Retinoic acid proximalizes level-specific properties responsible for intercalary regeneration in axolotl limbs

The blastema cells of regenerating amphibian limbs inherit from their parent limb cells a memory of their axial levels of origin (positional memory, Carlson, 1975; Wolpert, 1971). Positional memories define regenerate boundaries, and limb axes can be viewed as a series of such memories, with each level-specific set determining the pattern of the base of a potential regenerate (Stocum, 1984).

We have shown that a discontinuity in proximodistal (PD) positional memory, made by grafting an extra blastema to a regenerating axolotl limb, can be eliminated by a sorting process in vivo (Crawford & Stocum, 1988). To make the discontinuity, wrist or elbow blastemas were grafted perpendicular to the blastema-stump junction of a host hindlimb that was regenerating from the mid-thigh. The grafted blastemas displaced distally to their equivalent level of origin on the host regenerate, where they developed as they would have in situ. This displacement behaviour, which we have termed ‘affinophoresis’, indicates the existence of level-specific differences in the affinity of blastema cells for one another that mediate the recognition, and elimination, of positional disparity.

Retinoic acid (RA) proximalizes PD positional memory in regenerating urodele limbs (Niazi & Saxena, 1978; Maden, 1982; Thoms & Stocum, 1984). Using affinophoresis as an assay, we found that RA coordinately proximalized blastema cell affinity and positional memory (Crawford & Stocum, 1988). Wrist and elbow blastemas of RA-treated animals formed structures proximal to their level of origin, in addition to what they normally would have regenerated, and their distal displacement on the host regenerate was proportional to the degree of pattern proximalization. These observations indicated that the cellular basis of positional memory is directly related to the differential affinity properties of blastema cells.

A different mechanism of eliminating pattern discontinuities in regenerating limbs is intercalary regeneration, which occurs after making a discontinuity in positional memory by juxtaposing the cross-sectional wound surfaces of two different axial levels. The intermediate structures normally present between the graft and host levels are then regenerated (Bohn, 1965; Iten & Bryant, 1975; Stocum, 1975).

Intercalary regeneration in the PD axis of amphibian limbs is polarized in two ways. First, it occurs only in a proximal to distal direction, the cells of the host level providing the material for the intermediate structures, while the graft redifferentiates according to its level of origin (Pescitelli & Stocum, 1980; Maden, 1980). Second, discontinuity in positional value is recognized only when a blastema is grafted to a more proximal level. The discontinuity created when a blastema is grafted to a more distal level goes unrecognized (Stocum & Melton, 1977). The situation in the amphibian limb is therefore different to that in the insect limb, where intercalation can occur from either graft or host tissue, and after either distal to proximal or proximal to distal transplants of leg segments (Bohn, 1976; Bart, 1988).

Although intercalary regeneration and affinophoresis represent different corrective responses to pattern discontinuity, it seems likely that both processes would use the same differential affinity-based mechanism to recognize disparity in positional value. This hypothesis predicts that RA would abolish the capacity of a blastema to evoke intercalary regeneration, provided the positional memory of the blastema were proximalized to a level equal to or exceeding that of the more proximal level to which it was grafted. This prediction was borne out by the results of the experiments reported here, in which RA-treated wrist and elbow or ankle and knee blastemas were unable to evoke intercalary regeneration when homografted to a midstylopodial amputation level.

Larval axolotls (Ambystoma mexicanum) were provided by the Indiana University axolotl colony or were obtained from spawnings of our laboratory stock. They were reared individually in 50% Holtfreter solution in waxed paper cups, at room temperature (21°C), and fed freshly hatched brine shrimp or frozen brine shrimp every other day. The animals weighed 2 to 5-5 g and were 5 to 9 cm in length at the time of amputation.

Fig. 1 illustrates the experimental design. Animals were anaesthetized in Benzocaine (Sigma) dissolved in 100%

Holtfreter solution (0·007% w/v), and placed on mats of Benzocaine-soaked filter paper. Donor forelimbs were amputated through the wrist or elbow and donor hindlimbs through the ankle or knee. Donor RA-treated animals were injected via microlitre syringe with 100 μg of all-trans RAg⁻¹ body wt, 4 days postamputation, as described previously (Kim & Stocum, 1986a,b). The RA was dissolved in dimethylsulphoxide (DMSO) at a concentration of 50 μg μl^{− 1}. Donor control animals were injected with 2 μ l of DMSOg⁻¹ body wt, 4 days postamputation. Retinoic acid and DMSO were from Sigma Chemical Co.

Donor limbs were allowed to regenerate to the late bud stage (staging according to Stocum, 1979). The blastemas were then homografted ipsilaterally to the wound surface of a host limb freshly amputated through the midstylopodial level, aligning the axes of graft and stump. To distinguish between donor- and host-derived tissues in the regenerates, donor blastemas were exchanged between forelimbs and hindlimbs of albino or white axolotls and dark axolotls, using pigmentation differences between donor and host, and skeletal differences between forelimbs and hindlimbs, as markers. Some of the white animals used were eyeless mutants; this mutation, however, has no effect on regeneration.

Operated limbs were allowed to regenerate for 5-7 weeks, after which they were fixed in Gregg’s fixative and stained in toto for cartilage by the method of Fox (1982), with methylene blue substituted for Victoria Blue B. In order to preserve melanin pigment in the skin, the bleaching step was omitted from the staining procedure, except in a few cases. Donor Limb skeletons were distinguished from host limb skeletons by the different numbers, shapes and patterns of skeletal elements (for a detailed description, see Pescitelli & Stocum, 1980). The forelimb has a 4-digit hand, and a pronounced anterior bend at the elbow, whereas the hindlimb has a 5-digit foot, and a slight dorsal to ventral curvature, but no anterior bend at the knee.

Figs 1 and 2 summarize the results, excluding cases in which the grafts failed to survive. Two ankle and one elbow-level DMSO-treated grafts were completely resorbed, followed by the regeneration of normal host hindlimbs and forelimbs. All of the 26 surviving grafts developed as they would have in situ, having the donor pattern and shape of skeletal elements (with minor deviations from normal). The missing intermediate segments were regenerated, and their skeletal elements had the shapes and patterns of the host limb, indicating that the presence of the graft had evoked intercalary regeneration of the intermediate structures from the host amputation level (Figs 3–5). The pigment pattern of the regenerate also indicated the derivation of the intermediate structures from the host limb. The junction between pigmented and unpigmented skin coincided to a high degree with the skeletal graft-host junction (Figs 3–5).

In the RA-treated series, seven grafted blastemas resorbed completely, followed by regeneration of normal host limbs. 58 grafts survived, 52 (90%) of which were proximalized to the level of the girdle or proximal stylopodium. These grafts developed into regenerates that were complete limbs having donor character. No intercalary regeneration from the host amputation level was observed in these cases.

Fig. 6 shows a chimaeric limb resulting from grafting a RA-treated wrist blastema from a dark animal to the mid-thigh of a white animal. The photograph is a negative print, which makes the donor skin appear iridescent. The regenerate is clearly a complete forelimb, indicating that RA proximalized the positional memory of the blastema to the level of the proximal upper arm. Figs 7-12 illustrate methylene-blue-stained whole mounts of RA-treated regenerates. Each regenerate has the morphology, pattern and polarity of skeletal elements typical of a donor limb regenerating from the proximal end of the stylopodium or from the girdle. In most of these cases, the junction between pigmented and unpigmented skin coincided with the graft-host junction defined by skeletal morphology and pattern but in a few cases the host pigmentation pattern was observed to lie over skeletal elements of clearly donor origin or vice versa (Fig. 9). Hence as reported previously, skin pigmentation by itself is not always a wholly reliable marker (Stocum, 1980).

RA-induced proximalization of positional memory creates a PD-reversed disparity of half a stylopodial segment at the graft-host junction (from juxtaposition of a girdle or proximal stylopodial positional value of the graft immediately distal to the midstylo-podial value of the host). Nevertheless, no recognizable distal to proximal intercalary regeneration took place from the donor tissues. No stylopodial structures exhibiting reversed PD polarity, which would have been indicative of intercalation from the donor, were observed.

Six of the RA-treated regenerates developed hypo-morphically, with unidentifiable cartilage condensations. Two white-dark chimaeric limbs with hypo-morphic regenerates are shown in the negative print of Fig. 13. The boundary between donor and host tissue is sharply defined by skin pigmentation. Fig. 14 is a methylene-blue-stained hypomorphic specimen illustrating the unidentifiable cartilage formed by hypomorphic regenerates. No intercalary regeneration was observed in any of the cases with hypomorphic regeneration.

The objective of this study was to determine whether a RA-treated regeneration blastema loses its ability to evoke intercalary regeneration when it is grafted to a level proximal to its level of origin. The results clearly indicate that RA treatment does abolish this ability. DMSO-treated control blastemas derived from the wrist and ankle or elbow and knee levels evoked intercalary regeneration of the deleted intermediate structures, when grafted to a midstylopodial amputation level. The control regenerates were chimaeric, because the donor blastema formed the structures it would have in situ and the host stump contributed the intermediate structures. This result confirms the results of similar experiments in which triploid blastemas were grafted to more proximal levels of diploid limbs (Pescitelli & Stocum, 1980), and indicates that DMSO by itself has no effect on the ability of proximally shifted blastemas to evoke intercalary regeneration.

By contrast, none of the regenerates on limbs that had received grafts of RA-treated blastemas were chimaeric. Instead, they were of strictly donor character, indicating their origin solely from the grafted blastema. Furthermore, the RA-treated grafts formed complete limbs (i.e. were proximalized to the level of the proximal stylopodium or girdle), rather than just the parts that were cut off. This degree of proximalization created a half-segment, PD-reversed, pattern discontinuity in these limbs by establishing a proximal stylopodial or girdle positional value immediately distal to a midstylopodial value. The discontinuity, however, was not filled in by intercalary regeneration.

The lack of intercalary regeneration following the RA-induced creation of a PD-reversed pattern discontinuity is a result identical to that observed after grafting an untreated blastema to a more distal amputation level. For example, grafting a native midupper arm blastema to the tarsus level of the hindlimb creates a very large PD-reversed pattern discontinuity. No intercalary regeneration takes place, however, and the regenerate simply forms what it would have in situ, i.e. a distal half stylopodium and a whole zeugopodium distal to the host tarsus (Stocum & Melton, 1977). Hence RA-treated wrist and ankle or elbow and knee blastemas are equivalent to native blastemas derived from the proximal end of the stylopodium or from the girdle.

A few RA-treated donor blastemas developed hypomorphically on their host limbs. Hypomorphic or inhibited regeneration routinely occurs when animals are given doses of RA that exceed the dose causing maximum proximalization (Thoms & Stocum, 1984; Niazi et al. 1985; Kim & Stocum, 1986a,b). Thus the dose of RA used here may have exceeded the optimum in a few animals, resulting in their hypomorphic regeneration. Nevertheless, the positional value of the blastemas in these cases was either proximalized, or erased, or other cell functions necessary for intercalary regeneration were inhibited by RA, since intercalary regeneration from the host amputation level was not observed. Hypomorphism in itself is not a likely cause of the failure of the graft to evoke intercalation, because double-anterior-half blastemas, which give rise to symmetrical hypomorphic regenerates, nevertheless support intercalary regeneration when grafted to more proximal levels (Stocum, 1981).

Our results show that the cellular mechanism that recognizes the disparity between non-neighbouring cells and initiates intercalary regeneration, is coordinately proximalized with positional memory (regenerate pattern). Hence, we conclude that this recognition mechanism and positional memory are directly related. Since the mechanism that recognizes the positional disparity leading to distal displacement during affinophoresis is also proportionally proximalized with regenerate pattern by RA (Crawford & Stocum, 1988), it is probable that both affinophoresis and intercalation use the same mechanism to recognize pattern discontinuity. As discussed elsewhere (Crawford & Stocum, 1988), the molecular components of this mechanism are most likely associated with the cell surface, and RA probably affects them by modifying genomic transcription, through the mediation of two nuclear retinoic acid receptors (Petkovich et al. 1987; Giguere et al. 1987; Brand et al. 1988; Benbrook et al. 1988; Hashimoto et al. 1988).

In our interpretations, we have assumed that positional memory is a function of individual blastema cells. An alternative view is that individual blastema cells do not possess a positional memory. Rather, the PD level at which the regenerate begins is determined by how much blastemal mesenchyme is available to be acted upon by pattern-specifying mechanisms prior to redifferentiation. This amount is, in turn, controlled by the rate of proliferation of blastema cells (Faber, 1976). In support of this idea is the fact that upper arm blastemas are larger than lower arm blastemas at the onset of redifferentiation (Iten & Bryant, 1973; Stocum, 1979). Furthermore, DeBoth (1970) reported that zeugopodial elements were formed in a small number of cases, when several wrist blastemas (which singly would form hands) were massed together in the eyesocket. Thus, in this view, exogenous RA would proximalize regenerate pattern by increasing the rates of blastema cell proliferation or extending the period of proliferation, prior to redifferentiation.

Several things militate against this view. First, we have been unable to repeat DeBoth’s results by massing wrist blastemas in dorsal fin tunnels (D.L.S. unpublished data). Second, RA actually inhibits mitosis of dedifferentiated limb cells for 7-10 days prior to the emergence of a proximalized blastema (Maden, 1983; Kim & Stocum, 1986b). Third, the undifferentiated blastemas of RA-treated animals appear to be no larger, and are probably smaller, than comparable control stages, although the density of their mesenchymal cell population may be somewhat higher than in controls (Kim & Stocum, 1986b). Finally, there is the question of what it is that controls the level-specific growth of the blastema? Thus we think it unlikely, but cannot strictly rule out, that RA proximalizes regenerates by affecting blastema cell proliferation. The affinophoresis and intercalation assays can both be used for a more rigorous test of this idea.

The cellular behaviours associated with affinophoresis and intercalary regeneration subsequent to the disparity recognition event represent two different ways of eliminating pattern discontinuities and both are consistent with Steinberg’s (1963, 1978) differential adhesion model; i.e. they restore maximum cohesiveness between like cells. Which one is used would depend on which path is energetically most favourable in reducing or eliminating the discontinuity, within the mechanical constraints imposed by the experimental conditions. Distal displacement is favoured when wrist or elbow blastemas are grafted to the dorsal surface of the blastema-stump junction of a hindlimb regenerating from the mid-thigh, but intercalary regeneration at an angle to the longitudinal axis of the host limb occurs under these conditions, in a minority of cases in which distal displacement fails (Crawford & Stocum, 1988). These facts suggest that intercalary regeneration is triggered if sufficient exchange toward a normal neighbour pattern does not take place. In accord with this idea are observations that if corrective rotation to reestablish a normal neighbour pattern does not take place after 180° rotation of amphibian embryonic limb discs and larval regeneration blastemas, supernumerary limbs arise through the interaction of positionally disparate graft and host tissues (Harrison, 1921; Maden & Turner, 1978).

What determines that intercalary regeneration will occur when displacement fails to restore a normal neighbour pattern? First, some threshold disparity that can be sensed by the recognition mechanism probably must be equalled or exceeded. In rotated limb regeneration blastemas, for example, the minimum angle of rotation required to produce a supernumerary appears to be 20° (Maden & Turner, 1978). Second, it is likely that the length of time a particular disparity is sensed by two adjacent cells is important. Two groups of cells having different positional values and undergoing relative displacement would not stay in contact with one another long enough to trigger intercalary regeneration but, if they stopped displacing for any reason before reducing the disparity below the threshold, the resulting increased duration of the discontinuity signal between them may trigger intercalation. What the molecular mechanisms are that measure signal duration and translate the signal into the process of displacement or intercalation remain to be determined.

We thank Jo Ann Cameron for suggestions and critical comments on the manuscript. This research was supported by NIH Grant HD 12659.