4 DISCUSSION

Why did the protists not obey the rule of increased species richness with increasing habitat diversity? Our findings show how discrete microhabitat niches lead to compositional heterogeneity of microbial communities within tree canopies, and ultimately within whole ecosystems (Peay, Kennedy, & Talbot, 2016; Cregger et al. , 2018). Application of taxon-specific primers ensured an exhaustive coverage of the investigated protistan taxa, a crucial precondition to minimise the number of taxa absent due to undersampling (Supplementary Figure 2). Furthermore, a study comparing the outputs of general and group specific primers has shown that structuring effects of the environment on protistan communities could only be seen in the data sets generated with specific primers (Lentendu et al ., 2014). With data obtained this way, we could report a further increase in the diversity of investigated protistan taxa, as it was already demonstrated by Fiore-Donno et al. (2018). Thus, the majority of detected OTUs account for a large amount of undescribed diversity (Figure 1) and may represent so far uncharacterised lineages, especially in the taxon of oomycetes, as only 34% of the OTUs were 97%–100% similar to any known sequence.
A world‐wide survey on forest litter to investigate patterns of diversity in a group of testate amoebae observed that this common group of terrestrial protists behave like macroscopic organisms. Documented community structures were strongly correlated with climatic and physicochemical parameters but also with geographical barriers (Laraet al. , 2015). We observed Canopy communities to be strikingly different in terms of beta diversity compared to litter and soil communities on the ground, and as expected, different microhabitats within tree canopies were further colonized by distinct protistan communities, especially by Cercozoa (Figure 3A). However, the hypothesised rule of increased species richness with increasing habitat diversity could not be confirmed due to the almost ubiquitous distribution of protistan taxa (Figure 5). Our data exemplify that microorganisms do not always obey ecological rules assigned to multicellular organisms. Moreover, our findings are in contrast to patterns observed for bacteria (Lundberg et al. , 2012; Ottesenet al. , 2013; Wagner et al. , 2016), epifoliar fungi (Gilbert, Reynolds, & Bethancourt, 2007) or lichens (Boch et al. , 2013; Marmor et al. , 2013). Most of these groups are directly dependent on specific environmental conditions or the availability of resources offered by the habitat (Vandenkoornhuyseet al. , 2015; Sasse, Martinoia, & Northen, 2018). However, high turnover rates and the ability of protists to form cysts as resting stages imposes less constraints on their distribution over large distances due to a reduced likelihood of isolation and high diversity of source pools for local community assembly (Fenchel & Finlay, 2004; Bahram et al. , 2016). Accordingly, habitat diversity strongly favoured certain protistan taxa in terms of relative abundance, but the effect of OTU richness on community composition was negligible due to their almost ubiquitous distribution.
Mahé et al. (2017) hypothesised that protists in soils may be a subset from the canopy that have rained down from above; a pattern confirmed for leaf endophytic fungi in one-year old beech litter of temperate forests (Guerreiro et al. , 2018). On the other hand, there is growing evidence of cercozoan species particularly adapted to life in the phyllosphere (Dumack et al. , 2017; Flues, Bass, & Bonkowski, 2017, Sapp et al. , 2018). Cercozoan phyllosphere communities, isolated from fresh canopy leaves, were overall indeed surprisingly similar to leaf litter communities on the ground (Figure 3A). However, this cannot be confirmed for protists in general, as differences in oomycete communities between phyllosphere and ground litter showed strikingly different patterns. While fresh canopy leaves and ground litter had highest OTU richness of Cercozoa (Figure 5A), ground litter contained a significantly depleted diversity of oomycetes (Figure 2B). Interesting were the resembling patterns of beta diversity between phyllosphere and deadwood Cercozoa and oomycetes which deserve further attention in future studies. The small, but significant differences of oomycete communities between tree species (Figure 4B, Supplementary Table 5) might be explained by differences in host specificity, since oomycetes are well known to contain specific pathogens infecting leaves, stems and roots of forest trees (e.g. Rizzo & Garbelotto, 2003; Lehtijärvi et al. , 2017). Apparently, tree species do not shape their associated protistan communities to the same degree as bacteria, where even different genotypes of the same tree species can show distinct spatial patterns in their colonizing bacterial communities (Redford et al. , 2010; Leff et al. , 2015; Cregger et al. , 2018). Heterotrophic (or organotrophic) bacteria experience a direct selection pressure by differences in nutritional resource composition between plant microhabitats (Thapa et al. , 2017), while the proportion of bacterivorous Cercozoa at the next trophic level lack this direct dependency on the host-tree species (Figure 4A).
Compared to the highly specific bacterial communities of tree bark, mosses and lichens (Aschenbrenner et al. , 2017), canopy protists appear to rather depend on microhabitat characteristics. This is best exemplified within the cryptogamic epiphytes. Lichen and the two moss taxa harboured quite similar protistan communities (Figure 3). These epiphytes are characterised by rapidly changing conditions with rapid swelling and storage of moisture from morning dew and after rainfall to severe dryness at sunshine (e.g. Jonsson et al. , 2014; Benítezet al. , 2018) and to a certain degree may act as environmental filters selecting specific protistan communities.
Protistan communities of arboreal soil samples showed high variability, spanning from moss-like communities to soil-like communities (Figure 3). Importantly, this indicates that protistan communities resembling those of mineral soil are not restricted to the forest floor. Community variability in arboreal soil might be due to the varying degree of decay of the sampled material and its distinct physico-chemical properties (Nadkarni et al. , 2002), and further strengthens our hypothesis that increasing habitat richness may result in increasing compositional heterogeneity of protistan communities.

Conclusions

Beta diversity of protists was solely driven by differences in the relative abundance of OTUs, because almost all taxa were ubiquitously distributed among tree crowns and soil of the floodplain forest. Accordingly, species richness did not increase with habitat diversity as hypothesized, although strong differences in beta diversity between protistan communities of the forest floor and tree crowns, and among microhabitats within tree crowns, demonstrate strong differences in relative abundance. Different tree species had a surprisingly low influence on protistan community assembly; even the mostly plant-parasitic oomycetes did not show a high degree of host-specificity. However, a high number of OTUs from canopy communities could not be assigned to any known sequence, giving evidence that protistan communities of tree canopies are largely understudied. Both strata show unique protistan communities, indicating no top-down relationship of investigated protistan taxa in trees. Thus, our findings illustrate that the diversity of soil protists is solely shaped by the habitat itself, from which no conclusions regarding the total diversity of the canopy can be drawn. The occurrence of only a few specialist OTUs does not imply functional homogenization at the community level across microhabitats, but rather indicates increasing functional diversity to a greater extent than increasing OTU richness with increasing habitat diversity (Ofek-Lalzar et al. , 2014). Future studies will be needed to further address this hypothesis, supplementing functional traits to the taxonomical assignment of the investigated protistan OTUs.