INTEGRATIVE IDENTIFICATION OF ALTERNATIVE BIOMARKERS FOR NON-INVASIVE STRESS AND WELFARE ASSESSMENT IN TELEOST

Fish welfare assessment has traditionally relied on glucocorticoids, yet their context dependent interpretation and limited specificity highlight the need for complementary indicators. This work investigated dehydroepiandrosterone (DHEA), an emerging biomarker with potential modulatory effects against glucocorticoids, alongside its ratio with cortisol in a freshwater species under acute stress and in a marine species under chronic stress exposure. In rainbow trout (Oncorhynchus mykiss), acute stress triggered parallel cortisol increases in blood and fin tissue, supporting fins as a minimally invasive alternative matrix for hormone evaluation. DHEA showed tissue-specific variability, particularly in fin and muscle, and was influenced by sex and maturity, suggesting a role in reproductive processes. In European sea bass (Dicentrarchus labrax), chronic boat noise did not provoke strong endocrine responses, likely reflecting habituation or stimulus insufficiency. Still, sex-related patterns of DHEA were evident, with higher concentrations in males, especially in scales and gonads, reinforcing the gonads as a possible site of synthesis. The correlation between the hormone levels in scales and gonads highlights scales as a promising, minimally invasive tool for assessing reproductive status. In parallel, artificial intelligence–based behavioral tracking using DeepLabCut (DLC) was applied to both multi-animal and single-animal experiments. In complex group settings, classical metrics such as speed, distance, and inter-individual spacing have proven to be valuable behavioral biomarkers. Complementary measures of cortisol, glucose, and lactate emphasized the benefit of integrating behavioral and physiological data. In controlled single-animal setups, DLC enabled high-resolution tracking, allowing exploration of novel metrics such as harmonic swimming patterns and millimeter-scale proximity analysis, broadening the range of behaviors accessible for welfare monitoring. These findings highlight the potential of combining minimally invasive physiological indicators with AI-driven behavioral analysis. Such integrative frameworks can provide precise, scalable, and context-sensitive tools to advance fish welfare assessment in both research and aquaculture.