The Turkey Lake Watershed forms a cascading lake system that drains into Lake Superior and has historically served as a key research site for studies of acid rain and climate change. Although its microbial communities have received some attention, no previous work had explicitly examined its viruses. Viruses are integral components of aquatic microbial communities, where they influence biogeochemical cycling, primary production and bacterial mortality by infecting and lysing host cells. The authors argue that viral ecology offers a way to extend the environmental legacy of the watershed by identifying biological mechanisms through which long-term changes in acid deposition, climate, hydrology, carbon chemistry and forest disturbance propagate through microbial food webs.

The sampling design focused on Big Turkey Lake, a dimictic and oligotrophic lake located roughly 50 kilometres north of Sault Ste. Marie, Ontario. Biomass samples were collected in July 2018, monthly from May to August 2019, and in January 2020. One litre of water was collected in triplicates at Secchi depth and vacuum filtered through Whatman GF/C filters, then kept frozen until processing. For the January 2020 sample, taken during ice cover, water was drawn from about 25 centimetres below the ice. Only a single replicate from each time point was sent for sequencing, and the July 2019 sample was later excluded from analysis because very low coverage and sequencing depth caused it to fail quality control checks.

In the laboratory, DNA was extracted with the Qiagen DNeasy PowerSoil kit and sequenced on an Illumina MiSeq platform with 250-base paired-end reads. Raw reads, totalling 2.8 to 4.54 million per sample, were processed through the Metagenome-atlas pipeline for quality control and assembly, using MetaSPAdes as the assembler. Viral contigs were identified with a combination of VIBRANT and VirSorter2 within the WhatThePhage pipeline, with a minimum contig length of 1,500 base pairs, and contigs flagged as viral by at least one of the two tools were retained. Taxonomic classification was performed with geNomad, and relative abundance was calculated from the number of viral contigs in each sample.

The results showed a viral assemblage dominated by two major classes. Caudoviricetes, the tailed bacteriophages, were the most abundant group, accounting for 71.8 percent of total viral contigs and appearing in every sample. Megaviricetes, large double-stranded DNA viruses belonging to the nucleocytoplasmic large DNA viruses, which can infect a diverse range of eukaryotic organisms, were also present in all samples but at a lower relative abundance of 9.1 percent. This pattern matches earlier work in Lake Ontario and Lake Erie, where bacteriophages of the formerly recognized Myoviridae family made up roughly 80 percent of detected viral sequences while phycodnaviruses contributed a much smaller share.

The relative abundances of the two classes shifted markedly with season. Megaviricetes peaked during the summer months of June and August 2019, at 27.7 and 22.6 percent respectively, but fell to 2.0 percent in May 2019 and 2.1 percent in January 2020. Caudoviricetes showed the reverse trend, dropping to 35.5 percent in August 2019 and rising to approximately 85 percent in both May 2019 and January 2020. Because summer water temperatures at the lake typically exceed 20 degrees Celsius while winter brings freezing conditions and ice cover, and because microbial host abundance and composition are known to respond to temperature in aquatic systems, the authors conclude that temperature plays a central role in structuring the cell-associated viral community. Within Megaviricetes, most sequences could be assigned to family level, chiefly Phycodnaviridae, which infect eukaryotic algae, and Mimiviridae, which infect protists; both families were more abundant in June and August 2019 and declined in the winter sample.

Deeper taxonomic resolution proved difficult for the bacteriophages. Following the 2023 abolishment of the families Myoviridae, Siphoviridae and Podoviridae, many bacteriophage sequences now lack a corresponding family assignment, and in Big Turkey Lake most Caudoviricetes contigs could not be resolved beyond the class. Even in August 2019, the sample with the highest proportion of deeper assignments, only 18 percent of Caudoviricetes sequences received family-level taxonomy. The families that could be assigned, such as Straboviridae and Kyanoviridae and the order Crassvirales, appeared in only a few samples and at low abundance, with Crassvirales reaching a maximum of 5 percent in June 2019. About 19 percent of all viral contigs could not be classified at all, which the authors interpret as a sign that many viruses in the lake are potentially novel. Rarefaction analysis indicated that sequencing depth was sufficient to cover the full extent of the cell-associated viral diversity, supporting the interpretation that the observed differences reflect true temporal variation rather than sampling artefacts. Two virophage signatures, satellite viruses that co-infect with large algal viruses, were detected in May and June 2019, though at low relative abundance.

To infer hosts, the team ran all 799 viral contigs through the iPHoP pipeline, a machine-learning tool designed to predict bacterial and archaeal hosts from viral genomic sequences. Hosts could be predicted for 19.5 percent of contigs, most with confidence scores of 90 to 96 percent. Viruses predicted to infect members of the phylum Bacteroidota appeared in all samples. Predictions for Cyanobacteriota and Pseudomonadota hosts were found in every sample except January 2020; the absence of cyanobacterial hosts in winter is consistent with a reported near-complete absence of cyanobacteria in the ice-covered months, while the absence of Pseudomonadota predictions in January was unexpected, since that phylum reportedly made up more than half of the bacterial community at that time. The authors offer two possible explanations: either the pipeline failed to detect these associations in the winter sample, or Pseudomonadota in the lake are free of cell-associated viral infection in winter, with their viruses residing in the free viral fraction that this study did not capture. Hosts from the phylum Verrucomicrobiota dominated the predictions in August 2019 and January 2020, at 40 and 60 percent of predictions respectively, but were absent from the May and June predictions. Notably, some July 2018 contigs were predicted to infect the cyanobacterial genera Cyanobium, Vulcanococcus and Tolypothrix, which the authors report as the first identification of possible virus signatures associated with these genera.

The AMG analysis identified 62 candidate genes across the samples, of which 44 were classed as potential AMGs with one viral signature in a flanking region and 18 as high-confidence AMGs with viral signatures on both flanks. Nucleotide metabolism genes were the most abundant category, with 23 genes in total, including guanine reductase and ribonucleoside-diphosphate reductase, which can bolster nucleic acid biosynthesis during viral replication. Folate metabolism genes such as thymidylate synthase and dihydrofolate reductase were also detected. Energy metabolism genes included a potential photosystem II PsbK protein gene, a nitrogen fixation protein gene found in May 2019, and genes involved in the TCA cycle and iron-sulfur cluster insertion. Genes related to phosphorus cycling, specifically phoH, were found across all five samples; given the lake’s oligotrophic status and the prior identification of phosphorus-scavenging bacterial taxa in the watershed, the authors suggest this gene could serve as a phage diversity marker in the lake, as it has in marine environments, and that the findings are consistent with coupling between watershed-scale transport of dissolved organic matter and phosphorus, seasonal microbial succession, and host-virus interactions. A high-confidence AMG for lipid metabolism, 1,2-diacylglycerol beta-glucosyltransferase, involved in membrane lipid biosynthesis, was detected only in the January 2020 sample, an intriguing winter-specific pattern given that some bacteria adapt to cold by increasing membrane lipid production.

Finally, the team binned viral contigs into 103 viral bins spanning all five samples, of which six met the CheckV completeness threshold of above 90 percent and were considered draft viral metagenome-assembled genomes. These vMAGs ranged from 11 to 45 kilobase pairs and all aligned with Caudoviricetes phages in the IMG/VR database. Four could be assigned host phyla: two to Bacteroidota, one to Cyanobacteriota and one to Pseudomonadota. Two vMAGs contained integrase coding regions, indicating a possible temperate lifecycle, two carried tRNA genes, and one contained ribosomal protein L18 and a 23S rRNA/tRNA-pseudouridylate synthase, a reminder that some bacteriophages encode ribosomal components. The authors caution that their work covered only cell-associated viruses, not free viruses in the water column or those in sediments, and that with limited sampling it is difficult to identify which specific ecological, biological and physicochemical parameters drive the observed AMG distributions. Even so, they conclude that the cell-associated viral community in Big Turkey Lake is temporally dynamic, dominated by Caudoviricetes with a large unclassified fraction, and that novel viral genomes can be assembled even from small amounts of data derived from the cellular microbial fraction, establishing a foundation for future studies of viruses in this watershed.

Subject of Research: Viral community composition, predicted hosts and auxiliary metabolic genes in a dimictic lake in the Turkey Lake Watershed, Northern Ontario

Article Title: Cell-associated viral community composition and its functional potential in a dimictic lake on the Canadian shield

Article References: Collis, C. C., Cameron, E. S., Müller, K. M., Emelko, M. B., Tromas, N., & Nissimov, J. I. (2026). Cell-associated viral community composition and its functional potential in a dimictic lake on the Canadian shield. Frontiers in Microbiology, 17, 1941165. https://doi.org/10.3389/fmicb.2026.1941165

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1941165

Keywords: viral ecology, metagenomics, Big Turkey Lake, Turkey Lake Watershed, Caudoviricetes, Phycodnaviridae, auxiliary metabolic genes, vMAGs, host prediction, freshwater microbiology, phoH, seasonal dynamics