Dynamic alterations in rumen bacterial community and metabolome characteristics in response to feed nutrient levels of cashmere goats

Background: Dietary energy and protein play important roles in rumen fermentation. However, the comprehensive impacts of dietary energy and protein on rumen bacterial composition and ruminal metabolites were largely unknown. Therefore, the objective of the current study was to investigate the changes in rumen bacterial community and metabolites in response to the diets with simultaneous changes of dietary energy and protein levels in Shaanbei white cashmere goats (SWCG). Methods: A total of 12 ruminal samples were collected from SWCG, which were divided into two groups, including high-energy and high-protein (Group H; crude protein, CP: 9.37% in dry matter; metabolic energy, ME: 9.24 MJ/kg) or control (Group C; CP: 8.73%; ME:8.60 MJ/kg) groups. The experiment lasted for 65 days, including 10 days for adaptation. 16S rRNA gene sequencing and quantitative real-time polymerase chain reaction (qRT-PCR) were performed to identify the rumen bacterial community. Metabolomics analysis was done to investigate the rumen metabolites and the related metabolic pathways in Groups C and H. Results: The study observed 539 genera belonging to 30 phyla, which were distributed throughout the rumen samples. The high-energy and high-protein diets increased the relative abundance of phylum Bacteroidetes and genera Prevotella_1 and Succiniclasticum, while decreased the number of phylum Proteobacteria (p<0.05). Among the 24 differential metabolites (VIP>1.0, p<0.05) detected in this study, the dominant differential metabolites were amino acids, peptides and analogs. Tyrosine metabolism played an important role among the 9 main metabolic pathways. Correlation analysis revealed that Prevotella_1 showed strong positive correlation with 5-methoxyindole-3-acetic acid (r=0.601, p<0.05) and catechol (r=0.608, p<0.05). Succiniclasticum was positively correlated with 2-ketoadipate (r=0.741, p<0.01). Conclusions: Our ndings revealed trimethylchlorosilane; PCA: principal component analysis; OPLS-DA: orthogonal correction partial least squares discriminant analysis; SPSS: Statistical Package for the Social Sciences; ANOVA: one-way analysis of variance; VIP: variable importance in the projection.


Background
Shaanbei white cashmere goat (SWCG) is a local breed in the northern Shaanxi province of China and the total population of SWCG exceeds 10 million. SWCG is well-known for cashmere wool and meat, which are the most important economic sources of the local farmers [1,2]. Traditional grazing management is mainly dependent on natural pastures, which are limited in the extremely harsh winter. Hence, nutritional management, especially the choice of dietary nutrient levels, is important to promote the growth of goats [3].
The rumen is a complex microbial ecosystem in ruminants. It can ferment feedstuffs to volatile fatty acids (VFAs), microbial proteins and vitamins, which play important roles in animal health and production [4][5][6]. Among the microbiota, bacteria are the most abundant, diverse and metabolically active species in the rumen [7,8]. Bacterial community in the rumen are linked to various factors, such as animal diet, breed, age health and geographic region [9,10]. Diet is the major determinant of the microbial composition in the rumen [11].
The functions of the rumen microbiota make ruminants highly adaptable to various diets [12]. The energy and protein levels in the diets are the most restrictive factors for ruminal microbial growth [13,14]. Dietary protein is utilized to synthesize microbial protein (MCP) for host utilization [15][16][17]. However, protein overfeeding increases the excreted nitrogen from urine and feces, which causes environmental pollution [18] and economic losses [19]. Previous studies have reported that dietary energy can promote protein to synthesize MCP [20][21][22]. The effective way to improve average daily weight gain (ADG) and production performance in cattle is increasing dietary energy levels under the same concentration of forage ratio [5].
Previous studies showed that the phenotypic traits of ruminants were affected by rumen microbiota, whose functions could be re ected by the ruminal metabolites [23][24][25]. However, most studies have only focused on the single change of dietary protein/energy and few reports studied the comprehensive effects of dietary energy and protein on the rumen bacterial composition and rumen metabolites [3,16,18,22,26,27].
In our previous study, we detected that high levels of dietary energy (metabolic energy, ME: 9.24 MJ/kg) and protein (crude protein, CP: 9.37% in dry matter (DM)) could signi cantly enhance the ADG, dressing percentage and eye muscle area of SWCG [28]. The primary objective of this study was to investigate the changes in the rumen bacterial diversity by 16S rRNA gene sequencing and qRT-PCR, and the metabolites

Animals and sampling
A total of 12 SWCG (aged 8 months, an average initial body weight of 24.5 ± 1.87 kg, six males and six females) were selected and fed at Diqingyuan farm (37.6 °N, 108.79 °E) located at Yulin, Shannxi Province, China. Based on their diet types, all goats were randomly allocated into two groups, including high-energy and high-protein (Group H; crude protein, CP: 9.37% in dry matter; metabolic energy, ME: 9.24 MJ/kg) and control (Group C; CP: 8.73%; ME:8.60 MJ/kg) groups (n = 6 per group). The experimental diets were formulated based on the Feeding Standard of Meat-Producing Sheep and Goats (NY/T816-2004, China). In addition, the ratio of dietary energy to protein and the ratio of dietary forage to concentrate in the two groups were not changed. (See Additional le 1: Table S1).
The rumen samples were used for metabolomics analysis. For each sample, 200 µL rumen uid sample, 200 µL liquid methanol and 20 µL L-2-Chlorophenylalanine (CAS#: 103616-89-3, ≥ 98%) (1 mg/mL in H 2 O) as an internal standard were sequentially added to the 1.5 mL Eppendorf (EP) tubes. The mixture was vortexed for 10 s and then centrifuged at 13000 rpm for 15 min at 4 °C. After centrifugation, 370 µL of the supernatant was transferred to a 2 mL GC/MS glass vial and dried in a vacuum concentrator without heating. After evaporation, 80 µL of methoxy amination hydrochloride (20 mg/mL in pyridine) was added to the sample and incubated for 30 min at 80 °C. Meanwhile, 100 µL of the N,O-Bis (trimethylsilyl) tri uoroacetamide (BSTFA) with 1% trimethylchlorosilane (TMCS) were added to the sample aliquots and the mixture was incubated at 70 °C for 1.5 h.
Derivatized samples were analyzed using the Agilent 7890B gas chromatograph system (Agilent, USA) coupled with the LECO Chroma TOF PEGASUS HT (LECO, USA) [35]. Injecting 1 µL aliquot of the analyte into splitless mode, helium was used as the carrier gas. The injection, transfer line and ion source temperatures were 280, 270 and 220 °C, respectively. The mass spectrometry data were performed in fullscan mode with 50-500 m/z at 20 scans/s after a solvent delay of 6.1 min.
The extraction of raw peaks, ltering and calibration of the baseline data, peak alignment, deconvolution analysis, peak identi cation and peak area integration were performed by Chroma TOF 4.3X software [36]. The content of each component was calculated by the peak area normalization method. Principal component analysis (PCA) and orthogonal correction partial least squares discriminant analysis (OPLS-DA) were conducted using SIMCA software (V14.1).

Statistical analysis
Statistical differences were performed by Statistical Package for the Social Sciences (SPSS) version 20.0 (SPSS Inc., Chicago, IL, USA). Data were analyzed using one-way analysis of variance (ANOVA). Signi cant differences between Groups C and H were calculated using Student's t-test. For the GC-TOFMS data, differential metabolites between two groups were identi ed combining variable importance in the projection (VIP) > 1 and p < 0.05. Signi cant correlations between rumen bacteria and metabolite variables were assessed by Spearman correlation analysis, if the correlation coe cients (r, in absolute values) were above 0.55. The statistical signi cance was set at p < 0.05.

Results
Diversity, richness and similarity of the ruminal bacterial communities A total of 826,727 16S rRNA gene sequences were obtained from 12 different samples with 61,658 rare ed sequencing reads per sample. Group C exhibited the highest number of unique sequences (667 OTUs), followed by Group H (35 OTUs). Approximately 71% of the total OTUs (1704 OTUs) were shared among two groups (Fig. 1A). The rarefaction curves (Fig. 1B) reached the saturation plateau and the indices of Good's coverage were above 0.99 (See Additional le 1: Table S3), indicating that the sequencing depth was reasonable. ACE (Fig. 1C) and Chao (Fig. 1D) indices were signi cantly decreased when the goats were fed with high energy and protein diets in Group H (p < 0.05), while Shannon and Simpson indices had no signi cant effects (p > 0.05) (See Additional le 1: Table S3).

Quantitative Real-time Pcr Analysis
According to 16S rRNA gene sequencing data, the differences in the number of Bacteroidetes (phylum level) and Prevotella (genus level) between Groups C and H were further veri ed by absolute qRT-PCR. As shown in Table 2, the number of Prevotella and Bacteroidetes in the rumen of Group H was signi cantly increased (p < 0.05) compared with Group C. Table 2 In uence of different nutrient levels in the diets on the number of bacteria a .

Functional Predictions Of Rumen Bacteria
The potential functions of the bacterial community in the rumen of SWCG were predicted by the PICRUSt2 based on 16S rRNA gene sequencing data. At KEGG level 1, metabolism-related pathways had the highest abundance (> 50%). Compared with Group C, the rumen bacteria of Group H were predicted to have signi cantly higher capability of in uencing metabolism and genetic information processing and lower capability of in uencing environmental information processing, cellular processes and human diseases (p < 0.05) (See Additional le 1: Table S6). At KEGG level 2, the highest relative abundance was carbohydrate metabolism. In addition, the abundances of genes belonged to carbohydrate metabolism, energy metabolism, nucleotide metabolism, glycan biosynthesis and metabolism, biosynthesis of other secondary metabolites, translation, and replication and repair were signi cantly higher in Group H than Group C. The abundances of genes involved in lipid metabolism, membrane transport and signal transduction were signi cantly higher in Group C compared with Group H (Fig. 3, See Additional le 1: Table S7).

Metabolic Pathways Of Differential Metabolites
In order to provide a comprehensive view of the differential metabolites between Groups C and H, pathway analysis was visualized in Fig. 5. The varied rumen microbial metabolites between Groups C and H were identi ed to be mainly involved in the 9 main metabolic pathways, including beta-alanine metabolism; tyrosine metabolism; pantothenate and CoA biosynthesis; sphingolipid metabolism; glutathione metabolism; glycerophospholipid metabolism; pyrimidine metabolism; tryptophan metabolism; and arginine and proline metabolism. These pathways are mainly involved in amino acids metabolism, lipid metabolism and nucleotide metabolism. Additionally, among these metabolic pathways, tyrosine metabolism has the largest impact.

Correlation Analysis Between Rumen Bacteria And Rumen Metabolites
Based on Spearman correlation analysis (|r| > 0.55 and p < 0.05), we constructed the correlation networks between the bacterial genera in Groups C and H, respectively. As shown in Additional le 2: Figure S2A and Figure S2B, 171 and 79 edges were observed in Group C and Group H, respectively, which indicated that the relationships between the bacterial genera in Group C were more complex than those in Group H. The comprehensive relationships between ruminal bacterial genera were observed in this study (See Additional le 3: Table S9). Among them, Prevotella_1 was positively correlated with Succiniclasticum (r = 0.580, p < 0.05) and Ruminococcus_2 (r = 0.651, p < 0.05). Selenomonas_1 was positively correlated with Prevotellaceae_UCG-004 (r = 0.78, p < 0.01) We determined the relationships between the differential metabolites and the top 50 bacterial communities at the genus level ( Fig. 6 and Additional le 4:  Table 3 Significant differential metabolites between Groups C and H (VIP>1.0; p<0.05).

Discussion
The effects of feed nutrient levels on growth performance, carcass characteristics and serum biochemical indices of SWCG have been reported in our previous study [28]. In brief, compared with Group C, ADG (95.37 vs. 81.06 g; p < 0.05), dressing percentage (48.79% vs. 44.04%; p < 0.05) and eye muscle area (21.72 vs. 19.55 cm 2 , p < 0.05) were signi cantly increased in Group H, which indicated that the higher dietary energy and protein levels remarkably improved the growth performance and carcass characteristics of goats in our previous study. In addition, other studies reported that the serum biochemical parameters were sensitive indicators of health status of animals [37][38][39]. In our previous study, few differences in serum biochemical indices between Groups C and H (p > 0.05) suggested the similar healthy status of goats in the two groups [28].
Comparison of the composition and differences of ruminal bacterial communities As mentioned above, diets with high energy and protein levels in Group H effectively promoted the growth performance and carcass characteristics of goats. Meanwhile, the digestion and absorption of these diets were closely related to the rumen bacteria [40][41][42]. A previous study reported that the changes in the ruminal microbiota could help promote the ADG of goats [5]. Therefore, we determined the differences of rumen bacterial communities of SWCG with simultaneous changes of dietary energy and protein levels in this study.
In this study, 16S rRNA gene sequencing was used to assess the rumen bacterial community in SWCG. Liu et al. [43] and Tapio et al. [44] reported that the richness of the bacterial community were in uenced by diets. Our results also revealed low bacterial richness (ACE and Chao indices) with the increasing levels of dietary energy and protein, while no signi cant changes were detected in the bacterial diversity (Shannon and Simpson indices). In line with the previous studies [45,46], this study revealed that Bacteroidetes, Firmicutes and Proteobacteria were the most dominant phyla in the two groups. These bacterial phyla were the core microbiota in the rumen and their structural compositions were unchanged regardless of feeding different types of diets [47]. Among the thirty phyla detected in this study, the abundance of Bacteroidetes increased signi cantly in Group H, which might be related to the protein degradation function of this phylum [48].
The effects of dietary nutrient levels on the bacterial population at the genus level were also detected in this study. Similar to the results of decreased bacterial richness in Group H, the high energy and protein levels of diets reduced the complexity of rumen bacterial interactions. Among the genera detected in the present study, Prevotella_1, belonging to the Bacteroidetes, was the most abundant bacteria in both the rumens of goats fed with different diets, which was consistent with the previous reports [3,5]. In addition, the population of Prevotella_1 was signi cantly increased in Group H. The difference of abundances between the groups might be related to bacterial functions. The genus Prevotella is mainly involved in protein metabolism [3,5,49]. Wang et al. [3] reported that the number of Prevotella_1 was increased when the animals fed with the high protein diet. Succiniclasticum mainly participates in fermenting succinate to propionate, which is the most precursor of glucose in ruminants [50,51]. In our study, the relative abundance of Succiniclasticum signi cantly increased in the Group H. Similarly, the number of Succiniclasticum increased in the high-concentrate diet/high-energy diet group in previous reports [5,11]. Bacteroidetes and Prevotella were selected in this study to verify the differences between two groups by absolute qRT-PCR and the results agreed with those by 16S rRNA gene amplicon sequencing. Furthermore, the previous studies have reported that the level of dietary protein were positively correlated with the relative abundance of Prevotella [26,52] and the increasing levels of protein could promote the growth of cellulolytic bacteria [18]. In this context, the cellulolytic bacteria-Succiniclasticum and Ruminococcus_2 [27,53] were positively associated with Prevotella_1 in our study.

Functional Prediction Of The Ruminal Bacteria In Swcg
Whether the changes in the bacterial community structures would lead to functional differences were detected by PICRUST2. Consistent with Miao et al. [54] and He et al. [55], we found that the abundances of metabolism were the highest in the rumen at KEGG level 1. The differences of diet affected the KEGG pathways of bacteria [56], hence, the integrated results demonstrated an overall increase in genetic information processing and metabolism, and an decrease in human diseases when goats fed with high energy and protein diets. Liu et al. [57] also reported that the KEGG pathways involved in carbohydrate metabolism were highly enriched in the microbiota of individuals fed with high energy diets, which was consistent with our study. Furthermore, the increasing number of Prevotella_1 in Group H of this study and the involvement of the genus in energy metabolism, nucleic acid metabolism and glycan biosynthesis and metabolism [52] might explain the increase of the above pathways in Group H. Although PICRUSt2 approach was utilized to predict the rumen bacterial functions, this method did not accurately detect the related function due to the limited number of sequencing studies in ruminants [58].

Comparison Of The Composition And Differences Of Ruminal Metabolites
Microbiota interacts with numerous physiological functions in the host through its metabolic products [59]. Thus, we used the GC-TOFMS analysis to explore the metabolic functions of ruminal microbiota.
The main metabolites were amino acids, peptides and analogues in this study, which was consistent with the previous reports [43,60]. Amino acids in the rumen are the key precursors for protein and polypeptides synthesis and are mainly obtained from the dietary proteins and microproteins [61].
According to the OPLS-DA results, a clear difference of ruminal metabolites was demonstrated between Groups C and H. These results con rmed that ruminal metabolites were closely related to the composition of diets [62].
Previous studies reported that uracil concentration in the rumen was increased with high concentrate diets [63,64], which was consistent with our study. In addition, correlation analysis revealed that uracil was positively correlated with Ruminococcus_2. The increased concentration of uracil in the rumen of SWCG in Group H might re ect that some bacterial nucleic acids were rapidly degraded to uracil by Ruminococcus_2 [65]. 5-oxoproline (pyroglutamic acid) could be the intermediate product in the glutathione cycle and its concertation was negatively correlated with the concentration of antioxidantglutathione [63,66,67]. Based on the correlation analysis, Lachnospiraceae_ND3007_group might decrease the concentration of 5-oxoproline and more glutathione was produced in Group H to promote antioxidative capacity. Similarly, previous study also found that the concentration of 4hydroxyphenylacetic acid was decreased when cows were fed with high-quality forage [68]. Microbiota degrade dietary protein to tryptophan, which could be later converted into melatonin [69]. Melatonin (Nacetyl-5-methoxytryptamine) as an effective antioxidant [70] could be ultimately oxidized to 5methoxyindole-3-acetic acid [71]. Hence, the level of 5-methoxyindole-3-acetic acid in Group H with higher protein diets was signi cantly higher than that in Group C. In this study, catechol as an antioxidant [72,73] was positively related to the relative abundances of Ruminococcus_2 and Prevotella_1, which might imply that high energy and protein levels in Group H could enhance the catechol concentration by these two genera. Xue et al. [74] reported that the content of spermidine increased in the group of high concentrate diets compared with moderate concentration of diets, which is in line with our study. Furthermore, correlation analysis in this study revealed that spermidine had high positive relationships with Selenomonas_1, Ruminococcaceae_NK4A214_group, Lachnospiraceae_NK3A20_group, Prevotellaceae_UCG-004 and Prevotellaceae_NK3B31_group. Spermidine is an organic compound widely used as an antioxidant [75]. The upregulation of spermidine observed in Group H of this study might enhance the antioxidative capacity in the rumen of this group by the above genera. These data implied that the diets with high energy and protein levels could improve the ruminal antioxidative capacity.
Based on the metabolomic analysis, we found that signi cantly different metabolites were involved in lipid metabolism and nucleotide metabolism, and this result was also identi ed by PICRUSt2 analysis.
Tyrosine metabolism played an important role among those 9 main metabolic pathways in this study, which was also detected by Ferguson et al. [76]. Furthermore, the enriched abundances of beta-alanine, arginine and proline metabolism in this study were related to their functions. Beta-alanine could be metabolized into acetic acid and its concentration is positively associated with the amount of starch and readily available carbohydrate [43]. Arginine and proline involved in RNA synthesis and protein glycosylation are necessary for cellular function [75]. Additionally, we observed that the changes in the concentrations of differential metabolites were correlated with pyrimidine metabolism, which was associated with the dietary protein. Dietary nitrogen from protein could be degraded and reused by the microbiota in order to synthesize microbial nucleic acids [43,77].

Consent for publication
Not applicable.

Availability of data and material
All data generated or analyzed are available from the corresponding author on request.

Competing interests
The authors declare no competing nancial interests.    Table S6. Predicted functions at level 1 of the rumen bacterial microbiota. Table S7. Predicted functions at level 2 of the rumen bacterial microbiota. Table S8. The mean of relative quantitative values of differential metabolites (VIP>1.0; p<0.05).
Additional le 4: Table S10. sheet 1. The correlation between bacterial genera and differential metabolites; sheet 2. The p-value among bacterial genera and differential metabolites  Differences in bacterial metabolism function at KEGG level 2 between Groups C and H using PICRUSt2.  Metabolome view map of the deferential metabolites (VIP>1, p<0.05) identi ed in rumen from goats fed with the diets with different energy and protein levels. The large size indicates high pathway enrichment, and dark color indicates high pathway impact values.