Inhibition of Sonic hedgehog signaling leads to posterior digit loss in Ambystoma mexicanum: Parallels to natural digit reduction in urodeles


  • Geffrey F. Stopper,

    Corresponding author
    1. Department of Ecology and Evolutionary Biology, Yale University, New Haven, Connecticut
    • Department of Ecology and Evolutionary Biology, Yale University, 165 Prospect Street, New Haven, CT 06520
    Search for more papers by this author
  • Günter P. Wagner

    1. Department of Ecology and Evolutionary Biology, Yale University, New Haven, Connecticut
    2. Peabody Museum of Natural History Faculty Affiliate, Yale University, New Haven, Connecticut
    Search for more papers by this author


Molecular mechanisms patterning the tetrapod limb, including anterior-posterior axis determination involving Sonic hedgehog (Shh), have received much attention, particularly in amniotes. Anterior-posterior patterning in urodele amphibians differs radically from that of amniotes in that it shows a pronounced anterior-to-posterior sequence of digit development. In contrast, amniotes develop their digits almost simultaneously with a slight posterior-to-anterior polarity. Here we use cyclopamine, an inhibitor of the Hedgehog signaling pathway, to investigate the role of Shh in anterior-posterior patterning in the urodele limb. Inhibition of Shh signal transduction affects digit number long before their morphological appearance. In accordance with the apparently derived order of digit development in urodeles, exposure reproducibly removes digits in a posterior-to-anterior sequence, the inverse of their developmental sequence. This pattern of digit loss mimics the order of digit loss in natural variation. We suggest that variation in Shh expression and/or signal transmission may explain natural variation in digit number in urodeles. Developmental Dynamics 236:321–331, 2007. © 2006 Wiley-Liss, Inc.


The tetrapod limb is an excellent and well-studied model for investigating the development and evolution of morphological characters. We have come to understand many details of limb development mechanisms—the roles of individual genes and gene products in morphological patterning and how they interact with other genes and gene products. Most of these data are from the amniote model species mouse and chicken. It is unclear how general these mechanisms are among the tetrapods, and understanding the extent to which they are general or variable is of great importance (Stopper and Wagner,2005). Comparatively little is known about mechanisms of limb development in the other main branch of tetrapods, the amphibians, though we do know of many instances of variation in developmental mechanisms between the amniotes and amphibians, and within each of these groups (Stopper and Wagner,2005). Despite all that we know about the genetics of limb development in model organisms, and all that we know about variation in molecular mechanisms of limb development, cases for mechanistic changes that are causal in the evolution of morphology are rare. Further elucidation of developmental mechanisms in limbs with differing morphologies should help to shed light on how changes in mechanisms of development cause evolutionary changes in morphology.

One major difference in limb development that exists among tetrapods is that of the order of digit development. Amniotes and anurans develop their digits in relative synchrony; digit cartilage condensation and bone chondrification usually start with the penultimate posterior digit, which is quickly followed by development of the most posterior digit and the more anterior digits developing in a posterior-to-anterior sequence (Shubin and Alberch,1986). Urodeles, however, develop their most anterior two digits first, followed by the posterior digits in an anterior-to-posterior sequence (Shubin and Alberch,1986; Wake and Shubin,1998). Considering the current understanding of tetrapod relationships and the phylogenetic distribution of these modes of development, it is likely that the shared amniote and anuran pattern is ancestral, and the urodele pattern is derived (Shubin,1995; Vorobyeva and Hinchliffe,1996; Wagner et al.,1999; Stopper and Wagner,2005). There are several hypotheses that have been posed to explain the novel order of digit development in urodeles. One hypothesis states that this difference can be explained by a polyphyletic origin of lissamphibians (Holmgren,1933; Jarvik,1980; Hanken,1986). In this case, anurans are thought to be more closely related to amniotes than they are to urodeles (Fig. 1A). This hypothesis, however, is refuted by the vast majority of morphological and molecular studies; most evidence supports monophyly of lissamphibians (Fig. 1B), with anurans and urodeles being equally closely related to amniotes, and more closely related to each other than either is to amniotes (Milner,1988; Hedges et al.,1990; Trueb and Cloutier,1991; Marshall and Schultze,1992; Cannatella and Hillis,1993; Eernisse and Kluge,1993; Larson and Dimmick,1993; Ahlberg and Milner,1994; Duellman and Trueb,1994; Zardoya and Meyer,2001). Another hypothesis suggests that urodeles evolved the heterochronic shift to earlier anterior digit development as an adaptation for aquatic larvae to interact with substrate (Wake and Marks,1993). A third hypothesis posits that an ancestor of all extant urodeles had limbs reduced to two digits—those homologous to digits 3 and 4 of ancestral tetrapods—and then subsequently evolved novel, later-developing posterior digits that have no homologues in amniotes or anurans (Wagner et al.,1999). Here we investigate one major aspect of anterior-posterior patterning, the Hedgehog signaling pathway, and ask what role Shh plays in the derived mode of digit development in urodeles.

Figure 1.

Phylogenetic hypotheses for limbed tetrapods. A: Alternate hypothesis supporting lissamphibian paraphyly. B: Currently predominant hypothesis supporting lissamphibian monophyly. The completely limbless amphibian order Gymnophiona is excluded, as their phylogenetic placement does not affect limb evolution interpretations made here.

The patterning of tetrapod limbs is largely directed by signaling across the three major axes: proximal-distal, dorsal-ventral, and anterior-posterior. Signaling along the anterior-posterior axis is mediated by the zone of polarizing activity (ZPA). The ZPA lies at the posterior margin of the limb bud, near the distal end. The Sonic Hedgehog (Shh) protein is expressed in, and secreted from, the ZPA. While the exact nature of the ZPA's activity is not fully understood, one model poses that Shh diffusion from this region forms a posterior-to-anterior gradient in its concentration, with the highest levels at the posterior (Saunders and Gasseling,1968; Riddle et al.,1993). Expansion along the anterior-posterior axis is directed by Shh and the anterior-posterior identity of any region of the limb is thought to be specified by the amount of Shh received by that region. Multiple lines of evidence suggest that the shh expression domain and the general anterior-posterior patterning function of Shh in limbs has been conserved across tetrapods (Riddle et al.,1993; Chang et al.,1994; Endo et al.,1997; Imokawa and Yoshizato,1997; Torok et al.,1999; Hanken et al.,2001). In urodeles, including Ambystoma mexicanum, although the limb shh domain appears to be slightly smaller and more proximal than in non-urodele tetrapods, shh expression is very similar in timing and location to that of non-urodele tetrapods (Imokawa and Yoshizato,1997; Torok et al.,1999).

To better understand the functions of the members of the Hedgehog family during limb development in urodele amphibians, and to analyze the fine-scale temporal dynamics of the patterning properties of Hedgehog gene products, we use cyclopamine to inhibit Hedgehog signaling during five different time windows of limb development in axolotl (A. mexicanum). It was previously shown that, when administered during limb regeneration in A. mexicanum, cyclopamine causes varying reductions in digit number based on its concentration (Roy and Gardiner,2002). In our experiments, cyclopamine exposure generally causes reductions in the anterior-posterior extent of limb development, including digit loss, and failures of limb element separation. Cyclopamine exposure causes digit reduction in the exact opposite order of their development (i.e., the posterior digits are affected first). This pattern mimics the pattern of natural variation in urodele digit reduction.


Animal Husbandry and Cyclopamine Exposure

A. mexicanum eggs were purchased from the Indiana Axolotl Colony. Upon arrival, each individual was isolated in 15 ml 20% Holtfretter's solution. Housing containers were plastic 2-ounce Solo brand soufflé cups (Highland Park, IL). Temperature of the housing room remained at 20° ± 1.3°C. Each cup was covered with a matching Solo brand clear plastic lid that was perforated with two small holes (approximately 2-mm diameter) in order to reduce evaporation while allowing air exchange. Eggs were force-hatched by removing egg casing at approximately stages 35–40 (Bordzilovskaya et al.,1989). Holtfretter's solution and housing containers were completely replaced daily. After hatching, animals were fed approximately 20–50 brine shrimp each, once per day, between one and three hours before water and container change. Lids were removed from containers between feeding and water change. Salt water was removed from brine shrimp by filtering through a fine plastic mesh screen, then scooping shrimp into the axolotl's media with a wooden toothpick.

Cyclopamine powder (11-Deoxyjervine; Toronto Research Chemicals Inc.) was dissolved in 100% ethanol to a concentration of 5 mg/ml and this stock solution was stored at −20°C. A similar volume of 100% ethanol was stored at −20°C. During periods of cyclopamine exposure, after Holtfretter's solution changes, 3 μl of the stock solution of cyclopamine was added to each experimental animal's Holtfretter's solution. For corresponding control animals, 3 μl of the 100% ethanol stock was added instead. Each experimental or control animal underwent a period of 10 days of exposure to cyclopamine or ethanol, respectively. Pilot experiments indicated that exposure to cyclopamine for the entirety of limb development had a marked negative effect on survival and developmental rate. In the current experiment, exposures to cyclopamine were restricted to time windows of 10 days each. Exposures were started at one of five different developmental stages, which approximately correspond to each of stages 45–49 (Nye et al.,2003). More specifically exposures were started, in order of youngest to oldest, when (1) forelimb bud length was equal to forelimb bud width, (2) forelimb bud length was twice forelimb bud width, (3) anterior-posterior differentiation began to be accentuated as the anterior distal tip of the limb became slightly extended more distally than the posterior end, and a slight bulge appeared along the posterior margin of the limb, but prior to the formation of a cleft between digits 1 and 2, (4) a recessed cleft appeared in the interdigital membrane between digits 1 and 2, and (5) an interdigital cleft appeared between digits 2 and 3. A total of 20 experimental and 20 control animals were used with 4 experimental and 4 control animals in each stage-matched group. Additionally, for each of the five stages, a single individual was exposed to a doubled concentration of cyclopamine by adding 6 μl of the cyclopamine stock solution instead of 3 μl. There were no control individuals for these five animals. Aside from the lack of controls and the doubled dose of cyclopamine, procedures for these five individuals were exactly the same as those for all other individuals.

Each animal, upon emergence of the second digit of the hindlimb, was over-anesthetized with MS222, fixed in 4% paraformaldehyde at 4°C overnight, then rinsed twice for 10 min each in Phosphate Buffered Saline, rinsed twice for 10 min each in 100% methanol, and transferred to a final change of 100% methanol and stored at −20°C. Specimens were then cleared and stained by a modification of the method of Hanken and Wassersug (1981) by trypsin digestion, treatment with alcian blue, and clearing in 0.5% KOH.

In many of the experimentally produced limb phenotypes, two terminal phalanges could represent either the presence of two individual digits or the bifurcation of a single digit. Considering the limb phenotypes observed, in many cases it would be very difficult to reliably make this distinction. For simplicity and objectivity of classification in this analysis, every terminal phalange will be counted as a “digit.” Four experimental individuals were fixed earlier than the rest, one due to premature death, and three to document the effects of cyclopamine exposure on ontogeny. For these individuals, digit number was counted as the number of digits or obvious digit condensations present at the time of fixation, which, in our judgment, is the same number that would have ultimately developed in each of these eight limbs.

mRNA and cDNA Preparation and Semi-Quantitative RT-PCR

Ten animals were prepared for semi-quantitative RT-PCR by rearing as above, but were not exposed to ethanol or cyclopamine. Individuals were sacrificed as above, stored overnight at 4°C in RNA later (Ambion), then moved to −20°C. Forelimbs were dissected off of each individual, and mRNA was isolated using a NucleoSpin RNA II purfication kit (BD Biosciences). From each limb mRNA preparation, cDNA was written with an Advantage RT-for-PCR Kit (BD Biosciences) using an oligo(dT) primer.

Polymerase Chain Reactions (PCRs) were carried out for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and shh using sequence-specific primers (GAPDH forward: 5′-CTGGCGTGTTCACCACAATC-3′, GAPDH reverse: 5′-CAGTGGAGGCTGGAATGATG-3′, shh forward: 5′-GACCAAGAGAACACCGGAGC-3′, shh reverse: 5′-CAATGAATGTGGGCCTTGG-3′). PCR reactions were a total of 50 μl each: 42 μl Reddy Mix Master Mix (ABgene), 4 μl cDNA, and 2 μl each 10-μM primer (to a final primer dilution of 0.4 μM each primer). PCR reactions were run on a Perkins Elmer Thermal Cycler, with 3 min initial denaturation at 92°C, followed by cycling (25–30 cycles for GAPDH, 30–50 times for shh) 30 sec at 92°C, 30 sec at 61°C, and 1 min at 72°C, with a final extension after cycling of 3 min at 72°C. Relative levels of product of each gene were measured by running 10 μl each PCR product on a 1.5% agarose gel at 100 V, and visualizing intensities by Ethidium Bromide/UV-illumination of the Reddy Mix Dye in a BioRad transillumination system with Quantity One software version 4.2.1. Sensitivity of this protocol for detecting real differences in starting shh cDNA concentrations was confirmed by running PCRs on serial dilutions made from a single cDNA stock solution.

Museum Specimens

Cleared and stained museum specimens of urodeles were obtained for the analysis of patterns of digit number reduction (see Fig. 5). Pseudobranchus striatus specimen MVZ198685 was redrawn with permission from an unpublished drawing by David Wake. Siren lacertina specimen YPM103 was obtained from the Yale Peabody Museum of Natural History, and Amphiuma pholeter specimen UF28812 was obtained from the Florida Museum of Natural History.

Figure 5.

Evolved forelimb reductions in digit numbers in urodeles. Anterior is left. First limb on the left (A) is Siren lacertina drawn from specimen YPM103. Second from left (B) is the dwarf siren Pseudobranchus striatus redrawn with permission from a drawing of specimen MVZ198685 courtesy of David Wake. Third from the left (C) is Amphiuma tridactylum and fourth from the left (D) is Amphiuma means. Both amphiumids have been redrawn from Stoudemayer (1949) with permission from the publisher. Fifth from the left (E) is Amphiuma pholeter, drawn from the left forelimb of specimen UF28812. This museum specimen was cleared and stained for bone, but cartilage was not stained. Calcified bone is clearly visible in this specimen (solid lines). Outlines of continuous cell condensations can be discerned that connect these calcified elements (dashed lines), though separations of elements within these condensations are not discernable, and may not exist. Because of the apparent lack of separation of these elements, we have included the humerus in this drawing (E), while all other drawings exclude the humerus (A–D).


Cyclopamine Exposures

We exposed developing larvae of the species A. mexicanum to cyclopamine to block signaling through the hedgehog pathway. Exposures were conducted during five time windows of limb development, with exposures starting at times corresponding approximately to each of stages 45–49 as defined by Nye et al. (2003), and lasting 10 days each (Fig. 2). Because A. mexicanum larval development rate is relatively invariable among individuals, the full range of developmental stages during which hedgehog signaling was knocked down is very consistent among individuals that share a common stage of exposure initiation (i.e., a group of individuals that were all at the same stage at the start of exposure were at the same older stage at the end of exposure). Four individuals were exposed to a concentration of 1 μg/ml cyclopamine at each of the five stages. Additionally, at each of the five exposure stages, one individual was exposed to a doubled concentration of 2 μg/ml cyclopamine, totaling 25 exposed individuals. For each experimental individual exposed to cyclopamine, a control individual at the same stage of development was exposed to a volume of ethanol equal to the volume that was used in the experimental treatment to dissolve the cyclopamine. Hindlimbs develop long after forelimbs, completely outside of the windows of cyclopamine exposure, and were morphologically unaffected by cyclopamine exposures. With the exception of one animal that died during the experiment (one experimental animal) and a few that were fixed immediately after exposure (three experimental and three control animals), all animals were fixed at or about the stage when the hindlimbs had started to develop two distinct digits (Stage 54–55; Nye et al.,2003).

Figure 2.

Timing of cyclopamine exposure. Days are represented on the x-axis and exposure group on the y-axis. Shaded boxes represent days during which animals of that group were exposed to cyclopamine in experimental groups or ethanol in controls. Exposures were started at stages that correspond approximately to Nye et al. (2003) stages 45 (early), 46 (early middle), 47 (middle), 48 (late middle), and 49 (late).

Nearly all ethanol-exposed control limbs (39 of 40) were morphologically indistinguishable from normal axolotl limbs. The other control limb (1 of 40) had a normal stylopod and autopod, but its zeugopod had an unusual bending and fusion of elements. We take this as an anomalous result, as it is not similar to any phenotype produced in the experimental individuals, and so is produced in only 1 of 90 limbs that were exposed to ethanol or ethanol with cyclopamine. It is, therefore, apparent that ethanol at the levels to which these individuals were exposed does not affect limb development.

The morphologies resulting from the doubled, 2 μg/ml, dose of cyclopamine fell at the more severely affected end of the range of morphologies exhibited by those exposed at 1 μg/ml for the same stage. None of these five individuals exposed to the 2 μg/ml, however, resulted in morphologies that are different from those resulting from exposure to 1 μg/ml at the corresponding stage. For this reason, unless otherwise stated, the results for both concentrations of cyclopamine exposure are grouped together.

The limb phenotypes resulting from cyclopamine exposure exhibited small ranges of variation in phenotype within exposure groups and some overlap with phenotypes from neighboring stage intervals of exposure (Fig. 3). In most cases, even when no obvious effect was observed on the anterior-posterior patterning of the limb, some limb elements appear to be fused to other elements. These fusions, or failures of separation, make identification of individual limb elements difficult in some of the experimental phenotypes.

Figure 3.

Limb morphologies during exposure, phenotypes produced by cyclopamine exposure, and counts of digits on experimental limbs. Left column (AE) contains photographs of representative limbs at the start of cyclopamine exposure. Next column (FJ) shows photographs of control limbs at the end of exposure to represent the developmental time period of normal development during which cyclopamine was administered. Limb outlines are traced in white to make morphologies more obvious (A–J). Photographs in the third column (KO) represent the most reduced phenotype produced in each exposure group, and those in the fourth column (PT) represent the least reduced phenotype produced in each exposure group. Photographs in the third and fourth columns (K–T) are of animals after clearing and staining with Alcian blue. The far right column of graphs shows the total count of limbs from each exposure group that ultimately developed each sum of digits, with digit number on the x axis and number of limbs on the y axis. All scale bars = 1 mm. A-P, anterior-posterior.

The earliest time window of cyclopamine exposure started shortly after limb budding when the limb bud was approximately as long as it was wide (Fig. 3A). Exposure at this stage produced limbs with one digit (7 of 10; Fig. 3K) or two digits (3 of 10; Fig. 3P). Fusions occurred variably in all elements distal to the proximal zeugopod, including one case of a single zeugopodial long bone that is clearly broader than either the wild type radius or ulna alone (Fig. 3P).

The next time window of cyclopamine exposure started when limb bud length was approximately twice its width (Fig. 3B). Resulting phenotypes were similar to those of the earliest exposure, ranging from one digit (4 of 10; Fig. 3L) to two digits (6 of 10; Fig. 3Q), producing more limbs with two digits than the earliest exposure window. One limb bears two clearly distinct digits that articulate proximally with a single nodular carpal element (Fig. 3Q). Several limbs developed zeugopodial elements that appear as a broad single element or radius and ulna that did not completely separate (Fig. 3Q).

The middle time window of exposure started as the anterior distal end of the limb began to extend slightly more distally than the posterior distal end and a slight bulge appeared along the posterior margin of the limb, but prior to the formation of a cleft between digits 1 and 2 (Fig. 3C). Resulting limbs bore two digits (6 of 10; Fig. 3M) or three digits (4 of 10; Fig. 3R). In most cases where a third digit developed (3 of 4), the most posterior digit was very small (equal to or smaller in size than the metacarpal of its anterior neighbor) and did not exhibit the close articulation with carpal elements that was exhibited by all other distinct digits in the experiment (Fig. 3R). Several of these limbs also exhibited a single broad zeugopodial element (Fig. 3M) or incompletely separated radius and ulna (Fig. 3R).

The fourth time window of exposure started just as a cleft formed distally between the first two digits (Fig. 3D). Limbs exposed to cyclopamine during this time window produced three digits (4 of 10; Fig. 3N) or four digits (6 of 10; Fig. 3S). In one three-digited limb, the carpal elements, though somewhat fused, are distinct enough that all of the normal carpal elements are identifiable (Fig. 4B). Normally, the metacarpals of digits 1 and 2 articulate with the basale commune and digits 3 and 4 articulate with distal carpals 3 and 4, respectively. In the case of this three-digited limb, the metacarpals of digits 1 and 2 articulate with the basale commune, and the metacarpal of digit 3 articulates with distal carpal 3, as normal. There is, however, no fourth digit to articulate with distal carpal 4 (Fig. 4B). Therefore, topologically it is digit 4 that is missing.

Figure 4.

Detailed view of carpals in series of cyclopamine-induced digit number reductions. Limbs are cleared and stained with Alcian blue. Limbs are from late exposure (A), late middle exposure (B), early middle exposure (C), and early exposure (D). Individual carpal elements are indicated by arrowheads. All scale bars = 1 mm.

The latest time window of exposure to cyclopamine started at the first sign of the budding of digit 3 from the posterior of the limb (Fig. 3E). All limbs treated at this stage clearly develop four distinct digits (10 of 10; Fig. 3O,T). Despite developing a normal number of digits, these limbs still exhibited fusions between carpals and fusions between phalanges or phalanges and metacarpals, though in no case were digit elements fused to carpal elements. In five of the eight limbs that were allowed to develop fully, the intermedium was fused to the radius, but not to the ulna, while these limbs exhibit no examples where the intermedium fused to the ulna and not the radius.

sonic hedgehog Semi-Quantitative RT-PCR

Semi-quantitative RT-PCRs were performed during multiple stages of normal A. mexicanum limb development to quantify the expression profile during normal limb development. Whole-limb mRNA was obtained and cDNA constructed by reverse transcription. These cDNA pools were diluted to produce levels of the PCR product of the ubiquitously expressed standard (glyceraldehyde-3-phosphate dehydrogenase, GAPDH) that were equal among samples or slightly declining in concentration with later developmental stages. Intensities of shh RT-PCR results were slightly inconsistent from one reaction to another, but averaging over several reactions shows the highest concentration at stage 46 (Nye et al.,2003), the earliest investigated stage, and a steady decline of shh PCR product with progressively later stages of limb development, with no detectable product at these standardized dilutions after stage 50.

In order to test whether cyclopamine effects on phalange fusion could be due to Shh function during digit development, one limb bud at stage 52 (Nye et al.,2003) was dissected into two pieces, a distal piece that included the entirety of digits 1 and 2 with part of the distal carpus connecting them, and a proximal piece consisting of the rest of the limb from the proximal end of the humerus to the carpus, including the buds of digits 3 and 4. Using our RT-PCR protocol with higher concentrations of cDNA than used for our investigation of the developmental profile of shh expression, we were able to detect shh expression in the proximal portion of this limb, but were unable to detect it in the distal portion containing digits 1 and 2. The PCR product from the proximal portion was sequenced to confirm its identity as shh.


Lack of Skeletal Element Separation

Besides sonic hedgehog, two other hedgehog paralogues are known to have existed in Osteichthyes at the origin of the crown group (Zardoya et al.,1996b). In tetrapods, these are called desert hedgehog and indian hedgehog (in amniotes; ihh) or banded hedgehog (in amphibians; bhh). Ihh, like Shh, is known to function during limb development (St-Jacques et al.,1999). Ihh in mammals appears to prevent chondrocytes from entering their hypertrophic phase, maintaining them in a phase of proliferation. Chondrocytes in mouse ihh null mutants experience early hypertrophy, resulting in limbs with shortened long bones and a failure of many skeletal elements to separate from one another (St-Jacques et al.,1999). These mutants also show that Ihh is necessary for proper osteoblast development in endochondral bone. Human brachydactyly type A1, shown to be caused by mutations in ihh, is characterized by the shortening of digits by fusion or loss of phalanges (Gao et al.,2001). Similar expression profiles of ihh in chick suggest that its function is probably conserved between chick and mouse (Vortkamp et al.,1996). The orthology of ihh and bhh is well-established (Zardoya et al.,1996a, b; Stark et al.,1998), and bhh expression patterns in the anuran Xenopus suggest that the patterns of ihh/bhh limb expression, and probably function, are an ancestral property of tetrapods (Moriishi et al.,2005).

All of the hedgehog paralogues signal through the same pathway (Carpenter et al.,1998). This pathway includes two transmembrane proteins, patched (Ptc) and Smoothened (Smo). Smo activates downstream transcription, but its activation is suppressed by Ptc. When Ptc is directly bound by a Hedgehog protein, however, the repression of Smo by Ptc is removed, thereby activating transcription of the downstream targets of the pathway. Cyclopamine is thought to act through direct binding to the Smoothened protein (Chen et al.,2002). Thus, exposure to cyclopamine should knock down both of the Hedgehog signals, Shh and Ihh/Bhh. In our experiments, therefore, we are unable to unequivocally separate the phenotypic effects of blocking Shh from the effects of blocking Ihh/Bhh. It is notable, however, that our cyclopamine-induced phenotypes conform to what would be expected from blocking both Shh and Ihh/Bhh if the ancestral tetrapod functions of Shh and Ihh/Bhh are maintained in A. mexicanum. Therefore, we suggest that both failure of skeletal element separation and proximal-distal reductions in skeletal element size are due to blocking of Ihh/Bhh functions, and that reductions in the anterior-posterior extent of limb development, including reduction in digit number, are due to blocking of Shh functions.

Functional Dynamics of Sonic Hedgehog in Anterior-Posterior Axis and Digit Development

Digit number is specified long before their morphological appearance.

Transduction of the Shh signal through the downstream members of its pathway appears to occur very quickly, as the effect of ectopic Shh on processing of Gli3, one of the downstream targets of the Hedgehog pathway, is detectable within one hour (Wang et al.,2000). Experiments with cyclopamine in chicken limb buds showed that the knockdown of the signaling pathway is complete within 6 hr of cyclopamine exposure (C. Tabin, personal communication).

By controlling the extent of anterior-posterior limb development, Shh plays a significant role in determining digit number, and has been implicated in the determination of digit identity (Drossopoulou et al.,2000; Litingtung et al.,2002). In the current experiments, the most obvious effect of cyclopamine is reduction of the number of digits. In the latest stage of exposure, the normal complement of four digits develops in every limb (10 of 10), despite the fact that cyclopamine exposure was started only at the first external sign of the budding of digit 3, several days before digit 4 buds. The next earliest stage of exposure produces four digits in the majority of the exposed limbs (6 of 10) with cyclopamine exposure starting just after the a cleft forms between digits 1 and 2, several days before the morphological appearance of digit 3, and at least 10 days before the morphological budding of digit 4 (Figs. 2, 3D). From these results, we conclude that the limb has already initiated developmental programs downstream of Shh to develop the full complement of four digits even before the third digit is morphologically apparent.

Digit number is specified sequentially over an extended period of development.

A. mexicanum and other urodeles develop the digits of an individual limb far more asynchronously than in other tetrapods, and in a derived anterior-to-posterior sequence. Amniotes and anurans develop all of the digits of an individual limb nearly simultaneously within a paddle-like distal domain. Within this field, digit elements begin to chondrify in close succession with a sequential bias from posterior to anterior. In the derived urodele state of prolonged, and reverse-ordered, digit differentiation, digit number could either (1) be specified in a short time window similar to that of amniotes and anurans, with a progressive delay in the morphological manifestation/differentiation of more posterior digits, or (2) be specified over an extended period of time parallel to the extended timing of digit morphogenesis. If the former model is correct for A. mexicanum limb development, we would expect it to be difficult to precisely time cyclopamine exposures to produce any specific number of digits, with few experimental animals displaying phenotypes intermediate between one digit and four digits. If the latter model is a more accurate representation of A. mexicanum limb development, we expect that it would be possible to repeatedly produce experimental phenotypes that bear an intermediate number of digits, between one and four. Considering the repeatable, sequential, and experimentally separable nature by which our experiments inhibit digit development, it appears that the digits of A. mexicanum are specified sequentially over a long temporal window that parallels the extended time period of their morphological appearance.

Sequential digit loss occurs in a posterior-to-anterior sequence.

In one three-digited limb from the late middle exposure, though carpals exhibit some degree of fusion, distal carpal 4, which normally articulates with the metatarsal of digit 4, is present, but articulates with no digit (Fig. 4B). This indicates that in this three-digited form, it is digit 4 that is missing.

In one two-digited limb from the early middle exposure, two well-formed and distinct digits articulate proximally with a single distinct carpal element (Fig. 4C). This is the same as the case of the articulation of normal digits 1 and 2 with the basale commune. Thus, it appears that these two remaining digits are topologically digits 1 and 2, and it is digits 3 and 4 that are missing. The earliest two exposure windows produced many limbs with only a single digit. In all of these, a single proximal-distal series of elements exists. As an example, we describe here a single limb from the earliest exposure window (Fig. 4D). Most proximally, a single stylopodial element is present. Only a single zeugopodial longbone is formed, followed distally by one nodular carpal element, identifiable as such by being significantly shorter than the long bones of the zeugopod and the digits. Distal to this are three more long elements, with the most distal resembling a terminal phalange. Again, because of the relative lack of identifying characters on the digits of A. mexicanum, it is difficult to identify this digit, and its two phalanges suggest that it could be digit 1, 3, or 4. The zeugopodial element, however, suggests that the remaining elements are anterior structures. In normal limbs, the radius articulates proximally with the anterior of the distal humeral head, whereas the ulna's anterior proximal end articulates with the distal end of the humeral head. The remaining zeugopodial element in this limb clearly has the position and proximal articulation of the normal radius. In most of the limbs with only a single digit, the radius can be similarly identified, and in no case where there is a single zeugopodial element does the articulation clearly resemble that between a normal humerus and ulna. These data suggest that it is the most anterior structures, and therefore topologically digit 1, that remain in experimental limbs that bear only a single digit.

These combined data suggest that axolotl digit loss occurs in a posterior-to-anterior sequence with progressively earlier exposure to cyclopamine. Thus, relatively late downregulation of Hedgehog signaling selectively inhibits the development of the most posterior digit, and progressively earlier downregulation removes more digits in a posterior-to-anterior sequence.

shh is expressed later in A. mexicanum limb development than previously shown.

In a previous report on A. mexicanum, Torok et al. (1999) used in situ hybridization to investigate the expression profile of shh throughout forelimb development. The latest stage at which expression in the ZPA was reported in forelimb development was approximately stage 48 (Nye et al.,2003) just after a cleft forms between digits 1 and 2 (Torok et al.,1999). Surprisingly, at much later stages, approximately stage 52, as digit 4 is just starting to bud from the posterior of the limb, while still showing no detectable shh in the ZPA, their probes showed reactivity in the metacarpals of digits 1 and 2. This was interpreted as a cross-reactivity with ihh/bhh transcripts (Torok et al.,1999). To test this interpretation, we used RT-PCR to detect shh expression in a limb from the same stage. We were unable to detect shh expression in a preparation of the limb including all of digits 1 and 2 and a portion of the anterior distal carpus connecting them, consistent with the idea that the reactivity found by Torok et al. (1999) was not reactivity with shh transcripts. Unexpectedly, however, we were able to detect shh expression in an mRNA preparation from the remaining, more proximal, portion of the same limb.

The data presented here show that Shh is acting very early in limb development to specify digit number, much earlier than the morphological appearance of digits (see above). It is, therefore, surprising that we were able to detect shh expression at such a late stage, when the morphological budding of digit 4 is already visible. Even inhibition of Hedgehog signaling at stages much earlier than this (our late stage of exposure) produces four digits. The skeletal abnormalities that do occur at this late stage of exposure, only some abnormal element fusions, are more likely explainable as resulting from an inhibition of Ihh/Bhh function than Shh function. It is unclear what role shh might be playing at such a late stage in development.

Experimental Phenotypes Are Comparable to Evolutionary Digit Reduction in Urodeles

Limb reduction is very common within tetrapods. In some groups, such as caecilians and most snakes, forelimbs and hindlimbs are lost completely. Adult cetaceans (whales, dolphins, and porpoises) have no hindlimbs (Thewissen et al.,2006). Many other groups have evolved reductions in digit numbers. This is obvious in many mammalian species, especially ungulates, and in many groups of reptiles (Greer,1991; Shapiro,2002). Several clades of salamanders also have reduced limbs (Stoudemayer,1949; Duellman and Trueb,1994). These include both amphiumids and sirens. Amphiumid species bear three digits (Amphiuma tridactylum), two digits (Amphiuma means), or one digit (Amphiuma pholeter) on their forelimbs. While sirens lack hindlimbs, some (genus Siren) retain the ancestral amphibian state of four digits on their forelimbs. Dwarf sirens (genus Pseudobranchus), however, have only three digits on their forelimbs.

Experiments have shown that inducing early cessation of cell proliferation in urodele limbs can cause limb reductions that closely resemble some reductions in nature (Alberch and Gale,1983, 1985), although the most severe reduction produced was the loss of a single digit. Our cyclopamine-induced morphologies, however, comprise a much more drastic range of reduction, with adult morphologies ranging from one to four digits. As was the case with the limb reductions induced by the inhibition of cell proliferation, many of our experimentally-induced limb reductions closely resemble instances of evolutionary limb reductions within urodeles (Figs. 4, 5). As in A. mexicanum, the autopod of most urodeles has a large distal carpal, the basale commune, which articulates with the metacarpals of both digits 1 and 2, and sometimes 3 (Shubin et al.,1995; Wake and Shubin,1998; Figs. 3U, 5A). The metacarpal of digit 2 usually articulates proximally almost exclusively with the basale commune. The metacarpal of digit 1 articulates with the basale commune, and variably with a more anterior-proximal carpal element. And the metacarpal of digit 3 occasionally articulates proximally with the basale commune to some degree (Fig. 5A). The position of the basale commune in relation to digits can, therefore, be used to identify the digits remaining in limbs with reduced digit numbers. A. tridactylum and A. means, though they display different numbers of digits on their forelimbs (three and two, respectively), both retain only three carpal elements (Fig. 5C,D). The pattern of articulation of their remaining digits with the single large distal carpal, the basale commune, suggests that it is the posterior digits that are lost in both cases, digit 4 in A. tridactylum (Fig. 5C), and digits 3 and 4 in A. means (Fig. 5D). The dwarf siren (Pseudobranchus striatus; Fig. 5B), which has three digits, also retains only three carpal elements. The pattern of digit articulation with the basale commune, which is very similar to that of A. tridactylum (Fig. 5C), suggests that it is digit 4 that is lost in P. striatus (Fig. 5B). Thus, it appears that evolutionary digit reduction in the urodeles is proceeding with posterior-to-anterior polarity.

Cyclopamine-induced reductions to three digits exhibit a range of carpal morphologies. At the least-reduced end of the spectrum of three-digited morphologies, limbs have a long digit 3 and retain a carpus with clearly identifiable representatives of every normal carpal element (Fig. 4B). At the most reduced end of the range of three-digited morphologies, a very small digit 3 exists, usually not articulating closely with a carpal element, and there are as few as three carpal elements (Fig. 3R and data not shown). These experimental reductions to three digits with three carpals thus resemble P. striatus (Fig. 5B) and A. tridactylum (Fig. 5C). Two-digited experimentally-induced morphologies that retain two full digits usually have three carpal elements (Fig. 4C), very similar to those in A. means (Fig. 5D). There is, therefore, a great deal of morphological similarity between the morphologies of urodele forelimbs with natural digit reduction and reductions produced in A. mexicanum by cyclopamine-induced downregulation of Hedgehog signaling. Furthermore, both natural and cyclopamine-induced reductions appear to proceed by loss of digits in a posterior-to-anterior sequence, the opposite polarity of the order of developmental appearance of digits, and therefore in accordance with Morse's law (Morse,1872). This implies, as was suggested to be the case for reduced digits in lizards in the genus Hemiergis (Shapiro et al.,2003), and evolutionary hindlimb reduction in cetaceans (Thewissen et al.,2006), that early downregulation in sonic hedgehog expression or reduced transduction could be responsible for the pattern of natural reduction in digit number within urodeles.

A. pholeter has only a single digit on its forelimbs (Fig. 5E) and hindlimbs. This reduced limb retains a distinct radius and ulna. In contrast, our cyclopamine-induced single digit morphologies have only a single zeugopodial element (Fig. 4D). This retention of two zeugopodial elements suggests that there is still a substantial amount of Shh functioning in the development of the A. pholeter zeugopod and that this natural reduction probably occurred by some mechanism other than a simple temporal change in Shh function. This does not necessarily rule out changes in Shh dynamics as a mechanism for this reduction, as our experiments sought to change only the temporal dynamics of Shh function, not Shh spatial patterns. It is possible that the A. pholeter digit reduction was instantiated by a change in the spatial dynamics of Shh function, with ancestral levels retained proximally in specifying the zeugopod, but reductions in the level of more distal shh expression or Shh signal transduction specifying digit number.


Through cyclopamine exposures and semi-quantitative RT-PCR, we show that, in A. mexicanum (1) digits are specified long before their morphological appearance, (2) digit number is specified over an extensive period of development, (3) digit reductions due to progressively earlier Shh inhibition occur in a posterior-to-anterior sequence, and (4) shh is expressed much later than was previously shown. Furthermore, the phenotypes resulting from these experiments suggest that instances of evolutionary reduction of digit number in some urodeles could occur due to an early cessation of Shh expression. Unfortunately, although we know a great deal about sonic hedgehog expression and function in tetrapods, detailed data on the temporal dynamics of its function throughout limb development do not exist from any tetrapod that retains the ancestral order and pattern of digit development. Without these comparable data, it is impossible to draw conclusions about how the temporal dynamics have changed along the urodele lineage in concert with the change in the order of digit development. When compared to the timing of digit development in urodeles, ancestral tetrapods develop their digits quickly, in relative synchrony. Considering this relative synchrony of digit development in the ancestral tetrapod pattern, it is quite unlikely that Shh functioned in the specification of anterior-posterior pattern and digit number over the very protracted time period of limb development that we find in A. mexicanum. Therefore, it is likely that the transition from the ancestral pattern to the derived pattern of digit development in urodeles involved an extension of the time period during which Sonic hedgehog is functioning in the specification of digit number.


We thank Stéphane Roy for sharing cyclopamine exposure methods, David Wake for sharing urodele carpal morphology data for the Pseudobranchus striatus specimen MVZ198685, and Cliff Tabin for sharing unpublished data on molecular response to cyclopamine exposure. We are grateful to the Yale Peabody Museum of Natural History and the Florida Museum of Natural History for providing specimens (Siren lacertina specimen YPM103 and Amphiuma pholeter specimen UF28812, respectively). We also thank Stephen Stearns and two anonymous reviewers for comments and suggestions that improved the quality of this report.