Compatible limb patterning mechanisms in urodeles and anurans.

We have experimentally tested the similarity of limb pattern-forming mechanisms in urodeles and anurans. To determine whether the mechanisms of limb outgrowth are equivalent, we compared the results of two kinds of reciprocal limb bud grafts between Xenopus and axolotls: contralateral grafts to confront anterior and posterior positions of graft and host, and ipsilateral grafts to align equivalent circumferential positions. Axolotl limb buds grafted to Xenopus hosts are immunologically rejected at a relatively early stage. Prior to rejection, however, experimental (but not control) grafts form supernumerary digits. Xenopus limb buds grafted to axolotl hosts are not rejected within the time frame of the experiment and therefore can be used to test the ability of frog cells to elicit responses from axolotl tissue that are similar to those that are elicited by axolotl tissue itself. When Xenopus buds were grafted to axolotl limb stumps so as to align circumferential positions, the majority of limbs did not form any supernumerary digits. However, in experimental grafts, where anterior and posterior of host and graft were misaligned, supernumerary digits formed at positional discontinuities. These results suggest that Xenopus/axolotl cell interactions result in responses that are similar to axolotl/axolotl cell interactions. Furthermore, axolotl and Xenopus cells can cooperate to build recognizable skeletal elements, despite large differences in cell size and growth rate between the two species. We infer from these results that urodeles and anurans share the same limb pattern-forming mechanisms, including compatible positional signals that allow appropriate localized cellular interactions between the two species. Our results suggest an approach for understanding homology of the tetrapod limb based on experimental cellular interactions.

INTRODUCTION 9 It has long been recognized that an understanding of developmental principles may enhance our understanding of evolutionary processes (Goodwin, 1982). Historically, studies of phylogeny and homology have been based on analyses of morphological similarities and differences. However, it is not the pattern itself but the genetic information that underlies the generative principles of pattern formation and morphogenesis that is inherited. Ultimately, therefore, a complete understanding of evolutionary processes will depend on an understanding of the generative processes of form and their evolution.
The tetrapod limb represents an opportunity to begin integrating developmental and evolutionary concepts, in part because of the historical background of morphological studies and in part because of the extensive experimental analyses of the mechanisms of limb outgrowth and patterning. The wide diversity seen in limb structure among living and fossil tetrapods is usually interpreted as adaptive variation imposed on a common ground plan inherited from a common ancestor. Nevertheless, differences both in the pattern of skeletal structures and the sequence of their differentiation have been interpreted to indicate that there may be a dichotomy in basic limb pattern and patterning mechanisms within the tetrapods, with anurans and amniotes on one side and urodeles on the other (Holmgren, 1933;Jarvik, 1980; see reviews by Shubin and Alberch, 1986;Hanken, 1986). This interpretation has been used to support the idea that urodele limb structure and development are unique among tetrapods (Holmgren, 1933;Jarvik, 1980). It is the issue of whether or not a dichotomy in limb patterning mechanisms exists within the tetrapods that we address in this paper. 9 The well-studied phenomenon that allows for a direct test of the similarity of developmental mechanisms between tetrapods is the ability of limb cells to make supernumerary limbs in response to tissue rearrangements that bring about positional disparities (see Bryant et aL, 1987). Hence in urodeles, anurans, chicks, and mammals, when anterior and posterior limb cells are confronted, position-dependent growth and patterning involving communication between cells results in the formation of supernumerary limbs (see Maden, 1981;Wanek et al., 1988). We have used the formation of supernumerary limbs as an assay to determine whether the limb cells of a urodele (Ambystoma mexicanum) and an anuran (Xmmpus laevis) are able to communicate with each other in a position-dependent way. Limb buds of the two species were reciprocally transplanted to create positional disparities. Our results show that position-dependent interactions occur between the cells of the two species and that their cells can cooperate to build recognizable limb skeletal elements. We conclude 0012-1606/89 $3.00 Copyright 9 1989 by Academic Press, Inc. All rights of reproduction in any form reser~'ed.
29-i that urodeles and anurans share compatible intercellular positional signals that are utilized in a common limb patterning mechanism.

MATERIALS AND METHODS
Experiments were performed on Mexican axolotls (A.
me:"~ica~zu'm) and South African clawed toads (X laevis) spawned at the UCI Developmental Biology Center. Anima!s were reared at room temperature (20~ in saline, and were changed to fresh saline and fed three times a week. Axolotls were kept in 25% Holtfreter's solution and were fed tubifex worms. Xenopus larvae were kept in 10% Steinberg's solution with 1% humie acid and were fed nettle powder. Axoloti and Xe~to~s larvae were matched for size and stage of their hind Hmb buds. Xe~oN~s larvae were at stage 52-53 (Nieuwkoop and Faber, 1975) and axolotl larvae were also at a stage just prior to d~git formation in the hind limb. Matched pairs of animals were anesthetized in 1:4000 MS222, and transferred to 20% Steinberg's solution with 100 U/ml penicillin and 50 #g/ml streptomycin sulfate. The distal one-third to one-half of the hind limb buds were amputated and then grafted reciprocally to the limb stumps of the other species (Fig. 1). Grafts were made either ipsilatera!ly to maintain normal orientation (control grafts) or eontraiateraliy to reverse the anterior-posterior orientation of graft and host while main:aining the normal dorsal-ventral orientation (experimental grafts). Grafts were allowed to heat in place for 10-15 rain afver which time axolotl hosts were placed in individual oneliter plastic boxes containing 25% Holtfreter's solution, and XenoI.~s hosts were returned to similar containers of 10% Steinberg's solution with 1% humic acid. Developing limbs were examined and drawn three times a week using a camera Iueida. Limbs representing various developmental stages were collected from 6 days to sev-AA PP AP FIO. 1. Diagram of the grafting procedures used in this study. Developing limb buds were reeiproeaily grafted between young larvae of Xenopv..~ lae.vis (upper, shaded) and AmNdsto*na mexicanum (lower, unshaded). Experimental eontralateral grafts (a) result in positional confrontations between anterior and posterior limb tissues of host and donor, while control ipsilaterai grafts (b) do not. erat weeks after grafting when digit formation was judged to be complete or (in Xenopus hosts) when the graft tissues began to exhibit signs of immunological rejection by the host. All limbs were preserved in aqueous Bouin's fixative, stained with either victoria blue B or aieian blue, and cleared in mothy1 salieylate for whole-mount analysis. In scoring the final pattern of digits,, only elements that clearly articulated with more proximal elements were counted. Minor bifurcations without segments were not coun~ced. Selected limbs were later embedded in para~n, sectioned, and stained with hematoxylin and eosin or Mallory's triple stain for histoiogieal analys~s.

Xenopus Grafts onto Axolotl Hosts
A total of 53 successful grafts of Xenopus limb buds onto axototl host limb bud stumps was performed, of which 27 were contralateral (experimental) and 26 were ipsilateral (control) grafts. The grafts became vaseularized within a few days of operation. In 6 eases, we observed that the Xenopus graft was displaced by autonomous growth from the axolotl stump, and these cases are not included in the results below.
An analysis of the final pattern of digits of wholemount preparations of 18 experimental and 20 control limbs revealed that all experimental limbs formed supernurnerary axolotl digits (Fig. 2a), whereas the rna-joriLy of control limbs (60%) formed no axolotl digits ( Fig. 2b; Table 1). Taken as a whole, experimental limbs produced nearly four times as many supernumerary axolotl digi~cs as control limbs (Tab]e 1). In addition, all experimental limbs were complete in the proximal-distal axis (Fig. 2a), whereas in the majority of the control limbs (i.e., those that did not form supernumerary digits) the host was truncated proximally (Fig. 2b). The supernumerary digits that were formed by a minority of the control limbs (Table 1) probably arose as a result of small positional mismatches between the host and the ipsilateral graft. Such mismatches presumably also occur when grafting within species, but are most likely resolved by back rotation to align positional values of host and graft (see Harrison, 1921). Back rotation in this grafting combination might be precluded by the enormous difference in cell size between axolotls and Xen~pus .
In the Xenoy~s graft/axolotl host combination, the Xenopus grafts neither grew extensiveiy nor formed well-developed digits, despite the fact that they became well vascuiarized and innervated (Fig. 3). However, in both whole-mount and histological preparations, where the dramatic difference in celt size between ~he two species allows for unambiguous identification, it is clear DEVELOPbIENTAL BIOLOGY VOLUME 131, 1989 FIG. 2. (a) Whole-mount skeletal preparation of an axolotl host/Xenopus donor experimental limb with seven supernumerary axolotl digits, including two that have fused proximally. This limb also has supernumerary tarsals. (victoria blue/methyl salicylate), x16. (b) Whole-mount skeletal preparation of an axolotl hest/Xenopus donor control limb. The Xenopus graft has formed articulated skeletal elements (arrow) and no axolotl skeletal structures have formed distal to the host stylopodium (victoria blue/methyl salicylate). • that the grafts were not rejected but remained healthy and formed small, articulated limb cartilages (Figs. 2b and 4). In many cases the Xenopus and axolotl cartilage cells were smoothly integrated into single elements (Fig. 4b).
Pronounced supernumerary outgrowths formed anterior and posterior to the graft-host junction within 1 to 2 weeks after grafting. Histological examination of such early outgrowths showed that the majority (13) were composed primarily of axolotl cells; in four cases a small tongue of Xenopus cells also projected into the outgrowth. Two additional outgrowths were composed of Xenopus cells alone and three were chimeric and consisted of approximately equal amounts of Xenopus and axolotl tissue (Fig. 5). As can be seen from Fig. 6, tissues from both species are actively growing at this stage, and the chimeric border is sharp. Multiple outgrowths from six older limbs were examined histological!y, and in these the contribution pattern had changed such that axolotl cells predominated and eventdally formed all of the supernumerary structures.  • appeared to be healthy and rapidly growing. However, all axolotl grafts were eventually rejected by Xenopus hosts. With this limitation, weare nevertheless able to report a clear difference in the behavior of control and experimental axolotl grafts. In both controls and experimentals, axolotl digits appeared in an anterior to posterior sequence, as they do in normal ungrafted limbs. In the grafted limbs, however, rejection occurred prior to formation of all digits, although on average three or four well-developed axolotl digits formed before rejection. In those limbs prepared for whole-mount analysis, the major difference between experimental and control grafts is that in experimental grafts, an additional digit formed anterior (relative to the graft) to the normal axolotl digit I in 15 of 17 cases (Fig. 7). This additional digit was unambiguously identified in every case as a supernumerary axolotl digit 2 based on three independent criteria: its position relative to the other identified axolotl digits, its time of development (i.e., after the formation of the normal axolotl digits 2 and 1), and its articulation pattern as observed in whole-mount preparations (i.e., the metatarsal articulates on the most anterior tarsal element along with the metatarsals of the normal axolotl digits 2 and 1). In addition, one of these experimental grafts formed a second supernumerary axolotl digit (digit 3) on the anterior edge of the graft. In contrast, none of the 20 control grafts examined as whole mounts formed any supernumerary axolotl digits (Fig. 8).

Axolotl Grafts onto Xenopus Hosts
Axolotl tissues are readily identified in both wholemount and histological preparations by the much larger size of the axolotl cells (Fig. 9). This cell size difference between axolotl and Xenopus enabled us to conclude that half of the supernumerary axolotl digits contained some cartilage elements, including joints, that were chimeric, with the edge closest to the Xenopus host consisting of Xenopus cells (Fig. 9a). This conclusion was confirmed by subsequent histological examination of four of the whole mount preparations (Fig. 9b). This chimerism was always "appropriate" in that axolotl tarsals formed by the graft (e.g., radiale and tibiale), recognizable by their position and patterns of articula-tion, were confluent with the equivalent Xenopus host element (Figs. 7, 8, and 9b). We have also histologically examined the early changes at the graft-host junction in three control and three experimental limbs fixed after 1 week. Only the experimental combinations showed evidence of lateral outgrowths anterior or posterior to the graft-host junction. Out of a total of four outgrowths, three were clearly chimeric with distinct graft-host boundaries (Fig. 10) and the other was composed only of axolotl cells.

DISCUSSION
In this paper, we have experimentally tested the similarity of limb pattern-forming mechanisms in urodeles and anurans. A previous analysis of the cellular contribution to supernumerary limbs in Xenopus provided indirect evidence that the patterning mechanism in Xenopus is the same as that in axolotl (Muneoka and Murad, 1987). To directly test the compatibility of patterning mechanisms between urodeles and anurans, we compared the results of two kinds of reciprocal limb bud grafts between Xe~zopus and axolotls: contralateral grafts to confront anterior and posterior positions of graft and host, and ipsilateral grafts to align equivalent circumferential positions. Previous studies have consistently shown that positional disparities created by grafts of whole or portions of developing or regenerating limb tissues result in the formation of supernumerary structures in each of a wide variety of organisms including urodeles, anurans, chicks, mammals, and insects (see French et aL, 1976;Maden, 1981~ Wanek et a~, 1988, whereas grafts that do not result in positional disparities do not form supernumerary structures. Our premise is that if the mechanism controlling growth and pattern specification is the same in urodeles and anurans, then experimental (contralateral) grafts between axolotls and Xenopus will develop supernumerary structures and control (ipsilateral) grafts will not.  share the same limb patterning mechanism including compatible intercellular patterning signals. The differences in structure and sequence of development that have been previously described between ur0deles and anurans (Holmgren, 1933;Shubin and Alberch, 1986) cannot, therefore, reflect differences in the basic limbforming mechanism. Direct evidence that amniotes share a common limb patterning mechanism including compatible intercellular patterning signals was obtained by Fallon and Crosby (1977), who grafted pieces of limb bud tissues from reptiles, mammals, and birds into limb buds of chicks. Our results, in conjunction with those of Fallon and Crosby (1977), suggest that all tetrapods share the same basic limb patterning mechanism inherited from a common ancestor. Furthermore, we can conclude that intercellular positional signals are compatible among amniotes, on the one hand, and between the two groups of limbed amphibians on the other. The common, basic patterning mechanism is apparently ancient, and appears to be based on a fundamental property of cell interaction that is characteristic of epimorphic systems in both vertebrates and invertebrates (Bryant and Simpson, 1984).