Genomic control of behaviour is complex

The genomic architecture of behaviour is typically complex, encompassing the action of genetic variation in DNA sequence, through to the factors that regulate gene expression. Furthermore, behaviours are typically influenced by many interactions between gene networks and internal and external environmental stimuli (Mackay 2009). The study of behavioural genetics has made a transition towards disseminating the genomic architecture involved in behaviour since it is known that genes show interactions, such as regulatory or modulatory effects (Boake et al. 2002; Reif & Lesch 2003; Mackay 2009). The results from Chapters 2-6 demonstrated many genes were involved in the control of behaviour (Chapter 4), which are potentially distributed across the genome (Chapter 3), indicating complex genomic control. Moreover, there was evidence of pleiotropic effects of particular genes. In addition, potential epistatic effects may explain small changes in genes involved in the hypothalamic-pituitary-interrenal (HPI) axis (Chapters 5 & 6). Differences in genetic diversity among stress responsiveness, aggression and boldness were also found (Chapter 2) suggesting that trout may have a genetic propensity to win fights but not to be bold. Overall, these results may contribute to the knowledge of the complex genomic association with individual behaviour as well as genomic effects involved with behavioural syndromes.

Pleiotropy, where a gene may have effects in many phenotypes, may be implicated when the expression of particular genes in one behaviour are the same genes known to be associated in other behaviours or phenotypes (Anholt & Mackay 2004; Barendse et al. 2009; Edwards et al. 2009b). For example, in Chapter 4, neuromedin S was identified as a novel candidate gene associated with aggression, whereas it was previously implicated in feeding behaviour (Ida et al. 2005; Miyazato et al. 2008). Similarly, proopiomelanocortin (POMC) was upregulated in aggressive trout. POMC is a precursor to adrenocorticotrophic hormone (ACTH), which leads to the release of cortisol, the principal hormone secreted during stress in many animals including trout (Sloman et al. 2001). As such, POMC is a crucial part of the HPI axis, which is involved in the control of stress responsiveness and aggression. Moreover, pleiotropy may be evident in Chapters 5 and 6, where similar genes were studied in the brains of fish that had experienced aggression and stress. Whilst there were some weak changes in HPI genes, some of these changes were similar between aggressive fish and those that had received an acute stressor. The vasotocin receptor V1a showed similar expression patterns in aggressive fish as in stressed fish. Not only does this support pleiotropy due to similarities in gene expression in the brain of these two behaviours, but

also due to the role of vasotocin and its receptors in learning and memory (de Kloet 2010). However, other genes, mineralocorticoid (MR) and ependymin (EPD) did not show similar patterns between stress and aggression tests, which indicate that only particular genes potentially exhibit pleiotropic effects. These Chapters highlight, that while the study of individual candidate genes is important, the study of networks of interacting genes may be more informative when studying complex behavioural traits.

Using transcriptomes may be more informative than single candidate genes to determine the number of genes involved in a complex trait. In Chapter 4 many genes were correlated with aggressiveness. Moreover, transcriptomic data provides initial networks from which gene interactions may be studied and enable a move towards exploring networks of interacting genes, which may have a small effect upon phenotype and that combine to produce a specific behavioural outcome (Anholt & Mackay 2004). Using this transcriptome data, it may be possible to map epistatic effects of genes, where the expression of certain genes or gene products regulates the expression of other genes (Reif & Lesch 2003). In Chapter 5, some genes may be under epistatic control of other factors. For example, arginine vasotocin (AVT), the agonist of V1a suppressed aggression in rainbow trout, where this may be regulated by serotonin (5-HT), which suppresses AVT (Backström & Winberg 2009). Thus a lack of gene expression changes, in genes expected to be involved in stress and aggression, may be due to the control of genes not studied in these chapters.

Of course, behaviour is not solely under the control of genes and it is well-known that genes are both inherited and environmentally responsive for most phenotypes, but this is a relatively new concept in behavioural ecology (McGuffin et al. 2001; Dick & Rose 2002; Bell 2009). Gene transcripts can have a biological role in causing behavioural phenotypes, as they link phenotypic variation with gene expression (Boake et al. 2002). However, the transcript differences seen in Chapters 4-6 may be expressed because of the specific previous experiences or environmental effects influencing a response in gene transcripts of individual fish. Indeed, behaviour is known to alter depending upon environment, for example, boldness in rainbow trout was dependent upon food availability and predation threat (Thomson et al. 2012). Furthermore, rainbow trout alter their boldness based upon previous experience or observation (Frost et al. 2007). Similarly, a winning experience altered the type of aggressive behaviour used to initiate fights in the cyprinid fish Rivulus marmoratus (Hsu & Wolf 2001). Gene-environment interactions also affect behaviour, for example, three transcriptional profiles of aggression in Drosophila melanogaster differed between highly and less aggressive flies but showed little overlap (Bendesky & Bargmann 2011). It is often the case that behaviour is affected by gene-gene interactions as well as gene-

environment interactions. For example the interactions of serotonin with monoamine oxidase a and rearing environment affected behavioural responses to a human intruder in rhesus macaques, Macaca mulatta (Kinnally et al. 2010). These studies suggest that transcript profiles from this thesis should be considered within the context of the experimental environment in mind. Furthermore, by emulating natural environments or by controlling previous experiences during the life of a fish, transcriptome studies questions about the context-dependent nature of behaviour may be approached.