Size dependence during the development of the amphibian foot. Colchicine-induced digital loss and reduction

In 1949 Bretscher reported a series of experiments which showed that local treatment of the early limb bud of the frog Xenopus laevis with the mitotic inhibitor colchicine resulted in the reduction, and even loss, of digits. These experiments were further elaborated and discussed in Bretscher & Tschumi (1951) and Tschumi (1953; in this paper the development was inhibited using chloroethylamine). These authors interpreted their results in terms of competition for cells among the various skeletal elements during chondrogenesis. These competitive interactions, again according to Bretscher & Tschumi, in a limb bud with an artificially reduced number of cells resulted in some welldefined patterns of digital reduction. These regularities were particularly evident in the fact that some digits were affected more frequently than others. For example, digits 3 and 4 were very rarely affected, while digit 1 (the ‘thumb’) was always the most reduced. Based on that information, they pointed out the potential phylogenetic implications of their results and drew an analogy between their experimentally observed patterns of digital reduction and loss and the phyletic trend toward limb reduction observed in some groups of reptiles and mammals.

(At that time, Bretscher & Tschumi did not know of any anuran which had lost either a complete digit or single phalanges.) Rensch (1959), Waddington (1962) and Devillers (1965) reported Tschumi’s results and their potential implications for evolutionary studies. However, their work has been mostly forgotten and it has had very little impact in the literature on evolution or development. Nevertheless, it was an appealing system in which any developmental perturbation that resulted in a limb bud with a smaller number of primordial cells could indirectly result in the loss or reduction of digital elements. This evolutionary issue is dealt with in a related paper (Alberch & Gale, 1983).

These experiments also posed some interesting developmental questions deserving closer scrutiny. One, in particular, is the relationship between size and pattern formation in embryonic processes. Most theoretical models of pattern formation (e.g. Turing, 1952; Kauffman, Shymko & Trabert, 1978; Newman & Frisch, 1979; Murray, 1981; Meinhardt, 1982) show some degree of size dependence, i.e., the final pattern is dependent on the size of the embryonic field at the time of the process. In fact, some models, like Kauffman et al. (1978) and Newman & Frisch (1979), use tissue growth as an organizing agent in their developmental systems. Conversely, Cooke (1979, 1981, 1982), Tam (1981), Maden (1981a) and Dan-Sohkawa & Sato (1978) among others have presented experimental evidence that some pattern formation processes are scale independent. Obviously, these two views need not be mutually exclusive. Some embryonic processes might be largely scale independent while others might not. Even the same system could be scale independent at some stage in its development and not at some other.

Bretscher & Tschumi envisioned amphibian limb development as dependent on the size (roughly equivalent to number of undifferentiated cells) of the embryonic limb field. Other authors have also shown, since then, that the treatment of developing limbs with various mitotic inhibitors triggers loss and/or reduction of elements (e.g. Kieny, 1975; Scott, Ritter & Wilson, 1977; Raynaud, 1981). Patterns similar to the ones reported by Bretscher & Tschumi resulted regardless of the experimental taxa used (amphibians, reptiles, birds or mammals). That is, the loss of elements was not chaotic but it exhibited well-defined regularities. Most of these authors, however, were interested in mechanistic problems and did not emphasize the comparative aspects of their results. Obviously not all teratogens induce digital or phalangeal loss (see Tickle & Wolpert, 1981, for a review on teratogenic effects in limb development). For example, Scott et al. (1977,1980), Klein, Scott & Wilson (1978) and Scott (1981) have discussed other drugs that induce polydactyly by affecting the timing of onset in the physiological necrosis that characterizes the normal development in higher vertebrates. This would not apply to our system since cell death does not appear to play a role in amphibian limb development (Cameron & Fallon, 1977a).

In this paper we report on a series of experiments where we modify and expand on Bretscher & Tschumi’s experimental protocol. We locally treated with colchicine the limb buds of the anuran, Xenopus laevis, and the urodele, Ambystoma mexicanum, at various stages of limb development. Unlike Bretscher & Tschumi, we focused on the differentiation of individual skeletal elements rather than on external size measurements. Also, we controlled for specific developmental stage at the time of treatment. Bretscher & Tschumi did not specifically state when they treated their Xenopus, although their illustrations seem to indicate that they used stage-52 animals (Nieuwkoop & Faber, 1956; Fig. 1). We give a detailed quantitative account of the experimentally induced morphologies. The experimental results corroborate the ones obtained by Bretscher & Tschumi and pose interesting questions relating to current models of limb morphogenesis and to qualitative differences between urodele and anuran limb development.

For this study, laboratory strains of Xenopus laevis (bred in our laboratory or pre-limb-bud-stage tadpoles purchased from Carolina Biological Supply) and Ambystoma mexicanum (eggs obtained from the Axolotl Colony, Indiana University) were used. The axolotls were raised individually in glass jars, in 10 % Steinberg’s solution, while the Xenopus were raised in small groups in Modified Holtfreter’s Solution (50 % Holtfreter’s with 0·1 g/1 MgSO₄, 7H₂O added). The axolotls were kept separated to prevent cannibalistic leg damage thereby avoiding the digital anomalies due to regeneration.

In this experiment, both the axolotls and the Xenopus laevis were treated using a method slightly modified after Bretscher (1949). We staged the Xenopus according to the normal table of Nieuwkoop & Faber (1956) and the axolotls based on the number of externally visible toes. Larvae were then anaesthetized in a 1: 7000 concentration of ethyl m-aminobenzoate and placed, left side up, in a small cavity carved into a paraffin-filled Petri dish. The head and tail of the animal to be treated were covered with wet paper towels to prevent desiccation but not so tightly as to cause suffocation. Under a dissecting microscope, the left hindlimb bud was gently perforated in four evenly distributed points with a sharpened tungsten needle. (This was necessary to help the rapid diffusion of the colchicine into the limb bud). A small piece of tinfoil was placed under the limb bud isolating it from the underlying body wall. Next, the limb bud was covered with a filter paper (approximately 1·5 mm²) saturated with a 1:2000 solution of colchicine (Sigma Co.). The larvae were treated for 25 min. After that period the piece of filter paper and the tinfoil were removed and the animals returned to Holtfreter’s Solution. Most of the animals (over 90 %) survived this treatment. The right foot remained unaffected and developed normally, thus being used as a control for each animal. In addition, the same experimental manipulations were repeated on control limb buds with the exception that the filter paper was soaked in Holtfreter’s Solution instead of colchicine. When the control foot had fully developed, the animal was preserved in 10 % buffered formalin. At that time, which corresponded to the metamorphic climax in Xenopus (N&F stage 60), all elements have undergone endochondral ossification. The axolotls, a neotenic salamander which does not undergo metamorphosis, were allowed to grow well beyond (about one to two months) the termination of the development in the control limb. Consequently, at the time of preservation the axolotls were undergoing endochondral ossification, with no further development occurring in the experimental limb. To further prove this point, we allowed some Xenopus to completely metamorphose and grow as froglets prior to preservation. The experimental limb showed the same affected morphology as in the animals preserved at metamorphic climax. A list of specimens treated and their developmental stage at the time of operation is given in Table 1.

Preserved animals had their soft tissue cleared with KOH after trypsin treatment and stained in to to with Alcian blue 8GX (Sigma Co.) for mucopolysaccharides (cartilage) and alizarin red S (Sigma Co.) for calcium deposits (bone) (method slightly modified after Wassersug, 1976).

To test for any degree of cell death resulting from the colchicine treatment, as well as its spatial distribution, two methods were used: 1) Vital staining with neutral red (Sigma Co.). One hour after treatment, two Xenopus were placed in a solution of neutral red (following procedure outlined by Cameron & Fallon, 1977a) for 90 min followed by rinsing with dechlorinated tap water. The two hindlimb buds of each animal were excised, observed under a microscope for differential staining, fixed in 10 % buffered formalin and mounted on microscope slides with Permount for further reference. The same procedure was repeated with another two specimens 24 h after treatment. 2) Light microscopy of histological sections. Both hindlimb buds of four experimental animals preserved 2 and 24 h after treatment were embedded in paraplast, serially sectioned (10 pm) and stained with toluidine blue O (e.g. Humason, 1979). Differences in cell morphology, mitotic activity and cell death between experimental feet and opposite control feet were studied.

To quantify differences in growth rates between the experimental and control limbs, camera-lucida outlines of both feet of six Xenopus laevis were drawn at the time of treatment. This was repeated every 5 to 7 days for the following 9 weeks. The area of each foot was computed each week by use of a digitizer. Thus we obtained individual longitudinal growth curves.

Ontogenetic series of both species were prepared by preserving groups of three axolotls at 5-day intervals and five Xenopus at 7-day intervals throughout hindlimb development. The preserved animals were then staged and cleared and stained by techniques referred to above.

The timing of differentiation of the various metatarsal and phalangeal elements during Xenopus and Ambystoma normal limb development is shown in Table 2. The data are based on specimens cleared in toto and stained with Alcian blue. Therefore, we score the presence of an element only after chondrogenetic condensation has occurred and cartilage matrix mucopolysaccharides are being secreted. In Table 2 we report the number of elements present at every stage. For example, at stage 53 in Xenopus, the metatarsal (MT) of only the fourth digit is differentiated, while at stage 54+, the metatarsal of the first digit is just starting to chondrify while digits 2 through 5 have metatarsals and digits 3 and 4 already have differentiating first phalanges. The numbers refer to phalanges from proximo to distal level. A dot is placed over the symbol to indicate that chondrogenesis is just beginning.

In axolotls (Table 2A), digits 1 and 2 differentiate almost in synchrony (Fig. 2A). They are sequentially followed by digit 3 (Fig. 2B), 4 and 5 (Fig. 2C). With the exception of the first two toes, each digit differentiates before the first phalange of the following digit starts to condense. Digit 5 is clearly the last to differentiate; all other digits have completely differentiated before the first phalange of digit 5 is formed.

The clearly asynchronous and anteroposterior sequential development of the axolotl foot contrasts with the more simultaneous proximodistal differentiation observed in Xenopus (Table 2B). At stage 52 the Xenopus limb bud is still a field of mesenchymal cells (Fig. 1). The metatarsals of digits 3 and 4 start to differentiate at stage 53; they are soon followed by the differentiation of the metatarsals of digits 2 and 5. Digit 1 is clearly the last digit to differentiate.

In addition, we encountered a pattern of proximodistal phalangeal differentiation in the axolotl which we did not observe in Xenopus. As indicated in Table 2A, the appearance of the phalangeal elements in digits 3 and 4 is not in simple sequence. In these toes, a second phalange appears in the form of a continuous cartilaginous rod that subsequently splits into two elements in digit 3, or three in digit 4 (Fig. 2D). This exception in the sequence of proximodistal differentiation has also been pointed out by Smith (1978) in regenerating urodele limbs. In addition, Maden (1981a) discusses the fact that the digits form before the wrist. This same phenomenon can be observed in Fig. 2A.

We quantitatively describe the differences in growth patterns between the untreated control limb vs. the colchicine-treated experimental limb in Xenopus.

Some representative curves to show the variation in longitudinal growth encountered are illustrated in Fig. 3. While there is a marked difference in rates of growth among individuals, all normal growth curves are characterized by a period of approximately exponential growth during the stages of digit differentiation (53-57). On the other hand, the colchicine-treated limb bud stops growing for a varying length of time. In most cases the period of arrestment of absolute growth lasts several weeks at 18 °C. (More detailed measurements on these aspects of the dynamics of growth and how they relate to differentiation are presently being performed in our laboratory.) After this somewhat surprisingly long period, the experimental feet recover and begin to grow at gradually increasing rates. No catch-up growth (Williams, 1981) has been observed. Consequently, as is shown in the figures, the treated foot at the time of termination of the developmental process is substantially smaller than the control.

Colchicine is a well-known mitotic inhibitor affecting microtubule assembly (Hooper, 1961; Borisy & Taylor, 1967). Preliminary histological results have shown that after colchicine treatment no mitotic activity is observed in the treated limb bud. Cells, as early as 2h after the treatment, exhibit a rounded morphology and are closely packed together (Fig. 4). Their morphology contrasts with the opposite untreated limb bud (Fig. 1). The untreated limb-bud cells show a typical mesenchymal morphology with abundant extracellular matrix. No localized cell death was observed in the experimental limb by vital staining with neutral red. The mesoderm stained and destained uniformly. Histological sections, however, showed some leucocytes, pycnotic dark-staining inclusions, and extracellular debris usually associated with cell death (Fig. 4). However, in agreement with the vital staining, we did not observe any welldefined regions of necrosis. Therefore, we tentatively conclude that, although cell death occurred, it was not localized in any particular area of the developing limb bud. Detailed measurements of the effect of colchicine on cell dynamics are in progress.

A summary of the morphologies obtained by treating the limb bud with colchicine is presented in Table 1 and Fig. 5. Table 1 is a list of all specimens treated, while in Fig. 5 we graphically summarize these data. Every box in Fig. 5 represents a digital element. For Xenopus, the normal phalangeal formula is: digit 1 (metatarsal, MT, plus 2 phalanges); digit 2 (MT + 2 phalanges); digit 3 (MT + 3 phalanges); digit 4 (MT -I-4 phalanges); digit 5 (MT + 3 phalanges) or 2,2,3, 4, 3. The complete formula for the axolotl is 2,2, 3,4,2. The percentage of treated individuals that have lost a given element is shown within the corresponding box. For example, 68 % of the experimental Xenopus (21 specimens) lost the fourth phalange of digit 4. One axolotl and two Xenopus were unaffected, i.e., they exhibited a normal phalangeal composition in the experimental foot, although a substantial overall size reduction was present as compared with the control foot.

Of the 38 experimental Xenopus, there were seven feet from which phalangeal formulas could not be reliably taken. Three of the unusable feet had fused elements. The homologies of the digits of three feet, two with only two digits (Nos. X-10, X-ll) and one with three (No.X-9), could not be reliably ascertained and were excluded from our analysis. Another specimen lost all the distal elements beyond the tibia and fibula in the treated foot (X-12). The homologies of the toes in the remaining Xenopus with missing digits were elucidated on the basis of presence of claws and positioning with respect to tarsal elements. A particularly unusual specimen (X-7), used in the experimental results, had lost the tibia and fibula while retaining all the more distal elements with the exception of two terminal phalanges.

Fig. 5 shows two types of foot reduction in both Xenopus and Ambystoma ‘. loss of a whole digit and loss of individual elements. In Xenopus, digit 1 is lost much more frequently than any other digit. In contrast, axolotls lost only digit 5. Both Xenopus and axolotls showed digital loss in about one third of the cases. However, loss of a whole toe is not necessarily accompanied by extreme proximodistal reduction in the remaining elements. This phenomenon is illustrated in Figs 6 and 7.

Loss of individual elements also shows a different pattern in Xenopus than in Ambystoma. In Xenopus the terminal elements of digits 1, 4 and 5 are most frequently lost, 90 %, 68 % and 65 % of the cases respectively. The first phalange (58 %) of digit 1, the third phalange of digit 3 (35 %), the second phalange of digit 2 (35 %), and the second phalange of digit 5 (26 %) are the following three most frequently lost elements. In axolotls, the terminal phalange of digit 4 is by far the most frequently lost element (93 % of the cases). The terminal phalange of digit 1 (57 %) has the second highest percentage. Besides those two elements, the axolotl foot is only repeatedly affected at three other phalanges (excluding the loss of the whole of digit 5 (36 % of the cases)): the second phalange of digit 2 (36%), the third phalange of digit 3 (43 %) and the third phalange of digit 4 (43%).

There has been a recent surge of interest in the connections between the processes of growth and pattern formation during development (e.g. see Maden, 1981a; Summerbell, 1981a; Cooke & Summerbell, 1981; Cooke, 1982) such as the dependence of mechanisms of pattern specification on the scale and dimensions of the embryonic field. We relate the results reported in this paper to this general issue. We propose that the effect of colchicine is due to its action in reducing the number of mesenchymal cells in the developing limb bud. This is simply a working hypothesis consistent with our experimental results. We are presently testing it through detailed analysis at the cellular level, using standard histological techniques and autoradiographic labelling. We have shown that colchicine causes the treated limb to be much smaller than the control limb (Fig. 3). The observed differences in overall size are concomitant with a reduction in number of mesenchymal (and prechondrogenic) cells in the experimental limb bud. This initial reduction in cell number is subsequently amplified during development due to the properties of the exponential growth curves that characterize embryonic cell proliferation (e.g. Bertalanffy, 1960; Katz, 1980). These changes affect subsequent processes of morphogenesis and pattern specification resulting in varying degrees of loss of skeletal elements.

We believe that temporary arrest of mitotic division is the main role of colchicine, rather than colchicine having a more specific role, for example, through affecting structural gene expression. Our contention is supported by reports that other mitotic inhibitors that affect cell proliferation through very different physiological pathways (specifically chloroethylamine (Tschumi, 1953), vinblastine (Kieny, 1975) or cytosinearabinoside (ARA-C) (Raynaud, 1981) have a very similar effect, that is, an ordered sequence of digital reduction. In addition, Schmalhausen (1925) showed that even malnutrition or abnormally high temperatures at critical stages during limb morphogenesis retarded the development of the postaxial portions of the limbs of the axolotl. Presently, we are repeating the same experiments that we report here but using ARA-C instead of colchicine. If our hypothesis is correct, we expect to obtain the same results.

Colchicine does not terminally stop differentiation events at the time of administration. Differentiation does occur after the treatment. For example, limb buds treated at stage 53 (see results in Table 1) do not result in limbs with only three metatarsals (the number of elements present at stage 53; from Table 2). Our data also show that our experimental perturbations affect two processes of pattern formation independently. These processes are: one, the determination of the anterioposterior elements (differentiation of digits) and two, the proximodistal differentiation of phalanges. This dissociation between the determination of the anteroposterior vs. the proximodistal axis is a fact that agrees with current models of limb pattern formation (e.g. Hinchliffe & Johnson, 1980; Stocum & Fallon, 1982; Tickle & Wolpert, 1981). We discuss the implications of our results separately for each of these axes.

As previously pointed out by many authors on comparative (e.g. Holmgren, 1933; Saint-Aubain, 1981; Hinchliffe & Johnson, 1980) or experimental grounds (Maden, 1981b), there are significant differences between frogs and salamanders in the sequence of appearance of the toes. In salamanders, at least based on the axolotl data (Table 1A), there is a welldefined anteriorposterior sequence of differentiation, as follows: Digits (1-2) → 3→ 4→ 5. Conversely, in anurans, the sequence seems to occur from the centre towards the periphery, (3-4) → (2-5) → 1. In both cases, experimental reduction of the size of the limb bud via colchicine results in the loss of the element that appears last in normal ontogeny. 36 % of the axolotls treated lost the 5 th toe (the only digit completely lost in our sample), and 32 % of the Xenopus lost their first toe. It is not obvious how to reconcile these qualitatively different ontogenetic patterns and results with models proposing a single organizing centre, like the zone of polarized activity (ZPA), described in chick morphogenesis (see Summerbell & Honig (1982) for a recent review). Cameron & Fallon (1977b) report the presence of a posteriorly located region with ZPA-like effects in Xenopus and Slack (1977) argues for the presence of an anteroposterior polarizing region in urodele limbs.

Regardless of the mechanistic model, our data clearly show: one, the last digit specified in ontogeny is the first to be experimentally affected; and, two, there is no correlation between loss of a whole toe and loss of terminal phalanges. This second point is illustrated by the fact that specimens with four toes do not necessarily lose many phalanges on the other digits (Table 1). If digital loss occurred by global truncation of ontogeny, one would expect the following formulas (from Table 2): Axolotl -2-2-3-1-X; Xenopus - X-MT-l-l-MT (X = lost). Examination of Table 1 clearly shows this not to be the case. (Xenopus that have lost the first toe have more phalanges in the remaining digits (Table 3).) In fact, six out of eight experimental Xenopus have lost less than two additional terminal phalanges and two specimens have a full complement in the remaining four toes (Table 3). The same lack of correspondence is found in axolotls (Table 4), although in this case the difference is in the opposite direction, i.e., instead of having more elements, they have lost more elements in the first and second digits than would be expected. For example, Fig. 7 shows an example of a ‘4-toed’ Ambystoma that has lost the terminal phalanges in the first and fourth toes. This independence between whole digital loss vs. distal phalange loss is also evident in the phylogenetic data (Alberch & Gale, 1983 and in preparation).

Two possible mechanistic interpretations can be given to the results on the proximodistal pattern of differentiation. One, the number of mesenchymal cells in the limb bud is reduced and chondrogenesis is dependent on a critical minimum number of mesenchymal cells (Newman, 1977). Therefore, as development proceeds in limbs with reduced number of cells, the system runs out of available cells prior to the differentiation of the terminal elements. The second would be along the lines of the ‘progress zone’ model (Summerbell, Lewis & Wolpert, 1973) based on positional information. This model argues that the fate of cohorts of cells is sequentially determined at the apical tip of the limb where cells are actively proliferating. Cells acquire their fate (‘positional value’) as a function of time spent in the apical region of the limb bud. After several cell divisions, cells are ‘pushed’ back relative to the apical tip and then differentiate into a specified element. Therefore, there is a direct relationship between number of divisions of mesenchymal cells in the apical growth zone of the limb bud and the number of skeletal elements laid down proximodistally (Lewis, 1975, and related X-irradiation experiments (Wolpert, Tickle & Samford, 1979; Summerbell, 1980)). There are subtle, but conceptually important, differences between these two hypotheses. In the first, the number of skeletal elements is dependent on a threshold value in absolute number of mesenchymal cells, while in the secopd the dependence is on number of cell divisions.

Summerbell (1981a) has thoroughly reviewed the relationships between growth, size and pattern in the developing chick limb bud. However, as shown by Maden & Goodwin (1980) and Maden (1981b), it is dangerous to extrapolate from avian to amphibian development, since the systems appear to behave quite differently regarding their regulative properties. Our results are in agreement with most models of limb morphogenesis, where the pattern is specified by either number of cell divisions (Summerbell et al. 1973), a monotonic gradient (e.g. Tickle, Summerbell & Wolpert, 1975; Summerbell, 1981b; Cooke & Summerbell, 1981) or a more complicated diffusion-reaction model that generates a prepattern of regions of various levels of morphogen concentration (e.g. Wilby & Ede, 1975; Newman & Frisch, 1979; Meinhardt, 1982). Without intention to enter into a discussion of the pros and cons of each of the models proposed by these authors, it suffices to say that they are all to some degree boundary dependent, i.e., a change in the size of the domain of interaction can result in a qualitative change in global pattern. In addition, there are comparative studies on genetic mutants that illustrate that changes in number of digits are correlated with changes in the dimensions of the early limb bud, for example, the oligosyndactyly (Os) mutation in mice (Griineberg, 1963). Similarly, work with teratogens (e.g. the work by Messerle & Webster (1982) on cadmium-treated limbs) shows that limb defects associated with digital loss are correlated with limb primordia of reduced dimensions. In addition, Sewertzoff (1931), Raynaud, Gasc & Renous-Lacuru (1974) Raynaud et al. (1975) and Raynaud (1976) have shown that in species of lizards characterized by digital loss the limb bud is comparatively smaller, a fact that Raynaud (1976) has argued is due to a reduction in the migration of mesodermal cells into the limb bud primordia.

Given this perspective, which stresses the importance of scale in limb morphogenesis, it is not surprising that one of the best documented roles of the two major organizing centres in limb morphogenesis, the AER and the ZPA, is to regulate mitotic activity (see Summerbell, 1981/?, for review). Furthermore, if our contention is correct, one would expect a fair degree of invariance in the absolute size of the limb bud at the time of pattern specification among species sharing the same limb skeletal pattern regardless of the final size of the organism. This hypothesis is not in agreement with Maden (1981a), who conclusively showed that urodele limb regeneration is size independent. One would have to argue that there are differences in this respect between limb regeneration and normal development. This would not be surprising, since the problem of size dependence vs. independence is probably a quantitative rather than qualitative issue. Some developmental systems are more stable to perturbations in boundary conditions than others. As Maden (1981Z?) has shown, the concepts of mosaic vs. regulative development are part of a continuum which varies ontogenetically and phylogenetically.

In conclusion, we have reconfirmed the qualitative differences between the ontogenies of anurans and urodeles by direct examination and by studying their response to the same experimental stimulus. Our results also suggest that pattern specification in the amphibian limb is dependent on either number of undifferentiated cells or absolute dimensions of the limb bud, and that pattern specification mechanisms along the anterior-posterior and proximodistal axes are independent to some degree.

We would like to thank the Axolotl Colony, University of Indiana, for sending the Ambystoma mexicanum embryos. Anna Haynes and M. Maden critically read the manuscript and offered useful comments. We thank Laszlo Meszoly for the illustrations and A. Coleman for the photographic work. Catherine McGeary provided assistance in the final preparation of the manuscript. The work was supported by NSF Grant DEB-81-20917.