DF-Phos: Prediction of Protein Phosphorylation Sites by Deep Forest.

Phosphorylation is the most important and studied post-translational modification (PTM), which plays a crucial role in protein function studies and experimental design. Many significant studies have been performed to predict phosphorylation sites using various machine-learning methods. Recently, several studies have claimed that deep learning-based methods are the best way to predict the phosphorylation sites because deep learning as an advanced machine learning method can automatically detect complex representations of phosphorylation patterns from raw sequences and thus offers a powerful tool to improve phosphorylation site prediction. In this study, we report DF-Phos, a new phosphosite predictor based on the Deep Forest to predict phosphorylation sites. In DF-Phos, the feature vector taken from the CkSAApair method is as input for a Deep Forest framework for predicting phosphorylation sites. The results of 10-fold cross-validation show that the Deep Forest method has the highest performance among other available methods. We implemented a Python program of DF-Phos, which is freely available for non-commercial use at https://github.com/zahiriz/DF-Phos Moreover, users can use it for various PTM predictions.

Challenges in supplying empirical proof for predictions derived from Species Distribution Models (SDMs): the case of an invasive cyanobacterium.

Species distribution models (SDMs) calibrated with bioclimatic variables revealed a high probability for range expansion of the invasive toxin producing cyanobacterium, Raphidiopsis raciborskii to Sweden, where no reports of its presence have hitherto been recorded. While predictions focused on the importance of climate variables for possible invasion, other barriers to dispersal and successful colonization need to be overcome by the species for successful invasion. In this study, we combine field-based surveys of R. raciborskii (microscopy and molecular analysis using species-specific primers) of 11 Swedish lakes and in-silico screening of environmental DNA using 153 metagenomic datasets from lakes across Europe to validate the SDMs prediction. Field-based studies in lakes with high/low predicted probability of occurrence did not detect the presence of R. raciborskii, and in-silico screening only detected hints of its presence in 5 metagenomes from lakes with probability ranging from 0.059 to 0.825. The inconsistencies between SDMs results and both field-based/in-silico monitoring could be due to either sensitivity of monitoring approaches in detecting early invasions or uncertainties in SDMs that focused solely on climate drivers. However, results highlight the necessity of proactive monitoring with high temporal and spatial frequency.

Large-scale phylogenomics of aquatic bacteria reveal molecular mechanisms for adaptation to salinity.

The crossing of environmental barriers poses major adaptive challenges. Rareness of freshwater-marine transitions separates the bacterial communities, but how these are related to brackish counterparts remains elusive, as do the molecular adaptations facilitating cross-biome transitions. We conducted large-scale phylogenomic analysis of freshwater, brackish, and marine quality-filtered metagenome-assembled genomes (11,248). Average nucleotide identity analyses showed that bacterial species rarely existed in multiple biomes. In contrast, distinct brackish basins cohosted numerous species, but their intraspecific population structures displayed clear signs of geographic separation. We further identified the most recent cross-biome transitions, which were rare, ancient, and most commonly directed toward the brackish biome. Transitions were accompanied by systematic changes in amino acid composition and isoelectric point distributions of inferred proteomes, which evolved over millions of years, as well as convergent gains or losses of specific gene functions. Therefore, adaptive challenges entailing proteome reorganization and specific changes in gene content constrains the cross-biome transitions, resulting in species-level separation between aquatic biomes.

Hypersaline Lake Urmia: a potential hotspot for microbial genomic variation.

Lake Urmia located in Iran is a hypersaline environment with a salinity of about 27% (w/v). Metagenomic analyses of water samples collected from six locations in the lake exhibited a microbial community dominated by representatives of the family Haloferacaceae (69.8%), mainly those affiliated to only two genera, Haloquadratum (59.3%) and Halonotius (9.1%). Similar to other hypersaline lakes, the bacterial community was dominated by Salinibacter ruber (23.3%). Genomic variation analysis by inspecting single nucleotide variations (SNVs) and insertions/deletions (INDELs) exhibited a high level of SNVs and insertions, most likely through transformation for abundant taxa in the Lake Urmia community. We suggest that the extreme conditions of Lake Urmia and specifically its high ionic concentrations could potentially increase the SNVs and insertions, which can consequently hamper the assembly and genome reconstruction from metagenomic reads of Lake Urmia.

High-resolution metagenomic reconstruction of the freshwater spring bloom.

BACKGROUND: The phytoplankton spring bloom in freshwater habitats is a complex, recurring, and dynamic ecological spectacle that unfolds at multiple biological scales. Although enormous taxonomic shifts in microbial assemblages during and after the bloom have been reported, genomic information on the microbial community of the spring bloom remains scarce. RESULTS: We performed a high-resolution spatio-temporal sampling of the spring bloom in a freshwater reservoir and describe a multitude of previously unknown taxa using metagenome-assembled genomes of eukaryotes, prokaryotes, and viruses in combination with a broad array of methodologies. The recovered genomes reveal multiple distributional dynamics for several bacterial groups with progressively increasing stratification. Analyses of abundances of metagenome-assembled genomes in concert with CARD-FISH revealed remarkably similar in situ doubling time estimates for dominant genome-streamlined microbial lineages. Discordance between quantitations of cryptophytes arising from sequence data and microscopic identification suggested the presence of hidden, yet extremely abundant aplastidic cryptophytes that were confirmed by CARD-FISH analyses. Aplastidic cryptophytes are prevalent throughout the water column but have never been considered in prior models of plankton dynamics. We also recovered the first metagenomic-assembled genomes of freshwater protists (a diatom and a haptophyte) along with thousands of giant viral genomic contigs, some of which appeared similar to viruses infecting haptophytes but owing to lack of known representatives, most remained without any indication of their hosts. The contrasting distribution of giant viruses that are present in the entire water column to that of parasitic perkinsids residing largely in deeper waters allows us to propose giant viruses as the biological agents of top-down control and bloom collapse, likely in combination with bottom-up factors like a nutrient limitation. CONCLUSION: We reconstructed thousands of genomes of microbes and viruses from a freshwater spring bloom and show that such large-scale genome recovery allows tracking of planktonic succession in great detail. However, integration of metagenomic information with other methodologies (e.g., microscopy, CARD-FISH) remains critical to reveal diverse phenomena (e.g., distributional patterns, in situ doubling times) and novel participants (e.g., aplastidic cryptophytes) and to further refine existing ecological models (e.g., factors affecting bloom collapse). This work provides a genomic foundation for future approaches towards a fine-scale characterization of the organisms in relation to the rapidly changing environment during the course of the freshwater spring bloom. Video Abstract.

Genome-resolved insight into the reservoir of antibiotic resistance genes in aquatic microbial community.

Aquatic microbial communities are an important reservoir of antibiotic resistance genes (ARGs). However, distribution and diversity of different ARG categories in environmental microbes with different ecological strategies is not yet well studied. Despite the potential exposure of the southern part of the Caspian Sea to the release of antibiotics, little is known about its natural resistome profile. We used a combination of Hidden Markov model (HMM), homology alignment and a deep learning approach for comprehensive screening of the diversity and distribution of ARGs in the Caspian Sea metagenomes at genome resolution. Detected ARGs were classified into five antibiotic resistance categories including prevention of access to target (44%), modification/protection of targets (30%), direct modification of antibiotics (22%), stress resistance (3%), and metal resistance (1%). The 102 detected ARG containing metagenome-assembled genomes of the Caspian Sea were dominated by representatives of Acidimicrobiia, Gammaproteobacteria, and Actinobacteria classes. Comparative analysis revealed that the highly abundant, oligotrophic, and genome streamlined representatives of taxa Acidimicrobiia and Actinobacteria modify the antibiotic target via mutation to develop antibiotic resistance rather than carrying extra resistance genes. Our results help with understanding how the encoded resistance categories of each genome are aligned with its ecological strategies.

Genome-resolved analyses show an extensive diversification in key aerobic hydrocarbon-degrading enzymes across bacteria and archaea.

BACKGROUND: Hydrocarbons (HCs) are organic compounds composed solely of carbon and hydrogen that are mainly accumulated in oil reservoirs. As the introduction of all classes of hydrocarbons including crude oil and oil products into the environment has increased significantly, oil pollution has become a global ecological problem. However, our perception of pathways for biotic degradation of major HCs and key enzymes in these bioconversion processes has mainly been based on cultured microbes and is biased by uneven taxonomic representation. Here we used Annotree to provide a gene-centric view of the aerobic degradation ability of aliphatic and aromatic HCs in 23,446 genomes from 123 bacterial and 14 archaeal phyla.  RESULTS: Apart from the widespread genetic potential for HC degradation in Proteobacteria, Actinobacteriota, Bacteroidota, and Firmicutes, genomes from an additional 18 bacterial and 3 archaeal phyla also hosted key HC degrading enzymes. Among these, such degradation potential has not been previously reported for representatives in the phyla UBA8248, Tectomicrobia, SAR324, and Eremiobacterota. Genomes containing whole pathways for complete degradation of HCs were only detected in Proteobacteria and Actinobacteriota. Except for several members of Crenarchaeota, Halobacterota, and Nanoarchaeota that have tmoA, ladA, and alkB/M key genes, respectively, representatives of archaeal genomes made a small contribution to HC degradation. None of the screened archaeal genomes coded for complete HC degradation pathways studied here; however, they contribute significantly to peripheral routes of HC degradation with bacteria. CONCLUSION: Phylogeny reconstruction showed that the reservoir of key aerobic hydrocarbon-degrading enzymes in Bacteria and Archaea undergoes extensive diversification via gene duplication and horizontal gene transfer. This diversification could potentially enable microbes to rapidly adapt to novel and manufactured HCs that reach the environment.

mOTUpan: a robust Bayesian approach to leverage metagenome-assembled genomes for core-genome estimation.

Recent advances in sequencing and bioinformatics have expanded the tree of life by providing genomes for uncultured environmentally relevant clades, either through metagenome-assembled genomes or through single-cell genomes. While this expanded diversity can provide novel insights into microbial population structure, most tools available for core-genome estimation are sensitive to genome completeness. Consequently, a major portion of the huge phylogenetic diversity uncovered by environmental genomic approaches remains excluded from such analyses. We present mOTUpan, a novel iterative Bayesian method for computing the core genome for sets of genomes of highly diverse completeness range. The likelihood for each gene cluster to belong to core or accessory genome is estimated by computing the probability of its presence/absence pattern in the target genome set. The core-genome prediction is computationally efficient and can be scaled up to thousands of genomes. It has shown comparable estimates to state-of-the-art tools Roary and PPanGGOLiN for high-quality genomes and is capable of using genomes at lower completeness thresholds. mOTUpan wraps a bootstrapping procedure to estimate the quality of a specific core-genome prediction, as the accuracy of each run will depend on the specific completeness distribution and the number of genomes in the dataset under scrutiny. mOTUpan is implemented in the mOTUlizer software package, and available at github.com/moritzbuck/mOTUlizer, under GPL 3.0 license.

Connectivity of Fennoscandian Shield terrestrial deep biosphere microbiomes with surface communities.

The deep biosphere is an energy constrained ecosystem yet fosters diverse microbial communities that are key in biogeochemical cycling. Whether microbial communities in deep biosphere groundwaters are shaped by infiltration of allochthonous surface microorganisms or the evolution of autochthonous species remains unresolved. In this study, 16S rRNA gene amplicon analyses showed that few groups of surface microbes infiltrated deep biosphere groundwaters at the Äspö Hard Rock Laboratory, Sweden, but that such populations constituted up to 49% of the microbial abundance. The dominant persisting phyla included Patescibacteria, Proteobacteria, and Epsilonbacteraeota. Despite the hydrological connection of the Baltic Sea with the studied groundwaters, infiltrating microbes predominantly originated from deep soil groundwater. Most deep biosphere groundwater populations lacked surface representatives, suggesting that they have evolved from ancient autochthonous populations. We propose that deep biosphere groundwater communities in the Fennoscandian Shield consist of selected infiltrated and indigenous populations adapted to the prevailing conditions.

Candidatus Methylumidiphilus Drives Peaks in Methanotrophic Relative Abundance in Stratified Lakes and Ponds Across Northern Landscapes.

Boreal lakes and ponds produce two-thirds of the total natural methane emissions above the latitude of 50° North. These lake emissions are regulated by methanotrophs which can oxidize up to 99% of the methane produced in the sediments and the water column. Despite their importance, the diversity and distribution of the methanotrophs in lakes are still poorly understood. Here, we used shotgun metagenomic data to explore the diversity and distribution of methanotrophs in 40 oxygen-stratified water bodies in boreal and subarctic areas in Europe and North America. In our data, gammaproteobacterial methanotrophs (order Methylococcales) generally dominated the methanotrophic communities throughout the water columns. A recently discovered lineage of Methylococcales, Candidatus Methylumidiphilus, was present in all the studied water bodies and dominated the methanotrophic community in lakes with a high relative abundance of methanotrophs. Alphaproteobacterial methanotrophs were the second most abundant group of methanotrophs. In the top layer of the lakes, characterized by low CH4 concentration, their abundance could surpass that of the gammaproteobacterial methanotrophs. These results support the theory that the alphaproteobacterial methanotrophs have a high affinity for CH4 and can be considered stress-tolerant strategists. In contrast, the gammaproteobacterial methanotrophs are competitive strategists. In addition, relative abundances of anaerobic methanotrophs, Candidatus Methanoperedenaceae and Candidatus Methylomirabilis, were strongly correlated, suggesting possible co-metabolism. Our data also suggest that these anaerobic methanotrophs could be active even in the oxic layers. In non-metric multidimensional scaling, alpha- and gammaproteobacterial methanotrophs formed separate clusters based on their abundances in the samples, except for the gammaproteobacterial Candidatus Methylumidiphilus, which was separated from these two clusters. This may reflect similarities in the niche and environmental requirements of the different genera within alpha- and gammaproteobacterial methanotrophs. Our study confirms the importance of O2 and CH4 in shaping the methanotrophic communities and suggests that one variable cannot explain the diversity and distribution of the methanotrophs across lakes. Instead, we suggest that the diversity and distribution of freshwater methanotrophs are regulated by lake-specific factors.

Energy efficiency and biological interactions define the core microbiome of deep oligotrophic groundwater.

While oligotrophic deep groundwaters host active microbes attuned to the low-end of the bioenergetics spectrum, the ecological constraints on microbial niches in these ecosystems and their consequences for microbiome convergence are unknown. Here, we provide a genome-resolved, integrated omics analysis comparing archaeal and bacterial communities in disconnected fracture fluids of the Fennoscandian Shield in Europe. Leveraging a dataset that combines metagenomes, single cell genomes, and metatranscriptomes, we show that groundwaters flowing in similar lithologies offer fixed niches that are occupied by a common core microbiome. Functional expression analysis highlights that these deep groundwater ecosystems foster diverse, yet cooperative communities adapted to this setting. We suggest that these communities stimulate cooperation by expression of functions related to ecological traits, such as aggregate or biofilm formation, while alleviating the burden on microorganisms producing compounds or functions that provide a collective benefit by facilitating reciprocal promiscuous metabolic partnerships with other members of the community. We hypothesize that an episodic lifestyle enabled by reversible bacteriostatic functions ensures the subsistence of the oligotrophic deep groundwater microbiome.

Distinct microbial community along the chronic oil pollution continuum of the Persian Gulf converge with oil spill accidents.

The Persian Gulf, hosting ca. 48% of the world's oil reserves, has been chronically exposed to natural oil seepage. Oil spill studies show a shift in microbial community composition in response to oil pollution; however, the influence of chronic oil exposure on the microbial community remains unknown. We performed genome-resolved comparative analyses of the water and sediment samples along Persian Gulf's pollution continuum (Strait of Hormuz, Asalouyeh, and Khark Island). Continuous exposure to trace amounts of pollution primed the intrinsic and rare marine oil-degrading microbes such as Oceanospirillales, Flavobacteriales, Alteromonadales, and Rhodobacterales to bloom in response to oil pollution in Asalouyeh and Khark samples. Comparative analysis of the Persian Gulf samples with 106 oil-polluted marine samples reveals that the hydrocarbon type, exposure time, and sediment depth are the main determinants of microbial response to pollution. High aliphatic content of the pollution enriched for Oceanospirillales, Alteromonadales, and Pseudomonadales whereas, Alteromonadales, Cellvibrionales, Flavobacteriales, and Rhodobacterales dominate polyaromatic polluted samples. In chronic exposure and oil spill events, the community composition converges towards higher dominance of oil-degrading constituents while promoting the division of labor for successful bioremediation.

Freshwater Chlorobia Exhibit Metabolic Specialization among Cosmopolitan and Endemic Populations.

Photosynthetic bacteria from the class Chlorobia (formerly phylum Chlorobi) sustain carbon fixation in anoxic water columns. They harvest light at extremely low intensities and use various inorganic electron donors to fix carbon dioxide into biomass. Until now, most information on the functional ecology and local adaptations of Chlorobia members came from isolates and merely 26 sequenced genomes that may not adequately represent natural populations. To address these limitations, we analyzed global metagenomes to profile planktonic Chlorobia cells from the oxyclines of 42 freshwater bodies, spanning subarctic to tropical regions and encompassing all four seasons. We assembled and compiled over 500 genomes, including metagenome-assembled genomes (MAGs), single-amplified genomes (SAGs), and reference genomes from cultures, clustering them into 71 metagenomic operational taxonomic units (mOTUs or "species"). Of the 71 mOTUs, 57 were classified within the genus Chlorobium, and these mOTUs represented up to ∼60% of the microbial communities in the sampled anoxic waters. Several Chlorobium-associated mOTUs were globally distributed, whereas others were endemic to individual lakes. Although most clades encoded the ability to oxidize hydrogen, many lacked genes for the oxidation of specific sulfur and iron substrates. Surprisingly, one globally distributed Scandinavian clade encoded the ability to oxidize hydrogen, sulfur, and iron, suggesting that metabolic versatility facilitated such widespread colonization. Overall, these findings provide new insight into the biogeography of the Chlorobia and the metabolic traits that facilitate niche specialization within lake ecosystems.IMPORTANCE The reconstruction of genomes from metagenomes has helped explore the ecology and evolution of environmental microbiota. We applied this approach to 274 metagenomes collected from diverse freshwater habitats that spanned oxic and anoxic zones, sampling seasons, and latitudes. We demonstrate widespread and abundant distributions of planktonic Chlorobia-associated bacteria in hypolimnetic waters of stratified freshwater ecosystems and show they vary in their capacities to use different electron donors. Having photoautotrophic potential, these Chlorobia members could serve as carbon sources that support metalimnetic and hypolimnetic food webs.

The microbiome of the Black Sea water column analyzed by shotgun and genome centric metagenomics.

BACKGROUND: The Black Sea is the largest brackish water body in the world, although it is connected to the Mediterranean Sea and presents an upper water layer similar to some regions of the former, albeit with lower salinity and temperature. Despite its well-known hydrology and physicochemical features, this enormous water mass remains poorly studied at the microbial genomics level. RESULTS: We have sampled its different water masses and analyzed the microbiome by shotgun and genome-resolved metagenomics, generating a large number of metagenome-assembled genomes (MAGs) from them. We found various similarities with previously described Black Sea metagenomic datasets, that show remarkable stability in its microbiome. Our datasets are also comparable to other marine anoxic water columns like the Cariaco Basin. The oxic zone resembles to standard marine (e.g. Mediterranean) photic zones, with Cyanobacteria (Synechococcus but a conspicuously absent Prochlorococcus), and photoheterotrophs domination (largely again with marine relatives). The chemocline presents very different characteristics from the oxic surface with many examples of chemolithotrophic metabolism (Thioglobus) and facultatively anaerobic microbes. The euxinic anaerobic zone presents, as expected, features in common with the bottom of meromictic lakes with a massive dominance of sulfate reduction as energy-generating metabolism, a few (but detectable) methanogenesis marker genes, and a large number of "dark matter" streamlined genomes of largely unpredictable ecology. CONCLUSIONS: The Black Sea oxic zone presents many similarities to the global ocean while the redoxcline and euxinic water masses have similarities to other similar aquatic environments of marine (Cariaco Basin or other Black Sea regions) or freshwater (meromictic monimolimnion strata) origin. The MAG collection represents very well the different types of metabolisms expected in this kind of environment. We are adding critical information about this unique and important ecosystem and its microbiome.

Metagenomic discovery and functional validation of L-asparaginases with anti-leukemic effect from the Caspian Sea.

By screening 27,000 publicly available prokaryotic genomes, we recovered ca. 6300 type I and ca. 5200 type II putative L-asparaginase highlighting the vast potential of prokaryotes. Caspian water with similar salt composition to the human serum was targeted for in silico L-asparaginase screening. We screened ca. three million predicted genes of its assembled metagenomes that resulted in annotation of 87 putative L-asparaginase genes. The L-asparagine hydrolysis was experimentally confirmed by synthesizing and cloning three selected genes in E. coli. Catalytic parameters of the purified enzymes were determined to be among the most desirable reported values. Two recombinant enzymes represented remarkable anti-proliferative activity (IC50 <1IU/ml) against leukemia cell line Jurkat while no cytotoxic effect on human erythrocytes or human umbilical vein endothelial cells was detected. Similar salinity and ionic concentration of the Caspian water to the human serum highlights the potential of secretory L-asparaginases recovered from these metagenomes as potential treatment agents.

A Genomic Perspective Across Earth's Microbiomes Reveals That Genome Size in Archaea and Bacteria Is Linked to Ecosystem Type and Trophic Strategy.

Our view of genome size in Archaea and Bacteria has remained skewed as the data has been dominated by genomes of microorganisms that have been cultivated under laboratory settings. However, the continuous effort to catalog Earth's microbiomes, specifically propelled by recent extensive work on uncultivated microorganisms, provides an opportunity to revise our perspective on genome size distribution. We present a meta-analysis that includes 26,101 representative genomes from 3 published genomic databases; metagenomic assembled genomes (MAGs) from GEMs and stratfreshDB, and isolates from GTDB. Aquatic and host-associated microbial genomes present on average the smallest estimated genome sizes (3.1 and 3.0 Mbp, respectively). These are followed by terrestrial microbial genomes (average 3.7 Mbp), and genomes from isolated microorganisms (average 4.3 Mbp). On the one hand, aquatic and host-associated ecosystems present smaller genomes sizes in genera of phyla with genome sizes above 3 Mbp. On the other hand, estimated genome size in phyla with genomes under 3 Mbp showed no difference between ecosystems. Moreover, we observed that when using 95% average nucleotide identity (ANI) as an estimator for genetic units, only 3% of MAGs cluster together with genomes from isolated microorganisms. Although there are potential methodological limitations when assembling and binning MAGs, we found that in genome clusters containing both environmental MAGs and isolate genomes, MAGs were estimated only an average 3.7% smaller than isolate genomes. Even when assembly and binning methods introduce biases, estimated genome size of MAGs and isolates are very similar. Finally, to better understand the ecological drivers of genome size, we discuss on the known and the overlooked factors that influence genome size in different ecosystems, phylogenetic groups, and trophic strategies.

Cascading effects in freshwater microbial food webs by predatory Cercozoa, Katablepharidacea and ciliates feeding on aplastidic bacterivorous cryptophytes.

Heterotrophic nanoflagellates (HNF) are considered as major planktonic bacterivores, however, larger HNF taxa can also be important predators of eukaryotes. To examine this trophic cascading, natural protistan communities from a freshwater reservoir were released from grazing pressure by zooplankton via filtration through 10- and 5-µm filters, yielding microbial food webs of different complexity. Protistan growth was stimulated by amendments of five Limnohabitans strains, thus yielding five prey-specific treatments distinctly modulating protistan communities in 10- versus 5-µm fractions. HNF dynamics was tracked by applying five eukaryotic fluorescence in situ hybridization probes covering 55-90% of total flagellates. During the first experimental part, mainly small bacterivorous Cryptophyceae prevailed, with significantly higher abundances in 5-µm treatments. Larger predatory flagellates affiliating with Katablepharidacea and one Cercozoan lineage (increasing to up to 28% of total HNF) proliferated towards the experimental endpoint, having obviously small phagocytized HNF in their food vacuoles. These predatory flagellates reached higher abundances in 10-µm treatments, where small ciliate predators and flagellate hunters also (Urotricha spp., Balanion planctonicum) dominated the ciliate assemblage. Overall, our study reports pronounced cascading effects from bacteria to bacterivorous HNF, predatory HNF and ciliates in highly treatment-specific fashions, defined by both prey-food characteristics and feeding modes of predominating protists.

Correction to: Phage-centric ecological interactions in aquatic ecosystems revealed through ultra-deep metagenomics.

Following publication of the original article [1], the authors reported that an affiliation of the first author was missing.

Phage-centric ecological interactions in aquatic ecosystems revealed through ultra-deep metagenomics.

The persistent inertia in the ability to culture environmentally abundant microbes from aquatic ecosystems represents an obstacle in disentangling the complex web of ecological interactions spun by a diverse assortment of participants (pro- and eukaryotes and their viruses). In aquatic microbial communities, the numerically most abundant actors, the viruses, remain the most elusive, and especially in freshwaters their identities and ecology remain unknown. Here, using ultra-deep metagenomic sequencing from pelagic freshwater habitats, we recovered complete genomes of > 2000 phages, including small "miniphages" and large "megaphages" infecting iconic freshwater prokaryotic lineages. For instance, abundant freshwater Actinobacteria support infection by a very broad size range of phages (13-200 Kb). We describe many phages encoding genes that likely afford protection to their host from reactive oxygen species (ROS) in the aquatic environment and in the oxidative burst in protist phagolysosomes (phage-mediated ROS defense). Spatiotemporal abundance analyses of phage genomes revealed evanescence as the primary dynamic in upper water layers, where they displayed short-lived existences. In contrast, persistence was characteristic for the deeper layers where many identical phage genomes were recovered repeatedly. Phage and host abundances corresponded closely, with distinct populations displaying preferential distributions in different seasons and depths, closely mimicking overall stratification and mixis.