Posterior shift of Shh-Fgf signaling in axolotl limb regeneration drives sequential digit formation Authors: Saya Furukawa, Sakiya Yamamoto, Haruki Nakayama, Ayaka Ohashi, Akira Satoh Published: 2026 ABSTRACT SummaryDo conserved morphogen modules generate a given form only via a single spatiotemporal deployment, or can alternative deployments yield the same morphology? Axolotl limb regeneration provides a tractable test bed. SHH and FGF8, universally used in limb formation across vertebrates, also operate in axolotl, although the spatial domain of Fgf8 differs markedly from that in amniotes. A mutual Shh/Fgf feedback loop is likewise conserved. Here, we show that, in axolotl, the active domain of this loop shifts progressively posterior as digit formation proceeds. In step with this anterior-to-posterior displacement, digit-forming regions are sequentially induced posteriorly, explaining the axolotl’s reversed order relative to the amniote posterior-to-anterior sequence. These findings indicate that conserved molecular toolkits can overcome differences in spatial deployment to produce equivalent final limb architectures, demonstrating that there is not a single route to a target morphology. Introduction Is the cellular machinery that executes morphogenesis genuinely singular, or could alternative mechanisms generate the same form? Although developmental mechanisms are often deeply conserved—and therefore assumed to be rational and robust—biological systems frequently allow multiple operational principles to generate similar outcomes.1,2 In the context of morphogenesis, the process relies on the precisely coordinated spatiotemporal deployment of multiple signaling modules that regulate cell proliferation, migration, and differentiation. Because many of these regulatory architectures are highly conserved across species, it is often assumed that there is a single canonical route to a given morphological outcome. Whether this assumption is valid, however, remains unclear. To tackle this issue, limb morphogenesis in urodeles—an exception within the otherwise deeply conserved morphogenetic system—offers a powerful observational window. Across vertebrates—from fish fins to mammalian limbs—the core signaling modules specifying the anteroposterior (AP), proximodistal (PD), and dorsoventral axes are highly conserved. Numerous studies have also documented broadly conserved gene-expression programs and patterning logic that underline this shared architectural framework. For example, Sonic hedgehog (Shh) is consistently expressed at the posterior margin of the limb-bud mesenchyme, forming the zone of polarizing activity (ZPA), where it contributes to AP patterning.3,4,5,6,7,8,9 Likewise, Fgf8 is localized to the apical ectodermal ridge (AER), where it promotes PD outgrowth and sustains AER-ZPA interactions through reciprocal feedback with Shh.4,10,11 The robustness and evolutionary conservation of this AER-ZPA module have fostered the prevailing view that the highly conserved spatiotemporal deployment of multiple signaling modules constitutes the canonical route—or “the only correct answer”—for generating limb morphology. Yet urodeles such as axolotls (Ambystoma mexicanum) represent a striking exception. In axolotls, Shh is expressed posteriorly as in amniotes,8 whereas Fgf8 is localized to the anterior mesenchyme, placing the Shh/Fgf8 feedback loop entirely within the mesenchyme rather than across epithelial-mesenchymal compartments.12,13 Furthermore, digits develop in the reverse order relative to amniotes: Axolotls first develop the anterior (preaxial) digits, followed by the addition of more posterior (postaxial) digits, instead of following the posterior-to-anterior sequence characteristic of amniotes.14,15,16 Despite these differences in expression domains and in the sequence of digit formation, the overall limb anatomical architecture is highly comparable to that of amniotes. Taken together, these observations raise the possibility that urodeles generate a conserved limb architecture through a spatiotemporal program that diverges from the canonical route—suggesting that conserved limb forms may not necessarily arise from a single developmental pathway. In addition to these evolutionary developmental differences from amniotes, the unique context of limb regeneration provides an opportunity to probe more deeply how shared anatomical outcomes can arise from distinct initial conditions. As a central premise, regeneration reactivates many developmental gene-expression programs, and much of the underlying regulatory logic is thought to be conserved,17,18,19 yet proceeds from mature tissues and requires regeneration-specific events, including dedifferentiation, re-epithelialization, and nerve-dependent responses.20,21,22,23,24 Moreover, regenerating blastemas span a wide size range (∼0.5–8 mm) and exhibit allometric expansion of Shh and Fgf8 expression domains, whereas the effective diffusion ranges of their morphogens remain approximately constant.25,26 We previously suggested that the mesenchymal positioning of Fgf8 in urodeles contributes to maintaining near-constant SHH/FGF8 molecular ratios under such variable scaling conditions.25 Together, these features position axolotl regeneration as a system well suited to revealing how flexible—and potentially plural—the mechanisms generating conserved limb forms may be. Thus, axolotl limb regeneration provides a rare opportunity to test whether a conserved limb architecture requires one canonical route, or can instead be achieved through alternative developmental paths. Here, we account for the reversed order of digit formation in axolotls through the spatial dynamics of Shh/Fgf signaling. We identified within the blastema a high-activity domain of Shh/Fgf signaling, which we term ΩSF, characterized by elevated cell proliferation and specification of the chondrogenic region.25 In this study, we further show that ΩSF shifts posteriorly over time, and our data suggest that this displacement drives the sequential, anterior-to-posterior progression of digit formation. Axolotls’ mesenchymal expression of Fgf8—unlike the AER-localized pattern in amniotes—places the Shh/Fgf feedback loop entirely within the mesenchyme, potentially granting ΩSF increased positional mobility. These observations suggest that axolotls may use the conserved Shh/Fgf module through a urodele-specific spatiotemporal organization. Together, these findings provide a conceptual vantage point from which to reconsider whether conserved limb forms must follow a single developmental route, or whether alternative regulatory sequences may also give rise to equivalent morphologies. This perspective adds resolution to the consensus built from amniote studies and aids in extracting the core elements—the “backbone”—of the limb morphogenetic system while redefining the boundary conditions of its design space. Result A posterior shift of the Shh/Fgf feedback loop activity as digit differentiation progresses The Shh/Fgf feedback loop is a key factor in regulating digit formation during limb development and regeneration in axolotls.12,13,25 Previously, we reported that this loop promotes cell proliferation, with the highest cell density appearing between the Shh and Fgf8 expression domains in blastemas.25 Since the region with the highest cell density coincides with the site of the first-forming digit (digit II), the Shh/Fgf feedback loop has been suggested to lead the digit II formation by controlling the cell proliferation activities.25 Here, we examined whether this feedback loop also contributes to the formation of more posterior digits (digit III-IV (forelimb) or III-V (hindlimb)). We defined an early digit stage (EDS) as the stage at which a single cartilage condensation, marked by type II collagen (Col2) expression, was detected in the distal region (Figures 1A and 1B), with sparse Col2 signals observed anteriorly, likely corresponding to digit I (Figures 1A and 1B).To investigate gene expression after the EDS, we collected samples a few days later, in which two distinct cartilage condensations were observed on the distal section (Figure 1I), along with faint posterior Col2 signals likely representing digit III (Figure 1I). We defined this stage as the middle digit stage (MDS) (Figure 1A). In both EDS and MDS blastema samples, Shh was not detected on the distal sections, with its expression localized to more proximal sections. Therefore, we performed analysis on proximal sections where HoxA13, an autopod marker, was expressed (Figures 1D and 1K). As previously reported, the expression signals of Shh and Fgf8 become less detectable as limb regeneration progresses.25 Although their signals are weak, Shh and Fgf8 expression were consistently detected in the EDS blastema, with the accumulated Col2 signals positioned between the Shh and Fgf8 signals (Figures 1C–1E, 1F). On the MDS proximal section, the two cartilage aggregations merged into one (Figure 1J, black arrowhead). The Shh expression remained detectable in most samples (Figure 1L, L’, black arrowheads), while the Fgf8 expression was too weak to confirm their signals by in situ hybridization (ISH) (Figure 1M, M’). Hence, we performed quantitative RT-PCR (qRT-PCR) analysis (Figures 1P and 1Q). The results showed that Shh expression at MDS was not significantly reduced compared with EDS (Figure 1P), whereas Fgf8 showed a decreasing trend (p = 0.051) (Figure 1Q). However, comparing expression levels to intact limbs, both Shh and Fgf8 expression levels at MDS remained significantly higher than those in intact limbs (Figures 1P and 1Q). These results suggest that the Shh/Fgf feedback loop remains active at MDS although its activity declines as digit formation progresses.Figure 1A posterior shift of the Shh/Fgf feedback loop activity as digit differentiation progresses(A) Definition of the blastema stages.(B–G; I–N) Gene expression patterns visualized by ISH in EDS (b–g) and MDS (i–n) blastemas. Col2 (B, C, I, and J), HoxA13 (D and K), Shh (E and L), Fgf8 (F and M), Spry1 (G and N). (B) and (I) are sections in the distal level, and those in (C–G; J–N) are sections in the proximal level. Enlarged views of the boxed regions in (E), (F), (L), and (M) are shown in (E′), (F′), (L′), and (M′), respectively.(H and O) Virtually merged images, with (H) showing the merge of (C, E, and G) and (O) showing the merge of (J, L, and N). Open arrowheads in (B) and (I) indicate sparse Col2+ signals, and filled arrowheads in (B, C, I, and J) mark Col2+ cell aggregates. Scale bars, 400 μm (B and I); 100 μm (E′ and L′). The dotted lines in (B–G), (I–N), and the gray lines in (H and O) indicate the boundary of the blastema epithelium.(P and Q) Relative expression levels of Shh (P) and Fgf8 (Q) in blastemas measured by qRT-PCR. Statistical analysis was performed on biological replicates (n = 6–9 for each stage) using the Bonferroni correction. Each data point represents a technical replicate, and consistent trends were observed across independent biological replicates. Diamonds and triangles represent forelimb and hindlimb data points, respectively. Data are represented as mean ± SEM.(R) Blastema size used for the quantification of gene expression domains shown in (T–X).(S) Schematic illustrating the measurement methods used in (T–X).(T–V) Quantification of the relative position of the Spry1+ relative to the Col2+. The posterior end (T), center (U), and anterior end (V) of the Spry1+ were measured. The center of the Col2+ was set to 0 (indicated by the dashed magenta line); positions anterior to this reference are represented by negative values, and posterior positions by positive values.(W and X) Distances between the centers of the Shh+ and Col2+ (w) and the Shh+ and Spry1+ (X). Statistical analysis for (R; T–X) was performed using an unpaired two-tailed Student’s t-test. Data are presented as box-and-whisker plots (R; T–X). Sample sizes were n = 5 biological replicates for both EDS and MDS. A posterior shift of the Shh/Fgf feedback loop activity as digit differentiation progresses (A) Definition of the blastema stages. (B–G; I–N) Gene expression patterns visualized by ISH in EDS (b–g) and MDS (i–n) blastemas. Col2 (B, C, I, and J), HoxA13 (D and K), Shh (E and L), Fgf8 (F and M), Spry1 (G and N). (B) and (I) are sections in the distal level, and those in (C–G; J–N) are sections in the proximal level. Enlarged views of the boxed regions in (E), (F), (L), and (M) are shown in (E′), (F′), (L′), and (M′), respectively. (H and O) Virtually merged images, with (H) showing the merge of (C, E, and G) and (O) showing the merge of (J, L, and N). Open arrowheads in (B) and (I) indicate sparse Col2+ signals, and filled arrowheads in (B, C, I, and J) mark Col2+ cell aggregates. Scale bars, 400 μm (B and I); 100 μm (E′ and L′). The dotted lines in (B–G), (I–N), and the gray lines in (H and O) indicate the boundary of the blastema epithelium. (P and Q) Relative expression levels of Shh (P) and Fgf8 (Q) in blastemas measured by qRT-PCR. Statistical analysis was performed on biological replicates (n = 6–9 for each stage) using the Bonferroni correction. Each data point represents a technical replicate, and consistent trends were observed across independent biological replicates. Diamonds and triangles represent forelimb and hindlimb data points, respectively. Data are represented as mean ± SEM. (R) Blastema size used for the quantification of gene expression domains shown in (T–X). (S) Schematic illustrating the measurement methods used in (T–X). (T–V) Quantification of the relative position of the Spry1+ relative to the Col2+. The posterior end (T), center (U), and anterior end (V) of the Spry1+ were measured. The center of the Col2+ was set to 0 (indicated by the dashed magenta line); positions anterior to this reference are represented by negative values, and posterior positions by positive values. (W and X) Distances between the centers of the Shh+ and Col2+ (w) and the Shh+ and Spry1+ (X). Statistical analysis for (R; T–X) was performed using an unpaired two-tailed Student’s t-test. Data are presented as box-and-whisker plots (R; T–X). Sample sizes were n = 5 biological replicates for both EDS and MDS. Since it was revealed that the Shh/Fgf feedback loop remains active at MDS, we next examined where this feedback loop activity occurs. As Fgf8 expression at MDS was too weak to define its expression domain, we used Sprouty1 (Spry1), a known marker of Fgf signaling transduction,27 to estimate the region of Fgf signaling activity. At EDS, Spry1 signals largely overlapped with the Fgf8 expression domain and exhibited a dorsoventrally separated expression pattern (Figures 1F and 1G). Given that this dorsoventrally separated gene expression pattern is a characteristic of Fgf8 among FGF ligands in blastemas, Spry1 is likely to reflect Fgf8-dependent signaling predominantly. Although Spry1 may also respond to other mesenchymally expressed FGF ligands, such as Fgf10, its expression nonetheless delineates the spatial domain of active Fgf signaling in this region. A merged image showed that the Spry1 expression domain (Spry1+) was positioned approximately at the same AP level as the Col2 expression domain (Col2+). (Figure 1H). At MDS, however, the Spry1+ appeared in more posterior regions than Col2+ (Figures 1J, 1N, and 1O). To quantitatively assess this change, we measured the relative positions of gene expression domains using blastemas of similar size at the EDS and MDS (Figures 1R and 1S). The results showed that the posterior end and center of the Spry1+ domain relative to Col2+ were positioned significantly posteriorly at the MDS compared with the EDS, whereas the anterior end showed no significant difference (Figures 1T–1V). Additionally, we measured the distances between the center position of the Shh expression domains (Shh+) and the Col2+, as well as between the Shh+ and the Spry1+ (Figures 1W and 1X). The Shh-Col2 distance was significantly greater at the MDS than at the EDS, whereas the Shh-Spry1 distance showed no significant difference. This suggests that the posterior region adjacent to the Col2 signals expanded during the transition from the EDS to the MDS. Taken together, our quantitative analysis revealed that Fgf signaling activity, indicated by Spry1 expression, shifted posteriorly relative to presumptive digit II cartilage aggregates as digit differentiation progressed, suggesting a posterior shift of the Shh/Fgf feedback loop activity. Influence of blastema differentiation on the position of the Shh/Fgf feedback loop We next examined whether the posterior shift of the Shh/Fgf feedback loop activities was caused by changes in the differentiation state of blastema cells during regeneration. To test this hypothesis, we grafted either undifferentiated mesenchymal tissues from a middle bud stage blastema (control) or differentiating mesenchymal tissues from an MDS blastema into the middle bud stage blastema and compared the effects on the positions of the Shh and Fgf8 expression domains (Figure 2A). We used GFP transgenic animals as donors and grafted the tissues into wild-type (WT) animals to facilitate the identification of grafted tissues (Figures 2A and 2B). Samples were fixed 3 days post-grafting (dpg), and the positional relationships among Shh, Fgf8, and GFP signals were analyzed on transverse sections (Figures 2C–2Q). To exclude the potential effect of blastema size, host animals with blastemas of comparable size were used, as no significant difference was observed between groups (Figure 2C). Shh, Fgf8, and GFP signals were detected in three adjacent sections by ISH (for Shh and Fgf8) and immunofluorescence (for GFP) (Figures 2D–2F; 2H–2J). By comparing the merged images of these three sections (Figures 2G and 2K), it seemed that the Fgf8+ domain relative to the GFP signals in the MDS mesenchyme-grafted blastema (Figure 2K) was located more posteriorly compared with control samples (Figure 2G). To quantify this shift and specifically evaluate the effects of the grafted tissues, we measured the relative positions of the Shh+ and Fgf8+ domains with respect to the GFP-positive grafts (Figures 2L–2Q). The relative position of the Shh+ to the GFP signal did not significantly differ between conditions (Figures 2M and 2N). In contrast, the position of Fgf8+ significantly shifted posteriorly in samples with MDS mesenchyme grafts (Figures 2O and 2P). We also measured the distance between the centers of the Shh+ and the Fgf8+, which showed a significant decrease when MDS mesenchyme was grafted (Figure 2Q). To assess whether ECM changes accompanying blastema differentiation contribute to the altered positioning of gene expression domains, we performed analogous grafting experiments using decellularized mesenchymal tissue from MDS blastemas (Figure S1). In these samples, the Fgf8+ domain relative to the GFP-positive graft showed a modest posterior tendency, and the distance between the centers of the Shh+ and Fgf8+ domains was correspondingly reduced. However, these effects were weaker than those observed with intact MDS mesenchyme grafts and did not reach statistical significance compared with controls. These cell grafting results suggest that changes in the cellular composition of the blastema mesenchyme as regeneration progresses may contribute to the posterior shift of the Shh/Fgf feedback loop.Figure 2Blastema differentiation affects the position of the Shh/Fgf feedback loop(A) Experimental procedure. Mesenchymal tissues from the middle bud stage or MDS blastema of GFP-transgenic animals were grafted into the middle bud stage blastemas of wild-type animals.(B) Fluorescence stereomicroscopic image at 3 days post-grafting (3 dpg), showing the GFP+ grafted tissue in the central region of the regenerating blastema. Scale bars, 1 mm.(C) Blastema size used for the quantification of gene expression domains shown in (M–Q). Statistical analysis was performed using an unpaired two-tailed Student’s t-test. Data are presented as box-and-whisker plots.(D–F) Shh and Fgf8 expression and GFP signals in blastemas grafted with middle bud mesenchyme were visualized by ISH and immunofluorescence, respectively. An enlarged view of the boxed regions in (E) is shown in (E′).(G) Virtually merged image of (D), (E), (F).(H–J) Shh and Fgf8 expression and GFP signals in blastemas grafted with MDS mesenchyme were visualized by ISH and immunofluorescence, respectively. An enlarged view of the boxed regions in (I) is shown in (I′).(K) Virtually merged image of (H), (I), (J). Scale bars, 200 μm (D, E′, H, I′). The dotted lines in (D–F), (H–J), and the gray lines in (G, K) indicate the boundary of the blastema epithelium.(L) Schematic illustrating the measurement methods used in (M–Q).(M–P) Quantification of the relative position of the Shh (M and N) and Fgf8 (O and P) expression domains relative to the GFP+ domain. The center positions of Shh+ (M) and Fgf8+ (O), the posterior end of Shh+ (N), and the anterior end of Fgf8+ (P) were measured. The center of the GFP+ domain was set to 0 (indicated by dashed light green lines); positions anterior to this reference are represented by negative values and posterior positions by positive values.(Q) Distance between the centers of the Shh+ and Fgf8+ domains. Statistical analysis for (M–Q) was performed using a two-tailed Student’s t-test. Data are presented as box-and-whisker plots (M–Q). Sample sizes: middle bud mesenchyme, n = 6 biological replicates; MDS mesenchyme, n = 9 biological replicates, for c and M–Q. Blastema differentiation affects the position of the Shh/Fgf feedback loop (A) Experimental procedure. Mesenchymal tissues from the middle bud stage or MDS blastema of GFP-transgenic animals were grafted into the middle bud stage blastemas of wild-type animals. (B) Fluorescence stereomicroscopic image at 3 days post-grafting (3 dpg), showing the GFP+ grafted tissue in the central region of the regenerating blastema. Scale bars, 1 mm. (C) Blastema size used for the quantification of gene expression domains shown in (M–Q). Statistical analysis was performed using an unpaired two-tailed Student’s t-test. Data are presented as box-and-whisker plots. (D–F) Shh and Fgf8 expression and GFP signals in blastemas grafted with middle bud mesenchyme were visualized by ISH and immunofluorescence, respectively. An enlarged view of the boxed regions in (E) is shown in (E′). (G) Virtually merged image of (D), (E), (F). (H–J) Shh and Fgf8 expression and GFP signals in blastemas grafted with MDS mesenchyme were visualized by ISH and immunofluorescence, respectively. An enlarged view of the boxed regions in (I) is shown in (I′). (K) Virtually merged image of (H), (I), (J). Scale bars, 200 μm (D, E′, H, I′). The dotted lines in (D–F), (H–J), and the gray lines in (G, K) indicate the boundary of the blastema epithelium. (L) Schematic illustrating the measurement methods used in (M–Q). (M–P) Quantification of the relative position of the Shh (M and N) and Fgf8 (O and P) expression domains relative to the GFP+ domain. The center positions of Shh+ (M) and Fgf8+ (O), the posterior end of Shh+ (N), and the anterior end of Fgf8+ (P) were measured. The center of the GFP+ domain was set to 0 (indicated by dashed light green lines); positions anterior to this reference are represented by negative values and posterior positions by positive values. (Q) Distance between the centers of the Shh+ and Fgf8+ domains. Statistical analysis for (M–Q) was performed using a two-tailed Student’s t-test. Data are presented as box-and-whisker plots (M–Q). Sample sizes: middle bud mesenchyme, n = 6 biological replicates; MDS mesenchyme, n = 9 biological replicates, for c and M–Q. Given that changes in the differentiation state of blastema cells during regeneration have been suggested to cause posterior displacement of the Shh/Fgf8 domain, digit cartilage differentiation represents a reasonable candidate factor underlying this posterior shift. As we discussed later in discussion, differentiating cartilage might act as a physical barrier or an absorbent for SHH diffusion. To examine how the emergence of such obstacles affects SHH diffusion and the position of Fgf8 expression, we performed in silico analyses using a reaction-diffusion model (Figure 3).28 In this model, the spatiotemporal distribution of secreted SHH protein is described by a minimal reaction-diffusion equation that incorporates diffusion, degradation, and localized production: ∂tC = D∇2C-γC+S0χsource, where C represents the SHH protein concentration, D is the diffusivity, γ the degradation rate, S0 the synthesis rate, and χsource the indicator function of the SHH protein source region (the Shh gene expression domain). As a quantitative readout of the anterior extent of SHH protein diffusion, we defined the anterior end of the SHH protein domain as follows: For each image row, we identified the first pixel whose SHH concentration exceeded a fixed threshold (C > 0.001) and recorded its x-coordinate; the anterior end position was defined as the minimum of these values across all rows. After the system had largely reached a quasi-steady state of SHH distribution (at simulation frame 1000 out of 30,000), we introduced a circular obstacle (radius 10), representing a differentiating digit cartilage (Figure 3). The obstacle was implemented in two distinct conditions. First, it was modeled as an impermeable body with a no-flux (Neumann) boundary condition (Figures 3B–3D), representing a physical barrier to SHH diffusion. Under this condition, we simulated the anterior extent of SHH protein diffusion while varying key reaction-diffusion parameters: the diffusion coefficient D from 1.0 (purple) to 5.0 (red) (Figure 3B), the degradation rate γ from 0.1 (purple) to 0.4 (red) (Figure 3C), and the synthesis rate S0 from 0.1 (purple) to 5 (red) (Figure 3D). Second, it was modeled as an adsorbing body with partial-adsorption (Robin) boundary condition, -D∂nC = kC (Figure 3E), representing local SHH absorption. The strength of adsorption was controlled by the dimensionless parameter k. In the visualization, low values of k (approaching no adsorption) are shown in purple, whereas high values of k (corresponding to strong or effectively complete adsorption) are shown in red, with intermediate values represented by a graded color scale (Figure 3E, right). For intuition, we also visualized heat maps of SHH and of FGF8, which is transcriptionally responsive to Shh signaling activity (Figure 3A); the analysis itself used the presence map. Parameter sweeps (Figures 3B–3D) showed that—except under extreme parameter settings—the anterior limit of SHH diffusion shifted posteriorly after the obstacle appeared, which was likely responsible for the posterior displacement of the SHH-dependent Fgf8 expression domain. For example, in the diffusivity (D) sweep, a posterior shift was observed both at the minimal diffusivity (Figure 3B, purple trace) and at the maximal diffusivity (Figure 3B, red trace): In both cases, the SHH anterior end moved posteriorly after the obstacle appeared (after frame 1000). Similar trajectories were obtained when varying the degradation rate (γ) or synthesis rate (S0) (Figures 3C and 3D). Introducing surface adsorption (Figure 3E) advanced the timing of this posterior shift as the degree of adsorption increased, whereas the magnitude of the shift is similar to others (global minimum x), because rows near the dorsal/ventral margins that bypass the obstacle often determine that minimum. Taken together, these analyses support the view that progressive changes in the composition of mesenchymal tissue during cartilage differentiation shorten the effective diffusion range of SHH, thereby providing a plausible explanation for the posterior shift of the Shh/Fgf8 feedback loop activity. This expectation is consistent with the spatiotemporal course of digit differentiation, implying that SHH recedes posteriorly in step with digit formation.Figure 3Physical barrier- or absorbent-induced posterior shift of the SHH diffusion anterior boundary in 2D simulations(A) Snapshots from a 2D reaction–diffusion simulation showing SHH concentration (left), FGF8 concentration (middle), and the binarized SHH-positive area (right; threshold SHH >0.001) at the indicated frames (t/30,000). Dashed lines in the right panels indicate the quantified SHH anterior end. A circular body, representing either a physical barrier (B–D) or an adsorbing body (E), is introduced at t = 1,000 (dashed line in B–E). The bottom two rows of each column show an example simulation in which a circular body representing a physical barrier was introduced (middle panel), and the subsequent predicted outcome (bottom panel) is shown.(B–D) Trajectories of the SHH anterior end position along the anteroposterior axis before and after introduction of the physical barrier. (B) The diffusion coefficient D is varied from 1.0 to 5.0. (C) The degradation rate γ is varied from 0.05 to 0.40. (D) The synthesis rate S is varied from 0.10 to 5.0. All other parameters are held constant at D = 2.5, γ = 0.1, and S0 = 2.5. The vertical dashed line marks the time (frames) of physical barrier introduction.(E) Effect of surface adsorption (absorbing body). The absorbing body is introduced at t = 1,000. Left, SHH-positive masks at the indicated time (frames) for simulations with an adsorbing body (surface adsorption). Right, comparison of SHH anterior end trajectories across adsorption strengths, with D = 2.5, γ = 0.1, and S0 = 2.5 held constant. Line color indicates adsorption strength: purple, no adsorption (k = 0; no-flux); blue–green, intermediate adsorption strengths; red, the strongest adsorption condition. Physical barrier- or absorbent-induced posterior shift of the SHH diffusion anterior boundary in 2D simulations (A) Snapshots from a 2D reaction–diffusion simulation showing SHH concentration (left), FGF8 concentration (middle), and the binarized SHH-positive area (right; threshold SHH >0.001) at the indicated frames (t/30,000). Dashed lines in the right panels indicate the quantified SHH anterior end. A circular body, representing either a physical barrier (B–D) or an adsorbing body (E), is introduced at t = 1,000 (dashed line in B–E). The bottom two rows of each column show an example simulation in which a circular body representing a physical barrier was introduced (middle panel), and the subsequent predicted outcome (bottom panel) is shown. (B–D) Trajectories of the SHH anterior end position along the anteroposterior axis before and after introduction of the physical barrier. (B) The diffusion coefficient D is varied from 1.0 to 5.0. (C) The degradation rate γ is varied from 0.05 to 0.40. (D) The synthesis rate S is varied from 0.10 to 5.0. All other parameters are held constant at D = 2.5, γ = 0.1, and S0 = 2.5. The vertical dashed line marks the time (frames) of physical barrier introduction. (E) Effect of surface adsorption (absorbing body). The absorbing body is introduced at t = 1,000. Left, SHH-positive masks at the indicated time (frames) for simulations with an adsorbing body (surface adsorption). Right, comparison of SHH anterior end trajectories across adsorption strengths, with D = 2.5, γ = 0.1, and S0 = 2.5 held constant. Line color indicates adsorption strength: purple, no adsorption (k = 0; no-flux); blue–green, intermediate adsorption strengths; red, the strongest adsorption condition. Posterior shift of the proliferative region in relation to the Shh/Fgf feedback loop Given that the Shh/Fgf feedback loop promotes cell proliferation13 and that this loop shifts posteriorly during regeneration, we next examined whether the region of active cell proliferation also shifts posteriorly. To assess this, we compared the distribution of proliferative cells between EDS and MDS blastemas (Figure 4). In the same sections, Col2 expression, BrdU incorporation, and Hoechst staining were visualized (Figures 4A–4C; 4G–4I). The Col2 expression domains were computationally reconstructed into a three-dimensional (3D) image based on the image series obtained from these sections (Figures 4M and 4P), and three slices around the proximal region were extracted from the reconstructed data (Figures 4D and 4J). Heatmaps of BrdU+ cell density were generated (Figures 4E–4K; 4N and 4Q) and overlaid onto the reconstructed Col2 expression images (Figures 4F, 4L, 4O, and 4R). At EDS, BrdU+ cell density was high in both the posterior and anterior regions of the Col2+ (Figures 4A–4F; 4M–4O). However, at MDS, BrdU+ signals were more densely distributed in the posterior region compared with the anterior region (Figures 4G–4L; 4P–4R). To quantify this, we measured the ratio of BrdU+ cells to the total number of nuclei in the posterior or anterior region of the Col2+. The results showed a significantly higher BrdU+ cell ratio in the posterior region than in the anterior region at MDS, whereas no significant difference was observed at EDS (Figures 4S and 4T). These findings indicate that cell proliferation activity becomes biased toward the posterior region at MDS, consistent with the posterior shift of Shh/Fgf feedback loop activity.Figure 4Posterior shift of the proliferative region in relation to the Shh/Fgf feedback loop(A–C) Col2 expression, BrdU incorporation, and nuclei were visualized on identical sections of EDS blastemas. Enlarged views of the boxed regions in (B) and (C) are shown in (B′), (B″), (C′), (C″). The dotted lines indicate the boundary of the blastema epithelium.(D) A region encompassing the section shown in (A–C), including one distal and one proximal section, was extracted from the 3D reconstructed image in (M) and displayed in distal view. Magenta indicates Col2 expression.(E) BrdU+ cells from the same three consecutive sections as in (D) were extracted and color-coded according to their relative density.(F) Merged image of (D), (E).(G–I) Col2 expression, BrdU incorporation, and nuclei were visualized on identical sections of MDS blastemas. Enlarged views of the boxed regions in (H) and (I) are shown in (H′), (H″), (I′), and (I″). The dotted lines indicate the boundary of the blastema epithelium.(J) A region encompassing the section shown in (G–I), including one distal and one proximal section, was extracted from the 3D reconstructed image in (P) and displayed in distal view. Magenta indicates Col2 expression.(K) BrdU+ cells from the same three consecutive sections as in (J) were extracted and color-coded according to their relative density.(L) Merged image of (J), (K).(M and N) 3D reconstructed images of Col2 expression (M) and color-coded BrdU+ cells (N) in EDS blastemas.(O) Merged image of (M), (N).(P and Q) 3D reconstructed images of Col2 expression (P) and color-coded BrdU+ cells (Q) in MDS blastemas.(R) Merged image of (P), (Q). Scale bars, 200 μm (A, D, G, J, and M); 100 μm (B′, B″, H′, and H″).(S and T) Comparison of the ratio of BrdU+ cells in the anterior and posterior regions of the Col2+ domain in EDS (S) and MDS (T) blastemas. Statistical analysis was performed using a two-tailed Student’s t-test. Data are presented as box-and-whisker plots. Sample sizes (biological replicates): EDS anterior, n = 8; EDS posterior, n = 8; MDS anterior, n = 9; MDS posterior, n = 9. Posterior shift of the proliferative region in relation to the Shh/Fgf feedback loop (A–C) Col2 expression, BrdU incorporation, and nuclei were visualized on identical sections of EDS blastemas. Enlarged views of the boxed regions in (B) and (C) are shown in (B′), (B″), (C′), (C″). The dotted lines indicate the boundary of the blastema epithelium. (D) A region encompassing the section shown in (A–C), including one distal and one proximal section, was extracted from the 3D reconstructed image in (M) and displayed in distal view. Magenta indicates Col2 expression. (E) BrdU+ cells from the same three consecutive sections as in (D) were extracted and color-coded according to their relative density. (F) Merged image of (D), (E). (G–I) Col2 expression, BrdU incorporation, and nuclei were visualized on identical sections of MDS blastemas. Enlarged views of the boxed regions in (H) and (I) are shown in (H′), (H″), (I′), and (I″). The dotted lines indicate the boundary of the blastema epithelium. (J) A region encompassing the section shown in (G–I), including one distal and one proximal section, was extracted from the 3D reconstructed image in (P) and displayed in distal view. Magenta indicates Col2 expression. (K) BrdU+ cells from the same three consecutive sections as in (J) were extracted and color-coded according to their relative density. (L) Merged image of (J), (K). (M and N) 3D reconstructed images of Col2 expression (M) and color-coded BrdU+ cells (N) in EDS blastemas. (O) Merged image of (M), (N). (P and Q) 3D reconstructed images of Col2 expression (P) and color-coded BrdU+ cells (Q) in MDS blastemas. (R) Merged image of (P), (Q). Scale bars, 200 μm (A, D, G, J, and M); 100 μm (B′, B″, H′, and H″). (S and T) Comparison of the ratio of BrdU+ cells in the anterior and posterior regions of the Col2+ domain in EDS (S) and MDS (T) blastemas. Statistical analysis was performed using a two-tailed Student’s t-test. Data are presented as box-and-whisker plots. Sample sizes (biological replicates): EDS anterior, n = 8; EDS posterior, n = 8; MDS anterior, n = 9; MDS posterior, n = 9. Posterior digit induction by the posteriorly shifted Shh/Fgf feedback loop activity We subsequently examined whether the formation of posterior digits (III–IV in the forelimb and III–V in the hindlimb) depends on the activity of the posteriorly shifted Shh/Fgf feedback loop, by inhibiting this signaling loop with either the Hh signaling inhibitor cyclopamine or the Fgf signaling inhibitor SU5402. First, to confirm the effectiveness of these inhibitors, animals were treated with them, and gene expression levels in their blastemas were quantified (Figure 5A). Consistent with previous studies, the expression levels of both Shh and Fgf8 were significantly reduced by treatment with either cyclopamine or SU5402 (Figure 5A, A’).13,25 These results indicate that cyclopamine and SU5402 each effectively inhibit the Shh/Fgf8 feedback loop in axolotl blastemas. However, because SU5402 is a pan-FGFR inhibitor and may also block signaling mediated by FGF ligands other than FGF8, we used cyclopamine to inhibit this feedback loop at different time points during limb regeneration (Figures 5B–5F). When axolotls were treated with cyclopamine immediately after limb amputation, all samples developed spike-like structures, as previously reported, whereas the DMSO-treated controls exhibited normal limb regeneration (Figures 5C–5F). In limbs treated with cyclopamine from the EDS, at least two digits, presumably corresponding to digits I and II, formed, suggesting that these digits are already specified by the EDS and can continue to develop independently of the Shh/Fgf feedback loop. However, more than half of the forelimb samples and all hindlimb samples treated with cyclopamine from the EDS exhibited digit loss, suggesting that the role of the Shh/Fgf loop in digit specification had not yet been completed at this stage. Consistent with this, delaying cyclopamine treatment until the MDS markedly reduced the severity of the defect (Figures 5C–5F), suggesting that the Shh/Fgf feedback loop remains critical for digit induction after the EDS. When comparing forelimbs and hindlimbs, the number of digits formed was comparable under the same treatment conditions; however, because WT hindlimbs normally have one additional digit, the loss was effectively greater in hindlimbs. When animals were treated with SU5402 in the same manner, limb regeneration appeared to be more severely affected, with not only AP digit formation but also PD blastema elongation seemingly impaired (Figure S2). This broader phenotype is likely attributable to the fact that SU5402 is a pan-FGFR inhibitor and thus suppresses signaling mediated by FGF ligands other than FGF8. Indeed, a previous study reported that Fgf10 is expressed in the axolotl blastema mesenchyme.29 Therefore, to specifically examine the effects of disrupting the Shh/Fgf feedback loop, we used cyclopamine alone in the subsequent experiments. In summary, these findings suggest that the Shh/Fgf signaling continues to play an essential role in inducing posterior digits after EDS.Figure 5Posterior digit induction by the posteriorly shifted Shh/Fgf feedback loop activity(A) Relative expression levels of Shh (A) and Fgf8 (A′) in blastemas treated with DMSO, cyclopamine, or SU5402 were measured by qRT-PCR. Statistical analysis was performed on biological replicates (n = 3–4 per condition) using one-way ANOVA followed by Tukey’s post hoc test. Data are represented as mean ± SEM. Each data point represents a technical replicate, and consistent trends were observed across independent biological replicates.(B) Experimental procedure for stepwise cyclopamine treatment.(C) Alcian blue-stained images of regenerated forelimbs under each treatment condition.(D) Graph showing the number of digits formed in regenerated fore limbs under each condition.(E) Alcian blue-stained images of regenerated hindlimbs under each treatment condition.(F) Graph showing the number of digits formed in regenerated hindlimbs under each condition. Scale bars, 1 mm (C and E).(G and H) Animals were treated with DMSO (G) or cyclopamine (H) from the early digit stage (EDS) to the middle digit stage (MDS) and fixed at MDS. Col2 and Shh expressions were visualized by ISH. BrdU incorporation was detected by immunofluorescence, and nuclei were counterstained with Hoechst. Enlarged views of the boxed regions are shown in the right panels. Dotted lines indicate the boundary of the blastema epithelium. Scale bars, 200 μm (G and H).(I and J) Comparison of the ratio of BrdU+ cells in blastemas treated with DMSO or cyclopamine in the posterior (I) and anterior (J) regions of the Col2+ domain. Statistical analysis was performed using a two-tailed Student’s t-test. Data are presented as box-and-whisker plots. Sample sizes (biological replicates): DMSO posterior, n = 7; cyclopamine posterior, n = 8; DMSO anterior, n = 7; cyclopamine anterior, n = 8.(K) Immunofluorescence images of caspase3 staining. Scale bars, 200 μm.(L) Ratio of the Caspase3+ area to the total mesenchymal area in blastema sections. Statistical analysis was performed using a two-tailed Student’s t-test. Data are presented as box-and-whisker plots. Sample sizes (biological replicates): DMSO, n = 5; cyclopamine, n = 6. Posterior digit induction by the posteriorly shifted Shh/Fgf feedback loop activity (A) Relative expression levels of Shh (A) and Fgf8 (A′) in blastemas treated with DMSO, cyclopamine, or SU5402 were measured by qRT-PCR. Statistical analysis was performed on biological replicates (n = 3–4 per condition) using one-way ANOVA followed by Tukey’s post hoc test. Data are represented as mean ± SEM. Each data point represents a technical replicate, and consistent trends were observed across independent biological replicates. (B) Experimental procedure for stepwise cyclopamine treatment. (C) Alcian blue-stained images of regenerated forelimbs under each treatment condition. (D) Graph showing the number of digits formed in regenerated fore limbs under each condition. (E) Alcian blue-stained images of regenerated hindlimbs under each treatment condition. (F) Graph showing the number of digits formed in regenerated hindlimbs under each condition. Scale bars, 1 mm (C and E). (G and H) Animals were treated with DMSO (G) or cyclopamine (H) from the early digit stage (EDS) to the middle digit stage (MDS) and fixed at MDS. Col2 and Shh expressions were visualized by ISH. BrdU incorporation was detected by immunofluorescence, and nuclei were counterstained with Hoechst. Enlarged views of the boxed regions are shown in the right panels. Dotted lines indicate the boundary of the blastema epithelium. Scale bars, 200 μm (G and H). (I and J) Comparison of the ratio of BrdU+ cells in blastemas treated with DMSO or cyclopamine in the posterior (I) and anterior (J) regions of the Col2+ domain. Statistical analysis was performed using a two-tailed Student’s t-test. Data are presented as box-and-whisker plots. Sample sizes (biological replicates): DMSO posterior, n = 7; cyclopamine posterior, n = 8; DMSO anterior, n = 7; cyclopamine anterior, n = 8. (K) Immunofluorescence images of caspase3 staining. Scale bars, 200 μm. (L) Ratio of the Caspase3+ area to the total mesenchymal area in blastema sections. Statistical analysis was performed using a two-tailed Student’s t-test. Data are presented as box-and-whisker plots. Sample sizes (biological replicates): DMSO, n = 5; cyclopamine, n = 6. We further examined the effect of Shh/Fgf feedback loop inhibition on cell proliferation in the posterior region, where high proliferative activity is normally observed in the MDS blastema (Figure 4). Axolotls were treated with cyclopamine from the EDS to the MDS, and cell proliferation activity was analyzed (Figures 5G–5J). The ratio of BrdU+ cells in the posterior region of the Col2+ was significantly decreased compared with DMSO-treated controls, whereas no significant difference was observed in the anterior region (Figures 5I and 5J). This indicates that the Shh/Fgf signaling loop primarily promotes cell proliferation in the posterior region of the Col2+ at MDS, consistent with the posterior shift of Shh/Fgf signaling activity. In amniote limb development, Shh signaling has been suggested to inhibit programmed cell death in the mesenchymal region, influencing digit morphology.30 To examine whether the inhibition of the Shh/Fgf signaling affects apoptosis in axolotl blastemas, we analyzed Caspase 3 signals (Figures 5K and 5L). In both DMSO- and cyclopamine-treated blastemas, Caspase 3 signals were barely detectable with no significant difference between treatments. Similar results were obtained using TUNEL assays, which showed no significant difference between treatments (Figure S3). These findings suggest that the posterior shifted Shh/Fgf signaling activity continues inducing digit formation even after EDS by promoting cell proliferation in the posterior region. Extra digit formation by artificially induced Shh/Fgf8 expression Our results thus far have shown that the posteriorly shifted Shh/Fgf signaling loop induces posterior digit formation, and that the early termination of its activity by cyclopamine reduces the number of digits. These findings led us to hypothesize that extending Shh/Fgf signaling activity could result in extra digit formation. To test this hypothesis, we artificially extended the duration of Shh and/or Fgf8 expression by electroporation at MDS (Figure 6). We overexpressed Gfp, Shh+Fgf8, Shh, or Fgf8 in the posterior region of MDS blastemas (Figures 6A–6C). In all conditions, GFP fluorescence persisted at 10 days post-electroporation (dpe) (25 days post-amputation (dpa)), when posterior digit differentiation is underway, confirming sustained transgene expression during the digit formation stage (Figure S4). When the pCS2-AcGFP vector was electroporated, no extra digit was formed, although the one-digit loss occurred in some samples (N = 3/15) (Figures 6D and 6J). However, ectopic expression of both Shh and Fgf8 in the posterior region at MDS induced extra digit(s) in both forelimbs and hindlimbs (N = 5/16) (Figure 6E; 6H–6J). Additionally, Fgf8 electroporation alone also resulted in extra digit formation (N = 5/20), whereas ectopic Shh expression alone did not (N = 0/16) (Figures 6F, 6G, and 6J). This result aligns with our previous findings, which suggested that Fgf8 expression declines earlier than Shh during digit formation (Figures 1P and 1Q). It is likely that ectopic Fgf8 expression prolongs Shh/Fgf8 signaling loop activity by compensating for the reduction in endogenous Fgf8 expression. Together, the electroporation and inhibitor experiments (Figure 5) suggest that the duration of Shh/Fgf signaling loop activity is closely linked to the number of digits formed.Figure 6Extra digit formation by artificially induced Shh/Fgf8 expression in MDS blastemas(A–C) Bright-field (A), GFP fluorescence (B), and merged (C) stereomicroscopic images of blastemas at 3 days post-electroporation (dpe). Gene overexpression was induced in the posterior region of the MDS blastema (at 15 days post amputation (dpa)) using electroporation. Scale bars, 1 mm.(D–G) Time course images of hindlimbs overexpressing Gfp (D), Shh + Fgf8 (E), Fgf8 (F), or Shh (G). The upper side of each image corresponds to the anterior, and the lower side to the posterior. Filled white arrowheads indicate the formation of an extra digit. Scale bars, 2 mm (D).(H and I) Skeletal double-staining images of fully regenerated forelimbs (H) and hindlimbs (I) after co-overexpression of Shh and Fgf8. Black arrowheads indicate an extra digit. Scale bars, 2 mm.(J) Graph showing the proportion of regenerated limbs exhibiting each digit phenotype under the indicated electroporation conditions. Extra digit formation by artificially induced Shh/Fgf8 expression in MDS blastemas (A–C) Bright-field (A), GFP fluorescence (B), and merged (C) stereomicroscopic images of blastemas at 3 days post-electroporation (dpe). Gene overexpression was induced in the posterior region of the MDS blastema (at 15 days post amputation (dpa)) using electroporation. Scale bars, 1 mm. (D–G) Time course images of hindlimbs overexpressing Gfp (D), Shh + Fgf8 (E), Fgf8 (F), or Shh (G). The upper side of each image corresponds to the anterior, and the lower side to the posterior. Filled white arrowheads indicate the formation of an extra digit. Scale bars, 2 mm (D). (H and I) Skeletal double-staining images of fully regenerated forelimbs (H) and hindlimbs (I) after co-overexpression of Shh and Fgf8. Black arrowheads indicate an extra digit. Scale bars, 2 mm. (J) Graph showing the proportion of regenerated limbs exhibiting each digit phenotype under the indicated electroporation conditions. Finally, we attempted to induce extra digits in the intact limbs. If the establishment of the Shh/Fgf feedback loop is sufficient to induce digit formation, it is considered that ectopic Shh and/or Fgf8 expression can induce extra digits in the intact limbs. Thus, we electroporated Gfp, Shh+Fgf8, Shh, or Fgf8, as shown in Figure 6, in the posterior region of the intact limbs. To promote the regeneration/morphogenesis process, we removed the intact skin 3 days after the electroporation (Figure 7A). To maintain the expression in the long term, we performed additional electroporation 13 days post-first-electroporation (dpe) (Figure 7A). The electroporation successfully delivered exogenous gene expression locally (Figure 7B). When Gfp expression was introduced into the posterior region, neither bud nor extra-digit formation was observed, except in one case (Figures 7C left, 7D, and 7J; N = 23/24). When Fgf8 was introduced, some limbs formed a bud, but its growth was not sustained (Figures 7E and 7J; N = 5/16). Nevertheless, digit formation was observed in a few cases (Figure 7J; N = 4/16). For Shh electroporation, most limbs did not form any extra digits (Figures 7F and 7J; N = 15/16). However, when Shh+Fgf8 were co-electroporated, a bud was typically observed by 13 dpe (Figure 7C right), and most of these buds developed into extra digits (Figures 7G and 7J; N = 10/16). We also found that additional Fgf2 expression enhanced extra-digit induction in intact limbs. When Shh, Fgf2, and Fgf8 were introduced together into the posterior region, one to three extra digits were induced (Figures 7H–7J; N = 14/16). Given the mitogenic role of FGF2, its additional expression likely contributed to creating more space for digit formation. Lastly, we note that the extra-digit induction could be induced not only in the posterior region, but also in any place within the autopodial region (Figure S5). The exogenous activities of Shh- and Fgf-signaling in the anterior or inter-digital region gave rise to the extra digit formation. These findings suggest that Shh/Fgf signaling loop activity is sufficient to induce digit formation.Figure 7Extra digit formation by artificially induced Shh/Fgf8 expression in intact limbs(A) Experimental procedure.(B) Dorsal view of an electroporated forelimb at 3 days post-electroporation (before skin wounding). GFP signals were detected in the posterior region of the limb.(C) Time course images showing extra digit formation in forelimbs. The left panels show control limbs overexpressing Gfp, and the right panels show limbs overexpressing Shh and Fgf8.(D–H) Representative digit phenotypes of electroporated hindlimbs. An arrowhead in (E) indicates a transient bud formation that failed to continue growth, and an arrow in (G) indicates an induced extra digit. (H) Shh + Fgf2 + Fgf8 co-electroporated limb, showing multiple extra digits in the posterior region (arrows).(I) Skeletal double-staining image of the sample shown in (H). Scale bars, 2 mm (B, C, D, and H).(J) Graph shows the proportion of regenerated limbs exhibiting each digit phenotype under the indicated electroporation conditions. Extra digit formation by artificially induced Shh/Fgf8 expression in intact limbs (A) Experimental procedure. (B) Dorsal view of an electroporated forelimb at 3 days post-electroporation (before skin wounding). GFP signals were detected in the posterior region of the limb. (C) Time course images showing extra digit formation in forelimbs. The left panels show control limbs overexpressing Gfp, and the right panels show limbs overexpressing Shh and Fgf8. (D–H) Representative digit phenotypes of electroporated hindlimbs. An arrowhead in (E) indicates a transient bud formation that failed to continue growth, and an arrow in (G) indicates an induced extra digit. (H) Shh + Fgf2 + Fgf8 co-electroporated limb, showing multiple extra digits in the posterior region (arrows). (I) Skeletal double-staining image of the sample shown in (H). Scale bars, 2 mm (B, C, D, and H). (J) Graph shows the proportion of regenerated limbs exhibiting each digit phenotype under the indicated electroporation conditions. Discussion This study suggests that posteriorly shifting Shh/Fgf feedback loop activity drives digit formation in an anterior-to-posterior sequence during axolotl limb regeneration. We clarified that the position of the Shh/Fgf feedback loop activity relative to differentiating digit II shifts posteriorly in the period between the EDS and the MDS (Figure 1). Although the activity of this loop gradually declines, it continues to promote cell proliferation and digit differentiation beyond the EDS (Figures 4 and 5). Cyclopamine treatment further suggests that digit occurs sequentially rather than simultaneously: Digits II and I are specified early at the EDS and can proceed independently of Shh/Fgf activity, whereas digits III through V require its continued function (Figures 5B–5F). Integrating these findings with our previous work,25 we propose a model in which the attenuating and posteriorly shifting Shh/Fgf feedback loop sequentially induces posterior digits until its activity falls below the threshold required for further digit formation. Consistent with this model, extending the Shh/Fgf activity by exogenous Shh and Fgf8 expression at the MDS results in the formation of extra digit(s). Thus, we propose that our digit induction model provides a molecular mechanistic basis for the reversed, preaxial dominance of digit formation in urodeles. Previous studies have suggested that evolutionary shifts in limb polarity are associated with changes in the balance between GLI3-mediated repression and 5′ Hox gene activity, which biases digit-forming potential toward either the preaxial or postaxial side.31 While these findings point to an important role for the GLI3–HOX regulatory system in establishing limb polarity, how this system relates to the dynamically shifting Shh/Fgf feedback loop described here remains unclear. Resolving these relationships will require further investigation and is an important direction for future studies. Although this digit induction model is based on findings from limb regeneration, several results are consistent with previously reported observations during axolotl limb development. At developmental stage 49, when digits II and I are marked by Sox9 expression in the forelimb, Shh expression persists, while Fgf8 transcripts are undetectable by ISH but detectable by qRT-PCR and are present in both anterior and posterior compartments, paralleling our findings at the MDS (Figure 1).13 Moreover, stage-dependent inhibition of Shh signaling during limb development produces digit patterns consistent with those observed during regeneration (Figures 5B–5F).32 Thus, we propose that our digit induction model, providing a mechanistic basis for the reversed, preaxial dominance of digit formation in urodeles, can be applied to both limb development and regeneration processes. We also suggest that cartilage forming within the blastema is a key factor driving the posterior shift of Shh/Fgf activity. In our previous work, we showed that phalangeal differentiation begins in the region between Shh and Fgf8 expression domains,25 making it reasonable that nascent phalanges influence SHH diffusion. Two non-exclusive mechanisms may account for this effect by nascent phalanges: (1) a physical barrier to diffusion, and (2) adsorption of SHH by extracellular matrix (ECM) associated with differentiating cartilage. Because cartilage accumulates abundant ECM, this acellular matrix could impede molecular diffusion; moreover, ECM is well known to sequester specific secreted factors, including during limb development and regeneration.33,34,35,36 SHH binding to heparan sulfate proteoglycans (HSPGs) is also well established, and cartilage contains abundant HSPGs, making SHH adsorption by differentiating cartilage a reasonable expectation.37 Consistent with this idea, grafting tissue presumed to encompass the nascent phalangeal region induced a posterior shift of Fgf8 expression domains (Figure 2). This was not a transplant of fully formed cartilage, and the graft likely contained non-cartilaginous differentiating tissues. Pure cartilage could not be isolated at this stage because Col2 activity is relatively broad, and fully differentiated phalanges were unsuitable due to divergent properties (e.g., perichondrium differentiation). We acknowledge the technical limitations in graft characterization. Nonetheless, simulations indicate that either physical barrier or adsorption is sufficient to shorten the effective diffusion range of SHH, which likely accounts for the posterior displacement of the SHH-dependent Fgf8 expression domain and, in turn, supports the view that emergent obstacles posteriorize the Shh/Fgf8 feedback loop activity. Our study suggests that, although the spatial deployment of the Shh/Fgf8 feedback loop in urodeles differs markedly from that in amniotes, its core role in digit induction is conserved. In amniotes, SHH produced by the ZPA and FGF8 produced by the AER form a positive feedback loop, establishing epithelial-mesenchymal interactions. On the other hand, in urodeles, the Shh/Fgf8 feedback loop is established within mesenchymal tissues. In this study, we revealed that this mesenchymal Shh/Fgf feedback loop activity dynamically shifts posteriorly as chondrogenic differentiation proceeds. Despite this distinct spatial deployment, its core function appears conserved: shortening its duration caused digit loss, whereas prolonging it led to extra digit formation (Figures 5 and 6). Similar results have been reported in amniote limb development, where reduced Shh signaling exposure results in fewer digits, while prolonged Shh/Fgf signaling results in polydactyly.30,38,39,40 Together, these findings highlight both the conserved mechanisms of digit formation and an alternative mode of operation. A network centered on Shh/Fgf functions in both amniotes and urodeles, implying the use of shared or closely related genomic modules. Yet differences in their spatial deployment and expression patterns suggest that the “use” of modules within the limb field varies across lineages. Why amniotes adopted an AER-dependent mode, and what selective pressures favor it over non-AER configurations, remain unresolved. Future studies integrating comparative analyses of digit formation across vertebrates with investigations of cis-regulatory evolution will be essential to clarify how the spatial and temporal deployment of Shh/Fgf signaling has been shaped during the evolution of limb development and regeneration. Historically, studies of morphogenesis—including limb development —have emphasized a single route to a target form. The divergence of axolotl limb morphogenesis from that of amniotes provides a clear example of multi-route attainability. More broadly, the convergence of morphology during the phylotypic period supports the idea that identical or closely similar end forms can be achieved through multiple developmental trajectories, even when relying on the same molecular modules. These genome-encoded modules may therefore be more plastic than expected, allowing conserved modules to be redeployed in alternative ways to produce equivalent morphologies. Such a concept can be explained by the developmental system drift.41 In this context, the significance of our work lies in providing concrete evidence that an “alternative” spatial deployment of a conserved genetic toolkit can nevertheless converge on the same final morphology. Limitations of the study This study has some limitations. Although we focused on posteriorly shifting Fgf signaling during digit induction, we could not definitively identify the responsible FGF ligand. In some assays, Fgf8 expression was too weak to detect, and we therefore relied on Spry1 as a proxy for Fgf signaling activity. Consequently, our conclusions are limited to the role of posteriorly shifting Fgf signaling, rather than ligand-specific effects of Fgf8. In addition, despite the difference in final digit number between axolotl forelimbs and hindlimbs, we detected no significant differences in gene expression at either the EDS or MDS, nor in digit number following stepwise cyclopamine treatment under identical conditions. While this may indicate that digit induction mechanisms are largely shared between forelimbs and hindlimbs, with potential differences in SHH/FGF8 threshold sensitivity, the present data do not allow firm conclusions regarding the basis of fore-hind digit number differences. Furthermore, the present analyses were intentionally focused on the Shh/Fgf signaling loop. However, limb patterning is governed by a complex interplay of multiple regulatory systems, such as Hox gene networks, WNT, BMP, and other signaling pathways, GLI-SHH gradients, and Turing-like patterning models, among others. Future studies examining how these pathways interact with the posteriorly shifting Shh/Fgf feedback loop identified here, and how such interactions are coordinated during regeneration, will be essential for achieving a more comprehensive understanding of axolotl-specific mechanisms of limb patterning. Despite these limitations, our results support a model in which a posteriorly shifting Shh/Fgf feedback loop sequentially induces posterior digits, with the duration of its activity closely linked to the number of digits formed. Resource availability Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Akira Satoh (satoha@cc.okayama-u.ac.jp), Okayama University, Graduate School of Environmental, Life, Natural Science and Technology, Okayama, Japan. Materials availability This study did not generate new unique reagents. Data and code availability •This study did not generate new unique datasets.•The Python codes were shared as the supplemental data and will be shared by the lead contact upon request.•Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. •This study did not generate new unique datasets. This study did not generate new unique datasets. •The Python codes were shared as the supplemental data and will be shared by the lead contact upon request. The Python codes were shared as the supplemental data and will be shared by the lead contact upon request. •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Acknowledgments We are grateful to Ms. R. Iwata and T. Satoh for supporting office work and animal housing. We also thank Dr. M. Nagayama (Hokkaido Univ.) for giving us kind consultation on the simulation. This work was supported by the following funding sources: JSPS Kakenhi (A.S.: #24K02034, #24H01953), Grant-in-Aid for 10.13039/501100001691Japan Society for the Promotion of Science fellows (#23KJ1611 to S.F.), and 10.13039/100007449Takeda Science Foundation (# 2025022483 to A.S). Author contributions S.F. and A.S. contributed to building the concept of the experiment. Methodology was developed and established by S.F. and A.S. All data validation was performed by S.F., A.O., S.Y., and A.S. S.F. A.O., S.Y., and H.N. contributed to animal housing. Simulation was conducted by A.S. and A.O. The original draft was written by A.S. and S.F. Then, all authors (S.F., A.O., H.N., S.Y., and A.S.) contributed to reviewing and editing the manuscript. A.S. and S.F. played a role in the funding acquisition. Declaration of interests We have no competing interests. Declaration of generative AI and AI-assisted technologies in the writing process During the preparation of this work, we used ChatGPT3.5 (Thinking) in order to develop the python codes. After using this tool or service, we reviewed and edited the content as needed and take(s) full responsibility for the content of the publication. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesanti-BrdU (G3G4)Developmental Studies Hybridoma BankCat# G3G4; RRID: AB_2618097Anti-GFP (Green Fluorescent Protein) pAbMBLCat# 598; AB_591819Cleaved Caspase-3 (Asp175) AntibodyCell Signaling TechnologyCat# 9661; RRID: AB_2341188F(ab')2-Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488Thermo Fisher ScientificCat# A11017; RRID: AB_2534084Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488Thermo Fisher ScientificCat# A21206; RRID: AB_2535792Sheep Anti-Digoxigenin Fab fragments Antibody, AP ConjugatedRocheCat#11093274910; RRID: AB_514497BiologicalsamplesMiddle bud stage axolotl limb blastemaThis studyN/ALate bud stage axolotl limb blastemaThis studyN/AEarly Digit Stage axolotl limb blastemaThis studyN/AMiddle Digit Stage axolotl limb blastemaThis studyN/AAxolotl intact limbThis studyN/AChemicals,peptides, andrecombinantproteinsCyclopamineLKT LaboratoriesCat# C9710SU5402FUJIFILM Wako Pure Chemical CorporationCat# 193-16733Dimethyl sulfoxide (DMSO)Nacalai TesqueCat# 08904-14CriticalcommercialassaysTriPure Isolation ReagentRocheCat# 11 667 157 001PrimeScript II Reverse TranscriptaseTakaraCat# 2690AKAPA SYBR FAST qPCR Master MixKapa BiosystemsCat# KK4605DepositeddataPython code for 3D heatmap generationThis studyData S1Python code for simulating SHH and FGF8 dynamicsThis studyData S2Experimentalmodels:Organisms/strainsAmbystoma mexicanum: wildtypeAmphibian Research Center (Hiroshima University)N/AAmbystoma mexicanum: GFPAmphibian Research Center (Hiroshima University)N/AOligonucleotidesShh RNA probeIwata et al.,42N/AFgf8 RNA probeMakanae et al.,43N/AHoxA13 RNA probeSatoh et al.,44N/ASpry1 RNA probeThis studyN/ACol2 RNA probeMitogawa et al.,45N/ARecombinantDNApCS2-AcGFPFurukawa et al.,25N/ApCS2-Shh-p2A-AcGFPFurukawa et al.,25N/ApCS2-Fgf8This studyN/ApCS2-Fgf2-p2A-Fgf8-p2A-mCherryThis studyN/ASoftware andalgorithmsAmira SoftwareThermo Fisher Scientifichttp://www.fei.com/software/amira-3d-for-life-sciences/ImageJ (Fiji)Fijihttps://imagej.net/software/fiji/Olympus cellSens SoftwareOlympushttp://www.olympus-lifescience.com/en/software/cellsens/PythonPython Software FoundationVersion 3.13 REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesanti-BrdU (G3G4)Developmental Studies Hybridoma BankCat# G3G4; RRID: AB_2618097Anti-GFP (Green Fluorescent Protein) pAbMBLCat# 598; AB_591819Cleaved Caspase-3 (Asp175) AntibodyCell Signaling TechnologyCat# 9661; RRID: AB_2341188F(ab')2-Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488Thermo Fisher ScientificCat# A11017; RRID: AB_2534084Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488Thermo Fisher ScientificCat# A21206; RRID: AB_2535792Sheep Anti-Digoxigenin Fab fragments Antibody, AP ConjugatedRocheCat#11093274910; RRID: AB_514497BiologicalsamplesMiddle bud stage axolotl limb blastemaThis studyN/ALate bud stage axolotl limb blastemaThis studyN/AEarly Digit Stage axolotl limb blastemaThis studyN/AMiddle Digit Stage axolotl limb blastemaThis studyN/AAxolotl intact limbThis studyN/AChemicals,peptides, andrecombinantproteinsCyclopamineLKT LaboratoriesCat# C9710SU5402FUJIFILM Wako Pure Chemical CorporationCat# 193-16733Dimethyl sulfoxide (DMSO)Nacalai TesqueCat# 08904-14CriticalcommercialassaysTriPure Isolation ReagentRocheCat# 11 667 157 001PrimeScript II Reverse TranscriptaseTakaraCat# 2690AKAPA SYBR FAST qPCR Master MixKapa BiosystemsCat# KK4605DepositeddataPython code for 3D heatmap generationThis studyData S1Python code for simulating SHH and FGF8 dynamicsThis studyData S2Experimentalmodels:Organisms/strainsAmbystoma mexicanum: wildtypeAmphibian Research Center (Hiroshima University)N/AAmbystoma mexicanum: GFPAmphibian Research Center (Hiroshima University)N/AOligonucleotidesShh RNA probeIwata et al.,42N/AFgf8 RNA probeMakanae et al.,43N/AHoxA13 RNA probeSatoh et al.,44N/ASpry1 RNA probeThis studyN/ACol2 RNA probeMitogawa et al.,45N/ARecombinantDNApCS2-AcGFPFurukawa et al.,25N/ApCS2-Shh-p2A-AcGFPFurukawa et al.,25N/ApCS2-Fgf8This studyN/ApCS2-Fgf2-p2A-Fgf8-p2A-mCherryThis studyN/ASoftware andalgorithmsAmira SoftwareThermo Fisher Scientifichttp://www.fei.com/software/amira-3d-for-life-sciences/ImageJ (Fiji)Fijihttps://imagej.net/software/fiji/Olympus cellSens SoftwareOlympushttp://www.olympus-lifescience.com/en/software/cellsens/PythonPython Software FoundationVersion 3.13 Experimental model and study participant details Axolotls (Ambystoma mexicanum) were procured from the Hiroshima University Amphibian Research Center and kept in aerated water at 22°C. The axolotls were housed in aerated water at 22°C. No information about animal sex was recorded. Before all surgeries, axolotls were anesthetized using MS-222 (Sigma-Aldrich). Limbs were amputated at the middle of the zeugopod. We defined the late-bud stage blastema as previously reported.25 The early digit stage (EDS) blastema is defined as the stage at which a single cartilage condensation, marked by Type II collagen (Col2) expression, was detected in the distal region. The middle digit stage (MDS) blastema is defined as the stage in which two distinct cartilage condensations were observed on the distal section. All animal use was approved by the Animal Care and Use Committee, Okayama University (license #580 to A.S), and all animal experiments were conducted following the guidelines of Okayama University. Method details Sectioning, ISH, and immunofluorescence Samples were fixed with 4% paraformaldehyde for 1 day at room temperature. After a 30% sucrose/PBS treatment for ∼12 h, they were then embedded in OCT compound (Sakura Finetek). Frozen sections with a thickness of 14 μm were prepared using a Leica CM1850 cryostat. The sections were dried under an air dryer and kept at −80°C until use. For probe synthesis, target genes cloned on pTAC-2 plasmid (BioDynamics Lab Inc.) were amplified by PCR using M13 primers. PCR fragments were purified and used as an RNA probe template. Probe synthesis was performed by T7 or SP6 RNA polymerase (Takara). ISH was performed as previously reported.42 For ISH and BrdU double staining, ISH was performed first, and immunostaining was subsequently performed. Immunofluorescence for GFP and Caspase3 was performed on the adjacent sections, thereby avoiding protein degradation by the ISH procedure. Immunofluorescence of the sections was performed as described in previous study.42 The antibodies were as follows: anti-BrdU (G3G4, 1/300, DSHB), anti-GFP (MBL, #598, 1/400), anti-Cleaved Caspase3 (CST, #9661, 1/400), anti-mouse IgG Alexa 488 (Invitrogen, #A11017, 1/1000), anti-rabbit IgG Alexa 488 (Invitrogen, #A21206, 1/500). BrdU antigen retrieval was performed with 2 mol/L HCl for 30 min at room temperature. Antigen retrieval for Caspase3 was carried out in citrate buffer for 30 min at 100°C. Nuclei were visualized using Hoechst33342 (346–07951, Wako-Fuji film) after performing ISH and immunostaining. Images were captured using an Olympus BX61 fluorescence microscope. qRT-PCR analysis RNA preparation for RT-PCR was performed using TriPure Isolation Reagent (Roche). Reverse transcription was performed using PrimeScript II Reverse Transcriptase (Takara). The quantitative RT-PCR analysis was performed by StepOneTM (ThermoFisher) and KAPA SYBR FAST qPCR Master Mix (Kapa Biosystems). Primers are described below.EF1α forward; AACATCGTGGTCATCGGCCATEF1α reverse; GGAGGTGCCAGTGATCATGTTShh forward; TGTAGACCTATATTATTTGCShh reverse; GAAAGTCCCATTCACATCGTFgf8 forward; ATGCAAGCAAAGCATGTGAGFgf8 reverse; CTCTTGCCAATCAGCTTTCC EF1α forward; AACATCGTGGTCATCGGCCAT EF1α forward; AACATCGTGGTCATCGGCCAT EF1α reverse; GGAGGTGCCAGTGATCATGTT EF1α reverse; GGAGGTGCCAGTGATCATGTT Shh forward; TGTAGACCTATATTATTTGC Shh forward; TGTAGACCTATATTATTTGC Shh reverse; GAAAGTCCCATTCACATCGT Shh reverse; GAAAGTCCCATTCACATCGT Fgf8 forward; ATGCAAGCAAAGCATGTGAG Fgf8 forward; ATGCAAGCAAAGCATGTGAG Fgf8 reverse; CTCTTGCCAATCAGCTTTCC Fgf8 reverse; CTCTTGCCAATCAGCTTTCC Total RNA was extracted from the entire harvested blastema. For the intact limb samples (Figure 1), the autopod region was dissected and processed for RNA extraction. For the samples shown in Figure 5A, animals were treated with the indicated chemicals for 3 days before the blastemas were harvested and used for RNA extraction. Blastema mesenchymal tissue grafting Transgenic animals constitutively expressing GFP were used as donors. Donor blastemas were collected at either the middle bud stage (control) or the middle digit stage (MDS), and the epithelium was carefully removed using fine forceps in PBS. For the decellularized condition, the mesenchymal tissues, free of epithelium was treated with 2M Urea solution for 1 day. After washing with PBS, the tissue was used for the graft experiment. The mesenchymal tissue of each condition was then transplanted into the central region of the blastema of a wild-type host at the middle bud stage. A small incision was made in the central part of the host blastema using fine scissors, and the graft was gently inserted with forceps. To promote graft attachment, the operated animals were kept on ice for more than 3 h before being returned to normal housing conditions. Samples were collected and fixed 3 days after transplantation and subsequently used for ISH and immunofluorescence analyses. 3D reconstruction For the 3D reconstruction shown in Figure 4, Col2, BrdU, and Hoechst were stained on the same sections throughout all serial transverse sections of the blastema. The captured images for Col2 expression were reconstructed through volume rendering using Amira Software (version 6.4.0, Thermo Fisher Scientific; http://www.fei.com/software/amira-3d-for-life-sciences/) running on an iMacPro (CPU, 2.3 GHz Intel Xeon W; DRAM, 128 GB 2666 MHz DDR4; and graphics, Radeon Pro Vega 64 16368 MB). Gene expression signals were extracted using the segmentation tool on Amira. The BrdU signal images were converted into 3D TIFF files using Amira software. The signals were then filled using the Color Threshold tool in ImageJ (Fiji) and subsequently binarized with the Threshold tool. Connected signals were separated using the Binary tool, and the coordinate information of each signal was obtained with the Analyze Particles function. The coordinate information was processed using the following Python code to generate a three-dimensional heatmap: The exact python code used to generate the heatmaps was available as in Data S1. Cyclopamine and SU5402 treatments Animals were treated with 10 mL solutions for all chemical treatments. For these experiments, smaller animals (approximately 5 cm in length) were used. Working stocks of 8 mM cyclopamine (#C9710, LKT Labs, Inc.) and 30 mM SU5402 (#193–16733, Wako) were prepared in DMSO (#08904–14, Nacalai Tesque) and diluted to final concentrations of 8 μM and 30 μM in water, respectively. Control animals were treated with a solution containing an equivalent concentration of DMSO. For Figure 5A, animals having blastemas were treated with chemicals for three days and the blastema samples were harvested. For the experiments shown in Figures 5B–5F and Figure S2, cyclopamine or SU5402 treatment was initiated at three different stages: immediately after limb amputation, at the early digit stage (EDS), and at the middle digit stage (MDS). In addition, a control group was continuously immersed in DMSO throughout the entire experimental period. To synchronize the experimental timing among groups, animals that had not yet begun cyclopamine or SU5402 treatment were immersed in DMSO solution until the corresponding cyclopamine or SU5402 treatment started. For Figures 5G–5l, animals bearing blastemas at the EDS were treated with cyclopamine or DMSO and allowed to regenerate until the MDS, after which the blastema samples were collected. The chemical-containing water was refreshed daily for the short-term treatments (Figure 5A) and every two days for the long-term treatments (Figures 5B–5L). TUNEL analysis Samples were fixed with 4% paraformaldehyde for 20 min. DNA strand breaks were labeled using the MEBSTAIN Apoptosis TUNEL Kit Direct (No. 8445; MBL). Counterstaining was performed with Hoechst 33342 (No. 346–07951; Wako Fujifilm). Whole-mount Alcian blue staining Samples were fixed with 10%-Formaldehyde Neutral Buffer Solution (37152–51, Nacalai Tesque) for 1 day. Samples were then stained with Alcian blue solution (Wako, pH 2.0) and dissolved in 70% ethanol/1% HCl for 1 day after replacing the 10% formalin solution with 70% ethanol. The samples were washed with 70% ethanol, then placed in 100% ethanol. Skeletal double staining Samples were fixed in 100% ethanol for 1 day at room temperature, then stained with Alcian blue solution (Wako, pH 2.0) in 20% acetic acid (Nacalai Tesque) with 80% ethanol solution at 37°C overnight. Then, samples were washed in tap water several times and fixed in 10% Formaldehyde Neutral Buffer Solution (Nacalai Tesque) for 1 day. Samples were then stained with Alizarin red S (Nacalai Tesque) in the solution (4% KOH:10% Formaldehyde Neutral Buffer Solution = 2:3) at room temperature for 1 day. Finally, samples were placed in graded glycerol with 4% KOH for clearing. Electroporation The animals were anesthetized as described and the DNA solution was injected directly into the regenerating blastema or under the skin. Then, electroporation was performed on the area where the DNA was injected. The conditions of the electroporation were as follows: 20 V, 20 times, 50 ms pulse, 950 ms interval (NEPA21, Nepa gene). The injected DNA were as follows: pCS2-AcGFP, pCS2-Shh-p2A-AcGFP, pCS2-Fgf8 and pCS2-Fgf2-p2A-Fgf8-p2A-mCherry. Shh-p2A-AcGFP, Fgf8, and Fgf2p2A-Fgf8-p2A-mCherry were created as artificial synthetic genes and subcloned into pCS2 Vector. All plasmids were purified with a Genopure Maxi kit (Roche, Basel, Switzerland) and the final concentration was 2 μg/μL. Simulation We simulated SHH and FGF8 dynamics in a 2D limb-bud–like domain using finite differences to solve ∂tC = D∇2C-γC+Son a 300 × 100 grid (dx = dy = 1; dt = 0.05). The computational domain was an ellipse (semi-axes a = 25, b = 50) centered on the grid. SHH synthesis was restricted to the 100 rightmost interior pixels of the ellipse, and FGF8 synthesis occurred in an AER-like peripheral subset only where SHH exceeded 0.001. A circular, impenetrable obstacle (radius = 10 pixels) was introduced at frame 1,000 and positioned within the FGF8-positive region based on the current FGF8 field (threshold 0.1; xpos = xmax-0.3(xmax-xmin)); obstacle pixels were excluded from PDE updates to impose a no-flux boundary. Visualizations were produced with Matplotlib (continuous fields for SHH/FGF8 and a binary mask for SHH>0.001), and animations were created using FuncAnimation; selected frames (e.g., 990, 1010, 1500) were saved as PNGs. To quantify the anterior extent of SHH, for each image row we identified the first x at CSHH crossed 0.001 (with linear interpolation) and reported the minimum across rows. We performed parameter sweeps varying diffusion D (1.0–5.0 in 0.1 steps), degradation γ (0.10–0.40; 31 values), and synthesis S0(0.1–5.0 in 0.1 steps) for 0–1,600 frames, recording every 10 frames; trajectories of left-edge position vs. time were plotted in Matplotlib. The exact python codes used to generate the figures was available as in Data S2. Quantification and statistical analysis Measurement of signals on sections The measurement of gene expression domains was performed as previously described.25 Each image was rotated to align the AP axis horizontally, and the AP length of the blastema mesenchymal region was measured. The image was then binarized using the threshold tool in ImageJ (Fiji) to extract ISH or GFP immunofluorescent signals. The anterior and posterior coordinates of the binarized signals were obtained, and the center positions were calculated from these data. The Caspase3+ area shown in Figure 5L was quantified using the Analyze Particles tool in ImageJ (Fiji) on binarized images. To ensure consistency, all measurements were performed under identical exposure times and magnifications. Calculation of the ratio of BrdU+ cells To calculate the ratio of BrdU+ cells, a line perpendicular to the A–P axis was drawn at either the anterior or posterior edge of the Col2 expression domain. The numbers of BrdU+ cells and total nuclei located anteriorly or posteriorly to this line were manually counted using the Count tool in Photoshop.