Early regeneration genes: Building a molecular profile for shared expression in cornea–lens transdifferentiation and hindlimb regeneration in Xenopus laevis



Recent studies in Xenopus laevis have begun to compare gene expression during regeneration with that of the original development of specific structures (e.g., the hindlimb and lens), while other studies have sought differences in gene expression between regeneration-competent and regeneration-incompetent stages. To determine whether there are any similarities between the regeneration of different structures, we have used a differential screen to seek shared early gene expression between hindlimb regeneration and cornea–lens transdifferentiation in the Xenopus tadpole. We have isolated 13 clones representing genes whose expression is up-regulated within the first few days of both regenerating processes and which are not demonstrably up-regulated in the context of basic wound healing. Furthermore, all of these genes also show prominent late embryonic expression. The expression patterns and putative identities of all 13 genes are presented, and a model is considered that allows us to characterize and profile important changes in gene expression, which might be shared among various regenerating and developmental systems. Developmental Dynamics 230:615–629, 2004. © 2004 Wiley-Liss, Inc.


Regeneration entails several key events, including wound healing, cellular de-differentiation, proliferation, and specific changes in gene regulation to direct the differentiation and patterning of regenerated structures. The extent to which this process mirrors the original, embryonic formation of these structures is currently under investigation (Henry et al., 2002; reviewed by Tsonis, 2002; Henry, 2003; Stocum, 2004). While the expression of patterning genes and the ultimate differentiated state are largely unique to each tissue or organ regenerated, there may be early molecular processes, which are largely shared among regenerating systems in driving de-differentiation, establishing a renewing population of replacement cells, and initiating regeneration (e.g., hindlimb vs. lens regeneration).

Recent studies in Xenopus laevis have used differential display to identify up-regulated gene products in regenerating tails (Ishino et al., 2003), subtracted library screening for differential gene expression in transdifferentiating corneas (Henry et al., 2002), and in regenerating and nonregenerating hindlimbs (King et al., 2003). These studies demonstrate that early events in regeneration involve up-regulating large suites of genes, which include those necessary to initiate and maintain successful regeneration. Comparison between regenerating systems shows that some genetic events are shared, such as those involved in wound healing. For example, during wound healing and early regeneration of the limb and lens in Xenopus, matrix metalloproteinases such as MMP-9 are up-regulated and are thought to assist in removing extracellular matrix (ECM) components, which otherwise stabilize the differentiated tissues and, hence, inhibit cellular rearrangements and de-differentiation during regeneration (Carinato et al., 2000; Stocum, 2002; Henry, 2003).

Searches for common molecular profiles among different populations of cells have been used to identify shared markers of “stemness” in mouse embryonic, neural, and hematopoietic stem cells (Ramalho-Santos et al., 2002; Ivanova et al., 2002), although the design and interpretation of these comparisons is still under discussion (Fortunel et al., 2003; Ivanova et al., 2003). A similar strategy can be used to screen for shared genetic events at the beginning of regeneration in multiple systems, thus potentially identifying genes that initiate regeneration or control cellular de-differentiation. To pursue this objective, we chose to compare gene expression during the initiation of hindlimb regeneration and cornea–lens transdifferentiation in the frog Xenopus laevis. Presumably shared, up-regulated genes will reveal factors involved in wound healing, cellular de-differentiation, and perhaps early patterning of the regenerate.

Xenopus hindlimb buds are first visible during the late larval period at developmental stage 46 (stages of Nieuwkoop and Faber, 1956). Amputations through the limb bud up to stage 52 usually result in perfect regeneration of lost distal structures, but during later stages (53–60, late larval and metamorphic stages), this ability declines until the animals eventually fail to regenerate any complete structures and instead produce a nonfunctional, nonmuscular heteromorphic spike by stage 58 (Dent, 1962; Overton, 1963; Muneoka et al., 1986; Wolfe et al., 2000).

Limb amputation typically perturbs tissues that have differentiated to various extents; thus, limb regeneration cannot be a simple recapitulation of the process of limb development. Subsequent to amputation, wound healing must occur, followed by alteration of the stump tissues to produce the regeneration blastema. Previous studies have shown that the early regeneration blastema of Xenopus limbs contains a morphologically undifferentiated mesenchyme (Korneluk and Liversage, 1984; Anton et al., 1988; Khan and Liversage, 1990). At this stage, the regeneration blastema of the Xenopus hindlimb resembles the distal portion of a developing limb bud: a thickened apical epithelial cap devoid of basement membrane and dermis covers a largely undifferentiated mesenchyme. The mesenchyme undergoes proliferation to restore lost mass to the limb, and patterning occurs to direct the new cells into appropriate tissues of the regenerate. Previous studies have verified that, during this latter process, certain growth and patterning factors originally expressed during limb development are re-expressed in similar patterns in later-stage animals after amputation. These factors include fibroblast growth factors 8 and 10 (Christen and Slack, 1997; Yokoyama et al., 2001), sonic hedgehog (Endo et al., 1997), and several hox genes (with some unique regeneration-specific expression in Ambystoma; Gardiner et al., 1995; Torok et al., 1998; Carlson et al., 2001).

Xenopus tadpoles also possess the ability to regenerate the lens of the eye from the corneal epithelium by means of the process of cornea–lens transdifferentiation, originally described by Freeman (1963; reviewed by Henry, 2003). Before mid-metamorphosis in the frog, the outer cornea consists of two squamous epithelial layers with an underlying basement membrane. After removal of the lens, cells of the inner sensorial layer of the outer corneal epithelium become cuboidal in shape. A group of cells overlying the pupillary space then begin to aggregate and separate from the cornea, forming an epithelial vesicle. Cells located toward the vitreous chamber enlarge and elongate, ultimately losing their epithelial appearance and forming organized, parallel lens fiber cells. Ultimately, the new growing lens comes to lie in the optic cup to take the place of the original lens.

During late larval development, maturation of the cornea involves the arrival of migrating, mesenchymal neural crest cells to form the inner cornea (the corneal endothelium) as well as the synthesis and thickening of the collagen meshwork of the stroma between the inner and outer corneas (Hay, 1977). Maturation of the cornea during this time also results in a decreased competence to respond to lens-inducing signals (Freeman, 1963; Henry and Elkins, 2001). Furthermore, as the tadpole progresses through metamorphosis, it demonstrates an accelerated ability to heal wounds to the inner corneal endothelium; the intact inner cornea cuts off critical retinal inductive factors that support transdifferentiation (Bosco et al., 1979; Filoni et al., 1997; Henry and Elkins, 2001).

Morphogenetic events within the cornea, as well as profiles of gene expression, demonstrate that cornea–lens transdifferentiation largely recapitulates the original development of the lens from the presumptive lens ectoderm (Schaefer, 1999; Henry et al., 2002), although in a slightly delayed manner (Brahma and McDevitt, 1974; reviewed by Henry, 2003). One explanation for this delay is that the cells of the cornea overlying the pupillary space must undergo de-differentiation in response to lens-inducing signals from the retina. The alteration of differentiated state can be observed immediately in the morphological change of corneal cells from a squamous to a cuboidal shape. Furthermore, by the second day postlentectomy the cells of the placode change their nucleolar ratio to that more typically seen in lens epithelial cells (e.g., a higher percentage of these cells contain only one nucleolus rather than two; Freeman, 1963). Eventually, γ-crystallin mRNA becomes detectable in the vesicle by 4 days postlentectomy (stage four of cornea–lens transdifferentiation from Freeman, 1963; Brahma and McDevitt, 1974; Henry and Mittleman, 1995; Schaefer et al., 1999), which is not detectable in unoperated larval corneas. The strongest evidence implicating de-differentiation in this process is the observation that mature, postmetamorphic corneal transplants can also undergo transdifferentiation when directly exposed to retinal factors by means of transplantation into the optic cup (Filoni et al., 1997).

To date, the extant molecular markers for regenerating tissues consist primarily of cytoskeletal filament types, ECM components, degradative enzymes, and antigens associated with uncharacterized intracellular products (Géraudie and Ferretti, 1998). Furthermore, nearly all investigation into such markers has been conducted in urodele amphibians whose regeneration mechanics may differ from those of the anuran Xenopus. A more thorough molecular profile of gene expression, occurring specifically during the initiation of regeneration, will give us a better understanding of how amphibian regeneration is induced and maintained. Furthermore, this profile should help illuminate how cells de-differentiate to return to a multipotent state in preparation for regeneration.

Our goal has been to further our understanding of the intrinsic cellular responses necessary to initiate regeneration of the lens and distal hindlimb in Xenopus larvae. Presumably both processes use wound healing, de-differentiation, proliferation, and re-differentiation (or transdifferentiation) of nearby cells to replace lost structures, and we are investigating shared molecular events in these processes that occur before organ-specific determination and patterning. An advantage to using Xenopus for this study is its ability to regenerate disparate tissues such as the hindlimb and lens during the late larval period. Another well-characterized model of limb regeneration, the axolotl, does not regenerate the lens (Stone, 1967; Del Rio-Tsonis et al., 1995).

The present study has identified several shared gene products which are up-regulated during hindlimb regeneration and cornea–lens transdifferentiation. Whereas most of these represent completely novel genes, we have identified a neuronal leucine-rich repeat protein, a putative dimethyladenosine transferase (18S dimethylase), a kelch domain-containing protein, cytochrome oxidase subunit I, and a 16S mitochondrial rRNA among the recovered clones. We have characterized the expression of these genes during initial phases of both regenerative processes; additionally, none of them appears to be expressed during the basic process of wound healing. Furthermore, these factors are also expressed during the development of the eye and other organs during later embryonic stages. The potential roles of these and other genes recovered are discussed in this report. By examining the expression and function of genes for which expression is shared during initial regeneration events in multiple organs, we expect to develop a greater understanding of the general mechanisms by which amphibian transdifferentiation, regeneration, and development are initiated.


Screen for Genes Up-Regulated During Hindlimb Regeneration and Cornea–Lens Transdifferentiation

Approximately 16,000 colonies of a subtracted cornea–lens transdifferentiation library (Henry et al., 2002) were screened with radiolabeled cDNAs generated from regenerating hindlimbs. Positive colonies were picked again and subjected to secondary and tertiary screens using the same probe to eliminate false positives; 23 positive colonies were ultimately recovered. Ten of these were excluded based on size (less than 500 bp after vector trimming), and 13 clones were retained for further study. Each clone was sequenced and compared with known sequences at NCBI by using blastn and blastx searches. The highest probability matches from these searches are shown in Table 1. Clones W007, W009, and W010 have been identified as dimethyladenosine transferase, 16S mitochondrial rRNA, and cytochrome oxidase subunit I, respectively, based on these database searches. Additionally, two clones (W006 and W011) were putatively categorized as members of known protein families, which have been previously unidentified in Xenopus. The predicted translation of clone W006 indicates that it bears a strong similarity to neuronal leucine-rich repeat protein 1 and clone W011 contains several kelch repeat domains and, on this basis, is similar to a specific group of kelch-containing proteins in rat, mouse, and human (see Table 1 and Discussion section below). The other clones, W001, W002, W003, W004, W005, W008, W012, and W013, bear no similarity (or equivocal similarity within their untranslated regions) to known genes. However, all of these clones with the exception of W008 matched available expressed sequence tags (ESTs) in several developmental and adult tissue cDNA libraries (see Table 1). Longer contigs were assembled using these perfect EST matches, but yielded no new identity information in the NCBI databases (i.e., were still of “unknown” identity).

Table 1. Summary of Clones Recovered and Database Search Results
Clone IDGenBank accession no.Size (nt)Putative/known identityaPercent similaritybPutative/known functions and notesReferences
  • a

    Species of match and nucleotide/protein accession numbers from NCBI are given in parentheses.

  • b

    Similarity represents % similar (inferred amino acids) or identical (nucleotide) residues over the span of residues in the match.

  • c

    Multiple disparate hits over the same region, lacking an open reading frame, with comparable Blast scores suggest shared 3′UTR of these mRNAs and were not considered indicative of the identity of this gene.

  • d

    The original clone W011 was recovered in the reverse orientation; all sequence data and in situ hybridization results take the correct orientation of the gene into account.

  • e

    Blastn searches of this clone showed one or more perfect matches to ESTs in Xenopus laevis developing embryo and adult tissue libraries.

  • f

    This clone's sequence has been extended in silico with a 5′RACE fragment to generate additional coding sequence; the “size” column represents the longest contiguous sequence with these fragments assembled.

W002eCK8159911407Hypothetical protein FLJ20397 (H. sapiens NP_060272)77%/140 aaUnknown; predicted protein sequence from mRNAStrausberg et al. (2002)
W005eCK8159942315Epithelial sodium channel, α subunit (X. laevis XLU23535)93%/613 ntUnknownc-
   osteocalcin gene (X. laevis AY043179.1)92%/583 nt  
   hypothetical protein MGC53422 mRNA (X. laevis BC050257.1)92%/543 nt  
W006fCK8159951627Neuronal leucine-rich repeat protein (X. laevis BAA28530.1)57/289 aaSuggested role in neurogenesisHayata et al. (1998)
W007eCK8159961359Similar to dimethyladenosine transferase, 5′ end (X. laevis BG354624.1)100%/336 nt (partial clone)From Harland stage 19–23 Xenopus laevis cDNA libraryGrammer et al. (2001)
   Putative dimethyladenosine transferase (H. sapiens NP_055288.1)92%/313 aa (full length)18s rRNA dimethylase; rRNA maturationStanchi et al. (2001)
W009eCK815998634Mitochondrial DNA, complete genome (X. laevis XELMTCG)97%/620 ntMitochondrial 16S rRNA subunitRoe et al. (1985)
   also overlaps another clone from cornea-lens transdifferentiation library96%/244 ntClone D006; mitochondrial 16S rRNA subunitHenry et al. (2002)
W010eCK815999894Mitochondrial DNA, complete genome (X. laevis XELMTCG)97%/873 ntCytochrome c oxidase subunit 1Roe et al. (1985)
W011fCK816000d1818kelch domain containing 3 mRNA (X. laevis BC056132.1)99%/681 ntCoding sequence predicted from library screenStrausberg et al. (2002)
   Similar to kelch domain containing 3, testis intracellular mediator protein (Rattus norvegicus XP_343533.1)89%/132 aaStructurally homologous to recombination activation gene 2 (RAG2)Ohinata et al. (2003)

Based on predicted reading frames of the sequenced clones, it was concluded that all of the clones are 3′ partial cDNAs. 5′ rapid amplification of cDNA ends (RACE; Clontech, Palo Alto, CA) was conducted, isolating more 5′ sequence from clones W006, W011, and W012. RACE amplicons from these clones were included in GenBank submissions.

Expression During Development

In situ hybridization using digoxigenin-labeled riboprobes was conducted to ascertain the spatiotemporal expression of each clone during embryonic and early larval development. Stages examined included postgastrulae (stage 14) through early larvae (stage 37). Representative whole-mount samples at larval stages 24–37 from each clone are shown in Figures 1 and 2. All 13 clones showed expression patterns that included the lens of the eye, and many were found in the developing central nervous system and in various mesodermally derived tissues. Developmental expression patterns by tissue type are summarized in Table 2. The five clones that showed expression at gastrula or neurula stages were faintly represented throughout the developing neural plate/tube and in the presumptive placode-forming ectoderm at the anterior pole (see Table 2).

Figure 1.

Spatiotemporal expression of clones W001 through W007 during selected posthatching stages. Embryos and early larvae at the stages indicated in the top row were probed with “antisense” riboprobes against the clones indicated in the leftmost column. All specimens are shown with the anterior end to the left and the dorsal surface toward the top of the figure. Where indicated, white arrowheads indicate the position of the presumptive lens ectoderm or lens and black arrowheads indicate the position of the otic vesicle. The inset in O (W005) is a twofold enlargement of the boxed area in another specimen that underwent longer color development after in situ hybridization to reveal mRNA localization in discrete groups of migrating somite-derived cells (open arrowheads) described by Martin and Harland (2001), which presumably represent hypaxial myoblasts. Scale bar = 1 mm in U (applies to A–U), 0.5 mm in O inset.

Figure 2.

Spatiotemporal expression of clones W008 through W013 during selected posthatching stages. Embryos and early larvae at the stages indicated in the top row were probed with “antisense” riboprobes against the clones indicated in the leftmost column. All specimens are shown with the anterior end to the left and the dorsal surface toward the top of the figure. Where indicated, white arrowheads indicate the position of the presumptive lens ectoderm or lens and black arrowheads indicate the position of the otic vesicle. Open arrowheads indicate the position of the pronephros in F, I, and L. Prominent staining of the cloacal region by W009 is evident adjacent to asterisks in D–F. Scale bar = 1 mm in R (applies to A–R).

Table 2. Summary of Organ-Specific Gene Expression During Embryonic and Early Larval Development
inline image

Expression During Hindlimb Regeneration and Development

Tadpoles at stages 51–53 were amputated through the presumptive ankle joint of the left hindlimb and allowed to recover for 24, 48, and 72 hr. All 13 clones demonstrated prominent expression by means of in situ hybridization in the regeneration blastema, in the epithelium of the apical epithelial cap (AEC) or in the proliferating, mesenchymal cells of the blastema underlying the AEC. Representative samples from these groups of specimens are shown in Figure 3, and results for limb expression during regeneration and development are summarized in Table 3. Three distinct trends in expression were observed. Clones W001, W003, W005, and W007 displayed absent or minimal levels of expression in the regenerate before 48 hr postamputation, but this expression demonstrably intensified between 48 and 72 hr (Fig. 3A, C, E, and G, respectively; see also Table 3). The opposite temporal pattern—maximal expression in earlier blastemas (24/48 hr) diminishing through later regeneration (72 hr)—was observed with clones W002 and W013 (Fig. 3B and M, respectively; see also Table 3). The remaining seven clones did not show demonstrable changes in expression between time points; their blastema expression appears consistent throughout the initial 72 hr of hindlimb regeneration (see Fig. 3; Table 3). Although each clone demonstrated clear up-regulation within the regeneration blastema after amputation, several clones are also expressed in the normal, developing limb (see Table 3).

Figure 3.

Expression during hindlimb regeneration. Each clone's expression during hindlimb regeneration was examined using in situ hybridization 24, 48, and 72 hr after amputation through the left presumptive ankle joint at stages 51–53. The right limb in each case was also examined for expression in the unoperated, developing hindlimb. The paired limbs are shown with the distal (or amputated) end toward the top of the figure. Each sample may be staged by the morphology of the control limb on the right; differences in regeneration blastema expression were not observed among the three stages chosen for amputation. Clones used as riboprobe are shown as follows: A, W001; B, W002; C, W003; D, W004; E, W005; F, W006; G, W007; H, W008; I, W009; J, W010; K, W011; L, W012; M, W013. Each panel displays a representative limb pair from either a 24 hr or 48 hr regenerate (“early” time point) and a limb pair from a 72 hr regenerate (“later” time point). In some, but not all, cases background staining was observed at the base of a limb, corresponding to the site where the limbs were detached from the flank tissue after fixation (e.g., B, left limb pair); this staining was also observed in “sense” probe controls (not shown) and is an artifact relating to tissue preparation. N: Illustrations of limb morphology used to stage each set of limbs are shown; the dashed lines indicate amputation sites at each stage (illustrations modeled after Nieuwkoop and Faber, 1956). The stage at which each sample was amputated may be ascertained by comparison with these staging diagrams (recovery times were not sufficient for specimens to advance in developmental stage). Scale bar = 1 mm in N (applies to A–N).

Table 3. Summary of Expression During Limb Regeneration and Developmenta
Clone IDDevelopmental stageRegeneration blastemaWound Assayb
51525324/48 houra72 hour
  • a

    No differences in expression were observed between 24 and 48 hour hindlimb regenerates.

  • b

    See text for detail of wound assay.

  • M, proliferative Mesenchyme of the limb bud or regeneration blastema; AER, Apical Ectodermal Ridge (thickened epithelium overlying the limb bud mesenchyme); AEC, Apical Epithelial Cap (thickened epithelium overlying the regeneration blastema); + indicates that mRNA for the clone was detected; (+) indicates that mRNA for the clone was detected at a minimal level; − indicates that mRNA for the clone was not detected.


Expression During Cornea–Lens Transdifferentiation

Figure 4 shows sample in situ hybridization specimens from lentectomized tadpoles at stages 47–51. Expression of the genes was prominent in the outer cornea overlying the pupillary space within 24–96 hr after lentectomy. Two of these six clones, W002 and W011, were also detected in the lateral line organs of the head (Fig. 4B,F, hollow arrowheads). All 13 clones were examined by means of polymerase chain reaction (PCR) for representation in transdifferentiating and control cornea cDNAs using clone-specific primers. Results of the differential PCR analysis are shown in Figure 5 (see Experimental Procedures section). Five of the clones, W001, W002, W006, W012, and W013, are only represented in the transdifferentiating cDNA and not in control (unoperated) cDNA. The remaining eight clones appear to be expressed in both the transdifferentiating and control corneas. Although these latter clones have not been removed during the subtraction process, they appear to represent genes that are dramatically up-regulated during transdifferentiation and that are also present in the subtracted cDNA library (Henry et al., 2002). For instance, in situ hybridization (e.g., W003, W004, W007, and W011) shows demonstrable up-regulation in the transdifferentiating cornea (e.g., Fig. 4C–F). None of these genes is detected in control (unoperated) corneas by in situ hybridization (e.g., Fig. 4G).

Figure 4.

Expression of selected clones during cornea–lens transdifferentiation. mRNA expression was examined by means of in situ hybridization 24, 48, 72, and 96 hr after lentectomy in tadpoles at stages 47–51. Representative examples and regeneration times are shown as follows: A, W001 at 48 hr; B, W002 at 24 hr; C, W003 at 24 hr; D, W004 at 24 hr; E, W007 at 24 hr; F, W011 at 96 hr. White dotted lines illustrate the approximate boundary of staining in the cornea epithelium. Open arrowheads in B and F indicate prominent staining of lateral line organs by probes W002 and W011, respectively. G: The contralateral, unoperated lens from F (probed with W011). H: An eye 96 hr after lentectomy probed with a “sense” control riboprobe for W011. All specimens were examined during developmental stages 47–51 (Nieuwkoop and Faber, 1956); eye morphology does not demonstrably differ during these stages. Scale bar = 300 μm in H (applies to A–H).

Expression During Wound Healing

To determine whether the expression of these clones during regeneration and transdifferentiation was truly regeneration-specific or simply a response to wounding, we conducted a “wound assay” and looked for their expression by means of whole-mount in situ hybridization (as described by Carinato et al., 2000). Tadpoles at stages 50–53 were wounded in three locations: the posterior margin of the outer cornea was cut (as it is during lentectomy, but the inner cornea was left intact so as not to trigger regeneration), the hindlimb at the presumptive tarsal joint was cut halfway through its cross-sectional area, and a lateral incision was made through the flank ectoderm and underlying musculature (see Experimental Procedures section). Specimens were allowed to recover for 24 hr, as this time period has been sufficient for detecting expression of other genes implicated in wound healing (Carinato et al., 2000). No expression of any of the genes represented by these clones was detected at the wound sites (data not shown).


We have identified a small number of genes expressed early during both hindlimb and lens regeneration. Based on the number of colonies screened from the subtracted cornea–lens transdifferentiation library, we estimate the overall frequency of positives originally recovered at approximately 0.14% (see Experimental Procedures section), which represents a fairly low level of similarity between these two processes. It is also important to note that we started with a subtracted cDNA library enriched specifically for genes expressed during cornea–lens transdifferentiation. The percentage of clones recovered implies that the subset of shared genes between disparate regenerative processes is small; perhaps the early expression of these genes assists tissues in mobilizing their specific regeneration or transdifferentiation machinery.

Each of the 13 clones examined shows prominent expression in the developing embryo, suggesting that the initiation of regeneration involves re-deployment of developmental genes rather than the utilization of novel, “regeneration-specific” ones. However, specific tissues may use other, nondevelopmental genes during regeneration; for example, some previously isolated clones from the subtracted cornea–lens transdifferentiation cDNA library do not have demonstrable embryonic expression patterns (e.g., clones B010, B037, B044, B046, B070, B097, B111, C046; Henry et al., 2002; Walter et al., 2004). These clones may represent genes uniquely involved in transdifferentiation of the cornea. It is interesting that no members of this class of genes (e.g., not expressed in embryonic development) were recovered in the present screen. The shared re-deployment of embryonic genes during early regeneration events suggests that, to some extent, both regenerative processes involve a return through a previous, embryonic “node” of gene expression (Henry, 2003). Furthermore, the most prominent developmental expression of all 13 genes occurred fairly late in the developing tadpole, well after initial embryonic patterning (e.g., stages 28–37; see Table 2; Figs. 1, 2). The timing of this expression may indicate roles in organogenesis and differentiation, rather than global roles in embryonic pattern formation, although functional studies of each gene are needed to confirm this classification. The five clones from this study that could be at least putatively identified do have some known and predicted roles during development and regeneration, as are discussed below.

Neuronal Leucine-Rich Repeat Protein May Play a Role in Neurogenesis and Cranial Placode Formation and Regeneration

The only neuronal leucine-rich repeat protein (NLRRP) previously identified in Xenopus is XNLRR-1, which was classified based on its predicted protein sequence similarity to mammalian NLRRP-1 (Hayata et al., 1998). That investigation suggested that XNLRR-1 plays a role in neuronal cell surface interactions, particularly cell adhesion. Clone W006 from this study demonstrates a low percentage similarity at the protein level (57%) in its available coding region to XNLRR-1 but has no significant similarity to the gene at the nucleotide level, suggesting that it might be a novel member of this protein family in Xenopus. Additionally, the developmental expression pattern of W006 is very similar to that reported for XNLRR-1, with prominent staining in the brain and developing eye (Fig. 1Q–R; Hayata et al., 1998). Although leucine-rich repeat-containing proteins perform a variety of described functions, they share a role in protein–protein interactions, mediated by several leucine-rich repeat regions (Bormann et al., 1999). In regenerating axons of the zebrafish optic nerve after transection, transcripts of zfNLRR rise from undetectable levels to maximal expression within days of injury then drop back to undetectable levels after nerve growth (Bormann et al., 1999). Clone W006 demonstrates a similar pattern of expression during cornea–lens transdifferentiation; it is not detectable by PCR in control corneas but is up-regulated during the first 4 days of transdifferentiation (Fig. 5). Another leucine-rich repeat protein has been implicated in mediating Nogo-66 inhibition of axon regeneration in vitro (Fournier et al., 2001), emphasizing the varied roles that members of this protein family might play in regenerating systems.

Figure 5.

Transdifferentiating versus control cornea polymerase chain reaction (PCR) analysis. Each clone was amplified with specific primer pairs from transdifferentiating (“T”) and control (“C”) cornea cDNA, as described in the Experimental Procedures section. Each PCR amplicon shown matched its positive control PCR product (not shown). Note that no product was amplified from control cornea cDNA for W001, W002, W006, W012, and W013.

Dimethyladenosine Transferase Is an Uncharacterized Participant in Regeneration

Dimethyladenosine transferase (clone W007), also known as 18S rRNA dimethylase, was identified in a screen of human transcripts that matched yeast genes (Stanchi et al., 2001). This protein dimethylates two adjacent adenosine residues of pre-18S rRNA, and has been suggested to play an essential role in regulation of rRNA maturation. When partial mRNA of the Xenopus homolog of this gene was unilaterally injected into two-cell stage embryos, expression of the pan-neural marker nrp-1 was lost (Grammer et al., 2000). This result suggested that dimethyladenosine transferase plays a role in regulating neural tissue development, although no studies to date have examined its role in regeneration.

Mitochondria Play a Role in Regenerating Systems

It stands to reason that a tissue which is embarking on massive cellular rearrangement, proliferation, or changes in gene expression will require additional sources of energy. Previous work has indicated that regenerating hindlimbs in larval Bufo regularis demonstrate increased numbers of mitochondria in the muscle and cartilage (Ramadan et al., 1987). Furthermore, oxygen consumption and the activity of cytochrome oxidase increase during regeneration of muscle in axolotl limbs, while suppression of regenerative capacity by means of x-irradiation led to a drop in this activity (Teplits, 1975). These results predict that increased transcription of mitochondrial genes involved in respiration would be seen in regenerating tissues, as is the case with clone W010 in this study (see in situ hybridization of regenerating vs. control hindlimbs, Fig. 3J). Additionally, increased transcription of genes, which are important for protein synthesis, might also be expected during regeneration. Collins (1972) reported that nuclear rRNA gene amplification and concomitant increase in nuclear rRNAs occur during the initial stages of lens regeneration in the newt Triturus. The present investigation extends this finding by locating a mitochondrial rRNA, clone W009, which also shows increased transcriptional levels by means of in situ hybridization in regenerating hindlimbs (Fig. 3I). Other mitochondrial genes have been isolated from the cornea–lens transdifferentiation subtracted library, such as human mitochondrial ribosomal protein S2 and mitochondrial ribosomal protein L13 (clones H153 and H169, respectively; Henry et al., 2002), further suggesting a role for changes in mitochondrial gene expression during regeneration.

Kelch Domain Containing 3 Is a RAG-2–Like Protein Implicated in Meiotic Recombination

Clone W011 contains an open reading frame whose predicted protein sequence shares similarity with the latter two kelch domains of mammalian kelch domain containing 3 (Table 1). This protein in mouse, Peas, bears a similarity to recombination activation gene 2 (RAG-2) and is found in the spermatocytes of the testis; it is thought to play a role in meiotic recombination (Ohinata et al., 2003). Whereas clone W011 only bears a partial similarity to mammalian kelch domain containing 3, it shows nearly perfect identity to a long stretch of the close Xenopus homolog of this gene at the nucleotide level (see Table 1). This similarity, coupled with the transcript's location in the developing eye, ear, and central nervous system (Fig. 2J–L), regenerating hindlimb (Fig. 3K), and the placode-derived lateral line organs of the head and in the transdifferentiating cornea (Fig. 4F), suggests that its role in Xenopus may be very different from its predicted role in the mouse. Although this clone did not seem to bear a significant resemblance to recently classified BTB/kelch genes in Xenopus neural development (Haigo et al., 2003), its expression patterns suggest a role more similar to the reported roles of those genes in neural floor plate and placode development. Further investigations into this gene's role in development and regeneration of the lens and hindlimb are currently under way.

Model for Profiling Regeneration Gene Expression

All of the genes isolated in this study exhibit early expression, which is initiated within the first 24–72 hr of regeneration. Until functional studies are undertaken we can only speculate as to the roles of these and other genes in regeneration. One can, however, speculate on the classes of gene expression that might be associated with these processes. Figure 6 shows a diagrammatic model predicting such classes of genes in transdifferentiating tissues (modified from Henry, 2003). The profiles labeled “A” through “H” represent types of genes whose expression can be altered as cells switch from state “X” to state “Z,” such as from cornea epithelium to lens cells or from stump tissues to mesenchyme to restore missing limb structures during hindlimb regeneration. The shift between these states necessitates the establishment of a “transitional state” characterized by the de-differentiated progenitor cells that restore the missing structure. In the case of hindlimb regeneration, this state represents de-differentiation, proliferation, and differentiation of the blastema tissues; the mesenchymal cells of the blastema take on a phenotype and expression pattern reminiscent of the development of the original limb bud (reviewed by Stocum, 2004). During cornea–lens transdifferentiation, the “transitional” state might also signify regression through a previous, embryonic state or a direct shift in gene expression from cornea-specific to lens-specific genes (Henry, 2003). The transitional state might also be characterized as a “neutral” or repressed condition in which the genes defining the two differentiated cell types are not expressed (i.e., complete absence of state “X”- and “Z”-specific genes). Some genes, such as those illustrated as type “A” in Figure 6, demonstrate maximal postdevelopmental expression during the transitional state of regeneration. These genes could include those specific to regeneration and the establishment of the de-differentiated state, as well as re-expressed developmental regulatory genes such as Xotx2 and xSOX3 during cornea–lens transdifferentiation (Schaefer et al., 1999). Another role for type “A” genes could be in promoting proliferation of cell populations or repressing differentiated states. An example of a gene that may play these roles is found during newt limb regeneration in the re-expression of the gene msx-1; this transcription factor promotes cell proliferation, is inversely correlated with differentiated states, and is implicated in fragmentation and mononuclear myoblast formation among myotubes in the newt limb blastema (discussed by Brockes and Kumar, 2002). In Xenopus limb regeneration, the patterning factor gene sonic hedgehog is re-expressed in response to amputation in a pattern that resembles its initial, developmental expression (Endo et al., 1997). From this study, clones W001, W002, W006, W012, and W013 may play this role in the cornea, as they are undetectable in the control cornea but are up-regulated during transdifferentiation (Figs. 4, 5). In the hindlimb, clones W001, W003, W006, W009, and W010 appear to be expressed in both the regeneration blastema and in the contralateral, unamputated hindlimbs (see Fig. 3A,F,I,J controls and Results section above), suggesting that these genes may play a role in limb regeneration more akin to type “B” genes described below.

Figure 6.

Classes of gene expression during regeneration and transdifferentiation. The initial tissue state “X” and the final regenerated tissue state “Z” represent differentiated states, while the “Transitional State” (gradient shading) represents the process of de-differentiation and proliferation or transdifferentiation. Different types of gene expression that may be observed during these processes are shown by “A” through “H;” see Discussion section for detailed descriptions and examples of each class (modified from Henry, 2003).

Similar to the type “A” genes described above are genes whose transcripts are detectable at low levels in differentiated tissues but that rise dramatically during the transitional state. This pattern is depicted as type “B.” Examples of genes whose expression seems to follow this pattern are the T-box transcription factors NvTbox1 in newt forelimbs and Tbx4 in hindlimbs. In both cases, basal transcript levels of each gene were detected in the corresponding adult limbs, and in response to amputation higher transcript levels were induced, followed by a return to the basal level at the conclusion of regeneration (Simon et al., 1997; Khan et al., 2002). These results were interpreted to indicate a requirement for baseline expression of genes that confer positional information in the adult limb, thus predisposing the tissue to respond to regeneration–initiation signals. This principle may also be illustrated by pax-6 expression during Xenopus cornea–lens transdifferentiation (Schaefer et al., 1999; Mizuno and Mochii, 1999). Similarly, during Wolffian regeneration of the lens from the dorsal iris in the newt, baseline expression of pax-6 in eye tissues was dramatically increased during regeneration and, afterward, dropped back to low levels (Del Rio-Tsonis et al., 1995).

Certain genes expressed during the regeneration and transdifferentiation processes will be specific to either the initial or the final differentiated state. Types “C” and “D” indicate genes whose expression correlates with the initial state “X” while types “E” and “F” represent final state “Z”-specific genes. Examples of these differentiated state-specific genes may be found during the transdifferentiation of pigmented epithelial cells (PECs) in chick and human cultures into lens. When the cells lose their pigment they also lose expression of PEC-specific genes (Fig. 6, type “C”), while not yet expressing lens-specific ones (Fig. 6, type “E”), producing a transitional state known as de-differentiated PECs (dePECs; Eguchi and Kodama, 1993; Kodama and Eguchi, 1994). PEC-specific markers such as the melanosomal matrix protein 115 (mmp115) and the protease inhibitor pP344 (Mochii et al., 1988a, 1988b; Eguchi and Kodama, 1993) and the PEC-sustaining transcription factor Mitf (Mochii et al., 1998) are down-regulated as the PECs lose their pigment and enter the transitional dePEC state. Likewise, the lens-cell specific transcription factor L-Maf is an example of a PEC gene, which serves as a type “E” marker of lens state-specific expression; this factor is likewise expressed in several other vertebrate lenses, including Xenopus (Ogino and Yasuda, 1998). Additionally, lens-specific αA1, γ6, and δ1 (avian only) crystallins are not expressed during the transitional state (e.g., in dePECs), but transcripts of these genes have been used to indicate terminal lens-specific gene expression (Ogino and Yasuda, 1998; Schaefer et al., 1999). At the same time, dePECs express a type “A” gene, c-myc, which has been proposed to prevent dePECs from entering G0, thus maintaining the transitional state (Agata et al., 1989; Eguchi and Kodama, 1993).

Other subtle differences may be noted, such as the difference between type “C/D” and between type “E/F” relating simply to the timing of gene expression during the transitional state; some differentiation-specific genes may quickly be silenced or activated, while others may increase or decrease in their expression more gradually. Figure 6 predicts that overlapping expression of type “D” and “F” genes is possible, as has been shown in transdifferentiating striated muscle in the jellyfish Podocoryne carnea. Within 1 hr of the onset of transdifferentiation in these cells the transcript of Pax-B is detected and its expression continues throughout the transition into nerve and smooth muscle cells for several weeks (Gröger et al., 2000). During this same process, striated muscle myofilaments (i.e., type “D” products) are still detectable as late as 10 days after the initiation of transdifferentiation (Schmid et al., 1988), suggesting that gene products indicative of each differentiated state may overlap in their expression during the transitional state.

Genes that are down-regulated during the transitional state also represent potentially important molecular players in regeneration. These genes, represented by type “G,” are those whose products repress transdifferentiation, or which sustain differentiated cell states. One example of this phenomenon is found at the level of protein regulation rather than transcription. In the myotubes of newt, the active retinoblastoma (Rb) protein blocks the entry of nuclei into mitosis and sustains the differentiated muscle state (Zacksenhaus et al., 1996). However, in response to regenerative signals in vitro (serum stimulation), Rb becomes phosphorylated, loses its ability to block cell cycle reentry, and myotube nuclei begin proliferating (Tanaka et al., 1997). A similar pattern was observed in the newt: antibodies against hyperphosphorylated Rb detected increased levels in the transdifferentiating pigmented epithelial cells of the dorsal iris (Thitoff et al., 2003). This level of protein control resembles the pattern shown in type “G” and serves as a reminder that the categories of regeneration gene profiling shown in Figure 6 should be further scrutinized at the levels of translational regulation and posttranslational modification, in addition to transcriptional activity of the genes.

Other genes, whose expression might be found at relatively unchanged levels in intact and regenerating systems, have been relegated to type “H.” These genes may include “housekeeping genes” and others whose expression (or activity) does not vary between differentiated state “X,” the transitional state and differentiated state “Z.” The screens used in this study were designed to exclude genes of types “C,” “D,” “G,” and “H.” We specifically set out to identify genes likely to fall into the type “A,” “B,” “E,” or “F” categories, because the original subtracted cornea–lens transdifferentiation library is enriched for genes whose expression is initiated or dramatically up-regulated during the transitional state (Henry et al., 2002).

By comparing molecular profiles in regenerating populations from different tissues, it should become possible to elucidate shared mechanisms, which are important to regeneration in general. This investigation has examined the initial period of hindlimb regeneration and cornea–lens transdifferentiation to isolate clones of 13 different genes whose expression suggests roles in development as well as regeneration. Further study will be devoted to establishing the function of these genes in both processes.



Adult albino Xenopus laevis were obtained from Xenopus I (Ann Arbor, MI) and NASCO (Fort Atkinson, WI). Fertilized eggs and embryos were prepared according to Henry and Grainger (1987). Embryos and larvae were reared as described in Henry and Mittleman (1995) and staged according to Nieuwkoop and Faber (1956).


All surgical specimens were anesthetized in MS222 (1:2,000 dilution; Sigma, St. Louis, MO) until immobile. Surgery and recovery of specimens were conducted in 1/20x NAM. Specimens for hindlimb regeneration (tadpole stages 51–53) were amputated at the presumptive ankle joint and allowed to recover for up to 72 hr. Tadpoles at stages 47–51 were selected as specimens for cornea–lens transdifferentiation. The posterior half of the outer cornea was cut and reflected, followed by incision through the inner cornea and removal of the lens with sharp Number 5 Dumont forceps. The outer cornea was smoothed back into place, and the animals were allowed to recover for up to 4 days.

The wound assay was conducted on tadpoles at stages 50–53. Specimens were subjected to incision through the posterior half of the outer cornea, partial amputation through the posterior half of the presumptive ankle joint, or a lateral incision through the epidermis and underlying flank muscles. Animals were allowed to recover for 24 hr.

Library Screening

A subtracted cDNA library, enriched for genes expressed during cornea–lens transdifferentiation, was constructed as previously described (Henry et al., 2002), using stage 50–52 mRNA from transdifferentiating and unoperated corneas. Total RNA was collected from stage 51–53 regenerating hindlimb blastemas every 6 hr for the first 72 hr of regeneration and reverse transcribed using SuperScript II MMLV reverse transcriptase (Invitrogen, Carlsbad, CA), an oligo-dT primer and a SMART oligonucleotide adapter using the SMART III cDNA library construction kit (Clontech, Palo Alto, CA). First-strand cDNA was amplified by means of long-distance PCR per the kit's instructions. A total of 25 ng of regenerating hindlimb cDNA was used as a template for 32P-dCTP radiolabeling with the RadPrime protocol (Invitrogen).

Approximately 16,000 colonies of the subtracted cDNA library were plated and lifted onto nylon Osmonics membranes (82 mm, 0.45-μm pore size; GE Osmonics, Minnetonka, MN); lifts were then denatured and ultraviolet-crosslinked as described by Sambrook et al. (1989). The colony lifts were probed with regenerating hindlimb radiolabeled probe overnight at 68° in Church block (Church and Gilbert, 1984). Colonies that showed hybridization to the regenerating hindlimb probe were selected from the original plates and secondary and tertiary screens were conducted to eliminate false positives. Originally, 23 positive clones were identified by this method; 10 of these clones were excluded due to size (less than 500 bp after vector trimming). When all 23 original positive clones are considered, the frequency of clones recovered is slightly above 0.14%.

5′ RACE (SMART RACE cDNA Amplification Kit, Clontech) was used to isolate more 5′ sequences. Gene-specific, 5′-directed RACE primers and nested primers were designed for each clone and were used with poly-A RNA from stage 32–36 hatchling larvae, collected as described above, and purified by using the Micro-polyA Purist Kit (Ambion, Austin, TX). Three of the clones, W006, W011, and W012, were extended successfully, yielding additional fragments ranging from 1.5 to 2.0 kb. The RACE sequences were appended to the sequences of the appropriate clones and used in database searches (see Table 1).

In Situ Hybridization

Digoxigenin (DIG)-labeled antisense and sense RNAs were produced by digesting selected clones with NotI or SalI and transcribing with T7 and SP6 RNA polymerases (Invitrogen or Roche, Indianapolis, IN), respectively, following the manufacturer's instructions (Roche). In situ hybridization was performed as previously described (Harland, 1991), with the exception that 2 mg/mL glycine in phosphate buffered saline/Tween was used to terminate proteinase K digestion of specimens (Belo et al., 1997). Hybridized probe was detected by using anti–DIG-alkaline phosphatase (Roche) followed by the alkaline phosphate substrate BM-purple (Roche). Embryos (stage 12) through early larval stages (stage 37) were examined for expression during normal development. Hindlimb regenerates were probed at 24, 48, and 72 hr postamputation, and cornea–lens transdifferentiation specimens were probed at 24, 48, 72, and 96 hr postlentectomy. Wound assay specimens were probed at 24 hr postwounding (see Surgeries section above). Hindlimb specimens were subsequently bisected with a sharp scalpel blade and examined for staining in the apical ectodermal ridge and proliferating mesenchyme of the developing limb and in the AEC and proliferating mesenchyme of the regenerate limb.

Differential PCR Analysis

Based on the sequence analysis of individual clones, specific PCR primers were designed to amplify known fragments to check for the presence or absence of these sequences in plasmid DNA prepared from the two original cDNA libraries used to prepare the subtracted cDNA library (the control and transdifferentiating cornea libraries described above). The reactions each contained 2.5 mM MgCl2, 0.2 mM dNTPs (Roche), 0.2 μM each of the clone-specific forward and reverse primers, and 1 unit of Taq polymerase in a final volume of 50 μl of 1× PCR buffer (Invitrogen). One or 10 ng of input library cDNA was used in each PCR reaction. MgCl2 concentrations and annealing temperatures (Ta) were adjusted for optimal results starting with the recommended temperatures for each specific set of primers. Reactions were run through 30 cycles of the following: 30 sec at 94°C, 30 sec at Ta, and 3 min at 72°C, followed by a terminal 5-min extension at 70°C. Positive controls using 1 ng of purified plasmid containing the appropriate cloned cDNAs were run with every set of reactions. PCR products were resolved by using 1% agarose TAE gels (Sambrook et al., 1989).


The authors thank Matthew Elkins, Brian Walter, Maria Carinato-Cengel, Ronald Parchem, and Jane He for technical advice and assistance during this study.