Characterising the bacterial microbiota across the gastrointestinal tracts of dairy cattle: membership and potential function

The bacterial community composition and function in the gastrointestinal tracts (GITs) of dairy cattle is very important, since it can influence milk production and host health. However, our understanding of bacterial communities in the GITs of dairy cattle is still very limited. This study analysed bacterial communities in ten distinct GIT sites (the digesta and mucosa of the rumen, reticulum, omasum, abomasum, duodenum, jejunum, ileum, cecum, colon and rectum) in six dairy cattle. The study observed 542 genera belonging to 23 phyla distributed throughout the cattle GITs, with the Firmicutes, Bacteroidetes and Proteobacteria predominating. In addition, data revealed significant spatial heterogeneity in composition, diversity and species abundance distributions of GIT microbiota. Furthermore, the study inferred significant differences in the predicted metagenomic profiles among GIT regions. In particular, the relative abundances of the genes involved in carbohydrate metabolism were overrepresented in the digesta samples of forestomaches, and the genes related to amino acid metabolism were mainly enriched in the mucosal samples. In general, this study provides the first deep insights into the composition of GIT microbiota in dairy cattle, and it may serve as a foundation for future studies in this area.

Diversity, richness and composition of the digesta-associated bacterial communities across cattle GITs. Number of OTUs, Shannon diversity index and richness differed significantly among regions in the digesta samples (Table 1), and the Chao 1 value, number of OTUs and Shannon index in the forestomach were significantly higher than in the small and large intestines (Table 1). When bacterial composition of digesta microbiota among GIT regions was compared using the unweighted UniFrac distance, digesta-associated bacterial communities of the forestomach, small intestine and large intestine were separated spatially from each other ( Fig. 2A). An unweighted distance-based analysis of molecular variance (AMOVA) was used to assess the statistical significance of the spatial separation observed among various regions of the principal coordinate analysis (PCoA) plots. Statistically significant dissimilarities were observed across most regions with respect to bacterial diversity, with the exception of similarities between the cecum and the rectum (p = 0.052), the colon and the cecum (p = 0.179), the rectum and the colon (p = 0.485) and the jejunum and the duodenum (p = 0.123) (Table S2). In addition, digesta-associated bacterial communities in the forestomach and large intestine samples clustered close to one another, while those from the small intestine samples did not ( Fig. 2A).
At the phylum level, 21 bacterial phyla were identified in the digesta samples (Table S3). The majority of the sequences obtained belonged to Firmicutes (64.81%), Bacteroidetes (15.06%) and Proteobacteria (13.29%) (Fig. 3A). In addition, only Firmicutes, Bacteroidetes, Proteobacteria, Spirochaetae, Cyanobacteria and Tenericutes were found in all samples. Among the 10 GIT sites, the rumen and abomasum harboured most of the phyla and groups (19 phyla), while the lowest number of phyla was observed in the cecum (12 phyla). When bacterial composition was compared regionally, the phylum Firmicutes dominated all bacterial communities along the GIT except for in the duodenum, where Proteobacteria (45.6%) was predominant. Bacteroidetes was the second most prevalent in the forestomach, while Proteobacteria was the second most prevalent in digesta samples of the jejunum, ileum, cecum and colon (Fig. S2).
At the genus level, 542 taxa were observed throughout cattle GITs; however, 44.6% of all sequences were not identified at the genus level. For clarity and visualisation purposes, Fig. 4A presents the most abundant taxa (those with a relative abundance of ≥ 2% in at least one GIT region). Results showed that predominant genera in cattle GITs included Prevotella, Treponema, Succiniclasticum, Ruminococcus, Acetitomaculum, Mogibacterium, Butyrivibrio and Acinetobacter, as well as those unclassified derived from Peptostreptococcaceae (family), Ruminococcaceae (family), Enterobacteriaceae (family), Prevotellaceae (family), Clostridiales (order), Rikenellaceae (family) and Bacteroidales (order). In the digesta samples, the dominant taxa, Prevotella, unclassified Ruminococcaceae, unclassified Bacteroidales and unclassified Rikenellaceae were enriched significantly in the digesta of forestomach samples ( Fig. 4B and Table S4). The unclassified Enterobacteriaceae were enriched [false discovery rate (FDR) < 0.001] in the small intestine, cecum and colon, and a large proportion of Acetitomaculum, Ruminococcus and unclassified Lachnospiraceae were enriched (FDF < 0.001) in the jejunum. Proportions of Butyrivibrio were significantly higher in the abomasum, duodenum and jejunum than in other GIT regions. The unclassified Peptostreptococcaceae and Turicibacter were enriched (FDR < 0.001) in the ileum and large intestine, while the Clostridium were more enriched in the large intestine (FDR < 0.001) than in the forestomach and small intestine.
At the OTU level, the study identified 5,573 OTUs in the digesta samples (Fig. S3). Of them, one OTU, OTU-3825, classified in the family Enterobacteriaceae was the most dominant among the dominant OTUs (representing ≥ 2% of all sequences in at least one region in all samples), and its abundance was the greatest (FDR < 0.001) in the duodenum digesta samples ( Fig. S4 and Table S5).
Diversity, richness and composition of mucosa-associated bacterial communities across cattle GITs. Number of OTUs, Shannon diversity index and richness differed significantly among regions in mucosa samples (Table 1). In the mucosal tissue, the Chao 1 value and the number of OTUs in the forestomach were significantly higher than in the small intestine. No significant differences were observed in the Shannon index among regions, with the exception of the omasum and the cecum (Table 1). When bacterial composition of mucosal microbiota among various GIT regions were compared using unweighted UniFrac distance, the mucosa-associated communities in the forestomach clustered more closely to each other than did other members of the combined large intestinal communities (Fig. 2B). AMOVA testing of mucosal samples revealed that compositions of bacterial communities differed significantly across GIT regions (except for abomasum vs omasum, p = 0.082; ileum vs jejunum, p = 0.328; cecum vs colon, p = 0.949; cecum vs rectum, p = 0.224; and colon vs rectum, p = 0.585) (Table S2). At the phylum level, 22 bacterial phyla or groups were identified in mucosal samples (Table S6). The majority of the sequences obtained belonged to Firmicutes (42.22%), Bacteroidetes (21%) and Proteobacteria (17.56%), and Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, Tenericutes, Spirochaetae, Cyanobacteria and Lentisphaerae were found in all samples (Fig. 3A). Most of the phyla (18-20/22 phyla) were found among the 10 GITs sites. Firmicutes dominated all mucosa-associated bacterial communities along the GITs except for in the duodenum, where Proteobacteria (37.26%) was predominant. However, relative abundance of the second most predominant phyla varied markedly among mucosal tissue by GIT region (Fig. S2). Bacteroidetes was the second most prevalent phyla in mucosal samples of the reticulum, omasum, abomasum, colon and rectum, whereas Proteobacteria was the second most prevalent in mucosal tissues of the rumen, jejunum and ileum. Firmicutes and Spirochaetae were the second most dominant phyla in mucosal samples of the duodenum and cecum, respectively.
At the genus level, an analysis of the most-abundant taxa by sample revealed that abundances of Butyrivibrio, Campylobacter and Desulfobulbus were enriched significantly in mucosal samples of rumen and reticulum ( Fig. 4C and Table S7). In the abomasum, a large proportion of Mycoplasma, Acetobacter, unclassified Acetobacteraceae and unclassified Bifidobacteriaceae were enriched (FDR < 0.05). In the ileum, Turicibacter dominated (FDR < 0.001) compared with the forestomach, duodenum and jejunum samples, and in the cecum, Anaerovibrio were enriched significantly compared with the forestomach and small intestine samples. Relative abundance of unclassified Peptostreptococcaceae was higher in the ileum and large intestine (FDR < 0.001) than in the forestomach samples, while Acinetobacter was more enriched (FDR < 0.001) in the small intestine than in the forestomach samples. Treponema dominated represented as relative abundances on the Y-axis. Boxes with a star symbol above their whiskers are significantly different between the digesta and its corresponding mucosa in each GIT site at p < 0.05 using a t-test analysis. D: digesta samples; M: mucosal samples.
(FDR < 0.001) in the mucosa of the cecum and colon, and Prevotella was enriched significantly in the omasum and abomasum. Unclassified bacteria derived from the families Ruminococcaceae and Rikenellaceae were more enriched in the forestomach and large intestine than in the small intestine. In addition, unclassified Bacteroidales were more enriched in the forestomach and large intestine than in the small intestine samples, and unclassified Prevotellaceae dominated (FDR < 0.001) in the forestomach compared with the large and small intestine samples, except for those from the rectum.
At the OTU level, 6,327 OTUs were identified in mucosal samples (Fig. S3). Of them, OTU-332 classified to Acinetobacter was the most dominant among the dominant OTUs in the samples, and its abundance was the greatest (FDR < 0.001) in mucosal tissues of the duodenum and jejunum (Table S8 and Fig. S4). Comparison of diversity, richness and composition of bacterial microbiota between corresponding digesta and mucosa samples. When alpha diversity indices were compared between mucosal tissue and digesta of each GIT region, OTU values in the ruminal digesta were significantly higher than in the rumen epithelium (Fig. S5A). In contrast, the numbers of OTUs in mucosal tissues of the omasum, cecum, colon and rectum were higher (p < 0.05) than in their respective digesta. Mucosal tissues of the reticulum, omasum and large intestine exhibited higher (p < 0.05) Chao1 index values than in their respective digesta (Fig. S5B). The Shannon index was significantly higher for digesta of the rumen and reticulum than for their corresponding mucosal tissues, while mucosal tissues of the duodenum, ileum and large intestine had a higher Shannon index than their corresponding digesta (pv < v0.05) (Fig. S5C). In addition, a PCoA profile (Fig. 2C) and AMOVA analysis (Table S9) revealed significant differences in composition and structures of bacterial communities between the mucosal tissue and digesta of each GIT region. At the OTU level, the study found 5,040 OTUs across all samples, 533 exclusively in the digesta and 1,287 in the mucosa (Fig. S3).
At the phylum level, the proportion of Firmicutes in the digesta of the forestomach, jejunum and large intestine was significantly higher than in their corresponding mucosal tissues (Fig. 3B). The abundance of Bacteroidetes was higher (FDR < 0.001) in the digesta of the rumen and reticulum, while it was lower (FDR < 0.001) in the digesta of the duodenum, jejunum and large intestine when compared with their corresponding mucosal tissues. Mucosal tissues of the forestomach and rectum presented higher (FDR < 0.001) proportions of Proteobacteria than their corresponding digesta samples, while digesta of the duodenum showed a comparatively higher proportion of Proteobacteria.
At the genus level, in the rumen, the proportions of Prevotella, unclassified Ruminococcaceae, unclassified Rikenellaceae, unclassified Christensenellaceae and unclassified Bacteroidales were significantly higher in digesta samples than in mucosal samples, while rumen mucosa presented higher (FDR < 0.001) percentages of Butyrivibrio, unclassified bacteria, Desulfobulbus and Campylobacter (Fig.  S6A). In the small intestine, luminal samples presented a significantly higher abundance of unclassified Enterobacteriaceae and a lower abundance of Acinebacter compared with corresponding mucosal samples (Figs S6B, S6C and S6D). In the large intestine, unclassified Peptostreptococcaceae, Turicibacter and Clostridium were dominant in the lumen (Figs S6E, S6F and S6G), while Treponema and unclassified Ruminococcaceae were more abundant in the mucosa.

Total bacterial populations in the GIT of dairy cattle. Total bacterial populations in various GIT
regions were estimated with a real-time PCR analysis by measuring the total copy number of bacterial 16S rRNA genes. Regional sites affected bacterial density of dairy cattle significantly ( Table 2). Higher digesta-associated bacterial numbers (p < 0.05) were observed in the colon and omasum, and higher mucosa-associated bacterial numbers were present in the rumen and omasum ( Table 2). In addition, mucosa-associated bacterial densities in the jejunum, ileum, cecum, colon and rectum were significantly lower than in their corresponding digesta (Table 2). However, digesta-associated bacterial numbers in the duodenum were lower than mucosa-associated bacterial densities (p = 0.002).

Predicted molecular functions of bacterial microbiota.
To gain insight into the molecular functions of bacterial microbiota across cattle GITs, we used PICRUSt to predict the metagenomic contribution of the communities observed. PICRUSt predicts metagenomic potential by imputing the available annotated genes within a known sequenced database, such as the Kyoto Encyclopaedia of Genes and Genomes (KEGG) and the Clusters of Orthologs Groups (COGs) catalogue, based on the presence/ absence of OTUs in a 16S rRNA survey. With PICRUSt, one can calculate nearest sequenced taxon index (NSTI), which measures how closely related the average 16S rRNA sequence in an environmental sample is to an available sequenced genome. When this number is low, PICRUSt is likely to perform well in predicting the genomes of the organisms in an environmental sample. This study used the KEGG database to match the chosen reference OTUs, and the average NSTI for the 120 samples was 0.075. This low NSTI metric suggests that PICRUSt may perform well when predicting the molecular function of microbial communities in the GITs of dairy cattle. Using PICRUSt as a predictive exploratory tool, the present study inferred that 39 gene families were identified in the digesta and mucosa samples (Fig. 5A). A principal component analysis (PCA) on the relative abundance values of the KEGG pathways represented from the digesta and mucosal microbiota showed a clear distinction between the clustering of the intestine digesta and that of the forestomach samples (Fig. 5B). In the mucosal tissues, PCA observed a separate clustering of the forestomach and intestine samples, except for those of the rectum (Fig. 5C).
Of the 39 gene families, the majority of the genes belonged to membrane transport (17.82% in digesta-associated microbiota and 17.61% in mucosa-associated microbiota, respectively), carbohydrate metabolism (10.68% in digesta, 10.79% in mucosa), amino acid metabolism (8.36% in digesta, 9.13% in mucosa), replication and repair (7.70% in digesta, 7.62% in mucosa) and energy metabolism (4.69% in digesta, 4.84% in mucosa). Of the 39 gene families, 33 gene families in the mucosa-associated microbiota had significantly different abundances among GIT regions (Table S10), and the prevalence of 37 gene families in the digesta-associated microbiota was significantly different among GIT regions (Table S11). Of the five predominant gene families mentioned earlier, in the digesta samples, the relative abundances of the genes involved in carbohydrate metabolism and replication and repair in the microbiota of forestomach samples were significantly higher than in the microbiota of the cecum and colon (Fig. 5D and Table S12). The abomasum had the highest (FDR < 0.001) abundance of genes involved in amino acid metabolism, while the proportion of gene families involved in amino acid metabolism in the rectum was the lowest (FDR < 0.001) throughout the GIT (Table S12). In mucosal samples, there was a notable enrichment (FDR < 0.001) of genes related to amino acid metabolism in the duodenum when compared with those in the forestomach and large intestine. The microbiomes in the abomasum and the omasum were significantly enriched in categories associated with carbohydrate metabolism (Table S12).
Of the five predominant gene families, when compared with corresponding digesta samples, the abundance of genes related to membrane transport was significantly higher in the mucosa of the reticulum, while it was lower (FDR < 0.05) in the mucosa of the duodenum and rectum (Fig. S7A). The proportion of genes involved in carbohydrate metabolism was higher (FDR < 0.001) in the digesta of rumen and reticulum compared with their corresponding mucosa, while it was significantly lower in the digesta of duodenum and rectum compared with their corresponding mucosal samples (Fig. S7B). Genes related to replication and repair were significantly enriched in abundance in the ruminal and reticulum digesta-associated microbiota compared with the mucosa samples (Fig. S7C). Genes involved in amino acid metabolism were higher in percentage in mucosa-associated microbiota (except for in the abomasum, cecum and colon) (FDR < 0.001) than they were in their corresponding digesta samples (Fig.  S7D). In addition, genes related to energy metabolism were significantly higher in mucosa of the rumen, reticulum, cecum, colon and rectum compared with their corresponding digesta samples (Fig. S7E).

Discussion
This study characterised the compositions and phylogenetic distributions of digesta-and mucosa-associated microbiota in the GITs of Holstein dairy cattle. Results showed significant differences in bacterial richness and diversity, as indicated by the Chao 1 values and Shannon index, among GIT regions ( Table 1), indicating that GIT region exhibit strong determinants of the microbial community composition in dairy cattle. These results may be explained by the difference in the luminal pH values of different regions, as well as by the gut motility, redox potential, nutrient supplies and host secretions 10 . In addition, results from PCoA profiling and AMOVA analysis revealed significant differences among sections of the GIT, implying that GIT region is strongly determinant of microbial community structure.
Overall, findings of the present study revealed that the taxonomic groups represented within cattle GIT were Firmicutes, Bacteroidetes and Proteobacteria: These varied considerably among regions in abundance and in the number of genera composing them (Fig. 4A). In general, the digesta-associated microbiota of the forestomach exhibit greater relative abundances of Firmicutes and Bacteroidetes, whereas the small intestine and large intestine, except for the rectum, show higher relative abundances of Firmicutes and Proteobacteria (Fig. S2). Interestingly, the phylum Firmicutes, which was prominent in digesta-associated microbiota of the large intestine, was composed mainly of the genera Clostridium, Turicibacter and unclassified Peptostreptococcaceae (Table S4), which together reached up to 70% of the total reads in some samples. However, the dominant taxa belonging to Firmicutes in the digesta-associated forestomach microbiota were unclassified Ruminococcaceae, unclassified Rikenellaceae, unclassified Christensenellaceae and unclassified Lachnospiraceae (Table S4), which have been detected widely in the rumen and implicated as playing an important role in degradation of starch and fibre 11 . The fact that members of Ruminococcaceae, Christensenellaceae and Lachnospiraceae dominated in the forestomach while Turicibacter, Clostridium and members of Peptostreptococcaceae were enriched in the large intestine revealed significant differences in composition of the predominant microbial communities between the forestomach and large intestine.
The phylum Bacteroidetes was significantly less abundant in digesta-associated microbiota of the small and large intestine, whereas in the rumen, it was prominent and mainly composed of the genus Prevotella (Table S4), which reached up to 19.08% of the total reads in some digesta samples. This finding was consistent with a previous study by Stevenson and Weimer 12 , who reported that Prevotella were the most abundant (making up 17-50% of total bacterial reads) of all genera identified in rumen samples from lactating cows. High abundance of this genus in the rumen is thought to relate to the high genetic variability of this genus, which enables its members to occupy various ecological niches within the rumen 13 . However, the exact mechanism explaining the result that the genus Prevotella was less abundant in digesta-associated microbiota of small and large intestines is not clear yet, and further studies are required. The phylum Proteobacteria was significantly less abundant in digesta-associated microbiota of the forestomach, whereas its abundance was significantly increased in digesta samples of the small and large intestines, except for the rectum (Fig. 3A). The high abundance of Proteobacteria in digesta of the small intestine was attributed mainly to OTUs representing unclassified Enterobacteriaceae (Table S5). Unclassified Enterobacteriaceae refers to bacterial genera encompassed by the Enterobacteriaceae family, about the function of which little is known yet 14 . As these unclassified gut bacteria have not yet been evaluated functionally, the cause of the increase in abundance of unclassified Enterobacteriaceae in the digesta of the small intestines is not entirely clear, and future studies are needed to clarify these issues.
Gastrointestinal mucosa-associated microbiota could play important biological roles due to their close proximity to the animal host, but knowledge of their composition in cattle still is limited to calves 6,15 . The mucosal microbiota of calves has previously been investigated using culture-based methods and culture-independent DNA techniques 6,16 , and these studies revealed the presence of mucosa-associated bacterial communities with bacterial species that differed from those associated with the digesta of calves. In the present study, our data revealed significant differences in the diversity index of mucosa-associated bacteria communities among most GIT regions (Table 1). In addition, PCoA profiling and AMOVA analysis revealed significant differences in composition and structure of mucosa-associated bacteria communities among GIT sections. Thus, the present study has provided further evidence of mucosa-associated bacterial communities along the GITs of dairy cattle, which appears to support previously reported trends regarding gastrointestinal region-specific microbiota.
Among the regions of the GIT, in contrast to the rumen epithelium, the mucosa-associated bacterial microbiota of the small and large intestines has been studied infrequently. The present study highlighted the dominant presence of the aerobic bacterial genus Acinetobacter in small intestine mucosa. This finding is in accordance with previous works showing the presence of oxygen at the apical surface of the intestinal epithelial cells 17 , representing a possible mechanism of exclusion of strictly anaerobic, extremely oxygen-sensitive microorganisms 18 . In addition, the present study found a higher abundance of Butyrivibrio (belonging to Firmicutes) in the rumen and reticulum mucosal samples across the GITs of dairy cattle. As the mucosal butyrate producers release butyrate close to the epithelium, species of Butyrivibrio may enhance butyrate bioavailability for the host, which may be particularly useful in proliferating rumen and reticulum epithelium 19,20 . In addition, the present study's finding of the enrichment of the genus Treponema (belonging to Spirochaetae) in large intestine mucosa may have health implications. Previous studies revealed that Treponema spp., well adapted to oxidative stress 21 , are involved in a number of diseases of the skin or mucus membranes in several mammals 22 . Additionally, Treponema spp. are believed to be associated with ulcerative mammary dermatitis and bovine digital dermatitis in cattle and contagious ovine digital dermatitis in sheep [23][24][25] . Thus, enhanced Treponema in large intestine mucosa could have deleterious effects on hindgut health. Altogether, findings of the present study revealed that abundances of dominant genera in mucosa-associated microbiota varied considerably among GIT regions, and uniform distribution of the attaching bacterial compositions along the rumen, small intestine and hindgut is likely due to host-bacterium interactions in the mucosa.
Consistent with the findings in bovine calves using denaturing gradient gel electrophoresis, clone library and pyrosequencing 6,15 , the present study revealed a higher proportions of the predominant taxa Prevotella, unclassified Ruminococcaceae, unclassified Enterobacteriaceae and unclassified Peptostreptococcaceae in the digesta-associated microbiota (Fig. S6) and a larger percentage of Butyrivibrio, Acinetobacter and Treponema in the mucosa. Previous studies revealed that the genus Prevotella, unclassified Ruminococcaceae and unclassified Peptostreptococcaceae might play important roles in feed digestion 26,27 and that members of the genera Butyrivibrio, Acinetobacter and Treponema were more involved in epithelium proliferation and diseases 6,28 . These findings imply that sampling sites show strong determinants of microbial community structure and function of bacterial communities among regions.
The cattle gastrointestinal microbiome presents many physiological functions that are lacking in the host, and therefore, they can be considered essential to cattle life. To determine the potential functions of the gastrointestinal microbiota in the samples, the present study used PICRUSt to infer putative metagenomes from 16S rRNA gene profiles 8 . The study inferred that the most abundant functional categories were those corresponding to the functions of carbohydrate metabolism, amino acid metabolism, membrane transport and replication and repair. This is consistent with the general metabolic functions (such as carbohydrate, protein and amino acid metabolism) being essential for microbial survival 29,30 , and it is in line with the observations of other metagenomic studies in mice and humans [31][32][33] . Findings of the present study revealed significant differences in bacterial function among regions across the GIT (Fig. 5B,C). For example, genes relating to carbohydrate metabolism were more abundant in the digesta-associated microbiota of the forestomach than in that of the hindgut. The increased occurrence of these genes mirrors the increase in sequences affiliated with carbohydrate degradation, as discussed above, supporting the importance of these bacteria within the forestomach. In addition to these findings, the present study inferred that genes associated with amino acid metabolism were more enriched in the mucosa-associated microbiota of most GIT regions than in their corresponding digesta (Fig. S7D). One possibility is that the mucosal tissue provides a decreased supply of carbohydrates and that bacteria may derive energy from amino acid fermentation 15 . In addition, these results imply that the mucosal microbiota may be more necessary to amino acid degradation. In this study, next-generation sequencing on the Miseq platform and bioinformatics analyses were performed to investigate the GIT bacterial microbiota of dairy cattle. Results revealed that microbial communities of dissimilar composition and metabolic function occupied different regions within the GIT ecosystems of dairy cattle, while there were some shared traits across all microbiota. The forestomach, small intestine and large intestine were characterised by a specific microbial community, likely shaped by different physicochemical conditions, such as pH value. Specifically, the taxonomic shift between digesta-and mucosa-associated bacteria and from the rumen to the rectum within cattle GITs supports the assumption that digesta-associated microbiota might play an important role in feed digestion, while the mucosa-adherent microbiome may be involved in epithelium proliferation and diseases, indicating that the structure and function of mucosal and ruminal bacterial communities are distinct. Similarly, the present study inferred that genes related to amino acid metabolism were overrepresented in mucosal samples. These findings indicate that the microbial community associated with mucosa may be more necessary to amino acid degradation. In general, this research has revealed partially the high level of heterogeneity in species composition and functional capacities of microbial assemblages across the bacterial ecology system of the GIT, and these findings can be used to potentially modulate gastrointestinal microbiota and improve health, nutrient use and milk production in dairy cattle.

Materials and Methods
Ethical approval. The  Animals and sample collection. The study used six Holstein dairy cattle aged five years (body weight: 607 ± 55.6 kg; milk yield = 29.72 ± 4.7 kg/day, 140 to 189 days in lactation). The cows' diets were formulated to meet or exceed the energy requirements (at 24 kg/day dry matter intake) of Holstein cattle yielding 35 kg of milk/day with 3.50% milk fat and 3.10% true protein (Table S13). Diets were fed ad libitum as a total mixed ration to avoid the selection of dietary components. The cattle were fed at 07:00 and 18:00 h, one-half of the allowed daily ration at each feeding. The experimental period was 84 d; the first 83 d were used for diet adaptation and the last day was used for measurements. Throughout the experimental period, the cattle were housed in tie stalls and fed ad libitum to assure 5% orts, and they were given free access to fresh water during the trial.
On day 84, the cattle were slaughtered, and mucosal tissue and digesta samples were collected from the rumen, reticulum, omasum, abomasum, duodenum, jejunum, ileum, cecum, colon and rectum. Mucosal tissue samples (60 total) were rinsed 3 times with sterile, phosphate-buffered saline (PBS) (pH 7.0) to remove the digesta, cut into 4-5 mm 2 and immediately frozen in liquid nitrogen. The digesta of the rumen, reticulum, omasum, abomasum, duodenum, jejunum, ileum, cecum, colon and rectum were homogenised, separately. The pH values of the contents of each GIT segment were determined immediately using an Accumet gel-filled polymer-body combination pH electrode (Fisher Scientific; Fairlawn, NJ, USA). Then, the homogenised digesta from each GIT segment were sampled (60 total) and immediately frozen in liquid nitrogen. The remaining samples were centrifuged immediately at 2,000 × g, and the supernatants were stored at −20 °C until they were analysed for VFA. The amount of VFA was measured using capillary column gas chromatography (GC-14B, Shimadzu; capillary column: 30 m × 0.32 mm × 0.25 mm film thickness; column temperature = 110 °C, injector temperature = 180 °C, detector temperature = 180 °C) 27 .
DNA extraction, PCR amplification, illumina MiSeq sequencing and sequencing data processing. For DNA extraction, three gram (wet weight) of homogenised samples (120 total) of digesta and mucosal tissue from each GIT segment of each cow was used. DNA was extracted by a bead-beating method using a mini-bead beater (Biospec Products; Bartlesville, USA), followed by phenol-chloroform extraction 27 . The solution was precipitated with ethanol, and the pellets were suspended in 50 μ L of Tris-EDTA buffer. DNA was quantified using a Nanodrop spectrophotometer (Nyxor Biotech; Paris, France) following staining using a Quant-it Pico Green dsDNA kit (Invitrogen Ltd.; Paisley, UK). DNA samples were stored at −80 °C until further processing.
The V3-V4 regions of the bacteria 16S rRNA gene were amplified by PCR (95 °C for 2 min, followed by 25 cycles at 95 °C for 30 s, 55 °C for 30 s, 72 °C for 30 s and a final extension at 72 °C for 5 min) using primers 338F (5′-barcode-ACTCCTRCGGGAGGCAGCAG)-3′ and 806R (5′-GGACTACCVGGGTATCTAAT-3′), where barcode is an eight-base sequence unique to each sample. PCR reactions were performed in a triplicate 20 μ L mixture containing 4 μ L of 5 × FastPfu Buffer, 2 μ L of 2.5 mM dNTPs, 0.8 μ L of each primer (5 μ M), 0.4 μ L of FastPfu Polymerase and 10 ng of template DNA. Amplicons were extracted from 2% agarose gels and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences; Union City, CA, USA) according to the manufacturer's instructions and quantified using QuantiFluor ™ -ST (Promega; USA). Purified amplicons were pooled in equimolar and paired-end sequenced (2 × 250) on an Illumina MiSeq platform according to standard protocols 34 .
Raw fastq files were de-multiplexed and quality-filtered using QIIME (version 1.70) 35 , with the following criteria: 1) The 250 bp reads were truncated at any site receiving an average quality score < 20 over a 10 bp sliding window, discarding the truncated reads that were shorter than 50 bp; 2) Exact barcode matching, 2 nucleotide mismatch in primer matching, reads containing ambiguous characters were removed; 3) Only sequences that overlapped longer than 10 bp were assembled according to their overlap sequence. Reads that could not be assembled were discarded. OTUs were clustered with a 97% similarity cut-off using UPARSE (version 7.1 http://drive5.com/uparse/), and chimeric sequences were identified and removed using UCHIME 36 . The most abundant sequences within each OTU were designated as 'representative sequences' and aligned against the core set of Greengenes 13.5 37 using PYNAST 38 with the default parameters set by QIIME. A PH Lane mask supplied by QIIME was used to remove hypervariable regions from the aligned sequences. FASTTREE 39 was used to create a phylogenetic tree of the representative sequences. Sequences were classified using the Ribosomal Database Project (RDP) classifier with a standard minimum support threshold of 80% 40 . Sequences identified as chloroplasts or mitochondria were removed from analysis. Community diversity was estimated using the ACE, Chao1 and Shannon indices. The unweighted UniFrac distance method was used to perform a principal coordinate analysis 41 , and an unweighted distance-based analysis of molecular variance (AMOVA) was conducted to assess significant differences among samples using the programme MOTHUR v.1.29.0 42 .