Retinoic acid (RA) is a vitamin A (retinol) metabolite required for normal development. The availability of RA is regulated by production from retinol, catabolism, and binding (reviewed in Blomhoff and Blomhoff, 2006). Dietary retinol is first converted to retinal by alcohol dehydrogenases (ADH), and retinal is then further metabolised to RA by a group of retinal dehydrogenases (Raldh1-3, also known as aldehyde dehydrogenase 1a1-3 or Aldh1a1-3). During development, the rate-limiting step in RA synthesis is controlled by these Raldh enzymes, which exhibit tissue- and stage-specific expression. Binding to the developmentally regulated cellular retinoic acid–binding proteins (Crabp) 1 and 2 further refines the level of available RA, with Crabp1 acting to limit RA availability and Crabp2 enhancing the ability of RA to interact with DNA via its receptors (Boylan and Gudas, 1992; Budhu and Noy, 2002). RA is also actively degraded by members of the Cyp26 family of cytochrome p450 hydroxylase family, which are also developmentally regulated (Swindell et al., 1999).
RA is lipid soluble and is, therefore, able to diffuse into cells unhindered, where it can bind to RA receptors (RAR) and retinoid X receptors (RXR), which translocate directly to the nucleus as part of a heterodimer complex that interacts with genomic retinoic acid response elements (RAREs), activating target gene transcription (for a recent review, see Niederreither and Dolle, 2008). Importantly, direct measurement of RA has shown that regions of the developing mouse embryo that express Raldh2 are actively synthesising RA, whereas regions expressing Cyp26 are actively degrading it (Swindell et al., 1999). Therefore, gradients of RA can be envisaged, and the level of RA present in a particular cell at a particular time in development can regulate target gene expression accordingly (Reijntjes et al., 2005).
RA has the ability to act as a morphogen during development of the hindbrain (Guthrie, 1996), somites (Duester, 2007), and limbs (Lee et al., 2004), although whether it plays this role in vivo is still debated. Early experiments indicated the presence of a gradient of RA in the chick limb bud (Thaller and Eichele, 1987) and showed that RA applied exogenously could mimic the zone of polarising activity (ZPA), indicting a role in patterning the anterior to posterior (thumb to little finger) axis of the developing limb (Summerbell, 1983; Tickle et al., 1982). It is now known that ectopic RA does this by inducing shh expression in the ZPA, and that Shh is the anterior posterior morphogen (Niederreither et al., 2002; Noji et al., 1991; Wanek et al., 1991). RA has also been proposed to control proximal to distal patterning of the limb via direct regulation of the proximally expressed Meis homeodomain proteins, Meis1 and Meis2 (Mercader et al., 2000, 2005). Overexpression of these genes leads to truncation of the distal limb (Capdevila et al., 1999; Mercader et al., 1999). Targeted disruption of Meis1 results in death at around embryonic day 14.5 in mice (Hisa et al., 2004; Azcoitia et al., 2005). No limb phenotypes have been reported in these mice even though proximal-distal pattern is established and digits are apparent well before this stage (Martin, 1990). Compelling evidence for RA's involvement in the proximal distal patterning of the limb comes from examination of transgenic mice harbouring a reporter driven by retinoic acid response elements (RAREs), which show the existence of a gradient, and from cyp26b knockout mice, which have truncated limbs, a phenotype that can be reproduced with exposure to exogenous RA (Yashiro et al., 2004). Recently, however, experiments using RA supplementation in mice lacking both Raldh2 and Raldh3 have suggested an alternative role for RA in positioning the limbs (Zhao et al., 2009).
Amphibians demonstrate a rare ability to respond to amputation injuries by perfect replacement of the lost parts. While urodeles, such as the axolotl, can regenerate limbs as adults, anurans like Xenopus laevis can only do so for a limited time as tadpoles (Beck et al., 2009; Straube and Tanaka, 2006). In both cases, regeneration proceeds in a proximal to distal direction only, with faithful reproduction of the missing parts, indicating the presence of positional information along the proximal-distal (PD) axis. RA has been proposed as the morphogen responsible for this gradient of information during regeneration of the amphibian limb, and remains the only candidate. The influence of RA on amphibian limb regeneration was discovered in anurans in the late 1970s (Niazi and Saxena, 1978). RA was subsequently shown to have a concentration-dependent proximalising affect on the regenerating limb of axolotls, causing serial duplication of skeletal elements along the PD axis of the limb (Maden, 1982; Scadding and Maden, 1986a). The limbs were also ventralised and posteriorised, suggesting an affect on all three axes of asymmetry. The effects were dose-dependent to some degree with hypomorphic regenerates occurring at low and high doses of retinol palmitate and duplications at intermediate doses (Scadding and Maden, 1986a). PD and anterior-posterior (AP) duplications were also seen when the hindlimb of the anuran Xenopus laevis was treated with retinol palmitate in suspension following amputation at stages 51–53 (Scadding and Maden, 1986b). However, when the RA was applied locally using silastin implants, PD duplications were never observed (Scadding and Maden, 1986c). In both amphibians, duplications were not seen when developing limbs were treated by immersion or locally, suggesting that RA affects development and regeneration differently (Scadding and Maden, 1986a–c) and that a differential concentration of RA is needed in regenerating but not developing limbs to interpret PD and AP information. Furthermore, treatment of regenerating limbs with disulfuram or citral (inhibitors of RA production) can inhibit limb regeneration (Maden, 1998).
Despite this classical evidence that RA can act as a morphogen for limb development and regeneration, information is lacking on how RA gradients are formed and interpreted by the cells of the limb. Recently, we used a microarray approach as an unbiased way to identify genes that were differentially regulated during limb regeneration in Xenopus laevis (Pearl et al., 2008). Among the top candidates to which gene ontology could be assigned, we identified Crabp2 and Raldh2, both components of the RA pathway, suggesting that regulation of RA was strongly correlated with successful regeneration in anurans. Since the expression of these two genes had not been characterised in Xenopus limbs, and several other components of RA regulation (2 other Raldh genes, Cyp26 genes and Crabp1) were absent from the array, we determined to clone and study the regulation of these genes during development and regeneration of the limb to get a clear picture of whether RA could be acting as a morphogen. Our results strongly support the presence of dynamic RA gradients in development and regeneration of the Xenopus limb, and reveal a key difference between frog limbs, which make their own proximal RA, and chicken and mouse limbs, which do not. This difference could underpin the ability of frog limbs to regenerate at early stages.
Xenopus Limb Buds Have Intrinsic Proximal Raldh2 Activity During Early Limb Patterning
As in mice (Lee et al., 2004), the principal source of RA in developing Xenopus limbs is due to the activity of Raldh2. Raldh1 was not expressed in the limbs at any point throughout limb development (data not shown). Raldh2 expression was seen in the proximal hindlimb bud as early as Nieuwkoop and Faber stage (st.) 49 (Fig. 1A) and this staining was maintained until st. 54, but was absent from st. 55 limbs when all five digits are apparent in the autopod (Fig. 1B–G). This is in sharp contrast to mouse limb buds, which have no intrinsic Raldh activity and can only receive RA by diffusion from the adjacent body wall lateral plate mesoderm (Yashiro et al., 2004). Between st. 51 and 52, two new patches of cells began expressing Raldh2 in the anterior/proximal quadrant, on either side of the limb, within the future stylopod (Fig. 1D). These patches of expression were maintained throughout most of limb development but had declined by st. 55 (Fig. 1D–G). In the hindlimb, as digits began to emerge in the autopod at late st. 52, a third set of expression domains could be seen between the forming digits, initially between digits IV and III, III and II, and posterior of digit IV (Fig. 1D). As the hindlimb continued to develop, these domains of Raldh2 expression were seen between all five toes, with cells in the centre of the forming webs switching off Raldh2 so that expression flanks the forming cartilages rather than spanning the whole interdigital region (Fig. 1E–G). Zeugopod expression of Raldh2 first appeared in discrete patches of cells located posteriorly at st. 52 (Fig. 1D) and then anteriorly at st. 53 (Fig. 1E).
In forelimbs, Raldh2 could be detected in the proximal half of the mesenchyme concomitant with its appearance as a distinct anlagen at st. 50 (Fig. 1H). Expression of Raldh2 in other domains was slightly delayed relative to the hindlimbs, with anterior/proximal expression beginning at st. 53 (Fig. 1J) and interdigital expression clearly seen by st. 55 (Fig. 1K). This probably reflects the fact that hindlimbs develop slightly ahead of forelimbs morphologically, appearing approximately two developmental stages later. Expression in the forelimb zeugopod was not evident, but otherwise expression followed that of hindlimbs.
Raldh3 was not expressed until st. 52 when a small patch of anterior cells located about one third of the way down the hindlimb showed weak but specific expression (Fig. 2A). At st. 53, expression was detected in the autopod, in a small patch of cells anterior to the position where digit IV is forming and another patch further anterior, possibly corresponding to the future digit II (Fig. 2B). These autopodal patches were small, faint, and transient, positioned midway along digits V and III at st. 54 (Fig. 2D). In contrast, expression in the ventral muscles began at st. 53 (Fig. 2C) and remained strong, expanding to mark blocks of forming muscle on both dorsal and ventral sides by st. 54 to 55 (Fig. 2D, E).
Cyp26 Gene Expression During Limb Development Suggests Active Modulation of RA Levels
Cyp26a expression was absent from early limb buds, appearing transiently in the epithelium of the distal two thirds of the hindlimb at st. 53 as well as in a small patch of mesenchymal cells anterior to the position of the future digit IV (Fig. 3A,B). Once the autopod begins to form digits, Cyp26a expression was turned on in cells flanking the anterior side of each digit. At st. 54, Cyp26a expression was limited to cells anterior to digits IV and III, but by st. 55 it was expressed in all four interdigital regions (Fig. 3C,D). Two small stripes of expression also appeared at the proximal base of each digit with the exception of digit I. Finally, two small stripes of Cyp26a expression appeared in the anterior at st. 55.
Cyp26b expression appeared in late st. 51/early st. 52 hindlimb buds, and st. 52 forelimbs, in the mesenchyme corresponding to the future zeugopod (Fig. 4). Expression was strong in most mesenchymal cells of the distal half of the limb with the exception of the distal-most cells in hindlimbs and the posterior/distal-most cells in forelimbs (Fig. 4B,H). By st. 52, hindlimb expression was more diffuse across the proximal-distal (PD) axis and the mesenchyme immediately underlying the apical ectodermal ridge (AER) also began to express Cyp26b (Fig. 4C). At st. 53, expression in hindlimbs was strong in the forming digit IV and in mesenchyme underlying the AER from digit IV to digit I in the hindlimb (Fig. 2D). This pattern became clearer at st. 54, with digit IV strongly expressing Cyp26b, as well as cells immediately underlying the AER in a region extending from digit IV to I (between the black arrows in Fig. 4E). By st. 55, digits no longer showed clear Cyp26b expression and only the AER-flanking mesenchyme from digits III–I still retained Cyp26b (Fig. 4F). At st. 56, Cyp26b appears localised to the forming claws on digits III–I and to ventral expression in digits IV–I (Fig. 4G). In contrast, forelimb distal mesenchyme did not express Cyp26b and expression was also lacking in the digits from stage 53–55 (Fig. 4I–K). Later, digit and metacarpal expression could be seen clearly from the ventral side (Fig. 4L).
Crabp2 Expression in Xenopus Limbs Is Dynamic and Complex, and Differs Between Hind- and Forelimbs
We compared the expression of Crabp1 and −2 during limb development (Fig. 5). Crabp1 transcripts were not detected in limbs until skeletogenesis, when weak expression was seen in the region of the joints of digits I to III of the developing hindlimb (Fig. 5A). In contrast, Crabp2 expression was detected even in very early hindlimb buds, with strong expression in the distal mesenchyme ending at a sharp boundary midway along the length of the limb bud at st. 50 (Fig. 5B). The expression of Crabp2 was very dynamic and complex: by st. 51, transcripts had been cleared from the distal portion of the limb bud leaving a strongly expressing band of cells midway along the proximal distal axis (Fig. 5C), in the region of the future knee joint. At st. 52, Crabp2 expression was maintained at knee level, with additional expression in patches of cells in the proximal half of the limb (Fig. 5D). At St. 53, expression could still be detected at knee level, especially when viewed from the ventral side, with weaker expression in the forming autopod mesenchyme (Fig. 5E,F). This autopod expression presumably resolves into the interdigital expression, first seen at st. 54 (Fig. 5G), with transcripts absent from the interdigital webbing, but maintained on either side of the forming phalanges by st. 55 (Fig. 5H). Forelimb expression of Crabp2 differed markedly from that of hindlimbs, with early expression absent and the first detection of transcripts in the interdigital regions of the autopod at stage 54 (Fig. 5I,J).
Some RA Regulating Genes Are Up-Regulated in Limb Regeneration, But No New RA Is Synthesized
All seven markers for RA regulation were next examined for expression during regeneration of the hindlimb. Xenopus hindlimbs regenerate perfectly during early limb bud stages but this ability gradually declines over developmental time (Dent, 1962). We amputated st. 52 limbs at the level of the future knee and looked for expression of RA-related genes after 1, 3, or 5 days of regeneration (Fig. 6). Neither Cyp26a nor Crabp1 was expressed during regeneration over this time (data not shown). Cyp26b expression could not be detected in 1-day regenerates (Fig. 6A), but by 3 days was strongly up-regulated in mesenchymal cells of the blastema (Fig. 6B). After 5 days, the expression of Cyp26b was restricted to the central two thirds of the blastema and absent from the most proximal cells (Fig. 6C).
Interestingly, the binding protein Crabp2 was also strongly expressed in stump cells as early as 1 day after amputation (Fig. 6D–F). This corroborates data from our previous microarray screen for genes transcribed preferentially in 3-day post-amputation regenerating hindlimbs (Pearl et al., 2008). Crabp2 expression appears earlier than that of Cyp26b. However, this early expression is also seen in non-regenerating (heat-shocked noggin overexpressing N1 transgenic line) limbs of the same stage (Fig. 6D′–F′) and may in fact be residual expression rather than de novo, regeneration-specific expression. The pattern of Crabp2 expression in 1-day regenerates suggests residual developmental expression is in fact responsible (Fig. 6D). By 3 days, however, the expression of Crabp2 differed between regenerating and non-regenerating limbs, and was seen strongly in the distal mesenchymal blastema cells with the exception of a small patch at the extreme distal tip of the growing blastema (Fig. 6E,E′). By 5 days, the pattern of Crabp2 expression in regenerating limb buds was almost reciprocal to that of Cyp26b, with Crabp2 transcripts localised to the proximal blastema and extending only into posterior and anterior cells of the mid-blastema (Fig. 6F). Mammalian Crabp2 is known to have a RARE in its upstream regulatory region and to be inducible by RA (Astrom et al., 1994), and could be similarly regulated in Xenopus. In contrast, non-regenerating N1 limbs showed low initial expression of Crabp2 in the pseudoblastema at 3 days post-amputation (Fig. 6E′), but this had disappeared by 5 days (Fig. 6F′). Furthermore, the expression of Crabp2 in non-regenerating 3-day pseudoblastemas was more ubiquitous, with no zone of non-expression at the distal tip (compare Fig. 6E and E′).
There was no de novo expression of Raldh1-3 in regenerating hindlimb buds (Fig. 6G–I). However, as previously shown (Fig. 1), Raldh2 is expressed proximally in a patch of cells bordering the body wall, and in two patches of anterior/proximal cells. All these Raldh2-expressing domains are retained after amputation, suggesting that a local source of RA would be available to blastemal cells. If RA diffusion is unimpeded, then a gradient would be present in the stump tissue, with lowest levels in the posterior/distal cells, which may act to inform cells of their position.
In order to access how general the need for RA modulation during regeneration might be, we examined expression in older limbs and in tadpole tails. Crabp2, but not Raldh2, was also upregulated in st. 56 limbs following amputation at the knee level (see Supp. Fig. S1, which is available online). St. 56 limbs do not normally regenerate any obvious structures although a small blastema still forms early on. Crabp2 is expressed in the distal mesenchyme, in the same way as in regeneration competent limb stages. However, the absence of Crabp2 from the distal tip does not occur after amputation of the later, non-competent, limbs. Crabp2, then, clearly does not drive regenerative success, but its distribution within the blastema may be important for subsequent re-patterning. Xenopus laevis tadpoles can also regenerate their tails, except for a refractory period between st. 46 and 47 (Beck et al., 2003). Crabp2 and Raldh2 expression was also up-regulated during tail regeneration (Supp. Fig. S2) in the regeneration bud, specifically in the neural ampulla, bullet-like notochord cells, and mesenchyme. This differs from the situation in limb, where Raldh genes are not up-regulated, and any RA required for repatterning the limb must come from the stump tissues. Neither gene was expressed following amputation of refractory st. 47 tails or the tails of transgenic N1 animal when the noggin transgene was activated by heat shock, nor was expression induced by wounding the tail (Supp. Fig. S3). Our results indicate that tail regeneration requires de novo Raldh2 synthesis in order to produce RA within the blastema. This suggests a fundamental difference in the requirements for RA gradients in tail and limb regeneration in anurans.
A Proximal-Distal RA Gradient Could Be Present in the Early Developing Xenopus Limb
Our results in Xenopus show that RA synthesis by Raldh2 and catabolism by Cyp26b could be acting to generate a differential of RA across the developing limbs in Xenopus. As fore- and hindlimbs show the same initial pattern (although forelimbs lag behind by one or two stages and there are one or two differences, most notably a lack of early Crabp2 expression), the more familiar hindlimbs will be discussed here, and expression data for hindlimbs are summarised in Figure 7. Initially, up to st. 51, RA produced by the proximal Raldh2-expressing cells would be predicted to diffuse unimpeded across the whole limb bud, since there is no expression of Cyp26 genes or Crabp1 at this stage. At this point in development, cells of the future stylopod (upper limb) would be exposed to the highest levels of RA, with intermediate and low levels in the future zeugopod (lower limb) and autopod (foot), respectively. Only the PD axis is likely to be patterned by RA at this stage (Fig. 7A).
At st. 51.5, before the autopod becomes distinct, marked by the appearance of the future ankle joint, Cyp26b expression is seen for the first time in the distal third of the Xenopus hindlimb, creating a sink for RA. Cyp26b expression would be expected to reinforce and stabilise the gradient of RA from proximal to distal (Fig. 7B). Cyp26b is also expressed distally in chick and mouse limb buds, where it is thought to remove RA produced by Raldh2-expressing cells in the body wall lateral plate (reviewed in Niederreither and Dolle, 2008). Modulation of RA levels by Cyp26b is essential for normally developing limbs in amniotes, as shown by the phenotype of the Cyp26b knockout mouse (Yashiro et al., 2004). Cyp26b−/− mice have meromelic limbs, with 2–3 digits complete with nails but more proximal elements shortened and fused. Importantly, this phenotype is replicated by dosing mothers with RA during limb outgrowth. In Xenopus, Cyp26b may be required to delay development of the autopod, which begins to express Raldh2 at low levels at this stage.
Scadding and Maden have shown that developing Xenopus hindlimbs are more sensitive to exogenous RA from st 50–53 and forelimbs, which develop slightly behind, from 51–53 (Scadding and Maden, 1986b). Earlier treatment with RA tends to result in defects in the stylopod and zeugopod (phocomelia) whereas later treatments alter the patterning of the autopod, reducing the number of digits (oligodactyly) as well as the number of skeletal elements per digit (brachydactyly). From our gene expression data, we predict that in limb buds up to and including st. 50, RA will be produced in proximal mesenchymal cells and diffuse freely, creating a short-lived gradient across the PD axis. Supplying limbs with extra RA during this time may disrupt this gradient, leading to phocomelia. We note, however, that Scadding and Maden's data suggest that this only occurs with prolonged RA treatment of 7 days or more, which would be expected to extend into st. 51, suggesting that st. 50 limb buds may be refractory to RA. In this case, one could suggest that Raldh2 in the proximal third of the early limb bud is required only to switch on expression of Cyp26b in the distal cells at st. 51.5 in hindlimbs and st. 52 in forelimbs. Previous studies in chickens have shown that Cyp26b expression is upregulated in the presence of ectopic RA (Reijntjes et al., 2005). Therefore, it would be interesting to determine if Cyp26b expression is delayed or altered when early Xenopus limb buds are treated with RA.
Once Cyp26b expression begins in the distal half of the limb bud, any exogenous RA would be broken down and metabolised, presumably up to a point where the enzyme becomes overloaded. If Cyp26b is a target of RA, this may not occur until high doses are reached. Cyp26b is expressed distally in the presumptive autopod of Xenopus fore- and hindlimbs although its expression seems to be absent in the most distal-posterior cells. Since the protective affect of Cyp26b is cell autonomous, these distal-posterior cells would not be protected from exogenous RA and may be sensitive to disruption, leading to abnormal patterning or loss of digits V and IV and possibly III. Similarly, Cyp26b is not expressed in proximal cells at this stage and, therefore, RA could potentially disrupt the formation of the upper limb, leading to the observed phocomelia.
Binding proteins may also affect the function of the RA gradient. Crabp2 would be expected to assist in the transport of RA to the nucleus, and while it is dispensable for development in mice, the forelimbs of mutants do form small ectopic phalanges (postaxial polydactyly) (Fawcett et al., 1995). Crabp protein has been shown to be localised to the progress zone of developing chick limbs as well as being present in a gradient from anterior to posterior (Maden et al., 1989). At the time of this study, only one Crabp was known, our data suggests that this protein was likely to be Crabp2. In Xenopus early limb buds (<st. 51), Crabp2 is expressed across the distal half of early limb buds with the exception of the distal-most cells, probably marking the zeugopod (Fig. 7A). At st. 52 Crabp2 is excluded from the autopod (Fig. 5B) and proximal expression is very dynamic, resolving to the knee level at st. 53 where it may help to make the zeugopod and autopod distinct by enabling RA to activate transcription in the distal zeugopod. We do not observe an anterior to posterior gradient of Crabp2 in Xenopus limb buds by in situ hybridisation, suggesting that there is either a species difference or that protein and mRNA are distributed differently.
RA May Pattern the Anterior-Posterior Axis of the Zeugopod and Autopod
While the RA gradient set up by proximal Raldh2 and distal Cyp26b early in limb development may direct PD pattern of the limb, later expression of Raldh2 from st. 52 in the posterior zeugopod and Cyp26b in the anterior zeugopod is predicted to produce an evolving gradient with the highest levels of RA in proximal/posterior cells and the lowest levels in distal/anterior cells (Fig. 7C). We suggest that there may be a posterior to anterior gradient because of the presence of a posterior source of RA generated by Raldh2 and an opposing anterior sink comprised of Cyp26b-expressing cells. Posterior expression of Shh occurs much earlier than this (Christen and Slack, 1998 and our own observations) and so the posterior Raldh2 expression may reflect earlier Shh activity in this region. In the zeugopod region, therefore, RA could be directing both AP and PD pattern. Also beginning at st. 52, Raldh2 expression is seen for the first time in autopod cells, marking the future digit margins. Cyp26b is also present except at the very distal tip, and the posterior ZPA, suggesting that some AP pattern is also directed by RA in the autopod (Fig. 7C).
The role of RA in AP patterning of limbs in other species is also still unclear. In zebrafish, RA produced by Raldh2 in the somites is required for the establishment of the zebrafish pectoral fin field (Gibert et al., 2006). Raldh2 is expressed in posterior mesenchyme of developing pectoral fin and directs anterior posterior fates within the fin bud. The no-fin mutant, which lacks pectoral fins, is due to loss of functional Raldh2 (Grandel et al., 2002). While Raldh2 is not expressed in the mouse limb at all E9.5–E11 (Yashiro et al., 2004), mice lacking Raldh2 have no forelimb buds (Niederreither et al., 2002).
Complex Regulation of RA in Later Limb Development Sculpts the Autopod and Determines Claw Formation
The presence of Raldh2 and Crabp2 in interdigital regions suggests that RA is being both synthesised and transported to the nucleus. A likely role of RA is to limit the size of cartilage condensations within the autopod. Both Crabp2 and Raldh2 are also expressed interdigitally in mouse autopods (Niederreither et al., 1997; Ruberte et al., 1992), and syndactyly (fusion of the digits, extra phalanges) occurs in Crabp2 knockout mice (Lampron et al., 1995) suggesting that the normal function of RA at this stage is to keep developing digits separated. Raldh2 expression, and therefore RA synthesis, is restricted to interdigital regions in mice by the direct action of prochondrogenic BMPs, which attenuate Raldh2 expression (Hoffman et al., 2006).
In Xenopus, another RA synthesis gene, Raldh3, is transiently expressed in developing interphalageal joints, appearing first in digit IV, the first digit to form. By st. 54, the expression of Raldh3 is attenuated, suggesting that RA may play a role in limiting the formation of interphalangeal joints until digit growth reaches a certain point. Alternatively, Raldh3 could be acting to generate positional information locally. Expression of the inhibitory binding protein Crabp1 appears in the joints as Raldh3 is declining. It is tempting to speculate that Crabp1 acts to scavenge any RA from these future joint regions, allowing cartilage formation to proceed unimpeded, although Crabp1 knockout mice are described as entirely normal (Gorry et al., 1994), indicating possible redundancy or protection from teratogens.
The RA catabolic enzyme Cyp26b is expressed in cells immediately underlying the AER of Xenopus limbs. We observe a clear anterior bias to this expression from st. 53/54 in hindlimbs. The anteriormost digits I, II, and III retain expression of Cyp26b in this region even as the digits form, resolving over time to expression in cells corresponding to the future claws. Cyp26b is absent from this region in digits IV and V of the hindlimb and all digits of the forelimb, suggesting that claw formation in digits I–III is directed by Cyp26b in Xenopus. These claws are thought to be an innovation of amphibia (Maddin et al., 2009), suggesting co-option of Cyp26b to this role.
Cyp26b expression is also seen in a dynamic pattern corresponding to forming skeletal elements of the autopod. Cyp26b has a known role in the axial skeleton from loss of function studies in other vertebrates. The zebrafish homologue of cyp26b is expressed in condensing chondrocytes and osteoblasts, and cyp26b1 mutant fish (stocksteif) have reduced midline cartilage and hyperossification leading to vertebral fusion (Spoorendonk et al., 2008). Mice treated with R115866, an inhibitor of Cyp26 activity, have fused cervical vertebrae (Laue et al., 2008). Cyp26a is also seen in small anterior and posterior stripes flanking the position of the forming proximal phalanx in digits II–V. Cyp26a mutant mice exhibit sirenomelia and axial skeletal defects, but forelimbs develop normally (Abu-Abed et al., 2001; Sakai et al., 2001).
The main site of Cyp26a expression was found to be in the interdigital mesenchymal cells of hindlimbs. Corresponding expression was also noted in forelimbs, indicating active catabolism of RA in these regions. Interdigital expression of Cyp26a has not been reported in amniotes. Interdigital tissue is selectively removed by apoptosis in chickens, a process that is regulated by BMP7, which in turn is regulated by RA (Merino et al., 1999). Unlike amniotes, the interdigital region in Xenopus limbs does not undergo apoptosis (Cameron and Fallon, 1977). The presence of Cyp26a in the interdigital regions of tadpole hind and forelimbs may explain this difference.
A PD gradient of RA is likely to be re-established during appendage regeneration, and de novo RA synthesis occurs during tail but not limb regeneration.
In a recent microarray screen to identify genes that are differentially regulated during hindlimb regeneration, transcript levels of Raldh2, Adh1, and Crabp2 were found to be significantly different between regenerating and noggin-inhibited (non-regenerating transgenic N1 line) blastemas, 3 days after amputation at st. 52. Raldh2 and Adh1 were up-regulated in the non-regenerating limbs, whereas Crabp2 was up-regulated in regenerating limbs (Pearl et al., 2008). Coding sequence was amplified for all three genes using the sequence data used to design the probesets (Affymetrix, Santa Clara, CA). However, specific expression of Adh1 could not be reproduced by in situ hybridisation. Raldh2 was also not up-regulated during regeneration (Fig. 6G–I), and we explain the anomalous array result as due to the inadvertent capture of more proximal cells, expressing Raldh2, when harvesting pseudoblastemas as opposed to blastemas. Raldh1 and 3 were also absent from regenerates suggesting that no new RA is being actively made as a result of the process of limb regeneration.
Crabp2, however, was up-regulated in blastemal cells with expression absent from the distal tip. Crabp2 overexpression in cell culture makes the cells more sensitive to RA (Budhu and Noy, 2002), suggesting that any RA that reaches proximal cells from the stump may be bound by Crabp2 and transported to the nucleus. The expression pattern of Crabp2 suggests a role in determining proximal identity within the blastema, consistent with the results of classical experiments using RA to induce PD duplications. Also concordant with our results in Xenopus, an early study showed that Crabp protein increased during limb regeneration in the axolotl (Maden et al., 1989).
Further evidence for a gradient of RA in the blastema comes from the analysis of RA inhibitor expression. While Crabp1 and Cyp26a are absent from regenerates, Cyp26b is both strongly upregulated and exhibits graded expression within the blastema, reciprocating its embryonic expression in the early limb bud. Although the regenerating blastema is unable to synthesise RA, the upregulation of Cyp26b in the central blastemal cells (Figs. 6A–C, 8C) suggests that most if not all RA diffusing into the blastema from more proximal sites is actively destroyed. It is possible that cells in the proximal part of the blastema can “read” their position from the level of RA, and develop accordingly. More proximal cuts would be expected to cause exposure of these cells to higher levels of RA, as they are closer to the sources of RA in the stump. Perhaps the PD duplications arising from treatment of Xenopus hindlimbs with exogenous RA expose these cells to higher than normal levels, proximalising the cells at the margin (Scadding and Maden, 1986b). Cyp26b is not a known direct target of RA, but a RARE has been found in the Cyp26a promoter region (Loudig et al., 2000).
We propose a model for RA regulation during regeneration in which RA from the proximal stump cells diffuses freely following amputation of the distal limb bud. The presence of Crabp2 in the distal stump may assist in transport of RA to the nucleus at low levels, enabling sensing of RA remote from the source, and possibly acting to impart positional information to the cells that will form the blastema (Fig. 8). Cyp26b expression appears with the establishment of the blastema and is co-expressed with newly synthesised Crabp2 except at the distal tip, in mesenchymal cells immediately adjacent to the apical epithelial cap, the equivalent of the AER in regenerating limbs. We predict that no RA signalling will occur in the distal-most cells under the AER, but that low levels of RA escaping Cyp26b, or active metabolites, may regulate transcription with the assistance of Crabp2 in the rest of the blastema. Since there is a proximal source and a distal sink, an RA gradient seems likely across this axis. However, the central location of Cyp26b and the absence of Crabp2 suggests an almost concentric gradient, where levels of RA activity are lowest or absent in the central distal cells, under the apical epithelial cap, and highest in the anterior, posterior, and proximal cells (Fig. 8B, C). This differs from the early, simple PD gradient predicted to occur in very early limb buds during development and may suggest that the regenerating autopod is formed from the central distal cells directly under the apical epithelial cap, with more proximal elements arising from more peripheral cells.
While limb blastemas receive RA patterning information from the stump cells, tail blastemas may actively synthesise RA, since Raldh2 is strongly upregulated in regenerating, but not refractory stage, tails. This suggests a notable difference in the way that limb and tail blastemas detect their position within the appendage and, therefore, the fate of their cells. It is not clear how the tail blastema RA could generate or respond to positional information, but the repeated module nature of the tail means that positional errors are less critical or noticeable.
We have documented the expression patterns of several components of the RA signalling pathway during Xenopus limb development and regeneration. In the developing limb, the corresponding expression patterns suggest the existence of proximal to distal gradient of RA in early limb development. The binding protein Crabp2 may stabilise this early limb gradient in hindlimbs, although expression is absent from the smaller, later developing forelimbs at equivalent stages. Early Xenopus hindlimbs regenerate well, and although no new RA appears to be produced following amputation, we present evidence that residual proximal Raldh2 and opposing Cyp26b induced in the blastema could re-establish a PD RA gradient in the regenerating limb. We suggest that cells of the stump may read their position relative to the RA source in order to regenerate according to the appropriate level along the PD limb axis. Distally, we show evidence to suggest that the central portion of the blastema is likely to have the lowest RA levels and that these cells may represent the regenerating autopod, with more proximal elements arising from peripheral blastema cells. We find little or no evidence for RA patterning across the AP axis of the autopod, during either development or regeneration. Instead, we propose the existence of an AP gradient in the developing zeugopod and suggest that the role of RA in the autopod is to enable the formation of the digits. Fate mapping and reporter experiments will be required to confirm these predictions.
The N1 stable line of transgenic Xenopus has been previously described (Beck et al., 2003). Briefly, the animals contain a transgene comprised of two linked parts, the first containing X. laevis Noggin coding sequence under the control of the Hsp70 promoter, and the second the green fluorescent protein (GFP) coding sequence under the control of the lens specific promoter γ-crystallin. The line is derived from a single insertion founder made by sperm nuclear injection using the method of Kroll and Amaya (1996) modified as in Beck et al. (2003). The Noggin transgene is activated by heat shock as described below, and Noggin acts to block BMP signalling, which is essential for regeneration in frogs.
Amputation and Heat Shock
Culture of Xenopus laevis tadpoles has been described previously (Barker and Beck, 2009). Animals were staged according to Nieuwkoop and Faber, 1967. Briefly, stage (st.) 48 marks the emergence of hindlimbs, st. 50 the forelimbs, st. 52 the appearance of the ankle joint, and st. 53–54 the paddle stage. Hindlimb amputations at the level of the future knee were carried out at st. 52 under anaesthetic (1/4,000 w/v MS222, Sigma, St. Louis, MO), generally unilaterally on the right-hand side. Tail partial amputations involved the removal of one third of the tail at stage 50 under anaesthetic. Where applicable, transgene activation by means of heat shock was initiated by immersing tadpoles in water at 34°C for 30 min, followed by return to normal aquarium temperature (25°C).
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. The Committee approved all experiments under protocol AEC78/06.
In Situ Hybridisation
Note that limbs are presented with posterior uppermost as this results from the orientation of Xenopus tadpoles with anterior to the left and dorsal uppermost as per convention. This results in an apparent inversion of the AP axis during developing limb stages; shh and the ZPA are therefore localised on the upper side of the limb. Probes were synthesised as described in Lynch et al. (2010). Whole-mount in situ hybridisation of embryos and tadpoles was performed as previously described (Beck and Slack, 1998), with the following modifications. Heads, tails, and the viscera of tadpoles were removed post-fixation and before hybridisation, to reduce the target area for probes. Tadpoles at different developmental stages were treated with 10 μg/ml proteinase K as follows, with the time of treatment (min) in parentheses: st. 51 (15), st. 52 (20), st. 53 (25), st. 54 (30), blastemas (5), or with 20 or 40 μg/ml proteinase K for 30 min (st. 55–56 and st. 57, respectively).
The authors thank Amy Armstrong for frog colony care and past lab members Esther Pearl, Robert Trought, Donna Barker, Sarah Holman, and Tamsin Jones for advice and support. C.B. also thanks the three anonymous reviewers of this article for their expert, helpful, and constructive comments. This work was funded by a University of Otago Research Grant (2008) to C.B.