Vertebrate Limb Development and Malformations

Abstract

To understand limb abnormalities it is necessary to understand how the limb develops. The limb is the organ whose development is probably best understood. The limbs develop from small protrusions (the limb buds) that arise from the body wall of the embryo. Positioning and patterning the limb involves cellular interactions both between the ectoderm surrounding the limb bud and between the mesenchymal cells that form the core of the limb bud. As the limb grows out the cells acquire a positional value that relates to their position in the bud with respect to all three axes, proximo-distal, antero-posterior, and dorso-ventral. These positional values largely determine how the cells will develop such as what sort of cartilaginous elements they will form. The positional value of the cells is acquired in the progress zone at the tip of the growing bud. The time spent in the progress zone may determine the positional values along the proximo-distal axis, that is the formation of, for example the humerus, then the radius and ulna. Loss of the progress zone due to damage to the overlying apical ridge leads to truncations, and this progress zone model can also account for the effects of thalidomide. Position along the antero-posterior axis such as the character of the digits is by a signal from the polarizing region at the posterior margin of the limb and involves the signaling protein Sonic hedgehog. A signal from the dorsal ectoderm specifies the dorso-ventral axis. Hox genes that are transcription factors are expressed both along the body axis and in a complex pattern in the limb and may record positional value. Human mutations in these genes lead to limb abnormalities. Muscle cells have a separate origin from the cartilaginous cells and those that form connective tissue and tendons, and they migrate into the bud from the somites and are patterned by the connective tissue. Cell death separates the digits.

Main

Despite many genes having been identified that are responsible for congenital malformations, there are few examples where the malformation can be accounted for in terms of how the normal developmental processes have been altered^(1–4). The vertebrate limb is probably the best understood example of organogenesis. The vertebrate embryonic limb has the advantage of having a basic pattern that is initially quite simple, and in the chick embryo that provides a good model system, the limbs are easily accessible for surgical manipulation. Malformations due to mutations are quite easily identifiable in the mouse. The limbs develop from small protrusions (the limb buds) that arise from the body wall of the embryo. By d 10 the main features of chick limbs are well developed, particularly the cartilaginous elements (which are later replaced by bone), muscles, and tendons (Fig 1). The limb has three developmental axes: the proximo-distal axis runs from the base of the limb to the tip; the antero-posterior axis runs parallel with the body axis (in the human hand it goes from the thumb to the little finger, and in the chick limb that has three easily identifiable digits, from digit 2 to 4); the dorso-ventral axis runs from the back of the hand to the palm.

Figure 1

The embryonic chick wing 10 d after laying. By this time, the main cartilaginous elements have been laid down. They later become ossified to form bone. The muscles and tendons are also well developed at this stage but cannot be seen in this type of preparation. The three developmental axes of the limb are proximo-distal, antero-posterior, and dorsoventral, as shown in the top panel. The chick wing only has three digits that have been called 2, 3, and 4. Scale bar = 1 mm. (Used with permission from^{Ref. 4}).

The early limb bud has two major components: a core of loose mesenchymal mesoderm cells and an epithelial ectodermal layer. Most of the limb develops from the mesenchymal core, although the muscle cells of the limb have a separate lineage and migrate into the bud from the somites. At the tip of the limb bud is a progress zone of rapidly dividing and proliferating undifferentiated cells. The progress zone lies directly beneath a thickening in the ectoderm, the apical ectodermal ridge (Fig. 2). It is only when cells leave the progress zone that some begin to differentiate as, for example, cartilage cells and so as the bud grows, cartilaginous structures begin to appear in the mesenchyme in a proximo-distal sequence.

Figure 2

Cells acquire their positional value in the progress zone that is specified at the distal end of the bud by the apical ectodermal ridge. As the limb bud grows outward, cells in the progress zone proliferate and acquire a positional value. When they leave the zone, cartilage can begin to differentiate and the cartilaginous elements are laid down in a proximo-distal sequence, starting with the humerus. The position of the cells along the antero-posterior axis is specified by a signal from the polarizing region, which is located at the posterior side of the bud (Used with permission from^{Ref. 4}).

Patterning in the limb is dependent on cell-cell interactions. The early limb bud has considerable powers of regulation and pieces of it can generally be removed, rotated, or put in a different position without perturbing the final pattern. However, there are two regions in which this general rule does not apply. These are crucial organizing regions, and their removal or transplantation has profound effects. One of these regions is the apical ectodermal ridge at the tip of the limb; the other is a region at the posterior margin of the mesenchyme known as the polarizing region (Fig. 2).

Patterning of the limb, i.e. the spatial organization of cartilage, muscles, and tendons, can be considered in terms of a mechanism based on positional information⁽⁴⁾. Positional information refers to a property that cells acquire that relates to their position with respect to a boundary, for example a position could be specified by the concentration gradient of a diffusible molecule. This could give the cells a positional value. The cells then can interpret this positional value by differentiating in specific ways at different positional values and therefore generating a pattern of cellular differentiation. Cells would thus acquire positional information first and then interpret these positional values according to their genetic constitution and developmental history. It is, for example, the difference in their developmental history that makes wings and legs different, because their cells have the same positional values.

THE PROXIMO-DISTAL AXIS

The apical ectodermal ridge at the tip of the bud consists of closely packed columnar cells that are linked by extensive gap junctions. Their tight packing gives the ridge a mechanical strength that probably keeps the limb flattened along the dorso-ventral axis; the length of the ridge controls the antero-posterior width of the bud. The progress zone, lying beneath the apical ectodermal ridge, is a region of rapidly proliferating mesenchymal cells, which produces the initial outgrowth of the limb bud. It is also the region where limb cells acquire their positional information. The localization of the apical ectodermal ridge appears to involve the gene radical fringe, which is a homologue of the Drosophila gene fringe that is similarly involved in specifying a dorso-ventral boundary in the wing⁽⁵⁾. Radical fringe is expressed in the dorsal limb ectoderm before the formation of the apical ectodermal ridge. The apical ectodermal ridge develops at the boundary between cells that express radical fringe and those that do not.

Removal of the ridge from a chick limb bud by microsurgery results in a significant reduction in growth and in truncation of the limb, with distal parts missing. The proximo-distal level at which the limb is truncated depends on the time at which the ridge is removed, the earlier the ridge is removed, the more distal regions are lost. Any human malformation involving truncation should initially to be considered in terms of possible damage to the ridge.

A major signal from the ridge is provided by proteins of the FGF family. FGF-8 is expressed throughout the ridge and FGF-4 in the posterior region. FGF-4 protein can act as a functional substitute for an apical ridge, so if the ridge is removed and beads that release the growth factor are grafted into the tip of the limb in its place, more or less normal outgrowth of the limb continues⁽⁶⁾. Presumably FGF-4 can bind to the receptors for FGF-8.

Patterning along the proximo-distal axis is not well understood but the positional value of the cells along this axis may be specified by the time cells spend in the progress zone. As the limb bud grows, cells are continually leaving the progress zone. In the forelimb, for example, cells leaving first develop into the humerus and those leaving last form the tips of the digits. If cells could measure the time they spend in the progress zone, this would give them a positional value along the proximo-distal axis (Fig. 3). A timing mechanism of this sort is consistent with the experimental observation that removal of the apical ridge results in the disappearance of the progress zone and in a distally truncated limb.

Figure 3

A cell's proximo-distal positional value may depend on the time it spends in the progress zone. Cells are continually leaving the progress zone. If the cells could measure how much time they spend in the progress zone, this could specify their position along the proximo-distal axis. Cells that leave the zone early form proximal structures whereas cells that leave it last form the tips of the digits (Used with permission from^{Ref. 4}).

Evidence for a mechanism based on time comes from killing cells in the progress zone or blocking their proliferation at an early stage by, for example, X irradiation. The result is that proximal structures are absent but distal ones are present, and can be almost normal. Because many cells in the irradiated progress zone do not divide, the number of cells that leave the progress zone during each unit of time is much smaller than normal. Proximal structures are thus very small, or even absent. As the bud grows out the progress zone becomes repopulated by the surviving cells, so that only distal structures such as digits, are formed. Also consistent with this mechanism is the observation that a presumptive proximal region from a leg bud that would give rise to the femur is grafted into the progress zone of a wing bud. This acquires more distal positional values and develops as toes; it develops according to both its position and developmental history and so forms distal leg structures⁽⁴⁾.

It is possible that this could provide the basis of an explanation for the action of thalidomide in causing limb abnormalities that typically involve loss of proximal structures⁽⁷⁾. There is evidence that thalidomide blocks angiogenesis⁽⁸⁾ and this could cause cells to remain in the progress zone for much longer periods than normal because of reduced proliferation. Another possibility is that thalidomide causes a hematoma and cell death in the early bud; the surviving cells would proliferate and only later move out of the progress zone and only form distal structures.

THE ANTERO-POSTERIOR AXIS

The polarizing region of a vertebrate limb bud has signaling properties and when the polarizing region from one early wing bud is grafted to the anterior margin of another early wing bud, a wing with a mirror-image pattern develops: instead of the normal pattern of digits 2 3 4, the pattern 4 3 2 2 3 4 develops. The pattern of muscles and tendons in the limb shows similar mirror-image changes.

The additional digits come from the host limb bud and not from the graft, showing that the grafted polarizing region has altered the developmental fate of the host cells in the anterior region of the limb bud. The limb bud widens in response to the polarizing graft, which enables the additional digits to be accommodated; widening of a limb bud is always associated with an increase in the extent of the apical ectodermal ridge.

One way that the polarizing region could specify position along the antero-posterior axis is by producing a diffusible morphogen (Fig. 4). The concentration of morphogen could specify the position of cells along the antero-posterior axis with respect to the polarizing region located at the posterior margin of the limb. Cells could then interpret their positional values by developing specific structures at particular threshold concentrations of morphogen. Digit 4, for example, would develop at a high concentration, digit 3 at a lower one, and digit 2 at an even lower one. According to this model, a graft of an additional polarizing region to the anterior margin would set up a mirror-image gradient of morphogen, which would result in the 4 3 2 2 3 4 pattern of digits that is observed. If the action of the polarizing region in specifying the character of a digit is due to a diffusible signal, then when the signal is weakened, the pattern of digits should be altered in a predictable manner. Grafting small numbers of polarizing region cells to a limb bud results only in the development of an additional digit 2.

Figure 4

The polarizing region can specify pattern along the antero-posterior axis. If the polarizing region is the source of a graded morphogen, the different digits could be specified at different threshold concentrations of signal (left panels). Grafting an additional polarizing region to the anterior margin of a limb bud (center panels) would result in a mirror image gradients of signal, and thus the observed mirror image duplication of digits. The signal from a grafted polarizing region can be attenuated by grafting only a small number of polarizing region cells to the anterior margin of the limb bud (right panels), so that now only an extra digit 2 develops. (Used with permission from^{Ref. 4}).

The Sonic hedgehog gene is expressed in the limb polarizing region and Sonic hedgehog protein is known to be involved in numerous patterning processes, e.g., in the somites and the neural tube⁽⁹⁾. Chick fibroblast cells transfected with a retrovirus containing the Sonic hedgehog gene acquire the properties of a polarizing region; they cause the development of a mirror-image limb when grafted to the anterior margin of a limb bud, as will beads soaked in Sonic hedgehog protein. The bead must be left in place for 24 h for extra digits to be specified and the pattern of digits depends on the concentration of Sonic hedgehog protein in the bead. A higher concentration is required for a digit 4 to develop than for a digit 2.

Further evidence for the key role of Sonic hedgehog comes from the mouse mutation extra toes; in which there is an additional anterior digit and additional anterior expression of Sonic hedgehog. Polydactyly could thus result from anterior expression of the human version of Sonic hedgehog. However, mice in which the Sonic hedgehog gene has been knocked out develop proximal limb structures but lack distal ones, showing that the gene is not required for initiation of limb development. Moreover, retinoic acid applied locally or given systemically can also cause polydactyly⁽¹⁰⁾.

It is not clear that sonic hedgehog protein is itself the diffusible morphogen that specifies positional information along the antero-posterior axis. Several other molecules may also be involved in positional signaling along the antero-posterior axis⁽¹¹⁾. The bone morphogenetic growth factors BMP-2 and BMP-4 are present in a gradient within the progress zone, with the high point at the posterior margin of the limb; however, their function is not known but they could be involved in positional signaling. Local application of Sonic hedgehog protein induces expression of the BMP growth factors.

The Greigcephalo-polysyndactyly syndrome is autosomal dominant and results in polydactyly and craniofacial abnormalities; it is probably due to a mutation in the Gli3 gene that is involved in the transduction of the Sonic hedgehog signal⁽¹²⁾, i.e. the internal events that allow a Sonic hedgehog protein binding to a receptor on the cell surface and leading to changes in gene expression.

The polarizing region is also involved in the maintenance of the apical ridge probably via BMP-2 and BMP-4, and there is a positive feedback loop between Sonic hedgehog protein in the mesoderm and FGF-4 expression in the ridge. This interaction between FGF-4 and Sonic hedgehog can be shown by the induction of new limbs at ectopic sites⁽¹³⁾. Localized application of FGF-4 to the flank of a chick embryo, between the wing and leg buds, induces the production of FGF-8 in the ectoderm, and then the ectopic expression of Sonic hedgehog. The Sonic hedgehog protein then feeds back to induce expression of the embryo's own gene for FGF-4 resulting in the maintenance of the ridge and the outgrowth of an additional limb bud at this site. The type of limb is determined at the site of Sonic hedgehog expression. In general, additional wing buds develop from application of FGF-4 to the anterior flank, whereas leg buds develop from application to the posterior flank.

The development of a pattern of cartilaginous elements along the antero-posterior axis may involve mechanisms other than signaling by the polarizing regions. Evidence for this comes, for example, from observing the development of reaggregated limb buds after the mesenchymal cells of early chick limb buds have first been disaggregated and thoroughly mixed to disperse the polarizing region⁽¹⁴⁾. They are then reaggregated, placed in an ectodermal jacket, and grafted to a site where they can acquire a blood supply, such as the dorsal surface of an older limb (Fig 5). Limb-like structures develop from these grafted buds even though they have developed in the absence of a polarizing region. Several long cartilaginous elements may form in the more proximal regions of these abnormal limbs, although none of the proximal elements can be easily identified with normal structures. More distally, however, reaggregated hind limb buds develop identifiable toes. The fact that well-formed cartilaginous elements can develop at all in the absence of a discrete polarizing region shows that the bud has a considerable capacity for self-organization. In the digits that develop in hind limb reaggregates there is no sign of the correlation between Hox d gene expression and antero-posterior position seen in normal development as described below.

Figure 5

Reaggregated limb bud cells form digits in the absence of a localized polarizing region. Mesodermal cells from a chick leg bud are separated, mixed to disperse all the cells, including the polarizing region, and then reaggregated, placed in an ectodermal jacket, and grafted into a neutral site. Well-formed toes develop distally (Used with permission from^{Ref. 4}).

There may therefore be a mechanism in the limb bud that generates a basic pattern (a prepattern) of equivalent cartilaginous elements. These elements could then be given their identifies and further refined by response to positional information involving signals such as Sonic hedgehog. The mechanism for generating the prepattern could be based on a reaction-diffusion mechanism. In the wing, for example, a reaction-diffusion or related mechanism could result in a single peak in some morphogen forming in the proximal region of the limb, which would specify a prepattern for the humerus. More distally, alterations in the reaction-diffusion conditions, due to changes in proximo-distal positional information, could give rise to three peaks of the morphogen, giving the cartilaginous elements of the three wing digits. These prepatterns could then be modified by signals specifying antero-posterior and dorso-ventral positional information. In this reaction-diffusion model, polydactyly in humans could simply result from a chance widening of the limb bud. If a reaction-diffusion mechanism generates a periodic pattern of cartilage elements across the limb as the digits are forming, merely widening the limb bud by some small developmental accident would enable a further digit to develop.

THE DORSO-VENTRAL AXIS

There is a well-defined pattern along the dorso-ventral axis; muscles and tendons have a complex dorso-ventral organization. The development of a pattern along this axis has been studied by recombining ectoderm taken from left limb buds with mesoderm from right limb buds, so that the dorso-ventral axis of the ectoderm is reversed with respect to the underlying mesoderm. Distal regions, particularly the "hand" region, have a reversed dorso-ventral axis, with the pattern of muscles and tendons reversed and corresponding to that of the dorso-ventral axis of the ectoderm. The ectoderm can therefore specify dorso-ventral patterning in the limb. Genes controlling the dorso-ventral axis in vertebrate limbs have been identified from mutations in mice⁽¹⁵⁾. Mutations that inactivate the gene Wnt-7a, which encodes a secreted signaling protein of the Wingless family, result in limbs in which many of the dorsal tissues adopt ventral fates to give a double ventral limb, the two halves being mirror images. The Wnt-7a gene is expressed in the dorsal ectoderm (Fig. 6) and this suggests that the ventral pattern is the ground state and is modified dorsally by the dorsal ectoderm, the Wnt-7a gene playing a key role in patterning the dorsal mesoderm. Expression of the gene engrailed characterizes ventral ectoderm. Mutations in engrailed result in Wnt-7a being expressed ventrally, giving a double dorsal limb.

Figure 6

Traverse section through progress zone of a limb bud. The ectoderm controls dorso-ventral pattern in the developing limb. The gene Wnt-7a is expressed in the dorsal ectoderm and engrailed is expressed in the ventral ectoderm. The gene Lmx-1 is induced by Wnt-7a in the dorsal mesoderm and is involved in specifying dorsal structures (Used with permission from^{Ref. 4}).

One function of Wnt-7a is to induce expression of the LIM homeobox gene Lmx-1 in the underlying mesenchyme. This gene encodes a transcription factor, the expression of which specifies a dorsal pattern in the mesoderm. Ectopic expression of Lmx-1 in the ventral mesoderm results in the cells adopting a dorsal fate, resulting in a mirror-image dorsal limb. A structural relative of Lmx-1 in Drosophila, apterous, is involved in specifying the dorsal surface of the insect wing.

SPECIFICATION OF FORE- VERSUS HIND-LIMBS

The positional signals controlling limb patterning are the same in chick forelimbs and hindlimbs as well as in different vertebrates, but they are interpreted differently. A polarizing region from a wing bud, for example, will specify additional mirror-image digits if grafted into the anterior margin of a leg bud, where these digits are toes, not wing digits, as the signal is interpreted by leg cells. Similarly, a polarizing region from a mouse or human limb bud specifies additional wing structures when grafted into the anterior margin of a chick wing bud.

Forelimbs and hindlimbs differ because of the difference in their developmental histories, which are related to the position of the limb bud along the antero-posterior axis of the animal's body. Genes that play a key role in specifying the difference between fore- and hindlimbs are those related to Brachyury, a gene that plays a key role in early mesoderm specification. These are the Tbx genes and Tbx5 and Tbx4 are restricted in their expression to fore- and hind limbs, respectively^(16,17), and in addition another transcription factor, Pitx1, is expressed only in the hindlimb. Direct and striking evidence that such genes determine limb identity comes from expression of Pitx1 in the chick wing bud with the result that it develops as a leg and Tbx4 is expressed⁽¹⁷⁾. The Holt-Oram syndrome is due to a mutation in the Tbx5 and this results in a variety of defects in the upper limb and shoulder girdle⁽¹⁸⁾. Forelimb abnormalities in the ulnar-mammary syndrome are due to a mutation in Tbx3, which is expressed mainly in the forelimb⁽¹⁹⁾.

HOX GENE EXPRESSION AND POSITIONAL SPECIFICATION

Vertebrate Hox genes specify position along the antero-posterior axis of the body and may also provide positional values in the limbs. At least 23 different Hox genes are expressed during chick limb development⁽²⁰⁾. Attention has been largely focused on the genes of the Hox a and Hox d gene clusters, which are related to the Drosophila abdominal a gene These sets of genes are expressed in both forelimbs and hindlimbs, whereas the expression of Hox b and Hox c gene clusters is restricted to either the forelimb or the hindlimb bud, respectively.

The expression of Hox genes during limb development is dynamic; the pattern of expression of a particular gene can undergo significant changes as the limb bud grows. A particular stage can be used on illustrate the main features of Hox expression (Fig. 7) and shows for two of the complexes a correlation between gene expression and position along the proximo-distal and antero-posterior axes. Expression of Hox d9 through Hox d13 is sequentially initiated at the posterior margin of the limb in the progress zone, resulting in a nested pattern of expression of different Hox d genes along the antero-posterior limb axis. Thus Hox d9 expression covers the complete region where the Hox d genes are expressed, and successive genes occupy more and more posterior regions so that Hox d9-Hox d13 are all expressed in a small posterior region (lower panel Fig. 7). Expression of the Hox a genes is similarly sequentially initiated in the progress zone and ends up as a nested pattern of expression along the proximo-distal axis (upper panel Fig. 7). At this stage of limb development the proximo-distal sequence of expression of Hox a genes corresponds to the three main proximo-distal regions of the limb, which are: upper limb, where the humerus (or femur in the hind limb) forms, Hox a9 expressed; the lower limb, where the radius and ulna (or tibia and fibula) develop, Hox a9-11 expressed; and the wrist and digits, Hox a9-13 expressed.

Figure 7

Pattern of Hox gene expression in the chick wing bud. The Hox a genes (top panel) are expressed in a nested pattern along the proximo-distal axis, Hox a 13 being expressed most distally, although the Hox d genes (bottom panel) are expressed in a similarly nested pattern along the antero-posterior axis, Hox d13 being expressed most posteriorly (Used with permission from^{Ref. 4}).

If the Hox genes are involved in recording positional information, then experimental manipulations that lead to changes in the pattern of the limb skeletal elements should be preceded by a corresponding change in Hox expression domains. If a polarizing region is grafted to the anterior margin of a wing bud, there is indeed a change to a mirror-image pattern of Hox d expression. This occurs within 24 h of grafting, which is about the time required for the polarizing region to exert its effect.

The results of experiments designed to answer this question in which specific Hox genes are knocked out (totally inactivated or removed) are complicated; there is evidently no simple relationship between Hox gene expression and the pattern of the cartilaginous elements. There is no simple Hox code for recording in a combinatorial manner positional values in the limb. Knockouts of individual Hox genes in the mouse do not transform one digit into another. Instead, many bones in the hand are affected in both size and shape, and new elements may even develop. When more than one Hox gene is knocked out at the same time, the effects can be much more severe, and it seems that the Hox genes have an important influence on growth of the cartilaginous elements. Thus, knockout of both Hox a11 and Hox d11 results in the absence of both the radius and ulna. Over expression of Hox a13, which is normally expressed in the distal region of the limb bud, results in limbs in which the radius and the ulna are reduced in size. This suggests that they be transformed into small wrist-like elements, probably by a change in the control of cell multiplication. Over expression of Hox d13 results in the shortening of the long bones of the leg owing to Hox d13 affecting the rate of cell proliferation in the growing cartilaginous elements. These results show that expression of Hox genes can control the size of the cartilaginous elements in the limbs at both early and later stages^(21,22).

There is a very well-defined pattern of Hox gene expression along the mesoderm of the main body axis from which the wing and leg buds form, and the set of genes expressed in the fore- and hindlimb regions is different. That these patterns specify whether a wing or leg develops is supported by the changes in Hox gene expression when an FGF bead is placed in the flank, inducing a wing or a leg. The new pattern of Hox gene expression corresponds to that normally found in either the wing or the leg⁽¹³⁾.

Evidence that Hox genes are involved in specifying the position of the polarizing region comes from mouse embryos that express a transgene of Hox b8 in more anterior regions of the embryo than normal. In these mice, an extra polarizing region forms at the anterior margin of the forelimb buds, causing extra digits to develop. Evidence for the combined action of Hox genes in determining antero-posterior position of the forelimb bud comes from knockout mice lacking Hox b5 expression, in which the forelimbs develop at a more anterior level.

The involvement of Hox genes in human limb development is shown by the phenotype of human Hox gene mutations. A mutation in the human Hox d13 gene results in polydactyly and fusion of digits whereas a mutation in human Hox a13 results in anterior and posterior digits that are reduced in size⁽²³⁾.

MUSCLES AND TENDONS

Limb muscle cells have a different lineage to that of limb connective tissues, cartilage, and tendons. Cells that give rise to muscle migrate into the limb bud from the somites at a very early stage. After migration, the future muscle cells multiply and initially form a dorsal and a ventral block of presumptive muscle (Fig. 8). These blocks undergo a series of divisions to give the final muscle masses. Unlike the cartilage and connective tissue cells, which acquire positional information in the progress zone, the presumptive muscle cells, at least initially, do not acquire positional values, and are all equivalent. The muscle pattern is determined by the connective tissue into which the muscle cells migrate, rather than by the muscle cells themselves. A mechanism that could pattern the muscle is based on the prospective muscle-associated connective issue having surface or adhesive properties that the muscle cells recognize, resulting in the migration of muscle cells to these regions. The pattern of muscle could thus be determined by the pattern of muscle-associated connective tissue, which is presumably specified by mechanisms similar to those that produce the pattern of cartilage. If the pattern of connective tissue changed with time, the presumptive muscle cells would migrate to the new sites, and this could account for the splitting of the muscle masses.

Figure 8

Development of muscle in the chick limb. A cross-section through the chick limb in the region of the radius and ulna shortly after cartilage formation shows presumptive muscle cells present as two blocks; the dorsal muscle mass and the ventral muscle mass. These blocks then undergo a series of divisions to give rise to individual muscles (Used with permission from^{Ref. 4}).

The pattern of cartilage, tendon, and muscle in the limb may be specified by the same signals, as a polarizing region graft causes the development of a mirror-image pattern of all these elements⁽²⁴⁾. Each of the elements develops in its final position and there is little interaction between them⁽²⁵⁾. The mechanism whereby the correct connections between tendons, muscles, and cartilage are established has still to be determined. They simple seem to make connections with those muscles and tendons nearest to their free ends.

CELL DEATH

Programmed cell death by apoptosis plays a key role in molding the form of the chick and mammalian limb, especially the digits⁽²⁶⁾. The region where the digits form is initially plate-like, as the limb is flattened along the dorso-ventral axis. The cartilaginous elements of the digits develop from the mesenchyme at the correct positions within this plate. Separation of the digits then depends on the death of the cells between these cartilaginous elements. There is evidence for the involvement of BMP-4 in cell death; if the function of the BMP-4 receptor is blocked in the developing chick leg, cell death does not occur, and the digits are webbed.