Defining drinking water metal contaminant mixture risk by coupling zebrafish behavioral analysis with citizen science

Abstract

Contaminated drinking water is an important public health consideration in New England where well water is often found to contain arsenic and other metals such as cadmium, lead, and uranium. Chronic or high level exposure to these metals have been associated with multiple acute and chronic diseases, including cancers and impaired neurological development. While individual metal levels are often regulated, adverse health effects of metal mixtures, especially at concentrations considered safe for human consumption remain unclear. Here, we utilized a multivariate analysis that examined behavioral outcomes in the zebrafish model as a function of multiple metal chemical constituents of 92 drinking well water samples, collected in Maine and New Hampshire. To collect these samples, a citizen science approach was used, that engaged local teachers, students, and scientific partners. Our analysis of 4016 metal-mixture combinations shows that changes in zebrafish behavior are highly mixture dependent, and indicate that certain combinations of metals, especially those containing arsenic, cadmium, lead, and uranium, even at levels considered safe in drinking water, are significant drivers of behavioral toxicity. Our data emphasize the need to consider low-level chemical mixture effects and provide a framework for a more in-depth analysis of drinking water samples. We also provide evidence for the efficacy of utilizing citizen science in research, as the broader impact of this work is to empower local communities to advocate for improving their own water quality.

Figure 1

Conceptual map depicting the combination of main methods used in this study including; citizen science, an in vivo functional assay using the zebrafish model, and rigorous statistical cluster analysis. Also highlighted is the main finding, that alterations in biological outcomes (mortality, hatching, and activity) are highly dependent on chemical mixtures even at low concentrations.

Figure 2

Heat map, representing concentration of a given metal as a percent of its maximum contaminant level (MCL) in a drinking water sample; MCLs for arsenic (As)—10 ppb, cadmium (Cd)—5 ppb, lead (Pb)—15 ppb, and uranium (U)—30 ppb. Percent mortality and hatching inhibition at 5 dpf are also included. Samples are further categorized into exposures that resulted in no significant change in larval total distance (TD) traveled, significant hypoactivity, and significant hyperactivity relative to egg water controls. Blue colors indicate a lower percentage and reds indicate a higher percentage.

Figure 3

Dot plot representing total distance after a 1:1 sample/egg water exposure. Data is shown in the form of fold change difference between treatment relative to 100% egg water control against the amount of (a) arsenic (µg/L), (b) cadmium (µg/L), (c) lead (µg/L), and (d) uranium (µg/L) present in a given sample. Dashed red line represents current EPA maximum contaminant levels in drinking water. Samples that induced significant hyper or hypoactivity can be found in Supplemental Table SI, p value < 0.05, ANOVA, n = 24.

Figure 4

(a) Examples of principal component analysis (PCA) derived scatter plots that show significant clustering between control-like and hypoactive and control-like and hyperactive behaviors from metal combination inputs of Cr, Mn, Fe, Se, Cd, Sb, Ba, Pb, & U (combination 162) and Fe, Ni, Cu, Se, Pb, & U (combination 3921) respectively. (b) Dot plot representing the percent prevalence of a given metal in mixtures that resulted in significant clustering between control-like and hyperactive (red) as well as control-like and hypoactive (blue) behavior. (c) Bar graphs representing the top 15 combinations of two or three metals within mixture subsets that resulted in significant clustering between control-like and hypoactive and control-like and hyperactive behavior. Clusters were created via PCA and an F-statistic was calculated to determine significance, p value < 0.005.

Figure 5

Box plot representing (a) % mortality and (b) % hatching inhibition at 5dpf after a 1:1 sample/egg water exposure relative to 100% egg water control. Data is plotted based upon samples that significantly induced hypoactivity, hyperactivity, or no significance from egg water controls. Additional information regarding specific samples that induced significant hyper or hypoactivity can be found in Table S1, p value < 0.05, ANOVA, n = 24.