A guide to the natural history of freshwater lake bacteria

Abstract

Freshwater bacteria are at the hub of biogeochemical cycles and control water quality in lakes. Despite this, little is known about the identity and ecology of functionally significant lake bacteria. Molecular studies have identified many abundant lake bacteria, but there is a large variation in the taxonomic or phylogenetic breadths among the methods used for this exploration. Because of this, an inconsistent and overlapping naming structure has developed for freshwater bacteria, creating a significant obstacle to identifying coherent ecological traits among these groups. A discourse that unites the field is sorely needed. Here we present a new freshwater lake phylogeny constructed from all published 16S rRNA gene sequences from lake epilimnia and propose a unifying vocabulary to discuss freshwater taxa. With this new vocabulary in place, we review the current information on the ecology, ecophysiology, and distribution of lake bacteria and highlight newly identified phylotypes. In the second part of our review, we conduct meta-analyses on the compiled data, identifying distribution patterns for bacterial phylotypes among biomes and across environmental gradients in lakes. We conclude by emphasizing the role that this review can play in providing a coherent framework for future studies.

FIG. 1.

Timeline of 16S rRNA genes represented in the database. Lead author and journal abbreviation from papers contributing sequences are noted along the timeline. The yellow bar indicates the previous freshwater sequence collection review by Zwart and colleagues (238). Journal abbreviations are as follows: AEM, Applied and Environmental Microbiology; AME, Aquatic Microbial Ecology; EM, Environmental Microbiology; FEMS ME, FEMS Microbiology Ecology; FW Bio, Freshwater Biology; GeoMicro J, Geomicrobiology Journal; Mikro, Mikrobiologiia; L&O, Limnology and Oceanography; ME, Microbial Ecology; Micro, Microbiology; FEMS ML, FEMS Microbiology Letters; IJSEM, International Journal of Systematics and Evolutionary Microbiology; Polar Bio, Polar Biology; SAM, Systematic and Applied Microbiology. ¹, manuscript in press at Environmental Microbiology (141a) (includes GenBank accession numbers FJ827781 to FJ828505); ², GenBank accession numbers for the sequenced clones are HQ386253 to HQ386631; ³, GenBank accession numbers for the sequenced clones are FJ916807 to FJ916903 and HQ530565 to HQ532908, except HQ530583, HQ531638, and HQ532521. (Based on data from references , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , and .)

FIG. 2.

Stacked bar plot representing the distribution of sequences by phylum in the freshwater lake sequence database. Sequences were taxonomically classified by the RDP classifier (February 2010) (222) and were required to have 75% classifier confidence to be included as a representative of a phylum. Sequences with <75% confidence were termed unclassified. Phyla listed are present in both bar plots but are listed on only one side to aid in the visualization of the plots. ^a, the phylum Proteobacteria was split into its representative classes for sequences with a ≥75% confidence score at the class taxonomic assignment and unknown Proteobacteria for those sequences with confidence scores of <75%.

FIG. 3.

(A) Radial consensus phylogram of the freshwater lake lineages of the Actinobacteria. (B) Consensus phylogram of the tribes of lineages acI, acTH1, and acSTL. (C) Consensus phylogram of the tribes of lineages Luna1, acIII, Luna3, and acTH2. (D) Consensus phylogram of the tribes of lineages acIV and acV. All phylogenetic inference calculations were conducted with nearly full-length sequences (>1,300 nt) representing the lineages/clades/tribes. Frequency base filters were created to mask out highly variable positions, and representative sequences from members of the Archaea were used as the outgroup for all tree reconstructions. Bayesian clade credibility values and maximum likelihood bootstrap values (in parentheses) of >0.5 (50) are listed at each node. Tribe names are listed at the end of tree tips (trapezoids), with the number of sequences associated with each tribe listed in square brackets. Tree tips ending in ovals represent phylogenetic clusters that were named previously but did not meet the criteria to be called a tribe (≥2 sequences with ≥97% identity of ≥1,300 nt isolated from >1 lake). Current lineage and clade names are listed to the right of the phylograms and are highlighted in purple. The Linnaean taxonomy column is based on the RDP classification (222), where all sequences within the classified group must have ≥75% assigned confidence for the taxonomic group. Square brackets around a genus name indicate a candidate genus, and an asterisk indicates that the genus is described but not yet covered in the RDP classification. Alternate names from previous studies also covering the defined lineages/clades/tribes are listed. Probe coverage is based upon a perfect match to ≥75% of the sequences in a defined lineage/clade/tribe. References to past clades and probes are as follows: ^a, reference ; ^b, reference ; ^c, reference ; ^d, reference ; ^e, reference ; ^f, reference ; ^g, reference ; ^h, reference ; ⁱ, reference ; ^j, reference ; ^k, reference ; ^l, reference ; ^m, reference . *, tribe Luna1-A2 is associated with three “Candidatus” genera: “Candidatus Limnoluna,” “Candidatus Flaviluna,” and “Candidatus Rhodoluna” (73).

FIG. 3.

(A) Radial consensus phylogram of the freshwater lake lineages of the Actinobacteria. (B) Consensus phylogram of the tribes of lineages acI, acTH1, and acSTL. (C) Consensus phylogram of the tribes of lineages Luna1, acIII, Luna3, and acTH2. (D) Consensus phylogram of the tribes of lineages acIV and acV. All phylogenetic inference calculations were conducted with nearly full-length sequences (>1,300 nt) representing the lineages/clades/tribes. Frequency base filters were created to mask out highly variable positions, and representative sequences from members of the Archaea were used as the outgroup for all tree reconstructions. Bayesian clade credibility values and maximum likelihood bootstrap values (in parentheses) of >0.5 (50) are listed at each node. Tribe names are listed at the end of tree tips (trapezoids), with the number of sequences associated with each tribe listed in square brackets. Tree tips ending in ovals represent phylogenetic clusters that were named previously but did not meet the criteria to be called a tribe (≥2 sequences with ≥97% identity of ≥1,300 nt isolated from >1 lake). Current lineage and clade names are listed to the right of the phylograms and are highlighted in purple. The Linnaean taxonomy column is based on the RDP classification (222), where all sequences within the classified group must have ≥75% assigned confidence for the taxonomic group. Square brackets around a genus name indicate a candidate genus, and an asterisk indicates that the genus is described but not yet covered in the RDP classification. Alternate names from previous studies also covering the defined lineages/clades/tribes are listed. Probe coverage is based upon a perfect match to ≥75% of the sequences in a defined lineage/clade/tribe. References to past clades and probes are as follows: ^a, reference ; ^b, reference ; ^c, reference ; ^d, reference ; ^e, reference ; ^f, reference ; ^g, reference ; ^h, reference ; ⁱ, reference ; ^j, reference ; ^k, reference ; ^l, reference ; ^m, reference . *, tribe Luna1-A2 is associated with three “Candidatus” genera: “Candidatus Limnoluna,” “Candidatus Flaviluna,” and “Candidatus Rhodoluna” (73).

FIG. 4.

(A) Radial consensus phylogram of the freshwater lake lineages of the Bacteroidetes. (B) Consensus phylogram of the tribes of lineages bacI and bacIV. (C) Consensus phylogram of the tribes of lineages bacII, bacIII, bacV, and bacVI. All phylogenetic inference calculations were conducted with nearly full-length sequences (>1,300 nt) representing the lineages/clades/tribes. Frequency base filters were created to mask out highly variable positions, and representative sequences from members of the Archaea were used as the outgroup for all tree reconstructions. Bayesian clade credibility values and maximum likelihood bootstrap values (in parentheses) of >0.5 (50) are listed at each node. Tribe names are listed at the end of tree tips (trapezoids), with the number of sequences associated with each tribe listed in square brackets. Tree tips ending in ovals represent phylogenetic clusters that were named previously but did not meet the criteria to be called a tribe (≥2 sequences with ≥97% identity of ≥1,300 nt isolated from >1 lake). Current lineage and clade names are listed to the right of the phylograms and are highlighted in purple. The Linnaean taxonomy column is based on the RDP classification (222), where all sequences within the classified group must have ≥75% assigned confidence to the taxonomic group. Square brackets around a genus name indicate a candidate genus, and an asterisk indicates that the genus is described but not yet covered in the RDP classification. Alternate names from previous studies also covering the defined lineages/clades/tribes are listed. Probe coverage is based upon a perfect match to ≥75% of the sequences in a defined lineage/clade/tribe. References to past clades and probes are as follows: ^a, reference ; ^b, reference ; ^c, reference ; ^d, reference ; ^e, reference ; ^f, reference ; ^g, reference ; ^h, reference ; ⁱ, reference .

FIG. 5.

(A) Radial consensus phylogram of the freshwater lake lineages of the Alphaproteobacteria. (B) Consensus phylogram of the tribes of lineages alfI, alfII, and alfV. (C) Consensus phylogram of the tribes of lineages alfIII and alfIV. All phylogenic inference calculations were conducted with nearly full-length sequences (>1,300 nt) representing the lineages/clades/tribes. Frequency base filters were created to mask out highly variable positions, and representative sequences from members of the Archaea were used as the outgroup for all tree reconstructions. Bayesian clade credibility values and maximum likelihood bootstrap values (in parentheses) of >0.5 (50) are listed at each node. Tribe names are listed at the end of tree tips (trapezoids), with the number of sequences associated with each tribe listed in square brackets. Tree tips ending in ovals represent phylogenetic clusters that were named previously but did not meet the criteria to be called a tribe (≥2 sequences with ≥97% identity of ≥1,300 nt isolated from >1 lake). Current lineage and clade names are listed to the right of the phylograms and are highlighted in purple. The Linnaean taxonomy column is based on the RDP classification (222), where all sequences within the classified group must have ≥75% assigned confidence to the taxonomic group. Square brackets around a genus name indicate a candidate genus, and an asterisk indicates that the genus is described but not yet covered in the RDP classification. Alternate names from previous studies also covering the defined lineages/clades/tribes are listed. Probe coverage is based upon a perfect match to ≥75% of the sequences in a defined lineage/clade/tribe. References to past clades and probes are as follows: ^a, reference ; ^b, reference ; ^c, reference ; ^d, reference ; ^e, reference ; ^f, reference .

FIG. 6.

(A) Radial consensus phylogram of the freshwater lake lineages of the Betaproteobacteria. (B) Consensus phylogram of the tribes of lineages betI and betV. (C) Consensus phylogram of the tribes of lineages betII, betIII, betIV, and betVII. All phylogenic inference calculations were conducted with nearly full-length sequences (>1,300 nt) representing lineages/clades/tribes. Frequency base filters were created to mask out highly variable positions, and representative sequences from members of the Archaea were used as the outgroup for all tree reconstructions. Bayesian clade credibility values and maximum likelihood bootstrap values (in parentheses) of >0.5 (50) are listed at each node. Tribe names are listed at the end of tree tips (trapezoids), with the number of sequences associated with each tribe listed in square brackets. Tree tips ending in ovals represent phylogenetic clusters that were named previously but did not meet the criteria to be called a tribe (≥2 sequences with ≥97% identity of ≥1,300 nt isolated from >1 lake). Current lineage and clade names are listed to the right of the phylograms and are highlighted in purple. The Linnaean taxonomy column is based on the RDP classification (222), where all sequences within the classified group must have ≥75% assigned confidence to the taxonomic group. Square brackets around a genus name indicate a candidate genus, and an asterisk indicates that the genus is described but not yet covered in the RDP classification. Alternate names from previous studies also covering the defined lineages/clades/tribes are listed. Probe coverage is based upon a perfect match to ≥75% of the sequences in a defined lineage/clade/tribe. References to past clades and probes are as follows: ^a, reference ; ^b, reference ; ^c, reference ; ^d, reference ; ^e, reference ; ^f, reference ; ^g, reference ; ^h, reference ; ⁱ, reference ; ^j, reference .

FIG. 7.

(A) Consensus phylogram of the freshwater lake lineages of the Gammaproteobacteria. (B) Consensus phylogram of the freshwater lake lineages of the Verrucomicrobia. (C) Consensus phylogram of the freshwater lake lineages of OP10, the Fibrobacteres, and the Cyanobacteria. All phylogenic inference calculations were conducted with nearly full-length sequences (>1,300 nt) representing the lineages/clades/tribes. Frequency base filters were created to mask out highly variable positions, and representative sequences from members of the Archaea were used as the outgroup for all tree reconstructions. Bayesian clade credibility values and maximum likelihood bootstrap values (in parentheses) of >0.5 (50) are listed at each node. Tribe names are listed at the end of tree tips (trapezoids), with the number of sequences associated with each tribe listed in square brackets. Tree tips ending in ovals represent phylogenetic clusters that were named previously but did not meet the criteria to be called a tribe (≥2 sequences with ≥97% identity of ≥1,300 nt isolated from >1 lake). Current lineage and clade names are listed to the right of the phylograms and are highlighted in purple. The Linnaean taxonomy column is based on the RDP classification (222), where all sequences within the classified group must have ≥75% assigned confidence to the taxonomic group. Square brackets around a genus name indicate a candidate genus, and an asterisk indicates that the genus is described but not yet covered in the RDP classification. Alternate names from previous studies also covering the defined lineages/clades/tribes are listed. Probe coverage is based upon a perfect match to ≥75% of the sequences in a defined lineage/clade/tribe. References to past clades and probes are as follows: ^a, reference ; ^b, reference ; ^c, reference .

FIG. 8.

Habitat specificity of putative freshwater lake bacterial tribes. Two nearly full-length (>1,300 nt) 16S rRNA gene sequences from each tribe were chosen as representatives for a subsequent blastn analysis against the NCBI nonredundant (nr) database. Each tribe's sequence representatives were chosen based on obtaining a maximum or near-maximum within-tribe phylogenetic distance between the two representatives. The top 5,000 blastn hits were returned for each sequence representative, and the database sequence records with ≥97% identity were retained. The “isolation source” field was parsed from each sequence record, allowing categorization by the biome from which the sequence was obtained (freshwater, saltwater, or terrestrial) and by organism association (whether or not the sequence was isolated from a bacterium living on or within another organism). Two environments required special biome characterization: a brackish isolation source was considered half freshwater and half saltwater, and a sediment isolation source was considered half terrestrial and half aquatic (either freshwater or saltwater). Following categorization, the means of the biome and organism association results were calculated for each tribe. The percentage of sequences associated with each biome category for each tribe is visualized in ternary plots. Triangle vertices indicate a 100% relative abundance of the indicated biome for the plotted tribe. The symbol color for each tribe indicates the percentage of GenBank hits for the tribe that were noted as having been associated with another organism when the sequences were obtained (e.g., human skin or algal phycosphere). The symbol shape for each tribe reflects the mean number of GenBank hits with ≥97% sequence identity to the tribe reference sequences.

FIG. 9.

Bar plots indicating the relative recovery (number of tribe sequences/number of total sequences) of each tribe from all lakes and their prevalence (number of lakes from which sequences from a tribe were recovered/total number of lakes) in the 47 lakes. For reference, bars are aligned with an ultrametric tree of all freshwater lake tribes observed for the 47 lakes.

FIG. 10.

Sign and magnitude of Spearman rank correlation rho between the relative recovery of each tribe from each lake and environmental conditions (pH, the ratio between dissolved organic carbon and total phosphorus concentrations [DOC/TP], and absolute latitude [AbsLat]). Positive (blue) and negative (red) correlations between each tribe and each of the three variables are indicated. The size of the symbol indicates the magnitude of Spearman's rho. Environmental correlations are aligned with an ultrametric tree of all freshwater lake tribes observed for the 47-lake data set used for this analysis.

FIG. 11.

Detrended correspondence analysis (DCA) (94) to visualize patterns of tribe distribution along environmental gradients (pH, ratio between dissolved organic carbon and total phosphorus concentrations [DOC/TP], and absolute latitude [AbsLat]). The first axis explained ∼12% of the variance, and the second axis explained an additional 6%. All tribes were included in a single DCA, but each panel presents a limited number of tribes for clarity. (A) Actinobacteria; (B) Alphaproteobacteria; (C) Bacteroidetes; (D) Betaproteobacteria; (E) minor phyla.

FIG. 12.

(A) Time course of the relative contributions by the acI lineage and acI-A clade (A) and the acI-A2, acI-A4, acI-A5, and acI-A6 tribes (B) of the phylum Actinobacteria to the bacterial community of Lake Mendota (by percentage of total fluorescence), inferred from automated ribosomal intergenic spacer analysis (ARISA) according to methods described previously (141). Sampling was conducted biweekly for 4 years during the ice-off period of the lake. The curves are constructed as a moving average (n = 2) of the relative contribution data. For comparison, the tribe plots are scaled to the maximum fluorescence observed for each tribe.