Résumé
Genome scans of population differentiation identify candidate loci for
adaptation but provide little information on how selection has influenced
the genetic structure of these loci. Following a genome scan, we
investigated the nature of the selection responsible for the outlying
differentiation observed between populations of the marine mussel Mytilus
edulis at a leucine/arginine polymorphism (L31R) in the antimicrobial
peptide MGD2. We analysed DNA sequence polymorphisms, allele frequencies
and population differentiation of polymorphisms closely linked to L31R,
and pairwise and third-order linkage disequilibria. An outlying level of
population differentiation was observed at L31R only, while no departure
from panmixia was observed at linked loci surrounding L31R, as in most of
the genome. Selection therefore seems to affect L31R directly. Three
hypotheses can explain the lack of differentiation in the chromosomal
region close to L31R: (i) hitchhiking has occurred but migration and
recombination subsequently erased the signal, (ii) selection was weak
enough and recombination strong enough to limit the hitchhiking effect to
a very small chromosomal region or (iii) selection acted on a pre-existing
polymorphism (i.e. standing variation) at linkage equilibrium with its
background. Linkage equilibrium was observed between L31R and linked
polymorphisms in every population analysed, as expected under the three
hypotheses. However, linkage disequilibrium was observed in some
populations between pairs of loci located upstream and downstream to L31R,
generating a complex pattern of third-order linkage disequilibria which is
best explained by the hypothesis of selection on a pre-existing
polymorphism. We hypothesise that selection could be either balanced,
maintaining alleles at different frequencies depending on the pathogen
community encountered locally by mussels, or intermittent, resulting in
sporadic fluctuations in allele frequency.