Urodeles, such as newts and axolotls, have been intensively studied for their ability to regenerate missing body parts such as jaws, limbs, and tails, perfectly into adult life. Although higher vertebrates have lost that regenerative capacity completely, another amphibian, the African clawed frog Xenopus laevis, has retained some aspects of regeneration. Xenopus is able to regenerate limbs perfectly during larval stages but loses that ability with the progression of differentiation. Up to stage 53, when the footplate becomes visible, limbs regenerate perfectly after a proximal amputation. Later in development, the regeneration ability decreases gradually and the regenerates are imperfect with toes and lower limb parts missing. At stage 57, when ossification is almost complete, limbs have lost their regenerating ability entirely and resections heal like a wound, forming a stump (Dent, 1962). Xenopus tadpoles are also able to regenerate their tail before metamorphosis.
Studies in urodeles show that regeneration takes place in a series of conserved processes. In the first phase, the wound gets covered with a wound epithelium and the mature tissue starts to dedifferentiate and form a population of mitotically active cells. These undifferentiated and proliferating cells form an early bud blastema at the centre of the amputation plane (Gardiner et al., 1986; Ferretti, 2001; Brockes and Kumar, 2002). In the second phase, the blastema expands and grows to mid- and late-bud stages and becomes patterned through a process that is now considered to involve the same pathways as used during development (Muneoka and Bryant, 1982; Christen and Slack, 1998; Gardiner et al., 1999; Smith et al., 2000). In the third phase, the regenerate differentiates and forms all missing limb parts.
Hox genes are good candidates for playing a role in providing both the competence and the positional information during regeneration. The Abdominal-B-type Hox genes in particular have been implicated in playing that role in regeneration and have consequently been extensively studied during newt and axolotl limb and tail regeneration (Simon and Tabin, 1993; Beauchemin et al., 1994; Gardiner et al., 1995; Gardiner and Bryant, 1996; Torok et al., 1998; Khan et al., 1999; Carlson et al., 2001). To start to unravel the molecular differences that underlie the differences in the ability to regenerate missing body parts, we looked at temporal and spatial differences in expression of three of the Abdominal-B-type Hox genes, XHoxc10, XHoxa13, and XHoxd13. We describe the expression pattern of these three posterior Hox genes in developing limbs and compare them with their expression during hindlimb and tail regeneration in Xenopus laevis tadpoles.
Search for Posterior Hox Genes
Because of the lack of Abdominal B-type Xenopus Hox genes in the databases, we embarked on an extensive screen of expressed genes by using reverse transcription-polymerase chain reaction (RT-PCR). Six AbdB-type genes were found, members of all four paralog groups (Lombardo and Slack, 2001). Among them was XHoxa13, whose expression during normal development has been studied extensively (Endo et al., 2000; Lombardo and Slack, 2001). We have now obtained the complete coding region of XHoxa13 by 5′- and 3′-RACE. Over the whole of the protein sequence, XHOXA 13 is 73% conserved between mouse and human, whereas the homeodomain alone is 98% conserved (Fig. 1B). More recently, we have found two more posterior Hox genes, XHoxc10 and XHoxd13, on the NCBI database whose expression pattern have not been published previously. The full-length XHoxc10 was successfully cloned by RT-PCR with the 5′ primer designed against the published partial sequence (Grammer et al., 2000; accession no. BG354667) and the C-terminal primer against the XHoxc10 image clone (accession no. BG552812), which contains part of the homeobox and a long 3′ untranslated region (UTR). XHOXC10 is 74% and 68% homologous to axolotl and mouse/human, respectively. The homeodomain on its own, however, shows 100% and 98% conservation for axolotl or human/mouse, respectively (Fig. 1A). XHOXD13 (accession no. BE576752) is a full-length clone that shows 76% homology to the human and mouse homologs over the entire protein and 96% conservation within the homeodomain (Fig. 1C).
Expression in Embryos and Developing Limbs
The expression patterns in embryos and developing hindlimbs were studied by whole-mount in situ hybridisation. The patterns are very dynamic during tail bud stages but, in general, very similar to those in other vertebrates. XHoxc10 comes on first and is detected in stage 25 embryos in the central nervous system (CNS) and the mesoderm. At later tail bud stages, XHoxc10 expression can be seen from somite 8/9 to the tail bud tip in the CNS and the paraxial mesoderm (Fig. 2A). XHoxa13 and XHoxd13 are detected later in embryogenesis at around stage 27 and are restricted more posteriorly. Lombardo and Slack (2001) mapped the anterior boundary of XHoxa13 to somite level 14 in the CNS at stage 32, whereas the mesoderm expression is confined to the posterior wall of the tail bud. XHoxd13 has a wider expression domain in the CNS corresponding to somite level 13/14, whereas its mesoderm expression is also restricted to the posterior wall (Fig. 2E,I).
In the hindlimb, XHoxc10 is detected at stage 50 in the anterior middle section of the limb bud mesenchyme. The size and shape of the XHoxc10 expression domain does not change significantly during limb development. It stays proximal and anterior up to stage 55 (Fig. 2B–D). This expression is quite different from the axolotl, for which Hoxc10 is expressed in a much wider, distal domain and only shows a slight tendency toward a posterior to anterior gradient (Carlson et al., 2001). The expression pattern of XHoxa13 has already been published, and we are including it here only for comparison (Endo et al., 2000; Lombardo and Slack, 2001). XHoxa13 is expressed, as in other vertebrate limbs, most strongly in the distal mesenchyme of stage 50 and 51 limb buds (Fig. 2F) and is later detected across almost the entire footplate (Fig. 2G). It remains expressed in the tips of the developing toes at stage 55 (Fig. 2H). XHoxd13 shows a similar expression pattern as XHoxa13 in young limb buds, being distally restricted without an obvious anteroposterior gradient as reported in mouse and chick (Dollé et al., 1989; Yokouchi et al., 1991; Nelson et al., 1996). In later stages, XHoxd13 is still expressed across the whole footplate; however, the posterior two toes express XHoxd13 at a higher level (Fig. 2L), as described by Nelson et al. (1996) in later chick limbs.
Expression in Regenerating and Nonregenerating Limbs
We looked at temporal and spatial control of the expression pattern of these three Hox genes in regenerating stage 54 limbs. At this stage, all limbs have the ability to regenerate; however, regeneration is imperfect. Within 24 hr of resection, XHoxc10 is expressed very strongly in the stump mesenchyme just proximal to the amputation plane (Fig. 3A). As the blastema forms, XHoxc10 expression is confined to the blastema at early and medium bud stages (Fig. 3B,C). Note that this expression pattern is different from the pattern seen in developing limb buds. To control for specificity in regeneration, we also looked at nonregenerating stage 57 limb stumps (data not shown). No XHoxc10 expression is detected in these older, nonregenerating limbs, suggesting that XHoxc10 up-regulation is specific to regeneration. Similarly, XHoxa13 is detected distal to the plane of amputation in half of the stage 54 limbs after 24 hr and is more strongly expressed in all the blastemas by the early bud stage (Fig. 3E,F). XHoxa13 is detected in the mesenchyme of almost the entire blastema and highly up-regulated by mid-bud stage (Fig. 3G). We do not detect XHoxa13 expression in nonregenerating stage 57 limbs; hence, XHoxa13 is regeneration-specific as well (Fig. 3H). In contrast, XHoxd13 is not detected by in situ hybridisation until mid-bud stage when patterning of the blastema is under way and is then confined to the posterior distal part of the blastema (Fig. 3I–K). This asymmetric expression differs from the developmental pattern, which is uniform in the distal tip of the young limb bud. In addition, XHoxd13 is regeneration-specific, as it is not detected in nonregenerating limbs (Fig. 3L).
Expression in Regenerating and Wounded Tails
A similar picture is apparent for their expression during tail regeneration, with some variation in timing. All three genes are expressed in regenerating tails. Both XHoxc10 and XHoxd13 are expressed in the tip of the neural tube 1 day after amputation (Fig. 4A,G). At day 3 after amputation, XHoxc10 is located in the dorsal blastema, at the tip of the regenerating tail (Fig. 4B), whereas XHoxd13 expression domain is wider and covers the whole regenerating tail with strongest expression in the distal tip (Fig. 4H). Strong expression of XHoxa13 is seen in the proximal part of the blastema after 1 day (Fig. 4D) and in a graded pattern extending from the region just behind the growing tip of the regenerating tail to the site of amputation after 3 days (Fig. 4E).
Wounding vs. Regeneration
Previously, it has been reported that some of the Abdominal-B-type Hox genes also play a role in wound healing in nonregenerating systems (Stelnicki et al., 1998; Torok et al., 1998; Uyeno et al., 2001). In Xenopus, the initial response to close a skin wound or a wound caused by amputation at regenerative stages is expected to be the same, because no scar is formed in either process. Therefore, we looked for expression of these three Hox genes in limbs and in tails 1 day after wounding. Despite all three genes being re-expressed in 1-day tail blastemas, we did not detect any up-regulation of either Hox gene in wounded tails (Fig. 4C,F,I). This finding is consistent with the lack of expression of these genes in wounded or nonregenerating limbs (Fig. 3D,H,L). Hence, none of the three Hox genes under study appear to be involved in wound healing.
Posterior Hox Genes in Xenopus Limb Development
We describe the expression patterns of three posterior Hox genes, XHoxc10, XHoxa13, and XHoxd13. Little differences exist between the expression patterns of the latter two genes and published data for other vertebrates, including axolotl (Gardiner et al., 1995; Torok et al., 1998). In contrast, the XHoxc10 pattern diverges somewhat from the one seen in developing axolotl limbs but agrees with the pattern seen in the chick (Nelson et al., 1996). To date, no one has attributed a particular role to Hoxc10 in a functional assay; therefore, it is unclear whether this difference in expression has any significance for later regeneration capacity.
Development vs. Regeneration
Over the years, it has become clear that regeneration and development have a lot in common. Later regeneration events seem to recapitulate development by using the same genetic pathways that have been used to establish a limb or tail initially. Important genes for limb or tail development are reexpressed in regenerating appendices with a strikingly similar expression pattern (Endo et al., 1997, 2000; Christen and Slack, 1998; Torok et al., 1998; Yokoyama et al., 2001). Therefore, it is legitimate to compare different blastema stages with developing limb or tail buds. We propose that a 1-day and a 3-day tail blastema correspond to a stage 30 and stage 32/33 tail bud, respectively. For the limb, a mid-bud blastema is similar to a stage 51 limb bud in appearance and gene expression.
Regeneration-Specific Expression Patterns
Hox genes have long been identified as universal gene products that code for positional information of the primary and secondary axes of the developing vertebrate embryo. This finding makes them good candidates to encode positional information in regenerating systems as well. For this reason, they have been studied in regenerative systems from hydra to axolotl and newts (Shenk et al., 1993; Simon and Tabin, 1993; Torok et al., 1998; Gauchat et al., 2000; Carlson et al., 2001). In axolotl and newts, it has been shown that some of the Hox genes are expressed in the mature adult fore- or hindlimb (Simon and Tabin, 1993; Beauchemin et al., 1994; Savard and Tremblay, 1995). None of these studies, however, addressed the spatial distribution of the Hox genes in the adult limb. We have previously published spatial expression patterns of some of the posterior Hox genes during the stages when Xenopus is able to regenerate its limbs, consistent with the idea that Hox genes are able to provide the positional information needed after a resection at knee level (Lombardo and Slack, 2001).
In this study, we have found some differences in expression in regenerating limb and tail blastemas between Xenopus and axolotl. For Hoxc10 and Hoxa13 where direct comparison is possible (Gardiner et al., 1995; Endo et al., 2000; Carlson et al., 2001), we find differences in the first detectable XHoxc10 expression. In Xenopus, XHoxc10 comes on 24 hr after amputation in the distal limb stump, just proximal to the amputation plane, whereas Hoxc10 in axolotl is not detected until 5 or 6 days after amputation and then accumulates at the amputation plane in a very thin layer. Early and mid-bud expression look identical between the two species, and no differences are seen in mid-bud tail blastema expression. However, although Hoxc10 expression in axolotl seems to recapitulate normal development, the pattern in Xenopus is novel to regeneration and not seen in developing limb buds. In the developing limb bud, XHoxc10 expression is restricted to anterior and then anterior/proximal very early on. In contrast, in regenerating limbs, XHoxc10 expression is seen in the region where cells dedifferentiate and undifferentiated cells are recruited and, later on in the whole blastema, with no obvious anterior restriction. Note, however, that this expression pattern is the same as the pattern seen in developing and regenerating axolotl limbs. The difference in timing between species may arise because Xenopus regenerates its limbs and tail quicker than axolotl. The simplest explanation for this difference is a difference in the water temperature in which the two species are normally cultured. Alternatively, the difference could be attributed to the fact that Hoxc10 in axolotl is reexpressed, whereas in Xenopus it only needs to be up-regulated in stage 54 regenerating limbs. Other people have speculated about a prominent role for the Hoxc genes in regeneration (Simon and Tabin, 1993; Savard and Tremblay, 1995; Carlson et al., 2001). In axolotl, Hoxc10 is only expressed in the hindlimb during normal development but is reexpressed very quickly after amputation in the hindlimb and activated de novo in the forelimb. This novel expression in regenerating forelimbs makes Hoxc10 the only true regeneration-specific gene to date (Carlson et al., 2001). In newts, however, Simon and Tabin (1993) did not detect Hoxc10 expression in forelimb blastemas by Northern hybridisation analysis.
The Hoxa paralogs are expressed very early (1–2 days after resection) in regeneration in axolotl and have been called dedifferentiation markers for that reason (Gardiner et al., 1995). We also see XHoxa13 reactivation early in the regeneration process (tail blastemas express it very strongly after 1 day and half the amputated limbs show expression after 1 day), and no expression in either wounded limbs or tails. However, despite the similar timing, after 1–2 days Xenopus has progressed further in the regeneration process compared with axolotl and is well into early bud stage.
For XHoxd13, no other regeneration data exist. It is not reactivated until mid-bud stage when patterning of the blastema takes place. A similar story is reported for Hoxd11 in axolotl limb blastemas. Hoxd11 too, seems to have only a role in patterning and is not expressed during the first phase of regeneration (Torok et al., 1998). Of interest, the pattern seen in limb blastemas differs from the pattern in normal developing limbs, insofar that it is restricted to a distal/posterior region and does not spread over the entire anteroposterior axis as in normal development. Hence, in regeneration, XHoxd13 expression is more similar to the expression patterns seen in chick and mouse (Nelson et al., 1996; Torok et al., 1998). It is not known what importance this asymmetric expression pattern might have in regeneration.
Two Phases for Hox Genes in Regeneration
It has become increasingly clear that the Hox genes play several distinct roles in regeneration. In the early regenerate, their temporal control and their expression pattern are quite different from their developmental expression patterns. In this first phase, they seem to fulfil a unique role in setting up a regeneration blastema or having a role in wound healing. In the second phase during mid-bud stage, when the blastema gets patterned, their expression patterns become similar to the ones observed during normal development (Gardiner et al., 1995; Carlson et al., 2001). We also see differences in expression patterns between the early phase of regenerates and limb development, whereas the expression at mid-bud stage more closely resembles the developmental pattern. We propose a role as a “first phase Hox gene” for XHoxc10 in the dedifferentiation and cell recruitment process. Several lines of evidence support this idea. First, XHoxc10 has the right expression profile in blastemas to support such a role. Second, Hoxc10 was detected in regenerating axolotl forelimbs, although it is not expressed there during normal limb development and, therefore, must play an early role in regeneration but not in patterning (Gardiner et al., 1995).
Even though gene expression of the Hoxa genes during regeneration seems to be similar in axolotl and Xenopus, our interpretation varies slightly from one presented previously (Gardiner et al., 1995; Endo et al., 2000). Others have proposed a dual role for Hoxa13 during early regeneration, first in dedifferentiation and second in patterning. Because Hoxa13 specifies the autopod in normal development, they argue, the distal part of the regenerated limb is the first part to be respecified during regeneration, whereas intermediate parts are intercalated secondarily. This finding is in contrast to normal development, where it is accepted that the pattern is specified in a proximal-to-distal sequence. Also, in axolotl, the proximal structures of the regenerated limb differentiate before the distal ones. Furthermore, XHoxc10 is coexpressed with XHoxa13 in these distal blastemas while, during normal limb development, coexpression is minimal. For these reasons, we suggest that the early Hoxa13 expression has nothing to do with patterning or with specifying the autopod but more likely with the proliferation of the blastema. Only later in regeneration would Hoxa13 specify the autopod. Up-regulation of posterior Hox genes and, in particular, Hoxa genes has been implicated in proliferation in hematopoiesis and acute myeloid leukaemia (Bjornsson et al., 2001; Calvo et al., 2002). In C. elegans, a two-phase expression of mab-5, an Antennapedia homolog, has been reported to allow for proliferation in the first phase and patterning in the second one (Salser and Kenyon, 1996). In the second phase, by mid-bud blastema stage, Hoxa13 is restricted to the distal part of knee/elbow level blastemas but covers the whole blastema in ankle/wrist amputations in axolotl, which corresponds to the expected pattern if patterning in regeneration recapitulates patterning during limb development (Gardiner et al., 1995; Endo et al., 2000). We did not observe this distal restriction of XHoxa13 in amputations at knee level; however, our interpretation of the mid-bud blastema staining is consistent with the imperfect regeneration in Xenopus limbs. Stage 54 limbs amputated at knee level regenerate with intermediate structures (zeugopod) missing; therefore, even a proximal amputation regenerates mainly an autopod, which would explain XHoxa13 staining in the whole blastema.
The third Hox gene studied, XHoxd13, does not have an early phase expression and, therefore, may fulfil only a patterning role, similar to normal limb and tail development. There are other Hoxd genes in axolotl, however, which are expressed during the first phase of regeneration. Two of the more 3′ Hoxd genes, Hoxd8 and Hoxd10, are expressed very early in regenerating limbs or wounds putting them at the forefront of the regeneration process (Torok et al., 1998).
It does not look like a particular Hox cluster or a Hox paralog group fulfils a specific role during regeneration. Instead, it seems to be very complex with differences between species.
To get a full-length XHoxa13 clone, two specific primers targeted to the central region of XHoxa13 were designed. Forward primer: 5′ GCTACCTGGATATGCCCGTGG TTTC 3′, reverse primer: 5′ AGAGACGATTTCTAGAGGTGGGTTG 3′ were used to obtain longer 3′- and 5′-RACE-PCR clones (SMART-cDNA kit Clonetech). The RACE-PCR products were then cloned into the promega pGEM-T cloning vector and sequenced. Another pair of primers were designed (forward: 5′ GCTATGACAGCTTCAGTGCTCCTC 3′, reverse: 5′ GGGTTCGGCCTGTATGGGTT 3′) and used to amplify the full-length XHoxa13. This product subsequently was cloned into the pGEM-T vector and sequenced on both strands using the departmental sequencing service (ABI Prism 377). XHoxc10 and XHoxd13 ESTs were obtained in pCMV-sport6 vector from the UK HGMP Resource Centre. They were sequenced initially with SP6 and T7 standard primers and then internal primers: XHoxd13int: 5′ CGGCAA CGGATACTACAGCTG 3′. A full-length XHoxc10 was than obtained by RT-PCR from stage 35 cDNA by using a primer designed against the published N-terminal clone (5′ ACGCGTCCGAAATGTCATGT 3′; Grammer et al., 2000) (accession no. BG354667) and a C-terminal primer in the 3′-UTR (5′ CAGGAGGGCTACTTAACGGT 3′). The PCR product was then cloned into pGEMT and sequenced with SP6, T7, and two internal primers. Accession numbers are Xhoxc10, AY167741; Xhoxa13, AY167740; Xhoxd13, AY167742.
Limb and Tail Resections
Tadpoles were anaesthetised in 1/3,000 w/v MS222 in water and resectioned at knee level with a pair of iridectomy scissors at stage 54 (regenerating) or stage 57 (nonregenerating). For the wounding experiment, a small square (approximately 2 × 2 mm) of skin and muscle was removed with a dissecting needle at stage 54. Tadpoles were allowed to regenerate or heal for either 1, 3, or 5 days before they were killed in 1/200 MS222 and fixed in MEMFA. For tail resections, tadpoles were anaesthetised as stated above and subjected to 50% tail resection at stage 50–52. After recovery, they were cultured in sterile 1/10 NAM salts for 1–3 days before killing in 1/200 MS222 and fixation. All specimens were then subjected to in situ hybridisation.
Whole-mount in situ hybridisation for XHoxa13 has been previously described (Lombardo and Slack, 2001). Probes for XHoxc10 (Image 4203226) and XHoxd13 (Image 3399571) were made by transcription with T7 RNA polymerase after restriction digestion with EcoRI. Whole-mount in situ hybridisations were carried out as previously described (Harland, 1991) with modifications as in Pownall et al. (1996).
We thank Catherine Willoughby for some initial preliminary results.