Estimating heritability in honeybees: comparison of three major methods based on empirical and simulated datasets

The genetic contribution to phenotypic variation (namely the heritability) affects the response to selection. In honeybee, the haplodiploid sex determination does not allow the straightforward use of classical quantitative genetics methods to estimate heritability and genetic correlation. Nevertheless, specific methods have been developed for about 40 years. In particular, sibling analyses are frequently used with three main methods: an historical model using the average colony relatedness, a half-sibs/full-sibs model and the more recent animal model. We compared those three methods using experimental and simulated datasets to see which performs the best. Our experimental dataset is composed of 10 colonies with a total sample of 853 workers. All individuals were genotyped to reconstitute the pedigree, and phenotypic traits were measured: labial palpus and wing cubital veins lengths. We also simulated phenotypic datasets with varying levels of heritability, common environmental effect and genetic correlation between traits. The simulation approach showed that the average colony relatedness was highly biased in presence of common environmental effect whereas the half-sibs/full-sibs and the animal model gave reliable estimates of heritability. The animal model provided the greatest precision in genetic correlations. Using this latter method, we found that wing vein lengths had high heritabilities, allowing the use of those morphometric characters to discriminate between populations. On the contrary, the palpus length had lower heritability due to larger environmental variance and measurement error. Finally, significant genetic correlations among measured traits indicate that they do not evolve independently.