1. Developmental Biology
  2. Evolutionary Biology
Download icon

Fgf-signaling is compartmentalized within the mesenchyme and controls proliferation during salamander limb development

  1. Sruthi Purushothaman
  2. Ahmed Elewa
  3. Ashley W Seifert  Is a corresponding author
  1. University of Kentucky, United States
  2. Cold Spring Harbor Laboratory, United States
Research Article
  • Cited 0
  • Views 894
  • Annotations
Cite this article as: eLife 2019;8:e48507 doi: 10.7554/eLife.48507

Abstract

Although decades of studies have produced a generalized model for tetrapod limb development, urodeles deviate from anurans and amniotes in at least two key respects: their limbs exhibit preaxial skeletal differentiation and do not develop an apical ectodermal ridge (AER). Here, we investigated how Sonic hedgehog (Shh) and Fibroblast growth factor (Fgf) signaling regulate limb development in the axolotl. We found that Shh-expressing cells contributed to the most posterior digit, and that inhibiting Shh-signaling inhibited Fgf8 expression, anteroposterior patterning, and distal cell proliferation. In addition to lack of a morphological AER, we found that salamander limbs also lack a molecular AER. We found that amniote and anuran AER-specific Fgfs and their cognate receptors were expressed entirely in the mesenchyme. Broad inhibition of Fgf-signaling demonstrated that this pathway regulates cell proliferation across all three limb axes, in contrast to anurans and amniotes where Fgf-signaling regulates cell survival and proximodistal patterning.

https://doi.org/10.7554/eLife.48507.001

eLife digest

Salamanders are a group of amphibians that are well-known for their ability to regenerate lost limbs and other body parts. At the turn of the twentieth century, researchers used salamander embryos as models to understand the basic concepts of how limbs develop in other four-limbed animals, including amphibians, mammals and birds, which are collectively known as “tetrapods”. However, the salamander’s amazing powers of regeneration made it difficult to carry out certain experiments, so researchers switched to using the embryos of other tetrapods – namely chickens and mice – instead.

Studies in chickens, later confirmed in mice and frogs, established that there are two major signaling centers that control how the limbs of tetrapod embryos form and grow: a small group of cells known as the “zone of polarizing activity” within a structure called the “limb bud mesenchyme”; and an overlying, thin ridge of cells called the “apical ectodermal ridge”. Both of these centers release potent signaling molecules that act on cells in the limbs. The cells in the zone of polarizing activity produce a molecule often called Sonic hedgehog, or Shh for short. The apical ectodermal ridge produces another group of signals commonly known as fibroblast growth factors, or simply Fgfs.

Several older studies reported that salamander embryos do not have an apical ectodermal ridge suggesting that these amphibian’s limbs may form differently to other tetrapods. Yet, contemporary models in developmental biology treated salamander limbs like those of chicks and mice. To address this apparent discrepancy, Purushothaman et al. studied how the forelimbs develop in a salamander known as the axolotl.

The experiments showed that, along with lacking an apical ectodermal ridge, axolotls did not produce fibroblast growth factors normally found in this tissue. Instead, these factors were only found in the limb bud mesenchyme. Purushothaman et al. also found that fibroblast growth factors played a different role in axolotls than previously reported in chick, frog and mouse embryos. On the other hand, the pattern and function of Shh activity in the axolotl limb bud was similar to that previously observed in chicks and mice.

These findings show that not all limbs develop in the same way and open up questions for evolutionary biologists regarding the evolution of limbs. Future studies that examine limb development in other animals that regenerate tissues, such as other amphibians and lungfish, will help answer these questions.

https://doi.org/10.7554/eLife.48507.002

Introduction

Limb development is an ideal model to investigate how cellular and molecular networks exhibit plasticity or resilience during tetrapod evolution. Since the turn of the twentieth century the limb has endured as a fundamental model to investigate morphogenesis, cellular growth, and differentiation during embryonic development. From studies spanning comparative embryology through developmental genetics, limb development has provided deep insight into mechanisms underlying pattern formation, genotype-phenotype relationships, and the complexity of molecular networks (Swett, 1937; Wolpert, 1969; Zeller et al., 2009). Perhaps equally studied as a fully developed structure, the limb has also served as a model for evolutionary biologists seeking to reconstruct tetrapod ancestry and how micro- and macro-evolutionary changes might explain major events in vertebrate evolution such as the fin to limb transition (Fröbisch and Shubin, 2011; Shubin and Alberch, 1986).

The two most important embryonic models for modern limb development studies have been chicken and mice. Chicken embryos were integral to our understanding of limb bud outgrowth and morphogenesis aided by the availability of genetic limb mutants and the ability to discern skeletal defects in the wing caused by surgical manipulation (Saunders, 1948; Summerbell et al., 1973). With the advent of stable transgenesis, mice became the model of choice to investigate the molecular basis of limb development. Together, data from chicken, mice and frogs have been synthesized into contemporary models covering limb development in tetrapods (Zeller et al., 2009; Zuniga, 2015). In these models, all vertebrate limb buds utilize the same molecular network (e.g. Shh, Fgfs, Bmps, Wnts and retinoic acid) governed by Hox genes and controlled by two major signaling centers; the zone of polarizing activity (ZPA) and the apical ectodermal ridge (AER).

And then there are the urodeles. With tetrapod monophyly solidly supported by molecular and morphological data (Ahlberg and Milner, 1994; Marjanović and Laurin, 2013), for scientists investigating the evolution and development of vertebrate limbs, salamanders and newts have longed proved problematic (Holmgren, 1933). Although salamander and newt embryos were used to uncover key principles regarding specification of the prospective limb field and establishment of the primary limb axes (i.e., proximal-distal, anterior-posterior and dorsal-ventral) (Harrison, 1918; Stocum and Fallon, 1982; Swett, 1937), it was also observed that urodele limb development deviated from anurans and amniotes in at least two key respects: skeletal specification exhibited preaxial dominance (anterior elements form before posterior elements) rather than postaxial dominance (Shubin and Alberch, 1986) and urodele limb buds did not form an apical ectodermal ridge (AER) (Sturdee and Connock, 1975; Tank et al., 1977). In addition to these important developmental differences, adult urodeles differ from anurans and amniotes in their ability to completely regenerate an amputated limb. With these ideas in mind, we sought to investigate limb development in salamanders to determine whether morphological and molecular data support a unified model of limb development that includes or excludes urodeles.

Results

Digit specification and differentiation are uncoupled during axolotl forelimb development

In order to study axolotl limb outgrowth and axis specification, we first staged developing limbs on the basis of external morphology and skeletal chondrification at 20–21°C (Figure 1A–B and Figure 1—figure supplement 1). As the limb bud emerged from the flank, limb mesenchyme expanded directly above the body wall musculature and at no time during limb development did we observe an AER or thickening of the limb ectoderm (Figure 1C). In agreement with previous work in salamanders (Holmgren, 1933; Nye et al., 2003; Shubin and Alberch, 1986), we found that cartilage condensations of the limb skeleton formed proximal to distal, and within the zeugopod and autopod, anterior to posterior (preaxial dominance) (Figure 1D). While this was strictly true for the radius and ulna, alcian blue staining always demonstrated digit I and II forming together between stages 47 and 48 (Figure 1D). To further examine chondrogenesis in the limb, we identified axolotl Sox9 (Supplementary file 1) and used its expression to analyze condensing mesenchymal cells prior to cartilage formation (Wright et al., 1995) (Figure 1E and Figure 1—figure supplement 2). We first observed Sox9 expression at stage 45 in a broad proximal area of the axolotl forelimb bud (Figure 1E). Sox9 expression at stage 46 appeared in a centrally located domain corresponding to the future radius, ulna and carpals (Figure 1D–E). However, in contrast to the Alcian blue staining pattern, Sox9 expression in the presumptive digits appeared sequentially II-I-III-IV and this result was consistent across 13/15 (86.66%) limbs analyzed across stage 48 (Figure 1E and Figure 1—figure supplement 2). At stage 49, Sox9 expression expanded to clearly mark digit I along with digit II (Figure 1E and Figure 1—figure supplement 2). Thus, although chondrification of the limb skeleton proceeds anterior to posterior in the zeugopod, digit specification as marked by Sox9 expression in the autopod exhibits a pre-pattern that first emerges along the central axis marked by digit II.

Figure 1 with 2 supplements see all
Skeletal chondrification proceeds anterior to posterior within the zeugopod and autopod during axolotl forelimb development.

(A) Box plot depicting post-hatching limb stages at 20°C (n = 35 total). Central line within the box = median number of larvae for that stage; edges of the box = 25th and 75th quartiles and the whiskers = the interquartile range. (B) Normal, identifiable stages viewed dorsally with anterior (A) at the top and posterior (P) at the bottom (not to scale). For scaled limb stages see Figure 1—figure supplement 1. Key stage indicators used: st. 45: bending of limb bud at anterior body wall junction (bt1), st. 46: length doubling along proximodistal axis, st. 47: dorsoventral flattening of distal bud (flt), st. 48: notch (nt) appears separating digit 1 and 2, st. 49: ulnar bulge (ub) appears, st. 52: elbow bend appears (bt2) along with prominent separation of three digits and st. 54: digit four present. (C) H and E stained forelimb buds at stages 44 (C’–C”) and 45 (C’”) show no evidence of an AER and skeletal muscle (red arrows) underlies the emerging limb bud. White dotted lines (C”–C”’) delineate mesenchyme from ectoderm. Ectoderm covering the limb is consistent with flank ectoderm (1–2 cells thick). (D) Alcian blue-Alizarin red staining reveal preaxial pattern of chondrogenesis (n = 3–4/stage). Cartilage condensations abbreviated: humerus (hu), radius (r), ulna (ul), carpal (car), metacarpal (met) and digits 1–4 (I–IV). (E) Expression domains of the pre-condensation marker Sox9 during forelimb development. Sox9 is diffusely expressed at stage 45, whereas three discreet domains of Sox9 expression begin at stage 46 marking the ulna (ul), radius (r) and carpals (car). Sox9 expression in presumptive digit II is first observable at stage 47 and precedes digit I expression which occurs during late stage 48. Sox9 expression in the remaining digits occurs anterior to posterior.

https://doi.org/10.7554/eLife.48507.003

Shh-expressing cells in the presumptive ZPA are proximally restricted from the autopod and contribute exclusively to digit IV

Having established the spatiotemporal pattern of mesenchymal specification during skeletal formation, we next sought to investigate the expression of key genes involved in limb bud outgrowth and anterior-posterior patterning. For this, we optimized wholemount in situ hybridization such that we could accurately identify mesenchymal and ectodermal expression (Figure 2—figure supplement 1). During limb initiation in other tetrapods, Sonic hedgehog (Shh) is induced in the posterior limb mesenchyme (Riddle et al., 1993). Using an antisense-probe (~1000 bp) that was a gift from the Tanaka lab (IMP, Austria), we determined that Shh was first expressed at stage 45 in a posteriorly restricted domain and persisted through stage 49 in a posterior-proximal position (Figure 2A). Compared to anuran and other amniote forelimb buds where Shh expression precedes expression of Hoxd11, the onset of Shh expression in axolotls appeared relatively late and after Hoxd11 expression (Matsubara et al., 2017; Riddle et al., 1993) (Figure 2—figure supplement 2). While the posterior Shh expression pattern was consistent with previous reports using relatively short probes at several stages (Bickelmann et al., 2018; Imokawa and Yoshizato, 1997; Torok et al., 1999), we also observed Shh expression in an anteriorly restricted domain at stages 46, 47 and 48 (Figure 2A). To support our in situ data, we divided stage 46–49 limb buds into posterior and anterior halves and collected RNA (Figure 2—figure supplement 3A–C). Using qRT-PCR, we observed Shh expression from the anterior and posterior compartments (Figure 2—figure supplement 3D). We also isolated full-length Indian hedgehog (Ihh) and aligned it with Shh revealing that our RNA probe shared 42% similarity between sequences (Supplementary file 1). Furthermore, we generated an RNA probe to Ihh and found spatiotemporal expression domains distinct from Shh expression localized to areas of skeletal ossification (Figure 2—figure supplement 2).

Figure 2 with 3 supplements see all
Shh-expressing cells in the presumptive ZPA are proximally restricted from the autopod and contribute progenitors exclusively to digit IV.

(A–E) Dorsal views of stage 44–49 axolotl forelimbs with anterior (A) on top and posterior (P) on the bottom of each panel. Red arrows indicate expression. (A) Shh expression is first detected in a small posterior domain at stage 45 and persists through stage 49. Posterior expression is never detected in the autopod. An anterior domain of Shh expression is visible at stages 46–48. Ptch1 expression overlaps with and surrounds Shh expression posteriorly and anteriorly. (B) Gli1 expression is detected at stage 46 adjacent to, and slightly overlapping with, Shh and Ptch1 expression. Gli1 also exhibited a small anterior domain at stages 46–48. Gli3 is expressed distally across the anterior-posterior axis between stages 45 and 48 with very weak expression overlapping the ZPA. (C) Shh domain in a stage 46 axolotl forelimb, HH26 chicken wing and E11.5 mouse forelimb. Scale bar = 100 µm (axolotl), 200 µm (chicken and mouse). (D) Schematic representation of Shh domain (indigo) compared to mesenchymal condensations (dotted) and Alcian blue stained cartilage condensations (blue) in a stage 46 axolotl forelimb, HH26 chicken wing (Montero et al., 2017) and E11.5 mouse forelimb (Taher et al., 2011). Cartilage condensations abbreviated: humerus (hu), radius (r) and ulna (ul). (E) DiI injections within the approximate Shh domain at stage 46 using light and fluorescence microscopy. Fluorescently labeled cells were followed through stage 53. Inset images at stages 48 and 53 show the migratory route of the DiI labeled mesenchymal cells (red arrow) to digit IV (n = 34). Yellow arrows represent pigment cells (autoflorescence) and ectodermal cells that have picked up the DiI.

https://doi.org/10.7554/eLife.48507.006

We next examined the spatiotemporal expression of the hedgehog receptor Patched 1 (Ptch1) and effector molecule Gli1, both of which are direct targets of Shh-signaling (Figure 2A–B). Ptch1 expression was localized in two broad domains corresponding to the posterior and anterior domains of Shh expression (Figure 2A). While we observed Gli1 expression in an anterior domain similar to Ptch1 expression, the posterior expression of Gli1 appeared to localize more closely with Ihh (Figure 2B and Figure 2—figure supplement 2). Lastly, we examined the expression of the repressor form of Gli3 which serves to restrict Shh to the posterior of the limb bud in mice and chickens and found a broad expression pattern across the anterior-posterior axis that excluded the presumptive ZPA at stages 45, 46 and 47 (Figure 2 and Figure 2—figure supplement 3D).

Although Shh expression remains posterior in the forelimb buds of chicks and mice, it tracks distally into the future autopod where it maintains a close association with the AER (Figure 2C–D). In contrast, Shh expression in axolotl forelimb buds did not appear in the autopod (Figure 2A,C–D). Given this proximally restricted position of the Shh domain we asked whether Shh-expressing cells (or those close by) contribute to the digits. To track ZPA cells we injected DiI into the approximate position of Shh expression around stage 45 and monitored fluorescence till stage 53 (Figure 2E). Although our injections labeled Shh-expressing cells and nearby cells as well, we only observed cells migrating along the posterior limb margin and contributing to digit IV (Figure 2E). This data mimics labeling experiments in chick limbs where ZPA cells only give rise to the most posterior digit in the hindlimb and posterior margin of the forelimb (Towers et al., 2008; Towers et al., 2011). Thus, despite some late spatial expression differences, our data suggests a conserved role for Shh during forelimb development.

Amniote and anuran AER-specific Fgf ligands (8, 9, 17) are expressed exclusively in axolotl limb mesenchyme

Proximal-distal outgrowth of the limb bud and maintenance of Shh-signaling from the ZPA are regulated in tetrapods by the AER, and specifically by Fgf-signaling (Lewandoski et al., 2000; Mariani et al., 2008; Niswander et al., 1993; Saunders, 1948). Several Fgfs are expressed in the anuran and amniote AER (i.e. Fgf4, 8, 9, 17), but Fgf8 alone is required for cell survival and limb bud outgrowth (Lewandoski et al., 2000; Sun et al., 2002). Although an AER does not form during limb development in the direct developing frog Eleutherodactylus coqui (Gross et al., 2011) or the marsupial Monodelphis domestica (Doroba and Sears, 2010), AER-Fgfs are still restricted to, and expressed, in the ectoderm. Because salamanders lack a morphological AER, we asked if Fgf4, 8, 9 and 17 were expressed in the axolotl limb bud ectoderm. In contrast to anurans and amniotes, we found that Fgf8, Fgf9 and Fgf17 were solely expressed in the mesenchyme (Figure 3A–C and Supplementary file 1). We could not consistently detect Fgf4 during limb development and, when we did, it was expressed at very low levels in the mesenchyme only (Figure 3—figure supplement 1). At stage 44, we detected Fgf8 in a broad mesenchymal zone directly beneath the ectoderm (Figure 3A). Fgf8 expression persisted in the distal mesenchyme until stage 47 when it segregated into symmetrical domains within the dorsal and ventral mesenchyme (Figure 3A–B). Although Fgf8 appeared to exhibit an anteriorly restricted expression pattern at stage 44, using qRT-PCR we found Fgf8 expression in both the anterior and posterior compartment at stages 46, 47 and 49 (Figure 2—figure supplement 3D). Fgf9 and Fgf17 showed distally restricted expression at stages 45–46 and Fgf9 appeared to have an additional proximal expression domain at stage 46 (Figure 3C). Fgf17 appeared to have a posterior bias at stage 46 (Figure 3C). Consistent with what is known for amniotes and anurans, we found that Fgf10 was broadly expressed in the distal mesenchyme at stages 45–46 (Figure 3C and Supplementary file 1). We also examined expression of Fgf receptors 1 and 2 (FgfR1 and FgfR2). FgfR1 was first expressed weakly at stage 44 and became more proximally restricted during stages 45–46 (Figure 3D and Supplementary file 1). At stage 46, the proximal-anterior domain of FgfR1 overlaps with the Fgf9 domain (Figure 3C–D). FgfR2 showed weak proximal expression at stage 44 and was later expressed during stages 45–46 in a domain proximal to Fgf8, 9 and 17 (Figure 3A–E and Supplementary file 1). Lastly, we examined Gremlin1 expression in developing salamander limbs, and observed strong mesenchymal expression until digits began condensing at stage 48 (Figure 3A). We also detected Gremlin1 staining at stage 49 in the area destined to become digits III and IV (Figure 3A). Taken together, our expression analysis shows that key Fgf ligands normally expressed in the ectoderm during amniote and anuran limb development are instead, compartmentalized entirely in the limb mesenchyme (Figure 3E).

Figure 3 with 1 supplement see all
Amniote and anuran AER-specific Fgf ligands (8, 9, 17) are expressed exclusively in axolotl limb mesenchyme.

(A, C–E) Dorsal views of stage 44–49 axolotl forelimbs with anterior (A) on top and posterior (P) on the bottom of each panel. Red arrows indicate expression domains. (A) Gremlin1 and Fgf8 expression at forelimb stages 44–49. Gremlin1 is first expressed distally across the anteroposterior axis at stage 45. As the limb bud lengthens Gremlin1 expression becomes centralized at the developing zeugopod and remains strongly expressed through stage 47. Between stages 48 and 49 Gremlin1 expression becomes posteriorly restricted. Fgf8 is expressed exclusively in the mesenchyme (stages 44–48). Fgf8 expression is first detected at stage 44 with a slight anterior bias that expands distally until stage 46 and then shifts proximally. Fgf8 expression was not detected at stage 49. (B) Fgf8 expression at stage 46 begins to segregate dorsoventrally and ultimately separates into separate dorsal and ventral domains at stages 47–48. Anterior view of right limbs with dorsal side on top and ventral side on bottom. (C) Fgf9 shows distal expression at stages 45–46 with an additional proximal domain at stage 46. Fgf17 is expressed distally with a posterior bias at stage 46. Fgf10 maintains distal mesenchymal expression at stages 45–46. (D) FgfR1 and FgfR2 are expressed proximally at stages 44–46. (E) Schematic representation of expression patterns for Fgf8, Fgf9, Fgf17, Fgf10, FgfR1 and FgfR2 at stages 44–46. Black brackets show the ectodermal layer.

https://doi.org/10.7554/eLife.48507.010

Single-cell RNA-seq analysis supports spatial segregation of Fgf ligands and receptors

Our gene expression analysis suggested that Fgf ligands and their cognate receptors might be spatially segregated within the limb mesenchyme. To address this possibility, we analyzed single-cell RNA-seq (scRNA-seq) data from developing axolotl limbs that matched our limb stages 44 and 45 (see Materials and methods) (Gerber et al., 2018). Single-cell data acquired from these developing limbs included only mesenchymal cells (Gerber et al., 2018). Using principal component analysis, we first identified principal component 3 (PC3) as a model for the proximodistal (P-D) axis during stage 45 based on known proximal and distal marker genes (Figure 4A and Figure 4—figure supplement 1). Specifically, we found that proximal (Meis1 and Meis2) and distal markers (Fgf8 and Hoxd11) were expressed in cells on the opposite ends of contributors to PC3 (Figure 4A–B and Figure 4—figure supplement 1). Moreover, Etv4 and Dups6 which are direct targets of Fgf8 were among the top 20 genes contributing to the distal end (Hoxd11+) of PC3 (Figure 4A–B and Figure 4—figure supplement 1). Consistent with expression patterns revealed by in situ hybridization, Fgf ligands such as Fgf8, 9, 17 and 10 were all restricted to the distal end of PC3, while Fgf receptors FgfR1-4 were restricted to the proximal end (Figure 4A–B and Figure 4—figure supplement 1). Despite seeing FgfR1 expressed broadly across the modeled proximal-distal axis, our analysis showed that FgfR1 was expressed to a greater extent in cells at the proximal end of PC3 (Figure 4A–B). Statistical analysis confirmed that Fgf ligands (n = 12) and Fgf receptors (n = 5) were separated along PC3 (Welch Two Sample t-test, T = −4.1588, p=0.0038). We also asked whether Fgf ligands and receptors might still be co-expressed in some cells and found that where FgfR1 was expressed in the distal portion of PC3, some of these cells also expressed Fgf8, Fgf9 and Fgf17 (Figure 4—figure supplement 2). However, we found few to no cells co-expressing FgfR2-4 and Fgf8, 9 and 17 (Figure 4—figure supplement 2). Considered together with our spatiotemporal expression analysis, our scRNA-seq analysis supports compartmentalization of Fgf-signaling within the developing limb mesenchyme and largely points to cellular segregation of ligand-receptor interactions.

Figure 4 with 2 supplements see all
Fgf ligands and receptors are segregated along a modeled proximodistal axis.

(A) Gene expression levels on PCA plot (PC3 vs PC4) for stage 45 limb bud mesenchyme. PC3 models the proximodistal axis of limb bud development exemplified by the expression of Meis1 (proximal) and Hoxd11 (distal). (B) Contributions of axolotl genes to PC3 highlighting the opposite directions of proximal (Meis1) and distal (Hoxd11) markers and that Fgf ligands and receptors are separated along the proximal-distal axis. The sigmoidal curve follows a normal distribution (Shapiro-Wilks normality test, W = 0.95551, p<2.2e-16).

https://doi.org/10.7554/eLife.48507.012

Early inhibition of Fgf-signaling reduces cell proliferation and leads to loss of posterior digits

In amniotes and anurans, a Shh-Grem-Fgf signaling loop regulates proximodistal outgrowth and maintains limb bud mesenchyme in a proliferative, undifferentiated, and multipotent state (Globus and Vethamany-Globus, 1976; Reiter and Solursh, 1982; ten Berge et al., 2008; Towers et al., 2008). To identify if a similar signaling loop exists in salamanders, we tested the functional requirement for Shh- and Fgf-signaling during axolotl limb development. First, in order to explore mesenchymal Fgf-signaling, we used the broad spectrum Fgf-receptor inhibitor (SU5402) that selectively binds to the intracellular kinase domain thereby inhibiting downstream signaling (Mohammadi et al., 1997). We treated axolotl embryos with SU5402 beginning prior to limb bud outgrowth from the flank (at stage 39) and then harvested limbs at stages 45, 46 or 54 (Figure 5A). We administered SU5402 based on a dose-response study and selected a maximum dose that could be delivered continuously which was not toxic to the developing animals (see Materials and methods). Limb buds harvested at stage 45 did not show a significant difference in limb size between treatment and control animals (one-tailed student’s t-test, T = −1.68637, p=0.0514) (Figure 5—figure supplement 1A) whereas those harvested at stage 46 exhibited significantly smaller limbs (one-tailed student’s t-test, T = −8.7759, p<0.0001) (Figure 5B). Although we did find a small, but significant decrease in total animal length (snout to tail-tip length) following SU5402 treatment (one-tailed student’s t-test, T = −4.52, p=0.0007) (Figure 5—figure supplement 1B), this did not account for the size differences between treatment and control limbs (T = −8.1, p<0.44). To test the efficacy of Fgf-signaling inhibition, we determined expression of the Fgf-signaling targets Etv1 and Etv4 (Kawakami et al., 2003) (Figure 5C–E). In response to Fgf inhibition, we were unable to detect Etv1 expression in the limb bud at stages 45 or 46 and Etv4 expression was barely detectable (Figure 5C–D). qRT-PCR confirmed that both targets were significantly down-regulated at stage 46 (Figure 5E).

Figure 5 with 2 supplements see all
Early inhibition of Fgf-signaling reduces cell proliferation and leads to loss of posterior digits.

(A) Design for SU5402 and DMSO treatments. Capital letters refer to harvest stage and figure panel. Red ovals depict dorsal muscle blocks. (B) Limb bud size (area) at stage 46 shows a significant decrease in SU5402-treated larvae (one-tailed student’s t-test, T = −8.7759, asterisk represents p<0.0001, n = 5 for DMSO and n = 6 for SU5402). Error bars represent standard error of mean. Scale bar = 100 µm. (C–D) SU5402 treatment efficiently down-regulates Fgf-signaling targets Etv1 and Etv4 at stages 45 and 46. (E) qRT-PCR analysis of stage 46 limb buds from DMSO (control) and SU5402 treatments indicates down-regulation of Etv1, Etv4, Gremlin1 and Shh. In contrast, Ptch1 expression appeared unaffected in the treated limbs (two tailed t-test; Etv1: T = −19.13, p<0.0001; Etv4: T = −9.79, p=0.0006; Gremlin1: T = −4.14, p=0.014; Shh: T = −6.95, p=0.0022 and Ptch1: T = 2.44, p=0.070; n = 3 for DMSO and n = 3 for SU5402; asterisk depicts significant p-values; n.s = not significant). Relative gene expression depicted as 2-ΔΔCt values with GAPDH as the housekeeping gene. Error bars represent standard error of mean. (F–G) SU5402 treatments cause a down-regulation of Gremlin1 while Ptch1 expression remained unaffected at both stages. (H–I) SU5402-treated stage 46 limbs show a significant decrease in the EdU-positive cells. Lightsheet images are depicted as volume rendered red cell aggregates within a hand drawn limb mask (dorsal view with anterior A on top and posterior P on the bottom), mid-longitudinal sections (blue box = plane of section) of the volume rendered limbs and mid-transverse sections (green box = plane of section) of the volume rendered limbs (H). Abbreviations: dorsal (D) and ventral (V) in the mid-transverse sections. The in-set images on top of the mid-transverse sections depict the orientation of the limb and the plane of section. Cell proliferation is down-regulated throughout the limb, and very few proliferating cells were present in the proximal and distal parts of treated limbs (yellow arrows). Scale bar = 100 µm. (I) Statistical comparisons within control (DMSO) and treatment (SU5402) groups were determined by one-way ANOVA (Kruskal-Wallis tests). n = 3 for DMSO and SU5402. Horizontal bars represent median values; p<0.05. Asterisk depicts significant p-value. (J) Alcian blue stained DMSO (control) and SU5402-treated stage 54 limbs. (K) Drug-treated animals had smaller limbs with 73.5% lacking posterior digit IV (red star in J indicates position of missing digit). Scale bar = 1 mm.

https://doi.org/10.7554/eLife.48507.015

To determine if Fgf-signaling controls Shh-signaling, we quantified Shh expression in response to Fgf inhibition and found that it was reduced compared to control limbs (Figure 5E). Given the relatively small number of Shh-expressing cells at these early time points, we asked if reduced Shh transcription translated into reduced pathway activity as assessed by Ptch1 expression (Figure 5E–G). Using qRT-PCR and in situ, we found that Ptch1 was not reduced in treated limb buds suggesting that Shh pathway activity remained normal (Figure 5E–G). Interestingly, while Gremlin1 is inhibited by AER-Fgf-signaling in amniotes (Merino et al., 1999; Verheyden and Sun, 2008), we found that Gremlin1 expression was virtually eliminated in SU5042-treated limbs (Figure 5E–G).

Loss of AER-Fgf-signaling during mouse limb development does not affect cell proliferation, but produces smaller limbs and proximal truncation due to increased proximal cell death (Mariani et al., 2008; Sun et al., 2002). We used Lysotracker to mark dying cells (Mariani et al., 2008; Seifert et al., 2009) and during normal limb development we did not detect any dying cells (Figure 5—figure supplement 2). Similarly, when we inhibited Fgf-signaling we did not detect dying cells in limb buds (Figure 5—figure supplement 2). We next assessed if smaller limbs might have resulted from alterations in cell proliferation (Figure 5H–I). Inhibiting Fgf-signaling as above, we examined actively proliferating cells (EdU+) in treated and control stage 46 limb buds using lightsheet microscopy (Figure 5H–I). We calculated the proliferative population as a fraction of total limb volume and observed an 83% decrease in actively proliferating cells (one-way ANOVA, Kruskal-Wallis test, p<0.05) (Figure 5I). While the decrease in proliferating cells appeared proportionally across the anteroposterior axis, we noted that proliferating cells were nearly absent from the proximal limb bud and from a small distal domain in treated limbs (Figure 5H).

To determine the ultimate effect of Fgf inhibition on limb development, we assessed the effect of treating embryos from before limb outgrowth until all four digits appeared in control limbs at stage 54 (Figure 5J–K). Analyzing the forelimb skeleton, we found that inhibiting Fgf-signaling led to smaller, but nearly complete limbs which generally lacked posterior digit IV (Figure 5J–K). Specifically, 73.5% of SU5402-treated embryos exhibited this phenotype (25/34) while ~11% (4/34) had fewer than three digits (Figure 5I). In those cases where more than one digit was lost, the missing digits were the next most posterior in sequence. These results reveal that Fgf-signaling regulates mesenchymal proliferation, but not cell survival during axolotl limb development. Thus, inhibiting Fgf-signaling leads to smaller limbs and loss of posterior digits, a result consistent with colchicine treatment of developing axolotl and Xenopus larva (Alberch and Gale, 1983). Lastly, Fgf-signaling regulates Gremlin1 expression, whereas Shh-signaling is relatively independent of Fgf-signaling.

Sonic hedgehog signaling regulates Fgf8 expression during axolotl limb development

While our results above suggest that a Shh-Grem-Fgf signaling loop does occur during axolotl limb development, we sought to examine if Fgf-signaling was reliant on Shh-signaling. Using cyclopamine to inhibit Shh signal transduction, previous work showed that Shh-signaling controls anterior-posterior patterning of the zeugopod and autopod during axolotl limb development, a role consistent with ZPA function in other tetrapods (Stopper and Wagner, 2007). Cyclopamine-treated axolotl limbs phenocopied Shh-/- mouse limbs with significant proximodistal outgrowth, fusion of the radius/ulna, and almost complete loss of the autopod (Chiang et al., 2001). To analyze the interaction of Shh- and Fgf-signaling during limb bud outgrowth, we treated stage 39 larvae with cyclopamine for 10 days and analyzed the limb buds at stages 45 and 46 (Figure 6A). Limb buds harvested at stage 45 did not show a significant difference in limb size between treatment and control animals (one-tailed student’s t-test, T = 1.43, p=0.909) (Figure 5—figure supplement 1C), whereas at stage 46, limb buds from cyclopamine-treated larvae were significantly smaller compared to control treated limbs (one-tailed student’s t-test, T = 8.36, p<0.0001) and this effect was independent of (T = 0.03, p=0.975) a small, but significant decrease in animal length (one-tailed student’s t-test, T = 3.87, p<0.0016) (Figure 6B and Figure 5—figure supplement 1D). Analyzing Ptch1 expression by qRT-PCR and in situ hybridization, we found that cyclopamine treatment efficiently inhibited hedgehog signaling in limb buds (Figure 6C–E). Using qRT-PCR we saw a significant reduction in Fgf8 and Gremlin1 expression at stage 46 and using in situ we were unable to detect these genes at either stage 45 or 46 (Figure 6B–D). These data place Fgf8 and Gremlin1 downstream of Shh-signaling during axolotl limb development. We also tested whether Shh-signaling controlled cell survival and cell proliferation. Similar to our results using SU5042, when we inhibited Shh-signaling we did not observe cell death in developing limb buds (Figure 5—figure supplement 2). However, we did find that loss of Shh-signaling led to a 53% reduction in cell proliferation (Figure 6F–G). Whereas Fgf-inhibition led to proportionally smaller limbs, Shh-inhibition led to a dramatic loss of cell proliferation in the distal limb bud (one-way ANOVA, Kruskal-Wallis test, p<0.05) (Figure 6F). Together, our findings place Fgf8 downstream of Shh-signaling and suggest that Shh-signaling also controls cell proliferation, although more specifically in the distal limb bud where digit progenitors reside.

Sonic hedgehog signaling regulates Fgf8 expression and distal cell proliferation during axolotl limb development.

(A) Design for cyclopamine and ethanol treatments. Capital letters refer to harvest stage and figure panel. Red ovals depict dorsal muscle blocks. (B) Limb bud size (area) at stage 46 shows a significant decrease in cyclopamine-treated larvae (one-tailed student’s t-test, T = 8.36, p<0.0001, n = 6 each for ethanol and cyclopamine). Scale bar = 100 µm. (C) Real-time PCR analysis of Ptch1, Gremlin1 and Fgf8 in stage 46 limb buds from ethanol (control) and cyclopamine treatments (one tailed t-test; asterisk depicts significant p-values; Ptch1: T = 4.95, p=0.0039; Gremlin1: T = 5.97, p=0.002 and Fgf8: T = 11.22, p=0.0002; n = 3 for ethanol and cyclopamine). Relative gene expression depicted as 2-ΔΔCt values with RPL32 as the housekeeping gene. Error bars represent standard error of mean. (D–E) Cyclopamine treatment efficiently down-regulates Ptch1 and therefore Shh-signaling. Cyclopamine treatment also down-regulates Gremlin1 and Fgf8. (F–G) Cyclopamine-treated stage 46 limbs show a significant decrease in EdU-positive cells. Lightsheet images depicted as volume rendered red cell aggregates within a hand drawn limb mask (dorsal view with anterior A on top and posterior P on the bottom), mid-longitudinal sections (blue box = plane of section) of the volume rendered limbs and mid-transverse sections (green box = plane of section) of the volume rendered limbs. Abbreviations: dorsal (D) and ventral (V) in the mid-transverse sections. The in-set images on top of the mid-transverse sections depict the orientation of the limb and the plane of section. There is a domain specific loss of cell proliferation in the proximal and distal parts of the treated limbs (yellow arrows). Scale bar = 100 µm. (G) Comparisons within control (ethanol) and treatment (cyclopamine) groups were determined by one-way ANOVA (Kruskal-Wallis tests); n = 3 for ethanol and cyclopamine. Horizontal bars represent median values; p<0.05. Asterisk depicts significant p-value.

https://doi.org/10.7554/eLife.48507.018

Discussion

Comparing forelimb development between urodeles, anurans and amniotes

Urodeles were among the first vertebrates used to study limb field specification and morphogenesis (Harrison, 1918; Stocum and Fallon, 1982; Swett, 1937). Although chick and mouse embryos largely replaced urodeles as model systems to study limb development, a perpetual fascination with understanding the molecular basis of limb regeneration has resurrected interest into amphibian limb development (Gerber et al., 2018; Keenan and Beck, 2016; Stocum, 1975). Elucidating the cellular and molecular basis of limb development in urodeles and other amphibians also has important implications for our understanding of how limbs evolved (Alberch and Gale, 1983; Fröbisch and Shubin, 2011; Stopper and Wagner, 2005). Despite deep homology among tetrapod limbs, biologists have long recognized several unique aspects of urodele limb development (Holmgren, 1933). For example, although the limb field in urodeles is established in the gastrula, the limb bud does not emerge from the flank until much later in a free-swimming larva. Unlike amniotes, this situation demonstrates a level of autonomous development that is temporally de-coupled from limb specification and patterning of the main body axis (Stocum and Fallon, 1982). Retinoic acid (RA) generated from the somites in chicks and mice diffuses into the lateral plate mesoderm where it permits correct spatiotemporal induction of Tbx5 (Cunningham et al., 2013; Nishimoto et al., 2015; Stephens and McNulty, 1981). In contrast, during anuran limb bud development which occurs in larval animals (similar to urodeles), retinoic acid (RA) appears to be generated autonomously in the forelimb bud with raldh2 expressed proximally and cyp26b distally (McEwan et al., 2011). In addition to limb heterochrony, urodele limb buds do not form an apical ectodermal ridge (AER) (Sturdee and Connock, 1975; Tank et al., 1977) and exhibit preaxial dominance of the limb skeleton (Shubin and Alberch, 1986). Lastly, urodeles can regenerate an entire limb, something no other group of tetrapods can do as adults (Table 1). Thus, while these aspects of urodele limb development challenge our notion of an inclusive vertebrate limb development model (Zeller et al., 2009) they also beg the question; does the molecular machinery directing limb morphogenesis exhibit critical differences when compared to amniotes and anurans?

Table 1
Salient features of forelimb development between axolotl, Xenopus, chicken and mouse.
https://doi.org/10.7554/eLife.48507.019
FeatureAxolotlXenopusChickenMouseReferences
Autopod skeletal differentiationPreaxial dominance1Postaxial dominance1Postaxial dominance1Postaxial dominance11 Shubin and Alberch, 1986
Location of ZPA domainPosterior. Excluded from autopod1,*Posterior. Extends into autopod2Posterior. Extends into autopod3Posterior. Extends into autopod41 Torok et al., 1999
2 Endo et al., 1997
3 Riddle et al., 1993
4 Echelard et al., 1993
* This study
Contribution of ZPA cells to posterior digit(s)Yes (DiI labeling)*?Yes (DiI labeling)1
No (GFP grafting)2
Yes (Genetic labeling)31 Towers et al., 2008
2 Towers et al., 2011
3 Harfe et al., 2004
* This study
Shh-signaling during limb developmentKey mediator of anterior-posterior patterning1Key mediator of proximal-distal and anterior-posterior patterning2Key mediator of anterior-posterior patterning3Key mediator of anterior-posterior patterning4, 51Stopper and Wagner, 2007
2Stopper et al., 2016
3 Ros et al., 2003
4 Chiang et al., 1996
5 Chiang et al., 2001
Morphological AERNo1Transient2Yes3Yes41 Tank et al., 1977
2 Tarin and Sturdee, 1971
3 Saunders, 1948
4 Wanek et al., 1989
Molecular AERNo*YesYesYes* This study
Location of AER-specific Fgfs (4,8,9,17)Mesenchyme*AER1AER2, 3, 4AER2, 5, 6, 7, 8, 91 Christen and Slack, 1997
2 Mahmood et al., 1995
3 Duprez et al., 1996
4 Havens et al., 2006
5 Ohuchi et al., 1994
6 Crossley and Martin, 1995
7 Niswander and Martin, 1992
8 Sun et al., 2002
9 Sun et al., 2000
* This study
Location of Fgf receptorsFgfR1-4 expressed exclusively in the mesenchyme*?FgfR1IIIc, FgfR2IIIb, FgfR3IIIb expressed in the ectoderm and FgfR1IIIb, FgfR2IIIc, FgfR3IIIb/c and FgfR4 expressed in the mesenchyme1, 2FgfR1IIIb and FgfR2IIIb expressed in the ectoderm and FgfR1IIIc, FgfR2IIIc, FgfR3c and FgfR4c expressed in the mesenchyme3, 4, 51 Havens et al., 2006
2 Sheeba et al., 2010
3 Min et al., 1998
4 MacArthur et al., 1995
5 Ornitz and Itoh, 2015
* This study
Fgf-signaling during limb developmentControls cell proliferation and limb size*?Regulates proximodistal patterning, cell survival and cellular differentiation1, 2, 3,4Regulates proximodistal patterning, cell survival and cellular differentiation5, 61 Saunders, 1948
2 Summerbell, 1974
3 Janners and Searls, 1971
4 Dennis Summerbell, 1977
5 Sun et al., 2002
6 Mariani et al., 2008
* This study
Positive regulation of Gremlin1 by Fgf-signaling in the limbYes*?No1No21 Merino et al., 1999
2 Verheyden and Sun, 2008
* This study
Positive regulation of Shh-signaling by Fgf-signaling in the limbNo*?Yes1Yes21 Crossley et al., 1996
2 Lewandoski et al., 2000
* This study
Positive regulation of Fgf-signaling by Shh-signaling in the limbYes*?Yes1Yes21 Laufer et al., 1994
2 Harfe et al., 2004
* This study
Limb regenerationRegenerates complete limb through adulthood1, 2, 3Regenerates complete limb through larval stage 534Does not regenerate5Does not regenerate. *Regenerates digit tip61 Young et al., 1983a
2 Young et al., 1983b
3 Monaghan et al., 2014
4 Dent, 1962
5 Muneoka and Sassoon, 1992
6 Borgens, 1982

In this study, we examined several key aspects of salamander limb development as they relate to patterning and outgrowth of the tetrapod limb. In doing so, we considered salient features of salamander forelimb development as they compare to Xenopus, chickens and mice (summarized in Table 1). First, we confirm previous reports that axolotls lack a morphological AER. Second, using Sox9 expression, we show that digit specification occurs first along the metapterygial axis of the limb with digit II, and then proceeds postaxial with digits I, III and IV. However, using Alcian blue staining as a proxy for cartilage condensation, we show that digits I and II differentiate simultaneously and are followed in sequence by digits III and IV. Thirdly, we show that Shh is restricted posteriorly, and although Shh expression does not overlap with the autopod, ZPA cells contribute to digit IV. Together with cyclopamine experiments these data support Shh-signaling as a key mediator of anterior-posterior patterning. Fourthly, we focused on the cellular source of Fgf-signaling and find that, in contrast to anurans and amniotes where reciprocal signaling is compartmentalized between the limb ectoderm and mesenchyme, Fgf ligands (Fgf8, 9, 17, 10) and receptors (FgfR1-4) are all expressed solely in the mesenchyme. By functionally testing the requirement for Fgf-signaling using a broad Fgf-receptor antagonist (SU5042), we demonstrate that Fgf-signaling regulates limb size by controlling cell proliferation across all three limb axes. Again, this stands in contrast to anurans and amniotes where Fgf-signaling regulates cell survival and cellular differentiation along the proximal-distal axis. Another key finding from these experiments is that Fgf-signaling regulates Gremlin1, whereas Shh-signaling is maintained in the face of Fgf-inhibition. While these results strongly suggest that a Shh-Grem-Fgf signaling loop is not present during salamander limb development, they do show that Fgf8 expression is dependent on Shh-signaling. Together, our results show Shh-signaling from the ZPA has maintained its core function while de-coupling itself from Fgf-signaling and that Fgf-signaling has evolved to regulate cell proliferation in the limb.

Skeletal differentiation during axolotl forelimb development

Our analysis of skeletogenesis shows that axolotl digits appear to be specified in a different order than they differentiate. Taking advantage of subtle variations in within animal and between animal staging, our analysis using Sox9 expression shows a 2>1>3>4 pattern of digit specification. However, when we observed digit differentiation using Alcian blue to label condensing cartilage we always found digit I and II appearing together followed in sequence by digits III and IV. This pattern of digit differentiation is consistent with that observed by other investigators using histological preparations (Fröbisch, 2008; Shubin and Alberch, 1986) or Alcian blue (Nye et al., 2003). These findings support independent molecular control of digit specification and differentiation and hint at the wide diversity in ontogenetic and heterochronic shifts that have occurred in the limb during urodele evolution (Blanco and Alberch, 1992; Franssen et al., 2005). Similarly, Sox9 expression shows spatiotemporal variability across urodeles (Kerney et al., 2018) and this underscores the need for more comparative limb studies to better understand the ancestral condition. While preaxial vs. postaxial dominance of skeletal formation clearly separates urodele limb development from anurans and amniotes, it is likely this difference points to alterations in the upstream genetic control of digit specification involving Fgfs, Bmps, and retinoic acid (Montero et al., 2017; ten Berge et al., 2008).

A conserved role for Sonic hedgehog signaling during vertebrate limb development

With respect to anterior-posterior patterning, our data show that the members of the hedgehog signaling pathway are expressed in a mesenchymal pattern consistent with other tetrapods studied to date. However, our results also reveal three important differences. First, our data reveals that Shh expression appears relatively late during limb bud outgrowth and after Fgf8 and Hoxd11 are expressed. Second, salamander forelimb buds exhibit a proximal-anterior domain of Shh expression that emerges as the axolotl limb bud begins to bend posteriorly; a domain which is not found during normal limb development in amniotes and anurans. Third, posterior Shh expression corresponding to the ZPA remains in a relatively small proximal domain (Torok et al., 1999) rather than the elongated expression domain found in other tetrapods that extends the proximal-distal length of the autopod (Matsubara et al., 2017; Riddle et al., 1993; Shapiro et al., 2003). With respect to the first point, previous work has demonstrated that flank tissue surrounding the limb field plays an important role in specifying the anteroposterior axis and thus, the temporal appearance of Shh expression may be less important than its posterior induction (Stocum and Fallon, 1982). In chick limbs, Hoxd11 can be induced by Shh, but only in proximity to the AER or in the presence of AER-Fgfs (Laufer et al., 1994) and in Shh KO mice Hoxd11 is still expressed in the limb bud (Chiang et al., 2001). While intriguing, the transient appearance of an anterior Shh domain is difficult to explain. We did not detect an anterior necrotic zone (Figure 5—figure supplement 2), and Shh inhibition did not lead to increased cell death in this region. Moreover, Shh inhibition produced phenotypes consistent with other amniote models of limb development (Chiang et al., 2001; Scherz et al., 2007; Stopper and Wagner, 2007; Towers et al., 2008; Vargas and Wagner, 2009; Zhu et al., 2008). When Shh-signaling is inhibited in axolotls with cyclopamine, zeugopodial and autopodial skeletal elements are lost in a posterior to anterior direction that is dependent on when cyclopamine is administered (Stopper and Wagner, 2007). A similar phenotype occurs in chick embryos where elegant work demonstrated that Shh-signaling controls digit progenitor specification and limb bud growth (Towers et al., 2008). When coupled with the results of Shh inhibition (Stopper and Wagner, 2007), our observations that Shh-signaling controls distal cell proliferation and that Shh-expressing cells contribute to digit IV, our findings support a conserved role for Shh-signaling as it pertains to specifying digit progenitors during limb development across all tetrapods. Future experiments ectopically expressing Shh using beads or virus will serve as a means to test downstream targets of Shh-signaling, as will transgenic elimination of Shh during development.

Developing urodele limbs lack a morphological and molecular AER

In amniotes and anurans, limb bud outgrowth and proximal-distal patterning are controlled by the AER via expression of specific Fgf ligands (e.g. Fgf4, 8, 9, 17) (Cohn et al., 1995; Lewandoski et al., 2000; Mariani et al., 2008; Ohuchi et al., 1997; Sun et al., 2002). The AER maintains ZPA activity (Vogel and Tickle, 1993) and AER-Fgfs and Fgf-regulated Etvs are essential in inducing and maintaining posterior expression of Shh in the ZPA (Niswander, 2002; Zhang et al., 2009). Furthermore, Gremlin1 acts as an intermediary between these signaling centers where it relays signals from the ZPA to control Bmp-signaling and maintain the AER (Zeller et al., 2009; Zuniga, 2015). In stark contrast to anurans and amniotes, our data represents the first instance of a tetrapod lacking Fgf ligand and receptor expression in the developing limb bud ectoderm. Although previous studies suggested that Fgf8 was expressed early in salamander limb bud ectoderm (Han et al., 2001), a result consistent with studies in Monodelphis domestica (Doroba and Sears, 2010) and Eleutherodactylus coqui (Gross et al., 2011) which lack an AER but exhibit ectodermal Fgf8 expression, our findings during development show this is not the case. Moreover, our expression patterns for Fgf8 are consistent with recent data from regenerating axolotl limbs where Fgf8 expression was found restricted to the mesenchyme (Nacu et al., 2016). Consistent with our spatiotemporal expression results, analysis of scRNA-seq data from stage 45 axolotl limb buds (Gerber et al., 2018) showed similar patterns along a modeled proximodistal axis. Co-expression analysis of Fgf ligands and receptors at stage 45 provide evidence that some cells expressing FgfR1 also express several Fgf ligands. Given the relatively small number of cells analyzed in this dataset (<200), a more complete scRNA-seq analysis across multiple time points will help address the complexity of Fgf-signaling and whether mesenchymal cells secreting Fgf ligands respond in an autocrine fashion. Our experiments also call into question what role, if any, the ectoderm plays during urodele limb development. For instance, experiments in Pleurodeles waltl showed that early stage forelimb bud mesoderm could autonomously develop a fully formed limb when grafted under heterologous flank epidermis (Lauthier, 1985). A clear functional role of the ectoderm awaits genetic manipulation since manual removal of the ectoderm results in rapid regeneration.

Fgf-signaling regulates cell proliferation and limb size

Our experiments using the Fgf receptor antagonist SU5402 sought to test if spatial re-location of Fgf ligands and receptors affected the established function of Fgf-signaling during anuran and amniote limb bud development. SU5402 inhibition (beginning before limb bud outgrowth) revealed that the primary function of Fgf-signaling in axolotl limb development is to regulate cell proliferation throughout the limb bud. It does not, however, appear to control cellular differentiation, cell survival or proximodistal patterning. Removal of the AER in chickens (Saunders, 1948) leads to a stage dependent loss of skeletal elements in a proximal to distal direction and combined genetic deletion of Fgf8/Fgf4/Fgf9 in the mouse AER (Mariani et al., 2008) leads to a complete loss of the stylopod, zeugopod and autopod. In contrast, broadly inhibiting Fgf-signaling from the outset of axolotl limb development phenocopies urodele limbs treated with the mitotic inhibitor colchicine where in both cases, the most posterior digit fails to form in an otherwise smaller, but normal limb (Alberch and Gale, 1983). These results echo experiments in chickens where pharmacological inhibition of cell proliferation using trichostatin A, colchicine, or vinblastine leads to loss of anterior digits, which are the last to form (Towers et al., 2008). Broadly inhibiting Fgf-signaling also shows that Shh operates independently of Fgf-signaling, while our cyclopamine experiments show that Shh-signaling regulates Fgf8 expression. In addition, our finding that Gremlin1 is regulated by Fgf-signaling suggests further rearrangement of genetic interactions observed in amniotes (Lewandoski et al., 2000; Sun et al., 2002). Together, these data show that while the role of Shh-signaling from the ZPA is conserved in salamander limb development, movement of amniote/anuran AER-Fgf ligands to the mesenchyme was accompanied by a change in how Fgf-signaling regulates limb development. Ultimately, our findings support similar molecular players being deployed during limb development across tetrapods but demonstrate that a divergent molecular program in urodeles resides predominantly in one cellular compartment: the limb mesenchyme.

While it is tempting to speculate that mesenchymal compartmentalization of limb developmental signaling is somehow causally related to regenerative ability, available data suggests otherwise. Data from basal Actinopterygians (paddlefish), Chondrichthyians (catfish) and anurans show ectodermal-mesenchymal segregation of the Shh- and Fgf-signaling (Christen and Slack, 1998; Tulenko et al., 2017). Pectoral fin regeneration is an ancient feature of Actinopterygians, Sarcopterygians and Chondrichthyians suggesting that mesenchymal core signaling alone is not exclusive to regenerating species. On the flip side, anurans show subtle molecular differences during limb development compared to amniotes, especially along the dorsoventral axis (Christen and Slack, 1998). Thus, it is plausible that ectodermal-mesodermal compartmentalization was the first step toward developmental canalization that ultimately increased robustness in the limb program, specifically in the autopod. Future limb development studies using a diverse array of anurans and lungfish will help shed light on these questions as will continued studies in salamanders and newts.

Materials and methods

Animals and tissue harvest

Request a detailed protocol

Axolotls (Ambystoma mexicanum) were acquired from the Ambystoma Genetic Stock Center (Lexington, KY) and from our own laboratory colony. Chicken eggs (University of Kentucky, Department of Animal Sciences) were incubated to stage and mouse embryos were harvested from Swiss Webster mice (ND4, Envigo, Indianapolis, IN). All procedures were conducted in accordance with, and approved by, the University of Kentucky Institutional Animal Care and Use Committee (IACUC Protocol: 2013–1174). Axolotl embryos were kept at 20–21°C and reared in glass bowls (7.5’/6’/2.5’ dimensions) in 800 ml 20% Holtfreter’s solution. 15–20 larvae were kept in a single bowl and were fed with brine shrimps from ~3 weeks post-fertilization. Larvae used for drug treatments, proliferation, cell death assays, qRT-PCR and DiI labeling were reared in 6-well plates. Larvae used for forelimb staging and area/snout to tail-tip measurements, Alcian blue-Alizarin red staining, in situ hybridization, histology and proliferation assays were anesthetized using 1x benzocaine (Sigma) and fixed overnight in 4% paraformaldehyde at 4°C. Larvae used for cell death assay were rinsed gently in Hanks BSS four times, five mins each on a rocker, fixed overnight in 4% PFA post-Lysotracker treatment. Developmental stages were referenced against previously reported post-hatching stages (Nye et al., 2003) and were extended as outlined in Figure 1. Individual animals (n = 35) were examined every day to assess limb stage. Larvae used for qRT-PCR were anesthetized using 1x benzocaine (Sigma), limb tissue samples were snap frozen and stored at −80°C until further use. Chicken (HH25) and mouse (E11.5) embryos were harvested, fixed overnight in 4% paraformaldehyde at 4°C and processed for in situ hybridization experiments.

Alcian blue and Alizarin red staining

Request a detailed protocol

Larvae were fixed overnight, washed three times, 10 mins each in PBT (Phosphate Buffer Saline and 1% Tween 20), dehydrated in graded ethanol series 25%, 50%, 75% and stored in 100% ethanol at −20°C until further use. Dehydrated larvae were washed three times, 10 mins each in 1x PBS (Phosphate Buffer Saline), stained with 0.02% Alcian blue 8GX (Sigma Aldrich) in 70% ethanol and 30% glacial acetic acid for 3 hr to overnight. Stained larvae were rehydrated in graded ethanol series (100%, 75%, 50% and water 1 hr each) and stained with 0.1% Alizarin Red (Sigma Aldrich) 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).

Histology

Request a detailed protocol

Larvae were fixed overnight, washed twice in 1x PBS and stored in 70% ethanol at 4°C until further use. Larvae were then processed for paraffin embedding (Histo5 Tissue Processor, Milestone) and tissue samples were sectioned at 5 μm. H and E staining was done on deparaffinized and rehydrated sections. For bright-field visualization of limb buds, Mayer’s haematoxylin was used to counterstain the nuclei and coverslips were mounted with Cytoseal XYL (ThermoFisher, Waltham, MA).

Gene isolation and riboprobe synthesis

Request a detailed protocol

Coding sequences for axolotl genes were obtained from NCBI, (Bryant et al., 2017) or www.ambystoma.org (Supplementary file 1). Axolotl sequences were aligned with human homologs to locate the 3’UTRs. Primers were designed against (when possible) or close to 3’UTRs of the axolotl sequences. Briefly, RNA was extracted using Trizol reagent from stage 31 larvae, stage 32 larvae or regenerating early forelimb buds from 5 to 10 cm juveniles. cDNA was synthesized from 1 to 0.5 μg RNA using SensiFast cDNA synthesis kit. Coding genes sequences were amplified out using Advantage HD polymerase kit and amplified products were ligated into pGEMT-Easy vector (Promega) via T-A cloning using manufacture’s protocol. Plasmids were transformed into Max Efficiency DH5α cells (Invitrogen) and blue-white colonies were obtained. Colony PCR was done to confirm insert sequence size and positive colonies were picked for plasmid mini-prep (Qiagen). Plasmids were sent out for sequencing. Gene sequences and orientation of insertion into vector was verified and positive colonies were used for plasmid maxiprep (Zymo Research). Plasmids containing Shh and Fgf8 genes for chicken and mouse were a gift from the Cohn lab, University of Florida. Plasmids containing Keratin5 and Keratin17 genes for axolotl were a gift from the Satoh lab, Okayama University. Plasmids containing Fgf8 and Gli3 genes for axolotl were a gift from the Tanaka lab, IMP, Austria. ~ 20 μg of plasmid was linearized using specific restriction enzymes to obtain the sense and antisense probe templates.

Sense and antisense probes for axolotl genes were synthesized from 2.5 μg linearized plasmids using DIG RNA Labeling Kit (SP6/T7) (Roche). Sense and antisense probes for chicken and mouse genes were synthesized from 1 μg linearized plasmids using the same kit. In vitro transcription of probes was carried out for 3 hr to overnight at 37°C. Probes were treated with 1unit DNAse (Promega, CAT#M610A) for 1 min at 37°C and reaction was terminated with 2 μl DNAse stop solution (Promega, CAT#M198A). Probes were purified and eluted in 50 μl of nuclease-free water using mini Quick Spin RNA Columns (Roche) or RNeasy MinElute Cleanup Kit (Qiagen), run on 1% agarose gel to access quality and quantified using NanoDrop.

In situ hybridization experiments

Request a detailed protocol

Larvae/embryos fixed overnight were washed three times, 10 min each in PBT (Phosphate Buffer Saline and 1% Tween 20), dehydrated in graded methanol/PBT series 25%, 50%, 75% and stored in 100% at −20°C until further use. For all the experiments at least three larvae or embryos/stage/gene were used. Each axolotl larva was decapitated, and the bottom half of the trunk was amputated. Two axolotl larvae per stage were placed in a DNAse/RNAse free 2 ml tube and treated with 2 ml of each solution. For chicken and mouse, one embryo was placed in a similar 2 ml tube. Dehydrated larvae/embryos were rehydrated in a graded methanol/PBT series 75%, 50%, 25%, washed with PBT twice 5 min each, bleached with 6% H2O2/1x PBS for 1 hr, washed with PBT twice 5 min each. The above steps were done under ice-cold conditions. Larvae/embryos were permeabilized with 20 μg/ml Proteinase K (Roche) in PBS for 7–10 min (40 μg/ml Proteinase K was used for Sox9 and Ihh axolotl genes), washed with PBT twice for 5 min each, fixed with 0.2% Gluteraldehyde/4% paraformaldehyde and washed with PBT twice for 5 min each. The above steps were done at room temperature. Larvae/embryos were incubated overnight in hybridization buffer (5% Dextran sulphate, 2% blocking powder from Roche, 5X SSC, 0.1% TritonX, 0.1% CHAPS from Sigma Aldrich, 50% formamide, 1 mg/ml tRNA from Roche, 5 mM EDTA from Sigma and 50 μg/ml Heparin from Sigma) at 65°C. The tubes were replaced with fresh hybridization buffer, 0.1–1 μg of probe was added into each vial and incubated at 65°C for 2 days. High stringency washes were done with 2X SSC/0.1% CHAPS thrice for 20 min each, 0.2X SSC/0.1% CHAPS 4 times for 25 mins each and with KTBT (15 mM Tris-HCl pH 7.5, 150 mM NaCl, 10 mM KCl and 1% Tween 20) twice for 5 min each. Larvae were blocked with 20% goat serum in KTBT for 3 hr. Later, fresh blocking solution was added. An anti-Digoxigenin-AP, Fab fragment antibody (Roche) was added at 1:3000 dilution and incubated overnight at 4°C. Larvae were washed with KTBT 5 times for 1 hr each and then incubated in KTBT overnight at 4°C. Larvae were washed with NTMT (100 mM Tris-HCl pH 9.5, 50 mM MgCl2, 100 mM NaCl and 1% Tween 20) twice for 5 min each and incubated in NBT/BCIP (Roche) solution in NTMT (BCIP-0.17 mg/ml, NBT-0.33 mg/ml, 10% DMF) till a signal developed with minimum background staining. Limbs were photographed and larvae were washed with TE buffer (10 mM Tris HCl, pH 8 and 1 mM EDTA, pH eight made up in DEPC treated water) 3 times for 10 mins each and fixed in 4% PFA until further use.

Cryosectioning post in situ hybridization experiments

Request a detailed protocol

Larvae from Keratin5 and Keratin17 in situ hybridizations were washed in TE buffer (10 mM Tris HCl pH 8 and 1 mM EDTA pH 8) 3 times for 10 min each, fixed in 4% PFA for 1 hr and washed with 1x PBS 3 times for 5 min each. Larvae were transferred into 2 ml vials with 30% sucrose in 1x PBS and placed on a rotor for 20 mins till they sank to the bottom. Larvae were then placed in OCT for 25 min and frozen for cryo-sections. Cryo-sections were taken at 5 µm thickness, dried overnight at 37°C, fixed with 4% PFA for 10 min, washed with 1x PBS 3 times for 5 min each, treated with Hoechst solution (1:10,000 dilution), air dried, sealed with ProLong Gold antifade reagent (Invitrogen) and imaged.

qRT-PCR analysis

Request a detailed protocol

For drug experiments, post-hatch axolotl larvae were reared in 6-well plates in either 1.5% DMSO, 45 µM SU5402, 0.02% ethanol or 1 µg/ml cyclopamine till stage 46 (see below or treatment details). Whole limbs were dissected from the body wall, immediately snap frozen and stored at −80°C until RNA extraction. Three replicates were used for each condition (treatment/control) and each replicate represented a pool of limbs (both left and right) from 10 to 20 animals.

To validate the anterior versus the posterior expression of Shh, Fgf8 and Gli3 genes, larvae were reared in glass bowls (7.5’/6’/2.5’ dimensions) in 800 ml 20% Holtfreter’s solution. Whole limbs were dissected from the body wall, further dissected into anterior and posterior halves/compartments (Figure 2—figure supplement 3D), immediately snap frozen and stored at −80°C until RNA extraction. n = 2 or 3 was used and each set was pooled from 20 limbs (both right and left).

RNA was extracted using Trizol reagent (Invitrogen). cDNA was synthesized from 1 to 0.5 μg RNA using SensiFast cDNA synthesis kit. The following primers were used for qRT-PCR:

  • Shh Forward- ATTTTTAAGGACGAAGAGAACACCG,

  • Shh Reverse- CTTATCCTTACACCTCTGGGTCATT,

  • Fgf8 Forward- ATTAATTGTGGAAACGGACACCTTC,

  • Fgf8 Reverse- AATCAGCTTTCCCTTCTTGTTCATG,

  • Gli3 Forward- CATGGATGTGGTCGTTATTGATGTG,

  • Gli3 Reverse- GAGGTTATTTACGAGACCGACTGTC,

  • Etv1 Forward- TCTTGGAAGAGTTCTTCTGAGTCAT

  • Etv1 Reverse- CGTGTGAGAAATTGTAACGAGAGA

  • Etv4 Forward- ACTATGCATACGATTCAGATGTTCC

  • Etv4 Reverse- ATAGCCCTCCACGTTCATATACATT

  • Gremlin1 Forward- GGACACCCAGAATACTGAGCA

  • Gremlin1 Reverse- GTAGACCAATCGAAACATCCTGT

  • Ptch1 Forward- TGTAGATCTGCTCCAATGCAAAC

  • Ptch1 Reverse- CTGACCCGGAGTACTTGCAG

  • Tubulin alpha Forward- CCCAGGGCCGTGTTCGTC

  • Tubulin alpha Reverse- CCGCGGGCGTAGTTGTTG

  • GAPDH Forward- AAAAGGTCTCCTCTGGCTATGAC

  • GAPDH Reverse- AGGGCTATAAAAGAGCATTATCGAG

  • RLP32 Forward- AGGCTACTGGGAGTTTTAATAAGGA

  • RLP32 Reverse- AGATTACAGCACCCACTGTCTTTT

Tubulin alpha was used as the internal control/house-keeping gene for limbs obtained from larvae reared in 20% Holtfreter’s solution. GAPDH and RLP32 were used as the internal control/house-keeping genes for the DMSO/SU5402 and ethanol/cyclopamine experiments, respectively, since there was no significant fold change in the 2-Ct values (Schmittgen and Livak, 2008) (Supplementary file 2). 2-ΔΔCt method was used to calculate the fold change values between control (DMSO or ethanol) and treatment (SU5402 or cyclopamine) groups (Schmittgen and Livak, 2008).

DiI labeling of the proximal Shh domain

Request a detailed protocol

Post-hatch axolotl larvae were reared in 6-well plates in 3 ml Holtfreter’s solution. DiI (D-282, molecular probes) was dissolved in dimethyl formamide at 3 mg/ml concentration. Larvae between stages 45 and 46 were anesthetized in 1x benzocaine and approximately 5 nl (0.05–0.20 mm diameter) of DiI was injected into the approximate position of the posterior Shh domain. A total of 34 limbs were analyzed for this experiment. Images were taken immediately to confirm the domain specific restriction of the injection and fluorescence was tracked every 2 days till all four digits formed completely. All images were taken post-anesthesia in 1x benzocaine.

Analysis of single-cell RNA-seq data

Request a detailed protocol

Single-cell RNA-seq data from embryonic limb buds (Gerber et al., 2018, Table S7) were analyzed as follows: read counts from stages 40 and 44 (as reported in Gerber), and which correspond to our stages 44 and 45 (Prayag Murawala, personal communication) were normalized and used as input for principle component analysis. The top twenty loadings of the positive and negative axes of the first five principle components were inspected to identify principle components that segregated developmental axes markers on separate ends. No reliable anterior-posterior, or dorsal-ventral markers segregated in this manner in the current dataset. However, proximal-distal markers (Meis2 and Hoxd11) were segregated on opposite ends of principle component 3 in the PCA of cells from stage 44, but not 40. Due to the low expression levels of most genes of interest, raw read counts were used for visualizing gene expression levels on PCA plots by rescaling the log2 of the read count to span between 0 and 14. To determine co-expression of Fgf ligands and receptors, expression was defined as the presence of one or more reads mapping to the gene in question.

SU5402 and cyclopamine treatments

Request a detailed protocol

Axolotl larvae at pre-limb bud stage 39 were reared in 6-well plates and treated with 3 ml solutions for all drug experiments. A working stock of 3 mM SU5402 (Sigma) was made in DMSO and then diluted to 45 μM in 3 ml 20% Holtfreter’s solution per well. Control larvae were treated with an equivalent amount of DMSO (1.5%) in 20% Holtfreter’s solution. Larvae were kept in the dark and the solution was changed every three days. At three weeks post fertilization each larva was fed 20–30 brine shrimp every day.

Cyclopamine treatments were performed as previously described (Stopper and Wagner, 2007). A working stock of 5 mg/ml of cyclopamine was made in 100% ethanol and 0.6 μl from this stock was added into 3 ml 20% Holtfreter’s solution per well (1 μg /ml final concentration). An equal amount of 100% ethanol (0.02%) was added into control wells. Larvae were kept in dark and solutions were replenished every two days. At three weeks post fertilization each larva was fed 20–30 brine shrimps every day.

Limb area and snout to tail measurements

Request a detailed protocol

Limb area was measured for stage 45 and stage 46 larvae that were reared in 1.5% DMSO, 45 μM SU5402, 0.02% ethanol or 1 μg/ml cyclopamine. Snout to tail-tip lengths were measured for stage 46 larvae that were reared in 1.5% DMSO, 45 μM SU5402, 0.02% ethanol or 1 μg/ml cyclopamine. All measurements were made using Fiji software (NIH) after calibrations.

Cell proliferation assay using EdU labeling

Request a detailed protocol

Post-hatch larvae were reared in 6-well plates in 3 ml of either of the solutions: 20% Holtfreter’s solution, 1.5% DMSO, 45 μM SU5402, 0.02% ethanol or 1 μg/ml cyclopamine. Larvae were additionally treated with 0.1 μg/ml of EdU at stage 46 for 24 hr, fixed overnight in 4% PFA, washed with 1x PBS twice for 5 min, dehydrated in 1x PBS/methanol series (25%, 50%, 75% and 100% methanol, 5 min each) and stored in 100% methanol at −20°C until further use.

For EdU staining, all steps were performed at room temperature unless mentioned otherwise. Larvae were rehydrated backwards through the methanol series starting at 100% methanol and ending at 100% 1x PBS. This was followed by PBT washes for 5 min twice and 2.5% trypsin (Gibco) treatment for 10 min. The larval limbs were checked for clarity at this point. Larvae were washed with water for 5 min twice, treated with 20 μg/ml of proteinase K in PBT for 7–10 min, washed with water for 5 min twice, fixed in 100% acetone at −20°C for 10 min, washed with water for 5 min once, washed with PBT for 5 min once, incubated in fresh click reaction solution (1x TRIS buffer saline, 4 mM CuSO4 in 1x TRIS buffer saline, 2 μl Alexa-flour-594 Azide (Life technologies), 1 mM sodium ascorbate in 1x TRIS buffer saline) for 30 min on a rocker in the dark, washed with 1x PSB for 5 min thrice, incubated in DAPI (1:1000 dilution) for 30 mins, washed with 1x PSB for 5 min thrice, checked for fluorescence under a stereomicroscope and stored at 4°C in the dark till lightsheet imaging.

Cell death assay using LysoTracker

Request a detailed protocol

Post-hatch larvae were reared in 6-well plates in 3 ml of either of the solutions: 20% Holtfreter’s solution, 1.5% DMSO, 45 μM SU5402, 0.02% ethanol or 1 μg /ml cyclopamine. 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 hr at 20–21°C. Larvae were rinsed gently in Hanks BSS four times, 5 min each on a rocker, fixed overnight in 4% PFA, rinsed once in Hanks BSS for 10 min and dehydrated through a methanol series in Hanks BSS (50%, 75%, 80%, 100%, 5 min each step) to eliminate background staining and stored in 100% methanol at −20°C until imaging.

Microscopy and image analysis

Request a detailed protocol

Whole mount images for limb staging/size measurements, Alcian blue-Alizarin red staining, in situ hybridization, DiI experiments and apoptosis assays were taken on an SZX10 light microscope (Olympus, Tokyo, Japan) using a DP73 CCD camera (Olympus). Figure 1—figure supplement 1 depicts all different forelimb stages to scale. Forelimb images are presented unscaled for in situ hybridizations except where treatment/control limbs are presented. Images for H and E staining and cryo-sections (from Keratin5 and Keratin17 wholemount hybridizations) were taken on a BX53 microscope (Olympus, Tokyo, Japan) with DP80 CCD camera. Both the microscopes were equipped with CellSense software (CellSense V 1.12, Olympus corporation).

EdU stained stage 46 larvae were imaged using a Zeiss Lightsheet Z.1 (College of Arts and Science Imaging Centre, University of Kentucky). Larvae were embedded in 1% low melting agarose (Sigma) dissolved in 1x PBS, mounted in a glass capillary (2.15 mm inner diameter) and the glass capillary was placed in a chamber (specific for the 20x objective lens) filled with 1x PBS. Imaging was done with Zen software (Zeiss) and samples were excited using 561 nm and 488 nm lasers. 20x objective was used and images of all the limbs were taken at 0.75 zoom. Either right, left or both lightsheets were used based on the orientation of the limbs. Image processing and analysis was done with Arivis vision4D software (Arivis). The czi extension files from Zen software was imported into the Arivis vision4D software. An object mask was hand drawn at each z-planes based on the DAPI signal to outline the limb and create a limb mask for total limb volume calculations. The following pipeline was made to calculate red cell aggregate volume and total limb volume: Input ROI (current image set, all channels, scaling 50%), object mask (add the hand drawn object), denoising filter (mean, radius = 10), Intensity filter (radius = 9, K = 0, offset = −10, binary - false), denoising filter (median, radius = 10) and result storage (intensity threshold, range specified, allow holes while quantification). Volume values in μm3 and voxel counts were given as outputs.

Statistical analyses

Request a detailed protocol

All statistical analyses were made using JMP (version Pro 12.10, SAS Institute Inc). Box plots for forelimb staging were made using graph builder in JMP. The vertical line within the box represents the median number of larvae found at a specific stage and days post fertilization. The ends of the box represent the 25th and 75th quartiles and the whiskers on either side represent the interquartile range. For qRT-PCR data the 2-ΔΔCt method was used to calculate fold changes of genes between control (DMSO or ethanol) and treatment (SU5402 or cyclopamine) groups. Calculations for mean Ct values, ΔCt values for experimental and control groups, ΔΔCt values between experimental and control groups, 2-Ct and 2-ΔΔCt fold change values were made using Microsoft Excel and detailed in Supplementary file 2. Student’s t-test was used to calculate the significant changes in relative gene expressions between control and treatment groups. For limb size and snout to tail-tip measurements, student’s t-test was used to calculate the significant changes between control and treatment groups. Differences were considered significant if p<0.05. To analyze whether the decrease in limb size post drug treatment was due to the decrease in overall body length (snout to tail-tip length) we used a fit model ANOVA with treatment, overall body length and treatment*overall body length interaction.

To evaluate co-expression of Fgf ligands and receptors along the limb proximal-distal axis from single-cell RNAseq data, we utilized the fact that contribution to principle component three followed a normal distribution according to a Shapiro-Wilks normality test (W = 0.95551, p-value<2.2e-16). A Welch Two Sample t-test test was conducted to test the hypothesis that two populations (Fgf ligands and receptors) have the same mean contribution to principle component 3.

For lightsheet data, red cell aggregate volume/limb volume (%) was calculated in Microsoft Excel. Comparisons within control and treatment groups were determined by one-way ANOVA (Kruskal-Wallis tests).

Specific statistical information

Request a detailed protocol
Experiment/figure numberStatistics
Figure 5B: Limb size for stage 46 DMSO or SU5402 treatmentsUnpaired one-tailed Student’s t-test
• n = 5 (DMSO), n = 6 (SU5402)
• Mean (DMSO) = 0.1614, Mean (SU5402) = 0.072
• SD (DMSO) = 0.015582, SD (SU5402) = 0.0177539
• SEM (DMSO) = 0.006969, SEM (SU5402) = 0.007248
• p-value<0.0001
T = −8.775
• 95% confidence interval = −0.06636 to −0.11244
Figure 5E: qRT-PCR data for DMSO or SU5402 treated stage 46 limbsUnpaired two-tailed Student’s t-test
• n = 3 (DMSO), n = 3 (SU5402)
• Mean (DMSO) = 1.09 (Etv1), 1 (Etv4), 0.97 (Shh), 0.85 (Ptch1), 1.1 (Gremlin1)
• Mean (SU5402) = 0.28 (Etv1), 0.4 (Etv4), 0.37 (Shh), 1.25 (Ptch1), 0.47 (Gremlin1)
• SD (DMSO) = 0.055 (Etv1), 0.1 (Etv4), 0.06 (Shh), 0.18 (Ptch1), 0.07 (Gremlin1)
• SD (SU5402) = 0.05 (Etv1),. 04 (Etv4), 0.12 (Shh), 0.22 (Ptch1), 0.24 (Gremlin1)
• SEM (DMSO) = 0.03 (Etv1),. 06 (Etv4), 0.03 (Shh), 0.11 (Ptch1), 0.04 (Gremlin1)
• SEM (SU5402) = 0.03 (Etv1),. 02 (Etv4), 0.08 (Shh), 0.13 (Ptch1), 0.14 (Gremlin1)
• p-value=<0.0001 (Etv1), 0.0006 (Etv4), 0.0022 (Shh), 0.07 (Ptch1), 0.014 (Gremlin1)
T = −19.13 (Etv1), −9.79 (Etv4), −6.95 (Shh), 2.44 (Ptch1), −4.14 (Gremlin1)
• 95% confidence interval = −0.7 to −0.91 (Etv1),−0.43 to −0.77 (Etv4), −0.36 to −0.83 (Shh), 0.9 to −0.05 (Ptch1), −0.2 to −1 (Gremlin1)
Figure 5I: Cell proliferation in DMSO or SU5402 treated stage 46 limbsOne-way ANOVA (Kruskal-Wallis tests)
• n = 3 each for DMSO and SU5402
• Median: DMSO = 29.15, SU5402 = 4.75
• p-value=0.0495
Figure 5—figure supplement 1A and C: Limb size for stage 45 DMSO or SU5402 and ethanol or cyclopamine treatmentsUnpaired one-tailed Student’s t-test
• n = 15 (DMSO), n = 15 (SU5402)
• Mean (DMSO) = 0.082, Mean (SU5402) = 0.074
• SD (DMSO) = 0.016, SD (SU5402) = 0.012
• SEM (DMSO) = 0.004, SEM (SU5402) = 0.003
• p-value=0.0514
T = −1.68637
• 95% confidence interval = −0.019 to 0.002
• n = 6 (ethanol), n = 6 (cyclopamine)
• Mean (ethanol) = 0.071, Mean (cyclopamine) = 0.063
• SD (ethanol) = 0.012, SD (cyclopamine) = 0.006
• SEM (ethanol) = 0.002, SEM (cyclopamine) = 0.005
• p-value=0.909
T = 1.43
• 95% confidence interval = −0.004 to 0.02
Figure 5—figure supplement 1B and D: Snout to tail-tip length for DMSO or SU5402 and ethanol or cyclopamine treatmentsUnpaired one-tailed Student’s t-test
• n = 5 (DMSO), n = 6 (SU5402)
• Mean (DMSO) = 10.92, Mean (SU5402) = 10.29
• SD (DMSO) = 0.2, SD (SU5402) = 0.25
• SEM (DMSO) = 0.088, SEM (SU5402) = 0.104
• p-value<0.0007
T = −4.5
• 95% confidence interval = −0.316 to −0.950
• n = 6 (ethanol), n = 6 (cyclopamine)
• Mean (ethanol) = 10.5, Mean (cyclopamine) = 9.98
• SD (ethanol) = 0.15, SD (cyclopamine) = 0.3
• SEM (ethanol) = 0.06, SEM (cyclopamine) = 0.121
• p-value<0.0016
T = 3.87
• 95% confidence interval = 0.82 to 0.222
Figure 6B: Limb size for stage 46 ethanol or cyclopamine treatmentsUnpaired one-tailed Student’s t-test
• n = 6 (ethanol), n = 6 (cyclopamine)
• Mean (ethanol) = 0.16, Mean (cyclopamine) = 0.095
• SD (ethanol) = 0.008, SD (cyclopamine) = 0.016
• SEM (ethanol) = 0.003, SEM (cyclopamine) = 0.006
• p-value<0.0001
T = 8.36
• 95% confidence interval = 0.08 to 0.045
Figure 6C: qRT-PCR data for ethanol or cyclopamine treated stage 46 limbsUnpaired one-tailed Student’s t-test
• n = 3 (ethanol), n = 3 (cyclopamine)
• Mean (ethanol) = 1.12 (Ptch1), 0.94 (Gremlin1), 1.12 (Fgf8)
• Mean (cyclopamine) = 0.7 (Ptch1), 0.39 (Gremlin1), 0.22 (Fgf8)
• SD (ethanol) = 0.145 (Ptch1), 0.12 (Gremlin1), 0.14 (Fgf8)
• SD (cyclopamine) = 0.00.04 (Ptch1), 0.1 (Gremlin1), 0.03 (Fgf8)
• SEM (ethanol) = 0.08 (Ptch1), 0.07 (Gremlin1), 0.08 (Fgf8)
• SEM (cyclopamine) = 0.02 (Ptch1), 0.06 (Gremlin1), 0.02 (Fgf8)
• p-value=0.0039 (Ptch1), 0.002 (Gremlin1), 0.0002 (Fgf8)
T = 4.9 (Ptch1), 5.97 (Gremlin1), 11.21 (Fgf8)
• 95% confidence interval = 0.66 to 0.19 (Ptch1), 0.8 to 0.3 (Gremlin1), 1.1 to 0.68 (Fgf8)
Figure 6G: Cell proliferation in ethanol or cyclopmaine treated stage 46 limbsOne-way ANOVA (Kruskal-Wallis tests)
• n = 3 each for ethanol and cyclopamine
• Median: ethanol = 30.06, cyclopamine = 13.9,
• p-value=0.0495

References

  1. 1
  2. 2
    Size dependence during the development of the amphibian foot. Colchicine-induced digital loss and reduction
    1. P Alberch
    2. EA Gale
    (1983)
    Development 76:177–197.
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
    The mouse Fgf8 gene encodes a family of polypeptides and is expressed in regions that direct outgrowth and patterning in the developing embryo
    1. PH Crossley
    2. GR Martin
    (1995)
    Development 121:439–451.
  14. 14
  15. 15
    Reduction of the rate of outgrowth, cell density, and cell division following removal of the apical ectodermal ridge of the chick limb-bud
    1. Dennis Summerbell
    (1977)
    Journal of Embryology and Experimental Morphology 40:1–21.
  16. 16
  17. 17
    The divergent development of the apical ectodermal ridge in the marsupial Monodelphis domestica
    1. CK Doroba
    2. KE Sears
    (2010)
    The Anatomical Record: Advances in Integrative Anatomy and Evolutionary Biology 293:1325–1332.
    https://doi.org/10.1002/ar.21183
  18. 18
    Activation of Fgf-4 and HoxD gene expression by BMP-2 expressing cells in the developing chick limb
    1. DM Duprez
    2. K Kostakopoulou
    3. PH Francis-West
    4. C Tickle
    5. PM Brickell
    (1996)
    Development 122:1821–1828.
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
    Xenopus limb bud morphogenesis
    1. SR Keenan
    2. CW Beck
    (2016)
    Developmental Dynamics : An Official Publication of the American Association of Anatomists 245:233–243.
    https://doi.org/10.1002/dvdy.24351
  36. 36
    Early limb patterning in the direct-developing salamander Plethodon cinereus revealed by sox9 and col2a1
    1. RR Kerney
    2. J Hanken
    3. DC Blackburn
    (2018)
    Evolution & Development, 20, 10.1111/ede.12250, 29527799.
  37. 37
  38. 38
    Morphogenetic role of epidermal and mesodermal components of the fore-and hindlimb buds of the newt Pleurodeles waltlii michah. (Urodela, Amphibia)
    1. M Lauthier
    (1985)
    Archives of Biological Sciences 96:23–43.
  39. 39
  40. 40
    FGF-8 isoforms activate receptor splice forms that are expressed in mesenchymal regions of mouse development
    1. CA MacArthur
    2. A Lawshé
    3. J Xu
    4. S Santos-Ocampo
    5. M Heikinheimo
    6. AT Chellaiah
    7. DM Ornitz
    (1995)
    Development 121:3603–3613.
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
    The BMP antagonist gremlin regulates outgrowth, chondrogenesis and programmed cell death in the developing limb
    1. R Merino
    2. J Rodriguez-Leon
    3. D Macias
    4. Y Gañan
    5. AN Economides
    6. JM Hurle
    (1999)
    Development 126:5515–5522.
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
    Interplay between the molecular signals that control vertebrate limb development
    1. L Niswander
    (2002)
    The International Journal of Developmental Biology 46:877–881.
  56. 56
    Fgf-4 expression during Gastrulation, myogenesis, limb and tooth development in the mouse
    1. L Niswander
    2. GR Martin
    (1992)
    Development 114:755–768.
  57. 57
  58. 58
  59. 59
    The mesenchymal factor, FGF10, initiates and maintains the outgrowth of the chick limb bud through interaction with FGF8, an apical ectodermal factor
    1. H Ohuchi
    2. T Nakagawa
    3. A Yamamoto
    4. A Araga
    5. T Ohata
    6. Y Ishimaru
    7. H Yoshioka
    8. T Kuwana
    9. T Nohno
    10. M Yamasaki
    11. N Itoh
    12. S Noji
    (1997)
    Development 124:2235–2244.
  60. 60
    The fibroblast growth factor signaling pathway
    1. DM Ornitz
    2. N Itoh
    (2015)
    Wiley Interdisciplinary Reviews: Developmental Biology 4:215–266.
    https://doi.org/10.1002/wdev.176
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
    Evolutionary Biology
    1. NH Shubin
    2. P Alberch
    (1986)
    A morphogenetic approach to the origin and basic organization of the tetrapod limb, Evolutionary Biology, Boston, Springer, 10.1007/978-1-4615-6983-1_6.
  71. 71
    Evidence for a metameric pattern in the development of the chick humerus
    1. TD Stephens
    2. TR McNulty
    (1981)
    Development 61:191–205.
  72. 72
  73. 73
    Control of pattern formation in Urodele limb ontogeny: a review and a hypothesis
    1. DL Stocum
    2. JF Fallon
    (1982)
    Journal of Embryology and Experimental Morphology  69:7–36.
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
    A quantitative analysis of the effect of excision of the AER from the chick limb-bud
    1. D Summerbell
    (1974)
    Journal of Embryology and Experimental Morphology 32:651–660.
  80. 80
  81. 81
  82. 82
    Determination of Limb-Axes
    1. FH Swett
    (1937)
    The Quarterly Review of Biology 12:322–339.
    https://doi.org/10.1086/394535
  83. 83
  84. 84
  85. 85
    Early limb development of xenopus laevis
    1. D Tarin
    2. AP Sturdee
    (1971)
    Journal of Embryology and Experimental Morphology 26:169–197.
  86. 86
  87. 87
  88. 88
  89. 89
    Insights into bird wing evolution and digit specification from polarizing region fate maps
    1. M Towers
    2. J Signolet
    3. A Sherman
    4. H Sang
    5. C Tickle
    (2011)
    Nature Communications, 2, 10.1038/ncomms1437, 21829188.
  90. 90
  91. 91
  92. 92
  93. 93
    FGF-4 maintains polarizing activity of posterior limb bud cells in vivo and in vitro
    1. A Vogel
    2. C Tickle
    (1993)
    Development 119:199–206.
  94. 94
  95. 95
  96. 96
  97. 97
  98. 98
  99. 99
  100. 100
  101. 101
  102. 102

Decision letter

  1. Marianne E Bronner
    Senior and Reviewing Editor; California Institute of Technology, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your work entitled "Shh mediated FGF-signaling is compartmentalized in the mesenchyme and controls limb size during salamander development" for consideration by eLife. Your article has been reviewed by a Senior Editor, a Reviewing Editor, and two reviewers. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife in its current form.

As you can see below, both reviewers agree that the work is of potentially great importance. In our discussions, we all agreed that your work is an ambitious effort to examine in axolotls at least 4 aspects of limb development (AER signaling, ZPA signaling, dorsoventral patterning, and cartilage differentiation) that have each been studied for decades in chick and mouse. The data presented indicate a number of major differences in the mode and mechanisms of limb formation in salamander relative to chick, mouse, and other research organisms, challenging a number of current paradigms of limb development. Our problem with these findings is that the conclusions are overly bold given the relative superficiality of the analyses. For example, our biggest concern is that only Fgf8 was looked at when it has been shown by many that multiple Fgfs are involved in AER signaling. Because the argument about loss of ectodermal FGF signaling is central to the paper, it strikes us as essential to look at expression of the Fgfs that are involved in AER signaling (Fgf4/8/9/17) at minimum, and probably to look at distribution of the 4 FGF receptors., so they need to be thorough in their analysis.

We appreciate that addressing the points above is a significant amount of work. If you choose to address these and the other comments raised below and submit a more in-depth manuscript later on, we would be happy to re-examine the work. If, on the other hand, you would prefer to have the work published elsewhere as is we are enclosing the full text of the reviews, which we hope you will find useful. As you will see, both reviewers concurred on many aspects. To facilitate your evaluation of the enclosed reviews, we have identified those comments which are essentially the same. I will use R1 and R2 to identify the reviewers, followed by point numbers of the respective reviews.

• R1-1 = R2-9;

• R1-2 = R2-5 and R2-6;

• R1-4 = R2-4 and part of R2-3;

• Questions about the SU5402 treatments are discussed in R15, R2-3 and R2-14;

• Different but complementary questions about Shh expression are raised in R1-6 and R2-1;

• R1-7 = R2-2

Reviewer #1:

This study makes the bold claim that Urodeles, as exemplified by the axolotl Ambystoma mexicanum, exhibit distinct developmental features compared to other tetrapod species. An intriguing implication of this claim is that a difference in developmental program may be related to the ability of axolotls to regenerate limbs. The study covers a wide variety of features: (1) The morphology of the limb bud, including whether or not an AER forms, (2) the order of deposition of an Alcian blue positive matrix and the expression of Sox 9, Ihh; (3) order of digit formation, (4) Proximal distal patterning (via Fgf8 expression, SU5402 treatment) (5) A/P specification (via Shh expression, cyclopamine treatment) (6) Relationship between the P/D and A/P regulatory loops (via Grem. Ptch1 regulation), and (7) D/V patterning (via Wnt, En1 expression etc).

Unfortunately, the study handles each of these issues very superficially considering that there are at least 100 major studies examining each of these topics using chicken or mouse model systems over the course of the last 25 years. As a consequence, this reviewer believes that the bold conclusions drawn based on the data collected so far, while intriguing, are much too premature.

Essential revisions:

1) Just because there is no distal pseudo-stratified epithelial thickening (AER) doesn't mean an ectodermal signaling center is not there. There are other examples of amphibians that have no AER and still have AER-FGF expression.

2) Typically condensation is defined by histological coalescing of cells, however this study does not use histology to determine the order of digit condensation. If condensation is defined by proxies for cartilage differentiation the authors need to explain the variation in Alcian blue intensity for digits (order = 2?1? or vice versa, not clear from images 3, 4), sox 9 (2,1, 3, 4), and Ihh (1, 2? 3, 4). The authors should compare their more careful findings with those of Sue Mackem and colleagues (PMID: 18410737 and 21509901). Axolotls have 4 digits, is it digit 1 that is missing or digit 5?

3) The authors should be aware that colormetric RNA in situ hybridization is not as sensitive as other RNA detection methods (FISH or RNAscope). Thus, lack of color deposition, especially using whole mount preps, does not necessarily mean that no expression is present. For example, it may be the case that there is an early distal swath of Sox9 expression but that it present at too low levels to detect give the method used.

4) There are at least 22 Fgfs and 4 FGF receptors. In many comparisons between species, related growth factors have been shown to substitute for each other. For example, Sox10 appears to provide the function of mouse Sox9 in zebrafish. The expression pattern for Fgf8 shown in this study is reminiscent of the pattern of Fgf10 expression in chicken/mouse embryos. Without examining at least a handful of the most well-known FGF genes (Fgf4, Fgf10, Fgf9, Fgf2, Fgfr1, Fgfr2), it is premature to claim that axolotl limbs develop without needing an ectodermal source of FGFs.

5) SU5402 treatments typically result in a mild phenotype but based on their table of results CAN result in more severe digit loss phenotypes. This suggests that either the dose or timing of administration can be modulated to achieve phenotypes that match mouse KO phenotypes.

(Perhaps best compared to Fgfr1/2 combo KOs). In addition, given the long period between establishment of the limb field and limb outgrowth, it may be that one would need to administer SU5402 much earlier (prior to limb bud outgrowth) to inhibit proximal specification in these species.

6) The observation that Shh is also expressed in an anterior domain is interesting and should be examined. Given that cyclopamine administration gives the expected results, is this anterior domain not significant? Does the location of Shh expression correlate with areas of cell death at the anterior and posterior autopod margins?

7) The authors make the major claim that the FGF and Hh regulatory loops are not interlinked. This is based on examining a few read-outs by whole mount RNA in situ hybridization and is not comprehensively or quantitatively assessed. Furthermore, analysis of cell death is not presented, nor are genes or proliferation rates examined PRIOR to differences in limb bud size being evident.

8) Finally the analysis of D/V patterning genes seems tacked on and could be eliminated from the paper.

In its current state this reviewer recommends rejection of the paper as a thorough examination of the above issues would certainly take longer than the 2-3 months typically allotted for eLife submissions. Should the bold claims hold after more detailed analysis, (suggest focusing on FGF and HH pathways as most interesting and critical, while describing the order of condensation/differentiation is likely less revelatory), the study could be of great interest and significance.

Reviewer #2:

Summary:

In this fascinating paper, Purushothaman and Seifert examine limb development in the axolotl, a salamander that has long been the primary model for studies limb regeneration but has been largely overlooked for developmental studies. The lack of data on how axolotl limbs develop, compared to the developmental workhorses of chick and mouse, is a major problem because the data on axolotl limb regeneration are usually compared/contrasted with data on chick and mouse limb development, with an underlying assumption that limb development is similar across these species. Thus, the importance of the data presented here, which identifies a number of major differences in the mode and mechanisms of limb formation in salamander relative to chick, mouse, and other model and non-model organisms, is difficult to overstate. Their results not only highlight the manner in which developmental mechanisms have evolved in salamanders, but they also challenge some current paradigms of limb development that are based on work in chick and mouse, and they provide a developmental baseline for comparison with regeneration studies in the same species.

The authors begin by examining skeletal development and they validate some earlier work that showed differences in the spatiotemporal pattern of digit formation. They confirm the earlier observation of Sturdee and Connock, (1975) that the axolotl limb bud lacks an AER, and they go on to show that, unlike several other tetrapods that have been reported to lack an AER, axolotls show no ectodermal expression of Fgf8. This contrasts with the results of Han et al., (2001), who reported that Fgf8 is expressed in both the ectoderm and mesenchyme. Purushothaman and Seifert go to great lengths to detect Fgf8 in both tissues and they show, unequivocally in my opinion, that the Fgf8 expression domain is entirely mesenchymal.

Based on current models of limb development from mouse and chick, multiple FGFs (from the AER) positively regulate Shh, which then signals to the AER (directly) to maintain FGF expression, and to the limb mesenchyme, where it induces Gremlin1 expression. Mesenchymal Gremlin1 antagonizes Bmps in order to maintain FGF expression in the AER. The authors report that in axolotl limb buds, Shh expression was unaffected by suppression of FGF signaling using an FGF receptor antagonist, SU5402, which is a surprising finding because chick/mouse limb experiments would predict that near-complete suppression of FGF signaling would lead to reduced Shh expression. This result is made even more interesting by their discovery that Grem1 expression is virtually eliminated in SU5402-treated limbs. When Hedgehog signaling is blocked by cyclopamine treatment, Gremlin1 and Fgf8 expression is reduced, and the limb buds are smaller. The authors conclude that, in axolotl limb buds, FGF signaling is not required for Shh expression but is required for Gremlin expression and limb size, and that SHH controls distal limb patterning and growth by regulating Grem1 and Fgf8.

Finally, they turn to dorsoventral patterning and show that Wnt7a/Radical fringe and Engrailed1, which mark, respectively, the dorsal and ventral ectoderm in chick and mouse limbs, are expressed throughout the limb mesenchyme. This is another major difference between axolotl and the major model systems, and opens many new questions about the roles of those genes in dorsovental patterning of limb structures. Although not discussed, loss of dorsoventrally polarized expression also fits with loss of the AER. Polarized expression of these genes in limb ectoderm establishes the AER at the tip of the limb bud in chick and mouse. Furthermore, chick "Wingless" mutants form and then lose the AER, after which the dorsoventral markers lose their restricted expression (Ohuchi et al., 1997). It would be worth discussing whether the authors think that the dorsoventral change is a cause or consequence of the loss of an AER.

They conclude that the mesenchymal-epithelial (Shh-Grem-Fgf8) feedback loop that controls limb development in other tetrapods has been re-arranged in axolotls, where it is entirely mesenchymal and has evolved different regulatory interactions. In a rich but short discussion, the authors explore the implications of these findings for understanding genetic interactions in the limb, evolution of developmental mechanisms, and molecular control of regeneration. The brevity of the discussion does a disservice to the many sacred cows that this paper has toppled, and I would encourage them to break apart the results and discussion so that a proper airing of these ideas can be presented. Overall, the data are clear and convincing and the discoveries are likely to be of interest to a broad group of biologists in the areas of development, regeneration and evolution.

Essential revisions:

1) The Shh expression pattern is quite interesting and makes one whether the cells that form the digits ever see the Shh signal. Although the initial domain in the posterior part of the limb is typical of early limb buds of tetrapods and fishes, the subsequent domains are very different from the patterns reported for other tetrapods. They state, "Posterior expression is never detected in the autopod" (Figure 2 legend). If the digits progenitors arise from (or near) the Shh domain, then the role of Shh in digit development could be conserved. If the cells of the Shh domain are distant from the digit progenitors, then this would challenge the current paradigm of digit development. This could be tested quite easily by labeling the initial Shh domain (at stage 45) with an injection (or carefully applied crystals) of an indocarbocyanine dye, like DiI, and then examining the distribution of labeled cells several stages later in order to determine whether those cells contribute to the digits. Determining whether digit progenitors reside within the Shh domain would allow the authors to make a direct comparison with the Shh-expressing (ZPA) cells of mouse and chick limb buds, which contribute to posterior digits.

2) Because the SU5402 treatments show that FGF signaling does not appear to maintain Shh in axolotls, is there really is a FGF-Shh feedback loop in axolotl limb bud mesenchyme?

3) Do the authors think that the same mesenchyme cells that are expressing Fgf8 are also responding to Fgf8? Do they express an appropriate receptor? How does the topological shift in the position of the ligand, Fgf8, relate to the ability of SU5402, a FGF receptor antagonist, to produce such different outcomes in axolotl vs chick and mouse limbs? Do the authors think that there is combinatorial FGF signaling in axolotls (as there is in chick and mouse), but the ligands are produced and received by the mesenchyme? If so, then can they suggest an explanation for how the same Fgfs could regulate Shh when secreted by the AER(as in chick/mouse) but do not regulate Shh when produced by the mesenchyme? Chick experiments in which the AER is removed and FGF beads are implanted into the limb bud mesenchyme suggests that the source of the ligand is irrelevant. Have they considered the possibility that Fgf8 acquired a novel function during axolotl limb evolution and lost its roles as a Shh regulator and an outgrowth signal? The proliferation data suggest that at least some of these roles are conserved.

4) In several places, the Fgf8 result is generalized to the other Fgfs involved in AER signaling (e.g., Results and Discussion section: "Taken together, our expression data clearly shows that the Shh-Grem-FGF feedback loop is entirely expressed within the mesenchymal compartment during salamander limb development."). In chick and mouse limbs, at least 4 Fgfs (Fgf4, Fgf8, Fgf9, and Fgf17) are expressed in the AER and signal through mesenchymal FGF receptors. Whilst it would be informative to examine the expression patterns of these other Fgfs in axolotl limb buds to determine whether the Shh-FGF feedback loop is entirely in the mesenchyme (or if some of the Fgfs have remained in the epithelium) if they choose not to examine the expression of these other Fgfs, then they should tailor the discussion accordingly and at least mention the alternate possibilities. For example, they show that Fgf8 is expressed in the mesenchyme, but it is unknown whether or where other Fgfs are expressed in the limb. Thus, several scenarios are possible, including (a) Fgf8 is mesenchymal but Fgf4, Fgf9, and Fgf17 are expressed in the epithelium, (b) All 4 Fgfs are mesenchymal, (c) Fgf8 is mesenchymal and the other 3 Fgfs are not expressed. Without knowing which of these scenarios is true, one must be cautious and clear with the interpretation the SU5402 results since SU5402 treatment probably suppresses all FGF signaling. For example, scenarios (b) and (c) would support "a topological rearrangement of the Shh-Grem1-FGF loop", but scenario (a) would not because 3 of the 4 Fgfs would be expressed in ectoderm. To be clear, I am not insisting that the authors examine every FGF and FGF receptor in axolotl limbs, but they need to consider the complexity of FGF signaling in amniotes in their interpretation of the results. They could also cite other experimental work that supports their argument that the critical FGF signal is mesenchymal, such as the work of Lauthier (1985), who showed that newt limbs can develop even when the apical epithelium is removed from the mesenchyme.

5) Figure 1C. Looking at the alcian blue stained limbs, there are no visible digit condensations at stage 47, and at stage 48, three distinct digit condensations are stained. At stage 49, these 3 condensations remain visible (only I and II are labeled in the figure, but digit III is also stained). The initial condensation of digit IV is also visible at stage 49. Given that no digits are visible at stage 47, and three digits (I, II, III) appear at stage 48, what is the basis for the argument that digits form from anterior to posterior? (see the following point for continued discussion of this issue)

6) Figure 1E. I found the Sox9 in situ hybridizations to be less informative than the alcian blue preparations, possibly due to the level of background staining, although this could also be a result of slight staging differences between the alcian blue and Sox9 limbs. For example, although one can see three digits in the alcian blue limb at stage 48, only digit II is obvious in the Sox9 in situ hybridization at the same stage. Is the latter a slightly younger limb, or is the in situ just not as clear as the alcian blue preparation?

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "FGF-signaling is compartmentalized within the mesenchyme and controls proliferation during salamander limb development" for further consideration at eLife. Your revised article has been favorably evaluated by Marianne Bronner (Senior Editor), a Reviewing Editor, and two reviewers.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

Essential revisions:

1) Please break the Discussion section into subsections, including one that compares the findings in axolotl with chick and mouse.

2) Please add a table that compares limb development in axolotl with mouse/chick.

3) Please try, if possible, within a reasonable time frame, to address the reviewer comments below, though some were deemed non-essential (e.g. scRNAseq of limb ectoderm; re-imaging the immunofluorescence). If you cannot address certain comments, please explain why.

Reviewer #1:

In this revised submission, Purushothaman et al., compare limb development in the axolotl salamander to classical findings in chicken and mouse embryos. Some findings confirmed previous studies using more refined techniques, other findings were new. Most striking were the numerous differences seen in axolotl including pre-axial condensation, condensation first along the metapterygial axis, lack of a morphological AER, lack of AER-FGF expression (instead found at the distal mesenchyme), no requirement for FGF signaling for Prox./Distal patterning or cell death, anterior expression of Shh, requirement of Shh for distal proliferation but not cell death. These observations are carefully presented and clearly discussed and very interesting, however, Although the authors responded to many of the criticisms in the first round with additional experiments, I am conflicted as to whether this is a collection of interesting descriptive phenomena vs. a comprehensive mechanistic conceptual advance in the field.

Essential revisions:

1) It would be nice to see some RNA-ISH that is positive in the overlying ectoderm just to rule out the possibility that lack of staining is not due simply to a technical problem with doing RNA-ISH on axolotl ectoderm (some permeability problem). Another method would be to do RNA-ISH on a section.

2) Please show Fgf4 RNA-ISH and other 'data not shown' such as cell death assays even if not very high.

3) How come only scSeq in mesenchyme is shown? Could an analysis of the ectoderm be included? In no FGF ligands detected in this data set, this would be nice to report.

4) Lysotracker and TUNEL are coincident in mouse/chick. Are they in axolotol? Could cells be dying but this is not visible due to failure of phagocytic bodies?

5) The second paragraph of the Discussion section does a better job at summarizing the paper than Abstract, perhaps the abstract could be enhanced with some of the ideas here.

6) Even for the limb bud aficionado it would be useful to have a chart comparing mouse/chick results to findings in axolotl.

7) In Figure 5H and Figure 6—figure supplement 1 panel F, I can't tell which of the red patches are positive above background. Also, in the mid-sagittal slice the control limb bud shows almost all cells as positive, is this correct?

8) Why are Christensen et al., 2002; Han et al., 2001 incorrect in their observation of FGF expression in the AER?

Reviewer #2:

The authors are to be congratulated for the quality and scope of the revised manuscript. It is a rigorous and extremely thorough study. All of my concerns have been addressed satisfactorily.

The DiI results are quite interesting. They injected DiI into the early Shh domain of the limb bud and then let it grow, examining the localization of labeled cells at different stages. This allowed them to map the spatiotemporal movements of the initial Shh domain and to determine whether the putative ZPA cells end up in the digits (and if so, which digits). This was very informative. Relative to chick and mouse, axolotl limbs have somewhat crazy fingers and gene expression patterns, but this experiment identifies a deeply conserved feature – the contribution of Shh-expressing cells to the posterior digit(s).

Expansion of the FGF analysis to include all of the Fgfs expressed in the AER of amniotes (FGF 4, 8, 9, 17) as well as Fgf10 and Fgfr1 and Fgfr2, provides definitive evidence that the "AER Fgfs" of amniotes are expressed in the mesoderm of axolotl limb buds (with the possible exception of Fgf4, which was variable but never in the ectoderm). This is an unequivocal demonstration that axolotl limb buds lack the ectodermal domains of Fgfs that are found in the AER of amniotes.

The authors note that the proximal-distal segregation of FGF ligand and receptor expression is puzzling, but the results are clear. Moreover, the in situ hybridization patterns are validated by the scRNAseq data. It is also interesting to see that where Fgfr1 does extend into the distal domain, it is either in or near the cells expressing FGF ligands.

I am impressed by the manner in which the authors responded to the reviewers' criticism that some conclusions were overly broad in the previous version of the paper. Rather than just removing conclusions that were not entirely supported by the results, they investigated whether those conclusions would stand up to further testing. In some places, they were just a half step away from having data that would address the question, but in other cases, they went to great lengths to carry out the appropriate experiments. Where the results provided clear support for the hypothesis (e.g., the AER Fgfs are expressed in the mesenchyme of axolotls), they say so, and where more than one interpretation is possible, they provide conservative conclusions and acknowledge the open questions that remain. I very much like this version of the paper and think that it will be a valuable contribution of the fields of limb development and limb evo-devo, and it will provide important developmental context for regenerative studies.

https://doi.org/10.7554/eLife.48507.026

Author response

[…] We appreciate that addressing the points above is a significant amount of work. If you choose to address these and the other comments raised below and submit a more in-depth manuscript later on, we would be happy to re-examine the work. If, on the other hand, you would prefer to have the work published elsewhere as is we are enclosing the full text of the reviews, which we hope you will find useful. As you will see, both reviewers concurred on many aspects.

We appreciate the thorough reviews prepared by both reviewers and the summary provided by the reviewing editor. We have worked hard to address every point raised by the reviewers and address each of these points individual. Our revised manuscript contains a much more thorough analysis of the requested FGF ligands and receptors using WISH and scRNA-seq. We have included a more in-depth analysis of our inhibition experiments and now provide data on cell death and cell proliferation. We also use qRT-PCR to quantify gene targets of FGF- and Shh-signaling and provide an additional early time point. We have also reanalyzed our analysis of digit specification and differentiation. Lastly, we have reorganized our paper into a traditional format at the suggestion of R2 so as to provide a separate Discussion section at the end of the paper.

To facilitate your evaluation of the enclosed reviews, we have identified those comments which are essentially the same. I will use R1 and R2 to identify the reviewers, followed by point numbers of the respective reviews.

Using the similar points stated here, we refer to the comment below in response to R1 and then indicate in response to R2 that we have responded to the R1 comment.

R1-1 = R2-9;

Refers to AER structure.

We have corrected our language in some cases and include reference to direct-developing frogs and marsupials. Our data confirms that there is no MORPHOLOGICAL AER. Our new data covering the additional FGF ligands and receptors demonstrates there is not MOLECULAR AER.

R1-2 = R2-5 and R2-6;

Digit patterning and specification.

We agree with the reviewers and have completed a more detailed analysis using Sox9 and Alcian blue staining. We present a deeper analysis as a Figure 1—figure supplement 2.

R1-4 = R2-4 and part of R2-3;

FGF ligand and receptors,

We have addressed these comments using in situ hybridization and scRNAseq analysis. New data is presented in Figure 3 (FGF expression), Figure 4, Figure 4—figure supplement 1 and Figure 4—figure supplement 2 (scRNAseq data).

Questions about the SU5402 treatments are discussed in R15, R2-3 and R2-14;

We now address these in the text and methods. We performed dose response which yielded the concentration we used. We administered during limb field specification prior to limb budding and thus believe our results accurately reflect inhibiting FGF-signaling from the outset of limb development. New data is also presented in Figure 5—figure supplement 1.

Different but complementary questions about Shh expression are raised in R1-6 and R2-1;

Our analysis of Shh expression and signaling now includes an analysis of cell death and cell proliferation and we have added a DiL labeling experiment at the request of R2. We present this data in Figure 2 and Figure 6.

R1-7 = R2-2

To address these comments we performed a cell death analysis during normal development and in response to FGF and Shh inhibition. We do the same for cell proliferation and use qRT-PCRT to quantify changes in gene targets following inhibition. This new data is presented in Figure 5, Figure 5—figure supplement 2 and Figure 6.

Reviewer #1:

This study makes the bold claim that Urodeles, as exemplified by the axolotl Ambystoma mexicanum, exhibit distinct developmental features compared to other tetrapod species. An intriguing implication of this claim is that a difference in developmental program may be related to the ability of axolotls to regenerate limbs. The study covers a wide variety of features: (1) The morphology of the limb bud, including whether or not an AER forms, (2) the order of deposition of an Alcian blue positive matrix and the expression of Sox 9, Ihh; (3) order of digit formation, (4) Proximal distal patterning (via Fgf8 expression, SU5402 treatment) (5) A/P specification (via Shh expression, cyclopamine treatment) (6) Relationship between the P/D and A/P regulatory loops (via Grem. Ptch1 regulation), and (7) D/V patterning (via Wnt, En1 expression etc).

Unfortunately, the study handles each of these issues very superficially considering that there are at least 100 major studies examining each of these topics using chicken or mouse model systems over the course of the last 25 years. As a consequence, this reviewer believes that the bold conclusions drawn based on the data collected so far, while intriguing, are much too premature.

We fully appreciate all the detailed studies that have been done in chicken and mouse examining each of these issues during limb development. In light of this general comment, and the more detailed comments outlined below, we have expanded our analysis of FGF- and Shh-signaling in much greater detail and have reserved the D/V information for an additional paper.

Essential revisions:

1) Just because there is no distal pseudo-stratified epithelial thickening (AER) doesn't mean an ectodermal signaling center is not there. There are other examples of amphibians that have no AER and still have AER-FGF expression.

We agree with the reviewer on this point. In the original submission we refer to the cases of Eleutherodactylus coqui (direct-developing frog) and Monodelphis domestica (opossum), both of which lack a morphological AER but demonstrate ectodermal FGF expression. In response to the comment, we have now been more careful with our language and clearly link FGF-expression where we are discussing the AER and highlight the two examples of the above-mentioned animals not having an AER, but still possessing an ectodermal signaling center (subsection “Amniote and anuran AER-specific FGF ligands (8, 9, 17) are expressed exclusively in axolotl limb mesenchyme” and Discussion section). We also believe data showing expression for FGF ligands and receptors further addresses this point.

2) Typically, condensation is defined by histological coalescing of cells, however this study does not use histology to determine the order of digit condensation. If condensation is defined by proxies for cartilage differentiation the authors need to explain the variation in Alcian blue intensity for digits (order = 2?1? or vice versa, not clear from images 3, 4), sox 9 (2,1, 3, 4), and Ihh (1, 2? 3, 4). The authors should compare their more careful findings with those of Sue Mackem and colleagues (PMID: 18410737 and 21509901). Axolotls have 4 digits, is it digit 1 that is missing or digit 5?

In response to the reviewer, we attempted to further analyze coalescing cells in developing axolotl limb buds using paraffin sections and semi-thin resin embedded sections but were unable to produce sections that could resolve the issues raised by in situ and Alcian blue samples. Instead, we revisited each and every limb stained for Sox9 which we now present to the reader in a new supplementary figure (Figure 1—figure supplement 2). This data shows (pretty clearly in our opinion) that Sox9 expression specifies digit II before digit I which appears shortly afterwards. In fact, Sox9 expression can be seen emerging in the area of digit I in late stage 48 limb buds. The two digits are clearly stained at stage 49 for reference. In contrast, revisiting all of our Alcian blue stained limbs across these and later stages, we were unable to find an instance where digit II was differentiating before digit I. Therefore, we now present this data and draw the conclusion that the digits are specified in a slightly different temporal sequence than when they differentiate. Said another way, digits I and II differentiate at the same time even though digit I is specified before digit II. Rather than compare to the two Susan Mackem papers (which we are familiar with), we discuss how our findings support ontogenetic and heterchronic shifts that have occurred during urodele limb evolution. There is precedent for variation in the pattern of digit differentiation across urodeles, although no one has looked across many species using Sox9. With respect to the final point in this comment regarding digit identity, it is not our intention to get drawn into the controversy of digit identity, a debate which continues and is likely unresolvable without molecular markers for individual digits (see Towers et al., 2011). Thus, we have chosen not to address this in our manuscript.

3) The authors should be aware that colormetric RNA in situ hybridization is not as sensitive as other RNA detection methods (FISH or RNAscope). Thus, lack of color deposition, especially using whole mount preps, does not necessarily mean that no expression is present. For example, it may be the case that there is an early distal swath of Sox9 expression but that it present at too low levels to detect give the method used.

We appreciate that FISH and RNAscope have increased sensitivity compared to WISH. We also note that these methods have increased noise that creates problems when trying to demonstrate the absence of signal. With respect to the Sox9 result, while our new supplementary figure presenting all Sox9-stained limbs for stages 47-49 cannot rule out a faint swath of expression in digit I as it is being expressed in digit II, we would expect that the same faint signal would be present prior to our observation of Sox9 in digit II at stage 47. Thus, we do not believe that implementing an additional (and very costly) method would bring appreciable clarity to this point. We have been careful not to overstate our result which we believe should address this comment.

4) There are at least 22 Fgfs and 4 FGF receptors. In many comparisons between species, related growth factors have been shown to substitute for each other. For example, Sox10 appears to provide the function of mouse Sox9 in zebrafish. The expression pattern for Fgf8 shown in this study is reminiscent of the pattern of Fgf10 expression in chicken/mouse embryos. Without examining at least a handful of the most well-known FGF genes (Fgf4, Fgf10, Fgf9, Fgf2, Fgfr1, Fgfr2), it is premature to claim that axolotl limbs develop without needing an ectodermal source of FGFs.

As noted by the other reviewer we agree with this point. We now include WISH staining and scRNA-seq data for all the requested FGF ligands and receptors. Our data shows quite conclusively that none of the ligands or receptors are expressed in the ectoderm. These data also provide insight into segregation across the proximodistal and anteroposterior axes. We were unable to detect Fgf4 using in situ or in the scRNA-seq dataset.

5) SU5402 treatments typically result in a mild phenotype but based on their table of results CAN result in more severe digit loss phenotypes. This suggests that either the dose or timing of administration can be modulated to achieve phenotypes that match mouse KO phenotypes.

(Perhaps best compared to Fgfr1/2 combo KOs). In addition, given the long period between establishment of the limb field and limb outgrowth, it may be that one would need to administer SU5402 much earlier (prior to limb bud outgrowth) to inhibit proximal specification in these species.

We appreciate the point raised here which is a result reminiscent of so many chick limb studies using pharmacological agents. Although a result can be robust, sometimes unseen effects to animal health or growth can have an additive effect on the ultimate phenotype. We administered SU5402 based on a dose-response study and selected the max dose that was not toxic and did not appreciable effect total growth of the animals. We administered SU5402 at stage 39 which is prior to limb bud outgrowth when the limb bud is not visible in an effort to inhibit FGF-signaling as soon as it began. Lastly, even in the 4/34 cases where additional elements were lost (“more severe phenotypes as referred to in the comment”), the missing elements were lost in the same posterior to anterior direction, rather than the proximal truncation predicted from AER loss of function studies in chicken and mouse. Furthermore, our new data showing that FGF-signaling regulates cell proliferation throughout the limb provides additional supporting evidence for our SU5042 results. Thus, our results are perfectly consistent with results from salamanders, anurans and chickens that show digit loss as a consequence of inhibiting cell proliferation (Discussion section). In our interpretation, loss of additional elements in a few (4/34) cases likely resulted from the small additive effect of a reduction in whole animal growth that was independent of FGF-signal inhibition.

6) The observation that Shh is also expressed in an anterior domain is interesting and should be examined. Given that cyclopamine administration gives the expected results, is this anterior domain not significant? Does the location of Shh expression correlate with areas of cell death at the anterior and posterior autopod margins?

We too find the anterior domain of Shh expression intriguing which is why we were careful to look for additional anterior expression of Ptch1 and Gli1. These support that region as real expression. However, they do not speak to function. Therefore, at the suggestion of the reviewer, we used Lysotracker to analyze cell death throughout limb development. We did not, however, detect cell death in either the posterior or anterior domains. Likewise, our cell proliferation data does not show anything that specifically correlates with the anterior domain. In light of the cyclopamine results, we cannot at present assign a specific function to the anterior domain. It is feasible that the expression contributes to limb growth, but that when Shh-signaling is inhibited the effect is swamped out by the more serious effect on AP patterning and specification of skeletal progenitors.

7) The authors make the major claim that the FGF and Hh regulatory loops are not interlinked. This is based on examining a few read-outs by whole mount RNA in situ hybridization and is not comprehensively or quantitatively assessed. Furthermore, analysis of cell death is not presented, nor are genes or proliferation rates examined PRIOR to differences in limb bud size being evident.

We agree that our initial assessment could benefit from a deeper analysis. We now present additional evidence to determine if an FGF-Shh regulatory loop exists in salamander limbs. First, the reviewer makes a good point that we should have analyzed the limb buds for altered gene expression prior to differences in limb bud size. Thus, as the limb buds emerge from the flank at stage 44, we now present an analysis of stage 45 limb buds (Figure 5C,F and Figure 6D). In response to FGF inhibition (using SU5042) we show that Etv1, Etv4 and Gremlin1 expression are not present. In contrast, we show that Ptch1 expression persists in a domain almost identical to the control limb. This data mimics our results at stage 46 (when the limb size is reduced). In addition to WISH, we now quantify gene expression changes at stage 46 and show that Etv1, Etv4, Gremlin1 and Shh expression are significantly down-regulated. Again, our qRT-PCR data shows that Ptch1 expression does not decrease (and in fact is trending towards a significant increase). This shows that Shh-signaling is not regulated by FGF-signaling. We present the same type of analysis using cyclopamine to inhibit Shh-signaling (Figure 6A-E). Our qualitative (WISH) and quantitative (qRT-PCR) analysis shows that Fgf8 expression is regulated by Shh-signaling, as is Gremlin1 expression. In addition to expression analysis, we now include functional assessment of cell death and cell proliferation. Cell death did not change in response to either treatment. Our cell proliferation analysis, however, showed that FGF-signaling regulated this process throughout the limb. This demonstrates that in contrast to anurans and amniotes where FGF-signaling controls cell survival and differentiation, in urodeles FGF-signaling controls cell proliferation instead. We obtained similar data for cyclopamine accept that Shh-signaling appeared to regulate cell proliferation to a lesser extent than FGF-signaling and that this control was more concentrated in the distal mesenchyme where Fgf8 is expressed. In addition to presenting new data, we also discuss the implications for our findings in an expanded Discussion section.

8) Finally, the analysis of D/V patterning genes seems tacked on and could be eliminated from the paper.

We agree with the reviewer and have removed this data from the paper.

In its current state this reviewer recommends rejection of the paper as a thorough examination of the above issues would certainly take longer than typically allotted for eLife submissions. Should the bold claims hold after more detailed analysis, (suggest focusing on FGF and HH pathways as most interesting and critical, while describing the order of condensation/differentiation is likely less revelatory), the study could be of great interest and significance.

We believe the additional data and discussion addresses all of the reviewer’s comments and agree that the study will be an important contribution to the field.

Reviewer #2:

Summary:

[…] They conclude that the mesenchymal-epithelial (Shh-Grem-Fgf8) feedback loop that controls limb development in other tetrapods has been re-arranged in axolotls, where it is entirely mesenchymal and has evolved different regulatory interactions. In a rich but short discussion, the authors explore the implications of these findings for understanding genetic interactions in the limb, evolution of developmental mechanisms, and molecular control of regeneration. The brevity of the discussion does a disservice to the many sacred cows that this paper has toppled, and I would encourage them to break apart the results and discussion so that a proper airing of these ideas can be presented. Overall, the data are clear and convincing and the discoveries are likely to be of interest to a broad group of biologists in the areas of development, regeneration and evolution.

We would like to thank R2 for their thorough assessment of our paper. As indicated by the Reviewing editor, several of the points raised below were identical to those raised by R1. Thus, where applicable, we note our response above and provide additional comment when necessary.

Essential revisions:

1) The Shh expression pattern is quite interesting and makes one whether the cells that form the digits ever see the Shh signal. Although the initial domain in the posterior part of the limb is typical of early limb buds of tetrapods and fishes, the subsequent domains are very different from the patterns reported for other tetrapods. They state, "Posterior expression is never detected in the autopod" (Figure 2 legend). If the digits progenitors arise from (or near) the Shh domain, then the role of Shh in digit development could be conserved. If the cells of the Shh domain are distant from the digit progenitors, then this would challenge the current paradigm of digit development. This could be tested quite easily by labeling the initial Shh domain (at stage 45) with an injection (or carefully applied crystals) of an indocarbocyanine dye, like DiI, and then examining the distribution of labeled cells several stages later in order to determine whether those cells contribute to the digits. Determining whether digit progenitors reside within the Shh domain would allow the authors to make a direct comparison with the Shh-expressing (ZPA) cells of mouse and chick limb buds, which contribute to posterior digits.

We have responded above to the comment about the anterior domain of Shh in comment reviewer 1, comment 6. We present data in a series of in situs (summarized in a schematic) showing the posterior expression domain of Shh relative to similar stages in chick and mouse (Figure 2C). This shows that in salamanders, Shh expression remains outside of the autopod after limb bud elongation. While the functional significance of this extended expression remains unclear, we appreciate the reviewer’s suggestion to use DiL as a way to investigate conservation of the posterior domain with respect to establishing digit progenitors. We performed the suggested experiment by injecting DiL at stage 45 when Shh is first expressed in the posterior and then tracked the labeled cells in limbs until all the digits had formed at stage 53. We present this data in a new figure alongside expression for Shh, Ptch1, Gli1 and Gli3 (Figure 2). Our results mirror similar experiments in chick limbs showing that the region expressing Shh gives rise to the most posterior digit and that later Shh-signaling controls growth of the autopod. When viewed alongside our data using cyclopamine (and that previously published by Stopper and Wagner, 2007), it appears Shh-signaling controls digit progenitor specification and later distal cell proliferation. Similar to mouse and chick, digit 1 specification in salamanders is not regulated by Shh-signaling.

2) Because the SU5402 treatments show that FGF signaling does not appear to maintain Shh in axolotls, is there really is a FGF-Shh feedback loop in axolotl limb bud mesenchyme?

This is a good point. We have revised our manuscript accordingly. We still refer to the feedback loop as it applies to anurans and tetrapods and where we refer to the interaction of Shh and FGF in salamanders indicate that it is not a loop.

3) Do the authors think that the same mesenchyme cells that are expressing Fgf8 are also responding to Fgf8? Do they express an appropriate receptor? How does the topological shift in the position of the ligand, Fgf8, relate to the ability of SU5402, a FGF receptor antagonist, to produce such different outcomes in axolotl vs chick and mouse limbs? Do the authors think that there is combinatorial FGF signaling in axolotls (as there is in chick and mouse), but the ligands are produced and received by the mesenchyme? If so, then can they suggest an explanation for how the same Fgfs could regulate Shh when secreted by the AER(as in chick/mouse) but do not regulate Shh when produced by the mesenchyme? Chick experiments in which the AER is removed and FGF beads are implanted into the limb bud mesenchyme suggests that the source of the ligand is irrelevant. Have they considered the possibility that Fgf8 acquired a novel function during axolotl limb evolution and lost its roles as a Shh regulator and an outgrowth signal? The proliferation data suggest that at least some of these roles are conserved.

The reviewer raises good points with respect to signal source and receiver. In an attempt to clarify what is happening during salamander limb development with respect to FGF-signaling, we took two additional approaches. First, we used in situ hybridization to detect additional FGF ligands that act as AER signals in chick and mouse (i.e., Fgf4, 9 and 17) and the two primary FGF receptors (i.e., Fgfr1 and 2). We also looked at Fgf10 since it is normally expressed in the mesenchyme, but signals to FgfR2iiib expressed in the ectoderm. We now present all of this data in a new Figure 3. Two things are apparent from this data: (1) all of the ligands (except Fgf4 which we could not detect) are expressed in the distal mesenchyme and none are expressed in the ectoderm and (2) the cognate receptors are expressed mostly in proximal cells that do not appear to overlap with the ligands. One exception is Fgfr1 which appears to be expressed more broadly at stages 44 and 45, although even then it is restricted from the most distal mesenchyme. Second, we tried to address localization of FGF-signaling in the limb and whether the same mesenchymal cells that are expressing FGF ligands are also responding to FGF-signaling. To do this we analyzed a single cell RNA-seq (scRNAseq) dataset generated for two early stages of axolotl limb development that corresponded to our stages 44 and 45 (Gerber et al., 2019). We present the results from this analysis in a new Figure 4, and two supplementary figures. Using this dataset, we conducted a principal component analysis of the sequenced cells and used known genetic markers of the proximodistal axis to discover if a principal component could accurately model the proximodistal axis. This analysis revealed a proximodistal axis as defined by expression of Meis1 and Meis2 (proximal markers) and Hoxd11 (distal) (presented in Figure 4—figure supplement 1). Next, we conducted a co-expression analysis which showed that the FGF ligands and receptors are in fact, not co-expressed along the modeled proximodistal axis. This data is now presented alongside the in situs in Figure 4. We also asked whether Fgf8, 9, 10 and 17 were co-expressed in cells that also expressed any of the four FGF receptors. This analysis is presented in Figure 4—figure supplement 2 and shows that some cells expressing Fgfr1 also express these four ligands. However, few to no cells co-express the ligands and other receptors. Together, we believe these approaches present compelling evidence that distal mesenchyme cells expressing ligands are signaling to more proximally expressed receptors, but that some distal mesenchymal cells may be able to respond in an autocrine fashion.

With respect to the second point regarding an explanation for breaking the Shh-FGF feedback loop; as discussed above, Shh-signaling appears to maintain a conserved function with regard to anteroposterior patterning, specification of digit progenitor cells, and distal growth of the autopod. Presumably, movement of FGF-signaling to the mesenchyme occurred after the divergence of urodeles and anurans and our cell proliferation and cell death data suggest that FGF-signaling adopted a novel function in supporting growth of the limb rather than controlling cell survival and proximodistal pattern. We discuss this in relation to results from salamanders, anurans, and amniotes which show similar phenotypes when cell proliferation is broadly inhibited. How this happened in the context of evolution is much more difficult to answer and we are not sure it can be, at least not in the scope of this manuscript.

4) In several places, the Fgf8 result is generalized to the other Fgfs involved in AER signaling (e.g., Results and Discussion section: "Taken together, our expression data clearly shows that the Shh-Grem-FGF feedback loop is entirely expressed within the mesenchymal compartment during salamander limb development."). In chick and mouse limbs, at least 4 Fgfs (Fgf4, Fgf8, Fgf9, and Fgf17) are expressed in the AER and signal through mesenchymal FGF receptors. Whilst it would be informative to examine the expression patterns of these other Fgfs in axolotl limb buds to determine whether the Shh-FGF feedback loop is entirely in the mesenchyme (or if some of the Fgfs have remained in the epithelium) if they choose not to examine the expression of these other Fgfs, then they should tailor the discussion accordingly and at least mention the alternate possibilities. For example, they show that Fgf8 is expressed in the mesenchyme, but it is unknown whether or where other Fgfs are expressed in the limb. Thus, several scenarios are possible, including (a) Fgf8 is mesenchymal but Fgf4, Fgf9, and Fgf17 are expressed in the epithelium, (b) All 4 Fgfs are mesenchymal, (c) Fgf8 is mesenchymal and the other 3 Fgfs are not expressed. Without knowing which of these scenarios is true, one must be cautious and clear with the interpretation the SU5402 results since SU5402 treatment probably suppresses all FGF signaling. For example, scenarios (b) and (c) would support "a topological rearrangement of the Shh-Grem1-FGF loop", but scenario (a) would not because 3 of the 4 Fgfs would be expressed in ectoderm. To be clear, I am not insisting that the authors examine every FGF and FGF receptor in axolotl limbs, but they need to consider the complexity of FGF signaling in amniotes in their interpretation of the results. They could also cite other experimental work that supports their argument that the critical FGF signal is mesenchymal, such as the work of Lauthier (1985), who showed that newt limbs can develop even when the apical epithelium is removed from the mesenchyme.

We direct the reviewer to our responses reviewer 1, comment 4 and reviewer 1, comment 5 above.

5) Figure 1C. Looking at the alcian blue stained limbs, there are no visible digit condensations at stage 47, and at stage 48, three distinct digit condensations are stained. At stage 49, these 3 condensations remain visible (only I and II are labeled in the figure, but digit III is also stained). The initial condensation of digit IV is also visible at stage 49. Given that no digits are visible at stage 47, and three digits (I, II, III) appear at stage 48, what is the basis for the argument that digits form from anterior to posterior? (see the following point for continued discussion of this issue)

6) Figure 1E. I found the Sox9 in situ hybridizations to be less informative than the alcian blue preparations, possibly due to the level of background staining, although this could also be a result of slight staging differences between the alcian blue and Sox9 limbs. For example, although one can see three digits in the alcian blue limb at stage 48, only digit II is obvious in the Sox9 in situ hybridization at the same stage. Is the latter a slightly younger limb, or is the in situ just not as clear as the alcian blue preparation?

We direct the reviewer to our response to reviewer 1, comment 2

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Essential revisions:

1) Please break the Discussion section into subsections, including one that compares the findings in axolotl with chick and mouse.

2) Please add a table that compares limb development in axolotl with mouse/chick.

3) Please try, if possible, within a reasonable time frame, to address the reviewer comments below, though some were deemed non-essential (e.g. scRNAseq of limb ectoderm; re-imaging the immunofluorescence). If you cannot address certain comments, please explain why.

Following the essential revisions provided, we have broken up the Discussion section into subheadings and specifically included one where our findings in axolotl are compared with Xenopus, chick and mouse. In this same section we now reference a new Table 1 detailing these comparisons. In response to the specific comments from the two reviewers we have outlined our point-by-points replies below.

Reviewer #1:

[…] Essential revisions:

1) It would be nice to see some RNA-ISH that is positive in the overlying ectoderm just to rule out the possibility that lack of staining is not due simply to a technical problem with doing RNA-ISH on axolotl ectoderm (some permeability problem). Another method would be to do RNA-ISH on a section.

We understand the request for this piece of data. To demonstrate the reliability of our protocol at detecting ectodermal, as well as mesenchymal expression in developing axolotl limb buds, we now include an additional supplementary figure (Figure 2—figure supplement 1). This figure shows Msx2 staining (which is absent from the ectoderm) in the distal mesenchyme alongside in situs for two Keratin genes, Krt5 and Krt17. Staining for these genes is observed broadly across the surface ectoderm. To further show these genes are expressed in the ectoderm, we provide sections of our embryos where localization is clearly demonstrated in the overlying ectoderm (and is not expressed in mesenchymal cells).

2) Please show Fgf4 RNA-ISH and other 'data not shown' such as cell death assays even if not very high.

We appreciate the request to include these data which we now include as additional figure supplements. We now present limb buds negative for Fgf4 expression in Figure 3—figure supplement 1 which show no expression at stages 44-46. We have expanded Figure 5—figure supplement 2 to show the lack of cell death during normal development at stage 44 and 46 using Lysotracker Red. We also now include data for stage 46 limb buds from control and treatment animals from our SU5402 and cyclopamine experiments.

3) How come only scSeq in mesenchyme is shown? Could an analysis of the ectoderm be included? In no FGF ligands detected in this data set, this would be nice to report.

The scRNAseq data we analyzed and that was reported in Gerber et al., (2018) for limb buds excluded ectodermal cells. The authors of that paper did not release the data for the ectodermal cells they excluded. We had originally hoped to do this, but discovered they had excluded these cells.

4) Lysotracker and TUNEL are coincident in mouse/chick. Are they in axolotol? Could cells be dying but this is not visible due to failure of phagocytic bodies?

5) The second paragraph of the Discussion section does a better job at summarizing the paper than Abstract, perhaps the abstract could be enhanced with some of the ideas here.

We agree and we originally had an abstract that better reflected this summary in the Discussion section.

However, eLife only allows an abstract of 150 words and this length does not allow much detail. Thus, the current Abstract attempts to present several key findings instead of every one we detail in the paper.

6) Even for the limb bud aficionado it would be useful to have a chart comparing mouse/chick results to findings in axolotl.

We now include a table (new Table 1) that appears with the discussion. The table compares salient features of forelimb development between axolotl, Xenopus, chick and mouse.

7) In Figure 5H and Figure 6—figure supplement 1 panel F, I can't tell which of the red patches are positive above background. Also, in the mid-sagittal slice the control limb bud shows almost all cells as positive, is this correct?

We believe the reviewer is referring to the lightsheet images depicting cell proliferation in Figure 5 and Figure 6 (so not the supplement for Figure 6). Provided this is the case, all the red “patches” are in fact, EdU+ cells. There is no background. Because the lightsheet provides a threedimensional projection of all the data, variations in brightness refer to cells that are closer to the front of the projection. As for the slice the reviewer is referring too (mid-longitudinal), yes, almost all cells are positive for EdU at this stage.

8) Why are Christensen et al., 2002; Han et al., 2001 incorrect in their observation of FGF expression in the AER?

First, we inadvertently cited Christensen et al., 2002 along with Han et al., 2001. Han et al., 2001 present whole mount expression data and state that “Fgf8 is expressed in limb bud epidermis and mesenchyme of the budding limb”. They go on to state in their discussion, “In the present study, we have found that the FGF-8 expression domain is translocated from epidermis to mesenchyme gradually as limb development proceeds”. We simply state that this is not what we found because at no time do we observer ectodermal expression of any FGF ligands. The quality of their images makes it impossible to determine in what compartments Fgf8 is or is not expressed. So all we are left with is what they state in their paper. Christensen et al., 2002 state, “Fgf8 and 10, highly expressed in the amniote developing limb AER and mesoderm, respectively (Crossley et al., 1996; Ohuchi et al., 1997), were accordingly also expressed in axolotl developing and regenerating limbs”. Thus, they imply that Fgf8 is expressed in axolotl limb ectoderm. However, they do not actually show an in situ for developing limbs in that paper. Therefore, it is inappropriate to cite Christensen et al., 2002 since they do not present data to support or refute our data. Thus, we have removed the Christensen et al., 2002 citation.

https://doi.org/10.7554/eLife.48507.027

Article and author information

Author details

  1. Sruthi Purushothaman

    Department of Biology, University of Kentucky, Lexington, United States
    Contribution
    Conceptualization, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3974-4731
  2. Ahmed Elewa

    Cold Spring Harbor Laboratory, Cold Spring Harbor, United States
    Contribution
    Formal analysis, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1988-7970
  3. Ashley W Seifert

    Department of Biology, University of Kentucky, Lexington, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Investigation, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    awseifert@uky.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6576-3664

Funding

National Science Foundation (IOS -1353713)

  • Ashley W Seifert

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors thank Ryan Woodcock for help with axolotl bioinformatics, Elly Tanaka and Ina Stutzer for access to genomic data, James Monaghan and Malcolm Maden for discussion, Jakub Famulski and Douglas Harrison for assistance with lightsheet imaging and image processing. We are grateful to our anonymous reviewers for their extremely helpful comments.

Ethics

Animal experimentation: All procedures were conducted in accordance with, and approved by, the University of Kentucky Institutional Animal Care and Use Committee (IACUC Protocol: 2013-1174).

Senior and Reviewing Editor

  1. Marianne E Bronner, California Institute of Technology, United States

Publication history

  1. Received: May 16, 2019
  2. Accepted: August 19, 2019
  3. Version of Record published: September 20, 2019 (version 1)
  4. Version of Record updated: September 24, 2019 (version 2)

Copyright

© 2019, Purushothaman et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 894
    Page views
  • 95
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Developmental Biology
    2. Human Biology and Medicine
    Benjamin R Thomson et al.
    Short Report
    1. Developmental Biology
    2. Microbiology and Infectious Disease
    Paola Bardetti et al.
    Research Article