Selection for corticosterone
In total we have bred and assayed 1714 birds (866 males and 848 females) from 10 to 20 families per line per generation (Table 1). Figure 2 shows how the peak corticosterone titre changed over the duration of the selection experiment. Overall there were significant effects of selection pressure (line) (Wald = 148.75, d.f. = 2, P < 0.001), sex (Wald = 7.37, d.f. = 1, P = 0.007) and generation (Wald = 36.35, d.f. = 1, P < 0.001). The effect of family was highly significant (change in deviance because of family effect = 187.78, d.f. = 139, P < 0.001), as was replicate (change in deviance because of replicate effect = 35.93, d.f. = 1, P < 0.001). There was also a significant line by generation interaction, showing that the lines diverged over time (Wald = 17.12, d.f. = 2, P < 0.001; Fig. 2). There has been a significant difference between the mean corticosterone titre of the high and control (and high and low lines), since generation 2, as shown by significant contrasts between corticosterone selections regimes. The figure also reveals that there has been a significant reduction in corticosterone titre in the two control lines (control 1 b = −6.5, r2 = 0.94, n = 6 P < 0.01; control 2 b = −5.1, r2 = 0.83, n = 6, P < 0.05).
Figure 2. Response to artificial selection per generation for four generations of selection, for positive selection (triangle symbols) and negative selection (square symbols), control lines are shown by circular symbols, in two replicate lines in each direction as shown by solid lines with filled symbols and dashed lines with open symbols. Points are population mean corticosterone titre (±SE). The box shows whether there was a significant post hoc contrast (Tukey's test) between corticosterone selection regimes for each generation.
Download figure to PowerPoint
The significant effect of sex in previous analyses suggests that it is justified to separate the two sexes. It is noticeable that females have significantly higher corticosterone titres than males, as shown by a significant sex effect in the analysis shown above. However, when the analyses are separated by sex they show that there have been qualitatively similar changes in each of the two sexes. There were also significant effects of line (males Wald = 115.04, d.f. = 2, P < 0.001; females Wald = 44.21, d.f. = 2, P < 0.001), and generation (males Wald = 37.49, d.f. = 1, P < 0.001; females Wald = 27.56, d.f. = 1, P < 0.001). A significant line by generation interaction term existed for both sexes showing there had been significant divergence in corticosterone titre between the lines during the process of selection (males Wald = 7.13, d.f. = 2, P < 0.05; females Wald = 10.32, d.f. = 2, P < 0.01). Both sexes also showed significant effects of replicate (males change in deviance because of replicate effect = 24.75, d.f. = 1, P < 0.001; females change in deviance because of replicate effect = 9.52, d.f. = 1, P < 0.01) but the family effect was nonsignificant for both sexes (males change in deviance because of family effect = 87.00, d.f. = 74, 0.01 > P > 0.05; females change in deviance because of family effect = 102.18, d.f. = 66, n.s.) In males considered separately, there have also been significant differences in mean corticosterone titre between the high and control lines (and high and low lines), since generation 2. This is similar to those seen in the combined dataset, but the males in the high line were significantly different from males in the low line by generation 1. In females however, the effect of selection is less marked, with significant differences between the high and low lines only emerging by generation 3 and between high and control lines by generation 4. There have been similar changes in corticosterone titre in the two control lines with time in both sexes, although these have been more marked in females than in males (males: control 1 b =−6.5, r2 = 0.98, n = 6, P < 0.001; control 2 b = −4.0, r2 = 0.67, n = 6, n.s.; females: control 1 b = −7.1, r2 =0.87, n = 6, P < 0.05; control 2 b = −6.7, r2 = 0.87, n =6, P < 0.05).
The fact that there have been significant declines in corticosterone titre in the control lines suggests that there have been consistent changes in the animals, independent of the selection regime. To remove this unwanted generation effect, the response to selection in the selected lines has been expressed as the difference from their respective control lines. Figure 3 shows the response to selection in the combined dataset against the cumulative selection differential. There were effects observed in the high lines with a significant regression coefficients in response to selection for high levels of peak corticosterone in one line and a near significant effect in the other (high 1 r2 = 0.87, n = 6, P < 0.05; high 2 r2 = 0.35, n =6, n.s.). However, there was no significant effect observed in either of the low lines which both changed in parallel with the controls, with nonsignificant regression coefficients in both these lines (low 1 r2 = 0.13, n =6, n.s.; low 2 r2 = 0.28, n = 6, n.s.). The data from male birds show similar results, with a significant response to upward selection and no significant effect of downward selection (high 1 r2 = 0.79, n = 6, P < 0.05; high 2 r2 =0.12, n = 6, n.s.; low 1 r2 = 0.04, n = 6, n.s.; low 2 r2 =0.21, n = 6, n.s.). The results obtained from female birds were also qualitatively similar to the combined dataset (high 1 r2 = 0.49, n = 6, n.s.; high 2 r2 = 0.36, n = 6, n.s.; low 1 r2 = 0.13, n = 6, n.s.; low 2 r2 = 0.23, n = 6, n.s.). Greater downward selection was exerted on females (cumulative selection differentials males low line 1 = −60.3, line 2 = −58.8; females low line 1 =−71.9, line 2 = −79.3), while greater upward selection was exerted on males (cumulative selection differentials males high line 1 = 83.6, line 2 = 102.5; females high line 1 = 75.7, line 2 = 74.9).
Figure 3. Response to selection expressed as deviation from relevant control in each generation, in relation to the cumulative selection differential. Symbols as in Fig. 2. Lines represent regression lines for each replicate, note that the regressions for the low selection lines are not significant but have been added for consistency. Some jitter has been added to the points at zero.
Download figure to PowerPoint
Realized heritability is calculated as the regression coefficients from the relations between cumulative responses to selection and cumulative selection differentials (Falconer & Mackay, 1996; Lynch & Walsh, 2005). Table 2 summarizes these results and shows that there was significant heritability in corticosterone production in the high lines, but not in the low lines and that heritability estimates were similar in the two sexes.
Table 2. Realized heritability estimates (h2) ±SD for peak corticosterone response in high and low line zebra finches, calculated as the slope of the regression between cumulative selection differential and response to selection.
|Direction of selection||Replicate||All birds combined||Males only||Females only|
|High||1||0.24 ± 0.04||0.27 ± 0.04||0.20 ± 0.06|
|High||2||0.10 ± 0.03||0.09 ± 0.04||0.11 ± 0.05|
|Low||1||0.08 ± 0.02||0.04 ± 0.02||0.09 ± 0.02|
|Low||2||0.14 ± 0.02||0.21 ± 0.03||0.10 ± 0.02|
Correlated effects on testosterone
There was no significant difference in the plasma levels of testosterone in adult males in breeding condition in the different corticosterone selection lines although there were significant effects of generation (F1,206 = 21.54, P < 0.001), suggesting a decline in testosterone titre of 0.02 ± 0.001 nmol L−1 per generation that was similar in all lines. There was also a significant effect of replicate (F1,206 = 12.26, P < 0.001; replicate one being an average of 0.01 ± 0.001 nmol L−1 lower than replicate two) there are no significant effects of either line (F2,202 =0.35, P = 0.70) or line interacting with generation (F2,202 = 0.54, P = 0.58). When the line by generation interaction term is removed, the line term does not become significant (F2,204 = 0.20, P = 0.81). Therefore, the change in testosterone with time appears to have occurred in similar ways in the different lines (Fig. 4). If there has been a change in corticosterone titres over time and no change in testosterone titres over time, we should expect that there would be a significant corticosterone titre by generation interaction term to emerge in the data analysis of the birds for which we have data on both hormones. This is indeed the case (F1,191 = 17.10, P < 0.001).
Figure 4. Plasma testosterone titres in random samples of adult, breeding condition males from generations 2–4. Symbols as in Fig. 2. An average of 10 males were sampled per line per generation (range 5–17) representing about a third of the males in any generation.
Download figure to PowerPoint