Neural induction is the process that initiates nervous system development in vertebrates. Two distinct models have been put forward to describe this phenomenon in molecular terms. The default model states that ectoderm cells are fated to become neural in absence of instruction, and do so when bone morphogenetic protein (BMP) signals are abolished. A more recent view implicates a conserved role for FGF signaling that collaborates with BMP inhibition to allow neural fate specification. Using the Xenopus embryo, we obtained evidence that may unite the 2 views. We show that a dominant-negative R-Smad, Smad5-somitabun-unlike the other BMP inhibitors used previously-can trigger conversion of Xenopus epidermis into neural tissue in vivo. However, it does so only if FGF activity is uncompromised. We report that this activity may be encoded by FGF4, as its expression is activated upon BMP inhibition, and its knockdown suppresses endogenous, as well as ectopic, neural induction by Smad5-somitabun. Supporting the importance of FGF instructive activity, we report the isolation of 2 immediate early neural targets, zic3 and foxD5a. Conversely, we found that zic1 can be activated by BMP inhibition in the absence of translation. Finally, Zic1 and Zic3 are required together for definitive neural fate acquisition, both in ectopic and endogenous situations. We propose to merge the previous models into a unique one whereby neural induction is controlled by BMP inhibition, which activates directly, and, via FGF instructive activity, early neural regulators such as Zic genes. xenopus ͉ default model ͉ Smad5-sbn N eural induction is viewed as a decision made by gastrula ectodermal cells between neural and epidermal fates (1, 2). This process has been best studied in the Xenopus and chick embryos, which led to the emergence of distinct molecular models. The default model, based initially on Xenopus studies, has proposed that bone morphogenetic protein (BMP) inhibition is necessary and sufficient for neural induction (1). Studies in the chick have implicated additional instructive signals, among which FGF is an early and essential one (3, 4). However, one shared conclusion is that neural fate assignment requires the down-regulation of BMP signals (5, 6). What remains controversial is whether BMP inhibition could be sufficient for neural induction. Recently, we and others have introduced a paradigm to test the validity of the default model in frogs, which consists of micro-injection of cell-autonomously acting BMP inhibitors in ventral ectodermal cells of the 16-or 32-cell embryo (5, 6). Fate mapping combined to marker gene analysis indicate that these blastomeres normally give rise exclusively to epidermal cells (5). Those cells are competent for neuralization, but do not become neural if injected with Smad6 or a dominant-negative BMP receptor (5, 6). However, the epidermal-to-neural switch occurs when a low amount of FGF4 (called eFGF in the frog) is combined with those BMP inhibitors, supporting a combinatorial model (5, 6).It remaine...