The relative roles of genetic and non-genetic factors in determining immunocompetence in birds is controversial. Although theoretical models of sexual selection tend to assume that such traits have a high genetic component, previous empirical evidence has often proven equivocal. The results of our cross-fostering experiment support claims that a part of the variation in at least one aspect of immune response is caused by genes [25-29]. Since males therefore differ in genetically-based levels of this aspect of immunocompetence, females could conceivably target immunocompetence during mate choice as predicted by the CMIH hypothesis.
In addition to indicating that immunocompetence is heritable in zebra finches, our experiment found substantial positive genetic covariance between our indices of immunocompetence and body condition. We found that approximately 56% of genetic variation in our index of body condition may be explained by genetic variation in our index of immune response. Since secondary sexual ornaments are typically condition-dependent in zebra finches [39,40], females that select males on the basis of well developed ornaments are likely to gain genes for their offspring which confer higher levels of immunocompetence. Therefore, our findings are consistent with the hypothesis that the genes that determine parasite-resistance may be a major target of sexual selection in this species.
The significant genetic covariance between our indices of immunocompetence and condition implies that the same genes underlie a proportion of the variation in both traits. However, although females selecting males with well developed ornaments are likely to gain genes that confer higher levels of immunocompetence, our genetic analysis is not sufficient to conclusively demonstrate that there are some genes that affect our indices of both immunocompetence and body condition. It is possible that linkage disequilibrium, generated by selection for both traits, may also contribute to the genetic covariance we have found between these two traits. Distinguishing pleiotropy from linkage disequilibrium in such a species is difficult; one approach would be the development of a pleiotropic quantitative trait loci map of both traits, but such techniques are yet to be applied in wild passerine populations.
The magnitude of the genetic correlation between our indices of immunocompetence and condition is surprising because theory predicts that such heritable genetic variation should be eroded through selection. What factors could maintain such variation? One possible explanation is that a third trait not included in this analysis trades off with our indices of immunocompetence and condition [44]. For example, one aspect of growth rate displays a negative genetic correlation with our index of immunocompetence (DJG unpublished data), suggesting that a more complex model of resource allocation than the simpler two-trait system of our indices of immunocompetence and condition might need to be considered to understand the maintenance of genetic variance in these traits.
Our results should be interpreted cautiously as this study suffers from a number of limitations. First, although our reciprocal cross-fostering design is efficient at detecting additive genetic effects, it is less powerful in estimating other quantitative genetic components. Our results indicated that none of the other causal components of covariance that we estimated were found to be significant. On first inspection, this suggests that non-additive genetic and environment effects do not play a role in generating covariance between our indices of immunocompetence and body condition, but we caution that this interpretation would be premature. Because the cross-fostering experimental design and the subsequent analytical method that we have used are primarily designed to detect additive genetic components of variance and covariance [42], we cannot exclude the possibility that environmental and/or non-additive components do exist and that we have simply failed to detect to them. This possibility is highlighted by the covariance estimates in Table 3, many of which are large in magnitude and are only non-significant because of the very large standard errors. Under these circumstances, no firm conclusions about the absence of environmental or non-additive covariance between these two traits can be drawn.
A second limitation of cross-fostering designs that use full sib cross-fostering is that they can only control for those aspects of environmental variance that occur after the cross-foster manipulation itself. In the case of avian studies, chicks are typically cross-fostered within 48 hours of hatching. Hence, although such studies can estimate variation associated with later incubation and parental feeding, they cannot deal with variance in factors such as the way mothers provision eggs or anything that happens in first few hours in the nest. There is always a risk, therefore, that cross-foster studies will inflate the estimate genetic components because these also include pre-cross-foster environmental components. Nevertheless, in the case of our study such environmental covariance is unlikely to be solely responsible for the high positive genetic covariance between our indices of immunocompetence and condition because the sire-offspring observation component is large and positive, which is unaffected by this source of variation.
The third limitation of our study is the lack of information about the adults used to establish the breeding experiment. Because the adults were caught from wild populations we do not know whether they were related to one another, their immunological history, or whether the parental generation experienced selection, all of which could effect the pattern of variation among individuals. In addition, although field parent and laboratory offspring relationships have been used to estimate genetic components of variance, it is not clear how differences between the lab and field environments would affect the genetic estimates in a model as complex as the one employed here.
Finally, our study is also limited by the fact that we have only used a single index of body condition and a single index of immunocompetence. Using residual values of body weight on skeletal size is a widely used index of body condition in birds, but it is well established that this method is not without its limitations and dangers [24,45]. Such an index cannot, for instance, differentiate between different aspects of condition, such as fat deposition and muscle size, and it has been shown that such an index can retain an element of body size itself. We nevertheless used this measure for our genetic tests, firstly, because it the index of body condition that has been used widely in previous phenotypic studies and is therefore of particular interest to avian studies, secondly, because it has repeatedly been shown to be under selection in avian populations [46], and thirdly, because it is not currently possible to perform multivariate analyses (with body size and tarsus length as separate covariates) using the genetic framework employed here. It would nevertheless be interesting in the future to test for genetic correlations among alternative measures of body composition, and to perform multivariate analyses if the statistical techniques are developed. Similarly, like any other index of immune response, our single measure of cell-mediate immunity cannot provide information on all elements of vertebrate immunocompetence [18,47]. Again, we used this index in our genetic study because it has been widely employed in phenotypic tests [15,20,22,30,40], has been shown to be associated with important components of fitness such as survival [20,30], and there is even limited evidence that variation in this measure may be positive associated with variation humoral immune response [[48], but see [20]]. But this should not obscure the facts that future studies of other elements of immunity are required to obtain a comprehensive understanding of the genetic basis of parasite resistance, and that multivariate models would help to tease these apart. Ideally, a full genetic variance-covariance matrix is required for a series of indices of body condition and a series of indices of immunocompetence, but this is well beyond the scope of the current study.
Answers to all your Biology Questions
