Uptake of Phytoplankton-Derived Carbon and Cobalamins by Novel Acidobacteria Genera in Microcystis Blooms Inferred from Metagenomic and Metatranscriptomic Evidence

ABSTRACT

Interactions between bacteria and phytoplankton can influence primary production, community composition, and algal bloom development. However, these interactions are poorly described for many consortia, particularly for freshwater bloom-forming cyanobacteria. Here, we assessed the gene content and expression of two uncultivated Acidobacteria from Lake Erie Microcystis blooms. These organisms were targeted because they were previously identified as important catalase producers in Microcystis blooms, suggesting that they protect Microcystis from H₂O₂. Metatranscriptomics revealed that both Acidobacteria transcribed genes for uptake of organic compounds that are known cyanobacterial products and exudates, including lactate, glycolate, amino acids, peptides, and cobalamins. Expressed genes for amino acid metabolism and peptide transport and degradation suggest that use of amino acids and peptides by Acidobacteria may regenerate nitrogen for cyanobacteria and other organisms. The Acidobacteria genomes lacked genes for biosynthesis of cobalamins but expressed genes for its transport and remodeling. This indicates that the Acidobacteria obtained cobalamins externally, potentially from Microcystis, which has a complete gene repertoire for pseudocobalamin biosynthesis; expressed them in field samples; and produced pseudocobalamin in axenic culture. Both Acidobacteria were detected in Microcystis blooms worldwide. Together, the data support the hypotheses that uncultured and previously unidentified Acidobacteria taxa exchange metabolites with phytoplankton during harmful cyanobacterial blooms and influence nitrogen available to phytoplankton. Thus, novel Acidobacteria may play a role in cyanobacterial physiology and bloom development.

IMPORTANCE Interactions between heterotrophic bacteria and phytoplankton influence competition and successions between phytoplankton taxa, thereby influencing ecosystem-wide processes such as carbon cycling and algal bloom development. The cyanobacterium Microcystis forms harmful blooms in freshwaters worldwide and grows in buoyant colonies that harbor other bacteria in their phycospheres. Bacteria in the phycosphere and in the surrounding community likely influence Microcystis physiology and ecology and thus the development of freshwater harmful cyanobacterial blooms. However, the impacts and mechanisms of interaction between bacteria and Microcystis are not fully understood. This study explores the mechanisms of interaction between Microcystis and uncultured members of its phycosphere in situ with population genome resolution to investigate the cooccurrence of Microcystis and freshwater Acidobacteria in blooms worldwide.

INTRODUCTION

Interactions between microorganisms have profound impacts on global biogeochemistry by influencing microbial fitness, metabolism, and community composition. For example, many microbes use the waste products of others for growth or produce compounds required by other community members (1–3). A widely recognized example of such metabolic handoffs is in the interactions between phytoplankton and heterotrophic bacteria. Phytoplankton support heterotrophic bacterial growth by providing organic carbon and organic sulfur (4, 5). In addition, some heterotrophic bacteria can obtain cobalamin (vitamin B₁₂) by remodeling pseudocobalamin (6, 7), of which cyanobacteria are a major environmental source (8). In turn, the heterotrophic partners can improve phytoplankton growth by producing essential vitamins and growth factors such as cobalamins (4, 9), increasing the bioavailability of trace metal cofactors (10), regenerating nutrients from organic material (11–13), and detoxifying reactive oxygen species (14). Heterotrophic bacteria are known to impact phytoplankton fitness through the transfer of metabolites (10, 11, 15, 16) in a zone of close physical association termed the phycosphere (17, 18).

Interactions between heterotrophic bacteria and phytoplankton also influence competition between phytoplankton taxa (19). Thus, phycosphere interactions likely play a role in shaping successions of phytoplankton taxa (20) and may have implications at the level of whole ecosystems by modulating primary productivity and phytoplankton bloom formation (16, 18, 21). Interactions in the phycosphere can have both strain- and species-specific outcomes (11, 22), and the fitness impacts on phytoplankton have been linked to the exchange of specific metabolites in some interactions (9, 15, 16). Therefore, identifying the bacterial taxa associated with a given phytoplankton taxon and the metabolites exchanged between them can improve our understanding of phytoplankton physiology and competition between phytoplankton in natural assemblages with cooccurring bacteria, thereby improving modeling of ecosystem-wide processes such as primary productivity and harmful algal bloom formation (16, 18, 21).

Phytoplankton-bacterium interactions likely influence the development and phytoplankton community structure of Microcystis blooms, which degrade freshwater systems around the world (23). Metagenomic and metatranscriptomic studies have suggested complementary gene content (24) and expression (25) in Microcystis and cooccurring bacterial communities. Furthermore, Microcystis spp. grow in colonies that harbor heterotrophic bacterial communities that differ from the surrounding microbial communities (26–28) and differ both seasonally (27) and by Microcystis genotype (27, 29). Heterotrophic bacteria have previously been shown to impact competition between Microcystis and eukaryotic algae (19). Such impacts on the invasion and successions of phytoplankton taxa are likely important determinants of toxin concentrations in Microcystis blooms, which are substantially influenced by the relative proportions of toxin-producing and non-toxin-producing Microcystis spp. (30, 31). However, the impact of phycosphere bacteria on Microcystis growth and physiology remains uncharacterized, in part because many of the microbes associated with Microcystis colonies are yet to be cultured (26, 27, 32) and existing metatranscriptomic studies lack genome-resolved analysis of in situ communities (25, 33–35). Direct recovery of bacterial genomes from the environment can provide insights into the biochemical and ecological characteristics of these uncultivated organisms of interest.

This study focused on metagenome-assembled genomes (MAGs) of two novel, uncultivated Acidobacteria from a western Lake Erie cyanobacterial harmful algal bloom (CHAB) in the summer-fall of 2014. These Acidobacteria were responsible for a substantial fraction of catalase transcripts in a western Lake Erie Microcystis bloom community despite their low relative abundance (36). Therefore, both Acidobacteria may be important for H₂O₂ detoxification during CHABs, which is an essential service that heterotrophic bacteria provide to some cyanobacterial species (14) and has been proposed to influence Microcystis strain succession during CHABs (36). In addition, both genomes were reconstructed from phytoplankton- and particle-attached size fractions, and one of the Acidobacteria (genus Bryobacter) was previously identified in 25% of individual Microcystis colony phycosphere communities (27) and correlated with certain Microcystis genotypes (37). Together, these results suggest that the Acidobacteria interact with Microcystis and other phytoplankton in freshwater cyanobacterial blooms.

Here, we examined the gene content and expression of the Acidobacteria MAGs during the western Lake Erie cyanobacterial bloom in the summer-fall of 2014 to explore potential interactions with cyanobacteria and other phytoplankton. In addition, the abundance of both Acidobacteria in amplicon data sets from size-fractionated samples in a range of eutrophic environments was examined to determine their specificity to cyanobacterial blooms and Microcystis colonies. The data suggest that both genomes express genes involved in degradation of exopolysaccharide (EPS), organic acids, amino acids, and peptides but are only occasionally found in Microcystis bloom communities.

RESULTS AND DISCUSSION

Identification of two novel species of subdivision 3 Acidobacteria in Microcystis blooms.

Metagenome-assembled genomes (MAGs) of two Acidobacteria were obtained from metagenomic data from microbial communities collected during the summer-fall of 2014 (see Data Set S1 in the supplemental material). The samples spanned various environmental conditions and stages of Microcystis bloom development (36, 38, 39). These genomes were targeted because they highly expressed catalase-peroxidase genes (katG) during the cyanobacterial bloom (36) and were detected in particle- or phytoplankton-attached samples (27, 36) and thus were hypothesized to be associated with phytoplankton and important degraders of H₂O₂, which may influence the composition and development of cyanobacterial blooms (40).

Both genomes are nearly complete with low contamination and meet standards for high-quality draft genomes (Table 1) (41). The 16S rRNA gene sequences of MAGs CoA2 C42 and CoA8 C33 were classified as Paludibaculum and Bryobacter (see Materials and Methods) and were closest matches to 16S rRNA gene sequences from Paludibaculum fermentans and Bryobacter aggregatus strains (92.29 and 96.5%, respectively), which were both isolated from peat bogs (42, 43). Based on 16S rRNA gene similarity thresholds for species and genera (44), the percent similarity score for the CoA8 C33 MAG with Bryobacter aggregatus is above genus-level thresholds, and the percent similarity for CoA2 C42 MAG with Paludibaculum fermentans is below genus-level but above family-level thresholds. Phylogeny of 16S rRNA gene sequences placed both genomes within subdivision 3 Acidobacteria, Bryobacteraceae (Fig. 1A). While the CoA8 C33 MAG was placed as a sister lineage to Bryobacter aggregatus with high confidence, the specific placement of the CoA2 C42 MAG within subdivision 3 Acidobacteria had lower bootstrap support. These results support that genome CoA8 C33 is a novel species of Bryobacter, while genome CoA2 C42 likely represents a novel genus within the subdivision 3 Acidobacteria sister to Paludibaculum. Here, these species will be referred to as Bryobacter CoA8 C33 and acidobacterium CoA2 C42.

TABLE 1.

Quality information for Acidobacteria MAGs from western Lake Erie cyanobacterial blooms

BinID	Genus	Completeness (%)	Contamination (%)	GC %	Size (Mbp)	N 50	Gene count	16S no.	23S no.	tRNA no.
CoA8 C33	Bryobacter	98.21	0.87	60.89	5.0536	7,548	4,488	1	1	47
CoA2 C42	Unclassified	98.70	2.17	64.93	6.0673	38,916	5,163	1	1	47

FIG 1.

Taxonomic assignment of novel Acidobacteria MAGs based on 16S rRNA phylogeny and genomic average nucleotide identity (gANI). (A) Maximum likelihood tree of published, full-length Acidobacteria 16S rRNA gene sequences. The tree is rooted with the sequence from the deltaproteobacterium Geobacter metallireducens. Node labels show bootstrap support (n = 2,000). Branch lengths have no information. Subdivision 3 Acidobacteria are highlighted in green, subdivision 4 Acidobacteria are highlighted in gold, and subdivision 6 Acidobacteria are highlighted in purple. The leaf labels for the 16S rRNA sequences from the Lake Erie MAGs are colored red. (B and C) gANI comparisons of Bryobacter CoA8 C33 (C) and Acidobacteria CoA2 C42 (B) with published, high-quality Acidobacteria genomes in IMG (n = 63).

Whole-genome alignments also indicated that both genomes were most similar to Acidobacteria genomes from subdivision 3 (Fig. 1B and C). The CoA8 C33 genome was most similar to Bryobacter aggregatus strains, with genomic average nucleotide identity (gANI) and alignment coverage values (0.72 and 0.44, respectively) within the range of genus-level but below species-level cutoffs (45, 46), further supporting that it is a novel species of Bryobacter. The CoA2 C42 genome was most similar to a MAG from a drinking-water metagenome (IMG Gold Study identifier [ID]: Gs0114768). While subdivision 3 Acidobacteria have been recognized as numerically important in soils (47, 48) and present in the microbiome of marine sponges (49), freshwater (37), and marine waters (50), to our knowledge, this study represents the first detailed and targeted description of nearly complete Bryobacteraceae genomes from an aquatic environment.

Highly expressed genes are involved in ATP synthesis, respiration, biofilm adhesion, and chemotaxis.

In both genomes, the most highly expressed genes were involved in translation, secretion proteins, peptidases of unknown function, chaperonins, H₂O₂ detoxification, and ATP synthesis, along with hypothetical or uncharacterized proteins (see Fig. S3 and S4 in the supplemental material). Highly expressed genes encoding hypothetical proteins did not align with RNA gene sequences in NCBI and either had best hits to other hypothetical proteins or no significant hits to any proteins in the database, suggesting that they encode proteins of unknown function (Table S2). Both genomes have a complete tricarboxylic acid (TCA) cycle, cytochrome c oxidase, and nearly complete glycolysis and Entner-Doudoroff pathways (Fig. S5) and lack known pathways for carbon fixation and synthesis of bacteriochlorophyll and rhodopsin pigments, indicating that these organisms are aerobic chemoheterotrophs like known subdivision 3 Acidobacteria isolates (51). Highly expressed genes involved in biofilm adhesion, chemotaxis, flagellum biosynthesis, and motility (Fig. S6 and S7) indicate that these organisms are chemotactic.

Expression of genes involved in degradation of exopolysaccharides.

Functional annotation of expressed genes suggests that both Acidobacteria obtain carbon from breakdown of complex exopolysaccharides, including known phytoplankton products. Both genomes possess putative pectate-lyase, alpha-mannosidase, and xylan esterase exoenzymes to completely degrade homogalacturonan, mannose, and xylose polymers completely to the constituent monosaccharides (Fig. S5 and Data Sets S2 and S3 in the supplemental material). The Bryobacter MAG also contains genes for degradation of galactose and arabinofuranose polymers, while the Acidobacteria CoA2 C42 MAG contains genes for the degradation of alginate (Fig. S5). Both Acidobacteria genomes also expressed genes involved in the degradation of galacturonate monomers and other monosaccharides such as xylose, glucose, galactose, and mannose (Fig. S5). These monosaccharides along with uronic acid polymers make up the exopolysaccharide (EPS) mucilage encasing cyanobacterial cells (52–55), and bacterial degradation of Microcystis EPS has been observed in cocultures with heterotrophic bacteria (55), supporting the hypothesis that EPS in Microcystis colonies could be a substrate of the Acidobacteria exoenzymes. Eukaryotic phytoplankton also produce extracellular polysaccharides composed of these same constituents (56–58), which could potentially provide carbon to these organisms (56–59).

Evidence for uptake of low-molecular-weight organic compounds.

Metatranscriptomic data suggest that low-molecular-weight organic carbon is a source of carbon and nitrogen for both Acidobacteria and further support the hypothesis that both Acidobacteria participate in metabolic exchanges within phycosphere communities, which may include phytoplankton and other heterotrophic organisms (Fig. 2). Among the most highly expressed transporters in both Acidobacteria during the Microcystis bloom were concentrative nucleoside transporters (CNT) (Fig. S6 and S7), suggesting that they use nucleosides from the environment. There was also detectable expression of many genes putatively encoding amino acid, peptide, and polyamine uptake (Fig. 2). This indicates that the nitrogen demand for both Acidobacteria is likely met in part by uptake of organic nitrogen in dissolved amino acids, peptides, and nucleosides, which are important sources of nitrogen for bacterioplankton (60–62) and are derived from a wide variety of cell lysates and exudates, including those from phytoplankton (4, 13, 59, 63). Expression of amino acid oxidases (Fig. 2) suggests that both Acidobacteria deaminate amino acids to access nitrogen (61, 64, 65).

FIG 2.

Relative abundance of low-molecular-weight organic carbon transporters and enzymes related to their metabolism by Bryobacter CoA8 C33 (blue) and acidobacterium CoA2 C42 (red) associated with phytoplankton seston in the 4 August metatranscriptome from nearshore western Lake Erie station WE12. Relative abundance is expressed as reads mapped per kilobase of gene per million reads mapped to the respective genome (RPKM; rounded to the nearest whole number). A range of RPKM values indicates that multiple gene loci were predicted to encode the indicated reactions (includes gene duplications or genes encoding enzyme subunits), and only maximum and minimum RPKM values are shown.

The metatranscriptomic data also suggest uptake of phytoplankton exudate. Both Acidobacteria expressed genes to oxidize lactate and glycolate and dephosphorylate phosphoglycolate (Fig. 2), which are common exudates of cyanobacteria (63, 66, 67) and eukaryotic phytoplankton (68, 69). Lactate permease, which is involved in uptake of both lactate and glycolate (70), was expressed, along with transporters putatively involved in the uptake of other organic acids (Fig. 2). Among the known pathways for glycolate use in bacteria, both genomes lack genes in the glyoxylate cycle and methylaspartate cycle and the majority of genes in the ethylmalonyl coenzyme A (CoA) and 3-hydroxypropanoate cycles but possess and express genes in the serine pathway, although the final two genes of this pathway are missing in both genomes (Fig. 2 and Fig. S8). The presence of genes in the serine pathway perhaps indicates that the glyoxylate formed from the oxidation of glycolate is incorporated into amino acids (71–73) and the metabolism of C₁ compounds (73).

Evidence for regeneration of nitrogen from peptides and amino acids.

Transcripts for dipeptide and oligopeptide transporters, peptidases, and cyanophycinase-like proteins suggest that both Acidobacteria use extracellular peptides as a carbon and nitrogen source (Fig. 2). Expressed amino acid efflux transporters (Fig. 2) further suggest that peptide degradation may be linked to efflux of excess amino acids. Amino acid efflux allows biosynthesis of other amino acids to meet cellular demands (74–76) and is essential for maintaining balanced growth from degradation of oligopeptides (76–78). Thus, peptide degradation followed by amino acid efflux by bacteria could regenerate amino acids, which have been shown to enhance cyanobacterial growth and biomass in lakes (30) and cyanobacterial cultures (79–81).

Bacteria can also regenerate nitrogen from organic matter via excretion of ammonia when amino acids are the major sources of nitrogen for bacterial growth (60) or under carbon-limiting conditions (82, 83). In both Acidobacteria, expression of amino acid oxidases/deaminases (Fig. 2), which convert extracellular amino acids into ammonium, supports that these organisms regenerate ammonia from dissolved amino acids. Ammonium generated from dissolved organic matter (DOM) is used by phytoplankton and bacteria for growth (64, 65), community demand for ammonium can be high in Microcystis blooms (84), and ammonium uptake by Microcystis has been linked to ammonium regeneration from DOM by cooccurring microbes (25, 84). Together, these data support the hypothesis that bacterial excretion of ammonia is a potential source of nitrogen for Lake Erie cyanobacteria.

Evidence for cobalamin auxotrophy and uptake.

Gene annotation and expression data from the Acidobacteria genomes suggest that both organisms are auxotrophs of cobalamin. Some cobalamin-dependent enzymes are involved in critical cellular functions such as methionine and nucleotide synthesis (85). Thus, organisms that lack genes for cobalamin biosynthesis must obtain cobalamin released into the environment by cobalamin-producing organisms (9, 86). Both Acidobacteria lack the entire pathway for biosynthesis of the corrin ring structure common to all cobalamins, as in other heterotrophic bacteria cocultured with cyanobacteria (87). The Bryobacter genome expressed genes annotated as cobalamin transporters (Fig. 3A), but the Acidobacteria CoA2 C42 MAG lacked genes annotated as cobalamin transporters (see Data Set S2 in the supplemental material). Both genomes expressed genes encoding TonB-family proteins, which are required to energize the membrane for cobalamin transport (88, 89), and some of these were among the most highly expressed membrane-associated proteins in Acidobacteria CoA2 C42 (Fig. S6). Together, this suggests that neither of these Acidobacteria can synthesize cobalamins de novo and that both transport cobalamins from the environment into the cell.

FIG 3.

Relative abundance of transcripts from the Bryobacter CoA8 C33 genome involved in cobalamin transport (A), cobalamin-dependent genes and their cobalamin-independent counterparts (B), and cobalamin remodeling (C) in the 4 August metatranscriptome from western Lake Erie nearshore station WE12. Relative abundance is expressed as reads mapped per kilobase pair of gene per million reads mapped to the genome (RPKM).

Both Acidobacteria expressed cobalamin-dependent genes. A nrdJ-encoded class II ribonucleotide reductase and a metH-encoded methionine synthase were detected in the Bryobacter genome (Fig. 3B). Bryobacter CoA8 C33 lacks cobalamin-independent alternatives to nrdJ, which suggests that cobalamin is a requirement for this organism. Although Acidobacteria CoA2 C42 uses the cobalamin-independent alternatives to nrdJ, nrdA and nrdB (Data Set S2 in the supplemental material), and has the cobalamin-independent version metE (for which no expression was detected), it expressed metH (Table 2). Therefore, both organisms were likely using cobalamins during the cyanobacterial bloom.

TABLE 2.

Expression of B₁₂-dependent genes in Acidobacteria CoA2 C42 MAG

Gene	KEGG no.	KO	IMG annotation	Gene symbol(s)	4 August RPKM
2806999884	2.1.1.13	K00548	Methionine synthase (B12 dependent)	metH	216.80
2807002171	2.1.1.14	K00549	5-Methyltetrahydropteroyltriglutamate–homocysteine methyltransferase	metE	0
2807000579	2.5.1.17	K00798	Cob(I)alamin adenosyltransferase	cobA, pduO	93.77

We also detected expression of genes involved in remodeling cobalamin axial ligands in both Acidobacteria genomes. Variants of cobalamins differ in the chemical groups that make up the upper and lower axial ligands (85, 86), and because most microbial taxa exclusively use cobalamins with specific lower axial ligand structures (8, 86, 90), mechanisms to remodel cobalamins are necessary for microbes to convert the various exogenous cobalamin forms into the correct chemical forms needed for growth (6, 86, 91). Both genomes possessed genes for attaching an adenosyl group to the upper ligand (Fig. 3C and Table 2), but only the Bryobacter genome had genes for remodeling the lower axial ligand (Fig. 3C). Expression of cobalamin-remodeling genes in Bryobacter suggests that it can convert various chemical forms of cobalamin into the specific variant required as its enzyme cofactor.

We identified methylpseudocobalamin and two stereoisomers of hydroxopseudocobalamin in axenic cultures of Microcystis aeruginosa PCC 7806 and PCC 9806 (Fig. 4A; see also Fig. S9 to S11). Cobalamins containing a methyl group in the upper ligand are the active cofactors for methyltransferase reactions, including methionine synthesis (85), whereas cobalamins containing a hydroxyl group in the upper ligand are degradation products of biologically active pseudocobalamins (92). Cyanocobalamin and other forms of cobalamin (the lower axial ligand is 5,6-dimethylbenzimidazol [DMB]) were not detected in Microcystis cultures (Fig. 4B), consistent with previous studies of cyanobacteria (8, 86, 93). In addition to detecting pseudocobalamin in Microcystis biomass, we detected a complete pathway for production of pseudocobalamin in closed Microcystis genomes, which recruited reads at greater than 95% nucleotide identity from western Lake Erie metatranscriptomes (Fig. S12). The presence and expression of complete genetic pathways for pseudocobalamin biosynthesis in Microcystis genomes and communities, along with metabolomic analyses, support that Microcystis produces methylpseudocobalamin for growth as do other cyanobacteria (8, 86, 87, 94). Taken together with expression of cobalamin-remodeling genes (Fig. 3C) and genes that require cobalamin for nucleotide and methionine synthesis by Bryobacter CoA8 C33 (Fig. 3B) and the physical association of Bryobacter with Microcystis phycosphere colonies (27), these results support the hypothesis that Microcystis and other bloom-associated cyanobacteria (e.g., Synechococcus-Cyanobium [27]) that are known cobalamin producers (8, 94) are potential sources of cobalamins for Bryobacter CoA8 C33. The chemical form of cobalamins used by both Acidobacteria remains unknown, so it is possible that the Acidobacteria obtain cobalamins from other sources (91). Regardless, our data suggest that both Acidobacteria rely on other microorganisms to meet cobalamin demands.

FIG 4.

(A) Structure and LC-MS chromatograms of pseudocobalamin forms observed in Microcystis cultures and Spirulina powder. The Spirulina served as a positive control given that commercial standards for pseudocobalamin are not readily available. The compound labeled 1* has the same molecular weight and mass spectrum as compound 1 and is therefore a potential stereoisomer of hydroxopseudocobalamin that was observed only in Microcystis cultures. (B) Structure of cyanocobalamin and LC-MS chromatograms of a cyanocobalamin standard compared with Microcystis cultures and Spirulina powder, demonstrating its absence in the cyanobacterial samples tested here. Arrows indicate metal coordination in cobalamin structures. EIC, extracted ion chromatogram; NL, normalization value for extracted ion intensity.

Presence and relative abundance in 16S rRNA gene amplicon data sets.

To determine if these Acidobacteria regularly occur in, or are specific to, Microcystis-dominated blooms, we measured their abundance in published 16S rRNA gene amplicon data sets spanning a range of freshwater systems where Microcystis-dominated blooms occur (Table 3; see Materials and Methods). Amplicon sequences with high percent similarity (97%) to variable regions of the 16S rRNA gene sequences in both Acidobacteria MAGs were present at low relative abundance in western Lake Erie (Bryobacter, mean 0.072%, range 0 to 1.41%; acidobacterium CoA2 C42, mean 0.13%, range 0 to 0.69%) and other systems (Bryobacter, mean 0.006%, range 0 to 0.40%; acidobacterium CoA2 C42, mean 0.048%, range 0 to 0.72%). There was a weak but significant positive relationship between the relative abundance of both Acidobacteria taxa and Microcystis relative abundance in whole water samples collected from western Lake Erie cyanobacterial blooms (Fig. 5A and B). In 100-μm retentate samples from the summer-fall of 2014 in western Lake Erie, there were no significant relationships between the relative abundances of Microcystis and Acidobacteria CoA2 C42 or Bryobacter (Fig. S13). In other freshwater systems, there was also a significant positive correlation between Bryobacter relative abundance and Microcystis relative abundance (Fig. 5C). In contrast, acidobacterium CoA2 C42 had a significant negative correlation in other data sets, but the correlation was weak (Fig. 5D). The regression results indicate that strains related to both Acidobacteria species also occur in freshwater systems at times and locations where Microcystis is absent. However, in samples where Microcystis relative abundance was below 1%, Bryobacter relative abundance was typically lower (mean 0.001%), indicating that the highest Bryobacter relative abundance usually occurs during Microcystis blooms. Both Acidobacteria could also be absent while Microcystis was present at high abundance, suggesting that while both may occur in Microcystis blooms, they are not consistently present in all microbial communities associated with Microcystis blooms. An inconsistent cooccurrence between Bryobacter spp. and Microcystis is consistent with bacterial interactions with Microcystis being strain specific (27, 29, 37) and with uneven spatial and temporal distribution of Microcystis strains (95).

TABLE 3.

Description of public data sets mined for Acidobacteria 16S rRNA sequences in this study

NCBI accession no.	Location and sampling scheme	Material type	Size fraction(s)	Reference
PRJNA575023	Discrete sampling of eutrophic lakes around the globe	Bulk phytoplankton seston dominated by Microcystis	>100 μm	Cook et al., 2020 (24)
PRJNA386411	Lake Taihu, China, time series	Size-fractionated communities	>120 μm, 3–36 μm, 0.2–3 μm	Shi et al., 2018 (28)
PRJNA591360	Transects across the Laurentian Great Lakes in spring and summer	Free-living communities	0.22–1.6 μm	Paver et al., 2020 (124)
PRJNA479553	Monthly sampling of Nakdong River, South Korea	Whole water communities	>0.22 μm	Chun et al., 2019 (128)
PRJNA255432	Transect across the Laurentian Great Lakes	Whole water communities	>0.22 μm	Rozmarynowycz et al., 2019 (125)
PRJNA353865	Lake Champlain, Canada, time series	Whole water communities	>0.22 μm	Tromas et al., 2017 (127)
PRJEB14911	Lake Mendota, USA, time series	Whole water communities	>0.22 μm	Kara et al., 2013 (126)

FIG 5.

Percent abundance of Bryobacter and acidobacterium CoA2 C42 as a function of Microcystis percent abundance in whole water microbial community rRNA amplicon data sets from freshwaters. (A) Bryobacter OTU percent abundance versus Microcystis OTU percent abundance in V4 16S rRNA amplicon data sets collected during western Lake Erie cyanobacterial blooms. (B) acidobacterium CoA2 C42 OTU percent abundance versus Microcystis OTU percent abundance in V4 16S rRNA amplicon data sets collected during western Lake Erie cyanobacterial blooms. (C) The percent abundance of reads in published amplicon data sets that mapped to the 16S rRNA gene from the Bryobacter CoA8 C33 MAG versus the percent abundance of amplicon reads that mapped to 16S rRNA from Microcystis. (D) The percent abundance of reads in published amplicon data sets that mapped to the 16S rRNA gene from the acidobacterium CoA2 C42 MAG versus the percent abundance of amplicon reads that mapped to 16S rRNA from Microcystis. In all panels, the shaded area around the regression line indicates the regression standard error, error bars show the 95% confidence intervals determined from replicate filters (n = 4 or 8), and P values show the significance of the regression slope calculated with an F-test.

We assessed the abundance of both Acidobacteria groups in particle-attached microbial communities of various sizes, which may indicate physical association with phytoplankton, including Microcystis, which grow in large, buoyant colonies (96). The relative abundance of both Bryobacter and Acidobacteria CoA2 C42 amplicons was enriched in particle-attached communities (>100-μm retentate samples) during the 2014 western Lake Erie Microcystis bloom (Fig. 6). Similarly, although it was present in only August and September, Bryobacter was associated with the size fraction that contained the most Microcystis phytoplankton in Lake Taihu but was absent from smaller particles and free-living communities (Fig. 7). In contrast, acidobacterium CoA2 C42 was absent from large, Microcystis-containing aggregates and present in smaller particles and free-living communities in Lake Taihu (Fig. 7). Bryobacter was also absent from free-living communities throughout the Great Lakes, while acidobacterium CoA2 C42 was present (Fig. S14). A previous study identified Bryobacter in ~25% of Microcystis colonies sampled, while other Acidobacteria were largely absent (27), and another study found that Bryobacter relative abundance was significantly correlated with the relative abundance of certain Microcystis genotypes (37). Together with the relationships between Acidobacteria relative abundance and Microcystis relative abundance (Fig. 5), the relative abundance of Bryobacter in size-fractionated communities suggests that Bryobacter is present in some Microcystis blooms when conditions are favorable and physically attaches to Microcystis colonies, while acidobacterium CoA2 C42 likely facultatively colonizes other particles and is not specifically associated with Microcystis blooms.

FIG 6.

Percent abundance of Bryobacter and acidobacterium CoA2 C42 OTUs in size-fractionated samples from a western Lake Erie time series collected in the summer-fall of 2014. Bar colors depict size fraction.

FIG 7.

Percent abundance of Acidobacteria of interest and major Cyanobacteria taxa, Anabaena-Dolichospermum and Microcystis, in size-fractionated samples from a time series of Lake Taihu cyanobacterial blooms.

Conclusions.

This study reported insights into two novel, uncultured Acidobacteria species that are associated with phytoplankton seston in Microcystis blooms around the world. While both organisms were detected in Microcystis blooms, only Bryobacter was found directly associated with Microcystis colonies. The transcriptomic evidence supports the hypothesis that Bryobacter and acidobacterium CoA2 C42 use organic carbon and nitrogen from dissolved organic matter present in phytoplankton lysate or exudate and regenerate reduced N that may fuel growth of other microorganisms, including Microcystis, which can grow using reduced N (30, 97, 98). The data also suggest that both Acidobacteria use cobalamins released into the environment by other organisms, which may include Microcystis. The inferred reciprocal exchange of metabolites between Acidobacteria and phytoplankton suggests potential for a mutualistic relationship, but additional work is required to test this hypothesis.

MATERIALS AND METHODS

DNA and RNA extraction and sequencing.

Water samples were collected through the summer and fall of 2014 in western Lake Erie in order to monitor microbial communities and water chemistry throughout various stages of Microcystis bloom development (see references 36, 38, and 39 and Data Set S1 in the supplemental material for more detail about how the samples correspond to different times of year and Microcystis bloom development). Microbial samples were obtained from a 20-L depth-integrated water sample collected from the surface to 1 m above the lake bottom. Depth-integration was performed by collecting water with a peristaltic pump and slowly moving the length of the peristaltic pump tubing repeatedly up and down the water column until a 20-L carboy was filled completely. The integrated water column sampling method is critical to capture the full bloom community because Microcystis cells can migrate vertically throughout the water column (96, 99).

To focus on the Microcystis colony-associated fraction, 2 L of depth-integrated sample was filtered through a 100-μm-pore-size mesh, and the retentate was backwashed into a Falcon tube using altered BG-11 medium (Table S1). RNAlater was added in a 2:1 ratio with the backwash, which was then filtered onto a 1.6-μm-pore-size glass fiber filter with a syringe. The backwash was filtered onto the 1.6-μm filter immediately after resuspension in BG-11 medium, so there was likely little or no effect of BG-11 on microbial community composition. The filters were stored in a 2-mL cryovial with 1 mL of RNAlater and kept on ice during cruise transit. Upon arrival at the lab, the filters were frozen at −80°C until extraction. We cannot rule out the possibility that free-living bacteria adhered to particles or that particle-attached bacteria were flushed from particles during sample collection. However, previous work using western Lake Erie samples collected with the same methods showed that >100-μm particle-associated communities were often distinct from whole water communities and that the degree of similarity was correlated with Microcystis abundance (27). Therefore, attached microbial communities are distinct from free-living communities collected with these methods, and any similarity between >100-μm particle-associated and whole water samples is likely due to a high abundance of Microcystis colonies in the water column at the time of sampling rather than capture of free-living bacteria by particles retained during filtration.

Filters with collected biomass were thawed, folded with biomass facing inward, and rinsed with sterile phosphate-buffered saline (PBS) to remove RNAlater preservative. Filters were incubated in 100 μL Qiagen ATL tissue lysis buffer, 300 μL Qiagen AL lysis buffer, and 30 μL proteinase K for 1 h at 56°C on a rotisserie (Qiagen, Hilden, Germany). Cells were further lysed by vortexing in this lysis buffer for 10 min. Lysates were homogenized using a QIAshredder column, and DNA was purified from the filtrate using the Qiagen DNeasy blood and tissue kit according to the manufacturer’s standard protocol. The quantity and quality of DNA in each sample were determined using a NanoDrop Lite spectrophotometer (Thermo Scientific). DNA extracts were frozen at −80°C until analysis.

For RNA extraction, the filters were incubated in 600 μL Qiagen RLT+ buffer and 6 μL β-mercaptoethanol for 90 min on a rotisserie. The filters were then vortexed for 10 min and homogenized using a QIAshredder column. RNA was purified from the homogenized solution using the RNeasy kit according to the manufacturer’s standard protocol.

All sequencing was performed at the University of Michigan Sequencing Core. Paired-end DNA sequencing (2 × 125) was conducted on an Illumina HiSeq 2000 with V4 chemistry reagents with “low-input prep” using the Rubicon ThruPlex kit. RNA single-read sequencing (1 × 50) was performed on an Illumina HiSeq 2000 with V4 chemistry reagents. Before sequencing, RNA libraries were prepared with a 50/50 mix of plant and bacterial Ribo-Zero kits to remove rRNA sequences.

Metagenomic assembly.

Combined-sample assemblies (coassemblies) were generated with MEGAHIT (100) using kmin 21, kmax 141, and kmer step size of 12. We performed 3 coassemblies in total, choosing samples based on the abundance of the target Acidobacteria organisms estimated from 16S rRNA gene amplicon sequences. One coassembly was generated with the only two samples that yielded low-quality Acidobacteria genomes in pilot single-sample assemblies (Aug-4 and Aug-25 particle metagenomes from WE12). Another coassembly was constructed with all the samples in which Acidobacteria CoA2 C42 was present in corresponding amplicon sequence data sets (8 samples total), and another with the same 8 samples but with the read kmer coverage normalized to 20× prior to assembly with BBnorm in the BBTools package (101). Paired-end reads were quality and adapter screened and dereplicated with BBTools prior to coassembly (101). An additional single-sample MEGAHIT assembly was constructed on the particle size fraction sample from 4 August at WE12 following the same pipeline. This sample was chosen for assembly because it represented peak particulate microcystin (cyanotoxin) concentrations for this location (102), which is of particular societal importance because it is near the drinking water intake for Toledo, OH, a city that lost access to drinking water due to cyanotoxins in August during the year of sampling (103).

Genome binning.

Contigs were binned using a multialgorithm binning approach. Contigs were binned using differential coverage and tetranucleotide frequencies in CONCOCT (104) and Metabat2 (105) and with tetranucleotide frequencies alone using VizBin (106) and Emergent Self-Organizing Maps (ESOMs) (107). The contig size window for ESOM was 4 to 10 kbp and 2.5 to 10 kbp for the other binners. The resulting redundant bin data sets from each assembly were dereplicated using DASTool (108). For the single-sample assembly, differential coverage was estimated by mapping reads from 4 August WE12 and 25 August WE12 to the contigs. For coassemblies, differential coverage was estimated by mapping reads from the corresponding samples used to generate the coassemblies. Read mapping to contigs was performed using Bowtie2 (109).

The bins were refined manually in Anvi’o (110) using contig coverage and ward linkage clustering of tetranucleotide frequencies. Quality metrics of the refined bins were estimated using the lineage workflow in CheckM (111). All bins with contamination scores greater than 5% after refinement were eliminated from downstream analysis. Contamination scores of the final bins were considered while ignoring the amount of contamination due to strain heterogeneity. Redundant marker genes were considered to be from closely related strains if their shared amino acid identity was 95% or greater. This redundant bin data set was dereplicated using dRep (112) with 97% ANI and 60% alignment coverage cutoffs and skipping the MASH preclustering step. The final bin for the acidobacterium CoA2 C42 genome came from the 2-sample coassembly, and the final bin for the Bryobacter CoA8 C33 genome came from the 8-sample coassembly without kmer normalization.

Gene annotation and metatranscriptomic read mapping.

Gene calls and functional annotations were generated using the Integrated Microbial Genomes annotation pipeline (113), and membrane-associated proteins were predicted using TransportDB v. 2.0 (114). Genes annotated as iron complex transporters were compared to biochemically confirmed cobalamin transporters from Escherichia coli strain K-12 (UniProtKB accession numbers: P06129, P06609, and P06611) via protein BLAST v. 2.2.31+ (115). We excluded any significant hits in the results with alignment coverage less than 70% of the gene in the Acidobacteria genomes.

All predicted open reading frames were compared to proteins in the NCBI nonredundant protein database (as of 17 October 2018) via protein BLAST v. 2.2.31+ (115). Gene expression was determined by mapping metatranscriptomic reads to predicted gene sequences using nucleotide BLAST v. 2.2.31+ (115). Only alignments with percent identity of ≥95%, E value of ≤1 × 10⁻⁵, and alignment coverage of ≥80% of read length were counted. Some reads below the alignment coverage cutoff were counted if they mapped to either the start or stop end of the gene. The relative abundance of transcripts for each gene was calculated as reads mapped per gene kilobase per million reads mapped (RPKM), using total number of reads mapped to the appropriate genome. Reads were competitively mapped to both Acidobacteria MAGs and the following Microcystis reference genomes: Microcystis strain FD4, strain NIES 843, strain NIES 2481, strain NIES 2549, and strain PCC 7806SL.

Predicted gene calls, functional annotations, and metatranscriptomic gene mapping were used to infer in situ metabolism of Bryobacter and acidobacterium CoA2 C42. Due to the novelty of the Acidobacteria genomes reported here, most of the predicted protein-coding genes have low shared amino acid identities with published protein sequences with the same function (<70% shared identity to the best matches in many cases [Fig. S1]), so we present these results as putative functions and interactions of interest that require validation with future work. Because acidobacterium CoA2 C42 had a sufficient amount of reads only in the 4 August sample (Fig. S2), which coincided with an early peak in phytoplankton pigments at this station (102), this sample was the focus of reported RPKM values in the main text and figures.

16S rRNA gene phylogenetic and gANI analyses.

The 16S rRNA genes from each MAG were compared to the SILVA SSU database v. 138.1 (116) using the online SINA Aligner v. 1.2.11 (117) and classified using the approach of Wang et al. (118) in MOTHUR v. 1.43.0 (119). One MAG (CoA2 C42) was not binned with an rRNA operon, but an unbinned contig with the full rRNA operon was assigned to the bin by examining the assembly De Bruijn graph using Bandage v. 0.8.1 (120) and the paired-end mapping information. A maximum likelihood phylogenetic tree with published 16S rRNA genes from Acidobacteria available in NCBI (as of 8 November 2020) was computed with RAxML v. 8.2.4 using the GTRGAMMA nucleotide substitution model (121) and rooted and visualized using the Interactive Tree of Life webtool v 6.3 (122). The 16S rRNA genes were aligned using Clustal Omega v. 1.2.1 (123). Shared average nucleotide identity (gANI) was computed with whole-genome alignments of Acidobacteria genomes available in IMG (as of 5 November 2018) using the compare function in dRep v. 2.0.5 (112) without the MASH preclustering step. Genomes were included in the gANI analysis only if the completeness and the contamination calculated with the CheckM lineage workflow were above 90% and below 5%, respectively (111).

Amplicon data set mining.

To assess the frequency of occurrence of the Acidobacteria in Microcystis blooms, we searched for their presence in previously published rRNA amplicon data sets (24, 28, 102, 124–128), which are described in Table 3. To assess the MAGs’ abundance in western Lake Erie, the relative abundances of operational taxonomic units (OTUs) from a previously published abundance matrix were reported (27). For the Bryobacter CoA8 C33 MAG, the relative abundance of OTUs classified as Bryobacter was reported. For the acidobacterium CoA2 C42 MAG, the relative abundance of OTUs classified as Paludibaculum was reported if the 16S rRNA gene in the MAG aligned with the amplicon sequence with 97% or more shared nucleotide identity as determined via nucleotide BLAST v. 2.2.31+ (115).

Because other data sets used a range of different primer sets and did not provide abundance tables, we determined the abundance of each organism in these data sets by mapping amplicon reads to the 16S rRNA gene sequence in each MAG with BLAST v. 2.2.31+ (115). The relative abundances of Microcystis, Synechococcus, and Dolichospermum were similarly determined in these data sets by mapping amplicon reads to reference sequences. We mapped to full-length 16S rRNA gene sequences from Anabaena cylindrica PCC 7122, Microcystis aeruginosa PCC 7806SL, Microcystis aeruginosa PCC 9806, Synechococcus elongatus PCC 6301, and Synechococcus elongatus PCC 7942, as well as sequences assembled from Lake Erie metagenomes using EMIRGE (129) and classified as Microcystis, Anabaena, Dolichospermum, and Synechococcus using the classifier of Wang et al. (118) in MOTHUR v 1.43.0 (119). The relative abundance of each organism in each sample was calculated as the number of reads mapped for that given organism divided by the total number of reads in the data set.

Identification of pseudocobalamin in Microcystis cultures.

Two strains of Microcystis aeruginosa (PCC 7806 and PCC 9806) were grown on modified BG-11 growth medium (130) with the sodium nitrate concentration reduced to 2 mM in preparation for screening for pseudocobalamin production via liquid chromatography-mass spectrometry (LC-MS) analysis. The Microcystis strains were grown as batch cultures at room temperature under cool white fluorescent bulbs. The light intensity was kept between 30 and 60 μmol photons/m²/s by covering the lights with a single layer of neutral-density 0.3 filter screen (product 209R; Lee Filters, Burbank, CA). For each strain, 300 mL of late-log-phase culture was split into six 50-mL aliquots and harvested by centrifugation at 10,000 × g for 15 min, decanting liquid media, and freezing at −80°C until extraction.

Analysis of cell pellets was carried out using a previously published method (8), with some modifications. Briefly, cell pellets were resuspended in 6 mL of cold acetonitrile-methanol-water (40:40:20 ratio by volume) with 0.1% formic acid and transferred to 15-mL centrifuge tubes. The cells were lysed via bead beating with 250 mg each of 100- and 500-μm-diameter glass beads on a vortex mixer set to maximum speed for 40 s. Bead beating was performed three times, with samples resting on ice for 5 min between each treatment. The suspension was pelleted via centrifugation at 10,000 × g, and the supernatant was transferred to a round-bottom flask. The pooled supernatants from each strain were dried in a rotary evaporator under pressure of 0.3 × 10⁵ Pa at 30°C and then resuspended in a small volume of solvent A (described below) before LC-MS analysis. Extraction from ~6 g Spirulina powder was also performed as a positive control (8) following the same procedure described above for the Microcystis cells, with the exception that the powder was suspended in 10 mL of cold acetonitrile-methanol-water solution (40:40:20 ratio by volume). All extraction and processing steps were conducted under low-light conditions to minimize photodegradation of pseudocobalamin.

LC-MS analysis was carried out on a Thermo Scientific ultrahigh-pressure liquid chromatograph (UHPLC) coupled to a Q-Exactive Orbitrap high-resolution mass spectrometer equipped with an electrospray ionization (ESI) source and running in positive mode. Sample (5 μL) was injected onto a 2.6-μm Kinetex RP C₁₈ column (150- by 4.6-mm inside diameter) held at 25°C. The HPLC gradient used was 5% to 95% solvent B over 22 min, where solvent A consisted of 20 mM ammonium formate and 0.1% formic acid in water, and solvent B consisted of 0.1% formic acid in acetonitrile. MS data were collected over a mass range of 600 to 1,400 m/z, using data-dependent MS/MS analysis with 0.5-s dynamic exclusion enabled. Pseudocobalamin variants were identified by comparing the obtained compound masses and MS/MS spectra to previously reported literature values (8, 131) and the Spirulina extract.

Scripts and data availability.

All data tables and shell and R code used for the analysis are included on GitHub at the following web address: https://github.com/Geo-omics/WLE-Acidobacteria-Genomes. Acidobacterial MAG sequences and gene annotations are deposited and publicly available in the IMG database (IMG genome IDs: 2806310633 and 2806310632). Assembled MAG sequences are also available in NCBI (BioSample accession numbers SAMN20863144 and SAMN20863205). Raw read data sets are publicly available in NCBI SRA under BioSample accession numbers SAMN09102072 to SAMN09102087, and Whole Genome Shotgun projects have been deposited at DDBJ/ENA/GenBank under the accession numbers JAINDK000000000 and JAINDL000000000. Full metagenome assemblies from which the MAGs were derived are publicly available in IMG (IMG genome IDs: 3300028429 and 3300028430). Raw metabolomic data are submitted to the GNPS-MassIVE database under the following ID: MassIVE MSV000088058. Accession numbers are listed for each sample and assembly in Data Set S1 in the supplemental material.