ABSTRACT
A lack of information on the intestinal microbiome of neonatal calves prevents the use of microbial intervention strategies to improve calf gut health. This study profiled the taxonomic and functional composition of the small intestinal luminal microbiome of neonatal calves using whole-genome sequencing of the metagenome, aiming to understand the dynamics of microbial establishment during early life. Despite highly individualized microbial communities, we identified two distinct taxonomy-based clusters from the collective luminal microbiomes comprising a high level of either Lactobacillus or Bacteroides. Among the clustered microbiomes, Lactobacillus-dominant ileal microbiomes had significantly lower abundances of Bacteroides, Prevotella, Roseburia, Ruminococcus, and Veillonella compared to the Bacteroides-dominated ileal microbiomes. In addition, the upregulated ileal genes of the Lactobacillus-dominant calves were related to leukocyte and lymphocyte chemotaxis, the cytokine/chemokine-mediated signaling pathway, and inflammatory responses, while the upregulated ileal genes of the Bacteroides-dominant calves were related to cell adhesion, response to stimulus, cell communication and regulation of mitogen-activated protein kinase cascades. The functional profiles of the luminal microbiomes also revealed two distinct clusters consisting of functions related to either high protein metabolism or sulfur metabolism. A lower abundance of Bifidobacterium and a higher abundance of sulfur-reducing bacteria (SRB) were observed in the sulfur metabolism-dominant cluster (0.2% ± 0.1%) compared to the protein metabolism-dominant cluster (12.6% ± 5.7%), suggesting an antagonistic relationship between SRB and Bifidobacterium, which both compete for cysteine. These distinct taxonomic and functional clusters may provide a framework to further analyze interactions between the intestinal microbiome and the immune function and health of neonatal calves.
IMPORTANCE Dietary interventions to manipulate neonatal gut microbiota have been proposed to generate long-term impacts on hosts. Currently, our understanding of the early gut microbiome of neonatal calves is limited to 16S rRNA gene amplicon based microbial profiling, which is a barrier to developing dietary interventions to improve calf gut health. The use of a metagenome sequencing-based approach in the present study revealed high individual animal variation in taxonomic and functional abundance of intestinal microbiome and potential impacts of early microbiome on mucosal immune responses during the preweaning period. During this developmental period, age- and diet-related changes in microbial diversity, richness, density, and the abundance of taxa and functions were observed. A correlation-based approach to further explore the individual animal variation revealed potential enterotypes that can be linked to calf gut health, which may pave the way to developing strategies to manipulate the microbiome and improve calf health.
INTRODUCTION
The gut microbiome consists of millions of nonredundant genes that are necessary for microbial survival in the gastrointestinal tract environment, and the majority of these microbial genes (∼99%) are of bacterial origin (1). In addition to the housekeeping processes universal to microbial survival, various biosynthetic processes are also carried out by the microbial genome (2). Recently, the biosynthetic processes of gut microbiota have been used to explain the mechanisms underlying host-microbial interactions in humans (3). For example, the Firmicutes/Bacteroidetes ratio is high in obese individuals, and a high-energy harvesting capacity of Firmicutes relative to Bacteroidetes has been implicated as a causal factor contributing to the obese phenotype (3). Thus, a thorough understanding of microbial functions and taxonomic composition is vital when elucidating the mechanisms underlying host-microbial interactions.
Metagenome (genetic content of a microbial community) sequencing has been widely used to explore human microbiomes (4). Whole-genome sequencing of metagenomes provides information on the microbial gene composition and abundance that can be used to predict microbial functions as well as to profile taxonomic composition (5). Metagenomic analysis of the human infant gut microbiome has advanced our knowledge of the establishment of the early life gut microbiome (6). However, knowledge of the pre-ruminant intestinal microbiome is limited. Although studies have profiled the preweaned calf gut microbiota using 16S rRNA gene amplicon-based sequencing (7–9), only a few publications have used the whole-genome sequencing of metagenomes to report the preruminant rumen microbiome (10–12). Neonatal dairy calves are highly susceptible to infections by a variety of enteric pathogens that target the mucosal epithelium of the small intestine (13). The lack of information on the microbial functions in the intestinal tracts of neonatal calves, which are colonized by a diverse group of pioneer species (8), prevents the use of microbial intervention strategies to improve neonatal calf gut health. The small intestines of neonatal dairy calves harbor region-specific microbiota (8), and a recent study in mice demonstrated region-specific regulation of the intestinal epithelium by the gut microbiota (14). Linkages between early gut microbial composition and dairy calf health and growth have been reported in a study characterizing the fecal microbiota of preweaned calves (7). A follow-up study reported that oral administration of Faecalibacterium prausnitzii, a bacterium that is associated with calf health and growth (7), lowered the incidence of severe diarrhea and diarrhea-associated calf death compared to control calves (15). These findings together suggest that the early microbiome plays a crucial role in neonatal calf health and that early microbial intervention strategies can be effectively used to improve calf health. However, knowledge of the mechanisms underlying host responses to the early microbiome of preweaned calves is limited. Therefore, it is necessary to characterize the taxonomic and functional composition of the early intestinal microbiome and to elucidate the mechanisms by which the microbiome may alter mucosal immune and barrier functions to improve calf gut health.
In this study, whole-genome sequencing of the luminal metagenome and 16S rRNA gene amplicon sequencing of the mucosal-tissue-attached bacteria were used to characterize the taxonomic and functional composition of the small intestinal microbiomes of neonatal calves. Immune-related genes expressed in the neonatal calf ileum were then extracted from the whole transcriptomes of a previously published work conducted using the same animals (16) to explore whether the variations identified in the microbiome may also be reflected in the host mucosal immune system. The results from these analyses provide the first characterization of functions associated with the intestinal microbiome and provide a framework for investigating the functional consequences of early gut microbiota.
RESULTS
Postnatal changes in the taxonomic composition of the small intestinal microbiota.(i) Luminal microbiota. After quality filtering to remove host DNA and artificial reads from the raw sequences, we generated a 1.82 Gb data set (40,555,690 ± 4,656,914 reads, expressed as the mean ± the standard error of the mean [SEM]) from the small intestinal luminal microbiomes (Data Set S1), which were used to assign taxonomy and functions using the M5NR database and the SEED subsystems within the MG-RAST platform, respectively. Small intestinal luminal microbiomes of neonatal calves consisted of bacteria, archaea, fungi, protozoa, and viruses from the first week of life (Fig. S1; Data Set S2, sheet 1). Further analysis of bacteria, the dominant microbial group, revealed 21 phyla, 46 families, and 510 genera in total from all luminal microbiomes. Four bacterial phyla accounted for nearly 93% of all identified bacteria, with substantial variation in the abundance of phyla among individual microbiomes (Firmicutes, 11 to 80%; Bacteroidetes, 0.5 to 75%; Proteobacteria, 0 to 35%; and Actinobacteria, 1 to 85%) (Fig. 1A). These four main phyla were observed in all intestinal regions of all calves (n = 15), with the exception of Proteobacteria, which was absent from the proximal jejunum of one 6-week-old calf (6W). Individual animal variation in the bacterial composition was highest within the lower taxonomic hierarchy, and no single bacterial family or genus was detected in all of the intestinal samples from all animals.
Postnatal changes in the small intestinal microbiomes of preweaned dairy calves. (A) Individual calf variation of the relative abundance of four main bacterial phyla in the digesta-associated communities collected from three different regions of the small intestine at three different ages. Each data point represents an individual calf, and the bar represents the mean relative abundance of a small intestinal region or an age group. The upper panel presents all three age groups within a small intestinal region (15 data points/phylum, 3 charts for 3 intestinal regions), and the lower panel presents all three small intestinal regions within an age category (15 data points/phylum, 3 charts for 3 age groups). (B) The abundance of the small intestinal subsystems (level 1 microbial functions of the SEED hierarchy, only the abundant subsystems [mean relative abundance >1%] are presented). Numerical values represent individual calf IDs. 1W, 1-week-old calves; 3W, 3-week-old calves; 6W, 6-week-old calves. Each data series represents relative abundance of detected subsystems.
Among the bacterial genera detected in the neonatal calf intestine, Clostridium and Eubacterium were present in all jejunal microbiomes, whereas Prevotella and Bacteroides were observed in all ileal microbiomes. Bifidobacterium (5.7% ± 2.2%), Prevotella (7.5% ± 1.8%), Lactobacillus (15.4% ± 3.4%), Clostridium (5.6% ± 0.1%), Bacteroides (8.2% ± 2.6%), Streptococcus (4.6% ± 1.2%), and Eubacterium (2.6% ± 0.4%) were present in all sampled calves, although they were not detected in all intestinal regions surveyed (Fig. S2). For example, Eubacterium was detected in the ileum of all but one 1W calf. Significant temporal variations were evident when comparing the relative abundances of the detected taxa among different age groups within a small intestinal region (Table 1 ). For example, the relative abundance of Ruminococcus was significantly higher (P < 0.05) in older calves (3W and 6W) than 1W calves only in the proximal jejunum (Table 1). Similarly, the relative abundance of Lactobacillus was significantly lower (P < 0.05) in older calves than 1W calves in two jejunal regions but not in the ileum (Table 1). Among the observed significantly different bacterial taxa, many were different when comparing 1W calves with either 3W or 6W calves, but not between 3W and 6W calves (Table 1). There were no regional differences observed in the abundance of bacterial taxa along the calf small intestine either when comparing within an age group or when comparing all age groups together.
Temporal variations in the small intestinal digesta-associated metagenome (taxonomy and function) of the preweaned calves within each small intestinal region
Although temporal variations observed in the small intestinal microbiomes regardless of region were not statistically different, there were increasing (higher relative abundance in older calves) and decreasing (lower relative abundance in older calves) trends in the relative abundance of individual bacterial genera as calves aged (Fig. S3). For example, the mean relative abundances of Bifidobacterium (1W, 2.3% ± 1.0%; 3W, 5.0% ± 2.1%; 6W, 9.8% ± 6.9%) and Ruminococcus (1W, 0.8% ± 0.2%; 3W, 1.7% ± 0.1%; 6W, 2.4% ± 0.5%) increased with calf age (Fig. S3 in the supplemental material), whereas the relative abundances of Bacteroides (1W, 17.3% ± 5.6%; 3W, 3.8% ± 0.6%; 6W, 3.5% ± 0.8%) and Lactobacillus (1W, 21.9% ± 2.4%; 3W, 14.9% ± 10.3%; 6W, 9.4% ± 1.2%) decreased with calf age (Fig. S3).
(ii) Mucosa-attached microbiota.Amplicon sequencing of the V1-V3 region of the 16S rRNA gene revealed that the small intestinal mucosa-attached bacterial community comprised 6 phyla, 38 families, and 104 genera. Similar to the luminal bacteria, the relative abundance of each phylum varied greatly among individual calves (Actinobacteria, 0 to 93.3%; Firmicutes, 0 to 94.4%; Proteobacteria, 0 to 99.9%; and Bacteroidetes, 0 to 35.3%) (Data Set S2, sheet 1). No single bacterial genus was observed in all tissue samples analyzed. The predominantly observed mucosa-attached bacterial genera were Propionibacterium (7.3% ± 4.0%), Prevotella (7.0% ± 2.9%), and Lactobacillus (4.8% ± 2.0%) in 1W samples, Propionibacterium (12.2% ± 8.9%) and Prevotella (6.6% ± 3.3%) in 3W samples, and Bifidobacterium (8.5% ± 4.2%), Sharpea (4.6% ± 3.8%), and Prevotella (3.0% ± 1.4%) in 6W samples (Fig. S2). Prevotella was detected in at least one intestinal region of all calves sampled. When the relative abundances of the mucosa-attached bacterial groups were compared, there were no significant differences observed among either the intestinal region or calf age.
Diversity and richness of the small intestinal microbiota.The diversity (Shannon index) of the luminal microbiota increased with calf age and dietary changes, but this increase was not significantly different when comparing among age groups (Table 2). Although the total number of bacterial genera detected was not different among age groups, there was a significantly (P = 0.02) lower number of predominant genera (relative abundance of >1% in at least one sample) in the ileal lumen of 3W and 6W calves than in the age-matched jejunal communities (Table 2). When the bacterial diversity and richness (number of detected genera) were compared for mucosa-attached communities, they were not significantly different among either intestinal regions or age groups (Table 2).
Small intestinal bacterial diversity and richness of neonatal calvesa
Small intestinal bacterial densities in neonatal calves.Real-time PCR-based estimation (using DNA) revealed that the density of total bacteria and the proportion of Lactobacillus were significantly lower (P < 0.01) in the mucosa-attached communities compared to the luminal communities (Table 3). The individual animal variation was high in the estimated total bacterial density, ranging from 2.63 × 108 to 3.22 × 1011 16S rRNA gene copies/g of sample. Similarly, the density of Lactobacillus varied greatly, between 1.49 × 104 and 3.72 × 109 16S rRNA gene copies/g of sample, among individuals. The proportion of Bifidobacterium was affected by the three-way interaction effect of region by age by sample type (P = 0.02). The proportion of Bifidobacterium was significantly higher in the lumen of the proximal jejunum at 1W compared to those of 3W and 6W (Table 4). The proportion of Bifidobacterium in the proximal jejunum at 1W was also higher in the lumen than the respective mucosa-attached community (Table 4). Similar to total bacteria and Lactobacillus, the density of Bifidobacterium varied greatly among individual animals (1.19 × 107 to 2.21 × 1010 16S rRNA gene copies/g of samples). However, the coefficient of variation (CV) of Bifidobacterium (6.5%) was lower than that of total bacteria (18.5%) and Lactobacillus (18.8%).
Bacterial density in the small intestinal communities of preweaned calves
Bifidobacterium density in the small intestinal communities of preweaned calves
Postnatal changes in the functions of the small intestinal luminal microbiome.Functional assignment using MG-RAST revealed 29 subsystems/level 1 functions (Data Set S2, sheet 2) in the SEED hierarchy for the collective small intestinal microbiomes (all gut regions of all calves, from here onward we refer to them as the collective microbiomes). The functions of the collective microbiomes were dominated by genes involved in the metabolism of carbohydrates and proteins (Fig. 1B). The “respiration” subsystem was observed in all microbiomes, and it contained enzymes involved in fermentation, such as anaerobic glycerol-3-phosphate dehydrogenase (EC 1.1.5.3) and formate hydrogenlyase. Although other subsystems were not observed in all microbiomes, they were detected in more than 70% of the intestinal microbiomes generated in the study. The top subsystems identified from the collective microbiomes of the calf small intestine were “carbohydrate metabolism” (range, 0 to 13.7%; mean ± SEM = 10.5% ± 0.6%), “protein metabolism” (range, 0 to 19.5%; mean ± SEM = 9.7% ± 0.6%), “amino acids and derivatives” (range, 0 to 12.6%; mean ± SEM = 6.7% ± 0.4%), “phages, prophages, transposable elements, plasmids” (range, 0 to 95.4%; mean ± SEM = 9.3% ± 3.0%), “DNA metabolism” (range, 0 to 9.6%; mean = 4.8% ± 0.3%), “RNA metabolism” (range, 0 to 13.7%; mean ± SEM = 4.8% ± 0.3%), and “cofactors, vitamins, prosthetic groups, pigments” (range, 0 to 8.1%; mean ± SEM = 4.3% ± 0.3%).
Temporal changes in the relative abundances of the subsystems were observed when pairwise comparisons (1W versus 3W, 1W versus 6W, and 3W versus 6W) were performed within each gut region (Table 1). For example, the abundance of “protein metabolism” was higher in the older calves (3W, 6W) than 1W calves in the two jejunal regions but not in the ileum (Table 1). Similarly, a lower abundance of “phages, prophages, transposable elements, plasmids” in the older calves (compared to 1W calves) was also observed only in the two jejunal regions (Table 1). Once again, the temporal changes in the relative abundances of these subsystems were mainly observed when comparing 1W calves against older (3W and 6W) calves.
When the metagenomes generated from all three small intestinal regions and metagenomes generated from all three age groups were compared, there were no significant temporal or regional changes, respectively. However, the subsystems could be categorized into three patterns depending on the observed temporal trends of the relative abundance (Fig. S4). The abundance of essential functions for all microbiota, such as “protein metabolism,” “carbohydrate metabolism,” and “DNA and RNA metabolism,” as well as biosynthetic processes (“cofactors, vitamins, prosthetic groups, pigments”), increased from the first to the third week and then remained stable from the third week to the sixth week (Fig. S4). In contrast, the abundance of “phages, prophages, transposable elements, plasmids,” “membrane transport,” “virulence, disease, and defense,” “iron acquisition,” “respiration,” and “sulfur and potassium metabolism” decreased from the first to the third week and then remained stable from the third week to the sixth week (Fig. S4). However, the abundances of “motility and chemotaxis,” “cell wall and capsule,” “regulation and cell signaling,” “secondary metabolism,” “phosphorus metabolism,” and “stress response” remained stable throughout the first 6 weeks of life (Fig. S4).
In total, 185 level 2 functions in the SEED hierarchy were detected within the small intestinal microbiomes (proximal jejunum, 179; distal jejunum, 144; ileum, 183), with no single level 2 function shared among all microbiomes. We identified 16 and 5 level 2 functions present in all distal jejunum and ileum microbiomes, respectively, but no level 2 functions were conserved among all of the proximal jejunum microbiomes (Data Set S2, sheet 2). Microbial functions related to “protein biosynthesis,” “folate and pterines,” “DNA repair,” “lysine, threonine, methionine, and cysteine,” and “RNA processing and modification” were conserved in the distal jejunum and ileum. Temporal changes in the abundance of level 2 functions were evident through pairwise comparisons conducted within each small intestinal region (Table 1). When the collective metagenomes were compared, there were no significant temporal or regional differences observed in the abundance of level 2 functions. The abundances of “DNA repair,” “di- and oligosaccharides,” “carbohydrates,” “transcription,” “alanine, serine, and glycine,” and “pyridoxine (vitamin B6)” increased numerically with age (Fig. S4). In contrast, the abundances of “electron-donating reactions,” “protein biosynthesis,” “cell division and cell cycle,” “central carbohydrate metabolism,” “lysine, threonine, methionine, and cysteine,” and “ABC transporters” decreased numerically with age (Fig. S4).
Individual variation in the taxonomic and functional composition of the luminal microbiome.Principal-coordinate analysis (weighted and unweighted UniFrac distance matrices built using operational taxonomic units clustered at 97% similarity) resulted in no segregation of luminal microbiomes based on either small intestinal region or calf age (Fig. S5). We compared coefficients of variance (CV) within an individual animal to those of the relevant age group to understand the individual variation of luminal microbiomes (Fig. S6; Data Set S2, sheet 4). A lower intraindividual variation (variation among three intestinal regions) than the interindividual variation (variation among individuals within an age group) was evident in microbiomes of all age groups at the functional and taxonomic levels. Therefore, the taxonomic and functional compositions of the collective luminal microbiomes (45 metagenomes generated from all luminal samples) were further analyzed using a hierarchical clustering approach.
Use of the 27 most frequently detected bacterial genera (in at least half of the lumen microbiomes or ≥22 of 45 microbiomes) revealed that the majority of luminal microbiomes (24 of 45 microbiomes) were not correlated with one other based on the taxonomic composition. However, we observed two distinct clusters (Fig. 2A) among the highly correlated (R2 ≥ 0.9, P < 0.01) luminal microbiomes. One cluster consisted of microbiomes containing a higher relative abundance of Lactobacillus (ileal profiles of four calves: 1W, calf5 [72.4%]; 3W, calf7 [86.4%] and calf8 [85.8%]; 6W, calf12 [56.4%]; jejunal profiles of six calves: proximal jejunum: 1W, calf2 [40.4%], calf3 [23.0%], and calf5 [36.2%]; 6W, calf12 [28.0%]; distal jejunum: 1W, calf1 [38.7%], calf3 [14.9%], and calf5 [58.1%]; 3W, calf8 [14.9%]; 6W, calf12 [31.2%]) compared to all other microbiomes. Another cluster consisted of microbiomes containing a higher relative abundance of Bacteroides compared to all other microbiomes (0.01 to 5.5%; all three small intestinal regions of one 1W calf [calf4: proximal jejunum, 23.7%; distal jejunum, 41.7%; ileum, 56.3%], the distal jejunum of one 1W calf [calf2, 22.2%] and one 6W calf [calf11, 16.4%], and the ileum of two 1W calves [calf2, 45.8%; calf3, 33.9%], as well as one 3W calf [calf6, 17.9%]). Comparison of the relative abundance of bacterial genera between two clusters revealed that Bacteroides, Prevotella, Roseburia, Ruminococcus, and Veillonella were significantly lower in the Lactobacillus-dominant ileal microbiomes compared to the Bacteroides-dominant ileal microbiomes (Table S1).
Clustering of the collective microbial profiles based on taxonomic/functional similarities. (A) Clustering of calves based on microbial taxonomic composition. (B) Clustering of calves based on microbial functional abundance. Numbers on the x axis represent calf IDs. PJ, proximal jejunum; DJ, distal jejunum; IL, ileum. Abundances of bacterial genera between two function-based clusters are compared using Metastats, and data are presented as means ± the SEM. Clustering is based on the Spearman rank correlation coefficient between two samples generated either based on the relative abundance of 27 bacterial genera identified in at least 50% of samples or based on the relative abundance of subsystems. Each row and column represent an individual microbial profile generated from small intestinal samples (45 microbial profiles).
Hierarchical clustering of the collective metagenome using functional composition revealed a high correlation (R2 ≥ 0.9, P < 0.01) among the majority of luminal microbiomes (34 of 45 microbiomes), and we identified two distinct clusters among these highly correlated microbiomes (Fig. 2B). One cluster contained samples from 3W and 6W calves (cluster 1), whereas the other cluster contained samples from 1W and 3W calves (cluster 2). Comparison of these two clusters revealed that functions related to “protein metabolism” (cluster 1, 12.3% ± 0.5%; cluster 2, 9.0% ± 0.5%; P < 0.01), “amino acids and derivatives” (cluster 1, 7.9% ± 0.4%; cluster 2, 6.4% ± 0.6%; P = 0.06) and “nucleosides and nucleotides” (cluster 1, 5.8% ± 0.3%; cluster 2, 3.8% ± 0.3%; P < 0.01) were significantly enriched in cluster 1 compared to cluster 2. In contrast, functions related to “cell wall and capsule” (cluster 1, 3.1% ± 0.3%; cluster 2, 4.4% ± 0.3%; P = 0.01) and “sulfur metabolism” (cluster 1, 0.3% ± 0.05%; cluster 2, 0.8% ± 0.2%; P = 0.02) were enriched in cluster 2 compared to cluster 1. Comparison of the taxonomic composition of the microbiomes belonging to the two functional clusters revealed that the relative abundances of the predominant genera Bifidobacterium, Bacillus, Streptococcus, Lactococcus, and Corynebacterium were significantly higher in cluster 1 than cluster 2, while the abundance of sulfur-reducing bacteria (Desulfarculus, Dethiosulfovibrio, Desulfatibacillum, Desulfuromonas, Desulfurispirillum, and Desulfotalea) was significantly higher in cluster 2 than cluster 1 (Fig. 2B; Table S2). Furthermore, there were 19 and 16 additional genera that were relatively enriched in cluster 1 and cluster 2, respectively (Table S2).
Linking identified ileal bacterial taxonomy-based clusters to the host transcriptome.To explore whether the variations observed in the ileal microbiomes could also be reflected in the host at the transcriptome level, ∼3,000 transcripts involved in mucosal immune-related functions were extracted from a previously published ileal transcriptome data set generated from the same calves (16) (Data Set S3). Among these immune-related transcripts, 28 were upregulated (>2-fold change) in the calves with a higher abundance of Lactobacillus in the ileum than in those with a higher abundance of Bacteroides in the ileum (Data Set S2, sheet 5). Among these upregulated (enriched) transcripts, six were significantly different (P < 0.05) between the two groups, and five tended to be different (0.05 < P < 0.1). Similarly, 63 immune-related transcripts were upregulated in the calves with a high abundance of Bacteroides in the ileum compared to levels in calves with a high abundance of Lactobacillus in the ileum (Data Set S3, sheet 5). Among these enriched transcripts, 32 were significantly different (P < 0.05) between the two groups, while 15 tended to be different (0.05 < P < 0.1). GO enrichment of the transcripts upregulated in the ileal Lactobacillus-dominant calves revealed that they were involved in “leukocyte and lymphocyte chemotaxis,” “cytokine/chemokine-mediated signaling pathway,” and “inflammatory responses” (Fig. 3A; Data Set S2, sheet 6). Protein-coding genes CXCL9 (P = 0.12), CXCL10 (P < 0.05), CXCL11 (P < 0.05), CCL2 (P = 0.18), and CCL14 (P < 0.1) were mainly annotated to these enriched functions (Fig. 3B). In contrast to the ileal Lactobacillus-dominant calves, the ileal Bacteroides-dominant calves had functions enriched for processes related to “cell adhesion,” “response to stimulus,” “cell communication,” and “regulation of MAPK cascades” (Fig. 3A; Data Set S2, sheet 7). The main annotated protein coding genes for these enriched functions included PLCE1 (P < 0.05), CCL22 (P < 0.05), EGFR (P < 0.05), IL1RL1 (P = 0.05), IL1R2 (P = 0.10), IL17RD (P = 0.53), and ITGB6 (P = 0.15) (Fig. 3B).
Ileal mucosal immune-related genes and functions of calves belong to two taxonomic clusters. (A) Functions of the genes upregulated/enriched in taxonomic clusters. (Upper panel) Functions of the genes enriched in the ileal tissue of the Lactobacillus-dominant calves; (lower panel) functions of the genes enriched in the ileal tissue of the Bacteroides-dominant calves. Functional enrichment was performed using GO enrichment analysis in Gene Ontology Consortium. The P value indicates significance for the enrichment in the data set of the listed GO identifier. A P value close to zero, that is, a higher –log(P), means that the group of genes associated with the particular GO term is more significant and it is less likely the observed annotation of the particular GO term to a group of genes occurs by chance. (B) Main genes annotated to enriched immune functions. Each bar represents the fold change expression of annotated genes. The fold change is calculated by dividing the expression of genes (counts per million [cpm]) in Lactobacillus-dominant calves by the expression of genes (mean cpm) in Bacteroides-dominant calves for the Lactobacillus-dominant cluster and vice versa for the Bacteroides-dominant cluster. CXCL9, chemokine ligand 9; CXCL10, chemokine ligand 10; CXCL11, chemokine ligand 11; PLCE1, phospholipase C epsilon 1; CCL22, C-C motif chemokine ligand 22; EGFR, epidermal growth factor receptor.
DISCUSSION
Although functions of the human gut microbiome have been studied extensively through metagenomics approaches, such knowledge is very limited for ruminants, especially preruminants. The present study profiled the small intestinal microbiomes of healthy neonatal calves to gain a better understanding of the taxonomic and functional composition and temporal dynamics associated with the establishment of the gut microbiome during early life. Calf small intestinal microbiomes were mainly colonized by bacteria, although metagenome sequencing of the luminal content revealed the presence of archaea, fungi, protozoa, and viruses from the first week of life. The genus Methanobrevibacter dominated the small intestine archaeal population, whereas the family Microviridae dominated the viral (phage) population of the neonatal calf intestine (Fig. S1B). These observations are in agreement with human studies reporting a higher abundance of Methanobrevibacter (17) and Microviridae (18) in fecal microbiomes and the presence of Methanobrevibacter in preweaned dairy calves (19, 20). The presence of the fungal phyla Ascomycota and Basidiomycota (Fig. S1B), which were the most abundant fungi identified in the present study, was also reported in a metagenomics-based study conducted using fecal samples of healthy humans (21). The main protozoa observed in the neonatal calf small intestine belonged to the genus Cryptosporidium (Fig. S1B). C. parvum is the main parasitic protozoan that contributes to neonatal calf diarrhea; however, calves may harbor Cryptosporidium even in the absence of clinical signs or symptoms (22). This is the first study to present the composition of all microbial groups in the intestine of neonatal calves.
The increased diversity and richness in the small intestinal bacterial communities (both luminal and tissue) with increasing age suggest a progressive establishment of a complex microbiome during the first 6 weeks of life. Furthermore, total bacterial densities (qPCR-based quantification) of the small intestinal tissue and lumen environments of neonatal calves reached levels similar to those of weaned calves (23) from the first week of life. The calves used in this study were fed whole milk during the first week postpartum and whole milk with ad libitum calf starter supplementation thereafter. Thus, the observed changes in the luminal microbiomes were both age and diet driven. The dietary changes may have contributed substantially to the significant differences in microbial composition observed when comparing 1W versus 3W and 6W calves. The higher similarity between 3W and 6W calves further suggests that the luminal microbiome may have started to stabilize once solid feed was introduced. Regardless, a high individual animal variation in the composition and density of bacteria was evident throughout the neonatal period. In humans, the fecal microbiome begins to stabilize and increase in similarity among individuals around 3 years of age with the intake of a stable solid diet (24). Similarly, high individual animal variation in the rumen microbiota has been reported to decrease with increasing age in cattle (25). Thus, high individual variation in the intestinal microbiome may be a conserved trait in neonates of different species. Host genetics has also been shown to shape gut microbiomes, leading to an individualized gut microbiome in mice (26) and heritable bacterial taxa in humans (27–29). The impact of host genetics may be one of the factors contributing to the high individual variation observed in the gut microbiomes of neonates, whose diet is not yet stabilized.
Principal-coordinate analysis revealed no segregation of luminal microbial profiles in the present study. However, the correlation-based hierarchical clustering approach we used grouped the collective luminal microbiomes based on their taxonomic and functional compositions. Metagenome sequencing of the fecal microbiome has also been used to categorize the human population broadly into three enterotypes based on the individual microbial profile (30). Despite the small number of samples, our approach also resulted in two taxonomic clusters of microbiomes with either high Lactobacillus or high Bacteroides abundances in the ilea of the neonatal calves, suggesting potential enterotyping of neonatal calves. Comparisons of the ileal transcriptomes of the calves belonging to these two clusters further revealed variations in the expression of mucosal immune-related genes and enriched immune functions between them. These findings together suggest a link between potential enterotyping of neonatal calves and their mucosal immune status during early life.
The Lactobacillus-dominant calves in the ileal lumen exhibited a higher expression of CXCL9, CXCL10, and CXCL11 in the ileal tissue than did the Bacteroides-dominant calves in the ileal lumen. These proinflammatory chemokines have been shown to be upregulated in dendritic cells and human intestinal epithelial cells after exposure to Lactobacillus acidophilus (31, 32). However, such upregulation quickly returned to baseline levels one hour after exposure to L. acidophilus (32), suggesting a potential priming of immune responses by Lactobacillus during the first encounter. The priming effect of early Lactobacillus exposure may prepare the host for subsequent anti-inflammatory regulatory responses, which may be a crucial process in neonatal calves with a naive mucosal immune system. In contrast to these studies, infectious models (colitis and pneumovirus infection in mice) have demonstrated that Lactobacillus exerts an anti-inflammatory effect through the downregulation of CXCL10 and CXCL11 (33, 34). Therefore, these studies suggest that the beneficial effects of Lactobacillus may vary depending on the context, with a priming of immune responses upon initial encounter but a reduction of inflammatory responses during infections.
The increased expression of PLCE1, CCL22, and EGFR in the ileal tissue of Bacteroides-dominant calves suggests an increased capacity for T-cell-dependent responses in the guts of these calves, since these mitogen-activated protein kinase cascade-related genes are involved in the regulation of T-cell differentiation (35). Our transcriptome data from calves with a higher abundance of either Lactobacillus or Bacteroides in the ileal lumen revealed differential expression of genes that may be involved in the activation of mucosal immune responses. The identified enriched mucosal immune functions associated with taxonomy-based microbiome clusters may indicate that the calves with a “beneficial-bacterium-enriched microbiome” are resilient to enteric health issues such as diarrhea during early life. Thus, validation of the enriched genes in a larger population of healthy animals and/or use of enteric challenged models with phenotypic measurements will be required to determine whether Lactobacillus and Bacteroides represent enterotypes in the calf small intestine that are associated with specific immune functions and health outcomes.
Although taxonomic profiles of the luminal microbiomes were markedly different among individual calves, such differences were not directly reflected in functional profiles. This may be due to the observed higher abundance of universal microbial functions that are important for microbial survival, such as “protein biosynthesis,” “RNA processing and modification,” “DNA replication,” “central carbohydrate metabolism,” and “cell cycle.” Nevertheless, two distinct clusters based on functional abundance were observed from the collective luminal microbiomes of neonatal calves. Functional cluster 1 was enriched for functions related to the microbial protein metabolism, which has been suggested to play an important role in host health through the regulation of amino acid homeostasis in the gut (36).
The luminal microbiomes of functional cluster 1 had a higher abundance (69-fold) of Bifidobacterium than those in functional cluster 2, while those of functional cluster 2 had a higher abundance of sulfur-reducing bacteria (SRB) than functional cluster 1. A recent study has shown that Bifidobacterium utilizes milk protein effectively and requires cysteine, a sulfide amino acid, for growth (37). SRB also utilize cysteine as an energy source (38). Thus, the observed low abundance or absence of Bifidobacterium in the luminal microbiomes of functional cluster 2 calves may be due to a lack of cysteine for their growth as a result of competition with SRB. Dissimilatory sulfur reduction and cysteine degradation by SRB produce H2S, which can have detrimental effects on intestinal epithelial cells (38). An accumulation of H2S in the colon inhibits the activity of acyl coenzyme A dehydrogenase, which catalyzes the oxidation of butyrate in colonocytes and compromises epithelial barrier permeability (39). Bifidobacterium has been reported to promote intestinal epithelial barrier functions (40, 41), while SRB have been reported to have a negative impact on the epithelial barrier (39). Although these observed differences in bacterial composition might suggest potential differences in epithelial barrier functions between functional clusters 1 and 2, transcriptome data (16) revealed no difference in the expression of tight-junction genes (<2-fold change in the expression of genes belonging to the claudin family, occludin, zonula occludens, and junctional adhesion molecules, with P > 0.1) when comparing animals in these two clusters. Intestinal barrier integrity loss can be measured using changes in the expression of tight-junction proteins, epithelial cell damage, or the translocation of bacterial products (42). Therefore, assessing barrier function by using these alternative indicators may provide a more sensitive measurement of how compositional and functional variations in the gut microbiome impact epithelial barrier function.
A high rate of diarrhea-related neonatal deaths remains a major health problem in the cattle industry (43). Neonatal calf diarrhea is often treated with antibiotics, and often prophylactic antibiotics are given to calves during the first month to prevent colonization of the small intestine by pathogenic Escherichia coli (one of the diarrhea-causing bacteria) (44). With the increasing restrictions on the prophylactic use of antibiotics (44), it is important to explore other microbial manipulation techniques that can limit or prevent pathogen colonization of the calf small intestine. Feeding probiotics has been shown to reduce the frequency of pathogenic E. coli in the calf gastrointestinal tract (45), suggesting that promoting the colonization of beneficial bacteria may exclude pathogens. Limited knowledge of the calf microbiome, especially in the small intestine, is a barrier to developing such strategies to prevent or treat diarrhea. Thus, our findings on gut microbial markers linked with mucosal immune functions during early life may pave the way to developing strategies to manipulate the microbiome and improve calf health.
The taxonomic composition of the small intestinal bacteria observed using metagenome sequencing revealed that the luminal microbiome in the present study was similar to previous studies reporting a higher abundance of Firmicutes in the small intestine (8) and the fecal microbiome (7) of young calves using 16S rRNA amplicon sequencing. However, there were significant differences within the lower taxonomic hierarchy that could be attributable to differences in the microbial profiling tools (whole-genome sequencing versus 16S amplicon sequencing), calf age, and sample type (fecal versus small intestine). Sequencing the V1-V3 region of 16S rRNA genes revealed a higher abundance of Lactobacillus in the ileum and Sharpea in the proximal jejunum at 3 weeks of life (8). The use of metagenomic profiling of luminal content in the present study, however, revealed a higher abundance of Prevotella (17.9% ± 12.8%) in the ileum and Streptococcus (12.0% ± 6.1%) in the proximal jejunum of 3-week-old calves. In contrast to luminal communities, the mucosa-attached communities in the present study differed from those reported in a previous study, even at higher levels of the taxonomic hierarchy (8). In contrast to our previous study (8), the present study revealed a higher abundance of Actinobacteria (41.2% ± 15.1%) in the ileal tissues, followed by Firmicutes (26.8% ± 9.6%), Proteobacteria (20.1% ± 15.5%), and Bacteroidetes (11.8% ± 6.1%), and a higher abundance of Proteobacteria (66.5% ± 23.2%) in the jejunal tissues, followed by Firmicutes (14.0% ± 8.1%), Actinobacteria (11.0% ± 10.5%), and Bacteroidetes (8.0% ± 4.2%), at 3 weeks of age. The use of animals from different herds and feeding calves different starter rations may explain some of the observed differences between these two studies. In addition, the small intestinal tissue-attached bacterial community is closely scrutinized by the host mucosal immune system (46), which may also contribute to taxonomic variation in the microbial communities among different animals.
Conclusion.This study generated a detailed analysis of the taxonomic and functional development of the small intestinal microbiome in healthy preweaned calves using metagenome sequencing of lumen content. During this developmental period, age- and diet-related increases in microbially diverse populations, richness, and density were observed; however, the taxonomic composition (both tissue and lumen) varied widely among individual animals. The taxonomic comparison of luminal microbiomes classified calves as either Lactobacillus dominant or Bacteroides dominant in the ileum, while the functional comparison of luminal microbiomes revealed an antagonism between SRB and Bifidobacterium. These distinct taxonomic and functional clusters may provide a framework to further analyze interactions between the intestinal microbiome and the immune function, health, and growth of neonatal calves to improve calf gut health.
MATERIALS AND METHODS
Animal experiments and sampling.All experimental protocols were approved by the Livestock Care Committee of the University of Alberta (AUP00001012) and were conducted following the guidelines of the Canadian Council on Animal Care. This study used Holstein bull calves (n = 18) reared at the Dairy Research and Technology Center (DRTC), University of Alberta (Edmonton, Alberta, Canada), under the DRTC standard management practices. Calves used in the present study received 4 liters of colostrum/day during the first 3 days postpartum and 4 liters of whole milk/day from the fourth day up to the end of the trial. Calves had ad libitum access to calf starter (canola meal, soybean meal, wheat, barley, corn, peas, and vitamins and minerals; 29.5% crude protein; 27.1% starch [Wetaskiwin Co-Op Country Junction, Wetaskiwin, Alberta, Canada]) from the second to the sixth week of life. Calves used in the study had no records of respiratory or enteric diseases or use of treatments for any disease. Calves were humanely euthanized using a captive bolt gun stunning method and sampling was completed within 30 min after euthanization. Mucosal tissue and lumen (digesta) samples (proximal jejunum, distal jejunum and ileum) were collected from calves at week 1 (1W; n = 6), week 3 (3W; n = 6), and week 6 (6W; n = 6); snap-frozen; and stored at −80°C. First, digesta was collected into 50-ml Falcon tubes, and then mucosal tissue was rinsed three times with sterile phosphate-buffered saline (pH 7.2) before transfer into tubes. Sampling sites were maintained consistently among all animals. Ileal samples were collected 30 cm proximal to the ileocecal junction, distal jejunal samples were collected 30 cm proximal to the collateral branch of the cranial mesenteric artery, and proximal jejunal samples were collected 100 cm distal to the pylorus sphincter. Ten-centimeter intestinal segments were collected from each site in the middle of each segment aligned with the above-mentioned measurements.
Analysis of the small intestinal microbiomes.Total DNA was extracted from tissue and digesta samples using the repeated bead-beating plus column method (47). Briefly, samples were first ground using mortar and pestle with liquid nitrogen and ∼0.5 g of samples was measured into a 2-ml screw-cap tube (BioSpec, Bartlesville, OK) containing 0.5 g of 0.5-mm zirconia beads (BioSpec). The samples were then subjected to bead beating using a Mini-Beadbeater 8 (BioSpec) at 5,000 rpm for 3 min in the presence of a sodium dodecyl sulfate (SDS)-based lysis buffer. Impurities and SDS used during bead-beating steps were removed by ammonium acetate and nucleic acids were precipitated using isopropanol. Finally, recovered nucleic acids were purified using QIAamp Fast DNA stool minikit (Qiagen, Frederick, MD) according to the manufacturer’s instructions. DNA concentration was measured using Qubit 2.0 (Thermo Fisher Scientific, Massachusetts) and used in library preparation.
DNA libraries were prepared for metagenome sequencing (n = 5/age group, 45 small intestinal luminal samples; one animal was removed during the library preparation due to the low DNA amount) using the TruSeq DNA PCR-free library preparation kit (Illumina, CA). The genomic DNA was first normalized with a resuspension buffer to a final volume of 55 µl at 20 ng/µl, and 50 µl of the solution was transferred into a Covaris microTUBE (Covaris, Inc., MA) for fragmentation using a Covaris S2 focused-ultrasonicator. Then, the cleaned up fragmented DNA was subjected to end repair and size selection (350 bp), followed by the adenylation of the 3′ ends and ligation of the adaptor index. Each metagenomic library was then quantified using a Qubit 2.0 Fluorometer (Thermo Fisher Scientific, MA) and whole-genome sequencing of 100-bp paired-end reads was performed using HiSeq 2000 (Illumina, CA) and TrueSeq cluster kit V3 and TrueSeq SBS kit V3 sequencing kits at Génome Québec (Montréal, Canada).
Raw sequences (6.91 Gb) of metagenome sequencing were first demultiplexed (CASAVA version 1.8; Illumina) and uploaded into the MG-RAST metagenomic analysis server (48), version 3.3.9, and paired ends were joined before submitting for processing. Artificial replicates, host (bovine) DNA, and low quality (Phred score < 25) sequences were removed from the raw data, and the remaining good quality sequences (1.82 Gb; 26% of the generated raw data) were used to assign taxonomy and functions. The subsystems approach in the SEED hierarchy was used to assign microbial functions of the small intestinal microbiomes, and the taxonomic assignments were made using M5 nonredundant protein database (M5NR) within MG-RAST. A maximum cutoff E value of 1e–10, maximum identity of 70% and a maximum alignment length of 50 were used for data selection criteria for both taxonomy and function abundance analyses. Taxonomic and functional abundances of luminal microbial communities obtained from the MG-RAST platform were analyzed using metastats (49) to explore temporal (1 week versus 3 weeks, 1 week versus 6 weeks, and 3 weeks versus 6 weeks) and regional (proximal jejunum [PJ] versus distal jejunum [DJ]; PJ versus ileum [IL]; and DJ versus IL) changes in the abundance of the identified microbial taxa and functions.
DNA extracted from tissue samples was used in 16S rRNA gene-based amplicon sequencing (V1-V3 region) due to the higher abundance of host DNA in the genomic DNA extracted from tissue, which would be sequenced during the whole-genome sequencing. DNA extracted from tissue samples (n = 3/age group, 27 small intestinal tissue samples) was amplified using 27F and 338R primers containing barcodes (8), and purified PCR products (∼400 bp) were subjected to 454 sequencing (Roche GS-FLX Titanium) at Génome Québec. Data were analyzed using the Quantitative Insight into Microbial Ecology (QIIME) toolkit (50), following the removal of low-quality (Phred score < 25, length < 100 bp) and chimeric sequences as described by Malmuthuge et al. (8). Briefly, chimeric sequences were filtered out using usearch61 within the QIIME platform, and the remaining high quality sequences were used to perform operational taxonomic unit clustering at 97% similarity and taxonomic assignment using the Greengenes database (13_8 release).
Estimation of small intestinal bacterial densities using quantitative real-time PCR.Total bacterial, Lactobacillus, and Bifidobacterium densities (n = 6/age group, 54 luminal samples and 54 tissue samples) were estimated using previously published primers (51) and SYBR green chemistry (Fast SYBR Green Master Mix; Applied Biosystems) with a StepOnePlus real-time PCR system (Applied Biosystems, Foster City, CA). The copy number of 16S rRNA gene/g of fresh tissue or luminal content was then calculated using the equation described by Li et al. (52). All bacterial density data were analyzed using the MIXED procedure and repeated measurement experimental design in SAS (SAS 9.4; SAS, Inc., Cary, NC) with small intestinal region as the repeated measurement and animal as the experimental unit. Compound symmetry covariance structure was selected as the best fit by the Bayesian information criteria (BIC). The following statistical model was fitted to test the effect of age, gut region, and sample type on bacterial density: Yijkl = μ + Ai + Rj + Sk + (AR)ij + (AS)ik + (RS)jk + (ARS)ijk + eijkl, where Y is the bacterial density (total bacteria, Lactobacillus, Bifidobacterium), μ is the mean, A is the calf age, R is the small intestinal region, S is the sample type (tissue, content), and e is the residual error.
Clustering analysis of luminal microbiomes.We used a hierarchical clustering approach to obtain the associations among the luminal microbiomes using metagenomes generated from all small intestinal samples (45 samples), regardless of calf age or small intestinal region. The 27 most frequently detected (in at least half of the luminal samples or ≥22 of 45 samples) bacterial genera were used for the taxonomic composition clustering and 29 subsystems (level 1 functions) were used for the functional composition. First, a Spearman rank correlation analysis was performed using the cor() function in R to obtain a correlation (ρ) between each pair of small intestinal microbiomes. Then, a distance matrix containing the square root of ρ values was used to perform hierarchical clustering using the hclust() function within R package (V 3.4.1). The observed taxonomic and functional clusterings were then validated by assessing the clustering tendency using the get_clust_tendency() function in the R packages cluster and factoextra to calculate Hopkins statistics. Approximately unbiased (AU) probability values (P values) of the validated clusters were then calculated using a multiscale bootstrap resampling (1,000 bootstrap replications) in the R package pvclust (53) to identify distinct clusters among the highly correlated luminal microbiomes. The relative abundances of bacterial genera and microbial functions between two taxonomic clusters and two functional clusters were then compared using a nonparametric t test (Mann-Whitney-Wilcoxon) in R (3.3.1v) and a multiple test correction was performed according to the method described by Benjamini and Hochberg in 1995 (54) to identify the differences associated with microbial clusters.
Clustering tendency validation.Validation of clustering tendency revealed that Lactobacillus-dominant and Bacteroides-dominant clusters are significantly different clusters (Hopkins statistic, 0.1898534; Hopkins value close to 0 [far less than 0.5] means that data are significantly clusterable). AU probability values (P values) by multiscale bootstrap resampling was 100% for Lactobacillus-dominant cluster and 98% AU for Bacteroides-dominant cluster, indicating that these two clusters were strongly supported by the taxonomic composition data and that they were not randomly formed clusters. Similarly, the observed lower Hopkins statistic (0.1367969) for function-based clusters also suggested that observed two function-based clusters are significantly different. The AU for the protein metabolism-dominant cluster was 98%, and it was 97% for the sulfur metabolism-dominant cluster, implying that these two significantly different clusters were supported by the functional composition data.
Identifying host transcriptome differences in response to the microbiome.A total of 3,306 mucosal immune-related genes (counts per million >1 in at least 50% of the samples) extracted from a whole ileal transcriptome generated using the same calves (16) (Data Set S3) were used to explore the differences in the mucosal immune systems of the calves belonging to the two taxonomy-based clusters. Calves with either Lactobacillus-dominant ileal microbiomes or Bacteroides-dominant ileal microbiomes were chosen to compare the expression of immunity-related genes using a nonparametric t test (Mann-Whitney-Wilcoxon) in R (3.3.1V). Multiple test correction was performed according to the method described in Benjamini and Hochberg (54). Differentially expressed genes were declared at fold change >2 and a corrected P value of <0.05. All transcripts with >2-fold change between the two groups were then subjected to a GO enrichment analysis (http://www.geneontology.org/page/go-enrichment-analysis) to identify the enriched mucosal immune responses in the calves belonging to the two taxonomic clusters.
ACKNOWLEDGMENTS
L.L.G. received funding support from Alberta Livestock and Meat Agency, Ltd., Edmonton, Canada (project 2011F129R), and an NSERC discovery grant. N.M. holds a Banting Postdoctoral Fellowship and received funding support from the Alberta Innovates Doctoral Graduate Student Scholarship. G.L. received funding support from the Alberta Innovates Doctoral Graduate Student Scholarship. P.J.G. holds a Tier I CRC funded by the Canadian Institute for Health research (CIHR).
We thank the staff at the Dairy Research Technology Center (University of Alberta), M. Zhou, X. Sun, Y. Chen, E. Hernandez-Sanabria, and S. Urrutia for their assistance with sample collection and X. Sun and Y. Chen for their assistance with metagenomics library preparation and real-time PCR.
FOOTNOTES
- Received 17 October 2018.
- Accepted 3 January 2019.
- Accepted manuscript posted online 18 January 2019.
Supplemental material for this article may be found at https://doi.org/10.1128/AEM.02534-18.
- Copyright © 2019 American Society for Microbiology.
