Tokay gecko tail regeneration involves temporally collinear expression of HOXC genes and early expression of satellite cell markers

Abstract

Background

Regeneration is the replacement of lost or damaged tissue with a functional copy. In axolotls and zebrafish, regeneration involves stem cells produced by de-differentiation. These cells form a growth zone which expresses developmental patterning genes at its apex. This system resembles an embryonic developmental field where cells undergo pattern formation. Some lizards, including geckos, can regenerate their tails, but it is unclear whether they show a “development-like” regeneration pathway.

Results

Using the tokay gecko (Gekko gecko) model species, we examined seven stages of tail regeneration, and three stages of embryonic tail bud development, using transcriptomics, single-cell sequencing, and in situ hybridization. We find no apical growth zone in the regenerating tail. The transcriptomes of the regenerating vs. embryonic tails are quite different with respect to developmental patterning genes. Posterior HOXC genes were activated in a temporally collinear sequence in the regenerating tail. The major precursor populations were stromal cells (regenerating tail) vs. pluripotent stem cells (embryonic tail). Segmented skeletal muscles were regenerated with no expression of classical segmentation genes, but with the early activation of satellite cell markers.

Conclusions

Our study suggests that tail regeneration in the tokay gecko—unlike tail development—might rely on the activation of resident stem cells, guided by pre-existing positional information.

Background

Regeneration is the replacement of missing or lost tissue with a functional copy. It helps animals deal with injuries [1, 2] and is seen at different levels of biological organization and at different stages of the lifecycle [1]. Interestingly, humans have a very limited regeneration capacity [3, 4] and one of the goals of regeneration research is to develop new advances in regenerative medicine [5, 6]. Differences in the availability of stem cells might partly explain why regenerative capacity differs in different species [3, 7]. Regenerative capacity is also influenced by the characteristics of stem cells, potentials for dedifferentiation and trans-differentiation, the expression of genes associated with regeneration, the presence of epigenetic regulators, and immune system responses (reviewed in Ref. [3]).

Some vertebrates show an impressive capacity for regeneration. These include the zebrafish [8], salamanders [9], the axolotl [10], and lizards [11, 12]. Interestingly, regeneration in the axolotl limb [13,14,15,16] and zebrafish caudal fin [8] bears a number of similarities to the embryonic development of those structures. For example, an entire suite of developmental patterning genes is activated in the regeneration blastema. Furthermore, the tissue organization of the blastema is similar to an embryonic limb bud or tail bud because both of these structures are small, composed of a mass of undifferentiated cells, and have a distal growth zone (the progress zone in the case of the limb bud) [17, 18] where developmental genes are active.

The tokay gecko (Gekko gecko) is one of the many lizards showing tail regeneration. This typically follows “autotomy”, the process whereby the animal suddenly sheds its tail to distract a predator [19]. The regenerated lizard tail is a functional replica of the original but shows some anatomical differences [11, 20]. These include a continuous cartilage tube instead of caudal vertebrae, an ependymal tube in place of the spinal cord, and a slightly different pattern of scales [21, 22].

Some authors have argued that the regeneration blastema in lizards does not resemble the apical “growth zone” seen during regeneration in other vertebrates. Thus, in the green anole lizard, proliferating cells are distributed along the proximodistal axis of the regeneration blastema [23], and there is no apical (distal) growth zone [24]. However, other research suggests that some lizards do regenerate the tail via a typical growth zone (reviewed by Ref [25]). Thus, there is conflicting information about how lizard tail regeneration is organized. In this paper, we have explored the unusual pattern of tail regeneration in the tokay gecko by characterizing the genes and cell populations involved. By using a combination of bulk transcriptomics, single-cell sequencing, and performing an extensive panel of in situ hybridization experiments on tissue sections, we have characterized tail regeneration in the tokay gecko.

Results

A distinctive molecular signature at each stage of tail regeneration

We conducted mRNA-seq of the regenerating tail at different stages, namely 0, 4, 8, 16, 20, and 28 days post autotomy (dpa; N = 3 per stage; Fig. 1A). Note that the 0 dpa sample consists of adult tail tissue that has not undergone regeneration. Differential gene expression analysis was performed by comparing each stage with the previous adjacent stage (for example, stage 4 with stage 0, stage 8 with stage 4, and so on). The transcriptome sequencing shows that each time-period in the tail regeneration has a very different transcriptional profile. The most dramatic transcriptional change (in terms of number of genes) is at the beginning of regeneration (0–4 dpa) when 2565 genes show significant differential expression, 1241 of those being unique to that comparison (Fig. 1B,D). The second largest change is during blastema formation (8–16 dpa) where 355/1462 differentially expressed genes were unique to that comparison. For the other comparisons (4–8 dpa, > 16 dpa), the number of uniquely regulated genes is much less (Fig. 1D).

Fig. 1

Transcriptome analysis shows a distinctive gene expression profile at each stage of regenerating tail and embryonic tail. A Gecko samples used for transcriptome analysis (indicated by boxed areas). Note that we also included adult tail stump/0 dpa for transcriptome analysis (figure not shown). B Number of up- and down-regulated genes during tail regeneration stages. Differential gene expression analysis by stage-wise comparison. Three biological replicates each stage. Significant differential expression (DE) genes with adjusted p-value < 0.01 and L2FC ≥ 1. Statistical test results can be found in Additional file 6: Suppl_ 1. C Number of genes regulated in regenerating tail vs. embryonic tail. ET embryonic tail, RT regenerating tail. Three biological replicates each stage. Significant differential expression (DE) genes with adjusted p-value < 0.01 and L2FC ≥ 1. D UpSet of cross-state showing the comparison between different stages of regeneration and tail development. ET embryonic tail, RT regenerating tail. E Heatmap of 6817 significant differentially expressed genes (whole transcriptome) in embryonic and regenerating tails of the Tokay Gecko. F Heatmap of 236 significant differentially expressed genes (toolkit genes only) in embryonic and regenerating tails of the tokay gecko

We performed gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis using gene set enrichment analysis (GSEA) of all genes expressed at each stage (Fig. 2). Immune system-related GO terms are enriched at 4 dpa, while muscle activity-related and muscle development-related GO terms are depleted (Fig. 2B). Bone remodeling-related and epidermis development-related GO terms are also enriched at this stage. In the 8 dpa regenerating tail, only collagen fibril organization and cartilage development are enriched, while epidermis development is depleted. At 16 dpa, many developmental processes-related GO terms are enriched (e.g., pattern specification process, embryonic morphogenesis, WNT Signaling pathway, cell fate specification, response to BMP, somite development). By contrast, immune system-related GO terms are depleted. At 20 dpa, muscle-development-related GO are becoming enriched together with extracellular matrix organization terms, while the developmental-processes-related GO terms are maintained. Finally, at 28 dpa, the muscle-activity-related and muscle-development-related GO terms, that were depleted at 4 dpa, are now enriched.

Fig. 2

Gene set enrichment analysis reveals biological processes and pathways enriched across stages of tail regeneration and in the embryonic tail. A Biological process enrichment in regenerating tail vs. embryonic tail. B Biological processes enrichment at different stages of tail regeneration. C Pathway enrichment in regenerating tail vs. embryonic tail. D Pathway enrichment at different stages of tail regeneration. ET embryonic tail, RT regenerating tail. The more extensive results are shown in Additional file 7: Fig. S6, Additional file 8: Fig. S7, Additional file 9: Suppl_ 2 and Additional file 10: Suppl_ 3

The KEGG pathway analysis gives a similar pattern (Fig. 2D). Thus, immune-system-related pathways are enriched at 4 dpa, as are osteoclast differentiation, TNF signaling pathways, pathways in cancer, and steroid biosynthesis. At 8 dpa, some immune-system-related pathways, such as NFKB and IL-17 signaling pathways, together with pathways in cancer and the TNF signaling pathway are depleted in the of regenerated tail. The only two enriched KEGG pathways at this stage are complement and coagulation cascades and ECM receptor interaction. At 16 dpa, pathways related to stem cells and cancer development are enriched, together with melanogenesis and axon guidance, while immune-system-related pathways are depleted. Later on, only cardiac muscle contraction pathway that is enriched in 20 dpa of regenerated tail while maintaining pathways that are enriched in the previous stage. Finally, the suppression of immune-system-related pathways is still happening in the 28 dpa of regenerated tail.

The regenerating adult tail has a different transcriptome to the embryonic tail

In order to determine whether adult regeneration involves the same mechanisms as embryonic tail development, we performed mRNA-seq on the embryonic tail (16 days post oviposition [dpo]; N = 3, Fig. 1A; stage based on Ref. [26]) compared it with the regenerating tail at all stages (Fig. 1C). The heatmap of significantly differentially expressed genes (embryonic tail vs. all stages of regenerating tails) shows major differences in transcriptomes (Fig. 1E). Notably, genes in clusters 6 and 9 (Fig. 1E) are strongly expressed in the embryonic tail but not in any stages of the adult regenerating tail.

We then looked at the so-called developmental toolkit genes (developmental patterning genes). These genes encode transcription factors, signaling molecules, transmembrane receptors, and other proteins that play a crucial role in embryonic pattern formation. They constitute 484 genes according to Ref. [27]. We found 236/484 of these developmental patterning genes among the significant differentially expressed genes (i.e., including both up- and down-regulated) in all comparisons (Fig. 1F; see for example clusters 3 and 5 of the heatmap in Fig. 1F).

Then, to focus only on comparable stages of histogenesis, we compared the embryonic tail transcriptome with that of 20 and 28 dpa regenerating tails only. We chose this narrower comparison because the embryonic tail and regenerating tail at these stages all show early stages of tissue differentiation. The narrower comparison yielded 3983 differentially expressed genes (Fig. 1C).

Among these, 1996 were significantly enriched in the 16 dpo embryonic tail whereas 1987 genes were significantly enriched in the 20 dpa regenerated tail (Fig. 1C). Similarly, in the comparison of the embryonic tail vs. 28 dpa regenerating tail, 3637 genes were significantly different in expression. Of these, 1599 were significantly enriched in the embryonic tail, and 2038 genes were enriched in the 28 dpo regenerating tail (Fig. 1C). These two comparisons share 1446 differentially expressed genes. Of these, 534 genes are uniquely regulated in the embryonic tail vs. the 20 dpa regenerating tail, and 363 are in the embryonic tail vs. the 28 dpa regenerating tail (Fig. 1D). In summary, these data show major differences between embryonic tail and regenerating adult tail transcriptomes.

We also characterized gene ontology (GO) and KEGG pathways for the for both comparisons of the embryonic tail and the adult regenerating tail (Fig. 2A and C). Both comparisons have a similar GO result: the embryonic tail is dominated by developmental process-related terms, while the adult regenerating tails, of both stages, are dominated by immune-system-related terms and epidermis development-related terms (Fig. 2A). Major differences included, in the embryonic tail vs. 20 dpa regenerating tail, a comparable enrichment for genes associated with spinal cord development, anterior/posterior pattern specification, and embryonic pattern specification. By contrast, the same GO terms are dominant in the embryonic tail when compared with the 28 dpa regenerating tail. Some terms related to tissue morphogenesis or axonogenesis are equally enriched in the embryonic tail vs. 28 dpa regenerating tail. By contrast, these same terms are dominant in the embryonic tail if compared to the 20 dpa regenerating tail. Moreover, muscle structure development-related and extracellular matrix organization terms are dominant in the 28 dpa regenerating tail compared to the embryonic tail.

KEGG pathway is analysis consistent with the GO enrichment analysis. Thus, the embryonic tail is dominated by expression of developmental process-related pathways while the regenerating tail is enriched in immune system-related pathways (Fig. 2C). The term “pathways in cancer” is dominant in the 20 dpa regenerating tail vs. the embryonic tail (Fig. 2C). Finally, the term “signaling pathways regulating pluripotency of stem cells” is dominant in the embryonic tail vs. the 20 dpa regenerating tail comparison, while this term is equally enriched in embryonic tail vs. 28 dpa regenerating tail (Fig. 2C).

Different cell populations in the regeneration blastema and embryonic tail bud

To further test the hypothesis that adult regeneration is based on the reactivation of embryonic developmental pathways, we compared cell populations in the regenerating and embryonic tails. We used single-cell RNA sequencing (scRNA-seq) on the following six tissues: 3 dpo embryonic tail bud, 7 dpo embryonic tail with tail bud, and the regenerating tail blastemata of 16, 20, 24 and 28 dpa (Fig. 3A; N = 1 per sample). We identified 13 cell clusters (Fig. 3B) based on differentially expressed genes (Fig. 3D). Significantly, the embryonic tail bud did not, for the most part, cluster with the regenerated tail blastema, even though there are some mixed cell clusters (Fig. 3C).

Fig. 3

Different cell populations in the regeneration blastema and embryonic the tail bud. A Samples used for single-cell RNA sequencing (boxed area). Key: dpa, days post autotomy; dpo, days post oviposition. B UMAP plot showing cells in the regeneration blastema and embryonic the tail bud clustered by cell type. C UMAP plot of cells in the regeneration blastema and embryonic tail bud clustered by stages. D Genes making up the expression profiles of the different cell types in B. The gene symbols highlighted in blue represent human gene orthologs identified based on gene descriptions from the Sphaerodactylus townsendi gene identifier. The mapping of gene symbols and corresponding gene identifiers in Sphaerodactylus townsendi genome is provided in Additional file 11: Table S1. E Cell fate probability estimation of embryonic tail bud. E1, fate probabilities of embryonic tail bud cell clusters towards estimated terminal states; E2–E5 fate commitment of each cell cluster in regenerating tail blastema towards terminal states. F Cell fate probability estimation of regenerating tail blastema. F1, fate probabilities of regenerating tail blastema cell clusters towards estimated terminal states; F2–F6, fate commitment of each cell cluster in regenerating tail blastema towards terminal states. G Pseudotime analysis of regenerating tail blastema. G1, distribution of pseudotime values of all cell clusters of regenerating tail blastema; G2, distribution of pseudotime values of mesenchymal lineage cell clusters of regenerating tail blastema

We also estimated the terminal state, and cell fate probability, of cell clusters. In the embryonic tail bud, mesenchymal cells were identified as the initial state (Fig. 3E1). They have a high probability of becoming (as terminal states) chondrocytes and myocytes (Fig. 3E2 and E3), a medium-to-low probability of becoming endothelial cells (Fig. 3E4), and a very low probability of becoming nerve cells (Fig. 3E5). In the regenerating blastema, we estimated stromal cells (resident stem cells in mesenchymal lineages) and fibroblasts to be in the initial state (Fig. 3F1). However, the terminal states in the regenerated tail blastema predominantly develop from within their own lineage (Fig. 3F2–6). Notably, keratinocytes develop only from the keratinocyte cell cluster (Fig. 3F7). We explored these results further using pseudotime estimation (Fig. 3G1). In this analysis, keratinocytes were the earliest of all cell clusters (Fig. 3G1). However, among the mesenchymal lineages alone, the stromal cells were the earliest (Fig. 3G2).

Posterior HOXC genes show colinear activation in the regenerating tail

HOX genes are transcription factors that play a key role in embryonic axial patterning [28,29,30] and, potentially, in adult tissue regeneration [31, 32]. Genes from all HOX clusters were expressed in the embryonic tail, but only hoxa5, hoxa13, hoxb9, hoxb13, and hoxc13 were significantly upregulated during regeneration (Fig. 4A–D). There was minimal evidence in the bulk transcriptomes of any statistically significant upregulation of hox genes during regeneration (Additional file 1: Fig. S1). However, visual inspection of the graphs (Fig. 4A–D) appeared to show that the posterior HOXC genes hoxc10–13 might be enriched in the regenerating tail. Therefore, we used in situ hybridization on tissue sections to examine this in more detail at the tissue level.

Fig. 4

HOXC genes are activated during adult tail regeneration in a temporally collinear pattern. A–D Expression profiles of HOX genes in the embryonic tail and adult regenerating tail. E Caudal HOXC genes are sequentially expressed during tail regeneration. In situ hybridization on tissue sections. For high magnification views of these figures, see Additional file 3: Fig. S3. Key: myb, myoblast; myt, myotube; mys, myoseptum; epi, epidermis

We found that these posterior HOXC genes are all expressed in the developing skeletal muscle, and the wound epithelium, of the regenerating tail (Fig. 4E). However, the posterior HOXC genes differed in the stage of the onset of their expression (Fig. 4E). Thus, at 16 dpa, weak expression of hoxc10 was observed in the wound epithelium, and in scattered mesenchymal cells, which were myod⁺ myoblasts (Fig. 4E1, Additional file 2: Fig. S2). There was no expression of hoxc11–13 genes at this stage (Fig. 4E5, E9, and E13). At 20 dpa, we observed the expression of hoxc10–12 in both the wound epithelium, and in the metamerically arranged, developing skeletal muscles, consisting of myotubes separated by intermuscular septa (Fig. 4E2, E6, and E10). Hoxc13 was not expressed at this stage (Fig. 4E14). At 24 dpa, all posterior Hoxc genes are expressed in the wound epithelium and in newly formed skeletal myotubes with myosepta (Fig. 4E3, E7, and E11) even though Hoxc13 was weakly expressed (Fig. 4E15). At 28 dpa, only hoxc12 and hoxc13 expression can still be observed in the wound epithelium and regenerating muscles (Fig. 4E12 and E16). Further figures showing HOXC gene expression are provided in Additional file 3: Fig. S3.

Chondrogenesis in the regenerating tail is uncoupled from segmentation

The axial skeleton of the normal tail of the gecko is made up of segmentally arranged vertebrae. In the regenerating gecko tail, however, the vertebrae are replaced by a continuous, unsegmented cartilage tube [33]. The absence of vertebrae in the regenerated tail is a major unexplained phenomenon in gecko regeneration biology. We wanted to determine whether genes associated with chondrogenesis in the embryo (shh, sox9, wnt5a, wnt3, bmp1, bmp3) are expressed during tail regeneration. We also looked at the expression of pax1, which mediates notochordal signals for somite (sclerotome) and vertebra differentiation [34].

Our transcriptome data showed low expression of chondrogenesis-related genes during early regeneration compared to the embryonic tail, excepting wnt3a and bmp3, which had comparable expression levels to the embryo sample (Fig. 5A). At later stages, only wnt5a and bmp1 showed equal or higher expression than the embryonic tail, with pax1 expression remaining low at all stages. In situ hybridization revealed sox9 expression in the cartilage tube from 20–49 dpa and in myotubes at 20 dpa (Fig. 5B). At 49 dpa, sox9 was strongly expressed throughout the cartilage tube (Fig. 5B7 and B8). Wnt5a was expressed in the cartilage tube, myotubes, and dermis at 20 and 49 dpa (Fig. 5C). At 49 dpa, bmp3 was expressed in the proximal part of the cartilage tube (Fig. 5D). Col2a1, a cartilage marker, was expressed specifically in the cartilage tube at 24 dpa (Fig. 5E). Wnt3 was expressed at 20 and 30 dpa in myotubes, pre-cartilage of the cartilage tube, and epidermis (Fig. 5F).

Fig. 5

Chondrogenesis in the regenerating tail is uncoupled from segmentation. A Expression profiles of chondrogenesis genes in the embryonic tail (left) and during adult tail regeneration (right). One asterisk (*) shows significant differential expression in stage-wise comparison. Two asterisk (**) shows significant differential expression compared to embryonic tail 16 dpo. Significant differential expression (DE) genes with adjusted p-value < 0.01 and L2FC ≥ 1. B Expression of sox9 in the regenerating tail. epi epidermis, ct cartilage tube, my myotube, mys myoseptum. C Expression of wnt5a in the regenerating tail. D Expression of bmp3 at later stages of tail regeneration. E Expression of col2a1 in the 24 dpa regenerating tail. F Expression of wnt3a at 20 dpa regenerating tail

Tail regeneration in the gecko involves a unique pattern of skeletal muscle development

Unlike the unsegmented cartilage tube, the regenerating tail muscles are arranged segmentally. They form a longitudinal series of myomeres composed of myotubes, and separated at intervals by fibro-connective tissue myosepta [35, 36]. We wanted to examine whether muscle development in the regenerating tail resembles the development of axial muscles in the embryo. To do this, we looked at the transcriptome profiles of myogenesis genes, namely: pax3, pax7, myod1, myog, ckm, and tnc. We also wanted to know whether the segmentation of the muscles was based on the activity of lfng and hes7, two genes that are involved in embryonic somite segmentation [37, 38]. We also looked at paralogs of hes7, namely hes1, hes2, hes5, and hes6 to see if they have acquired a function in regeneration (e.g., by neofunctionalization).

In stagewise comparisons, pax3 shows statistically significant upregulation in the transcriptome at 20 dpa (Fig. 6A). Pax7, myod1, myog, and ckm all showed a significant increase in transcript abundance at stages 20–28 dpa. At 28 dpa, myog, ckm, and myod1 were all significantly upregulated. All of the somite segmentation genes (Fig. 6B) except hes2 showed lower expression in the regenerating tail than in the embryonic tail. The other major finding was that lfng was significantly upregulated at 28 dpa.

Fig. 6

Tail regeneration in the gecko involves a unique pattern of skeletal muscle development. A Expression profiles of myogenesis genes in the embryonic tail (left) and during adult tail regeneration (right). One asterisk (*) shows significant differential expression in stage-wise comparison. Two asterisk (**) shows significant differential expression compared to embryonic tail 16 dpo. Significant differential expression (DE) genes with adjusted p-value < 0.01 and L2FC ≥ 1. B Expression profiles of somite segmentation genes in the embryonic tail (left) and during adult tail regeneration (right). C Expression of pax3, pax7, and myod1 at 8 dpa in the regenerating tail. Key: adp adipose tissue, amc adult muscle cells, pbl pre-blastema, wep wound epithelium, epi epidermis. D Expression of pax3, pax7, and myod1 at 16 dpa of the regenerating tail. Key: myb myoblast. E Expression of myog, tnc, and ckm at 24 dpa in the regenerating tail. Key: myt myotube, mys myoseptum. F Expression of lfng in the regenerating tail. Key: drg dorsal root ganglion. G The expression of hes2 in the regenerating tail. Key: bvl blood vessel

We then used in situ hybridization to localize the gene expression. At 8 dpa, we found expression of pax3, pax7, and myod1 in scattered cells in the putative newly formed blastema beneath the wound epidermis (Fig. 6C). The expression of pax3 and pax7 can still be observed in the myod1⁺ myoblast clusters at 16 dpa (Fig. 6D). These clusters were present in the regions of the blastema that lacked col1a2 expression (Fig. 7C; Additional file 4: Fig. S4C). Col1a2 is a specific marker for the connective tissue of the early blastema; it is not strongly expressed in the adult tissue (Fig. 7C; Fig. S1C). Expression of myog, ckm, and tnc in myotubes first appeared at 24 dpa (Fig. 6E). Myotubes at this stage, like the myoblasts before them, lack col1a2 expression; however, col1a2 was expressed in the connective tissue myosepta (Additional file 4: Fig. S4C). Furthermore, lfng was expressed in myoblasts at 16 and 20 dpa, and myotubes at 24 and 28 dpa (Fig. 6F). Hes2 was expressed in myotubes and blood cells at 24 and 28 dpa (Fig. 6G); no expression of hes6 was seen in the regenerating tail (Additional file 2: Fig. S2).

Discussion

Caudal fin regeneration in the zebrafish [8], limb regeneration in the axolotl [13, 14], and tail regeneration in the anole lizard [23], all involve the reactivation of molecular pathways that are normally active during embryonic development. In the regenerate tail of adult geckos, vertebrae are replaced by a simple cartilage tube, while the neural tube is replaced by a sleeve of ependyma lining the cartilage tube. In evolutionary terms, this incomplete tail may be sufficient as a functional replacement for the lost tail part. The mechanism of regeneration in the gecko tail may therefore be imperfect, but nonetheless adaptive, given the major advantage of autotomy in escaping death by predator. We have examined the hypothesis that tail regeneration in the adult tokay gecko is a re-run of embryonic tail development (in terms of the genes and cell populations involved), as it may be in the regeneration of the axolotl limb and zebrafish caudal fin.

Our results support the idea that regeneration does not fully recapitulate embryonic development [23, 39, 40]. Instead, the regeneration process appears to be dynamic and adaptive to varying conditions as a response to injury [41, 42]. When we compared the embryonic tail and adult regenerating tail in the gecko, we found that their transcriptomes were quite different. Less than half of the development-related genes expressed in the embryonic tail were also expressed in the regenerating tail. Thus, the regenerating tail shares a limited subset of differentially expressed genes and pathways with the embryonic tail. Our scRNA-seq analysis reveals important differences in the cell populations in the regeneration blastema compared to the embryonic tail bud in the gecko. The regeneration blastema is dominated from 16 dpa onwards by fibroblasts. By contrast, the tail bud shows a substantial population of pluripotent stem cells.

The early regeneration transcriptome of the Tokay Gecko is dominated by immune-related genes and genes associated with wound epithelium formation and bone remodeling (Fig. 2). Our finding is in line with transcriptomic studies on the initial stage of the tail regeneration in the common wall gecko (Hemidactylus frenatus) [43], although we did find a many more differentially expressed genes in our study. Immune-related genes are essential during the early stages of regeneration [44,45,46] and bone remodeling genes are essential for bone regeneration [47, 48]. At the onset of blastema formation (16–20 dpa), we find mass upregulation of genes associated with developmental pathways, and down-regulation of immune-related genes. For example, genes in the WNT signaling pathway are enriched at this time. This pathway is not only involved in developmental patterning [49,50,51] but is also essential for tail regeneration in lizards, especially in promoting cell proliferation and proximodistal outgrowth in the blastema [52]. We also found that Hedgehog signaling pathways were enriched at this stage. Hedgehog signaling is critical for many developmental processes in vertebrates [53] and, in response to BMP signaling, is essential in initiating chondrogenesis in the regenerating lizard tail [33, 54]. Finally, the 28 dpa transcriptome is characterized by maturation and functional development of the skeletal muscle.

Based on cell fate probability analysis on scRNA-seq data, we find no common pool of mesenchymal precursors in the regenerating adult tail blastema. Rather, the terminal states develop mostly from their own resident stem cells (stromal cells), which are stem cells in adult mesenchymal tissues [55]. Furthermore, we find that the development of keratinocytes is different from that of from mesenchymal cell clusters. This difference makes it unlikely, we believe, that the epidermis contributes in any way to the blastema (for example, epithelial–mesenchymal transition).

Taken together, these findings support the concept of two major pathways of regeneration in vertebrates (Table 1). One is exemplified by caudal fin regeneration in zebrafish and limb regeneration in axolotls. In this type of regeneration, we suggest that the blastemas are relatively small—of the order of 1 mm. This, according to Wolpert, is the size of a classic developmental field, such as an embryonic limb bud or tail bud, in which positional values are specified de novo [56]. In this type of regeneration, and in a classical developmental field, development gene expression is early, distal, and occurs in a small area. For example, in zebrafish tail regeneration, there is an apical growth zone in which there is a distal-to-proximal gradient in differentiation [8], very much like a developmental field such as the progress zone of the embryonic limb bud [57].

Table 1 The tokay geckos and other lizards show a pattern of regeneration that is quite different from a classical developmental field. Note that the blastemas of the regenerating caudal fin, and the regenerating limbs in the axolotl, are similar in properties and size to the developmental fields of embryos. By contrast, the regenerating tail blastema of geckos and other lizards is quite different from the developmental field

The other pathway of regeneration in vertebrates is exemplified by tail regeneration in geckos and other lizards. This appears to differ from regeneration in other vertebrate models [23, 24], such as the urodele limb and zebrafish caudal fin. According to our observations, and examination of the literature, it seems that the regeneration blastema in lizards is much larger than a classical developmental field or the regeneration blastema of zebrafish caudal fin [8] and axolotl limb [13, 14]. However, a study of axolotl regeneration revealed that it is not the size that reduces the regenerative ability of the axolotl but rather changes taking place at metamorphosis and during the aging process [61]. Nevertheless, we assume that the characteristic features of lizard tail regeneration are the fact that its developmental gene expression and cell proliferation are distributed throughout the blastema [23] and not restricted to a small, apical growth zone. Further, those developmental genes that are expressed in tokay gecko tail regeneration are activated relatively late (Fig. 7A). Finally, another line of evidence that there is no apical growth zone in the tokay gecko blastema is that we found only col1a2⁺ connective tissue cells under the wound epidermis (Fig. 7C; Additional file 4: Fig. S4C).

Fig. 7

A summary of selected key features of tail regeneration in the tokay gecko. A Schematic overview of biological process enrichment in the transcriptome at selected stages of tail regeneration. The height of each line represents (schematically) the relative enrichment or the process. The greyscale drawings at the bottom represent the histological structure of the regenerating tail at that stage. B Schematic illustration highlighting the major differences in the dominant pathways in the regenerating tail vs. the embryonic tail. C There is no apical growth zone in the regenerating tail of the tokay gecko. The tissues immediately under the wound epithelium (pre-blastema) are dominated by col1a2⁺ connective tissue cells. Key: adp adipose tissue, amc adult muscle cells, bls blastema, bvl blood vessel, ep epidermis, ery erythrocytes, leu leukocytes, myb myoblast, pbl pre-blastema, scb wound scab, wep wound epithelium

We suggest there is no de novo specification of positional information in the blastema of the tokay gecko. Rather, regeneration depends on pre-existing positional values in the adult cells and tissues. This hypothesis was supported by HOX gene expression in the regenerating tail of the tokay gecko. Hox genes are known to be important in regeneration. In caudal fin regeneration in the zebrafish, for example, hoxc13a and hoxc13b are expressed early and distally in the blastema [58]. In this study, we find the HOXC genes show temporal collinearity, but not spatial collinearity, in the regenerating gecko tail, principally in the cells of the skeletal muscle lineage. HOX gene expression during embryonic tail development is spatially and temporally collinear [62, 63], meaning that HOX genes become transcriptionally active in a sequence reflecting their position along the chromosome.

In mammals, at least some HOX genes are expressed in the adult, often in the same position-specific pattern or “hox code” that they showed during embryonic development [64,65,66]. Furthermore, fibroblasts from adult humans [67] and skeletal muscle from adult mice [68] have “positional memory” that reflects their anatomical location. It would be interesting to know whether HOXC expression patterns in the regenerating tail are determined by existing positional information in the intact tail before it was shed. This could arise, for example, if there is differential expression of hox genes along the normal tail. It would also explain why the regenerating gecko tail replaces a similar number of muscle segments that were present before the tail was amputated. This is the case in the axolotl limb, which regenerates only those limb parts appropriate to the proximodistal level of the amputation plane (reviewed in Ref. [69]). A similar situation is seen with the zebrafish caudal fin regeneration in which only the appropriate missing parts are replaced [70]. Finally, we suggest that late expression of hoxc13 that we observed here represents the finalization of the patterning process in gecko tail regeneration since it is known that hoxc13 is essential for terminating tail development [71].

The absence of a development field in the regenerating tail blastema of the gecko is also consistent with the origin of the blastemal cells from the resident stem cells of the original tail. In the axolotl, de-differentiation of the existing mature tissues adjacent to the amputation site is a major source of precursor cells [13]. Existing fibroconnective tissue of the stump contributes mostly to muscle and cartilage in the regenerating limb of the axolotl [13]. De-differentiation has also been described in tail regeneration in the mourning gecko (Lepidodactylus lugubris) [72]. Our RNA velocity analysis on scRNA-seq data, however, are not consistent with the dedifferentiation of adult cells into stem cells; instead, they are consistent with the origin of the blastema from lineage-restricted resident stem cells. Furthermore, our pathway analysis based on the transcriptome data revealed that the WNT signaling and pluripotency pathways at the blastema stage do not promote pluripotency (Additional file 5: Fig. S5). Instead, the enriched genes in these pathways appear to drive differentiation. One could argue that these are only bioinformatic inferences. On the other hand, they are based on the transcriptomes of thousands of single cells from multiple stages of regeneration.

We propose a model in which the regenerating tail blastema in the tokay gecko is composed of multiple stem cells that were already committed to a particular lineage at the start of regeneration. For instance, our gene expression profiling showed that myogenesis during gecko tail regeneration involves an early activation of satellite cells (muscle stromal cells). This is supported by an early expression of the muscle lineage markers pax3, pax7 and myod1 at 8 dpa. Our reasoning is that pax7 and myod1 are co-expressed during satellite cell activation [73]. Based on previous studies [11, 74], and on our data, it is likely that cells that express pax3⁺, pax7⁺ and myod1⁺ will aggregate and differentiate into mononuclear myoblasts at 16 dpa. The early expression of pax3⁺ and pax7⁺ is also seen in the regenerating tail blastema of the green anole lizard Anolis carolinensis [23] and Schlegel’s Japanese gecko, Gekko japonicus [75].

The fact that the skeletal muscle in the regenerated lizard is arranged into segments (bundles of myotubes separated by myosepta) raises the question of whether regeneration involves a re-activation of embryonic segmentation genes. Such genes could include lfng and hes7 which are necessary for somitogenesis in the embryo [37, 76]. Our results suggest that only lfng is expressed in the muscle lineage of the regenerating gecko tail. Lfng is a modulator of the Notch signaling pathway [77], and this pathway is known for its role in the self-renewal of muscle resident stem cells (satellite cells) through direct regulation of pax7 [78]. Among the Hes genes, only hes2 was expressed in gecko tail regenerates, in the skeletal muscle and hemopoietic lineages. And even hes2 is unlikely to be involved in muscle segmentation because we only find it expressed at 24 dpa, after the muscles have already segmented.

Our suggestion, of a lack of Hes gene involvement in muscle segmentation in the regenerating gecko tail, is consistent with a study on axolotl tail regeneration [79]. That study found evidence that lfng⁺ multipotent stem cells (which they called “asomitic” stem cells) give rise to segmented structures during regeneration without the involvement of hes7 [79]. We suggest that muscle segmentation in the regenerating tail relies on positional memory in satellite cells; these “remember” the segmental pattern of the original muscle. This suggestion is supported by a study of mouse cell lines, showing that muscles and their associated satellite cells have positional memory of their developmental origin and anatomical location and that this positional memory is regulated by hoxa10 [68].

Another supporting evidence that there is no de novo patterning in regenerating blastema is the separate development of unsegmented cartilage tubes and segmented muscle. The de novo patterning during axolotl tail regeneration is able to create both segmented vertebrae and segmented muscles, even without the involvement of hes7 [79]. This is not what we see in the regenerating tail of the tokay gecko. The lack of segmentation of the cartilage tube in the gecko might be caused by the absence of pax1 expression, even though shh is expressed in the ependymal lining [33]. Pax1 is needed to mediate the notochordal signal, SHH, which induces sclerotome development [80] and is also essential for further development of the vertebral column [34]. Despite the uncoupling of development of the cartilage tube, and muscle regeneration, their patterning is nonetheless guided by Wnt signaling genes (wnt3 and wnt5a).

Given these differences between the gecko tail regeneration blastema, and other regeneration and developmental systems, it appears that tail regeneration in the gecko and other lizards is an evolutionarily derived process not described in other vertebrates. Our proposed model is that pattern formation in the blastema is based on the existing positional values of existing resident stem cell populations. This model is consistent with the finding that the cartilage tube in the regenerating lizard tail is derived from stromal cells resident in the vertebral periosteum [54] and with the finding that regenerating muscle might derive from stromal cells in the muscle of the tail stump [72].

Studies on animal regeneration have often been difficult to translate to human medicine because they apparently require the reactivation of developmental processes in a tiny growth zone and also require pluripotent stem cells—which are lacking in adult humans. Our study shows that regeneration can take place using resident stem cells which are present in adult human tissues. Therefore, further translation studies in mammalian regeneration could be informed by the single-cell analyses and other data reported here. Another potential translational link of our data might be with cancer research. It has already been shown that regeneration shares many pathways with cancer [25, 81]. In this study, we also find several pathways involved in cancer are enriched in the regenerating tokay gecko tail, especially early stages (4 dpa). The link between regeneration and cancer is enigmatic and worth exploring in future studies.

Conclusions

In conclusion, our study suggests that tail regeneration in the tokay gecko does not closely resemble embryonic tail development. Instead, our data suggest an alternative hypothesis: the activation of resident stem cells that then develop according to pre-existing positional information. Additionally, regeneration in the tokay gecko does not seem to occur through an apical growth zone, as seen in zebrafish and axolotl. Our findings provide valuable insights into the diversity of regenerative mechanisms in vertebrates. However, they are based on bioinformatics analysis and in situ hybridization. Therefore, further validation through functional studies, including gene knockdown or overexpression, lineage tracing experiments, and transplantation experiments, is necessary.

Methods

Animal collection and maintenance

Thirty-nine adult tokay geckos were captured in the vicinity of the Faculty of Biology, Universitas Gadjah Mada, Yogyakarta, Indonesia. They were acclimatized for 2 weeks, two males to one female, in a stainless-steel terrarium and fed various species of live crickets and sometimes cockroaches, twice a day. They were maintained at an ambient temperature and sprayed once a day with water. The geckos were also basked under natural sunlight once every 2 days for 1 h in the period 09:00–11:00. Twenty-one adult tokay geckos were used for collecting tissue from the internal organs and from regenerated tails, while the other 18 tokay geckos were bred to produce eggs for embryo sampling.

Experimentation and samples collection

Regenerated tail samples

Twenty-one adult tokay geckos were anesthetized with intramuscular ketamine injection (50–75 mg/Kg body weight) administered in the thigh prior to the autotomy procedure. Autotomy was performed by gently restraining the gecko manually while leaving the tail accessible. Using the thumb and index finger, a firm pinch was applied to the required autotomous region (every segment after the fifth caudal vertebra) of the tail causing the gecko to spontaneously shed its tail. The wound was painted with Betadine® iodine antiseptic solution but no dressing was applied. After the autotomy procedure was performed, the geckos were allowed to regenerate a new tail. The geckos were distributed into experimental groups based on the number of days post autotomy, each consisting of three geckos. After ketamine anesthesia, the tails of geckos in each group were re-autotomized 2 cm cranially from the previous autotomy site at 0, 4, 8, 16, 20, and 28 days post autotomy (dpa). The regenerated tail samples were washed and further handled in phosphate-buffered saline (PBS, 4 °C) and subsequently fixed with cold RNAlater® stabilization solution (Invitrogen).

Embryo samples

Eighteen adult tokay geckos were grouped and bred as described above. The terraria were checked daily for the presence of gecko eggs. The eggs were left in the terrarium for the required number of days of incubation at ambient temperature. The embryos were harvested into cold PBS and subsequently fixed with RNAlater® (4 °C).

Organ samples

Three adult tokay geckos were euthanized by an overdose of intramuscular ketamine (150 mg/Kg body weight, intramuscular injection). The geckos were then dissected to harvest organs, namely, the pancreas, liver, brain, heart, skeletal muscle, whole skin, kidneys, and lungs. These eight organ samples were cut up in cold PBS and placed in RNAlater®(4 °C).

RNA isolation and cleaning for transcriptome sequencing

The tissue/organ samples were taken from RNAlater® solution and washed with sterile PBS. They were then put into 1.5 mL Eppendorf tubes together with 1 mL Trizol and two steel balls (⌀ 4 mm). The tubes were then shaken with a TissueLyser II (Qiagen) at a frequency of 30 per second for 1–2 min. The resulting tissue homogenates were triturated with a hypodermic needle (21G = 0.8mm) attached to a 1-mL syringe until no tissue clumps were visible. The samples were then incubated for 30 min and subsequently centrifuged for 10 min at 12,000 RCF (relative centrifugal force).

The supernatants were transferred into new tubes and 0.2 mL of chloroform added to each tube. The sample mixture in the tubes was incubated for 2 to 3 min after 11 s of vigorous hand shaken and then centrifuged for 15 min at 16,000 RCF. The colorless aqueous phase of each mixture was transferred to a fresh tube and 0.7 mL of propan-2-ol (isopropyl alcohol, isopropanol) added to each tube. The samples were then centrifuged at 16,000 RCF for 15 min after 10-min incubation. The supernatants were discarded and the RNA pellet washed with 1 mL of 75% ethanol per tube. The RNA pellet-75% ethanol mixtures were centrifuged again for 5 min at 7500 RCF. The supernatants were discarded and the RNA pellets allowed to dry for 5–10 min. The dry RNA pellets were dissolved in 50 µL of RNase-free water and incubated at 55° C for 10 min. The total RNA of each sample was then cleaned with Qiagen RNeasy Mini Kit (Cat. No. 74104) according to the manufacturer’s protocols.

Transcriptome analysis

The library preparation and transcriptome sequencing were done by the European Molecular Biology Laboratory (EMBL), Heidelberg, Germany with in-house protocols. The sequencing platform used was Illumina NextSeq 2000 with a read length of 2 × 100 bp.

Transcriptome data analysis was carried out within the compute resources from Academic Leiden Interdisciplinary Cluster Environment (ALICE) provided by Leiden University. The sequencing data (paired-end reads, fastq files) were examined with FastQC to check the sequence quality. The sequence adapters were subsequently cut using Trim Galore [82] before transcriptome assembly. Later on, the transcriptome assembly was done using Trinity RNA-Seq de novo assembler [83, 84]. Finally, the transcriptome assembly results were then evaluated using rnaQUAST [85] with the Gekko japonicus genome as a reference, BUSCO [86, 87], and FastQC.

Functional annotation was initiated by predicting coding region within the Trinity assembly result using Transdecoder. The predicted protein fasta files were then searched against databases (NCBI, pFAM, Swissprot) using the Trinotate functional annotation pipeline [88]. The tools made use of NCBI Blast + , HMMER [89], TMHMM [90], and SignalP [91]. After functional annotation was completed, the transcripts of both regenerated and embryonic tail samples were quantified using Salmon 1.6.0 [92] with Tanuki pipeline. The quantification files were then analyzed with differential expression analysis with 3DRNAseq [93] using limma (R package) pipeline for differential expression comparison [94]. Gene Ontology and KEGG Pathways were analyzed using gene set enrichment analysis [95] with clusterProfiler 4.0 [96] of which results were visualized with ggplot2 (v3.4.0) [97].

Tissue dissociation and ACME fixation for single-cell sequencing

Samples were cut into pieces and were transferred into 15-mL tube containing cold 1% trypsin (Sigma Aldrich®; T7409), 1 mM EDTA (380 mg/L) in Hanks’ balanced salt solution, without Mg and Ca, and with phenol red (Sigma Aldrich®; H9394). Samples were then incubated for 90–120 min on the shaker at 200 rpm with trituration every 30 min. The samples were then centrifuged on 440 RCF (16 rad/min) at 4 °C for 5 min, and the supernatant was then discarded carefully. Acetic-methanol (ACME) solution was made based on García-Castro (2021) [98]. ACME solution without methanol (8.5 mL) was added to the tube containing samples. After 20 min of incubation, 1.5 mL of methanol was then added to each tube and incubated for another 20 min. The dissociated tissue of all samples was then strained with 50µm Corning® cell strainer. The cell suspensions were then centrifuged on 440 RCF (16 rad/min) at 4 °C for 5 min, and the supernatant was then discarded carefully. 1% BSA in PBS were added to resuspend and wash the cells and they were subsequently recentrifuged. One and a half mL of 1% BSA in PBS were the added to resuspend the cells. About 100 µL of dimethyl sulfoxide (DMSO) were added to each tube containing cell suspension which was then snap-frozen with liquid nitrogen.

Single-cell RNA sequencing analysis

The single-cell library preparation was done by the European Molecular Biology Laboratory (EMBL) Heidelberg, Germany, using 10x genomics v3 (28bp). The sequencing was done by using Illumina NextSeq 2000 with a read length of 2 × 50 bp.

Gene expression data were obtained using cellranger-6.1.2 pipelines carried out within the compute resources from the Academic Leiden Interdisciplinary Cluster Environment (ALICE) provided by Leiden University. The genome of Sphaerodactylus townsendi (GCF_021028975.2) [99] was used as a custom genome reference by utilizing “cellranger mkref” pipeline. The transcript quantification and gene expression data matrix were produced by using “cellranger count” pipeline.

Seurat (v5.0.0) [100] was used for the analysis of single-cell RNA-seq (scRNA-seq) datasets, which included standard quality control and pre-processing steps. To integrate datasets from all samples, we applied the reciprocal principal component analysis (RPCA) method, allowing us to minimize batch effects while preserving the biological signatures of each sample. The integrated dataset was then normalized, with cell cycle effects and mitochondrial gene percentages regressed out.

Following normalization, we performed linear dimensional reduction using principal component analysis (PCA). Cell clustering was conducted based on PCA values, and uniform manifold approximation and projection (UMAP) was used for non-linear dimensionality reduction. Marker genes for each cluster were identified through differential expression analysis using the Wilcoxon rank sum test, implemented via the limma package. These marker genes were then cross-referenced with human orthologues to facilitate cell-type assignment.

Cell-type assignments were refined through manual curation, incorporating predictions from the Clustermole package and the SCtype pipeline [101]. This curation process utilized multiple cell marker databases, including PanglaoDB [102], CellMarker [103], MSigDB [104], SaVanT [105], xCell [106], and TISSUES [107].

RNA velocity was estimated from the scRNA-seq data using the Velocyto.py pipeline [108]. The RNA velocity data were integrated with the processed scRNA-seq dataset to estimate cell fate probabilities. This was achieved by implementing the Generalized Perron Cluster Cluster Analysis (GPCCA) algorithm through the CellRank framework [109, 110], in conjunction with Scanpy and scVelo.

Probe synthesis

In this project, all gene names follow the style and nomenclature of the NCBI database for Podarcis muralis (Common Wall Lizard). There was no published annotated genome sequence available for the tokay gecko (Gekko gecko) at the time. The predicted regions of the Gekko gecko genome for the genes were aligned using the NCBI nucleotides database to find orthologues in other lizard genomes. Primers were designed based on the conserved regions of the aligned sequences using blast-n. Using these primers (with approximately 800 bp in length), each gene of interest was amplified by PCR (Sigma-Aldrich; Darmstadt, Germany) from Gekko gecko cDNA. The amplicons were cloned into a plasmid (Escherichia coli vector pCR™II-TOPO™) following the TOPO™ cloning reaction protocol (Invitrogen, Carlsbad). The bacterial clones were subsequently sequenced (BaseClear B.V., the Netherlands). Probe synthesis was carried by labeling antisense RNA probes with dioxygenin-11-UTP (Roche) according to the manufacturer’s protocols (Additional file 13: Suppl_5).

Histology and in situ hybridization

The regenerating tail blastema samples were rinsed in RNAse-free PBS and fixed in cold 4% paraformaldehyde in PBS. The fixed samples were dehydrated with a graded concentration series of methanol starting from methanol 70%, 80%, 90%, and 100% for 2 × 30 min. The samples were cleared 3 × 1 h (Neo-Clear®, Merck-Milliopore). The samples were infiltrated 2 × 1 h molten paraffin (62 °C) and 1 × 8–18 h. The samples we finally embedded with fresh molten paraffin and be solidified in room temperature.

The paraffin blocks were then sectioned longitudinally at 6–8 µm. The sections were placed into poly-L-lysine coated slides and left overnight at 37 °C. Adjacent sections were then processed in situ hybridization protocols based on Acloque et al. [111]. The sections were deparaffinized with xylene 3 × 7 min and rehydrated using ethanol series (100% 3 × 2 min; 90%, 80%, 70% for 1 min each). After that, pre-treatment was performed using PBST (phosphate-buffered saline with Triton) 3 × 5 min, proteinase K (10 ng/ml) 10 min, PBST 3 × 5 min, 4% paraformaldehyde (pFA) 10 min, and PBST 2 × 5 min. Then, the sections were covered with pre-hybridization mix (without probe) for ≥ 1 h at 60 °C and incubated with probe mix (1000–1500 ng/mL) at 60 °C overnight with a cover slip. Next day, the sections were washed using washing solution A (2 × SSC (standard sodium citrate) 0.1% CHAPS (3-[ (3-cholamidopropyl) dimethylammonio]−1-propanesulfonate), 50% Formamide) 3 × 30 min and TBST (tris (tris (hydroxymethyl)aminomethane)-buffered saline with Triton) 3 × 10 min. Anti-DIG-AP detection was carried out using 10% sheep serum in TBS (tris-buffered saline) for 1 h and continued with anti-DIG-AP (1:1000 concentration) in 10% sheep serum in TBST at 4 °C overnight. Finally, the sections were covered with TBST 3 × 10 min and NTT (NaCl Tris buffer with Tween) pH 9.5 3 × 10 min. Staining was conducted using BM purple and was checked regularly under the microscope up to 2 days. If the desired degree of color was achieved, the staining was stopped to prevent overstaining. The samples were then dehydrated using graded ethanol (70%, 80%, 90%, 3 × 100% fast), preserved using Eukit® and coverslip.

Statistical analysis

Statistical analysis for sequencing data was carried out using R or Python according to the analysis pipeline used for each data mentioned in method section. Differential gene expression analysis was performed by comparing each stage with the previous adjacent stage (for example, stage 4 with stage 0, stage 8 with stage 4, and so on). The log2 fold change (L2FC) of gene/transcript abundance was calculated based on contrast groups and significance of expression changes were determined using t-test. p-values of multiple testing were adjusted with Benjamini–Hochberg (BH) to correct false discovery rate (FDR). A gene/transcript was significantly DE in a contrast group if it had adjusted p-value < 0.01 and L2FC ≥ 1.