Genotypes and mitochondrial haplotypes
We successfully genotyped 7 to 14 microsatellite markers from of 324 samples (Table S1). Although the Bayesian clustering of individuals analysis (Npop=1) resolved to 31 genetic clusters (k=31, p=0.999, Appendix 1), both the spatial Bayesian clustering analysis of five localities (NPop=5, k=4, p=0,999; Appendix 2), and the spatial Bayesian clustering analysis of 12 spatio-temporal groups (Npop=12), detected only four genetically differentiated population clusters (Npop=12, k=4, p=0.999; Appendix 3). Additionally, spatial Bayesian analyses including all genotyped female individuals based on both the five localities and the 12 spatio-temporal groups, suggest restricted gene flow between the four clusters, and across generations.
Both the five localities and the 12 spatio-temporal groups analyses suggest that Sottunga and Seglinge-Kumlinge are genetically similar to one another (Appendix 3). In the Bayesian clustering of localities (NPop=5) analysis, Sottunga and Seglinge-Kumlinge populations made up one cluster (FST=0.014, p=0.042, Table 3), and Föglö, Finström and Saltvik each made up the three other clusters (Figure 1c). Although the clustering analysis of the 12 spatio-temporal groups in general support these same clusters, it also reveals that the early samples from Föglö and Finström, and all samples from Saltvik, group in the same cluster (red). The two island populations (Sottunga and Seglinge-Kumlinge) are always found to be genetically different from Föglö (FST=0.113, p=0.001, Table 3). The shared ancestry of the Sottunga and Finström samples is visible through detailed screening of the individual cluster dataset (Npop=1, Appendix 1). The samples collected in Sottunga belong to 13 genotypes, five of which (genotypes #9, 11, 14, 17 and 26) also include samples from Finström collected in 1992 and 1993. These same five genotypes include most of the samples from Kumlinge-Seglinge (N=32/42, 76%), and are also found in other localities (Appendix 1), suggesting shared ancestry within each of the populations.
In general, the admixture analyses revealed little gene flow and interbreeding between the four clusters (Figure 3). We observe some gene flow between Föglö and Sottunga, between Föglö and Saltvik (possibly as a result of dispersal through several generations, or stepping stone events), as well as between the two mainland populations of Saltvik and Finström (as a result of dispersal between the two neighbouring regions) (Figure 3). Additionally, consistent with Couchoux et al. (2016), inbreeding occurs in Åland at similarly low levels in each of the five localities, despite the mainland populations being of more connected, and of larger sizes. The inbreeding coefficient (Fis ), generally about 0,222 across Åland, ranges from 0,174 in Saltvik, and 0,258 in Föglö (Table 3); while reaching 0,206 in Sottunga.
The parasitoid H. horticola in Åland has experienced both global and local population crashes through the years (Figure 2), some of which resulted in detectable changes in local genetic structure. The genotype characterizing Föglö after the population crash of 2010 (Figure 1) is from a different genetic cluster than any of the three other clusters characterized from the other four populations. In contrast, other local crashes, for example in 1999 and 2006 in Finström, and in 1999 in Saltvik, did not affect the genotypic clusters of these populations. Unfortunately, we lack data to directly address the impact of bottlenecks that occurred in both Seglinge-Kumlinge and Sottunga.
Finally, in agreement with a previous study by Duplouy et al. (2015), the majority of the H. horticola parasitoids carry the C-mitotype (N=145, 67%), while the remaining specimens carry the T-mitotype (Table 1). Noticeably, the C-mitotype is prevalent in all localities (57% in Saltvik, 71% in Sottunga, 84% in Finström, 95% in Föglö) except Seglinge-Kumlinge (28%) (Table 1). These general patterns hold across the different spatio-temporal groups (Table 2).
Wolbachia infection status
Wolbachia was detected in samples from all five localities. The mean infection rate across our samples is 50%, which is consistent with previous work from Duplouy et al. (2015) across the entire Åland archipelago. Among the localities, Seglinge-Kumlinge shows the highest infection rate (95%), greatly contrasting with the lower infection rates of Finström (42%), Föglö (32%), Saltvik (59%), and Sottunga (23%) (Table 1).
Among all Wolbachia -infected wasp sampled for the present study, 57% carry the T-mitotype, which is also consistent with the intermediate infection rate from across the entire Åland archipelago, as characterized by Duplouy et al. (2015). However, the proportion of infected wasps carrying the T-mitotype differed among localities. In Saltvik, Finström and Föglö, only 60%, 33% and 0% ofWolbachia -infected wasps carry the T-mitotype, respectively. In contrast, despite the two islands showing contrasting proportions of specimens carrying the T-mitotype (see above), Sottunga and Seglinge-Kumlinge are once again more similar to each other than to any other population, with the majority of Wolbachia- infected wasps carrying the T-mitotype (75% and 79%, respectively) rather than the C-mitotype. Finally, the four most common genotypes in Seglinge-Kumlinge (Genotype -5 (N=13); -6 (N=6); -9 (N=6); and -17 (N=6)) are found in strict associations with Wolbachia- infected wasps, but all mainly carry the T-mitotype (mean of 77%; genotype -5 (54%), -6 (83%), -9 (100%), -17 (100%), respectively). This contrasts with the rest of Åland, including Sottunga, where these four genotypes are not strongly associated with Wolbachia , and more rarely in association with the T-mitotype than with the C-mitotype (mean of 21%; genotype -5 (11%), -6 (0%), -9 (29%), -17 (50%)). All other genotypes in the Seglinge-Kumlinge sample are represented by less than 3 samples, and thus it is difficult to determine patterns from them.