detected in fiber-rich diet [3]. Therefore, it is obvious to expect a decreasing abundance as dietary nutrient density increased in the present study.
Significant difference in Prevotella abundance was observed between steers fed H and L diets because of its function in starch fermenting and protein metabolism [35]. Similar result was observed in relative abundance of Prevotellaceae UCG-001, a genus belongs to the same family Prevotellaceae. However, Prevotellaceae UCG-003 showed the opposite results, which increased as dietary nutrient density decreased. These results regarding Prevotellaceae UCG-003 were similar to the report of Liu et al. [6] wherein higher proportion of Prevotellaceae UCG-003 was found in forage-based feeding group, indicating that different genera in the same family may involve in diverse metabolic function. Rikenellaceae RC9 gut group was previously regarded to be involved in the degradation of carbohydrates [36] and recent study in Holstein bulls has revealed the negative association between propionate concentration and Rikenellaceae RC9 gut group abundance [33]. In this study, the relative abundance of Rikenellaceae RC9 gut group decreased as dietary nutrient density increased which resembled to our previous research in feces [9], indicating that this genus may also be involved in degrading fiber in the rumen. The present study emerged a cumulative increase in Succiniclasticum abundance as dietary nutrient density increased, which is similar to previous report wherein high proportion of this genus was found in high-grain feeding pattern [6]. Previous study observed that Yak fed with more concentrate harbored higher abundance of Ruminococcaceae NK4A214 group and Ruminococcus 2 [6], which applied equally to Holstein steers fed with high density diet in the present study, indicating these two genera had the potential to degrade nonfibrous material. Ruminococcaceae UCG-011, a genus belonging to the family Ruminococcaceae, has been reported to be cellulose utilizer in rumen of beef cattle and higher proportion of this genus was observed in forage-based diet as compared to grain-rich diet [6], this is similar to our findings wherein relative abundance of Ruminococcaceae UCG-011 increased as the dietary nutrient density decreased.

Correlations between Ambient Temperature, Physiological and Digestive Characteristics, and Ruminal Bacterial Community
In this study, correlation analysis revealed the relationships between ambient temperature and rumen bacterial abundances. The superphylum Patescibacteria was commonly detected in anoxic environments and potentially involves anaerobic fermentative metabolisms [37]. Therefore, it is possible that Patescibacteria did exist in rumen because of anaerobic environment in rumen. In this study, Patescibacteria abundance was positively correlated with ambient temperature, indicating that Patescibacteria abundance in rumen could be influenced by ambient temperature, which could also be seen from higher abundance of this superphylum in phase 3 than in phase 1 and phase 2 ( Table 6). Previous study reported opposite variation trend in relative abundances of Bacteroidetes and Firmicutes from winter to summer [10], and similar results were also observed in the present study, with negative correlation between ambient temperature and Bacteroidetes and positive correlation between ambient temperature and Firmicutes. Interestingly, one of the plasma metabolic parameters, total protein, was observed to be positively associated with the abundance of the superphylum Patescibacteria. The concentration of total protein depends on the absorption of globulin and albumin [38]. The present results indicate that Patescibacteria may indirectly accelerate the absorption of total protein. To our knowledge, this is the first time to observe association between plasma total protein concentration and Patescibacteria abundance.
The intake of fiber (NDF and ADF) correlated positively with alpha diversity indexes. This could be explained from two aspects. On the one hand, more intake of fiber indicates greater degradable fiber and fiber digestibility ( Table 2), and greater fiber degradation indicates greater substrate and metabolite releases [39], thus the microbial populations have greater concentration of substrates for their living. On the other hand, greater intake of fiber carbohydrates keeps optimal ruminal acidity by stimulating rumination and saliva production [27], thus provides a favorable environment for the microbiota to grow. In addition, positive correlations were observed between NDF and ADF intake and Fibrobacteres, Kiritimatiellaeota, and Cyanobacteria. Among these phyla, Fibrobacteres is a typically cellulolytic phylum; Kiritimatiellaeota is potentially involved in fiber degradation; and Cyanobacteria was found to participate in degrading plant polysaccharides [1]. Succiniclasticum plays vital roles in degrading starch and transforming succinate into propionate [40], which explained the positive association between Succiniclasticum and propionate concentration. As described previously, propionate is the main product fermenting from nonstructural carbohydrates, and this type of carbohydrates are fermented rapidly and may restrict fiber fermentation [27], resulting in quite low ruminal pH which is unsuitable for microbial growth. Therefore, it is reasonable to see negative correlations between propionate concentration and bacterial diversity, as well as negative correlations between bacterial diversity and digestibility of NDF and ADF, and between bacterial diversity and rumen pH value. Moreover, propionate concentration correlated negatively with Rikenellaceae RC9 gut group abundance, which is very similar to the report of Wang et al. [33], indicating potential function of this genus in impeding propionate fermentation. A/P represents the ratio of structural carbohydrates to nonstructural carbohydrates in diet because of their main end product are acetate and propionate, respectively [13]. As described above, nonstructural carbohydrates generally ferment rapidly and result in low bacterial diversity due to lower ruminal pH, thus positive associations were observed between A/P and alpha diversity indexes. A/P was also positively correlated with Kiritimatiellaeota, proving again the involvement of this phylum in degradation of structural carbohydrates. Valerate is one of the products resulting from the degradation of branched amino acids and itself serves as an important precursor for synthesizing odd branched-chain fatty acids [6]. It has also been suggested that odd branched-chain fatty acids could be synthesized by elongating propionate [41]. Hence, it is reasonable to observe positive correlation between Succiniclasticum abundance and concentrations of propionate, isovalerate, valerate, and total BCVFA due to vital roles of this genus in fermenting succinate to propionate and degradation of protein [12,40].

Conclusions
In summary, both dietary nutrient density and fattening phase significantly impacted nutrient intake and digestion, rumen fermentation characteristics and rumen bacterial community. Besides, dietary nutrient density affected concentrations of ALT and urea N in plasma, rumen bacterial richness and evenness, while fattening phase influenced most of metabolic characteristics, except for ALT and triglyceride. The interaction effects between dietary treatment and fattening phase throughout the investigation indicate that continuous stimulations deepen the effect of dietary nutrient density on physiological and digestive characteristics, and bacterial community. Furthermore, correlation analysis indicates that diet and fattening phase, as well as ambient temperature, play crucial role in regulating digestive and physiological metabolisms and rumen bacteria.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-2607/8/3/335/s1, Figure S1: Dynamic changes in ambient temperature ( • C) and relative humidity (%) and temperature-humidity index semimonthly as fattening phases advanced. Figure S2: Shannon-Wiener curves based on Shannon index for high energy and high protein diet (H), moderate energy and moderate protein diet (C), and low energy and low protein diet (L).  The APC was funded by the same funds described above.