Since the beginning of the 1980s, stocks of European eel have been declining in most of their geographical distribution area. Many factors can be attributed to this decline such as pollution by xenobiotics released into the environment through agricultural, industrial and domestic activities. Because the New European Chemicals Legislation (REACh) is asking for alternatives to animal testing and reduction of animals sacrified in ecotoxicology and in accordance with conservation biology considerations, we have developed an appropriate and reproducible methodology to obtain a post-nuclear fraction of isolated European eel peripheral blood mononuclear cells (PBMC) in order to evaluate the toxicity of xenobiotics using a subproteomic approach. In a first study, we have studied the in vitro toxicity of perfluorooctane sulfonate (PFOS) in eel PBMC exposed during 48 h to sublethal concentrations (10 µg and 1 mg PFOS/L). A proteomic analysis using 2D-DIGE was performed to compare PBMC from the control group with cells exposed to the pollutant. A total of 48 proteins displaying significant changes in abundance were identified and categorized according to their functional classes. Besides providing clues on the cellular pathways mainly affected by PFOS, results allowed the identification of proteins rarely found in other ecotoxicological proteomic studies. These proteins could constitute potential biomarkers of exposure to PFOS in fish. In order to determine the specificity of the proteomic pattern observed after in vitro PFOS contaminations, we have completed the set of data with in vitro exposures to two other xenobiotics, dichlorodiphenyl-trichloroethane (DDT) and cadmium, using exactly the same methodologies as for the PFOS experiments. The aim of this new study was the discovery of protein expression signatures specific of different classes of pollutants. The identification of the proteins of interest by mass spectrometry allowed selecting four candidates for a minimal common signature between the three experiments. Moreover, 10 protein biomarker candidates belonging to diverse functional classes have been selected to develop an Integrated Biomarker Proteomic index (IBP). For the first time, the use of star plot graphs has been applied to proteomic data in order to allow visual integration of a set of early warning responses measured with protein biomarkers. IBP values, as well as the areas of star plots, could be used to provide information about global adverse environmental effects as well as about the pollutants involved. Lastly, the in vivo toxicological effects of PFOS on the whole animal have been investigated. For that purpose, the protein expression profiles in PBMC of yellow eels exposed in vivo to environmental PFOS concentrations (28 days of exposure to 1 or 10 µg PFOS/L), as well as after in situ samplings of fish from Belgian rivers displaying different levels of PFOS contamination, have been studied. The comparison of the in vitro, in vivo and in situ results allowed the identification of two proteins in common, plastin-2 and alpha-enolase, whose expression was found to be significantly affected by PFOS. Interestingly, the expression of these two proteins was also modified in gills of European bullhead (Cottus gobio) exposed in vivo to either 0.1 or 1 mg PFOS/L (Dorts et al., 2011), suggesting their potential use as biomarkers of PFOS exposure in fish species. Moreover, the recurrence of the main functional classes of proteins affected by PFOS exposure leads us to think that in vitro exposure of cells to pollutants might be useful in the prediction of the in vivo toxicity of these compounds.