Transgenic animals are essential research tools, whether to address basic biological questions or to develop preclinical models of human diseases. Their generation through the injection of naked plasmid DNA into the male pronucleus of a fertilized oocyte has been the standard practice for almost 3 decades. However, this approach is invasive, largely limited to the mouse and results in low rates of transgene integration. Although the inefficiency of pronuclear injection can be overcome in small animals by high-throughput screening for transgene integration, this becomes economically more challenging in larger animals, such as sheep, pigs or cows, because of the extreme costs associated with this procedure ($60'000-$300'000). Thus, new strategies to enhance the production and the variety of transgenic animals would be an important asset. One such approach has emerged through the use of lentiviral vectors, a gene delivery system that can efficiently transfer and stably integrate its cargo into a wide variety of cell types from numerous species, and has been successfully applied to generate transgenic mice, rats, pigs and cows. The increasing use of lentivector-mediated transgenesis warrants a full analysis of its modalities. As a step towards this goal, the present study aimed at characterizing the genotypic features of transgenic mice generated by lentiviral infection of zygotes. First, as an indirect method to measure the kinetics of lentiviral integration in early embryos, we examined the rates of transmission of individual proviruses from G0 transgenic mice to their G1 progeny using a PCR-based technique specific for each integrant. The mean rate of transmission of 44% for individual proviruses suggests that vector integration was most often established between the post-S phase of the one-cell stage and pre-S phase of the two-cell stage. We then moved on to define proviral integration site selection in lentivector-generated transgenic mice. Using LAM-PCR-mediated amplification and sequencing of the host genome-provirus junctions, we found that the frequency of proviral integration inside a gene was 1.75- and 1.88-fold higher in transgenic mice and 3T3 fibroblasts, respectively, than expected from the distribution of annotated genes in the mouse genome (Pmice .03e 09; P3T3 5.55e 16). Moreover, integration was not influenced by the transcriptional orientation of the target gene and tended to favour the middle of transcribed regions. We also observed a subtle difference in the integration patterns of lentiviral vectors in 3T3 fibroblasts and transgenic mice, with the latter exhibiting a slightly higher frequency of integration into repeats. Although this difference was statistically not significant, it might reflect the elevated transcriptional activity of repetitive elements in preimplantation embryos. Since proviruses favour intragenic localization, we analyzed G2 mice homozygous for specific lentiviral integrants using PCR-based techniques and observed that the frequency of homozygosity was below the expected mendelian transmission rate of 25%. This suggested that some of the homozygous G2 mice must be nonviable due to the potential of retrovirus integrants to inactivate target genes. However, the normal rates of G1 representation of proviruses that did not achieve homozygosity in G2 suggested an absence of toxicity in the heterozygous state. Processes governing the earliest steps of mammalian development are still incompletely understood. In particular, the stage at and modalities by which early blastomeres become committed to either the inner cell mass (ICM), which subsequently yields the embryo proper, or the trophectoderm, which serves as precursor of extra-embryonic tissues, is unclear. The long held "random" hypothesis for development of the embryonic-abembryonic axis states that, within these two structures, early blastomeres are equally represented. However, recent lineage-tracing experiments support a model of "developmental bias", in which the progeny of the early blastomeres is allocated in a biased fashion. In our study, we took advantage of the specific integration pattern of lentiviral vector – i.e. each lentiviral vector integrates the host genome at specific sites – to "fingerprint" individual blastomeres and to follow them during embryogenesis. For this, we injected lentiviral vectors into embryos at either the one-cell or the two-cell stage. The Southern Blot (SB) analysis performed at E12.5 revealed that the infection at the one-cell stage yielded integrants shared for their majority by both embryonic and extra-embryonic tissues (74% of similar bands), while the infection at the two-cell stage induced a drastically different picture, with only a minority of common integrants between both types of tissues (22% of similar bands). This suggests that early blastomeres – from the two-cell stage on – contribute differentially to these lineages, which argues against a purely stochastic model. However, we still obtained 22% of shared integrations after infection at the two-cell stage, with only 8 out of 27 embryos showing a complete segregation of the proviral integration patterns. This rules out a "fully determined" model. Thus, our observations rather support the "developmental bias" hypothesis.