Genomic signatures of Lake Erie bacteria suggest interaction in the Microcystis phycosphere

Microbial interactions in harmful algal bloom (HAB) communities have been examined in marine systems, but are poorly studied in fresh waters. To investigate HAB-microbe interactions, we isolated bacteria with close associations to bloom-forming cyanobacteria, Microcystis spp., during a 2017 bloom in the western basin of Lake Erie. The genomes of five isolates (Exiguobacterium sp. JMULE1, Enterobacter sp. JMULE2, Deinococcus sp. JMULE3, Paenibacillus sp. JMULE4, and Acidovorax sp. JMULE5.) were sequenced on a PacBio Sequel system. These genomes ranged in size from 3.1 Mbp (Exiguobacterium sp. JMULE1) to 5.7 Mbp (Enterobacter sp. JMULE2). The genomes were analyzed for genes relating to critical metabolic functions, including nitrogen reduction and carbon utilization. All five of the sequenced genomes contained genes that could be used in potential signaling and nutrient exchange between the bacteria and cyanobacteria such as Microcystis. Gene expression signatures of algal-derived carbon utilization for two isolates were identified in Microcystis blooms in Lake Erie and Lake Tai (Taihu) at low levels, suggesting these organisms are active and may have a functional role during Microcystis blooms in aggregates, but were largely missing from whole water samples. These findings build on the growing evidence that the bacterial microbiome associated with bloom-forming algae have the functional potential to contribute to nutrient exchange within bloom communities and interact with important bloom formers like Microcystis.


Introduction
Cyanobacterial harmful algal blooms (cHABs) occur annually in both freshwater and marine systems. These blooms have the potential to be disruptive to aquatic ecosystems due to both the scale of accumulated biomass and the release of secondary metabolites that have metabolic consequences for other organisms [1,2]. Microcystis is a pervasive genus of cyanobacteria that forms blooms on every continent except Antarctica [2]. Some species of Microcystis produce microcystins, potent hepatotoxins that can limit access to potable water [3,4]. The threats to ecosystem services and public health posed by cHABs have resulted in numerous and diverse mitigation and management strategies ranging from simple aeration of small freshwater systems [5] to application of chemicals [6,7] or barley straw [8]. In the last decade, considerable advancements have been made in the application of biotic solutions that inhibit cHABs, including those based on microorganisms, although these technologies have yet to be successfully validated beyond laboratory-scale, albeit environmentally relevant, studies [9]. Several bacteria have been identified that are capable of degrading microcystins produced by cyanobacteria or are algicidal [10][11][12]. Previous work suggests such antagonistic interactions may occur in the Microcystis phycosphere [13,14], a microenvironment that surrounds phytoplankton cells analogous to the rhizosphere in plant roots [15]. However, bacterial-phytoplankton interactions are not exclusively antagonistic, as evidence from marine phytoplankton studies suggests mutualistic relationships also exist [16]. The phycosphere provides a nutrientrich environment for heterotrophic bacteria due to the release of organic molecules by the phytoplankton, including dissolved organic carbon [17]. In the early phases of cell growth, phytoplankton release lower molecular weight molecules, such as amino acids and carbohydrates, while higher molecular weight molecules, such as polysaccharides, nucleic acids, and proteins, can be released into the phycosphere during lysis [17]. The release of molecules by phytoplankton can attract heterotrophic bacteria to the phycosphere, ultimately leading to a potential exchange of nutrients between the bacteria and phytoplankton [16,17]. The physiology of Microcystis spp. makes them well-suited for nutrient exchange with heterotrophic bacterial partners. Microcystis spp. are colonial cyanobacteria surrounded by an exopolysaccharide layer with which bacteria are tightly coupled [18][19][20]. Previously, multiple strains of bacteria have been found to impact the formation of Microcystis colonies and exopolysaccharide production [21]. Furthermore, the species of bacteria associated with the Microcystis phycosphere differ based on whether they are particle associated (> 10 μm) or free-living [22]. While we know that heterotrophic bacteria are closely associated with Microcystis colonies [23], the potential mechanisms of exchange between the partners in the freshwater cHAB phycosphere have yet to be characterized. In the current study we provide genomic data that support the previously proposed interaction by which heterotrophic bacteria (heterobionts) utilize the carbon released by Microcystis, while Microcystis may benefit from nutrient or vitamin products released by the heterotrophic bacteria [23].
Here, we report on the genomic content of five bacterial strains isolated from Microcystis aggregates in western Lake Erie in August 2017, focusing on the genetic potential for interaction with Microcystis. Access to the genomic information of Microcystis-associated heterotrophic bacteria has provided new insight into the potential microbial interactions and metabolic pathways that occur within Microcystis blooms, specifically that nutrient exchange may occur in the Microcystis phycosphere. To demonstrate the ecological relevance of these strains, we surveyed available metatranscriptomic data from Microcystis spp. blooms in North America and China. These observations show the potential for bidirectional, mutualistic interactions in the Microcystis phycosphere which could serve as a future target for cHAB mitigation.

Sample collection and environmental conditions
Surface samples were collected from four stations in the Western Basin of Lake Erie in August 2017 (WE02, WE04, WE13, and MB18) using 20 μm and 80 μm mesh plankton nets. These pore sizes were chosen to exclude free-living bacteria and enrich for those bacteria that are associated with Microcystis aggregates. Previous studies have characterized Microcystis-associated bacteria in size fractions ranging from � 3 μm up to 100 μm. Environmental conditions at the time of sample collection were reported by Boedecker et al. [24].
Isolate characteristics. Five isolates were targeted for full genome sequencing from a library of over 100 individual isolates generated from the Lake Erie bloom samples. These isolates were selected based on their N-utilization and pigment production capabilities. The five isolates selected to be sequenced were identified at the genus level via 16S rRNA and rpoB gene sequences from genomic data as an Exiguobacterium sp. (JMULE1), an Enterobacter sp. (JMULE2), a Deinococcus sp. (JMULE3), a Paenibacillus sp. (JMULE4), and an Acidovorax sp. (JMULE5) (S1-S5 Figs in S1 File). Exiguobacterium sp. JMULE1 is a gram-positive, rodshaped, motile bacterium that produces orange pigmented colonies. Members of this genus have previously been shown to impact colony formation in individual strains of Microcystis, both positively [25] and negatively [26], depending on the strains tested. Enterobacter sp. JMULE2 is a gram-negative, rod-shaped bacterium from the class Gammaproteobacteria. Multiple Enterobacter strains have been found to have microcystin degradation capabilities and induce cell aggregation by Microcystis [21,27,28]. Deinococcus sp. JMULE3 is a gram-positive rod-shaped bacterium belonging to the class Deinococci that produces pink-orange pigmented colonies; members of the Deinococcus-Thermus phylum have been previously identified in Microcystis metagenomes [13,29]. Paenibacillus sp. JMULE4 is a gram-negative, rod-shaped bacterium in the class Bacilli that produces endospores, and members of this genus have been previously identified in cyanobacterial bloom communities [30]. Isolates of Paenibacillus are commonly applied as algal bioflocculants, as they induce algal cell aggregation [31,32]. Acidovorax sp. JMULE5 is a gram-negative, rod-shaped bacterium belonging to the class Betaproteobacteria. Several strains of this genus have been isolated from samples of Microcystis, both in culture [33] and from the environment [34].
The microbiome of freshwater lakes and rivers is often dominated by members of the phylum Actinobacteria. While none of the bacteria isolated for this study were members of the Actinobacteria phylum, this is consistent with recent work demonstrating that this phylum is significantly depleted in populations closely associated with the Microcystis phycosphere [35]. In fact, members of the phyla Proteobacteria and Firmicutes are enriched in Microcystis aggregate samples compared to free water bloom samples in several studies [22,25,35,36]. It has been hypothesized that Actinobacteria likely do not rely on Microcystis-derived carbon due to actinorhodopsin activity [35,37]. Furthermore, one benefit to phycosphere bacteria may be protection from predation by zooplankton, to which ultramicrobacterial Actinobacteria are not as vulnerable [38,39].
Acidovorax sp. JMULE5 was the only isolate for which a single contig was obtained ( Table 1). Its closest sequenced relative, Acidovorax sp. KKS102 was originally isolated from soil and has been shown to degrade polychlorobiphenyl (PCB) (

Functional annotation
To better understand the dominant metabolic pathways encoded by each Lake Erie isolate, the protein families of each genome were annotated using the Subsystems (SEED) approach [49,50]. Overall, the five genomes contain the highest number of subsystems related to the Carbohydrates, Amino Acids and Derivatives, and Protein Metabolism categories (Fig 1). The genome of Paenibacillus sp. JMULE4 contained the greatest number of pathways related to Carbohydrates, while Exiguobacterium sp. JMULE1 and Deinococcus sp. JMULE3 contained a proportionately greater number of pathways related to Protein Metabolism (Fig 1). As isolates of Paenibacillus have been shown to be capable of degrading complex carbohydrates, this may be indicative of similar capabilities for the Lake Erie isolate of this genus [51, 52] and bacteria associated with Microcystis aggregates [22,36]. The Paenibacillus sp. JMULE4 genome also contained the most genes related to Dormancy and Sporulation, and this is the only isolate of the five observed to produce endospores (Fig 1).

Nitrogen utilization
Bacterial heterobionts are thought to be a source of nitrogen (N) to algae in the phycosphere [35,36,53]. In many ways this is self-evident, as respiration of biological materials from phytoplankton results in the loss of C (as CO 2 ) and residual, excess N [54, 55]. The presence of several different N-transformation genes in the bacterial genomes we examined suggests that these bacteria have the capability to act as an external source of ammonium for Microcystis. All the genomes except Exiguobacterium sp. JMULE1 contain genes for the reduction of nitrate and nitrite to ammonium ( Table 3; S2-S6 Tables in S1 File). Enrichment for this function has been previously identified in metagenomes generated from Microcystis aggregates in Lake Erie compared to whole water samples [35]. The denitrifying reductase gene clusters in the Acidovorax sp. JMULE5 genome includes genes for nitric and nitrous oxide reductase as well as cyanate hydrolysis (

PLOS ONE
N 2 [59, 60], however, the JMULE4 isolate only encodes a NifU-like protein, which is a nonessential protein for nitrogen fixation in organisms such as Dolichospermum (Anabaena) [61]. The role of these isolates in potentially providing reduced N to Microcystis likely comes from the breakdown other exogenous N sources, such as urea and nitrate.
Urease is an enzyme that catalyzes the hydrolysis of urea to ammonia and carbon dioxide (CO 2 ) [62]. In addition to serving as an N source for freshwater cyanobacteria, including Microcystis, the CO 2 released during urea hydrolysis also can act as a carbon source for Microcystis during periods of high biomass [63,64]. As the pH increases during bloom events, it becomes increasingly difficult for additional CO 2 to dissolve in the water. CO 2 availability can be impacted by static conditions (no aeration), which can cause Microcystis to stop growing [65]. The genomes of Enterobacter, Paenibacillus, and Acidovorax spp. contain genes encoding the alpha (ureC), beta (ureB), and gamma (ureA) subunits of the urease enzyme complex. The Deinococcus sp. JMULE3 genome contains genes for the alpha (ureC) and gamma (ureA) urease subunits. The Deinococcus sp. JMULE3 and Acidovorax sp. JMULE5 genomes contain all of the urease accessory genes (ureEFGD), while the Enterobacter sp. JMULE2 and Paenibacillus sp. JMULE4 genomes encode a subset of these accessory genes (ureEFD and ureFGD respectively). Enterobacter sp. JMULE2, Deinococcus sp. JMULE3, and Acidovorax sp. JMULE5 were all confirmed to be ureolytic by inducing a color change on urea slants and can grow with urea as their sole N source. Urease is a metalloenzyme that binds to and requires nickel to function [66,67]. The genomes of Enterobacter sp. JMULE2 and Acidovorax sp. JMULE5 contain the nickel-binding accessory genes ureJ and hupE. Genes for nickel incorporation proteins (hypAB) were identified in the Enterobacter sp. JMULE2 and Deinococcus sp. JMULE3 genomes. Due to its potential as a source of CO 2 , the breakdown of urea by heterotrophic bacteria in the phycosphere could act as a dual source of both carbon and N for Microcystis, as has been shown in marine systems [68].

Carbon utilization
The monosaccharide composition of Microcystis extracellular polysaccharides (EPS) has been extensively characterized, and multiple species of bacteria can use components of the Microcystis EPS as a sole carbon source [69][70][71]. The genomes of Paenibacillus sp. JMULE4 and Enterobacter sp. JMULE2 contain genes for xylose utilization, a key monosaccharide in the EPS of Microcystis [69,70]. In both freshwater and marine phycosphere communities, xylose is a known low molecular weight (LMW) carbohydrate source for associated bacteria [72][73][74]. Paenibacillus sp. JMULE4 genome contains genes for xylose isomerase, transporters, and binding components, while Enterobacter sp. JMULE2 has genes encoding XylFGHR proteins that allow for transcriptional regulation, xylose transport, and ATP binding (S2-S6 Tables in S1 File). All five of the isolates' genomes also contain genes for mannose utilization (S2-S6 Tables in S1 File), another component of the Microcystis EPS [69,70]. All of the isolates have the genetic capacity to produce mannose-6-phosphate (M6P) isomerase (S2-S6 Tables in S1 File), which converts M6P to fructose-6-phosphate (F6P), an intermediate of glycolysis. The genomes of Deinococcus sp. JMULE3, Paenibacillus sp. JMULE4, and Acidovorax sp. JMULE5 contain the glcD gene for glycolate dehydrogenase, indicating they likely have the ability to use algal-derived glycolate as a carbon source (S2-S6 Tables in S1 File). Glycolate is an organic carbon source produced from the oxygenase activity of RuBisCO during photorespiration by phytoplankton, with rates of excretion dependent upon the form of N available [73,75]. The potential utilization of glycolate by heterotrophic bacteria is indicated by the presence of the glcD gene, which encodes glycolate oxidase; this gene is now considered a biomarker for the ability to consume algal-derived carbon [76,77]. Bacterial utilization of glycolate released by phytoplankton has been examined in both marine systems and lakes [77,78].
To determine whether these organisms actively attempt to access glycolate pools during bloom events, we examined the genetic potential of bloom communities to use this C source in metagenomes from Lake Greenfield (Iowa, USA) and the active transcription of glcD in a set of transcriptomes from Lake Erie (North America) and Lake Tai (Taihu) and (Table 4). Signatures of Paenibacillus or Deinococcus glcD expression are largely non-existent (Table 4), indicating they were not actively using (or capable of using) glycolate during the bloom events sampled. While overall few reads recruited to glcD from the bloom metatranscriptomes, the greatest number of reads recruited to the Acidovorax JMULE5 glcD gene, indicating that there is some active transcription of this gene during bloom events by this species and other members of the Comamonadaceae during bloom events in Taihu and Lake Erie ( Table 4). The increased recruitment of reads from the Greenfield metagenomes is likely a function of samples being DNA rather than RNA, indicating the potential of these organisms to use glycolate, rather than active transcription. The low number of reads which recruited to the glcD sequences in these bloom libraries is likely a function of the sample collection, as these were all whole water samples rather than Microcystis-aggregates. Furthermore, little is known about seasonality of bacterial interactions in the phycosphere. There may be a specific bloom stage during which phycosphere bacteria may actively consume algal-derived carbon such as glycolate. These organisms are members of phyla that are significantly reduced in whole water samples compared to aggregates [14,35], where Actinobacteria are universally dominant in freshwater systems [14, 79,80]. Unfortunately, few if any datasets exist that measure gene expression specific to bacteria within Microcystis aggregates, although several recent studies have reconstructed bacterial functional potential within aggregates using metagenomics [22,25,35,36]. Furthermore, it is likely that the type of interactions between Microcystis and its associated bacteria may vary between synergistic or mutualistic and antagonistic depending on the stage of bloom development [74]. Taken together, the content of these isolates' genomes likely indicates carbon exchange in the Microcystis phycosphere.
The Enterobacter sp. JMULE2 genome contains genes encoding FepBCDEG proteins for the transport of ferric enterobactin (S3 Table in S1 File). Genes encoding the enterobactin biosynthesis pathway proteins EntBSH are also present (S3 Table in S1 File). Enterobactin siderophores are characteristic of the Enterobacteriaceae family and are amongst the strongest siderophores with a high affinity for iron [82]. The genomes of 115 Microcystis aeruginosa isolates in Genbank do not contain a gene encoding enterobactin esterase (fes), the enzyme necessary to remove iron from the enterobactin siderophore. If Microcystis spp. cannot use enterobactin, it is possible that iron scavenging by phycosphere bacteria could be competitive with their cyanobacterial host during specific phases of bloom development [74]. The Enterobacter sp. JMULE2 genome also contains iucA-D genes necessary for aerobactin synthesis (S3 Table in S1 File). Aerobactin siderophores do not have as strong an affinity for iron as enterobactins, but have an advantage for bacterial growth in iron-limited conditions [83]. Members of the bloom-forming genus Dolichospermum (formerly Anabaena) can use ferric aerobactin in culture, although it is not considered a robust iron donor for cyanobacteria [84].
The genome of Deinococcus sp. JMULE3 contains genes for isochorismate synthase (S3 Table in S1 File), a precursor of siderophores including enterobactin [85]. Isochorismate synthase is necessary for the synthesis of salicylic acid for plant defense [86]. The Paenibacillus sp. JMULE4 genome contains genes related to bacillibactin and anthrachelin siderophores (S3 Table in S1 File). The genome also contains genes for Feu A-C proteins for Fe-bacillibactin transport [87]. Bacillibactins are catechol-based siderophores that are structurally like enterobactin siderophores which are produced by different members of the Bacillus genus, including Bacillus anthracis [88]. These siderophores have also been described in a Paenibacillus honeybee pathogen [89]. Paenibacillus sp. JMULE4 genome contains genes for anthrachelin uptake transporters (S3 Table in S1 File).
The Ton and Tol transport systems are used to transport ferric-siderophore complexes and vitamin B 12 across the cell membrane [90]. The genomes of Acidovorax sp. JMULE5, Deinococcus sp. JMULE3, and Enterobacter sp. JMULE2 contain genes involved in the Ton and Tol transport systems. All three organisms also have genes for TonB-dependent receptors (S3

Production of auxins
Each of the five sequenced bacterial genomes encode genes for the biosynthesis of tryptophan, and four encode homologues of ipdC, the gene which encodes indole-3-pyruvate decarboxylase (S2-S5 Tables in S1 File; Fig 2). Sequences of ipdC fall into four clusters based on similarity to sequences of known function; Enterobacter sp. JMULE2 belongs to cluster I, encoding indolepyruvate decarboxylase, along with other members of the Enterobacter genus and close relatives such as Citrobacter (Fig 2). Acidovorax sp. JMULE5 falls in cluster II, whose members encode an α-keto decarboxylase, while Exiguobacterium sp. JMULE1 and Paenibacillus sp. JMULE 4 belong to cluster IV, which encodes a phenylpyruvate decarboxylase (Fig 2). IpdC and its homologues catalyze the second reaction in the indole-3-pyruvic acid (IPA) IAA synthesis pathway, converting IPA to indole-3-acetaldehyde (IAAld) (Fig 2). When supplemented with 5 mM tryptophan, all five isolates were confirmed to produce IAA via colorimetric assay, with a range of 3.3 μM (Exiguobacterium sp. JMULE1) to 47.3 μM (Acidovorax sp. JMULE5) after a 24-hour incubation (S6 Table in S1 File). The closest relative of JMULE1, Exiguobacterium sp. MH3, produces auxins that are hypothesized to play a role in growth-promoting activity in its aquatic plant host [44]. Production of auxins by native bacteria increases cell density of freshwater eukaryotic microalgae [91]. Tryptophan is an important precursor for indole-3-acetic acid (IAA), the main auxin that occurs in plants. Amin et al. (2015) found that a bacterial consortium promoted diatom cell division due to the secretion of an auxin synthesized from diatom-derived tryptophan. Microcystis aeruginosa NIES 843 has the genetic capability to produce tryptophan and could therefore serve in a similar role [16,92]. The production of auxins like IAA by these isolates suggests a possible important growth-promoting effect of the bacteria on Microcystis and other freshwater cHAB formers.

Quorum sensing and signaling
The Enterobacter sp. JMULE2 genome contains genes for AI-2 transport and processing. Quorum sensing, or cell-to-cell communication in bacteria, relies on the production of signaling molecules known as autoinducers [93]. Autoinducer 2 (AI-2) is an autoinducer produced by many different bacterial species. This lsrACDBFGE operon has also been described in Salmonella enterica and Escherichia coli [94,95], both members of the Enterobacteriaceae family with Enterobacter. In these organisms, the lsrACDB genes encode AI-2 transporter components while the rest of the genes in the operon are needed from processing AI-2 once it is internalized [95]. Its closest relative (based on rpoB identity), E. asburiae sp. L1 ( Table 3) is known to produce an array of quorum sensing molecules, including AHLs [41].

Vitamin production
Many algae, including some cyanobacteria [96][97][98], require vitamin B 12 for growth yet are unable to produce it and must rely on exogenous B 12 [97]. Microcystis requires vitamin B 12 for the methionine biosynthesis pathway. This pathway requires B 12 as a cofactor for a type-II MetH enzyme [48,97]. Microalgae can obtain vitamin B 12 directly from bacterial interactions [97]. Paenibacillus sp. JMULE4, Acidovorax sp. JMULE5, and Deinococcus sp. JMULE3 genomes all contain genes for cobalamin (vitamin B 12 ) synthesis (MTR, cobY, cobU, cobQ, bluB) (S1-S5 Tables in S1 File). Vitamin B 7 (biotin) is a cofactor that is essential for carboxylase enzymes, including acetyl coenzyme A (CoA) carboxylase which is used in the production of fatty acids. As with vitamin B 12 , cultures of Microcystis are supplemented with B 7 in the growth medium. All five of the bacterial genomes contain genes for biotin biosynthesis (S1-S5 Tables in S1 File). In marine systems, 22% of HAB forming organisms are vitamin B 1 (thiamine) auxotrophs [99]. The Enterobacter sp. JMULE2, Paenibacillus sp. JMULE4, and Acidovorax sp. JMULE5 genomes contain genes for vitamin B 1 synthesis (S1-S5 Tables in S1 File) and could provide Microcystis with this essential nutrient in natural populations.

Potential for interaction in the Microcystis physcosphere
The reductionist approach to understanding the dynamics of HABs is shifting toward a more dynamic model. No organism lives in isolation, including the phytoplankton which form HABs. The phycosphere, a potential hotbed for interactions between algae and their heterotrophic bacterial microbiome, can be considered a counterpart of the terrestrial rhizosphere. Within this microenvironment, exchange of nutrients and other compounds drive the mutualistic relationships between phytoplankton like Microcystis and their associated bacteria. Genomic analysis of five bacterial isolates from a 2017 Microcystis bloom in Lake Erie indicate these bacteria have the genetic potential for bidirectional exchange of nutrients and other growthpromoting compounds such as vitamins and hormones with their photosynthetic partner. The carbon-rich EPS produced by Microcystis contains the sugars mannose and xylose, which can be taken up and utilized by all five of the Lake Erie isolates (Fig 3). For decades, it has been posited that the bacteria associated with the Microcystis mucilage likely benefit from access to these various forms of carbon [38,100]. Less is known, however, about how Microcystis may benefit from this close association with heterotrophic bacteria. One potentially important mechanism of exchange may be the bidirectional exchange of carbon. Bacteria respire CO 2 , as well as produce it as a byproduct of the hydrolysis of urea. During peak bloom conditions, Microcystis populations in Lake Erie have increased transcription of genes involved in carbon concentration, suggesting the potential for CO 2 limitation in dense bloom populations [101]. Respiration by associated bacteria could provide a supplemental source of CO 2 for Microcystis (Fig 3) [102]. Many phytoplankton, including Microcystis require an exogenous source of vitamins B 12 , B 1 , and B 7 [97,99]. Phycosphere bacteria have been identified as potential sources of these vitamins for marine bacteria, and four of the five Lake Erie isolates may serve in this capacity during Microcystis blooms (Fig 3). In addition to providing vitamins, these bacteria may also provide reduced N to Microcystis during periods of N stress, which is common in systems such as Lake Erie during peak bloom conditions (Fig 3). Interestingly, all five of the sequenced Lake Erie isolates can produce the plant hormone IAA when supplemented with tryptophan. IAA has been shown to have growth promoting effects on both freshwater and marine algae [16,91], and may have an important role in the mutualistic exchange that occurs between Microcystis and its associated bacteria (Fig 3). The genetic capacity of these bacteria to provide critical vitamins and other nutrients provides new insight into the role that biotic interactions may have in the development of Microcystis blooms. Further culture studies will reveal the mechanisms which underlie these hypothesized multidimensional interactions illustrated in the genomic potential of these five bacterial isolates.

Sample collection, isolation, and identification
Water samples were collected from Lake Erie on August 9, 2017 from the Ohio State University Stone Lab R/V Gilbraltar III. Samples were taken from four different stations: WE02 (N 41˚45.777', W 83˚12.931'), WE04 (N 41˚49.634', W 83˚11.659'), WE13 (N 41˚44.619', W 83˚08.081'), and MB18 (N 41˚44.886', W 83˚24.061'). 20 μm and 80 μm mesh plankton nets were used to collect samples from each site to ensure only those bacteria tightly associated with the Microcystis colonies would be isolated. 150 mL were collected and stored on ice until transport back to the laboratory. Both general (LB agar) and selective (CT medium with 100 μM urea agar) media were used for the isolation of bacterial samples [63]. Bacterial isolates are maintained on CT-TY, CT medium with 1 g/L tryptone and 1 g/L yeast extract [103]. Roughly 500 μL of each water sample was plated onto each medium and incubated at 26˚C and 32˚C, for 48 hours (LB/CT-TY) or seven days (CT-urea). Single colonies were re-streaked onto new plates of the respective media until the isolates were pure as confirmed by microscopy.
Isolates that were grown on urea-supplemented CT were tested for urea utilization capabilities. Briefly, the isolates were inoculated into urea broth and urea agar slants (Hardy Diagnostics) and observed for color change to indicate urease activity via ammonia production. To extract DNA for ureC screening, turbid overnight cultures were pelleted at 17949 x g. The pellet was resuspended in 500 μL of sterile water and heated at 95˚C in a dry bath for 15 minutes. After heating, the tubes were centrifuged again for one minute. The supernatant containing the DNA template was used for PCR amplification of the ureC gene using primers IGKAGNPforward (5'ATHGGIAARGCIGGIAAYCC3') and HEDWGA-reverse (5'IGYICCCCART CYTCRTG 3') (modified from Collier et al. [104]). The PCR program was as follows: 94˚C for one minute, 25 cycles of 94˚C for 30 seconds, 53˚C for 30 seconds, and 72˚C for 45 seconds, with a final extension of 72˚C for 10 minutes.

DNA extraction and sequencing
Isolates were grown from a single colony in CT-TY broth for 48 hours at 26˚C or 32˚C (JMULE4). The DNeasy UltraClean Microbial Kit (Qiagen) was used for DNA extraction according to manufacturer's instructions. A NanoVue Plus (GE Healthcare) was used to check the quantity and purity of the DNA. The genomic DNA was sent to Genewiz (South Plainfield, NJ, USA) for sequencing on the PacBio Sequel System. The PacBio SMRTbell library was prepared according to the manufacturer's instructions. The SMRTbell libraries were then sequenced with the PacBio Sequel System.
The SMRTLink suite was used to demultiplex the sequence libraries that were generated from the PacBio Sequel platform. These demultiplexed sequence files (BAM and FASTQ) were then provided by the GENEWIZ sequencing facility for assembly and annotation.

Genome assembly and annotation
The genomes were assembled using the PacBio de novo assembly pipeline on the CLC Genomics Workbench plugin CLC Genome Finishing Module (Qiagen) using default parameters. First, the raw reads were imported into CLC Genomics Workbench and corrected for sequencing errors and untrimmed adapters. The error-corrected reads were then assembled into contigs with the "de novo Assemble PacBio Reads" tool. The corrected reads were then mapped to the contigs to close gaps and join contigs and subsequently mapped to the larger contigs.
The genomes were annotated with the NCBI prokaryotic genome annotation pipeline (PGAP) [105] and RAST [106,107]. The annotated genomes were viewed on the RAST SEED Viewer Amino acid alignments were generated using T-coffee (http://tcoffee.crg.cat/apps/tcoffee/) and phylogeny was calculated using the Maximum-Likelihood method with a Bootstrap test of phylogeny (1000 iterations) in Mega X [108][109][110].

Confirmation of IAA production
Bacterial isolates were tested for the capability to produce the auxin indole-3-acetic acid (IAA). The isolates were inoculated in 1mL of CT-TY broth fortified with 5 mM L-tryptophan and incubated while shaking at 26˚C, with the exception of 32˚C for Paenibacillus, for 48 hours. 200 μL of axenic Microcystis and 2 μL of broth containing each isolate were added to a 96-well plate in triplicate. 200 μL of the isolate alone were also added in triplicate to the 96-well plate. The well plate was then left to incubate for 24 hours at 26˚C. 150 μL of Salkowski's reagent (0.5 M FeCl 3 and 70% perchloric acid) were added to each well in the dark [102]. The plate was incubated in the dark for 30 minutes before measuring absorbance at 530 nm to observe color change and compare to a standard curve.

Prevalence in environmental data
To determine whether these isolates are present during bloom conditions, we analyzed sequence libraries from blooms that occurred in three locations: Taihu (China) [103], Lake Erie (North America) [4,103], and Lake Greenfield (North America). Recruitments were performed in CLC Genomics Workbench with a similarity fraction of 0.8 and a length fraction of 0.5 to capture closely related organisms [104]. All recruited reads were then classified via blastn in CLC Genomics Workbench, and those that did not match each isolate at the family-or genus-level based on e-value were excluded. The Lake Greenfield samples were collected on 20 July 2018 from the dock and a second site at a drainage pipe. The samples were collected during a period when the town of Greenfield detected microcystin in the drinking water supply. One liter of water was filtered through Sterivex units and kept on ice until frozen at -20˚C (~2 hours). Samples were extracted using the DNEasy1 PowerWater1 (Qiagen) extraction kit [105]. Genomic DNA was sent to GeneWIZ1 for library preparation and sequencing on the Illumina HiSeq4000 platform to generate 150 bp paired end reads. Reads are available at the NCBI SRA under BioProject PRJNA610583.