3.1 Introduction
4.1.3 Measuring intrinsic selection
The study by Kruuk (1997) suggests that embryonic mortality may be used as a measure of fitness in Bombina hybrid zones. In general, the reproductive strategy in amphibians is adapted to high mortality rates in the egg and larval stage of the life cycle. Amphibians typically produce a large number of eggs most of which are destined not to survive. High mortality rates suggest that early stages of the life cycle may be seen as an arena of natural selection. Travis et al. (1987) found that almost all embryonic mortality in the treefrog Hyla crucifer occurred during gastrulation and neurulation, at a stage in
which interactions between maternally inherited cytoplasmic elements and the zygote is most intense. This effect will probably be magnified by the degree of divergence between the parental genomes. In addition, early hybrid dysfunction may be related to incompatibilities between nuclear parental alleles in the zygote that are likely to cause severe developmental problems and result in the abortion of the embryo.
Kruuk (1997) measured embryonic mortality in egg batches taken from the field in Pescenica. However, she related it to the mean hybrid index of the respective adult population. This might be in error, since the genotypic range of adults in a site may not represent the gene pool of the individuals that actually breed there. Adults may visit a site before moving on to a more suitable one in which they actually reproduce. For example, Vines (2002) found that the parents he inferred from egg data were significantly more B. variegata-like than the overall adult sample in temporary sites, implying strong breeding site preference. Therefore, intrinsic selection is better investigated at the level of families rather than at the population level. In this Chapter, I relate embryonic mortality to hybrid indices across families to test for a correlation between fitness and heterozygosity.
Second, to test for intrinsic selection within families, I take an approach in which I infer the parental genotypes from the family and compare the observed number of offspring per possible genotype to the expectations from Mendelian segregations without selection. This test for intrinsic selection within families can only identify effects that are physically linked to the loci in question, because correlations between genotype and phenotype within families always imply physical linkage. I refer to the entity of the inferred parental genotypes as “joint parental genotype”. It is the listing of the most likely parental genotypes across loci and families, where genotypes are coded by the number of B. variegata alleles present. For example, the occurrence of all three genotypes (0, 1 and 2) within a family suggests that both parents were heterozygous at this locus (1,1). The joint parental genotype across all four loci might be, for example, ((1,1)(0,0)(1,2)(0,2)). Note that for unlinked marker loci one may not determine which per-locus genotype came from which parent.
There are two ways in which intrinsic selection may act. First, alleles of one taxon may have an intragenomic selective advantage and preferentially be passed on to the next generation, resulting in a shift in segregation ratios, but not in heterozygote deficit. The selective advantage of single alleles may be determined in a heterozygous parent during
gamete production before any zygote is formed (akin to meiotic drive) or afterwards through selection against individual zygotes carrying a certain proportion of alleles of the “wrong” taxon. Meiotic drive favors a certain allele irrespective of the genetic background, whereas selection in the zygote depends on the allelic state of other genes in the genome. This mechanism is de facto frequency dependent because it disfavors alleles of the “minority taxon”. Note that the occurrence of meiotic drive is not very likely a prominent force in a hybrid zone. If consistent patterns across marker loci are found one would have to assume that each one of them is linked to a driver locus despite the fact that same-taxon suppressor alleles need to have recombined away from the driver alleles and that drivers occur in only one taxon. Second, intrinsic selection may act against heterozygous individuals due to genetic incompatibilities, leading to heterozygote deficit in the offspring through selective deaths in early developmental stages.
Before any analysis of intrinsic selection can be attempted, it is important to exclude genotyping errors and errors that arise from mistakenly analyzing mixed families or the same family more than once. In Bombina in particular, the identification of true families is not trivial due to the adults’ spawning habits. During spawning, pairs in amplexus move around the water body, and the females often deposit the eggs in several locations. Eggs are usually attached to plants which are sometimes limiting, especially in ephemeral puddles. Therefore, an egg batch that appears homogenous may in fact contain eggs from more than one family, while full siblings may be distributed over several separate egg batches. Genotyping errors and undetected mixed families would widen the range of genotypes present in a family. This would erroneously increase the overall number of heterozygous parents inferred from the data. A genotyping error may only become evident as a locus-specific, single aberrant genotype within a family, which I refer to as a “singleton”. Initially, I assume Hardy-Weinberg equilibrium and test whether the observed frequency of singletons across the data set agrees with Mendelian expectations. An excess could have two explanations: genotyping errors or non-random segregation, which can be tested with repeated laboratory analyses. The presence of one family in various egg batches is indicated when compatible joint parental genotypes occur across all loci. In most sites, information on a highly variable microsatellite marker locus that allows family assignments is available, which enables the identification of repeatedly sampled and of mixed families. In two sites without genotypes for this locus, finding true families is more indirect. There, mixed families
can be detected by a significant association between allelic states at different loci, since no such associations are expected between physically unlinked loci among full siblings. To test for this, I apply two different statistical methods described below. Finally, I infer the maximum likelihood estimate of segregation ratios within loci across all families. This estimate will be used to compute the expected ratio of heterozygotes, which is compared to the observed number, to test for heterozygote deficit in the egg families.
4.2
Methods
4.2.1 Selection of sites
For the adult survey presented in Chapters 2 and 3, a wide range of sites was visited over both seasons. In this survey, all available egg batches were sampled, which gave a set of 34 sites. For the study presented in this and the next Chapter, it would not be worthwhile to genotype eggs in sites with almost pure B. variegata adults. Therefore, I concentrated the analysis on 14 sites that i) produced a reasonable number of egg batches over the season, ii) produced surviving tadpoles at the end of the season for the cohort study in Chapter 5, iii) covered a broad range of habitat types and iv) contained a wide range of adult genotypes. An estimate for adult genotypes may be obtained in the field from their belly pattern as described in Chapter 2, which facilitates the choice of suitable sites in the field.