Divergent stem cell mechanisms govern the primary body axis and appendage regeneration in the axolotl

Abstract

Exploring the fundamental mechanisms of organ regeneration is crucial for advancing regenerative medicine. The axolotl tail represents an opportunity to study regeneration of the primary axis including segmented muscle, vertebrae, and skin. During tail development, muscle stem cells (MuSCs) displayed expected specificity to the muscle lineage. Tail amputation, however, induced expansion of MuSC potential yielding clonal contribution to muscle, connective tissue including cartilage, pericytes, and fibroblasts. This expanded potential was not observed during limb regeneration, and cross-transplantation showed that these differences in potential are likely intrinsically determined. Single-cell RNA sequencing profiling revealed that tail MuSCs revert to an embryonic mesoderm–like state. Through genetic manipulation involving the overexpression of constitutively active transforming growth factor–β (TGF-β) receptors or Smad7 (antagonist of TGF-β signaling) in MuSCs, we demonstrated that the levels of TGF-β signal determine the fate outcome of MuSCs to connective tissue lineage or muscle, respectively. Our findings illustrate that variation in MuSCs may represent a fundamental difference between regeneration of primary axis versus appendage.

Muscle stem cells regain multipotency via TGF-β signaling during axolotl tail regeneration.

INTRODUCTION

The axolotl regenerates multiple, complex organs, including the tail and limb (1–3). These consist of similar mesodermal tissues including muscle, dermis, and skeleton, but it is unknown whether the same stem cell mechanisms are implemented to regenerate these two body regions. Previous studies in the limb found that lineage-restricted progenitor/stem cells contribute to the blastema and give rise to corresponding tissues in regeneration (4, 5). The tail is of particular interest as it represents the regeneration of the primary axis. Amputation of the tail results in the regeneration of a functional structure that consists of all primary body axis tissues, including muscle, connective tissues (CTs), and spinal cord (6, 7). However, in comparison to the limb, our understanding of the cell origins of tail regeneration remains limited.

Muscle stem cells (MuSCs) are generally considered a unipotent stem cell that regenerates muscles after injuries, although a contribution to other cell types is sometimes observed (8–11). In axolotls, fate mapping of genetically labeled Pax7⁺ MuSCs or labeled muscle tissue (muscle plus MuSCs) has consistently demonstrated their contribution exclusively to muscle during limb regeneration (5, 12, 13). The tail myotomes also harbor Pax7⁺ MuSCs (7, 12, 14), but their fate during tail regeneration has not been studied. From a developmental perspective, limb and tail MuSCs originate separately from the ventrolateral and dorsomedial portions of early somites (15–18). They then migrate into the lateral plate or are situated in somites, giving rise to abaxial (limb) and primaxial (tail) muscles, respectively (17). The primary body axis is considered an ancient feature that evolutionarily emerged much earlier than the appendage/limb (19, 20), and its regeneration has been observed in invertebrate chordate amphioxus (21). Here, we followed the fate of Pax7⁺ MuSCs, including clonal analysis during axolotl tail regeneration and unexpectedly found that the tail MuSCs, in contrast to their limb counterparts, expand their potency to contribute to multiple CT lineages via dedifferentiation to an embryonic-like mesodermal progenitor state. The levels of transforming growth factor–β (TGF-β) signaling regulate the fate outcome of MuSCs during tail regeneration, linking an injury signal with reversion to an early multipotent mesodermal cell–like state.

RESULTS

Pax7⁺ cells produce CT lineages during tail regeneration but not during development in juvenile axolotls

To identify the cell source contributing to the primary body axis during tail regeneration, we first bred Pax7:ER^T2CreER^T2 transgenic axolotls, in which expression of a membrane-tagged cherry (memCherry) followed by an inducible Cre is driven by the endogenous Pax7 promoter (12), with CAG:GFP-STOP-Cherry line (22) to establish Pax7:ER^T2CreER^T2/CAG:GFP-STOP-Cherry double transgenic (dTG-Pax7) line (Fig. 1A). Both previous studies and our data demonstrated that the dTG-Pax7 system is highly faithful, showing no leakiness before tamoxifen treatment and enabling specific reporter conversion upon tamoxifen administration (figs. S1 and S2) (12). Before tamoxifen induction, faint membrane-localized CHERRY (memCHERRY) signals were observed in Pax7⁺ cells, confined to the spinal cord and MuSCs. Upon tamoxifen treatment, excision of the floxed-STOP cassette activated cytosolic CHERRY expression under the strong CAGGS promoter. This recombination event resulted in permanent labeling and lineage tracing of Pax7⁺ cells and their progeny, with a marked increase in fluorescence intensity and the emergence of newly formed cytosolic CHERRY⁺ myofibers (fig. S2) (12). These distinct differences allow unambiguous discrimination between background and recombined signals. In all subsequent experiments, we exclusively trace these recombined cytosolic CHERRY signals, referring to them simply as “CHERRY.” Immunohistochemistry on sections showed that about 30% PAX7⁺ MuSCs were converted to CHERRY expression at 14 days post–tamoxifen treatment (fig. S3). Consistent with the distribution of PAX7-expressing cells (fig. S4), after 14 days or 6.5 months of tracing, the tamoxifen-converted CHERRY signal was restricted to the spinal cord and muscle compartments of dTG-Pax7 axolotls as expected (Fig. 1, B to E, and figs. S5 and S6, A and B). Within the muscle compartment, the converted cells were positive for PAX7, myocyte enhancer factor 2C (MEF2C), and myosin heavy chain (MHC), confirming their muscle lineage identity (fig. S5) (13).

Fig. 1. Pax7⁺ cells give rise to multilineages during juvenile axolotl tail regeneration.

(A) Strategy for genetic tracing of Pax7⁺ cells in Pax7:ER^T2CreER^T2/CAG:GFP-STOP-Cherry dTG (dTG-Pax7) axolotls. The system includes three fluorescent components: green fluorescent protein (GFP) (CAGGS-driven, not used for lineage tracing), memCHERRY (faint, membrane-localized in Pax7⁺ cells), and CHERRY (bright, cytosolic, CAGGS-driven, activated by 4-OHT–induced Cre to permanently label Pax7⁺ cells and progeny for lineage tracing). (B, D, and F) Scheme of lineage tracing of Pax7⁺ cells in 2.5-cm dTG-Pax7 axolotl tails during development (B and D) and regeneration (F). (C, E, and G) Pax7⁺ cells produce multilineages during tail regeneration, but not during tail growth. Representative images of CHERRY fluorescence in the tail of dTG-Pax7 axolotls at 14 days (d) [(C) n = 30], 6.5 months [(E) n = 6)] after tamoxifen treatment, or 30 days postamputation following tamoxifen treatment [(G) n = 30)]. (H to K) Molecular characterization of the progeny of Pax7⁺ cells. Immunofluorescence for CHERRY (red) and PRRX1 [green in (H)], SOX9 [green in (I)], CD34 [green in (J)], SOX10 and TUJ1 [green and white in (K)], combined with 4′,6-diamidino-2-phenylindole (DAPI, blue) staining on longitudinal sections of 30-day tail regenerates from the tamoxifen-converted dTG-Pax7 axolotls (n = 6). The rectangles [in (C), (E), and (G) to (K)] are depicted at a higher magnification in their respective counterparts. Yellow dashed lines, the plane of amputation; white dashed lines, shape of the tail. Arrows, CHERRY⁺ progeny from converted Pax7⁺ cells in tail regenerates; arrowheads, CHERRY⁺ pericytes. Scale bars, 2 mm [(C) and (E)]; 1 mm (G); 500 μm [(H) to (K)]. See also figs. S1 to S8. dpa, days postamputation.

To map the fate of Pax7⁺ cells during regeneration, we amputated the tails of tamoxifen-converted dTG-Pax7 axolotls and followed the progeny of CHERRY⁺ cells (Fig. 1F). We observed that, in addition to the spinal cord and muscle, abundant CHERRY⁺ cells appeared throughout the regenerated fin and cartilage tissues at 30 days postamputation (Fig. 1G and fig. S6C). These CHERRY⁺ cells exhibited various morphologies different from typical muscle lineage cells (Fig. 1G and fig. S6, D to G). To define the identity of these cells, we further characterized the markers associated with converted CHERRY⁺ cells on longitudinal sections of 30-day tail regenerates by immunohistochemistry. We found that converted Pax7⁺ cells gave rise to CHERRY⁺ peripheral nerves, Schwann cells, fin fibroblasts, chondrocytes, and pericytes, identified by the expression of markers TUJ1, SOX10, PRRX1 and SOX9, or by their physical locations (Fig. 1, H to K) (22–25). Except for nerve fibers and a few Schwann cells, no labeled CT lineages were detected in uninjured dTG-Pax7 axolotls at 14 days after tamoxifen treatment (fig. S6, H to N). Quantification showed that, at 14 days post–tamoxifen treatment, CHERRY⁺ cells comprised primarily PAX7⁺ MuSCs and MEF2C⁺ myocytes, with only a small fraction of SOX10⁺ Schwann cells. In contrast, more than 20% (22.3%) of all CHERRY⁺ cells were of nonmuscle lineages, including PRRX1⁺ fibroblasts, SOX9⁺ chondrocytes, pericytes wrapping around CD34⁺ blood vessels, and SOX10⁺ Schwann cells at 30 days postamputation (fig. S6, L to N). These nonmuscle lineages accounted for 6.4, 11.8, 7.7, and 7.1% of their respective total cell populations in tail regenerates (fig. S6O).

We next examined the contribution of MuSCs during limb development and regeneration in dTG-Pax7 axolotls. Similar to observations in the tail, at 14 days post–tamoxifen treatment, CHERRY-converted cells were restricted to the PAX7⁺ cells and their muscle derivatives (5, 12) but were absent from PRRX1⁺ CT cells (fig. S7, A and B). Quantification revealed that approximately 30% of PAX7⁺ MuSCs were converted to CHERRY expression at this time point (fig. S3). Consistent with previous findings (5, 12), these converted CHERRY⁺ cells contributed exclusively to muscle, with no incorporation into CT lineages during limb regeneration (fig. S7, C and D). Similarly, developmental tracing of Pax7⁺ MuSCs revealed their strictly myogenic fate during both short-term (2 weeks) and long-term (6.5 months) periods of limb development and growth (fig. S8). These data show that the acquisition of extra CT lineages by Pax7⁺ cells is exclusive to tail regeneration, with no instances observed during tail and limb development or limb regeneration in juvenile axolotls.

MuSCs are the source contributing to CTs during tail regeneration

During development, Pax7 expression is associated not only with myogenic lineages but also with the dorsal neural tube and neural crest (12, 26–28). We next sought to pinpoint the Pax7 cell type that generates multilineages during tail regeneration. Without amputation, CHERRY-converted cells in dTG-Pax7 axolotls include neural stem cells (NSCs), MuSCs, and their derived muscles (fig. S5, A and B). We therefore created various CreER^T2 transgenic lines to trace these relevant cell types. Characterization of newly established Sox2:CreER^T2/CAG:DeadGFP-STOP-Cherry (29) or MCK:ER^T2CreER^T2/CAG:GFP-STOP-Cherry dTG lines (dTG-Sox2; dTG-MCK) revealed faithful labeling and CHERRY conversion in desired cell types (figs. S9, A to E, and S10). Fate mapping of NSCs in the nervous system or mature muscles showed that neither Sox2⁺ NSCs nor MCK⁺ muscle fibers gave rise to CT or muscle lineages during tail regeneration (fig. S9, F to J, and S11). We found Schwann cell contributions arising from CHERRY-converted Sox2⁺ NSCs (fig. S9, G to J). These results suggest that the previously observed Schwann cells are likely derived from Pax7⁺ NSCs, while CT lineages potentially originated from Pax7⁺ MuSCs in dTG-Pax7 axolotls.

To precisely determine the origin of CHERRY⁺ CT cells in dTG-Pax7 axolotls during tail regeneration, we chose Myf5, another marker expressed in myogenic stem/progenitor cells (30–32), and generated Myf5:ER^T2CreER^T2 transgenic axolotls. Tracing the progeny of Myf5⁺ cells in Myf5:ER^T2CreER^T2/CAG:DeadGFP-STOP-Cherry dTG axolotls (dTG-Myf5) revealed that converted CHERRY⁺ cells exclusively produced Pax7⁺ MuSCs and MEF2C⁺ or MHC⁺ muscles, but not spinal cord NSCs, 14 days after tamoxifen treatment in the uninjured tail (Fig. 2, A and B, and fig. S12). However, following amputation, these CHERRY⁺ cells gave rise to CT lineages, including pericytes wrapping around CD34⁺ blood vessels, PRRX1⁺ fibroblasts, and SOX9⁺ cartilage chondrocytes in 30-day tail regenerates of dTG-Myf5 axolotls (Fig. 2, B to D, and fig. S13, B, D, and F). In contrast, all these CT lineages were absent in dTG-Myf5 tail tissue at 14 days post–tamoxifen treatment before amputation (fig. S13, A, C, and E). Quantification showed that, at 14 days post–tamoxifen treatment, CHERRY⁺ cells consisted primarily of PAX7⁺ MuSCs and MEF2C⁺ myocytes. In contrast, at 30 days postamputation, approximately 18.3% of all CHERRY⁺ cells belonged to CT lineages, comprising 12.3% PRRX1⁺ fibroblasts, 4.4% SOX9⁺ chondrocytes, and 1.7% pericytes (fig. S13, G and H). These results demonstrate that Myf5⁺ myogenic stem/progenitor cells are capable of being committed to multilineages and, together with the Pax7 tracing data, imply that MuSCs are the potential source giving rise to CT lineages during tail regeneration.

Fig. 2. Myf5⁺ MuSCs give rise to CT lineages during tail regeneration.

(A and B) Strategy (A) and experimental design (B) for genetic labeling and tracing of Myf5⁺ cells in Myf5:ER^T2CreER^T2/CAG:DeadGFP-STOP-Cherry dTG (dTG-Myf5) axolotls. (C and D) Tamoxifen-converted Myf5⁺ cells contribute to CTs during tail regeneration. Representative images of CHERRY (red) fluorescence in the tail (C) and immunofluorescence for CHERRY (red), PRRX1 (green), and DAPI staining on tail cross-sections (D) of tamoxifen-converted dTG-Myf5 axolotls at 30 days postamputation (dpa) (n = 8). SC, spinal cord in D(b). The rectangles are depicted at a higher magnification in their respective counterparts. Yellow and white dashed lines indicate the plane of amputation and shape of the tail, respectively. Yellow arrows, CHERRY⁺ fibroblasts; white arrows, CHERRY⁺ pericytes; arrowheads, CHERRY⁺ cartilage. Scale bars, 2 mm (C); 100 μm (D). See also figs. S9 to S13.

It has been reported that MuSCs are a heterogeneous population (8, 33, 34). Live imaging revealed that tamoxifen-converted CHERRY⁺ cells in dTG-Pax7 axolotls actively migrated into the newly formed tail blastema during regeneration, giving rise to distinct clusters potentially resembling clones of muscles and CTs (fig. S14). To investigate the potency and heterogeneity of Pax7⁺ MuSCs, we conducted single-cell transplantation (10) and mapped their lineage during tail regeneration (Fig. 3A). We dissociated tail muscle tissue from tamoxifen-converted dTG-Pax7 axolotls to isolate CHERRY⁺ single cells, of which 87.7% (93 of 106) were positive for PAX7 by immunostaining (Fig. 3B). A freshly isolated single CHERRY⁺ cell was then transplanted to the tail of a host d/d axolotl to trace its lineage during regeneration (Fig. 3, C, D, and I, and fig. S15, A, D, and G). Following amputation, 16% (32 of 200) of transplanted host axolotls contained CHERRY⁺ cells in 30-day tail regenerates. Live imaging and immunohistochemical analysis on cross sections of the tail regenerates revealed contributions to single or multiple lineages of muscle, chondrocytes, and fibroblasts (Fig. 3, E to H, J, and K, and fig. S15, B, C, E, F, and H to J). Of the 32 successfully transplanted hosts, 21.9% (7 of 32) contributed to multiple lineages of muscle, cartilage, and fibroblast; 15.6% (5 of 32) contributed to cartilage and fibroblast; while 28.1% (9 of 32), 15.6% (5 of 32), and 18.8% (6 of 32) only muscle, cartilage, or fibroblast, respectively (Fig. 3L). These results demonstrate that tail Pax7⁺ MuSCs or their derived progenitors involved in regeneration are heterogeneous, with various differentiation potencies, showing myogenic or CT-genic unipotency, or dual myogenic and CT-genic multipotency.

Fig. 3. Clonal analysis reveals the heterogeneity and multipotency of MuSCs during tail regeneration.

(A) Scheme of single MuSC transplantation and fate mapping during tail regeneration. Pax7:ER^T2CreER^T2/CAG:GFP-STOP-Cherry dTG (dTG-Pax7) axolotls were treated with 4-OHT to label MuSCs. CHERRY⁺ single cells were isolated from tail muscle compartments 14 days post-tamoxifen and transplanted into d/d hosts. Hosts underwent tail amputation 2 days later, and CHERRY⁺ cell fates were analyzed in 30-day regenerated tails. (B) PAX7 expression in isolated CHERRY⁺ single cells. Immunofluorescence for CHERRY (red) and PAX7 (white), and DAPI on isolated single cells from muscle compartments of tamoxifen-converted dTG-Pax7. (C) Merged and CHERRY-only fluorescence images of a CHERRY⁺ single cell loaded into a capillary pipette for transplantation. (D to K) Transplantation and fate mapping of single CHERRY⁺ cells during tail regeneration. Bright field (BF), CHERRY (red) fluorescence, and merged images showing a transplanted CHERRY⁺ cell settling in d/d host tails (D and I) and its derived progeny in regenerated tail (E and J). (F to H and K) Molecular characterization of the progeny from transplanted single cells shown in (E) and (J). Cross-sections of regenerated tails immunostained for CHERRY (red) with MHC [green (F and K)], SOX9 [green (G)], PRRX1 [green (H)], and DAPI (n = 32). (L) Quantification of single MuSC fates during tail regeneration. The rectangles are depicted at a higher magnification in their respective counterparts. Yellow and black dashed lines indicate the plane of amputation and shape of spinal cord, respectively. Black arrows, transplanted single cells; white arrows, CHERRY⁺ muscles; white arrowheads, CHERRY⁺ fibroblasts; empty arrowheads, CHERRY⁺ chondrocytes. Scale bars, 1 mm [(C), (E), and (J)]; 500 μm [(D) and (I)]; 50 μm in [(B), (F) to (H), and (K)]. See also figs. S14 and S15.

Considering the differences observed in MuSCs between the tail and limb during regeneration, we next investigated the lineage and potency of limb MuSCs at the single-cell level. To this end, we carried out single-cell grafting of freshly isolated limb MuSCs to the tail, or tail MuSCs to the limb of d/d hosts and followed the fate of the transplanted cells (Fig. 4A). We found that all grafted limb MuSCs (11 of 11) exclusively produced muscles, even in the tail regeneration environment (Fig. 4B). These data further revealed that MuSCs in the limb, unlike their tail counterparts, are unipotent. We found that tail MuSCs gave rise to both muscle and CTs in the limb environment during regeneration (Fig. 4, C and D). These experiments indicate that the variations in differentiation potency in MuSCs are intrinsically determined, and MuSCs keep a memory of their original differentiation potency during regeneration, even when situated in a different environment.

Fig. 4. Reciprocal grafting of limb and tail MuSCs reveals potency variations.

(A) Scheme for reciprocal grafting of single converted CHERRY⁺ satellite cells between the limb and tail, and fate mapping during limb and tail regeneration. Host d/d axolotls were amputated 2 days posttransplantation (2 dpt) of tamoxifen-converted, single CHERRY⁺ satellite cell, and analyzed at 35 or 81 days postamputation (dpa). (B to D) Bright-field (BF) and CHERRY (red) fluorescence images showing the progeny of reciprocal transplanted MuSCs [limb to tail, (B); tail to limb, (C and D)] during regeneration. Note that a single grafted limb MuSC produces only muscle lineages (n = 11) during tail regeneration (B). However, a single grafted tail MuSC gives rise to both muscle and CT lineages (fibroblasts and chondrocytes, n = 3) (D), or only CT lineages (pericytes and chondrocytes, n = 7) (C) during limb regeneration. The rectangles are shown at higher magnification as single-channel CHERRY images. Solid and empty arrowheads indicate fibroblasts and chondrocytes; white and yellow arrows indicate muscle fibers and pericytes, respectively. Scale bars, 2 mm (B); 1 mm [(C) and (D)].

MuSCs give rise to CT lineage through intermediate progenitors

To elucidate the transition routes underlying the multipotency acquisition of Pax7⁺ MuSCs during axolotl tail regeneration, we collected tamoxifen-converted CHERRY⁺ cells and their progeny by fluorescence-activated cell sorting (FACS) from uninjured and regenerating tails of dTG-Pax7 axolotls (fig. S16A), and carried out single-cell RNA Smart-seq2 sequencing (scRNA-seq) (35). We obtained a total of 1914 cells, with an average of ~7500 genes detected per cell (fig. S16B).

We then analyzed scRNA-seq data by Louvain clustering (36) and annotated 20 subtypes, including cell types in the muscle and neuronal lineages such as MuSCs (Pax7 and Myf5), myoblasts (Myog and Myf5), neural progenitors (Pax7, Gfap, and Sox2), Schwann cells (Sox10), and neurons (Neurod6); cell types in the CT lineages such as fibroblasts (Prrx1, Pdgfra, and Lum) and chondrocytes (Sox9, Pdgfra and Lum) (Fig. 5, A and B, and fig. S16C). We also uncovered two subtypes that exhibited great similarity to fibroblasts (Prrx1, Pdgfra and Lum) but also expressed MuSC markers Pax7 and Myf5 (Fig. 5, A and B), therefore named as intermediate cell cluster 1 (IM1) and 2 (IM2). We separated cells from all samples into three compartments on Uniform Manifold Approximation and Projection (UMAP) according to gene expression similarity (fig. S16D), named “regions 1 to 3” (Fig. 5C). MuSCs, IMs, and fibroblasts fell into region 1. To explore the temporal dynamics of the identified cells throughout tail regeneration, we further subdivided seven samples into four phases according to the kinetics of MuSCs and IMs (Fig. 5C). We found that before amputation, consistent with previous fate mapping data, nothing other than muscle lineage existed in region 1 at phase 1 (0 days postamputation) (Fig. 5C). IM1 and IM2 sequentially emerged at phase 2 (2, 4, and 6 days postamputation) and III (10 and 15 days postamputation), concomitant with a decline of MuSCs and myoblasts, and a burst of fibroblasts and chondrocytes, accordingly (Fig. 5C). At phase IV (33 days postamputation), close to the completion of tail regeneration, MuSCs and myoblasts in region 1 reverted to a state resembling that at 0 days postamputation, coinciding with the vanishing of IMs (Fig. 5C). We therefore designated region 1 as the “regeneration active zone.”

Fig. 5. MuSCs give rise to CT lineages via intermediate progenitors during tail regeneration.

(A) UMAP of single-cell clustering from Smart-seq2 data. Twenty cell types were annotated, including Cho, chondrocyte; Sch, Schwann cell; Endo, endothelial cell; Ery, erythroid cell; Epe, ependymoglial cell; Epi1, epithelial cell 1; Epi2, epithelial cell 2; Fib1, fibroblast 1; Fib2, fibroblast 2; Fib3, fibroblast 3; IM1, intermediate cell 1; IM2, intermediate cell 2; Ker, keratinocyte; Mac1, macrophage 1; Mac2, macrophage 2; MuSC, muscle stem cell; Myo, myoblast; Neu, neuron; NeuP, neural progenitor cell, with one unknown cell type designated as “Mix.” (B) Violin plot displaying expression levels of representative markers in each annotated cell type. (C) UMAP visualization of cells from all sampling stages (left) and dynamics of cell types at different phases (right). Three regions with lineage-related cells are delimited by polygons and labeled as regions 1 to 3. Note that major activities occur in region 1, designated as the regeneration active zone. (D) Heatmaps displaying stemness potential (left), Pax7 expression levels (middle), and the merge of top 20% stemness potential and Pax7 expression levels ≥3 (right) in cells within the regeneration active zone. (E) RNA velocity streamline plots predicting lineage transitions among myogenic-related and CT cells at early [left, 0, 2, and 4 days postamputation (dpa)] and late (right, 6, 10, 15, and 33 days postamputation) regeneration stages, respectively. (F) Trajectory inference showing dual routes of MuSC, committed to myoblast at 0 days postamputation (development), whereas switched to IM1 at 2 and 4 days postamputation (early regeneration). See also figs. S16 and S17.

To determine the origin and descendants of IMs in regeneration, we performed a stemness score calculation (37) and uncovered that IM1 and IM2, similar to MuSCs, exhibited a higher stemness score and Pax7 expression (Fig. 5D and fig. S16E). We then performed RNA velocity analysis (38) and revealed a transition from MuSCs to IM1 at the early stage of regeneration (Fig. 5E). Trajectory analysis (39) also showed that MuSCs differentiated into myoblasts before amputation (0 days postamputation) but switched to IM1 during early regeneration (2 and 4 days postamputation) (Fig. 5F and fig. S17). The emergence of IM1 appeared to occur at the expense of MuSCs, as evidenced by a dramatic decline of MuSCs and myoblasts (Fig. 5C), whereas, at the late stage of regeneration, we noticed a shift from IM1 to IM2, followed by differentiation into CT cells or replenishment of the consumed MuSCs (Fig. 5, C and E). In conclusion, our data indicate that IMs, as a regeneration-induced stem/progenitor population, are derived from MuSCs and act as the direct source contributing to CT lineages during tail regeneration.

The state and lineages of IM progenitors resemble that of early embryonic multipotent mesodermal cells

We next assessed the ground state of MuSC-derived IMs. Previous reports have shown that limb fibroblasts and brain ependymoglial cells in axolotls revert to a more primitive, embryonic-like state during regeneration (23, 40). Therefore, we wondered whether IMs share similar gene expression profiles and lineages with certain types of early embryonic cells, e.g., mesodermal cells. It has been reported that a subpopulation of Pax7-expressing mesodermal cells gives rise to trunk adipocytes, dermis, and muscles during early mouse embryogenesis (e.g., E9.5 and E10.5), but become lineage-restricted to generate only muscles as development proceeds (e.g., E12.5) (27). Whole-mount in situ hybridization on early axolotl embryos and immunohistochemical analysis on sections showed Pax7 expression in neural crest, neural tube, somitic and presomitic mesoderm cells (Fig. 6A and fig. S18A), consistent with observations in rodents (27). We then treated dTG-Pax7 embryos with a consecutive or single dose of tamoxifen at stages ranging from 15 to 44, to map the fate of Pax7⁺ cells in the tail of 2-cm axolotls during development (Fig. 6B and fig. S18B). In addition to muscle and neuronal lineages, Pax7⁺ cells made substantial contributions to CT lineages. CHERRY⁺ cells converted at stage 15 to 25 by a single-pulse tamoxifen produced abundant CHERRY⁺ progeny, including dermis, fibroblasts, and pericytes in the fin tissue of axolotl larvae, while those converted at stage 30 or 44 yielded fewer or no CHERRY⁺ progeny (Fig. 6C and fig. S18, C to F). This transient CT lineage commitment in Pax7⁺ cells during early embryogenesis might resemble the induction of CT lineages observed in MuSCs during juvenile axolotl tail regeneration. Collectively, our results suggest that a population of Pax7⁺ embryonic progenitor cells (EPrCs) within the neural crest or mesoderm are able to transiently give rise to CT lineages during axolotl embryonic tail development.

Fig. 6. The state and lineages of intermediate cells (IMs) resemble that of early embryonic multipotent mesodermal cells.

(A) Embryonic expression of Pax7. Whole-mount in situ hybridization of axolotl embryos at stages (st.) 15, 20, 25, and 30. Empty arrowheads, arrowheads, and arrows indicate Pax7 expression in the neural crest, somite, and presomatic mesoderm, respectively. (B) Scheme for genetic tracing of Pax7⁺ cells at early embryonic stages. dTG-Pax7 embryos were treated with a single dose of tamoxifen at the indicated stages, and analyzed in 2-cm axolotls. (C) Lineages of early embryonic Pax7⁺ cells. Representative CHERRY (red) fluorescence images of tails from 2-cm dTG-Pax7 axolotls treated with tamoxifen at indicated stages. (n = 3 to 5 embryos per stage). (D) Scheme of Smart-seq2 sample collection at embryonic development stages. (E) PCA plot showing cell type similarity. EPrCs overlay with IMs, which indicates IMs from regeneration resembling the state of EPrCs from embryos. EPrC, early embryonic progenitor cell. (F and G) Spatial visualization of EPrCs on Stereo-seq section of stage 25 axolotl embryos. Dark red, EPrC scRNA data projected to Stereo-seq data; magenta, Sox10⁺ neural crest cells; light green, mesoderm areas. Note that EPrC is mainly located in the mesoderm area as determined by ssDNA staining (F). (H) Integration of Smart-seq2 and embryonic Stereo-seq data. EPrCs and cells from the regeneration active zone primarily clustered with mesodermal cells, but not neural crest cells. Meso, mesoderm; NC, neural crest. (I) Bubble plot exhibiting the expression of neural crest and mesoderm marker genes in annotated cell types from Smart-seq2 data. Scale bar, 1 mm (A); 500 μm (C); 100 μm [(F) and (G)]. See also figs. S18, S19, and S25.

We next attempted to evaluate the likeness of EPrCs and IMs. To this end, we consecutively treated dTG-Pax7 embryos with tamoxifen from stage 15 to 25 and isolated tamoxifen-converted CHERRY⁺ cells containing EPrCs at stage 35 for Smart-seq2 sequencing (Fig. 6D). We then integrated and clustered these data with scRNA-seq data from regeneration samples and uncovered three main cell types, including neural progenitors, neurons, and a previously uncharacterized cell cluster from traced embryos. This cluster was drawn in the regeneration active zone and expressed Pax7 and Prrx1 (fig. S19, A to C), speculating its EPrC identity. A constellation plot showed intensive interplays among EPrCs, MuSCs, and IMs, demonstrating shared expression profiles among these cell types (fig. S19D). Moreover, principal components analysis (PCA) demonstrated an overlap between EPrCs and IMs on the plot (Fig. 6E), suggesting that IMs partially recapitulate the state of EPrCs, with those from 10 to 15 days postamputation showing the closest proximity (fig. S19E).

As previously observed, EPrCs that contributed to CT lineage were potentially derived from neural crest or mesodermal cells (Fig. 6, A to C, and fig. S18). To define the origin and localization of EPrCs, we performed Stereo-seq on cross section of stage 25 axolotl embryos and projected EPrCs to the spatial transcriptomics data (41). Such projection showed that EPrCs were primarily located within the mesoderm region, as determined by single-stranded DNA (ssDNA) staining on the Stereo-seq section (42) (Fig. 6, F and G). In addition, by extracting mesoderm and neural crest cells from the spatial transcriptomics data and using integrative clustering on extracted spatial segmented cells and all scRNA data, we observed that IMs and EPrCs were predominantly situated in the embryonic mesoderm zone, separated from the Sox10⁺ neural crest area on the UMAP (Fig. 6H). Consistently, marker genes relevant to the mesoderm (43), but not neural crest cells (44), were expressed in IMs and EPrCs (Fig. 6I). Together, these data suggest that both IMs and EPrCs closely resemble the state of early mesodermal cells, implying that tail MuSCs revert to a multipotent embryonic mesodermal cell–like state to produce both muscle and CT cells during regeneration.

TGF-β signaling pathway regulates the cell fate switch of myogenic MuSCs

To identify a molecular determinant modulating the cell fate switch of MuSCs toward CT lineages, we further subclustered MuSCs and IMs according to their differentiation potential (Fig. 7A) and obtained nine subtypes committed to fibroblasts (IM1_Fib1 and IM2_Fib2), muscle lineages (IMs_MuSC and IM2_Myo), or uncommitted. We then used these nine subtypes, combined with Fib1, Fib2, and Myo, major downstream cell types relevant to fibroblast or muscle lineages, to analyze pathway activity guiding the cell fate switch (45). We observed that these cells were grouped into two segments (Fig. 7B). IMs committed to fibroblast and muscle, situated in the upper and lower segments, respectively, displayed the most divergent activity on the heatmap. Among them, the Wnt, TGF-β, and MAPK pathways emerged as the top three activated pathways.

Fig. 7. TGF-β signaling regulates the cell fate switch of MuSCs.

(A) UMAP visualization of refined subclusters within the regeneration active zone. MuSC, IM1, and IM2 are partitioned into nine subclusters based on shared gene expression programs between progenitors and their progeny. Each cell type, e.g., IM2, is subdivided into subclusters committed to myoblasts (IM2_Myo), fibroblast 2 (IM2_Fib2), or uncommitted (IM2_UC). (B) Heatmap showing the scores of predicted pathway activity in nine defined subclusters from (A), combined with Fib1, Fib2, and Myo. (C and D) SB-431542 inhibits the production of fibroblasts during tail regeneration. Images of CHERRY fluorescence of 26-day tail regenerates treated with SB-431542 (D) and control (C) (n = 3 each). (E and F) Quantification of CHERRY signal area in the fin and regenerated fin length following SB-431542 treatment (n = 3). dpt, days posttreatment. Error bars, SEM; statistical analysis was conducted using Student’s t test (E) and one-way analysis of variance (ANOVA) (F); **P < 0.01, ****P < 0.0001; ns, not significant. (G) p-SMAD2 and SMAD2 immunoblots of regenerated tails treated with SB-431542 and control (n = 3 each). Glyceraldehyde-3-phosphate dehydrogenase immunoblot, loading control. (H and I) Quantification of signal intensity from p-SMAD2 and SMAD2 immunoblots in (G) (n = 3 each). Student’s t test; **P < 0.01. (J) Boxplot of the scores of endogenous TGF-β activity in MuSCs. Note that scores are attenuated in IM2 committed to myoblasts (IM2_Myo), in contrast, elevated in IMs to fibroblasts (IM1_Fib1, IM2_Fib2), compared to that in MuSCs at 0 days postamputation (dpa). (K and L) Cell fate trajectory predictions following in silico perturbation of TGF-β signaling. Activation of TGF-β diverts cells from muscle-related lineages to fibroblast-related lineages (K), whereas its suppression diverts cells from fibroblast-related lineages to muscle-related lineages (L). Rectangles in (C) and (D) are shown at higher magnification. Scale bars, 1 mm [(C) and (D)]. See also fig. S20.

On the basis of the above analysis, we next experimentally investigated the roles of these pathways in regulating the cell fate switch of Pax7⁺ MuSCs during regeneration. We first applied drugs IWR-1 (46), SB-431542 (47), or U0126 (48) to dTG-Pax7 axolotls to inhibit Wnt, TGF-β, or mitogen-activated protein kinase (MAPK) pathways, respectively, and traced the fates of tamoxifen-converted CHERRY⁺ cells during regeneration under each condition. We found that inhibition of the TGF-β pathway with SB-431542, evidenced by the down-regulation of phosphorylated SMAD2 (p-SMAD2) (Fig. 7, G to I), notably reduced the production of fibroblasts from CHERRY⁺ MuSCs and shortened the fin length during tail regeneration—a phenotype not observed upon the suppression of the Wnt or MAPK pathways (Fig. 7, C to F, and fig. S20). In line with the drug inhibition results, bioinformatic analysis of TGF-β activity (38) on the identified subclusters committed to different lineages revealed that the score of TGF-β signaling activation tended to increase in IM1_Fib1 and IM2_Fib2 committed to fibroblast lineage, but decrease in IM2_Myo committed to the muscle lineage, when compared to MuSC_0 days postamputation (Fig. 7J). Furthermore, in silico genetic perturbance analysis (38) suggests that MuSC preferentially differentiates to fibroblast upon TGF-β activation (Fig. 7K). In contrast, attenuation of TGF-β signaling promotes muscle lineage (Fig. 7L). Collectively, our data suggest that TGF-β, rather than Wnt or MAPK, is a potential regulator of the MuSCs fate switch.

Considering the potential side effects of drugs on regeneration, we next carried out a genetic approach to manipulate TGF-β signaling. We expressed constitutively active (ca), dominant negative (dn) formats of Tgfb receptor 1 (Tgfbr1) (49, 50) or Smad7, an antagonist of TGF-β signaling (51), in Pax7⁺ cells to study their roles in tail regeneration. To this end, we first validated these constructs in cell culture. We found that transfection of plasmids encoding Smad7, but not that encoding dnTgfbr1 inhibited TGF-β activity, while expression of caTgfbr1 promoted TGF-β activity, indicated by the level of p-SMAD2 (fig. S21). We then created Pax7:ER^T2CreER^T2/CAG:BFP-STOP-caTgfbr1/Smad7-Cherry dTG axolotls (dTG-caTgfbr1; dTG-Smad7) (Fig. 8, A and B) and investigated the cell fate commitment of Pax7⁺ MuSCs during tail regeneration upon overexpression of caTgfbr1 or Smad7. After tamoxifen treatment, we observed the CHERRY conversion in the NSCs of the dorsal spinal cord and MuSCs (figs. S22 and S23) and the overexpression of caTgfbr1 or Smad7 in converted cells in dTG-caTgfbr1 and dTG-Smad7 lines, compared to the control dTG-Pax7 axolotls (fig. S24). Following amputation, we noted that compared to the control (Fig. 8, C and F), when expressing Smad7, the production of CHERRY⁺ CT cells was nearly completely blocked, and only muscle lineage was visualized in the tail regenerates (Fig. 8, D and G). In contrast, the expression of caTgfbr1 directed the converted CHERRY⁺ cells toward the CT rather than the muscle lineage (Fig. 8, E and H). These results demonstrate that TGF-β signaling alters the balance of the cell fate switch of MuSCs toward muscle or CT lineages during tail regeneration.

Fig. 8. Modulating TGF-β activity leads to the cell fate switch of MuSCs during tail regeneration.

(A and B) Strategy for modulating TGF-β activity in Pax7⁺ cells and cell fate mapping. Two double-transgenic axolotl lines were generated: Pax7:ER^T2CreER^T2/CAG:BFP-STOP-Smad7-Cherry (dTG-Smad7) (A) and Pax7:ER^T2CreER^T2/CAG:BFP-STOP-caTgfbr1-Cherry (dTG-caTgfbr1) (B). In these lines, expression of Smad7 or constitutively active Tgfb receptor 1 (caTgfbr1), together with (cytosolic) Cherry driven by the CAGGS promoter, is specifically induced in Pax7⁺ cells upon tamoxifen administration. (C to E) Genetic inhibition or activation of Tgfb attenuates CT or muscle lineages, respectively, in Pax7⁺ MuSCs during tail regeneration. Images of the progeny of tamoxifen-converted CHERRY (red) fluorescence of 21-day tail regenerates upon expression of caTgfbr1 [(E) n = 10], Smad7 [(D) n = 7], and control [(C) n = 5]. The rectangles are shown at higher magnification [in C(a) to E(c)]. Note the complete loss of CT or muscle lineages upon expression of Smad7 or caTgfbr1, respectively. (F to H) Immunofluorescence for CHERRY (red), MHC (green), or PRRX1 (green), combined with DAPI staining, on cross sections of 21-day tail regenerates of control (Ctrl) dTG-Pax7, dTG-Smad7, and dTG-caTgfbr1 axolotls. Arrows, muscles; arrowheads, fibroblasts; empty arrowheads, chondrocytes; asterisks, spinal cord. Yellow and white dashed lines indicate the plane of amputation and the shape of the tails, respectively. Scale bars, 1 mm [(C) to (E)]; 50 μm [(F) to (H)]. See also figs. S21 to S24.

DISCUSSION

Here, we genetically traced the progeny of Pax7⁺ MuSCs, combined with fate mapping of transplanted single MuSC, and found that differentiation potency varies in MuSCs during primary body axis and appendage regeneration. MuSCs in the axolotl tail revert to a multipotent state resembling embryonic mesodermal cells and give rise to muscle and CT lineages during regeneration. Furthermore, we showed that the multipotency commitment of MuSCs is modulated by TGF-β signaling levels and demonstrated that controlling TGF-β activities led to the switch between muscle and CT lineages during tail regeneration (Fig. 9).

Fig. 9. Fate regulation of MuSCs during axolotl tail regeneration.

Schematic summary illustrating distinct fates of MuSCs during tail development and regeneration. (i) During tail development, MuSCs give rise exclusively to muscle tissue. (ii) During embryogenesis, mesodermal located, Pax7⁺ embryonic progenitor cells (EPrCs) exhibit multipotency and generate both muscle and CT derivatives. (iii) During tail regeneration, MuSCs reacquire a multipotent, embryonic-like state and generate both muscle and CT lineages. TGF-β signaling regulates the lineage switch between muscle and CT fates: High TGF-β activity drives MuSCs toward intermediate (IM) cells that give rise to fibroblasts, pericytes, and chondrocytes, whereas low TGF-β activity promotes their differentiation into muscle.

What causes the difference in MuSCs between the tail and limb during regeneration? From an evo-devo perspective, the primary body axis (tail) and appendage (limb) exhibit the following variations. First, the establishment of the primary axis occurs as early as the gastrulation stage in most vertebrates (52, 53). In axolotls, the formation of the somites starts at about stage 20, followed by the appearance of a protrusion-like tail bud at stage 21, whereas limb bud development is initiated at about stage 40 (53). Second, a muscle-containing appendage limb evolutionarily emerged much later than the primary body axis tail (19, 20, 54). Consequently, the cells that contribute to tail and limb formation could be potentially different (16–18). MuSCs in the tail and limb originate from the dorsomedial and ventrolateral parts of the somites. Moreover, the muscle tissues, including MuSCs of the tail and limb, are embedded, respectively, in a presomitic mesoderm- and lateral plate mesoderm–derived CT environments (16–18). These differences might be the primary rationale for our discovery that tail MuSCs, in contrast to their limb counterparts, contribute additionally to CT cells during regeneration. It will be important to further identify the key intrinsic or extrinsic factors resulting in the potency variations between tail and limb MuSCs during regeneration.

The multipotency of MuSCs was initially observed under in vitro cell culture conditions, where they were found to generate adipocytes and osteocytes, besides myocytes (55, 56). Subsequent studies revealed that MuSCs were also capable of producing similar multilineages during mouse embryonic development (11, 27, 57, 58); however, they seldom commit to multilineages spontaneously during homeostasis or regeneration (8). In our study, by applying genetic fate mapping, particularly, tracing the progeny of freshly isolated single-transplanted MuSCs, which provides the most direct evidence of heterogeneity and multipotency, we observed a unique induced multipotency conversion of MuSCs during tail regeneration and a potency variation between limb and tail MuSCs. This induced multipotent state of MuSCs, named IMs, exhibits a mixed feature of MuSCs and CTs, expressing molecular markers from each cell type. The phenotype, observed in MuSCs during tail regeneration, aligns remarkably well with the “cyclical fate restriction” hypothesis proposed by Kelsh and colleagues (59), which suggests that specialized cells can dedifferentiate into a multipotent cycle through several potential fates and eventually commit to a single fate. Previous findings in newts showed that the transplantation of in vitro expanded MuSCs gives rise to adipocytes, chondrocytes, and myocytes during limb regeneration (60). In that study, MuSCs from newts were isolated in the absence of genetic labeling, and the experiment was not carried out at the single-cell level. Therefore, the purity of the MuSCs could not be precisely determined. Moreover, the MuSCs underwent in vitro expansion before transplantation, potentially altering their properties and complicating the interpretation of their in vivo behavior. Furthermore, despite both being salamanders, the axolotls and newts use distinct strategies to regenerate limb muscles. Axolotls exclusively rely on resident MuSCs, whereas newts can regenerate muscle from progenitors derived from muscle dedifferentiation (13, 61). This divergence may endow the intrinsic differences in MuSCs between these two species. It is worth systematically investigating the differentiation potential of MuSCs between the limb and tail from the newt in development and regeneration environments.

A recent preprint from Masselink and colleagues (62) demonstrated that Lfng⁺ multipotent stem cells, a subtype of CT cells at the myotendinous junction, can give rise to CT progeny and additionally muscle tissue during axolotl tail regeneration. By analyzing marker gene expression in cell types linked to the Lfng⁺ multipotent stem cells, we identified that fibroblast 3 from our data exhibited relatively high expression of Lfng, Meox1, Scx, and Mkx, showing great similarity to Lfng⁺ multipotent stem cells (fig. S25), suggesting the potential cross-talk between the phenomena observed in these two studies. Lfng⁺ multipotent stem cells function differently from previously observed CT lineage during limb regeneration, in which labeled CT only produces CT cell fate (23). Together with our work, these two studies both focused on the tissue dynamics during tail regeneration and found that MuSC and CT behave differently during tail and limb regeneration, likely representing a fundamental difference between the primary body axis (tail) and appendage (limb) during regeneration. Moreover, both MuSCs and CT originate from mesodermal tissue at early developmental stages. These similarities raised the question of whether the Lfng⁺ multipotent stem cells described in Masselink et al. (62) and the Pax7⁺ MuSCs–derived IM cells identified in our study represent complementary sources for regeneration or sequential stages within a dedifferentiation-redifferentiation progression during regeneration. Several lines of evidence from our analyses support the interpretation that these two lineages correspond to distinct yet complementary sources. First, Pax7⁺ MuSCs in our study uniquely dedifferentiate into IM cells during regeneration. Such regeneration-induced intermediate state was not observed in uninjured tails. In contrast, the Lfng⁺ multipotent stem cells contribute to both muscle and CT lineages during development and regeneration and do not appear to generate a regeneration-specific transitional population analogous to IM cells. Second, lineage tracing and scRNA-seq analyses indicate that Pax7⁺ MuSCs and Lfng⁺ progenitors are molecularly distinct, as Lfng⁺ cells do not express Pax7, while MuSCs lack expression of Lfng, Meox1, Mkx, and Scx. Third, RNA velocity analyses revealed that fibroblast 3, despite its transcriptional similarity to Lfng⁺ cells, arises only from IM2 cells at later stage of regeneration. Overall, our study of Pax7⁺ MuSCs and the Masselink study of Lfng⁺ multipotent stem cells delineate two distinct cell populations operating in different biological contexts, providing a valuable, although still incomplete, understanding of axolotl tail regeneration. Together, findings from these two studies revealed plasticity within the mesoderm lineages of MuSCs and CT induced by regenerative signals. This could be a complementary security mechanism to guarantee complete regeneration of lost cell types after primary axis injury.

Although early studies showed that elevated TGF-β signaling impairs myoblast differentiation (63, 64), little is known about its role in the cell fate commitment of MuSCs during regeneration. A recent study demonstrated that TGF-β signaling plays a negative role in muscle fusion during regeneration (65). In our study, via analysis of scRNA-seq of the progeny of MuSCs during tail regeneration, combined with conditional regulation of key TGF-β signaling components, we identified that the level of TGF-β signaling balances the fate of MuSCs during tail regeneration.

Overall, the differentiation potential and lineage variation of MuSCs during tail and limb regeneration may represent a fundamental distinction between primary body axis and appendage regeneration in axolotls and may also extend to other vertebrate species. Exploring the intrinsic and extrinsic mechanisms governing this variation in MuSCs, and perhaps in other cell types, across these two regenerative contexts may provide a deeper understanding to develop strategies for regenerative medicine.

MATERIALS AND METHODS

Axolotl husbandry

Axolotls (Ambystoma mexicanum) were originated from Ambystoma Genetic Stock Center (https://ambystoma.uky.edu/genetic-stock-center/) and bred and maintained locally in freshly dechlorinated tap water at 20°C, with daily feedings. Animals were anesthetized in 0.03% (w/v) benzocaine (Sigma-Aldrich, E1501-500G) before imaging, surgery, and sample collection. All animal procedures were conducted in compliance with Chinese animal welfare legislation, approved by the Biomedical Research Ethics Committee of Guangdong Provincial People’s Hospital (approval number KY2024-192-01).

Molecular cloning

Tol2-MCK-ER^T2-Cre-ER^T2-PA

This construct was cloned using the Gibson assembly method from three fragments: MCK, Tol2-PA-Tol2, and ER^T2-Cre-ER^T2. The MCK fragment was digested from pUC118-MCK-Cre-polyA (13) using enzymes BstEII and SbfI. The Tol2-PA-Tol2 and ER^T2-Cre-ER^T2 fragments were generated by polymerase chain reaction (PCR) amplification with primer pairs PA-Tol2-F and Tol2-R, and ERCreER-F and ERCreER-R, respectively, using template Tol2-COL2A1:ER^T2-Cre-ER^T2-T2A-EGFP-nls (22, 66).

pGEMT-Sox2-ORF-memCherry-T2A-Cre-ER^T2-PA

This plasmid was constructed via Gibson assembly from three fragments: vector backbone, Cherry-T2A, and Cre. The vector backbone was generated by digesting plasmid pGEMT-Sox2-ORF-P2A-memCherry-T2A-ER^T2-Cre-ER^T2-PA (12) with SbfI and BamHI to remove Cherry-T2A-ER^T2-Cre. Cherry-T2A and Cre fragments harboring homologous arm were PCR amplified with primer pairs Sox2-F1 and Sox2-R1, and Sox2-F2 and Sox2-R2, respectively, using pGEMT-Sox2-ORF-P2A-memCherry-T2A-ER^T2-Cre-ER^T2-PA as the template.

pGEMT-Myf5-P2A-memCherry-T2A-ER^T2-Cre-ER^T2-PA

To generate this construct, the Myf5 bait sequence, including vector homologous arms, was amplified by PCR using primers Myf5-F, Myf5-R, pGEMT-Myf5-F, and P2A-Myf5-R from axolotl genomic DNA. The fragment was then cloned via Gibson assembly into the plasmid pGEMT-Sox2-ORF-P2A-memCherry-T2A-ER^T2-Cre-ER^T2-PA (12), which had been opened with MluI and SphI.

Tol2-CAGGS:loxp-BFP-nls-STOP-loxp-Tgfbr1/caTgfbr1/dnTgfbr1/Smad7-T2A-Cherry-PA

To generate Tol2-CAGGS:loxp-BFP-nls-STOP-loxp-T2A-Cherry-PA vector, the BFP-nls harboring homologous arms was amplified by PCR using primers BFP-nls_HR-F and BFP-nls_HR-R from a synthesized BFP-nls fragment (Genscript), and assembled using in-fusion cloning method into the plasmid Tol2-CAGGS:LoxP-GFP-STOP-LoxP-T2A-Cherry-PA (22, 66) linearized with AgeI and BamHI. Tgfbr1 and Smad7 open reading frame (ORF) were amplified by reverse transcription PCR (RT-PCR) from total RNA prepared from tail fin tissue of the axolotl and Xenopus, using primers Tgfbr1-F, Tgfbr1-R, and Smad7-F, Smad7-R (51). Next, Tgfbr1 and Smad7 harboring homologous arms were amplified by PCR using previously obtained Tgfbr1 and Smad7 ORF as templates, with primers Tgfbr1_HR-F, Tgfbr1_HR-R, and Smad7_HR-F, Smad7_HR-R. They were then inserted into the vector Tol2-CAGGS:loxp-BFP-nls-STOP-loxp-T2A-Cherry-PA using the in-fusion cloning method (Vazyme, C115-01) to generate Tol2-CAGGS:loxp-BFP-nls-STOP-loxp-Tgfbr1/Smad7-T2A-Cherry-PA. Plasmids Tol2-CAGGS:loxp-BFP-nls-STOP-loxp-caTgfbr1/dnTgfbr1-T2A-Cherry-PA were further constructed by PCR amplification with the corresponding mutation primers (K234R-F, K234R-R, or T206D-F, T206D-R), using Tol2-CAGGS:loxp-BFP-nls-STOP-loxp-Tgfbr1-T2A-Cherry-PA as template with a high fidelity DNA polymerase (Vazyme, P515-03) (49). All final constructs were verified by Sanger sequencing. The primer sequences used for cloning are listed in table S1.

Transgenesis

The knock-in transgenic axolotls “Sox2:CreER^T2” [tm(Sox2^t/+:Sox2-P2A-memCherry-T2A-Cre-ER^T2)^Fei] and “Myf5:ER^T2CreER^T2” [tm(Myf5^t/+:Myf5-P2A-memCherry-T2A-ER^T2-Cre-ER^T2)^Fei] were generated according to previously published CRISPR-Cas9–mediated knock-in protocols (67) with the following targeting constructs and guide RNAs (gRNAs): pGEMT-Sox2-ORF-memCherry-T2A-Cre-ER^T2-PA, pGEMT-Myf5-P2A-memCherry-T2A-ER^T2-Cre-ER^T2-PA, Sox2-gRNA#5 (5′-GTGCCGGGCTCGTCCATCAA-3′) (12), and Myf5-gRNA (5′-GCAGAAGACCTCAAGTGCTT-3′). CAS9 protein was purchased from Editgene (EDE0008-C5), and gRNAs were synthesized by Genscript.

The transgenic axolotls “MCK:ER^T2CreER^T2” [tgTol2(Mmu.MCK:ER^T2-Cre-ER^T2)^Fei], “CAG:BFP-STOP-caTgfbr1-Cherry” [tgTol2(CAGGS:loxp-BFP-nls-STOP-loxp-caTgfbr1-Cherry)^Fei], and “CAG:BFP-STOP-Smad7-Cherry” [tgTol2(CAGGS:loxp-BFP-nls-STOP-loxp-Smad7-Cherry)^Fei] were generated using a transposon-based method (22, 66).

The following published axolotl strains were also used: d/d (control strain), Pax7 knock-in axolotl “Pax7:ER^T2CreER^T2” [tm(Pax7^t/+:Pax7-P2A-memCherry-T2A-ER^T2-Cre-ER^T2)^Etnka] (12, 66), transgenic line “CAG:GFP-STOP-Cherry” [tgSceI(CAGGS:LoxP-GFP-STOP-LoxP-Cherry)^Etnka] (22, 66), and transgenic line “CAG:DeadGFP-STOP-Cherry” [tgSceI(CAGGS:LoxP-DeadGFP-STOP-LoxP-Cherry)^Etnka], in which a frameshift mutation in the green fluorescent protein (GFP) sequence abolished GFP fluorescence (29). Nomenclature is according to a previous publication (68).

Drug treatment

Tamoxifen (Sigma-Aldrich, H7904, 4-OHT) was dissolved in dimethyl sulfoxide (DMSO, Sigma-Aldrich, D8418) at 20 mM for the stock solution. For embryonic tamoxifen treatment, embryos at different stages were incubated with a single dose of tamoxifen at 0.5 to 1 μM for 12 to 24 hours or continuously treated with tamoxifen during the stages indicated. Juvenile axolotls (2.5 cm) were bathed in 3 μM tamoxifen twice, with a 1-day interval between treatments. Following tamoxifen treatment, both embryos and juvenile animals were subjected to further analysis.

SB-431542 (MedChemExpress, HY-10431) and IWR-1 (MedChemExpress, HY-12238) were dissolved in DMSO at 30 mM stock solutions, and U0126 (MedChemExpress, HY-12031) was dissolved in absolute ethanol. At 3 days postamputation, axolotls (3 to 4 cm) were kept in 25 μM SB-431542, 25 μM IWR-1, or 40 μM U0126 (46–48), with DSMO or ethanol as controls. Drugs were replaced daily, and fins were imaged under a fluorescent stereomicroscope at 2 weeks or the indicated time posttreatment.

Whole-mount in situ hybridization

Albino axolotl embryos at stages 15, 20, 25, and 30 were fixed in PFA overnight at 4°C. Subsequently, they were washed with methanol (100, 90, 75, 50, and 25%) diluted with PTW [diethyl pyrocarbonate–treated 0.1× phosphate-buffered saline (PBS) with 0.2% Tween], followed by PTW wash. The samples were then permeabilizing by proteinase K (4 to 8 μg ml⁻¹) for 2 to 4 min at room temperature, followed by hybridization containing Pax7 antisense RNA probes (500 ng ml⁻¹) (12). Unbound RNA probes were removed by washing with posthybridization and monoclonal antibody (mAb) buffer several times. Afterward, the samples were incubated with anti–digoxigenin–alkaline phosphatase (AP) antibody (1:4000, Roche, 11093274910) diluted in blocking buffer. The signals were detected using BM purple substrate (Roche, 11442074001), and samples were stored in 50% glycerol diluted with PBS for imaged using an Olympus SZX10 microscope.

Paraffin and cryosectioning

Embryos at stages 15, 25, 30, and 35, or surgically excised tail or fin tissues from juvenile axolotls, were fixed in MEMFA (MOPS-EGTA-MgSO₄-formaldehyde) at 4°C overnight. For paraffin sections, the embryonic samples were dehydrated using an ascending series of methanol (50, 75, 90, and 100%) for 2 hours each, followed by dehydration in 100% ethanol for 30 min. They were then processed in xylene twice for 15 min each before being embedded in paraffin. Paraffin sections of 6-μm thickness were prepared for subsequent analysis. For cryosections, fixed tissues were dehydrated with 30% sucrose, embedded in OCT compound (Sakura, 4583), and cryosectioned to 10-μm thickness using a CryoStar NX70 cryostat. Sections were stored at −20°C for further analysis.

Immunofluorescence

Paraffin sections were deparaffinized before immunofluorescence staining. Both deparaffinized sections and air-dried cryosections were washed briefly with PBST (0.3% Triton X-100). Antigen retrieval, when necessary, was performed at 100°C for paraffin sections or 85°C for cryosections for 10 min in Citrate Antigen Retrieval solution (Sangon, E673001-0100). Following a wash with PBST to remove the retrieval solution, sections were blocked in 5% donkey serum (Solarbio, ASL050) and incubated overnight at 4°C with primary antibodies diluted in antibody dilution buffer (Genefist, GF1600-1). The primary antibodies used in this study included mouse–anti-TUJ1 (R&D, MAB1195), rabbit–anti-CD34 (Abcam, ab81289), rabbit–anti-MEF2C (Santa Cruz Biotechnology, sc-313), rabbit–anti-SOX2 (GeneTex, GTX124477), rabbit–anti-PRRX1 (a gift from E. M. Tanaka’s laboratory), rabbit–anti-SOX10 (Abcam, ab264405), mouse–anti-PAX7 (DSHB), rat–anti-CHERRY (Invitrogen, M11217), rabbit–anti-SOX9 (Merck, AB5535), rabbit–anti-Laminin (Sigma-Aldrich, L9393), mouse–anti-MHC [fluorescein isothiocyanate (FITC)–coupled, DSHB], and rabbit–anti–p-SMAD2 (CST, 18338 T). After a PBST wash, the sections were incubated for 2 hours at room temperature with secondary antibodies and 4′,6-diamidino-2-phenylindole (Sigma-Aldrich, D9542). The secondary antibodies used were Cy3-AffiniPure donkey anti-rat immunoglobulin G (IgG, the Jackson ImmunoResearch, 712-165-153), Cy5-conjugated goat anti-rabbit IgG (Thermo Fisher Scientific, A10523) and Alexa Fluor 647 donkey anti-mouse IgG (the Jackson ImmunoResearch, 715-606-150). Sections were mounted with anti-fade mounting mediums (Bio-Rad, BUF058C) and imaged as described below.

In situ hybridization

RNA probes were synthesized using the T7 RNA Polymerase in vitro transcription kit (NEB M0251L) following the manufacturer’s instructions. RT-PCR products for RNA probe preparation were amplified using primers Tgfbr1-probe-F, Tgfbr1-probe-R, and Smad7-probe-F, Smad7-probe-R (listed in table S2). The sections were washed with PBST, followed by prehybridization and incubation in hybridization buffer containing Tgfbr1 or Smad7 RNA probes (660 ng ml⁻¹). The sections were sequentially washed with SSC (Sangon, B548110-0200) and TNE buffer, then treated with ribonuclease A (RNase A, Sigma-Aldrich, R4642), and further washed with mAb and blocking buffer. Subsequently, they were incubated with anti–digoxigenin-AP antibody prepared in blocking buffer. Signal was detected with BM purple solution after washes with mAb and AP buffer. The reaction was stopped using cold PBS containing 1 mM EDTA. Sections were mounted with an anti-fade mounting medium and imaged as described below.

Single-cell transplantation

Donor dTG-Pax7 axolotls (2.5 cm) were bathed in 3 μM 4-OHT twice for 1 day each time, with a 1-day interval between treatments. Two weeks post–tamoxifen treatment, muscle tissues from the tail region between the 11th and 13th myotomes posterior to the cloaca were collected for isolation of converted CHERRY⁺ MuSCs using collagenase (Sigma-Aldrich, C7657-25MG). Under an Olympus Stereomicroscope (SZX16), CHERRY⁺ single cells were picked using a pulled capillary pipette (Sutter, BF120-69-10) and then carefully transplanted into the incision at the 12th myotome posterior to cloaca of the host d/d animals. Transplanted d/d hosts were placed in a wet petri dish for 30 min to allow for settlement of the transplanted single cell and then transferred to clean water. After 2 days of recovery, hosts were amputated within 500 μm posterior to the transplanted CHERRY⁺ cells. Reciprocal grafting of limb and tail MuSCs followed the same procedure. When regeneration was complete, regenerates containing CHERRY⁺ cells were imaged and collected for further analysis.

Cell culture

NIH3T3 cells were cultured at 37°C, 5% CO₂, and 95% humidity in Dulbecco’s Modified Eagle’s Medium (Gibco, C11995500BT) supplemented with 10% fetal bovine serum (Hyclone, SH30396.03) and 1% penicillin-streptomycin (Gibco, 15140-122). Cells were split to 30% density using 0.25% trypsin (Gibco, 25200-072) and transfected with Tol2-CAGGS:loxp-BFP-nls-STOP-loxp-caTgfbr1/dnTgfbr1/Smad7-T2A-mCherry-PA alone or cotransfected with CAGGS:Cre plasmid using Lipofectamine (Thermo Fisher Scientific, L3000-015) at 70% confluency following the manufacturer’s guidance. The expression of the transfected constructs and the efficiency of Cre-mediated recombination were evaluated by fluorescence imaging. The effects of caTgfbr1, dnTgfbr1, and Smad7 on the TGF-β pathway were investigated by Western blotting.

Western blotting

Proteins were extracted from transfected cells or SB-431542–treated axolotl tissues using a commercial lysis buffer (Affinibody, AIWB-012). After centrifugation at 12,000 rpm at 4°C for 20 min, the supernatant was collected, and protein concentration was measured by the BCA protein quantification kit (Vazyme, E112-01) following the manufacturer’s protocol. A total of 30 μg of protein per group was loaded to 4 to 12% SDS–polyacrylamide gel electrophoresis gel. Gel electrophoresis was conducted at 160 V for 60 min. Proteins were transferred to polyvinylidene difluoride membrane (Immobilon, IPVH00010) via semidry transfer at 20 V for 15 min. The membrane was blocked in 5% bovine serum albumin (Sigma-Aldrich, V90933) for 1 hour at room temperature. Primary antibodies used included rabbit anti–p-SMAD2 (1:1000, CST, 18338 T), rabbit–anti-SMAD2 (1:1000, CST, 5339 T) and mouse anti–glyceraldehyde-3-phosphate dehydrogenase (1:25,000, Sigma-Aldrich, G8795), which were incubated at 4°C overnight. After thoroughly washing, the membranes were incubated with secondary antibodies rat anti–mouse–horseradish peroxidase (HRP, 1:10,000, the Jackson ImmunoResearch, 115-035-068) or donkey anti–rabbit-HRP (1:10,000, the Jackson ImmunoResearch, 711-035-152) for 1 hour at room temperature. Then, after TBST (0.1% Tween) washing for three times, the signal was visualized using the NcmECL Ultra kit (Ncmbio, P10100), and images were acquired using the Touch Imager (E-BLOT).

Imaging

Bright-field and fluorescence images of axolotls and NIH3T3 cells were acquired using an Olympus stereomicroscope (SXZ16) equipped with a fluorescent illuminator (FLuoCa). Fluorescence images of sections and cells were captured with a confocal microscope (LSM980), using 10× or 20× objectives. In situ hybridization sections were imaged with a slice scanner (Danjier, Pannoramic MIDI).

Statistical analysis

At least three independent experiments were performed for each quantification. All quantifications were performed with at least three animals, with a minimum of three representative and distinct sections being imaged and counted per animal. Immunohistochemistry was carried out on a series of cross sections from each axolotl tail, starting from the 12th tail myotome (myotome immediately caudal to the cloaca is defined as the first myotome) toward caudal direction, with 1-mm intervals between each section. Statistical analysis was performed using GraphPad Prism 9.0.0 software. The results were displayed as mean value ± SEM. The two-tailed unpaired Student’s t test was used to compare means between two independent groups. For comparisons among multiple groups, one-way analysis of variance (ANOVA) was applied, followed by Dunnett’s test or Fisher’s least significant difference post hoc test where appropriate. P < 0.05 was considered statistically significant.

Animal preparation and cell dissociation for smart-seq2

Tails of 3 to 4 cm dTG-Pax7 axolotls were amputated at the 12th myotome posterior to the cloaca, collected as 0 days postamputation samples. Regenerating tails were sampled from approximately one myotome anterior to the amputation plane at various stages: 2, 4, 6, 10, 15, and 33 days postamputation. Embryos were treated with 1 μM 4-OHT from stage 15 to 25, and their tails were collected at stage 35. Juvenile tail samples were sheared using a surgical scissor after a brief wash with 0.8× PBS, then transferred to 1 ml of dissociation solution (1× Liberase, 0.1 U μl⁻¹ deoxyribonuclease I, 0.8× PBS) and incubated in a 37°C water bath for 50 min with gentle shaking every 5 min. Embryonic tails were incubated in the same solution for 15 min. After complete dissociation, the cell suspension was filtered through MACS SmartStrainers 70 μm (Miltenyi Biotec, 130-098-462) to remove larger tissue debris, followed by centrifugation at 300g, 4°C for 5 min to reduce blood cells. The resulting cells were resuspended in 500 μl of 0.8× PBS for scRNA-seq.

FACS and Smart-seq2 sequencing

The cell suspension was processed with a FACS machine (BD FACSAria III), and single CHERRY⁺ cells were isolated into 384-well plates. Total RNA from CHERRY⁺ cells was extracted with 0.4% Nonidet P40 Substitute (Roche, 11332473001), followed by reverse transcription, cDNA fragmentation, and amplification. The amplified PCR products were purified and sequenced using the Illumina NovaSeq 6000 sequencer (35).

Sample collection and spatial transcriptomics experiment

Stage 25 embryos were collected, embedded in chilled Tissue-Tek OCT (4583, Sakura, Torrance, CA), rapidly frozen in liquid nitrogen–cooled isopentane, and stored at −80°C. Tissue sections, 20 μm thick, were collected on the Stereo-seq chip surface and fixed in methanol at −20°C for 30 to 40 min. The sections were stained with ssDNA mixture consisting of 1× Qubit ssDNA reagent (Invitrogen, Q10212) and RNase inhibitor (0.05 U ml⁻¹) diluted in 5× SSC. The nuclear signal and track lines were captured by a Motic, Custom PA53 FS6 Scan microscope with a 10× objective under the FITC channel. Sections were then permeabilized using 0.1% pepsin (Sigma-Aldrich, P7000) in 0.01 M HCl buffer at 37°C for 18 min. RNA released from the permeabilized tissue was subjected to reverse transcription, followed by cDNA fragmentation and amplification. The resulting PCR products were purified for DNA Nanoball generation and sequenced on the MGI DNBSEQ-T10 sequencer. Detailed experimental procedures were previously described (40, 42).

Downstream scRNA-seq analysis

Smart-seq2 single-cell data process pipeline

Quality control of the Smart-seq2 single-cell data was performed using fastqc (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) based on the nf-core pipeline guidelines (69). The fastq data were then aligned to the Axolotl genome (AmexG_v6.0-DD) and the transcriptome version of AmexT_v47 using STAR (2.7.10b) (70). Gene expression levels were quantified by featureCounts (1.22.2) (71) and transformed into a Seurat (4.1.0) object (72). Cells expressing fewer than 4000 genes were filtered out and then normalized using the SCTransform function. Nearest neighbors were identified through PCA embedding computation, and a UMAP was constructed on the basis of the first 30 principal components. Single-cell clustering was performed using the Louvain algorithm (36) with the resolution parameter set to 0.8. Last, marker genes for each cluster were identified using the FindAllMarkers with min, pct set to 0.25, and logfc. threshold set to 0.25.

RNA velocity analysis and in silico genetic perturbation

scRNA velocity analysis was performed by Dynamo (1.3.2) (38). Reads of Smart-seq2 were aligned to the axolotl genome via STAR, and spliced, unspliced, and spanning reads were annotated by “velocyto run-smartseq2” (0.17.16) in each cell (73). Loom files were combined by applying the combine function in the Loompy package (3.0). The dataset was bifurcated into early regeneration (0, 2, and 4 days postamputation) and late regeneration (6, 10, 15, and 33 days postamputation) stages. The anndata object was constructed using Dynamo’s read_loom function. The single-cell metadata and UMAP information within the anndata object facilitated consistent display of plots with single-cell data. The dynamo object from an anndata was created using the dynamo.pp.recipe_monocle function by top 3000 genes, incorporating essential genes achieved through the ‘genes_to_append’ parameter. For RNA velocity analysis, various functions were used, including dynamo.tl.dynamics, dynamo.tl.moments, dynamo.tl.reduceDimension, dynamo.tl.gene_wise_confidence, and dynamo.tl.confidence_cell_velocities. Directional visualization of pseudo-lineage transitions was performed by the streamline_plot function. Changes in primary cell flow direction following suppression or activation of the TGF-β signaling pathway were predicted by dynamo.pd.perturbation function.

Single-cell trajectories reconstruction

Cell differentiation trajectories in the primary cell types during the early regeneration stage were generated by Monocle2 (2.24.1) (39). Expression data and metadata from the relevant cells were extracted by the subset function, and the monocle2 object was constructed by the newCellDataSet function. To identify genes associated with pseudotime, the differentialGeneTest function was used and genes with a q value <10 ×⁻⁴ were selected for downstream analysis. For dimensionality reduction, the DDRTree approach was reinforced by corrected gene counts and the percentage of mitochondrial RNA using the residualModelFormulaStr parameter. Cell sorting was performed by the orderCells function with MuSCs as the root state. The BEAM function was applied to identify pseudotime genes in designated cell types and two transitional paths, gene expression of both paths across pseudotime was visualized by plot_genes_branched_heatmap. GO enrichment for genes in each hierarchical cluster was analyzed via DAVID (https://davidbioinformatics.nih.gov/).

Cell differentiation status and pathway activity score prediction

CytoTRACE (0.3.3) was used to predict the single-cell differentiation state (37). The top 200 most positively correlated genes were selected as gene count signatures (GCSs), and the estimate of the GCS vector was then iteratively refined by leveraging local similarity between cells and applying a two-step smoothing procedure. Such a process resulted in a score between 0 and 1 for each cell, representing the predicted order of cells by their relative differentiation status, where values reflect pluripotency. The CytoTRACE function was invoked with the expression matrix as input, which was normalized to counts per million. PROGENy R package was used to predict pathways activity scores in different cell types (45), and activation of 14 pathways was retrieved. The expression matrix was normalized using the SCTransform function, and the axolotl gene index was converted to human orthologous genes while specifying the species as “Human.”

IM subcells identification

The IntegrateData function of Seurat was applied to predict the progeny of stem cells (IM1 and IM2) by integrating single-cell data from multiple sample stages (74), based on the anticipation that a stem cell undergoing differentiation will display partially similar gene expression patterns to its descendant cell types. IM1 and IM2 cells were divided into subsets according to their distinct differentiation tendencies. Data integration was using the first 20 principal components and 3000 highly variable genes, with the clustering resolution set to 0.7.

Constellation plot analysis

A Constellation plot (75) was made to illustrate the interconnections among cell types. Node size reflects the cell number proportion of the respective cell type, and the position coordinates for the corresponding cell type were determined by the UMAP centroid of the respective cell cluster. Edges in the plot represent relationships between two nodes, where wider edges indicate a closer relationship. The edge width at the node equals to the proportion of the nearest neighboring cells belonging to the opposite node, with the standard of identifying the 30 nearest neighbors by the top 50 principal components of a single cell as input to RANN::nn2. Only edges with a fraction greater than 0.02 were displayed.

Downstream spatial transcriptomics data analysis

Spatial transcriptomics raw data processing

Stereo-seq sequencing files were generated from MGI DNBSEQ-T1 sequencer. Read 1 contained coordinate identity (CID) and molecular identity (MID) sequences [CID: 1 to 25 base pair (bp), MID: 26 to 35 bp], while read 2 contained the cDNA sequence with 100 base pairs. Standard raw data were processed via SAW (76) (Standard Analysis Workflow) (v5.5.3). Reads with a length of less than 30 bp after trimming the Poly(A) sequence were filtered. The remaining clean reads were aligned to the AmexG_v6.0-DD genome (77). Reads overlapping more than 50% with exon regions were counted as exon transcripts. Reads overlapping less than 50% with exon regions but overlapping with the adjacent intron sequences were annotated as intron transcripts, otherwise as intergenic transcripts. Reads mapped to the transcriptome but on the opposite strand of their annotated gene were counted as antisense transcripts. Unique molecular identifiers (UMIs) with the same CID and gene locus were collapsed, allowing a 1-bp mismatch for PCR errors. Then, exon reads were used to generate CID-containing expression profile matrices. Register and extraction of tissue coverage areas were performed by the register and tissuecut model of SAW (76). Last, expression matrices within tissue regions were specified for further analysis.

Cell segmentation of spatial transcriptomics data

The ssDNA staining images were used to identify cell regions of the Stereo-seq section (40, 42). CellProfiler (v4.2.6) was applied to automatically recognize and analyze signal strength for single-cell feature extraction (78). The range of pixel units was set between 16 and 60 to fit different cell sizes globally. A lower bound of 0.2 was applied to treat weaker signals as noise and discard them. Expression matrices were generated in each labeled cell of each section.

Mesoderm lineage judgment

Stereo-seq was used on axolotl embryo samples to elucidate the lineage origin of Pax7⁺ cells during embryonic development. On the basis of nuclear staining images, mesoderm regions in embryos were identified. Cells from the spatial transcriptomics data within these regions, as well as cells expressing the neural crest marker Sox10, were extracted for further analysis. The Seurat IntegrateData function was applied to integrate Smart-seq2 and Stereo-seq data, where 4000 feature genes and the first 30 principal components were incorporated. RCTD was used to map the annotated Smart-seq2 data with the R package spacexr (v2.2.1) (41). EPrC cells were identified and then spatially visualized using a custom script.

Supplementary Materials

This PDF file includes:

Figs. S1 to S25

Tables S1 and S2