A monoclonal antibody detects a difference in the cellular composition of developing and regenerating limbs of newts

Adult urodele amphibians such as newts and axolotls can regenerate their limbs after amputation. Several lines of evidence support the view that regeneration and development of the urodele limb proceed by similar mechanisms. During development a bud of undifferentiated mesenchymal cells appears on the flank of the embryo and progresses through a series of stages to form the mature limb (reviewed by Hinch-cliffe & Johnson, 1980). The morphogenesis of a regenerating limb proceeds in a similar manner (reviewed by Wallace, 1981). After amputation the wound heals by formation of a specialized epithelium and a mass of undifferentiated progenitor cells, called a blastema, appears beneath it. The early blastema can be considered equivalent to the early bud in several respects. The blastemal cells differentiate and replace the missing parts of the limb by reiterating the developmental stages at the level of tissue organization and patterning of the limb. A two-dimensional electrophoretic analysis of the major proteins synthesized by mesenchymal cells of the developing bud and the regenerating blastema revealed only minor differences (Slack, 1982). Under certain experimental conditions designed to investigate the rules governing pattern formation, developing buds and regenerating blastemas behave similarly. Both respond to apposition of distal grafts to proximal stumps by intercalary regeneration of missing parts (Iten & Bryant, 1975; Stocum, 1975; Maden & Goodwin, 1980). Furthermore, both respond to inversions of the transverse axes by forming supernumerary limbs (Iten & Bryant, 1975; Tank, 1978; Wallace & Watson, 1979; Maden & Goodwin, 1980; Maden, 1980). In fact the regenerating blastema can substitute for the developing limb bud, and vice versa, in the generation of supernumerary limbs (Muneoka & Bryant, 1982, 1984).

On the other hand, there is evidence of differences between developing limb buds and regenerating limb blastemas. (1) Scadding & Maden (1986) showed in the same animal that morphogenesis of a regenerating forelimb blastema and a developing hindlimb bud is differentially affected by treatment with vitamin A (retinol palmitate). (2) The specific origin of the cells that form the regenerating blastema of urodeles has long been a matter of controversy, with much of the evidence favouring the view that the cells arise by dedifferentiation of mature tissues, in clear contrast to development (for example, Hay & Fischman, 1961; Namenwirth, 1974; reviewed by Hay, 1974; Wallace, 1981). (3) The role of the nervous system in growth control is clearly different during development and regeneration. Early limb outgrowth and development are initiated before the limb becomes innervated. In contrast, the dependence of limb regeneration on the nervous system has been recognized since the work of Todd (1823). In a comprehensive series of experiments, Singer demonstrated that limb regeneration in amphibians is dependent on the quantity of axons at the plane of amputation and that either motor or sensory axons can meet this requirement (reviewed by Singer, 1952, 1974). Denervation appears not to affect the initial appearance of blastemal cells but stops blastemal cell proliferation, thereby arresting regeneration (Singer, 1952; Kelly & Tassava, 1973; Tassava, Bennet & Zitnik, 1974; Mescher & Tassava, 1975; Maden, 1978).

One antibody, called 22/18, has provided new information about the cellular composition of the blastema, including the origin and nerve-dependent proliferation of a subset of blastemal cells. The antibody was raised by immunizing mice with a particulate fraction of blastemal cells from the adult newt, Notophthalmus viridescens (Kintner & Brockes, 1985). Screening of the resulting clones by immuno-fluorescence on sections of the early regenerating limb identified the 22/18 reagent by virtue of its reactivity with blastemal cells but not mature tissues. The 22/18 antigen is intracellular and filamentous (Fekete, Ferretti, Gordon & Brockes, 1987), appearing first in the processes of Schwann cells and interstitial cells of loose connective tissue at 36-48 h after amputation (Gordon & Brockes, 1987). By the second week after amputation, it is also expressed by muscle tissue and approximately 80% of the blastemal cells. The number of positive cells declines thereafter and reactivity with the antibody is no longer detectable some weeks before limb patterning and growth are complete (Kintner & Brockes, 1985). Most of the blastemal cells were expressing the 22/18 antigen during the time when the nerve was reported to exert its greatest influence on regeneration as assayed by the response of the regenerate to denervation (Schotté & Butler, 1944; Butler & Schotté, 1949). Interestingly, 22/18 was shown to be an excellent marker for a subset of blastemal cells whose division is dependent on the nerve (Kintner & Brockes, 1985; Brockes & Kintner, 1986).

With these facts in mind and, in particular, the utility of 22/18 as a marker of nerve-dependent cells in the early blastema, it was of obvious interest to determine if the 22/18 antigen is expressed during limb development. This paper presents evidence for a clear difference in the cellular composition of regenerating blastemas and developing limb buds as detected by immunostaining with 22/18. We suggest that this difference may be related to the role of the nerve and its accompanying Schwann cells, and discuss some future experiments to test this hypothesis.

Pleurodeles waltl embryos were supplied by Xenopus, Ltd within two days of spawning. N. viridescens embryos were bred in the laboratory from a colony of adult newts supplied by Xenopus, Ltd. Larvae were raised individually in aged tap water in plastic Petri dishes at room temperature and were fed three times per week. Bait consisted of live brine shrimp (Artemia) which were hatched in sea water and rinsed with tap water. Animals were staged according to Gallien & Durocher (1957).

Animals were anaesthetized by immersion in 1:6000 tricaine (methane sulphonate) in sterile distilled water. Prior to stage 40, the right limbs were amputated approximately midway along the proximal-distal axis using tungsten needles. At later stages, right limbs were amputated through the midhumerus using iridectomy scissors. Camera-lucida drawings of the limbs were made immediately before and after amputation and also just before fixation to assess the progress of regeneration. Animals were placed into sterile tap water following amputation and resumed swimming within a few minutes. The rate of limb regeneration tended to decrease as the animals aged and therefore survival times varied between 4-10 days. Amputation at the later stages often resulted in contraction of the soft tissues, leaving the humerus protruding from the cut surface. However, the epidermis healed over the exposed bone within 12 h. When such limbs were examined by histology at several days after amputation, the blastemas were observed to form around the circumference of the distalmost humerus (also described by Stocum, 1979). Animals were processed before the amputated limbs reached a late-bud stage (according to Stocum, 1979). Fewer than 10% of the animals were eliminated from the sample because of poorly formed blastemas.

It was necessary to remove the gelatinous capsules of embryos that were processed prior to hatching by gentle teasing with fine forceps. All larvae were anaesthetized in tricaine, then immersed for 1 h in fixative containing 0·05 % paraformaldehyde, 0·lM-lysine, 0·01 M-sodium periodate diluted either in Hepes-buffered saline (120mM-NaCl, 3mM-KCl, 0·1 ntM-CaCl₂, 5mM-Hepes, pH 7·2) or an 80% solution of Gibco phosphate-buffered saline (PBS), pH 7·4. Animals were transferred through graded sucrose solutions and embedded in gelatin/sucrose at 4°C (Godsave, Wiley, Lane & Anderton, 1984). They were trimmed, oriented and frozen onto microtome chucks cooled with liquid N₂. Blocks were stored at -65 °C until sectioned with a cryostat at 5-10μm at -25°C. Slides were acid cleaned, dipped in 1:1 ethylene glycol: egg whites and air dried at 37°C. Sections were collected sequentially onto five sets of slides so that adjacent sections could be reacted with different combinations of antibodies. In all but a few cases the larvae were sectioned transversely and all of the sections were saved from the rostral tip to the caudalmost extent of the limbs. Sections were stored in air-tight containers at -65°C for up to several months.

Sections were thawed at 37°C and in some cases they were fixed to the slide by dipping in periodate-lysine-para-formaldehyde fixative for 5min. They were thoroughly rinsed before staining and between each antibody step with PB (0·1 M-phosphate buffer, 0·02% sodium azide, pH 7·4) with or without 0·05 % Tween 20. .All antibodies were diluted in 10 % fetal or newborn calf serum in PB and incubated on the sections for 45min to 1h at room temperature.

For each animal, one set of slides was double labelled with the monoclonal antibodies, 22/18 (IgM) and 12/101 (IgG; a newt muscle marker) according to Kintner & Brockes (1984). In brief, sections were incubated simultaneously with 22/18 ascites (1:75 or 1:150, depending on the batch) and 12/101 ascites (1:100), followed by rabbit antj-mouse-IgM antiserum (Miles; 1:200), and then simultaneously with rhodamine-conjugated goat anti-rabbit-Ig antiserum (Cappel; 1:250), fluorescein-conjugated goat anti-mouse-IgG (Cappel; 1:150) and 1·25μgml⁻¹ Hoechst dye 33258 (bisbenzimide). Sections were postfixed in acidalcohol (95% acetic acid in absolute ethanol), rinsed and coverslipped with DABCO (l,4-diazabicyclo[2.2.2]octane) in 90% glycerol, 10% PB, pH 8·6.

A second set of sections was double labelled with 22/18 and a rabbit antiserum (called NP70; gift of D. Bray) directed against the low molecular weight (70x10³M_r) subunit of neurofilament protein (Jacob, Choo & Thomas, 1982). The neurofilament staining was improved by pre-treatment with acid-alcohol, but the 22/18 staining was drastically reduced on sections pretreated in this way. Therefore the following protocol was adopted: incubation with 22/18 ascites followed by fluorescein-conjugated goat anti-mouse-IgM (Miles; 1:100), then a 5 min treatment with acid-alcohol, then NP70 (1:100) followed by rhodamine-conjugated goat anti-rabbit Ig (Miles; 1:250). Sections were postfixed and coverslipped as described above.

We also tried a variety of other antibodies in double-labellingexperiments with 22/18. If the other antibody was a mouse monoclonal of the IgG subclass, such as RT-97 (directed against the high molecular weight subunit of neurofilament protein, Wood & Anderton, 1981; gift of B. Anderton), then we used a protocol similar to that de-scribed above for 22/18 versus 12/101. If the other reagent was a rabbit antiserum such as anti-glial fibrillary acidic protein (anti-GFAP; gift of M. Noble) or anti-vimentin antiserum (Jacob et al. 1982; gift of D. Bray), then we used a protocol similar to that described for 22/18 versus NP70. In some cases, sections were stained with only one primary antibody, including 22/18,22/31 (a mouse IgM monoclonal antibody; Kintner & Brockes, 1985), and Leu7 (gift of A. Mudge). Leu7 is a mouse IgM monoclonal antibody, also known as HNK-1, originally isolated as a marker for human natural killer cells (Abo & Barch, 1981) but later shown also to recognize the myelin-associated glycoprotein of Schwann cells and other neural glycoconjugates (Schuller-Petrovic, Gebhart, Lasmann, Rumpold & Kraft, 1983; McGarry, Helfand, Quarles & Roder, 1983; Kruse, Mailhammer, Wernecke, Faissner, Sommer, Goridis & Schachner, 1984; Ilyas, Quarles & Brady, 1984; Inuzuka, Quarles, Noronha, Dobersen & Brady, 1984). In adult newts it serves as a good marker of Schwann cells in the limb (Gordon & Brockes, 1987). The primary antibodies were detected by the protocol described above using a rabbit anti-mouse IgM second layer followed by rhodamine-conjugated goat anti-rabbit antiserum.

Controls for 22/18 staining included elimination of 22/18 from the staining protocol or substitution of IgM antibodies of unrelated specificity in place of 22/18. In general, background fluorescence was present throughout the embryo but was clearly distinguishable from a positive signal. The pronephric tubules and the gut gave the highest back-ground fluorescence. Autofluorescent cells were evident by virtue of their visibility under both filter systems and because their colour was different from either rhodamine or fluorescein. The tissue was viewed by epifluorescence or Nomarski optics and photographed as described by Kintner & Brockes (1985).

Tracings (X200 magnification) made using a drawing tube helped in comparing the staining patterns of different antibodies on adjacent sections. The distribution of both NP70-stained axons and 12/101-stained muscles (including myoblasts) was determined from camera-lucida tracings of every fifth section through 14 normal forelimbs at stages 35, 36, 38, 39, 40, 42 and 45. The more limited staining at earlier stages was assessed by direct observation of immunostained sections.

The 22/18 antibody was initially isolated and characterized using blastemas of N. viridescens (Kintner & Brockes, 1985). As a species, N. viridescens is difficult to breed in captivity - our own colony unexpectedly started breeding for one season, yielding a small crop of embryos. For the present developmental study, we screened several different urodele species with the intention of finding an alternative to Notophthalmus. One or more adult newts of each species (including Amby stoma mexicanum, Triturus cristatus, Triturus helvéticas and P. waltl) were amputated through the upper forelimb and allowed to regenerate to the early- to mid-bud stage (Iten & Bryant, 1973). The limbs were tested for 22/18 reactivity as previously described (Kintner & Brockes, 1985). Only P. waitl gave a similar staining pattern within the blastema (Fig. 1) to that reported previously for N. viridescens. In the axolotl, 22/18 cross reacted with the epidermis but failed to stain blastemal cells. Faint staining was observed in the severed nerves of the two species of Triturus, but no staining of blastemal cells or skin was evident. As a result of these observations, embryos and larvae of both N. viridescens and P. waltl were used in this study. There was essentially no difference in the results obtained for the two species either during normal development or following amputation.

This paper is particularly concerned with the expression during development and larval regeneration of 22/18, and its spatial and temporal relationship to nerve and muscle. Results obtained with other antibodies are presented as ancillary observations as these were generally not studied systematically throughout development.

Peripheral nerves were found to be labelled earlier in development by an antiserum directed against the low molecular weight neurofilament subunit (NP70) than by a monoclonal antibody against the high molecular weight neurofilament subunit (RT-97). The sequential expression during development of the neurofilament subunits (low and medium followed by a phosphorylated form of the high molecular weight subunit) has been reported previously for several vertebrate species (Shaw & Weber, 1982; Willard & Simon, 1983; Pachter & Liem, 1984; reviewed by Willard, Meiri & Glicksman, 1985). In the adult N. viridescens blastema, the distal tips of regenerating axons were more extensively stained using NP70 compared with RT-97 (data not shown). In the results that follow, all descriptions of limb innervation refer to observations made using the NP70 antiserum.

Because a general description of limb innervation and its relationship to cytodifferentiation is lacking for the newt, this is presented first followed by a description of 22/18 reactivity during development. The urodele limb undergoes considerable growth before nerves can be detected among the mesenchymal cells. At the earliest stages (31-32) of outgrowth, a thin nerve bundle lies adjacent to the bud but does not penetrate into the limb mesenchyme. There is no evidence of overt cytodifferentiation of either cartilage or muscle. Cartilage prepatterning is not observed when the tissue is viewed with either Nomarski optics (Fig. 2A) or a nuclear dye stain. The muscle-specific antibody, 12/101, stains the differentiating myoblasts in the adjacent flank, but does not stain cells in the bud (Fig. 2B). From the outset, mesenchymal cells can be stained with an antiserum to vimentin and with a monoclonal antibody, 22/31 (Kintner & Brockes, 1985), which gives a staining pattern very similar to anti-vimentin (Fig. 2C).

By stage 35, the bud has elongated and a few axons are present in its proximal part. A ventrally directed nerve bundle enters the limb first, followed shortly afterwards by a smaller dorsal branch emerging from the brachial plexus. At that time, a few 12/101-positive myoblasts are found near the junction between the nerve bundles. Cartilage prepatterning is apparent as a swirling of mesenchymal cells in the central part of the bud.

Throughout limb morphogenesis the nerves were found in a more distal distribution than muscle. In contrast, weak cartilage prepatterns were evident in distal parts of the limb prior to the arrival of nerves. Because many fine nerve branches had sprouted by the notch-stage (38), it was not possible to reconstruct accurately the innervation pattern from semiserial sections. We focused instead on the relative amount of nerves at a given proximal-distal level and also on the distalmost extent of growing neurites at this and subsequent stages.

Cartilage prepatterning was present up to the base of the notch at stage 38-39. The muscles of the shoulder region were clearly differentiated as discrete entities. Beyond this, two main clusters of myoblasts, a dorsal and a ventral group, were present in the proximal half of the limb. Although the main nerve branches were also confined to the proximal half, several finer bundles extended about 150-200 pm beyond the 12/101-positive myoblasts.

By the two-digit stage (40-41) the limb is innervated along most of its length, although the axons in the distal 25 % are quite fine and do not form discrete bundles. Axons have barely penetrated into the proximal part of the forming digits. At later stages the number of axons continues to increase, with progressively more fibres entering the distal limb. Individual skeletal elements are well formed by stage 45.

Initially the entire embryo was sectioned and stained in order to assess the overall distribution of 22/18. The staining of the larval epithelium was most striking and provided a convenient internal control with which to compare the staining of mesenchymal cells within the limb bud. The simple epithelium of the skin was difficult to preserve with such light fixation conditions so that many cells sloughed off during processing. It is for this reason that some of the photographs show only partial staining on the limb periphery, where the epithelial cells would be expected to lie. The staining of larval epithelium is in sharp contrast to results obtained from adult N. viridescens (Kintner & Brockes, 1985) and P. waltl (Fig. 1) where the stratified squamous epidermis is clearly not stained by 22/18.

IN the developing brain and spinal cord, 22/18-positive cells that resembled radial glial cells were observed (Fig. 3A,B). Double labelling with anti-GFAP antiserum showed extensive overlap in the staining patterns of the two antibodies. There were regions where the staining with 22/18 was brighter than with anti-GFAP, and vice versa, suggesting that the two antibodies were not necessarily recognizing the same antigen. In the adult newt brain, the 22/18-positive cells were clearly a subset of the cells stained with anti-GFAP. In the larva a few cells weakly double labelled with the two antibodies were also observed in the trunk between the aorta and the pronephric tubules, and in the gut. These cells were most evident at the earliest stages (28-30).

Very strong staining with 22/18 was observed in the lens (Fig. 3C,D) at early stages and later along the wall of the aorta. Longitudinal sections through several late-staged larvae (stage 45 and beyond) revealed that positive cells stopped short of the heart and also extended only a short distance past the dorsal bifurcation of the aorta (Fig. 3E,F). No 22/18-positive cells were observed in the aorta of adult newts.

Prior to stage 36, examination of the limb buds relied primarily on unamputated animals. At later stages, the control limbs of animals whose right limbs had been amputated (see below) were also used. These data were supplemented with a few post-stage-36 larvae which were not subjected to amputation. A total of 67 unamputated limbs ranging from stage 30 through stage 52 were used.

At all of the stages examined, fewer than 1 % of the limb mesenchymal cells were strongly stained with 22/18. A typical example of the staining pattern of 22/18 is shown in Fig. 2D. The epidermal cells of the limb bud are stained, as well as sparse filamentous staining in the limb mesenchyme. This pattern is in marked contrast to that obtained using another mouse monoclonal of the IgM subclass, 22/31, which stained all of the mesenchymal cells and nothing in the epidermis (compare Fig. 2C with 2D). In other cases there were essentially no 22/18-positive filaments detectable in the mesenchyme. A weak expression of the 22/18 antigen which is below the resolution of the immunofluorescence technique obviously cannot be ruled out, although brightly stained cells were observed in adult blastemas processed in parallel with the larval tissue and so the difference between them is unambiguous. When present, 22/18-positive staining appeared to be associated almost exclusively with cell processes, such as that shown in Fig. 2D, and was more prevalent in hindlimbs than in forelimbs. Only in rare cases was it possible to identify positive staining around the cell body of a limb mesenchymal cell. Usually it was still not possible to identify the stained cell as any differentiated type. They clearly did not represent pigment cells. In one case at stage 38; labelling with 22/18 closely overlapped with an area in which early myoblasts were differentiating, as evidenced by staining with 12/101. Several cells appeared to be double labelled by the two reagents. This result was obtained on both limbs of one N. viridescens larva, but could not be reproduced upon examination of 15 other limbs of similar stages from either N. viridescens or P. waltl.

After the limb was innervated, weak 22/18 staining was sometimes observed in association with areas through which the nerves were travelling. We suspect that these weakly positive cells could represent Schwann cells, but were unable to verify this because our marker of Schwann cells in adult limbs, Leu-7, was not specific in the larvae. The problem with Leu-7 was not an absence of staining (which might be interpreted as an absence of Schwann cells), but rather that more cells were labelled than would be expected for a Schwann cell-specific marker. Leu-7 labelled the spinal ganglia and peripheral nerves, but extended also to adjacent areas. An alternative possibility is that the 22/18-positive cells were associated with blood vessels which may be following a course similar to the nerves. It was not possible to test this due to the absence of an appropriate marker.

Because 22/18 reactivity was so limited during early limb outgrowth, we were able to assess the ontogeny of 22/18 expression in response to amputation of developing buds. The earliest stage at which we were able to amputate the limb bud and still retain a visible stump was approximately stage 34 for forelimbs, and stage 45 for hindlimbs. In each case, the right, amputated limb was compared with the left, control limb to assess both the intensity of 22/18 staining and the relative number of positive cells or cell processes. At all stages, the amputated limbs appeared to regenerate; a few animals were left to survive amputation and these regenerated normal limbs within 3-4 weeks.

Amputation of either fore- or hindlimbs at early stages does not provoke a 22/18 response when compared with the contralateral control (Fig. 4A,B). In these cases the nerve fibres have entered the limb, but most have not yet reached the plane of amputation. Forelimbs that were amputated between the notch stage (38) and the two-digit stage (40/41) and examined 3-6 days later show a few 22/18-positive cells at the plane of amputation. In several cases, we were able to verify that a few nerve fibres were also present near the amputation plane. Examination of adjacent or double labelled sections stained for 22/18 revealed a close spatial relationship between the nerve and the 22/18-positive cells. After the fore-limbs had reached a three-digit stage (44), amputation through the midhumerus resulted in a robust 22/18 response (Fig. 4C,D). Many blastemal cells were brightly labelled by 22/18 and were often located close to nerve fibres (Fig. 5). The distalmost 100 μm of the blastema tends to have fewer positive cells than the region closer to the plane of amputation where nerve bundles are more prevalent. It was not possible to determine if the 22/18-positive cells originated from Schwann cells because of the lack of a Schwann cell marker for these stages (see above). The results of these experiments are summarized in Fig. 6.

The difference between the 22/18-response of early- and late-stage limb buds is most graphically demonstrated within a single animal. For example, by stage 46 a fourth digit is beginning to form on the forelimb whereas the hindlimb has only reached a midbud stage of development. Unilateral limb amputation at this stage results in a 22/18 response in the amputated forelimb compared with the contralateral control. In contrast, there is no difference between the amputated and control hindlimbs, with both showing a small amount of 22/18 reactivity that does not differ from unamputated animals.

The staining with 22/18 has revealed a clear difference in the cellular composition of developing and regenerating limbs of newts. Fewer than 1 % of the limb mesenchymal cells were brightly labelled with the antibody during development. This contrasts sharply with limb regeneration in adults where as many as 80% of the blastemal cells are 22/18-positive at the early- to mid-bud stages. It is not clear what would be a comparable developmental stage, perhaps the pre-limb-bud stage, when the mesenchymal cells are just beginning to condense on the flank of the embryo. Examination of such limb rudiments did not reveal 22/18-positive staining; positive cells were restricted to the nonlimb tissues described above, including the skin, lens, aorta and glia of the central and enteric nervous systems. At slightly later stages, a few 22/18-positive cells were sometimes observed within the limb bud mesenchyme, particularly in the hindlimb. The present analysis cannot distinguish between an absence of the 22/18 antigen during limb development and a situation where the antigen is present but requires post-translational modification (e.g. phosphorylation) before it can be recognized by the antibody.

This distinction between the cellular composition of buds and blastemas can be considered in relation to the results of Muneoka & Bryant (1982, 1984) which led them to conclude that the rules governing pattern formation are the same during development and regeneration of axolotl limbs. Our results on newts show that at stages similar to those used by Muneoka & Bryant in their grafting experiments, buds and blastemas are clearly different with respect to 22/18 expression. Although species differences pose an obvious difficulty when comparing such studies, we suspect instead that the disparate results emphasize the difference between the processes involved in blastema formation and growth (as detected by 22/18), and those that control subsequent pattern formation.

The staining of tissues other than limb does not lend itself to simple explanation. At the moment, we know little about the function of the 22/18 antigen, its molecular characteristics or the epitope recognized by the antibody. The intracellular distribution of the antigen in situ and in cultured blastemal cells is indicative of an intermediate filament protein, or perhaps an intermediate filament-associated protein (Fekete et al. 1987). Very little is known about the functional role of this large and varied group of proteins which share at least one antigenic determinant (Pruss, Mirsky, Raff, Thorpe, Dowding & Anderton, 1981). Other antibodies tend to be more specific, usually recognizing one or more proteins within a class, but not cross reacting with those of a different class (e g. Wang, Fischman, Liem & Sun, 1985). If 22/18 is directed against intermediate filament protein(s), it has the unusual property of recognizing some but certainly not all cells which are known to contain intermediate filaments of very different classes, including GFAP, vimentin and keratins specific to simple epithelia. It is also possible that the antibody may be cross reacting with unre-lated antigens, in which case the staining of different tissues may have no functional relevance. In any case, it may be noteworthy that the ventral wall of the aorta has been identified as a source of haematopoietic stem cells in developing chicks (Dieterlen-Lievre & Martin, 1981). Thus the cells of this structure may have some unique and unusual properties that would be lost at later stages of development. In this light, it is interesting that 22/18 stains the wall of the aorta in larval but not adult newts. The presence of GFAP-containing glia in the gut has been observed in the rat (Jessen & Mirsky, 1980,1983). It is unclear why 22/18 should stain a subset of these cells or a subset of GFAP-positive glial celts in brain and spinal cord of newt.

Amputation of the limb at early stages of development yields a 22/18-negative blastema which is capable of regenerating a normal limb. At later stages, after the limb is innervated, amputation yields a 22/18-positive blastema with similar regenerative capacity. When an early-staged hindlimb and a late-staged forelimb were both amputated in a single animal, only the forelimb blastema developed reactivity to 22/18. This result suggests that the development of a 22/18 response to amputation cannot be explained by the appearance of a circulating factor.

At least two interpretations, somewhat related, could be advanced to explain the developmental transition of an amputated limb bud from a 22/18-negative to a 22/18-positive state. One possibility is that expression of the 22/18 antigen is related to a change whereby embryonic regulation gives way to epimorphic regeneration. A second possibility is that the nerve plays a critical role in the origin and subsequent proliferation of 22/18-positive blastemal cells. These two possibilities are schematically represented in Fig. 7 and will be discussed in turn.

The term ‘regulation’ has been used to describe a variety of morphogenetic phenomena that occur following the removal of tissue from different developing systems. For the present discussion, we will use a definition suggested by Slack (1980): ‘regulation’ refers to a process whereby the fates of primordial cells within an embryonic field are altered by removal of a part of the field, so that the remaining cells give rise to a greater range of structures than normal. Pre-forelimb bud rudiments of early urodele embryos exhibit regulative behaviour when half of the tissue is removed (Harrison, 1918; Slack, 1980). At later stages, amputated larval limbs are assumed to under-go the same process of epimorphic regeneration that occurs in adult limbs (e.g. Stocum, 1979). That is, the missing parts of an amputated limb are re-formed by the proliferation of undifferentiated blastemal cells that were derived from mature tissues. Between these two stages, the developing limb is at an intermediate point where some of the cells are still undetermined, while other cells are differentiating for the first time. Amputation at such stages is followed by the formation of a perfectly normal limb (Maden & Goodwin, 1980). The issue is whether this occurs by regulation or regeneration. Slack has proposed that the difference between regulation and regeneration corresponds to a difference between the response of undetermined rudiments and differentiated organs. He has emphasized the importance of dedifferentiation of mature tissue to yield a regenerating blastema, which then relies on the adjacent stump tissue to direct its subsequent morphogenesis.

When the developing limb bud regenerates after amputation, it must be doing so by a process which differs from that occurring in differentiated larval or adult limbs, since the 22/18 antigen is not expressed at the early stages. This observation may reflect the distinctions proposed by Slack in that the early 22/ 18-negative response to amputation would correspond to a situation in which there are still a sufficient number of pluripotent cells available that can alter their prospective fates. The later 22/18-positive response would correspond to a situation in which most of the cells are already determined so that now the missing parts must be reconstituted by cells that have to revert to a pluripotent condition. During this process, the cells, as in the adult limb, express the 22/18 antigen.

A second possible explanation for the ontogeny of the 22/18-response to amputation may be related to the arrival of nerves (axons plus Schwann cells) in the developing bud. Previous observations on adult limbs lend support to this suggestion. In the adult forelimb, the nerve has been shown to contribute 22/18-positive cells to the blastema and also to be required for the subsequent proliferation of these cells. In fact, these two roles of the nerve can be dissociated. The Schwann cells of the nerve are induced to express the 22/18 antigen after amputating the limb (Kintner & Brockes, 1985) or sectioning the nerve at the brachial plexus. If the limb is maintained in a denervated state, the Schwann cells continue to express the antigen for at least 6 weeks (Gordon & Brockes, 1987) . If the limb is amputated and then denervated, 22/18 is expressed by Schwann cells and blastemal cells, but their division is severely reduced (Kintner & Brockes, 1985). Both a cellular contribution to the blastema by the nerve (Maden, 1977) and a mitogenic effect of the nerve (reviewed by Brockes, 1984) have been previously described. The additional information gained by using 22/18 was that the mitogenic effect was most pronounced for only a subset of blastemal cells and that some of these originate from Schwann cells.

The proximity of the nerve to the 22/18-positive cells in regenerating larval limbs suggests that as in the adult, some of these cells might be derived from Schwann cells. Unfortunately, there is little information about the temporal relationship between the arrival of axons and the arrival of Schwann cells at different proximodistal levels of a developing urodele limb. It is thought that peripheral axons are naked when they first leave the cord and that Schwann cells migrate down the axons some time later (Harrison, 1924; Speidel, 1964; but see also Weston, 1963), although in the absence of axons or an appropriate cell marker, undifferentiated Schwann cells would be very difficult to distinguish from undifferentiated mesenchymal cells. The majority of Schwann cells arise in situ by local proliferation along the length of the axon (Asbury, 1967) rather than by continued migration from the neural crest. Electron microscopic studies of growing neurites at the base of the developing Xenopus limb indicate that they are devoid of Schwann cells at their distal tips, but that Schwann cells may be as close as 50 μm from the tip (Prestige & Wilson, 1980). The proximity of the Schwann cells may be related to the fact that nerves arrive in the vicinity of the limb prior to limb outgrowth, but do not immediately penetrate the developing bud in Xenopus and chick (Lamb, 1974; Hamburger, 1975; Bennett, Davey & Uebel, 1980). In the present study, we were not able unequivocally to identify early limb Schwann cells with any of several immunological markers that we tried, including Leu-7. Thus, we can only surmise that urodeles resemble other vertebrates in that the arrival of the Schwann cells in the limb may be slightly delayed relative to the axons. If the initial 22/18 response is due to induction of the antigen in Schwann cells, then the presence of nerves (as detected by neurofilament antibodies) would only serve as an approximate indicator of the expected location and overall number of 22/18-positive cells. The initial expression of 22/18 in response to amputation occurs several stages after nerve fibres first penetrate the limb. At these stages, the nerve bundles are quite thick in proximal parts of the limb, but generally taper down to a few ‘pioneer’ fibres in the distal parts. If the intensity of the 22/18 response is related to the number of Schwann cells at the amputation plane, then the level of the amputation would be an important variable. This might account, in part, for the variable onset of the 22/18 response in different animals.

In addition to providing a source of 22/18-positive cells in the regenerating larval limb, the nerve might also be controlling the proliferation of such cells by provision of a growth factor as it does in adult regeneration. If true, then regeneration (or regulation) would be nerve independent at early stages of development and would gradually acquire nerve dependency as more fibres sprout into the limb. A gradual shift from nerve-independent to nerve-dependent proliferation would also be manifest by an increase in the proportion of 22/18-positive blastemal cells since, as mentioned above, 22/18 is a good marker for this property in adult regeneration. At some point, the system becomes absolutely dependent on the nerve, since amputation of a denervated larval forelimb after the fourth digit has formed arrests regeneration. Prolonged denervation after amputation leads to progressive resorption of the limb, often as far as the shoulder (Butler & Schotte, 1941; Schotte & Butler, 1941).

If the limb develops in the near or total absence of innervation (the so-called ‘aneurogenic limb’), then it remains nerve independent for regeneration (Yntema, 1959). Furthermore, if such aneurogenic limbs are transplanted to normal hosts and are subsequently innervated by the host brachial nerves, then they quite abruptly become dependent on the nerve when amputated (Thornton & Thornton, 1970). We predict that, like the preinnervated developing limb, the blastema of an aneurogenic limb will not express the 22/18 antigen in response to amputation (see also discussion by Brockes, 1984). Likewise, the blastemal cells that arise will not be dependent on the nerve-derived growth factor but will respond to non-neural growth factors. We further predict that after innervation of the aneurogenic limb, 22/18-positive cells will be generated which can only respond to the nerve-derived growth factor. That is, the limb will have been restored to its normal state of nerve dependence after amputation. Experiments are currently under way to test the prediction that an amputated, aneurogenic limb regenerates via a 22/18-negative blastema. If it does, then the acquisition of the 22/18 response to amputation does not control the transition between regulation and regeneration, but rather represents the imposition of the nerve-dependent system of growth control.

We thank Frank Wilson and Susan Hall for their assistance with animal care and Brian Boycott, Malcolm Maden and Jonathan Slack for critical reading of the manuscript. Brian Anderton, Dennis Bray, Anne Mudge and Mark Noble kindly supplied us with antibodies. We thank Herman Gordon for his help with the pilot experiments and Chris Kintner who isolated and characterized the monoclonal antibodies 22/18, 22/31 and 12/101. D.M.F. is a Smith-Kline and French Fellow of the Life Sciences Research Foundation.