Fgf10 mutant newts regenerate normal hindlimbs despite severe developmental defects

Abstract

In amniote limbs, Fibroblast Growth Factor 10 (FGF10) is essential for limb development, but whether this function is broadly conserved in tetrapods and/or involved in adult limb regeneration remains unknown. To tackle this question, we established Fgf10 mutant lines in the newt Pleurodeles waltl which has amazing regenerative ability. While Fgf10 mutant forelimbs develop normally, the hindlimbs fail to develop and downregulate FGF target genes. Despite these developmental defects, Fgf10 mutants were able to regenerate normal hindlimbs rather than recapitulating the embryonic phenotype. Together, our results demonstrate an important role for FGF10 in hindlimb formation, but little or no function in regeneration, suggesting that different mechanisms operate during limb regeneration versus development.

Results and Discussion

Fibroblast Growth Factor 10 (FGF10) is well known to be essential for amniote limb development (1, 2). Fgf10 transcripts are expressed throughout the limb mesenchyme, where FGF10 stimulates Fgf8 expression in the apical ectodermal ridge (AER) (3). Fgf10 knockout in mice is lethal and leads to severe truncation of both fore- and hindlimbs. In axolotl, Fgf8 is expressed in the limb mesenchyme rather than the epidermis (4–7), suggesting that FGFs may work differently in urodele amphibians. In contrast to amniotes, urodeles can regenerate their limbs throughout adulthood. While it has been suggested that adult limb regeneration may recapitulate development, it is unclear whether or not FGFs function in similar ways during development and regeneration.

To address the role of FGF10 in limb development and regeneration, we tested its function in the newt Pleurodeles waltl. First, we examined the expression patterns of Fgf8 and Fgf10 using hybridization chain reaction (HCR) in limb buds at the early limb bud stage and blastemas at the late bud stage. Transcripts of both were predominantly expressed in the mesenchyme (Fig. 1 A and B) (5–7), although at apparently lower levels in the blastema than the limb bud. Bulk RNA-seq confirmed that Fgf8 and Fgf10 were expressed in both fore- and hindlimbs (Fig. 1C).

Fig. 1.

Fgf10 mutants have defects only in the hindlimb. (A) In situ HCR of Fgf8 and Fgf10 in the forelimb (FL, st. 34) and hindlimb (HL, st. 40) bud. (B) In situ HCR of P63, Fgf8, and Fgf10 in the HL blastema at the late bud stage. P63 is expressed in the epidermis. The dashed line shows the epithelial and mesenchymal boundary. (C) Fgf8 and Fgf10 expression level in the FL (st. 33 to 34) and HL (st. 39 to 40). (D) Confirmation of phenocopy of Fgf10 crispants using different gRNAs. (E) The genotype of the FL and HL of Fgf10 crispant (gRNA3). Mutant alleles and their occupancy rates are shown corresponding to the FL and HL of each crispant (#1-3). (F) The phenotype of Fgf10 mutant. The arrow and arrowhead show a normally developed FL and a defective HL, respectively. Note that the HLs are not missing above the knee, but one digit is present and bends ventrally. The dashed line indicates the intact digit. (G) Sanger sequence of the wild type and Fgf10 mutant. (H) Fgf8 expression in the HL bud of Fgf10 mutants. (I) Volcano plot for differential gene expression of Fgf10 mutant versus wild type in the HL bud. (J) MicroCT image of the HL of the adult Fgf10 mutant. (a) Control, Fgf10 F1 hetero. (b) Fgf10 F2 mutant. The bracket, arrowhead, and asterisk show the stylopod, fibula/tibia, and digits, respectively.

Next, we performed Fgf10 knockouts using CRISPR-Cas9. In contrast to mouse knockouts, the forelimb of Fgf10 crispants developed normally; however, the hindlimb was defective with digit number reduced to one or two (Fig. 1D). This phenotype was reproduced with multiple gRNAs (Fig. 1D), with a somatic mutation rate of more than 93% in both fore- and hindlimbs (Fig. 1E), inferring that differences in phenotype between them were not due to mosaicism. Furthermore, we generated Fgf10 F2 mutants with a 2-bp deletion (Fig. 1 F and G). Importantly, Fgf10 mutants exhibited either the same phenotype as crispants (longitudinal defect, fewer digits, and absence/hypoplasia of zeugopods, Fig. 1F) or even more severe phenotypes, completely missing the zeugopod and autopod (terminal transverse defect). Even with severe hindlimb defects, forelimbs developed normally. Possible explanations for forelimb versus hindlimb differences include different FGF signaling pathways and/or functional redundancy, as previously suggested for other developmental processes (8, 9).

To investigate downstream genes, hindlimb buds from Fgf10 mutants were individually collected and processed by RNA-seq. Fgf8 (Fig. 1H), other FGF pathway genes, and Sall1 (Fig. 1I) were significantly downregulated in Fgf10 mutants. Sall1 is expressed in distal mesenchymal cells of the limb bud in Xenopus (10), mouse (11), and the regenerating hindlimb in Xenopus (12). Sall1/Sall3 have redundant activity, and the double-mutant mouse exhibits loss of digits in the autopod (13). In addition, Hoxa/d11 and Runx1 were significantly decreased (Fig. 1I). Deletion of Hoxa/d11 results in zeugopod malformations (14), whereas the transcription factor Runx1 is essential for osteogenesis (15). F2 mutants had fewer digits and lacked fibula or tibia formation, contrasting with Fgf10 heterozygotes with proper osteogenesis (Fig. 1J). Based on these results, we suggest that Fgf10 plays two roles: first in autopod patterning by coordinating FGF pathway and Sall genes and second in zeugopod formation through Hox11 and Runx genes.

Given that limb development and regeneration share many common properties, we next examined the role of FGF10 in regeneration. To this end, we amputated either the forelimb and hindlimb of Fgf10 crispants or the hindlimb of F2 mutants. As expected, the forelimbs of Fgf10 crispants regenerated normally. Surprisingly, the hindlimbs of Fgf10 crispants and mutants regenerated normal or near-normal structures after amputation rather than recapitulating their developmental morphology (Fig. 2). Whereas mutants developed severely defective hindlimbs, 61% (n = 11/18) of mutants regenerated completely normal zeugopods and autopods after amputation at the stylopod (Fig. 2 A and B). Overall, regeneration led to an increase in the number of digits in 72% (n = 23/32) of mutants (Fig. 2C). A higher regeneration score was obtained by stylopod amputation, a more proximal site, suggesting that stronger intercalation evokes regeneration (16). The regenerated limb of Fgf10 mutants ossified the fibula and tibia normally (Fig. 2D), suggesting that the regenerated hindlimb is comparable to the wild type.

Fig. 2.

The defective hindlimb in Fgf10 mutants was restored to normal by regeneration. (A) HL regeneration of Fgf10 mutants amputated at the stylopod (#1) or zeugopod (#2). (B) Recovered HL defect by regeneration (Left) and unamputated side (Right) of #1. (C) Summary of restored HL in Fgf10 mutants. The HL which has a severely truncated zeugopod was amputated at the stylopod (first row), and the other was amputated at the zeugopod (second row). The asterisk indicates that these newts did not form blastemas. (D) MicroCT image of regenerated HL of Fgf10 mutant. The bracket, arrowhead, and asterisk show the stylopod, recovered fibula/tibia, and digits, respectively. (E) Expression of limb development-related genes in the Fgf10 mutant blastema. (F) Summary of differential expression analysis of limb development-related genes in the Fgf10 mutant limb bud or blastema compared with the wild type. The down and right arrows indicate significant downregulation (p < 0.05) and no difference, respectively. (G) A model for development and regeneration-specific program in the urodele hindlimb.

To examine downstream gene expression, we performed RNA-seq in hindlimb blastemas of Fgf10 mutants. The expression levels of Fgf10-regulated genes were comparable between the wild type and mutant, suggesting that downstream gene expression was restored (Fig. 2E). Other limb development-related genes including FGF signaling pathway genes were also similar between the wild type and mutant (Fig. 2F). Thus, developing and regenerating urodele amphibian limbs utilize differential Fgf8 induction mechanisms. Reciprocal and feedback loop regulation between Fgf10 expressed in the mesenchyme and Fgf8 expressed in the AER are essential for amniote limb development (1, 2). However, both of these Fgfs are expressed in the mesenchyme of urodeles (4–7), highlighting potential developmental differences between urodele amphibians and amniotes.

One important difference between development and regeneration is the presence of nerves. Nerve-derived factors are known to play a critical role in regeneration (17, 18). These findings together with our results raise the possibility that direct induction of Fgf8 by regeneration cues including nerve-derived factor(s) rather than FGF10 may be key to limb regeneration in urodeles (Fig. 2G).

Our findings highlight both similarities and differences between limb development and regeneration. Bryant and colleagues proposed that the early phase of limb regeneration may involve regeneration-specific steps. Subsequently, there is a gradual transition to an autonomous limb growth and patterning program similar to that in development (19). We find that most of the downstream genes required for limb development do not differ in the blastema of Fgf10 mutants. Accordingly, it is possible that a limited number of genes, including Fgf10, have differential effects at early stages of regeneration, whereas later events are parallel to development (Fig. 2G). Importantly, even though the hindlimb of Fgf10 mutants display marked abnormalities, our results show that the hindlimb has “developmental plasticity” and is able to regenerate a normal limb from a developmentally defective one.

Materials and Methods

P. waltl were obtained from a breeding colony. Experimental details are provided in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Sequencing data have been deposited in the NCBI BioProject (PRJDB15083) (20). All other data are included in the manuscript and/or SI Appendix.