Deliverable D5.1 Rules to use information from the gut microbiota to improve feed efficiency

Objectives

As many studies have shown, the gut microbiota plays an important role in the metabolism, nutrition, and immune system of monogastric animals. This suggests that microbial communities inhabiting the gastrointestinal tract could also influence feed efficiency. The objective of this deliverable was to evaluate the genetic effects influencing the composition of the gut microbiota as affect by a wide range of environmental conditions and physiological factors (i.e., maternal transmission, feeding regime, farm management, presence of antibiotics in the feed, sex, heat, and humidity). The aim was to propose criteria and methodologies to use variability in gut microbiota as a heritable trait affecting feed efficiency.

Rationale

Results obtained in this deliverable rely on the analysis of five experimental designs conducted in rabbits and pigs and performed by INRA and IRTA:

•  A factorial design conducted at INRA allowed disentangling the maternal transmission of gut microbiota (neonatal environment) from the direct genetic effect of the animal in a cross-fostering trial between and within rabbit lines selected (INRA G10) or not (INRA G0) for feed efficiency (300 rabbits per line in three batches).
• The interaction between feeding regime (i.e., ad libitum or restricted) and the genetic control of the gut microbiota composition was studied in relation to post-weaning growth and feed efficiency in a rabbit line (451 Caldes kits) selected by IRTA for high average daily gain after weaning.
• The relationship between microbiota composition and apparent digestibility coefficients was evaluated in 60 castrated male pigs from Duroc, Large White, and Pietrain breeds raised in the experimental facility of INRA. Pigs were fed alternatively a low-fiber (LF) and a high-fiber (HF) diet during four successive 3-week periods from 11 to 23 weeks of age.
• The impact of climate and heat challenge on microbiota composition was evaluated in female and castrated male pigs from a backcross design between Large White and Creole breeds. A total of 522 and 531 genetically-related pigs were raised either under a temperate or tropical climate, respectively, from 11 to 23 weeks of age. In the temperate climate, the pigs were exposed to an acute heat stress (HS) of 29°C during 3 weeks.
• Sex and diet effects and association between feed efficiency and faecal microbial composition were investigated in 160 commercial three-breed crossbred pigs from Topigs Norsvin.

All animals were genotyped and their faeces (pigs) or caecal content (rabbits) were collected.

The microbial composition, diversity and richness of gut microbiota was characterized by means of Illumina sequencing of 16S rRNA gene amplicons (V4-V5 hypervariable regions) in a MiSeq platform. Raw paired-ended sequences were processed with QIIME/QIIME2 software by discarding the low quality and the chimeric sequences. Filtered sequences were assembled into contigs and then clustered into OTUs (Operational Taxonomic Units; contigs sharing a 97% of similarity)/ASVs (Amplicon Sequence Variants; contigs sharing a 99% of similarity). The OTU/ASV table was filtered at sample (discarding those with less than 5,000 filtered contigs) and OTU/ASV (discarding those with less than 0.01% counts across all samples) levels. This table was normalized using the Cumulative Sum Scaling (CSS) normalization yielding the normalized abundances of 931/792 OTUs/ASVs. The taxonomic affiliation was obtained using the RDP train set 15 with the utax algorithm (pigs and INRA rabbits) or the Greengenes reference database gg_13_5_otus/99 (IRTA rabbits).

Differences in caecal microbial composition, diversity, and richness were assessed from a univariate perspective (i.e., bootstrap analysis of variance fitting a model with a factor combining the farm where the animal was raised, the batch, the feeding regime, and the presence or absence of antibiotics in rabbit feed) and from a multivariate perspective using: (1) Principal Variance Component Analysis, (2) sparse Partial Least Squares - Discriminant Analysis (sPLS-DA) and (3) machine learning algorithms (Support Vector Machine) selecting a certain number of OTUs/ASVs that allow the best discrimination of samples according to a given factor.

Kruskal-Wallis tests, glm methodologies, or generalized Friedman rank sum tests with replicated blocked (package muStatsv1.7.0) were used to point out OTUs with significant abundancy differences between conditions (i.e., sex, temperature, diet, maternal effect, and line effect).

Genetic determinism of the caecal microbiota of rabbits (IRTA) has been assessed by fitting two mixed models (i.e., a genetic model and another model without the additive genetic effect) for 14 representative traits of rabbit caecal microbiota (i.e., the relative abundances of 8 phyla, 4 a-diversity indexes and the first 2 principal components computed from the relative abundance phyla table). To test the existence of a genetic determinism of caecal microbiota, different model choice criteria were used: (1) Deviance Information Criterion (DIC) and (2) Bayes Factor (BF). Variance components were computed (gibbs2f90) for the 14 traits. In INRA rabbits, the variance components (i.e., linear animal mixed model) and QTL analyses (i.e., fixed effect test of each of the 161 033 SNP) were performed using the GEMMA software.

Prediction of growth and feed efficiency traits from microbiota data were evaluated using Partial Least Square Regression method (PLSR).

The relationships between OTUs and the zootechnical traits were characterized by the maximal information coefficient (MIC) index and the background noise was estimated by the maximal value of random permutations.

Mixed linear models were used to estimate jointly host genetic effects and microbiota effect on growth and feed efficiency traits.