/ . Embryol. exp. Morph. Vol. 70, pp. 197-213, 1982 197Printed in Great Britain © Company of Biologists Limited 1982

Axial organization of the regenerating limb:asymmetrical behaviour following

skin transplantation

By M. MADEN1 AND K. MUSTAFAFrom the Developmental Biology Division,

National Institute for Medical Research, London

SUMMARY

An extensive series of skin grafting operations has been performed to investigate axialorganization in the regenerating axolotl limb. Semicircular cuffs of skin from either anterior,posterior, dorsal or ventral surfaces were exchanged between right and left limbs therebycreating limbs with double anterior, double posterior, double dorsal or double ventralskin, all with normal internal tissues. Both fore and hindlimbs were used at both upper andlower limb levels. Following amputation through the grafted region the resulting regenerateswere analysed both by whole-mount cartilage staining to observe the pattern of digits andby serial sectioning to observe the pattern of muscles. There were clear asymmetries inability to produce duplications - posterior to anterior grafts resulted in a consistently highfrequency of digital duplications, whereas anterior to posterior grafts produced very few.Similarly, dorsal to ventral grafts resulted in a good frequency of muscle duplications,whereas ventral to dorsal grafts did not. Such asymmetrical behaviour is not predicted bymost models involving local cell:cell interactions and the significance of the results fortheories of pattern formation is discussed.

INTRODUCTION

The analysis of pattern formation in regenerating limbs has, like otherdevelopmental systems, relied on techniques of transplanting tissues andobserving the resultant disruption of pattern. For example, when regenerationblastemas are grafted to contralateral limb stumps to reverse either one of thetransverse axes of the limb or grafted upside down on the same stump toreverse both transverse axes, supernumerary limbs are induced (Iten & Bryant,1975; Bryant & Iten, 1976; Tank, 1978 a; Wallace & Watson, 1979; Maden,1980; Tank, 1981; Maden, 1982). Details of the structure and handedness ofsuch supernumerary limbs has provided many insights into the way in whichthe transverse axes of the limb are organized (French, Bryant & Bryant, 1976;Stocum, 1980; Slack, 1980a; Maden & Mustafa, 1982)

Another method of analysing axial organization which has the potential of

1 Author's address: Division of Developmental Biology, National Institute for MedicalResearch, The Ridgeway, Mill Hill, London NW7 1AA U.K.

198 M. MADEN AND K. MUSTAFA

providing more precise information, as evidenced by work on the insect leg(French, 1978; 1980), is to transplant individual tissues to various positions onthe limb circumference. Thus the roles of skin, muscle and bone have beeninvestigated (Carlson, 1974; 1975a, b\ Tank, 1979) with the result that skinseems to be the most morphogenetically active component. The greater precisionof such a method has revealed significant differences between the generalconclusions from blastemal transplantations and tissue transplantations. Firstly,concerning the anteroposterior axis, a difference in the power of anterior andposterior skin to induce supernumerary limbs has been demonstrated. Slack(19806) grafted anterior half circumferences of skin to the posterior side andobtained normal limbs after amputating through the graft, whereas the converse(posterior to anterior) resulted in mirror-imaged supernumerary limbs. Nosuch asymmetry had previously been suspected because blastemal transplant-ations involving the anteroposterior axis produce two supernumeraries, one atthe anterior pole and one at the posterior pole. Secondly, concerning thedorsoventral axis, Carlson (1974) grafted skin from one limb to the contra-lateral side, thereby inverting only the dorsoventral axis, and obtained normallimbs. Similarly, grafting dorsal skin to ventral or vice versa, analogous toSlack's experiments on the anteroposterior axis, produced normal limbs too(Carlson, 1974; Tank, 1979). This prompted Carlson to conclude that thedorsoventral axis is not polarized in the regenerating axolotl limb. This con-tention is in stark contrast to the results from blastemal transplantation whereinversion of the dorsoventral axis produces supernumerary limbs (Tank, 1978 a;Maden, 1980; 1982).

Now, much information on the organization of the dorsoventral axis hasbeen gained by serially sectioning supernumerary limbs to examine musclepatterns (Maden, 1980, 1982; Maden & Mustafa, 1982). It is conceivable thatthe apparently normal limbs obtained in the experiments of Carlson and Tankalthough not producing supernumeraries, have abnormal muscle patternsanalogous to the mirror-imaged cartilage patterns revealed in the antero-posterior axis. Therefore, in an effort to resolve the above contradictions andto re-examine the effects on the dorsoventral axis, we carried out similar skingrafting operations. Both upper and lower, fore and hindlimbs of axolotls wereused and the resulting regenerates both stained for cartilage and seriallysectioned. The results reveal that the dorsoventral axis is indeed affected byskin grafts and that asymmetries in both axes exist. The significance of theseobservations for theories of pattern formation is discussed.

MATERIALS AND METHODS

The experiments were performed on 70-80 mm axolotls, Ambystoma mexi-canum, under MS222 anaesthesia. Half-circumference segments of skin wereexchanged between contralateral pairs of limbs and sutured in place with

Axial organization of the regenerating limb 199

la lb

lc Id

Fig. 1. Diagram showing the experimental design. In the upper diagram exchangeof anterior and posterior half-skin cuffs between left and right upper forelimbs isshown, resulting in left limbs having double anterior skin and normal internaltissues (a) and right limbs with double posterior skin and normal internal tissues(b). In the lower diagram a similar exchange between dorsal and ventral half-skincuffs is shown, resulting in limbs with double dorsal skin (c) or double ventralskin (d). These four operations were also performed at the lower forelimb, upperhindlimb and lower hindlimb levels.

0-2 Ethilon monofilament suture thread (Ethicon Ltd.). Four categories ofoperations were produced: double anterior - Series l - (F ig . la), doubleposterior - Series 2 - (Fig. 1 b), double dorsal - Series 3 - (Fig. 1 c) and doubleventral - Series 4 - (Fig. 1 d), and each performed at four different levels:upper forelimb, lower forelimb, upper hindlimb and lower hindlimb. Tenanimals were used for each series making 160 operations in all.

200 M. MADEN AND K. MUSTAFA

One week after operating, when the grafts had healed perfectly in place,limbs were amputated through the grafted region. They were allowed to re-generate for at least 8 weeks and the resulting regenerates fixed in neutralformalin. They were stained in Victoria Blue to reveal cartilage elements.Regenerates from the double dorsal and double ventral series were then preparedfor sectioning at 10 ptm and the slides stained with haematoxylin and eosin.Muscle patterns were analysed as previously described (Maden, 1980, 1982).

RESULTS

Cartilage patterns - the anteroposterior axis

Series 1. Anterior skin -> posterior

Forelimbs. Grafting anterior skin in place of posterior skin had no significanteffect on the cartilage structure of the regenerates whether the operations wereperformed at upper or lower forelimb levels (Table 1, rows 1 and 2); tworegenerates from the lower limb series had a bifurcated digit 4 resulting in afew extra phalanges and the remainder were indistinguishable from normallimbs (Fig. 2).

Hindlimbs. The effects of grafting anterior skin to posterior were morenoticeable in the hindlimb. At the upper hindlimb level four regenerates werenormal (five digits, Fig. 3), three were hypomorphic and three had super-numerary elements (Table 2, row 1). The three cases with supernumeraryelements each had an extra digit 3 on the posterior side of the limb (Fig. 4).

At the lower hindlimb level there were no normal regenerates (Table 2,row 2). Three had extra digits on the posterior side, as in Fig. 4 and theremaining six were hypomorphic. Each of these hypomorphics had threedigits (Fig. 5) and a reduced number of tarsals. Unlike those at the upperlimb level where the hypomorphics simply had digits missing, their patternof digits here was not normal - they were clearly not digits 1, 2 and 3 becausedigit 3 has four phalanges (see Fig. 3). They were therefore identified as digits1, 2 and 2 or 1, 2 and 1. In some cases the pattern of tarsals looked mirror-imaged and thus it was concluded that these six hypomorphic regenerateswere double anterior in structure. Thus the frequency of duplication in thisseries was 100%.

Thus, in the forelimb, anterior to posterior skin grafts had no significanteffect. In the hindlimb supernumerary digits were produced. Clear mirror-imaged reduplicates only occurred in the lower hindlimb series and then onlyconsisted of three digits.

Series 2. Posterior skin -> anterior

Forelimbs. In contrast to the previous series, grafting posterior skin in placeof anterior skin had a major effect on regenerate morphology (Table 1, rows3 and 4). At the upper arm level, seven out often cases produced supernumerary

Tab

le 1

. C

arti

lage

str

uctu

re o

f th

e re

gene

rate

s fro

m f

our

type

s of

ski

n gr

afti

ng o

pera

tion

in u

pper

and

low

er fo

reli

mbs

(The

nor

mal

num

ber

of d

igits

in t

he f

orel

imb

is 4

.)

Car

tilag

e st

ruct

ure

of r

egen

erat

es i

n te

rms

of d

igit

num

ber

Fore

limb

leve

l

Upp

erL

ower

Upp

erL

ower

Upp

erL

ower

Upp

erLo

wer

Ope

ratio

n

A->

PA

-*P

P-»

AP

^AD

-*V

D-»

VV

-»D

V->

D

Surv

ivin

gca

ses

10 10 10 10 9 8 9 8

4*

* O

ne o

f th

ese

limbs

had

sup

ernu

mer

ary

fore

arm

ele

men

ts a

nd c

arpa

ls, b

ut o

nly

4 di

gits

.3'

Tab

le 2

. C

arti

lage

str

uctu

re o

f th

e re

gene

rate

s fro

m f

our

type

s of

ski

n gr

afti

ng o

pera

tion

in u

pper

and

low

er h

indl

imbs

(The

nor

mal

num

ber

of d

igits

in t

he h

indl

imb

is 5

.)

Car

tilag

e st

ruct

ure

of r

egen

erat

es i

n te

rms

of d

igit

num

ber

Hin

dlim

b Su

rviv

ing

r—le

vel

Ope

ratio

n ca

ses

110

Ǥ $ s I"U

pper

Low

erU

pper

Low

erU

pper

Low

erU

pper

Low

er

A->

PA

-*P

P->

AP

->A

D-*

VD

-»V

10 9 10 9 9 9 9 9

1 —

4 2 6 9 7 7

3 1—

1 1—

1

—1 1 3 — — —

— 1 2 4 — — 1 1

— — 4 1 — —

202 M. MADEN AND K. MUSTAFA

2 3

4 3

3/4

Axial organization oj the regenerating limb 203structures with limbs ranging from five to nine digits and at the lower armlevel the frequency was even higher - nine out of ten.

Some of the upper arm supernumeraries were mirror-imaged limbs in theanteroposterior axis as one might expect (see Fig. 6). However, the majoritywere more complicated in that the extra elements were displaced ventrally(never dorsally). This resulted in a boomerang-shaped limb (viewed end on)with the ventral surfaces of the normal and supernumerary hands facing eachother. Presumably this represents a flat mirror-imaged duplicate which hasbeen twisted during development.

Regenerates from lower arm amputation levels were more uniform, all beingsimple mirror-images in the anteroposterior axis. The nine supernumerariesfell into the categories described by Slack (1980Z?) namely, duplicates with adigit pattern of 4 3 2 1 2 3 4 (Fig. 6) or 4 3 2 2 3 4, partial duplicates lackingone posterior extremity, e.g. 3 2 1 2 3 4 (Fig. 7) or duplicates with serialrepetition, such as 2 2 1 2 3 4. There were no hypomorphic limbs and onewas normal.

Hindlimbs. In the hindlimb too, posterior to anterior skin grafts had a pro-found effect upon regenerate morphology (Table 2, rows 3 and 4). At theupper level eight out of ten limbs produced supernumeraries ranging fromseven to ten digits and at the lower level every limb was reduplicated with sixto nine digits.

The upper level regenerates were nearly all double limbs with nine or tendigits (Fig. 8). Half of them were complicated, as in the case of the forelimbs,by digits being displaced ventrally with the ventral surfaces of the two limbsfacing each other. The remainder were straight-forward mirror images asdescribed above, that is with a digital pattern of 5 4 3 2 1 2 3 4 5 or partialduplicates lacking one posterior extremity or duplicates with serial repetition.This same pattern of duplication was found in lower hindlimb regenerateswith half of the regenerates having digits displaced in the dorsoventral axisand the remainder with normal duplications (Fig. 9).

Fig. 2. Typical forelimb regenerate from Series 1 - grafting anterior skin toposterior. After amputating either through the upper or lower forelimb levelsregenerates were normal as shown here. Digit numbers are marked, x 14.Figs. 3-5. Three types of hindlimb regenerate from Series 1 - grafting anterior skinto posterior.Fig. 3. Normal regenerate from the upper hindlimb level. Digit numbers aremarked, x 14.Fig. 4. Regenerate from the upper hindlimb level with one extra digit on theposterior side (left) making six digits in all. The two posterior digits could beeither digits 3 or 4, digit 5 being absent, x 14.Fig. 5. Hypomorphic regenerate from the lower hindlimb level with only seventarsals and three digits. The mirror-image nature of the regenerate is apparent,particularly the digits which were assigned the numbers 1, 2, 1. x 16.

204 M. MADEN AND K. MUSTAFA

8

7

4 4

Figs. 6-9. Regenerates from Series 2 - grafting posterior skin to anterior.

Fig. 6. Lower forelimb level duplicated regenerate with a digital pattern of4 3 2 1 2 3 4. x 16.

Fig. 7. Lower forelimb level partial duplicate with one posterior extremity missing,making a digital pattern of 3 2 1 2 3 4. x 16.

Fig. 8. Upper hindlimb level full duplicate with a digital sequence of 5 4 3 2 1 2 3 4and a very small 5th digit, x 14.Fig. 9. Lower hindlimb level duplicate with a digital sequence of 5 4 3 2 2 3 4 5.xl4.

Axial organization of the regenerating limb 205

10 11Fig. 10. Regenerate from Series 3 - dorsal skin to ventral. At both upper and lowerforelimb and hindlimb levels the regenerates were mostly normal. The normalregenerate shown here is from the upper forelimb level, x 14.Fig. 11. Regenerate from Series 4-ventral skin to dorsal. As in Series 3 mostregenerates were normal, that shown here is also from the upper forelimb level.xl4.

Thus, in the forelimb and hindlimb, grafting posterior skin to anterior hasa major effect on pattern formation in the regenerates, causing the productionof mirror-imaged duplicated limbs with up to ten digits in 70-100% of thecases. Clearly then, there is an asymmetry between this situation and theconverse graft of anterior skin to posterior described in Series 1 where thevast majority (85 %) of regenerates were normal or hypomorphic. Those thatwere mirror-imaged duplicates were hypomorphic too.

Series 3. Dorsal skin -» ventral

Forelimbs. Grafting dorsal skin in place of ventral skin had very little effecton the cartilage structure of the regenerates whether the operations wereperformed in the upper arm or lower arm (Table 1, rows 5 and 6). Of theregenerates, 88% (15 out of 17) were normal (Fig. 10). The hypomorphiccase recorded in Table 1 was one which produced a two-digit anterior half

206 M. MADEN AND K. MUSTAFA

limb and the hypomorphic case was a limb with an extra digit 4 which protrudedventrally below the normal one.

Closer scrutiny of these supposedly normal regenerates revealed slight ab-normalities in carpal shape and in many cases digits bent dorsally instead ofventrally.

Hindlimbs. Similar results were obtained on the hindlimb (Table 2, rows 5and 6). Of upper and lower regenerates, 83 % (15 out of 18) were normal, onelimb had one extra digit and two were hypomorphic.

Series 4. Ventral skin -> dorsal

Forelimbs. Grafting ventral skin in place of dorsal skin also had no effecton the cartilage structure of the regenerates from either upper or lower armlevels (Table 1, rows 7 and 8). 100% of the regenerates were normal (Fig. 11).

Hindlimbs. Of the hindlimbs, 78 % (14 out of 18) were normal. Three limbsproduced extra digits, all on the anterior edge of the regenerate and displacedin the dorsoventral axis. Even the two with eight digits had poorly developedsupernumerary digits which could not be identified.

From the results of the last two series we might conclude, as Carlson (1974)did, that the dorsoventral axis of the regenerating limb of axolotls is notpolarized. However, the above method of scoring cartilage elements is onlyconcerned with the anteroposterior axis which was not altered by such operations.

Fig. 12. Muscle patterns in the normal forelimb at the metacarpal level. On thedorsal surface, above each metacarpal is a small, crescent-shaped muscle, theextensor digitorum brevis (arrows). On the ventral surface is a large mass of musclecomposed of about 16 separate muscles fused together. The ventral muscles of digit4 (far left) have split off from the main body of muscle in preparation for theseparation of digit 4. x 60.Fig. 13. Muscle patterns in a regenerate from Series 3 - dorsal skin to ventral.In contrast to Fig. 12 it can be seen that instead of a large mass of muscle onthe ventral surface there are four crescent-shaped extensores digitorum breves whichmirror-image those present as normal on the doisal surface. This limb is clearlydouble dorsal in structure, x 55.Fig. 14. Muscle patterns in another regenerate from Series 3. Here the normalcomplement of four extensores digitorum breves are on the dorsal surface of themetacarpals. On the ventral surface of the two digits on the left there are alsoe.d.b. present (arrows). Beneath the other two digits are normal ventral muscles(v). Thus the left half of this regenerate is mirror-imaged (double dorsal) and theright half is normally asymmetric, x 50.Fig. 15. Regenerate from Series 4-ventral skin to dorsal. Here the only asym-metrical digit is the far right one with an e.d.b. (arrow) on the dorsal surface andventral muscles underneath. The remaining three digits have ventral muscles onboth the ventral and dorsal surfaces which fuse in the mid-line to completely sur-round the digits in muscle. No e.d.b. are present in those three double ventraldigits, x 52.

Axial organization of the regenerating limb 207

CO

C\J

208 M. MADEN AND K. MUSTAFA

Table 3. Muscle patterns in regenerates from Series 3 and 4

(The reduplicated cases are scored below according to the number ofindividual digits in the limb with reduplicated muscle patterns.)

Limb

Fore

Hind

Fore

Hind

Level

UpperLowerUpperLowerUpperLowerUpperLower

Operation

D-*VD-»VD-*V

V-»DV->DV->D

Survivingcases

98999899

Normal

275477

00 00

Pattern of muscles

Number of reduplicated

5 4 3

— 2 —

2— — 1— — 1

— — —

2

3

12

—

1

digits

1

212211

1

Any dorsoventral pattern duplication, equivalent to anteroposterior redupli-cation, would not be apparent unless a separate supernumerary was generated.Consequently the dorsoventral organization of these regenerates, in terms oftheir muscle patterns, was examined in serial sections.

Muscle patterns - the dorsoventral axis

The normal muscle pattern of axolotl limb has been described before, usingthis technique of reconstructed serial sections (Maden, 1980a, b). A detaileddescription will therefore not be repeated here; suffice it to show a typicallynormal section through the metacarpal level in Fig. 12. The ventral surfaceof the hand is characterized by a group of about 16 muscles (Grim & Carlson,1974) together forming a solid mass of muscle, while the dorsal muscles haveseparated into four distinct extensores digitorum breves. These dorsal muscleshave a characteristic crescent shape and sit on the top of the carpals anddigits.

Series 3. Dorsal skin -> ventral

Forelimbs. Despite the fact that all the upper arm regenerates looked normalwhen stained for cartilage, seven out of nine had varying degrees of duplicationin their pattern of muscles (Table 3, top row). Three were perfect doubledorsal limbs (Fig. 13), identical in structure to those that have been previouslydescribed (Maden, 1980a, b), that is with crescent-shaped extensores digitorumbreves on both sides of the limb. The ventral surface is totally devoid of themass of muscle normally present. The remaining four limbs had partial dupli-cation, that is either half of the limb (two digits) was double dorsal and theother half normal (Fig. 14) or just one digit had duplicated muscle patterns.

Axial organization of the regenerating limb 209At the lower arm level the clear effect seen above was not repeated. Here

only one limb showed any sign of duplication Table 3, row 2) and that hadonly one digit affected even then.

Hindlimbs. At the upper hindlimb level, four of the nine regenerates hadduplicated muscle patterns (Table 3, row 3) one of which was fully doubledorsal, the other three partial. At lower levels, five out of the nine regenerateshad double dorsal muscle patterns (Table 3, row 4), all of which were partiallyduplicated.

Thus, after grafting dorsal skin to ventral, three of the levels tested gavebetween 45 % and 80 % of double dorsal regenerates, the exception being thelower forelimb level.

Series 4. Ventral skin -> dorsalForelimbs. Only three limbs out of a total of 17 showed any signs of muscle

duplication at the upper and lower forelimb levels (Table 3, rows 5 and 6).The duplications produced in this series were such that double ventral musclepatterns were produced. There were no extensores digitorum breves present,instead large expanses of muscle both dorsally and ventrally which fused inthe mid-line to surround the digits in solid muscle (Fig. 15). One limb hadthree digits with double ventral muscle (Fig. 15) and the other two had justone digit duplicated.

Hindlimbs. In the hindlimbs too the frequency of duplication was very lowwith only 2 limbs out of 16 showing any signs (Table 3, rows 7 and 8). Oneregenerate had one digit with double ventral muscles and the other had twodigits so duplicated.

DISCUSSION

It is apparent from the results presented above that there are distinct asym-metries in behaviour following the grafting of skin from one side (eitheranterior, posterior, dorsal or ventral) to the other. Posterior skin grafted toanterior gave a very high frequency of duplicated regenerates while the con-verse operation gave a very low frequency. Similarly, dorsal skin grafted toventral gave a moderate frequency of duplicated muscle patterns while theconverse did not. To assist in further comparisons, the frequency of duplicationin each series is recorded in Table 4 where the frequency is not calculatedsimply in terms of the number of limbs affected, but rather by consideringeach individual digit. This provides a valid comparison between the antero-posterior axis and the dorsoventral axis in terms of their ability to causeduplications. Thus Table 4 reveals that posterior skin has the greatest abilityto do so, followed by dorsal skin, anterior skin and lastly ventral skin. It iscomforting to discover that this conclusion is exactly the same as that reached byLheureux (1975). He performed skin grafts on Pleurodeles forelimbs such that only

210 M. MADEN AND K. MUSTAFA

Table 4. Frequency of reduplication in each series scored in terms of theproportion of the total number of digits regenerated that were reduplicated

Limb Level Operation

Total Totalnumber of number of

digits digits re-regenerated duplicated

Re-duplication

(%)

Averagere-

duplication(%)

Fore

Hind

Fore

Hind

Fore

Hind

Fore

Hind

UpperLowerUpperLowerUpperLowerUpperLowerUpperLowerUpperLowerUpperLowerUpperLower

P-^AP-*AP^AP->AD->VD->VD->VD-»VA->PA-*PA->PA-*PV->DV-*DV-*DV->D

39385042

3433

4045

4040

4433

3632

5047

11232925

161

79

00

311

41

21

28605860,

473

17-520

00

733.

113

42.

52

22

10

one polar quality (dorsal, ventral, anterior or posterior) was present at theamputation plane by rotating strips of skin through 90°. The ability to inducesupernumerary elements was highest with posterior skin (49%), then dorsal(28%), anterior (18%) and least with ventral skin (7%): even the frequenciesare virtually identical to those in Table 4. Similarly the highest frequency ofhypomorphic limbs was found with anterior skin, exactly as reported here.However, it is important to note that Lheureux's analysis was performed onlyon cartilage elements in the regenerates and not on muscle patterns. PerhapsPleurodeles are more efficient at producing distinct supernumeraries afterdorsoventral inversions than axolotls.

The analysis of muscle patterns in the regenerates has provided importantnew data from the supposedly' normal' limbs. Had the frequency of duplicationsimply been recorded in terms of cartilage elements then the above resultswould have confirmed Carlson's conclusion that the dorsoventral axis of theregenerating axolotl limb is not polarized. There is no reason to suppose thata mirror-imaged muscle pattern is not an equally valid duplicate to a mirrorimage in the anteroposterior axis revealed by whole-mount cartilage staining.It would be equivalent to an AP duplicate with a digital sequence of 4 3 3 4.Thus the dorsoventral axis is polarized in the regenerating axolotl limb, but

Axial organization of the regenerating limb 211not to the same degree of'strength' as the anteroposterior axis when measuredin terms of degree of duplication (Table 4).

The scope of these experiments, in performing them on both fore andhindlimbs at upper and lower limb levels, was such that some inconsistencieswere bound to occur, yet these were few. In Series 1 (anterior skin -> posterior)all forelimb regenerates were normal, but very few hindlimbs were. Only atthe hindlimb lower level were any mirror-image duplicates found and theyonly consisted of three digits. In Series 3 (dorsal skin -> ventral), at the fore-limb lower level only one muscle reduplicate appeared in contrast to the goodfrequency at the other three levels. On the other hand, Series 1 and 4 wereentirely consistent, the former resulting in a uniformly high frequency of goodfive- to 10-digit duplicates and the latter being uniformly inert.

Consistency with other results is variable. The results of Lheureux (1975)have already been mentioned and those of Slack (19806) are also identical.However, inconsistencies arise in considering the results of Carlson (1974)and Tank (1979). The fact that neither author reported duplicates (or veryfew) after dorsal or ventral skin grafting has been resolved in that they didnot section the regenerates to examine muscle patterns. More significant is thelack of any difference between anterior and posterior skin in their ability toinduce duplication (Carlson, 1974) and yet another type of result, the highproportion of no regeneration after such operations (Tank, 1979); both remainunresolved.

What conclusions can be drawn from the distinct asymmetries between theopposite sides of the two transverse axes of the regenerating limb after skingrafting? Models involving straight-forward local cell-cell interactions are in-capable of explaining such asymmetries. In the polar coordinate model (French,Bryant & Bryant, 1976) for instance, transplanting tissue to the other side ofthe circle, e.g. from 9 to 3, should not produce a different result from theconverse operation, from 3 to 9, when it interacts with the underlying internaltissues. Both should intercalate the same missing values, but this is not whathappens in practice. Furthermore, it is not appropriate to explain the asym-metries in terms of the results of experiments on double half limbs (Bryant &Baca, 1978; Stocum, 1978; Tank, 19786). In this case two half limbs of thesame pattern (e.g. two anterior halves or two posterior halves) are constructedand on amputation, double posteriors regenerate more pattern than doubleanteriors. Whilst initially this result seems similar, the two are not readilycomparable at all because in the double-half limb experiments there are nopoints of incongruity around the limb circumference and the regenerates, notsurprisingly, rarely produce more than the normal number of digits in total(even though they may be mirror imaged) and usually considerably less. In theexperiments reported here, generating positions of maximum incongruity wasprecisely the point and the subsequent regeneration involves the production ofsupernumerary elements - indeed some regenerated almost two whole limbs

212 M. MADEN AND K. MUSTAFA

(Figs. 6-9). Clearly such supernumerary tissue must be produced by the inter-action between the grafted skin and the underlying internal tissues.

The only model which addresses this question of asymmetrical behaviour isthat proposed by Slack (19806) following his anterior and posterior skingrafting experiments. This considers the anteroposterior axis in terms of aseries of territories in which sets of biochemical switches are either on or off.At the posterior edge all the switches are on, at the anterior edge all are off.The dominance of higher switches over lower switches so that any posteriorterritory contains the information for making all the more anterior territories,but not vice versa, explains the anteroposterior asymmetry. This concept caneasily be supplemented by proposing a similar scheme which operates in thedorsoventral axis with the region where all the switches are on at the dorsaledge and all the switches off at the ventral edge. Furthermore variation in theprecise location of territories within the cross section of the limb could explainthe differences in behaviour between different levels of the same limb when thesame graft has been performed (e.g. Series 1 and 3) and the differences betweenfore and hindlimbs (e.g. Series 1 and 3). Clearly, many more tests of competingmodels of pattern formation during limb regeneration need to be performedbefore any be accepted wholeheartedly, but the asymmetrical behaviour of skingrafts revealed above severely strain the credulity of some.

REFERENCESBRYANT, S. V. & BACA, B. A. (1978). Regenerative ability of double-half and half upper arms

in the newt, Notophthalamus viridescens. J. exp. Zool. 204, 307-324.BRYANT, S. V. & ITEN, L. E. (1976). Supernumerary limbs in amphibians: experimental

production in Notophthalmus viridescens and a new interpretation of their formation.DevlBiol. 50, 212-234.

CARLSON, B. M. (1974). Morphogenetic interactions between rotated skin cuffs and under-lying stump tissues in the regenerating axolotl forelimb. Devi Biol. 39, 263-285.

CARLSON, B. M. (1975a). Multiple regenerating from axolotl limb stumps bearing cross-transplanted minced muscle regenerates. Devi Biol. 45, 203-208.

CARLSON, B. M. (19756). The effects of rotation and positional change of stump tissuesupon morphogenesis of the regenerating axolotl limb. Devi Biol. 47, 269-291.

FRENCH, V. (1978). Intercalary regeneration around the circumference of the cockroach leg.J. Embryol. exp. Morph. 48, 53-84.

FRENCH, V. (1980). Positional information around the segments of the cockroach leg. J.Embryol. exp. Morph. 59, 281-313.

FRENCH, V., BRYANT, P. J. & BRYANT, S. V. (1976). Pattern regulation in epimorphic fields.Science 193, 969-981.

GRIM, M. & CARLSON, B. M. (1974). A comparison of morphogenesis of muscles of theforearm and hand during ontogenesis and regeneration in the axolotl {Ambystoma mexi-canurri). I. Anatomical description of muscles of the forearm and hand. Z. Anat. EntwGesch.145, 137-148.

ITEN, L. E. & BRYANT, S. V. (1975). The interaction between the blastema and stump in theestablishment of the anterior-posterior and proximo-distal organisation of the limbregenerate. Devi Biol. 44, 119-147.

LHEUREUX, E. (1975). Nouvelles donnees sur les roles de la peau et des tissus internes dans

Axial organization of the regenerating limb 213la regeneration du membre du Triton Pleurodeles waltlii Michah (Amphibien Urodele).Wilhelm Roux' Arch, devl Biol. 176, 285-301.

MADEN, M. (1980). Structure of supernumerary limbs. Nature, Lond. 287, 803-805.MADEN, M. (1982). Supernumerary limbs in amphibians. Am. Zool. 22, (in the Press).MADEN, M. & MUSTAFA, K. (1982). The structure of APDV supernumerary limbs and a

hypothesis of their formation. (In preparation.)SLACK, J. M. W. (1980a). A serial threshold theory of regeneration. J. theor. Biol. 82, 105-

140.SLACK, J. M. W. (19806). Morphogenetic properties of the skin in axolotl limb regeneration.

J. Embryol. exp. Morph. 58, 265-288.STOCUM, D. L. (1978). Regeneration of symmetrical hindlimbs in larval salamanders.

Science 200, 790-793.STOCUM, D. L. (1980). Intercalary regeneration of symmetrical thighs in the axolotl, Amby-

stoma mexicanum. Devi Biol. 79, 276-295.TANK, P. W. (1978 a). The occurrence of supernumerary limbs following blastemal trans-

plantation in the regenerating forelimb of the axolotl, Ambystoma mexicanum. Devi Biol.62, 143-161.

TANK, P. W. (19786). The failure of double-half forelimbs to undergo distal transformationfollowing amputation in the axolotl, Ambystoma mexicanum. J. exp. Zool. 204, 325-336.

TANK, P. W. (1979). Positional information in the forelimb of the axolotl: experimentswith double-half tissues. Devi Biol. 73, 11-24.

TANK, P. W. (1981). Pattern formation following 180° rotation of regeneration blastemasin the axolotl, Ambystoma mexicanum. J. exp. Zool. 217, 377-387.

WALLACE, H. & WATSON, A. (1979). Duplicated axolotl regenerates. / . Embryol. exp. Morph.49, 243-258.

{Received 27 November 1981, revised 15 February 1982)