A fundamental biological question is how the body plan is laid down during embryonic development and how precise arrangements of specialized cells and tissues arise. The vertebrate limb has a complex anatomy and is an excellent model in which to address this question. Vertebrate limbs develop from small buds of apparently homogeneous unspecialized mesenchyme cells that are encased in the ectoderm. As the buds grow out from the body wall, these unspecialized cells begin to differentiate into various tissues of the limb — including the cartilage and, later in development, the bone — that make up the skeleton. The limb skeleton consists of a defined number of bones of characteristic size and shape that are arranged in a specific pattern. The anatomy of the limb can be described with respect to three orthogonal axes, proximo–distal, from shoulder to finger tips, antero–posterior, from thumb to little finger, and dorso–ventral, from the back of the hand to the palm (Fig. 1). The processes that regulate limb formation are highly coordinated; for example, each digit forms in a specific place. But how is this pattern of cell differentiation controlled?
Figure 1 | The three main axes of the human hand.
The diagram shows the three main axes, proximo–distal, antero–posterior and dorso–ventral, of a human hand. Tightly regulated processes during embryonic development ensure that the thumb arises at one edge of the hand, whereas the little finger arises at the other.
For the last 40 years or so, this problem and, in particular, how digit pattern arises, has been tackled by applying the concepts of positional information1, 2. According to these concepts, cells are first informed of their position in the limb bud and, as a result, they acquire a positional value that encodes this information. In a second step, these values are interpreted, leading to the formation of the appropriate structure at that position. Positional information across the antero–posterior limb axis is provided by signalling of the polarizing region at the posterior margin of the early limb bud. Classical experiments on chick embryos3, 4 (Box 1) led to the identification of this region and the proposal that the polarizing region produces a morphogen which diffuses, over time, into adjacent limb tissue to give a concentration gradient. Cells at different positions across the limb bud would be exposed to different concentrations of the morphogen — cells nearest the polarizing region, at the posterior of the limb, would be exposed to high concentrations of the morphogen, whereas cells further away, at the anterior of the limb bud, would be exposed to low concentrations. Therefore, the local concentration of the morphogen could provide information about position across the antero–posterior axis. These experiments also indicated that a ratchet-type mechanism is in operation, such that the anterior positional values can be promoted irreversibly to more posterior positional values, and the most posterior positional values are then remembered.
In the last 15 years, the molecular basis of limb development has begun to be unravelled through the identification of the signals that are generated in the polarizing region and the discovery of molecules that are produced in response to these signals. These molecular advances come mainly from genetic studies in mice (Mus musculus), although many originated in fruitflies (Drosophila melanogaster). Recent sophisticated analyses in mouse embryos, which link gene expression and cell fate in developing limbs, have stimulated the re-evaluation of the mechanisms of digit patterning and have highlighted the particular problem of providing positional information in a growing organ. In this article, I begin by outlining the embryological studies that revealed the signalling properties of the polarizing region and then discuss the molecular basis of polarizing signalling in the developing limb bud, with particular emphasis on the secreted molecule, sonic hedgehog (SHH). I conclude by discussing the outstanding challenges in identifying how antero–posterior values are encoded molecularly and how they are ultimately translated into digit anatomy.
Embryology of digit patterning
Digits form at relatively late stages of vertebrate limb development. By this time, the small bud has grown substantially and changed shape so that the main regions of the future limb can be made out (Box 2). However, experiments on chick embryos have shown that cells are specified to form limbs, long before any buds are visible, and that both the antero–posterior and the dorso–ventral polarity of the limb are already established (reviewed in Ref. 5). Fibroblast growth factors (FGFs) and Wnt-signalling molecules are involved in limb initiation and budding and control the limb-specific expression of genes that encode members of the Tbx family of transcription factors6.
An early event in bud formation is the development of the apical ectodermal ridge. The apical ridge forms at a compartment boundary between the dorsal ectoderm and the ventral ectoderm — this ensures that the limb buds form at the sides of the body7. Continued outgrowth of the bud depends on FGF signalling by the apical ridge8. Another important event in limb-bud development is the formation of the polarizing region that controls the antero–posterior pattern of distal structures. Once equipped with an apical ectodermal ridge and a polarizing region, the limb bud can develop autonomously.
Insights from chick embryos. The normal chick wing has three digits known as 2, 3 and 4 (Box 2), and it has been proposed that digit patterning involves signalling between the cells of the polarizing region and the adjacent limb-bud cells3. Grafting the polarizing region from one chick wing bud to the anterior (opposite) side of a second bud results in a dramatic change in digit pattern; six digits develop instead of three, with the extra set of digits in mirror-image symmetry with the normal set, giving rise to the pattern 4 3 2 2 3 4 (Box 1). Grafts of the posterior margin of mammalian limb buds to chick wing buds have also been shown to have polarizing activity (for example, see Ref. 9). The extra digits induced in this case are, nevertheless, chick digits, which indicates that, although the signal is the same, the interpretation differs.
But how does the polarizing region produce such a pattern? The results of many chick embryological experiments are consistent with the idea that the polarizing region produces a long-range morphogen that specifies antero–posterior positional values. Polarizing activity can be detected in the posterior region of the chick wing bud from early stages until the stage that the digits start to form10. Long-range signalling by the polarizing region operates over a few hundred 
m (about 10–30 cells; Box 1). Therefore, it is likely that positional values are specified in the early limb bud. Indeed fate maps show that digits arise from the posterior region of the early limb bud, which then expands to fill the digital plate11, 12. Fate maps do not provide information about commitment, but, because there is experimental evidence for positional memory (Box 1), the prevailing model has been that polarizing-region signalling sets up a morphogen gradient, which specifies antero–posterior positional values in the early bud (Fig. 2a). This set of initially tightly packed positional values then becomes distributed across the limb bud as the bud grows, and later dictates the development of each digit primordium.
Figure 2 | Models for specifying antero–posterior positional values in the chick wing and the mouse limb.
![Figure 2 : Models for specifying antero|[ndash]|posterior positional values in the chick wing and the mouse limb. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nrm/journal/v7/n1/thumbs/nrm1830-f2.jpg)
a | Morphogen-gradient model in the chick wing in which sonic hedgehog (SHH) concentration specifies antero–posterior positional values. The numbers represent the positional values for each chick wing digit. All positional values are specified in the early limb bud according to a morphogen gradient (pink shaded area) and the values are then remembered. The digit primordia develop approximately 24 hours after the positional values are fixed. The timeline illustrates the specification of positional values in the early limb bud from the onset of SHH expression on the basis of a time course of induction of extra digits by SHH beads21. Positional values are progressively promoted (between 16 and 24 hours) as the gradient of SHH becomes established. b | A new model for the specification of antero–posterior values in the mouse limb that involves both the concentration and the length of exposure to SHH35. The numbers represent positional values for each mouse digit. Shh is expressed at embryonic day (E)9.75. By E10.5, the SHH concentration gradient (shaded area) that was established in the early limb bud has led to the specification of the positional value of digit 2, and contributes to the specification of digit 3. The development of digit 1 is not dependent on SHH signalling. By E11.5, digit 3 is specified according to both the concentration and the length of exposure to SHH, but digits 4 and 5 are specified according to differences in the length of exposure to SHH. The timeline shows that, according to this model, posterior digits will be specified at later stages of development than anterior digits.
Polarizing-region grafts change digit number in addition to pattern. One of the consequences of grafting a polarizing region to the anterior margin of the wing bud is an increase in the width of the bud to accommodate the extra digits4. This is accompanied by an increase in length of the apical ectodermal ridge. It was postulated a long time ago that the polarizing region regulates production of an apical-ridge maintenance factor by the mesenchyme cells in the posterior region of the bud13. It has also been suggested that signalling through the polarizing region might have a direct effect on cell proliferation14, because changes in cell proliferation have been detected prior to changes in ridge length.
Morphogen gradient mechanism
The vitamin-A derivative, retinoic acid, was the first defined signalling molecule to be identified as having the ability to induce mirror-image duplications in the chick wing15 (Box 1). Although it was shown that retinoic acid is readily diffusible and functions in a concentration-dependent fashion15, the main role of endogenous retinoic acid in polarizing signalling is now thought to be the induction of the expression of the SHH gene16. It should be noted that retinoic acid also has other roles in limb-bud initiation17 and in patterning of the proximal part of the limb18.
Identification of candidate morphogens. SHH is expressed in the polarizing region of both the chick16 and the mouse limb buds, and immunohistochemical studies19 and a biological assay — based on the ability of SHH to induce differentiation in a cell line20 — indicate that the SHH protein diffuses some distance away into the limb bud. Beads soaked in SHH protein cause concentration-dependent changes in digit pattern and the timing of digit induction in the chick wing bud matches that seen with polarizing-region grafts. A lag of approximately 14 hours occurs before any extra digits form, and is followed by sequential formation of the 3 digits between 16 and 24 hours21. Promotion of positional values over time was demonstrated directly by tracing the behaviour of cells at known distances from a SHH bead with a lipophilic dye that can be applied to the membranes of cells to mark them. When the SHH bead is removed at 16 hours, cells at a distance of 130 m (about 13 cell diameters) from the bead participate in forming an extra digit 2. When SHH treatment is extended, cells at 130 m from the bead participate in forming more posterior digits. One interpretation of these data is that it takes about 24 hours to establish a stable SHH gradient across the anterior region of the chick wing bud. This provides an estimate of the time that is required to establish a SHH gradient in the posterior of the limb bud during normal development (Fig. 2a).
There is a positive-feedback loop involving FGF signalling from the apical ridge that maintains SHH expression in the polarizing region8 ; SHH, in turn, maintains expression of FGF genes, including FGF4, in the apical ridge (Box 3). In Shh-/- mouse embryos, the distal regions of the limbs are very reduced — at best, one digit develops22 — and this can be understood in terms of the failure to maintain the positive-feedback loop.
The role of GLI genes. Important components of the hedgehog (Hh) signalling pathway were first discovered through genetic studies in fruitflies, but it is now clear that these components have been largely conserved during evolution (reviewed in Ref. 23). The three vertebrate GLI proteins, GLI1, GLI2 and GLI3, are the transcriptional effectors of SHH signalling. In the absence of the SHH ligand, the GLI2 and GLI3 proteins are processed to short repressor forms that translocate to the nucleus and repress SHH target genes. By contrast, in the presence of the SHH ligand, full-length GLI proteins — in particular GLI1 and GLI2 — translocate to the nucleus where they function as activators to induce expression of SHH target genes, including the gene encoding GLI1.
Analysis of mouse embryos shows that Gli3 has an important role in limb patterning. Gli1-/- and Gli2-/- mouse embryos have normal limbs, but Gli3-/- mouse limbs are polydactylous and have approximately eight unpatterned digits24, 25. In the normal limb, low levels of the short repressor form of GLI3 (GLI3R) are present in the posterior region, from which the series of patterned digits develop, and high levels of GLI3R are present in the anterior region, which does not give rise to digits26 (Box 3). On the basis of these data, it has been suggested that the main role for SHH signalling is to relieve GLI3 repression in the posterior region of the limb bud. In the absence of SHH — for example, in Shh -/- embryos — the short GLI3R will prevail throughout the limb bud and therefore digit development will be severely reduced.
It is unclear whether the balance between the levels of the full-length activator form of GLI3 (GLI3A) and GLI3R forms the basis for a graded response to SHH concentration, and subsequently leads to the specification of antero–posterior positional values, or if just the levels of GLI3R in different regions of the limb bud are important. Other mouse mutants with many unpatterned digits, in which novel vertebrate components of the Hh signalling pathway (such as the intraflagellar transport proteins) are affected, also have defective GLI3 processing27, 28. The increase in digit number in all these mutants is due to widespread expression of the gremlin gene, which is regulated by GLI3 repression29, and is now known to encode a bone morphogenic protein (BMP) antagonist — the apical-ridge maintenance factor (Box 3).
The series of many unpatterned digits, which develop in the absence of GLI3 function in mouse mutants, is reminiscent of the digit pre-pattern that has been proposed to function in combination with the morphogen gradient2. According to this proposal, a series of digit condensations, a pre-pattern, is specified by a wave-like distribution of a morphogen that is generated by a reaction–diffusion mechanism, with the peaks corresponding to the condensations. A gradient of another morphogen — for which SHH is a good candidate — then provides each peak with a positional value and a digit identity. The number of peaks that are generated by the reaction–diffusion mechanism depends on the width of the limb. Therefore, the fact that these polydactylous mouse mutants have broader limbs could explain why they have more digits; furthermore, that the balance between GLI3A and GLI3R, and/or the absolute levels of GLI3R, are abnormal could explain why the digits are unpatterned.
Does SHH function as a morphogen? A critical issue to elucidate is whether target genes, for which repression is alleviated by SHH, encode positional values, or if specification of positional values is indirect and involves the production of other signalling molecules. Genes encoding BMPs are expressed in early limb buds. BMP2 is expressed at the posterior margin of the chick limb that overlaps with the polarizing region. Ectopic expression of BMP2 can be induced at the anterior margin of a chick wing when SHH is applied21. BMPs could therefore be involved in the specification of positional values, and could function locally or diffuse across the limb to form a gradient. However, addition of BMP2 to the anterior margin of the chick wing buds produces only small changes in digit pattern30, which indicates that BMPs might only be able to specify positional values in cells already 'primed' by SHH. Experiments in chick wing buds further support this idea31, whereas other experiments indicate that BMP signalling might also operate at the anterior of the limbs during the early stages of limb development, possibly antagonizing SHH signalling32. In addition to these extracellular signalling molecules that are associated with the polarizing region, there is a gradient of gap-junctional communication across the antero–posterior axis of the chick limb33. Interfering with gap-junctional communication in the cells within the polarizing region, and also in the responding cells, affects polarizing activity34. It therefore seems that direct cell–cell interactions might also be involved in the specification of positional values.
Timing mechanism
Another way of specifying positional values is through a timing mechanism — for example, through the length of exposure to a signal — and such a mechanism has been proposed for laying down the pattern along the long axis of the limb (see below). Recently, fate maps of cells in mouse embryos35, 36 revealed that the length of time that cells are exposed to the highest concentrations of SHH might contribute to digit patterning. This has led to a detailed model for specification of positional values for each mouse digit that integrates both concentration and length of exposure to SHH35. Specification of digit 1 is probably SHH-independent because the single digit that develops in the limbs of Shh-/- mice most closely resembles digit 1 (Ref. 22); low concentrations of SHH specify the positional value that leads to the formation of digit 2, whereas both time and concentration specify digit 3; cells are specified to form digits 4 and 5 according to the length of exposure to SHH (Fig. 2b).
The fate maps, which revealed the importance of timing in the mouse were made using activation of 
-galactosidase to mark the cells that expressed either Shh (Ref. 35) or Gli1 (Ref. 36). Staining for -galactosidase can be undertaken at different stages of limb development and this procedure allows the contributions that are made by marked cells to the digits to be assessed. The gene encoding GLI1 is a known target of SHH signalling (Box 3), therefore, marking of cells that express Gli1 allows the fate of cells that have responded to SHH to be traced. Cells that expressed Shh between days 10 and 11 of development and were located at the very posterior of the early limb bud were found to contribute to digits 3–5. However, cells that expressed Gli1 — these include Shh-expressing cells and a rim of cells anterior to the Shh-expression domain — contributed to digits 2–5 in a graded manner, with the highest number of marked cells being found in the posterior digits. These data confirm the results from the fate maps of early chick wing buds and indicate that, in the mouse as in the chick, there is considerable expansion of posterior tissue as the limb grows out. It is suggested, however, that, at this time, only the positional value that leads to formation of digit 2 has been specified by a low concentration of SHH (Fig. 2b). Cells that were marked in later mouse limb buds, between 11–12 days of development, were also found in the digits. Cells that had expressed Shh contributed to digits 3 and 4, but cells that had expressed Gli1 contributed to digits 2–5. These data show that cells that contribute to digits 3, 4 and 5 have been exposed to high SHH concentrations for progressively longer periods of time than cells that give rise to digits 1 and 2. It has therefore been suggested that digits 3, 4, and possibly 5, will not have been specified until this time (Fig. 2b).
Experimental evidence also indicates that SHH concentration alone might not specify positional values. Alterations in the amount and spatial distribution of SHH in the mouse limb buds, through manipulation of a gene that affects SHH diffusion35, resulted in the loss of a middle digit. If SHH concentration was responsible for the specification of the positional values, it would be expected that the most posterior digit would be lost.
Manipulation of the duration of SHH signalling would directly test the importance of time. Mouse mutants in which SHH signalling is curtailed are available — the naturally occurring mouse-limb-deformity mutant37 and the gremlin knockout38. In both of these mutants, digit number is reduced and posterior digits appear to be lost. Developing limbs of closely related Australian lizards that exhibit varying degrees of evolutionary limb reduction39, are also characterized by shorter durations of Shh expression. Interestingly, this seems to be related to reductions in digit number rather than changes in pattern. In both examples, however — the mouse-limb-deformity mutants and the Australian lizards — levels of Shh expression are also reduced.
The importance of sustained SHH signalling for mouse digit patterning seems at odds with the chick embryological experiments that indicated that anterior-wing-bud cells remember their exposure to a polarizing signal40, even when the signal has ceased (Box 1). Whether addition of SHH to the anterior margin of a chick wing bud induces SHH expression or not is still unclear, as the results are conflicting. It is also unclear how this timing model could be applied to the digits of the chick wing. Recent work indicated that the most anterior digit in the chick wing, digit 2, arises in a SHH-independent fashion41, but the homologies between the individual digits in the chick and the mouse are still controversial42.
Patterning and growth. The limb bud grows considerably during the patterning process and several models have been put forward to explain how patterning along the proximo–distal axis of the limb is generated as the limb bud grows out. According to the long-standing progress-zone model43, the length of time that the cells spend at the distal tip of the limb bud, and their exposure to signals (such as FGFs) from the apical ectodermal ridge, determines the proximo–distal positional value. Therefore, in the early limb bud, only the most proximal positional values will have been specified, and progressively more distal positional values will be generated over time as the limb bud grows out. More recently, it has been suggested that proximo–distal pattern44 is already specified in the early limb bud by an as-yet-unknown mechanism, and that the role of FGFs is to expand this pattern. Therefore according to this model, a complete set of positional values will already have been specified in the early limb bud. At later stages, the more proximal positional values have expanded, but the more distal positional values are still closely packed.
The progress-zone model43 has recently been challenged and, instead, an early-specification model has been proposed44, which is similar to the morphogen gradient for antero–posterior pattern. It is possible that the mechanisms that specify the positional values along the antero–posterior and proximo–distal axes of the limb are more similar than was previously appreciated. Alternatively, it is possible that both mechanisms — the morphogen gradient and timing — might operate in both axes. It is unclear whether either of these mechanisms operates along the dorso–ventral axis.
Antero–posterior positional values
One of the most important questions is how antero–posterior positional values are encoded molecularly. Early studies indicated that homeobox (Hox) genes (Box 4) could be important, but orthologues of fruitfly wing-patterning genes are also probable candidates.
Hox genes. A striking nested set of 5' Hoxd-gene-expression domains is established across the antero–posterior axis of the limb-forming region, with Hoxd13 expression at the very posterior45. Recently, these genes, together with 5' Hoxa genes, have been shown to have a role in initiating Shh expression in the polarizing region46. SHH signalling is then required to maintain this pattern of Hoxd gene expression, with cells at different positions across the antero–posterior axis expressing different combinations of Hoxd genes (Box 3) as the limb bud grows out. When a polarizing region or a bead soaked in retinoic acid, SHH or BMP2 is grafted onto the anterior margin of a chick wing bud, 5' Hoxd genes are ectopically expressed, which mirrors the pattern that is normally seen posteriorly (for example, see Ref. 47). Furthermore, HOXD11 misexpression leads to either an extra digit 2 in the wing, or changes in toe morphology that indicate posteriorization48. These data are consistent with the idea that 5' Hoxd genes might encode position across the antero–posterior axis. However, the recent demonstration that anterior expression of Hoxd12 (Ref. 49) and other 5' Hoxd genes50 leads to ectopic Shh expression provides a more likely explanation for the induction of extra digits and pattern changes.
An enormous effort has been devoted to assessing the functions of 5' Hoxd and 5' Hoxa genes in the developing limb by making single, double and even triple mouse knockouts51. This has revealed that paralogous genes are required for the development of each of the main segments along the long axis of the limb, with Hoxd13 and Hoxa13 being responsible for digit development. Therefore, it seems likely that ectopic expression of Hoxd genes in the early stages of development after application of polarizing-region signals in chick wing buds simply reflects the establishment of a new limb axis that will develop digits.
Fruitfly wing-patterning genes. Insights into other genes that could encode antero–posterior positional values comes from the similarities between the signalling cascade in vertebrate limbs (where SHH regulates expression of Bmp2) and that in fruitfly wings (where Hh regulates expression of decapentaplegic (dpp), a relative of the Bmp2 gene52 ). In the fruitfly wing, gene targets of DPP signalling that could encode positional values have been identified and include the transcription factors Optomotor blind (Omb; also known as Bifid), Spalt and Iroquois (also known as Araucan). Particular combinations of these transcription factors contribute to the specification of individual wing veins53. DPP signalling leads to overlapping domains of Spalt and Omb expression in the fruitfly wing. The spalt domain represents a response to the local DPP concentration, whereas the omb domain represents an expansion of a population of cells in which transcripts/proteins and memory of an earlier response to DPP persists54. The observation that these two mechanisms appear to operate together in patterning the fruitfly wing is intriguing in light of the idea that both concentration and time operate in patterning the vertebrate limb.
The vertebrate orthologues of the fruitfly genes Tbx2, Tbx3 (omb orthologues), Sall (spalt orthologues) and Irx (Iroquois orthologues) are expressed in vertebrate limb buds (for example, see Refs 32,55,56) and there is evidence that TBX2 and TBX3 might encode posterior positional values in vertebrate limbs. Tbx2 and Tbx3 are expressed in anterior and posterior stripes in early limb buds and the posterior stripe of expression has been shown to depend on polarizing signalling32, 57 (Box 3). Overexpression of TBX2 and TBX3 has been reported to result in posteriorization of chick toes57, although others report a shift in limb position58. In addition, human patients with mammary-ulnar syndrome, caused by TBX3 haploinsufficiency, have posterior limb defects59. Although these studies on Tbx genes are an encouraging start, considerable efforts will be required to uncover the positional code for antero–posterior limb pattern to an extent that is comparable to the positional code that has been deciphered for other tissues, such as the dorso–ventral pattern of the neural tube (the forerunner of the central nervous system)60.
Interpretation of positional values
The last step in pattern formation is the morphogenesis of the digits through interpretation of the positional values. Evidence from studies on chick legs indicates that the morphogenesis of each individual digit involves local interactions61. During this time, the digital plate is very broad and SHH expression is reduced and soon disappears. A change occurs from the global patterning system that operates in the limb bud, in which SHH has a pivotal role, to a series of patterning systems, which regulate the morphology of individual digits after the expression of SHH is extinguished. The idea of separate patterning systems for each digit is supported by the observation that manipulations at this late stage have local effects; the anatomy of a single digit changes independently from that of other digits61, 62 (Fig. 3).
Figure 3 | Cell–cell signalling in the digital plate.
![Figure 3 : Cell|[ndash]|cell signalling in the digital plate. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nrm/journal/v7/n1/thumbs/nrm1830-f3.jpg)
a | The experimental manipulation that was described in Ref. 62. A bead (red circle) soaked in sonic hedgehog (SHH) was placed in the second interdigital space (IDS) of a chick leg at a stage when the digital primordia that will form the four toes had developed. The digit primordia are still joined by soft tissue, which will later undergo programmed cell death to separate them. b | The skeleton of the foot that developed 5 days after the manipulation shown in a. One of the toes (indicated by 3*) adjacent to the bead, which would normally be toe digit 2, is elongated and has an extra phalanx. It resembles toe digit 3 and shows that the toe has been posteriorized. The other toe digits 1, 3 and 4 have normal morphology.
Grafting experiments showed that endogenous signals that control individual digit morphologies come from interdigital tissue61; toe morphology can be both anteriorized (phalanx number reduced) or posteriorized (phalanx number increased). Although these effects are known to be mediated through BMPs (because co-implantation of beads soaked in BMP antagonists prevent toe elongation61) the gene targets of BMP signalling are currently unknown. Hoxa13 and the four contiguous genes in the Hoxd gene cluster, Hoxd10–Hoxd13, are co-expressed throughout the digital plate, and a global regulatory region that controls digital expression has been identified63. One possibility is that this late phase of Hox gene expression might be involved in interpreting positional values to give digit identity, but no changes in the expression of Hoxd10–12 were detected following digit manipulations61.
SHH-soaked beads that are implanted between the digits also produce posteriorization (Fig. 3), possibly by mimicking the signalling of another vertebrate Hh, Indian hedgehog (IHH), which is expressed at this stage. SHH-soaked beads prolong FGF signalling in the apical ectodermal ridge. Interfering directly with FGF signalling by either enhancing it or abrogating it, can lead either to the formation of extra phalanges or to a reduction in phalanx number, respectively62. Interestingly, toes that are induced from these manipulations have normal tips, which indicates that there is a special programme that controls the formation of a limb tip when FGF signalling is switched off. This might explain why even the rudimentary digit of Shh-/- mouse embryos is finished off by a claw22.
Conclusions
This review shows that a picture, albeit somewhat hazy, of the signalling cascades that are involved in digit patterning is beginning to emerge. The first cascade operates in the limb bud where SHH signalling has a pivotal role, and the second cascade operates in individual digit primordia and involves BMP signalling. Although a number of genes with potential roles in antero–posterior patterning have been recently identified, we are still facing the challenge to fill in the considerable gaps in our current understanding by identifying other genes that are involved in this process. Another challenge is to gain a deeper understanding of the cellular basis of limb development by finding new ways to visualize extracellular signalling molecules and measure their concentrations and the cellular responses to them. Finally, information about antero–posterior position must be coordinated with information about proximo–distal and dorso–ventral position, and fed into the interpretation process with information about whether the organ is a wing or a leg. A really detailed focus on the mechanisms of antero–posterior digit patterning might be relevant to understanding how positional information operates along the other two axes of the limb. The knowledge of how the limb develops should cast light on the basis of human congenital limb defects and the general principles will be relevant to tissue engineering and devising new approaches for tissue repair. In addition, many of the molecules that control cell activities during limb development have been also implicated in cancer.