A standard set of six experiments performed on the limb buds of two species of Anurans - Rana temporaria and Xenopus laevis - are described. The experiments are limb-bud amputation, distal to proximal shifts, proximal to distal shifts, inversion of the dorsoventral axis inversion of the anteroposterior axis and inversion of both axes. The results are compared to those previously reported for Urodeles and chicks to determine whether any principles of vertebrate limb development can be formulated. It appears that the proximodistal axis becomes increasingly mosaic from the Urodeles through Anurans to chicks. In the transverse axes however, there is much more uniformity of behaviour in the production of supernumerary limbs. The relation between the type of limb development (regulative or mosaic) and the subsequent regenerative powers of the adult limb is discussed.

To formulate universal laws of development we need to compare the results of experiments on many organisms throughout the animal kingdom. As far as vertebrate limb development is concerned we cannot yet make such generalisations largely because of the way in which this subject has been investigated. It is usual to study in great detail the most readily available laboratory organism of the day rather than to study in lesser detail a wider range of animals. Thus, work performed in the early part of this century mostly employed Urodeles, particularly Ambystoma species (Harrison, 1921; Swett, 1927), whereas more recently the chick limb bud has become the most studied system (see for example Ede, Hinchliffe & Balls, 1977). This situation is particularly unfortunate since the development of these two types of limb seems to be quite different - the Urodele limb is highly regulative (Harrison, 1918; Slack, 1980; Maden & Goodwin, 1980), whereas the chick limb is much more mosaic (Warren, 1934; Summerbell & Lewis, 1975; Summerbell, 1977).

Furthermore, because different experimenters have asked different questions of developing limbs there has not been a consistent set of experiments which has been performed on the limb buds of these animals. A previous publication (Maden & Goodwin, 1980) began an attempt to provide such standardisation by reporting the results of a set of operations on Ambystoma mexicanum limb buds. The data for the same experiments on chick limb buds can be extracted from the literature for comparison (see Discussion). In the present paper, the results of that same set of experiments performed on the limb buds of two species of Anurans, Rana temporaria and Xenopus laevis are described. Apart from permitting more valid generalisations to be made about vertebrate limb development, this work is also relevant to the question of whether the mechanisms of limb development dictate the subsequent regenerative capabilities of the limb. In other words, is the adult chick limb (and thus by inference the adult human limb) incapable of regeneration because of the mosaic nature of its development and, conversely, can the adult Urodele limb regenerate because of the regulative nature of its development?

The experiments were performed on Rana temporaria obtained as spawn from local ponds and reared in the laboratory and Xenopus laevis bred from laboratory stock. The hind limb buds were used since these are much more accessible than forelimbs in Anurans. Larvae were anaesthetised in 1:3000 MS222. Two stages of limb buds were used, those of Rana temporaria (Fig. 1 a and b) being equivalent to limb bud stages IV and V of Rana pipiens (Taylor & Kollros, 1946 - no staging of Rana temporaria to include the limb buds could be found) and for Xenopus stages 51 and 52 of Nieuwkoop & Faber (1967) (Fig. 1c and d). Most operations, that is limb-bud removal (Series I) and all the rotations (Series IV-VII), were performed at the earlier of the two stages. The distal to proximal grafts (Series II) were performed at the later of the two stages and for proximal to distal grafts (Series III) whole limb buds of the earlier stages were grafted to distal levels of later-stage buds. Thus the methodology was intentionally identical to previous work on Ambystoma mexicanum (Maden & Goodwin, 1980) for comparative purposes.

Fig. 1

Stages of limb development when the operations were performed. The dots represent melanophores in the limb buds, (a) Early stage of Rana temporaria equivalent to stage V of Rana pipiens (Taylor & Kollros, 1946) when the rotations (Series IV-VII) and simple amputations (Series I) were performed by cutting at the level marked A. (b) Later stage of Rana temporaria equivalent to stage VI of Rana pipiens when the distal to proximal (Series II) and proximal to distal (Series III) grafts were performed, (c) Stage-51 limb buds (Nieuwkoop & Faber, 1967) of Xenopus laevis when the rotations (Series IV-VII) and simple amputations (Series I) were performed by cutting at the level marked A. (d) Stage-52 limb buds of Xenopus laevis when the distal to proximal (Series IV) and proximal to distal (Series III) grafts were performed. Bar = 1 mm.

Fig. 1

Stages of limb development when the operations were performed. The dots represent melanophores in the limb buds, (a) Early stage of Rana temporaria equivalent to stage V of Rana pipiens (Taylor & Kollros, 1946) when the rotations (Series IV-VII) and simple amputations (Series I) were performed by cutting at the level marked A. (b) Later stage of Rana temporaria equivalent to stage VI of Rana pipiens when the distal to proximal (Series II) and proximal to distal (Series III) grafts were performed, (c) Stage-51 limb buds (Nieuwkoop & Faber, 1967) of Xenopus laevis when the rotations (Series IV-VII) and simple amputations (Series I) were performed by cutting at the level marked A. (d) Stage-52 limb buds of Xenopus laevis when the distal to proximal (Series IV) and proximal to distal (Series III) grafts were performed. Bar = 1 mm.

The operations were performed in air and the correct orientation of limb buds was easily verified in these two species thanks to the fortuitous presence of a patch of melanophores on the dorsal surface of the bud (Fig. 1). After transplantation the animals were placed at 4°C for 5 – 10 min to facilitate sticking of the graft (usually a clot of blood served as glue), and then returned to water. Those larvae whose grafts did not survive were discarded the next day. During the subsequent weeks the animals were regularly observed and after 5 – 8 weeks the limbs were fixed and stained with Victoria blue to reveal their skeletal structure.

Series I. Limb-bud removal

Whole left limb buds were removed at level A in Fig. 1 a and c for transplanting to the contralateral side in Series V and VI. The severed left stumps thus served as experimental material to examine their regulative powers. In Rana, a total of 38 limbs were stained, 36 of which produced perfect 5-digit limbs (Table 1, Fig. 2). The remaining two limbs had relatively minor defects in the form of a shortened and curved tibia and fibula.

Table 1

Summary of the results of three types of operation on the proximodistal axis of axolotl (data from Maden & Goodwin, 1980), Rana and Xenopus (results reported here) and chick (data from various sources noted below) limb buds assembled into one table for comparative purposes

Summary of the results of three types of operation on the proximodistal axis of axolotl (data from Maden & Goodwin, 1980), Rana and Xenopus (results reported here) and chick (data from various sources noted below) limb buds assembled into one table for comparative purposes
Summary of the results of three types of operation on the proximodistal axis of axolotl (data from Maden & Goodwin, 1980), Rana and Xenopus (results reported here) and chick (data from various sources noted below) limb buds assembled into one table for comparative purposes
Fig. 2

A perfect hind limb of Rana temporaria which regenerated following limb-bud removal (Series I). Vicotria blue staining. f = femur; t f = tibia and fibula; c = calcanéum; a = astragalus; 12 3 4 5 = digits.

Fig. 2

A perfect hind limb of Rana temporaria which regenerated following limb-bud removal (Series I). Vicotria blue staining. f = femur; t f = tibia and fibula; c = calcanéum; a = astragalus; 12 3 4 5 = digits.

This excellent ability of Rana to regulate for limb-bud extirpation is not paralleled by Xenopus. Here a total of 58 cases were examined, only 22 of which (38%) produced perfect limbs (Table 1, Fig. 3). Eleven cases gave nothing more than a femur which terminated about one half of the way down its length; this was presumably the level at which the bud was removed. The remaining 25 limbs were defective in one of two ways. The majority (15) had varying degrees of proximodistal deletions in the form of shortened or deleted elements (Fig. 4). The other 10 limbs were classified as half-limbs, that is they were proximodistally complete but only contained one element in the zeugopodium and carpals and only 2 – 4 digits (Fig. 5). Nine of these were identified as posterior halves and one anterior half limb. This strange result could be due to cutting inaccuracies, that is, amputating the limb bud on a slant rather than at right angles to the proximodistal axis. To avoid confusion they are not included in Table 1.

Fig. 3

A perfect hind limb of Xenopus laevis which regenerated following limb-bud removal (Series I) .Victoria blue staining. See Fig. 2 legend for symbols.

Fig. 3

A perfect hind limb of Xenopus laevis which regenerated following limb-bud removal (Series I) .Victoria blue staining. See Fig. 2 legend for symbols.

Fig. 4

A limb of Xenopus which regenerated after limb-bud removal. By comparing this with Fig. 3 it can be seen that the tibia and fibula are missing.

Fig. 4

A limb of Xenopus which regenerated after limb-bud removal. By comparing this with Fig. 3 it can be seen that the tibia and fibula are missing.

Fig. 5

Another limb of Xenopus which regenerated after limb-bud removal. Here there is a femur, fibula, calcanéum and digits 3, 4 and 5, i.e. a posterior half limb.

Fig. 5

Another limb of Xenopus which regenerated after limb-bud removal. Here there is a femur, fibula, calcanéum and digits 3, 4 and 5, i.e. a posterior half limb.

Series II. Distal to proximal transpositions

For these experiments, the slightly later-stage limb buds were used (Fig. 1 b and d), from which thick slices were removed by making two cuts (at B and C), removing the slice and replacing the tip on the proximal stump. In this fashion at least 50% of the mass of the exposed limb bud was removed. Although the precise level of the cuts was not known beforehand, it can be deduced from the cases described below where no intercalation occurred that the proximal cut was at the level of the mid-distal femur and the distal cut at the ankle. Thus the tissue removed was at least all of the zeugopodium.

A total of 50 Rana limb buds were so treated, 21 of which (42%, Table 1) produced perfect 5-digit limbs (see Fig. 2) despite the removal of such a large amount of tissue. Since the limb buds were frequently examined to ensure that the grafted tip remained in place we can conclude that these limbs exhibited perfect intercalary regulation. The remaining 29 limbs showed various degrees of defect which were grouped into two categories (Table 1). Thirteen (26%) had gross proximodistal deletions with the complete zeugopodium missing. The other 16 (32%) were classified as ‘attempted intercalation’ because they had all the segments present but with some (either the calcanéum and astragalus or tibia and fibula) shorter than normal (Fig. 6).

Fig. 6

One result of grafting a distal tip of a Rana limb bud onto a more proximal level (Series II). Here no whole segments are missing as in many other cases (e.g. Fig. 7 for Xenopus), but there is a defect at the ankle. This was classified as attempted intercalation.

Fig. 6

One result of grafting a distal tip of a Rana limb bud onto a more proximal level (Series II). Here no whole segments are missing as in many other cases (e.g. Fig. 7 for Xenopus), but there is a defect at the ankle. This was classified as attempted intercalation.

Again, this ability of Rana to regulate for tissue removal is not shown by Xenopus, where every one of the ten limbs operated upon revealed proximodistal deletions with whole segments missing (Table 1). In all cases the calcanéum, astragalus and 5 digits were present and these abutted onto the humerus; occasionally the distal ends of the tibia and fibula were present too (Fig. 7). In no case was there any sign of attempted intercalation of the zeugopodium.

Fig. 7

The result of distal to proximal grafting on Xenopus (Series II). No limbs intercalated and here the very distal end of the tibia and fibula are present (tf) and these join directly onto the mid femur (f) level.

Fig. 7

The result of distal to proximal grafting on Xenopus (Series II). No limbs intercalated and here the very distal end of the tibia and fibula are present (tf) and these join directly onto the mid femur (f) level.

Series III. Proximal to distal transpositions

Whole limb buds at the later stage (Fig. 1 b and d) severed at level C were grafted to distal levels (level B). In Rana a total of 20 such operations were performed. In all but one case the grafted limb bud did not develop well and remained as a miniature whole or part limb on the end of the host limb (at the level of the calcaneum and astragalus). Only that one limb developed well-grown repeated segments. The reason for the lack of development of the graft is not clear since there were no obvious problems in matching the two cut surfaces during the operations. This is the only series out of the several hundred operations reported here in which this phenomenon was noted. It is thus less likely that this can simply be attributed to incomplete healing and incorporation of the graft, perhaps a more profound developmental reason is the cause.

With Xenopus, on the other hand, although only five buds were operated upon each produced good limbs with fully grown repeated segments (Fig. 8). The sequence of elements which resulted was femur, tibia and fibula (host), femur, tibia and fibula and foot (graft).

Fig. 8

The result of grafting a whole limb bud onto distal levels (Series III) in Xenopus. Here there are two limbs tandemly repeated in the sequence femur, tibia and fibula, femur, tibia and fibula, calcanéum and astragalus and digits.

Fig. 8

The result of grafting a whole limb bud onto distal levels (Series III) in Xenopus. Here there are two limbs tandemly repeated in the sequence femur, tibia and fibula, femur, tibia and fibula, calcanéum and astragalus and digits.

Series IV. Controls

For this and all the subsequent series the earlier stage limb buds were used (Fig. 1 a and c).

In this series the limb buds were amputated at level A and replaced in the same orientation to serve as controls for the following axial inversion series. In Rana 20 limb buds were so treated, 17 of which produced perfect 5-digit limbs (see Fig. 2). The remaining three displayed curved and shortened tibiae and fibulae, a phenomenon seen frequently throughout the following series. One of these three also has one extra digit.

In Xenopus ten control limb buds were operated upon and all produced perfect 5-digit limbs (see Fig. 3) without any signs of the level of the cut being apparent.

Series V. Dorsoventral inversion

By grafting left limb buds onto right stumps, the dorsoventral axis can be inverted whilst the anteroposterior axis remains normally oriented. This operation resulted in a high proportion of supernumerary limb induction in Rana (Table 2). Of the 28 operated limbs 17 produced supernumeraries (61%), 15 of which were double (Fig. 9) and 2 single. They bifurcated mostly at the knee, but occasionally higher in the femur or lower at the ankle. Of the total of 32 supernumeraries 78% had 4 or 5 digits and all of these were of stump handedness (right). The position of origin of these supernumeraries was rather confusing. Whilst the majority arose at the dorsal or ventral poles, some were displaced into adjacent quadrants. Surprisingly, four limbs had supernumeraries which arose at the anterior or posterior poles of the host limb; perhaps this was caused by the rotation of the graft after the operation.

Table 2

Summary of the results of the three types of axial inversions on the limb buds of axolotl (data from Maden & Goodwin, 1980), Rana, Xenopus (results reported herd) and chick (data from Saunders, Gasseling & Gfeller, 1958) assembled into one table for comparative purposes

Summary of the results of the three types of axial inversions on the limb buds of axolotl (data from Maden & Goodwin, 1980), Rana, Xenopus (results reported herd) and chick (data from Saunders, Gasseling & Gfeller, 1958) assembled into one table for comparative purposes
Summary of the results of the three types of axial inversions on the limb buds of axolotl (data from Maden & Goodwin, 1980), Rana, Xenopus (results reported herd) and chick (data from Saunders, Gasseling & Gfeller, 1958) assembled into one table for comparative purposes
Fig. 9

Dorsoventral inversion (Series V) of Rana limb buds. From the ventral pole of the limb arises a 5-digit right-handed supernumerary (SJ, but with some phalanges missing. In the middle is the left limb bud (G) which was grafted onto the right stump. From the dorsal pole of the limb arises a perfect 5-digit right-handed supernumerary (S2). Both supernumeraries are M thus of stump handedness and bifurcate at the knee.

Fig. 9

Dorsoventral inversion (Series V) of Rana limb buds. From the ventral pole of the limb arises a 5-digit right-handed supernumerary (SJ, but with some phalanges missing. In the middle is the left limb bud (G) which was grafted onto the right stump. From the dorsal pole of the limb arises a perfect 5-digit right-handed supernumerary (S2). Both supernumeraries are M thus of stump handedness and bifurcate at the knee.

In Xenopus, by contrast, the frequency of supernumerary induction was much lower, being only 15% (3 out of 20). Each was a single supernumerary, which bifurcated in the femur, two at the dorsal pole and one at the ventral pole. Two were perfect 5-digit limbs of stump handedness (Fig. 10) and one had only the posterior 3 digits. This contradicts the report by Cameron & Fallon (1977) that Xenopus does not produce DV supernumeraries.

Fig. 10

Dorsoventral inversion of Xenopus limb buds (Series V). Here only single supernumeraries (S) were produced. This one was a perfect 5-digit limb of stump handedness arising at the ventral pole.

Fig. 10

Dorsoventral inversion of Xenopus limb buds (Series V). Here only single supernumeraries (S) were produced. This one was a perfect 5-digit limb of stump handedness arising at the ventral pole.

Series VI. Anteroposterior inversion

Grafting left limb buds onto right stumps in normal dorsoventral orientation results in the inversion of the anteroposterior axis. In Rana, this operation resulted in a good frequency of induction of supernumeraries − 14 out of 26 limbs or 54% (Table 2). As in the previous series the vast majority of these (11 in all) were double supernumeraries (Fig. 11) and the remaining three, single. Again they bifurcated in the vicinity of the knee. The majority (22 out of 25) were well formed supernumeraries with either 4 or 5 digits and all of these were of stump handedness. Most of them arose around the anterior or posterior poles of the host limb, but four limbs produced supernumeraries at the dorsal or ventral poles which, as in the previous series, confused the issue somewhat. Further complicating the situation was one peculiar supernumerary which was double anterior in structure.

Fig. 11

An anteroposterior inversion (Series VI) of Rana limb buds which resulted in two supernumeraries. At the anterior position is a right-handed supernumerary (S1), then the grafted left limb, then at the posterior position a right handed supernumerary (S2). The digital sequence from top to bottom is 12 3 4 5 (S1) 5 4 3 21 (graft) 12 3 4 5 (S2). Thus both supernumeraries are of stump handedness.

Fig. 11

An anteroposterior inversion (Series VI) of Rana limb buds which resulted in two supernumeraries. At the anterior position is a right-handed supernumerary (S1), then the grafted left limb, then at the posterior position a right handed supernumerary (S2). The digital sequence from top to bottom is 12 3 4 5 (S1) 5 4 3 21 (graft) 12 3 4 5 (S2). Thus both supernumeraries are of stump handedness.

In Xenopus a lower frequency was produced − 30% (6 out of 20 - Table 2). All were single as in the DV Xenopus series and all except one appeared at the anterior pole. Only two were good 4-or 5-digit limbs (Fig. 12).

Fig. 12

Anteroposterior inversion of Xenopus limb buds (Series VI). Here a 4-digit right-handed supernumerary (S) appeared at the anterior pole of the grafted left limb bud (G).

Fig. 12

Anteroposterior inversion of Xenopus limb buds (Series VI). Here a 4-digit right-handed supernumerary (S) appeared at the anterior pole of the grafted left limb bud (G).

Series VII. Dorsoventral arid anteroposterior inversion

In this series limb buds were amputated and replaced upside down, thus reversing both limb axes. In Rana, a total of 58 such operations were performed and 39 of these (67%-Table 2) produced supernumerary limbs which were remarkable for two phenomena. Firstly, eleven supernumeraries were double posterior limbs. They did not arise in any consistent location with respect to the host axes, they could be either single or present in conjunction with another normal supernumerary (two double posteriors were never produced) and had either 3, 4 or 5 digits (Figs. 13 and 14). Secondly, five cases of triple supernumeraries arose. Fig. 15, for instance, is a right limb which has a total of 18 digits. It consists of the grafted right limb, a ventral left-handed supernumerary, a posterodorsal right-handed supernumerary and a posterodorsal left-handed supernumerary.

Fig. 13

A double posterior supernumerary (S) resulting from APDV inversion of Rana limb buds (Series VII). It arose at the posterior pole and has a mirror-imaged calcanéum (C) and a digital sequence of 5 4 4 5. Above, is the rotated left limb bud.

Fig. 13

A double posterior supernumerary (S) resulting from APDV inversion of Rana limb buds (Series VII). It arose at the posterior pole and has a mirror-imaged calcanéum (C) and a digital sequence of 5 4 4 5. Above, is the rotated left limb bud.

Fig. 14

A double posterior supernumerary (S) resulting from APDV inversion of Rana limb buds (Series VII). It has two calcanea, a digital sequence of 5 4 3 4 5, and arose in the posterior-ventral quadrant. Above is the rotated left limb.

Fig. 14

A double posterior supernumerary (S) resulting from APDV inversion of Rana limb buds (Series VII). It has two calcanea, a digital sequence of 5 4 3 4 5, and arose in the posterior-ventral quadrant. Above is the rotated left limb.

Fig. 15

A triple supernumerary resulting from APDV inversion of Rana limb buds (Series VII). The limb has a total of 18 digits. In the middle is the rotated right limb (G). To the left is a 5-digit left-handed supernumerary (S1) which arose at the ventral pole. To the right is a V-shaped double supernumerary (S2), one left-handed and one right-handed with their ventral surfaces facing. There are two astragali (o), a central fused calcanéum and 8 digits with digits 4 and 5 being fused. This arose in the posterodorsal quadrant.

Fig. 15

A triple supernumerary resulting from APDV inversion of Rana limb buds (Series VII). The limb has a total of 18 digits. In the middle is the rotated right limb (G). To the left is a 5-digit left-handed supernumerary (S1) which arose at the ventral pole. To the right is a V-shaped double supernumerary (S2), one left-handed and one right-handed with their ventral surfaces facing. There are two astragali (o), a central fused calcanéum and 8 digits with digits 4 and 5 being fused. This arose in the posterodorsal quadrant.

The remaining limbs produced approximately equal numbers of either single or double supernumeraries (Fig. 16) from all quadrants of the host limb-no particular position of origin was preferred. As before, most bifurcated at the knee. Where they could be identified, pairs of supernumeraries were always found to be of opposite handedness. Many, however, were unidentifiable since the digits did not curve. It is possible that these could be mirror-imaged in the dorsoventral plane as has been found in regenerating axolotl limbs after 180° rotation (Maden, 1980a). This is currently being investigated by studying the muscle patterns in serial sections.

Fig. 16

A classical double supernumerary from APDV inversion of Rana limb buds (Series VII). In the middle is the right graft (G). Above is a 5-digit left-handed supernumerary (S1) in the posteroventral quadrant and below is a 5-digit right-handed supernumerary (S2) in the anteroventral quadrant, thus making a pair of opposite handedness.

Fig. 16

A classical double supernumerary from APDV inversion of Rana limb buds (Series VII). In the middle is the right graft (G). Above is a 5-digit left-handed supernumerary (S1) in the posteroventral quadrant and below is a 5-digit right-handed supernumerary (S2) in the anteroventral quadrant, thus making a pair of opposite handedness.

In Xenopus, rotating the limb bud 180° resulted in a low frequency of induction of 20% (4 out of 20, Table 2). One of these was a double supernumerary (Fig. 17), the other three single (Fig. 18). They bifurcated at the knee or higher, only one was a perfect 5-digit supernumerary (of opposite handedness to the stump), one was a 4-digit limb and the other three only developing 2 digits. No structures resembling the double posterior supernumeraries of Rana were produced.

Fig. 17

A double supernumerary resulting from APDV rotation of Xenopus limb buds. The grafted left limb bud (G) is still upside down. A perfect 5-digit right-handed supernumerary (S1) arose in the posterodorsal quadrant and a posterior half 2-digit supernumerary (S2) arose in the anterodorsal quadrant.

Fig. 17

A double supernumerary resulting from APDV rotation of Xenopus limb buds. The grafted left limb bud (G) is still upside down. A perfect 5-digit right-handed supernumerary (S1) arose in the posterodorsal quadrant and a posterior half 2-digit supernumerary (S2) arose in the anterodorsal quadrant.

Fig. 18

A single 4-digit supernumerary resulting from APDV rotation of Xenopus limb buds (Series VII). Note the calcanéum and astragalus of the supernumerary (S) abut directly onto the proximal end of the tibia and fibula of the rotated limb producing a considerable proximodistal disparity in levels.

Fig. 18

A single 4-digit supernumerary resulting from APDV rotation of Xenopus limb buds (Series VII). Note the calcanéum and astragalus of the supernumerary (S) abut directly onto the proximal end of the tibia and fibula of the rotated limb producing a considerable proximodistal disparity in levels.

Many of the rotated limbs displayed proximodistal deletions at the level of rotation with, for instance, the distal part of the femur or the proximal tibia and fibula missing. In Fig. 18 the calcanéum and astragalus of the 4-digit supernumerary abut directly onto the knee of the rotated limb. This is in contrast to the other two rotation series (V and VI) where the level of rotation could only be detected by a sudden inversion of limb structure, rather than by deleted tissue. Thus, this type of defect seems to be related to the complete inversion of limb buds.

The proximodistal axis

When the limb buds of axolotls (Maden & Goodwin, 1980) and other Urodele species (Harrison, 1918) are removed a perfect limb nevertheless develops. On the other hand there is universal agreement (Lillie, 1904; Shorey, 1909; Peebles, 1911; Spurling, 1923; Saunders, 1948; Hampé, 1959) that the chick limb produces truncations following such an operation. Between these two extremes are the Anurans. Rana temporaria almost always produced perfect limbs here and earlier stages of other species of Rana do likewise (Byrnes, 1898). But Xenopus laevis produced perfect limbs in less than half the cases (Table 1). This same gradation in regulative ability is evident in the results of distal to proximal grafts (Series II) in which a large slice of the limb bud is removed.

Axolotls undergo perfect intercalary regulation (Maden & Goodwin, 1980), the chick limb bud produces limbs with severe deletions (Summerbell, 1977; Kieny & Pautou, 1977) and in between are the Anurans. Rana undergoes intercalation in just under half the cases, but Xenopus does not at all (Table 1). The converse experiment, however, that of grafting a whole limb bud to distal levels (Series III) produces uniformity of results - serially repeated limbs are the norm (Table 1).

Not only do these different species vary in their responses to the experiments described above, but in Xenopus and Rana, which are of intermediate regulative ability, even members of the same species do not behave consistently. For instance, in Rana, after distal to proximal grafts some limbs intercalated perfectly, some did so partially and yet others did not at all. This phenomenon further confounds the search for general rules of development.

An important conclusion which emerges from this study of Anuran limb development is that the ability to regulate for limb-bud amputation is not related to the ability to regulate for intercalary deletions. If the two were strictly dependent then Rana, which always regenerates the limb bud, should always replace such intercalary deletions, but the latter only occurs in less than half the cases. Conversely, Xenopus which regenerates limb buds in 38% of the cases should be able to intercalate deletions in a certain proportion, but it cannot. In further support of this conclusion the Xenopus distal to proximal grafts of Series II were performed on right limb buds while the left limb buds of the same animals were simply amputated at the proximal level. In several animals the left limb buds regenerated complete limbs yet their contralateral partners did not intercalate for proximodistal deletions.

As an explanation of this unrelatedness it is possible that the regeneration of limb buds is primarily a property of the epidermis whereas regulation for proximodistal deletion is a property of the mesoderm. Regeneration of limb buds would thus depend on whether or not epidermis can close the wound, and reinitiate developement. Indeed, the terminal defects found after limb-bud amputation can be duplicated in each of these organisms by permanently removing the epidermis (Triton - Balinsky, 1935; Xenopus-Tschumi, 1957; chick-Saunders 1948; Summerbell, 1974.) Although the epidermis rapidly heals over the cut limb bud in chick, the apical ectodermal ridge (AER) is not regenerated (Saunders, 1948). To explain the results reported here on this basis we must assume that the AER is only regenerated in a certain percentage of cases in Xenopus and always in Rana, a suggestion which can easily be tested by performing the relevant histology.

The latter case, that regulation for proximodistal deletion is a property of the mesoderm, seems highly likely since this phenomenon takes place between two cut edges of mesoderm a long way from the ectodermal covering at the tip of the limb bud. Therefore we can conclude that axolotl, Rana, Xenopus and chick display decreasing regulative ability (or increasing mosaicism) in the mesoderm of the limb bud (Table 1), at least as far as the proximodistal axis is concerned.

What is the significance of these divergent properties of vertebrate limbs for current models of development in the proximodistal axis? One model, developed from studies using the chick limb, is the progress zone model (Summerbell, Lewis & Wolpert, 1973). It hypothesises that change in positional value takes place within the progress zone at the tip of the limb bud and outside, the cells rapidly lose this ability. It can very satisfactorily explain the mosaic results, that is chick limb-bud removal, proximodistal deletion experiments in the chick and Xenopus, and the proximodistal deletion observed after limb-bud amputation in Xenopus (Table 1). But this model cannot in its present form be modified to describe the regulative behaviour of axolotl and Rana limb buds. On the other hand an averaging model such as that developed for the regenerating adult limb (Maden, 1977) is highly regulative and can explain very well the normal proximodistal sequence of limb development (Maden, unpublished), the varied results of limb-bud amputation (provided this depends on the regeneration or not of the apical epidermis) and the proximodistal intercalation of axolotl and Rana. But the problem with this model lies in preventing interaction between neighbouring cells and so mosaic behaviour cannot be explained unless other assumptions are made. One solution is to use a model with two cell-state parameters rather than just one as in the model above, and vary one of them from species to species, or in the case of the Anurans, particularly Rana, vary them from individual to individual (Meinhardt & Gierer, 1980; Summerbell, unpublished).

The transverse axes

In contrast to the increasing mosaicism of axolotl, Rana, Xenopus and chick limb buds in the proximodistal axis, experiments on the transverse axes reveal a dramatic uniformity of results across these vertebrate groups. With one possible exception (dorsoventral inversion of the chick limb) the net effect of inversion of one or both transverse axes is to produce supernumerary limbs. The variables concern the number of supernumeraries, their frequency, position and structure (Table 2).

DV inversion. Frequency

After dorsoventral inversion Rana produces the highest frequencies of supernumeraries, nearly all being double, followed by axolotl and Xenopus. The low frequency recorded in Table 2 for chick refers to the occurrence of bidorsal and biventral wings after such an operation (Saunders, Gasseling & Gfeller, 1958). Similar structures are occasionally observed in regenerating limbs and could represent a supernumerary fused with the inverted graft. However, general opinion seems to be that chick limbs do not produce supernumeraries after dorsoventral inversion. Whether this is due to absence of outgrowth because there is no AER on the dorsal or ventral surfaces or because of a lack of interaction in the mesoderm is unknown.

Position

Supernumeraries appear either at dorsal or ventral poles with singles showing no distinct preference for either position. In axolotl and Rana some are found at anterior and posterior positions but in these cases the orientation of the graft had changed. Structure: In all cases where handedness can be determined DV supernumeraries are of stump handedness.

These observations force us to consider how the dorsoventral axis is organized during development, a subject which has received very little attention compared to the anteroposterior axis. The few experiments on the chick limb (Pautou & Kieny, 1973; McCabe, Errick & Saunders, 1974) have merely concluded that dorsoventral polarity resides in the mesoderm, but can be modified by the ectoderm during certain phases of limb development. However, there exists enough evidence from studies on much earlier stage Urodele embryos to suggest that there is a dorsal organizer analogous to the more well-known posterior organizer which behaves in the same fashion following axial reversal (Hollinshead, 1936; Swett, 1938). The presence of both regions is obligatory for full and complete development of the limb, as recently shown by Slack (1980).

The experiments on the dorsoventral axis described here are perfectly adequately explained by the hypothesis that there is a dorsal organizer which is the source of a diffusible morphogen with properties analogous to, yet distinct from, the posterior organizer (Slack, 1977a,b; Tickle, Summerbell & Wolpert, 1975). A totally different concept, yet equally competent to explain these results, is the polar coordinate model (French, Bryant & Bryant, 1976) which considers the transverse axes of the limb to be a single circumferential field rather than two orthogonal gradients. We shall see below how these two hypotheses fare in explaining the remaining results.

AP inversion. Frequency

Here, the axolotl has the highest frequency of supernumeraries closely followed by the chick and Rana, with Xenopus again the least frequent. All except Xenopus can produce double supernumeraries, with Rana again invariably doing so. Position: The position of single supernumeraries varies along with their frequency. In Xenopus most singles are at the anterior pole, in axolotl either anterior or posterior and in chick and Rana at the posterior pole. Double supernumeraries are, of course, at both anterior and posterior poles, but some in Rana and axolotl are found at dorsal and ventral positions. As above, in these cases the orientation of the graft had changed. Structure: In all cases where handedness can be determined AP supernumeraries are of stump handedness. This, as in the previous series seems to be a general rule.

The organization of the anteroposterior axis has been the subject of extensive investigation and theorizing, with the greatest emphasis being on the concept of a posterior zone as a source of a diffusible morphogen which is responsible for organization in this axis (Tickle et al. 1975; Slack, 1977a, b). More recently, the polar coordinate model (French et al. 1976) has become an alternative description, and has, in particular, been applied to the wealth of data on the chick limb bud (Iten & Murphy, 1980). As in the previous series, both hypotheses are equally capable of explaining these results on the inversion of the anteroposterior axis.

APDV inversion. Frequency

After rotating limb buds 180° to reverse both axes, the chick and Rana have the highest frequencies of induction of Supernumeraries, followed by the axolotl and, again, Xenopus being least frequent. However, Cameron & Fallon (1977) reported a much higher figure for Xenopus - 76% - which would put it ahead of all four organisms. Each can produce single or double supernumeraries but most significant are several cases of triple supernumeraries in Rana. These are also occasionally seen in regenerating axolotl limbs (Maden, unpublished) and in the stick insect (Bart, 1971). In the latter, triple supernumeraries formed in almost an identical number of cases (14 vs 13% here). Position: The axolotl shows a distinct preference for supernumeraries to appear in the anteroventral quadrant, although they do appear in other positions as well. In the chick the majority formed at the anterior and posterior poles, but 10% arose at the dorsal or ventral poles. In Rana and Xenopus neither seem to show a preference for any particular position although Cameron & Fallon (1977) noted that supernumeraries only arose in Xenopus at the anterior or posterior poles. It is difficult to draw any general conclusions from this data. Structure: In the chick two of the seven pairs of supernumeraries generated were of the same handedness as each other and the remaining five pairs were opposite. It is very interesting to note that although their bone structure appeared normal, the feather patterns were mirror imaged, a phenomenon that only occurred after 180° rotation and not after AP inversion (Saunders, et al., 1958). These supernumeraries may not be normal in the dorsoventral axis. In the axolotl the supernumeraries were not of good enough quality to determine handedness. Some had digits that did not curve and so could have been mirror imaged in the dorsoventral axis as has been found in regenerating limbs of axolotls (Maden, 1980a). So, too, in Xenopus where most supernumeraries looked normal after cartilage staining, but in a few the digits did not curve. Rana broke all the rules. Of the five triple supernumeraries, the handedness of four of them could not be completely determined because of straight digits. The one that could had two limbs of the same handedness as the stump and the other opposite. Two of the double supernumeraries whose structure could be determined were of the same handedness as each other and six of them were opposite Eleven supernumeraries were double posterior of varying degrees of mirror imaging and could appear either on their own or accompanied by a normal supernumerary. And in addition many were of indeterminate handedness because their digits did not bend and thus could have been mirror imaged in the dorsoventral plane.

The only general conclusion to emerge from this analysis is that APDV supernumeraries can arise in many positions, can be of normal or mirror imaged structure and perhaps the plane of mirror imaging is species dependent. However, what is quite clear is that the situation after APDV inversion is much more complicated than after AP or DV inversion.

Diffusion models which, in their present state of development concentrate only on the anteroposterior axis, are quite incapable of explaining the above results. This is because an APDV inversion is not equivalent to an AP inversion since the resultant structures are vastly different. Thus a double-gradient model needs to be developed and its detailed behaviour tested to determine whether it can match this diversity of experimental results.

On the other hand, the polar coordinate model goes a considerable way to explaining these results. The occurrence of triple supernumeraries and the lack of constant positions of origin suggests that the original evenly spaced form of the clockface adopted for the insect leg is more appropriate. Pairs of supernumeraries should, however, be of opposite handedness, which they are not necessarily. Mirror-imaged structures are explicable by this model, but specific additional predictions appear as unfortunate side effects. These are that supernumeraries mirror imaged in the anteroposterior axis should appear only at the dorsal and ventral poles, mirror images in the dorsoventral axis should appear only at the anterior and posterior poles and that mirror images should also appear at a certain frequency after AP and DV inversions. These predictions do not appear to hold, but perhaps additional assumptions can be built into the model to facilitate a better fit to the experimental findings.

Development and regeneration

What is immediately apparent from the above analysis is how remarkably similar the results of most of these experiments on developing limb buds are to those on regenerating limbs. In the transverse axes, for instance, DV inversion gives supernumerary limbs of the same handedness as the stump at dorsal or ventral poles (Tank, 1978; Wallace & Watson, 1979; Maden, 1980a) as does AP inversion, but at anterior or posterior poles (Iten & Bryant, 1975; Tank, 1978; Wallace & Watson, 1979; Maden, 1980a). 180° rotation gives supernumeraries of variable structure, either normal, double-dorsal or double-ventral in various combinations (Maden, 1980a). Although the precise structure of supernumeraries from 180° limb-bud rotations has not yet been examined it is possible that they too are abnormal in this respect as noted above. Thus we can conclude that as far as the transverse axes are concerned development and regeneration seem to be governed by the same developmental rules. It is thus valid to attempt to produce theories which not only describe vertebrate limb development, but regeneration as well (e.g. French et al. 1976).

However, this uniformity is not the case in the proximodistal axis. Only axolotl limb buds behave in the same fashion as regenerating limbs which undergo perfect intercalary regulation after distal to proximal shifts (Iten & Bryant, 1975; Stocum, 1975; Maden, 1980b). Rana, Xenopus and chicks are increasingly less regulative in this axis, which confounds the search for generalised rules of vertebrate limb development.

With these conclusions in mind it is interesting to consider whether organisms that regenerate their limbs in adult life show any consistent differences in developmental behaviour from those that cannot regenerate. In the transverse axes this is not the case since all seem to behave in a similar fashion during development. But in the proximodistal axis there is a correlation. Axolotls regenerate throughout their life and show complete regulation in the proximodistal axis. Rana temporaria can regenerate for a short period after limbs have developed (Polezhayev, 1946) and show intermediate regulation. Xenopus loses regenerative ability before limb development has terminated (Dent, 1962) and is mosaic in the proximodistal axis. Finally, chick limbs, which can never regenerate, are the most mosaic of all. Therefore, the capacity of adult animals to regenerate limbs reflects the persistence of embryonic regulative properties throughout the developmental period and into the adult form, with no transition to the mosaic condition occurring. Stimulation of limb regeneration in mammals would thus require a reawakening of the regulative capacity, which may be impossible if mosaicism results from a state of the mesoderm, rather than a deficiency in the epidermis or the lack of some systemic factor.

I would like to thank Katriye Mustafa for excellent technical work throughout this study and Dennis Summerbell for putting me right during many discussions on the manuscript.

Balinsky
,
B. I.
(
1935
).
Selbstdifferenzierung des Extremitatenmesoderms im Interplant
.
Zool. Jb. (Allgemeine Zoologie u.Physiologie der Tieré)
54
,
327
348
.
Bart
,
A.
(
1971
).
Morphogenèse surnuméraire au niveau de la patte du phasme Carausius morosas Br
.
Wilhelm Roux Arch. EntwMech. Org
.
166
,
331
364
.
Byrnes
,
E. F.
(
1898
).
On the regeneration of limbs in frogs after the extirpation of limbrudiments
.
Anat. Anz
.
15
,
104
107
.
Byrnes
,
E. F.
(
1904
).
Regeneration of the anterior limbs in the tadpoles of frogs
.
Wilhelm Roux Arch. EntwMech. Org
.
18
,
171
177
.
Cameron
,
J.
&
Fallon
,
J. F.
(
1977
).
Evidence for polarizing zone in the limb buds of Xenopus laevis
.
Devi Biol
.
55
,
320
330
.
Dent
,
J. N.
(
1962
).
Limb regeneration in larvae and metamorphosing individuals of the South African clawed toad
.
J. Morph
.
110
,
61
77
.
Ede
,
D. A.
,
Hinchliffe
,
J. R.
&
Balls
,
M.
(
1977
).
Vertebrate limb and somite morphogenesis
.
Symp. Br. Soc. Devi Biol
. no.
3
Cambridge University Press
.
French
,
V.
,
Bryant
,
P. J.
&
Bryant
,
S. V.
(
1976
).
Pattern regulation in epimorphic fields
.
Science
193
,
969
981
.
Hampé
,
A.
(
1959
).
Contribution a l’étude du développement et de la régulation des déficiences et des excédents dans la patte de l’embryon de poulet
.
Archs Anat. microsc. Morph, exp
.
48
,
345
478
.
Harrison
,
R. G.
(
1918
).
Experiments on the development of the forelimb of Amblystoma, a self-differentiating equipotential system
.
J. exp. Zool
.
25
,
413
461
.
Harrison
,
R. G.
(
1921
).
On relations of symmetry in transplanted limbs
.
J. exp. Zool
.
32
,
1
136
.
Hollinshead
,
W. H.
(
1936
).
Determination of the dorsoventral axis of the fore limb in Amblystoma tigrinum
.
J. exp. Zool
.
73
,
183
194
.
Iten
,
L. E.
&
Bryant
,
S. V.
(
1975
).
The interaction between the blastema and stump in the establishment of the anterior-posterior and proximo-distal organisation of the limb regenerate
.
Devi Biol
.
44
,
119
147
.
Iten
,
L. E.
&
Murphy
,
D. J.
(
1980
).
Pattern regulation in the embryonic chick limb: supernumerary limb formation with anterior (non-ZPA) limb bud tissue
.
Devl Biol
.
75
,
373
385
.
Kieny
,
M.
&
Pautou
,
M-P.
(
1977
).
Proximo-distal pattern regulation in deficient avian limb buds
.
Wilhelm Roux’ Arch, devl Biol
.
183
,
177
191
.
Lillie
,
F. R.
(
1904
).
Experimental studies on the development of organs in the embryo of the fowl (Gallus domesticas)
.
Biol. Bull. mar. biol. Lab., Woods Hole
.
1
,
33
54
.
Maden
,
M.
(
1977
).
The regeneration of positional information in the amphibian limb
.
J. theor. Biol
.
69
,
735
753
.
Maden
,
M.
(
1980a
).
Structure of supernumerary limbs
.
Nature
287
,
803
805
.
Maden
,
M.
(
1980b
).
Intercalary regulation in the amphibian limb and the rule of distal transformation
.
J. Embryol. exp. Morph
.
56
,
201
209
.
Maden
,
M.
&
Goodwin
,
B. C.
(
1980
).
Experiments on developing limb buds of the axolotl Ambystoma mexicanum
.
J. Embryol. exp. Morph
.
57
,
177
187
.
MacCabe
,
J. A.
,
Errick
,
J.
&
Saunders
,
J. W.
(
1974
),
Ectodermal control of the dorsoventral axis in the leg bud of the chick embryo
.
Devl Biol
.
39
,
69
82
.
Meinhardt
,
H.
&
Gierer
,
A.
(
1980
).
Generation and regeneration of sequence of structures during morphogenesis
.
J. theor. Biol
.
85
,
429
450
.
Nieuwkoop
,
P. D.
&
Faber
,
J.
(
1967
).
Normal Table o/Xenopus laevis (Daudin)
.
Amsterdam
:
North-Holland
.
Pautou
,
M.P.
&
Kieny
,
M.
(
1973
).
Interaction ecto-mésodermique dans l’établissement de la polarité dorso-centrale du pied de l’embryon de poulet
.
C. r. hebd. Séanc. Acad. Sci
.,
Paris D
277
,
1225
1228
.
Peebles
,
F.
(
1911
).
On the interchange of limbs of the chick by transplantation
.
Biol. Bull, mar. biol. Lab., Woods Hole
20
,
14
18
.
Polezhayev
,
L. W.
(
1946
).
The loss and restoration of regenerative capacity in the limbs of tailless Amphibia
.
Biol. Rev
. pp.
141
147
.
Saunders
,
J. W.
(
1948
).
The proximo-distal sequence of origin of the parts of the chick wing and the role of the ectoderm
.
J. exp. Zool
.
108
,
363
403
.
Saunders
,
J. W.
,
Gasseling
,
M. T.
&
Gfeller
,
M. D.
(
1958
).
Interactions of ectoderm and mesoderm in the origin of axial relationships in the wing of the fowl
.
J. exp. Zool
.
137
,
39
74
.
Shorey
,
M. L.
(
1909
).
The effect of the destruction of peripheral areas on the differentiation of the neuroblasts
.
J. exp. Zool
.
7
,
25
64
.
Slack
,
J. M. W.
(
1977a
).
Determination of anteroposterior polarity in the axolotl forelimb by an interaction between limb and flank rudiments
.
J. Embryol. exp. Morph
.
39
,
151
168
.
Slack
,
J. M. W.
(
1977b
).
Control of anteroposterior pattern in the axolotl forelimb by a smoothly graded signal
.
J. Embryol. exp. Morph
.
39
,
169
182
.
Slack
,
J. M. W.
(
1980
).
Regulation and potency in the forelimb rudiment of the axolotl embryo
.
J. Embryol. exp. Morph
.
57
,
203
217
.
Spurling
,
R. G.
(
1923
).
The effect of extirpation of the posterior limb bud on the development of the limb and pelvic girdle in chick embryo
.
Anat. Rec
.
26
,
41
56
.
Stocum
,
D. L.
(
1975
).
Regulation after proximal or distal transposition of limb regeneration blastemas and determination of the proximal boundary of the regenerate
.
Devl Biol
.
45
,
112
136
.
Summerbell
,
D.
(
1974
).
A quantitative analysis of the effect of excision of the AER from the chick limb-bud
.
J. Embryol. exp. Morph
.
32
,
651
660
.
Summerbell
,
D.
(
1977
).
Regulation of déficiences along the proximal distal axis of the chick wing bud: a quantitative analysis
.
J. Embryol. exp. Morph
.
41
,
137
159
.
Summerbell
,
D.
&
Lewis
,
J. H.
(
1975
).
Time, place and positional value in the chick limb-bud
J. Embryol. exp. Morph
.
33
,
621
643
.
Summerbell
,
D.
,
Lewis
,
J. H.
&
Wolpert
,
L.
(
1973
).
Positional information in chick limb morphogenesis
.
Nature
244
,
492
496
.
Swett
,
F. H.
(
1927
).
Differentiation of the amphibian limb
.
J. exp. Zool
.
47
,
385
439
.
Swett
,
F. H.
(
1938
).
Experiments upon the relationship of surrounding areas to polarization of the dorsoventral limb-axis in Ambystoma punctatum (Linn
.).
J. exp. Zool
.
78
,
81
100
.
Tank
,
P. W.
(
1978
).
The occurrence of supernumerary limbs following blastemal transplantation in the regenerating forelimb of the axolotl, Ambystoma mexicanum
.
Devi Biol
,
62
,
143
161
.
Taylor
,
A. C.
&
Kollros
,
J. J.
(
1946
).
Stages in the normal development of Rana pipiens larvae
.
Anat. Rec
.
94
,
7
23
.
Tickle
,
C.
,
Summerbell
,
D.
&
Wolpert
,
L.
(
1975
).
Positional signalling and specification of digits in chick limb morphogenesis
.
Nature
254
,
199
202
.
Tschumi
,
P. A.
(
1957
).
The growth of the hindlimb bud of Xenopus laevis and its dependence upon the epidermis
.
J. Anat
.
91
,
149
173
.
Wallace
,
H.
&
Watson
,
A.
(
1979
).
Duplicated axolotl regenerates
.
J. Embryol. exp. Morph
.
49
,
243
258
.
Warren
,
A. E.
(
1934
).
Experimental studies on the development of the wing in the embryo of Gallus domesticus
.
Am. J. Anat
.
54
,
449
485
.