Changes in beta diversity through time
In order to assess the beta diversity between surface and deep soils, plots having zero MOTUs in at least one depth were removed, corresponding to 4.28% (Bacteria), 14.28% (Eukaryota), 6.43% (Mycota), 39.28% (Collembola), 13.57% (Insecta) and 37.85% (Oligochaeta) of total plots.
GLMMs allowed us to detect changes in the beta-diversity of communities between surface and deep soil. Differences in community composition between the two depths decreased with time since glacier retreat for Bacteria, Eukaryota, Mycota and Insecta, indicating homogenization of communities, while we did not detect significant changes through time for the beta-diversity of Collembola and Oligochaeta (Fig. 3; Supplementary Table S5). Collembola and Oligochaeta were also the taxa for which the largest number of sites were discarded because of a lack of MOTUs. Overall, our models did not show significant changes in the turnover or nestedness components of the beta diversity measures through time, with the only exception of Oligochaeta, for which nestedness between surface and deep soils tended to increase through time (Supplementary Table S5 and Fig. S2).
Within each deglaciated foreland, the structure of communities was primarily related to time since glacier retreat (Fig. 4). Time significantly affected community structure for Bacteria, Mycota and Eukaryota (PERMANOVA: p< 0.05; Table 1); the amount of variance explained by time ranged from 2.4% to 5.7%. For Bacteria, Mycota, Eukaryota, as well as for Insecta, community structure also differed significantly between soil depths, but the explained variance was smaller (< 1%; Table 1). For none of the groups, we detected a significant interaction between time and soil depth (Table 1), suggesting that the effect of time was consistent between surface and deep soils. Differences in multivariate dispersions were never significant between soil depths, but were significant across time except for Collembola (Table 1). Bacterial community structure was the most strongly related to time and depth (R 2 = 5%; Table 1). Differences among deglaciated forelands were marked but tended to follow similar trends across the taxonomic groups (Fig. 4).
Based on the specificity and fidelity of each MOTU, 86 were identified as indicators (47 Bacteria, 34 Mycota and five Eukaryota; Table S6). For Bacteria, 22 taxa were strongly associated with young foreland soils, including members of the genera Roseiflexus ,Herbaspirillum , Novosphingobium that exhibited particularly high IndVal, while no one was strictly associated to the intermediate ages. Seventeen taxa of Bacteria were indicators of both surface and deep soil layers in older forelands, including members from the genera Actinoallomurus and Ferrimicrobium that showed the highest IndVal. Six taxa were indicators of the deep soil layers at both intermediate and old age. For Eukaryota, five taxa were considered as indicators; three were fungi related to old soils, while one mite (genus Gamasina ) was associated with the intermediate age class. For Mycota, 18 taxa were indicators of both surface and deep layers in older forelands, including members of the genus Cladophialophoraand the family Glomeraceae. Ten Mycota taxa were indicators of both surface and deep layers in young forelands while intermediate forelands contained less indicators, with only five taxa. Only one Mycota taxon was representative of a specific soil layer (the MOTU identified as Golovinomyces sordidus , associated to the surface layer of young forelands).