Comparative untargeted metabolome analysis of ruminal fluid and feces of Nelore steers (Bos indicus)

We conducted a study to identify the fecal metabolite profile and its proximity to the ruminal metabolism of Nelore steers based on an untargeted metabolomic approach. Twenty-six Nelore were feedlot with same diet during 105 d. Feces and rumen fluid were collected before and at slaughter, respectively. The metabolomics analysis indicated 49 common polar metabolites in the rumen and feces. Acetate, propionate, and butyrate were the most abundant polar metabolites in both bio-samples. The rumen presented significantly higher concentrations of the polar compounds when compared to feces (P < 0.05); even though, fecal metabolites presented an accentuated representability of the ruminal fluid metabolites. All fatty acids present in the ruminal fluid were also observed in the feces, except for C20:2n6 and C20:4n6. The identified metabolites offer information on the main metabolic pathways (higher impact factor and P < 0.05), as synthesis and degradation of ketone bodies; the alanine, aspartate and glutamate metabolisms, the glycine, serine; and threonine metabolism and the pyruvate metabolism. The findings reported herein on the close relationship between the ruminal fluid and feces metabolic profiles may offer new metabolic information, in addition to facilitating the sampling for metabolism investigation in animal production and health routines.

, 2). The uracil metabolite found in ruminal fluid (R_Ur) presented a negative correlation with fecal metabolites, such as organic acid (pyruvate, valerate, and phenylacetate) and amino acids (threonine, and proline). The uracil in feces (F_Ur) was positively correlated to organic acids (butyrate, propionate, 3-Hydroxybutyrate, and 3-Phenylpropionate) and amino acids (glucose, ornithine, glutamate, valine, sarcosine, glycine, and leucine) in rumen. However, some ruminal fluid metabolites were correlated with fecal uracil, and the same fecal metabolites were correlated with uracil present in the ruminal fluid. Among these metabolites, are noteworthy to mention: organic acids (acetoacetate, isovalerate, and lactate) and amino acids (alanine, lysine, methionine, and isoleucine). In order to further understand the usefulness of the detected metabolites by the 1 H NMR method, we performed pathway analysis to associate the metabolites to their corresponding pathways. The functional analysis of ruminal and fecal metabolites indicated that most metabolites were involved in more than one pathway (Fig. 4). In addition, the same metabolic pathways were observed for both ruminal fluid and feces (P < 0.05), except for www.nature.com/scientificreports/ the tyrosine metabolism, histidine metabolism, and beta-alanine metabolism which were uniquely significant in the rumen ( Table 2). The results of the enrichment analysis showed that pantothenate and aminoacyl-tRNA biosynthesis; alanine, aspartate and glutamate metabolism; valine, leucine and isoleucine biosynthesis; butanoate metabolism; glyoxylate and dicarboxylate metabolism; arginine biosynthesis; pantothenate and CoA biosynthesis; citrate cycle (TCA cycle); glycine, serine and threonine metabolism; pyruvate metabolism; synthesis and degradation of ketone bodies; and glutathione metabolism were significantly enriched (P < 0.05) in both bio-samples. Among these, the pathways with highest impact were synthesis and degradation of ketone bodies; alanine, aspartate and glutamate metabolism; glycine, serine and threonine metabolism; and pyruvate metabolism.

Discussion
In the present study, we investigated the metabolome of two areas of the gastrointestinal tract of Nelore steers, rumen and fecal ampulla, and the relationship between these environments using 1 H NMR and GC-FID. To date, the ruminal and fecal metabolome of Nelore cattle had not been described in the literature. The ruminal fluid involves hundreds of microorganisms, which interact with each other and with the host, while also degrading plant material 17 . The metabolic composition of feces can aid in clarifying this complex interplay between ruminants and their rumen ecosystem 21 , since this biological matrix contains information on the host, microbiota and feed components. Our metabolome data revealed a wide diversity of metabolites in ruminal fluid and feces due to the extraordinary activity of microorganisms, as previously described 12,21,23,24 . However, the comparison between rumen and the rectal ampulla also revealed that the Nelore cattle ruminal fluid metabolome is a little more diverse compared to the feces metabolome. Possibly certain metabolites were not detected in feces, due to high rumen solubility and degradability, although most metabolites (n = 49) have been observed in both bio-samples. These results confirm that the rumen metabolism is closely associated with the fecal metabolism.
In the present study, we observed a wide variety of compounds involved in multiple biochemical processes in rumen. The detected organic acids and amino acids indicate a high diversity of compounds in both biosamples. The organic acids are fermentation products in rumen 25 . Furthermore, many microorganisms in rumen   15 . Additionally, the microbiota is capable of amino acid synthesis through the use of acetate, propionate, among others to obtain carbon and nitrogen compounds as nitrogen sources [26][27][28] . In a recent study, Foroutan et al. 24 described the ruminal fluid from Bos taurus cattle using NMR, partially agreeing with the metabolic profile reported herein for Bos indicus. These differences are mainly due to breed, environment, handling and nutrition, as the identification and concentration of many ruminal fluid metabolites are strongly affected by these traits. Some signals in the ruminal fluid and fecal spectra could not be identified (unknown metabolites) in the used database. According to Almeida et al. 17 , uncharacterized molecules may have multiple origins and may be secreted by microorganisms, plants, and the host. Thus, further exploration of the rumen and fecal environment is required to completely understand the metabolome dataset and characterize metabolites not yet documented.
One particularly interesting point is the fact that the ruminal fluid of Nelore steers contains certain metabolites not observed in the fecal metabolome. Some of these metabolites can be quickly utilized during rumen fermentation, being degraded, catalyzed or metabolized, and thus not being detected in feces. It may be assumed, for example, that in the present study some dietary nutrients were degraded to maltose and later catalyzed by maltase and degraded to glucose in the rumen. As maltose has an energetic function, we propose it may have been rapidly used by microorganisms, and, therefore, was not able to accumulate in rumen, preventing its detection in fecal samples. Fumarate was also detected only in ruminal fluid, possibly due to the reduction of this metabolite into succinate in the rumen, which resulted in higher succinate concentrations in the fecal samples. Histidine is well-known to be extensively degraded by rumen microorganisms for use in microbial growth 29 . Ribose can be produced in the rumen from the degradation of nucleic acids from food sources or from dead bacterial cells, although it does not accumulate in the rumen, being rapidly metabolized by certain microorganisms.
Some metabolites (maltose, histidine, ribose, caffeine, nicotinate, imidazole, 4-Hydroxybutyrate and 4-Hydroxy-3-methoxymandelate metabolites) were observed for the first time in our study. Foroutan et al. 24 evaluated the chemical composition of the ruminal fluid of Bos taurus through different technological platforms, and did not observe maltose, caffeine, imidazole, 4-Hydroxybutyrate and 4-Hydroxy-3-methoxymandelate metabolites. Similarly, Saleem et al. 12 did not detect the 4-Hydroxy-3-methoxymandelate metabolite. O'Callaghan et al. 30 reported the same for the rumen fluid of cows submitted to different pasture feeding systems. Recently, a study assessing the fecal metabolome in Jiulong Yak bovines using 1 H NMR also reported the absence of these metabolites (4-Hydroxybutyrate, histidine, maltose, ribose, caffeine, nicotinate, 4-Hydroxy-3-methoxymandelate and imidazole), except for fumarate 21 . Thus, the differences between ruminal fluid metabolic profiles may be attributed to different breed, nutritional interventions and compound availability, which can intervene as elementary factors in the biochemical engineering of rumen and reflect directly on fecal metabolites.
The present study also demonstrated the nicotinate was detected only in rumen. While nicotinate is widely distributed in feed, this compound can have limited availability in rumen 31 , yet several bacterial species have the ability to synthesize this compound 30 . Nicotinate is intimately involved in the energy metabolism and may play a significant role in urea synthesis, which is a non-protein N source for rumen microorganisms 31 .
A chemometrics analysis within a biological context indicates that the compounds observed in the feces spectra were significantly altered when compared to rumen fluid. This clustering is mainly due to metabolite The letters indicate the pathways (see Table 2).  32 , and fecal metabolites correlations may partially reflect the metabolic processes that occur in rumen. As expected, the most abundant molecules were acetate, propionate and butyrate, representing the main plant material catabolism products obtained by the rumen microbiome. In this regard, these metabolites were also represented in feces by a portion of rumen fluid compound concentrations. The result concerning these metabolites corroborate the findings reported by Eom et al. 23 for dairy cattle ruminal fluid. Foroutan et al. 24 also observed these metabolites as the most abundant in Bos taurus ruminal fluid, although, butyrate was present in higher concentrations compared to propionate. According to Saleem et al. 12 , starch-rich diets increase the availability of free glucose in rumen which, in turn, promotes the growth of most bacteria, leading to greater production of volatile fatty acids like acetate, propionate and butyrate. Furthermore, Zhang et al. 33 reported that acetate and propionate levels were affected in cows with high-yield compared to low-yield milk production. Furthermore, Shabat et al. 16 reported that acetate and propionate concentrations were significantly higher in efficient animals compared to inefficient individuals. Ellis et al. 34 and van Gastelen et al. 35 positively associated acetate with phenotypic and environmental characteristics, such as methane emission (CH 4 ). In this context, our study also indicated that these metabolites are of paramount importance for fecal Nelore steer metabolomics.
Recent studies have demonstrated a direct relationship between Nelore cattle rumen and fecal microbiomes 20,36 . In addition, the association between bacterial and fungal rumen, small intestine, cecum and feces microbiota and the feed efficiency phenotypes has also been reported for this breed 37 . Just as the fecal microbiome was studied and compared to the ruminal microbiome, the integration of metabolome data can also contribute to animal production. Interestingly, the representation of rumen metabolites in feces may assist and simplify future studies, in order to facilitate non-invasive field data collection. In other words, the fecal metabolome partially reflects the metabolic processes that occur in rumen and allows researchers to obtain information and, potentially, a phenotype description based on less invasive sampling procedures. Table 2. Results from ruminal fluid and feces pathway analysis of Nellore steers. a Letter corresponds to the information shown in Fig. 3. b Total cmpd corresponds to total number of compounds in the pathway. c Hits correspond to the actually matched number from the user uploaded data. d P correspond to the original P-value calculated from the enrichment analysis. e − log(p) correspond to the P-value logarithm. f Impact correspond to the pathway impact value calculated from pathway topology analysis. www.nature.com/scientificreports/ Another interesting observation arising from this study is the high glutamate concentrations in ruminal fluid and feces. Glutamate biosynthesis occurs from alpha-ketoglutarate, which may potentially use branched-chain amino acids, such as valine, leucine and isoleucine 38 . Thus, in the present study, glutamate synthesis in rumen may have contributed to the decreased availability of certain amino acids.
Furthermore, we observed that valine, pyruvate, glucose, lactate and ethanol concentrations were increased in the fecal samples. In rumen, the valine biogenesis uses pyruvate and isobutyrate as the main substrates 39 . Indeed, pyruvate is the product of glucose degradation by the glycolytic pathway 40,41 . In addition, pyruvate can be transformed into lactate by lactate dehydrogenase or converted into acetyl coenzyme-A and formate by pyruvate formate-lyase 42 , and, finally, acetyl-CoA can be reduced to ethanol. High concentrations of ethanol in the gastrointestinal tract have been reported as leading to significant consequences for the host 43 , although the ethanol concentrations observed herein were lower than reported as harmful to ruminants.
Succinate can be produced by many microorganisms in rumen, being this metabolite rapidly utilized by Succiniclasticum or metabolized by propionate-producing bacteria, which prevents its accumulation 44,45 . Accordingly, Xue et al. 41 reported that decreased succinate concentrations can be explained by increased Succiniclasticum. Thus, in the present study increased succinate concentrations in feces may be due to the high production and reduced use of this compound in the rumen.
Xanthine, hypoxanthine, ornithine and uracil are intermediate degradation products of nucleic acids by rumen bacteria 46 . These compounds are rapidly degraded in rumen 47 , which may explain the lower concentrations of these metabolites in the feces. Alanine is also a product of the death of both gram-negative and grampositive bacteria 48 , although the concentration of this compound was not significantly different between the assessed bio-samples.
The apolar metabolite results demonstrated a diversity of fatty acids in ruminal fluid and feces due to microorganism activity. Unsaturated fatty acids can have toxic effects for ruminal bacteria and decrease fiber digestibility 49,50 . Thus, despite the fact that fatty acids are not used as an energy source by microorganisms, a portion of the ruminal microbiota exhibits mechanisms to hydrolyze and biohydrogenate dietary lipids 51,52 . According to Saleem et al. 12 and Bryszak et al. 53 , the dietary fatty acid composition also influences the biohydrogenation pathways performed by microbial population in the rumen. Therefore, the fatty acid flow leaving the rumen is higher than the dietary fatty acid intake 54 , and this process plays a vital role in energy production and storage, while also influencing lipid concentrations of final products, such as milk or meat 55 .
Stearic (C18:0) and palmitic acid (C16:0) were the most abundant apolar metabolites in the bio-samples assessed in the present study, and were reported as present at high concentrations in ruminal fluid by Saleem et al. 12 , Szczechowiak et al. 56 and Bryszak et al. 53 . Palmitic acid is the end product of the biohydrogenation of palmitoleic acid (C16:1) although C16:0 also can be metabolized by rumen microbes, and thereby converted to other fatty acids 57 . Similarly, stearic acid is produced from the biohydrogenation of 18-carbon unsaturated fatty acids, like C18:2 and C18:3 5,58 , which are the main polyunsaturated fatty acids in ruminant diets, possibly supporting the results reported herein. In addition, stearic acid represents the majority of intestinal fatty acids in the rumen when an efficient biohydrogenation process takes place.
Our results also indicated that fatty acids concentrations in feces were higher than in the rumen. Individual fatty acids digestibility and absorption are not related to their intake, so it is possible that the amount of fatty acids reaching the duodenum may affect fatty acid solubility and incorporation into micelles, resulting in decreased digestibility and absorption. This may also be due to the microbial synthesis of fatty acids in the large intestine, which are probably not absorbed 54 .
The importance of fatty acids in the rumen and feces is positively related to feed efficiency phenotypes, as reported by Artegoitia et al. 6 , where a higher concentration of pentadecanoic acid (C15:0) was reported in steers with lower average daily gain. The reduction of methane emissions has also been associated to lauric (C12:0), myristic (C14:0) and linoleic (C18:2n6) acids in the review performed by Toprak et al. 59 . Furthermore, rumen fluid lipids have been almost exclusively applied to dairy cattle, with little research conducted on beef cattle. For this reason, easier and more practical strategies in obtaining the apolar metabolite information in feces can aid in understanding rumen lipid metabolism and improve animal production, as the fatty acids fecal content of Nelore exhibits a strong potential to reflect rumen digestion conditions.
The correlation network analysis was applied to compare the ruminal and fecal metabolic profiles shows that ruminal fluid exhibited correlations with 43 fecal metabolites, while feces exhibited correlations with 60 ruminal metabolites. These findings suggest that correlations are mainly centered in the uracil metabolite from ruminal fluid and feces samples from Nelore steers. Uracil is a nucleic acid base found in the RNA and in bacterial degradation products in the rumen. Increases in the concentrations of this metabolite in rumen have been observed when feeding bovines high-grain diets 11,33,43,46 . Increases have also been reported in the rumen of the high-yield dairy cows 32 and in low residual feed-intake steers 60 . Uracil may, therefore, be an indicator for genetic microbial or plant material turnover in the rumen 60 , also increasing the synthesis and availability of crude microbial protein 30,61,62 . Consequently, this metabolite seems to have a significant impact on the biodiverse ruminal ecosystem, directly reflecting the metabolic feces profile.
One of the most interesting observations from this study is that the different metabolites detected in ruminal fluid and feces were specifically correlated to the same metabolic pathways as verified by the pathway impact analysis. In summary, we have characterized the ruminal and fecal metabolome biodiversity of Nelore for the first time. This study offers a new comprehensive insight into the biochemical mechanisms of ruminal fluid and its association with fecal samples, and also demonstrates that most metabolites are common to both environments. In general, these data emphasize the importance of gaining a better understanding of the biochemical functions of the bovine gastrointestinal tract. In addition, although our findings have shown a close relationship between the metabolic ruminal profile and feces in the Nelore breed. Additional studies that promote alterations www.nature.com/scientificreports/ in rumen are required to verify whether these modifications are also reflected in the feces and provide insights into animal production strategies.

Material and methods
Production of experimental animals and sample collection. All experimental procedures were conducted in accordance with animal welfare and humane slaughter guidelines and were approved by the EMBRAPA Livestock Science Ethics Committee on Animal Experimentation, São Carlos, São Paulo (Protocol No. 09/2016). A population of 26 contemporaneous uncastrated Nelore steers (Bos indicus) was placed in an experimental feedlot. Animals averaging 329.5 ± 34.2 kg of initial body weight and aged 20-21 months old were allocated to two collective pens containing 13 animals/pen for 105 d, of which the first 15 were exclusively for animal adaptation, followed by the growth and finishing stages. The animals received a diet consisting of corn silage (72.8%), soybean meal (3.06%), corn grains (21.4%), protected fat (1.19%), urea (0.59%) and Confinatto N235 Agroceres Multimix® (0.91%) twice a day.
After the finishing phase, fecal samples of the rectal ampulla were collected from each animal, kept on ice for approximately 2 h and stored at -80 °C for the metabolomic assays. The animals were then sent for slaughter at a final weight of approximately 477.3 ± 41.5 kg at 23-24 months of age, in accordance to the Humane Slaughter of Cattle guidelines. During the slaughter, the ruminal fluid of each animal was also collected, immediately immersed in liquid nitrogen and then stored at − 80 °C for the metabolomic assays.

Polar metabolite identification and quantification.
The resulting 1 H NMR spectra were processed and analyzed using the Chenomx NMR Suite Professional software package version 6.0 (Chenomx Inc., Edmonton, AB, Canada). The identified metabolites were also defined based on public databases, e.g., the Bovine Metabolome Database (BMDB, http:// www. cowme tdb. ca), Human Metabolome Database (HMDB, http:// www. hmdb. ca) and the Biological Magnetic Resonance Data Bank (BMRB, www. bmrb. wisc. edu). For quantification, individual metabolite peaks were integrated and quantified relative to the "Electronic REference To access in vivo Concentrations 2" (ERETIC2) signal experiment, performed using a 2 mM Sucrose standard, while the fixed receiver gain (RG) was used for the samples.
Sample preparation of apolar metabolites. The total fat content of ruminal fluid and feces were determined according to Folch method 63 with few modifications. In brief, 500 µL of ruminal fluid or 500 mg of feces were used for lipid extraction with chloroform and methanol (2:1 v/v). After the extraction procedure, the samples were centrifuged at 13,000 × g, organic phase was collected, the solvent was evaporated under a gentle nitrogen flow and the total weight of lipids was estimated gravimetrically. Extracted lipids were transmethylated to fatty acid methyl ester (FAME) using boron trifluoride (BF 3  . The chromatographic conditions were as follows: injector at 250 °C, injected volume 1 µL with split ratio set to 1:20, and the carrier gas was helium at 1.0 mL min −1 . The column oven was programmed as follows: 150 °C for 5 min, followed by an increment of 10 °C min −1 to 250 °C hold for 10 min. FID detector operated at 260 °C. The quantification (mg g −1 of total lipids) was performed using the internal standard method (C23:0) as described by Visentainer 65 . Biological samples were analyzed in technical triplicate.
Statistical analyses. The polar and apolar metabolites were analyzed by MIXED procedure available in the SAS statistical program (SAS Institute, Cary, NC, USA, 2011). The statistical model included pen as fixed effect and initial and final weight of feedlot as covariates. The metabolites results were deemed significant when P < 0.05. The obtained 1 H NMR data had a binning of 0.04 ppm applied and were transformed into a data matrix using the MNova software. Next, data were analyzed in the MetaboAnalyst 4.0 platform (http:// www. metab oanal yst. ca), using a Principal Component Analysis (PCA) and Partial Least Squares Discriminant Analysis (PLS-DA). Data preprocessing enrolled data no sample normalization, and the Pareto scaling was used (mean-centered and divided by the square root of standard deviation of each variable). Principal component was used for the discrimination of the analyzed ruminal fluid and feces samples. Specifically, for the PLS-DA, LOOCV was applied as the cross-validation method ( Supplementary Fig. 1). The accuracy and variable importance in projection (VIP) were also assessed to determine the performance and importance features of the analysis, respectively. Spearman's correlation analysis of ruminal and fecal metabolites was performed using the Cytoscape software (http:// www. cytos cape. org). The correlation results were plotted when P < 0.05. Based on the Kyoto Encyclopedia of Genes and Genomes (KEGG, http:// www. kegg. jp) network diagrams were construct between metabolites and the impact factor of the topology analyses of metabolic pathways was graphically presented using Metabo-Analyst 4.0 66 .