Microalgae commonly used as feed for bivalves, Pavlova lutheri (P), Isochrisis affinis galbana (T) and Chaetoceros calcitrans forma pumilum (Cp), were fed to Pacific oyster Crassostrea gigas to assess their nutritional value for larval development and metamorphosis during two experiments. Monospecific, bispecific and trispecific diets were firstly evaluated during 3 weeks from D larvae to young postlarvae. Then bispecific diets, based on different T and Cp proportions, were assessed during a similar period. Concurrently, ingestion was studied through the whole larval and postlarval development for each diet and/or diet mixture. Because lipids are assumed to be a key nutrient for bivalves, biochemical analysis was undertaken on the second set of trials focused on fatty acids and sterols. Compared to the other diet mixtures (mono and plurispecific diet) TCp induced the best larval growth performance (13.2 mu m day(-1)), a high larval survival (98%) but did not result in higher metamorphosis (72%). In contrast, monospecific diet P was the poorest for larvae with low growth and low survival. When varying T and Cp proportions, best larval developments were induced with 25T/75Cp and 50T/50Cp diets, though quite similar to that obtained with 75T/25Cp. In contrast, unbalanced diets (95T/5Cp and 95Cp/5T) led to low larval performances. In addition, grazing experiences showed preferential uptake of microalgae with P < PT << T << Cp << TCp =PCp=PTCp. For mixed diets a low daily consumption (< 10 000 microalgae per larvae) was noted during the first week followed by a second phase (next 8-10 days) with a sharp increase and regular intake, reaching 90 000 microalgae per larvae per day. Finally, a marked drop (40 000 microalgae per larvae) was observed at the beginning of metamorphosis from days 20 to 21. Principal component analysis between main fatty acids (19) and sterols (7) detected in larvae and postlarvae was used to discriminate profiles according to diets and/or metamorphosis competence. The correlation circle representation showed that the 26 variables are well explained by these combined variables (78%) with a repartition along the first principal component according to diets with a gradient from 5T/95Cp to 95T/5Cp. In contrast, postlarvae and larvae were discriminated on the second principal component while no relationships were found between competent and incompetent larvae. (c) 2006 Elsevier B.V. All rights reserved.