Bacteria plays a critical role in the ecosystems of freshwater lakes driving global biogeochemical cycles. As such, numerous surveys of bacterial communities within freshwaters have been conducted, including those examined by Newton et al. (2011) and more recent studies (e.g., Oh et al., 2011; Jones et al., 2012; Han M. et al., 2016; Hayden and Beman, 2016; Lee et al., 2016; Llorens-Marès et al., 2016; Yang et al., 2016; Linz et al., 2017; Ren et al., 2017; Salmaso et al., 2018). Differences in the bacterial communities of these lakes can vary significantly (Newton et al., 2011) as local environmental factors often play a role in structuring these communities (Hayden and Beman, 2016). The community structure of a single lake can vary over time, over the course of a single day (Jones et al., 2012) as well as over years (Linz et al., 2017). Nevertheless, many bacterial groups within freshwaters have a global distribution: Proteobacteria, Actinobacteria, Bacteroidetes, Cyanobacteria, and Verrucomicrobia (Zwart et al., 1998; Newton et al., 2011). Dominant members of these phyla include the Proteobacteria genus Acinetobacter (Lee et al., 2016), the acI cluster of Actinobacteria (Newton et al., 2007; Neuenschwander et al., 2018), and the Bacteroidetes genera Sphingobacteriales, Sediminibacterium, Fluviicola, and Flavobacterium (Salmaso et al., 2018). These communities, however, can be impacted by disruptions, such as cyanobacterial blooms (Eiler and Bertilsson, 2004; Scherer et al., 2017). Furthermore, bacteria can be introduced into a community via effluent from water treatment plants (e.g., Escherichia coli and Enterococcus) and stormwater runoff (e.g., Pseudomonas and Sphingomonas). Such introductions can include species harboring antibiotic resistance genes into the lake ecosystem (Drury et al., 2013; Marti et al., 2013; Fisher et al., 2015; Saarenheimo et al., 2017; Chu et al., 2018; Lorenzo et al., 2018).

The Great Lakes are one of the planet’s largest freshwater ecosystems (Fuller et al., 1995) and as dynamic as coastal oceans (Rao and Schwab, 2007). Not unsurprisingly, microbial surveys of the Great Lakes have found that they share common bacterial diversity with other freshwater habitats (Mueller-Spitz et al., 2009; Mou et al., 2013; Wilhelm et al., 2014; Fisher et al., 2015; Malki et al., 2015a; Newton and McLellan, 2015; Sible et al., 2015). As the Great Lakes are the source of drinking water for millions of North Americans, protection is of critical concern in the face of the many known effects of urbanization on water quality (Foley et al., 2005). This concern is perhaps most evident within Chicago, whose city and surrounding suburbs sprawl along the Lake Michigan shoreline. Lake Michigan waters are routinely monitored (Whitman and Nevers, 2008; Wong et al., 2009; Haack et al., 2013; Oster et al., 2014; Wijesinghe et al., 2015), and the incidence of illness attributed to these waters has been extensively studied (Dorevitch et al., 2011; DeFlorio-Barker et al., 2018). Aside from prompt and thorough testing, contamination via stormwater and fecal pollution has been a consistent concern for all cities along the Lake Michigan shore (Bower et al., 2005; Whitman and Nevers, 2008; Wong et al., 2009; Fisher et al., 2015; Templar et al., 2016). Despite these concerns, the study of Newton and McLellan (2015) found that, at least in the Milwaukee area, the dispersal of bacteria introduced from the urban environment was limited.

Freshwater pseudomonads may be persistent or ephemeral members of the community. Dispersal of pseudomonads from agricultural to non-agricultural environments may be the result of the water cycle (Morris et al., 2008, 2014). As a whole, pseudomonads are often not abundant within freshwater environments (Newton et al., 2011). When pseudomonads are abundant within freshwaters, their source is typically associated with storm waters (Hoadley and McCoy, 1966; Oliveri et al., 1977; Selvakumar and Borst, 2006; Fisher et al., 2015). Culture-independent, 16S rRNA gene amplification surveys of freshwaters, rarely detect pseudomonads indicative of their absence/low abundance. In our prior surveys of the Lake Michigan nearshore waters, Pseudomonas was rarely detected via 16S rRNA sequencing (Malki et al., 2015a; Sible et al., 2015). Nevertheless, diverse species of Pseudomonas have been isolated from freshwaters (Drucker and Panasyuk, 2006; Morris et al., 2008; Pietsch et al., 2017), including the Great Lakes Erie, Ontario and Superior (Bennett, 1969; Chatterjee et al., 2017). Furthermore, bacteriophage (phage) infectious of Pseudomonas species have been routinely found within Lake Michigan nearshore waters (Malki et al., 2015b; Watkins et al., 2018), thus suggesting the presence of the bacterial host. This motivated the study presented here in which Pseudomonas species found within Chicago’s nearshore waters of Lake Michigan were sequenced and characterized.

Water collected from the Chicago nearshore waters of Lake Michigan in the summer of 2009 and 2016 were plated on six different medias. The plated samples were monitored over 5 days in which the number of bacterial colonies was counted. As the medias used selected for different species of culturable bacteria, variation between the growth rates and total growth between media was expected. The “less-selective” medias – LB, Eugonics and EMB – always exhibited more colonization than the “more-selective” Streptococcus, API, and Centrimide medias; this is true across collection days and sampling sites (Supplementary Data Sheet 2). Distinct colonies were selected from the plates and identified by 16S rRNA gene sequencing (Figure 1). The genera identified were representative of species which are known to be present within freshwaters and more critically are amenable to the conditions of culture used in this study. The most abundant genus detected was Pseudomonas with several different species identified. Given the number of the pseudomonads recovered, we sequenced the genomes of 15 representatives of the diversity detected, each with a unique 16S rRNA gene sequence. Twelve of the isolates came from samples taken directly from the water, two from puddles on the beach, and one from the soil of the beach. Sequencing statistics, including the number of reads generated, the number of contigs, N50 scores, and coverage can be found in Supplementary Table 2. Table 1 summarizes the 15 genomes sequenced, including their GenBank accession numbers.

As prior studies have noted that the 16S rRNA gene sequence can be insufficient in distinguishing species within the genus, we employed two methods to determine the species present. First, Pseudomonas species can be differentiated based upon the sequence homology of four housekeeping genes: the 16S rRNA gene sequence, gyrB, rpoB, and rpoD (Gomila et al., 2015). Figure 2A shows the concatenated sequence phylogenetic tree for the 15 genomes sequenced here and complete RefSeq Pseudomonas genomes, with Cellvibrio japonicus as the outgroup. Second, the ANI was calculated for the 15 genomes sequenced here, RefSeq Pseudomonas genomes, and the C. japonicus outgroup (Figure 2B). ANI values ranged between 78.5711% (RefSeq genomes P. savastanoi pv. phaeseolicola 1448A and P. aeruginosa PAO1) and 99.9975% (P. fulva str. LB-090714 and P. fulva str. 57B-090714). Eight Pseudomonas species were identified as well as three isolates for which a species designation could not be made. Two of these species, Pseudomonas sp. MB-090624 and Pseudomonas sp. LB-090624, clade with P. putida (Figure 2A). These two species have a pairwise ANI of 88.3160% and are more similar to each other than any of the other Lake Michigan isolates (Figure 2B). Of the RefSeq sequences included in this analysis, they are most similar to P. putida KT2440 (87.8185–90.0128% ANI); this is nevertheless below the 95% threshold expected of strains belonging to the same species. The concatenated sequence phylogenetic tree and ANI analysis, coupled with BLAST results for the individual contigs (Supplementary Table 1), informed the species classifications made. In the case of P. jessenii str. LBp-160603, further investigation was required; the gene annotations for this strain were compared to those previously identified for differentiating P. jessenii from other close relatives (Verhille et al., 1999) (Supplementary Table 3). Furthermore, upon submission of the 15 genome sequences, NCBI confirmed species designation via ANI against the type strains and proxytype strains that are already in GenBank (Federhen et al., 2016).

The 15 Pseudomonas genomes sequenced here vary in size from 4.01 to 7.08 Mbp, encoding 3812 to 6342 protein sequences (Table 1). Genome annotations can be found in Supplementary Table 4 and in GenBank (see Table 1 for accession numbers). No evidence of the CRISPR-Cas system was found. Annotations were compared identifying core subsystems present within the genomes as well as strain specific pathways predicted by the RAST web service (Aziz et al., 2008) (Supplementary Table 5). Comparative analysis revealed 175 subsystems that varied in their presence/absence amongst the 15 strains (Supplementary Table 6). From this list we probed deeper to identify strain- and group-specific attributes distinguishing the Lake Michigan Pseudomonas strains. While all genomes contained genes associated with ammonia assimilation, strains varied in the presence of functionality associated with other roles of nitrogen fixation (Supplementary Table 5). Variation between the strains was also observed with respect to sources for sulfur and carbon (Supplementary Table 5). Several of the strains also included gene annotations for the synthesis of achromobactin, aerobactin, and pyoverdine siderophores (Table 2). Evidence for the synthesis of all three of these types of siderophores was found within the genome of Pseudomonas sp. MB-090624. In contrast, the genome of P. mosselii LBp-160603 does not include any genes associated with siderophore synthesis. Siderophore production was evaluated in the lab using a universal siderophore assay (see section “Materials and Methods”). Siderophore production was prominent within P. koreensis str. 57B-090624, P. fulva str. LB-090714, P. fulva str. 57B-090714, and P. alcaligenes str. MB-090714 (Supplementary Data Sheet 3).

Genome annotations were further mined for virulence factors and defense mechanisms. All 15 strains encoded for genes associated with bacteriocin (colicin E2 and V) resistance, multidrug efflux pumps, and resistance to fluoroquinolones; resistance to Fosfomycin, Vancomycin, Spectinomycin, and Streptothricin was also detected in several of the strains (Supplementary Table 6). Sensitivity was tested for three of the antibiotics, Ciprofloxacin (a fluoroquinolone), Fosfomycin, and Vancomycin, via the Kirby-Bauer (KB) disk diffusion susceptibility method. As shown in Table 3, all of the strains tested exhibited sensitivity to Ciprofloxacin. However, all of the Lake Michigan Pseudomonas genome annotations included gene(s) associated with resistance to fluoroquinolones. While resistance to Fosfomycin was predicted for several strains, resistance was only observed for one of the strains predicted: the P. protogens str. MB-090624. Vancomycin resistance was observed for all of the strains tested (Table 3), although this resistance was only identified in three of the strains: P. jessenii str. LBp-160603, P. protogens str. MB-090624, and P. protogens str. MB-090714 (Supplementary Table 6).

Given prior evidence of co-selection of antibiotic and heavy metal resistance in freshwater bacteria (see review, Di Cesare et al., 2016), we next turned our attention to heavy metal resistance genes within the strains. All 15 strains encoded genes conferring resistance to arsenic, cobalt-zinc-cadmium, and copper. Resistance to chromium was also identified for the Pseudomonas strains, with the exception of. the P. fulva and P. mosselii strains, and the P. chlororaphis LBs-160603 strain encoded genes associated with resistance to zinc. Genes associated with resistances to mercury were only annotated for the genome of Pseudomonas sp. MB-090624. To test resistance to mercury, this strain as well as one of the P. fulva and one of the P. mosselii strains were grown in the presence of Hg(II). Even in the presence of 100 μM of mercury, no effect was detected on bacterial fitness within the Pseudomonas sp. MB-090624 strain (Figure 3A). Resistance to copper was also tested for nine of the Pseudomonas strains, all of which included a coding region annotated for copper tolerance (Figure 3B). Four of the strains, P. koreensis str. 57B-090624, Pseudomonas sp. LB-090624, P. mosselii str. LB-090624, and Pseudomonas sp. 57B-090624, saw only minimal reduction in fitness in the presence of high levels of copper (4 mM).

All 15 of the genome annotations produced by RAST included phage-related genes. To investigate further, each genome was examined using PHASTER (Arndt et al., 2016). Twenty-two intact, 11 incomplete, and one questionable phage sequence (prophage) was identified amongst the 15 genomes. Only Pseudomonas sp. LB-090624 did not contain an intact prophage, containing a single incomplete prophage. The prophages identified here exhibited homology to prophages of Pseudomonas genomes in GenBank. The one exception being the prophage identified for the P. alcaligenes str. MB-090714 sequence; it had only modest sequence homology to an uncultured Caudovirales phage (Supplemental Table 7). Focusing on the intact phages identified, 19 were unique. The intact prophage sequences are provided in Supplementary Data Sheet 4. Two of the integrated prophages identified within the P. koreensis strains, 35.2 and 37.3 kbp in length, were similar (>95%) to each other and a prophage identified within the sequence of Pseudomonas moraviensis BS3668 (GenBank: LT629788) (Figure 4A). Two prophages 34.5 kbp (Figure 4B) and 53.7 kbp (Figure 4C) in length (subsequently referred to as Pfulva1 and Pfulva2, respectively) were identified in both P. fulva strains. While their sequences were identical within the two P. fulva strains isolated here, they were distinct from sequence within the nr/nt database (Supplementary Table 7). Moreover, no sequence homology was identified between the Pfulva1 and Pfulva2 prophages. The annotated Pfulva1 sequence includes an integrase, C-5 cytosine-specific DNA methylase, and terminase, amongst structural phage genes, lysis proteins (holin and spanin), and hypothetical proteins. Pfulva2 encodes several virulence factors, including a Clp protease, DNA methyltransferase, and antitoxin (Supplementary Tables 8, 9). The presence of these two prophages were further confirmed via PCR amplification.

In 2014 and 2015 we conducted high-throughput 16S rRNA sequencing surveys of the nearshore waters of Lake Michigan, which included the Montrose and 57^th Street Beaches, and detected little or no pseudomonads (Malki et al., 2015a; Sible et al., 2015). Thus, Pseudomonas is likely not a dominant constituent of these waters. Even so, several different species of Pseudomonas have been isolated from other Great Lakes waters via targeted gene sequencing (Chatterjee et al., 2017). Here we have begun a comprehensive catalog of this diversity via whole genome sequencing. Through culture-dependent methods we were able to isolate 11 different Pseudomonas species from the Chicago nearshore waters. P. chlororaphis, P. fulva, P. jessenii, P. koreensis, P. mosselii, and P. protegens, often found in soil, can protect associated plants from fungal and bacterial pathogens (Xie et al., 2003; Deora et al., 2010; Ramette et al., 2011; Liu et al., 2014; Deng et al., 2015; Wu et al., 2018) through the production of diverse molecules (see review, Haas and Défago, 2005). The Pseudomonas species isolated here do not commonly cause disease in humans. P. alcaligenes, P. fulva, and P. mosselii, also commonly found in soil, are rare opportunistic human pathogens (Valenstein et al., 1983; Almuzara et al., 2010; Leneveu-Jenvrin et al., 2013; Suzuki et al., 2013). We did not isolate any strains of the well-studied human pathogen P. aeruginosa. This was initially surprising given that P. aeruginosa has been detected and/or isolated from stream, river, and lake freshwaters (Pellett et al., 1983; Römling et al., 1994; Pirnay et al., 2005; Khan et al., 2007; Selezska et al., 2012; Magalhães et al., 2016), is prevalent within urban wastewaters (Slekovec et al., 2012), and is the host for phages we previously isolated from these waters (Malki et al., 2015b). However, prior sampling of Lake Erie waters similarly did not find P. aeruginosa either (Chatterjee et al., 2017).

Whole genome sequencing of the Lake Michigan Pseudomonas isolates provides insight into their evolutionary relationships. As Figure 2 shows, these strains represent a diversity of species within the Pseudomonas tree. As prior studies have found that 16S rRNA sequences alone cannot distinguish between Pseudomonas species (Mulet et al., 2012; Gomila et al., 2015), two methods were employed here: the concatenated housekeeping genes and whole genome ANI. Only the two strains identified as P. protegens, the two strains identified as P. koreensis, and the two strains identified as P. fulva have a pairwise ANI value exceeding the threshold demarking species delineation (Richter and Rosselló-Móra, 2009). While phylogenetic and ANI analyses both identified the P. fulva strains here within the P. putida group, they do not clade with the RefSeq P. fulva 12-X strain. Prior studies evaluating the ANI (e.g., Han N. et al., 2016; Peña et al., 2016) and phylogeny for this genus (Gomila et al., 2015) have likewise found the P. fulva 12-X genome distant from the P. putida group. Until quite recently, this was the only complete genome for the species available. Earlier this year, another complete genome was produced: P. fulva FDAARGOS_167 (NZ_CP014025), isolated from sputum of a Cystic Fibrosis patient (per the genome’s GenBank record metadata). We found that this new P. fulva strain also has an ANI value less than that expected for two representatives of the same species (Supplementary Data Sheet 5), suggesting that P. fulva 12-X is a distinct species (i.e., not fulva). The three isolates for which a strain determination could not be made – Pseudomonas sp. MB-090624, Pseudomonas sp. LB-090624, and Pseudomonas sp. 57B-090624 – may be representatives of new species.

Investigation of the annotations generated for the 15 Lake Michigan genomes identified unique features including metabolic and nutrient utilization pathways (Supplementary Table 6). Prevalent amongst the Lake Michigan Pseudomonas strains is the presence of gene annotations for the synthesis of siderophores (Table 2). We were able to verify siderophore production within four of the 14 strains containing annotations for siderophores using the universal siderophore assay (Supplementary Data Sheet 3). It is important to note that siderophore production within the other 10 strains is likely; however, we were unable to cultivate these strains on the universal siderophore assay media. Although these strains were cultivated routinely under the same laboratory conditions, the universal siderophore media may exclude a vital nutrient/resource for growth. Moreover, it is well-known that species-specific siderophore systems exist amongst the pseudomonads (Meyer et al., 2002). Siderophore production within most of the pseudomonads from Lake Michigan suggests a possible role in enhancing iron abundance within the water column, similar to speculations for siderophore-producing bacteria within marine surface waters (see review, Ahmed and Holmström, 2014). Furthermore, recent evidence found that siderophores play a role in the interactions between co-occurring Pseudomonas strains and likely drive diversification in nature (Butaitė et al., 2017).

The Lake Michigan pseudomonads encode for several antibiotic resistance genes (Supplementary Table 6). Fluoroquinolone resistance was identified within the annotations of all 15 Lake Michigan Pseudomonas strains. Prior studies have noted its prevalence amongst clinical Pseudomonas samples (Horii et al., 2005; López-Causapé et al., 2018) as well as aquatic isolates (Adachi et al., 2013; Johnning et al., 2015; Chu et al., 2018). KB tests performed here using the fluoroquinolone Ciprofloxacin, however, revealed that all of the strains were actually sensitive to the antibiotic (Table 3). The annotated fluoroquinolone resistance clearly does not confer resistance to Ciprofloxacin. The Lake Michigan pseudomonads may be resistant to other fluoroquinolones not tested here, as a prior study found P. aeruginosa strains vary in their sensitivity to the fluoroquinolones Ciprofloxacin, Levofloxacin, and Moxifloxacin (Grillon et al., 2016). While the genome annotations also indicated that some strains encode for resistance to Fosfomycin and Vancomycin, KB diffusion assays were often found to contradict these predictions (Table 3). KB assays found that all 15 of the strains are resistant to Vancomycin, nine exhibiting no inhibition of growth. Therefore, the annotations alone cannot definitively characterize a strain’s sensitivities to antibiotics. Correlating mutations within these coding regions and confirmed sensitive and resistant strains is necessary to improve annotations and antibiotic resistance predictions.

As previous research has found, heavy metal resistance genes can cause positive selection of antibiotic resistance genes (Baker-Austin et al., 2006; Pal et al., 2015; see review Di Cesare et al., 2016). The Lake Michigan Pseudomonas strains encode for resistance of several heavy metals. In the bioinformatic study of Pal et al. (2015) investigating co-occurrence patterns of resistance genes, the only metal resistance genes commonly co-occurring with antibiotic resistance genes were for mercury resistance. Only one of the Lake Michigan Pseudomonas strains, Pseudomonas sp. MB-090624, was shown to tolerant of high levels of mercury (Figure 3A). Mercury emissions remain a concern in the Chicago area (Sullivan and Mason, 1998; Gratz et al., 2013); mercury emissions within the area did not see strong regulation until 2006 with the Illinois EPA mercury and Clean Air Interstate Rule. Recently, an increased level of mercury in Lake Michigan fish has been observed (Blukacz-Richards et al., 2017). The multiple heavy metal resistance genes identified within the pseudomonads may prove beneficial within these urban waters in response to potential contaminations.

As our group’s prior work in these waters has focused on isolation and characterization of phages, and in particular Pseudomonas infecting phages (Malki et al., 2015b; Watkins et al., 2016, 2018), we were not surprised to find prophages within the 15 Lake Michigan Pseudomonas genomes. The P. koreensis prophage, identified in isolates from sampling sites 17 mi apart on two separate collection dates, may be (or have been at that time) a frequent member of the Lake Michigan nearshore community. We may hypothesize this to be true for the Pfulva1 and Pfulva2 prophages as well (Figure 4B,C). The P. fulva prophages most closely resemble Pseudomonas phage φCTX and the unclassified Cvm10 virus Enterobacter phage Arya, respectively. It is important to note that these similarities are tenuous; only ∼50% of these prophages exhibit sequence homology to a GenBank record. Novel prophages with < 50% sequence homology to any annotated phage or prophage sequence were also identified within the P. alcaligenes str. MB-090714, P. koreensis str. 57B-090624, P. protogens str. MB-090624, Pseudomonas sp. 57B-090624, and Pseudomonas sp. MB-090624 genomes (Supplementary Table 7). This hints at a large reservoir of phages within freshwaters that have yet to be discovered. Recent efforts have begun to isolate and characterize freshwater phages, in particular Pseudomonas infecting phages (e.g., Mäntynen et al., 2015; Lu et al., 2017).

Pseudomonas is a cosmopolitan genus, and while not commonly abundant within freshwater environments (Newton et al., 2011), it has been isolated from numerous freshwater lakes including the Great Lakes (Bennett, 1969; Chatterjee et al., 2017). Here we have presented the first whole genome sequences of pseudomonads from the Great Lakes. Given the anthropogenic impact on the Chicago nearshore waters, it is not possible to determine if these isolates are constituents of the microbial community or ephemeral members recently introduced. Nevertheless, pseudomonads’ ability to breakdown and utilize a wide variety of organic compounds serves them well, especially in varied conditions such as Chicago winters when the nearshore waters often freeze. As the study presented here shows, a variety of pseudomonads have inhabited the coastal waters of Lake Michigan.

MB extracted and prepared isolates for sequencing. RS, BH, MA, AA, MKh, PP, TR, SSi, SSa, EX, TH, and CP performed genome assemblies and annotations. MKo and AC collected samples. MB and LM performed phenotype assays. MB, LM, and CP conducted genome sequence analyses and drafted the manuscript. All authors contributed to the editing of the manuscript and approve of the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.