Characterisation of brain genetic markers for fish welfare and stunning assessment: an integrated in silico and methodological approach

Fish welfare assessment in aquaculture remains constrained by the lack of objective, molecularly grounded tools applicable across commercially relevant species. Brain gene expression offers a mechanistically informative complement to conventional welfare indicators, yet no systematic characterisation of candidate markers has been performed across the key aquaculture teleosts. This thesis presents an integrated in silico and methodological framework for the identification and characterisation of brain genetic markers for fish welfare and pre-slaughter stress assessment in Dicentrarchus labrax, Sparus aurata, and Oncorhynchus mykiss, using Danio rerio and Cyprinus carpio as reference species. Whereas the computational characterisation spans this five-species panel, the experimental component reported here is deliberately narrower: a single-species, two-treatment pilot in which brain and plasma serotonin were quantified in Dicentrarchus labrax following ice immersion or percussive stunning, providing a first experimental realisation of the framework rather than a full validation across species and stressors. In addition, the regional brain expression of two immediate-early genes (c-fos and egr1) was profiled by RT-qPCR across four brain regions, providing a region-resolved transcript-level complement to the serotonin measurements; neither gene showed a statistically robust regional difference. In this context, "stunning assessment" denotes retrospective, tissue-level molecular corroboration of the pre- and peri-slaughter stress state a humane stun is intended to prevent complementary to real-time electrophysiological (EEG) and behavioural indicators of consciousness, not a replacement for them. The intended application is therefore batch-level: post-slaughter sampling to audit and compare the efficacy of stunning protocols across a production line, rather than per-animal decision-making at the point of kill. Of 25 initial candidate genes, 24 were tractable across the panel (arc was excluded owing to a lack of retrievable orthologous sequence). The 24 genes span seven neurobiological stress-response pathways (immediate early genes, HPI axis, serotonergic, catecholaminergic, HSP/ER stress, apoptosis, and neuroinflammation) and were subjected to nine sequential analyses: multiple sequence alignment, nucleotide conservation scoring, maximum likelihood phylogenetic reconstruction, protein-level conservation analysis, evolutionary selection-pressure estimation (dN/dS), species-specific RT-qPCR primer design, in silico primer-specificity verification, RNA-seq expression-evidence mining, and promoter regulatory-element scanning. Species-level database gaps are reported per gene in Chapter 4. After excluding pairwise comparisons with saturated synonymous sites (the majority, given the deep teleost divergences), 23 of the 24 genes showed dN/dS below 1 (range 0.031–0.462), consistent with purifying selection across the panel; nr3c1b was excluded, retaining no unsaturated pair. Protein conservation was highest for the molecular chaperones hspa5 (94.7%) and hsp90aa1 (92.4%) and lowest for neuroinflammatory cytokines. Species-specific RT-qPCR primer pairs were designed for the three aquaculture species using Primer3, meeting standard RT-qPCR design criteria (amplicon 100–250 bp, Tm 60 ± 2 °C, GC 40–60 %); 70 pairs were retained after excluding two non-viable nr3c1b combinations (no retrievable sequence in D. labrax, and a duplicated/non-specific sequence in O. mykiss). In silico specificity verification (Primer-BLAST, including exon-junction spanning), amplification efficiency, and reference-gene stability remain to be established experimentally and are specified as the first validation step. Across the 70 retained pairs, predicted melting temperatures clustered tightly at approximately 60 ± 0.5 °C, supporting compatibility with a common thermal-cycling protocol.