In October of 1996, a Gymnodinium breve bloom occurred in shellfish harvesting waters of Alabama, Mississippi and Louisiana, Gulf of Mexico, USA. Bloom densities reached 5.6x10(5) cells liter(-1) and bloom residence at shellfish sampling stations ranged from 3 to 28 days. Brevetoxin-2 dominated G. breve toxin profiles in bloom seawater extracts. Shellfish toxicity, assessed by mouse bioassay, exceeded the guidance level for up to 75 days after the bloom had dissipated. Cytotoxicity assays and mouse bioassays showed similar temporal patterns of shellfish toxicity, but the two methods differed in estimations of brevetoxin-3 equivalent toxicity by a factor of 93 to 1. LC-ESI-MS showed the temporal patterns in shellfish toxicity reflected metabolism of G. breve toxins. The molecular ions m/z 1004, 1017 and 1033 dominated LC-ESI-MS spectra of toxic chromatographic fractions from the extracts and were identified as brevetoxin metabolites on the basis of LC-APCI-MS-MS. The discrepancy between cytotoxicity and mouse bioassay estimates of brevetoxin-3 equivalent toxicity resulted from the difference in extraction efficiency of solvents used in the respective methods and the relative sensitivity of the assays to toxin metabolite mixtures present in the extracts. The normalized cytotoxicity assay showed 75% agreement with mouse bioassay positive test samples and 64% agreement with mouse bioassay negative test samples. Published in 1999 by John Wiley & Sons, Ltd.