TTC - Shape Of Nature
Download File --->>> https://urlin.us/2tkLum
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Surprisingly little is known about the genetic architecture of body size in natural populations of Drosophila melanogaster. Using both generation means and triple-test-cross analyses, we investigated the genetic architecture of wing size (an indicator of body size) and wing shape in a naturally occurring body size cline. For wing size, we found significant epistatic genetic variance and evidence of past directional selection for increased body size. While wing shape also exhibits significant epistatic genetic variance, there was no indication of directional selection, suggesting instead a history of optimizing selection. Our results support the idea that epistatic variance may be more common in natural populations than was once suspected. Also, our results suggest substantial directional selection on wing size but not shape.
In addition to wing area, we also analysed wing shape. Although a number of investigators have documented natural variation in wing shape, little is known about the evolutionary genetics of wing shape within species. Weber (1990) found evidence consistent with optimizing selection on wing shape in natural populations, while Bitner-Mathé & Klaczko (1999) found high heritability for shape in D. mediopunctata. Despite recent advances in the understanding of the aerodynamics of the wings of smaller insects (Lehmann & Dickinson, 1998), virtually nothing is known about the functional implications of variation of wing morphology (Grodnitsky, 1999). This makes any inferences about the action of natural selection on shape difficult. Further investigation into the genetics and selective forces affecting shape would help to determinine whether or not natural variation is adaptive.
In a previous study, we investigated clinal variation in wing area using a generation means analysis (Gilchrist & Partridge, 1999). In the present work, we have extended our earlier study, using both a triple-test-cross analysis and a generation means analysis to investigate clinal variation in wing shape as well as wing size. Together, these designs allow the additive, dominance and epistatic components affecting the traits to be estimated. Using this information, we can characterize the genetic architecture of the traits and draw inferences about the type of natural selection that has acted on the traits.
We used wing area as an estimate of body size, because the two characters are highly correlated (Reeve & Robertson, 1952). Wing measurements were performed as detailed in Gilchrist & Partridge (1999). Briefly, wing images (one per fly) were captured using a compound microscope-mounted video camera. Using the OBJECT-IMAGE program (Vischer, 1998), coordinates of 10 wing landmarks were recorded and the wing area calculated based on the polygon shown in Fig. 1. From the same landmark data we also extracted the principle components (PCs) of shape variation for use as shape variables. PCs were computed using covariance matrices after the landmark data had undergone a Procrustes superimposition, a process of reflection, scaling (using centroid size) and rotation that produced a superimposition minimizing deviations from the overall mean shape. The process is described in detail in Dryden & Mardia (1998).
We measured the genetic components of both wing size and shape using two complementary biometrical designs. The first, a generation means analysis, was performed on the cross between two extreme populations from a natural body-size cline. This analysis estimated net genetic effects on the mean phenotypic values of the hybrid generations (as detailed in Kearsey & Pooni, 1996). The second experimental design, the triple-test-cross (TTC) design, analysed genetic variance components, rather than means, in the cross between isogenic lines from the cline ends (Kearsey & Jinks, 1968; Kearsey, 1980). These two approaches (i.e. a means analysis and a variance analysis) are complementary because they measure different aspects of underlying gene action and interaction (Jinks, 1979; Fenster et al., 1997). As the means analysis measures net genetic effects on means while the TTC measures variance components, their parameters are not correlated. Thus, for example, when the effects of increasing and decreasing dominant alleles are spread evenly between the two populations, the means analysis will show a zero net dominance (i.e. ambidirectional dominance), whereas a TTC analysis may detect significant dominance variance (because the variance is unaffected by the net direction of dominance). Alternatively, when dominance is directional rather than ambidirectional, the means analysis has the advantage of showing the direction of the dominance, something that cannot be inferred easily from the variance analysis. Similarly, the means analysis may detect epistatic interactions near fixation, while the variance resulting from the same alleles may be very small. Either analysis taken alone may produce an ambiguous indication of the genetic architecture, but together provide a clearer picture on the genetic architecture of the trait.
This analysis was performed on data from Gilchrist & Partridge (1999), where models describing wing size, but not wing shape, were presented. The generation means analysed were the two parental populations and their F1, F2 and back-cross generations. Keeping reciprocals separate, 14 generations were raised and wing traits measured. Using a weighted least-squares analysis (Kearsey & Pooni, 1996; Lynch & Walsh, 1998), composite additive ([a]), dominance ([d]), digenic epistatic ([aa], [ad] and [dd]) and maternal effects were calculated. Goodness-of-fit tests (using χ2-values) were then used to determine the model (incorporating some or all of these composite effects) that best described the observed generation means. Changes made to the method used in Gilchrist & Partridge (1999) involved a new parameter accounting for the additive effect of the X chromosome, [aX], that differs between male and females. In addition, the sampling variance was calculated separately for each generation, accounting for possible effects of variation in within-generation variance for different genotypes (Robertson & Reeve, 1953).
The transformation grids describing the shape variation measured for female wings. The shape variation shown represents the distortions necessary to move from the larger parental wing shape to the smaller parental wing shape for PC1 (upper) and PC2 (lower). The reference points are the same as those indicated in Fig. 1.
Our basic shape variables were PC1 and PC2, calculated separately for each sex. The only significant correlation with wing area occurred in females for PC2 (r=0.35[281], P < 0.001). This allometric variation could be the result of two possible underlying causes: (i) an underlying developmental constraint, i.e. the genetic mechanism controlling shape imposing particular shape variation as wing size increases; or (ii) parallel selection with wing area. If PC2 represents a developmental constraint, we have no expectation for its genetic architecture. If it is selected with size, then PC2 may have a similar genetic architecture resembling that of wing area. PC1, however, was not consistently correlated with size. Accordingly, it is unlikely to be subject either to developmental constraint or size-related selection. PC1 may represent a shape component free to respond to natural selection.
For both PC1 and PC2, the means analysis provided no sufficient model for either sex (Table 2), indicating the presence of higher order interactions and/or linkage. In both sexes there were significant [a], [ax], [aa] and [axax] components and maternal effects. In contrast to wing area, X-linked alleles appear to have a much larger effect on shape. Notably, there were no significant [d] or [dd] effects apparent for either PC in either sex. In the variance analysis (Table 3), PC1 showed significant epistatic variance only in females. The partitioning did not detect the predominant form of epistatic variance, and therefore the significant [aa] effects detected in the means analysis were not reflected in the variances. A possible explanation for this discrepancy is that the additive-by-additive component is the consequence of a large number of individually small interactions between many genes, a situation that can result in significant [aa] but nonsignificant Vaa (Kearsey & Pooni, 1996). Using a North Carolina Design III (shown for PC1 in Table 4), estimates of Va and Vd were obtained as for wing area (Table 5). In the case of the males, these estimates were unbiased because there was no significant epistasis, while the female estimates were biased to some degree by the small amount of epistatic variance. Vd for PC1 was not significant in males and Vd was small and only marginally significant in females. Given that dominance for PC1 is either small or absent, the genetic architecture of PC1 is predominantly additive, suggesting a history of optimizing rather than directional selection. In both sexes, the heritability of PC1 was high (Table 5).
Estimates of Va and Vd were also obtained for PC2, although these estimates are necessarily biased because of the highly significant epistatic effects on the variances (Tables 3 and 5). For PC2, male Vd was again not significantly different from zero, while for females Vd was again small but significant (P=0.04). This suggests that, for males at least, the absence of dominance effects on the means was in fact because of the absence of dominance variance, not ambidirectional dominance. Again, the low levels of dominance variance are suggestive of optimizing rather than directional selection on wing shape. 59ce067264