Pairwise correlations were used to measure the relationships between hormone
levels, between receptor-gene expression levels and between hormone levels and receptor-gene expression levels in the fishes’ brains and gonads. The hormone levels were log transformed to fit the normal distribution. To avoid an inflated chance of Type 1 error, the α level for this analysis was adjusted to be more conservative (P <
0.0024) by using the Bonferroni adjustment.
Multiple linear regressions and logistic regressions were used to examine the relationships between hormone levels (CORT & T) and contest behavior (latency to movement, initiating displays, latency to display, initiating gill displays, latency to gill display, initiating attacks, latency to attack, escalation and contest duration) in the 1st experience training. The hormone levels, latency to movement, latency to display, latency to gill display, latency to attack and contest duration were log transformed to fit the normal distribution. Because individuals with a first winning/losing experience would become more/less aggressive in subsequent contests (ET3W always initiated contests in their second and third experience training; ET3L never initiated contest in their second and third experience training), I only used contest behavior in the first experience training. I also analyzed the winning and losing individuals’ behaviors separately, because they faced different opponents (STW/STL) and were expected to behave differently. The focal fish’s standard length, last contest experience, strain type and standard length of trainer were included in the model as control factors.
Multiple linear regressions were also used to examine the influence of the type of contest experience and decay time on receptor-gene expression levels. Log transformed hormone levels (CORT & T), fish’s standard length, last contest experience and strain types were include in the model as control factors.
Multiple linear regressions were used to examine the relationships between contest behaviors in the 1st experience training and receptor gene expression levels.
Because the contest behaviors in experience trainings were highly correlated with
each other, including them in the models at the same time would result in a multicollinearity problem. The importance of each of these behaviors was therefore tested separately. The correlation between contest behaviors and receptor-gene expression was tested, and standard length, last contest experience and strain type were included in the model as control factors. JMP (v. 5.0.1; SAS Institute Inc., Cary, NC, U.S.A.), a commercial statistical package, was used for all the statistical analyses in this study.
Results
A total of 298 individuals were used in this study. The mean value of ΔCt of each receptor genes were as follows: In brain: AR (10.26±1.01), ERα (1.60±1.39), ERβ (1.41±1.41), GR (2.29±1.12), 5-HT1AR (0.95±1.32); In gonad: AR (3.52±0.90), ERα (3.24±0.60), ERβ (2.91±0.42), GR (5.35±0.57), 5-HT1AR (7.12±0.58). The mean value of hormone levels was T (942.77±823.96) and CORT (151.78±135.49).
The receptor-gene expression levels and hormone levels for each treatment are shown in supplemental data (Table 1-6).
3.1 The Relationship between Baseline Hormone Levels and Contest Behaviors in the 1
stExperience Training
Baseline CORT and T levels did not appear to have a significant influence on the contest behavior in the 1st experience training of those fish allocated to receive winning experiences (ET1W`& ET3W, matched with standard losers), (Table 4). Strain type, however, did have a significant effect on their contest behavior: the SLC8E strain took less time than average to initiate moves (P = 0.039), displays (P = 0.015) and attacks (P = 0.044). Individuals whose last contest experience was a win took longer to initiate gill displays (P = 0.019).
For those fish allocated to receive losing experiences (ET1L & ET3L, matched with standard winners), individuals with higher baseline T levels were more likely to initiate gill displays (P = 0.014) and persisted longer before eventually retreating (P = 0.019), although CORT levels did not have a significant relationship with any of the behaviors examined (Table 5). Larger individuals were also more likely to initiate gill displays (P = 0.050). Individuals whose last contest experience was a win were more
likely to escalate a contest (P = 0.032).
3.2 Effect of Experience Type and Decay Time on Post-Experience Receptor-Gene Expression Levels
3.2.1 Brain Tissue Receptor-Gene Expression Levels
Experience type had a significant effect on AR gene expression levels in the individual’s brains (P = 0.036, Table 6). Individuals with three losing experiences had significantly lower AR gene expression levels than the control group, which received one no-experience (P = 0.007). In addition, decay time effects were discovered in AR, ERα, ERβ and 5-HT1AR gene expression levels. Because the regression model’s interaction terms relating experience and time decay (experience × time-decay) to these receptor gene expression levels were not significant (all interaction terms P ≥ 0.718), it may be that the decay-time effect was caused by handling disturbance, rather than the experimental treatments. The results also showed that the receptor-gene expression levels related negatively with baseline CORT levels (P ≤ 0.017; except the relationship between CORT and AR, P = 0.114, and the relationship between CORT and GR, P = 0.051) but positively with baseline T levels (P ≤ 0.017).
3.2.2 Gonad Tissue Receptor-Gene Expression Levels
Experience type had a significantly affect on ERα gene expression levels in the individual’s gonads (P = 0.047, Table 7). Individuals with three losing experiences had significantly lower ERα gene expression levels than the control group which received one no-experience (P = 0.006). Decay time effects were also discovered in AR and ERα gene expression levels, but, as with the fishes’ brains, the fact that the
regression model’s interaction terms relating experience and time decay to these receptor gene expression levels were not significant (all interaction term P ≥ 0.353) suggests that this decay-time effect might also be caused by handling disturbance. The analysis also showed that the receptor-gene expression levels related negatively with CORT levels (P ≤ 0.014; except the relationship between CORT and AR, P = 0.131 and the relationship between CORT and ERα, P = 0.075) but positively with T levels (P ≤ 0.001; except the negative relationship between T and AR, P < 0.001, and the relationship between T and ERβ, P = 0.177)
3.3 Relationship between the Contest Behaviors in the 1
stExperience Training and Post-Experience Receptor-Gene Expression Levels in Brain Tissue
Among the individuals allocated to receive winning experiences (ET1W`& ET3W, matched with standard losers), those which took longer to initiate moves and displays had significantly lower GR gene expression levels (initiate moves, P = 0.033; initiate displays, P = 0.043, Table 8). Interestingly, among the individuals allocated to receive losing experiences (ET1L & ET3L, matched with standard winners), those which were more likely to initiate displays, gill displays and took longer to retreat from contests had significant higher ERα (initiating displays, P < 0.001; initiating gill displays, P <
0.001; contest duration, P = 0.013), ERβ (initiating displays, P < 0.001; initiating gill displays, P < 0.001), GR (initiating displays, P < 0.001; initiating gill displays, P = 0.036) and 5-HT1AR (initiating displays, P < 0.001; initiating gill displays, P = 0.016) gene expression levels (Table 9). The relationships between contest behavior and AR gene expression levels, however, were not significant, although losing experiences did affect AR gene expression levels in brain tissue.
3.4 The Relationships between Baseline Hormone Levels and Post-Experience Receptor-Gene Expression Levels
Table 10 shows that receptor-gene expression levels are highly correlated with other receptor-gene expression levels. The gene expression levels of ERα, ERβ, GR and 5-HT1AR were positively correlated in both brain and gonad tissue (P ≤ 0.002;
except the relationship between ERβ and 5-HT1AR in brain, P = 0.008). The relationships between AR and other receptor-gene expression levels were often negative in gonad (P < 0.001; except ERβ in gonad; r = 0.140, P = 0.016) and were more significant in gonad tissue (P < 0.001) than in brain tissue (P ≤ 0.639).
CORT levels did not have a strong relationship with the gene expression level of the receptors examined (P ≥ 0.035), expect for its negative relationship with ERβ in gonad (P < 0.001). T levels, on the other hand, were closely related with the gene expression level of the receptors (exceptions: AR, GR and 5-HT1AR in brain, P ≥ 0.005; ERβ in gonad, P = 0.906). The significant relationships between T levels and receptor-gene expression levels were mainly positive (P < 0.001), except for the negative relationship between T and AR in gonad tissue (P < 0.001). In addition, T levels were positively correlated with CORT levels (P < 0.001).
Discussion
My results show that individuals with three losing experiences had significantly lower AR (in brain tissue) and ERα (in gonad tissue) gene expression levels than individuals in the control group, with one no-experience. Also, baseline hormone levels (T) and receptor-gene expression levels (ERα, ERβ, GR and 5-HT1AR) were positively correlated with individuals’ contest behavior in the 1st experience training, particularly in those individuals given losing experience(s).
The results do not, however, clearly establish that hormone/serotonin receptors might be the physical mechanism underlying winner and loser effects: neither one losing experience nor either of the winning experience treatments had a significant effect on hormone/serotonin receptor-gene expression levels. The following paragraphs discuss the results, the possible reasons for them and ideas for future research in more detail.
4.1 Effect of Experience Type on Post-Experience Receptor-Gene Expression Levels
My study discovered that neither one losing experience nor either of the winning experience treatments had a significant effect on hormone/serotonin receptor-gene expression levels. However, recent studies (e.g. Marini et al. 2006; Fuxjager et al.
2010) indicated that not only circulating hormone levels, but also receptor-gene expression levels are highly associated with contest experience. For instance, in California mouse, Fuxjager et al. (2010) found that three winning experiences significantly increased the expression levels of the AR gene in the medial anterior bed nucleus of the stria terminalis (BNSTma region), nucleus accumbens (NAcc) and ventral tegmental area (VTA); this region is a key brain area that controls social
aggression. While in Sprague-Dawley rats, Marini et al. (2006) discovered that GR gene expression levels were found significantly increasing in rat hippocampus 30 hr after receiving a forced losing experience. These results suggested that winning or losing experiences may affect receptor-gene expression in some brain regions;
however, in K. marmoratus, we did not find any significant changes of brain receptor gene expression levels in ET1L or ET3W. These different results might be a consequence of difference in study animals, difference in target brain regions or differences in experimental design:
(1) Differences in experimental design. Experimental design might be an important factor behind the different results. In my study, the focal individuals were divided from their trainer immediately after the contest was resolved. While in Marini’s study (2006), the focal individuals were left in visual, auditory and olfactory contact with the dominant individuals for 30 min after the contest was resolved; this might not only have enhanced the magnitude of social defeat stress for the focal individuals but also possibly had a different effect on brain receptor-gene expression levels. For instance, research on European starling (Sturnus vulgaris) showed that long-term stress reduced glucocorticoid receptor-gene expression in the hypothalamic paraventricular nucleus and mineralocorticoid receptor-gene expression in the hippocampus (Dickens et al. 2009), while short-term stress did not. It suggested that the changes of GR gene expression might be not only caused by losing experience, but also by long-term social defeat stress. Comparison with Fuxjager’s study (2010), I measured “post-experience” receptor gene expressions levels, while they measured
“post-encounter” (post-fight) receptor gene expression levels. For this reason, the changes of AR gene expression levels in their study may be caused by winning experience or contest interaction in the subsequent encounter. It is possible that the AR gene/protein expressions increased because of the elevated aggressive interaction
in contests (encounter stage), rather than because of the winning experiences. The results of the study, unfortunately, could not demonstrate a direct link between contest experiences and the changes of receptor gene/protein expression levels.
(2) Differences in target region (whole brain & specific brain regions). Different brain regions might mediate different behaviors and respond differently to contest experience. Fuxjager et al. (2010), for instance, focused on the brain regions that control the output of antagonistic behavior and social aggression in mice (e.g. nucleus accumbens, lateral septum, medial anterior bed nucleus of the stria terminalis, medial amygdale, anterior hypothalamus, ventrolateral subnucleus of the ventromedial hypothalamus, ventral premammillary nucleus, ventral tegmental area, dorsal periaqueductal gray), while Marini et al. (2006) focused on the brain region (e.g.
hippocampus) that was described as particularly sensitive to the effects of social defeat stress and stress hormones in rats. My study, however, checked gene expression levels in the whole brain rather than in a specific brain region, because the brain of K. marmoratus is too small to allow separation into different regions. If different brain regions respond differently to winning/losing experiences, the effect of contest experiences on receptor gene expression levels across the whole brain would be diluted and become hard to detect. Although my study did not discover any effect of a losing contest experience or three winning contest experiences on receptor-gene expression, the results still revealed significant relationships between receptor-gene expression levels and aggressive behavior, which suggests that the effect of contest experience on receptor gene expression may not be diluted even when measured over the whole brain.
(3) Differences in study animal (rodents & fish). Another difference between my study and those quoted above is the study organism: the previous studies all used rodents (California mice & Sprague-Dawley rats) whereas I used a hermaphroditic
fish, K. marmoratus. Because the physiological structures of rodents and fish are totally different (especially their brains and gonads), they may respond differently to the same experiences. For instance, androgen (methyl-testosterone) would efficiently change the sex of K. marmoratus, from hermaphrodite to male (Kanamori et al. 2006);
for this reason, circulating androgen levels or AR sensitivity in this fish’s gonads must be restricted below a specific threshold in order to prevent sex change, while the rodent’s physiological condition might be not restricted in this way.
The result that individuals with three losing experience had significant lower AR gene expression levels may be not caused by losing experience, because individuals with one losing experience did not have significant lower AR gene expression (or a non-significant trend) than individuals in the control group. This result may be caused by social defeat stress. Contest interaction is a kind of stress for contestants, especially for losing indivuduals (social defeat stress), and it induces animals’ quick but strong stress response (Miczek et al. 2002). This suggests that individuals with three losing experiences might be under more stress than those with one losing experience. Past studies have shown that social defeat stress may cause changes in receptor gene expression in brain tissue (e.g. Marini et al. 2006). Although the effect of social defeat stress on AR gene expression in brain tissue remains unclear, there are some in vitro data revealing a relationship between stress response and AR gene expression. For instance, Burnstein et al. (1995) discovered that stress hormone, glucocorticoids, down-regulated AR mRNA expression levels and decreased AR protein levels in COS 1 cell lines. Based on this finding, it is possible that the reduction in AR gene expression levels in those individuals with three losing experiences was caused by their strong stress response rather than their losing experience.
My study also showed significantly lower ERα gene expression in the gonads of
individuals given three losing experiences than in those of the control group. There is little published research about the relationship between winning/losing experiences and receptor gene expression levels in gonads; studies have, however, revealed that social defeat stresses caused by agonistic contests suppress gonad function or reproductive ability (e.g. Tilbrook et al. 2000). This result suggests that the stress induced by serial losing experiences affects gonad function and also possibly suppresses gonad receptor-gene expression levels, although no direct linkage between losing experiences and gonad ERα gene expression levels has been discovered.
4.2 The Relevance of the Fish’s Baseline Hormone Levels and Post-Experience Receptor-Gene Expression Levels in Brain to the Fish’s Contest Behaviors
My results showed that baseline hormone levels were highly correlated with the fish’s contest decisions when facing a much larger opponent (i.e. standard winner), in training for losing experience: individuals with higher baseline T levels were more likely to initiate gill displays and persisted longer before retreating. This result is similar to that of Earley & Hsu (2008), who found that pre-fight T levels related positively to contest initiation in the smaller opponent. Positive associations between T levels and aggression are also reported for other animal species (e.g. in rats, Caldwell et al. 1984; in sparrows, Wingfield & Hahn 1994; in humans, Archer 2006) and in studies using castration or exogenous hormone injections (Francis et al. 1992;
Ruiz-de-la-torre & Manteca 1999). Social mixing male lambs with testosterone enantate injections, for instance, were more aggressive than control individuals (Ruiz-de-la-torre & Manteca 1999); castrated male cichlid fish (Haplochromis burtoni;
Francis et al. 1992) with an accompanying reduction in T levels had significantly
lower aggression than sham-operated males.
In my results, however, no positive correlations were discovered between T levels and aggressive behavior in individuals facing smaller, submissive opponents (i.e. standard loser) when in training for winning experience. This may be because standard losers behaved submissively and usually quickly retreated from the experience training even without focal individuals initiating gill displays or launching attacks. Therefore the behavior in training of focal individuals might not precisely reflect the endocrine states.
My study also indicated that not only baseline hormones but also hormone/serotonin receptor-gene expression levels (ERα, ERβ, GR and 5-HT1AR) in the fish’s brains were positively correlated with those individuals’ aggressive behavior in experience training, particularly in individuals given losing experiences (fighting with larger habitual winners). Research in rodents has discovered that ERα expression levels are positively correlated with aggressive behavior, while ERβ expression levels are negatively or non-correlated with aggressive behavior. For instance, Trainor et al.
(2007) found that increases in aggressiveness in short-day treatment mice accompanied increased ERα gene expressions and decreased ERβ gene expressions in LS/BNST brain regions. Previous studies also indicated that ERα gene knockout male mice were less aggressive, whereas ERβ gene knockout male mice exhibited normal or increased aggression (Ogawa et al. 2000). In my results, ERα gene expression was positively correlated with individuals’ aggressive behavior as in the previous studies, while ERβ gene expression, unlike the previous research, was also positively correlated with individuals’ aggressive bahavior. Most of above research, however, focused on aggressive behavior in rodents; whether the affect of ERα and ERβ aggression in this fish is the same as in rodents is still unclear. It is possible that both ERα and ERβ gene expression levels have a positive effect on aggressive behavior in
this fish.
5-HT1AR is another receptor which has been found to be associated with aggressiveness. Classified by the expression location on neuron cells, there are two kinds of 5-HT1AR. One type is located on postsynaptic cells; the other type, located on presynaptic cells are called autoreceptors. Previous research indicated that the expression level of postsynaptic 5-HT1AR might be negatively correlated with aggressiveness, while autoreceptor expression levels might be positively correlated with aggressiveness. For instance, Caramaschi et al. (2007) discovered that the postsynaptic 5-HT1AR agonist increases receptor sensitivity, and decreases aggressiveness in rodents; the autoreceptor agonist decreases the firing rate of presynaptic cells and suppresses 5-HT synthesis and thereby increases aggressiveness in rodents. In my results, the 5-H1A receptor gene expression levels were positively correlated with individuals’ aggressiveness. Although I could not distinguish between the expression of these two kinds of receptor genes, it may be that the more aggressive individuals had higher autoreceptor gene expression levels in presynaptic cells than the less aggressive individuals.
In my results the expression of GRs genes was positively correlated with individuals’ aggressive behavior. GRs have usually been found to correlate with social stress response rather than with aggressiveness (e.g. Marini et al. 2006; Dickens et al.
2009); however, some studies suggest a positive relationships between GRs expression and aggressiveness. For example, in research on lizards (Anolis carolinensis), blocking GR by mifepristone reduced aggressive attacks/displays in the
early stages of fighting (Summers et al. 2005). This result demonstrates that GR gene expression levels might be positively correlated with aggressiveness, which is similar to the relationships observed in my study.
4.3 Other Possible Mechanisms Mediating Winner and Loser Effects
In addition to changes in levels of hormone/serotonin receptor gene expression, there are other mechanisms which might mediate winner/loser effects. Winning and losing experience might change individuals’ behavior by learning and memory processes, which would be mediated by a different set of physiological mechanisms.
The capability of fish to change its aggressiveness or willingness to fight as a result of learning has been demonstrated (e.g. Hollis 1984; Hollis 1999; Carpenter & Summers
The capability of fish to change its aggressiveness or willingness to fight as a result of learning has been demonstrated (e.g. Hollis 1984; Hollis 1999; Carpenter & Summers