A first characterization of the microbiota-resilience link in swine

Abstract

Background: The gut microbiome plays a crucial role in understanding complex biological mechanisms, including host resilience to stressors. Investigating the microbiota-resilience link in animals and plants holds relevance in addressing challenges like adaptation of agricultural species to a warming environment. This study aims to characterize the microbiota-resilience connection in swine. As resilience is not directly observable, we estimated it using four distinct indicators based on daily feed consumption variability, assuming animals with greater intake variation may face challenges in maintaining stable physiological status. These indicators were analyzed both as linear and categorical variables. In our first set of analyses, we explored the microbiota-resilience link using PERMANOVA, α-diversity analysis, and discriminant analysis. Additionally, we quantified the ratio of estimated microbiota variance to total phenotypic variance (microbiability). Finally, we conducted a Partial Least Squares-Discriminant Analysis (PLS-DA) to assess the classification performance of the microbiota with indicators expressed in classes.

Results: This study offers four key insights. Firstly, among all indicators, two effectively captured resilience. Secondly, our analyses revealed robust relationship between microbial composition and resilience in terms of both composition and richness. We found decreased α-diversity in less-resilient animals, while specific amplicon sequence variants (ASVs) and KEGG pathways associated with inflammatory responses were negatively linked to resilience. Thirdly, considering resilience indicators in classes, we observed significant differences in microbial composition primarily in animals with lower resilience. Lastly, our study indicates that gut microbial composition can serve as a reliable biomarker for distinguishing individuals with lower resilience.

Conclusion: Our comprehensive analyses have highlighted the host-microbiota and resilience connection, contributing valuable insights to the existing scientific knowledge. The practical implications of PLS-DA and microbiability results are noteworthy. PLS-DA suggests that host-microbiota interactions could be utilized as biomarkers for monitoring resilience. Furthermore, the microbiability findings show that leveraging host-microbiota insights may improve the identification of resilient animals, supporting their adaptive capacity in response to changing environmental conditions. These practical implications offer promising avenues for enhancing animal well-being and adaptation strategies in the context of environmental challenges faced by livestock populations. Video Abstract.

Keywords: Animal-health; Gut microbiota; Pigs; Resilience.

Fig. 1

Process for obtaining the four different resilience indicators. Part A represents animals with the least variability, while Part B represents animals with higher variability. Each part includes the following representations: A1, B1 the predicted values (in red) and the observed values (in green); A2, B2 the periods of consecutive negative errors highlighted in red; A3, B3 local minima depicted in red and local maxima depicted in yellow

Fig. 2

Density plot of distribution of the four different phenotypes divided by breed: Duroc (DR), Landrace (LR), Large White (LW). The dotted line represents the mean value for each breed of each phenotype: lag of 1 day of residual (Lag1), natural logarithm of residual variance (LnVar) area under the curve for periods with largest consecutive negative errors (MaxArea), sum of residual’s local minima (SumMin)

Fig. 3

Plot representing the Spearman correlation between the four resilience indicators and other productive phenotypes. Resilience indicator were lag of 1 day of residual (Lag1), natural logarithm of residual variance (LnVar) area under the curve for periods with consecutive negative errors (MaxArea), and sum of residual’s local minima (SumMin). Productive phenotypes were average feed daily consumption (FCD) and average body weight (Weight) estimated in the period of which the four resilience indicators were estimated (99–140 days). Phenotype collected before slaughter as percentage of muscle (Muscle), and backfat and intramuscular fat (IMF) and body weight (Final_Weight). Correlation was calculated for all animals (ALL) or for animals of same breed Duroc (DR), Landrace (LR), and Large White (LW)

Fig. 4

Scatter plot representing linear regression of the four resilience indicators (x-axis) on the two α-diversity measures (y-axis): Shannon (A) and inverse Simpson (B). The four resilience indicators were Lag of 1 day of residual (Lag1), natural logarithm of residual variance (LnVar) area under the curve for periods with largest consecutive negative errors (MaxArea), sum of residual’s local minima (SumMin). Coefficient of regression (R) and P-values (p) of each regression were reported in each plot

Fig. 5

Barplot illustrating the absolute value of the log fold change (LFC) abundances for significantly abundant amplicon sequence variants (ASVs), for the indicators of natural logarithm of residual variance (LnVar) and area under the curve for periods with largest consecutive negative errors (MaxArea). ASVs with positive LFC are represented by yellow bars, while those with negative LFC are represented by light-blue bars. ASVs are grouped based on the genera to which they belong, and when multiple ASVs belong to the same genus, the mean of all ASVs within that genus is reported. The number of ASVs per genus class is indicated near each bar

Fig. 6

Barplot illustrating the absolute value of the log fold change (LFC) abundances for the ten most significantly abundant KEGG pathways, for the indicators of natural logarithm of residual variance (LnVar) and area under the curve for periods with consecutive negative errors (MaxArea). KEGG pathways, with positive LFC, are represented by yellow bars, while those with negative LFC are represented by light-blue bars

Fig. 7

Shannon alpha diversity was assessed for the four resilience indicator were lag of 1 day of residual (A) natural logarithm of residual variance (B) area under the curve for periods with largest consecutive negative errors (C) and sum of residual’s local minima (D). The x-axis represents the resilience classes: lower (L), medium (M, as the control group), and higher (H). Above the plot, the P-values of the Kolmogorov-Smirnov test between the classes are reported above the boxplot. The shape of the point represents different breed within each class

Fig. 8

Heatmap illustrating the absolute value of the log fold change (LFC) abundances for the significantly abundant amplicon sequence variants (ASVs), with the indicators of natural logarithm of residual variance (LnVar) and area under the curve for periods with consecutive negative errors (MaxArea). The ASVs are grouped based on the genera to which they belong. The x-axis represents the LFC when comparing the resilient animal class with control groups (lfc_L), and the higher resilient animals class with control groups (lfc_H)

Fig. 9

A Barplot representing the proportion of the phenotypic variance of the four phenotypes explained by the nested effect of sire (S) in red, microbiome (M) in yellow, pen (PEN) in light blue, and the residual error in grey (E). The four indicators are lag of one day of residual (Lag1), natural logarithm of residual variance (LnVar), area under the curve for periods with the largest consecutive negative errors (MaxArea), and sum of residual’s local minima (SumMin). B Violin plot of the posterior density distributions for the microbiability parameter (as the proportion of variance explained by the microbial effect), corresponding to the four resilience indicators mentioned above. The red dot represents the median of each posterior density distribution. The red line represents the HPD₉₅ interval, which provides a measure of the uncertainty associated with the microbiability estimate

Fig. 10

A A two-dimensional Partial Least Square-Discriminant Analysis (PLS-DA) score plot was constructed using three classes of resilience (lower (L) in yellow, medium (M) in light blue as the control group, and higher (H) in grey). The plot represents the distribution of the samples based on the first two components in the model. Each point’s shape changes according to the breed to which the animal belongs. PLS-DA was performed for the four indicators that are lag of 1 day of residual (Lag1), natural logarithm of residual variance (LnVar), area under the curve for periods with the largest consecutive negative errors (MaxArea), and sum of residual’s local minima (SumMin). B Receiver operating characteristic (ROC) analysis was performed to discriminate between the three classes of resilience mentioned above (L,M,H) for the four indicator of resilience mentioned above (Lag1, LnVar, MaxArea, and SumMin). C Confusion matrix depicting the performance of Lag1, LnVar, MaxArea, and SumMin in predicting the resilience classes (L,M,H). The x-axis represents the categories of true values, while the y-axis represents the categories of predicted values