Dietary wheat and reduced methane yield are linked to rumen microbiome changes in dairy cows

Abstract

Fermentation of pasture grasses and grains in the rumen of dairy cows and other ruminants produces methane as a by-product, wasting energy and contributing to the atmospheric load of greenhouse gasses. Many feeding trials in farmed ruminants have tested the impact of dietary components on feed efficiency, productivity and methane yield (MeY). Such diets remodel the rumen microbiome, altering bacterial, archaeal, fungal and protozoan populations, with an altered fermentation outcome. In dairy cows, some dietary grains can reduce enteric methane production. This is especially true of wheat, in comparison to corn or barley. Using a feeding trial of cows fed rolled wheat, corn or barley grain, in combination with hay and canola, we identified wheat-associated changes in the ruminal microbiome. Ruminal methane production, pH and VFA concentration data together with 16S rRNA gene amplicon sequences were used to compare ruminal bacterial and archaeal populations across diets. Differential abundance analysis of clustered sequences (OTU) identified members of the bacterial families Lachnospiraceae, Acidaminococcaceae, Eubacteriaceae, Prevotellaceae, Selenomonadaceae, Anaerovoracaceae and Fibrobacteraceae having a strong preference for growth in wheat-fed cows. Within the methanogenic archaea, (at >99% 16S rRNA sequence identity) the growth of Methanobrevibacter millerae was favoured by the non-wheat diets, while Methanobrevibacter olleyae was unaffected. From the wheat-preferring bacteria, correlation analysis found OTU strongly linked to reduced MeY, reduced pH and raised propionic acid levels. OTU from the genera Shuttleworthia and Prevotella_7 and especially Selenomonadaceae had high anti-methane correlations. An OTU likely representing (100% sequence identity) the fumarate-reducing, hydrogen-utilising, rumen bacterium Mitsuokella jalaludinii, had an especially high negative correlation coefficient (-0.83) versus MeY and moderate correlation (-0.6) with rumen pH, strongly suggesting much of the MeY suppression is due to reduced hydrogen availablity. Other OTU, representing as yet unknown species from the Selenomonadaceae family and the genera Prevotella_7, Fibrobacter and Syntrophococcus also had high to moderate negative MeY correlations, but low correlation with pH. These latter likely represent bacterial species able to reduce MeY without causing greater ruminal acidity, making them excellent candidates, provided they can be isolated, for development as anti-methane probiotics.

Fig 1. Boxplots illustrating distributions of MeY, 16s amplicon abundance, Shannon diversity and Chao1 richness compared to diet across the herd of 32 cows.

Fig 2. Stacked barplot of the abundance in counts per million (cpm) of the 25 most abundant bacterial plus archaeal families from 32 rumen samples.

Also shown is the mean +/- sd CH4 yield (MeY) as gm CH4 per kg DMI for each group of cows and the diet of the group. # indicates cows treated with antibiotics before the feeding trial (Mastalone/wt9543, Yodimaspen/wt9534, CepravinLC/cn2319).

Fig 3. PCA.

Principal Components Analysis of rumen sequences using: A: abundance of families derived from unclustered DNA sequences aligned to the Silva 16s taxonomy database, B: abundance of OTU created using the Swarm2 clustering algorithm. For this analysis the cows were assigned to methane yield groups (g CH4/kg dry matter intake, MeY) as follows: low (MeY < 15), medium (15 < MeY < 25) or high (MeY > 25). Also see S1B Fig.

Fig 4. Differential abundance of OTU: Non-wheat vs wheat + MeY.

Volcano scatter plot of the result of the Spearman Rank Correlation vs the edgeR differential abundance analysis contrasting wheat diet rumen OTU vs non-wheat. Y axis = methane yield (MeY) correlation, X axis = log2 fold change (logFC). Shown are the bacterial OTU where the P-value for the MeY correlation was less than 0.001 and mean abundance at least 128cpm. All archaeal OTU are shown. Grey discs represent OTU where the abundance change between diets was not significant (false discovery rate > 0.02). For ease of visualisation genus labels are shown only for OTU near the extremities of the plot or to highlight certain taxa such as the archaea.

Fig 5. Correlation of OTU vs MeY, pH and ruminal acids.

Correlogram illustrating the Spearman Rank Correlation analysis of the abundance of 83 OTU versus each other and MeY (CH4 yield), ruminal pH, D-lactate, L-lactate, acetic acid, propionic acid, iso-butyric acid, butyric acid, iso-valeric acid, valeric acid, hexanoic acid and heptanoic acid. The 83 OTU are those with an abundance of at least 1000cpm in at least 5 cows, a significant correlation with MeY (P<0.01) and a logFC (log2 fold change) magnitude of at least 2.5 when comparing a wheat to a non-wheat diet. Also included are the 2 most abundant methanogenic archaea. Correlation coefficients are rendered as coloured circles (Red negative or Blue positive) if P<0.001. Colour intensity reflects degree of correlation.