Experimental model of microcystin accumulation in the liver of Oreochromis niloticus exposed subchronically to a toxic bloom of Microcystis sp.

Although accumulation of the liver toxin microcystin in phytoplanktivorous fish has been demonstrated in captive fish and in natural ecosystems, the relation between microcystin in ingested algae and the pattern of buildup of microcystin in fish is poorly known. In this month-long study performed at a Brazilian fish farm, 45 mature Oreochromis niloticus were fed daily with fresh seston periodically dominated by toxic Microcystis sp. Microcystin was measured daily in the food and every 5 days in liver and muscle samples. Control fish received a diet of seston that was low in toxic cyanobacteria. Initially, in treatment ponds, microcystin available for fish increased from 6.5 to 66.9 ng microcystin fish(-1)day(-1), which was accompanied by an increase from 5.5 to 35.4 ng microcysting liver(-1). Microcystin in muscle was below our detection limit of 4 ng g tissue(-1) for the entire study. In the bloom phase, available microcystin reached its highest concentration (4450 ng MC fish(-1)day(-1)) then decreased to 910 ng microcystin fish(-1)day(-1) on day 31. During this period, microcystin reached its highest concentration of 81.6 ng MC g liver(-1) and stayed high until the end of the experiment. A model based on rapid uptake, saturation, and exponential loss was built with these experimental results, and verified with data from the literature. Our model showed that accumulation was up to 50% of ingestion at low doses, but at intermediate doses, the onset of elimination led to a decline of liver burden. Although the accumulation rate confirms the high contamination potential of microcystin, it was balanced by a high depuration rate and this efficient systemic elimination may explain the tolerance of these fish to toxic blooms in the wild.

Microcystin accumulation in liver and muscle of tilapia in two large Brazilian hydroelectric reservoirs.

The objective of this study was to measure levels of the toxin microcystin in different tissues of fish known to feed on cyanobacteria during toxic bloom events. Wild Nile and redbreast tilapia (Oreochromis niloticus and Tilapia rendalli) were sampled from the catch of artisanal fishermen at eutrophic stations of Funil and Furnas reservoirs in southeastern Brazil. Phytoplankton communities in the two reservoirs were quite different taxonomically, but not dissimilar in microcystin content (200 microg g dry weight (DW) seston(-1) at Funil, 800 microg gDW seston(-1) at Furnas). All of the 27 fish sampled contained microcystin, ranging from 0.8 to 32.1 microg g liver(-1) and from 0.9 to 12.0 ng g muscle(-1). Most microcystin variants found in seston were also found in fish liver. T. rendalli had the lowest concentration in both tissues when compared to O. niloticus. In both reservoirs, one of every four fish sampled, always O. niloticus, had a level of microcystins beyond the World Health Organization tolerable daily intake (8 ng g tissue(-1)) and represented a risk for consumers. It is possible that closer study of inter-species variability in toxin burden in cyanobacteria-impacted water bodies will permit the development of guidelines for fish consumption that will better protect public health.

Some misconceptions about the spurious correlation problem in the ecological literature.

It is a common misconception that correlations between variables that share a common term are statistically invalid. Although the idea that such relationships are wholly or partially spurious was rejected decades ago by statisticians, ecologists continue mistakenly to exclude legitimate hypotheses on this basis. Besides directing attention to the statistical literature on the subject, we briefly reconsider the problem from 3 viewpoints: first, the confusion between spurious correlation and spurious inference, second, the problem of concept familiarity and definition, with particular reference to the self-thinning rule for plants, and third, a legitimate concern with measurement error of shared variable components.