Discussion
In this study, we used a distinct transition zone in vegetation and soil moisture, as the framework to analyze how different OTU generation methods affect the detection of a shift in the composition of microeukaryotic soil communites. Interestingly, the sharp transition in plant community, with lower richness in wet compared to mesic-dry soils, was not associated with a difference in observed richness for the corresponding microeukaryotic soil communities. However, both above and belowground community composition were significantly different in wet and mesic-dry soil moisture regimes. We demonstrate that different OTU generation methods applied to the same long amplicon eDNA dataset affect the documented composition of soil microeukaryotic communities. Similarly, earlier studies have reported that large scale ecological patterns are recovered irrespective of the OTU or ASV generation method or clustering threshold (for short read data) (Glassman & Martiny, 2018), sequencing technology (Furneaux et al., 2021) or sampling effort (Castle et al., 2019). We conclude that large scale ecological patterns are robustly recovered irrespective of the OTU or ASV generation method applied. However, for studies focused on the particular members of these contrasting communities, the OTU generation method selected significantly affects the phylogenetic resolution and detection of taxa. For instance, our results show that inference of ASVs with DADA2 (here OTU_A) captures less than 30% of all reads, providing information on intra-species genetic variation only for abundant taxa while rare taxa, including entire phylum-level lineages, remain undetected. The over-all estaimated OTU richness was also lower for OTU_A compared to the cluster-based methods. When comparing OTU generation methods, others have found contrasting patterns, with higher richness captured with ASVs, the equvivalent to OTU_A in this study, compared to clustering (Glassman & Martiny, 2018). Differences between our results and those of Glassman and Martiny (2018) can be attributed to the earlier study´s shorter amplicon (only ITS2 for fungi) and sequening depth generated by Illumina, compared to our long amplicon sequencing with lower depth using PacBio. In studies using short read amplicons, denoising increased overall richness by capturing intraspecies genetic variation (Callahan et al., 2016). However, when applied to long read amplicons from diverse communities, intra-species variation can only be captured for the most abundant taxa. While OTU accumulation curves saturated for all methods, we found that increasing the number of samples would have increased the number of detected taxa at the site. Due to soil heterogeneity and spatial community turnover, increasing the number of samples rather than the sequenincing depth increases the estimated alpha diversity even in well-mixed, managed agricultural soils (Castle et al., 2019). The same pattern was previously observed in forst soils from West Africa (Meidl et al., 2021), highlighting the importance of optimizing sampling effort versus sequencing depth to obtain a good representation of the alpha diversity.
Single-linkage clustering with a distance threshold of 2%, on the other hand, captures most reads in OTU_Ss that correspond closely to broadly accepted fungal species-level sequence similarity across the ITS region, suggesting that this method provides an acceptable proxy for species richness. We anticipate that phylogenetic resolution of species and genus-level relationships could have been improved by the generation of a hybrid tree that included ITS2 alignments to resolve relationships within each GH_90 lineage, in a manner similar to ghost-tree (Fouquier et al., 2016). Such tree could have been used to generate phylogenetic species hypotheses (ref) to analyze community composition and generate species richness estimates for these communities. Apart from sequence clustering approaches, extraction and amplification biases remains as a major filtering step for analysis of total microeukaryptic soil communites. Althought the primers we used have no known biases against Glomeromycota, we obtained low read abundance for this group, despite known high abundance of Gomeromycota spores at the site. This apparent contradiction may be explained by the low copy number of around ten rDNA operones in this phylum (Maeda et al., 2018) compared to other fungi that may harbor hundreds to thousands of copies (Lofgren et al., 2019). In addition to copy number variation, length difference in the rDNA, especially ITS, can introduce bias both during PCR and sequencing (Tedersoo et al., 2015), rendering this type of data far from quantitative, especially when applied to broad phylogenetically groups such as micoeukaryotes.
Our study also provides a first insight into the belowground diversity of a meadow known for its rich plant community (Sernander, 1948; Zhang, 1983; Zhang & Hytteborn, 1985). Studies that aim to simultaneously characterize communities of both protists and fungi have often found that fungi dominate the sequenced microeukaryotic communities, e.g., in tropical forest soil (Tedersoo et al., 2018) and soils from different habitats in temperate regions (Tedersoo & Anslan, 2019). In previous studies using the exact same primers, sequenced soil communities from ectomycorrhizal dominated forests in Sweden and West Africa have been almost completely dominated by reads taxonomically assigned to kingdom Fungi (Furneaux et al., 2021; Kalsoom Khan et al., 2020; Meidl et al., 2021). The dominance of protists in the sequenced microeukaryotic community indicates that these soil systems are particularly suitable for diverse communities of protists. High soil moisture may be one explanation, but other factors like plant community, pH and total nitrogen have also been associated with high abundance of protists in soil (Oliverio et al., 2020). We anticipate that future studies may hold many interesting discoveries of hitherto unknown diversity at this site.