Attenuation of bone morphogenetic protein signaling during amphibian limb development results in the generation of stage-specific defects



Caroline W. Beck, Department of Zoology, University of Otago, P.O. Box 56, Dunedin, New Zealand. T: +64(0)34794109; F: +64(0)34797584; E:


The vertebrate limb is one of the most intensively studied organs in the field of developmental biology. Limb development in tetrapod vertebrates is highly conserved and dependent on the interaction of several important molecular pathways. The bone morphogenetic protein (BMP) signaling cascade is one of these pathways and has been shown to be crucial for several aspects of limb development. Here, we have used a Xenopus laevis transgenic line, in which expression of the inhibitor Noggin is under the control of the heat-shock promoter hsp70 to examine the effects of attenuation of BMP signaling at different stages of limb development. Remarkably different phenotypes were produced at different stages, illustrating the varied roles of BMP in development of the limb. Very early limb buds appeared to be refractory to the effects of BMP attenuation, developing normally in most cases. Ectopic limbs were produced by overexpression of Noggin corresponding to a brief window of limb development at about stage 49/50, as recently described by Christen et al. (2012). Attenuation of BMP signaling in stage 51 or 52 tadpoles lead to a reduction in the number of digits formed, resulting in hypodactyly or ectrodactyly, as well as occasional defects in the more proximal tibia-fibula. Finally, inhibition at stage 54 (paddle stage) led to the formation of dramatically shortened digits resulting from loss of distal phalanges. Transcriptome analysis has revealed the possibility that more Noggin-sensitive members of the BMP family could be involved in limb development than previously suspected. Our analysis demonstrates the usefulness of heat-shock-driven gene expression as an effective method for inhibiting a developmental pathway at different times during limb development.


Tetrapod limb development has been extensively studied in amniote model organisms such as chickens and mice (Towers & Tickle, 2009; Zeller et al. 2009) and, in humans, mutations that alter limb patterning offer useful landmarks for clinical geneticists. Classical and molecular studies have combined to generate the current model of limb development, where two main signaling centers interact with each other to produce a patterned limb (Towers & Tickle, 2009). One of these signaling centers, the apical ectodermal ridge (AER), regulates the proximal to distal (PD) pattern and is positioned by the restricted expression of different genes in the dorsal and ventral compartments (Altabef et al. 1997). The activity of the AER depends on localized signaling regulated by several members of the fibroblast growth factor (FGF) signaling family, including Fgf8. The other main signaling center, the zone of polarizing activity (ZPA), regulates the anterior to posterior (AP) pattern, most obvious in the number and identity of the autopodal digits (Tickle et al. 1975). The ZPA is positioned by the activity of Hand2 and the 5′ Hox genes (Tarchini et al. 2006; Galli et al. 2010), and expresses the morphogen Shh (Riddle et al. 1993). The ZPA and AER interact with each other by means of a reinforcing positive feedback loop (Laufer et al. 1997). This self-regulatory loop has been extensively investigated in both chickens and mice, and depends on the activity of Gremlin 1 (Grm1), a secreted bone morphogenetic protein (BMP) inhibitor that is induced by Sonic Hedgehog (Shh) in the early limb bud mesenchyme and acts via BMP regulation to maintain expression of FGFs in the AER (Capdevila et al. 1999; Zuniga et al. 1999; Khokha et al. 2003). Later in limb bud outgrowth, the increasing FGF levels reach a threshold that triggers a second, inhibitory loop, where high FGF levels switch off Grm1, breaking the autoregulatory loop and resulting in its own eventual downregulation and the subsequent demise of the AER, which in turn brings limb bud outgrowth to a halt (Verheyden & Sun, 2008; Benazet & Zeller, 2009). This model is often assumed to hold true for all tetrapod vertebrates (for a review highlighting differences, see Stopper & Wagner, 2005). However, we know correspondingly little of the pathways that control amphibian limb development.

Early frog limb buds look very much like those of chickens and mice, they are small mesodermal bulges surrounded by an epithelial jacket. Molecular markers have shown that a shh-expressing ZPA and an fgf8-expressing cryptic AER (due to its cryptic nature in amphibians, the equivalent signaling center is called the apical epidermal cap, or AEC) are present in these limb buds, suggesting that the AP and PD patterning signals are the same as in amniotes (Christen & Slack, 1997, 1998). However, the establishment of dorso-ventral (DV) patterning in Xenopus limbs appears to differ from amniotes, fgf-8 expression is not seen prior to limb bud emergence (Christen & Slack, 1998) and the mechanisms that lead to limb bud positioning and outgrowth are unknown. Whereas in other vertebrates the limbs develop concurrently with other organs, in Xenopus the development of limbs accompanies metamorphosis, and limb buds emerge from swimming, feeding larvae. Another peculiarity of Xenopus is that hindlimbs develop slightly ahead of forelimbs, the opposite of amniotes and indeed urodele amphibians like the axolotl, whose lifecycle also includes a limbless larval stage. Recent evidence from the first insertional mutant Xenopus (Xenopus de Milo) to be characterized suggests that the kidneys provide signals that enable the forelimbs to develop, and that unique processes induce limb bud formation in the flank at forelimb and hindlimb levels (Abu-Daya et al. 2011).

There is compelling molecular genetic evidence from amniotes that BMP signaling has different functions at different time-points in limb development (Benazet et al. 2009). Early inhibition leads to the formation of distally truncated limbs, whereas later inhibition causes polydactyly and increased number of digits (Benazet et al. 2009). Xenopus limb development is of particular interest because early limb buds possess the ability to regenerate perfectly (Dent, 1962); therefore, pattern regulation in the frog limb appears to be more flexible than the amniote limb. We set out to examine the stage-specific roles of BMP signaling during limb development in the frog Xenopus laevis, by transiently inhibiting BMP signaling using a transgenic line of X. laevis containing the BMP antagonist Noggin under the control of a heat-shock promoter (Beck et al. 2003). Noggin was chosen because it is known to inhibit BMPs 2, 4 and 7, which are thought to be the key players in limb development (Francis et al. 1994; Zimmerman et al. 1996; Pizette et al. 2001; Bandyopadhyay et al. 2006). Previously, we have shown that noggin and bmp4 are expressed reciprocally in the mesenchyme of the developing Xenopus limb autopod, with noggin expression corresponding to forming digits and bmp4 to interdigital regions (Christen & Slack, 1998; Beck et al. 2006). Bmp2 is expressed in the apical epidermis, and distal AP mesenchyme (Beck et al. 2006) The expression of other potential Noggin targets in developing Xenopus limbs is unknown.

We hypothesized that the role of the BMP pathway would change throughout limb development in Xenopus because of evidence in amniotes of its time-dependent interactions with the key limb bud signaling centers, the ZPA and AER. To examine this hypothesis, we inhibited BMP signaling for a week during early limb development and subsequently examined the limb phenotypes resulting from this inhibition. We also determined the expression of key limb development markers to determine the mechanistic cause of limb defects following BMP inhibition. Finally, we show evidence that a broader range of potentially noggin-sensitive BMP ligands are expressed in developing Xenopus limbs than was previously suspected.

Materials and methods

Culture of X. laevis tadpoles

Xenopus laevis embryos and tadpoles were cultured as previously described (Barker & Beck, 2009), and staged according to Nieuwkoop & Faber (1967). All animal experiments were subject to New Zealand's animal welfare standards for vertebrates and were reviewed by the University of Otago Animal Ethics Committee, who approved all experiments under protocols AEC78/06 and 56/09.

Whole-mount skeletal (bone and cartilage) staining

Skeletal staining was carried out using an adaptation of the method described by Newman & Dumont (1983). Tadpoles were fixed in 4% formaldehyde/phosphate-buffered saline (PBS) for at least 24 h. After rinsing in PBS, viscera were removed manually before bleaching the tadpoles in 6% H2O2 to remove most of the pigment. Air bubbles were removed by brief exposure to a vacuum. Tadpoles were washed with 70% ethanol, and subsequently stained for cartilage using alcian blue stain (0.001% alcian blue in 40% acetic acid and 60% ethanol) overnight. Tadpoles were then destained in 70% ethanol, and soaked in 1% KOH for 2–3 days to clear tissues. Bone was then stained using alizarin red (10% alizarin red S, 1% KOH), before returning samples to 1% KOH overnight.

In situ hybridization and probes

Xenopus laevis fgf8, sox9 and shh were cloned from pooled cDNA isolated from embryos at stages 13, 25 and limb buds at stages 50–54. Tissues had been stored in RNAlater (Qiagen) and RNA was extracted with Trizol following the manufacturers' instructions. cDNA was generated using the Superscript III system, following manufacturers' instructions. fgf8, shh, gdf5 and sox9 were amplified using primers containing ClaI restriction sites at the 5′ end of the forward primers and XbaI sites at the 5′ end of the reverse primers and ligated into pBSIIKS+. Bmp7.1 and hoxd11 were amplified with specific primers and cloned into PCR4-TOPO (Invitrogen). Primers are listed in Table S1, except for hoxd13, which has been previously described (Christen et al. 2003). Antisense in situ hybridization probes were synthesized by run-off transcription with T7 or T3 RNA polymerase, precipitated with 7.5 m LiCl overnight and purified by centrifugation and washing with 70% EtOH, before resuspending in 10 mm EDTA. Whole-mount in situ hybridization was performed as previously described (Beck & Slack, 1998), with modifications for limb buds as in McEwan et al. (2011).

Transgenic line N1

The transgenic line used here has been previously described (Beck et al. 2006). The line is descended from a female founder with a single insertion of a double transgene -Hsp70:noggin1- γ-crystallin:GFP. Inhibition of BMP signaling was achieved by activating noggin expression in our transgenic line, by heat-shocking tadpoles daily for 1 week. This regime has been previously shown to induce high, biologically active levels of noggin transcripts (Beck et al. 2006). Heat-shocks were performed by transferring the tadpoles to water at 34 °C for 30 min and subsequently returning them to 25 °C. Tadpoles at stage 48 were heat-shocked for 5 days instead of 7 days, to increase survival rates. Once inhibition of BMP signaling was complete, tadpoles were either killed with MS222 and fixed at the appropriate stages for in situ hybridization, or grown to stage 58 for limb skeletal analysis.


Statistical analysis used unpaired t-tests to compare means of digit number for each of the four limbs to the corresponding limb in the wild-type (WT) stage 50, heat-shocked cohort. Differences were defined as significant if P < 0.05, and highly significant if P < 0.0001. Raw data for some of the categories are available as Table S2.

Bioinformatic identification of BMP family members in the developing limb bud

Deep sequencing (Illumina HiSeq, paired end) was carried out to compare the transcriptomes of stage 51 hindlimb buds divided into three equal-sized sections along the PD axis (distal, mid and proximal). Sequence reads were assembled using Trinity (Grabherr et al. 2011) and annotated with Blast2GO (Conesa et al. 2005). The number of reads for each gene was normalized for library size. BMP family members were deemed to be expressed in stage 51 limbs if the ‘total’ number of reads across all libraries was > 20. Protein sequence was obtained for each member and aligned using Genious software to determine the presence or absence of an Asparagine residue at amino acid 445. The presence of the N455 indicates a likely Noggin target (Seemann et al. 2009).


Limb defects result from temporary attenuation of BMP signaling at stages 50–54

Tadpoles were subjected to daily heat-shock treatments in order to inhibit the BMP signaling pathway for 1 week, with treatment beginning at different stages of limb development. Sibling WT tadpoles were subjected to the same heat-shock protocol to control for any effects of the heat-shock procedure itself. Limb defects were initially assessed in stage 58 tadpoles by counting the number of digits present on each of the four limbs, and confirmed after subsequent skeletal staining of fixed tadpoles. Stage 58 was used as an endpoint because forelimbs develop under a skin flap and only emerge at this stage of development. Once this has occurred, tadpoles rapidly complete metamorphosis, and affected animals are likely to find swimming difficult once the tail has regressed.

In X. laevis, normal developing forelimbs have four digits, numbered II–V (Satoh et al. 2006; Fig. 1A), and hindlimbs have five digits, I–V, with the anterior three digits bearing claws (Fig. 1B). No effect on the number of fore- or hindlimb digits was seen when BMP signaling was inhibited in transgenic N1 tadpoles starting at stage 48 and spanning stage 49 (Fig. 1D,E, n = 15). Similarly, in WT siblings that do not carry the transgene, normal numbers of digits were observed on all limbs and in all cases (n = 16) following heat-shock treatments from stage 50 and spanning stage 51. However, the number of forelimb digits was significantly reduced when BMP signaling was inhibited from stage 50 in tadpoles carrying the N1 transgene, when compared with their WT siblings (P = 0.0046 left forelimb and P = 0.0058 right forelimb), with the average number of digits being 3.1 for either forelimb (n = 25). Similarly, the number of digits on each hindlimb was reduced to an average of 2.8 for left and 2.9 for the right hindlimb (n = 25). The number of hindlimb digits was highly significantly different from WT animals (P < 0.0001).

Figure 1.

Digit development and the effect of noggin overexpression during limb development. (A, B) Normal Xenopus laevis forelimb (A) and hindlimb (B) at stage 58 stained for bone (dark gray) and cartilage showing the arrangement of the skeletal elements. (C) Stage series of X. laevis limb development, adapted from Nieuwkoop & Faber (1967). (D, E) Bar graphs showing the total number of formed digits, regardless of size, found on each limb after 1 week of daily heat-shock transgene induction beginning at the indicated stage of development in N1 or wild-type (WT) tadpoles, scored at stage 58. Note that in cases where multiple limbs were formed, only the digits on the original limbs were counted. Error bars show standard error (SEM) and * indicates significant difference compared with the corresponding WT limb (P < 0.05), ** indicates very strong significant difference (P < 0.0001). (D) Forelimbs, note that the normal number of digits per limb is four. (E) Hindlimbs, note that the normal number of digits per limb is five. LFL, left forelimb; LHL, left hindlimb; RFL, right forelimb; RHL, right hindlimb. n = 16 (WT), 15 (N1 st.48), 25 (N1 st.50), 14 (N1 st.51), 16 (N1 st. 52), 22 (N1 st.54).

orelimb digits were even more severely affected with later BMP inhibition. When BMP was inhibited from stage 51 (spanning 52), the average number of forelimb digits was 1.9 (n = 14), and when inhibition began at stage 52 (spanning 53), an average of 1.3 digits per left forelimb and 1.4 per right forelimb was observed (n = 16). Both treatments were significantly different from WT controls (P < 0.0001). By stage 54, digits are starting to form in the forelimb, and inhibition of BMP from this stage, spanning stage 55, did not result in significant changes to the number of digits on each forelimb (average = 3.9 digits, n = 22).

A similar picture was seen in hindlimbs, although the number of digits developing was similar after treatments beginning at stages 50, 51 or 52. BMP inhibition starting from stage 51 resulted in an average of 2.6 digits per foot (n = 14), whereas treatments beginning at stage 52 resulted in the development of an average of 2.3 toes per left foot and 2.4 per right foot (n = 16). Again, digit number resulting from either treatment was highly significantly different from control WT tadpoles (P < 0.0001). Hindlimb digits can be observed from stage 54 onwards, and inhibition of BMP from this stage resulted in only a small reduction in digit number, with an average of 4.7 toes per foot (n = 22). Despite this apparently small reduction, the number of toes was still significantly different from WT controls (P = 0.023).

Noggin overexpression can result in the formation of ectopic limbs

The most dramatic phenotype observed after overexpression of the BMP inhibitor Noggin during limb development was the formation of ectopic limbs (Fig. 2). This phenotype has recently been described and characterized by others (Christen et al. 2012). Ectopic limb formation was somewhat dependent on genetic background, occurring frequently in some tadpole cohorts but absent from others. Ectopic limbs were recorded in two of the four cohorts of tadpoles used in the study. The number of ectopic limbs varied from a single extra limb to three extra limbs, with no more than two limbs in total developing at one site. Ectopic limbs were only ever observed in N1 tadpoles exposed to heat-shock treatments from stage 50, spanning stage 51, slightly later than those described by Christen et al. (2012). WT limbs treated in the same way never developed ectopic limbs. The average number of limbs observed per tadpole was 4.7 ± 0.20 (Fig. 2A).

Figure 2.

Ectopic limbs can form when tadpoles overexpress noggin during early limb bud stages. (A) Bar graph showing the incidence of ectopic forelimbs and hindlimbs in tadpoles scored at stage 58. The sample number (n) for each stage is in parentheses under the x-axis, the actual number of cases is shown on the bars. (B) Ventral view of seven-legged tadpole. (C, D) Skeletal staining or tadpole in (B) viewed from ventral (C) and dorsal (D) aspects. (E, F) In situ hybridization of fgf8 expression in wild-type (WT) (E) and N1 (F) stage 51 tadpole hindlimb. Arrowheads indicate duplicated AERs. Hindlimb buds are oriented with proximal to the left and posterior uppermost. d, dorsal; f, forelimb; h, hindlimb; l, left; r, right; v, ventral.

The most commonly observed ectopic limb phenotype (six cases, n = 25) was two extra hindlimbs. Two further cases had an ectopic hindlimb on the left side only. The ventral (ectopic) pair of limbs tended to appear normal, but posture suggested low muscle tone in these limbs. Five toes with normal AP polarity were often present. In one case, the left ventral limb was synpolydactylous, with six digits, and the right had just two digits, and a single digit was seen on the left limb in another case. Dorsal hindlimbs also had abnormal posture and appeared hyperdorsalized, lacking the less pigmented skin characteristic of the ventral limb surface (Fig. 2B). Ectopic forelimbs were seen less frequently than hindlimbs, (four cases, n = 25) and were only ever present on one side of the animal (Fig. 2B–D). Two cases had the normal number of digits (four), one had three digits and one had two digits.

The ectopic limbs have been extensively examined for limb patterning gene expression by Christen et al. (2012), who confirmed that the limb duplication arises from expansion of the dorsal domain, resulting in formation of a ventrally placed secondary limb with normal DV pattern, and in dorsalization of the original limb, with each limb maintaining its own AER expressing fgf8 (Fig. 2E,F).

Overexpression of Noggin during mid limb bud de-velopment causes a reduction in the number of digits (hypodactyly)

In many cases, BMP inhibition during limb development resulted in the formation of a hypomorphic autopod comprising of a single digit with one metacarpal or metatarsal and one phalanx (Fig. 3). When this occurred, it was always bilateral, with no asymmetry observed between left and right limbs. Forelimbs were more frequently affected than hindlimbs, and the number of occurrences increased from 16 (stage 50) to 57% (stage 51), and affected 75% of tadpoles when BMP signaling was inhibited from stage 52 (Fig. 3A). One or two phalanges were present on the single forelimb digit possibly suggesting that it has a digit IV or V identity (Fig. 4B). However, in affected hindlimbs, claws were most often present (100% of stage 50, 83% of stage 51 and 63% of stage 52 treated limbs had a claw on the single digit), suggestive of an anterior, digit I or II identity (Fig. 3C). Skeletal elements were normally proportioned, although in some cases, a second, considerably shortened metatarsal was revealed on the hindlimb autopods after skeletal staining (Fig. 3D). The tibiale and fibulare were both present in some limbs (Fig. 3D), whereas a single bone was seen in others (Fig. 3C). Hypodactylous limbs were not found to be missing any more proximal elements, and limbs were appropriately proportioned, suggesting that only the autopod is affected by loss of BMP signaling between stages 50 and 53.

Figure 3.

Overexpression of noggin can cause hypodactyly. (A) Bar graph showing the occurrence of the single digit phenotype as a percentage in fore- and hindlimbs. The actual numbers of animals with this phenotype are shown inside the bars. The number of cases for each treatment (n) is in parentheses below the x-axis. (B–D) Skeletal staining of limbs. (B) Forelimb of stage 52 treated N1 tadpole, showing single metacarpal and two phalanges, viewed from the dorsal side. (C) Ventral view of the hindlimbs of a stage 52 treated N1 tadpole showing a single clawed digit on each limb, with a single metatarsal and two short phalanges. In this tadpole, the tibiale and fibulare are represented by a single bone of unknown identity. (D) Ventral view of hindlimb foot region of stage 52 treated N1 tadpole indicating a single complete metatarsal and a second, vestigal metatarsal. This example has just one phalange with a claw. c, claw; h, humerus; mc, metacarpal; mt, metatarsal; p, phalanges; ru, radioulna; tf, tibia-fibula; v, vestigal metatarsal.

Figure 4.

Overexpression of noggin at stage 50 leads to reduction of fgf8 expression in the hindlimb AER. In situ hybridization of tadpole hindlimb buds for the AER/AEC marker fgf8 (dark blue staining). Tadpoles were heat-shocked from stage 50 to 51, and fixed at various stages. WT denotes wild-type animals and N1 denotes sibling animals containing the N (Hsp70:noggin-γ-crys:GFP) transgene. (A) WT hindlimb bud at stage 52. (B) N1 hindlimb at stage 52, box indicates enlarged region of the AER in B′. (C) WT hindlimb at stage 53. (D) N1 hindlimb at stage 53 showing much reduced fgf8 expression. Box indicates region shown in D′. (E) WT hindlimb at stage 54. Roman numerals indicate the approximate positions of the forming digits. (F) N1 hindlimb at stage 54, box indicates the region shown in F′. Arrowheads show the extent of the fgf8 expression in presumptive AER in each case. Scale bar: 0.5 mm (in A, A–F). Tadpoles are shown with anterior to the left and dorsal uppermost, so that limbs are viewed dorsally, and oriented posterior up and proximal to the left.

Overexpression of Noggin leads to reduced expre-ssion of the AEC marker fgf8

To understand how overexpressing noggin from stage 50 leads to loss of digits and a hypodactylous phenotype, we performed in situ hybridization on stage 50 heat-shocked hindlimbs fixed at different stages of development (Fig. 4). In WT limbs, fgf8 expression marks the cells of the cryptic AER, or AEC (Christen & Slack, 1997). Heat-shocked WT tadpoles fixed at stage 52 showed strong expression of fgf8 throughout the AEC, with stronger staining in the anterior part (Fig. 4A). By stage 53, as digit IV begins to form, fgf8 expression is limited to the AEC from just posterior to digit IV to the region covering future digit I (Fig. 4C). The mesenchyme that will form digit V does not have fgf8-positive epithelial cells covering it at this stage. By stage 54, when all five digits can be observed, the AEC has regressed so that the fgf8 staining is restricted to the AEC of digits II and I, with only faint staining in the AEC overlying the tips of digits III and IV (Fig. 4E). In contrast, limb buds of N1 tadpoles that have had noggin expression elevated from stage 50 showed much reduced fgf8 staining due to earlier loss of fgf8 from the anterior AEC. By stage 52, these limb buds have acquired a characteristic exaggerated anterior flexion at the wrist level, and fgf8 expression is only seen in the anterior half of the AEC, and even this appears much weaker than in stage-matched WT (Fig. 4B,B′). By stage 53 this expression is limited to the AEC-covering digit IV (Fig. 4D,D′), and by stage 54, digit IV has extended, resulting in a characteristic pointed shape to the autopod (Fig. 4F,F′). Failure of digits I–III to develop appears to correlate with a loss of fgf8 expression in the AEC cells flanking the anterior mesenchyme. Failure of digit V to develop may result from loss of shh expression, normally found in the posterior mesenchyme in cells corresponding to the ZPA that are fated to end up in digit V. Supporting this, in N1 tadpoles heat-shocked at stage 50, shh expression was found to be absent (Fig. S1).

To confirm the identity of the hypodactylous digit as digit IV, we looked at expression of sox9, which labels developing endochondral cartilage condensations (Fig. 5). Compared with WT (Fig. 5A,C), N1 transgenic limbs that had been heat-shocked to activate noggin expression from stage 50 showed a clear reduction in the number of strongly sox9-positive condensations detected at stages 53 or 55, with only digit IV showing apparently normal sox9 expression (Fig. 5B,D). Activation of noggin expression at the earlier stage 48 did not result in a reduction in observed sox9 condensations at stage 54 (Fig. 5E,F), confirming phenotypic observations.

Figure 5.

Overexpression of noggin at stages 50–51 results in a reduction in the number of digit cartilage condensations. In situ hybridization of tadpole hindlimb buds for the early cartilage marker sox9 (dark blue staining). Tadpoles were heat-shocked from stage 50 to 51, except for (E), which was heat-shocked at stages 48–49 as a control, and fixed at stages 53–55. Roman numerals indicate approximate positions of the forming digits. (A) Wild type (WT) hindlimb at stage 53, note clear sox9 expression in condensing cartilages of digits V, IV and III. (B) N1 limb at stage 53 shows expression of sox9 is restricted to digit IV. (C) By stage 55 all five digits show clear sox9 expression in cartilages of WT hindlimb autopod. (D) In contrast, in the stage 55 N1 hindlimb only digit IV clearly expresses Sox9. (E, F) This effect is stage specific, when stage 48–49 N1 tadpoles are heat-shocked, overexpression of noggin does not alter the formation of the digits. Compare WT (E) with N1 (F) stage 54 limb, note digit I has not yet developed. Scale bar: 0.5 mm (in A, A–F). Tadpoles are shown with anterior to the left and dorsal uppermost, so that limbs are viewed dorsally, and oriented posterior up and proximal to the left. (G, H) In situ hybridization of N1 stage 50 heat-shocked tadpole hindlimbs for markers of digit identity at stage 54. (G) hoxd13 expression in all digits including digit IV. (H) hoxd11 expression. Scale bar: 0.5 mm (in G, G and H).

While the single digit morphologically resembles the anterior digit I, it has developed from the position of digit IV, and also shows the hox code of a posterior digit, expressing both hoxd13 and hoxd11 (Fig. 5G,H).

Noggin overexpression in later limb buds results in ectrodactyly

The most common phenotype observed in hindlimbs following transgenic inhibition of BMP signaling from stage 52 and spanning stage 53 was an apparent ectrodactyly, or split foot, which was seen in 50% of cases at this stage (n = 16; Fig. 6A). In all eight of these animals, hindlimb ectrodactyly was bilateral. The two digits were clearly separated, with reduced skin webbing between them. The more posteriorly located digit had a claw, suggesting anterior polarity, and the other was observed to be clawless. On closer examination, following skeletal preparation, the clawed toe was seen to consist of a single metatarsal and short phalanx, whereas the more anteriorly located toe was formed from a metatarsal, which appeared to have no distal growth plate. Often, a central short metatarsal was also present, again with no distal growth plate evident (Fig. 6B,C). Ectrodactyly was also seen in the hindlimbs of animals treated at stage 50; however, of six such cases (n = 25), three were dorsal ectopic limbs and two were unilateral.

Figure 6.

Overexpression of noggin in later limb buds results in hindlimb ectrodactyly as well as more proximal defects. Bar graph showing the occurrence of ectrodactyly phenotype as a percentage in fore- and hindlimbs. The actual numbers of animals with this phenotype are shown inside the bars. The number of cases for each treatment (n) is in parentheses below the x-axis. (B) Dorsal view of the left hindlimb of N1 animal heat-shocked at stage 52. Two metatarsals and a single claw have formed an ectrodactylous autopod. (C) Ventral view of N1 animal heat-shocked at stage 51 showing left hindlimb. The ectrodactylous limb is composed of two metatarsals: one with a distal claw, and a short central metatarsal. (D) Ectrodactyly of the forelimb following activation of noggin expression at stage 51. Ventral view of animal showing left hindlimb. Two digits are fused at the distal phalange. (E) Bar graph showing the frequency of zeugopod defects in the hindlimbs of N1 animals heat-shocked at different stages of limb bud development. The actual numbers of animals with this phenotype are shown inside the bars. The number of cases for each treatment (n) is in parentheses below the x-axis. (F) Example of shortened tibia-fibula, animal heat-shocked at stage 51 viewed from the dorsal side showing both hindlimbs. The tibia-fibulas on each limb are U-shaped and hypoplastic. (G) Example of absent tibia-fibula. Animal heat-shocked from stage 52 and viewed from the ventral side to show the right hindlimb. The tarsus, tibaile, fibulare and femur are intact, but the tibia-fibula is missing. Red staining indicates ossified bone and blue indicates cartilage. c, claw; f, femur; h, humerus; mc, metacarpal; mt, metatarsal; p, phalange(s); ru, radioulna; t, tarsals; tf, tibia-fibula; WT, wild-type.

Forelimbs were less frequently affected by ectrodactyly, peaking in stage 51 treatments where 31% were affected (n = 16). In contrast to the hindlimbs, the two digits were not well separated, often with soft tissue syndactyly. Both digits possessed phalanges, and most closely resembled digits IV and V (Fig. 6D).

Noggin overexpression can cause more proximal defects in hindlimbs, with missing or shortened tibia-fibula

Most of the defects seen in noggin-overexpressing limb buds, regardless of treatment stage, affected the autopod, sparing the more proximal limb elements. However, shortening or absence of the tibia-fibula was also observed at low frequency in hindlimbs with earlier noggin expression (stages 50–52; Fig. 6E). Tadpoles with tibia-fibula defects had unusual hindlimb posture that resulted either from a shortened, U-shaped tibia-fibula (hypoplastic, Fig. 6F) or an absent tibia-fibula (aplastic, Fig. 6G). These defects in the tibia-fibula were associated with a hypoplastic autopod, with reduction in the number of digits to two or one per hindlimb, along with brachydactyly of the digits. Curiously though, the other bones and cartilages of the foot: the tarsals, tibiale and fibiale, appeared to be unaffected (Fig. 6F,G). Therefore, this is not a straight case of hypoplasia of the distal hindlimb structures. Unlike the other phenotypes affecting the autopod, the affect on the tibia-fibula was often unilateral, but there was no clear trend to left or right limbs being affected (four cases right only, two cases left only). A corresponding defect in the forelimb radioulna was never observed.

Brachydactyly, or shortened digits, results from over-expression of Noggin in paddle stage limbs

When BMP inhibition was induced in transgenic animals beginning at stage 54, when the limb autopod is developing digits, the most common defect was brachydactyly, or shortened digits (Fig. 7). In this cohort, 86% of hindlimbs and 73% of forelimbs were affected (n = 22) in at least one digit (Fig. 7A). When skeletal preparations were made, the shortened digits could be seen to be the result of missing elements, most often of the distal phalanges (Fig. 7B–D). Growth plates, seen as regions of non-ossified cartilage at either end of the digit skeletal elements, were occasionally noted to be absent (Fig. 7B), along with corresponding joints. In some cases, phalanges failed to develop altogether, with the metacarpal or tarsal terminating abruptly at the expected length (Fig. 7C). Anterior digits were most likely to be absent (Fig. 7D; Table 1). Vestigal digits, where a partial metacarpal was seen, also occurred at low frequency in this group of animals (Fig. 7E; Table 1). Due to the loss of distal phalanges, hindlimb claws were rarely seen. Claws were most likely to be present on digit I (4/22), and only in a single case were three claws observed on digits I, II and III and none of these digits exhibited brachydactyly.

Table 1. Percentage of cases of isolated brachydactyly in each digit of the stage 54 treated cohort (n = 22)
  1. n/a, not applicable, as Xenopus forelimb only has digits II–V.

Figure 7.

Overexpression of Noggin during limb development leads to brachydactyly. (A) Bar graph showing the frequency at which brachydactyly (short digits) was observed in the hind- and forelimbs of wild-type (WT) and transgenic N1 tadpoles. The actual numbers of animals with this phenotype are shown inside the bars. The number of cases for each treatment (n) is in parentheses below the x-axis. (B–F) Skeletal preparations of fore- and hindlimb autopods of tadpoles following activation of the Noggin transgene at stage 54. Digits are labelled I–V according to standard nomenclature, where digit I is the most anterior digit. Red staining indicates bone formation and blue staining marks cartilage. (B) Mildly affected forelimb. Digit II is normal, digit III has one missing phalanx and one very small malformed phalanx, digit IV is as digit I but has also lost the metacarpal growth plate and joint (asterisk), and digit V is normal. (C) More severely affected forelimb, with phalanges missing from all four digits and digit VI having no phalanges. (D) Example of hindlimb with four digits. Digit II has only a metatarsal, digits III and V have a single phalanx, and digit IV has two missing phalanges. (E) Hindlimb with vestigal digit V, near normal digit II with only the claw absent, two phalanges missing from digit III and all four phalanges missing from digit IV. All are ventral views except C (dorsal). ru, radioulna; t, tarsals.

Hindlimb brachydactyly was also noted in association with other phenotypes caused by earlier Noggin overexpression, for example with ectrodactyly in Fig. 6B,C. These cases could not be accurately assigned a digit number and are not included in Table 1, which only shows isolated cases in the stage 54 treated cohort.

Analysis of the stage 51 limb bud transcriptome indicates there may be more noggin targets in developing limbs

As part of an ongoing study, we have sequenced the transcriptome of three dissected regions along the stage 51 hindlimb PD axis. Our analysis of these data indicates that in addition to gdf5, bmp2, bmp4 and bmp7, there may be other family members potentially playing a role in limb development in Xenopus (Fig. 8A,B). A likely homolog of bmp5 showed very high expression in early limb buds, with transcripts detected in a graded fashion from PD. bmp3b, gdf6, bmp8b and bmp6 were also identified. The protein products of these additional BMP family members are predicted to be sensitive to Noggin inhibition, as determined by the presence of a conserved Asparagine at amino acid 445. The primarily distal expression patterns of bmp2 and bmp4 have been previously described in Xenopus limb development (Christen & Slack, 1998; Beck et al. 2006). gdf5 expression in Xenopus was localized to forming joints in a PD temporal fashion as expected from its expression in amniote vertebrates (data not shown). Bmp7 expression also appears to resemble that seen in other vertebrate limbs (Fig. 8B–F), however, we note an early proximal restricted expression (Fig. 8B) that has not been previously noted.

Figure 8.

Potential targets of Noggin inhibition in early Xenopus limb buds. (A) Stacked column graph representing normalized relative expression, as a percentage of the total reads for that gene, of all detectable BMP family members as determined by transcriptome analysis of stage 51 hindlimb bud proximal, middle and distal thirds. (A′) Bar graph of the number of raw reads for each transcript found in all three libraries, and gives an indication of the relative transcript number. (B–F) In situ hybridization of bmp7 in hindlimbs. Early on, expression is very restricted proximally (B), but later expression is more uniform with a concentration of transcripts in the distal mesenchyme and AER as well as posterior mesenchyme (C, D). Interdigital expression predominates at later stages (E, F). Limb buds are oriented with proximal to the left and posterior uppermost, and digits are marked with roman numerals.


Attenuation of BMP signaling results in defects in the distal limb skeleton

In our experiments, distal limb defects resulted from transient loss of BMP signaling between developmental stages 50 and 54. By stage 50, the hindlimb and forelimb bud are well established, consisting of a rounded bag of mesenchymal cells encased in a morphologically featureless epithelial epidermis. The signaling centers of the AEC and ZPA are already established, as can be revealed by in situ hybridization for the fgf8 and shh genes, respectively. Interfering with BMP signaling at earlier stages does not affect the formation of these two signaling centers (data not shown) and the limbs develop normally (Fig. 1). This suggests that BMPs are not important for the establishment of limb polarity in Xenopus. Despite this, attenuating BMP signaling generated specific phenotypes at stages 50–54. Our data suggest that maintenance of the ZPA and AER are dependent on BMP signaling at these stages, and that the patterning of the autopod, but not the more proximal limb skeleton, is consequently affected.

Ectopic limb formation can result from Noggin over-expression

Ectopic limbs were by far the most dramatic phenotype observed in the Noggin overexpressing limbs and were caused by activation at stages 50–51. A recent publication by Christen et al. (2012) using the same N1 transgenic line focused on the limb duplications, which they found to be induced at stage 49. It is not clear how this phenotype arises, as the animals have an established single limb bud at the time of noggin overexpression, but there are a number of possible causes of ectopic limbs, or polymelia.

Ectopic limbs can result from implantation of FGF protein-soaked beads into the flank of chicken embryos, between the existing fore- and hindlimb fields (Cohn et al. 1995), but the ectopic limbs seen in our studies and those of Christen et al. (2012) are truly polymelic, arising from the same trunk level as the regular limbs. Noggin-induced ectopic limbs also differ from the polydactylous, mirror image AP duplications that arise from transplantation of the ZPA to the anterior of the chicken limb (Tickle et al. 1975), or from ectopic expression of Shh in the anterior limb bud mesenchyme (Riddle et al. 1993). Here, the entire limb is reproduced, the limb pairs are positioned dorsal and ventral relative to one another, with the dorsal pair appearing hyperdorsalized. This suggests a bifurcation of the limb along the DV axis early on in limb development, but after formation of the limb buds. A very recent, more in-depth analysis by Christen et al. (2012) supports this model. Notably, the same authors showed that ectopic noggin could also induce extra pectoral fins in zebrafish, suggesting conservation of this mechanism at least among anamniotes. Furthermore, knockout of the BMP receptor Bmpr1a specifically in limb bud mesenchyme in mice resulted in a mild DV patterning defect (Ovchinnikov et al. 2006). There is currently no evidence for a ventrally restricted BMP ligand in the limb bud mesenchyme of any tetrapod, but here we have identified a larger than suspected cohort of potential noggin-sensitive BMP family members in the stage 51 limb transcriptome, and it is possible that bmp5, bmp3b or one of the other ligands we have identified acts to maintain the ventral compartment.

Hypodactyly resulting from attenuated BMP sig-naling can be explained by failure of Sox9-positive cartilage condensations

Hypodactyly, most often the formation of a single hindlimb digit with a claw, was a commonly observed phenotype when BMP signaling was transiently blocked at stage 50 or 51. In these limb buds, fgf8 expression in the AER was reduced, suggesting that maintenance of this signaling center is defective. However, in classic experiments performed in chickens, early loss of the equivalent AER, by ablation, results in the formation of a proximally truncated limb whereas later ablation of the AER causes more distal truncation (Saunders, 1948; Summerbell, 1974). Furthermore, loss of Fgf8 in mice leads to loss of proximal limb elements with distal ones developing almost normally, thanks to rescue by Fgf4, which is expressed from a later stage (Lewandoski et al. 2000; Moon & Capecchi, 2000). Therefore, early failure of the AER by itself cannot explain our hypodactylous phenotype.

In hypodactyly mice, Akiyama and colleagues have shown that sox9 expression can rescue normal limb development, suggesting that loss of Sox9 is the primary cause of hypodactyly (Akiyama et al. 2007). Sox9 is a target of BMP signaling: in BmpRIa−/− and BmpRIb−/− double mutant mice no Sox9 is expressed and no endochondral bones form (Yoon et al. 2005). BMP is needed for endochondral bone and signals via smads1 and 5 (Retting et al. 2009). Sox9, in turn, directly activates col2a1 expression in chondrocytes (Bell et al. 1997). Conditional mutants of sox9 (Akiyama et al. 2002) have shown that this gene is needed for all stages of cartilage formation: condensation, proliferation and maturation of chondrocytes. Our results suggest this is also true in Xenopus, with loss of BMP signaling leading to failure of sox9-positive cartilage condensations to form. The persistence of a single digit may result from the timing of the digit cartilage condensations: in Xenopus, digit IV develops first (Nieuwkoop & Faber, 1967) and may be able to escape the loss of Sox9, whereas the other digits are prevented from forming.

Knockout of shh in mice has revealed that limbs develop normally up to knee/elbow level, but there is loss of the zeugopod AP polarity and hypodactyly of the autopod (Chiang et al. 2001). A single digit is formed, which in forelimbs is composed of a single distal cartilage element, whereas in hindlimbs the single digit has well-formed phalangeal elements and joints, plus a nail. In both shh−/− mice and our noggin-expressing tadpoles, this single hindlimb digit has two phalanges, suggestive of a digit I identity. However, analysis of hoxd11-13 expression in shh−/− mice showed loss of late-phase hoxd11 expression, further supporting the digit I identity (Chiang et al. 2001) In Xenopus limbs treated with the shh inhibitor cyclopamine, the resulting single digits also showed a hoxd13 + hoxd11- phenotype, typical of digit I (Satoh et al. 2006). In contrast, in our BMP-inhibited tadpoles, both hoxd11 and hoxd13 were present in the single digit, indicating a more posterior identity (Fig. 5G,H). These results are puzzling, as a claw, normally restricted to digits I–III, always formed on the single digit. The hypodactylous limb caused by BMP attenuation in mid limb bud development therefore has a mixed identity, developing from the normal position of digit IV, with a digit IV hox code and an eventual digit I morphology. Interestingly in regenerating froglet forelimbs, in which shh signaling is not re-established, the single spike that forms is hoxd11-positive as well, indicating a posterior identity (Satoh et al. 2006).

Noggin beads placed interdigitally can cause anterior homeotic transformation of the digits in developing chicken hindfeet. It was initially proposed that Noggin acted via BMP inhibition (Dahn & Fallon, 2000), but conditional mouse knockouts of bmp2/4/7 showed no such transformations (Bandyopadhyay et al. 2006), suggesting that Noggin was inhibiting another ligand such as Gdf5 or 6.

Hypodactyly can also be caused by defective hox gene function. Hypodactyly (hd) mice have a 50-bp deletion in the hoxa13 gene, and a phenotype that is more severe than loss of hoxa13 and seems to act in a dominant negative fashion (Mortlock et al. 1996; Post et al. 2000). In these mice, only the distal limb is affected, and there is an increase in apoptosis and a corresponding reduction in chondrogenesis (Robertson et al. 1996). Hoxa13 and Hoxd13 both interact with the BMP signaling intermediate Smad5 (Williams et al. 2005), and can regulate transcription of Smad5 targets. In our experiments, hoxd13 expression across the autopod did not appear to be affected by the loss of BMP signaling (Fig. 5G).

Brachydactyly phenotypes result from late-stage inactivation of BMP signaling in the autopod

Brachydactyly, or shortening of the digits, can arise from failed growth of the autopod elements or from failed development of the distal elements (Temtamy & Aglan, 2008). When tadpoles were induced to overexpress Noggin in late limb development, the predominant phenotype was brachydactyly arising from failure to form the correct number of phalanges, with metacarpals and tarsals less affected. Brachydactyly is closely associated with altered BMP signaling in humans. Point mutations in the Noggin coding region result in brachydactyly type B (BDB), in which there is a deficiency of the distal elements of the fingers and toes (Lehmann et al. 2007). These Noggin mutations are not loss of function and are predicted to still bind Gdf5. Our results suggest that these mutations could result in a more stable or active version of Noggin protein. Another class of brachydactyly (BDA2) results from either mutation of gdf5 (Ploger et al. 2008), dominant negative mutation of the receptor BmpR1B (Lehmann et al. 2003) or duplication of a conserved regulatory element in the Bmp2 gene (Dathe et al. 2009). In fact, all documented examples of isolated brachydactyly can be linked directly or indirectly to altered BMP signal transduction (Temtamy & Aglan, 2008).

Concluding remarks

Using a heat-shock-activated transgenic line of X. laevis, the BMP inhibitor Noggin was overexpressed during different time periods of limb development. Remarkably, different phenotypes were obtained in cohorts treated at different stages. Early limb buds were unaffected, suggesting that BMPs are dispensable for limb initiation and the formation of the two main signaling centers, the AER and ZPA. At mid limb bud stages we observed either completely duplicated limbs or loss of AP pattern in the autopod resulting in ectrodactyly or hypodactyly, with the single digit having a confused identity. During later limb development, ectopic noggin caused brachydactyly resulting from failure to form distal elements of the autopod. Transcriptome analysis has indicated that a larger cohort of BMP family ligands are employed in limb bud development than previously thought, which could serve to explain the complex and varied phenotypes observed by noggin overexpression.


The authors would like to thank Amy Armstrong for frog colony care and maintenance. The work was supported by a University of Otago Research Grant to C. B.

Author contributions

C. B. wrote the manuscript. T. J. and C. B. performed experiments. R. D. performed bioinformatics analysis of the limb bud transcriptome.