Human poisoning due to consumption of seafood contaminated with phycotoxins is a worldwide problem, and routine monitoring programs have been implemented in various countries to protect human consumers. Following successive episodes of unexplained shellfish toxicity since 2005 in the Arcachon Bay on the French Atlantic coast, a national research program was set up to investigate these atypical toxic events. Part of this program was devoted to fit-for-purpose cell-based assays (CBA) as complementary tools to collect toxicity data on atypical positive-mouse bioassay shellfish extracts. A collaborative study involving five laboratories was conducted. The responses of human hepatic (HepG2), human intestinal (Caco2), and mouse neuronal (Neuro2a) cell lines exposed to three known lipophilic phycotoxins-okadaic acid (OA), azaspiracid-1 (AZA1), and pectenotoxin-2 (PTX2)-were investigated. A screening strategy composed of standard operating procedures and a decision tree for dose-response modeling and assay validation were designed after a round of "trial-and-error" process. For each toxin, the shape of the concentration-response curves and the IC(50) values were determined on the three cell lines. Whereas OA induced a similar response irrespective of the cell line (complete sigmoid), PTX2 was shown to be less toxic. AZA1 induced cytotoxicity only on HepG2 and Neuro2a, but not on Caco2. Intra- and inter-laboratory coefficients of variation of cell responses were large, with mean values ranging from 35 to 54 % and from 37 to 48 %, respectively. Investigating the responses of the selected cell lines to well-known toxins is the first step supporting the use of CBA among the panel of methods for characterizing atypical shellfish toxicity. Considering these successful results, the CBA strategy will be further applied to extracts of negative, spiked, and naturally contaminated shellfish tissues.