3.1 | Population genetic structure
The FST values were relatively low (ranging from 0 to 0.055) indicating a weak genetic structure between the different populations (Table S1). A dendrogram constructed from the FST (Figure S1) shows that the populations initially clustered in pairs. However, sampling location is not a hierarchical level of the population structure since only the populations sampled in location 1 in 2011 and 2013 clustered together. The populations in location 3 were sampled in the same plantations in 2011 and 2013 but the plantations had been replanted between the two years and the differentiation observed could thus result from genetic drift after recolonization by P. fijiensis . Absence of clear hierarchical structure reinforced our decision to consider data from samples collected in 2011 and 2013 as spatial replicates but not as time series.
The nucleotide diversity (π), Watterson’s theta (Θ) and Tajima’s D estimated for each population were similar in populations sampled in the same year (in 2011 or in 2013) with Pool-Seq data. In populations sampled in 2011, the median values of π and Θ ranged from 0.0001 to 0.0019 and from 0.0003 to 0.0017, respectively. In populations sampled in 2013, the median values of π and Θ ranged from 0.0049 to 0.0072 and from 0.0040 to 0.0061, respectively. The greater genetic diversity detected in populations sampled in 2013 is certainly due to the fact that more individuals were sampled in 2013. However, Tajima’s D estimates were similar in the two sampling years, ranging from -0.429 to 0.055 and -0.316 to -0.005 in populations sampled in 2011 and 2013, respectively. For individual sequencing data, the median values of π and Θ were 0.0005 for both populations and both indices. Tajima’s D median was 0.0006 for population CU1S2 and 0.0005 for population CU1R2. The estimated values for all the indexes were lower with the individual sequences, probably due to the smaller number of SNPs detected. The data obtained by sequencing individuals were also used to estimate linkage disequilibrium (LD) decay over the distance separating two SNPs in each population. The distance corresponding to a LD decay of 50% was around 4 Kb when all the scaffolds constituting the core genome ofP. fijiensis were taken into consideration (Figure S2). Overall, no difference in the LD decay pattern was observed between the two populations (CU1S2 and CU1S2). The association became random in sites located 500 kb apart whatever the population.