In the past years, it has been observed that some compounds present in our environment can disturb the reproduction and development of animals like fishes, birds, or reptiles by interfering with their endocrine system. Indeed, these endocrine disrupting compounds (EDC) can mimic or antagonize the effects of hormones, alter the pattern of synthesis and metabolism of hormones or modify hormone receptor levels. These substances represent a risk for wildlife, and possibly for humans. Up to now, endocrine disruption was mainly evaluated for vertebrates and aquatic organisms and for oestrogeniclike substances. However, soil invertebrates, which play an important role in soil functioning, have rarely been considered. Moreover, as their endocrine system differs substantially from those of vertebrates (estrogens do not seem to regulate endocrine functions in invertebrates), other substances than can mimic invertebrate hormones should be taken into account. In this sense, insect growth regulators (IGR), which are third generation insecticides specially developed to interfere with insect endocrine system, are interesting compounds. These substances are supposed to have a high specificity for insect pest and a low toxicity for non-target organisms. In the first part of this study, the sublethal effects of six IGR (methoprene, fenoxycarb, precocene II, tebufenozide, hexaflumuron and teflubenzuron) were evaluated on the non-target soil arthropod Folsomia candida. The collembola F. candida represents an integral and beneficial part of the soil ecosystem. It is an euedaphic (subsurface) species which plays an important role in soil respiration and decomposition processes and is therefore vulnerable to the effects of soil contamination. This ecologically relevant organism is one of the most appropriate invertebrate test species for the assessment of environmental quality. It is recommended as test organism by the international standard ISO 11267. The 28-days reproduction tests conducted according to this protocol show that F. candida is affected by the chosen IGR. The most toxic compounds were the two chitin synthesis inhibitors, teflubenzuron and hexaflumuron, with an EC50 of 0.05 mg/kg (dw) for teflubenzuron and an EC50 of 0.6 mg/kg for hexaflumuron. These concentrations are probably environmentally relevant (toxicity/exposure ratios <5) and thus show that Collembola population are at risk. Endocrine disruptive damage induced by pollutants are often not detected by classical toxicity test based on acute or chronic exposure as they can be latent and not manifested until later in life (damage can even arise in the second or third generations of individuals). Therefore, multigeneration tests were developed and conducted with the collembola F. candida. Effects of four of the chosen IGR were evaluated on the second generation of collembola according to different types of exposure of the parental generation F0. In the first set-up, the F0 generation was exposed for 28 days to the pollutant. Eggs and juveniles from the F1 generation (generated by the F0) were also in contact with the pollutant. In the second set-up, only the F0 generation was exposed to the compound for 10 days. Neither the eggs nor the juveniles from the F1 generation were exposed to the pollutant. In both cases, the F2 generation was never in contact with the toxic substance. Multigeneration tests conducted show that some of the tested IGR (methoprene and teflubenzuron) have an impact on several generations of collembola, although only the F0 generation was exposed. Moreover, the complementarity of the two experimental set-up (28 days and 10 days) provide important data on the endpoints affected by these compounds which cannot be obtained with classical reproduction or mortality tests. In the second part of this study, as it is established that a specific and consistent protein response is given by organisms exposed to chemical stressors, a toxicoproteomic approach was conducted with F. candida to assess the impact of the chosen IGR on its protein pattern. Two-dimensional gels of F. candida were realized and the protein expression profile of 10-day old juvenile collembola was established. As the reproducibility of the protein pattern was quite good, reference mapping was possible and the protein expression profiles of collembola unexposed and exposed to precocene II and methoprene was compared. Only weak alterations of protein expression were observed. It is possible that the use of pools of whole organisms to realize the 2-D gels have masked some of the induced or repressed proteins. The feasibility to identify collembola proteins by MS analysis was determined. Three proteins were identified but the scores obtained were relatively low, probably because of the lack of homologous sequences in the protein databases. The sequencing of F. candida genome would facilitate further identifications. In conclusion, the classical reproduction and multigeneration tests conducted show that the non-target soil arthropod Folsomia candida is affected by the chosen insect growth regulators at concentrations that are probably close to environmental levels. Moreover, the transgenerational effects of some of these IGR was demonstrated. This can have crucial consequences on the populations of collembola or other non-target soil invertebrates. The disappearance of beneficial soil organisms may affect soil fertility and soil community balance and could lead to serious environmental damage. This should be considered in environmental risk assessment. However, there is no evidence that the observed effects can be considered as endocrine disruption. The elucidation of this question would need molecular approaches. The use of proteomics to obtain a specific response (a "protein expression signature") of F. candida exposed to chemicals seems to be a promising method to acquire early warning response of exposure to identify toxicants in environmental risk assessment. Moreover, as F. candida is a widely used test organism to assess the impact of pollutants in the environment, the obtention of specific protein response to pollutant exposure and in a further time the identification of these proteins would complement the data already available with classical ecotoxicological testing.