Isolation and Characterization of Lake Erie Bacteria that Degrade the Cyanobacterial Microcystin Toxin MC-LR

Abstract

Microcystin-LR (MC-LR) is a cyclic hepatotoxin produced by cyanobacteria, including Microcystis sp. and Planktothrix sp. Harmful algal blooms (HABs) in Lake Erie have become a major human health concern in recent years, highlighted by the August 2014 city of Toledo, Ohio, municipal water “do not drink” order that affected nearly 500,000 residents for 3 days. Given that microcystin degrading bacteria have been reported from HAB-affected waters around the world, we hypothesized that MC-LR degrading bacteria could be isolated from Lake Erie. To test this hypothesis, 13 water samples were collected from various Lake Erie locations during the summers of 2014 and 2015, MC-LR was continuously added to each water sample for 3 to 5 weeks to enrich for MC-LR-degrading bacteria, and MC-LR was quantitated over time. Whereas MC-LR was relatively stable in sterile-filtered lake water, robust MC-LR degradation (up to 19 ppb/day) was observed in some water samples. Following the MC-LR selection process, 67 individual bacterial isolates were isolated from MC-LR degrading water samples and genotyped to exclude potential human pathogens, and MC-LR degradation by smaller groups of bacterial isolates (e.g., groups of 22 isolates, groups of 11 isolates, etc.) was examined. Of those smaller groups, selected groups of four to five bacterial isolates were found to degrade MC-LR into non-toxic forms and form robust biofilms on siliconized glass tubes. Taken together, these studies support the potential use of isolated bacterial isolates to remove MC-LR from drinking water.

Introduction

Harmful algal blooms (HABs) occur in a variety of aquatic environments and are characterized as large aggregations of naturally occurring photosynthetic bacteria that release neurotoxic and hepatotoxic compounds (cyanotoxins) into water bodies (Harke et al., 2016; Preece et al., 2017). Changes in climate and eutrophication have greatly increased both the incidence and toxicity of HABs over the past few decades, highlighting that human health risks may continue to increase in the future (Bullerjahn et al., 2016; Michalak et al., 2013). Cyanobacteria, including Microcystis and Planktothrix, are well documented to thrive in eutrophic bodies of water and release various compounds, including the hepatotoxic microcystins (MCs) (Davis et al., 2015; Rinta-Kanto and Wilhelm, 2006; Steffen et al., 2017). MCs are cyclic, seven amino acid-containing structures, with the generic structure cyclo(D)-Ala-X-(D)-erythro-β-methyl-iso-Asp-Y-ADDA-(D)-iso-Glu-N-methyldehydro-Ala; X and Y indicate the two variable L-amino acid positions in MCs, while ADDA is (all-S,all-E)-3-Amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid (Rinehart et al., 1988). While over 100 MC variants have been documented to date (Puddick et al., 2014), each with differing levels of toxicity (Chorus and Bartram, 1999), MC-LR is the most toxic, most common, and most closely linked to liver cancer and other diseases in both humans and animals (Carmichael, 1994; Preece et al., 2017; Yu, 1995). MC-LR exerts its harmful effects by binding to type 1 and 2A protein phosphatases (PP1 and PP2A, respectively) in the liver, resulting in excessive phosphorylation. Loss of these functions can result in death in mammals and liver cancer, hemorrhage and hypovolemic shock in humans (Cohen and Cohen, 1989; MacKintosh et al., 1990; Yu, 1995).

HABs have been documented in most parts of the world as highlighted by the World Health Organization (WHO) (http://www.who.int/water_sanitation_health/publications/2011/dwq_guidelines/en/), with notable bloom events in China (Gan et al., 2012), Australia (Bourne et al., 1996), Spain (Moron-Lopez et al., 2017), and Argentina (Valeria et al., 2006). In the United States, HABs have been reported across the country, including large and/or persistent blooms in Florida (Lapointe et al., 2017), California (Kurobe et al., 2018), Ohio (Steffen et al., 2017), Iowa, Maryland, Minnesota, New York, North Carolina, Oregon, Virginia, Wisconsin, Kansas, Montana, Utah and Texas (Backer et al., 2013). For decades, eutrophication of Lake Erie has resulted in large and toxic HABs that have negatively affected recreational activities, led to anoxic/hypoxic conditions that kill fish and other wildlife, and fouled drinking water (Bullerjahn et al., 2016; Davis et al., 2015; Michalak et al., 2013; Steffen et al., 2017). Indeed, during the summer of 2014, a toxic HAB in Lake Erie released high concentrations of MCs into the city of Toledo, Ohio’s municipal water inlet, resulting in MC-LR levels in the finished drinking water exceeding the World Health Organization’s guideline level for safe drinking water (1 μg/L; 1 part per billion [ppb]). As a result of MC levels > 2 ppb in finished drinking water, the city of Toledo issued a 3-day “do not drink” advisory that negatively impacted more than 400,000 residents and hundreds of businesses (Bullerjahn et al., 2016; Steffen et al., 2017). In June 2015, the U.S. Environmental Protection Agency (EPA) issued a 10-day drinking water health advisory guideline for microcystins, which outlined no more than 0.3 ppb/day for children < 6 years old and no more than 1.6 ppb/day for school-aged children (> 6 years old) and adults (https://www.epa.gov/sites/production/files/2017-06/documents/microcystins-report-2015.pdf).

The ability of MCs to persist in bodies of water for an extended period is due to its stable cyclic structure. Previous studies have demonstrated that the half-life of MCs in natural water with direct sunlight is 90-120 days/meter water depth (Welker and Steinberg, 2000). MCs have demonstrated resistance to various physical and chemical processes including sunlight, extreme pH, and high temperatures (Rastogi et al., 2014; Wormer et al., 2010). Due to the stability of MCs, municipal water treatment plants have had to implement a number of treatment options for removal including ozonation, powdered activated carbon, sedimentation, sand filtration, and chemical coagulation to remove MCs from the finished product (Chang et al., 2014; Lahti, 1997). Of these processes, chemical coagulation followed by flocculation removes only small amounts of cyanobacterial toxins (Himberg et al., 1989). By comparison, methods such as chemical coagulation, powdered activated carbon, sand filtration, and sedimentation, particularly when used in combination, have proven to be relatively effective at MC removal (Lahti, 1997). However, MC removal can be reduced by the presence of other naturally occurring organic materials in the water (Lahti, 1997). In addition, studies have shown that ozonation of MCs result in the production of formaldehydes (Chang et al., 2014), that spent-powdered activated carbon must be properly disposed of because of high concentrations of MCs according to the Ohio EPA (epa.ohio.gov/portals/28/documents/HAB/AlgalToxinTreatmentWhitePaper.pdf), and that all of these MC removal processes have doubled or tripled water costs (Steffen et al., 2017). Given these issues, there is a clear need to explore safer, more efficient, and cost-effective methods to remove MCs from drinking water.

The first evidence that naturally occurring bacteria could degrade MCs was published by an Australian group that harvested samples from a Microcystis aeruginosa-dominated river HAB. Following incubation of those samples in the presence of M. aeruginosa extracts (including MC-LR), that group isolated a Sphingomonas sp. that was able to utilize MC-LR as a sole source of nutrients (Bourne et al., 1996; Jones et al., 1994). In follow-up studies, that same group identified the mlrABCD operon, encoding three intracellular hydrolytic enzymes and a putative oligopeptide transport protein, which are responsible for MC-LR degradation in this Sphingomonas sp. (Bourne et al., 2001). MC-LR degradation was found to occur in a stepwise manner, with a microcystinase (encoded by mlrA) initially cleaving cyclic MC-LR (LD₅₀ of 75 µg/kg when injected into mice intraperitoneally) (Harada et al., 2004) into a linearized form (160-times less toxic than cyclic MC-LR), a serine hydrolase (encoded by mlrB) cleaving linear MC-LR into a tetrapeptide (20-times less toxic than cyclic MC-LR), and a metalloprotease (encoded by mlrC) further degrading the tetrapeptide into smaller fragments (Bourne et al., 1996; Bourne et al., 2001) such as ADDA (LD₅₀ 133 times greater than cyclic MC-LR) (Harada et al., 2004). Finally, to evaluate the potential utility of Sphingomonas sp. to biodegrade MC-LR from contaminated water, sand filters inoculated with Sphingomonas sp. were found to remove 90% of MC-LR from river water within 2 days, compared with uninoculated sand filters (Bourne et al., 2006). MC-LR degradation is not unique to Sphingomonas sp. or to Australia, as a recent review highlighted that researchers around the world (e.g., Argentina, Brazil, China, Finland, Japan, Poland, Saudi Arabia, Thailand, United Kingdom) have isolated a diverse array of bacteria capable of degrading various MC variants, including MC-LR, MC-RR, MC-YR, MC-LF, and MC-LW (Li et al., 2017).

Because all of the aforementioned studies were performed great distances from the Great Lakes and because major differences in water chemistries, nutrient loading, and microbial compositions exist among eutrophic water bodies, it is possible that Lake Erie may contain distinct MC degrading bacteria and/or MC degradation enzymes. One previous study attempted to address this possibility by collecting water samples from a 2010 Lake Erie HAB, incubating those samples with MC-LR, quantitating MC-LR levels during a 48-hour incubation, and performing metagenomic sequencing to correlate bacterial taxonomy with potential MC degradation pathways (Mou et al., 2013). While results from that study indicated potential MC degradation by Lake Erie bacteria, a number of unanswered questions remain. First, the short duration (48 hours) of MC incubation likely only allowed for bacterial gene regulation changes and not substantive changes in microbial populations that degrade MC. As such, it is unclear what bacterial strains or populations were correlated with MC reductions. Second, it is unclear if observed reductions in MC concentrations were due to bacterial adsorption or if true enzymatic degradation of MC occurred. Third, although metagenomic sequencing revealed increased abundance of some phyla in MC-containing cultures in that study, only one genera, Methylotenera, increased in abundance in MC-containing cultures and individual MC degrading bacteria were not isolated. As such, it appears that previously described MC degrading bacteria are unlikely to degrade MC in Lake Erie and we do not know what Lake Erie bacteria strains or populations could degrade MC-LR. Finally, putative mlrABCD genes were not significantly detected in any of those Lake Erie samples, leaving major questions about what bacteria and enzymatic pathways may exist in Lake Erie to potentially degrade MC-LR.

The goal of this study was to rigorously enrich for and isolate MC-LR degrading bacteria from Lake Erie HAB events. We hypothesized that, because of increasing numbers and sizes of HAB events, naturally occurring bacteria exist in Lake Erie that degrade MC-LR. Following a rigorous MC-LR selection process, we isolated individual bacterial isolates and confirmed MC-LR degradation by groups of bacterial isolates. These results suggest that MC-LR degrading bacteria could be used to remove MC-LR from drinking water supplies.

Methods

Origin of samples and selection for MC-LR degrading bacterial isolates

Western basin of Lake Erie water samples

All samples from the western basin of Lake Erie were generously collected by Dr. Tom Bridgeman, University of Toledo. Six samples (Fig. 1A) were collected from visible cyanobacterial blooms in the western basin of Lake Erie on July 31, 2015, from the following sampling locations: GR1 (N 41°82.078, W 83°18.56), MB18 (N 41°74.207, W 83°40.16), 7M (N 41°73.340, W 83°29.71), 8M (N 41°78.901, W 83°35.56), 4P (N 41°75.045, W 83°10.35), and CRIB (N 41°70.3, W 83°26.3). GR1, MB18, 7M, 8M, and 4P collection sites were described in a previous publication (Bridgeman et al., 2013). The CRIB collection site is the water intake crib for the city of Toledo, which was engulfed by a large HAB in August 2014, resulting in the Toledo water crisis (Bullerjahn et al., 2016). Water samples were collected approximately 6 to 12 inches below the water surface into autoclaved, amber, wide mouth, 1-L HDPE bottles with screw-top lids (Nalgene). Samples were stored in coolers the day of collection and transported to the laboratory at the end of the day. Water samples were stored at room temperature (approximately 21°C) overnight and processed the following day. To remove large particulates, approximately 800 mL of each water sample was filtered through 1-µm glass microfiber filters (Whatman) into separate, sterile 2-L glass bottles with magnetic stirbars. Negative control water samples, without any viable organisms, were prepared by first pre-filtering each water sample using a 1-µm filter, then filter-sterilizing each water sample using a 0.22-µm filter (Foxx Life Sciences). All water samples, including negative controls, were maintained with continuous stirring (to simulate wave movement) at room temperature (approximately 20°C to 22°C) with 8h:16h light:dark cycles. MC-LR (Cayman Chemical) was suspended in 200 proof (100%) ethanol at 1 mg/mL and stored at −20°C until use. MC-LR, 50 µg/L (ppb), was added to each water sample every 3 to 4 days for approximately 4 weeks, with 1-mL aliquots collected at the beginning of the experiment (day 1) and before and after each MC-LR addition. All sample aliquots were transferred to sterile 5-mL glass tubes, sealed with parafilm, and stored at −20˚C until analyzed for MC-LR concentrations.

Fig. 1. Repeated MC-LR additions select for MC-LR degrading bacteria.

(A) Water samples were collected from visible cyanobacterial bloom events in the western basin of Lake Erie during the summers of 2014 and 2015 at the indicated sites. (B) Differences in bacterial isolate colony morphologies at the beginning [raw water] and end [Week 3] of the experiment. Representative colonies from CRIB sample are shown. (C) Enumeration of bacterial numbers [colony forming units; CFU] during the selection process. Samples were plated and enumerated at the beginning [raw water] of the experiment and once per week for 3 weeks. (D) MC-LR degradation, as measured by ADDA ELISA, in 6 samples from the western basin of Lake Erie. Blue arrows indicate additions of 50 µg/L (ppb) MC-LR.

Sandusky Bay water samples

All Sandusky Bay water samples were generously collected by Drs. George Bullerjahn and Robert Michael McKay, Bowling Green State University. One Sandusky Bay water sample 1163-A (N 41˚46.9, W 82˚71.5) was collected from a visible cyanobacterial bloom on June 8, 2015 (Davis et al., 2015). Other Sandusky Bay samples were collected from visible cyanobacterial bloom events on June 29, 2015 at the following sampling locations: ODNR1 (N 41˚47.73, W 82˚73.97), ODNR2 (N 41˚47.98, W 82˚78.28), BC115 (N 41˚45, W 82˚72.9), ONDR4 (N 41˚45.73, W 82˚89.86), and ODNR6 (N 41˚45.33, W 82˚96.07). All Sandusky Bay water samples were collected, processed, and supplemented with MC-LR as described above for the western basin of Lake Erie samples. The only exception was that MC-LR was added every 3 to 5 days for 15 days. Because of previous reports noting that high nitrogen and phosphorus levels contribute to HABs (Bullerjahn et al., 2016; Davis et al., 2015; Michalak et al., 2013), nitrogen (NH₄Cl and NaNO₃, 5 μM each) and phosphorus (NaH₂PO₄, 1 μM) were added to all Sandusky Bay water samples at the beginning of the study.

Stone Laboratory water sample

One water sample from a visible cyanobacterial bloom was collected near the Ohio State University Stone Laboratory on Gibraltar Island on August 15, 2014. This sample was kindly provided by Dr. Justin Chaffin, Ohio State University, and was stored at −80°C until use. The water sample, named hereafter as ‘Stone Lab,’ was slowly thawed in a 4°C cooler over the course of 48 hours, before being processed and supplemented with MC-LR as described above for the western basin of Lake Erie samples. MC-LR, 100 µg/L (ppb), was added to both negative control (filter-sterilized lake water) and Stone Lab water samples every 3 to 5 days for 33 days.

Quantitation of bacterial numbers in water samples

For western basin of Lake Erie water samples, aliquots were removed at the beginning of the experiment (day 1; before MC-LR addition) and approximately every 7 days during the MC-LR selection process. Aliquots were 10-fold serially diluted in PBS and plated in duplicate onto Difco R2A agar (BD) containing 25 µg/L MC-LR. R2A media was used, as it was developed to study bacteria commonly found in potable water. Plates were incubated at 23°C for 3 to 4 days, colonies were counted and averaged, and CFU/mL were calculated based on the dilution factor.

MC-LR quantitation by ELISA

MC-LR was quantitated using the Abraxis Microcysins and Nodularins ELISA. All samples were processed following the Ohio EPA Total (Extracellular and Intracellular) Microcystins – ADDA by ELISA, Analytical Methodology (version 2.0, January 2015). Briefly, all samples were subjected to three rapid freeze-boil cycles (using a dry-ice/ethanol bath and boiling water bath), filtered using 0.7-μm glass fiber syringe filters (Environmental Express), and stored in glass tubes. Processed samples were diluted in molecular grade water (Corning), based on predicted MC-LR concentrations, so that the final MC-LR concentration in each sample was theoretically within the range of standards supplied in the Microcysins and Nodularins ELISA kit. Samples were analyzed in triplicate or quadruplet. ELISA plates were analyzed using a plate reader (FLUOstar Omega), and MC-LR concentrations were calculated following the manufacturer’s instructions.

Isolating potential MC-LR degrading bacterial isolates

After incubation of each water sample with MC-LR for 21 to 33 days, aliquots were serially diluted in PBS and plated onto R2A media to isolate individual colonies. Selected dilutions were plated based on previous bacterial enumeration studies, which correlated optical density measurements at 600 nm (OD₆₀₀) with bacterial numbers. Plates were incubated as described above and colony phenotypes recorded (e.g., color, size, shape, texture). Individual bacterial isolates were replica plated to ensure clonality and individual bacterial isolates were suspended in BG-11 growth media (Gibco) containing 15% glycerol before archiving at −80°C for future use. A total of 67 individual bacterial isolates were isolated from MC-LR degrading cultures and archived during this study (Table S1).

16S rRNA genotyping of potential MC-LR degrading bacteria

Individual bacterial isolates were plated and incubated as described above; DNA was extracted from each isolate using either TRIzol reagent (Invitrogen), following manufacturer’s instructions, or phenol:chloroform. PCR amplification was performed using GoTaq Green Master Mix (Promega), 125 ng of bacterial isolate DNA, and 16S rRNA primers 8F (5’-AGAGTTTGATCCTGGCTCAG-3’) and 1492R (5’-GGTTACCTTGTTACGACTT-3’) (Turner et al., 1999). PCR amplification conditions were as follows: 95˚C for 2 minutes; followed by 40 cycles of 95°C for 30 seconds, 40°C for 30 seconds, and 72°C for 60 seconds; 72˚C for 7 minutes; and a final hold at 4°C. All PCR reactions were separated on 1% agarose gels and visualized to confirm correct product size (approximately 1400 bp). PCR products were purified using the Qiaquick PCR cleanup kit (Qiagen), following the manufacturer’s instructions. DNA concentration was quantitated using a Nanodrop (Thermo Scientific), and DNA sequencing was performed (Eurofins) using both forward and reverse 16S rRNA primer pairs. All sequencing results were queried using BLASTn analysis to the GenBank databank. In all cases, assignment of genus (and species) for each isolate (Table S1) was based on E values of 0.0 and nucleotide identity scores ≥ 99%.

Testing MC-LR degradation by randomly assorted bacterial isolates

Each of the 67 MC-LR degrading bacterial isolates were randomly assigned to the following groups to assess MC-LR degradation: three groups of 22 to 23 bacterial isolates/group; six groups of 11 bacterial isolates/group; 16 groups with four to five bacterial isolates/group; and 25 individual bacterial isolates (Table S1). Bacterial isolates were individually plated and incubated as described above. Bacteria were suspended in sterile-filtered lake water to approximately the same OD₆₀₀, and equal volumes of each bacterial suspension were inoculated into 50-mL conical tubes containing 30 mL of sterile-filtered lake water from Lake Erie, with foam Identi-Plugs (Jaece) at the top of each tube to allow for gas exchange. Lake water for filter sterilization was collected during non-bloom months (April, May, October, December) and at least 4 days after any local rain event to minimize nutrient loading from tributaries. The lake water was harvested from either Nickel Plate Beach in Huron, Ohio (N 41°39.64, W 82°54.36) or Metzger Marsh in Curtice, Ohio (N 41°64.08, W 83°24.66), frozen at −20°C until use, slowly thawed in a 4°C cooler over the course of 48 hours, filter-sterilized using 0.22-µm filters as described above, and stored at 4°C for experimental use. The starting OD₆₀₀ for each culture was approximately 0.1. Liquid cultures were incubated at 23°C with constant rotation at 215 rpm and were maintained on 12h:12h light:dark cycles using a grow light (BloomBoss). A 0.5-mL aliquot was removed from each culture at the beginning of the experiment (day 1), approximately 45 µg/L (ppb) MC-LR was added to each culture every 3 to 5 days, samples were collected both before and after MC-LR addition, and MC-LR levels were quantitated for 12 days by ELISA. Negative controls consisted of 30 mL of filter-sterilized lake water without bacterial inoculation. All culture aliquots were transferred to sterile 5-mL glass tubes, sealed with parafilm, and stored at −20˚C until analyzed for MC-LR concentrations by ELISA. Experiments were performed twice to confirm reproducibility.

Biofilm assessments

Selected bacterial isolates (see Table S1 for descriptions of 22 isolates-A, 11 isolates-C, 5 isolates-K, 5 isolates-L, 5 isolates-N, BC115 SP, ODNR4 P, 1163B LW, and ODNR1 LW) were plated as described above, suspended in filter-sterilized lake water, the OD₆₀₀ was adjusted to approximately 0.1, and 8 mL of each bacterial suspension was added to sterile 14-mL silica glass tubes (with foam plugs to allow for air exchange). Negative controls consisted of 8 mL of filter-sterilized lake water without bacterial inoculation. Liquid cultures were performed in triplicate and incubated as described above. Once per week, 2 mL from each culture was removed and replaced with 2 mL of fresh filter-sterilized lake water. After 4 weeks of incubation, all liquid was carefully removed from each tube. Tubes were gently washed three times with 3 mL of PBS, 3 mL of 100% methanol was added to each tube, tubes were incubated for 15 minutes, methanol was carefully removed, and tubes were air-dried for 30 minutes. Each tube was stained with 5 mL of staining solution (1% wt/vol crystal violet and 25% vol/vol ethanol) for 15 minutes, the staining solution was removed, tubes were gently rinsed with 5 mL of deionized water, and tubes were air-dried overnight. The following day, 5 mL of 33% glacial acetic acid was added to each tube, incubated for 15 minutes, and biofilm formation was quantitated by OD₅₇₀ measurements. Experiments were performed twice to confirm reproducibility.

Retesting MC-LR degradation by selected groups of bacterial isolates

Based on preliminary MC-LR degradation results, three groups of four to five bacterial isolates (4 isolates-K, 4 isolates-L, and 5 isolates-N), each consisting of four to five isolates each (Table 1), were selected for further analysis. Individual bacterial isolates were plated as described above; one loopful of each bacterial isolate was inoculated into 70 mL of filter-sterilized lake water, with the starting OD₆₀₀ adjusted to 0.1. An aliquot was collected from each flask at the beginning of the experiment (day 1). Flasks were incubated for 16 days, as described above. Approximately 16 µg/L (ppb) MC-LR was added to each flask every 3 to 5 days and aliquots were collected both before and after each MC-LR addition. All culture aliquots were stored at −20°C until analyzed for MC-LR concentrations. Experiments were performed twice to confirm reproducibility.

Table 1.

Groups of four to five bacterial isolates that degrade MC-LR.

4 isolates-K	4 isolates-L	5 isolates-N
Flectobacillus major	Pseudomonas lutea	Agrobacterium albertimagni
Leadbetterella byssophila	Pseudomonas putida	Flectobacillus major
Pseudomonas hunanensis	Runella slithyformis	Porphyrobacter sp.
Pseudomonas parafulva	Sphingobium yanoikuyae	Pseudomonas fluorescens
		Sphingobium yanoikuyae

MC-LR quantitation by Mass Spectrometry

Samples were analyzed using a Thermo Scientific TSQ Quantiva™ triple quadrupole mass spectrometer with UltiMate™ 3000 ultra high performance liquid chromatography (UPLC). The same MC-LR used in all experiments above was used as a control for mass spectrometry analyses. Using electrospray ionization (ESI) positive mode in selective reaction monitoring (SRM) and single ion monitoring (SIM) modes, analytes were separated using a Thermo Hypersil Gold column (2.1 × 50 mm, 1.9 µm particle size) at a flow of 0.35 mL/min using a binary gradient. Mobile phase A consisted of 0.1% formic acid in water and mobile phase B consisted of 0.1% formic acid in acetonitrile. The gradient started with mobile phase B held at 35% from 0 – 1 min and increased from 35 to 60% mobile phase B from 1 to 4.75 min. The column was then washed from 4.76 to 6.5 min at 98% mobile phase B and then allowed to equilibrate back to 35% mobile phase B from 6.51 to 9.5 min. The injection volume was 10 µL for quantitative results in SRM mode and 20 µL for SIM, and the column oven was kept at a constant temperature of 35°C. Source settings were as follows: ion voltage 3.5 kV, ion transfer tube temperature at 260°C, vaporizer temperature at 375°C, sheath gas at 25 arbitrary units, auxiliary gas at 15 arbitrary units, and sweep gas at 1 arbitrary unit.

PCR detection of mlrABC genes in MC-LR degrading bacterial isolates

Individual bacterial isolates were plated onto R2A agar plates and incubated at 23°C for 3 to 4 days. DNA was extracted from each isolate using standard phenol:chloroform extraction procedures. PCR amplification of mlrA, mlrB, or mlrC was performed using GoTaq Green Master Mix, 125 ng of bacterial isolate DNA, and each of the following primer pairs: mlrA F1 (5’-GACCCGATGTTCAAGATGCT-3’) and mlrA R3 (5’-CTCCTCCCACAAATCAGGAC-3’); mlrB F1 (5’-ATCCGCACCTATCTGCCTGAC-3’) and mlrB R2 (5’-GTCGCCATAGCCTTGCCAG-3’); or mlrC F1 (5’-GCTTGATCGTCGAACATTGATGG-3’) and mlrC R2 (5’-CGGCATGGCGAAGGCAC-3’). Sphingomonas sp. ACM3962 (kindly provided by Steve Wilhelm’s lab, University of Tennessee, Knoxville) was used as a positive control, as it has been demonstrated to contain all three mlr genes (Bourne et al., 2001). PCR conditions were as follows: 95˚C for 120 seconds; followed by 30 cycles of 95˚C for 30 seconds, 53˚C for 30 seconds (for mlrA; 61˚C for mlrBC), and 72˚C for 60 seconds; 72˚C for 7 minutes; and a final hold at 4°C. All PCR reactions were separated on 0.8 to 1% agarose gels and visualized to confirm correct product size (800 bp for mlrA, 400 bp for mlrB, 300 bp for mlrC). Agarose gels were imaged using a Mega Ultra Lum gel imager. Any incorrectly sized amplicons (e.g., isolate ODNR1 LW amplified a product using the mlrC primers) were purified using the QIAquick PCR cleanup kit. DNA concentration was quantitated using a Nanodrop and bidirectional DNA sequencing was performed (Eurofins) using the respective primer pair.

Results

Selection of MC-LR degrading bacteria from Lake Erie water samples

To enrich for and isolate MC-LR degrading bacteria from Lake Erie, 13 water samples were collected from visible cyanobacterial bloom events at various locations throughout the western basin of Lake Erie, including Sandusky Bay, during the summers of 2014 and 2015 (Fig. 1A). In the laboratory, samples were pre-filtered to remove large particulates, then MC-LR was added every 3 to 5 days for approximately 3 weeks to select for bacteria that could degrade MC-LR. For samples collected from the western basin of Lake Erie, selection was assessed three different ways. First, colony morphologies were compared at the beginning and end of the experiment based on the hypothesis that the MC-LR selection process would result in phenotypic changes in the bacterial population. While primarily small, white colonies were observed from raw lake water, a diverse array of colony sizes, colors, and shapes was apparent after 3 weeks of incubation with MC-LR (Fig. 1B), indicating a change in the bacterial population during 3 weeks of MC-LR exposure. Second, bacterial numbers were quantitated once/week during MC-LR incubation based on the hypothesis that bacteria which could degrade MC-LR would proliferate/increase in numbers, but bacteria which could not degrade MC-LR during the 3-week selection process would decrease in numbers and/or die. A 1.6- to 2.8-log increase in bacterial numbers was observed in all Lake Erie water samples when comparing the raw lake water sample with samples collected after 3 weeks of MC-LR incubation (Fig. 1C), providing evidence that bacteria could use MC-LR as an energy source. The diverse sampling locations around the western basin of Lake Erie (Fig. 1A) were reflected by the diverse replication rates of bacteria continuously exposed to MC-LR for 3 weeks including: sample 7M demonstrated rapid (1.7-log) bacterial replication between the raw lake water sample and week 1 of MC-LR incubation but no additional bacterial replication during subsequent weeks; sample MB18 demonstrated continuous bacterial replication during 3 weeks of MC-LR incubation, with a 2.8-log total increase in bacterial numbers by week 3 (Fig. 1C). Finally, MC-LR levels were quantitated before and after each MC-LR addition (three total additions). Because all samples were harvested from HAB events, each sample contained 3.2–11.9 ppb MC-LR at the beginning (day 1) of the experiment (Fig. 1D). Despite adding 150 ppb total MC-LR to each water sample over the course of 21 days, the final MC-LR concentration in each sample was between 5.6 to 54.1 ppb, indicating 95.9 to 144.4 ppb total MC-LR degradation or 4.6 to 6.9 ppb/day average daily degradation. Interestingly, samples 7M, MB18, and CRIB continuously degraded MC-LR, while samples 8M and GR1 required an adaptation period (13 days for 8M; 6 days for GR1) before MC-LR degradation was observed. Finally, sample 4P degraded MC-LR after each of the first two MC-LR additions but did not degrade MC-LR following the final MC-LR addition. MC-LR degradation also was observed in Sandusky Bay water samples, with similar variations in degradation rates among collection sites (data not shown).

Based on the encouraging MC-LR degradation data above, we randomly selected the Stone Lab sample (Fig. 1A) for a more thorough and long-term analysis of MC-LR degradation, where approximately 70 ppb MC-LR was added every 3 to 4 days and MC-LR degradation was assessed for 32 days (9 total MC-LR additions). To assess whether hydrolytic enzymes were present in the original sample that could abiotically degrade MC-LR, an aliquot of the Stone Lab water sample was filter-sterilized (0.22 µm) and equal amounts of MC-LR were added to this control sample (negative control) according to the same schedule. For the negative control sample (filter-sterilized lake water), MC-LR levels generally increased during the experiment, with 632 ppb MC-LR quantitated on day 32 (Fig. 2). Increasing MC-LR concentrations in the negative control sample during the experiment indicated that MC-LR was relatively stable during this 32 day experiment and under our culturing conditions, and any changes in MC-LR concentrations in parallel cultures likely would be due to bacterial degradation. In contrast, 16 ppb MC-LR was detected in the Stone Lab water sample on day 32, indicating 616 ppb total MC-LR degradation and 19 ppb/day average degradation (Fig. 2). Similar to the 8M and GR1 samples (Fig. 1D), the Stone Lab sample appeared to undergo a 4-day adaptation period before MC-LR degradation began (Fig. 2). However, from day 5 onward, the Stone Lab sample consistently degraded MC-LR to almost undetectable levels every 3 to 4 days (Fig. 2). Taken together, changes in colony morphologies, increases in bacterial numbers, and dramatic reductions in MC-LR concentrations strongly suggested that 3 to 4 weeks of continuous incubation with MC-LR resulted in the selection of MC-LR degrading bacteria from Lake Erie.

Fig. 2. Stone Lab water sample continuously and robustly degrades MC-LR.

MC-LR degradation, as measured by ADDA ELISA, from a water sample that was collected from a visible cyanobacterial bloom near the Ohio State University Stone Laboratory on Gibraltar Island (see Fig. 1A for location) in August 2014. Blue arrows indicate MC-LR additions of approximately 70 µg/L (ppb). A negative control water sample (Neg Ctrl) was filter-sterilized and incubated identical to the Stone Lab sample. The solid grey line is a linear regression analysis of MC-LR in the Neg Ctrl. The dashed red line indicates changes in MC-LR concentrations immediately after each MC-LR addition and 3 to 4 days later.

Randomized groups of bacterial isolates degrade MC-LR

Because it was impractical to isolate and archive all bacterial isolates from all 13 water samples, we instead selected between three to 13 bacterial isolates from each water sample, primarily based on differences in colony morphology among samples. In total, we isolated and archived 67 bacterial isolates (Table S1). To more closely examine bacterial diversity, all 67 isolates were individually genotyped by 16S rRNA gene sequencing. DNA sequence analysis revealed 42 distinct genus-species, with 13 genus-species being represented more than once (i.e., differences in subspecies or strain could not be determined based on 16S rRNA sequence alone; Table S1). A total of 12 different families (Moraxellaceae, Rhizobiaceae, Caulobacteriaceae, Bacillaceae, Cytophagaceae, Flexibacteriaceae, Flavobacteriaceae, Rhodobacteraceae, Comanonadaceae, Microbacteriaceae, Sphingomonadaceae, Pseudomonadaceae) were represented in our archived isolates, highlighting that a wide variety of potential MC-LR degrading bacterial isolates was selected and isolated (Table S1). For safety considerations in future biodegradation studies, 16S rRNA genotyping also allowed us to identify and exclude any potential human pathogens. To our knowledge, only two of our bacterial isolates are associated with human disease. One, Microbacterium trichothecenolyticum, was noted to be rarely isolated from catheter infections in leukemia patients (Lau et al., 2002). A second, Novosphingobium aromaticivorans, produces metabolites that are associated with primary biliary cholangitis, but the etiological agent of this autoimmune liver disease still is unknown (Tanaka et al., 2018).

Despite encouraging MC-LR degradation from lake water samples containing large numbers of different bacteria (Figs. 1D and 2), our ultimate goal was to identify a single bacterial isolate, or small group of isolates, that degrade MC-LR from water samples. Thus, we next examined whether smaller groups of bacterial isolates were capable of degrading MC-LR. Each of the 67 archived bacterial isolates was randomly assigned to groups of either 22 to 23 isolates/group, 11 isolates/group, four to five isolates/group, or individual isolates (see Table S1 for group assignments and isolates tested). In general, MC-LR degradation by 22 isolates/group (Fig. 3A), 11 isolates/group (Fig. 3B), four to five isolates/group (Fig. 3C), or individual isolates (Fig. 3D) was not as robust as MC-LR degradation in our initial studies (Fig. 1D and 2). For cultures containing 22 isolates/group, MC-LR degradation was detected in group A (22 isolates-A) between days 8 and 12 (10 ppb degradation) and in group B (22 isolates-B) between days 1 and 5 (2 ppb degradation; Fig. 3A). For cultures containing 11 isolates/group, no MC-LR degradation was detected for any group at any time (Fig. 3B). For cultures containing four to five isolates/group, MC-LR degradation was detected in group L (4 isolates-L) between days 8 and 12 (23 ppb degradation) and in group N (5 isolates-N) between days 1 and 5 (6 ppb degradation) and between days 5 and 8 (12 ppb degradation; Fig. 3C). Finally, although MC-LR degradation capability was assessed for 25 individual bacterial isolates (Table S1, Fig. S1, and data not shown), the most promising MC-LR degradation was observed from isolate GR1 GB between days 1 and 5 (4 ppb degradation) and between days 5 and 8 (6 ppb degradation) and isolate SL OY between days 1 and 5 (6 ppb degradation; Fig. 3D). For some bacteria-containing cultures (e.g., 22 isolates-A and 11 isolates-C), MC-LR levels were unexpectedly higher at the end of the experiment than in the negative control samples (Fig. 3), leading us to speculate that some of our groups of bacterial isolates may be degrading MC-LR into breakdown products which result in artificially high MC-LR values when using the commercial ELISA.

Fig. 3. MC-LR degradation by groups of bacterial isolates and individual isolates.

(A) 22 isolates/group, (B) 11 isolates/group, (C) four to five isolates/group, and (D) individual bacterial isolates were incubated without (Neg Ctrl) or with MC-LR for 12 days. Blue arrows indicate MC-LR additions of approximately 45 µg/L (ppb). MC-LR concentrations were quantitated by ADDA ELISA. The solid grey lines are a linear regression analysis of MC-LR in the Neg Ctrl. The dashed blue lines indicate quantitated MC-LR immediately after each MC-LR addition and 3-4 days later.

MC-LR degrading bacteria form biofilms

Given the above results demonstrating that Lake Erie bacterial isolates can degrade MC-LR from lake water, and previous studies demonstrating that microcystin-degrading bacteria could degrade MC-LR in biologically active sand filters (i.e., biofilters; (Bourne et al., 2006), we next tested the ability of our MC-LR degrading bacterial isolates to form biofilms – which would be an essential requirement for biofilter development and testing. Cultures of 22 isolates/group, 11 isolates/group, four to five isolates/group (Table S1), and individual bacterial isolates were grown in filter-sterilized lake water, and biofilm formation was assessed after 28 days. Negative controls consisted of filter-sterilized lake water only. Given the well-known ability of Pseudomonas aeruginosa to form biofilms on various surfaces (Harmsen et al., 2010), P. aeruginosa was grown in filter-sterilized lake water as a positive control. In general, nearly all of the MC-LR degrading bacterial isolates, either individually or in groups, formed biofilms as robustly as P. aeruginosa (Fig. 4). Interestingly, there did not appear to be any correlation between MC-LR degradation (Fig. 3) and biofilm formation (Fig. 4). Similarly, there did not appear to be any correlation between bacterial diversity (i.e., groups of 22 isolates, groups of 11 isolates, individual isolates) and biofilm formation. Robust biofilm formation by all of our MC-LR degrading bacterial isolates, either individually or in groups, strongly supports their use in future biodegradation and biofilter studies.

Fig. 4. Biofilm formation by MC-LR degrading bacteria.

Filter-sterilized lake water was either not inoculated (Neg Ctrl), inoculated with different groups of bacterial isolates, or inoculated with individual bacterial isolates (Table S1).

Mass Spectrometry Analysis of MC-LR Degradation

MC-LR degradation by bacterial isolates was re-tested with the following considerations. First, with respect to the future development of biofilters, the use of individual bacterial isolates, or a small number of bacterial isolates, is preferable to large groups of bacterial isolates. Smaller groups of bacterial isolates have inherent safety and quality control advantages over larger and more complex groups of bacterial isolates. Based on encouraging MC-LR degradation in previous studies (Fig. 3C), we selected three different groups of four to five isolates (4 isolates-K, 4 isolates-L, and 5 isolates-N; Table 1) to degrade MC-LR. Second, we used mass spectrometry (MS) analysis to more accurately quantitate MC-LR levels and specifically detect MC-LR degradation products in each culture. Compared to our previous MC-LR degradation analyses using ELISA (Figs. 1D, 2, and 3), MS analysis is much more accurate and allows for precise quantitation of both intact cyclic MC-LR and MC-LR breakdown products, which was important given some of the unexpected MC-LR levels detected in our previous degradation studies (Fig. 3).

All cultures, including negative control cultures (not inoculated), were grown in filter-sterilized lake water. MS analysis indicated that approximately 16 ppb MC-LR was added to each culture four times during the 16-day experiment, with approximately 64 ppb MC-LR quantitated in the negative control culture on day 16. By comparison, MS analysis of the 4 isolates-K, 4 isolates-L, and 5 isolates-N indicated total (ppb) and daily (ppb/day) MC-LR degradation of 30.8 (1.9/day), 14.3 (0.9/day), and 63.8 ppb (4.0/day), respectively (Fig. 5A). Differences in MC-LR degradation were not due to differences in bacterial numbers, as bacterial enumeration and OD₆₀₀ assessments at the end of the experiment estimated that each culture contained approximately 3.5 × 10⁹ CFU.

Fig. 5. Mass spectrometry analysis of MC-LR degradation.

Filter-sterilized lake water was either not inoculated (Neg Ctrl) or inoculated with groups containing four to five bacterial isolates (Table S1). (A) Blue arrows indicate MC-LR additions of approximately 16 µg/L (ppb). MC-LR concentrations were quantitated by mass spectrometry (MS). The solid grey lines are a linear regression of quantitated microcystin congeners in the Neg Ctrl. The dashed blue lines indicate changes in MC-LR concentrations after each MC-LR addition. (B-C) Secondary ion mass spectrometry [SIM] analysis of 5 isolates-N culture MC-LR breakdown products. (B) SIM chromatogram from Day 1 aliquot, immediately after MC-LR addition. 995 m/z peak at 1.5 minutes indicates native, cyclic MC-LR. (C) SIM chromatogram from Day 16 aliquot, 3 days after MC-LR addition. 615 m/z peak at 1.6 minutes indicates tetrapeptide biodegradation product and 1013 m/z peak at 2.5 minutes indicates linearized MC-LR. 995 m/z was not detected.

Given the dramatic MC-LR degradation by the 5 isolates-N group, we next used secondary ion mass (SIM) spectrometry analysis to detect MC-LR biodegradation products from those samples. A mass range from 130 to 1100 m/z was searched, targeting known MC-LR biodegradation products, including the ADDA fragment (332.22 m/z), the tetrapeptide (615.14 m/z), and the linearized form of MC-LR (1013.56 m/z) (Bourne et al., 1996; Yang et al., 2014). When day 1 samples were analyzed (immediately after MC-LR addition), no degradation products were present (i.e., only the native, cyclic 995 m/z molecule was detected; Fig. 5B). However, at later time points (e.g., day 16), cyclic MC-LR (995 m/z) was absent, but defined peaks for the tetrapeptide (615.14 m/z at 1.6 minutes) and the linearized form of MC-LR (1013.56 m/z at 2.5 minutes) were readily detected (Fig. 5C). These results not only confirm our earlier MC-LR degradation findings (Fig. 1D, 2, 3) but, more importantly, also provide detailed identifications of MC-LR biodegradation products from Lake Erie bacteria.

mlrABCD degradation genes are not present in Lake Erie bacterial isolates

Previous studies in Australia and China found that MC degrading bacteria degrade MC-LR by a series of enzymes encoded by the mlrABCD operon (Bourne et al., 2001). However, given the geographical and limnological differences between the water sources for those MC-LR degrading isolates and Lake Erie, we assessed whether our isolated MC-LR degrading Lake Erie bacterial isolates contained any of the mlr genes. Given that we observed the most robust MC-LR degradation from 4 isolates-K, 4 isolates-L, and 5 isolates-N groups (Fig. 5A) and that the 5 isolates-N group was found to biodegrade MC-LR into linearized MC-LR and tetrapeptide breakdown products (Fig. 5C), we attempted to PCR-amplify mlrA, mlrB, and mlrC from each of the 13 isolates making up the three groups (see Table 1). mlrA, mlrB, and mlrC primers were designed based on conserved regions within all available sequences in the NCBI database. Primers for 16S rRNA served as an internal control for DNA integrity and PCR amplification (Fig. 6). Sphingomonas sp. ACM3962, a MC-LR degrading bacteria isolated from Australia that previously has been shown to contain the mlrABC genes (Bourne et al., 2001), served as a positive control for our mlrA, mlrB, and mlrC amplifications. While a 16S rRNA product was amplified for Sphingomonas sp. ACM3962 and all 13 of our MC-LR degrading isolates, mlrA, mlrB, and mlrC only were amplified from Sphingomonas sp. ACM3962 (Fig. 6). Although an incorrectly sized mlrC product was amplified from isolate ODNR1 LW (genotyped as Pseudomonas putida by 16s rRNA sequencing; Table 1), DNA sequencing of this product did not align to any known mlrC nucleotide sequences. Instead, BLASTn analysis of this amplicon revealed 83% nucleotide identity (E value 4e-84) to a sequence fragment from the Pseudomonas umsongensis strain BS3657 chromosome 1 genome assembly. BLASTx analysis of the same sequence returned hits for a number of Pseudomonas sp. methyl-accepting chemotaxis proteins (E values 4e-46 to 3e-04; nucleotide identities 98% to 38%). Given the incorrect size of this DNA fragment and ambiguous results, we do not believe that this aberrant amplicon is a cryptic mlrC gene. While it remains possible that a highly divergent mlrABCD operon, or mlrABCD genes, exist(s) in our MC-LR degrading bacterial isolates, negative amplification results strongly argue against the existence of the mlrABCD operon in MC-LR degrading bacteria in Lake Erie.

Fig. 6. PCR amplification of known MC-LR degradation genes, mlrABC, from MC-LR degrading bacterial isolates isolated in this study.

DNA was purified from either Sphingomonas sp. ACM3962 (Sphingo; positive control) or 13 bacterial isolates from groups 4 isolates-K, 4 isolates-L, and 5 isolates-N. PCR was performed using primers for either 16S rRNA, mlrA, mlrB, or mlrC. A negative control sample (Neg Ctrl) contained molecular grade water only.

Discussion

Over 30 different MC degrading bacterial strains have been documented in eutrophic water bodies all over the world, representing a wide range of phylogenic classes including α-Proteobacteria, β-Proteobacteria, γ-Proteobacteria, Actinobacteria, and Bacilli (Bourne et al., 1996; Gan et al., 2012; Kormas and Lymperopoulou, 2013; Lemes et al., 2015; Li et al., 2017; Moron-Lopez et al., 2017; Valeria et al., 2006; Zhang et al., 2017). However, surprisingly little is known about MC degrading bacteria in Lake Erie, despite the increasing incidence of Lake Erie HABs over the past few decades, culminating in the August 2014 city of Toledo, Ohio, 3-day drinking water ban due to unsafe MC levels (Bullerjahn et al., 2016; Steffen et al., 2017). In comparison to previously identified MC degrading bacteria, we did not detect any Aeromonas sp., Arthrobacter sp., Bordetella sp., Brevibacterium sp., Burkholderia sp., Delftia acidovorans, Methylobacillus sp., Ochrobactrum sp., Paucibacter sp., Pseudomonas aeruginosa, Paucibacter toxinivorans, Ralstonia solanacearum, Sphingopyxis sp., Sphingosinicella microcystinivorans, Stenotrophomonas sp., or other previously identified genus-species. We did identify 2 Novosphingobium sp. (Table S1), synonymous with Sphingomonas sp., which were associated with some MC-LR degradation (Fig. 3). However, neither of these isolates was observed to degrade MC-LR individually (Fig. S1). One previous study used metagenomic analysis to identify potential MC-LR degrading bacteria in Lake Erie (Mou et al., 2013). While those researchers identified a wide variety of bacterial phyla which increased in abundance in the presence of 15 ppb MC-LR, MC-LR only was added once, and samples only were exposed to MC-LR for 48 hours. By comparison, the present study subjected scum samples to a rigorous enrichment process, whereby scum samples were collected from multiple locations throughout the western basin of Lake Erie, collected during multiple years, repeatedly exposed to high levels of MC-LR (16 to 70 ppb) for > 12 days, and exposed to light-dark cycles to simulate the natural lake environment. Also, MC-LR levels were quantitated from lysed samples to examine free, bound, and internalized MC-LR; MC-LR degrading bacteria were isolated and retested for MC-LR degradation in small groups; and MC-LR biodegradation products were identified by mass spectrometry analysis, confirming that biodegradation was actually occurring.

MC-LR reductions in our experiments could be attributed to many factors, including bacterial adsorption of MC-LR, bacterial internalization of MC-LR, MC-LR complex formation with organic or inorganic materials in lake water, MC-LR degradation by abiotic/secreted enzymes, or bacterial biodegradation of MC-LR. Here, we used multiple approaches to verify that our selection process resulted in the enrichment for and isolation of bona fide MC-LR degrading bacteria from Lake Erie scum samples. First, we demonstrated dramatic phenotypic changes in bacterial colony morphologies between the beginning and end of the experiment (Fig. 1B), suggesting bacterial population changes during 4 weeks of continuous MC-LR additions. Second, we demonstrated that bacterial numbers increased during the 4-week MC-LR selection process (Fig. 1C), indicating that distinct bacterial populations were able to replicate via the degradation of MC-LR. Third, our results demonstrated substantial heterogeneity in MC-LR biodegradation rates (e.g., degradation/day; total degradation/experiment) by Lake Erie bacterial populations (Figs. 1D, 2, 3, and 5A; Table S1), highlighting that additional studies are needed to fully understand the microcystin biodegradation capability of Lake Erie bacteria for practical applications (i.e., biofilters). Fourth, although there were occasional anomalies in MC-LR concentrations in filter-sterilized lake water (i.e., negative control samples), MC-LR levels generally increased to expected levels in all negative controls (Figs. 2, 3, and 5), indicating that any quantitated MC-LR decreases in bacterial-containing samples were most likely due to intracellular bacterial degradation and not to soluble enzymes or substrate interference in lake water. Fifth, any MC-LR adsorption to bacteria or MC-LR internalization without degradation was accounted for by the use of the Ohio EPA total microcystins methodology for all samples, which included three sequential freeze-boil steps to lyse bacteria and release MC-LR (see Methods). Finally, mass spectrometry identification of well-known and substantially less-toxic MC-LR biodegradation products, including linearized MC-LR (1013 m/z) and the tetrapeptide fragment (615 m/z), from our bacterial isolate cultures provides further support that our rigorous selection process resulted in the isolation of microcystin-degrading bacterial isolates.

As noted above, the goal of this study was to identify a single bacterial isolate, or small group of isolates, that degrade MC-LR from water samples. To accomplish this, we started by testing MC-LR degradation with lake water samples containing complex communities of bacteria (and other organisms), isolated potential MC-LR degrading bacterial isolates, and subsequently re-tested MC-LR degradation by less complex groups of bacterial isolates. We observed the most robust microcystin biodegradation from diverse groups of bacterial isolates, including the Stone Lab water sample (Fig. 2) and from cultures containing four to five isolates/group (Fig. 5A). We also tested MC-LR degradation by 25 individual bacterial isolates but the results were not encouraging. Although it remains possible that MC-LR biodegradation may be demonstrated from untested individual isolates in our archived samples, or in ongoing and future studies, the lack of biodegradation by individual isolates thus far indicates that two or more Lake Erie bacterial species may be necessary for MC-LR biodegradation. Based on our findings, studies currently are underway to better understand how groups of four to five bacterial isolates degrade MC-LR or if less complex groups (e.g., three isolates/group, two isolates/group) are capable of degrading MC-LR and other MC variants. Indeed, it has been suggested by other researchers that more than two bacteria may be needed for MC degradation, where one bacterial species initiates MC cleavage, followed by co-metabolism of remaining breakdown products by other bacterial species (Jones and Orr, 1994). Detailed studies also are being performed to examine if certain bacterial isolates outcompete other isolates during MC degradation experiments. While we did perform some qualitative assessments to confirm that four to five distinct isolates were present in selected MC-LR degradation studies (Fig. 5), this analysis was very cursory and more quantitative measures are needed to carefully assess if different bacterial isolates outcompete other isolates during incubations with MC-LR.

There are a number of limitations of our studies. First, bacterial plating for isolation and enumeration was performed on R2A medium. We fully acknowledge that some lake bacteria may not be culturable on R2A medium, may grow slowly on R2A medium, or may not be culturable on any laboratory medium. As such, identification of the complete repertoire of MC degrading bacteria using laboratory media is limited and bacterial enumerations (Fig. 1C) may not truly reflect MC degrading population numbers. We hope to overcome these limitations by using metagenomic sequencing of lake samples, with and without MC, in future studies. Second, it is possible that incubation in the presence of MC-LR for 3 to 5 weeks altered colony morphologies or induced mutations that altered colony phenotypes. Third, the Stone Lab sample (Fig. 2) was frozen at −80°C before being examined for MC degradation. Freezing could have substantially affected the types and numbers of bacteria present in that sample which are capable of degrading microcystin. Fourth, it is difficult to conclusively prove that our MC degrading bacterial isolates use MC-LR as a primary nutrient source. Indeed, there are a number of complicating factors in each culture, including potential differences in rates of bacterial death (and release of nutrients) and presence of natural lake nutrients in each sample that could have contributed to differences in bacterial numbers and/or MC degradation rates.

The results from this study have a number of potential translational outcomes. It is possible that given the increasing frequencies and severities of HABs, naturally occurring MC degrading bacteria also could be steadily increasing in numbers in Lake Erie, and other HAB affected waters, and that large blooms of MC degrading bacteria could natively degrade MCs before they enter drinking water systems. However, high concentrations of MC, and other HAB toxins year after year, throughout the U.S. and world, argue against this possibility. Indeed, while Lake Erie HABs generally are short-lived (e.g., 1 to 2 weeks), the advantage of our experimental approach was that we exposed HAB samples to high levels of MC-LR for 3 to 5 weeks to enrich for and apply strong selective pressure for MC degrading bacteria. Next, our isolated MC degrading bacteria theoretically could be directly added to HABs to degrade MCs but this is unlikely to be practical given the large size of many HABs, including those in Lake Erie. Finally, one of the most important practical outcomes from this project is that microcystin degrading bacteria potentially could be used in biologically active sand filters (i.e., biofilters) to remove and degrade MC-LR from drinking water. Although previous studies have explored the ability of biofilters to degrade MC-LR (Bourne et al., 2006; Ho et al., 2010; Ho et al., 2006; Hoefel et al., 2009), recent studies and/or application of these findings appear to be lacking. If successfully developed, biofilters could be a cost-effective and environmentally friendly supplement to, or replacement for, conventional water treatment processes (e.g., activated charcoal, ozonation, chlorine, etc.). We fully acknowledge that biofilters may possess their own limitations and regulatory difficulties, including: [1] Safety – any future biofilter should not contain pathogenic bacteria. Of the 67 bacterial isolates sequenced in this study (42 distinct genus-species), only two were remotely associated with human disease. Importantly, those two bacterial isolates were not found to degrade MC-LR, either individually or in groups (Table S1 and data not shown) and are unlikely to be included in future studies; [2] Biofilter viability – MC-LR degrading bacteria must be able to form robust biofilms on sand particles to avoid being washed off and should survive in sand filters with no added nutrients. Here, we demonstrated that selected groups of bacterial isolates, such as 5 isolates-N, formed biofilms on siliconized glass tubes (Fig. 4), degraded MC-LR (Fig. 5A), and performed both functions in filter-sterilized lake water, indicating that they may be good candidates for future biofilter studies; [3] Microcystin-degradation rates – any future biofilter should be able to degrade microcystin levels up to, and ideally exceeding, those currently plaguing municipal water utilities. During the 2014 city of Toledo water crisis, 2.5 ppb MC-LR was detected in finished water (Bullerjahn et al., 2016) and publically available data from the Ohio EPA indicates that > 5 ppb MC-LR was detected in multiple raw water samples from Lake Erie during the summer of 2017. Here, we demonstrated that our isolated bacterial isolates could degrade between 0.9 and 19 ppb/day MC-LR (Figs. 1D, 2, 3 and 5). Together with biofilm data, these results indicate that our MC-LR degrading bacteria potentially could be used in future biofilter development studies.

Conclusion

This study found that naturally occurring Lake Erie bacteria could be used to remove and degrade the microcystin toxin MC-LR from water supplies. To accomplish this, we collected Lake Erie water samples from HAB events during different years, continuously added MC-LR to each water sample for 3-4 weeks to select for MC-LR degrading bacteria, quantitated MC-LR before and after each MC-LR addition, and isolated individual MC-LR-degrading bacterial isolates from each culture at the end of each experiment. In addition to demonstrating MC-LR degradation by large groups of unknown/undefined bacteria, this study genotyped individual bacterial isolates to exclude potential human pathogens and re-examined MC-LR degradation by smaller, defined groups of bacterial isolates, proving that groups of four to five isolates could degrade MC-LR. In addition, we used mass spectrometry analysis to identify MC-LR breakdown products, confirming that biodegradation occurs and verifying that non-toxic breakdown products could be generated by Lake Erie bacteria. Finally, all of our isolated MC-LR degrading bacterial isolates, either individually or in groups, were found to form robust biofilms, supporting their potential use in future biodegradation or biofilter studies.