Genetic relationships between efficiency traits and gut microbiota traits in growing pigs fed a conventional or a high fiber diet

In pigs, the gut microbiota composition plays a major role in the process of digestion, but is influenced by many external factors, especially diet. To be used in breeding applications, genotype by diet interactions on microbiota composition have to be quantified, as well as their impact on genetic covariances with feed efficiency (FE) and digestive efficiency (DE) traits. This study aimed at determining the impact of an alternative diet on variance components of microbiota traits (genera and alpha diversity indices), and estimating genetic correlations between microbiota and efficiency traits for pigs fed a conventional (CO) or a high fiber (HF) diet. Fecal microbes of 812 full-siblings fed a CO diet and 752 pigs fed the HF diet were characterized at 16 weeks of age by sequencing the V3-V4 region of the 16S rRNA gene. A total of 231 genera were identified. Digestibility coefficients of nitrogen, organic matter and energy were predicted analyzing the same fecal samples with near infrared spectrometry. Daily feed intake, feed conversion ratio, residual feed intake and average daily gain (ADG) were also recorded. The 71 genera with less than 20% of zeros were retained for genetic analyses. Heritability (h2) of microbiota traits were similar between diets (from null to 0.38 ± 0.12 in the CO diet and to 0.39 ± 0.12 in the HF diet). Only three out of the 24 genera and two alpha diversity indices with significant h2 in both diets had genetic correlations across diets significantly different from 0.99 (P < 0.05), indicating limited genetic by diet interactions for these traits. When both diets were analyzed jointly, 59 genera had h2 significantly different from zero. Based on the genetic correlations between these genera and ADG, FE and DE traits, three groups of genera could be identified. A group of 29 genera was favorably correlated with DE and FE traits, 14 genera were unfavorably correlated with DE traits, and the last group of 16 genera had correlations close to zero with production traits. However, genera favorably correlated with DE and FE traits were unfavorably correlated with ADG, and vice versa. Alpha diversity indices had correlation patterns similar to the first group. In the end, genetic by diet interactions on gut microbiota composition of growing pigs were limited in this study. Based on this study, microbiota-based traits could be used as proxies to improve FE and DE in growing pigs.


62
In recent years, gut microbiota has been highlighted as a critical partner of feed and digestive efficiency 63 of pigs, as reviewed by Gardiner et al. (2020) and Patil et al. (2020). Indeed, microbiota composition differs for influence the genetic architecture of microbiota composition traits for pigs reared in a temperate or a tropical 85 environment (Gilbert et al., 2020). However, in this latter study the effect of climate, feed and housing could 86 not be disentangled. To be used in practice, it is critical to estimate the impact of genotype-by-environment 87 5 (GxE) interactions on the genetic relationships between genera abundances on one hand, and feed and digestive 88 efficiency traits on the other hand. 89 The objective of this study was to evaluate the genetic relationships between gut microbiota composition 90 and efficiency traits with a focus on the effect of diet on genetic parameters. A large dataset of closely related 91 pigs tested in the same farm with two diets with contrasted dietary fiber contents that significantly affected the  During the growing-finishing phase, the two sets of pigs were fed one of the two-phases dietary 120 sequences. A growing type of diet was first distributed, then a five-days transition was organized at 16 weeks 121 of age and a finishing diet was provided until the end of the test (slaughter body weight of 115kg  129 For each animal, average daily gain (ADG), daily feed intake (DFI), feed conversion ratio (FCR), and 130 residual feed intake (RFI) were computed between 35 and 115 kg, namely. The ADG was computed as the 131 ratio between body weight gain and number of days on test. The FCR was calculated as the ratio between DFI 132 and ADG, and was expressed in kg/kg. The RFI was determined using a multiple linear regression of DFI on 133 ADG, lean meat percentage, carcass yield and average metabolic body weight considering data from the two 134 diets jointly in the linear regression, as described in Déru et al. (2020). A spot collection of fecal samples was 135 carried out at 16 weeks of age just before feed transition, for digestive efficiency determination and microbiota 136 7 composition analyses. For each pig, feces were collected in a piping bag and manually homogenized. To 137 determine digestibility coefficients (DC), about 50 g of feces were stored in a plastic container at -20°C until 138 further analyses. Samples were freeze-dried and ground with a grinder (Grindomix GM200, Retsch). Individual 139 DC of energy, nitrogen and organic matter were then predicted based on near infrared spectrometry (NIRS) 140 analyses of these samples. Both the methodology to predict DC traits and procedures to validate predictions are Dutscher, France) were filled in with homogenized feces and were immediately frozen in liquid nitrogen. They 143 were then stored at -80°C until analysis.

144
Pigs that experienced health problems during the test period or had incomplete feed intake data, in equal 145 proportion in the two diets, were discarded from the analysis. In total, 1,663 pigs had FCR, ADG and DFI 146 performances, 1,595 had RFI measurements and 1,242 pigs had NIRS-based predictions of DC traits.

147
Microbiota DNA preparation and sequencing 148 A total of 1,564 fecal samples were used for ribosomal 16S DNA gene sequencing and analysis, with

161
Run quality was internally checked using PhiX (a library used as a control for Illumina sequencing runs), and 162 8 each pair-end sequence was assigned to its sample using the bcl2fastq Illumina software. the Chao1 richness estimator were calculated using the following formulas:

189
Genera abundances were normalized using a decimal logarithmic transformation after adding one of 190 each count. For genetic analyses, only genera with less than 20% of zeros in the whole dataset were retained in 191 order to limit deviations from the linear mixed model assumptions, which represented 71 out of the 231 genera.

192
The elementary statistics for these genera are given in Supplementary Table 2   of a given genera were defined as the sum of the heritability estimate and the corresponding empirical threshold.

223
Heritabilities within non-overlapping CI were considered as significantly different at 90% or 95%, as for a Z-  Finally, bivariate genetic analyses were carried out to estimate genetic correlations between heritable 256 microbiota traits, using the same linear mixed models and the same methodology as previously presented.

258
Variance components of microbiota traits within diet 259 Heritability estimated for microbiota traits within each diet are presented along with their 95% confidence 260 interval in Supplementary Table S3. Based on the bootstrap approach, all heritabilities larger than 0.123 in the 261 CO diet and 0.136 in the HF diet were declared significantly different from zero (P < 0.05). Heritability of 262 genera abundances were in the same range in the two diets, i.e. from null to 0.38 ± 0.12 in the CO diet and to 263 12 0.39 ± 0.12 in the HF diet, respectively. For each genus, heritabilities estimated within each diet are represented 264 in Figure 1. Out of the 71 genera, 24 genera had significant heritability estimates in both diets, 23 genera had 265 significant heritabilities only in one diet, whereas 24 genera were not heritable in any of the diets. None of the 266 71 genera had heritability estimates significantly different between diets considering 95% confidence intervals.

267
Reducing the confidence intervals to 90% (with 90% empirical thresholds equal 0.09 for the CO diet and 0.10 268 for the HF diet), heritability estimates of four genera were suggested as different between diets: Campilobacter, 269 Dialister, Fusicateribacter, and Ruminococaceae_UGC002, the first one having larger h² in the CO diet,   286 From the previous results, we considered that genetic by diet interactions were limited and microbiota 287 traits could be considered as the same trait in the two diets. Then, heritabilities were estimated for the relative    higher digestive efficiency were moderately to strongly correlated with diversity indices. On the contrary, the 331 genera associated with higher growth rate and poorer digestive efficiency were unfavourably genetically 332 correlated with gut microbiota diversity indices. Genera that were weakly correlated both with growth rate and 333 digestive efficiency were also weakly to moderately correlated with diversity indices.  To improve feed and digestive efficiency traits that are costly to measure, the present study shows that 467 an alternative to direct selection could be to select genera abundances or an alpha diversity index, which have  In conclusion, gut microbiota analysis is a promising approach to improve the feed and digestive 491 efficiency of growing pigs, which could be applied to pigs fed a range of diets from CO to alternative diets

669
The 5% threshold was obtained by bootstrapping genera abundances to individuals and determining heritability for each reassignment,   Figure 5. Genetic correlations 1 between sixty heritable genera considering data of growing pigs fed the conventional or 692 high fiber diets. Vertical bars show limits between groups obtained from the hierarchical analysis of genetic correlations 693 between these genera and production traits. 694 695 1 Correlations were represented in colors when they were significantly different from zero considering alpha=32%, and were left blank 696 otherwise. Genetic correlations that could not be estimated due to convergence problems of the algorithm are marked with an X. 697 698 32 Figure 6. Genetic correlations 1 between gut microbiota diversity indices and genera abundances considering data of 699 growing pigs fed the conventional or high fiber diets. Vertical bars show limits between groups obtained from the 700 hierarchical analysis of genetic correlations between these genera and production traits.