The finding that the implanted ZPA converts cells in the anterior nondigit region into only digit 3 and digit 4 suggests that the anterior nondigit region cannot organize digit 2. This idea is supported by the following results. First, the anterior nondigit region never gave rise to mario expression and digit 2 formation when it was joined to the digit 3–4 region (Fig. 6). Second, the anterior nondigit region with grafted ZPA sometimes formed digit 3 and digit 4, but this combination never formed digit 2 (Fig. 7). Not only does the ZPA have the ability to reorganize digit 2 in the anterior nondigit region, but this anterior region is also likely to have a property intolerant of digit 2 formation. When the digit 2 region was transplanted into the anterior nondigit region, the grafted tissue failed to express either hoxa13 (an autopod marker) or mario (a digit 2 marker), and the cartilage element derived from the graft lost morphological features of a digit (Fig. 8A). It is thought that this defect is not due to a long distance of the graft position from the ZPA, because this graft autonomously maintained marker gene expression in the flank region. Because the anterior nondigit region is located next to the digit 2 region, this inhibitory effect may be involved in limitation on the space of the field for digit formation, by restricting the position of digit 2 (Fig. 8M). The ANZ, one of the programmed cell death regions in the developing limb bud, is located in the anterior nondigit region and in the anterior base of wrist at later stages. The mario-expressing region appears to be next to the ANZ (Fig. 8G,H), although it remains unclear whether these two domains are overlapped or not. Interestingly, Xt mice, which show an anteriorly extended digit region, exhibit a significant decrease in cell death in the ANZ (Aoto et al., 2002), and talpid2 chick mutants, whose digit region are wider than that of wild-type and they lack the normal pattern of cell death in the ANZ and PNZ (Dvorak and Fallon, 1991). These findings suggest a relationship between the ANZ and the space for digit formation. Down-regulation of bmp4, whose expression domain is located in the nondigit region (Francis et al., 1994; Francis-West et al., 1995; Yokouchi et al., 1996), was also shown in the Xt and talpid2 limbs (Aoto et al., 2002; Bastida et al., 2004). Taken together with the fact that the bmp4 expressed there is necessary for apoptotic cell death (Yokouchi et al., 1996; Guha et al., 2002), these results suggest that bmp4 is a good candidate for this inhibitory effect. Mario expression was remarkably down-regulated after application of high concentration of BMP4 (Fig. 8J), and cell death increased in this case (Fig. 8I). On the other hand, FGF2, which has an effect opposite to that of BMPs (Niswander and Martin, 1993), extended the expression domain of mario (Fig. 4D). FGF2 application has been reported to diminish the cell death in the ANZ, resulting in additional digit 2s (Riley et al., 1993; Montero et al., 2001). These findings are compatible with the idea that bmp4 and the ANZ in the anterior nondigit region specify the anterior border of the digit region (Fig. 8M). However, the digit 2 region transplanted into the anterior edge of a bud lost molecular and morphological features of digit 2 but was not excluded by cell death. It was reported that an application of a low concentration of BMP4 (0.01 mg/ml) cannot cause extension of the necrotic zone over the distal mesenchyme in the limb (Bastida et al., 2004). Our anterior application of 0.01 mg/ml of BMP4 also caused only a faint increase of dead cells around the bead (Fig. 8K), although mario expression was down-regulated also in this case (Fig. 8L), suggesting that a certain downstream rather than cell death also mediates BMP signal for limiting mario expression into the digit 2 region. It is therefore possible that a signaling cascade other than programmed cell death in the anterior nondigit region also contributes to specification of the anterior border of the digit region (Fig. 8M).