Study Sites and Sampling

Samples were collected from August 2015 to July 2016 from five drinking-water plants with a history of elevated microcystin concentrations in their untreated source waters (Ohio Environmental Protection Agency, 2015a). Samples were collected either from source waters, from water skimmed from the surface of the sand/anthracite filters, or, when access was available, from the solid medium of the sand/anthracite filters. To collect the solid-medium samples, water was drained from the filter well; a sterile scoop was used to take multiple samples across the well from the top 3–4 inches; and samples were combined as a composite. Two plants obtain their water from storage reservoirs supplied by inland streams (sites Inland 1 and 2); one plant sources water from Lake Erie and is on a Lake Erie island (Lake Erie 1); and two plants source water from Lake Erie and are on the Western Basin shoreline (Lake Erie 2 and 3) (table 1). Samples were transported on ice and analyzed within 24 hours of collection in microcosm experiments.

Microcosm Setup and Incubation

Seven batch microcosm experiments were done to isolate and identify microorganisms from source waters or sand filters that were able to biodegrade microcystin-LR (MC–LR), the most common and toxic variant in the environment. Experiments were done in 500-milliliter (mL) flasks containing either source waters or supernatants from sand/anthracite filters along with buffer and nutrient amendments (table 2). For pretreatment of solid-medium samples from sand/anthracite filters, 20 grams of sample was added to 200 mL of sterile, deionized water with a nutrient and salt amendment, modified from Sandman and Ecker (2014) (table 2). The suspension was placed on a rotary shaker for 45 minutes to remove attached microorganisms, then allowed to settle for 1 minute; the resultant supernatant was then filtered through a 5-micrometer (µm) filter to remove large particles, and the filtrate was used in subsequent experiments. Microcystin-LR was added in concentrations ranging from 18 to 80 µg/L. Flasks were prepared with the following treatments: (1) unfiltered source water with MC–LR added, (2) source water or sand/anthracite supernatant filtered through a 5-µm filter with MC–LR added, (3) source water or sand/anthracite supernatant filtered through a 0.2-µm filter with MC–LR added (abiotic control), and (4) source water or sand/anthracite supernatant with no MC–LR added (background control).

Flasks were incubated in a shaking water bath at 28 degrees Celsius (°C) for 14 days. Aliquots (12 mL) were removed at various intervals (0, 1, 2, 4, 8, and 14 days), stored in a −80 °C freezer, and analyzed by enzyme-linked immunosorbent assay (ELISA) in batch for microcystin concentrations. At each time step, a second aliquot (500 microliters [μL]) was mixed with approximately 100 μL of sterile glycerol in a 1.5-mL microcentrifuge tube and frozen at −80 °C for later culture analysis. Based on observed decreases in microcystin concentrations, aliquots were selected to later be cultured to identify and isolate potential microcystin degraders. Pure culture isolates were (1) screened for the ability to metabolize microcystin as the only carbon source and (2) tested for the presence of the mlrA gene. Based on the abilities of isolates to metabolize microcystin as the only carbon source, a select group of the most promising microcystin biodegraders were tested for biofilm formation potential and identified by 16S rRNA sequencing.

Microcystin Concentration Measurement

Microcosm aliquots were analyzed for microcystins by ELISA based on procedures described in Ohio Environmental Protection Agency (2015b). Aliquots were freeze-thawed three times at −70 °C/80 °C to lyse cells and release intact microcystin, filtered through 0.45-µm glass-fiber filters, and added to a microtiter plate kit with ELISA reagents per manufacturer’s instructions (Microcystin-ADDA ELISA kit, Abraxis LLC, Warminster, Pa.). The immunoassay is a direct competitive ELISA in which a toxin in a sample and an antibody conjugate compete for binding sites of antibodies in solution; then, upon the addition of a substrate reagent to the solution, a color signal is generated. Each aliquot was analyzed in duplicate, with each aliquot’s absorbance value read by microplate reader at 450 nanometers, from which the mean absorbance value was calculated. The quality-assurance/quality-control measures include initial demonstration of capability, along with establishment of benchmarks, such as a method detection limit (at 0.3 µg/L), calibration standards and standard curve requirements, a laboratory diluent blank, a positive control, and acceptable coefficients of variation in replicates (USGS Ohio Water Microbiology Laboratory, accessed November 2017 at https://oh.water.usgs.gov/micro_qaqc_methods_for_cyanobacterial_toxins.htm).

Culture Isolations

In microcosms where the concentration of microcystin decreased over time, frozen culture aliquots from 0-time, 1-day, 4-day, and 8-day flasks were thawed, and 100 µL of the culture aliquot was spread-plated on R2A agar plates (BD Difco, Franklin Lakes, N.J., Fisher, catalog no. 218263). After 7 to 14 days incubation at 28 °C, the plates were inspected for phenotypic differences in colony morphology and for an increase in the same colony morphology between the 0-time/1-day plates and the 4-day/8-day plates. When differences in colony abundance or colony types were observed, approximately 10 individual colonies from 4-day/8-day plates were selected for isolation and enrichment as potential biodegraders. Colonies were isolated on lysogeny broth (LB) agar plates (BD Difco, Fisher, catalog no. DF0446-07-5) by three-phase streak, grown in LB broth for 3–5 days at 28 °C, and then frozen in LB broth plus 20-percent glycerol for further testing. Isolates were assigned a name based on the abbreviated name of the site where the bacteria were collected, followed by the alphabetical letter in order of isolation from the R2A agar plates.

Presence of the mlrA Gene

Isolates from six of seven microcosm experiments were screened for the presence of mlrA, a gene associated with microcystin biodegradation (Bourne and others, 2001). Pure culture freezer stocks were grown in tryptic soy broth (TSB) (TSB, Becton-Dickinson, Franklin Lakes, N.J.) at 28 °C until turbid (up to 2 days). After sufficient growth was achieved, 200 µL of each culture was extracted by using a silica-based DNA extraction kit (Qiagen, Hilden, Germany, catalog no. 69506). Quantitative polymerase chain reaction (qPCR) was used to test for the presence of the mlrA gene by using a primer-and-probe set with run conditions listed in Hoefel and others (2009). Each culture extract was run in duplicate, and synthetic plasmid standards (Integrated DNA Technologies, Coralville, Iowa, custom gene synthesis) were included on each qPCR plate. In addition, a matrix spike was performed on a subset of cultures to ensure there was no matrix inhibition. Standard curve information for qPCR analysis is reported in appendix 1, table 1.2.

Screening for Evidence of Microcystin Biodegradation

The isolates cultured from flasks that showed observed decreases in microcystin concentrations were tested for MC–LR biodegradation potential by means of a colorimetric microplate method that measures cellular respiration, adapted from Manage and others (2009). For enrichment cultures, frozen isolates were inoculated into 10 mL of enrichment media. Initial enrichments for biodegradation tests included (1) minimal medium with MC–LR and (2) TSB with 0, 20, 100, or 2,000 µg/L MC–LR; however, both MC–LR alone and increased concentrations of MC–LR with TSB exhibited inhibitory effects on isolate growth (data not shown). As a result, enrichment media used for MC–LR microplate biodegradation tests were (1) TSB prepared per manufacturer’s recommendations and (or) (2) half-strength TSB with 5 µg/L MC–LR. The enrichment medium used was a compromise of ensuring expression of MC–LR degradative genes of interest while still yielding sufficient microbial growth for degradation experiments.

After 48 hours of incubation at 28 °C, cells were centrifuged at 1,000 × g (gravitational force) for 15 minutes, and the pellet was washed with the amended buffer (table 2). After the wash procedure was repeated, cultures were incubated overnight in a shaking water bath at 28 °C with gentle agitation in 5 mL of buffer to remove exogenous sources of carbon. To account for different rates of growth during enrichment, cell density was estimated on a microplate reader (BioTek Instruments, Winooski, Vt., USA, Model EPOCH2) and normalized between cultures (A₅₉₀=0.35). Diluted isolates were then assessed for the ability to replicate when available sources of carbon were limited to MC–LR.

The Biolog MT2 MicroPlate (Biolog, Hayward, Calif.) method detects evidence of cellular respiration through proportional reduction of a tetrazolium dye, which yields a purple color whose absorbance is read by a microplate reader spectrophotometrically at 590 nanometers. This method requires the user to supply a carbon source to allow for cellular respiration; therefore, production of this color change when exogenous sources of carbon are limited to MC–LR allows the user to infer degradation in a high-throughput format. Wells were inoculated with diluted isolates in triplicate and received MC–LR at a final concentration of 20 µg/L, corresponding to the no-contact recreational advisory level set by the Ohio Environmental Protection Agency at the time of the study (Ohio Environmental Protection Agency, 2015a). Paired blanks of the isolates, receiving no MC–LR, were also prepared in triplicate and used to account for background absorbance levels. Plates were incubated at 28 °C and read spectrophotometrically at regular intervals for up to 96 hours post-inoculation. To quantify the biodegradation potential of each isolate, mean absorbance at each time step in the background control was subtracted from mean absorbance in the isolate test wells. Escherichia coli (American Type Culture Collection, 700891) and a strain of Stenotrophomonas rhizophila capable of degrading MC–LR (obtained from Dr. Xiaozhen Mou, Kent State University, Kent, Ohio) were included in each analysis and served as negative and positive controls, respectively.

Ten isolates, identified as promising MC–LR biodegraders through the MT2 MicroPlate screening process, were selected for further identification and testing. Isolates with high peak absorbance values were selected, and if several isolates showed similar absorbance patterns across time-step intervals, one isolate was selected to represent a group with similar patterns.

After incubation in an additional MT2 MicroPlate, with MC–LR as the sole carbon source, the 10 isolates were tested to confirm biodegradation through ELISA analysis. Isolates were analyzed in seven replicate wells—with E. coli and a strain of S. rhizophila serving as negative and positive controls, respectively—and incubated at 28 °C, then read spectrophotometrically at regular intervals for up to 96 hours after inoculation. The microcystin concentration at 0-time, and 8-day was analyzed by ELISA.

Sequencing for Taxonomic Identification of Isolates

The 10 selected isolates and 14 additional isolates (selected randomly to include all sampling sites) were taxonomically identified by 16S rRNA sequencing. Pure culture freezer stocks were grown in TSB at 28 °C until turbid (up to 2 days). After sufficient growth was achieved, 200 µL of each culture was extracted using a silica-based DNA isolation kit (Qiagen, catalog no. 69506). Using universal 16S primers 27F and 1492R (Weisburg and others, 1991), polymerase chain reaction (PCR) testing was run for each extracted isolate. PCR products were run on a fast gel electrophoresis system (Lonza, Basel, Switzerland, catalog no. 57067) to confirm product length. The PCR products were purified by using a rapid PCR purification kit (Qiagen, catalog no. 28104) and then sent to a laboratory for Sanger sequencing (Sanger and others, 1980) (GENEWIZ, South Plainfield, N.J.). The returned sequences were run through the Ribosomal Database Project (Wang and others, 2007) to determine isolate genus (and possible species) identifications.

Biofilm Production Assay

The 10 selected isolates were tested for biofilm formation potential based on procedures described in Djordjevic and others (2002) and Merritt and others (2005). Pure culture freezer stocks were grown overnight in TSB for 24 hours at 28 °C. Pseudomonas aeruginosa 10145 was used as a positive control, and 10-percent TSB was used as a negative control. A diluted culture was prepared (100 µL of culture with 9.9 mL of 10-percent TSB), and 200 mL was added into 8 replicate wells of a 96-well plate. The initial absorbance (optical density of 595 nanometers [OD₅₉₅]) of each well was recorded, and the plate was incubated for 48 hours at 28 °C. The 48-hours OD₅₉₅ of each well was read, media and planktonic cells were removed by gentle pipetting, and the wells were dried for 45 minutes. After the wells were dry, the attached biofilm cells were stained with 250 µL of 1-percent crystal violet for 20 minutes. After excess crystal violet was removed and the wells were washed three times with sterile deionized water, the plate was air dried. After 24 hours, the dye was solubilized with 250 µL of 95-percent ethanol; the contents were transferred to a new microtiter plate; and the OD₅₉₅ of crystal violet remaining in each well was read. The mean absorbance of eight wells was calculated. To quantify potential biofilm production for each isolate, the mean destained absorbance of the isolate was subtracted from the mean destained absorbance of the 10-percent TSB negative control.

Data Management and Statistical Analysis

Raw data reported herein and used for statistical analysis are available in the USGS data release by Francy and Cicale (2024). Spearman’s correlation coefficient was used to assess the relation between the destained absorbance and the 48-hour growth absorbance in biofilm tests. Statistical comparisons among different isolates for biofilm formation were made by using notched boxplots. Notched boxplots are useful for comparing medians among multiple groups (Chambers and others, 1983; McGill and others, 1978). If the notches for two boxes do not overlap, there is strong evidence (approximately 95-percent confidence) that the medians differ.

Microcystin Degradation in Microcosms With Samples From Drinking-Water-Plant Source Waters and Sand/Anthracite Filters

Concentrations of microcystins in samples collected from source waters and the water or solid medium from sand filters ranged from less than 0.3 to 5.5 µg/L (table 1). Results from three microcosm experiments showed at least a 60-percent reduction of microcystin by 14 days (fig. 2A through C); the percentages of microcystin remaining in Inland 1 (fig. 2A) and Lake Erie 2 (fig. 2B) source waters showed similar patterns. For unfiltered source water (blue circles in fig. 2), microcystin degradation was nearly complete at 4 days, with 7.0 percent and 11.3 percent remaining in the inland 1 and Lake Erie 2 source waters, respectively. For the 5-µm-filtered source waters (green squares), results indicated increases in microcystin concentrations at 2 or 4 days—possibly due to inherent method variability of ELISA—followed by microcystin decline and nearly complete degradation at 8 days for Inland 1 (12.3 percent) or 14 days for Lake Erie 2 (14.1 percent). In the abiotic controls, the percent of microcystin remaining after 14 days was 102 percent for Inland 1 and 70.4 percent for Lake Erie 2. Degradation was slower and not as complete in the Lake Erie 1 sand/anthracite filter medium, with 40 percent of microcystin remaining after 14 days. There was no degradation in the background controls, where initial microcystin concentrations were less than 1.0 µg/L.

Isolation and Identification of Potential Microcystin Biodegraders

From the microcosm experiments, 97 colonies were selected, isolated, and frozen (app. 1, table 1.1). Of those isolates, 11 were unable to grow on either LB or TSB and therefore were not further tested. Eighty-six isolates were screened for the presence of the mlrA gene. The mlrA gene was not detected in any of the isolates tested, and no matrix inhibition was found.

Sixty-seven isolates were successfully enriched with or without MC–LR and screened for MC–LR biodegradation potential on MT2 microplates, with MC–LR as the only carbon source. TSB and half-strength TSB with 5-µg/L MC–LR enrichment medium resulted in the highest peak absorbance for 31 and 36 isolates, respectively (app. 1, table 1.1). The time to reach peak absorbance (representing MC–LR degradation) varied by isolate and ranged from 1 to 96 hours. Isolate peak absorbance values ranged from 0.024 absorbance units (AU) at 6 hours for a Lake Erie 2 isolate (LE2.d) to 1.0176 AU at 24 hours for a Lake Erie 1 filter isolate (LE1F.r) (app. 1, table 1.1). Peak absorbance values for the E. coli negative control were 0 AU for E. coli enriched on TSB with 5 µg/L MC–LR and 0.060 AU for E. coli enriched on TSB alone. Absorbance values less than or equal to 0.060 AU for isolates grown on TSB alone, therefore, are unlikely to indicate any biodegradation potential (these values are indicated in app. 1, table 1.1). The average peak absorbance for all 67 isolates was 0.146 AU.

Ten isolates with high peak absorbance values (greater than 0.1 AU) in the MC–LR MT2 microplate screening were selected for biofilm testing and sequencing; these isolates were considered “promising” microcystin biodegraders in that they were capable of respiration when exogenous sources of carbon were limited to MC–LR (table 3). The promising isolates were from both the Inland and Lake Erie sites; seven isolates were from source waters, and three were from the water or solid medium from sand/anthracite filters. Isolates exhibited different rates and patterns of growth, indicating different rates of MC–LR degradation (fig. 3). Of the three isolates from a Lake Erie plant sand/anthracite filter (fig. 3A), two had similar growth patterns and showed peak absorbance values at 96 hours (LE1f.h and LE3F.g), and one showed the highest peak of any isolate at 24 hours (LE1F.r). The four isolates from Inland 1 had different growth patterns (fig. 3B), and absorbance values peaked at 96 hours for three isolates and 48 hours for one isolate (Inland1.b). The isolate Inland1.n had a similar growth pattern to that of the positive control, S. rhizophila. The three remaining isolates were from different sites and showed different peak absorbance values at different time steps (fig. 3C).

The 10 promising isolates and 14 additional isolates (randomly selected to include all sampling sites) were identified by sequencing. The 24 isolates were identified as follows (app. 1, table 1.1): 8 Pseudomonas spp., 4 Aeromonas veronii, 3 Acidovorax spp., 3 Acinetobacter johnsonii, 2 Comamonas spp., 1 Ensifer adhaerens, 1 Chryseobacterium jejuense, 1 Variovorax spp., and 1 Pimelobacter simplex. The 10 selected isolates included isolates identified as Pseudomonas spp., Aeromonas veronii, Acidovorax spp., Acinetobacteria johnsonii, and Chryseobacterium jejuense (table 3).

The 10 promising isolates were screened in a second MT2 microplate spiked with 5 µg/L of MC–LR as the only carbon source for confirmation of degradation. Concentrations of MC–LR were analyzed by ELISA at the beginning and at the end of the 8-day incubation period (fig. 4). Microcystin concentration decreased for the positive control, S. rhizophila, from 4.80 to 2.53 µg/L, and the microcystin concentration of the negative control, E. coli, decreased from 3.51 to 2.13 µg/L (fig. 4A). Initial microcystin concentrations of isolate wells ranged from 3.00 to less than 5.09 µg/L. Final concentrations of isolate wells ranged from 2.13 to 4.62 µg/L (fig. 4A). The positive control and two isolates (LE2.r, Inland1.d) had less than 60 percent of their initial concentrations of MC–LR remaining at the end of the 8-day incubation period (fig. 4B).

Biofilm Formation Potential

The 10 promising isolates were screened for biofilm formation in 8 replicate wells of a 96-well plate. For this procedure, after 48-hour growth (“48-hour growth absorbance”), planktonic cells were removed, and the remaining biofilms were stained, washed, and destained (“destained absorbance”). No significant correlation was found between the destained absorbance and 48-hour growth absorbance (rho=0.18, p-value=0.5567, data not shown); the lack of statistical correlation indicates that biofilm formation was independent of cell growth. Biofilm production was quantified by subtracting the mean destained absorbance of the isolate from the mean destained absorbance of the negative control (table 3). Higher biofilm production is associated with higher destained absorbance values. Five isolates (Inland1.n, LE2.r, Inland2.g, Inland1.d, and LE1F.r) showed potential for biofilm production in that the median destained absorbance in replicate wells was significantly greater than the median destained absorbance of the 10-percent TSB negative control (fig. 5); however, for Inland1.n, the difference was not as large as for the other isolates. Because planktonic cells were removed before staining/destaining, the positive values for the five isolates demonstrate the potential for biofilm production and represent cells that remained attached to the bottom of the plate that are likely biofilm producers.