Fig. 7. Model for the role of Smad5 signalling in amnion homeostasis. (A) The squamous, bilayered wild-type amnion at the early somite stage is composed of a mesoderm layer facing the exocoelom that expresses Twist, whereas an epithelial cell layer that expresses AP-2 lines the amniotic cavity. The chimera analysis demonstrates that Smad5 is essential in the mesoderm, where it functions in a non-cell-autonomous fashion, because mutant mesoderm cells can elicit thickening of wild-type amnion ectoderm. Reciprocally, however, we cannot conclude that Smad5 functions also non-cell-autonomously and/or cell autonomously in the ectoderm. (B) At the early somite stage the Smad5^m1/m1 mutant amnion displays an altered cellular organization with a zone of primarily thickened mesoderm but also thickened ectoderm. The mutant amnion ectoderm is positive for SSEA-1. The anterior amnion ectoderm remains cuboidal (squares), and is in continuity with anterior surface ectoderm and anterior neural tissue. Robust ectopic expression of Bmp2 and Bmp4 in mutant amnion is associated with increased Smad1/8 and ERK1/2 signalling and expression of periostin, Tbx2, fragilis and Msx2. The differential but overlapping ectopic expression patterns of all these Bmp-target genes suggest that cells in the mutant amnion sense different levels of Bmp signalling. How these different levels of Bmp signalling are established in the mutant amnion remains to be explored. (C) The transformation of the amnion to a complex tissue with different Bmp-sensitive cell types, such as e.g. an Oct4⁺, AP⁺, SSEA-1⁺ and fragilis⁺ region reminiscent of PGC-competent epiblast, Flk1-expressing endothelial cells and -globin-expressing cells (see Discussion). The ectopic and appreciable numbers of AP⁺ cells arise suddenly from the 5S stage onwards, which argues strongly for their in situ change in fate. The AP⁺ cells are localized near the ectodermal component of the aggregate, whereas haematopoietic and endothelial cells are positioned at the exocoelomic side. Our data suggest that Smad5 is the predominant Bmp signal-mediator in the amnion, and controls Bmp expression levels and amnion homeostasis tightly. Upon Smad5 removal the control of Bmp expression is released, which results in an excess of Bmp, in take-over by non-Smad5 pathways and ultimately in gain of Bmp signalling defects.