Differential intestinal microbiome response to heat stress in two rabbit maternal lines: a comparative analysis using Random Forest, BayesC, and PLS-DA.

Heat stress affects livestock productivity and health, particularly in rabbits, due to their physiological vulnerabilities. This study explores the relationship between environmental conditions, genetic lines backgrounds, and soft fecal microbiota. A 2 × 2 factorial design was used, involving 2 maternal rabbit lines: A (standard longevity line) and LP (high longevity line), exposed to heat stress and thermal comfort. Samples were analyzed with multiple models to assess the impact of heat stress on microbiota by comparing microbial diversity and evaluating the classification performance of Random Forest, Partial Least Squares Discriminant Analysis (PLS-DA), and Bayesian Regression (BayesC). Heat stress influenced microbial diversity in both lines, increasing alpha diversity and driving significant beta-diversity shifts (2.3% variance, P < 0.001). This could be due to intestinal barrier disruption, which facilitate pathogen proliferation. The high longevity line LP exhibited higher richness under thermal comfort, whereas heat stress equalized these differences between lines, possibly due to increased pathogen proliferation in the low longevity line A. These differences in response to heat stress may be influenced by the crosstalk between microbiota and host genetics, shaping distinct adaptive mechanisms in each line. Prediction accuracy and key selected variables distinguishing between lines A and LP varied across thermal conditions, with the area under the curve exceeding 0.92 under heat stress and 0.87 in thermal comfort. This reflects different microbiome regulations between the 2 lines under heat stress. Potential stress-associated taxa such as Erysipelatoclostridium and Monoglobus were more abundant in the low longevity line A. These results highlight LP's higher longevity and expected resilience, while line A's susceptibility is reflected in a higher abundance of heat stress-associated taxa in the latter. This underscores soft fecal microbiota as a potential biomarker for heat stress resilience and emphasizes the role of host-microbiota interactions in mediating genetic-environmental responses. Additionally, this study highlights the value of combining modeling approaches, which enhance accuracy and reveal key taxa driving heat stress responses. Among the models tested, PLS-DA achieved the highest accuracy, while Random Forest identified a smaller yet biologically relevant subset of taxa, providing valuable phylogenetic and taxonomic insights.