Multiple effects of inbreeding in populations undergoing directional selection

Background/Question/Methods

Plants are expected to be under strong selection for defense against insect herbivores.  The ability of populations to respond to selection is determined by the level of genetic variation within populations.  In randomly mating populations, the response is determined exclusively by additive genetic variation.  Randomly mating plant populations are probably the exception rather than the rule, however.  Plants vary widely in their mating-systems, and plant populations tend to be highly structured, resulting in high levels of biparental inbreeding.  In populations with high levels of inbreeding, non-additive genetic variation (dominance and epistasis) can account for much of the phenotypic variation, constraining or perhaps enhancing the ability to respond to selection.  We addressed the effects of inbreeding on the direct and indirect responses to selection with the following experiment.

We collected seed from 500 maternal plants from each of two geographically separated populations of Mimulus guttatus in California: Snell Valley (SV) and Don Pedro Reservoir (DP).  We randomly chose seed from 150 families from SV and DP to start base populations for artificial selection experiments in the greenhouse.  Three replicate base populations were established from both SV and DP.  Bivariate selection on trichome density (an important defense against herbivores) and stigma-anther separation (an important determinant of selfing rates in Mimulus) was then imposed on each base population in a factorial manner: trichome density (high or low) x stigma-anther separation (narrow or wide).  We imposed a highly selfing mating-system on those lines experiencing selection for narrow stigma-anther separation and a fully outcrossed mating-system on those selected for wide stigma-anther separation.  A random-mated, fully outbred control line was included for each replicate base population.

Results/Conclusions

After three generations of selection, traits in all populations diverged significantly from unselected controls.  The magnitude of the response of stigma-anther separation was significantly reduced in the inbred lines.  For population DP, the response to selection for reduced trichome density in inbred lines was almost twice the magnitude seen in outbred lines.  In population SV, date of first flower, a trait with a strong genetic correlation with trichome density, showed a predicted correlated response to selection in outbred lines but showed no response in inbred lines.  In both populations morphological and fitness traits uncorrelated with either trichome density or stigma-anther separation showed large phenotypic changes in inbred lines.  These results highlight the importance of nonadditive genetic variation in producing unpredictable responses in inbreeding populations experiencing directional selection.