Relationship between dose, distance and time in Sonic Hedgehog-mediated regulation of anteroposterior polarity in the chick limb

A fundamental step in vertebrate limb development is the establishment of limb pattern that leads to the correct arrangement of differentiated cells and tissues. Signalling by the zone of polarizing region (ZPA), which is located in the distal posterior mesenchyme, ensures that each digit arises in its correct position. When the ZPA of a chick wing bud is grafted to the anterior margin of a second chick wing bud, this results in mirror-image duplicated patterns of digits (Saunders and Gasseling, 1968). In the normal chick wing, the digits are 234 (anterior to posterior) while, following a ZPA graft, the digit pattern is 432234. From results of extensive grafting experiments, it has been shown that behavior of cells (e.g. identity of the digit that they will form) depends on distance from the ZPA (Tickle et al., 1975). Shh is a good candidate for a ZPA signal. Shh transcripts map to the ZPA (Riddle et al., 1993) and grafts of Shh-expressing cells induce digit duplication (Riddle et al., 1993). All signalling activity appears to reside in the aminoterminal peptide of Shh (Shh-N) (Marti et al., 1995a; Roelink et al., 1995; Fan et al., 1995), including the ability to specify additional digits in the wing bud (Lopez-Martinez et al., 1995).

Signalling by the ZPA is dose-dependent as the identity of additional digits specified depends on number of ZPA cells grafted. For example, grafts that include only small numbers of ZPA cells specify digit 2 while increasing numbers of cells give an additional digit 3 and finally digit 4 (Tickle, 1981). Further, in both chick and mouse limb buds, reduced levels of Shh expression are correlated with absence of posterior digits (Yang and Niswander, 1995; Parr and McMahon, 1995). This suggests that the number of Shh-expressing cells is related to digit specification. In one model that can account for such dosedependent signalling, a morphogen produced by the ZPA diffuses across the bud to set up a concentration gradient (Tickle et al., 1975). According to this model, high concentrations of morphogen would specify posterior digits and low concentrations of morphogen more anterior digits. Such signalling is long-range and would act over a distance of 200 μm or so (Honig, 1981). There is some evidence that the ZPA can act at a distance and induce formation of a digit 2 in cells that have never been next to the ZPA (Honig, 1981). A short-range signal could also act in a dose-dependent fashion if, upon exposure to the signal, cells themselves produced the same signal but at an attenuated level. A second model suggests that the ZPA activates a series of different short-range signalling molecules and that each signal controls local patterning at a particular threshold concentration (Riddle et al., 1993). In this model, one signal would control digit 4 identity, another digit 3 and yet another digit 2. It is also possible that the ZPA triggers the expression of secondary signals that interact to pattern the anteroposterior (A-P) axis. In this scenario, none of the secondary signalling molecules on their own would be able to mimic the function of the ZPA in specifying digit pattern.

Cells may respond in several ways to polarizing activity. They may respond to the highest concentration of signal that they experience or they may sum the total amount that they see over a certain period. Thus, length of exposure to the morphogen as well as its concentration at any one time may be important in limb patterning. This is illustrated by experiments in which a ZPA graft was removed after different time periods; grafts left in place for short intervals induced no additional digits but grafts left in place for about 16 hours induced an additional digit 2 and more posterior digits after longer periods (Smith, 1980). However, it is not clear whether the same cells are promoted to more posterior fates after longer exposure to ZPA signalling or that different digits are formed from distinct populations.

A major question that has not been resolved is whether Shh acts directly or indirectly to specify additional digits. Direct immunostaining only detects Shh in close association with expressing cells (Marti et al., 1995b) arguing against long-range diffusion. In other systems, Shh signalling appears to have both local and long-range components (Marti et al., 1995a; Roelink et al., 1995; Fan et al., 1995; Ericson et al., 1995). The best evidence for direct long-range effects comes from studies of motor neuron specification (Ericson et al., 1996). However, in the limb there is evidence that Shh may activate secondary signals. For example, Shh-expressing cells lead to activation of Fgf-4 in the anterior ectodermal apical ridge and Bmps in the anterior mesenchyme (Niswander et al., 1994; Laufer et al., 1994).

Here we have explored the role of Shh in patterning the chick limb. Our results are most consistent with a model in which the concentration of Shh is the primary determinant of A-P polarity in the limb. Our results also suggest that Shh is unlikely to act as a long-range signal directly regulating all digit identities.

Heparin-acrylic beads were soaked in 1 mg/ml FGF-4 as previously described (Niswander et al., 1993).

To prepare Shh beads, a 3 μl drop of Shh protein solution (see text for concentrations) was placed on a bacteriological grade Petri dish. Stock solutions were stored at 4°C and dilutions made in Tris chloride/sodium chloride buffer. 10-15 Affigel CM beads (200-250 μm in diameter) were rinsed in Tris chloride/sodium chloride buffer and then transferred to the 3 μl drop and allowed to soak in Shh for between 1 and 2 hours at room temperature.

A DNA fragment containing mouse Shh-coding sequences (amino acids 25-198) (Echelard et al., 1993) was cloned into the expression vector pEt11d 25 so that the Shh coding sequences were downstream of sequences encoding six histidine residues and a recognition sequence for the restriction protease thrombin. In these experiments, the hexahistidine sequences were not removed. The protein was induced and purified as previously described (Marti et al., 1995a). Following purification, Shh protein was extensively dialyzed against storage buffer (20 mM Tris-HCl (pH 7.9), 250 mM NaCl, 1 mM DTT, 5% (v/v) glycerol), concentrated and stored at −80°C.

COS cell lines that stably express Shh were obtained by transfection with pcDNA1-Shh followed by Neo selection. Grafts of mixtures of Shh-expressing cells and chick embryo fibroblasts (CEF) were prepared from confluent cultures. CEF or COS cells from one 100 mm dish were trypsinized and resuspended in 10 ml M199 medium supplemented with 9% fetal calf serum and 2% chicken serum. The cells were counted and Shh-expressing COS cells were mixed in known proportions with CEF cells. The cells were then centrifuged at 1000 revs/minute for 5 minutes and resuspended in 0.1 ml of medium (around 5×10⁷/ml). A 30 μl drop of resuspended cells was placed in a 35 mm Petri dish which was flipped upside down to form a hanging drop and then incubated for about 3 hours at 37°C. The cell aggregates that formed were cut into small pieces (100-200 μm in diameter containing about 3×10⁴ cells) and grafted to chick limb buds.

Fertile White Leghorn chicken eggs were purchased from SPAFAS, Inc (Norwich, CT) or from Poyndon egg farm, Hertfordshire, UK. Staging of chick embryos was according to Hamburger and Hamilton (1951). Shh beads or cell pellets were placed under a loop of apical ectodermal ridge at the anterior margin of wing buds of stage 19/20 embryos and where necessary secured with a platinum staple. In a small series of experiments, Shh beads were placed at the apex or at the posterior margin of stage 19/20 chick wing buds. In another series of experiments Shh beads were implanted at the anterior margin of stage 20 wing buds and, at intervals between 12 hours and 24 hours after implantation, beads were removed with a sharp needle, embryos were left to continue to develop to 10 days and skeletal patterns were assessed as described (Yang and Niswander, 1995). Wing buds were also fixed at the same time intervals in 4% paraformaldehyde (PFA) for whole-mount in situ hybridization to detect Fgf-4, Hoxd-13 and Bmp-2 expression. In a series of experiments to test for induction of polarizing activity, anterior mesenchyme adjacent to implanted Shh beads (beads soaked in 16 mg/ml Shh) was removed either 16 or 24 hours after bead implantation and grafted to a new host under the anterior apical ectodermal ridge region of the wing bud. Control polarizing region grafts from manipulated or contralateral wings of the embryos were performed as previously described (Tickle, 1981). A few embryos were also left to grow up to 10 days with implanted Shh beads in place and out of a total of 11 embryos more than half had wing duplications with additional digits 4 and 3 (Table 2B).

In situ hybridization was performed as previously described (Wilkinson, 1992; Nieto et al., 1996).

DiI and DiA injections were performed through glass micropipettes using a pressure injector (Picospritzer, General Valve Company). A dilution of 3 mg/ml in dimethyl formamide was used for each dye and a spot of dye was injected (diameter of 20-25 μm) to mark cells at different distances from the implanted Shh beads. Shh beads were soaked in 8 mg/ml or 16 mg/ml Shh and grafted at the anterior margin of the wing bud as above. Injections were performed 1 hour after bead implantation and distances were measured with an eyepiece graticule and with reference to the somites. Most embryos were fixed 96 hours later (while some were fixed immediately after the injections or after 36 hours) in 4% PFA. Photography was performed using a Nikon microscope with rhodamine filter while all bud widths were measured under a dissecting microscope using an eyepiece graticule.

Clones encoding Shh-N or full-length Shh lacking the signal peptide (amino acids 1-24) were cloned into the expression vector AP-tag4 (Cheng et al. 1995) to generate two constructs (AP::Shh-N and AP::Shh) which express either alkaline phosphatase (AP)-tagged Shh-N or Shh. In both cases, Shh-N or Shh was fused in frame to the C terminus of AP in AP-tag4. To generate a membrane-tethered Shh-N, cDNA sequences encoding Shh-N were fused with those encoding the transmembrane protein CD4 in the expression vector pcDNA3 (Invit-rogen). This fusion protein (Shh-N::CD4) appends the CD4 sequences to the signalling moiety of Shh.

To produce AP-tagged Shh proteins or membrane-tethered Shh-N, AP::Shh-N, AP::Shh or Shh-N::CD4 were transiently transfected into COS cells with Lipofectamine (Life Technologies, Inc) according to manufacturer’s instructions. The TCA-precipitated medium and cell lysate were analyzed by SDS-PAGE and immunoblotted with anti-Shh antibodies as previously described (Bumcrot et al., 1995). Suramin treatment was performed as described (Bumcrot et al., 1995).

COS cells transfected with constructs expressing AP-tagged Shh-N or Shh were harvested 48 hours post-transfection, mixed with 10% CEF cells and allowed to aggregate for 12 hours (Yang and Niswander, 1995). Cell pellets about 150 μm in diameter were implanted under the anterior apical ectodermal ridge of stage 20 chick wing buds. Operated chick embryos were incubated for 24 hours. Diffusion of AP-tagged Shh proteins was detected by whole-mount AP staining as described (Cheng and Flanagan, 1994).

To test whether the number of Shh-expressing cells governs the A-P identity of digit duplications, pellets of COS cells expressing full-length Shh were diluted with different proportions of primary chick embryo fibroblasts and grafted to the anterior margin of stage 19/20 wing buds (Table 1A; Fig. 1A). With 95% Shh-expressing COS cells, an additional digit 4 was specified in most cases (digit patterns 43234, 4334) (Fig. 1Aa); with between 20% and 80% Shh-expressing cells, a digit 3 is formed most often (Fig. 1Ab,c); whereas with 10% only an additional digit 2 was formed (Fig. 1Ad). Thus the number of Shh-expressing cells grafted determines the identity of the additional digits that are specified.

To directly test the dose effects of Shh, beads soaked in different concentrations of a bacterially expressed aminoterminal peptide of Shh (Shh-N) were implanted beneath the anterior apical ectodermal ridge of stage 19/20 chick wing buds (Table 1B). With beads soaked in high concentrations of Shh (16 mg/ml, 14 mg/ml, 8 mg/ml or 5 mg/ml), most wings had an additional digit 4 (Fig. 1Ba). With beads soaked in 2 mg/ml or 1 mg/ml Shh, an additional digit 3 was the most frequently observed digit duplication and no additional digit 4s were produced (Fig. 1Bb). When beads soaked in 0.75 mg/ml Shh were implanted, however, the most frequently observed additional digits obtained were digit 2s (Fig. 1Bc). Although beads soaked in lower concentrations of Shh occasionally gave an extra digit 2, most wings had a normal digit pattern (Fig. 1Bd). These results show that low concentrations of Shh can lead to the formation of the anterior digit, digit 2, and its induction can occur without inducing more posterior digits.

When a bead soaked in 16 mg/ml is removed 24 hours after implantation and placed at the anterior margin of a wing bud of a new host, additional 2s (two cases) or an additional digit 3 (one case with pattern 3234) can be specified in that host. This indicates that Shh protein in the grafted bead is still active after being incubated in chick embryos for 24 hours and that the absence of digit 2 in wings treated with the highest Shh concentrations are likely to be due to high levels of Shh throughout the bud.

To test how cells in different parts of a wing bud respond to Shh protein, Shh beads were also implanted either apically or posteriorly. When beads soaked in either 1 mg/ml or 16 mg/ml Shh (concentrations that usually give additional digit 3s or 4s, respectively, when placed anteriorly) were grafted at the apex of wing buds, an extra digit 3 could be formed in the region where the bead was placed, resulting in a 2334 pattern (3/3 cases, 16 mg/ml Shh; 3/6 cases, 1 mg/ml Shh). When beads soaked in high concentrations of Shh were placed posteriorly, most wings had a normal digit pattern (5/5 cases, 16 mg/ml Shh; 3/4 cases, 1 mg/ml Shh). Thus Shh only induces additional digits to form when the bead was grafted outside of the resident ZPA.

In almost all wings treated with Shh, proximal elements were thick and short. In 8% of normal limbs (i.e. limbs with no digit duplications) that developed after Shh application anteriorly (Fig. 1Bd) and in limbs with Shh beads implanted posteriorly, the shape of the humerus and occasionally the radius/ulna were distorted. One possibility is that Shh could mimic Indian hedgehog (Ihh), another vertebrate hedgehog gene that is expressed in early cartilage (Bitgood and McMahon, 1995; Vortkamp et al., 1996), and thereby influence cartilage morphogenesis (Vortkamp et al., 1996).

Digit specification depends not only on signal strength but also on length of exposure to the signal. When grafts of the ZPA are removed after 16 hours, only an additional digit 2 is specified. However, when grafts are left in place for longer, posterior digits are also formed (Smith, 1980). In order to find out whether length of exposure to Shh affects digit specification, beads soaked in 16 mg/ml Shh were implanted at the anterior margin of wing buds and then removed at different times (Table 2A). When beads were removed at 12 hours and 14 hours after implantation, no additional digits were formed. In contrast, when beads were removed at 16 hours, an additional digit 2 developed. Progressively longer periods resulted in more extensive duplications so that after a 24 hour treatment, an additional digit 4 was formed. Thus there appears to be two phases to digit specification; a phase lasting around 14 hours in which no irreversible changes in digit pattern are induced and a second promotion phase, about 8 hours long, in which additional digits are progressively specified. A high concentration of Shh on beads (16 mg/ml) applied for a shorter time (16 hours) can give the same results as low concentration Shh beads (0.75 mg/ml) applied for a longer time. However, when exposure of the wing bud to low Shh beads was prolonged, by replacing a low Shh bead at 16 hours with a second bead freshly soaked in low Shh, this failed to reproduce the effect of a higher concentration Shh bead. In only 1 out of 11 cases, was a digit 3 formed and the pattern of remaining limbs was either 2234 or normal (Table 1B).

In neural tube specification, application of Shh protein can itself induce Shh expression (Marti et al., 1995a). When we examined Shh expression in limb buds treated with Shh beads, no detectable Shh expression was induced in anterior mesenchymal cells near Shh beads (Fig. 1Ca,b; Table 4). Thus we can exclude an autoinductive Shh relay as a patterning mechanism.

In order to determine which cells in the limb form extra digits in response to Shh beads, DiI and DiA injections were performed to mark mesenchymal cells at different distances from the bead. Beads were soaked in either 8 mg/ml or 16 mg/ml of Shh protein and the injections were performed 1 hour later. Cells under the apical ridge near the posterior edge of the bead were marked 90 μm away from the bead and then at intervals up to 480 μm away from the bead. In some wing buds, just a single mark of DiI was applied; in other cases, two spots, one of DiI and one of DiA, were applied at different distances from the bead in the same bud.

Fig. 2A shows an example of a wing bud with an Shh bead injected at time 0. Two spots of dye which are about 20-25 μm in diameter can be seen in the bud. Fig. 2B shows a similar bud about 36 hours later; marked cell populations have given rise to two thick stripes of labelled cells extending from the apical ridge. Table 3 shows the position in which cells initially marked near the bead, end up relative to the anterior margin of duplicated wing buds at 96 hours and indicates the proliferation that has occurred to generate the new digits. At this time, the condensing cartilage of the digits can be seen and contribution of labelled cells to the extra digits assessed (Fig. 2C-F).

Most of the duplicated wings had a pattern of 4334 as expected, while others had extra 3s (pattern 334 and 3334). As shown in the summary diagram at the top of Fig. 2, an additional digit 4 came from cells initially within a 90 μm radius of the bead edge, while the additional digit 3 came from cells initially between 90 μm and 300 μm away; the normal digit 3 came from cells that lie more than 480 μm from the bead (Fig. 2E,F). In contrast, in control wing buds in which buffer-soaked beads were implanted, cells initially marked between 90 and 195 μm from the beads did not contribute to distal mesenchyme, while cells initially marked between 300 and 430 μm contribute to digit 2 (Fig. 2C; see also Table 3). Therefore anterior cells not normally destined to form digits can be induced to give rise to digits following implantation of a Shh bead.

To find out whether posterior digits arise by promotion from cells specified first as anterior digits, beads soaked in 16 mg/ml Shh were implanted, injections were performed as before but beads were removed 16 hours after implantation, which gives the digit pattern 2234. In these limb buds, cells labelled between 90 and 250 μm away from the bead contribute to the extra digit 2 while labelled cells between 300 and 430 μm contribute to the normal digit 2 (Fig. 2D). Thus the same cells initially (at a mean distance of about 130 μm from the bead) can give rise to either a digit 2 or a more posterior digit according to the length of time that the Shh bead has been left in place. However, it should be noted that the most posterior digit (the anterior digit 4 in a 4334 duplication) does not appear to arise by promotion and came from cells initially within 90 μm of the bead (i.e. less than 9-cell diameters away). Cells marked at 90 μm end up nearly twice as far away from the anterior margin when the Shh bead is left in place compared with when the bead is removed at 16 hours and this is related to the formation of an extra digit 4 (Table 3).

In a few cases, cells were marked either anterior or proximal to the bead, not immediately adjacent to the apical ridge. These labelled cell populations remained close to the bead and, although they expanded a little, they remained fairly rounded and did not end up in digits.

Previous experiments using existing antibody reagents have not detected endogenous Shh protein outside the polarizing region where it is processed (Marti et al., 1995b) but this technique may not be sufficiently sensitive. To develop a more sensitive assay for localizing Shh in limb buds, two constructs were generated which were tagged identically by the addition of alkaline phosphatase at the N terminus of the secreted Shh (Fig. 3). When alkaline phosphatase (AP) is attached to full-length Shh, this generates AP-tagged Shh-N* which is modified so that it is associated with cell membrane presumably due to addition of cholesterol (Porter et al., 1996). In addition, we engineered a second form that produces an unmodified AP-tagged Shh-N. Both tagged forms are processed equally well in producing cells in culture but the latter is secreted whilst the former is on the surface of the cells (Fig. 4). Activities of the tagged fusion proteins produced from Shh-N* and from Shh-N were determined by examining resulting digit patterns after grafting pellets of COS cells producing the two forms of Shh. In both cases, digit 2 was duplicated (n>5). No difference in polarizing activity could be detected between the two tagged forms of Shh although tagging the protein with alkaline phosphatase does attenuate polarizing activity because cells transfected with nonAP-tagged Shh-N or Shh induce in most cases digits that are more posterior in character.

Despite the fact that cells producing AP::Shh-N* and AP::Shh-N have similar polarizing activities, AP::Shh-N* behaves differently in the limb compared to AP::Shh-N. No significant diffusion of AP::Shh-N* could be detected by directly monitoring alkaline phosphatase activity while AP::Shh-N diffused extensively across the limb within 20 hours (Fig. 5A,D). Thus we can detect diffusion of an AP-tagged form of Shh, suggesting that long-range diffusion gradients of large peptides are at least a possibility in limb development. However, normal modification of Shh-N* is likely to prevent longrange diffusion. Furthermore, by whole-mount antibody staining, we found that Shh-N coated on the bead diffused in a broad region in the limb mesenchyme (Fig. 6B) whereas Shh-N* produced by COS cells expressing full-length Shh does not appear to diffuse (Fig. 6A). The expression of Ptc in both cases correlates with that of Shh. Induction of Ptc expression by Shh-expressing cells is localized to regions immediately adjacent to the cell graft (Fig. 6D) while the domain of Ptc expression induced by Shh-N released from the bead is fairly broad and symmetric to that of the endogenous Ptc (Fig. 6E). Nevertheless, in both cases, digit 3 and 4 were duplicated. These results taken together suggest that normal processing of Shh leads to a non-diffusible protein. One possibility is that diffusion of Shh is not necessary to polarize the limb because polarizing activity of both tethered and non-tethered Shh appears to be essentially the same.

To rule out the possibility that the broad diffusion observed is a result of cell migration from the graft, COS cells were labelled with DiI before being implanted into the chick limb bud to allow us to detect their final destination. Twenty hours after grafting, embryos were harvested, stained with AP substrate and sectioned through the limb buds. A striking difference in distribution of AP::Shh-N and modified AP::Shh-N* was observed. The distribution of AP::Shh-N as detected by AP staining extends far beyond that of DiI staining which marks the position of the grafted cells (Fig. 5B,C). In contrast, DiI and AP staining colocalises within the grafted cells that express AP::Shh-N* (Fig. 5E,F). These results indicate that cell migration from the graft did not occur and cannot account for the observed broad diffusion. They also confirm that tagged Shh-N* processed by a cell and placed into the environment of the limb bud does not diffuse to any appreciable level.

Though we have shown that Shh does not appear to diffuse in vivo, it is possible that an extremely low level of diffusion is sufficient to account for its activity at a distance. In this case, we reasoned that if Shh became completely membrane-tethered, its long-range activity should be abolished. To test this, we fused Shh-N to the integral membrane protein CD4 (Shh-N::CD4) in an attempt to generate a membrane-anchored Shh-N (Fig. 7). We first demonstrated that Shh-N::CD4 expressed in COS cells is completely membrane-tethered. Shh-N::CD4 is detected only in the cell lysate from cells transfected with Shh-N::CD4 and was not found in the medium even in the presence of suramin, indicating that it is membrane-tethered (Fig. 8). The activity of Shh-N::CD4 was then determined by examining resulting digit patterns after grafting pellets of COS cells as described above. We found that cell aggregates with 90-80% transfected COS cells were able to induce mirror-image digit duplication in a pattern 4334 (n=3, Fig. 9A,B). These data demonstrated that Shh-N when membrane-anchored is able to induce the most posterior digit, digit 4, an indication that it is fully active. Furthermore, this localized Shh-N signal is able to specify ectopic digits with at least two different positional values i.e., 4334 (n=3, Fig. 9A,B) or 32234 (n=2, Fig. 9C). In addition, when the percentage of transfected COS cells was reduced to 20%, the resulted digit pattern became 2234 in most cases (n=3, Fig. 9D,E). Occasionally, a digit pattern 3234 (n=1, Fig. 9F) was observed. This doseresponse relationship is similar to that observed using the wild-type Shh. Taken together, these results indicate that Shh-N when completely membrane-tethered exhibits a similar activity to that of wild-type Shh and suggest that Shh acts short range.

If Shh is not a direct long-range signal, how may the dose-dependent properties of Shh in limb patterning be explained? Two possibilities are that other signalling molecule(s) may be activated in limb bud cells in response to Shh or that cooperation between Shh and other signalling molecules could lead to dose-dependent digit specification and long-range effects on cells. Fgf-4 was induced in anterior apical ridge following grafts containing Shh cells and appears to be independent of cell number (data not shown). In contrast, activation of Bmp-7 in anterior mesenchyme was dose-dependent (Table 4; Fig. 10); an extensive domain of Bmp-7 expression was induced with high numbers (95%) of Shh-expressing cells (Fig. 10b). With fewer Shh-expressing cells (20%), activation of Bmp-7 in anterior mesenchyme was only weakly detectable (Fig. 10d).

In contrast, expression of Bmp-7 in anterior apical ectodermal ridge was very strong in all wings grafted with Shh-expressing cells (Fig. 10d). Hox-d13 and Hox-d11 gene expression also displayed a dose-response to Shh (Fig. 10f-l).

Changes in gene expression (Fig. 11A-C) occur at the start of the promotion phase suggesting that expression of at least some of these genes may be involved in digit specification (Table 4). By 16 hours after application of Shh protein, the normal domains of expression of Fgf-4, Bmp-2 and Hoxd-13 were all extended anteriorly and ectopic expression was clearly seen at 20-24 hours (Fig. 11D-F).

To explore whether expression of these genes or other as yet unidentified signalling molecules operate downstream of Shh as polarizing signals, we investigated whether anterior tissue next to a Shh bead acquires polarizing activity. Either 16-17 hours or 20-24 hours after implantation of the bead, we removed mesenchyme next to the Shh-soaked bead and grafted it to the anterior margin of a wing bud (stage 20). In 19 out of 21 cases, the grafted wings were normal, whereas 2 had a small cartilaginous fragment (blip) (Table 2B). In contrast, 50% of grafts of the polarizing region from manipulated or contralateral limb buds gave duplications with additional posterior digits (Table 2B). Therefore although Shh-soaked beads induced duplications, adjacent cells have no detectable polarizing activity when assayed 16-24 hours after bead implantation.

Our main finding is that Shh can produce dose-dependent changes in digit pattern. Thus the primary determinant of antero-posterior pattern in the limb is likely to be the concentration of Shh. Cells from a maximum distance of about 300 μm from a Shh source can be recruited to form additional digits but Shh processed by cells does not appear to be able to diffuse this far. Shh-N released from beads and Shh processed by COS cells exhibit dramatic difference in diffusibility in the limb mesenchyme. This correlates with the much broader domain of Ptc expression induced by the widely diffused Shh-N as compared to that induced by the apparently non-diffusible Shh-N*. Ptc has been suggested to be an indicator of exposure of cells to Hh signalling in both Drosophila and vertebrates (Perrimon, 1995; Goodrich et al., 1996). Therefore, since polarizing activities of diffusible and non-diffusible forms of Shh are similar, the critical action of Shh appears to be short range. Consistent with this, a membrane-tethered Shh exhibits similar activity to that of wild-type Shh in digit induction although we cannot rule out the possibility of some uncharacterized and undetected proteolytic cleavage of membrane-tethered Shh in vivo. Short-range signalling by Shh could still directly specify each digit in a dose-dependent fashion if production of Shh was propagated by a relay mechanism with progressive attenuation. However, this possibility is excluded because application of Shh does not induce Shh expression. Therefore, we conclude that it is unlikely that endogenous Shh produced by the polarizing region in the chick limb bud directly governs the behavior of cells at different distances from the polarizing region but instead it acts short range.

If the Shh signal does not directly act on cells to determine which digit they will form, there must be other downstream signal(s) activated by Shh that act in a dose-dependent manner. We showed that Bmps are activated in a dose-dependent fashion in response to Shh-expressing cells. However, it has not been possible to duplicate the effects of Shh signalling on digit pattern by application of BMP-2 protein (Francis et al., 1994) and grafts of cells expressing Bmp-2 also have only weak polarizing activity (Francis-West et al., 1995; Duprez et al., 1996). Furthermore, we have not been able to detect polarizing activity in cells that have been adjacent to an Shh bead. These observations suggest that Shh may activate more than one secondary signal and that continual Shh input may be required to produce secondary signal(s).

An important feature of Shh signalling is its time dependency. Irreversible digit pattern changes are not brought about until 16 hours after Shh application. Digit 2 is specified first and then progressively more posterior digits are specified over the next 8 hours or so. Since the mean number of digits remains almost constant but the character of additional digits changes with time, this suggests that cells, which would have initially formed anterior digits, are promoted with continued exposure to form more posterior digits (see also model by Eichele et al., 1985). Our data with DiI- and DiA-labelled cell populations indicate that this indeed appears to be the case; cells that will participate in forming a more posterior digit when the Shh bead is left in place, form digit 2 when the bead is removed early. In the promotion phase, cells may respond according to the highest concentration of signalling molecule(s) that they see. Another possibility is that cells integrate over time the strength of a Shh signal or a Shh-dependent signal to which they have been exposed. In this model, cells in the limb would not sense morphogen concentration at a particular time but instead continuously interpret the morphogen signal. Precedence for this comes from recent observations of positional signalling in frog embryos (Gurdon et al., 1995).

Specification of a series of digits across the anteroposterior pattern of the limb appears to involve a cascade of signals generated at the posterior margin of the limb bud. Retinoic acid can activate Shh expression in a dose-dependent fashion (Yang and Niswander, 1995) and we have shown that Shh signalling in turn is dose-dependent. Blocking of retinoic acid synthesis at pre-limb bud stages disrupts limb initiation (Stratford et al., 1996; Helms et al., 1996), while inactivation of Shh does not prevent limb initiation but instead results in limb truncation (Chiang et al., 1996). This suggests that retinoic acid may function in early polarizing signalling and Shh may signal later. Our bead removal experiments also indicate that continuous Shh signalling does not appear to be essential for patterning the most distal parts of the skeleton since these are not specified until at least 24 hours after bead removal (Summerbell, 1974). The idea that Shh is not required continuously is consistent with the observation that, when Shh expression is reduced in chick limb buds as a consequence of inhibiting retinoic acid synthesis at the bud stage (Stratford et al., 1996), digit development is normal. Also, some digit development still occurs in the ld mutant which has weak and transient Shh expression (Haramis et al., 1995; Kuhlman and Niswander, 1997).

We have shown here that Bmps are activated in a dosedependent manner by Shh at the time that digit promotion occurs. It is possible that establishment of a morphogen gradient occurs during this short phase of digit promotion rather than during the first priming phase. Thus, it is possible that secondary signals such as BMPs act as dose-dependent signals to promote digit specification once the limb bud has been primed in some way by a short-range Shh signal. It should be noted that the cellmarking experiments show that the most posterior digits come from cells very close to the bead. Diffusion of BMPs could perhaps extend the range of a polarizing signal and perhaps even specify the most anterior digit in the absence of Shh priming as has been observed by grafting Bmp-2-expressing cells to anterior mesenchyme (Duprez et al., 1996). Since the potential ability of BMP to specify posterior digits may depend on prior Shh priming and cells primed by Shh may produce several different BMPs and signalling molecules other than BMPs, this could explain, for example, why BMP-2 protein (or Bmp-2 expression) alone cannot provide a complete polarizing signal.

Finally, it should be noted that it has proved very difficult in a number of systems to distinguish between long-range and short-range signalling. Although in chick spinal cord development, it now seems that floor plate and motor neuron development involves both long-range and short-range Shh signalling (Ericson et al, 1996). In insect wing development, direct longrange signalling has been demonstrated by expression of mutated receptors that block formation of structures at a distance from the signalling source (Lecuit et al., 1996), while there is also evidence for simultaneous short-range signalling (Nellen et al., 1996). In Xenopus, different experiments have implicated both long-range and short-range signalling in mesoderm specification (Reilly and Melton, 1996; Jones et al., 1996). Specification of limb digits also appears to be just as complex and to involve multiple signals and there may also be different downstream targets that differ in their requirements for activation.

We thank Hwai-Jong Cheng and J. Flanagan for the generous gift of the AP-tag4 vector. Work in C. T.’s laboratory was carried out with equipment purchased by a donation from Astor Foundation. G. Drossopoulou and N. Vargesson were supported by the Medical Research Council and G. Drossopoulou also by an Anatomical Society Research Studentship, D.Duprez by the BBSRC. Work in A. P. M.’s laboratory was supported by a grant (NS33642) from the NIH. Y. Yang is supported by a postdoctoral fellowship from the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation, P.-T. Chuang by one from the Leukemia Society of America, D. Bumcrot by one from the Molecular Dystrophy Research Association and E. Marti by a NIH Fogarty Fellowship. Work in L. N. laboratory was supported by MSKCC support grant. We thank Martin Cohn for help with Shh whole mounts, Anne Crawley for help with photography and Professor Lewis Wolpert for his comments on the manuscript.