Joint analysis of microsatellites and flanking sequences enlightens complex demographic history of interspecific gene flow and vicariance in rear-edge oak populations

Inference of recent population divergence requires fast evolving markers
and necessitates to differentiate shared genetic variation caused by
ancestral polymorphism and gene flow. Theoretical research shows that the
use of compound marker systems integrating linked polymorphisms with
different mutational dynamics, such as a microsatellite and its flanking
sequences, can improve estimation of population structure and inference of
demographic history, especially in the case of complex population
dynamics. However, empirical application in natural populations has so far
been limited by lack of suitable methods for data collection. A solution
comes from the development of sequence-based microsatellite genotyping
which we used to study molecular variation at 36 sequenced nuclear
microsatellites in seven Quercus canariensis and four Q. faginea rear-edge
populations across Algeria. We aim to decipher their taxonomic
relationship, past evolutionary history and recent demographic trajectory.
First, we compare the estimation of population genetics parameters and
simulation-based inference of demographic history from microsatellite
sequence alone, flanking sequence alone or the combination of linked
microsatellite and flanking sequence variation. Second, we apply random
forest approximate Bayesian computation to identify which of these
sequence types is most informative. Whereas analysing microsatellite
variation alone indicates recent interspecific gene flow, additional
information gained by integrating nucleotide variation in flanking
sequences, by reducing homoplasy, suggests ancient interspecific gene flow
followed by drift in isolation instead. The weight of each polymorphism in
the inference also demonstrates the value of linked variations with
contrasted mutation dynamic to improve estimation of both demographic and
mutational parameters.