Combined Danio rerio embryo morbidity, mortality and photomotor response assay: a tool for developmental risk assessment from chronic cyanoHAB exposure

Abstract

Freshwater harmful algal blooms produce a broad array of bioactive compounds, with variable polarity. Acute exposure to cyanotoxins can impact the liver, nervous system, gastrointestinal tract, skin, and immune function. Increasing evidence suggests chronic effects from low-level exposures of cyanotoxins and other associated bioactive metabolites of cyanobacterial origin. These sundry compounds persist in drinking and recreational waters and challenge resource managers in detection and removal. A systematic approach to assess the developmental toxicity of cyanobacterial metabolite standards was employed utilizing a robust and high throughput developmental Danio rerio embryo platform that incorporated a neurobehavioral endpoint, photomotor response. Subsequently, we applied the platform to cyanobacterial bloom surface water samples taken from temperate recreational beaches and tropical lake subsistence drinking water sources as a model approach. Dechorionated Danio rerio embryos were statically immersed beginning at four to six hours post fertilization at environmentally relevant concentrations, and then assessed at 24 hours and 5 days for morbidity, morphological changes, and photomotor response. At least one assessed endpoint deviated significantly for exposed embryos for 22 out of 25 metabolites examined. Notably, the alkaloid lyngbyatoxin–a resulted in profound, dose-dependent morbidity and mortality beginning at 5 μg/L. In addition, hydrophobic components of extracts from beach monitoring resulted in potent morbidity and mortality despite only trace cyanotoxins detected. The hydrophilic extracts with several order of magnitude higher concentrations of microcystins resulted in no morbidity or mortality. Developmental photomotor response was consistently altered in environmental bloom samples, independent of the presence or concentration of toxins detected in extracts. While limited with respect to more polar compounds, this novel screening approach complements specific fingerprinting of acutely toxic metabolites with robust assessment of developmental toxicity, critical for chronic exposure scenarios.

Graphical abstract

1. Introduction

Freshwater cyanobacterial harmful algal blooms (cyanoHABs) represent a complex milieu of cyanobacterial cells, associated biomass and secondary metabolites (Chlipala et al., 2011; Dittmann et al., 2013; Dittmann et al., 2015; Merel et al., 2013). Among the oldest living organisms, cyanobacteria are photosynthetic, typically slow-growing, organisms that occupy diverse ecological niches (marine, terrestrial and freshwater) (Dittmann et al., 2015; Moreira et al., 2013). Cyanobacteria produce a wide range of natural products with intricate structures and bioactive properties (Chlipala et al., 2011; Dittmann et al., 2013; Dittmann et al., 2015; Merel et al., 2013; Moreira et al., 2013). They generate cyanotoxins under poorly understood environmental cues (Buratti et al., 2017; Ibelings et al., 2015; Merel et al., 2013). Anthropogenic influences on their ecological niches abound and are exacerbated by human population growth. These influences include excessive nutrient loading into surface and coastal watersheds, changing trends in global temperatures, and alterations in water cycles and availability. These influences contribute to environmental conditions conducive for proliferation of microscopic cyanobacteria. Visually stunning blooms appear as biomass on the water surface or form mats on bottom sediment (Merel et al., 2013; Paerl and Otten, 2013; Perovich et al., 2008; Visser et al., 2016). In turn, these blooms impact water quality and produce acutely toxic metabolites of immediate and direct concern for communities that rely on surface waters for recreation or drinking water (Farrer et al., 2015; Ibelings et al., 2015; Pirez et al., 2013; Zhang et al., 2009).

1.1. Classes of cyanotoxins and environmental risk.

Cyanotoxins vary in structure and, thus, toxicological effects. Microcystins (MCs) constitute a family of over 240 (Spoof and Catherine, 2017) identified cyclic hepatotoxic heptapeptides. Saxitoxins and anatoxin-a are potent alkaloid neurotoxins. These three families are produced by a wide variety of cyanobacteria and are frequently screened for in freshwater cyanoHABs by watershed managers (Buratti et al., 2017; Ibelings et al., 2015; Merel et al., 2013). The hepatotoxic nodularins (cyclic peptides) and cylindrospermopsins (CYNs) (alkaloids) also may be screened for in blooms where known cyanobacterial producers proliferate, while less focus has been placed on potent dermatotoxins and gastrointestinal toxins, such as lyngbyatoxin-a (alkaloid) or aplysiatoxin (Manganelli et al., 2012; Merel et al., 2013; Metcalf and Codd, 2009). Lipopolysaccharides (potential irritants to any exposed tissues) and the amino acid β-methyl-amino-L-alanine (BMAA) with potential links to neurodegenerative diseases (such as ALS, Parkinson’s, and Alzheimer’s), may be produced across most genera of cyanobacteria (Metcalf and Codd, 2009; Dittmann et al., 2013, Merel, et al., 2013).

Diverse genera of cyanobacteria produce additional peptides, polyketides, alkaloids, lipids, and terpenes, often identified through biopharmaceutical endeavors to exploit natural products for their bioactive properties (Dittmann, et al. 2015; Moreira, et al. 2013). These bioactive properties pose unknown health consequences for humans chronically ingesting them in potable water. Table S1 summarizes acutely toxic cyanobacterial metabolites produced in cyanoHABs. Table S2 outlines additional classes of linear and cyclic peptides, synthesis pathways, and inhibitory targets. These targets are key in cellular processes and confound chronic health risk presented by blooms. In vivo and in vitro, MC exposures results in carcinogenic, neurological, immunological, reproductive, and developmental effects (Funari and Testai 2008, Martinez Hernandez et al., 2009; Li et al., 2011; Buratti, Manganelli et al., 2017; He et al., 2017). Two classes of peptides, aeruginosamides and cyanopeptolins, previously presumed nontoxic, now have documented cytoxicity and neurotoxicity, respectively (Gademann et al., 2010; Faltermann et al., 2014). Pharmacological targets and effects of other categories of compounds are largely unexplored (Welker and von Dohren, 2006).

1.2. Guidelines and challenges in monitoring human health risks from cyanobacterial blooms.

Given the complex milieu of natural bioactive compounds that comprise blooms and the nonpoint source pollution that contributes to them, regulatory agencies have avoided traditional routes of regulation and compliance as occurs with manmade compounds. Instead, they have opted to provide and modify provisional guidelines and action levels for potable and recreational surface waters. The World Health Organization (WHO) established a provisional guideline of 1 μg/L MC-LR for drinking water sources based on available toxicological data for MC-LR (WHO 2003a, 2017); the WHO also provides guidance for likelihood of adverse health effect for recreational water based on cyanobacterial cell counts, chlorophyll-a and MC-LR (WHO 2003b). Numerous countries have adopted WHO guidelines, while some have secured more stringent and specific guidance. As of 2015, with growing concern and awareness, the U.S. EPA set an action level limits for 10-day health advisories for drinking water for MCs and CYNs, with recommended levels at 0.3 μg/L and 0.7 μg/L, respectively, for pre-school age children (<6 years) and infants (U.S. EPA Health Advisory 2015). It also provides recreational use guidelines of 8 μg/L and 15 μg/L (Table S3). Several U.S. states have adopted lower thresholds because of heightened concern around potential developmental; a recent report submitted to the International Joint Commission by the Health Professionals Advisory Board (Feb 27, 2017), invoked an even more stringent guideline of 0.01 to 0.03 μg/L (Miller et al., 2017). Additionally, as of 2017, twenty-one U.S. states enacted their own recreational water action level based on either cell counts or specific toxicant measurements. (https://19january2017snapshot.epa.gov/nutrient-policy-data/guidelines-and-recommendations_.html)

In part, enforced regulatory limits have not been implemented as public water authorities vary considerably in resources dedicated to monitoring surface source waters for drinking water intake, as well as recreational waters. Communities and public water officials are limited by the cost, time, and expertise required to screen samples. The gold standard for MCs and many other algal toxins is high performance liquid chromatography with tandem mass spectrometry (HPLC-MS/MS). This requires high startup cost and analytical expertise, as well as reference standards. User-friendly commercially available enzyme-linked immunosorbent assays (ELISAs) can be costly, have variable reported cross reactivity and cannot distinguish between the over 240 structural congeners (Kaushik and Balasubramanian, 2013; Pirez et al., 2013; Moreira et al., 2014). Similar methodological challenges exist for anatoxins, cylindrospermopsins, lyngbyatoxins, nodularins and saxitoxins, with far fewer laboratories equipped with certified reference standards, established protocols, and expertise (Osswald et al., 2007; Singh et al., 2012; Kaushik and Balasubramanian, 2013; Moreira et al., 2013). In addition, health risk from blooms can vary significantly over time and space and thus necessitate strategic sampling to adequately estimate risk (Salmasco et al., 2017).

1.3. Aquatic developmental model for human health risk from cyanobacterial metabolites.

A combination of analytical tools and aquatic models for toxic effect is needed to adequately protect public health, particularly in regions where blooms persist in drinking water sources, where no other alternate water source exists, and where resources to treat water are limited (Westrick et al., 2010; Pantelic et al., 2013; Roegner et al., 2014; Vlad et al. 2014). Combining analytical methods with aquatic toxicology models can substantially enhance a critical evaluation of efficacy and success of water treatment of bloom-laden source water. Given the number of samples and compounds to be screened over time and space to adequately assess risk to a human population, a high throughput bioassay is needed for implementation. Mechanism-based assays, such as protein phosphatase inhibition, have narrow application with specific compounds. Ecotoxicology (macrophyte, microalgae and invertebrate) models can help to critically evaluate ecosystem risk (Bownik, 2016) from blooms, but do not adequately address potential higher order organisms developmental health effects and have variable success as predictive models for human toxicity (Pietsch et al., 2001; Wiegand and Pflugmacher, 2005; Torokne et al., 2007; Acs et al., 2013; Herrera et al., 2015). Higher vertebrate aquatic models have emerged as improved bioassays, but a major limitation for the application to MCs has been the 100- to 1000-fold discrepancy in toxicological effect and morbidity observed in piscine spp. that likely co-evolved with the class of cyanobacterial metabolites ( Boaru et al., 2006; Chen et al. 2012; Chen et al 2016). Organic anion transporting polypeptide (OATP) expression involved in uptake and adenine triphosphate-binding cassette (ABC) transporters involved in efflux and detoxification appear to both be integral to defining susceptibility of toxicity of various species (Boaru et al., 2006; Steiner et al., 2014; Steiner et al., 2016). Developmental Danio rerio have identified significant morphological and behavioral changes upon exposure to diverse categories of cyanobacterial compounds and extracts (Berry et al., 2007; Berry et al., 2009; Sychrová et al., 2015; Jonas et al., 2014; Jaja-Chimedza et al., 2015; Jaja-Chimedza et al., 2017; Priebojová et al., 2019; Walton et al., 2014). Given this promise, we endeavored to apply a high throughput platform to assess dose dependent developmental risk through morbidity, mortality and behavioral endpoints to address the conundrum of diverse risk presented by cyanoHABs.

1.4. Danio rerio development and photomotor response.

With 70% gene homology to humans, zebrafish also exhibit highly conserved anatomy and physiology with humans (Howe et al., 2013). In particular, neurodevelopment parallels that of humans (Bailey et al., 2013; Bugel et al., 2014; Nishimura et al., 2016). Spontaneous tail contractions, regional innervation of muscles and concurrent development of photoreceptors in the hindbrain begin at approximately 19 to 29 hours post fertilization (hpf) as a normal photomotor response (PMR) in zebrafish (Kokel et al., 2013). Early exposure to exogenous chemicals have been shown to alter the prototypic response to a quick pulse of light at this stage (Kokel and Peterson, 2011; Kokel et al., 2013). Continuing through larval development, zebrafish exhibit consistent patterns of locomotor activity between light and dark photoperiods; presumably an evolutionary adaptive response, application of light results in cessation or slowing of movement, while darkness results in a peak increase in movement, followed by decline or acclimation. Previous work has harnessed these alterations to identify abnormal responses (hypoactivity or hyperactivity) in zebrafish exposed to environmental toxicants (Bailey et al., 2013; Bugel et al., 2014; Nishimura et al., 2016). Recent work has suggested a weakened response and hypoactivity in the dark during the initial period following exposure to MC-LR (Tzima et al., 2017).

A high throughput Danio rerio embryo screen including a PMR assay has proven sensitive in early detection of other classes of chemical compounds with developmental effects, prior to any visible morphological changes or mortality (Reif et al., 2016). Given the heterogenous developmental and chronic health concerns presented by cyanoHABs and need for a robust, high-throughput, bioassay to compliment analytical fingerprinting of source waters contaminated by cyanoHABs, we endeavored to systematically screen commercially-available cyanotoxin standards and representative cyanobacterial-derived bioactive peptides for developmental effects in Danio rerio embryos, including morbidity, mortality, and PMR. We then applied the approach to hydrophilic and hydrophobic bloom components from the Rio de La Plata waters at recreational beaches in Montevideo, as well as to temporal and spatial surveillance of subsistence source community drinking water from fishing communities in Kisumu Bay, Lake Victoria, Kenya.

2. Materials and Methods

2.1. Preparation of cyanobacterial metabolite standards.

Standards of MC-LR, -LA, -LW, -LF, -LY, -RR, -YR, -WR, [D-Asp3]MC-LR, [D-Asp3]MC-RR, and MC-HilR isolated from Microcystis aeruginosa (analytical standard grade, ≥95% HPLC) were purchased from Enzo Life Sciences (Farmingdale, NY, USA). Anatoxin-a (synthetic, ≥98% TLC), CYN (isolated from Cylindrospermopsis raciborskii, ≥95% HPLC), nodularin (isolated from Nodularia spumigena, ≥95% HPLC) were also purchased from Enzo Life Sciences. Aeruginosamides B and C, anabaenopeptins B and F, cyanopeptolin 1040 MB, microginin 690, micropeptin 1106, and oscillamide Y (synthetic, HPLC grade, purity ≥95%) were purchased from LKT Labs (St. Paul, Minnesota, US). L-BMAA hydrochloride (≥97% NMR) was purchased from Sigma Aldrich (St. Louis, Missouri, US). Lyngbyatoxin-a was obtained in kind from Drs. Patrick Videau and Benjamin Philmus (Oregon State University, Department of Pharmaceutical Sciences) and prepared as previously described (Videau et al., 2016). All standards were purchased or obtained as dried powder and dissolved in 100 μl dimethyl sulfoxide (DMS0). Stock solutions were prepared in 0.1 mg/mL solution and stored for no longer than two weeks. Dilutions (0.5% DMSO) were made at 5000, 500, 50, 5, and 0.5 μg/L to represent a range of realistic bloom scenario concentrations.

2.2. High throughput Danio rerio embryo morbidity and mortality screen.

Adult wild-type zebrafish (Tropical 5D) were raised at the Sinnhuber Aquatic Research Laboratory at Oregon State University, Corvallis, Oregon. Adults were maintained in 100-gallon tanks in standard laboratory conditions of 28 °C, 14-hr light/10-hr dark photoperiod in reverse osmosis water supplemented with Instant Ocean™, and spawned for routine embryo collection and staging (Kimmel et al., 1995; Westerfield et al., 1995). Per previously described protocol, a custom automated dechorionator was utilized on embryos at 4 hpf with Pronase solution (31.77 mg/mL; Roche, Indianapolis, IN, USA) and then embryos were individually placed into 96-well plates (one per well) with 100 μL of embryo media (Mandrell et al., 2012). Embryos were exposed to metabolites or extract beginning at 4 to 6 hpf and remained exposed in original solution for five days. A Hewlett Packard D300 Digital Dispenser was programmed to dilute stock solutions in DMSO into 10-fold concentration ranges from 5000 to 0.5 μg/L with control columns replicated on every plate. For each standard or sample tested, two 96-well plates were prepared with 32 embryo replicates per dilution of standard or sample. Each population-based exposure (n=32 per concentration or dilution) was carried out in duplicate. Final concentration of DMSO was 0.5% (v/v) in all wells, including control wells. Solvent controls experiments were also carried out to during this time period with no differences between control media 0.5% (v/v) embryos. Exposed plates were wrapped in parafilm to prevent evaporation and in aluminum foil to prevent light exposure, then incubated at 28°C for five days with no change of media. Embryos were assessed at 24 hpf for four developmental toxicity endpoints (mortality, developmental progression, spontaneous movement after light touch, and notochord malformation) and then at five days post fertilization (dpf) for mortality and 17 morbidity endpoints, utilizing the Zebrafish Acquisition and Analysis Program (ZAAP) (Reif et al., 2016; Truong et al., 2014; 2016).

2.3. Embryo photomotor response (EPR) and larval photomotor response (LPR) assays.

At 24 hpf, PMR was also assessed utilizing a custom built photomotor response analysis tool (PRAT) using previously described methods (Noyes et al., 2015, Reif et al., 2016). In brief, embryo 96-well plates were placed in a dark box and then video recorded during a 30 second dark phase (background), a pulse of light (excitation 1) followed by nine seconds of darkness and a second pulse of light (excitation 2), and finally 10 seconds of dark (refractory period). At 5 dpf, the embryo 96-well plates were placed in a ViewPoint Zebrabox behavior testing system (Viewpoint Life Sciences, Inc., France) at 28°C and exposed to light-dark cycles and locomotor movement recorded per well, as previously described (Knecht et al., 2016; Noyes et al., 2015), with slight modification. In brief, video tracking software was utilized to collect track short and large distance movements of individual larvae and integrate over 60s intervals through repeated cycles of light acclimation, dark stimulation, and dark acclimation at three-minute intervals. Assays were carried out at approximately the same time in the morning to avoid temporal variation. Total movement, along with variation between light and dark phases, were compared between treatments and controls (RCoreTeam 2017).

2.4. Statistical analysis.

Each population-based exposure (n=32 per concentration or dilution) was carried out in duplicate for standards and triplicate for environmental samples. All statistical analyses were carried out using custom R scripts (R Core Team 2017). The morbidity and morphological data was recorded as binary incidences of presence or absence of endpoint outcome from ZAAP, in association with plate and well location. Any chemicals having an incidence rate greater than three standard deviations from the mean in control wells, or control wells with less than 80% mortality at 5 days post fertilization, were identified as outliers and rerun. Response for each chemical endpoint was determined by calculating the lowest observable effect concentration (LOEC) in ug/L, or the concentration at which incidence significantly exceed a threshold over the control or background incidence rate. Given the binary outcomes and treatments in separate wells, a series of Bernoulli trials were carried out (n=32), so a binomial test was used to estimate the LOEC significance threshold. To account for variability in background incidence rates across compounds, the significance threshold (x) was determined for each compound-endpoint pair as: F (x; n_c,e, p_c,e) = P(X>x) ≤ 0.05, where n _c,e is number of controls for this compound and endpoint pair and p _c,e is observed incidence or positive responses in controls for this compound and endpoint pair. For more in-depth discussion of the statistical approach, see Truong et al. (2014, 2016) and Zhang et al. (2017).

The statistical frame work described in depth by Reif et al. (2016) and Knecht et. al (2017) was applied to the EPR and LPR data, respectively. The statistical analysis for EPR considers only the Background (21-29 s), Excitatory (32-39 s) and Refractory (42-48 s). The overall pattern within each interval was compared to the intervals negative control activity using a combination of percent change (50% peak difference) and a Kolmogorov-Smirnov test (Bonferroni-corrected p-value of 0.05) The LPR movement data was analyzed for each light phase (light or dark). The area under the curve of the total movement is computed for each treatment. The area under the curve for each treatment was then statistically compared to the control using a two-way ANOVA (p<0.01) and at least 50% change relative to control as a cut-off for significant effect. Animals dead or malformed at the 120 hpf timepoint were excluded from the LPR data analysis.

2.5. Collection of surface scum at beaches and extraction.

Surface samples from seasonal Microcystis scums from the Uruguayan coast of the Rio de La Plata, at Montevideo (Pirez et al., 2013; Bonilla et al., 2015) were collected in 2015 and 2016. Archived samples were selected based on cyanobacterial biomass and ELISA detection of MCs. The biomass was concentrated using a phytoplankton net of 20 μm mesh and lyophilized. Lipophilic and hydrophilic extracts were prepared from archived samples, as previous described (Berry et al., 2007) with slight modification. Briefly, 100 mgs dry weight of freeze-dried biomass were extracted twice initially with 10 mL CHCl₃ and then with 10 mL 30% ethanol (EtOH), respectively. EtOH (hydrophilic) and CHCl₃ (hydrophobic) extracts were then centrifuged, syringe filtered (0.2μm; Millipore, Burlington, MA), dried, and re-suspended in 1mL of 100% and 30% ethanol, respectively. This procedure was carried out in triplicate for each sample; one aliquot from the extract remained for screening through previously published methods by MALDI-TOF (Roegner et al., 2014) and a broad specificity ELISA (Pirez et al., 2013) at the Chemistry Faculty of Universidad de la República (Montevideo, Uruguay), while the other two aliquots of each extract were partitioned for screening by Danio rerio developmental assay and HPLC-MS/MS, respectively. All samples were frozen upon receipt and screened with a several-week time frame to minimize any temporal variation. Projected and initial bloom and extract cyanotoxin screening was utilized to prepare dilutions within environmentally relevant ranges for Danio rerio embryo developmental toxicity screening purposes.

2.6. Collection of cyanoHAB tropical lake drinking water.

As a part of a larger monitoring and global health study to examine subsistence fishing community health risk in Kisumu Bay, Lake Victoria, Kenya, surface water samples were collected from lakeside villages with source waters for daily drinking and use routinely impacted by blooms. Samples (n=6) were selected from two distinct seasons (rainy November 2015 and dry season in February 2016), in which blooms were present. Physicochemical parameters, cyanobacterial cell counts and abundance, and chorophyll-a were also determined, following previously published methods (Sitoki et al., 2012). For intracellular cyanotoxin determination and chlorophyll-a, water was collected in amber-colored glass 250 mL bottles and kept on ice. Within 12 hrs., 50-100 ml of surface water were passed through GFC Filters (Sigma Aldrich, St. Louis, MO), which were dried in an oven at 40 degrees C for 48 hrs. Filters were then wrapped in aluminum foil and frozen until analysis. Simultaneous filters were taken for chlorophyll-a measurements, along with nutrient analysis at Kenya Marine Fisheries Research Institute (KMFRI), Kisumu, Kenya.

For cyanotoxin extraction, modified from Beversdorf et al. (2017), filters were cut into four pieces using sterilized scissors and suspended in 5% acetic acid in water. Cells were lysed on the filters using freeze-thaw cycles for 30 min at −80°C C and thawing 10 min at 50°C with vortexing between cycles (Metcalf and Codd, 2000). After the final thaw, methanol (MeOH) was added to 67% with acetic acid at 5%. Filters were sonicated in 67% MeOH with acetic acid at 50°C for five minutes and then centrifuged. The supernatant was transferred to a scintillation vial. The remaining filter and biomass was in each tube was washed again with 100% MeOH and 5% acetic acid with vortexing; after centrifuging, the supernatant was added to the scintillation vial. The full extract was dried with nitrogen gas at 37°C. Dried extracts were re-suspended in 1 ml of 70% MeOH and spiked with ¹³C-phenylalanine as an internal standard prior to analysis via LC-MS/MS.

2.7. HPLC-MS/MS analysis of extracts.

Lyngbyatoxin-a analysis was conducted on lake water extracts courtesy of Dr. Benjamin Philmus at Oregon State University, as previously described (Videau et al., 2016). The following MCs and cyanopeptides were quantified from lake water samples using LC-MS/MS: MCs-LR, -YR, -RR, -LA, Dha7-MC-LR, nodularin, cyanopeptolins 1041, 1020, 1007, anabaenopeptins A, B, and F, and microginin 690. The analysis was conducted as previously described (Beversdorf et al., 2017). Mass-to-charge ratios for anabaenopeptin B and desmethyl-MC-LR are 837.544 > 70, 837.544 > 201.4 and 981.531 > 103.2, 981.531 > 135.3, respectively.

3. Results

3.1. Metabolite standards alone caused limited morbidity and mortality in Danio rerio.

Of the sixteen cyanotoxins and eight bioactive representative cyanopeptides tested, only CYN, D-asp3-MC-RR, L-BMAA, and lyngbyatoxin-a resulted in any significant mortality at 24 hpf (n=32), relative to control. CYN, D-asp3-MC-RR, L-BMAA, lyngbyatoxin-a, MC-LR and MC-RR were the only five compounds that resulted in significant relative mortality at 5 days post fertilization (Figure S1). However, of these, only lyngbyatoxin-a exhibited dose-dependent toxicity with a lowest observable effect concentration (LOEC) at 5 μg/L. None of the following metabolites caused any significant mortality: MC-LA, -YR, -LW, -LF, -LY, -WR, -MC HilR, D-Asp3 MC-LR, anatoxin-a, nodularin, and additional linear and cyclic peptides, aeruginosamide B, aeruginosamide C, anabaenopeptin B, anabaenopeptin F, cyanopeptolin 1040 MB, microginin 690, micropeptin 1106, oscillamide Y.

Alongside morbidity at 24 hpf and 5 dpf, three morbidity endpoints at 24 hpf, and 17 morbidity developmental endpoints are depicted in a heat map depicting LOEC (Figure 1). In addition to those metabolites causing appreciable mortality at 24 hpf and 5 dpf, anabaenopeptin B, cyanopeptolin 1040 MB, microginin 690, anatoxin-a, MC-LA, and MC-hilR, exhibited some nonspecific effects, with the primary endpoint being pericardial or yolk sac edema (PE or YSE, respectively). Twelve of 17 cyanotoxins and three out of eight additional peptides had at least one morphological endpoint significantly altered, with the exceptions of MC-LY, -YR, -WR, and D-Asp3 MC-LR.

Figure 1. Heat map of lowest observable effect concentration (LOEC) of cyanotoxins and unmonitored metabolites.

Darker coloration indicates more potent effect level. If not otherwise stated, toxicology endpoints were determined at 120 hpf. Each chemical’s LOEC was the lowest concentration that induced statistically significant effects in each endpoint compared to control.

3.2. Lyngbyatoxin-a immersion results in high mortality and progressive developmental impact.

Immersion in lyngbyatoxin-a resulted in mortality and progressive developmental effects at the LOEC (Figure 2). With 10 out of 17 morphological endpoints affected at 5 dpf, mortality was observed in a dose-dependent fashion down to 5 μg/L within 24 hpf. Significant mortality of 11.4%, 15.3%, and 15.1%, 61.9% and 70.8% were observed at the concentrations of 5, 50, 500 and 5000 μg/L, respectively. High incidence of morphological malformations in remaining individuals at 50 μg/L was also observable with any affect occurring in 18.4% at 50 μg/L and 31.4% at 500 μg/L. Other frequently observed population malformations at 50 and 500 μg/L included: YSE (8.2 and 22.5%), axis (8.8 and 17.5%), eye (8.2 and 22.5%), snout (8.2, 22.5%), jaw (8.2, 22.5%), otic (4.9 and 5.0 %), PE (13.4 and 35 %), pectoral fin (6.9 and 17.5%), caudal fin (3.9%), and swim bladder (7.84 and 2.5%).

Figure 2. Dose dependent (0.5, 5, 50, 500 and 5000 μg/L) cumulative morbidity and mortality for lyngbyatoxin-a at 24 hpf and 5 dpf.

Standard provided in kind by the Philmus Laboratory. Each concentration had 32 replicates distributed across two 96-well plates, including the control wells. Percent incidence of mortality or morbiditys is depicted. Stock solutions were maintained in DMSO and serial dilutions were performed with a digital dispenser (HP Tecan D300e); DMSO concentration (v/v), including for controls, was uniformly maintained at 0.5%. DMSO solvent controls and media controls were also compared with no significant difference in mortality or morbidity observed.

3.3. Photomotor Response (PMR) is altered at 24 hpf with select metabolites.

With inclusion of 24 hpf PMR in morbidity and mortality screen, 16 out of 17 cyanotoxin standards and six out of eight representative linear and cyclic standards resulted in developmental effects. In contrast to morphological alterations and morbidity observed predominantly with established cyanotoxins, low levels of eight cyanobacterial metabolites altered PMR significantly at 24 hpf (Table 1). Interestingly, six of eight of unmonitored peptides (aeruginosamide B, aeruginosamide C, cyanopeptolin 1040 MB, microginin 690, micropeptin 1106, oscillamide Y) altered PMR, yet only five out of 17 (MCs – LY, -RR, -WR, –YR, and lyngbyatoxin-a) established cyanotoxins significantly altered PMR response. The majority of compounds resulted in hypoactivity, with the notable exceptions of oscillamide Y, MC-LY, and lyngbyatoxin-a. With respect to lyngbyatoxin-a, a 24 hpf PMR hyperactive perturbation was observed at 50 μg/L and is predictive of progressive morbidity and mortality observed at 5 dpf in exposed larvae. Figure 3 depicts a representative hypoactive (microginin 690) and a hyperactive PMR response curve (lyngbyatoxin-a).

Table 1. Photomotor response assay at lowest observed effect concentration for standards.

Represents statistically significant (p<0.05) alteration in excitatory phases at 24 hpf with at least 50% change compared to control. No significant alterations were noted in either baseline or refractory, or in any photoperiods at 5 dpf. Hypo- or hyperactivity indicated.

Standard	LOEC for 24 hpf PMR	Activity
aeurignosamide B	0.5 μg/L	Hypo
aeurignosamide C	500 μg/L	Hypo
cyanopeptolin 1040 MB	5 μg/L	Hypo
microginin 690	0.5 μg/L	Hypo
micropeptin 1106	0.5 μg/L	Hypo
oscillamide Y	500 μg/L	Hyper
Lyngbyatoxin-a	50 μg/L	Hyper
microcystin-LY	0.5 μg/L	Hyper
microcystin-RR	5000 μg/L	Hypo
microcysin-WR	500 μg/L	Hypo
microcystin-YR	5 μg/L L	Hypo

Figure 3. Photomotor response curves exhibiting hyperactivity and hypoactivity.

Mean total movement (y-axis) is charted over time for each relevant concentration of metabolite (μg/L). The red line depicts first and second flash of light and demarcates the excitatory phase.

3.4. Microcystins from Microcystis aeruginosa blooms predominant in Hydrophilic portion

Bloom extracts were retrieved from archived samples from Montevideo, Uruguay. Samples originally screened by ELISA for total MCs for beaching monitoring purposes were extracted as described with chloroform and ethanol. Figure S2 provides the MALDI spectra from the extracts of samples originally screened by ELISA and indicates the likely presence of diverse MCs. Significant potential MC peaks in the distinct extracts are denoted by mass. Based on initial approximate quantitation by ELISA the 2015 sample (“EtOH and CH₃Cl₃ extract 1”) was estimated at 60.78 ng MCs/mg and 0.13 ng MCs/mg for the hydrophilic and hydrophobic portions, respectively, while the 2016 sample (“EtOH and CH₃Cl₃ extract 2”) was estimated to have 137.60 ng MCs/mg and 0.06 ng MC/mg for each elution. Projected exposures for zebrafish ranged peaked at approximately 30 μg/L and 300 μg/L for the EtOH extracts, while the CH₃Cl₃ extractions were less than 0.01 μg/L.

3.5. Hydrophobic extract exhibits profound developmental impacts despite only trace microcystins detected.

While dose dependent morbidity and mortality effects were observed at the two highest dilutions of the CH₃Cl₃ (hydrophobic) extracts from bloom material (Figure 4a), with extract from a 2016 beach monitoring having a more potent effect, no morbidity or mortality was observed in the MC- containing EtOH (hydrophilic) extracts (Figure 4a). Notably, the more diverse physiologic endpoints at 5 dpf were observed in the extract from 2016, including YSE, PE, axis, caudal fin and swim bladder abnormalities. No MCs or metabolites were detected in the CHCl₃ (hydrophobic) extracts when screened via HPLC-MS/MS after transport; the ELISA method had detected trace amounts, which would have translated to the highest concentration in the range tested having 0.063 and 0.031 μg/L for total MCs. In contrast, both EtOH (hydrophilic) extracts had concentrations of MCs exceeding both recreational and drinking guidelines (measured MC concentrations of 17.1 and 308.4 μg/L, Table 2); there was no significant morphologic change or mortality observed in these extracts, despite the combination of metabolites.

Figure 4. a) Heat map of developmental toxicity LOEC at 24 hpf and 5 dpf for Rio de La Plata extracts b) Dose dependent PMR for EtOH (hydrophilic) and CH₃Cl (hydrophobic) extract 2.

The no dilution (x) contained concentrations for each metabolite as delineated in Table 3, with serial dilutions or concentrations, following accordingly. At x (no dilution), EtOH extract 2 contained 308.4 μg/L MCs in along with Cyanopeptolins 1007, 1041, and 1020 at concentrations of 57.4, 13.5, 2.3 μg/L, respectively.

No significant dose dependent effects for morbiditiy and mortablity were observed in EtOH extracts despite detected levels of toxins, while no MCs or other metabolites tested were detected in the CH₃Cl extracts.

Table 2. HPLC-MS/MS detection of cyanobacterial metabolites (μg/L) in ethanol extracts (hydrophilic) from Rio de La Plata beach surface scum.

Concentrations represent highest dilution prepared for Danio rerio immersion experiments. No metabolites were detected in the CHCl₃ extracts (hydrophobic).

Cyanobacterial Metabolite*	m/z	Rio de la Plata EtOH extract 1	Rio de la Plata EtOH extract 2
		μgL	μgL
Microcystin-LR	995.619 >127.1, 135.3	10.85	180
Nodularin	825.522 > 135.3, 103.2	0	0
Microcystin-YR	1045.633 > 127.1, 135.3	4.415	23.05
Microcystin-LA	910.617 > 135.3, 776.4	0.0221	1.61
Microcystin-RR	520 > 70.1, 135.1	0.585	95.15
[Dha7 (desmethyl)]- Microcystin-LR	981.531 > 103.2, 135.3	1.27	8.61
Total Microcystins	NA	17.1	308.4
Anabaenopeptin B	837.544 > 70, 201.4	0.199	0
Anabaenopeptin F	851.757 > 175.1, 201	1	0
Anabaenopeptin A	844.532 > 84.3, 637.4	0	0
Microginin 690	691.368 > 343.1, 510.2	0	0
Cyanopeptolin 1007	1007.54 > 776.3, 989.6	3.59	57.4
Cyanopeptolin 1041	1042.528 > 828.3, 1024.5	0	13.45
Cyanopeptolin 1020	1021.6 > 776.4, 989.6	1.15	2.34

^*No metabolites were detected in CHCl₃ extract

^**Values represent highest concentration in serial dilution

Table 3 shows LOEC for 24 hpf and 5 dpf PMR responses in all extracts. Notably, LPR (5 dpf) was significantly altered for all the extracts at a range of concentrations. In contrast, EPR (24 hpf) was only altered in the CHCl₃ extracts (Figure 4b) from the recreational beach water sample with alterations of the baseline (hypoactive) and excitation phase (hypoactive). The EPR alterations at 24 hpf coincided with LPR morphological changes and morbidity at 5 dpf, as has been described with other compounds (Noyes et al., 2015).

Table 3. Photomotor response assay at lowest observed effect concentration at 24 hpf and 5dpf for all extracts.

Baseline, excitatory, and refractory LOEC were assessed at 24 hpf and all, dark, and light activity LOEC at 5 dpf, with at least 50% change from control. Hypo- or hyperactivity indicated and value in parenthesis reflect response at corresponding lowest dilution of sample of bloom (p<0.01) where x (or no dilution of the sample) contained 17.1 and 308.4 μg/L MCS in EtOH extracts 1 and 2, respectively, along with detectable Anabaenopeptins and Cyanopeptolins, and no MCs detected in the CHCl₃ extracts (see Table 2).

Extract from Rio De La Plata, Uruguay	24 hpf PMR LOEC			5 dpf PMR LOEC
(type of sampling site)	Baseline	excitatory	refractory	all	dark	light
Rio de la Plata CHCl3 extract 1	n/a	n/a	n/a	0.01× (hyper)	0.01× (hyper)	0.01× (hyper)
Rio de la Plata CHCl3 extract 2	no dilution	0.1× (hypo)	n/a	no dil (hyper)	0.01× (hyper)	n/a
Rio de la Plata EtOH extract 1	n/a	n/a	n/a	0.1× (hyper)	0.1× (hyper)	0.01× (hypo)
Rio de la Plata EtOH extract 2	n/a	n/a	n/a	0.01× (hyper)	0.01× (hyper)	no dil (hyper)

^*n/a signifies no significant perturbation of PMR

^**(hyperactive or hypoactive response indicated)

3.6. Lake Victoria community source waters indicate developmental toxicity risk through multiple seasons and despite variable toxin production.

PMR data from cyanobacterial extracts from three communities on Lake Victoria in two separate seasons indicates a developmental risk from source water both in presence of MCs both above and below the WHO provisional guidelines and U.S. EPA health action levels for children. MCs were detected in community source waters, but only exceeded WHO and U.S. EPA regulatory guidelines in two samples, while cell counts and chlorophyll-a levels consistently fell within the WHO action level of moderate to high risk for recreational activities. Table S4 provides the following for each sample: total cyanobacterial cell count per mL, WHO category of risk for recreational use based on cell count, chlorophyll a (μg/L), and the concentration of each metabolite (μg/L) detected in source waters, representing the highest concentration of exposure in extracts. It also includes whether a hypoactive or hyperactive response occurred with EPR at 24 hpf and LPR at 5 dpf and the dilution of extract at which an LOEC was observed. No morbidity or mortality was observed with direct immersion in any of the serial dilutions of extracts from Lake Victoria waters, despite exceeding the WHO provisional guideline for drinking water in two of the six samples and exceeding EPA health action levels for children in three out of the six samples. However, the sample with the greatest concentration of MCs (15.85 μg/L of LR, 17.75 μg/L total MCs) did result in hypoactivity during the excitatory phase with no dilution. Yet all six samples from both rainy and dry seasons resulted in altered LPR, even in samples diluted beyond theoretical concentrations in actual water. The majority of the alterations were hypoactive response during light or dark periods or both, however, hyperactivity was observed, particularly, from samples taken when cyanobacterial cell counts were deemed high risk. The cyanobacterial cell counts and chlorophyll-a across seasons were representative of trends found in eleven lakeside villages from October 2015 through May 2016 (publication in review).

4. Discussion

The diversity of cyanoHABs and the compounds they can release into the water column challenge recreational surface water monitoring and drinking water facilities as no single form of remediation or treatment completely addresses and eliminates all cyanotoxins, and we simply lack data on aforementioned bioactive compounds (Westrick et al., 2010; Pantelic et al., 2013; Roegner et al., 2014a; Vlad et al., 2014). In this manuscript, we present a high throughput and automated format that allows for relatively quick observation of toxicological endpoints (mortality or observable developmental endpoints) and perturbations in PMR as a representative neurobehavioral endpoint. The automated platform has the benefit of comparative application across a wide array of compounds, multiple extracts of environmental samples, and spatiotemporal samples of regional blooms impacting drinking water infrastructure or recreational waters. Used in tandem with existing analytical approaches, such as HPLC-MS and ELISA, the platform can be utilized as an accessory means to assess developmental risk from chronic exposures and to further fingerprint unidentified chronic toxicants associated with cyanobacterial biomass. Breakthrough into finished multi-stage treated drinking water presents a very eminent threat, particularly in the face of prolonged bloom situations. The shutdown of the city water source for Toledo, Ohio, in November 2013, and Salem, Oregon, in July 2018, suggests the vulnerability of modern treatment facilities (Cha and Stow, 2015; McCarty et al., 2016). It is paramount to understand both acute toxicants and chronic health risks in bloom-impacted surface waters, and to integrate developmental toxicology models into monitoring and surveillance to accurately assess efficacy of mitigation and water treatment strategies. In infrastructure limited regions, chronic exposures in populations lacking adequate drinking water treatment or alternate water supplies (Roegner et al., 2014b) have largely been ignored (Merel et al., 2013a; Merel et al., 2013b). Furthermore, the perturbations in Danio rerio survival, development, and PMR, particularly among less well studied metabolites, contributes substantially to the emerging body of literature identifying the prevalence of and unearthing chronic risks from individual cyanobacterial metabolites and supports calls particular attention to public health risk assessment for children or other vulnerable populations (Weirich et al., 2014).

Our study expands upon previous use of the Danio rerio embryo toxicity model to identify mortality or teratogenicity of cyanotoxins (Cazenave et al. 2006; Purdie et al. 2009a, 2009b; Tzima et al., 2017; Wang et al., 2005), as well as of cyanobacterial cultures and environmental samples (Berry et al., 2007; Berry et al., 2009; Sychrová et al., 2015; Jonas et al., 2014; Jaja-Chimedza et al., 2015; Jaja-Chimedza et al., 2017; Priebojová et al., 2019; Walton et al., 2014), with the notable inclusion of sensitive neurobehavioral endpoints. In agreement with previous publications, exposure to routinely monitored cyanotoxins, such as MCs or CYN, result in limited mortality, particularly within the range of environmentally relevant concentrations (Berry et al., 2007; Berry et al., 2009). This is a shortcoming of this approach – less lipid soluble compounds do not readily diffuse across membranes (de Koning et al. 2015; Song et al., 2011), and mortality has only been observed with methods to bypass membrane transport. Microinjection of MC-LR, a protein phosphatase inhibitor, into Danio rerio embryos at nanomolar concentrations, results in dose-dependent toxicity, likely through direct impact on phosphorylation cascades and blocking of blastomere coherence through alterations in distributions of B-catenin and cadherins (Wang et al., 2005). Similarly, microinjection of extracts with MCs from Planktothrix agardhii into medaka embryos resulted in dose dependent mortalities, in direct contrast with P. agardhii extracts with no MCs (Lecoz et al., 2008). Cylindrospermopsis raciborskii and Aphanizomenon ovalisporium extracts with CYN yielded similar mortality when microinjected (Berry et al., 2008), but not when performed in static immersion. While useful for mechanism-focused studies, the time and skill needed for microinjection limits the approach with more hydrophilic compounds. However, hydrophobic components can present a greater challenge for analytical approaches, illustrating the advantage of incorporating this supplemental screen.

As in mammalian species, the distribution and expression of OATPs influences congener-specific susceptibility and organ targeting of MCs in zebrafish (Steiner et al., 2016), and may result in overall reduced susceptibility in piscine species (Boaru et al., 2006). In our study, we did observe mortality at LOEC over five days post fertilization for CYN, MC-LR, and MC-RR in static immersion, which may be due to differences in dechorionation techniques lending for increased permeability, variability in standards used, or biological differences between strains of zebrafish with those in previous work. We did not see any difference between controls with and without DMSO; however, we cannot completely eliminate the role of solvent in altering membrane permeability (de Koning et al., 2015). In contrast, the significant dose dependent toxicity, morbidity, and PMR perturbations observed with lyngbyatoxin-a suggest a need to delineate exact mechanisms of effect, to pursue investigation with other structural variants, and to reassess overall risk from these compounds both to humans and aquatic species, including distribution and prevalence in freshwater and marine blooms, and accumulation in seafood. These findings illustrate the value of exploring this approach with more lipophilic compounds.

Chronic low exposures of zebrafish and mice to MCs has resulted in altered gene transcription, development, and growth patterns (Qiao et al., 2013; Xie et al., 2015; Tzima et al., 2017). We found EPR alterations at 24 hpf with both established (MCs -LY, -RR, -WR, and YR, and lyngbyatoxin-a) and unmonitored metabolites (aeruginosamide B and C, cyanopeptolin 1040 MB, microginin 690, micropeptin 1106, and oscillamide Y), which may predict similar changes in gene expression and later development. Future work could examine gene transcription patterns in relationship to metabolite exposure and neurobehavioral endpoints. Interestingly, lower concentrations of MC-RR and [D-Asp3] MC-RR resulted in mortality, as compared to MC-LR, despite the relatively established greater potency of MC-LR in mammals; previous work with MC-RR at similar concentrations has suggested thyroid endocrine disruption and growth impairment at early time points, as well (Xie, et al., 2015). In addition to MC-RR, the more lipophilic MCs (LY, WR, YR) resulted in perturbations in PMR. For CYN, the lowest concentration in our study (0.5 μg/L) led to appreciable mortality. Previous work with CYN exhibited dose dependent lethality upon microinjection, despite no pathological changes or appreciable mortality following immersion in concentrations up to 50 μg/mL (Berry et al., 2009); there could a bi-modal pattern of toxicity. The PE and YSE found in this study further support a non-specific mechanism of CYN toxicity. As both the mechanism of toxicity and uptake of CYN is poorly understood (inhibition of protein synthesis, oxidative stress), further exploration at the lower end of the concentration range is warranted.

L-BMAA (with observed mortality at 24 hpf and 5 dpf in this study) has been implicated in having a role in neurodegenerative disease. Previous work found increased mortality, delayed hatching, decreased heart rate and edema at immersion at higher concentrations (Purdie, et al., 2009a, Purdie et al., 2009b); supplementation with carbonate amplified developmental, cardiovascular and neuromuscular effects, presumably because of the compound’s excitatory effect as an analogue in the glutamate pathway. As a broad array of cyanobacteria are reportedly capable of producing L-BMAA, additive or potentiating effects could be further investigated with this robust screening platform.

In addition to expanding our understanding of chronic developmental risk from cyanotoxins, we have demonstrated how this approach can lend insight into developmental risk from newly identified or emerging isolated metabolites, not currently well understood. Embryos exposed to the cyclic serine protease cyanopeptolin 1040 MB and linear zinc-metalloproteinase inhibitor microginin 1106 had notable edema and swim bladder alterations at 5 dpf. Previous work with cyanopeptolins exposure in zebrafish has demonstrated changes in gene transcripts related to DNA damage recognition and repair, circadian rhythms, and responses to light (Faltermann et al., 2014). As these categories of metabolites are increasingly documented to co-occur with MCs (Beversdorf et al., 2017; Carneiro et al., 2012; Gadermann et al., 2010; Neumann et al., 2000; Tonk et al., 2009), additive or synergistic effects on neurodevelopment become a concern, particularly at chronic, low exposure levels. Such combinatorial effects may result in the discrepancy in developmental toxicity observed in Danio rerio exposed to bloom extracts with no detectable cyanotoxins in contrast with individual standards alone (Berry et al., 2007; Berry et al., 2009; Lecoz et al., 2018; Pietsch et al., 2001). Future work utilizing this robust platform could explore the effect of mixture exposures across the broad array of metabolites, or combine fingerprinting of metabolites, in the presence and absence of MCs, in environmental samples.

We utilized this platform to further probe the differential toxicity of the hydrophilic (EtOH) and hydrophobic (CH₃Cl) components of environmental samples. Previous work utilizing the zebrafish embryo assays has resulted in teratogenic and developmental effects of diverse categories of compounds including retinoic acids and terpenes, among others (Berry et al., 2007; Berry et al., 2009; Walton et al., 2014; Jaja-Chimedza, et al., 2015; Jaja-Chimedza et al., 2017). Multiple studies have suggested that the lipophilic fractions of cyanoHABs may have more potent developmental effects than the hydrophilic, MC-containing, components of blooms (Berry et al., 2007; Berry et al., 2009), presumably due to the ability to more easily pass across membranes; however, our approach is unique in employing extremely sensitive neurobehavioral endpoints, and the ability to simultaneously screen a number of samples in a high throughput manner. Current cyanoHABs surveillance, mitigation, and treatment does not address these other categories of metabolites, and they may play a role in nonspecific reported symptoms in surveillance of cyanoHAB exposure in recreational settings (Backer et al., 2008; Backer et al. 2010; Levesque et al., 2014). The combined embryo screen platform, including PMR, provides a means to screen for activity of these types of compounds in tandem with existing analytical tools to provide a more robust public health assessment of risk, from long term exposures.

More broadly, we applied this platform to distinct sets of environmental samples in partnership with routine monitoring and our findings provide additional insight for risk assessment purposes. The high throughput method for screening for developmental toxic effect in the context of environmental surveillance in which spatial and temporal variability can be high is particularly useful. In samples from recreational beach monitoring at the Rio de La Plata, we demonstrated potency of hydrophobic portions of bloom extracts compared to hydrophilic portions (Berry et al., 2007; Berry et al., 2009), and future work could try to isolate and identify specific metabolites that induce EPR and LPR changes, yet priority may focus on looking at intervention strategies to reduce the presence of bloom, MCs, and developmental toxicological observed in hydrophobic fractions. The LPR perturbations in the EtOH (hydrophilic portion) at 5 dpf suggest potential for cumulative effects from MCs or other unidentified components, including the categories of peptides explored in this manuscript. Future work in the development of this approach and tool could evaluate how robustly 24 hpf PMR alterations predict 5 dpf morbidity and mortality through screening of numerous environmental samples through both extraction methods. As such, the 24-hour EPR assay might serve as a rapid screening method for early detection of developmental toxicity of cyanoHABs associated with less well characterized components.

Finally, we applied the platform to a chronic subsistence exposure scenario to assess developmental risk from drinking water. Nearshore fishing villages in the Kisumu Bay Region, Lake Victoria, utilize lake water for washing, bathing, cleaning, fishing, and as the principal water source for drinking (manuscript in review). Bloom extracts from all three communities, during both short rains and subsequent dry season, showed a significant alteration in PMR response at 5 dpf (Table S4) from filtered samples, despite no significant morphological abnormalities or mortality. In addition, developmental risk persisted even when cyanotoxins were less than 0.01 μg/L, but cell counts remained high. With livelihoods and homes at the edge of the lake, these communities simply do not have a feasible alternate fresh water source. The evidence of chronic developmental risk coupled with year-round use of bloom-contaminated lake waters by these communities emphasize the importance of developing high-throughput methods to critically evaluate efficacy of mitigation strategies and water treatment to protect the health of the most vulnerable of the populations, including pregnant women and children.

5. Conclusion

In conclusion, we applied a high throughput developmental toxicity embryo assays that included neurobehavioral endpoints to a broad array of cyanobacterial metabolites and environmental samples of cyanoHABs. We systematically tested available standards across a range that encompassed environmentally feasible concentrations, and then took representative samples from temporal and spatial monitoring and surveillance in distinct field environments: a) cyanobacterial bloom from seasonal temperate recreational beach waters and b) tropical surface water monitoring for subsistence setting in which both drinking and recreational exposures occur. We propose that, by pairing this developmental toxicity platform with analytical fingerprinting, the chronic human health implications of unidentified bloom components can be further delineated and the spatial and temporal monitoring of blooms substantially enriched. We also provide novel toxicity data with respect to lyngbyatoxin-a and several cyanopeptides that co-occur with MCs. In addition, the morbidity, mortality and PMR resulting from exposures to cyanobacterial bloom components in distinct monitoring scenarios supports existing literature suggesting that other bloom components present a cumulative developmental risk. Utilized in combination with existing detection methods, high throughput zebrafish developmental and PMR response assays provide a robust approach for future mixture assessment and with which to gauge efficacy of management and source water treatment interventions. These findings provide a tool for chronic risk assessment for water resource managers and public health officials grappling with emerging and increasing risks from cyanoHABs.

Highlights.

• Diversity of toxicants in cyanoHABs challenge chronic risk evaluation

• Characterization of human developmental risk from chronic exposure is limited.

• A zebrafish embryo model with behavioral light response can expand risk evaluation.

• Metabolites other than cyanotoxins and hydrophobic portions impact behavioral responses.

• This zebrafish assay complements analytical tools when applied to environmental samples.