The utility of ASVs for community ecology
All the above conclusions about the predominant processes shaping community composition of soil microarthropods across the Troodos forests were very similar when based on OTUs or ASVs, contributing to the broader discussion about whether OTUs should be replaced by ASVs in metabarcoding studies (Callahan et al., 2017; Porter & Hajibabaei, 2020). As recent read-filtering methods have overcome the need of clustering to account for amplification and sequencing errors, the user-defined OTUs (traditionally used as proxies of species-level entities) could be redundant. However, the exclusive use of ASVs could affect the biological conclusions drawn from biodiversity analyses, as patterns of haplotypic diversity can reflect demographic attributes of populations, and do not always coincide with diversity patterns at the species level (Martin et al., 2021). In our system, while β diversity estimates at ASV and OTU levels were correlated and likely shaped by the same ecological processes as described above, we observed distinct α diversity (richness) patterns between them, which were explained by statistically significant differences in geographic or topoclimatic predictors. OTU richness per site was primarily explained by topoclimatic conditions, with assemblages hosting fewer OTUs as elevation and precipitation increased and temperature decreased (Table 1; Figure S3), thus following the general rule of declining species richness with increasing elevation (Rahbek, 1995), commonly interpreted as an outcome of environmental filtering driven by temperature or productivity gradients (Graham et al., 2014; Peters et al. 2016). In contrast to OTUs, ASV richness varied significantly along a longitudinal axis with local communities harbouring more haplotypes westwards (Table 1; Figure S3), which might be interpreted as a signature of higher on average intraspecific genetic diversity in the Western part of the mountain range, which has historically been less affected by anthropogenic disturbance (Delipetrou et al., 2008). This finding is in accordance with population genetic studies of forest trees that observed high genetic diversity in the western populations of Pinus brutia (Eliades, Aravanopoulos, & Christou, 2018) and Cedrus brevifolia (Eliades, Gailing, Lenemann, Fady, & Finkeldey, 2011). This is potentially a consequence of the local topography facilitating the maintenance of higher effective population sizes in this region during the Pleistocene climatic oscillations, and/or of less intensive historical human impact (livestock grazing and logging) than in Eastern Troodos (Eliades et al., 2018). The decline of haplotypic diversity in soil microarthropod assemblages eastwards may therefore indicate incipient biodiversity loss, as genetic variation tends to be eroded more quickly than species diversity under scenarios of global change (Balint et al., 2011). However, such differences between species and haplotype diversity were not reflected in patterns of community uniqueness, as our ASV- and OTU-based estimates of LCBD (local contribution to β diversity) were correlated and similarly explained by topoclimatic variation (Table 1; Figure S3), without any clear signature of historical contingencies, as those likely affecting longitudinal haplotypic richness patterns across Troodos. Our results therefore highlight the complementarity of OTUs and ASVs for community metabarcoding, as such side-by-side comparisons can help to detect processes that produce uncoupled patterns between the two levels of diversity (e.g., Reisch & Schmid, 2019).
Our ASV-level analyses were facilitated by the application of the metamate tool that utilised local and public reference sequence databases to discard non-authentic ASVs and retain only true biological sequence variants (Andújar et al., 2021). Although applying the most stringent filtering in metamate might have caused the removal of valuable rare biological haplotypes, appreciable intraspecific genetic variation (on average 2.35 ASVs per OTU) was still retrieved and produced reasonable haplotype diversity patterns as explained above. Based on our results, we advocate stringent ASV filtering, as it can provide informative datasets without compromising the required reliability for haplotype-level metabarcoding. Even if we cannot be fully confident that all erroneous haplotypes were filtered out, as the performance of the approach depends on the completeness of the reference sequence catalogue (Andújar et al., 2021), the future incorporation of intraspecific genetic data in local reference databases will provide further confidence and the opportunities for intra-OTU analyses of ASV variation (Elbrecht et al., 2018; Zizka, Weiss, & Leese, 2020).