ANTERIOR D ^A V 0 E ^ ■ T r T A R A POSTERIOR
zone of polarising activity
Wnt-7a
— Lmx-1 — en-1 — shh
R em odelling o f the limb, in particular to give rise to the digits, is a partial consequence o f cells dying. The limbs are essentially patterned in a miniature form , so that by just after halfway through gestation, they are tiny m odels o f the adult self, which then undergo extensive growth.
Limb specification and induction
The factors responsible for the formation o f tw o pairs o f limbs at the appropriate level on the axial skeleton are thought to involve hom eobox-containing genes and growth factors. Initially in the chick there are ill-defined boundaries o f H o x b 9 , H o x c9 and H o x d 9 in the prospective w ing region, but these soon become restricted such that H o x b 9 and H o x c9
form the posterior boundary o f the forelimb and Hoxd9 is expressed throughout the w ing bud (Cohn et al., 1997). The flank region expresses H o x b 9 and H o x c9 but not H o x d 9
and the leg region expresses H oxc9 and Hoxd9, but not Hoxb9. These three H o x genes appear to play a crucial role in the positioning o f the limb buds (Cohn et a l , 1997). Other data from the chick embryo demonstrates that beads soaked in FG Fs - F G F -1, FG F-2 or FG F-4 - and placed in the flank between stages 13 and 17 result in ectopic limbs being formed (Cohn et at., 1995). The extent o f re-programming o f H o x 9
distribution and the eventual type o f limb formed, w ing or leg, is dependent on the position o f the bead in the flank region (Cohn et al., 1995; Cohn et al., 1997). W hilst aU FG Fs so far tested can ectopically induce limb outgrowth it seem s that the crucial initiating FGF m ay be F G F -10 which is expressed by the prospective limb m esenchym e at the appropriate time; moreover, ectopic expression o f F G F -10 induces the formation o f a com plete limb via FGF-8 in the AER (Ohuchi et al., 1997).
Limb patterning
• Proximo-distal patterning
The AER is thought to play both a mechanical and biochemical role in limb developm ent. The thickening o f the distal tip o f the limb maintains the shape o f the limb bud, while flattening it in a D -V direction. Signals from the AER are important for maintaining the progress zone (PZ), a region o f undifferentiated cells at the distal part o f the limb bud (Summerbell et al., 1973). The cells in the PZ appear to “count time” and as they are progressively released from this zone, due to population pressure, they are allocated more and more distal positional values, with the m ost distal encoding the distal most phalanges. The PZ m esenchym e also signals back to the AER to maintain its existence and activity, in a reciprocal fashion.
The hom eobox gene M s x -\ is expressed by the mesenchym al cells at the tip o f the developing limb in the progress zone (Hill et al., 1989; Robert e t al., 1991). It is thought to play a role in maintaining cells in an undifferentiated and proliferative state (Kostakopoulou et al., 1996). A feedback mechanism exists between the AER and
M s x - \ expression, and if the ridge is removed then M s x - \ expression is lost. In the
limbless mutant which has severely truncated lim bs, the AER does not form and M s x - l
expression fails to be maintained in the limb m esenchym e, but can be restored if a normal AER is grafted onto the limb bud (Coelho and Kosher, 1991; Robert e t al., 1991).
It seem s very likely that members o f the FGF family are responsible for the AER signals since they are expressed in the ridge and able to substitute for the AER in ridge removal experiments (Nisw ander et a i , 1993; Fallon et al., 1994). FG F-8 is expressed throughout the AER (Mahmood et al., 1995) w hile FG F-4 is expressed later and restricted posteriorly (Niswander and Martin, 1992). FG F-2, how ever, is expressed in both the ectoderm and the m esoderm (Savage e t a i , 1993).
• A ntero-posterior patterning
Thirty years ago “cut and paste” experiments revealed that grafting o f posterior limb bud m esenchym e to the anterior side o f the limb resulted in outgrowth o f a mirror im age limb. These experiments demonstrated that A-P polarity in the limb is signalled by a small group o f posterior cells that release a morphogen and set up a gradient across the limb. These cells, known as the zone o f polarising activity (ZPA), have recently been show n to express the short range diffusible factor Shh (Riddle et al., 1993). The importance o f
Shh is illustrated in the loss o f function m ouse mutant which results in a tmncated limb with absence o f distal structures (Chiang et al., 1996). This phenotype is thought to result from a disruption in the feedback signalling between FG F-4 in the ridge and shh in the ZPA (Laufer et al., 1994; Niswander et al., 1994).
The zinc finger genes G //1-3 are expressed in the limb m esenchym e and are potential mediators o f hedgehog signalling (Alexandre et al., 1996). GUI is expressed in the posterior m esenchym e, overlapping Shh expression, while GU2 and GU3 are found throughout the limb m esenchym e except in the Shh domain. If Shh expression is experimentally manipulated then transcription o f the Gli genes is also affected, with GUI
being upregulated in areas o f Shh expression and GU3 is downregulated (Marigo et al.,
1996).
It seem s that Shh also activates the expression o f BM Ps in the ZPA and cells adjacent to them (Francis et al., 1994). Bm p2 and B m p l are expressed in the posterior m esenchym e (Francis et al., 1994; Francis-W est et al., 1995). B m p A is expressed throughout the chick w ing bud m esenchym e. These signalling m olecules appear to regulate expression o f the hom eobox genes, in particular the H O X D cluster (Dollé et al., 1989; D ollé et al.,
1991; Izpisùa-Belmonte et al., 1991). The resulting expression pattern o f H o x d genes
{Hoxd9 through to H o x d l3 ) in overlapping domains across the distal end o f the developing limb bud, in domains corresponding to their 3 ’ to 5 ’ chrom osom al location in the cluster, may be the means by which limb m esenchymal cells record and act on their
A -P positional cues designated to them by the morphogen gradient (Izpisùa-Belm onte et a l , 1991).
• Dorso-ventral patterning
The epithelium overlying the limb bud expresses a number o f genes som e o f w hich are restricted to either the dorsal or ventral compartment, and may be critical in D -V polarity. The gene W n t-lâ , which encodes a short range diffusible factor, is expressed in the dorsal ectoderm o f the developing m ouse limb and in mice null for this gene, double ventral limbs result suggesting that Wnt-la. is crucial for specifying the dorsal state (Parr and M cM ahon, 1995). Wnt-la. appears to induce the dorsal m esenchym e to express the LIM -hom eodom ain containing gene Lm x-1 and viral overexpression o f Lm x-l results in the formation o f a double dorsal limb (Riddle e t a l , 1995; V ogel et a l , 1995).
C onversely, the transcription factor Engrailed-l (E n -\), w hich is expressed by the ventral ectoderm, must play a ventralising role since E n -\ null mice have a dorsalised limb phenotype (Loom is et a l , 1996). A new candidate gene with a possible role to play in D -V patterning is Radical fringe (R-fng) which is expressed in the dorsal limb ectoderm o f chick embryos (but curiously not in the m ouse) and is repressed by E n -\. In the chick, it seem s that the AER forms precisely at the boundary between cells that express R-fng and those that do not (Laufer et a l , 1997a; Rodriguez-Esteban et a l ,
1997).
Program m ed cell death in the limb
Subsequent to the limb patterning mechanisms described above the limb bud is further remodelled and sculpted by programmed cell death or apoptosis. There are four areas o f programmed cell death in the developing limb varying in size according to the species
concerned (Hinchliffe and Johnson, 1982). These were described before programmed cell death had been fully characterised and were thus confusingly called necrotic zones. The first tw o regions are found on the proximal edges o f the limb bud and are called the anterior and posterior necrotic zones (A N Z and PN Z). These zones are particularly pronounced in the bird wing where there is no digit one and a dramatically reduced digit five. The third defined area o f cell death is found between the tibia and fibula (or the radius and ulna in the forelimb). The fourth zone o f programmed cell death is that in the interdigital w eb regions. In the chick and the m ouse interdigital cell death occurs throughout each interdigit, while in the duck the interdigital cell death is more restricted to the distal margins leading to webbing (Saunders and Fallon, 1967). There is a precise correlation between the appearance o f fragmented internucleosomal D N A and the characteristic morphological signs o f apoptosis in the chick leg bud (Garcia-Martinez et
a /., 1993).
Studies in the chick suggest that catastrophic interdigital apoptosis is triggered by signals from the overlying limb ectoderm since removal o f small patches o f this ectoderm , prior to the onset o f death, blocks apoptosis. Cells appear to default to a cartilage phenotype, since cartilage nodules, and even extra digits, are subsequently seen in the normal interdigital space (Hurle and Ganân, 1986; Hurle et a l , 1989). Similar experim ents were carried out in the duck leg bud and while ectopic cartilages were form ed, they were a different shape and few er extra digits were ultimately produced (Macias et al., 1992). This is consistent with the fact that there is less undifferentiated m esenchym e programmed to die. Prevention o f programmed cell death in the interdigital regions by removal o f ectoderm has been extrapolated to the PNZ. Local removal o f ectoderm results in the failure o f cell death in this region (Brewton and MacCabe, 1988).
W hen the AER is removed from the limb bud prior to the onset o f programmed cell death, about stage 30 in the chick, programmed cell death fails to occur in the interdigital areas and ectopic cartilage is found (Hurle and Ganân, 1986; Hurle and Ganân, 1987).
If the AER is removed after stage 30 then programmed cell death proceeds as normal (Hurle and Ganân, 1986).
B M P s - members o f the TG F-p superfamily - are expressed in the interdigital spaces prior to the initiation o f and during programmed cell death (B M P-2, BM P-4; Francis et al., 1994: BM P-7; Luo et al., 1995). A number o f experiments have investigated the role played by BM Ps in this process, particularly in the light o f the fact that B M P -4 was show n to trigger cell death in the hindbrain (Graham et al., 1994). Beads soaked in B M P -4 and implanted in the interdigital tissue were found to accelerate cell death and digit separation (Ganân et al., 1996). Additionally, when B M P-2 and B M P-4 proteins were added to m esenchym al cells in vitro apoptosis was induced (Yokouchi e ta l . , 1996). Furthermore, if BM P activity is blocked at the level o f the receptor by expressing a dominant negative type I BM P receptor (dnB M PR -lB ) in chick then interdigital cell death is dramatically reduced and webbed feet were formed (Zou and N isw ander, 1996). It w as even suggested that the reason why a duck has webbed feet was because it lacked expression o f B M P-2, B M P-4 and BM P-7 (Zou and N isw ander, 1996), but this statement has since been retracted (Laufer e t a l , 1997b).
Experiments with heparin beads soaked in TG F-p 1 and T G F -p2 inserted into the interdigital tissue resulted in the inhibition o f programmed cell death, with cartilage or extra digits being form ed (Ganân et a l , 1996). Local administration o f FG F-2 and FGF- 4 also prevents apoptosis in the interdigital tissue (Macias et a l , 1996). In the light o f this result and what is know n about FGF expression patterns in the lim b, it has been suggested that interdigital programmed cell death is initiated when AFR function stops (M acias et a l , 1996).
H ow are the d ea d cells cleared?
N ot only do cells die during the footplate sculpting process, but they also have to be cleared away. The efficiency o f this clearance is just as important as the death itself. Previous studies on the remodelling limb in avian and mammalian em bryo, not only describe the distribution o f programmed cell death, but speculate as to how these dying cells are cleared away in embryos. Generally it was felt that neighbouring mesenchymal cells were responsible for engulfing the cellular debris (Saunders and Fallon, 1967; Ballard and H olt, 1968). H ow ever, more recent studies have show n large numbers o f “professional” phagocytes, macrophages, present in the zones o f interdigital cell death from the earliest stages when programmed cell death is observed (H opkinson-W oolley et al., 1994). These cells are clearly sw ollen with apoptotic bodies. Later studies in the Hammertoe mutant m ouse, where there is less cell death, reveal a tight correlation between the numbers o f dead cells and the numbers o f macrophages (Zakeri et al.,
1994). Thesedata suggests that macrophages are “key players”, perhaps the only players, responsible for clearing programmed cell death in the developing limb.