Sonic hedgehog is Essential for Proximal-Distal Outgrowth of the Limb Bud in Salamanders

The developing forelimb has been a foundational model to understand how specified progenitor cells integrate genetic information to produce the tetrapod limb bauplan. Although the reigning hypothesis is that all tetrapods develop limbs in a similar manner, recent work suggests that urodeles have evolved a derived mode of limb dvelopment. Here, we demonstrate through pharmacological and genetic inactivation of Sonic hedgehog (Shh) signaling in axolotls that Shh directs expansion and survival of limb progenitor cells in addition to patterning the limb across the proximodistal and antero-posterior axis. In contrast to inactivation of Shh in mouse or chick embryos where a humerus, radius, and single digit develop, Shh crispant axolotls completely lack forelimbs. In rescuing limb development by implanting SHH-N protein beads into the nascent limb field of Shh crispants, we show that the limb field is specified in the absence of Shh and that hedgehog pathway activation is required to initiate proximodistal outgrowth. When our results are examined alongside other derived aspects of salamander limb development and placed in a phylogenetic context, a new hypothesis emerges whereby the ability for cells at an amputation plane to activate morphogenesis and regenerate a limb may have evolved uniquely in urodeles.

Despite this apparent conservation, previous work has shown that salamanders lack an AER (Sturdee and Connock, 1975;Tank et al., 1977) and that at some point during amphibian evolution localization of Fgfs and Fgf receptors shifted to the limb mesenchyme where they now control limb size but are largely dispensable for limb development (Purushothaman et al., 2019). At the very least, these data suggest that the molecular logic of tetrapod limb development may not be absolutely conserved among all tetrapods. Here, we asked whether reduced prominence of Fgf signaling during salamander limb development might be offset by an increased reliance on Shh signaling to control limb field progenitor proliferation, survival, outgrowth, and patterning. To test our hypothesis, we pharmacologically inhibited Shh signaling in axolotls using cyclopamine or the highly specific smoothened antagonist BMS-833923. We also genetically inactivated Shh using CRISPR/Cas9 and analyzed limb development in Shh crispants. We asked whether SHH-N or FGF-8b protein could stimulate limb development outside the limb field as has been observed in chick embryos (Cohn et al., 1995) and whether the limb field formed in the absence of Shh signaling.

Small-Molecule Smoothened Antagonist BMS-833923 Inhibits Proximodistal Outgrowth of the Limb Bud in Axolotl Embryos
To test the hypothesis that Shh signaling coordinates expansion of limb field progenitor cells and proximodistal outgrowth of the limb bud in salamanders, we first inhibited Shh signaling throughout early limb development using a pharmacologic approach. Previous studies using cyclopamine during axolotl limb development provided evidence that Shh functions primarily to pattern the anteroposterior axis following expansion of limb bud progenitors (Stopper and Wagner, 2007). Curiously, recent work exploring Shh signaling during zebrafish development and caudal fin regeneration revealed that a highly specific smoothened antagonist, BMS-833923 (hereafter BMS) (Akare et al., 2014) more potently and specifically inhibited hedgehog signaling compared to cyclopamine (Armstrong et al., 2017). Therefore, we treated pre-limb bud stage larvae (stage 39) with ethanol, cyclopamine, or BMS for 10 days (initiation and expansion phase) and harvested limbs at stages 46 and 54 ( Figures 1A,B and Supplementary Figures S1A-C). Analyzing skeletal differentiation at stage 54, we found cyclopamine treatment primarily affected anteroposterior patterning, with nearly all (~88%) resultant limbs possessing a humerus, single fused radius/ulna, and at least one digit ( Figure 1B and Supplementary Figures S1A, S1B, S1D). These results mirrored previous limb development studies in urodele and amniote embryos using the same concentration of cyclopamine (Omnell et al., 1990;Scherz et al., 2007;Stopper and Wagner, 2007). In stark contrast, 92% of the BMS-treated larvae had no limbs with only a small bump covering the scapula where the humerus would normally articulate or no bump at all ( Figure 1B and Supplementary Figures S1C, S1D).
Examining pre-chondrogenic limbs at stage 46, we observed that treatment with cyclopamine or BMS caused a significant decrease in limb bud size, although small limb buds still formed in both treatment groups ( Figure 1C). To assess the degree to which these drugs inhibited Shh signaling, we assessed expression of the direct target gene Patched1 (Ptch1) and downstream target genes, Gremlin1 (Grem1) and Fibroblast growth factor 8 (Fgf8) in stage 46 limb buds ( Figures 1D,E). Using in situ hybridization and qRT-PCR, we observed that Ptch1 was significantly downregulated by BMS treatment, Grem1 was significantly downregulated in BMS and cyclopamine-treated limbs, and Fgf8 was significantly downregulated in BMS compared to cyclopamine-treated limbs ( Figures 1D,E). qRT-PCR for downstream targets of Fgf signaling, ETS transcription factor family genes Etv1 and Etv4 showed that Etv4 was significantly downregulated in both the drug treatments ( Figure 1E). These data supported that BMS was more effective at inhibiting Shh signaling compared to a max dose of cyclopamine. Although BMS drug treatment ultimately resulted in no proximodistal outgrowth, analysis of initial limb bud size and expression of direct downstream targets at stage 46 revealed that neither drug completely inhibited Shh signaling when treated at stage 39. Although we used a max dosage of cyclopamine that embryos could tolerate without lethality (twice the concentrations used in previous amphibian studies), to rule out any potential for delayed activity of cyclopamine (compared to BMS), we exposed embryos to the two drugs prior to limb field formation ( Figure 1F). Embryos treated at neural fold stage 19/20 with cyclopamine still developed a humerus and radius, whereas BMS completely inhibited limb formation ( Figures 1F-H). These data demonstrate that BMS more completely inhibits the Shh signaling pathway compared to cyclopamine when used on salamander embryos and that Shh signaling regulates the earliest stages of axolotl limb development similar to phenotypes recovered from Shh inactivation during pectoral fin development in zebrafish (Neumann et al., 1999). In addition, the incomplete inhibition that we observed using cyclopamine allowed us to confirm that Shh also governs anteroposterior patterning across the limb mesoderm after bud outgrowth (Towers et al., 2008).
Differences across the drug treatments raised the possibility that previous cyclopamine studies may have overlooked an early role for Shh signaling in chick limbs. To test this idea, we inhibited Shh signaling during chick limb development using BMS beginning at stage HH14 to exclude the possibility of hedgehog pathway activation prior to the treatments (Figures 1H-J Supplementary Figures S2A-S2D). In contrast to our results in axolotls, chicken embryos treated with BMS developed limbs with a stunted stylopod, zeugopod, and autopod similar to the most severely affected pair of wings in a previous study using cyclopamine (Scherz et al., 2007) ( Figure 1J, Supplementary Figures S2A-S2D). These findings substantiate that Shh signaling in amniote limbs functions primarily to pattern the anteroposterior axis where it acts subordinately to the AER which controls proximodistal outgrowth and skeletal differentiation ( Figure 1H, Supplementary Figures S2B, S2D). Our findings in axolotls, however, show that Shh signaling first coordinates expansion of limb progenitor cells and proximodistal outgrowth of the limb bud in salamanders ( Figure 1H).

Axolotl Shh Crispants Completely Lack Forelimbs
To further interrogate the function of Shh signaling during salamander limb development, we genetically inactivated Shh in fertilized axolotl zygotes using CRISPR/Cas9. By designing three complementary guide RNAs to the Shh locus and separately injecting these into fertilized single cells, we recovered overlapping and robust mutant phenotypes using all three guide RNAs, thereby ruling out the chances of off-target effects (Figures 2A,B Table  S2). Out of 300 embryos injected with either Shh guide RNA#1, 2, or 3, 67 survived and >80% of the F0 crispants that we screened presented a range of severe craniofacial defects, including partial to complete cyclopia, caudal truncations, and a curved body axis (Figures 2A,B); phenotypes that were similar to those observed in Shh-null mice (Chiang et al., 1996;Chiang et al., 2001). Although complete cyclopia occurred in relatively few F0 crispants (3%), 87% of the crispants exhibited eyes that were positioned with little to no interocular separation ( Figures  2A,B). This resulted in a tight correlation between reduced interocular distance and the appearance of a bulge on the front of the head similar to the formation of a proboscis in Shh-null mice (Chiang et al., 1996). Shh mutants also had smaller heads and lacked most of the anterior craniofacial skeleton including jaws ( Figure 2A).
In addition to these defects, almost all Shh crispants completely lacked forelimbs, a phenotype similar to zebrafish Shh mutants but in contrast to Shh-null mice, and chick embryos treated with BMS ( Figure 1J and Figures 2B,C) (Chiang et al., 1996;Chiang et al., 2001;Neumann et al., 1999). Owing to a lack of jaw structures in the knockout animals that precluded them from eating, we analyzed the limb skeletons at stage 48 just prior to the onset of feeding ( Figure 2C). At stage 48, 100% of control larvae showed a chondrifying humerus, radius, and ulna, whereas 82% of the Shh  Scherz et al., 2007) and BMS-treated limbs at HH30-32 (n = 4 per treatment). Scale bar = 1 mm. Error bars: SEM; and asterisk: significant p-values. hu, humerus; r, radius; ul, ulna; D, digit; sty, stylopod; zeu, zeugopod; and aut, autopod. Frontiers in Cell and Developmental Biology | www.frontiersin.org April 2022 | Volume 10 | Article 797352 knockouts showed only elements of the pectoral girdle ( Figure 2C). Using a previously published guide RNA against Tyrosinase (Tyr) (Fei et al., 2018), we observed loss of pigmentation in all injected animals, but, otherwise, normal embryos; a result that reinforced the specificity of CRISPR/Cas9 in axolotl embryos (Supplementary Figures S5A-S5E). Next, we analyzed the limb field area in Shh knockout animals at stage 46 where limb buds should form and observed almost no outgrowth over the condensing pectoral skeleton compared to elongate limb buds observed in wild-type animals ( Figures 2C,D). This outgrowth defect was even more pronounced than in our BMS treatments with ectoderm almost directly covering the scapula. Last, we analyzed downstream Shh targets in the limb field of Shh crispants at stage 46. Although we observed strong expression of Ptch1, Grem1, and Fgf8 in control limb buds, we were unable to detect expression for these target genes in the forelimb fields of Shh crispants ( Figure 2D). Together with our BMS experiments, these results demonstrate that Shh signaling is required to stimulate expansion of forelimb bud progenitors and control proximodistal outgrowth of the limb bud.

Embryonic Flank Tissue Outside the Limb Field is Refractory to Limb Outgrowth Signals in Axolotl Embryos
In chicken embryos, implantation of FGF protein is sufficient to induce a limb from uncommitted flank tissue supporting Fgf signaling at the apex of a molecular limb program that can induce a secondary limb field (Cohn et al., 1995). To ascertain whether Shh signaling could alone trigger limb bud outgrowth, we grafted SHH-N protein (in 1× PBS with 0.1% BSA) infused Affi-Gel beads into the right flank of stage 37-39 wild-type axolotl embryos, several days before forelimb buds emerge from the forelimb field and monitored for development of an extra limb bud at stage 46 ( Figure 3A). On the contralateral (left) side of these embryos, we implanted 1× PBS with 0.1% BSA soaked control beads as control for the bead implant ( Figure 3A). In either case, we did not observe the development of ectopic limb buds from the flank sites where we implanted SHH-N or control beads ( Figure 3B). In contrast to the flank sites with implanted beads, normal forelimb buds were evident at 3 days post first bead-implantation FIGURE 2 | Axolotl Shh crispants completely lack forelimbs. (A,B) Phenotypes like body axis, position of the eye, and Alcian blue/Alizarin red staining of craniofacial structures of CRISPR control and Shh crispant larvae (n = 23 for CRISPR control and n = 67 for Shh crispant larvae for body axis and eye position analysis, n = 10 each for CRISPR control and Shh crispant larvae for Alcian blue/Alizarin red staining of craniofacial structures). Green dotted line, tail length; red dotted line, distance between the eyes; red asterisk, loss of anterior cranio-facial structures. Scale bar = 1 mm (for body axis) and 500 µm (for eye position and cranio-facial structures). (B,C) Alcian blue/Alizarin red staining for stage 48 limbs of CRISPR control and Shh crispant larvae (n = 10). Scale bar = 500 µm. (D) In situ hybridization for genes Ptch1, Grem1, and Fgf8 in CRISPR control and Shh crispant larvae at stage 46 limbs (n = 3 or 4 per treatments). Red arrows, expression domains. Scale bar = 100 µm. m, meckel; hh, hypohyale; bb1, basibranchial 1; hb1, hypobranchial 1; bb2, basibranchial 2; hb2, hypobranchial 2; ch, ceratohyal; and 4 cb1-4, ceratobranchials.
Frontiers in Cell and Developmental Biology | www.frontiersin.org April 2022 | Volume 10 | Article 797352 (dpi) ( Figure 3B). This result suggested that flank mesoderm was already committed to a non-limb fate but did not rule out the possibility that another factor could induce an ectopic limb bud. Fgf8 is the endogenous inducer of chick limb formation and implanting FGF-8b soaked beads can induce an ectopic limb bud and development of a complete limb. Similar to our experiment with SHH-N, we did not observe ectopic forelimb buds in response to FGF-8b soaked beads ( Figure 3C), and thus, these data supported that limb field progenitors are specified at precise positions very early during salamander development (Stocum and Fallon, 1982) and that neither SHH nor FGF8 could induce secondary limb fields in stage 37 embryos.

Sonic hedgehog Controls Limb Bud Outgrowth in Axolotls
On the basis of these results, we next asked whether the forelimb field was specified in Shh crispants and, if so, whether implantation of SHH-N protein could induce forelimb formation from forelimb field progenitors. We implanted beads just prior to when limb buds would normally emerge, i.e., stage 39 ( Figure 4). Stage 39 axolotls were characterized by longer and branched gills, distinct cloaca, pigmented eyes, and flanks, and these features were used to approximately stage in the control and Shh crispants (Schreckenberg and Jacobson, 1975). Affi-Gel blue beads infused with SHH-N protein were implanted into the position of the forelimb field (somites 3-5), whereas beads containing 1× PBS + 0.1% BSA were implanted into the left contralateral forelimb field as controls and beads were replaced once every 3-4 days. Although none of the beads on the control flank stimulated limb development, in seven of nine animals, limbs emerged in response to SHH-N (n = 2 showed nubbin like outgrowth, n = 3 showed progression to stage 45 limb bud, and n = 2 showed progression to stage 46/47 limb bud) ( Figure 4). Five of the nine limbs that emerged past a nubbin and showed proximodistal outgrowth, three exhibited cartilage formation at the humerus level ( Figure 4). Because of the inability of the crispants to feed, we could not take the limbs out far enough to determine whether the entire limb skeleton formed. However, these data do demonstrate that forelimb progenitor cells are competent to respond to exogenously delivered SHH protein that is sufficient to stimulate expansion of forelimb progenitor cells, formation of a forelimb bud, proximodistal outgrowth of the limb bud, and skeletal differentiation of the developing limb.

Sonic hedgehog Controls Cell Proliferation and Cell Survival During Axolotl Limb Development
Although literature in chicken and mouse limbs show that AER-Fgf signaling regulates cell proliferation, cell survival, Schematic representation of the embryonic bead implantation experiment using stage 37 axolotl embryos. Colored circles represent the right limb field (red), relative implantation site of protein-soaked beads (blue) or 1× PBS/0.1% BSA soaked beads (yellow). (B,C) SHH-N and FGF8 protein-soaked beads were grafted into the right flank lateral to the somites (blue arrow), and 1× PBS/0.1% BSA beads were implanted into symmetrical positions in the left flank (yellow arrow). Three days post first bead implantation (dpi), a second pair of beads were implanted into the same location as the first beads, and, consequently, a third pair were implanted at 6 dpi. Evidence of ectopic limb development was tracked until stage 46 (for SHH-N bead) or stage 49 (for FGF-8b beads), and, in either treatment condition, we did not observe evidence of limb outgrowth. Red asterisks mark emergence and growth of the normal forelimbs. Scale bar = 1 mm.
Frontiers in Cell and Developmental Biology | www.frontiersin.org April 2022 | Volume 10 | Article 797352 and limb outgrowth (Mariani et al., 2008;Summerbell, 1977;Janners and Searls, 1971), our BMS-treated and Shh crispant embryos suggested that Shh signaling regulated these processes in salamanders. To address this possibility, we quantified total cell proliferation in stage 45 limb buds across all treatment groups using light-sheet microscopy (Purushothaman et al., 2019). Compared to control limb buds, cyclopamine-treated limbs showed a decrease in proliferating cells only at the distal tip of the limb and did not show a significant decrease in total EdU+ cells ( Figures 5A,C). In contrast, BMS-treated embryos and Shh crispant embryos showed a significant decrease among total EdU+ cells compared to control limbs ( Figures 5A-C).
Although the fraction of total proliferating cells as a function of total limb volume was not different between the cyclopamine and BMS treatments at stage 45, it was evident from the lightsheet images that the BMS-treated limbs were significantly smaller and contained fewer mesodermal cells ( Figure 5A). Shh crispants had very few cells visible beneath the ectoderm because there was no outgrowth of the limb field ( Figures 5B,C).
Next, we analyzed cell survival using LysoTracker to label dying cells within developing stage 45 limbs (Mariani et al., 2008;Seifert et al., 2009). BMS-treated and Shh crispant larvae showed LysoTracker-positive cells throughout the limb buds, whereas none of the control larval limbs showed LysoTracker-positive cells ( Figure 5D). Cell death was most prominent in the proximal and distal ends of BMS-treated limbs, whereas dying cells were present beneath the ectoderm on the flanks of Shh crispants ( Figure 5D). Together, these results from Shh crispants and BMS-treated axolotl larvae support a model where Shh signaling seems to simultaneously control cell proliferation and cell survival in mesodermal progenitors of the limb field.

DISCUSSION
Our study supports a model where molecular components present in pectoral fin and amniote forelimb buds are deployed uniquely during salamander limb development. Specifically, our results demonstrate that Shh signaling is essential for proliferation, survival, and expansion of forelimb field progenitor cells to form a salamander forelimb bud. As the limb bud emerges from the flank, Shh signaling stimulates Fgf8 and downstream Fgf signaling targets in limb mesoderm, which support some cell proliferation at the distal tip of the limb bud but do not significantly contribute to proximodistal outgrowth of the limb (Purushothaman et al., 2019). Our results also uncovered that cyclopamine does not completely inhibit Shh signaling in salamanders when used at a maximum, non-lethal concentration for embryos and revealed that BMS is a more effective hedgehog inhibitor in salamander embryos. Genetic inactivation of Shh using CRISPR/Cas9 in newly fertilized zygotes confirmed our Frontiers in Cell and Developmental Biology | www.frontiersin.org April 2022 | Volume 10 | Article 797352 8 results using BMS and in combination with bead implantation experiments further revealed that Shh is not required to specify forelimb progenitor cells. Together, our data demonstrate that the molecular regulation of forelimb development has independently evolved in salamanders away from a reliance on reciprocal mesenchymal-epithelial Fgf signaling for limb bud outgrowth and proximodistal patterning: the arrangement present in anuran and amniote limbs and in actinopterygian pectoral fins ( Figure 6).
Pectoral fin and forelimb development in actinopterygian and amniote embryos, respectively, rely on ectodermal-mesodermal cross-talk between Fgf-and Shh signaling. In amniote embryos where ZPA/AER cross-talk is present, proximodistal outgrowth of the limb bud is Shh-independent with Shh functioning as a cell survival factor, maintaining Fgf signaling in the AER, and coordinating anteroposterior patterning (Laufer et al., 1994;Niswander et al., 1994;Sanz-Ezquerro and Tickle, 2000;Kraus et al., 2001;Ros et al., 2003;Harfe et al., 2004;Scherz et al., 2007). Although Shh signaling does appear to regulate cell proliferation in chick embryos (Towers et al., 2008), it remains unknown whether this is an indirect effect mediated through Fgf signaling as occurs in zebrafish (Prykhozhij and Neumann, 2008)-a situation that is likely in light of our BMS results with chicken embryos. Proliferation of the fin/limb mesoderm and proximodistal expansion of the bud appears reliant on Fgf signaling from the AER, which also maintains Shh signaling from the ZPA Fischer et al., 2003;Nomura et al., 2006;Mariani et al., 2008). For instance, although zebrafish shh mutants completely lack pectoral fins, fin buds emerge normally, only later to regress as ectodermal Fgf signaling is lost (Neumann et al., 1999). Similarly, in Shh −/− mice, limb buds emerge smaller, but otherwise normally with an intact AERexpressing Fgf8 (Chiang et al., 2001). However, without Shh signaling from the ZPA, the AER disappears, cell proliferation decreases, and posterior elements are lost (Chiang et al., 2001). It would appear that loss of Shh produces a different phenotype in fishes and amniotes (i.e., no pectoral fins versus elongate forelimbs with a humerus, radius, and single digit). However, the retention of skeletal elements in amniotes stems from the extent to which Fgf signaling is maintained in the absence of Shh signaling to sustain mesodermal proliferation and offset cell death reinforcing the reliance on Fgf signaling for fin and limb development in these groups. In fishes and amniotes, Fgf signaling is required for early expansion of fin/limb bud mesoderm where genetic disruption of AER-Fgfs leads to a complete absence of forelimbs/pectoral fins (Xu et al., 1998;Fischer et al., 2003;Nomura et al., 2006;Mariani et al., 2008). Thus, Fgf signaling from the ectoderm controls expansion of the fin/limb progenitor pool until such time that differentiation of the fin/limb skeleton begins.
In contrast to fishes and amniotes, our work demonstrates that ectodermal-mesodermal cross-talk disappeared in salamanders when Fgf signaling moved into the mesoderm (Purushothaman et al., 2019)-an event that also led to an erosion of the tight linkage between the two signaling pathways and an increased reliance on Shh signaling. Importantly, one consequence of this shift was that limb development in urodeles became largely independent of Fgf signaling. Inhibiting Fgf signaling in salamanders produces a relatively normal limb (minus a digit) that results from an overall small decrease in cell proliferation (Purushothaman et al., 2019), whereas inhibiting Shh signaling in this study resulted in a complete lack of forelimbs. Although our results using BMS or genetic inactivation of Shh make clear that limb bud cells cannot survive without Shh signaling, the emergence of small limb buds in response to BMS supports that a miniscule amount of Shh signaling is sufficient to stimulate cell proliferation among forelimb field progenitors-a result that could have manifested indirectly through residual Fgf signaling ( Figure 1E). However, limb buds never emerged in our Shh crispants where Shh was genetically inactivated. Implantation of SHH-N loaded beads into Shh crispants demonstrated that the molecular limb program in salamander limb progenitor cells is tuned to Shh signaling as a primary input. Thus, Shh signaling was sufficient to induce limb morphogenesis from forelimb progenitors in Shh crispants. In contrast, implantation of SHH-N beads (or FGF-8b beads) into flank FIGURE 6 | Limb regenerative ability may have evolved uniquely in caudates. Comparison of developmental mechanisms among tetrapods has revealed that urodeles possess a derived mode of limb development compared to anurans and amniotes including, absence of a morphological and molecular AER, mesenchymal restriction of canonical AER-Fgfs, absence of a positive feedback loop between Fgf and Shh signaling, and reliance on Shh-signaling to regulate proximal-distal outgrowth via cell proliferation and survival (orange line). Correlating with this unique mode of limb development, urodeles are the only extant tetrapods capable of complete limb regeneration at any post-embryonic life stage (larva, juvenile, or adult). One possible explanation for this tight correlation is that limb regenerative ability evolved within caudates as their mode of limb development was modified from the ancestral tetrapod state to reside within the mesenchymal tissue compartment (red asterisk). This opposes the hypothesis that limb and fin regenerative ability have a common origin among Sarcopterygii-with anurans and amniotes having lost limb regenerative ability (Nogueira et al., 2016). This hypothesis is based on common gene expression data derived from regeneration of cartilaginous elements in lungfish fins, a process similar to the regeneration of a cartilage spike in amputated Xenopus limbs.
Frontiers in Cell and Developmental Biology | www.frontiersin.org April 2022 | Volume 10 | Article 797352 mesoderm of wild-type embryos prior to when forelimb buds emerge could not induce ectopic limbs at stages 37-39. These experiments support that forelimb progenitors are specified much earlier during salamander development (Stocum and Fallon, 1982) and that these cells wait until endogenous Shh is activated, which, in turn, jumpstarts cell proliferation and activates downstream signaling pathways important for limb morphogenesis. Future studies examining molecular players that may lie upstream of Shh (e.g., RA and Tbx5/4) will provide important evidence as to which signals actually establish the field and secondarily induce Shh expression in these cells. Classical experiments in newts showed that pieces of otic placode implanted into the flank could induce an ectopic limb (Balinsky, 1925) and the otic placode expresses Fgf8 among other factors. This raises the possibility that, although salamander limb development may have shifted its reliance on Fgf signaling for morphogenesis, it still may be a key factor for establishing the limb field. Given the most recent phylogenetic models for relationships among extant tetrapods (Alexander Pyron and Wiens, 2011;Marjanović and Laurin, 2019), our findings support that urodeles possess a derived program for limb development relative to anurans and amniotes ( Figure 6). This condition would also seem to deviate from the Shh-Fgf cross-talk that is required for fish pectoral fin development as discussed above. This molecular reorganization appears to occur after the establishment of limb field progenitors (Tulenko et al., 2013) as CRISPR/Cas9 knockout studies in newts show that upstream regulators of limb/fin bud initiation like Tbx5 are conserved among vertebrates (Suzuki et al., 2018). The results of this study in combination with our previous results showing movement of Fgf signaling to the mesoderm (Purushothaman et al., 2019) and the work showing that the core long-range Shh enhancer (ZRS) is dispensable for limb development (Suzuki et al., 2018) support that the molecular circuitry during limb outgrowth has been reconfigured in urodeles. Interestingly, a study in medaka (Oryzias latipes) showed that, in addition to the canonical ZRS, a shadow ZRS (sZRS) controls Shh expression in fish (Letelier et al., 2018) and only deletion of both enhancers resulted in the complete loss of Shh expression and loss of pectoral fins (Letelier et al., 2018). Considered together, results from these studies point toward the existence of a shadow enhancer for Shh in urodele limbs as well. With the loss of a morphological and molecular AER (Purushothaman et al., 2019), the limb development program in urodeles relocated into the mesoderm with a role for the ectoderm diminished (Piatt and Kusner, 1960;Lauthier, 1985).
The derived nature of the urodele limb development program raises an intriguing hypothesis: the ability of adult cells to regenerate a functional limb may have evolved in concert with alterations governing the cellular and molecular control of limb development ( Figure 6). In this scenario, adult limb regenerative ability emerged in urodeles after they diverged from anurans, a scenario contrary to the current view where adult limb regeneration was present in the tetrapod ancestor only to be lost in all the other major extant lineages (Fröbisch et al., 2014;Darnet et al., 2019a). Support for this new hypothesis comes from regeneration studies in several vertebrate models. First, experiments in salamanders and newts have shown that the regenerative capacity of blastema cells (the proliferative mass of mesenchymal cells at the amputation plane) lies within these cells in that they exhibit self-organizing properties allowing them to undergo morphogenesis even when transplanted to foreign sites (Stocum, 1984). Thus, they are reliant on signaling pathways that they themselves generate, not necessarily on reciprocal mesenchymal-epithelial signaling as seen during anuran and amniote limb development. Second, Shh signaling is required for blastemal cells to undergo regeneration, although functional studies using cyclopamine during normal limb regeneration in axolotls have yielded contradictory phenotypes not dissimilar from the disparity revealed between our cyclopamine and BMS treatments (Roy and Gardiner, 2002;Nacu et al., 2016). Although an earlier study showed that Shh signaling just controls anteroposterior patterning during limb regeneration (Roy and Gardiner, 2002), a later study using almost double the concentration of cyclopamine showed that Shh signaling was required for regeneration where it maintains Fgf8 expression that is crucial for regenerative outgrowth (Nacu et al., 2016). Moreover, the canonical AER-specific Fgfs (Fgf8, 9, and 17) were re-expressed in the limb mesenchyme during normal limb regeneration. These results suggest Fgf signaling may play a more prominent role during regeneration where it functions in a positive feedback loop with Shh signaling (Nacu et al., 2016). Third, although Xenopus can regenerate limb buds and even parts of limbs during limb development (Dent, 1962), they cannot regenerate limbs following metamorphosis and instead generate unpatterned cartilage spikes. Limb regenerative ability is lost as the limb skeleton differentiates during development and is tied, in part, to an inability of blastemal cells to re-activate Shh and Fgf10 (Yokoyama et al., 2000). Shh is not reactivated in the mesenchyme at regeneration-incompetent stages nor is ectodermal Fgf8. This echos the failure of chicken limb buds to regenerate where they are unable to restore an AER and Fgf signaling (Kostakopoulou et al., 1996). Although a recent study suggested that the loss of limb regenerative ability in Xenopus was due to an inability of limb cells to reprogram to a more developmental state (aka dedifferentiation) (Lin et al., 2021), a previous study in Xenopus showed that partially restoring limb regenerative ability required limb bud cells but only when those cells overexpressed ß-catenin and were exposed to SHH and FGF10 (Lin et al., 2013). These studies reinforce that blastemal cells require the capacity to activate self-organizing potential through the expression of self-sustaining signaling pathways and underscore the key role that patterning mechanisms play for regeneration to occur.
Although some investigators have argued on the basis of comparative gene expression data that an ancestral regeneration program exists for fins and limbs (Nogueira et al., 2016;Darnet et al., 2019b;Lu et al., 2019), these studies do not account for the fact that common gene expression profiles obscure functional relationships between signaling pathways. Many comparative studies also fail to account for interspecific differences in age or growth mode when assessing regeneration, two traits that likely regulate the availability of cells to participate in regeneration. As Frontiers in Cell and Developmental Biology | www.frontiersin.org April 2022 | Volume 10 | Article 797352 such, our results underscore the need to expand limb development studies across a more diverse array of vertebrates, especially anamniotes. For instance, although the relatively few molecular studies in anurans support a mode of limb development more aligned with amniotes, the data also suggest that an alternative dorsoventral patterning system may be in place (Christen and Slack, 1998). In addition, although anurans appear to exhibit conserved ectodermal-mesodermal cross-talk (Wang and Beck, 2014), studies in the direct-developing frog Eleutherodactylus coqui indicate that a morphological AER is not required for compartmentalized ectodermal Fgf8 expression (Richardson et al., 1998;Gross et al., 2011). A recent study showing the absence of Fgf8 during the development of bowfin pectoral fins with an AER further supports the plasticity in the limb/fin molecular program (Thompson et al., 2021). Relative to amniote limbs that develop with input from the somites, amphibian limbs exhibit delayed development and a high degree of self-organization in that transplantation of limb buds to other parts of the body produce a relatively normal limb (Stocum and Fallon, 1982), an ability which may help support limb regeneration (Galis et al., 2003). Although other tetrapods exhibit high degrees of self-organization in the limb field, our results offer that shifting the limb program entirely into the mesenchyme and toward a more pronounced reliance on Shh signaling to coordinate limb outgrowth may have permitted the self-organizing behavior of limb progenitors to execute patterning and, ultimately, functional regeneration.

Animal Husbandry and Tissue Harvest
Axolotls (Ambystoma mexicanum) (albino and wild type) were acquired from our own laboratory colony. Chicken eggs (University of Kentucky, Department of Animal Sciences) were incubated to required stages. All procedures were conducted in accordance with, and approved by, the University of Kentucky Institutional Animal Care and Use Committee (IACUC Protocol: 2013-1,174). For detailed methodology of animal husbandry and tissue harvest, refer to the work of Purushothaman et al. (2019). Axolotl larvae were reared at 20°C-21°C and larvae used for drug treatments, Alcian blue/Alizarin red staining, whole mount in situ hybridizations, cell proliferation, and cell death assays were anesthetized using 1× benzocaine (Sigma) and fixed overnight in 4% paraformaldehyde (PFA) at 4°C. For qRT-PCR, larvae were anesthetized using 1× benzocaine, and limb tissue samples were snap-frozen and stored at −80°C until further use.

Drug Treatments on Axolotl Larvae and Chick Embryos
Drug treatments on stage 39 axolotl larvae were done according to Purushothaman et al. (2019). Larvae were reared in six-well plates and kept in dark throughout the experiment. A working stock of cyclopamine (5 mg/ml; Sigma) and BMS (5mg/ml; Cayman Chemical, Ann Arbor, MI, USA) was prepared in 100% ethanol and 0.6 µl from this stock was added into 3 ml of 20% Holtfreter's solution per well (1 μg/ml final concentration). An equal amount of 100% ethanol (0.02%) was added into control wells. Higher concentrations tested for BMS were lethal. For earlier drug treatments, de-jellied axolotl embryos were treated with ethanol, cyclopamine, or BMS at neural fold stage (stage 19/20) for 10 days. The solutions were replenished every 2 days and treatments lasted for 10 days. For all drug experiments, survival rate (compared to controls), overall animal health (i.e., that they are feeding, swimming, no buoyancy issues, and well-formed gills with adequate blood supply), and body length were monitored. For the drug concentrations administered in this study, we found no significant difference for these monitoring criteria between treatment and control animals.
Drug treatments on chicken embryos were done at HH14/15. Five milliliters of albumen was removed from the bunt end using a 5-ml syringe and 22G1½ needle. The eggs were windowed, and vitelline membrane around the limb bud was removed. The embryos were treated with 5 µl of solution (1 mg/ml) of BMS in DMSO or DMSO (control) followed by 200 µl of Ringers solution with Pen-strep (100 U/ml). The window was covered with a scotch tape, and the eggs were reincubated until harvest at HH30/32.

Alcian Blue and Alizarin Red Staining
Alcian blue and Alizarin red staining on axolotl larvae was done as previously described in Purushothaman et al. (2019). Fixed axolotl larvae were dehydrated in graded ethanol series and stained with 0.02% Alcian blue 8GX (Sigma Aldrich, St. Louis, MO, USA) in 70% ethanol and 30% glacial acetic acid for 3 h to overnight. Larvae were rehydrated in graded ethanol series and then stained with 0.1% Alizarin red (Sigma Aldrich, St. Louis, MO, USA) in 1%KOH overnight. Larvae were cleared in 1%KOH/glycerol series: 3KOH:1glycerol (imaged when cleared), 1KOH:1glycerol (1 day) and 1KOH:3glycerol (stored at room temperature). A subset of 10 larvae each were used from 23 (control) and 67 (Shh guide RNA injected) larvae for Alcian blue and Alizarin red staining to analyze craniofacial defects and skeletal elements within the limbs.

qRT-PCR Analysis
Stage 39 axolotl larvae were reared in six-well plates in either 0.02% ethanol, cyclopamine (1 μg/ml), or BMS (1 μg/ml) until stage 46. Whole limbs were dissected from the body wall, immediately snapfrozen, and stored at −80°C until RNA extraction. n = 3 was used for each condition and each replicate represented a pool of limbs (both left and right) from 10 to 20 animals.
RNA was extracted using TRIzol reagent (Invitrogen), cDNA was synthesized from 0.5 to 1 µg of RNA using SensiFast cDNA synthesis kit, and qRT-PCR was performed using iTaq Universal SYBR Green Supermix (Biorad) (refer Supplementary Table S1 for primer sequences). Melting curve was analyzed to confirm primer specificity.
Rlp32 were used as the internal control/house-keeping genes for the experiments, respectively, since there was no significant fold change in the 2 −Ct values (Schmittgen and Livak, 2008). 2 −ΔΔCt method was used to calculate the fold change values between control (ethanol) and treatment (cyclopamine or BMS) groups (Schmittgen and Livak, 2008).

Guide RNA Synthesis
Protocol for guide RNA synthesis was partially adapted from the work of Fei et al. (2018). The mRNA coding sequence for Shh gene was accessed from https://www.axolotl-omics.org/ search. DNA template oligos for guide RNA synthesis were designed using cloud-based informatics platform Benchling and oligos (20-mer or 18-mer) with high on-target and offtarget scores were selected. Three DNA template oligos for separate Shh guide RNAs were ordered from IDT with a 5′ adapter and T7 promoter at the 5′ end and a 3′ overhang sequence complementary to the constant sequence at the 3′ end [refer to the work of Fei et al. (2018) for schematic diagram and Supplementary Table S1 for sequences]. The DNA template oligo for Tyrosinase (Tyr) guide RNA ordered from with 5′ adapter, T7 promoter sequence, GG nucleotides at the 5′ end, and a 3′ overhang sequence complementary to the constant sequence at the 3′ end [refer to the work of Fei et al. (2018) for schematic diagram and Supplementary  Table S1 for sequences]. The DNA template oligo for Tyr guide RNA was adapted from the work of Fei et al. (2018).
The DNA template oligo was amplified using the Phusion DNA polymerase kit (NEB, cat# M0530S). Refer to Table S1 for primer sequences. The reaction mix and PCR reaction were as follows: PCR Reaction.
The PCR product was checked on a 1% agarose gel for single bands, purified (in 20 µl of water) using QIAquick PCR purification kit, and quantified. The purified product was used for guide RNA synthesis (in vitro transcription) using Ambion MegaShortscript Kit T7 (cat#AM1354). The in vitro transcription step was as follows: The reaction mix is incubated at 37°C overnight, and guide RNAs were precipitated by phenol/chloroform method as follows: the reaction mix was transferred into a 2-ml vial, 115 µl of water, and 15 µl of Ammonium acetate stop solution (from Ambion MegaShortscript Kit T7) and 500 µl of phenol + 500 µl chloroform mix were added, mixed until an emulsion formed, and centrifuged at 13,000 rpm for 1 min; aqueous phase was transferred into a fresh 1.5-ml vial; 2 volumes of ethanol was added and mixed well, chilled at −20°C for 15 min, and centrifuged at 13,000 rpm for 15 min; supernatant was carefully discarded; tubes were allowed to dry under the hood; and the RNA pellet was resuspended in~20 µl of water. The integrity of the guide RNAs was checked on a gel, quantified, and stored in −70°C (as 2 µl of aliquots) until microinjections.

Microinjections
Protocol for microinjection was partially adapted from the work of Fei et al. (2018). The following injection mix was freshly prepared once the female started laying eggs: The glass capillary tubing (OD = 1 mm, ID = 0.58 mm, length = 7.5 cm) was pulled (Sutter instruments Co.; settings: heat, 600; pull, 50; vel, 120; time, 165), and the injection mix was loaded into it. The needle tip was calibrated using a stage micrometer so as to inject a volume of 5 nl into each fertilized egg. For control injections (CRISPR control), injection mix minus guide RNA was injected into each single-cell fertilized egg.
The single-cell fertilized eggs were sterilized using 70% ethanol for 20 s, rinsed and dejellied in 1× MMR/Pen-strep solution, and transferred to 1× MMR/Pen-strep + 20% Ficoll solution for microinjections. Post injection, the eggs were transferred into fresh 1× MMR/Pen-strep + 20% Ficoll for 2 h and then transferred into 0.1× MMR/Pen-strep + 5% Ficoll for 24 h at 18°C. The healthy embryos were transferred into 24-well plates with 0.1× MMR/Pen-strep solution and not disturbed for 7 days. Fresh 0.1× MMR + Pen-strep solution was added on the eighth day, and solution was replenished once in 2 days until final harvest.

Genotyping
Protocol for genotyping was partially adapted from the work of Fei et al. (2018). For genomic DNA extraction, 1-mm tail clips from CRISPR control and guide RNA-injected larvae were snapfrozen in 1.5-ml vials and stored at −80°C until further use. Later, in to the 1.5-ml vials, 100 μl of 50 mM NaOH was added and incubated at 95°C for 20 min and 10 μl of 1 M Tris (pH 7.5) was added, mixed well, spun, and quantified. The following PCR was performed to amplify the gene locus: (Refer to Supplementary Table S1 for genotyping primer sequences).
The PCR product was checked on gel to verify single bands, quantified, and sent out for NGS (Amplicon-EZ, Genewiz) to sequence each gene locus.
To test the limb forming potential of the presumptive trunk in wild-type larvae, a tungsten wire was used to make an incision into the posterior ectoderm where the control and treated (SHH-N or FGF8) beads were slid into the presumptive trunk on the left and right sides of the larvae, respectively. The beads were grafted when larvae reached stage 37. The same procedure was used after 3 and 6 days to implant a second and third bead following the first insertion. Live larvae were imaged at 0, 3, and 6 dpi (days post initial bead insertion). Experimental larvae were harvested once larval limbs reached stage 46/49. Limb rescue bead experiments were done on larvae injected with either Shh guide RNA#3 or Shh guide RNA#1, 2, and 3 with evident axolotl Shh crispant phenotypes like curved body axis, partial to complete cyclopia, and no limb outgrowth. The SHH-N beads were grafted when the CRISPR control larvae reached stage 37. A nick was made at the approximate limb field position using a tungsten needle, and beads were grafted securely into the nicks. Protein-soaked bead was grafted into the right limb field, and control bead was grafted into the left limb field. Beads were replaced once in 3-4 days and experimental larvae were harvested once CRISPR control larval limbs reached stage 46/48.

Cell Death and Cell Proliferation Assays
Refer to the work of Purushothaman et al. (2019) for detailed protocols of cell proliferation and cell death assays in axolotls. Post-hatch larvae were reared in six-well plates in 3 ml of either of the solutions: 0.02% ethanol, cyclopamine (1 μg/ml), or BMS (1 μg/ml) in 20% Holtfreters solution. CRISPR control and Shh crispants were reared in six-well plates in 3 ml of 0.1× MMR/Penstrep solution. For cell proliferation assay, larvae were additionally treated with EdU (0.1 mg/ml) for 24 h when control larvae (ethanol-treated or CRISPR control) reached stage 45, fixed overnight in 4% PFA, dehydrated in 1× PBS/ methanol series, and stored in 100% methanol at −20°C until further use. Larvae were rehydrated backward through graded methanol series starting at 100% methanol and ending at 100% 1× PBS, treated with 2.5% trypsin (Gibco) for 10 min, permeabilized with proteinase K (20 μg/ml) in PBT for 7-10 min, fixed in 100% acetone at −20°C for 10 min, incubated in fresh click reaction solution [1× TRIS buffer saline, 4 mM CuSO 4 in 1× TRIS buffer saline, 2 µl of Alexa-flour-594 Azide (Life technologies), 1 mM sodium ascorbate in 1× TRIS buffer saline] for 30 min on a rocker in the dark, incubated in DAPI (1: 1,000 dilution) for 30 min, checked for fluorescence under a stereomicroscope, and stored at 4°C in the dark until lightsheet imaging.
For cell death assay, larvae were transferred into 24-well plates and treated with 200 µl of 5 µM LysoTracker Red DND-99 (molecular probes) in Hanks' BSS for 45 min to 1 h at 20°C-21°C, fixed overnight in 4% PFA, dehydrated through a graded methanol/Hanks' BSS series and stored in 100% methanol at −20°C until imaging.

Microscopy and Image Analysis
Whole mount images of limbs for Alcian blue/Alizarin red staining, limb size measurements, in situ hybridization, bead experiments, cell death assays, body axis analysis, and eye position and pigmentation analysis were taken on an SZX10 light microscope (Olympus, Tokyo, Japan) using a DP73 CCD camera (Olympus). The microscope was equipped with CellSense software (CellSense version 1.12, Olympus Corporation).
EdU-stained stage 45 larval limbs were imaged using a Zeiss Lightsheet Z.1 (College of Arts and Science Imaging Center, University of Kentucky). Refer to the work of Purushothaman et al. (2019) for detailed protocol of lightsheet microscopy for axolotl limb buds. Zen software (Zeiss) was used for imaging, and samples were excited using 561and 488-nm lasers. Arivis vision4D software (Arivis) was used for image processing. For total limb volume calculations, an object mask was hand drawn at each z-plane on the basis of the DAPI signal to outline the limb. Red cell aggregate volume and total limb volume were calculated using the previously standardized protocols, and volume values in cubic micrometers and voxel counts were given as outputs.

Fiji Analysis
Melanocyte pigmentation in the eyes of stage 46 tyrosinase crispants and control larvae was measured using Fiji software (NIH) after calibrations. Eye pigmentation was measured as pixel intensity and the ranges were as follows: pixel intensity = 221 to 148 (high), 147 to 74 (moderate), and 73 to 0 (low).

Statistics
All statistical analyses were performed using JMP (version Pro 12.10, SAS Institute Inc.) and Microsoft Excel. Bar and pie diagrams were made using Microsoft Excel.
For limb size between ethanol, cyclopamine, and BMS, one-way ANOVA followed by Tukey-Kramer HSD post hoc test was performed. Differences were considered significant if p < 0.05.
For qRT-PCR data, the 2 −ΔΔCt method was used to calculate fold changes of genes between ethanol, cyclopamine, and BMS groups. Calculations for mean Ct values, ΔCt values for treatment and control groups, ΔΔCt values between treatment and control groups, 2 −Ct and 2 −ΔΔCt fold change values were done using Microsoft Excel. One-way ANOVA followed by Tukey-Kramer HSD post hoc test was performed. Differences were considered significant if p < 0.05.
For light-sheet data, red cell aggregate volume/limb volume (%) was calculated in Microsoft Excel (red cells = EdU-positive cells). Post arcsin conversion, comparisons between control and treatment groups (ethanol vs. cyclopamine vs. BMS and CRISPR control vs. Shh crispant) were done by one-way ANOVA followed by Tukey-Kramer HSD post hoc test.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Materials, further inquiries can be directed to the corresponding author.

ETHICS STATEMENT
The animal study was reviewed and approved by the University of Kentucky Institutional Animal Care and Use Committee (IACUC Protocol: 2013-1174).

AUTHOR CONTRIBUTIONS
Conceptualization: SP and AS. Methodology: SP and BL. Investigation: SP and BL. Visualization: SP, BL, and AS. Supervision: AS. Writing-original draft: SP and AS. Writing-review and editing: SP and AS.