The Limb Bud Shh-Fgf Feedback Loop Is Terminated by Expansion of Former ZPA Cells

See allHide authors and affiliations

Science  16 Jul 2004:
Vol. 305, Issue 5682, pp. 396-399
DOI: 10.1126/science.1096966


Vertebrate limb outgrowth is driven by a positive feedback loop involving Sonic Hedgehog (Shh), Gremlin, and Fgf4. By overexpressing individual components of the loop at a time after these genes are normally down-regulated in chicken embryos, we found that Shh no longer maintains Gremlin in the posterior limb. Shh-expressing cells and their descendants cannot express Gremlin. The proliferation of these descendants forms a barrier separating the Shh signal from Gremlin-expressing cells, which breaks down the Shh-Fgf4 loop and thereby affects limb size and provides a mechanism explaining regulative properties of the limb bud.

Many key developmental decisions are made in response to secreted factors emanating from defined organizers or signaling centers. Whereas much has been learned about how signaling centers form, much less attention has been given to an equally important question: What is the mechanism that terminates their activity, assuring that the potent factors they produce do not interfere with subsequent development once they have fulfilled their function?

Two classic signaling centers are found in the limb bud. The distal outgrowth of the limb bud requires the activity of several members of the fibroblast growth factor (Fgf) family that are synthesized in a signaling center at the tip of the limb bud known as the apical ectodermal ridge (AER) (1, 2). Fgf signaling also serves to maintain the expression of another critical secreted factor in the limb bud, Sonic hedgehog (Shh) which is produced by a second signaling center, the zone of polarizing activity (ZPA) in the posterior mesenchyme (1, 3). Shh is responsible for patterning the anterior-posterior axis of the limb (4) and also acts to maintain the expression of several Fgf's in the overlying AER, including Fgf4 (1, 3), Fgf9, and Fgf17 (5). Shh maintains Fgf4 (and presumably Fgf9 and Fgf17) by up-regulating Gremlin in the adjacent mesenchyme (6, 7). Gremlin is a bone morphogenetic protein (Bmp) antagonist (8) that prevents Bmp's from down-regulating Fgf4 (9, 10). Importantly, the fourth member of the Fgf family expressed in the AER, Fgf8, is not directly dependent on Shh for its transcription (11). However, in the absence of Gremlin, the AER itself becomes disorganized and Fgf8 is down-regulated (12). Thus, the signaling centers in the posterior and distal tip of the limb bud are interdependent for their activity, and this positive feedback loop is required for producing a normal limb structure.

The signaling loop between Shh and Fgf's operates throughout limb development until embryonic day 6 (E6) (stage 27) (13), when Fgf4 and Gremlin cease to be expressed and Shh is down-regulated [Shh expression ceases at E7.5 (stage 29)] (4). Concomitantly, the rate of cell proliferation in the limb decreases (14). As critical as the feedback loop is to patterning, it is equally critical that it be terminated. This is highlighted by the effect of ectopically providing Shh late in limb development after the ZPA no longer exists. Fgf8 expression is transiently maintained at the distal tips of each digit by factors secreted from the forming skeletal elements (15). If Shh is ectopically provided at this stage, Fgf8 expression is prolonged (15), the digit rays continue to grow, additional phalanges are produced, and the limb becomes longer than normal (15, 16). Thus, the breakdown of the Shh-Fgf feedback loop is critical to control the size of the limb. To try to understand how this breakdown occurs, we overexpressed Shh, Gremlin, and Fgf4 before E6 (stage 27) and asked if any of them maintain the other genes in the loop after their normal down-regulation. A bead soaked in Fgf4 implanted in the posterior mesenchyme of E5 (stage 25) chick limb buds maintained Shh up-regulation for 48 hours until E7 (stage 28) (16 out of 27 limbs), when expression of Shh in contralateral controls was no longer detectable (Fig. 1, A and B), suggesting that a loss of responsiveness to Fgf4 is not the mechanism by which the feedback loop breaks down. Similarly, viral misexpression of Gremlin with an RCAS virus injected at E4 (stage 23) leads to a continued up-regulation of Fgf4 (8 out of 26 limbs) and Shh (4 out of 10 limbs) 72 hours later at E7 (stage 28) (Fig. 1, C to F), indicating that Gremlin responsiveness is intact as the feedback loop degenerates. In contrast, a bead soaked in Shh and implanted in the posterior limb at E5 (stage 25) cannot maintain Gremlin (0 out of 12 limbs) or Fgf4 (0 out of 10 limbs) expression 24 hours later (Fig. 1, G to J) This suggests that a loss in the ability of mesenchymal cells to respond to Shh by up-regulating Gremlin is the mechanism by which the Shh-Fgf feedback loop breaks down.

Fig. 1.

The Shh-Fgf feedback loop is disrupted between Shh and Gremlin. (A and B) An Fgf4 bead (A) maintains Shh expression in chick limbs at E7 (stage 28), unlike controls (B). (C to F) Ectopic Gremlin viral expression maintains both Fgf4 (C) and Shh (E) expression at E7 (stage 28) in chick limbs at higher levels than controls [(D) and (F)]. (G to J) A Shh bead implanted into the posterior chick limb is unable to maintain either Gremlin (G) or Fgf4 (I) expression at E7 (stage 28) as in controls [(H) and (J)]. (K) A Shh bead maintains ectopic Gremlin expression in the anterior at E7 (stage 28). The diagram on the bottom right shows the step of the feedback loop being tested in each experiment.

In control experiments, a Shh bead was implanted into the anterior limb at the same stage, and it ectopically maintained Gremlin expression (9 out of 16 limbs) (Fig. 1K). Thus, although Shh responsiveness is lost in the posterior limb, it is maintained in the anterior limb. This is less surprising in the context of previous studies showing that there is a zone in which Gremlin expression is excluded in the posterior limb (7, 17). From E4 to E5.5 (stages 23 to 26), this zone of exclusion increasingly expands, such that a gap opens between the domains of Shh and Gremlin expression (Fig. 2, A and B). It is the cells within this domain that are unable to express Gremlin in response to ectopic Shh.

Fig. 2.

Shh descendants are unable to express Gremlin. (A to C) Double in situ hybridizations for Gremlin (purple) and Shh (brown). At E4 (stage 23) in the chick, the domains of Gremlin and Shh expression are close together (A), but by E5.5 (stage 26) a gap has opened between them (B). The situation is similar in the mouse at E11.5 (C). (D) Shh descendants detected by β-galactosidase staining of a Shh::CRE;R26R mouse at E11.5 closely resemble the domain of posterior Gremlin exclusion. (E to J) An in situ hybridization in mouse limbs for Gremlin [(E) and (F)], β-galactosidase staining for Shh descendants [(G) and (H)] and the merged images [(I) and (J)]. At E11.5, Gremlin expression is contiguous but nonoverlapping with Shh descendants [(E), (G), and (I)]. Non-Shh descendants that are mixed into the Shh descendant domain express Gremlin, showing Gremlin repression is cell autonomous [(F), (H), and (J)].

In a separate recombinase-based fate mapping study, we marked the descendants of cells that at one time expressed Shh in the ZPA (18, 19). Former Shh-expressing cells expand anteriorly to encompass a domain substantially larger than the ZPA, similar to the Gremlin-negative domain at a similar stage in the mouse (Fig. 2, C and D). To test whether the Gremlin-negative domain is indeed the population of cells that formerly expressed Shh, sections of such a mouse limb were stained both for β-galactosidase to visualize Shh descendants and by in situ hybridization to detect Gremlin expression at E11.5. As previously observed, the Gremlin-expressing cells were only found in superficial layers subjacent to the dorsal and ventral surface ectoderm (Fig. 2E) (12, 17). Gremlin-expressing cells lie contiguous to Shh descendants but do not overlap with them (Fig. 2, E, G, and I). The conclusion that the former Shh-expressing cells cannot express Gremlin is particularly convincing because in some sections, groups of non-Shh descendants are mixed within the zone of Shh descendants (Fig. 2H). These cells express Gremlin, whereas neighboring Shh descendants do not (Fig. 2, F, H, and J), suggesting that the inability to express Gremlin is a cell-autonomous property of Shh descendants and not a general response to a posteriorly localized factor. This cell autonomy was further confirmed in transplantation experiments (fig. S1).

These data suggest a model to explain the termination of the Shh-Fgf feedback loop during limb development. Cells that express Shh and their descendants are unable to express Gremlin. Initially, this is not a problem because Shh protein diffuses (20) and activates target genes such as Ptc1 (21) at a substantial distance from the ZPA. However, as limb outgrowth proceeds, the proliferation of Shh descendants serves as a barrier between the source of Shh and cells that are able to produce Gremlin in response. By E6 (stage 27), the cells competent to express Gremlin become too distant from the source of Shh to receive an adequate dose of the protein to produce Gremlin, leading to a breakdown of the Shh-Fgf feedback loop. Consistent with this, Fgf4 expression is first lost from the posterior end of the AER (fig. S2).

To obtain direct evidence that the down-regulation of the loop is caused by the barrier of Shh descendants between Gremlin and Shh expressing cells, we removed a wedge of tissue containing Shh descendants from a chick limb bud and stapled the remaining posterior cells (including the cells actively producing Shh) to the remaining anterior limb at E5 (stage 25). We reasoned that removing these cells might put the responsive tissue back within the range of Shh, thereby preserving the feedback loop. Indeed, by eliminating the tissue between Shh-expressing and Gremlin-competent cells, the feedback loop continues to function. Shh (three out of six limbs), Gremlin (seven out of nine limbs), and Fgf4 (5 out of 15 limbs) are each up-regulated when compared with the unoperated contralateral limb for at least 24 hours after the endogenous Shh-Gremlin-Fgf feedback loop normally breaks down (Fig. 3, A to H). Thus, in the absence of this surgical manipulation, the barrier formed by the expansion of the former Shh-expressing cells provides an innate stopping mechanism for the Shh-Fgf4 feedback loop.

Fig. 3.

The Shh-Fgf4 signaling loop is prolonged by the removal of Shh descendants, leading to size regulation. (A to H) By removing a wedge of Shh descendants in E5 (stage 25) chick limbs and stapling the ZPA still expressing Shh to the anterior cells, Shh (A), Gremlin (C), and Fgf4 [(E) and (G)] expression is maintained compared with control limbs, [(B), (D), (F), and (H)] showing that the expansion of Shh descendants blocks the Shh-Fgf4 signaling loop. (G) and (H) are higher magnifications of (E) and (F), respectively. (I to M) Limbs in which Shh descendants are removed and the ZPA stapled to anterior cells show size regulation (I) to become comparable in anterior-posterior and proximal-distal patterning and size to contralateral control limbs (J). Limbs were injected with two adjacent dots of DiI (arrows) (K), and then either the Shh descendants were removed and the remaining tissue stapled (L) or the DiI-injected limbs were left unoperated (M). The cells in between the dots expand in the operated limbs, showing that Shh-expressing cells continue to proliferate until the distance between Shh and Gremlin expression lies outside the range of Shh diffusion again, thereby giving size regulation. (N to P) Ectopic expression of Gremlin in the posterior limb bud to E10 (stage 36) with the use of a virus, detected with an antibody, 3C2, against viral MA antigen (arrow) (P), gives rise to elongated growth of the posterior tissue (N) when compared with contralateral controls (O). Length of the ectopic growths were measured from the tip of the ectopic soft tissue to the proximal end of digit 4, which still forms cartilage.

In these experiments, a wedge containing as much as a third of the distal limb mesenchyme is removed, and the resulting limbs are wild type both in structure and in size (Fig. 3, I and J). This is an example of regulative development, the general phenomenon that embryos are extremely resilient; in spite of considerable experimental manipulations, normal embryos can develop (22). Yet, how regulative growth is achieved has not received much attention. The model presented here provides a potential explanation: Once the ZPA is pinned to the remaining limb tissue, without a barrier to Gremlin induction, the feedback loop is restored and the posterior cells resume expansion. This continues until the barrier again reaches the critical width at which the loop breaks down. Because this width depends on the diffusion distance of Shh and not on the amount of tissue removed, the development is regulative, and hence the limbs grow to their normal extent. Consistent with this model, expansion of the posterior tissue is indeed restored following surgery, as can be seen in the gap which once again is established between the sites of Shh and Gremlin expression (compare Fig. 3, A and C). The renewed expansion of the posterior tissue was also directly verified with injections of 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) (Fig. 3, K to M).

The mechanism by which the Shh-Fgf feedback loop is terminated in the limb also provides insight into size control in limb outgrowth. Controlling body size and establishing the proper relative sizes of the constituent body parts and organs are critical aspects of embryonic development. In spite of its importance [with a few exceptions, such as the control of muscle size by myostatin (23)], little is known about the control of differential size in vertebrates (24). As the limb bud grows out, the digit rays split into individual phalangeal elements in a proximal-to-distal sequence based on an intrinsic developmental program, such that new cartilaginous elements are formed distally, with a predefined spacing, as long as the digit ray keeps growing. The length of time that this process continues determines the length of the digits. Thus, dolphin digit rays grow for a longer period and form more phalanges than the limbs of land mammals (25).

As noted earlier, digits continue to grow and additional phalanges form when Shh and Fgf's are exogenously maintained independent of the normal requirement for positive feedback (15, 16). Artificially maintaining Gremlin expression in the distal posterior limb also results in a continuation of Shh and Fgf4 expression (Fig. 1, C to F). As a consequence, Gremlin-infected posterior digit primordia are 13.4% longer than the contralateral digit 4 (Fig. 3, N to P; five out of seven limbs). In these infected limbs, no chondrogenesis occurs in the extra outgrowth because of the Bmp-antagonistic effects of Gremlin. The expectation would be that if Gremlin expression were prolonged but did not proceed indefinitely (because of the viral promoter in this experimental setting), condensation would proceed normally, and longer digits would be formed.

Why are the former Shh-expressing cells refractory to Gremlin induction? It is possible that this nonresponsiveness is imposed by transcription factors expressed in the ZPA cells and their descendants. This is consistent with our finding that cells that never expressed Shh but are adjacent to or mixed with former Shh-expressing cells are capable of activating Gremlin. It is also possible that extremely high levels of Shh signaling, such as those normally attained only in an intracellular autocrine fashion by cells producing Shh, render cells unable to express Gremlin. This latter possibility may be supported by consideration of the situation where ectopic Shh protein is applied to the anterior limb bud (26). Recognizable posterior digits (albeit not as perfect as with a ZPA graft) are formed in response to Shh protein beads, which may imply that there is an induction of a Gremlin-refractory cell population by exceptionally high levels of Shh signaling adjacent to the beads. Alternatively, the termination of growth of these limbs could reflect the limited amount of Shh protein in the bead. Another series of experiments seemingly relevant to the current study involves grafts of irradiated ZPA cells. However, these experiments need to be reevaluated with modern tools (supporting online material text).

We focused on Fgf4 as the AER component of the Shh-Gremlin-Fgf feedback loop. As noted above, Fgf8, Fgf9, and Fgf17 are also expressed in the AER. Fgf4, Fgf9, and Fgf17 have been shown to depend on Shh for their transcription (5), whereas Gremlin is indirectly responsible for maintaining Fgf8 expression by affecting AER organization (12). Hence, the loss of expression and/or down-regulation of all four of these genes is likely to result when the expansion of the posterior limb bud exceeds the ability of Shh to maintain Gremlin. This provides a unique mechanism for terminating the activity of two key signaling centers in the limb, with important implications for the little-understood properties of regulative development and size control during embryogenesis.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 and S2


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article