Table 1.
Dietary composition of total mixed ration.
Fig 1.
Relative abundance of (A) major and (B) minor phyla and (C) their differential abundances. (A) Relative abundance of major phyla defined as those phyla found at greater than 3% relative abundance and graphed as relative abundance ± SE. (B) Minor phyla defined as those found below 3% relative abundance present in sample types. (C) Phyla that are significantly differentially abundant compared with grab samples. Graphed as coefficients with a 95% confidence interval from the corncob model. Families with negative coefficients for a sample type are expected to have a lower relative abundance when compared to the grab samples while positive coefficients suggest a higher relative abundance in that sample type compared to grab samples.
Fig 2.
Differences in estimated alpha diversity among sample types.
(A) DivNet estimate of Shannon diversity plotted as mean with 95% confidence intervals and (B) mean breakaway estimate of species richness with 95% confidence intervals. The x-axis ticks represent samples from each cow on different days. Both the richness and evenness of fecal samples were lower than all other rumen sample types (P ≤ 0.001). Stomach tube and liquid strained samples had lower evenness than grab samples (P ≤ 0.001). Solid and stomach tube samples were estimated to have fewer species than grab samples (P = 0.02 and P ≤ 0.001, respectively).
Fig 3.
Beta diversity as weighted UniFrac distances between samples.
To faithfully reflect the variance in the coordinates, the height-to-width ratio was based on the ratio between the corresponding eigenvalues.
Fig 4.
Double principal coordiant analysis (DPCoA) of the Bray-Curtis distances among samples.
DPCoA is a phylogenetic ordination method and that provides a biplot representation of both (A) samples and (B) taxonomic categories. Note that while the biplots are a square shape to fit the page the CS1 explains roughly twice the variation of CS2 similar to what is seen in Fig 1. The 1st axis discrimates fecal from rumen samples while the 2nd axis separates liquid strained samples from other rumen sample types. Samples that have larger scores on CS1 have a subset of taxa from Bacteroidetes and Firmicutes that is different than rumen samples. Liquid strained samples have lower values on CS2 suggesting they are distinguished from other rumen sample types by taxa in the phylum Kiritimatiellaeota and family Prevotellaceae. The DPCoA predictes that fecal samples will have a lower abundance of both Lachnospiraceae and Prevotellaceae and greater abundance of Ruminococcaceae compared to samples from the rumen.
Fig 5.
Significant differences in the relative abundance of specific bacterial families.
Relative abundance of (A) Prevotellaceae (B) Ruminococcaceae and (C) Lachnospiraceae as modeled by corncob. Points are the estimated relative abundance and bars are a 95% prediction interval for each cow on different days of sampling.
Fig 6.
Families that were significantly differentially abundant across sample type compared with grab samples.
Graphed as coefficients with a 95% confidence interval calculated from the corncob model. Families with negative coefficients for a sample type are expected to have a lower relative abundance when compared to the grab samples while positive coefficients suggest a higher relative abundance in that sample type compared to grab samples.
Fig 7.
Significant differences in the relative abundance of specific bacterial families between fecal and grab samples.
Fecal samples had significantly higher relative abundance of (A) Peptostreptococcaceae, (B) Akkermansiaceae, (C) Bacteroidaceae, compared to grab samples. Also, there was significantly lower relative abundance of (D) Veillonellaceae, (E) Bacteroidales_BS11_gut_group and (F) Spirochaetaceae compared to grab samples. Points are the estimated relative abundance and bars are a 95% prediction interval for each cow on different days of sampling.