Different microbial genera drive methane emissions in beef cattle fed with two extreme diets

Abstract

The ratio of forage to concentrate in cattle feeding has a major influence on the composition of the microbiota in the rumen and on the mass of methane produced. Using methane measurements and microbiota data from 26 cattle we aimed to investigate the relationships between microbial relative abundances and methane emissions, and identify potential biomarkers, in animals fed two extreme diets - a poor quality fresh cut grass diet (GRASS) or a high concentrate total mixed ration (TMR). Direct comparisons of the effects of such extreme diets on the composition of rumen microbiota have rarely been studied. Data were analyzed considering their multivariate and compositional nature. Diet had a relevant effect on methane yield of +10.6 g of methane/kg of dry matter intake for GRASS with respect to TMR, and on the centered log-ratio transformed abundance of 22 microbial genera. When predicting methane yield based on the abundance of 28 and 25 selected microbial genera in GRASS and TMR, respectively, we achieved cross-validation prediction accuracies of 66.5 ± 9% and 85 ± 8%. Only the abundance of Fibrobacter had a consistent negative association with methane yield in both diets, whereas most microbial genera were associated with methane yield in only one of the two diets. This study highlights the stark contrast in the microbiota controlling methane yield between animals fed a high concentrate diet, such as that found on intensive finishing units, and a low-quality grass forage that is often found in extensive grazing systems. This contrast must be taken into consideration when developing strategies to reduce methane emissions by manipulation of the rumen microbial composition.

Keywords: beef cattle; concentrate-based diets; enteric methane emissions; microbiota by diet interaction; zero-grazed grass diet.

FIGURE 1

(A) Sample plot from a Discriminant by projection of latent structures analysis (DA-PLS) fitted to discriminate amongst the fresh cut grass (Grass) and high concentrate (TMR) fed animals based on 28 clr-transformed microbial genera abundances. The model was built with a single component and presented a miss-classification rate of 0%. For visualization purposes, sample plot is based on two components. (B) Microbial genera identified by DA-PLS which presented differential abundances between grass and TMR (probability of the difference of being different from 0 ≥ 0.95). Differences are expressed in units of clr-transformed abundances. Full details of the PLS analysis and linear models can be found in Supplementary Table 1.

FIGURE 2

Data distribution of methane emissions [expressed as (A) methane production (g/day) or (B) methane yield (g/kg DMI)] and the natural log-ratio between archaea and bacteria abundances within animals offered two contrasting diets, high concentrate (TMR) or fresh cut grass (Grass). Methane and natural log-ratio between archaea and bacteria abundances was pre-corrected by fixed effects of breed, and body weight as a covariate.

FIGURE 3

Results from linear projection of latent structures regression (PLS) models aiming to predict methane yield (g/kg DMI) based on 28 or 25 clr-transformed microbial genera in fresh cut grass (Grass) or high concentrate (TMR) fed animals. Full details of the PLS models can be found in Supplementary Table 2. (A) Sample plots (both PLS models were built with one single component, but two components were fitted only for visualization purposes). (B) After a threefold cross-validation procedure repeated 20 times, methane yield was predicted with 85 ± 8% (Grass) and 66.5 ± 9% prediction accuracy.

FIGURE 4

Visualization of linear associations between clr-transformed abundances of microbial genera identified with PLS analysis and methane yield in fresh cut grass (Grass, in blue) or high concentrate (TMR, in orange) fed animals. Full details of the analysis can be found in Supplementary Table 2. (A) Six microbial genera were selected in the PLS model in both diet groups, although only Fibrobacter was associated with methane yield in the same direction. (B,C) Microbial genera presenting a linear regression coefficient with a probability of being different from 0 ≤ 0.95 on methane yield only under TMR (B) or grass (C) diets.