Patterning Systems From One End of the Limb to the Other

Developmental Cell, Vol. 4, 449 458, April, 2003, Copyright 2003 by Cell Press Patterning Systems From One End of the Limb to the Other Review C. Tickle Division of Cell and Developmental Biology School of Life Sciences University of Dundee Dundee DD15EH United Kingdom A combination of embryology and gene identification has led us to the current view of vertebrate limb development, in which a series of three interlocking patterning systems operate sequentially over time. This review describes current understanding of these regu- latory mechanisms and how they form a framework for future analysis of limb patterning. Vertebrate limb development is a powerful model for studying how structures develop in the right places in the embryo and how precise arrangements of differentiated cells and tissues are generated. One reason for this is the rich embryology, accompanied in the last 10 years or so by the accelerating discovery of some of the genes involved (reviewed by Capdevila and Izpisua Belmonte, 2001). However, a new era is looming in which there will be a vast increase in the number of genes that will have to be taken into account. It is therefore critical to hold on to the broad picture so that new information can be digested, synthesized, and interpreted in a robust framework. The progress zone model, one of the longstanding models for limb development, has been chal- lenged recently (Dudley et al. 2002; Sun et al., 2002). In addition to this current controversy, there is increasing evidence that development of the complete vertebrate limb pattern encompasses a series of three interlocking patterning systems, and this might have evolutionary implications. Sonic hedgehog (Shh) signaling is central to patterning in the vertebrate limb bud and recent work on the molecular mechanism provides a new slant on the role of a classical signaling region (Litingtung et al., 2002; te Welscher et al., 2002). Vertebrate Limb Embryology and the Classical View of Pattern Formation Embryological descriptions of vertebrate limb development have documented the appearance of differentiated limb structures in sequence along the long axis (prox- imo-distal axis; shoulder-fingers). In the chick embryo, the first signs of limb development are thickenings of the lateral plate mesoderm around 3 days of incubation (Figure 1A), and these thickenings soon develop into well-defined buds (Figure 1B). As the buds elongate, cells in regions of the bud nearest to the body wall (proximal regions) begin to differentiate. The vertebrate limb skeleton consists of a number of elements that are first laid down as cartilage. Figure 2A is a whole mount of an embryonic chick wing stained to show the cartilage elements that comprise shoulder Correspondence: c.a.tickle@dundee.ac.uk girdle and humerus (upper arm), radius and ulna (lower arm), and wrist and digits (hand). Each digit itself also consists of a series of cartilage elements, typically a metacarpal followed by phalange(s). Cartilage differentiation can be visualized by staining with Alcian Green (Figure 2A) or by S 35 incorporation, and this reveals that proximal skeletal elements, such as humerus in the forelimb and femur in the hindlimb, emerge first, and distal ones, such as digits, emerge later (see for example Hinchliffe, 1977). Cells at the tip of the limb bud remain undifferentiated. As the bud continues to elongate (Figure 1C), more distal skeletal elements differentiate pro- gressively until the complete skeleton is laid down in cartilage and the region of undifferentiated cells at the very tips of the digits disappears. Until recently, the pervading view of limb development was that specification of structures also took place progressively as the buds grow out in a proximo-distal se- quence slightly in advance of cell differentiation. It was proposed that this progressive specification of the dif- ferent parts of the limb along the proximo-distal axis is controlled by a timing mechanism a developmental clock that operates in the zone of undifferentiated mesenchyme that is preserved at the limb bud tip during outgrowth (Summerbell et al., 1973). This region was named the progress zone. The length of time that cells spend in the progress zone specifies proximo-distal in- formation and the positional value acquired as a result would be fixed at the time a cell leaves the zone (Figure 3). Detailed models were drawn up that took account of the fact that, as this progressive specification takes place, the limb is also increasing considerably in size (Figure 1; Lewis, 1975). Recently the progress zone model has been reexam- ined (Dudley et al., 2002), and it has been proposed instead that specification of the different parts of the limb along the proximo-distal axis has already occurred in the early bud at the stage illustrated in Figure 1 and not progressively in a sequence linked to cell differentiation (see also Sun et al., 2002). According to this new view, cells in the early bud have already acquired their proximo-distal positional values and then would remember them through subsequent development (Figure 3), although no mechanism for specifying these values was suggested. Furthermore it was proposed that subse- quent limb bud outgrowth would serve just to expand these specified precursors (Dudley et al., 2002). At some stage during this expansion, cells would become irre- versibly determined in a proximo-distal sequence. There have already been several reviews (Duboule, 2002; Saunders, 2002, Tickle and Wolpert, 2002; Wol- pert, 2002) offering arguments for and against these two rather different models. These will not be rehearsed again here but it is worth bearing in mind that, as yet, there is no direct evidence for early specification of the most distal limb structures. In addition, the early specification model cannot readily account for preferen- tial loss of proximal structures when chick wing buds are irradiated (Wolpert et al., 1979). One would have to assume that a particular phase in the differentiation and/

Developmental Cell 450 Figure 1. Chick Wing Development (A) Presumptive wing bud region marked by slight thickening of lateral plate mesoderm. (B) Wing bud. (C) Elongated wing bud with broad plate at tip in which primordia of digits are forming. All specimens at same magnification. Development from (A) to (C) takes approximately 48 hr. Figure 2. Chick Wing Skeleton and Patterning Systems Involved (A) Whole mount of 10-day-old chick wing stained to show skeletal elements laid down in cartilage. Main regions of the wing indicated below. (B) Interlocking patterning systems that operate sequentially over time to specify the different parts of the limb. LPM, lateral plate mesoderm. or elongation phase of the nascent limb segments is particularly sensitive to cell killing, whereas in the prog- ress zone model such loss would be expected following cell death throughout the limb bud because cells would spend longer in the progress zone in order to repopulate it (Wolpert et al., 1979). In both early specification and progress zone models, limb bud outgrowth is essential for elaborating the pat- tern along the proximo-distal axis. Outgrowth is medi- ated by a set of reciprocal epithelial-mesenchymal inter- actions between the thickened apical ectodermal ridge that rims the distal tip of the limb bud and the underlying mesenchyme (Figure 4). One piece of evidence showing that the apical ridge produces outgrowth signals is that grafts of ridge to the dorsal surface of a second limb bud induce a new outgrowth. Positional information along the proximo-distal axis must be integrated with positional information with respect to the two other axes of the limb, antero-posterior (a-p, thumb-little finger) and dorso-ventral (d-v, back of hand-palm) in order to ensure the proper limb pattern in three dimensions (Wolpert et al., 1975), and two other sets of interactions have been identified in early chick limb buds (Figure 4). Dorso-ventral patterning of the limb is controlled by signaling of the ectoderm covering the sides of the bud (Figure 4). Thus the supernumerary limb that develops following a graft of the apical ridge to the dorsal surface of a limb bud has a double-dorsal pattern. Position along the antero-posterior axis is specified by a signaling interaction between the polarizing region, a region of mesenchyme at the posterior margin of the bud and the adjacent mesenchyme cells at the bud tip (Figure 4). Signaling molecules associated with each of these three sets of interactions have now been identified: fi- broblast growth factors with Apical ectodermal ridge, Wnt-7a with dorsal ectoderm and retinoic acid, Sonic hedgehog, and, more controversially, Bone morphogenetic protein-2 with the polarizing region (Figure 4). Genes encoding various transcription factors that might register resultant positional values have also been identified, including genes from Hoxa and Hoxd complexes, the Tbx family, and Lmx1. One possible extension of the early specification model is that cells in the limb bud have already acquired not only proximo-distal positional values but also antero-posterior and dorso-ventral positional values. At the other extreme, building on the progress zone model, one could envisage that cells acquire all their positional values over time as the limb bud grows out. There are problems with both these possibilities; with respect to the first, it is not clear whether one could establish all the necessary positional values for the detailed limb pattern with the small number of cells available; with the second, the limb will be too large to allow, for example, specification of antero-posterior values across the entire limb while the digits are forming (see Figure 1). Therefore it seems likely that positional information along all three axes is not specified at the same time and even that different mechanisms may operate at different times during limb development. Indeed, one could view limb development as consisting of three sequential episodes: the first taking place in lateral plate mesoderm and being part of the antero-posterior patterning process along the main body axis, resulting in specification of the proximal part of the limb, e.g., shoulder girdle and humerus; the

Review 451 Figure 3. Progress Zone Model Contrasted with Early Specification Model Numbers represent positional values along the proximo-distal axis. In progress zone model, these values are generated as cells leave the progress zone (shaded) at the tip of the limb bud. In early specification model, these values have already been assigned in the early limb bud. second taking place within the limb bud equipped with limb signaling region first identified by Saunders (Saunits own signaling center, resulting in specification of the ders and Gasseling, 1968) in the posterior margin of lower limb, e.g., radius and ulna, establishment of digit the early chick wing bud, can produce antero-posterior number and positional information for digit identity; and pattern changes in distal structures when grafted to the third and final episode involving digit primordia, re- another host wing bud. Both mirror-image duplicated sulting in specification of the different parts of the digits digit patterns and fore-arm pattern changes result (Figure 2B). sometimes three bones form in the forearm or the forearm is symmetrical with two ulnae. In contrast, in an Patterning of the Proximal Part of the Limb extensive series of grafting experiments at a range of in Lateral Plate Mesoderm different stages, Wolpert and Hornbruch (1987) never There are several lines of evidence that specification of obtained duplications of the humerus and they sugthe proximal part of the limb (shoulder girdle and hupart gested that the polarizing region does not pattern this merus in forelimbs, pelvic girdle and femur in hindlimbs) of the limb. should be considered as a separate process. Classical The significance of these embryological observations fate maps of the chick wing bud (Saunders, 1948) has recently been thrown into relief by detailed examina- showed that the humerus, unlike the bones in the lower tion of the limb phenotype of Shh / mouse embryos. arm and eventually the digits, develops at the base of Sonic hedgehog (Shh) is expressed in the polarizing the bud almost parallel to the long axis of the body with region (Riddle et al., 1993) and is almost certainly the its proximal end near the anterior of the limb bud and molecule responsible for polarizing activity (the ability its distal end at the posterior, and this has been convery to induce additional digits). While distal structures are firmed in more recent studies (Vargesson et al., 1997). reduced in limb buds of Shh / mouse embryos, Results of grafting experiments have also highlighted proximal structures shoulder girdle and humerus and the fact that the humerus may be patterned in a different pelvic girdle and femur develop relatively normally way to more distal bones. The chick wing polarizing (Chiang et al., 2001; Kraus et al., 2001; Lewis et al., region or zone of polarizing activity (ZPA), the classical 2001). Furthermore, in Shh / embryos, the humerus has been described as being trapped in the body wall (Kraus et al., 2001). Recently a chicken mutant, oligozeugodactyly (ozd), has been described in which distal limb development is very impaired and there is no detectable Shh expression in the limb buds (Ros et al., 2003). These observations suggest that Shh signaling is required for proper patterning of only distal limb. All of the above suggest the possibility that the proximal-distal axis of the shoulder girdle and humerus/pelvic girdle and femur is initially specified in lateral plate mesoderm as part of the process of antero-posterior patterning of the main body axis independently of Shh signaling. Interestingly, comparative studies on dogfish fin development (Tanaka et al., 2002) raise the intriguing possibility that separate patterning of proximal limb might have evolutionary significance. Shh expression in developing fin buds of dogfish embryos could not be Figure 4. Classical View of Vertebrate Limb Patterning in Relation detected, suggesting that cartilaginous fish might not to the Three Main Axes of the Limb yet have acquired a limb bud signaling system. Interest- Upper left, dorsal view of early limb bud indicating apical ectodermal ingly, the ozd chicken mutant lacks Shh expression only ridge and polarizing region and associated signaling molecules. in the limb buds (Ros et al., 2003) and a limb-bud-spe- Dotted line shows plane of section through limb bud. Section viewed cific Shh regulatory element has been identified recently bottom left; arrows from ectoderm represent signaling from ectoin the Sasquatch mouse mutant (Lettice et al., 2002). derm to mesenchyme. Dorsal ectoderm and associated signal indicated. Upper right indicates the three main limb axes. Pr, proximal; There is now substantial evidence that retinoic acid D, Distal; A, Anterior; P, posterior. is involved in patterning the proximal part of the limb

Developmental Cell 452 Figure 5. Series of Patterning Systems that Operate Sequentially Over Time with Associated Expression Patterns of Genes Encoding Signaling Molecules throughout Vertebrate Limb Development P-D, proximo-distal; A-P, antero-posterior; D-V, dorso-ventral. anterior limb mesenchyme (Riddle et al., 1993). Retinoic acid has now also been shown to play a role in establishing the polarizing region during normal limb development in both chick and mouse embryos. Treatment of presumptive limb forming regions of the lateral plate mesoderm of chick embryos with inhibitors of retinoic acid synthesis or with synthetic molecules that act as retinoic acid receptor antagonists showed that retinoic acid is essential for establishing the polarizing region in the posterior of the developing limb bud and inducing Shh expression (Stratford et al., 1996; Helms et al., 1996). More recently, a similar conclusion has been reached for mouse forelimbs from analysis of retinoic acid-rescued Raldh2 / embryos (Niederreither et al., 2002). Shh comes to be expressed at the posterior margin of the limb bud because factor(s) such as D-Hand that control the response of cells to retinoic acid signaling are posteriorly restricted (Charité et al., 2000; Fernandez-Teran et al., 2000). Thus retinoic acid signaling not only specifies proximal limb but also provides a link to the next phase of patterning that takes place in the limb bud. (Figure 5). The presence of retinoic acid in presumptive limb bud regions and at the base of limb buds was demonstrated some time ago in transgenic mice in which the RAR retinoic acid responsive promoter drives LacZ (Rossant et al., 1991). More recently, work Patterning of Lower Limb and Establishment on chick embryos showed that Raldh2, the gene that of Digit Pattern in the Limb Bud encodes retinaldehyde dehydrogenase 2, the enzyme The phenotype of Shh / mouse embryo limb buds that oxidizes retinal to retinoic acid, is expressed in shows that Shh signaling of the polarizing region in the lateral plate and somatic trunk mesoderm just prior to limb bud is critical for proper development of distal parts limb bud formation, while Cyp26, a gene that encodes of the limb. In absence of Shh signaling, structures distal a P450 enzyme that metabolizes retinoic acid, is ex- to the elbow/knee are very reduced, with at best a single pressed slightly later than Raldh2 in ectoderm overlying rudimentary digit developing in the leg (Chiang et al., the limb forming region and persists in distal ectoderm 2001; Kraus et al., 2001; Lewis et al., 2001). There has as the limb bud grows out (Swindell et al., 1999). been increasing support in recent years for the idea that The importance of retinoic acid signaling in early limb Shh signaling fulfills two roles in the limb bud bud development has recently been demonstrated by specification of digit number and specification of antero- elegant experiments with Raldh2 / mouse embryos. In posterior position and hence digit identity. Data from these embryos there is no sign of limb development but two recent papers reinforce this idea (Litingtung et al., it is not clear whether this is due to the fact that the 2002; te Welscher et al., 2002) and build on the discovery embryos die early. However, retinoic acid supplementation that there is a gradient of repressor and activator forms of the maternal diet rescues mutant forelimb devel- of Gli3 (a bifunctional transcription factor that is essen- opment in a dose- and time-dependent fashion, showing tial for mediating hedghehog signaling; Méthot and that there is a specific time window during which retinoic Basler, 2001) in the limb bud (Wang et al., 2000; Figure acid is required (Niederreither et al., 2002). 6). Shh prevents processing of Gli3 to its repressor form Likely targets of retinoic acid in the proximal part of and diffusion of Shh from the polarizing region sets up the limb are the Meis genes, and experiments in chick a concentration gradient across the bud. This results embryos suggest that Meis 2 may play a role in develop- in dose-dependent relief of inhibition imposed by high ment of proximal limb structures (Mercader et al., 1999; levels of the repressor form of Gli3 and thus allows Capdevila et al., 1999). Meis 2 was first identified as a expression of target genes, which execute its two downstream retinoic acid-inducible gene and both Meis 1 and 2 are functions, leading to development of a patterned expressed in lateral plate mesoderm and are later ex- series of digits from the posterior part of the limb bud pressed throughout early limb buds. As the bud continues (Figure 6). to grow out, Meis expression recedes from the bud The data that lead to this conclusion came from com- tip and becomes confined to a proximal domain due to paring consequences of functionally inactivating Shh, repression by Fgf signaling from the apical ridge (Mer- Gli3, and both Shh and Gli3 together. Gli3 / mutant cader et al., 2000). This domain of Meis expression is limbs were already known to have an increased number associated with the region of the bud that gives rise to of digits of uniform morphology and this was thought structures proximal to the elbow/knee. When Meis-1 to be due to an ectopic patch of Shh expression at the was expressed throughout chick limb buds, this had a anterior margin of the mutant limb buds. But in proximalizing effect although mainly more distal parts Gli3 / Shh / double mutant embryos, limbs with many of the limb pattern were affected (Mercarder et al., 1999; digits with undefined antero-posterior pattern still develop. Capdevila et al., 1999). These findings can be neatly explained by differ- Retinoic acid was the first signaling molecule with ences detected in the balance of the repressor and activator polarizing activity to be identified (Tickle et al., 1982) forms of Gli3 (Litingtung et al., 2002). In the normal and this activity is due to induction of Shh expression in limb, Shh prevents processing of Gli3 to its repressor

Review 453 specify digit number (reviewed by Saunders, 1977). This factor has now been shown to be the Bmp antagonist, gremlin (Zuniga et al., 1999), whose expression is inhibited by high levels of the repressor form of Gli3 (Litingtung et al., 2002; te Welscher et al., 2002). It seems most likely that gremlin acts by antagonizing Bmp signaling in the ridge, thus allowing expression of Fgf4 and thus ridge function to be maintained (see later). Digit identity, the type of digit that develops in a particular position, is related to the strength of Shh signaling. Previous experiments in both chick and mouse limb buds showed that specification of the most posterior digits requires the highest levels of Shh signaling. Thus application of high concentrations of Shh protein to the anterior margin of chick wing buds is necessary to induce the most posterior digit, digit 4 (Yang et al., 1997), while in mouse embryos the first digit to be lost when Shh signaling in the endogenous polarizing region is attenuated is the most posterior digit (Lewis et al., 2001). Furthermore, since the effects of Shh application to chick wing buds are not only dose dependent but also time dependent, these observations suggest that, during normal limb bud development, the positional value of cells exposed to Shh is progressively posteriorized. This fits with the idea that the rudimentary anterior digit formed in Shh / hind limb buds (and in the ozd chicken mutant) develops from cells that have not been posteriorized (Figure 6). In terms of the recent data (Wang et al., 2000; Litingtung et al., 2002), posteriorization would require establishment of the activator Gli3:repressor Gli3 gradient. There is some tantalizing evidence from experiments on chick wing buds that the Shh signal itself might not directly specify digit identity but act via Bmps, in particu- lar Bmp2, which is expressed posteriorly in normal buds and can be induced anteriorly in response to Shh application (Drossopoulou et al., 2000). Application of Bmp-2 protein can promote posteriorization of digits in duplica- tions initiated by a short exposure of anterior mesenchyme to Shh signaling (Drossopoulou et al., 2000). This can now be interpreted in terms of the status of Gli3. It seems likely that the short treatment with Shh is sufficient to inhibit processing of Gli3 to the repressor form and set up a gradient of activator Gli3:repressor Gli3 in anterior cells on which Bmp signals could subse- quently act. One of the targets of polarizing region signaling in the limb bud is Tbx3. Tbx3 is a transcription factor of Tbox family and is related to the product of the Drosophila gene, optomotor blind (omb). In addition, Tbx3 haploin- sufficency has been linked with the human condition ulnar-mammary syndrome, in which predominantly pos- terior defects occur in the upper limb. Tbx3 is expressed both posteriorly and anteriorly in vertebrate limb buds. Recent analysis in both chick and mouse shows that posterior Tbx3 expression is positively regulated by Shh, and also by Bmp2 in chick limb buds, mirroring a signal- ing cascade in Drosophila wing development that regu- lates omb expression (Tumpel et al., 2002). This study also revealed that anterior Tbx3 expression in the chick limb bud is under the control of Bmp4 signaling and negatively regulated by Shh. Thus in the limb bud the posterior Shh signaling system seems to be opposed by an anterior Bmp4 signaling system and the same Figure 6. Gli3 Processing in Normal and Mutant Limb Buds (A) Processing of Gli3 in normal limb development. Repressor form of Gli3 is present anteriorly, represses anterior expression of Fgf4, gremlin, and Hoxd13 and development of digits. Shh posteriorly prevents processing of Gli3 to repressor form, thus allowing posterior expression of Fgf4, gremlin, and Hoxd13 leading to development of a polarized set of digits from posterior region of the limb bud. (B) Gli3 processing in Shh / limb bud. Repressor form of Gli3 pre- dominates throughout the limb bud. Transitory and weak expression of Fgf4, gremlin, and Hoxd13 at very posterior margin of the limb leads to formation of a single anterior digit in leg. (C) Gli3 processing in Shh / Gli3 / limb buds. In absence of Gli3, expression of Fgf4, gremlin, and Hoxd13 extends anteriorly and morphologically identical digits form across the limb bud. form in a concentration-dependent fashion, thus establishing a gradient of Gli3 activity in the posterior region of the limb; with a high ratio of activator Gli3:repressor Gli3 at the very posterior with the balance progressively tipping to a low ratio of activator Gli3:repressor of Gli3 more anteriorly. In the absence of Shh, Gli3 is processed to its repressor form throughout the limb and this shuts down distal development more or less completely (Figure 6). This contrasts with the normal limb bud, in which high levels of the repressor form of Gli3 shut down just the anterior half. The downstream consequences of Shh signaling are specification of digit number and specification of anteroposterior position leading to digit identity and target genes involved have been identified. Digit number is related to the bud width, which is determined by the length of the apical ectodermal ridge and the length of the apical ridge is controlled by the mesenchyme. It was postulated for a long time that the polarizing region must produce an apical ridge maintenance factor in order to

Developmental Cell 454 opposing set of signals operates in dorso-ventral pat- ventralizes the forearm muscle pattern, suggesting that terning of the neural tube. Lmx1 acts as a transcriptional activator in patterning The polarizing region and Shh signaling persist until this region of the limb (Rodriguez-Esteban et al., 1998). digit primordia appear. By this time, the posterior part One interpretation of the very restricted change in dorso/ of the limb, which has been under the influence of Shh, ventral patterning in Wnt7a / mouse limbs is that other has expanded to occupy the entire width of the limb signals maintain Lmx-1 expression in dorsal mesoderm bud. Bmps expressed in interdigital mesenchyme be- in proximal limb regions. tween digit primordia (Laufer et al., 1997) seem likely to Wnt7a signaling in limb bud stages also seems to be the legacy of Shh signaling, which will then be used fulfill another role in maintaining Shh expression and the in the final patterning process that takes place in each limbs of Wnt7a / mice often lack the most posterior digit primordium (Figures 2B and 5). Apical Ridge Signaling throughout Development of the Limb The apical ridge is present throughout all the phases of limb development. It is established in the ectoderm over presumptive limb lateral plate mesoderm, maintained during limb bud stages, and persists until late into devel- opment over the tips of the digits (Rubin and Saunders, 3, toe 3, 4, and toe 4, 5. Manipulations at this late stage 1972). Classical experiments showed that removal of of development, including insertion of barriers to bisect the apical ridge at any stage during limb development digital primordial, removal of interdigital mesenchyme, produces a truncation and the level of the truncation and grafts of digit primordia into different interdigital depends on the time at which the ridge is removed spaces, can produce digits with either increased num- (Saunders, 1948; Summerbell, 1974). Part of the recent bers of phalanges or with fewer phalanges suggesting controversy about the mechanism of proximo-distal patthat digit primordia can be anteriorized or posteriorized. terning is whether these truncations are the result of This contrasts with the rules operating in limb bud pat- cell death or to failure to maintain the progress zone terning, in which, under the influence of the polarizing (Dudley et al., 2002; reviewed Duboule 2002), and this region and Shh signaling, digits can only be posterior- reawakens an old debate (reviewed by Saunders, 1977). ized. Dahn and Fallon (2000) have suggested, based on Irrespective of these different interpretations, the rethe fact that application of noggin, a Bmp antagonist, moval experiments show that the ridge is required for can alter digit morphogenesis, that levels of local Bmp proper development of all parts of the limb. signaling control digit primordia development, but it is It is well established that outgrowth signals from the not yet clear exactly how this is accomplished and what apical ridge comprise members of Fgf family and genes gene targets are involved. encoding several different Fgfs are expressed in complex dynamic patterns during limb development. Initially, Dorso-Ventral Patterning along the Limb the apical ridge over lateral plate mesoderm expresses Patterning of Individual Digits in Digit Primordia Evidence that each digit primordium has its own patterning mechanism comes from recent work mostly on chick leg development (Dahn and Fallon, 2000). This showed that each individual digital ray is under the influence of immediately adjacent tissues and that its subsequent morphology can be altered independently of the other rays. Each toe of the chick leg has a different number of phalanges: toe 1, having 2 phalanges, toe 2, digit (Parr and McMahon 1995). This role in maintaining Shh expression may explain why Wnt-7a is expressed so early in limb bud development. There is rather little information about how dorso-ventral Fgf8, and then later in limb buds, Fgf4, Fgf9, and Fgf17 patterning is accomplished in presumptive limb region are expressed just in the posterior part of the ridge of lateral plate mesoderm, in the limb bud, and finally in (reviewed Martin, 1998). Still later when the digit primordia digit primordia, except that Wnt7a acts as a dorsalising are forming, Fgf4 expression disappears from the factor (Figure 5). Wnt7a expression is initiated very early ridge in both chick and mouse limb buds (Duprez et al., in ectoderm overlying presumptive limbs even before 1996; Niswander and Martin, 1992) but Fgf8 expression the earliest signs of apical ridge formation (Altabef and persists in the ridge rimming the tips of the primordia. Tickle, 2002) and is apparently maintained in dorsal ectoderm A recent paper surveyed expression of a large number throughout limb bud development and into digi- of Fgf genes in developing limbs and describes Fgf9 tal plate stages (Figure 5). Yet, surprisingly, in Wnt7a / expression in the apical ridge persisting at later stages mice, dorso-ventral pattern changes are only detected than Fgf4 (Hajihosseini and Heath, 2002). Thus different in the paws (Parr and McMahon, 1995). phases of limb development are associated with expres- There is evidence that Wnt7a signaling in dorsal ecto- sion of different Fgfs (Figure 5). derm controls expression of Lmx-1, a gene that encodes Rather little attention has been paid to the mecha- a transcription factor, in underlying dorsal mesenchyme, nisms that determine the initial length of the apical ridge but there are conflicting data about whether Wnt7a and/ that expresses Fgf8 in the pre-bud stage. Recently it or dorsal ectoderm is necessary to maintain Lmx1 ex- has been shown that in Pitx1 / and Pitx1 / Pitx2 / pression (Vogel et al., 1995; Riddle et al., 1995). This mouse embryos the early ridge is very short (Marcil et issue is relevant to whether dorso-ventral positional val- al., 2003). Pitx1 is thought to play a role in specifying ues are specified in the early bud and then remembered hindlimb identity, while Pitx2 is implicated in specifying or whether dorso-ventral specification is ongoing through- left-right asymmetry of lateral plate mesoderm. Interestingly, out development. Expression of a construct containing in the double mutants, in which the ridge is shorter, the engrailed repressor domain fused to the Lmx1 ho- proximal limb structures fail to develop. This essential meodomain in dorsal mesenchyme of chick wing buds role of the early ridge in development of these proximal

Review 455 structures, which, as outlined earlier, arise parallel to the apical ridge and that these levels specify proximodistal the main body axis may help to explain the puzzle of positional value. Yet another hypothesis is that an why the limb bud is so large even though only the poste- Fgf gradient produced by Fgf diffusing from the apical rior part gives rise to distal structures (Figure 6). ridge specifies position along the proximo-distal axis of One of the striking features of the patterns of expres- the limb bud (Papageorgiou and Almirantis, 1996). Such sion of the various Fgfs is that Fgf4, Fgf9, and Fgf17, in a gradient could perhaps operate in the early specifica- particular Fgf4, are expressed predominantly in the limb tion model. The gradient hypothesis has been tested bud patterning stage and furthermore only in the posterior with respect to establishing Hoxa gene expression pat- part of the ridge. It was discovered several years ago terns in chick wing buds, and there is evidence that Fgf that Shh normally maintains Fgf4 expression in posterior initiates and maintains Hoxa13 gene expression in the ridge and in turn that Fgf4 maintains Shh expression absence of the ridge (Vargesson et al., 2001). However, posterior mesenchyme (Laufer et al., 1994; Niswander application of Fgf to the intact bud altered neither timing et al., 1994). This property of Fgf4 may also be shared nor extent of Hoxa gene expression as would be predicted with Fgf9 and Fgf17. Interestingly, as mentioned earlier, from the model, suggesting that other factors are fate maps show that digit primordia come from posterior involved. limb bud (Saunders, 1948; Bowen et al., 1989; Vargesson A recent model for somite formation purposes that a et al., 1997) where Fgf4 together with Fgf9 and Fgf17 is clock mechanism producing oscillations interacts with expressed. Furthermore apical ridge fate maps show a gradient in Fgf8 signaling at a threshold level that that the posterior ridge of early buds, the part of the constitutes a wave front of activation (Dubrulle et al. ridge expressing Fgf4, expands to make up the entire 2001; reviewed by Varsiliauskas and Stern, 2001). Simi- ridge by the time the digit primordia are being laid down liarly, in the limb both an Fgf gradient and oscillations (Vargesson et al., 1997). Therefore it seems likely that it may be involved and operate together and/or at different is the length of the Fgf4-expressing ridge that deter- times. mines the number of digit primordia that form. Interestingly, the recent studies, on Gli3 / and Shh / Gli3 / Integrated Positional Information mutant limb buds show that Fgf4 is expressed through- Combinations of Fgf signaling with retinoic acid, Shh, out the ridge of early limb buds rather than being posteri- and Bmps have all been shown to induce Hox gene orly restricted and this correlates with the increased expression changes in developing chick wing buds (Dunumber of digits that develop (Litingtung et al., 2002; te prez et al., 1996). In somites, the number of oscillations Welscher et al., 2002). that cells experience under the influence of Fgf before The functional significance of these patterns of Fgf they form a somite has been suggested to be connected expression has been dissected by inactivation of indi- to activation of Hox gene expression and thus provide vidual Fgf genes, some of them conditionally for exam- a measure of position along the main body axis (Dubrulle ple, Fgf4 and Fgf8 because these Fgfs are required et al., 2001; Zakany et al., 2001). early: some double knockouts have also been made. As There is now overwhelming evidence that 5 genes in might be expected from the overlapping patterns of Hoxa (Hoxa9 through Hox a13) and Hoxd (Hoxd9 through expression, the interpretation is complicated by functional Hox d13) clusters play crucial roles in limb development, redundancy, and when either Fgf4 or Fgf9 or Fgf17 with Hoxa13 and d13 being responsible for digital devel- is inactivated in the ridge, normal limbs result (see Sun opment and Hoxa11 and Hoxd11, radius and ulna/fibula et al., 2002). In contrast, when Fgf8 is conditionally inac- and tibia. The 5 Hoxd genes are first expressed very tivated from presumptive limb bud stages onward (see early in lateral plate mesoderm, while the 5 Hoxa genes for example Lewandoski et al., 2000) both proximal and are activated in sequence beginning in lateral plate distal deletions occur with proximal structures being mesoderm and extending into the bud stage; 5 genes most affected. When double knockouts are made in in both clusters are also expressed at later stages during which both Fgf8 and Fgf4 are inactivated throughout patterning of digit primordia. This early activation of limb bud development, no limb structures develop, while 5 Hox gene expression could be considered to be a when Fgf8 and Fgf4 are inactivated from limb bud stages consequence of early specification, as suggested in the onward the limbs have normal proximal structures but recent model. Hoxa13, however, seems to be expressed lack some elements in the digits. It has been suggested too late to fit this idea. It is also clear that all the cells that the various limb phenotypes obtained in these Fgf- at the tip of the early limb bud expressing Hoxd13 do deficient mice can be more readily interpreted in terms not enter digit primordia but are left behind as the limb of an early specification model (Sun et al., 2002). grows out and stop expressing Hoxd13. This would only So how is Fgf signaling related to specification of fit with the early specification model by assuming that structures along the proximo-distal axis? According to as cells leave the influence of Fgf signaling they become the progress zone model, Fgf signaling in the elongating proximalized. limb bud and in digit primordia could be related to a A careful and comprehensive study by Tabin and colleagues developmental clock perhaps based on oscillations. (Nelson et al., 1996) distinguished three different When cells leave the influence of Fgf signaling, their phases of Hox gene expression during limb development. positional value would become fixed according to the In their analysis, they emphasized the corresponpositional number of oscillations they have experienced. A second dence of these phases of expression with specification hypothesis suggested by a bootstrap model (Meinhardt, of each of the proximo-distal limb segments (Nelson et 1983) is that continuing interactions between mesen- al., 1996). However, it is striking that the dramatic chyme and apical ridge throughout limb bud outgrowth changes in Hox gene expression that they describe cor- lead to increasing levels of Fgf production over time by respond more closely to the transitions discussed here,

Developmental Cell 456 i.e., the transition between patterning in lateral plate Acknowledgments mesoderm and in the limb bud and the transition be- I thank Angie Blake for her help in preparing this manuscript and tween patterning in the limb bud and in digit primordia. Figures 2, 5, and 6, Kevin Beattie for Figure 1, Nick Cole for help Interestingly, studies on zebrafish fin development with Figures 3 and 4, and Andy Bain for help with references. I thank showed that the phase of Hox gene expression associ- members of the lab and J.P. Couzo for stimulating discussions. My ated with the hand of higher vertebrates is not seen own research on limb pattern formation is supported by the MRC in teleost fin development (Sordino et al., 1995). One and The Royal Society. speculation is that the absence of this late phase in Hox gene expression in zebrafish reflects the fact that the References hand is unique to tetrapods. Detailed dissection of the Altabef, M., and Tickle, C. (2002). Initiation of dorso-ventral axis regulation of expression of the four most 5 genes in during chick limb development. Mech. Dev. 116, 19 27. the Hoxd cluster and the neighboring gene Evx1 in mice Bowen, J., Hinchliffe, J.R., Horder, T.J., and Reeve, A.M. (1989). The has identified an element located outside the Hoxd clus- fate map of the chick forelimb-bud and its bearing on hypothesized ter that drives expression specifically in the hand and developmental control mechanisms. Anat. Embryol. (Berl.) 179, recent work has given insights into the functional signifi- 269 283. cance of the organization of the complex (Kmita et al., Campbell, G. (2002). Distalization of the Drosophila leg by graded 2002). EGF-receptor activity. Nature 418, 781 785. Capdevila, J., and Izpisua Belmonte, J.C. (2001). Patterning mechanisms Conclusions controlling vertebrate limb development. Annu. Rev. Cell Dev. The view of limb development put forward here high- Biol. 17, 87 132. lights a sequence of patterning systems in lateral plate Capdevila, J., Tsukui, T., Rodriquez Esteban, C., Zappavigna, V., and Izpisua Belmonte, J.C. (1999). Control of vertebrate limb outmesoderm, limb bud, and digit primordium. An imporgrowth by the proximal factor Meis2 and distal antagonism of BMPs tant feature is the interaction between signals in succes- by Gremlin. Mol. Cell 4, 839 849. sive patterning systems so that the systems interlock Charité, J., McFaqdden, D.G., and Olson, E.N. (2000). The bhlh and information flows from one system to the next. Pat- transcription factor dhand controls Sonic hedgehog expression terning of proximal structures could differ significantly and establishment of the zone of polarizing activity during limb from that of more distal structures occurring with respect development. Development 127, 2461 2470. to the axes of the body rather than the axes of Chiang, C., Litingtung, Y., Harris, M.P., Simandl, B.K., Li, Y., Beachy, the limb bud. Furthermore, the patterning of indivdual P.A., and Fallon, J.F. (2001). Manifestation of the limb prepattern: digit primordia might be a separate process. All of these limb development in the absence of sonic hedgehog function. Dev. considerations need to be taken into account in models Biol. 236, 421 435. of how the limb pattern is laid down and eventually Dahn, R.D., and Fallon, J.F. (2000). Interdigital regulation of digit identity and homeotic transformation by modulated BMP signaling. translated into anatomy. Science 289, 438 441. Recent studies of how the different segments of the Drossopoulou, G., Lewis, K.E., Sanz-Ezquerro, J.J., Nikbakht, N., Drosophila leg are formed (Campbell, 2002; Galindo et McMahon, A.P., Hofmann, C., and Tickle, C. (2000). A model for al., 2002) suggest parallels with vertebrate limb develop- anteroposterior patterning of the vertebrate limb based on sequenment. Furthermore, the patterns of gene expression in tial long- and short-range Shh signalling and Bmp signalling. Development third instar leg discs of Drosophila are similar to those 127, 1337 1348. in fully segmented cricket leg buds (Inoue et al., 2002). Duboule, D. (2002). Making progress with limb models. Nature 418, One possible parallel between insects and vertebrates 492 493. is the finding that, in the Drosophila leg, as suggested Dubrulle, J., McGrew, M.J., and Pourquie, O. (2001). FGF signaling here for vertebrate limbs, there is a separate mechanism controls somite boundary position and regulates segmentation for patterning the most distal part, the tarsus. Further- clock control of spatiotemporal Hox gene activation. Cell 106, more this mechanism is based on graded signaling via 219 232. Dudley, A.T., Ros, M.A., and Tabin, C.J. (2002). A re-examination of the epidermal growth factor receptor, a receptor tyroproximodistal patterning during vertebrate limb development. Nasine kinase receptor that is the same class of receptor ture 418, 539 544. that Fgfs activate. Another striking feature of Drosophila Duprez, D., Kostakopoulou, K., Francis-West, P.H., Tickle, C., and leg development is that proximo-distal extension is fol- Brickell, P.M. (1996). Activation of Fgf-4 and HoxD gene expression lowed by proximal restriction of expression of genes by BMP-2 expressing cells in the developing chick limb. Developthat were previously expressed more widely. A possible ment 122, 1821 1828. example of this in vertebrate limb development is the Fernandez-Teran, M., Piedra, M.E., Kathiriya, I.S., Srivastava, D., proximal restriction of the domain of Meis gene expresanterior-posterior Rodriguez Rey, J.C., and Ros, M.A. (2000). Role of dhand in the sion from the distal region of the limb bud during elongation. polarization of the limb bud: implications for the Finally, in Drosophila leg development, the tip, Sonic hedgehog pathway. Development 127, 2133 2142. which expresses Distal-less, is specified very early. Verpatterning Galindo, M.I., Bishop, S.A., Greig, S., and Couso, J.P. (2002). Leg driven by proximal-distal interactions and EGFR signal- tebrate Distal-less genes are expressed in the apical ing. Science 297, 256 259. ectodermal ridge of vertebrate limb buds and recently Hajihosseini, M.K., and Heath, J.K. (2002). Expression patterns of digit abnormalities were reported in Dlx5 / Dlx6 / fibroblast growth factors-18 and -20 in mouse embryos is suggestive mouse embryos (Robledo et al., 2002). It is not clear of novel roles in calvarial and limb development. Mech. Dev. 113, how far one can push these similarities but the emerging 79 83. knowledge about insect leg development may suggest Helms, J.A., Kim, C.H., Eichele, G., and Thaller, C. (1996). Retinoic some new avenues for exploring vertebrate limb devel- acid signaling is required during early chick limb development. Development opment. 122, 1385 1394.

Review 457 Hinchliffe, J.R. (1977). The chrondogenic pattern in chick limb mor- feedback loop coordinates growth and patterning in the limb. Nature phogenesis: a problem of development and evolution. In Vertebrate 371, 609 612. Limb and Somite Morphogenesis. D.A. Ede, J.R. Hinchliffe, and M.J. Papageorgiou, S., and Almirantis, Y. (1996). Gradient model de- Balls, eds. (Cambridge: Cambridge University Press), pp. 293 310. scribes the spatial-temporal expression pattern of Hoxa genes in Inoue, Y., Mito, T., Miyawaki, K., Matsushima, K., Shinmyo, Y., the developing vertebrate limb. Dev. Dyn. 207, 461 469. Heanue, T.A., Maron, G., Ohuchi, H., and Noji, S. (2002). Correlation Parr, B., and McMahon, A. (1995). Dorsalizing signal Wnt-7a required of expression patterns of homothorax, dachshund and Distal-less for normal polarity of D-V and A-P axes of mouse limb. Nature 374, with the proximodistal segmentation of the cricket leg bud. Mech. 350 353. Dev. 113, 141 148. Riddle, R.D., Johnson, R.L., Laufer, E., and Tabin, C. (1993). Sonic Kmita, M., Fraudeau, N., Herault, Y., and Duboule, D. (2002). Serial hedgehog mediates the polarizing activity of the ZPA. Cell 75, 1401 deletions and duplications suggest a mechanism for the collinearity 1416. of Hoxd genes in limbs. Nature 420, 145 150. Riddle, R.D., Ensini, M., Nelson, C., Tsuchida, T., Jessell, T.M., and Kraus, P., Fraidenraich, D., and Loomis, C.A. (2001). Some distal Tabin, C. (1995). Induction of the LIM homeobox gene Lmx1 by limb structures develop in mice lacking Sonic hedgehog signaling. WNT7a establishes dorsoventral pattern in the vertebrate limb. Cell Mech. Dev. 100, 45 58. 83, 631 640. Laufer, E., Nelson, C.E., Johnson, R.L., Morgan, B.A., and Tabin, C. Robledo, R., Rajan, L., Li, X., and Lufkin, T. (2002). The Dlx5 and (1994). Sonic hedgehog and Fgf-4 act through a signaling cascade Dlx6 homoebox genes are essential for carniofacial, axial, and apand feedback loop to integrate growth and patterning of the develpendicular skeletal development. Genes Dev. 16, 1089 1101. oping limb bud. Cell 79, 993 1003. Rodriguez-Esteban, C., Schwabe, J.W., Pena, J.D., Rincon-Limas, Laufer, E., Pizette, S., Zou, H., Orozco, O.E., and Niswander, L. D.E., Magallon, J., Botas, J., and Belmonte, J.C. (1998). Lhx2, a (1997). BMP expression in duck interdigital webbing: a re-analysis. vertebrate homologue of apterous, regulates vertebrate limb out- Science 278, 305. growth. Development 125, 3925 3934. Lettice, L.A., Horikoshi, T., Heaney, S.J., van Baren, M.J., van der Ros, M.A., Dahn, R.D., Fernandez-Teran, M., Rashka, K., Caruccio, Linde, H.C., Breedveld, G.J., Joosse, M., Akarsu, N., Oostra, B.A., N.C., Hasso, S.M., Bitgood, J., Lancman, J.J., and Fallon, J.F. (2003). Endo, N., et al. (2002). Disruption of a long-range cis-acting regulator The chick oligozeugodactyly (ozd) mutant lacks sonic hedgehog for Shh causes preaxial polydactyly. Proc. Natl. Acad. Sci. USA 99, function in the limb. Development 130, 527 537. 7548 7553. Rossant, J., Zirngibl, R., Cado, D., Shago, M., and Giguere, V. (1991). Lewandoski, M., Sun, X., and Martin, G.R. (2000). Fgf8 signalling Expression of a retinoic acid response element-hsplacz transgene from the AER is essential for normal limb development. Nat. Genet. defines specific domains of transcriptional activity during mouse 26, 460 463. embryogenesis. Genes Dev. 5, 1333 1344. Lewis, J.H. (1975). Fate maps and the patterns of cell division: a calculation for the chick wing bud. J. Embryol. Exp. Morphol. 33, Rubin, L., and Saunders, J.W., Jr. (1972). Ectodermal-mesodermal 419 434. interactions in the growth of limb buds in the chick embryo: constancy and temporal limits of the ectodermal induction. Dev. Biol. Lewis, P.M., Dunn, M.P., McMahon, J.A., Logan, M., Martin, J.F., 28, 94 112. St-Jacques, B., and McMahon, A.P. (2001). Cholesterol modification of sonic hedgehog is required for long-range signaling activity and Saunders, J.W. (1948). The proximo-distal sequence opf origin of effective modulation of signaling by Ptc1. Cell 105, 599 612. the parts of the chick wing and the role of the ectoderm. J. Exp. Zool. 108, 363 404. Litingtung, Y., Dahn, R.D., Li, Y., Fallon, J.F., and Chiang, C. (2002). Shh and Gli3 are dispensable for limb skeleton formation but regudevelopment. Saunders, J.W. (1977). The experimental analysis of chick limb bud late digit number and identity. Nature 418, 979 983. In Vertebrate Limb and Somite Morphogenesis. D.A. Ede, J.R. Hinchliffe, and M.J. Balls, eds. (Cambridge: Cambridge Marcil, A., Dumontier, E., Chamberland, M., Camper, S.A., and University Press), pp. 1 24. Drouin, J. (2003). Pitx1 and Pitx2 are required for development of hindlimb buds. Development 130, 45 55. Saunders, J.W. (2002). Is the progress zone model a victim of prog- Martin, G.R. (1998). The roles of FGFs in early development of verteress? Cell 110, 541 543. brate limbs. Genes Dev. 12, 1571 1586. Saunders, J.W., and Gasseling, M.T. (1968). Ectodermal-mesenchy- Meinhardt, H. (1983). A bootstrap model for the proximodistal patchymal Interactions, R. Fleischmeyer and R.E. Billingham, eds. (Bal- mal interactions in the origin of limb symmetry. In Epithelial-Mesentern formation in vertebrate limbs. J. Embryol. Exp. Morph. 76, 139 146. timore: Williams & Wilkins), pp. 78 97. Mercader, N., Leonardo, E., Azpiazu, N., Serrano, A., Morata, G., Sordino, P., van der Hoeven, F., and Duboule, D. (1995). Hox gene Martinez, C., and Torres, M. (1999). Conserved regulation of proxiexpression in teleost fins and the origin of vertebrate digits. Nature modistal limb axis development by Meis1/Hth. Nature 402, 425 429. 375, 678 681. Mercader, N., Leonardo, E., Piedra, M.E., Martinez, A.C., Ros, M.A., Stratford, T., Horton, C., and Maden, M. (1996). Retinoic acid is and Torres, M. (2000). Opposing RA and FGF signals control proxi- required for the initiation of outgrowth in the chick limb bud. Curr. modistal vertebrate limb development through regulation of Meis Biol. 6, 1124 1133. genes. Development 127, 3961 3970. Summerbell, D. (1974). A quantitative analysis of the effect of excision Méthot, N., and Basler, K. (2001). An absolute requirement for Cubitus of the AER from the chick limb-bud. J. Embryol. Exp. Morphol. interruptus in Hedgehog signaling. Development 128, 733 742. 32, 651 660. Nelson, C.E., Morgan, B.A., Burke, A.C., Laufer, E., DiMambro, E., Summerbell, D., Lewis, J.H., and Wolpert, L. (1973). Positional infor- Murtaugh, C., Gonzalez, E., Tessarollo, L., Parada, L.F., and Tabin, mation in chick limb morphogenesis. Nature 224, 492 496. C. (1996). Analysis of Hox gene expression in the chick limb bud. Sun, X., Mariani, F., and Martin, G.R. (2002). Functions of FGG signal- Development 122, 1449 1466. ling from the apical ectodermal ridge in limb development. Nature Niederreither, K., Vermot, J., Schbaur, B., Chambon, P., and Dolle, 418, 501 507. P. (2002). Embryonic retinoic acid synthesis is required for forelimb Swindell, E.C., Thaller, C., Sockanathan, S., Petkovich, M., Jessell, growth and anteroposterior patterning in the mouse. Development T.M., and Eichele, G. (1999). Complementary domains of retinoic 129, 3563 3574. acid production and degradation in the early chick embryo. Dev. Niswander, L., and Martin, G.R. (1992). Fgf-4 expression during gastrulation, Biol. 216, 282 296. myogenesis, limb and tooth development in the mouse. Tanaka, M., Munsterberg, A., Anderson, W.G., Prescott, A.R., Hazon, Development 114, 755 768. N., and Tickle, C. (2002). Fin development in a cartilaginous fish and Niswander, L., Jeffrey, S., Martin, G., and Tickle, C. (1994). A positive the origin of vertebrate limbs. Nature 416, 527 531.