Cell fate in the chick limb bud and relationship to gene expression

The vertebrate limb is an excellent model for studying mechanisms of pattern formation (e.g.: Tabin, 1991; Duboule, 1992, 1995; Eichele and Tickle, 1994; Cohn and Tickle, 1996). In spite of considerable progress in understanding the interactions involved, basic parameters remain unclear. While several fate maps exist (Saunders, 1948; Stark and Searls, 1973; Lewis, 1975; Summerbell, 1976; Bowen et al., 1989), they are not always consistent with one another and detailed information on the fate of small groups of cells is not available either for the mesenchyme or, in particular, the apical ectodermal ridge (AER). A reliable fate map of the mesenchyme would be valuable for dealing with a number of issues; to show what part(s) of the bud give rise to specific structures, e.g. the digits; to see how defined populations of cells maintain or alter expression of genes, for example, Hox genes which have a dynamic pattern of expression; to show how the bud expands, which could be relevant to evolutionary ideas about the origin of amniote limbs (Shubin and Alberch, 1986; Coates, 1991, 1995; Sordino et al., 1995).

The wing and leg buds of the chick embryo develop from lateral plate mesenchyme, grow out from the body wall and lay down limb structures in a proximodistal direction. The antero-posterior axis of the limb is patterned by the polarising region (or ZPA, zone of polarising activity), which is located at the posterior margin of the limb (see Eichele and Tickle, 1994 for review). The product of the Sonic hedgehog (Shh) gene can mediate polarising activity (Riddle et al., 1993) and has a role in setting up anteroposterior pattern via other gene products, such as those of the HoxD genes (Duboule, 1995). Patterning along the proximodistal axis depends on signalling from the apical ectodermal ridge which is a thick epithelial covering, rimming the limb bud. Removal of the apical ridge results in truncation of the limb (Saunders, 1948; Summerbell, 1974). Fibroblast growth factors, FGF-2, FGF-4 and FGF-8, are expressed in the apical ridge and can rescue limb bud outgrowth and patterning following ridge removal (Fallon et al., 1994; Niswander and Martin, 1992; Niswander et al., 1993; Mahmood et al., 1995; Crossley et al., 1996). FGF maintains the progress zone (Summerbell et al., 1973) lying beneath the apical ridge where cells are proliferating. When cells fall out of the progress zone, their positional identity is determined and they differentiate accordingly. Hence cells leaving the zone late form more distal structures than cells leaving early.

Genes at the 5 ′ part of the HoxD complex have been shown to have a restricted expression in the mesenchyme along the anteroposterior axis of early limb buds suggesting involvement of these genes in the patterning process (see Tabin, 1991; Morgan and Tabin, 1994; Duboule, 1992, 1995 for review). In particular Hoxd-13 has a restricted expression in the posterior distal part of the early limb bud. However later, the gene is expressed throughout the prospective hand plate (Dolle et al., 1989). Hence it has been suggested there are two phases of HoxD gene expression, early and late, and each phase has a different function (Morgan and Tabin, 1994; Goff and Tabin, 1997). There may even be a third phase, intermediate between early and late phases (Nelson et al., 1996). We have investigated whether the later expression pattern of Hoxd-13 is due to expansion of the cell population that was originally expressing Hoxd-13 or to induction of gene expression in new cells.

Fgf-4 transcripts are restricted to posterior apical ridge in early limb buds (Niswander and Martin, 1992; Niswander et al., 1994), but are found throughout the apical ridge as development progresses (Duprez et al., 1996). FGF-4 can substitute for the outgrowth signal from the apical ridge (Niswander et al., 1993) but may also play a specialised role in a feedback loop that maintains Shh expression in posterior mesenchyme (Niswander et al., 1994; Laufer et al., 1994). We have followed the fate of Fgf-4 expressing cells in early apical ridge to find out whether the ridge expressing Fgf-4 at early stages gives rise later to the entire apical ridge or whether non-expressing apical ridge cells activate transcription of Fgf-4.

We have constructed fate maps of the limb bud using the lipophilic dyes DiI and DiA. These dyes can be administered easily in a non-invasive way and this allows a detailed, com-prehensive and accurate fate map to be produced. We have used this technology to update previous fatemaps and ask questions that it has not been possible to address previously. We have investigated, for example, the relationship between gene expression domains and cell behaviour. The results are relevant to evaluating the proposal that an anterior bending of the primitive vertebrate limb was central to the evolution of digits in limbs of higher vertebrates.

Operations were carried out on fertilised White Leghorn chicken embryos at stage 20/21 (Hamburger and Hamiliton, 1951). Eggs were incubated at 37°C, for between 48 or 96 hours after the operation, depending on the experiment. Some embryos were fixed at time zero to check the accuracy of the position of the label.

DiI (1,1-dioctadecyl-3,3,3 ′,3 ′-tetramethylindo-carbocyanine per-chloride; Molecular Probes) is a vital dye and member of the car-bocyanine dye family. DiI is a highly fluorescent lipophilic dye that labels the cell membrane and has been widely used for studying cell fate. It is passed on to progeny of labelled cells but does not leak to neighbouring cells and is not toxic (Honig and Hume, 1986, 1989). DiI (3 mg/ml in dimethylformamide) was administered, by pressure-injection using a picospritzer (General Valve), via a micropipette with a tip opening of 3 μm, made using a thin walled 1 mm diameter borosilicate capillary pipette in a Flaming Brown Micropipette puller (Model P-87). This method labels between 130-180 cells, as found by frozen sectioning of control limbs (data not shown). For double labelling experiments a second member of the carbocyanine dye family was also used, DiA (4-Di-16-Asp, D-3883 Molecular Probes).

In all these experiments, cells in the right limb bud were labelled, and measurements were taken of the DiI dot size and position at the time of administration and after the incubation period. The average initial size of the DiI injected dot was 25-50 μm (n=35).

Limb buds were removed from embryos, fixed in 4% paraformalde-hyde overnight and then mounted as wholemounts onto slides, under a coverslip with a solution of glycerol/PBS/Dabco. Limb buds were viewed within 1-2 days after fixation using a Nikon Optiphot 2 micro-scope with fluorescence attachment. Wholemounts were pho-tographed in black and white using Ilford HP5 film rated at 400ASA, or using colour slide film EPY400.

Detailed analysis of the 48-hour and 96-hour limb buds was carried out using the computer program ‘DIGIT’ on a BBC Archimedes 310 computer, to calculate area and percentage values for each labelled cell population.

The protocol followed was as described by Nieto et al. (1996). The Hoxd-13 probe was kindly provided by Denis Duboule and the Fgf-4 probe kindly provided by Lee Niswander.

DiI was injected into mesenchyme at different anteroposterior levels in wing buds, using the somites and somite boundaries as reference points, and the fate of labelled mesenchyme cells was followed for either 48 or 96 hours. The wing was divided into thirds (Fig. 1A). The distance between each injection position was 150 μm and the amount of DiI injected was approximately the same each time, giving a small patch 25-50μm in diameter, containing approximately 150 cells, as shown by sections of control, time zero specimens.

After 48 hours incubation, a dense region of labelled mes-enchyme cells was observed, with a minority of brighter scattered cells lying outside this main region. We do not know the identity of these scattered cells. They could be macrophages engulfing dead or dying DiI labelled cells or they could be a population of more mobile mesenchyme cells,which, as they appear to be brighter than cells in the main core, divide less. These scattered cells appeared to be more numerous anterior to the main region of labelled cells irre-spective of the position in which cells were labelled. However in 96-hour specimens, there were not as many scattered cells. The position, size and shape of the core of labelled cells was used to produce the fate maps (Fig. 1). Our results are compiled from 392 wing and 40 leg buds.

A marked difference was found, after 48 hours incubation, between the behaviour of populations of cells labelled directly beneath the apical ridge at stage 20, in the anterior third of the bud compared to the middle and posterior thirds (Fig. 1B). Average expansion for each site of injection at 48 hours is shown in Fig. 2A and at 96 hours in Fig. 2B.

The most anterior cell populations (opposite somite 15, 15/16, 16) remained in proximal positions, expanded little compared to other regions and remained adjacent to the ectoderm occupying 2% of the limb area (n=7, Figs 1B, 3A). For cells opposite somite 16/17, expansion was greater and the label was distributed beneath the ectoderm in a proximodistal direction, occupying 3% of the limb area (n=2). In the middle third, the labelled cells formed wide streams running proxi-modistally between apical ridge and proximal regions of the bud, occupying around 20% of the limb area (n=35); moreover in the distal limb these labelled streams fanned out in an anterior direction towards the anterior tip (see asterisk Figs 1B, 3B). In the posterior third, the labelled cells formed long, wide proximodistal streams running from the apical ridge proxi-mally, occupying 22% of the limb area (n=23), but no anterior shift was observed (Figs 1B, 3C-E). Thus, although the posterior third produces a greater proportion of the total area, it is the middle third that fills the anterior tip of the wing bud and contributes to much of the handplate (Figs 1B, 2A). The 96-hour fatemaps (Figs 1G, 4G-K) show that digit 2 arises from subapical cells on a level with somite 17/18 (Fig. 4H), digit 3 from subapical cells on a level with somite 18/19 (Fig.4J) and digit 4 from subapical cells on a level with somite 19/20 (Fig. 4K). Note that an anterior shift of cells into the anterior handplate was also observed at 96 hours (Fig. 4I).

To test directly the amount of cell mixing within the wing bud, two different subapical mesenchyme populations were labelled and their fate followed. The main domains did not intermingle when labelled populations were 150 μm apart initially (Fig. 5E). Thus, when populations of cells were labelled subapically at, for example, somite 18 with DiI, which fluoresces red, and at somite 18/19 with DiA, which fluoresces green, there were distinct non-overlapping territories of the densely labelled cells, distally. Some mingling of cells occurred between scattered cells lying outside the main domain of each labelled cell population proximally (Fig. 5E).

DiI was injected into mesenchyme at three proximal locations in stage 20 embryos; midbud (500 μm from the somites; Fig. 1A,C); base of bud (250 μm from the somites; Fig. 1A,D) and adjacent to the somites (50-75 μm from the somites; Fig 1A,E). Average expansion for each site of injection is shown in Fig. 2A.

Labelled mesenchyme midway between the apical ridge and bud base (500 μm from the somites) at stage 20, was generally found within the limb bud 48 hours later (Fig. 1C). Cell populations labelled within the anterior third, showed some expansion, occupied 2% of the limb area (n=2) and remained in a proximal position (Figs 2A, 3F). Cell populations labelled in the middle third showed greatest expansion, occupying 10% of the limb area (n=15) and forming streams of cells in the central region of the limb, lying in the prospective forearm region (Figs 2A, 3G). In the posterior third, labelled cell populations showed large expansion occupying 8.5% of the limb area (n=7) and forming streams running into the prospective wrist area (Fig. 3H).

Labelled mesenchyme at the base of the bud (250 μm from the somites) at stage 20, was generally found at the base of the limb 48 hours later (Fig. 1D). Cell populations labelled within the anterior third remained proximal and did not expand much, occupying 2.5% of the limb area (n=3; Figs 2A, 3J). Labelled cell populations in the middle third also remained proximal but showed greater expansion than anterior cell pop-ulations occupying 8% of the limb area (n=18; Fig. 3K). Posterior populations of labelled cells formed streams running along the posterior margin into presumptive forearm and occupied 9.5% of the limb area (n=8; Figs 2A, 3L,M). There was a large amount of overlap and intermingling between neighbouring labelled populations of cells, and the labelled populations did not always keep in the same relative positions.

Labelled mesenchyme adjacent to the somites, in the body wall (50-75 μm from the somites), at stage 20, remained more or less next to the somites, outside the limb, 48 hours after injection (Fig. 1E). In the anterior third, labelled populations ended up proximal and anterior to the humerus and scapula and expanded almost uniformly over the 48-hour incubation period, occupying 2.5% of the total limb area (n=3; Figs 2A, 3N). Cell populations labelled more posteriorly, in the middle third, underwent greater expansion, which was still fairly uniform and were found proximal to the humerus, occupying 7% of the limb area (n=17; Fig. 3O,P). Very posterior populations of cells formed short proxi-modistal streams running from the somites along the posterior margin into the central region of the limb, occupying 8.5% of the limb (n=11). The labelled cell populations overlapped, and in particular populations in middle and posterior thirds did not stay in the same relative positions as when initially labelled. For example a population of cells labelled on a level with somite 18/19, ended up in a proximal position adjacent to cells labelled opposite somite 19 (Figs 1E, 5F).

DiI was injected subapically into the stage 20/21 chick leg, again using the somites and somite boundaries as reference points. As in the wing, populations of cells labelled in the anterior third (i.e. opposite somite 26, 26/27, 27, 27/28) expanded to form short stripes of labelled cells which remained proximal (Fig. 1H). In contrast populations of cells labelled in the middle third of the leg bud (i.e. opposite somite 28, 28/29, 29, 29/30) expanded considerably, remained distal and their progeny formed streams of labelled cells running between apical ectodermal ridge and proximal regions of the leg. An anterior bending of cell populations labelled in the middle third was observed toward the anterior tip of the leg, although the shift did not appear to be as great as in of the middle third of the ridge showed some expansion and remained proximal (Fig. 4B). Moreover by 48 hours, these labelled cells appeared to have moved out of the apical ridge, into non-apical ridge ectoderm on the anterior margin of the wing bud (Fig. 4B). The cell population labelled in the middle apical ridge (i.e. opposite somite 17/18 and 18) showed greater expansion over the 48 hour incubation period, remaining within and occupying approximately 25% of the the wing (Fig. 4F). Populations of labelled cells in the posterior third of the leg bud (i.e. opposite somite 30, 30/31, 31, 31/32 and 32) also expanded, fanning out posteriorly and forming streams running from the apical ridge into the prospective foreleg region (Fig. 4L). Mesenchyme cells on a level between somites 28 and 29 gave rise to digit 1, somite 29/30-30 to digit 2, somite 30/31 to digit 3 and somite 31 to digit 4.

Cells in the apical ridge of a stage 20 wing bud were DiI labelled at different anteroposterior positions, using somites as reference points. Both basal cells and periderm cells were probably labelled. The labelled cells were initially in a patch 25 μm in diameter (Fig. 4A) containing 50-75 cells, and after 48 hours incubation, these formed a fairly compact ribbon of labelled cells in the ridge. It was also noted that cells labelled in the apical ridge, remained in the apical ridge or ectodermal layer at all times; none were found in the mesenchyme. Average expansion for each site of injection is shown in Fig. 2C.

Labelled cells in anterior third and anterior part final total length of the apical ridge (Fig. 4C). Posterior pop-ulations of labelled ridge cells showed the greatest expansion (Fig. 4D) coming to occupy approximately 50% of the final total area of the apical ridge (Fig. 2C) suggesting a posterior bias in proliferation as in the mesenchyme. At the borders of the labelled populations, it was noted that some labelled cells could be seen outside the main labelled domain, appearing to intermingle with cells in the non-labelled area of apical ridge (see also Fig. 5C).

To investigate specifically the extent of cell mixing within the apical ectodermal ridge, two populations of apical ridge cells were labelled and the distance between them measured immediately after injection and again after 48 hours incubation. When the apical ridge was labelled with DiI at somite 18, and DiA on a level with somite 18/19, the spots were found on average to be 153 μm apart at time zero (n=6). After 48 hours incubation the labelled populations had expanded as expected from the fatemap, and individual domains were easily visible, and were still around 100 μm apart (Fig. 5C). Although some cells were observed intermingling with unlabelled cells (Fig. 5D). This suggests that there is some local but no widespread translocation of cells within the apical ridge.

In order to examine relative movements of apical ridge and underlying mesenchyme, populations of apical ridge cells were labelled with DiI on a level with a particular somite, and cells in the mesenchyme at the same somite level were labelled with DiA and the relative positions of labelled cells in mesenchyme and apical ridge examined 48 hours later (Fig. 5B). When mesenchyme and apical ridge were labelled at the same level at different positions between somite 17/18 and somite 19/20, the labelled apical ridge ended up further anterior than the mesenchyme, by up to 250 μm (Fig. 5A). Injections on a level with somite 17, indicated that there was no anterior displacement of the ridge and mesenchyme and ectoderm appeared to stay in register, but at somite 16 the mesenchyme remained more proximal than labelled ectoderm. Thus the apical ridge does not stay in register with the mesenchyme but generally shifts anteriorly (Fig. 5B).

Hoxd-13 transcripts are restricted to the posterior distal region of the wing mesenchyme at stage 20/21, the anterior limit of expression is on a level with somite 18 (Fig. 6B), but 48 hours later, Hoxd-13 transcripts are found in prospective hand plate (Fig. 6F). The relationship between expression of Hoxd-13 and cell lineage was investigated by marking cells in stage 20 wing buds outside the expression domain, at the boundary of the expression domain and within the domain. For each set of labelled positions relative to the perimeter of the Hoxd-13 domain, cells were marked at distal end, at proximal end and midway along the perimeter boundary (Fig. 6A). The accuracy with which these marks were placed was confirmed by wholemount in situ hybridisation immediately after observing the fluorescence pattern (Fig. 6A). In such time zero specimens, labelled cells were found to be respec-tively on the border (3/4 cases), above (3/3 cases) or within (3/3 cases) the Hoxd-13 expression domain (Fig. 6A,B). Limbs were incubated for 48 hours after labelling, the position of the labelled cells recorded and then wholemount in situ hybridisation was carried out for Hoxd-13. Cells labelled at stage 20 just outside the perimeter of the Hoxd-13 domain were not incorporated into the domain (n=5; Fig. 6C,F). Cells labelled at the perimeter of the Hoxd-13 domain at stage 20 remained on the boundary distally 48 hours later but proximally fell out of the domain (n=5; Fig. 6D,F). Distally, cells labelled just inside the perimeter of the Hoxd-13 domain remained in the domain but proximally ended up outside the domain (n=4; Fig. 6E,F).

Fgf-4 transcripts are found in the posterior of the apical ridge at stage 20/21, the anterior limit of expression is on a level with somite 17/18 (Fig. 6H). Fgf-4 appears to be expressed throughout the apical ridge at stage 24 but at around stage 26 expression disappears (Niswander and Martin, 1992; Duprez et al., 1996). To investigate the relationship between Fgf-4 expression and cell lineage populations of apical ridge cells were labelled with DiI on the anterior boundary of the Fgf-4 expressing part of the ridge and both just inside and just outside the domain of expression at stage 20/21 (Fig. 6G). The accuracy of the labelling was confirmed by wholemount in situ hybridisation after observing the fluorescence pattern and the labels were found in the appropriate locations in 10/12 cases. Limbs were incubated for 24 hours, the position of labelled cells recorded and then whole-mount in situ hybridisation for Fgf-4 was carried out to compare the rela-tionship between expression and the labelled cells. Cells labelled on the Fgf-4 expression border remained on the Fgf-4 expression border 24 hours later (n=3; Fig. 6J). Cells,labelled just anterior to the expression domain, remained anterior to the expression domain (n=3; Fig. 6I) and cells labelled within the expression domain (n=3) remained within the domain 24 hours later (Fig. 6K).

We have produced detailed fate maps for small groups of cells in stage 20 chick wing bud mesenchyme and apical ectoder-mal ridge. A fate map for the apical ridge has not previously been drawn and the fate map for the wing mesenchyme is more comprehensive than previous fatemaps. At stage 20, the majority of cells that will make up the humerus are not in the protruding part of the limb bud. In contrast, radius and ulna arises from mesenchyme 250 μm to 500 μm from the somites; digits and handplate arise from subapical mesenchyme, specif-ically digit 2 from cells on a level with somite 17/18, digit 3 from cells on a level with somite 18/19 and digit 4 from cells on a level with somite 19/20 (Figs 1G, 4G-K).

Our fatemap for chick limb buds is in reasonable agreement with previous fatemaps (Saunders, 1948; Stark and Searls, 1973; Bowen et al., 1989) in that we show that limb elements arise from the posterior two-thirds of the early limb bud. The anterior third of the bud does not expand very much and is destined to form parts of the shoulder joint and upper humerus (see also Bowen et al., 1989). There are some differences between our conclusions and those of Stark and Searls (1973), who followed mesenchyme implants labelled with tritiated thymidine and concluded that the majority of the humerus arises from inside the stage 20 limb bud. They also suggested that the digits arise in a slightly more anterior position than our data would suggest. Rowe and Fallon (1981) however, arrived at the same conclusion as us about the levels at which cells give rise to the digits, based on a quite different approach in which most of the apical ridge was removed and specific small segments of ridge were left in place. The only available fatemap of a mouse hindlimb bud suggests several features in common with our chick limb fatemaps (Muneoka et al., 1989). Most of the distal parts of the limb comes from the posterior of the mouse hindlimb bud and that the femur region arises from the body wall (Muneoka et al., 1989).

Populations of mesenchyme cells labelled subapically show greater expansion and occupy a greater percentage of the limb bud after 48 hours than proximally labelled mesenchyme (Fig.1I). Hornbruch and Wolpert (1970) showed that the mitotic index was high for distal regions of the limb bud but considerably lower for proximal regions. Expansion of distal mesenchyme on a level with somite 18 is 70-fold more than initial injection size (Fig. 2A). If cells double each time they divide, then this amount of expansion can be produced from 6 cell divisions in 48 hours, hence 8 hours per cell cycle. Other workers have shown cell cycle times of between 8 and 13 hours for the mesenchyme (Janners and Searls, 1970; Hornbruch and Wolpert, 1970; Searls and Janners, 1971; Cooke and Summer-bell, 1980). We found that cell populations labelled subapically remained in contact with the apical ridge and extended proximally into the limb bud. Both Saunders (1948) and Bowen et al. (1989) noted in some cases that cell populations labelled near the apical ridge were found in proximal positions which we did not find, but they both used carbon particles which may not remain associated with the cells initially labelled.

There are marked differences in expansion of subapical mes-enchyme at different anteroposterior levels across the limb bud. Posterior and apical mesenchyme expands much more than anterior mesenchyme. It seems unlikely that local cell death can account for this differential expansion since there are necrotic zones both anteriorly and posteriorly in the developing bud (Saunders et al., 1962). Cooke and Summerbell (1980) have shown that there tends to be a higher S phase labelling index posteriorly in the bud suggesting a higher proliferation rate pos-teriorly (Fig. 1I). This could be due to influences from polarising region and posterior apical ridge. Posterior ridge expresses high levels of FGFs and in particular Fgf-4. Li et al. (1996) recently reported altered patterns of expansion of posterior proximal cell populations in response to FGF-2 in chick wing buds, and K. Kostakopoulou et al. (personal communication) found that labelled posterior cell populations expanded distally in amputated limb buds when a source of FGF-4 was applied. Our results also show a striking differential expansion of subapical mesenchyme across the anteroposterior axis. Cells labelled subapically in the middle third of the limb bud fan out into the anterior tip (Fig. 1B). Bowen et al. (1989) also observed anterior fanning out of marked cells in a similar position at stages 22 and 24. This anterior shift could be due to the fact that the posterior third expands considerably while the anterior part of the limb expands very little causing overgrowth of the middle third. Cell populations in the posterior third of the limb bud give rise to wide, straight tracts of cells with no anterior bending.

We also noted that subapical cell populations in the bud keep in register (Fig. 1B), whereas proximal cell populations show more overlap and relative displacements (Fig. 1D). Relative displacements of proximal cell populations may be allowed because patterning of these cells may already have been specified irreversibly.

Our fatemap of the apical ridge shows that cells in the posterior two-thirds at stage 20 make up the entire apical ridge later in development, and that labelled cells fall out of the apical ridge anteriorly. Cell death appears to be evenly distributed through-out the ridge at stage 20 (Todt and Fallon, 1984). When limb duplications are induced by placing beads soaked in retinoic acid underneath the apical ridge at stage 20, the apical ridge is longer than normal and covers the induced structure (Lee and Tickle, 1985). It seems likely that the longer apical ridge is due to the maintenance of cells in the apical ridge that would normally leave. This is consistent with the idea that the apical ridge is produced from a finite population of cells early in development. Experiments in which quail apical ridge was grafted in place of chick apical ridge show that the apical ridge behaves autonomously and cells are not recruited from either dorsal or ventral ectoderm (Saunders et al., 1976).

We have shown that the apical ridge does not keep in register with the underlying mesenchyme. This disparity appears to be due to differences in expansion between mesenchyme and apical ridge, with the apical ridge generally expanding more in an anterior direction. We observed intermingling of labelled and unlabelled cells, but only at the borders of the labelled domain, suggesting that local cell rearrangements may occur as the ridge expands.

Hoxd-13 has a dynamic expression pattern and initially tran-scripts are restricted to posterior distal mesenchyme and later appear throughout most of the prospective handplate. We have found by labelling populations of wing cells and following their fate over 48 hours, that some cells that were expressing Hoxd-13, switch off Hoxd-13 and no longer express it. Hoxd-13 can be reactivated in proximal cells by locally applying FGF-4 (K. Kostakopoulou, personal communication). Thus it seems likely that cells are left proximally and stop expressing Hoxd-13 because they are too far away from the influence of apical ridge signals as the bud grows out. We have also shown that the anterior shift in Hoxd-13 expression to occupy most of the handplate, up to the interdigit region between digit 2 and digit 3, is in concert with the fatemap and can be accounted for by the way that the limb bud grows. Thus, there appears to be no initiation of Hoxd-13 expression in anterior, previously non-expressing cells and this argues against the idea that Hoxd-13 gene expression in the hand plate represents a distinct separate phase of expression (Morgan and Tabin, 1994; Nelson et al., 1996). Hoxd-13 gene expression is believed to be involved in patterning, proliferation and identity of the digits (Morgan and Tabin, 1994; Duboule, 1995; Fromental-Ramain et al., 1996; Zakany and Duboule, 1996). Our findings do not preclude the possibility that Hoxd-13 has different functions throughout development, involved in growth, patterning and identity early in development (Morgan and Tabin, 1994; Zakany and Duboule, 1996) and in growth and differentiation later (Morgan and Tabin, 1994; Duboule, 1995; Yokouchi et al., 1995; Goff and Tabin, 1997) especially as Hoxd-13 is expressed on it’s own initially, but later in concert with Hoxa-13. In contrast to our findings for Hoxd-13, Hoxa-13, which also begins to be expressed posteriorly and then comes to occupy the distal tip of the limb, has been shown by DiI experiments to be activated in previously non-expressing cells (Nelson et al., 1996). Thus, it appears that elaboration of HoxA and HoxD expression patterns occurs by different mechanisms.

Fgf-4 is expressed in posterior cells of the apical ridge at stage 20 but is expressed more widely throughout the apical ridge as development proceeds. Cells labelled anterior to the expression domain at stage 20 are not incorporated into the Fgf-4 domain after 24 hours incubation. Therefore no new cells are induced to express Fgf-4 and it is the original population of expressing cells in the posterior two-thirds of the apical ridge that expand and fill the apical ridge in late development. Between stage 20 and stage 24 the apical ridge Fgf-4 expression domain extends to a more anterior limit than the mesenchymal Hoxd-13 domain (anterior limit of Hoxd-13 is on a level with somite 18, anterior limit of Fgf-4 is on a level with somite 17/18). Mesenchyme cells have been shown to move towards an FGF source (Muneoka et al., 1989; K. Kostakopoulou et al., personal communication) and it is possible that Fgf-4 signalling to the mesenchyme may contribute to the expansion of Hoxd-13 expressing cells into the handplate.

It has been suggested that evolution of the handplate is due to an anterior bending of the metapterygial axis of the limb at the level of the digital arch and digits then arising from the distal side of the arch i.e. the original posterior side (Shubin and Alberch, 1986; Coates, 1991, 1995). Sordino et al. (1995) showed that early in development Zebrafish fin and mouse limb buds both display similar HoxD and HoxA gene expression patterns. Fin buds however do not display the late limb bud HoxD gene expression pattern which, in amniotes, corresponds to most of the prospective handplate. They suggest therefore that the developing hand/foot plate of the amniote limb has no homologous structures in the Zebrafish fin and conclude that evolution of the autopod (handplate) depended on a second phase of Hox gene expression which was brought about by bending of the metapterygial axis. We find that subapical cell populations in the middle third fan out anterodistally into the anterior tip (see also Bowen at al., 1989). This does not appear to be consistent with the model of Shubin and Alberch (1986), which suggests that mesenchyme at the posterior margin should expand across the anteroposterior axis of the handplate and contribute to each of the digits.

N. V. is the recipient of an MRC Studentship. J. D. W. C. is funded by the Wellcome Trust. We thank the Astor foundation for funding equipment essential for the project and thank Dr Ketan Patel for help and advice with whole-mount in situ hybridisation.