Two drinking water treatment plants in Ohio, United States, were selected based on annual cyanobacterial bloom intensity. The first was the Collins Park Water Treatment Plant in Toledo, which is a historically high-bloom area in western Lake Erie (Ohio Department Health et al., 2020) and the second was the City of Painesville Water Treatment Plant, a low- or no-bloom area in central Lake Erie (Supplementary Figure S1). Water samples were collected once weekly during the summer and fall (May–November) of 2013 and 2014 from the Toledo and Painesville sites, totaling 58 and 60 samples, respectively. Samples were collected from the source water (sampled at the drinking water intakes within Lake Erie) and from the finished drinking water at each plant. Collections were done in sterile bottles and shipped in a cooler on ice. For toxin measurements, amber glass vials were used. The samples reached The Ohio State University (Columbus, Ohio, United States) for laboratory analysis within 15 h of sample collection. In the lab, the collected source water (200 mL) was filtered using the polycarbonate membrane (0.45 μm pore size, Millipore, Burlington, MA, United States) for molecular analyses, and filters were kept at −20°C and further analyses were performed as soon as possible.

Conventional water quality parameters (water temperature, turbidity, hardness, and pH) and total chlorine in finished water were measured routinely on-site according to the Standard Methods (Rice et al., 2012) by the two water treatment plants and the data were obtained. Chlorophyll-a and phycocyanin were measured using the AquaFluor^® Handheld fluorometer (Turner Designs, California, United States). Total phosphorus (TP) and soluble reactive phosphorus were measured using the USEPA-accepted Method 8190 (Hach PhosVer 3 with Acid Persulfate Digestion). Total nitrogen (TN) in the higher range (0.23–13.50 mg/L NO₃⁻-N) and the lower range (0.01–0.5 mg/L NO₃⁻-N) were analyzed using Method 10206 (Hach dimethylphenol method) and Method 8192, respectively. Total organic carbon (TOC) was analyzed at Alloway Laboratory (Marion, Ohio, United States) using the USEPA 451.3 (U.S. EPA, 2005).

Total MCs in water samples were measured using enzyme-linked immunosorbent assay (ELISA) kits (Abraxis, Warminster, PA, United States) with MC-LR (0.15–5 μg/L) as a working standard. The absorbance was read at 450 nm on the Dynex MRX microplate reader (Dynex Technologies. Inc., Chantilly, VA, United States). The detection range of the assay is 0.15–5.0 ng/mL with a detection limit of 0.10 ng/mL. All the ELISA assays were performed in triplicate.

DBPs (TTHMs and HAA5) in finished water were quantified with gas chromatography–mass spectrometry (GC–MS) using the USEPA-524.2 method.¹ TTHMs included bromoform (CHBr₃), chloroform (CHCl₃), bromodichloromethane (CHCl₂Br), chlorodibromomethane (CHClBr₂). HAA5 included monchloroacetic acid (CH₂ClCOOH, MCAA), dichloroacetic acid (CHCl₂COOH, DCAA), trichloracetic acid (CCl₃COOH, TCAA), monobromoacetic acid (CH₂BrCOOH, MBAA) and dibromoacetic acid (CHBr₂COOH, DBAA). DBPs were quantified at Alloway Laboratory (Marion, Ohio, United States).

We quantified concentrations of total and MC-producing Microcystis by targeting the Microcystis phycocyanin intergenic spacer (PC-IGS) and microcystin synthetase gene B (mcyB), using published approaches with slight modifications (Klase et al., 2019). For Microcystis PC-IGS, the PCR mixture (20 μL) in duplicate contained TaqMan universal PCR master mix (10 μL) (Life Technologies, United States), each primer (188F/254R, 300 nM), PC-IGS probe (100 nM), additional MgCl₂ (1.25 mM) and DNA template (2 μL). The PCR profile followed an initial cycle of 50°C for 2 min, 95°C for 10 min, 50 cycles at 95°C for 30 s, 56°C for 1 min, and 72°C for 30 s. For Microcystis mcyB, the PCR protocol was identical to that for PC-IGS, except we used 900 nM of each mcyB-specific-primer (30F/108R) and 250 nm of mcyB probe. The standard working curve was generated for each assay according to the standard Microcystis PC-IGS and mcyB DNA standard (pGEMT-PC-IGS and mcyB), respectively. The PC-IGS probe had a 5′ end of the fluorescent reporter, 6FAM, and a 3′ end of the non-fluorescent quencher (NFQ) that was attached with a minor groove binder (MGB) moiety (MGB-NFQ). In contrast, the mcyB probe contained a 5′ end of the fluorescent reporter, instead. We estimated the proportion of the potentially MC-producing genotype in the Microcystis population according to the ratio between Microcystis mcyB and PC-IGS in percentage unit (%).

We targeted three antibiotic resistance genes (tetQ for tetracycline resistance, sul1 for sulfonamide resistance, and bla_KPC for carbapenem resistance), total bacteria (targeting 16S rRNA gene), Legionella species (targeting the 5S rRNA gene), as well as a mobile genetic element [MGE, class 1 integron-integrase gene (intI1)] (Krøjgaard et al., 2011; Jia et al., 2014; Yin et al., 2017; Gupta et al., 2019; Klase et al., 2019) using the QX200 droplet digital PCR system (ddPCR, Bio-Rad, Hercules, CA, United States). For quantification of total bacteria, tetQ, sul1, and intI1, the ddPCR mixture (20 μL) contained 2X EvaGreen supermix (Bio-Rad), 200 nM of each primer, DNA template, and RNase-/DNase frees PCR water. To quantify KPC and Legionella, the ddPCR mixture (20 μL) contained 2X supermix for probes, 200 nM of each primer, 200 nM of the probe, DNA template, and RNase-/DNase free PCR water. With 20 μL of PCR mixture, droplets were generated using the Droplet generator (Bio-Rad) with droplet generation oil. PCR was performed using a thermal cycler (C1000 touch thermal cycler, Bio-Rad, Hercules, CA, United States), following previous studies (Yin et al., 2017; Gupta et al., 2019; Klase et al., 2019). After the PCR reaction, concentrations of targeted genes were analyzed using the Droplet Reader and QuantaSoft software (Bio-Rad Hercules, CA, United States).

Data were explored using scatterplots, time-series plots, box plots and bar charts. Means and standard deviations (SD) were calculated and data are presented as the mean ± SD. A one-way analysis of variance (ANOVA) was used for the spatial difference in the variables with a significance level set at p < 0.05. After ANOVA, Tukey Honest Significant Difference tests were used to determine if the means were different (p < 0.05) between the groups. All analyses were conducted using SPSS 24.0 (SPSS, Chicago, IL, United States).

The severity of blooms in the source water from western (Toledo) and central (Painesville) Lake Erie was examined (Figures 1–3) as were source water chemical parameters and water temperatures. The mean concentrations of chlorophyll-a and phycocyanin in the source water from western Lake Erie (bloom site) were 8.32 μg/L and 37.4 μg/L, respectively (Figure 1). The mean concentrations of chlorophyll-a and phycocyanin in the source water from central Lake Erie (control site) were 2.92 μg/L and 14.8 μg/L, respectively (Figure 1). The mean concentrations of two bloom indicators in the bloom site were significantly higher than that from the control site (p < 0.05). Similarly, significantly higher concentrations of total Microcystis (PC-IGS gene) and toxin-producing Microcystis (mcyB gene) were observed in the bloom site source water (p < 0.05) (Figure 2). Specifically, the mean concentration of the PC-IGS gene (gene copies/mL) was 5.57 × 10⁴ in Toledo source water (bloom site) vs. 8.79 × 10¹ in the Painesville source water (control site). The mean concentration of the mcyB gene (gene copies/mL) was 1.48 × 10⁴ in the bloom site vs. 1.32 × 10¹ in the control site. In regard to the cyanotoxin measurement, the mean concentration of microcystin in the bloom site source water was 1.65 μg/L which was also higher than the control site (not detected) (Figure 3). These results show that the intensity and frequency of blooms at the bloom site (Toledo) was obviously higher than the control site (Painesville).

Among water chemistry parameters, concentrations were consistently higher at the bloom sites; whereby mean total phosphorus levels for the two-year study period were 140 μg/L and 90 μg/L for the Toledo and Painesville DWTP source water samples. The mean total organic carbon was 0.36 mg/L (Toledo) and 0.24 mg/L (Painesville) and the mean total nitrate was 0.64 mg/L and 0.33 mg/L in the Toledo and Painesville source water samples, respectively. In addition to chemistry parameters, the mean temperatures of source water samples were 19.6°C and 16.1°C for Toledo and Painesville, respectively.

To evaluate the potential association between cyanobacteria and Legionella, the concentrations of Legionella were examined in the source water from the western Lake Erie and the control site for 2 years. The concentrations of Legionella (gene copies/mL) were 2.49 × 10² (ranging from 5.10 × 10⁰ to 5.09 × 10³) in the bloom source water and 1.63 × 10² (ranging from 4.66 × 10¹ to 3.85 × 10²) in the control source water. The associations between cyanobacteria and Legionella in the source water are summarized in Figure 4. An apparent significant relationship was observed in western Lake Erie between the concentration of the molecular markers for total cyanobacteria and Legionella spp. (F = 89.82, p = 0.001), but there was no relationship observed between the markers for total cyanobacteria and Legionella spp. observed in source water samples from the non-bloom site in Painesville (F = 0.31, p = 0.5814).

The concentrations of DBPs (TTHMs and HAA5) in the finished water from the bloom vs. control sites are summarized in Figure 5. The mean concentrations of TTHMs were 22.43 μg/L from the bloom site and 14.03 μg/L from the control site, and the mean TTHMs concentration from the bloom site was significantly higher than the control site (p < 0.05). The mean concentrations of HAA5 were 8.86 μg/L from the bloom site and 9.06 μg/L from the control site, and the mean HAA5 concentrations between the two locations were not significantly different (p > 0.05). Figure 6 shows that TOC levels in finished water were not statistically different between the bloom vs. control sites (p > 0.05). The mean concentrations of TOC were 1.42 mg/L from the bloom site and 1.72 mg/L from the control site. Furthermore, microcystin was not detected in finished water samples from either location.

Concentrations of ARGs [tetracycline (tetQ), sulfonamide (sul1), carbapenem (bla_KPC) resistance genes] were quantified in source and finished waters. The concentrations of all the ARGs were significantly lower (p < 0.05) in the finished water vs. the source water at bloom and control locations (Figure 7). In comparing the mean concentrations (gene copies/100 mL) of ARGs in the finished water, the Toledo finished water mean ARG concentrations were 3.70 × 10² (tetQ), 3.26 × 10² (sul1), and 1.91 × 10² (bla_KPC). The mean concentrations (gene copies/100 mL) of ARGs in the finished water from Painesville WTP were 2.43 × 10² (tetQ), 4.31 × 10² (sul1), and 4.69 × 10² (bla_KPC).

Concentrations of MGE (class 1 integron-integrase gene [intI1]) and the total bacteria gene (16S rRNA) were also significantly reduced by drinking water treatment. After water treatment processing, the mean concentrations of the MGE (gene copies/100 mL of intI1) in the finished water were 4.19 × 10⁰ in Toledo and 0.82 × 10⁰ in Painesville which represented significant (p < 0.01) reductions from their source water (Figure 8). Similarly, concentrations of the total bacterial gene in the finished water of the bloom and control sites (4.56 × 10³ in Toledo, 3.55 × 10⁴ in Painesville) were significantly lower (p < 0.01) than in their source water (6.48 × 10⁸ in Toledo, 9.00 × 10⁹ in Painesville). While drinking water treatment resulted in significantly lower concentrations of total bacteria, ARGs, and MGE, the relative abundance of ARGs among total bacteria (16S rRNA) increased significantly (p < 0.05) at both the bloom and control sites (Figure 9).

Two drinking water treatment plants, drawing from Lake Erie as their source water, were chosen in Toledo (a high-bloom region in western Lake Erie) and Painesville (a low- or no-bloom area in central Lake Erie) to examine water quality challenges beyond cyanotoxins when utilizing bloom-impacted source water. In assessing blooms, with an emphasis on cyanobacteria, multiple water quality parameters were employed to evaluate source and finished water quality. Upon characterizing bloom conditions, as expected, there were significant differences in chlorophyll-a, phycocyanin, total Microcystis, MC-producing Microcystis, and MC between the two locations. These differences were supported by the underlying water chemistry and temperatures enabling greater primary productivity by cyanobacteria as warmer water temperatures and higher concentrations of nutrients (P and N) are linked to cyanobacteria bloom formation, particularly when water temperatures approach or exceeds 20°C (Paerl and Huisman, 2008; Stumpf et al., 2012).

As predicted, bloom conditions manifested in source water during the warm months, mainly from July, August and September at the Toledo water intake (bloom site). Figures 1–4 show corroboration between bloom indicators (chlorophyll-a and phycocyanin), Microcystis densities, and ultimately cyanotoxins. Three of the source water MC results from Toledo exceeded the provisional 2015 Ohio contact recreation advisory threshold (6 μg/L) and one sample exceeded the revised and current (post-2019) Ohio and U.S. EPA contact recreation advisory threshold (8 μg/L standard) (Kasich et al., 2015; U.S. EPA, 2019). Notably, although MC levels were elevated in Toledo’s source water, they remained undetectable in finished water samples. Of note, the Microcystis concentrations indicated by PC-IGS genetic marker measurement in the source water in Toledo were similar to those in a Japanese Lake Mikata where Microcystis blooms occurred (Yoshida et al., 2007).

Once cyanobacterial blooms occur, blooms can impair surface water supplies used for drinking water. Blooms are associated with increases in TOC, turbidity, taste and odor compounds, and precursors for DBP formation (Nguyen et al., 2005). As hypothesized, TOC and turbidity levels were elevated in the bloom site source water relative to the control site (Painesville). Algal-derived organic carbon can be a significant source of DBP precursors for drinking water treatment facilities (Nguyen et al., 2005).

While DBP precursors can be elevated in source water, the presence of DBPs in finished drinking water varies according to the source water characteristics (e.g., temperature, pH, natural organic matter, etc.) and the processes used in water treatment. TOC concentrations are often used to predict DBP formation because TOC is a precursor to DBP formation. In this study, we investigated not only temporal variation in TOC (DBP formation precursor), but also the concentration of two commonly regulated DBPs, TTHMs and HAA5, in drinking water. TTHMs and HAA5, considered potentially carcinogenic, are the most important groups of DBPs (Rodriguez et al., 2004). The primary drinking water regulations for the U.S. mandates that the maximum acceptable levels of DBPs are 80 μg/L for TTHMs and 60 μg/L for HAA5 (U.S. EPA, 2023). Our results show that TOC concentrations in Toledo source water were higher than Painesville source water, and it is likely that the increase in TOC is related to cyanobacteria densities (Cory et al., 2016). As TOC levels were higher in the source water at Toledo, so were TTHMs and HAA5 in the finished water relative to the Painesville finished water. While DBPs were quantifiable, none of the samples exceeded primary standards set by the U.S. EPA, indicating that the current TOC removal processes at the two DWTPs can effectively control common DBPs.

Beyond altering source and finished water quality, cyanobacteria blooms in western Lake Erie occur in the context of a complex microbial ecosystem and can alter microbial communities.

One type of change was the observed apparent positive relationship between Legionella spp. abundance and Microcystis in the source water for the Toledo DWTP (bloom site), which was present in both study years, but non-existent in the source water of the Painesville DWTP that served as a control site. Recent research from western Lake Erie (near Toledo [bloom site]) focusing on bacterial diversity from non-cyanobacteria showed clear differential responses among non-cyanobacteria to cyanobacteria abundance (estimated by chlorophyll-a) (Berry et al., 2017). Among bacteria that are linked to an increase in abundance during bloom conditions, there is evidence that the density of Legionella is correlated with Microcystis and eutrophication (Numberger et al., 2022). When looking at similar warm season water samples from three inland Ohio lakes, which like Western Lake Erie exhibit eutrophic conditions, Lee et al. (2016) documented Legionella spp. and the Legionellacea as being among the most abundant and ubiquitous bacteria present in Ohio inland lakes.

While Legionella species are common in various natural and human-made aquatic environments, some species, mostly L. pneumophila, may cause legionellosis, which is a serious pulmonary infection established in persons following the inhalation of particles of contaminated aerosolized water. While there is some evidence that certain cyanobacteria can stimulate the growth of Legionella (Świątecki and Zdanowski, 2007; Bergeron et al., 2015), there are few studies primarily focusing on Legionella and cyanobacteria densities. Tison et al. (1980) suggested an intimate association between Legionella spp. and cyanobacteria (Fisherella). Their observation indicated that Legionella could use algal extracellular products as its carbon and energy sources (Tison et al., 1980; Berendt, 1981). A previous study also reported that symbiotic interactions between Legionella and cyanobacteria may help the colonization of aquatic environments (Carvalho et al., 2007). Several factors (e.g., temperature, pH, and concentrations of nutrients and ions) and possibly products produced by other non-cyanobacterial bacteria, such as earlier arriving Betaproteobacteriales (van der Kooij et al., 2018), which are also abundant in Lake Erie during blooms (Berry et al., 2017), may all contribute to Legionella growth in aquatic environments as part of an ecological succession linked to bloom conditions of certain cyanobacteria species, like Microcystis.

While the potential for an ongoing increased global incidence of legionellosis due to a warmer climate has been described (Walker, 2018), the role of environmental waters warrants additional study. In Ohio, the incidence rate of legionellosis has consistently been 1.9 to 2.6 times greater than the U.S. average in recent years (Han, 2021). Recently, aerosols associated with roadway exposures have been considered as part of the increased incidence (Han, 2021), but of particular interest here is how anticipated future increases in eutrophic waters experiencing further eutrophication and warming globally may contribute to increased aerosol-related Legionella exposures, which could occur from boat wake (water skiing), fountains, etc. Beyond warmer water, there may be possible synergisms occurring in warming eutrophic waters as high concentrations of phosphates can enable substantial growth of cyanobacteria and increased turbidity and/or aggregated biofilm materials which can prompt or enhance the growth of Legionella and other Gammaproteobacteria (Cai et al., 2013; Lee et al., 2016; Srivastava et al., 2016). While the Gammaproteobacteria include many biofilm formers that may benefit from increased turbidity, these bacteria are also associated with an abundance of antibiotic resistance genes (Zhang et al., 2021).

Adding to the complexity of bloom ecology beyond Legionella densities are other broader emergent concerns regarding the establishment, maintenance, or promotion of antibiotic resistance bacteria (ARB) and ARGs in the source water (Volk and Lee, 2022), finished water (Li et al., 2016), and the distribution system (Xi et al., 2009; Xu et al., 2016). When the source water is impacted by bloom conditions, a convergence of numerous selective pressures emerge in the microbial community, including interspecies competition in the source water coupled with impacts from the water treatment processes, such as activated carbon use and disinfection. Xu et al. (2020b) demonstrated that when MCs were present an increase in ARGs was observed in the DWTP.

In this study, while hypothesized there would be differences in the densities of MGEs and ARGs in the source water samples from bloom and control sites, no differences were observed for MGEs or the ARGs for tetracycline, sulfonamide, and carbapenem resistance (Figure 7). At both study locations, the DWTP processes significantly reduced the amount of MGEs and ARGs from source water to finished water. The differences that emerged were specific to relative abundances of ARGs and MGEs. At both DWTPs, while reduced in overall density, the total bacterial density was reduced even more greatly by the treatment processes. In comparison of the two DWTPs, the MGEs increased in relative abundance more in the Toledo DWTP than the Painesville DWTP.

Previous studies have revealed that commonly used disinfection technology can enrich ARB and spread ARGs (Shi et al., 2013). Water processing, including filtration and chlorination, remove most bacteria; however, extracellular stress can promote the replication of plasmids in bacteria. For example, chlorination might increase the copy number of plasmids in the surviving bacteria cells, resulting in the higher relative abundance of ARGs in treated water (Shi et al., 2013). While densities of ARGs and MGE were higher in the source water than the finished water, a limitation of this study was that it did not study the water distribution system, which can have higher levels of bacteria than the finished water due to regrowth potential of bacteria in the distribution system (Xi et al., 2009). If the source water selected for greater survival of biofilm formers preferring higher phosphate levels, corrosion control measures for the distribution system using phosphate may play a role in the regrowth of antibiotic resistant bacteria (Kappell et al., 2019; Kimbell et al., 2020). As expected, the finished water which included corrosion control had higher total P levels than the source water at the bloom site and control site (Supplementary Table S1).

The main scope of the present study was source water and finished water in Lake Erie region, thus tap samples or samples from within the distribution system were not included in this study, but we suggest that future study includes examining the distribution systems of bloom-impacted community water systems since it can contribute great knowledge regarding the public health implications of biofilm formation, ARGs, MGEs, and potential bacterial regrowth. This study adds to the body of evidence that some efforts should be made to monitor ARG concentrations before, during, and after drinking water treatments. As part of monitoring for ARGs and Legionella, methods reliant on culturable Legionella or other bacteria (e.g., Pseudomonas spp., Mycobacterium spp.) may provide benefits, but may underreport densities of some bacteria resistant to disinfection when hosted inside free-living Amoeba, which warrants a need to use a PCR-based approach (Calvo et al., 2013).

Complicating future studies on ARGs and MGEs in finished water and distribution systems are not only disinfectants but also disinfection byproducts. While this study illustrated that chlorine-related DBPs (THMs and HAA5) were elevated during bloom conditions, yet meeting regulatory thresholds, we suggest to characterize other DBPs that are likely elevated in cyanobacteria-impacted waters. Since there is considerable evidence that THMs and HAAs are linked to ARGs in the finished water and distribution system (Lv et al., 2014; Li and Gu, 2019; Zhang et al., 2023), we recommend to assess unregulated N-DBPs which likely are less abundant than THMs and HAAs, but are more common in bloom-impacted waters, and able to elicit greater cytotoxicity (Fang et al., 2010; Liu et al., 2020) in a future study. In addition, emerging literature indicates that future studies on DBPs ought to consider a broader range of DBPs for also assessing potential human health risks (Li and Mitch, 2018; Kali et al., 2021). Thus, in bloom-impacted waters there would likely be benefits from assessing N-DBPs which are associated with the chlorination of Microcystis aeruginosa (Fang et al., 2010). These same N-DBPs may also be of interest for improving understanding of antibiotic resistance phenomena in treated waters.