Automated Subdaily Sampling of Cyanobacterial Toxins on a Buoy Reveals New Temporal Patterns in Toxin Dynamics

Temporal variability of toxins produced by cyanobacteria in lakes is relatively unknown at time scales relevant to public health (i.e., hourly). In this study, a water quality monitoring buoy was outfitted with an automated water sampler taking preserved samples every 6 h for 68.75 days over a drinking water intake. A total of 251 samples were analyzed by tandem mass spectrometry for 21 cyanotoxin congeners in 5 classes producing 5020 data points. Microcystins (MCs) were the most abundant toxins measured (mean ± sd = 3.9 ± 3.3 μg/L) followed by cyanopeptolins (CPs) (1.1 ± 1.5 μg/L), anabaenopeptins (APs) (1.0 ± 0.6 μg/L), anatoxin-a (AT-A) (0.03 ± 0.06 μg/L), and microginin-690 (MG-690) (0.002 ± 0.01 μg/L). Advanced time series analyses uncovered patterns in cyanotoxin production. The velocity of cyanotoxin concentration varied from -0.7 to 0.9 μg/L/h with a maximum positive velocity just prior to peak toxin concentration during nonbloom periods. A backward-looking moving window of variance analysis detected major increases in cyanotoxin concentration and predicted the two greatest increases in MC. A wavelet analysis identified a significant ( p < 0.01) 2.8-4.2 day periodicity in toxin concentration over a ∼25 day period during peak toxin production, which is partially explained by easterly wind velocity ( R = -0.2, p < 0.05). Diversity in congener profiles was explored with principle component analysis showing that cyanotoxin dynamics followed a seasonal trajectory where toxin profiles were significantly clustered (ANOSIM R = 0.7, p < 0.05) on a daily basis. Variability in toxin profiles was strongly correlated with time ( R = -0.8, p < 0.001) as well as the C:N ratio of the toxin pool ( R = 0.17, p < 0.05). The methods employed here should be useful for uncovering patterns in cyanotoxin dynamics in other systems.