Summary: Factors shaping community patterns of microorganisms are controversially discussed. Physical and chemical factors certainly limit the survival of individual taxa and maintenance of diversity. In recent years, a contribution of geographic distance and dispersal barriers to distribution patterns of protists and bacteria has been demonstrated. Organismic interactions such as competition, predation and mutualism further modify community structure and maintenance of distinct taxa. Here, we address the relative importance of these different factors in shaping protists and bacterial communities on a European scale using high‐throughput sequencing data obtained from lentic freshwater ecosystems. We show that community patterns of protists are similar to those of bacteria. Our results indicate that cross‐domain organismic factors are important variables with a higher influence on protists as compared with bacteria. Abiotic physical and chemical factors also contributed significantly to community patterns. The contribution of these latter factors was higher for bacteria, which may reflect a stronger biogeochemical coupling. The contribution of geographical distance was similar for both microbial groups.

High-throughput sequencing technologies are lifting major limitations to molecular-based ecological studies of eukaryotic microbial diversity, but analyses of the resulting millions of short sequences remain a major bottleneck for these approaches. Here, we introduce the analytical and statistical framework of sequence similarity networks, increasingly used in evolutionary studies and graph theory, into the field of ecology to analyze novel pyrosequenced V4 small subunit rDNA (SSU-rDNA) sequence data sets in the context of previous studies, including SSU-rDNA Sanger sequence data from cultured ciliates and from previous environmental diversity inventories. Our broadly applicable protocol quantified the progress in the description of genetic diversity of ciliates by environmental SSU-rDNA surveys, detected a fundamental historical bias in the tendency to recover already known groups in these surveys, and revealed substantial amounts of hidden microbial diversity. Moreover, network measures demonstrated that ciliates are not globally dispersed, but are structured by habitat and geographical location at intermediate geographical scale, as observed for bacteria, plants, and animals. Currently available ‘universal’ primers used for local in-depth sequencing surveys provide little hope to exhaust the significantly higher ciliate (and most likely microbial) diversity than previously thought. Network analyses such as presented in this study offer a promising way to guide the design of novel primers and to further explore this vast and structured microbial diversity.

Network analyses of biological communities allow for identifying potential consequences of climate change on the resilience of ecosystems and their robustness to resist stressors. Using DNA metabarcoding datasets from a three-year-sampling (73 samples), we constructed the protistan plankton co-occurrence network of Lake Zurich, a model lake ecosystem subjected to climate change. Despite several documentations of dramatic lake warming in Lake Zurich, our study provides an unprecedented perspective by linking changes in biotic association patterns to climate stress. Water temperature belonged to the strongest environmental parameters splitting the data into two distinct seasonal networks (October–April; May–September). The expected ecological niche of phytoplankton, weakened through nutrient depletion because of permanent thermal stratification and through parasitic fungi, was occupied by the cyanobacterium Planktothrix rubescens and mixotrophic nanoflagellates. Instead of phytoplankton, bacteria and nanoflagellates were the main prey organisms associated with key predators (ciliates), which contrasts traditional views of biological associations in lake plankton. In a species extinction scenario, the warm season network emerged as more vulnerable than the cold season network, indicating a time-lagged effect of warmer winter temperatures on the communities. We conclude that climate stressors compromise lake ecosystem robustness and resilience through species replacement, richness differences, and succession as indicated by key network properties.

Discovery of novel diversity in high-throughput sequencing studies is an important aspect in environmental microbial ecology. To evaluate the effects that amplicon clustering methods have on the discovery of novel diversity, we clustered an environmental marine high-throughput sequencing dataset of protist amplicons together with reference sequences from the taxonomically curated Protist Ribosomal Reference (PR²) database using threede novoapproaches: sequence similarity networks, USEARCH, and Swarm. The potentially novel diversity uncovered by each clustering approach differed drastically in the number of operational taxonomic units (OTUs) and in the number of environmental amplicons in these novel diversity OTUs. Global pairwise alignment comparisons revealed that numerous amplicons classified as potentially novel by USEARCH and Swarm were more than 97% similar to references of PR². Using shortest path analyses on sequence similarity network OTUs and Swarm OTUs we found additional novel diversity within OTUs that would have gone unnoticed without further exploiting their underlying network topologies. These results demonstrate that graph theory provides powerful tools for microbial ecology and the analysis of environmental high-throughput sequencing datasets. Furthermore, sequence similarity networks were most accurate in delineating novel diversity from previously discovered diversity.

Microbial planktonic communities are the basis of food webs in aquatic ecosystems since they contribute substantially to primary production and nutrient recycling. Network analyses of DNA metabarcoding datasets emerged as a powerful tool to untangle the complex ecological relationships among the key players in food webs. In this study, we evaluated co‐occurrence networks constructed from time‐series metabarcoding datasets (12 months, biweekly sampling) of protistan plankton communities in surface layers (epilimnion) and bottom waters (hypolimnion) of two temperate deep lakes, Lake Mondsee (Austria) and Lake Zurich (Switzerland). Lake Zurich plankton communities were less tightly connected, more fragmented and had a higher susceptibility to a species extinction scenario compared to Lake Mondsee communities. We interpret these results as a lower robustness of Lake Zurich protistan plankton to environmental stressors, especially stressors resulting from climate change. In all networks, the phylum Ciliophora contributed the highest number of nodes, among them several in key positions of the networks. Associations in ciliate‐specific ^subnetworks resembled autecological species‐specific traits that indicate adaptions to specific environmental conditions. We demonstrate the strength of co‐occurrence network analyses to deepen our understanding of plankton community dynamics in lakes and indicate biotic relationships, which resulted in new hypotheses that may guide future research in climate‐stressed ecosystems.

Operational taxonomic units (OTUs) are usually treated as if they are internally uniform in environmental metabarcoding studies of microbial and macrobial eukaryotes, even when the OTUs are being used to infer biogeographic patterns. The OTUs constructed by the program Swarm have underlying network topologies in which nodes represent amplicons and edges represent 1 nucleotide differences between nodes. Such networks can be exploited to search for biogeographic patterns within each OTU. To do this, here we used an available protistan metabarcoding dataset consisting of the hypervariable V4 region of the 18S rRNA locus amplified from 77 lakes collected across Norway and Sweden. The 82 most abundant and wide-spread OTUs constructed by Swarm were evaluated using shortest path, assortativity, and geographical analyses. We found that while pairs of amplicons from the same lake were usually connected directly to each other within the OTUs, these pairs of amplicons from the same lake did not form assortative clusters within the OTUs, and amplicons were not more connected with other amplicons occurring in neighboring lakes than expected by chance. This new approach to looking at within-OTU is applicable to other metabarcoding datasets and we provide code to perform these analyses.