Flowers show important structural variation as reproductive organs but the evolutionary forces underlying this diversity are still poorly understood. In animal-pollinated species, flower shape is strongly fashioned by selection imposed by pollinators, which is expected to vary according to guilds of effective pollinators. Using the Antillean subtribe Gesneriinae (Gesneriaceae), we tested the hypothesis that pollination specialists pollinated by one functional type of pollinator have maintained more similar corolla shapes through time due to more constant and stronger selection constraints compared to species with more generalist pollination strategies. Using geometric morphometrics and evolutionary models, we showed that the corolla of hummingbird specialists, bat specialists, and species with a mixed-pollination strategy (pollinated by hummingbirds and bats; thus a more generalist strategy) have distinct shapes and that these shapes have evolved under evolutionary constraints. However, we did not find support for greater disparity in corolla shape of more generalist species. This could be because the corolla shape of more generalist species in subtribe Gesneriinae, which has evolved multiple times, is finely adapted to be effectively pollinated by both bats and hummingbirds. These results suggest that ecological generalization is not necessarily associated with relaxed selection constraints.

Most forests worldwide are renewed by natural processes, the exception being artificially planted woodlands, which account for 7% of the world's forests (FAO, 2015). The remaining primary and secondary forests are regenerated by means of seedlings resulting from open pollination between mature trees. Natural processes dominated by competition for light, water and nutrients, and environmental disturbances, exert a strong influence during the early stages of development of the renewed forest stand. Human activities, in the form of silvicultural regimes of various intensities, depending on tree species and country, subsequently superimpose over natural selection regimes, except in primary forests. Thus, most existing forests are subjected to both natural and human-mediated

Anticipating the evolutionary responses of long-lived organisms, such as trees, to environmental changes, requires the assessment of genetic variation of adaptive traits in natural populations. To this end, high-density markers are needed to calculate genomic relatedness between individuals allowing to estimate the genetic variance of traits in wild populations. We designed a targeted capture-based, next-generation sequencing assay based on the highly heterozygous pedunculate oak (Quercus robur) reference genome, for the sequencing of 3 Mb of genic and intergenic regions. Using a mixed stand of 293 Q. robur and Q. petraea genotypes we successfully captured over 97% of the target sequences, corresponding to 0.39% of the oak genome, with sufficient depth (97×) for the detection of about 190,000 SNPs evenly spread over the targeted regions. We validated the technique by evaluating its reproducibility, and comparing the genomic relatedness of trees with their known pedigree relationship. We explored the use of the technique on other related species and highlighted the advantages and limitations of this approach. We found that 92.07% of target sequences in Q. suber and 70.36% of sequences in Fagus sylvatica were captured. We used this SNP resource to estimate genetic relatedness in the mixed oak stand. Mean pairwise genetic relatedness was low within each species with a few values exceeding 0.25 (half-sibs) or 0.5 (full-sibs). Finally, we applied the technique to a long-standing issue in population genetics of trees regarding the relationship between inbreeding and components of fitness. We found very weak signals for inbreeding depression for reproductive success and no signal for growth within both species.