Retinoic acid (RA), the active vitamin A derivative, is an important developmental signaling molecule in vertebrates. In this study, we have assessed whether minimal numbers and/or specific distributions of RA-producing cells can support normal mouse embryonic development. Retinaldehyde dehydrogenase 2 (RALDH2) is the main RA-synthesizing enzyme acting during development. We have generated an embryonic stem (ES) cell line homozygous for an Raldh2 gene disruption, and have analyzed chimeric embryos with various contributions of wild-type cells. Whereas embryos almost completely derived from Raldh2(-/-) cells phenocopy the corresponding germline null mutants, the presence of even small numbers (<10%) of wild-type cells can rescue most of the morphogenetic defects, including embryonic turning and axial elongation, and left-right looping of the heart tube. No consistent bias in the distribution of wild-type cells was observed in the phenotypically rescued Raldh2(-/-) chimeras. Analysis of an RA-sensitive transgene indicates that RA can diffuse from wild-type cells and elicit a widespread transcriptional response in Raldh2-deficient cells. Our results show that few wild-type RA-producing cells, even when present in apparent random distributions, can support early morphogenesis of the mouse embryo. However, the Raldh2(-/-) chimeric fetuses display lung abnormalities, persistent truncus arteriosus, and abnormal myocardial differentiation, showing that subsequent RA-dependent events cannot be fully rescued by the mosaic presence of wild-type cells.