Pdgfab/Pdgfra-mediated chemoattraction guides the migration of sclerotome-derived fibroblast precursors in zebrafish

Abstract

In vertebrates, the sclerotome is a transient embryonic structure that gives rise to various tissue support cells, including fibroblasts. However, how fibroblast precursors are guided to diverse tissues remain poorly understood. Using zebrafish, our lab has previously shown that sclerotome-derived cells undergo extensive migration to generate distinct fibroblast subtypes, including tenocytes along the myotendinous junction and fin mesenchymal cells in the fin fold. The pan-fibroblast gene platelet-derived growth factor receptor a (pdgfra), which has been implicated in cell migration across various contexts, is specifically expressed in the sclerotome and its descendants. Loss of functional Pdgfra in a pdgfra gene-trap mutant results in severe defects in the migration of sclerotome-derived cells, leading to a dose-dependent loss of tenocytes and fin mesenchymal cells. By combining live imaging and mosaic labeling with a membrane-bound dominant-negative tool, we demonstrate that Pdgfra acts cell-autonomously to regulate the migration of sclerotome-derived cells. In the absence of ligand pdgfab, which is expressed in the medial somite, sclerotome-derived cells fail to migrate medially, resulting in a loss of tenocytes, although they can migrate normally toward the fin fold and generate fin mesenchymal cells. Strikingly, localized expression of Pdgfab in pdgfab mutants can direct the migration of sclerotome-derived cells to both normal and ectopic locations, suggesting a chemoattractive role for the Pdgfab ligand. Together, our results demonstrate that Pdgfab/Pdgfra-mediated chemoattraction guides the migration of sclerotome-derived fibroblast precursors to specific locations, where they diversify into distinct fibroblast subtypes.

During vertebrate development, fibroblast precursors migrate from the sclerotome to specific tissues, but the mechanisms that guide these precursors remain unclear. This study shows that Pdgfab/Pdgfra-mediated chemoattraction directs sclerotome-derived cells to distinct locations, enabling their differentiation into specialized fibroblast subtypes in zebrafish.

Introduction

Almost every organ in our body contains connective tissue, with fibroblasts being the main cellular component. Known primarily for producing the extracellular matrix, fibroblasts also perform various other functions, including secreting growth factors, creating niches for stem cells, modulating immune response, and contributing to wound healing [1]. The involvement of fibroblasts in such a wide range of processes raises the question of whether they are a single cell type. Thanks to single-cell sequencing technologies, many groups have reported that fibroblasts are substantially heterogeneous in many organs, including the heart, skin, skeletal muscle, intestine, bladder, and synovial joints [2–5]. Although fibroblasts arise from all three germ layers during embryonic development [6], little is known about the molecular mechanisms controlling how fibroblast precursors migrate to distinct locations and differentiate into diverse fibroblast subtypes.

The somite is a transient embryonic structure that gives rise to the trunk of vertebrates. In amniotes, the sclerotome is a sub-compartment formed in the ventral part of the somite [7]. Its development depends on signaling molecules derived from the notochord, with Sonic Hedgehog (SHH) being a key factor for its induction [8]. The sclerotome compartment later gives rise to the axial skeleton, including the vertebral body, ribs, and neural arches, as well as their associated tendons and ligaments [9]. Our previous studies show that the zebrafish sclerotome has a unique bipartite organization in the ventral and dorsal parts of the somite, specifically marked by the expression of nkx3.1 [10]. Live imaging, lineage tracing, and single-cell RNA sequencing experiments demonstrate that sclerotome progenitors undergo extensive migration to generate diverse fibroblast subtypes, including tenocytes, perivascular fibroblasts, and fin mesenchymal cells [10–12]. Although dispensable for sclerotome induction, active SHH signaling is required for the migration of sclerotome-derived cells toward the notochord [10]. However, the exact mechanisms controlling this migration remain unexplored.

One molecular pathway often associated with fibroblasts is the platelet-derived growth factor (PDGF) signaling pathway. This pathway comprises four secreted ligands (PDGFA, PDGFB, PDGFC, and PDGFD) and two tyrosine kinase receptors (PDGFRA and PDGFRB) [13,14]. PDGFRA is known as a pan-fibroblast marker [15] and controls the migration of various cells during embryonic development, such as cranial and cardiac neural crest cells in mouse and zebrafish [16,17], myocardial precursors in zebrafish [18,19], and mesendodermal cells in Xenopus and zebrafish [20,21]. All these migratory events are mediated by the ligand PDGFA, which is thought to function as a chemoattractant for the PDGFRA-expressing cells, although this has been inferred from its expression pattern. Homozygous Pdgfra knock-out mice die at E14-E16 with a cleft palate phenotype, consistent with the role of Pdgfra in neural crest cell migration [22]. These Pdgfra mutants also display various defects in vertebral neural arches, which are derived from the sclerotome [22,23]. In contrast, homozygous Pdgfa knock-out mice are viable but die by week 6 postnatally, likely due to respiratory issues [24]. Pdgfa mutants display early somite patterning defects similar to those observed in Pdgfra null mice and Pdgfa somite expression promotes rib and vertebral development [25]. These findings suggest that the PDGFA/PDGFRA signaling pathway plays a conserved role in regulating cell migration and chemotaxis across diverse cell populations, including during the development of sclerotome-derived lineages. A similar function is observed in Drosophila, where the PDGF/VEGF receptor homolog PVR directs the migration of border cells in response to its ligand, PDGF/VEGF factor 1 (PVF1), during ovarian development [26,27].

In this study, we describe the critical role of Pdgfra signaling in guiding the migration of sclerotome-derived fibroblast precursors. By combining mutant analysis with in vivo imaging, we demonstrate that pdgfra is required cell-autonomously for the migration of sclerotome-derived cells to different locations, where they differentiate into various fibroblast subtypes. The ligand pdgfab is expressed in the medial somites in response to SHH signaling. Loss of pdgfab specifically impairs the migration of sclerotome-derived cells toward the notochord. Mosaic expression analysis further reveals that the Pdgfab ligand functions as a chemoattractant, guiding the migration of pdgfra-expressing cells. Our study suggests that region-specific expression of the Pdgfab ligand provides a localized cue to direct the migration of Pdgfra-expressing fibroblast precursors, thereby facilitating fibroblast diversification.

Results

Characterization of a pdgfra gene-trap mutant

We previously showed that the zebrafish sclerotome has a bipartite organization occupying both the ventral and dorsal region of each somite along the trunk [10]. Cells originating from both sclerotome domains undergo extensive migration, with their migratory directions influencing cell fate and contributing to the diversification of distinct fibroblast populations [10,28]. Double fluorescent in situ hybridization (FISH) revealed that pdgfra, encoding PDGF receptor a, was co-expressed with sclerotome marker nkx3.1 in both dorsal and ventral sclerotome domains, as well as sclerotome-derived cells around the notochord at 24 hours post-fertilization (hpf) (Fig 1A). This result is consistent with our previous finding that pdgfra is broadly expressed across all fibroblast subtypes derived from the sclerotome [12]. Since pdgfra has been shown to regulate the migration of cardiomyocytes [18] and cranial neural crest cells [16] in zebrafish, we asked whether pdgfra is also required for the migration of sclerotome-derived cells. To test this, we characterized a pdgfra gene-trap line [19] with an insertion in the intron upstream of exon 17, which encodes the transmembrane domain (Fig 1B). This gene-trap line is predicted to produce a nonfunctional truncated Pdgfra protein, lacking the transmembrane and kinase domains, fused with mRFP (PdgfraΔTK-mRFP) (Fig 1B). We predicted that PdgfraΔTK-mRFP can bind to Pdgfra ligands in the extracellular space but is unable to signal due to the absence of transmembrane and kinase domains, thereby functioning as a dominant-negative decoy receptor. The gene-trap line is also predicted to generate ubiquitously expressed GFP fused with the Pdgfra transmembrane and kinase domains under the control of the β-actin promoter (Fig 1B). For simplicity, we designate wild-type, heterozygous, and homozygous pdgfra fish as pdgfra^+/+, pdgfra^mRFP/+, and pdgfra^mRFP/mRFP. Using region-specific probes, we confirmed that mRFP showed an identical pattern as the pdgfra5′ probe in pdgfra^mRFP/mRFP fish, while staining with GFP and pdgfra3′ probes resulted in similar ubiquitous patterns in homozygous fish (S1A and S1B Fig).

Fig 1. pdgfra is required for the migration of sclerotome-derived cells.

(A) Wild-type embryos at 24 hpf were co-labeled with pdgfra (cyan) and nkx3.1 (magenta). pdgfra and nkx3.1 are co-expressed in the dorsal sclerotome (short arrows), ventral sclerotome (long arrows), and sclerotome-derived notochord-associated cells (arrowheads). n = 12 embryos. (B) Schematics of the pdgfra gene-trap line. The exons of the pdgfra gene are color-coded: gray for extracellular coding exons, blue for the transmembrane coding exon, yellow for intracellular coding exons, and white for non-coding exons. The gene-trap insertion is flanked by two loxP sites and is located in the intron between exons 16 and 17. The gene-trap mutant is predicted to produce a secreted protein containing the extracellular domain of Pdgfra fused with mRFP (PdgfraΔTK-mRFP). (C) kdrl:EGFP; pdgfra^mRFP/+ embryos were injected with Alexa Fluor 647-conjugated ovalbumin (OVA-647) into the extracellular space at the blastula stage (4 hpf) and imaged at 55 hpf. Co-localization of OVA-647 (yellow) and PdgfraΔTK-mRFP (magenta) proteins in endothelial cells of the caudal vein plexus, labeled by the kdrl:EGFP reporter (cyan), are indicated by arrows. n = 20 embryos. (D) Expression analysis of sclerotome markers (nkx3.1 and pax9) and sclerotome-derived notochord-associated cell markers (nkx3.2 and pax1a) in embryos at 24 hpf from crosses of pdgfra^mRFP/+ Fish. The expression of all markers shows a dose-dependent reduction in the migrating sclerotome-derived cells (arrowheads) around the notochord across different mutant backgrounds. n = 25 embryos per staining. (E) Quantification of nkx3.1 expression area in the experiments showed in (D). The percentage of nkx3.1-stained area was measured within the 8-somite region located above the yolk extension. n = 14 (pdgfra^+/+), 23 (pdgfra^mRFP/+), and 14 (pdgfra^mRFP/mRFP) embryos. Data are plotted as mean ± SD. Statistics: One-way ANOVA with Tukey’s multiple comparison test. Asterisks representation: p < 0.001 (***) and p < 0.0001 (****). (F) pdgfra^mRFP/mRFP embryos injected with Cre mRNA at the one-cell stage, along with their uninjected controls, were stained for mRFP and nkx3.1 expression at 24 hpf. Cre-expressing pdgfra^mRFP/mRFP embryos show a complete loss of mRFP expression and restoration of migration of nkx3.1⁺ sclerotome-derived cells (arrowhead) at 24 hpf. n = 25 embryos per staining. (G) Expression analysis of embryos at 24 hpf from crosses of pdgfra^ref/+ fish. In pdgfra^ref/ref embryos, expression of both the sclerotome marker nkx3.1 and the sclerotome-derived notochord-associated cell marker nkx3.2 is absent in the migrating sclerotome-derived cells (arrowheads) around the notochord. n = 20 embryos per staining. Data for this figure are provided in S1 Data. Scale bars: (A, D, F, G) 100 μm; (C) 50 μm.

We next performed live imaging to analyze the distribution of different fusion proteins generated from the gene-trap line. While the GFP fusion protein was undetectable, we found that PdgfraΔTK-mRFP puncta were specifically enriched in endothelial cells of the caudal vein plexus (CVP), labeled by kdrl:EGFP (Fig 1C). Since endothelial cells do not normally express pdgfra and PdgfraΔTK-mRFP is predicted to be secreted, we hypothesized that scavenger endothelial cells [29] phagocytose the secreted PdgfraΔTK-mRFP fusion protein from the extracellular space. To test this possibility, we injected Alexa Fluor 647-conjugated ovalbumin (OVA-647) into the extracellular space of pdgfra^mRFP/+; kdrl:EGFP embryos at 4 hpf. Both PdgfraΔTK-mRFP and OVA-647 proteins displayed a similar puncta pattern in CVP endothelial cells at 55 hpf (Fig 1C). Together, our results suggest that the pdgfra gene-trap line produces a secreted PdgfraΔTK-mRFP protein, which is cleared from the extracellular space by scavenger endothelial cells.

pdgfra mutants show defects in the migration of sclerotome-derived cells

To determine the role of pdgfra in the migration of sclerotome-derived cells, we analyzed the expression pattern of sclerotome markers, nkx3.1 and pax9, in pdgfra gene-trap mutants at 24 hpf. Although both markers were normally expressed in the dorsal and ventral sclerotome domains, the expression of nkx3.1 and pax9 in sclerotome-derived notochord-associated cells was completely lost in pdgfra^mRFP/mRFP fish (Fig 1D). pdgfra^mRFP/+ fish also displayed a substantial reduction of nkx3.1 and pax9 staining around the notochord (Fig 1D), consistent with the prediction that PdgfraΔTK-mRFP likely functions as a dominant-negative decoy receptor. Indeed, quantification of nkx3.1 expression area in pdgfra^+/+, pdgfra^mRFP/+, and pdgfra^mRFP/mRFP embryos showed a significant reduction in a dose-dependent manner (Fig 1E). The sclerotome-derived cells around the notochord can be specifically labeled by the expression of nkx3.2 and pax1a [10]. We observed a similar dose-dependent reduction in the expression of both markers in the pdgfra mutant background (Fig 1D). The reduced migration could potentially result from increased cell death due to the loss of Pdgfra signaling in sclerotome cells. To test this possibility, we stained for the apoptotic marker cleaved caspase-3 (CC3) in pdgfra^mRFP mutant embryos in the background of the sclerotome reporter nkx3.1:Gal4; UAS:Kaede. We observed no significant change in the number of CC3⁺ cells in the trunk region or among Kaede⁺ sclerotome cells (S1C and S1D Fig), suggesting that Pdgfra signaling is not required for the survival of these cells. To further demonstrate that the migration defect is caused by the pdgfra gene-trap insertion, we injected Cre mRNA into one-cell stage pdgfra^mRFP/mRFP embryos to remove the loxP-flanked gene-trap cassette (Fig 1B). Early Cre expression resulted in the complete loss of mRFP expression in pdgfra^mRFP/mRFP fish (Fig 1F), suggesting that both copies of the insertion were successfully excised. Removal of the gene-trap insertion led to a substantial rescue of nkx3.1⁺ sclerotome-derived cells around the notochord in pdgfra^mRFP/mRFP fish (Fig 1F).

To validate our observations using the pdgfra gene-trap line, we analyzed a different pdgfra mutant known as refused-to-fuse (ref), which carries a point mutation that results in premature stop codon upstream of the transmembrane domain of pdgfra [18]. In pdgfra^ref/ref mutants at 24 hpf, the migration of sclerotome-derived cells was completely abolished, as indicated by the absence of nkx3.1 and nkx3.2 expression around the notochord (Fig 1G). In contrast to the migration defect observed in pdgfra^mRFP/+ embryos, pdgfra^ref/+ embryos exhibited normal migration of sclerotome-derived cells, indistinguishable from their wild-type siblings (Fig 1G), consistent with the recessive nature of the pdgfra^ref allele [18]. Together, our results suggest that sclerotome-derived cells require functional Pdgfra to migrate toward the notochord, and that the migration defects observed in pdgfra^mRFP/+ fish are likely due to the dominant-negative effect of the PdgfraΔTK-mRFP fusion protein, rather than haploinsufficiency.

Failed migration of sclerotome-derived cells impairs fibroblast development

The absence of sclerotome marker expression around the notochord suggests that cells fail to migrate out of the sclerotome domains. To directly visualize the migration of sclerotome-derived cells, we performed live imaging of pdgfra mutants from 22 to 36 hpf using the sclerotome reporter line, nkx3.1:Gal4; UAS:NTR-mCherry [10]. In pdgfra^+/+ controls, mCherry⁺ cells emerged from both the dorsal and ventral sclerotome domains and migrated extensively, occupying the medial region of the entire trunk (Fig 2A and S1 Movie). By contrast, both pdgfra^mRFP/+ and pdgfra^mRFP/mRFP embryos displayed severely reduced migration of mCherry⁺ cells, resulting in a marked reduction of cells surrounding the neural tube and notochord (Fig 2A and S1 Movie). These results suggest that pdgfra is required for the migration of sclerotome-derived cells.

Fig 2. pdgfra mutants show defects in the generation of sclerotome-derived cells.

(A) Snapshots from time-lapse imaging of the nkx3.1:Gal4; UAS:NTR-mCherry reporter in pdgfra^+/+, pdgfra^mRFP/+, and pdgfra^mRFP/mRFP backgrounds from 22 to 36 hpf. While mCherry⁺ cells (arrowheads) migrate normally in wild-type siblings, both heterozygous and homozygous pdgfra mutants display substantially reduced migration, with pdgfra^mRFP/mRFP mutants being the most affected. Note that the PdgfraΔTK-mRFP fusion proteins appear as bright dots (arrows) within the vasculature of pdgfra^mRFP/+ and pdgfra^mRFP/mRFP embryos. The corresponding movies are shown in S1 Movie. Time stamps are indicated in the hh:mm format. Transverse views of the final frames are shown on the right side for each genotype, with the neural tube (nt) and notochord (n) outlined by dotted lines. These are projected images of a 50 μm region centered on the dashed line. n = 4 embryos per group. (B) Expression of tenocyte markers (prelp and scxa), fin mesenchymal cell (FMC) marker (fbln1), and apical epidermal cell (AEC) marker (fras1) in pdgfra^+/+, pdgfra^mRFP/+, and pdgfra^mRFP/mRFP embryos at 48 hpf. The expression of both tenocyte (short arrows) and fin mesenchymal cell (notched arrowheads) markers is reduced in the pdgfra mutant background, while the apical epidermal cell marker (long arrows) remains unchanged. n = 25 embryos per staining. (C) Quantifications of tenocytes using the nkx3.1:Gal4; UAS:NTR-mCherry reporter in embryos with different pdgfra genotypes at 48 hpf. Each dot represents the average number of mCherry⁺ tenocytes along one MTJ, based on counts from three junctions of a single embryo. n = 7 (pdgfra^+/+), 10 (pdgfra^mRFP/+), and 8 (pdgfra^mRFP/mRFP) embryos. (D) Quantifications of fin mesenchymal cells in the major fin fold of embryos with different pdgfra genotypes at 48 hpf. Each dot represents the total number of fin mesenchymal cells from both the dorsal and ventral fin folds in the 8-somite region posterior to the end of the yolk extension in one embryo. n = 24 (pdgfra^+/+), 17 (pdgfra^mRFP/+), and 9 (pdgfra^mRFP/mRFP) embryos. All data are plotted as mean ± SD. Statistics: One-way ANOVA with Tukey’s multiple comparison test. Asterisks representation: p < 0.0001 (****). (E) Expression analysis of embryos at 48 hpf from crosses of pdgfra^ref/+ fish. The expression of both tenocyte marker prelp (short arrows) and fin mesenchymal cell (FMC) marker fbln1 (notched arrowheads) is reduced in pdgfra^ref/ref mutants. n = 20 embryos per staining. Data for this figure are provided in S1 Data. Scale bars: 100 μm.

Based on our previous work [28], we predicted that the sclerotome progenitors need to migrate out of the sclerotome domains to differentiate into distinct fibroblast subtypes. To test this, we stained pdgfra mutants at 48 hpf with fibroblast subtype-specific markers, including prelp and scxa for tenocytes and fbln1 for fin mesenchymal cells. In pdgfra^+/+ controls, tenocytes expressing prelp and scxa were present along the entire V-shaped myotendinous junction (MTJ) (Fig 2B). By contrast, pdgfra^mRFP/+ embryos almost completely lacked tenocytes in the dorsal half of the MTJ, while pdgfra^mRFP/mRFP fish showed further reduction of tenocytes, with only some remaining in the very ventral region of the MTJ (Fig 2B). Quantification using nkx3.1:Gal4; UAS:NTR-mCherry showed a dose-dependent decrease in tenocyte number across different mutant backgrounds (Fig 2C), reminiscent of our results with sclerotome markers (Fig 1D and 1E). Similar to tenocytes, we also observed a dose-dependent loss of fbln1⁺ fin mesenchymal cells in the fin folds (Fig 2B and 2D). In contrast to fbln1, we observed no difference in the staining of fras1, a marker for apical epidermal cells that are not sclerotome-derived (Fig 2B), suggesting that the fin fold compartment is formed normally. Similar to pdgfra^mRFP/mRFP embryos, pdgfra^ref/ref mutants showed a substantial reduction in prelp⁺ tenocytes and fbln1⁺ fin mesenchymal cells at 48 hpf (Fig 2E). Together, our results suggest that the migration defect caused by the loss of pdgfra compromises the generation of sclerotome-derived fibroblasts, including tenocytes and fin mesenchymal cells.

Pdgfra functions cell-autonomously in regulating the migration of sclerotome-derived cells

To validate the pdgfra mutant results and assess the autonomy of Pdgfra in the migration of sclerotome-derived cells, we generated two dominant-negative forms of the receptor. In the first construct, we used the heat shock promoter to drive the expression of the extracellular domain of Pdgfra fused with EGFP (hsp:pdgfraΔTK-EGFP), mimicking the secreted PdgfraΔTK-mRFP decoy receptor produced from the gene-trap line (Fig 3A). As a control, we expressed secreted GFP under the heat shock promoter (hsp:sec-GFP) [30] (Fig 3B). Induction of the secreted PdgfraΔTK-EGFP decoy receptor, but not the secreted EGFP, at 18 hpf, effectively abrogated the migration of nkx3.1⁺ sclerotome-derived cells at 24 hpf (Fig 3B and 3C). Next, we asked if expressing a membrane-tethered dominant-negative Pdgfra receptor would have the same effect. We generated a transgenic line, UAS:pdgfraΔK-EGFP, in which the UAS promoter controls the expression of a truncated Pdgfra with the kinase domain replaced by EGFP (Fig 3A). nkx3.1:Gal4; UAS:NTR-mCherry; UAS:pdgfraΔK-EGFP embryos showed a complete absence of mCherry⁺ sclerotome-derived cells around the neural tube and notochord in the medial trunk at 56 hpf (Fig 3D). These results suggest that a functional Pdgfra receptor is required for the migration of sclerotome-derived cells, confirming our pdgfra mutant results.

Fig 3. pdgfra is required cell-autonomously for sclerotome-derived cell migration.

(A) Schematics of endogenous Pdgfra protein and two different loss-of-function variants. Compared to wild-type Pdgfra protein (top), PdgfraΔTK-EGFP (middle) is a secreted decoy receptor with its transmembrane and cytoplasmic domains replaced by EGFP, while PdgfraΔK-EGFP (bottom) is a membrane-bound dominant-negative receptor with only the cytoplasmic domain replaced by EGFP. (B) Schematics of overexpression experiments. Wild-type embryos were injected at the one-cell stage with either hsp:sec-GFP or hsp:pdgfraΔTK-EGFP plasmids, heat shocked at 18 hpf, and stained and imaged at 24 hpf. (C) Expression of nkx3.1 in embryos at 24 hpf after overexpression of either secreted GFP or PdgfraΔTK-EGFP. Ectopic expression of PdgfraΔTK-EGFP, but not secreted GFP, impairs the migration of nkx3.1⁺ cells from the sclerotome (arrowhead). n = 20 embryos per staining. (D) Transgenic expression of PdgfraΔK-EGFP (cyan) in nkx3.1:Gal4; UAS:NTR-mCherry (magenta) background at 56 hpf. mCherry⁺ EGFP⁺ cells are mostly restricted in the dorsal or ventral regions of the trunk. A transverse view of a 50 μm region centered on the dashed line is shown on the right, with the neural tube (nt) and notochord (n) outlined by dotted lines. n = 5 embryos. (E) Mosaic expression of Lyn-EGFP or PdgfraΔK-EGFP (cyan) in nkx3.1:Gal4; UAS:NTR-mCherry (magenta) embryos at 50 hpf. In Lyn-EGFP-expressing embryos, EGFP⁺ cells are widely distributed in the trunk both around the notochord (arrowheads) and in the fin fold (notched arrowheads). By contrast, in PdgfraΔK-EGFP-expressing embryos, while mCherry⁺EGFP⁻ cells can be found around the notochord (arrowheads) and in the fin fold (notched arrowheads), EGFP⁺ cells (arrows) are mostly restricted in the dorsal or ventral regions of the trunk. Transverse views of a 50 μm region centered on the dashed line are shown on the right, with the neural tube (nt) and notochord (n) outlined by dotted lines. The corresponding confocal stacks are shown in S2 Movie. n = 10 embryos per group. (F) Quantification of the percentage of EGFP⁺ cells around the notochord from the experiments shown in (E). n = 7 (UAS:lyn-EGFP) and 16 (UAS:pdgfraΔK-EGFP) embryos. Data are plotted as mean ± SD. Statistics: Unpaired t test with Welch’s correction. Asterisks representation: p < 0.001 (***). Data for this figure are provided in S1 Data. Scale bars: 100 μm.

If Pdgfra is required cell-autonomously for the migration of sclerotome-derived cells, we would expect that cells expressing PdgfraΔK-EGFP fail to migrate, while neighboring non-expressing cells migrate normally. To test this directly, we performed low-dose plasmid injections to achieve mosaic expression of either UAS:lyn-EGFP (membrane-localized EGFP) or UAS:pdgfraΔK-EGFP in nkx3.1:Gal4; UAS:NTR-mCherry embryos. UAS:lyn-EGFP-injected fish showed a wide distribution of EGFP⁺ sclerotome-derived cells at 50 hpf, including cells around the notochord and fin mesenchymal cells in the fin fold (Fig 3E and S2 Movie). By contrast, in UAS:pdgfraΔK-EGFP-injected embryos, most of the mCherry⁺EGFP⁺ cells remained in either the ventral or dorsal region of the trunk, whereas mCherry⁺EGFP^- cells were distributed normally throughout the medial trunk (Fig 3E and S2 Movie). Compared to UAS:lyn-EGFP, expression of UAS:pdgfraΔK-EGFP resulted in a significant decrease in the percentage of EGFP⁺ cells around the notochord (Fig 3F). Together, these results suggest that Pdgfra acts cell-autonomously in controlling the migration of sclerotome-derived cells.

Hedgehog signaling regulates the migration of sclerotome-derived cells by controlling pdgfab expression

Mammalian PDGFRA receptor is regulated by PDGF ligands, including PDGFA and PDGFC [31]. To identify the ligands that regulate the migration of sclerotome-derived cells, we performed in situ hybridization to examine the expression pattern of the three candidate ligands, pdgfaa, pdgfab, and pdgfc. In wild-type embryos at 24 hpf, pdgfab displayed the most notable pattern in the medial trunk, with expression in the medial somites, motoneurons, and the hypochord (Fig 4A–4D). Meanwhile, pdgfaa was predominantly expressed in the dorsal spinal cord and the hypochord, while pdgfc was strongly expressed along the edge of the fin fold and in some anterior somites (Fig 4A). The expression of pdgfab in the medial region of the trunk suggests that Pdgfab might regulate the migration of sclerotome-derived cells to surround the notochord. Indeed, co-staining of pdgfab and pdgfra revealed a complementary pattern, where pdgfra⁺ sclerotome-derived cells were sandwiched between pdgfab⁺ somitic cells on one side and pdgfab-expressing motoneurons and hypochord on the other (Fig 4B). Double labeling with pdgfab and myod1, a myotome-specific marker, further confirmed that the medial part of the myod1⁺ myotome co-expressed pdgfab (Fig 4C).

Fig 4. Hh signaling controls sclerotome-derived cell migration by regulating the expression of pdgfab.

(A) Expression patterns of Pdgfra ligands in wild-type embryos at 24 hpf. pdgfaa expression is detected in the dorsal spinal cord (long arrow) and hypochord (arrowhead). pdgfab is expressed in the medial regions of the somites (arrow) and the hypochord (arrowhead). pdgfc expression is observed in anterior somites (arrow) and along the edge of the fin fold (notched arrowhead). n = 15 embryos per staining. (B–D) Wild-type embryos at 24 hpf were co-labeled with pdgfab (cyan) and pdgfra (B), myod1 (C), or ptc2 (D) (magenta). pdgfab is expressed in medial somites (arrows), motoneurons (arrowheads), and the hypocord (notched arrowheads). Sagittal views of merged channels and transverse views along the dashed lines—showing both individual and merged channels—are shown, with neural tube (nt) and notochord (n) indicated by dotted lines. n = 15 embryos per group. (B) pdgfab shows complementary expression to pdgfra. (C) pdgfab is expressed in the medial-most myod1⁺ somitic region (arrows) adjacent to the notochord. (D) pdgfab and ptc2 show co-expression in somitic cells (arrows) surrounding the notochord. (E) Expression analysis of pdgfab and nkx3.1 in wild-type embryos treated with DMSO or cyclopamine from 5 to 24 hpf. Cyclopamine treatment leads to near-complete loss of pdgfab expression in somites (arrow) and absence of nkx3.1⁺ sclerotome-derived cells (arrowhead) surrounding the notochord. n = 15 embryos per group. (F) hsp:pdgfab-2A-mCherry transgenic embryos and their wild-type siblings were heat shocked at 18 hpf and analyzed for pdgfab and nkx3.1 expression at 24 hpf. Compared to wild-type controls, hsp:pdgfab-2A-mCherry embryos show strong and ubiquitous pdgfab expression and impaired migration of nkx3.1-expressing cells (arrowhead) toward the notochord. n = 15 embryos per group. Scale bars: (A, E, F) 100 μm; (B, C, D) 50 μm (sagittal) and 25 μm (transverse).

The pdgfab expression in the medial somite is reminiscent of the pattern of Hedgehog (Hh) signaling target genes, such as ptc2 and gli1 [32,33]. Indeed, double labeling of pdgfab and ptc2 revealed a partially overlapping expression pattern, particularly in somitic cells close to the notochord, although pdgfab had a slightly broader expression domain than ptc2 (Fig 4D). The expression analysis suggests that pdgfab expression may be regulated by Hh signaling. To test this possibility, we treated wild-type embryos with DMSO or cyclopamine, a specific inhibitor of Hh signaling, from 5 to 24 hpf and stained them with pdgfab and nkx3.1. Cyclopamine-treated fish showed an almost complete absence of pdgfab expression in the medial trunk, accompanied by the loss of nkx3.1⁺ sclerotome-derived cells around the notochord (Fig 4E). These results suggest that Hh signaling regulates the migration of sclerotome-derived cells by controlling pdgfab expression. To determine how Pdgfab regulates the migration of sclerotome-derived cells, we overexpressed Pdgfab using a hsp:pdgfab-2A-mCherry transgenic line. Ubiquitous expression of pdgfab induced by heat shock at 18 hpf resulted in a complete loss of migrating nkx3.1⁺ sclerotome-derived cells at 24 hpf (Fig 4F). This result suggests that Pdgfab likely acts an instructive signal in controlling the migration of sclerotome derivatives.

pdgfab mutants display defective migration of sclerotome-derived cells to the notochord region

If Pdgfab is responsible for the migration of nkx3.1⁺ sclerotome-derived cells, we predicted that pdgfab loss-of-function mutants should partially phenocopy our pdgfra mutants. We characterized a nonsense mutant, pdgfab^sa11805, generated by the Zebrafish Mutation Project at the Wellcome Sanger Institute [34]. The pdgfab^sa11805 mutation introduces a premature stop codon, resulting in a truncated peptide of 111 amino acids (aa) instead of 230 aa. This shortened peptide is predicted to have the Pdgfab N-terminal domain, but it lacks the dimerization domain and most of the predicted receptor binding interface. For simplicity, we designate wild-type, heterozygous, and homozygous fish as pdgfab^+/+, pdgfab^+/−, and pdgfab^−/−, respectively. Using the nkx3.1:Gal4; UAS:NTR-mCherry reporter, we performed time-lapse imaging of sclerotome-derived cells from 24 to 34 hpf in different pdgfab mutant backgrounds (Fig 5A and S3 Movie). By 30 hpf, mCherry⁺ cells had dispersed throughout the medial trunk in pdgfab^+/+ controls, whereas in pdgfab^−/− embryos, only a few mCherry⁺ cells had begun to emerge from the sclerotome domains (Fig 5A and S3 Movie). At 34 hpf, pdgfab^−/− embryos showed a considerably less coverage of mCherry⁺ cells around the notochord compared to pdgfab^+/+ or pdgfab^+/− fish (Fig 5A and S3 Movie). However, the migration of mCherry⁺ sclerotome-derived cells toward the fin fold seemed unaffected in pdgfab^−/− embryos (Fig 5A and S3 Movie). This result suggests that pdgfab is specifically required for the migration of sclerotome-derived cells to the medial trunk, but not to the peripheral fin fold. Consistent with these findings, pdgfab^−/− embryos displayed substantially reduced staining of both nkx3.1 and nkx3.2 in sclerotome-derived cells around the notochord at 24 hpf (Fig 5B). Similarly, the expression of the tenocyte marker scxa was also greatly reduced in pdgfab^−/− fish at 48 hpf, whereas the expression of the fin mesenchymal cell marker fbln1 remained largely unaffected (Fig 5B). To confirm these observations, we quantified the number of tenocytes (Fig 5C) and fin mesenchymal cells (Fig 5D) at 56 and 48 hpf, respectively. Consistent with marker analysis, the number of tenocytes was significantly reduced in a dose-dependent manner in fish carrying different copies of the wild-type pdgfab allele, with averages decreasing from 19 per myotendinous junction in pdgfab^+/+ to 15 in pdgfab^+/− and 10 in pdgfab^−/− fish (Fig 5C). By contrast, there was no significant difference in the number of fin mesenchymal cells among different genotypes (Fig 5D). Notably, pdgfab^+/− embryos displayed a mild migration phenotype (Fig 5B and 5C), suggesting that the level of Pdgfab ligand is critical for the normal migration of sclerotome-derived cells. Together, our results suggest that the Pdgfab ligand primarily controls the medial migration of sclerotome-derived cells toward the trunk compartment, but not toward the fin fold. This region-specific role of pdgfab correlates well with its expression pattern shown above.

Fig 5. pdgfab is required for the migration of sclerotome-derived cells to the notochord.

(A) Snapshots from time-lapse imaging of the nkx3.1:Gal4; UAS:NTR-mCherry reporter in pdgfab^+/+, pdgfab^+/−, and pdgfab^−/− backgrounds from 24 to 34 hpf. In pdgfab^−/− embryos, the migration of mCherry⁺ cells (arrows) to the notochord is severely delayed compared to wild-type and heterozygous fish, while cells migrate normally to the fin fold to generate fin mesenchymal cells (notched arrowheads). The corresponding movies are shown in S3 Movie. Time stamps are indicated in the hh:mm format. Transverse views of the final frames are shown on the right side for each genotype, with the neural tube (nt) and notochord (n) outlined by dotted lines. These are projected images of a 50 μm region centered on the dashed line. n = 4 embryos per group. (B) Marker analysis in embryos with different pdgfab genotypes. pdgfab^−/− embryos lack nkx3.1 and nkx3.2 expression in sclerotome-derived notochord-associated cells (arrowheads) at 24 hpf. At 48 hpf, the tenocyte marker scxa (arrows) is greatly reduced in pdgfab^−/− embryos, while the fin mesenchymal cell marker fbln1 (notched arrowheads) remains unchanged across all genotypes. n = 25 embryos per staining. (C) Quantifications of tenocytes using the nkx3.1:Gal4; UAS:NTR-mCherry reporter in embryos with different pdgfab genotypes at 56 hpf. Each dot represents the average number of mCherry⁺ tenocytes along one MTJ, based on counts from three junctions of an individual embryo. n = 5 (pdgfab^+/+), 6 (pdgfab^+/−), and 5 (pdgfab^−/−) embryos. (D) Quantifications of fin mesenchymal cells in the major fin fold of embryos with different pdgfab genotypes at 48 hpf. Each dot represents the total number of fin mesenchymal cells from both the dorsal and ventral fin folds in the 8-somite region posterior to the end of the yolk extension in one embryo. n = 5 (pdgfab^+/+), 6 (pdgfab^+/−), and 5 (pdgfab^−/−) embryos. Data are plotted as mean ± SD. Statistics: One-way ANOVA with Tukey’s multiple comparison test. Asterisks representation: p < 0.001 (***), p < 0.01 (**), and p > 0.05 (ns, not significant). Data for this figure are provided in S1 Data. Scale bars: 100 μm.

Pdgfab acts as a chemoattractant in guiding the migration of sclerotome-derived cells

Next, we tested whether Pdgfab functions as a chemoattractant to guide the migration of Pdgfra-expressing cells from the sclerotome. To generate embryos with sparse pdgfab-expressing cells, we injected pdgfab^−/− mutants with a low dose of the hsp:pdgfab-2A-mCherry plasmid (Fig 6A). Injected embryos, along with uninjected controls, were heat shocked at 18 hpf and co-stained for nkx3.1 and mCherry probes at 24 hpf (Fig 6A). As shown in Fig 5B, in pdgfab^−/− controls, cells failed to migrate out of the sclerotome domains, resulting in no nkx3.1⁺ cells around the notochord at 24 hpf (Fig 6B and S4 Movie). By contrast, injected pdgfab^−/− embryos consistently exhibited migration of nkx3.1⁺ cells toward pdgfab-expressing cells (mCherry⁺) (Fig 6B and S4 Movie). We focused our analysis on embryos containing isolated mCherry⁺ cells or cell clusters. Strikingly, pdgfab expression in individual muscle cells (93%, 27/29 cases), notochord cells (100%, 16/16 cases), or cells in the spinal cord (83%, 10/12 cases) effectively attracted the migration of nkx3.1⁺ cells (Fig 6B and 6C and S4 Movie). By contrast, pdgfab expression in skin cells or cells in the dorsal spinal cord failed to trigger a robust migratory response of nkx3.1⁺ cells from the ventral sclerotome domain (S2 Fig and S5 Movie), suggesting that pdgfab⁺ cells need to be physically close to induce the migration of pdgfra-expressing cells. Interestingly, when pdgfab was ectopically expressed along the yolk extension, it induced abnormal ventral migration of nkx3.1⁺ cells (86%, 19/22 cases), resulting in a close juxtaposition of nkx3.1⁺ cells and the mCherry⁺ yolk epithelial cells (Fig 6B and 6C and S4 Movie). However, pdgfab expression in the ventral edge of the yolk extension failed to attract nkx3.1⁺ cells ventrally (S2 Fig and S5 Movie), likely due to the physical distance. To further characterize all cases with obvious chemoattraction (Fig 6C), we quantified the extent of migration of nkx3.1⁺ cells from the ventral edge of the somite (Fig 6D). Compared to somites in pdgfab^−/− embryos, pdgfab expression in muscles, notochord, or the spinal cord significantly rescued the migration of nkx3.1⁺ cells toward the notochord to a similar extent (Fig 6E). By contrast, pdgfab expression in the yolk extension resulted in the migration of nkx3.1⁺ cells in the opposite direction (Fig 6E). Together, our results demonstrate that Pdgfab functions as a chemoattractant in vivo, guiding the migration of Pdgfra-expressing sclerotome-derived cells.

Fig 6. pdgfab acts as a chemoattractant for sclerotome-derived cells.

(A) Schematics of the experimental design. pdgfab^−/− embryos were injected with the hsp:pdgfab-2A-mCherry plasmid at a low dose at the one-cell stage, heat shocked at 18 hpf, and double-stained using nkx3.1 and mCherry probes at 24 hpf. (B) Examples of nkx3.1 (cyan) and mCherry (magenta) expression in uninjected and injected pdgfab^−/− mutant embryos at 24 hpf. In sagittal views, the neural tube (long brackets), notochord (short brackets), and yolk extension (dotted lines) are labeled. In transverse views, the neural tube (nt), notochord (n), and yolk extension (y) are marked by dotted lines, and nuclear staining with Draq5 is shown. In uninjected controls (top), nkx3.1-expressing cells in the ventral sclerotome fail to migrate out to surround the notochord. By contrast, sparse expression of hsp:pdgfab-2A-mCherry in muscle, notochord, or spinal cord cells (mCherry⁺) substantially restores the migration of nkx3.1⁺ sclerotome-derived cells toward the notochord. Expression of hsp:pdgfab-2A-mCherry in the yolk extension leads to the ventral migration of nkx3.1⁺ cells toward the yolk. Localized mCherry⁺ pdgfab-expressing cells that attract nkx3.1⁺ sclerotome-derived cells are indicated by arrowheads. The corresponding confocal stacks are shown in S4 Movie. n = 10 (uninjected) and 80 (injected) embryos. (C) Quantification of the experiment shown in (B). The proportion of mCherry⁺ cells (or cell clusters) that display attraction of nkx3.1⁺ cells is shown for the tissues depicted in (B). The number of cases for each tissue is indicated above each bar. (D) Schematics of the quantification of the migration extent of nkx3.1⁺ cells in the experiment showed in (B). We measured the distance between the ventral border of the somite (position = 0) and the leading edge of the nkx3.1⁺ cells. In uninjected pdgfab^−/− mutants (“no migration”), the small distance represents the height of the ventral sclerotome domain. In hsp:pdgfab-2A-mCherry-injected pdgfab^−/− embryos, nkx3.1⁺ cells derived from the ventral sclerotome can migrate toward the notochord (“rescued”, positive migration) or the yolk extension (“ectopic”, negative migration). (E) Quantification of the migration extent of nkx3.1⁺ cells as depicted in (D) for each of the scenarios shown in (B). Only cases where attraction was observed are plotted here, except in the uninjected animals. n = 8 (uninjected), 27 (muscle), 16 (notochord), 10 (spinal cord), and 19 (yolk extension) cases from 10 (uninjected) and 80 (injected) embryos. Statistics: One-way ANOVA with Dunnett’s multiple comparison test. Asterisks representation: p < 0.0001 (****). Data for this figure are provided in S1 Data. Scale bars: 50 μm.

Discussion

In this study, we combine in vivo imaging with gain- and loss-of-function approaches to investigate how fibroblast precursors are guided to specific locations in the developing zebrafish embryo. Our experiments reveal three key findings (Fig 7). First, Pdgfra functions cell-autonomously in the migration of sclerotome-derived fibroblast precursors. Second, pdgfab is specifically required for the medial migration of sclerotome-derived cells. Third, localized Pdgfab expression acts as a chemoattractant in directing the migration of sclerotome-derived cells. Together, these results demonstrate that Pdgfra signaling mediates the chemoattraction of sclerotome-derived fibroblast precursors toward Pdgfab-expressing regions.

Fig 7. The migration of sclerotome-derived cells is regulated by chemoattraction mediated by Pdgfab/Pdgfra signaling.

(A) In wild-type embryos, nkx3.1⁺ (cyan) sclerotome-derived cells express the receptor Pdgfra (magenta), while the medial somites express the ligand Pdgfab (orange) in response to Hh signaling. pdgfra⁺ sclerotome-derived cells undergo extensive migration toward both the notochord in the medial trunk and the fin fold. (B) In pdgfra^−/− mutants (pdgfra^mRFP/mRFP and pdgfra^ref/ref), nkx3.1⁺ sclerotome-derived cells fail to migrate in any direction and remain in their original location in the ventral and dorsal parts of the somites. (C) In pdgfab^−/− embryos, pdgfra⁺ sclerotome-derived cells fail to migrate toward the notochord, although migration toward the fin fold remains unaffected. (D) In pdgfab^−/− embryos with mosaic pdgfab expression, nkx3.1⁺ sclerotome cells can migrate toward the ectopic pdgfab-expressing cells, either medially around the notochord or ventrally toward the yolk extension. When the pdgfab source is located near or in the notochord, it can completely rescue the migration of nkx3.1⁺ sclerotome cells toward the notochord. These diagrams depict a 2-somite region located above the yolk extension of a zebrafish embryo at 48 hpf. nt: neural tube; n: notochord.

Pdgfra is required cell-autonomously for the migration of sclerotome-derived fibroblast precursors

Previous studies in mice and chicks have shown that loss of PDGFRA function leads to various axial skeleton defects, suggesting compromised sclerotome development [22,23]. However, how PDGFRA signaling regulates the development of sclerotome-derived cells in vivo remains unclear. In our study, we provide three lines of evidence demonstrating that pdgfra is essential for the migration of sclerotome-derived cells in zebrafish (Fig 7A and 7B). First, pdgfra is specifically enriched in the sclerotome and its derivatives. Second, loss-of-function experiments, using either the pdgfra gene-trap mutant, the pdgfra^ref mutant, or overexpression of dominant-negative PdgfraΔTK-EGFP, result in a complete disruption of sclerotome-derived cell migration, accompanied by the loss of sclerotome derivatives such as tenocytes and fin mesenchymal cells. Third, mosaic inhibition of Pdgfra activity by membrane-tethered PdgfraΔK-EGFP specifically blocks the migration of PdgfraΔK-EGFP-expressing sclerotome cells but not non-expressing cells, indicating that Pdgfra functions in a cell-autonomous manner.

The pdgfra^mRFP gene-trap allele is predicted to produce a secreted, truncated Pdgfra receptor fused to mRFP (PdgfraΔTK-mRFP), lacking both the transmembrane and kinase domains. With an intact ligand-binding extracellular domain but no signaling capability, PdgfraΔTK-mRFP likely functions as a decoy receptor, sequestering Pdgf ligands in the extracellular space. Similar decoy receptors, such as interleukin-1 type II receptor (IL-1RII) and soluble vascular endothelial growth factor receptor-1 (sVEGFR-1), have been shown to inhibit their respective signaling pathways [35,36]. Although naturally occurring secreted short forms of PDGFRA have not been reported, muscle-resident fibro/adipogenic progenitors express a short isoform produced by an internal polyadenylation site [37]. This isoform functions as a membrane-tethered decoy receptor, attenuating PDGF signaling during muscle regeneration [37]. Consistent with the dominant-negative nature of a decoy receptor, heterozygous pdgfra^mRFP/+ embryos exhibit reduced migration and differentiation of sclerotome-derived cells compared to their wild-type siblings and heterozygous pdgfra^ref/+ fish. The mRFP signal from PdgfraΔTK-mRFP is undetectable in the extracellular space and only gradually appears later in the endothelial scavenger cells [29]. Similar vascular localization has been observed in other gene-trap lines targeting genes with secreted signals at their 5′ end [38], suggesting that these secreted truncated proteins are captured by endothelial scavenger cells.

Consistent with our findings, pdgfra has been recently shown to be required for the development of stromal reticular cells in zebrafish [39]. These cells, which support hematopoietic stem cells, are derived from the ventral sclerotome domain in the tail [28,40]. Using morpholino knockdown experiments, Murayama and colleagues show that pdgfra is required for the ventral migration of sclerotome cells to give rise to stromal reticular cells and fin mesenchymal cells, but is not necessary for dorsal migration toward the notochord [39]. This result contrasts with our findings that pdgfra is critical for the migration of sclerotome-derived cells in all directions, including toward both the notochord and the fin fold. This discrepancy may be due to the stronger phenotype caused by the dominant-negative nature of the pdgfra gene-trap mutant compared to the morpholino knockdown. Additionally, it is possible that the ventral migration of sclerotome cells toward the fin fold requires a higher level of Pdgfra activation, making it more sensitive to pathway inhibition.

Hh signaling-dependent pdgfab is required for sclerotome-derived cell migration toward the medial trunk

Previous studies in chick suggest that Hh signaling is essential for sclerotome development [41]. Our group has shown that in zebrafish, Hh signaling is dispensable for the induction of sclerotome domains but is essential for the migration and maintenance of marker expression in sclerotome-derived cells around the notochord [10]. Here, we show that pdgfab exhibits partially overlapping expression in the medial somite with other Hh target genes such as ptc2 and myod1. Inhibition of Hh signaling by cyclopamine results in a complete loss of pdgfab expression in the somites, accompanied by impaired migration of sclerotome-derived cells toward the notochord. This migration phenotype is recapitulated in pdgfab mutants, suggesting that Hh signaling indirectly regulates the migration of sclerotome-derived cells by controlling pdgfab expression. However, we cannot rule out the possibility of a parallel mechanism, in which Hh signaling-dependent adaxial cell specification and lateral migration [42,43] serve as prerequisites for triggering the migration of sclerotome cells toward the notochord [44].

A comparison of migration phenotypes between pdgfra and pdgfab mutants reveals two key differences. First, both pdgfra^mRFP/mRFP and pdgfra^ref/ref mutants show defects of sclerotome cell migration in all directions, leading to the loss of both tenocytes and fin mesenchymal cells (Fig 7B). By contrast, pdgfab mutants specifically show impaired sclerotome cell migration toward the notochord, while the migration of fin mesenchymal cell precursors into the fin fold remains largely unaffected (Fig 7C). We predict that sclerotome-derived cell migration is orchestrated by multiple Pdgfra ligands expressed in distinct regions. Based on our expression analysis, pdgfc is the only ligand expressed in the distal edge of the fin fold, the target area for fin mesenchymal cell precursors. Thus, it is plausible that Pdgfab acts as a chemoattractant to mediate migration toward the notochord region, while Pdgfc plays a similar role in the fin fold. Since we did not observe enhanced migration toward the fin fold in the absence of pdgfab, this suggests that additional mechanisms, such as contact inhibition as reported in neural crest cells [45], may regulate the number of fibroblast precursors in target regions.

Another difference between pdgfra and pdgfab mutants is that pdgfra embryos exhibit a more severe phenotype in the migration of sclerotome-derived cells toward the notochord, resulting in a greater reduction in tenocyte number. The expression of both pdgfaa and pdgfc in the medial trunk region suggests that these ligands may function redundantly with Pdgfab to mediate the migration of sclerotome-derived cells in this area. Similar functional redundancy has been observed between pdgfaa and pdgfab in zebrafish during pharyngeal arch artery morphogenesis [46] and primitive erythropoiesis [47].

Sclerotome-derived cell migration is regulated by chemoattraction through Pdgfab/Pdgfra signaling

Chemoattraction by PDGF ligands has been first described in vitro in experiments with smooth muscle cells and fibroblasts [48,49]. In vivo studies in Drosophila have further demonstrated a chemoattractive role of the homologous PDGF signaling pathway in border cell migration [26,27]. During early oogenesis in Drosophila, the oocyte expresses PDGF/VEGF-related factor 1 (PVF1), which acts as a chemoattractant to guide the migration of border cells expressing PVR, a receptor tyrosine kinase related to the mammalian PDGF and VEGF receptors [26,27]. Similar chemoattraction models have been proposed to explain the directed migration of both cardiomyocytes and neural crest cells in vertebrates, where PDGFRA-expressing cells migrate toward PDGFA ligand-expressing target areas [16,18,45]. Indeed, PDGFA-coated beads have been shown to direct the migration of PDGFRA-expressing neural crest cells in vivo in both zebrafish and mice [16,50], as well as in vitro in xenopus explants [45]. These findings suggest that PDGFA/PDGFRA signaling-mediated chemoattraction may be a conserved mechanism in regulating directed cell migration in various developmental contexts.

Previous studies have suggested a role for PDGFRA signaling in regulating the migration of sclerotome-derived cells in mice [22,23,25]; however, this function has been inferred rather than directly demonstrated due to technical limitations. In our current study, we provide in vivo evidence that Pdgfab acts as an instructive cue, likely functioning as a chemoattractant to guide the migration of Pdgfra-expressing fibroblast precursors in zebrafish. First, ubiquitous overexpression of Pdgfab completely abolishes the migration of sclerotome-derived cells. This suggests an instructive rather than permissive role for Pdgfab, as uniform distribution of the signal eliminates the directional cues necessary for guided migration. Second, in the absence of endogenous Pdgfab, mosaic expression of Pdgfab in medial cells, such as muscles, notochord, and ventral spinal cord cells, is sufficient to rescue the localized migration of sclerotome-derived cells to surround the notochord (Fig 7D). Although the notochord does not normally express pdgfab, it efficiently attracts sclerotome-derived cells, likely due to its position as their final migration destination [10]. This result supports a model in which the notochord can secrete functional Pdgfab protein, suggesting that the specific cell type producing the ligand is less critical than where sclerotome-derived cells encountered the protein. Our mosaic expression experiments further reveal that the secreted Pdgfab protein has a limited diffusion range, as pdgfab expression in distant tissues, such as the skin, dorsal spinal cord, or dorsal muscle fibers, is insufficient to trigger the migration of cells from the ventral sclerotome domain. This localized effect underscores the importance of spatially restricted Pdgfab expression for proper cell guidance. Ectopic Pdgfab expression in the yolk extension is sufficient to redirect sclerotome-derived cells to undergo abnormal ventral migration (Fig 7D). This result mirrors previous findings where ectopic PVF1 expression similarly redirects border cell migration in Drosophila [27]. Collectively, our findings strongly indicate that Pdgfab/Pdgfra signaling-mediated chemoattraction guides the migration of sclerotome-derived cells in zebrafish. Given that Pdgfa is expressed in the myotome in mice [25,51] and Pdgfra is expressed in the sclerotome in both mice and humans [51,52], we predict that the migration of sclerotome-derived cells in mammals is likely guided by a similar chemotactic mechanism.

In summary, our zebrafish studies demonstrate that Pdgfab functions as a localized chemoattractant cue to guide the migration of Pdgfra-expressing fibroblast precursors from the sclerotome (Fig 7). This work provides a framework for studying the development and diversification of distinct fibroblast subtypes during development. Our findings also suggest that similar PDGFA/PDGFRA signaling mechanisms may be employed in other contexts, such as recruiting fibroblasts during wound healing or regulating cell migration in pathological conditions like cancer metastasis [53]. Targeting PDGFA/PDGFRA signaling could represent a therapeutic opportunity to treat fibrosis-related disorders and cancer.

Materials and methods

Ethics statement

All animal experiments were conducted in accordance with the principles outlined in the current guidelines of the Canadian Council on Animal Care. All protocols were approved by the Animal Care Committee at the University of Calgary (#AC21-0102).

Zebrafish strains

Zebrafish strains used were raised under standard zebrafish welfare conditions [54]. The following lines were utilized: pdgfra^smh1300Gt (referred to as pdgfra^mRFP) [19], pdgfra^sk16 (referred to as pdgfra^ref) [18], pdgfab^sa11805 [34], Tg(hsp:pdgfab-2A-mCherry,cryaa:CFP)ca118 (referred to as hsp:pdgfab-2A-mCherry), Tg(kdrl:EGFP)la116 [55], TgBAC(nkx3.1:Gal4)ca101 [10], Tg(UAS:Kaede)s1999t [56], Tg(UAS:NTR-mCherry)c264 [56], and Tg(UAS:pdgfraΔK-EGFP)ca117. Fluorescent transgenic lines were screened by the expression of the respective fluorescent marker. The pdgfra^mRFP and pdgfra^ref lines were maintained as heterozygotes, as homozygous mutants are not viable beyond 14 dpf [18,19]. The pdgfab^sa11805 line was maintained as both heterozygotes and homozygotes.

Mutant genotyping

The genotype of embryos carrying the pdgfra^mRFP or pdgfab^sa11805 allele was determined using conventional PCR with allele-specific primers. Genomic DNA was extracted using the HOTSHOT protocol. Briefly, tissue from embryos or fin clips was dissociated in 50 mM NaOH at 95°C for 20 min, following by cooling on ice and neutralization with 1 M Tris-HCl (pH = 7.5). To identify the pdgfra^mRFP allele, the primers pdgfra-g-F3 (5′-TCGGACCACGAGTCTACATA-3′) and mRFP-1 (5′-GAGCCCTCCATGCGCACCTTGAA-3′) were used with an annealing temperature of 61°C, producing a single 805 bp band in homozygous mutant embryos. For wild-type embryos, the primers pdgfra-g-F3 and pdgfra-g2-R (5′-TGTGTCAACGCCACAACCTA-3′) were used with an annealing temperature of 61°C, yielding a single 575 bp band. PCRs from heterozygous embryos produce both corresponding bands.

The pdgfab^sa11805 allele has a 2 bp change from 5′-ATCTAC-3′ to 5′-ATtTAa-3′ at positions 333 and 336 of the coding sequence, resulting in a substitution of tyrosine (Y) at amino acid 112 in the wild-type allele with a premature stop codon in pdgfab^sa11805. We designed a one-step PCR assay [57] to distinguish the pdgfab^sa11805 allele from the wild-type allele. To identify the pdgfab^sa11805 allele, the primers Pdgfab_Geno3_Fmut (5′-CAAGACCCGGACGGTGCTT-3′) and Pdgfab_Geno2_R (5′-CAAACACACAAACCCGTGAC-3′) were used with an annealing temperature of 62°C, producing a single 171 bp band in homozygous mutant embryos. For wild-type embryos, the primers Pdgfab_Geno4_Fwt (5′-CAAGACCCGGACGGTGCTC-3′) and Pdgfab_Geno2_R were used with an annealing temperature of 62°C, also resulting in a single 171 bp band. PCRs from heterozygous embryos produce both corresponding bands.

The pdgfra^ref mutant line was genotyped by PCR analysis followed by restriction digest [18]. A 136 bp region flanking the pdgfra^ref locus was PCR amplified using the primer pair 5′-GTAGGTAAAAGTAAAGCTGGTA-3′ and 5′-CAAGGGTGTGTTGAACCTGA-3′. These primers introduce a KpnI restriction site in the wild-type allele but not in the pdgfra^ref allele. KpnI digestion of the wild-type PCR product yields fragments of 113 and 23 bp.

Generation of expression constructs

All expression constructs were assembled following the SLiCE protocol [58]. The hsp70:pdgfab-2A-mCherry plasmid was generated by replacing the pdgfaa sequence in the hsp70:pdgfaa-2A-mCherry plasmid [18] with the pdgfab coding sequence.

To generate the UAS:pdgfraΔK-EGFP plasmid, UAS:pdgfra was first constructed by replacing NTR-mCherry in the UAS:NTR-mCherry plasmid with the full-length pdgfra coding sequence (ENSDART00000103510.5). The cytosolic domain of pdgfra at position 2,838−4,397 of the coding sequence was then replaced with a linker sequence (5′-AGTCTCGGACCTGGACTCGGATCCGGA-3′) and an ATG-less EGFP sequence.

The hsp:pdgfraΔTK-EGFP plasmid was generated in two steps: First, pdgfraΔK-EGFP was inserted into hsp70:EGFP to replace EGFP. Second, the transmembrane domain coding sequence was removed using a PCR strategy [59,60].

Generation of transgenic lines

The Tg(hsp:pdgfab-2A-mCherry,cryaa:CFP) line was generated using the hsp70:pdgfab-2A-mCherry plasmid, which carries cryaa:CFP as a transgenesis marker. To generate a stable line, 40 pg of the hsp70:pdgfab-2A-mCherry plasmid was co-injected with 40 pg of tol2 mRNA into one-cell stage wild-type embryos. The injected embryos were screened at 3 dpf for cryaa:CFP expression in the lens. Positive embryos were raised to adulthood as potential founders. Stable transgenic lines were established by screening for cryaa:CFP expression in F1 embryos from these injected founders.

Similarly, the Tg(UAS:pdgfraΔK-EGFP) line was generated by co-injecting 40 pg of the UAS:pdgfraΔK-EGFP plasmid with 40 pg of tol2 mRNA into one-cell stage wild-type embryos. Injected embryos were raised to adulthood as potential founders. The stable Tg(UAS:pdgfraΔK-EGFP) line was established by crossing potential founders to hsp:Gal4 transgenic animals and selecting those that produced EGFP⁺ F1 embryos after heat shock.

Embryo injections

For most plasmid injections, 40 pg of plasmid DNA was co-injected with 40 pg of tol2 transposase mRNA into embryos at the one-cell stage. For mosaic experiments with UAS:pdgfraΔK-EGFP, 10 pg of the plasmid was injected. For mosaic experiments with hsp:pdgfab-2A-mCherry, 2 pg of the plasmid was injected without tol2 transposase mRNA. For Cre expression, 20 pg of Cre mRNA was injected into one-cell stage embryos obtained from intercrosses of pdgfra^mRFP/+ fish. To label scavenger endothelial cells, 1 nL of a 2 mg/mL solution of Alexa Fluor 647-conjugated ovalbumin (Invitrogen, O34784) was injected into the extracellular space of embryos at 4 hpf.

Heat shock induction

To induce heat shock promoter-driven genes, the embryos were transferred into a 2 mL tube and placed in a heat block set at 37°C for 30 min, at the desired developmental stage. Afterward, the embryos were transferred to E3 fish water and allowed to continue developing at 28.5°C until fixation.

Time-lapse imaging

Time-lapse imaging of zebrafish embryos was conducted using an Olympus FV1200 confocal microscope. Embryos at the desired developmental stage were anesthetized with tricaine and mounted in a 35 mm glass-bottom imaging dish. A small volume of E3 fish water containing phenylthiourea and tricaine was carefully added around the agarose to prevent pigment development and minimize embryo movement. The trunk region above the yolk extension was imaged, with Z-stack images captured at intervals specified in each figure. The images were then compiled and processed using Fiji software [61].

In situ hybridization and immunohistochemistry

Whole-mount in situ hybridization [62] and immunohistochemistry were performed following standard protocols. The RNA probes used were: fbln1, fras1, GFP, mCherry, mRFP, myod1, nkx3.1, nkx3.2, pax1a, pax9, pdgfaa, pdgfab, pdgfc, pdgfra, pdgfra5′, pdgfra3′, prelp, ptc2, and scxa. Double fluorescent in situ hybridizations were performed combining digoxigenin (DIG) and dinitrophenyl (DNP) labeled probes. For immunohistochemistry, the CC3 (Asp175) antibody (1:400, Cell Signaling, #9661) was used. Draq5 (1:10,000, Biostatus, DR50050) was used for nuclear staining. To obtain transverse views, stained embryos were manually sectioned using vibratome steel blades.

Drug treatment

Embryos at 5 hpf were treated in 100 μM cyclopamine (Cedarlane Labs, C988400-50) in E3 fish water. Control embryos were treated with 1% DMSO in E3 fish water. The treated embryos were maintained at 28°C until 24 hpf for analysis.

Quantifications

To quantify nkx3.1-stained area, stained embryos were imaged, and a region of interest comprising 8 somites above the yolk extension was defined. The image was then converted to black and white, a threshold was set, and the coverage ratio was calculated. These values were subsequently plotted on graphs.

For tenocyte counting, the nkx3.1:Gal4; UAS:NTR-mCherry reporter was used in the appropriate genetic background at the desired developmental stage. The number of tenocytes in three myotendinous junctions per animal was counted, and the average number per MTJ for each fish was then plotted on graphs.

To quantify fin mesenchymal cells, anesthetised animals at the desired developmental stage were observed under an inverted Zeiss microscope with a 20× objective. Fin mesenchymal cells, identified by their characteristic morphology, were counted both dorsally and ventrally in an area spanning 8 somites posterior to the end of the yolk extension. The counts were summed for each individual fish and plotted on graphs.

For the chemoattraction experiments, the extent of migration of the nkx3.1⁺ cells was determined by measuring the distance between the ventral edge of the somite and the leading nkx3.1⁺ cell beneath an mCherry⁺ cell. The measurements were performed using Fiji software [61].

Image deconvolution

Confocal images in Figs 4B–4D, 6B, and S2 were deconvolved using DeconvolutionLab2 [63]. Briefly, a simulated point spread function (PSF) was generated using the PSF Generator plugin with the Born and Wolf PSF model. Individual z-stack channels were then processed with their corresponding PSF files using the Richardson-Lucy algorithm for 50 iterations.

Statistical analysis

All the graphs were generated using the GraphPad Prism software. Data were plotted as mean ± SD. Significance was calculated using these statistic tests: One-way ANOVA with Tukey’s multiple comparison test (Figs 1E, 2C, 2D, 5C, 5D, and S1D), Unpaired t test with Welch’s correction (Fig 3F), and One-way ANOVA with Dunnett’s multiple comparison test (Fig 6E). p values: p < 0.0001 (****), p < 0.001 (***), p < 0.01 (**), p < 0.05 (*), and p > 0.05 (ns, not significant).

Supporting information

S1 Fig. Characterization of the pdgfra gene-trap line.

(A) Schematics of the pdgfra gene-trap mutant line. The exons of the pdgfra gene are color-coded: gray for extracellular coding exons, blue for the transmembrane coding exon, yellow for intracellular coding exons, and white for non-coding exons. This pdgfra gene-trap line is predicted to produce two mRNAs: the extracellular domain of pdgfra fused with mRFP (pdgfraΔTK-mRFP) under the control of endogenous pdgfra promoter, and GFP fused with the transmembrane and cytoplasmic kinase domains of pdgfra (GFP-pdgfraTK) under the control of the β-actin promoter. The locations of different region-specific probes used to characterize this mutant are shown as black lines aligned underneath the corresponding mRNA. (B) Expression analysis of the mRNAs produced from the pdgfra gene-trap line. pdgfra^+/+, pdgfra^mRFP/+, and pdgfra^mRFP/mRFP embryos were stained at 24 hpf using two probes against the endogenous pdgfra sequences (pdgfra5′ and pdgfra3′) and two probes against the exogenous inserts (mRFP and GFP). Only pdgfra^mRFP/+ and pdgfra^mRFP/mRFP embryos express mRFP and GFP. (C) Embryos from crosses of pdgfra^mRFP/+ and pdgfra^mRFP/+; nkx3.1:Gal4; UAS:Kaede fish were stained at 24 hpf with an antibody against cleaved caspase-3 (CC3, magenta), along with DRAQ5 (gray) for nuclear labeling. CC3⁺ cells are indicated by yellow arrows. (D) Quantification of CC3⁺ and CC3⁺Kaede⁺ cells in the trunk. n = 8 (pdgfra^+/+), 12 (pdgfra^mRFP/+), and 6 (pdgfra^mRFP/mRFP) embryos. Data are plotted as mean ± SD. Statistics: One-way ANOVA with Tukey’s multiple comparison test. Asterisks representation: p > 0.05 (ns, not significant). Data for this figure are provided in S1 Data. Scale bars: 100 μm.

(TIFF)

S2 Fig. Examples showing no chemoattraction in response to mosaic pdgfab expression.

pdgfab^−/− embryos were injected with the hsp:pdgfab-2A-mCherry plasmid at a low dose at the one-cell stage, heat shocked at 18 hpf, and double-stained using nkx3.1 (cyan) and mCherry (magenta) probes at 24 hpf. Examples of cases showing no chemoattraction are shown here. In sagittal views, the neural tube (long brackets), notochord (short brackets), and yolk extension (dotted lines) are labeled. In transverse views, the neural tube (nt), notochord (n), and yolk extension (y) are marked by dotted lines, and nuclear staining with Draq5 is shown. pdgfab expression in dorsal muscle fibers, skin cells, dorsal spinal cord cells, or yolk cells at the ventral edge of the yolk extension does not induce migration of nkx3.1⁺ cells from the ventral sclerotome domain. mCherry⁺ pdgfab-expressing cells are indicated by arrowheads. The corresponding confocal stacks are shown in S5 Movie. n = 2 (muscle), 2 (spinal cord), and 3 (yolk extension) cases from 80 (injected) embryos. Scale bars: 50 μm.

(TIFF)

S1 Movie. Time-lapse imaging of sclerotome-derived cells in wild-type and pdgfra gene-trap mutant backgrounds.

Embryos at 22 hpf were imaged every 17 min for 14 h. In wild-type siblings, sclerotome-derived cells above the yolk extension migrate from the ventral and dorsal sclerotome domains toward the notochord. In contrast, heterozygous and homozygous individuals exhibit reduced migration in a dose-dependent manner. Snapshots from this time-lapse are shown in Fig 2A. Time stamps are indicated in the hh:mm format. Scale bar: 100 μm.

(MP4)

S2 Movie. pdgfra is required cell-autonomously for sclerotome-derived cell migration.

Confocal stacks of embryos with mosaic expression of Lyn-EGFP or PdgfraΔK-EGFP shown in Fig 3E are compiled together. Scale bar: 100 μm.

(MP4)

S3 Movie. Time-lapse imaging of sclerotome-derived cells in wild-type and pdgfab mutant backgrounds.

Embryos at 24 hpf were imaged every 20 min for 10 h. In wild-type siblings, sclerotome-derived cells migrate from the ventral and dorsal sclerotome domains toward the notochord. pdgfab^+/− individuals show slight migration defects, while pdgfab^−/− mutants exhibit a more pronounced reduction in migration toward the notochord, though migration toward the fin folds remains largely unaffected. Snapshots from this time-lapse are shown in Fig 5A. Time stamps are indicated in the hh:mm format. Scale bar: 100 μm.

(MP4)

S4 Movie. Examples showing chemoattraction in response to mosaic pdgfab expression.

Confocal stacks from the examples shown in Fig 6B are compiled together. Scale bar: 50 μm.

(MP4)

S5 Movie. Examples showing no chemoattraction in response to mosaic pdgfab expression.

Confocal stacks from the examples shown in S2 Fig are compiled together. Scale bar: 50 μm.

(MP4)

S1 Data. This table contains all numerical data presented in the manuscript.

(XLSX)

Abbreviations

CC3

cleaved caspase-3

CVP

caudal vein plexus

PDGF

platelet-derived growth factor

PSF

point spread function

SHH

Sonic Hedgehog

Data Availability

All relevant data are within the paper and its Supporting information files.

Funding Statement

This study was supported by grants to P.H. from the Canadian Institutes of Health Research (PJT-169113 and OGB-198236, https://cihr-irsc.gc.ca/e/193.html) and Canada Foundation for Innovation John R. Evans Leaders Fund (Project 32920, https://www.innovation.ca). E.E.M. was supported by postdoctoral fellowships from the Alberta Children’s Hospital Research Institute (ACHRI, https://research4kids.ucalgary.ca), the Cumming School of Medicine (CSM, https://cumming.ucalgary.ca), and the Canadian Institutes of Health Research (MFE-201007, https://cihr-irsc.gc.ca/e/193.html). J.B. was supported by funding from the National Institutes of Health (R15HD108782 and P20GM130460, https://www.nih.gov). S.L. was supported by a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada (RGPIN-2022-03558, https://www.nserc-crsng.gc.ca/index_eng.asp). The funders play no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.