Sclerotome-derived PDGF signaling functions as a niche cue responsible for primitive erythropoiesis Open Access

Erythrocytes are the most common cell type in blood and are responsible for oxygen transportation and vascular morphogenesis during embryo development (Baron et al., 2012; Zhou et al., 2021). Erythrocytes arise in two successive waves of hematopoiesis, primitive and definitive. As the hematopoietic processes and molecular mechanisms in zebrafish are highly conserved in mammals, zebrafish has been regarded as a powerful model for the study of hematopoiesis (de Jong and Zon, 2005; Jagannathan-Bogdan and Zon, 2013). In mammals, the early hematopoietic wave takes place in the yolk sac, where the primitive erythrocytes are generated. In zebrafish, embryonic hematopoiesis originates from the intermediate cell mass (ICM) of the posterior lateral plate mesoderm and the rostral blood island located in the anterior lateral plate mesoderm (Davidson and Zon, 2004; Ellett and Lieschke, 2010). The second wave, known as definitive hematopoiesis, gives rise to the hematopoietic stem cells (HSCs), which facilitate lifelong replenishment of all major hematopoietic lineages. HSCs emerge from the aorta-gonad-mesonephros region in mammals and ultimately migrate to the mammalian fetal liver and the bone marrow, which function as niches for their proliferation and differentiation. This is equivalent to the HSC transition from the ventral wall of the dorsal aorta (DA) to the kidney marrow in zebrafish (Bertrand et al., 2010). The definitive hematopoietic niches serve important roles to maintain the self-renewal capacity of HSCs, protect them from exhaustion and regulate their differentiation, which is crucial for blood system homeostasis. These dedicated niches have been widely investigated in recent decades and comprise vascular endothelial cells and mesenchymal stromal cells, along with diverse molecular signals (Calvi et al., 2003; Morrison and Scadden, 2014; Rafii et al., 2015; Touret et al., 2022). Conversely, little is known about the components of the primitive hematopoietic niche and how they work in concert to regulate embryonic erythropoiesis.

The platelet-derived growth factor (PDGF) family comprises four different gene products (PDGFA, PDGFB, PDGFC and PDGFD), which exist in disulfide-linked dimeric forms. There are two receptors for PDGFs, PDGFRα and PDGFRβ (Pdgfra and Pdgfrb, respectively, in zebrafish) (Kazlauskas, 2017; Shim et al., 2010). Binding of PDGF ligands to PDGF receptors (PDGFRs) induces receptor dimerization, autophosphorylation on tyrosine residues and subsequent activation of downstream signaling cascades. PDGF signaling is one of the pivotal signaling pathways regulating cell proliferation, migration and differentiation. It has been demonstrated to promote the expansion of several hematopoietic cell types ex vivo (Su et al., 2001, 2002). In addition, aberrant activation of PDGF signaling is associated with hematopoietic malignancies, such as myeloid neoplasms with hypereosinophilia and large granular lymphocytic leukemia (Demoulin and Montano-Almendras, 2012; Yang et al., 2010). These findings indicate that PDGF signaling has a role in physiological and pathological hematopoietic processes. Interestingly, PDGF signaling has been recently reported to contribute to the definitive hematopoietic niche (da Bandeira et al., 2022; Damm and Clements, 2017). However, it is unknown whether PDGF signaling functions in primitive hematopoiesis. Here, we identify that PDGF signaling derived from the sclerotome acts as a niche cue that plays a role in embryonic erythropoiesis, facilitating the differentiation of erythroid progenitors and the formation of primitive red blood cells (RBCs).

In a previous study, we generated double mutants lacking pdgfαa (also known as pdgfaa) and pdgfαb (also known as pdgfab), the zebrafish orthologs of PDGFA, and found that the pdgfαa^−/−;pdgfαb^−/− double mutants have malformations of the pharyngal arch artery (Mao et al., 2019). Intriguingly, under brightfield microscopy, we observed that there appeared to be significantly fewer blood cells in the pdgfαa^−/−;pdgfαb^−/− embryos. To confirm this observation, o-dianisidine staining was applied at 48 h post fertilization (hpf) to detect the formation of RBCs by evaluating hemoglobin levels. Indeed, an obvious decrease in hemoglobin expression was found in pdgfαa^−/−;pdgfαb^−/− mutants (Fig. 1A), suggesting an important role of pdgfαa and pdgfαb in blood cell development. To further monitor erythroid populations, we crossed the Tg(gata1:dsRed) strain with pdgfαa^−/−;pdgfαb^−/− fish. A dramatic reduction in the number of circulating RBCs was observed in the mutants (Fig. 1B; Movie 1), whereas no obvious abnormalities of heart rhythm or contraction occurred. After deletion of pdgfαa and pdgfαb, the expression of the erythrocyte-specific markers gata1a (named gata1 for short) and hbbe3 was slightly decreased in ∼60% of embryos and markedly reduced in ∼40% of embryos at 22 hpf (Fig. 1C,D). These results imply that PDGF signaling is involved in primitive erythropoiesis.

Next, wild-type embryos were treated with inhibitor V, a selective PDGFR tyrosine kinase inhibitor, from the shield stage to the long-pec stage. As expected, a severe decrease in the number of RBCs was detected in the resulting embryos, as revealed by o-dianisidine staining (Fig. 1E). Decreased expression of gata1 confirmed the defective erythropoiesis in animals treated with inhibitor V (Fig. 1F).

To determine whether pdgfαa and pdgfαb are both involved in hematopoiesis, we examined the expression of gata1 and hbbe3 in pdgfαa^−/− or pdgfαb^−/− mutant embryos using in situ hybridization. No obvious changes were observed in primitive hematopoiesis at 14 hpf and 22 hpf (Fig. S1). This might be due to compensatory effects of pdgfαa and pdgfαb, as we have discussed in previous work (Mao et al., 2019). Collectively, these results indicate that both pdgfαa and pdgfαb play a crucial function in primitive erythropoiesis.

HSCs, a unique cell population arising in definitive hematopoiesis, are capable of self-renewal and multilineage differentiation. We next investigated HSC formation in pdgfαa^−/−;pdgfαb^−/− embryos so as to define the role of these genes in definitive hematopoiesis. Expression of the HSC markers runx1 and cmyb (also known as myb) was unaffected in pdgfαa^−/−;pdgfαb^−/− mutants (Fig. S2A,B). In addition, the thymus epithelial cell marker ccl25a and T lymphocyte marker rag1 were expressed normally in the double mutants (Fig. S2C,D). Taken together, these results suggest that pdgfαa and pdgfαb are dispensable for definitive hematopoiesis. However, PDGF signaling has been reported to be necessary for HSC specification by guiding the migration of neural crest cells to the DA, and HSCs have been found to be significantly reduced in pdgfra-deficient embryos (Damm and Clements, 2017). Interestingly, in our study, the expression pattern of the neural crest marker crestin remained unchanged upon depletion of both pdgfαa and pdgfαb (Fig. S2E). This could be explained by the existence of other PDGF ligands that activate PDGFRs during HSC formation (Andrae et al., 2008; Wiens et al., 2010).

The dramatic decrease in the number of RBCs in pdgfαa^−/−;pdgfαb^−/− double mutants raised the possibility that pdgfαa and pdgfαb inhibit erythroid progenitor proliferation or accelerate their apoptosis. To test this possibility, terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) and bromodeoxyuridine (BrdU) incorporation assays were performed. TUNEL staining revealed no obvious apoptotic signal in the erythrocyte progenitor cells of wild-type and pdgfαa^−/−;pdgfαb^−/− embryos (Fig. S3A,B). Meanwhile, the BrdU incorporation assays showed no apparent difference in erythroid progenitor proliferation between groups (Fig. S3C,D). Furthermore, similar results were observed in embryos treated with inhibitor V (Fig. S4). These data indicate that loss of pdgfαa and pdgfαb does not affect the proliferation and survival of erythrocyte progenitor cells.

We then sought to examine the emergence of erythrocyte progenitor cells in pdgfαa^−/−;pdgfαb^−/− mutants. Because hemangioblasts arise out of the ventral mesoderm and differentiate into hematopoietic cells and vascular endothelial cells (Vogeli et al., 2006), we first evaluated the expression of ventral mesodermal genes eve1 and vent and found that ablation of pdgfαa and pdgfαb did not affect the formation of ventral mesoderm (Fig. S5). Moreover, expression analysis of the hemangioblast markers npas4l and scl (also known as tal1) in wild-type and mutant embryos at the two-somite stage revealed a comparable hemangioblast population (Fig. 2A). However, expression of the erythroid progenitor markers gata1 and hbbe3 was reduced in pdgfαa^−/−;pdgfαb^−/− mutants at the 10-somite stage (Fig. 2B), indicating impaired differentiation of hemangioblasts into erythroid progenitors. In contrast, we observed no significant difference in expression of etv2 between the pdgfαa^−/−;pdgfαb^−/− mutants and the wild-type control at the 10-somite stage, suggesting that the development of endothelial progenitors is not affected by deletion of pdgfαa and pdgfαb (Fig. 2C). Moreover, no apparent vascular defect was observed in pdgfαa^−/−;pdgfαb^−/− embryos at later developmental stages (Fig. 2D,E). Consistent with the above observations, suppressing PDGF signaling with inhibitor V did not alter the formation of hemangioblast and the development of endothelial progenitors, but severely reduced the emergence of erythrocyte progenitors (Fig. S6).

Because primitive hematopoiesis consists of erythropoiesis and myelopoiesis, we next asked whether the generation of myeloid cells was also impaired when both pdgfαa and pdgfαb were deleted. The expression profiles of pu.1 (also known as spi1b), l-plastin (also known as lcp1) and mpx revealed that the myeloid cells, including macrophages and granulocytes, were fully developed in both wild-type and mutant embryos (Fig. 2F-I).

Together, these data suggest that pdgfαa and pdgfαb are specifically required for the differentiation of erythroid progenitors from hemangioblasts during primitive hematopoiesis, but are unnecessary for the development of endothelial and myeloid lineages.

Given the importance of pdgfαa and pdgfαb for early erythroid cell fate decisions, we asked whether the expression patterns of these ligands and their receptors are consistent with a role in primitive erythropoiesis. We first mapped the temporal-spatial distribution of pdgfαa and pdgfαb transcripts by performing whole-mount in situ hybridization (WISH) experiments. Expression of pdgfαa was observed in the central nervous system at 14 hpf and emerged in truncal epidermis at 18 hpf (Fig. 3A). Interestingly, pdgfαb transcript was enriched in the central nervous system and the ventral portion of the somite known as the sclerotome (Fig. 3B,C). On the other hand, the expression of pdgfra, the receptor gene of pdgfαa and pdgfαb, was mostly detected in the posterior lateral plate mesoderm at 14 hpf and then appeared in the ICM at later developmental stages (Fig. 3D,E). Furthermore, dual-color fluorescence in situ hybridization analysis showed that pdgfra was specifically expressed in the erythrocyte progenitor cells expressing gata1 during somitogenesis (Fig. 3F,G). Such spatial-temporal proximity of ligand and receptor expression implies that PDGF signaling derived from sclerotomal cells might play an important role in primitive erythropoiesis.

In order to verify the above hypothesis, we conducted rescue experiments in pdgfαa^−/−;pdgfαb^−/− mutants by overexpression of pdgfαb under the control of the promoter of the sclerotome-specific gene foxc1b (Clements et al., 2011; Qiu et al., 2016). After injecting the foxc1b:pdgfαb-P2A-EosFP plasmid and Tol2 transposase mRNA into pdgfαa^−/−;pdgfαb^−/− embryos, robust green fluorescence was observed specifically in sclerotomal cells at 14 hpf (Fig. 3H). Concurrently, the decreased expression of gata1 in pdgfαa^−/−;pdgfαb^−/− mutants was markedly recovered, as revealed by WISH and quantitative real-time PCR experiments (Fig. 3I-K).

Binding of PDGF ligands to their receptors triggers receptor dimerization and autophosphorylation on tyrosine residues. As expected, ablation of pdgfαa and pdgfαb significantly downregulated the phosphorylation level of PDGFRs in erythrocyte progenitors, as revealed by immunostaining experiments using an antibody that detected phosphorylated Pdgfra (p-Pdgfra), whose specificity was confirmed by western blotting analysis of inhibitor V-treated embryos (Fig. 3L,M; Fig. S7A). Importantly, the level of p-Pdgfra in pdgfαa^−/−;pdgfαb^−/− mutants was restored by overexpression of pdgfαb in the sclerotomal cells (Fig. 3L,M). Collectively, the above findings support the idea that PDGF signaling molecules from the sclerotome are essential for erythroid progenitor differentiation in the adjacent posterior lateral plate mesoderm.

It has been reported that pdgfra morpholino (MO)-injected embryos display normal primitive erythropoiesis (Damm and Clements, 2017). We further confirmed the efficiency of the MO that has been used to knockdown pdgfra expression (Damm and Clements, 2017). Indeed, injection of the pdgfra MO into wild-type embryos markedly decreased the level of p-Pdgfra and caused clear defects in the ventromedial migration of trunk neural crest cells (Fig. S7B,C). However, the expression of gata1 in the ICM region of pdgfra morphants was not obviously changed (Fig. S7D,G). In general, PDGFR proteins form homo- or hetero-dimers to be activated by PDGF ligands (Kazlauskas, 2017; Shim et al., 2010). Interestingly, we found that pdgfrb was expressed in the hypochord and in the erythroid progenitor cells located near its ventral region (Fig. S7E), indicating that pdgfrb might be relevant to erythropoiesis. We then injected a previously validated antisense MO targeting pdgfrb into wild-type embryos (Wiens et al., 2010). We found that although pdgfrb single morphants showed no defect in primitive erythropoiesis, pdgfra and pdgfrb double morphants exhibited significantly reduced expression of gata1 and hbbe3 (Fig. S7F,G). These data indicate that pdgfra and pdgfrb function together to promote primitive erythropoiesis, similar to their role in regulating craniofacial development (Fantauzzo and Soriano, 2016).

As PDGFRs have been reported to activate several downstream signaling cascades, including the AKT-PI3K and ERK1/2 pathways (Heldin, 2013; Hoch and Soriano, 2003), we sought to determine which pathway plays a primary role downstream of PDGFRs in primitive erythropoiesis. Firstly, phosphorylated AKT(Ser473) (p-AKT) and phospho-ERK1/2 (both Mapk3 and Mapk1 in zebrafish) were examined in pdgfαa^−/−;pdgfαb^−/− embryos. Immunostaining analysis revealed robust signals of phosphorylated AKT kinases in the ICM region of both wild-type and mutant embryos (Fig. S8). In contrast, the immunostaining signal of phosphorylated ERK1/2 in the ICM region was almost eliminated in pdgfαa^−/−;pdgfαb^−/− mutants as compared with that of control animals (Fig. 4A,B).

To verify the necessity of ERK1/2 signaling during erythrocyte development, the specific inhibitor U0126 was used to block the phosphorylation of ERK1/2 (Hawkins et al., 2008). Wild-type embryos treated with U0126 displayed drastically decreased expression of gata1 at 14 hpf, and exhibited severe defects in erythropoiesis at 48 hpf, as indicated by o-dianisidine staining (Fig. 4C,D). Notably, specific overexpression pdgfαb in the sclerotome effectively restored the levels of phosphorylated ERK1/2 in the pdgfαa^−/−;pdgfαb^−/− mutants (Fig. 4E,F). Importantly, when a constitutively active form of Mek1 (caMEK1; Bolcome and Chan, 2010), the upstream activator of ERK1/2, was overexpressed in pdgfαa^−/−;pdgfαb^−/− embryos, the emergence of erythroid progenitors and the development of erythrocytes was restored (Fig. 4G-L). Taken together, these findings indicate that ERK1/2 signaling acts downstream of PDGF signaling to regulate primitive erythropoiesis.

Based on these results, we propose a model for the function of PDGF signaling in embryonic erythropoiesis (Fig. 4M). PDGF ligands derived from sclerotome interact with and activate their receptors in the nearby lateral plate mesoderm, subsequently activating the downstream ERK1/2 pathway to eventually support erythroid progenitor differentiation. Thus, our study highlights the crucial role of sclerotome-derived PDGF signaling as a primitive erythropoietic niche cue.

It has been reported that deficiency of PDGFB or PDGFRβ leads to multisystem disorders in mice, including hematopoietic defects such as anemia and thrombocytopenia (Levéen et al., 1994; Soriano, 1994). More specifically, PDGF signaling has been demonstrated to help the establishment of the definitive hematopoietic niche (da Bandeira et al., 2022; Damm and Clements, 2017). In our study, we found that knockout of both pdgfαa and pdgfαb in zebrafish caused a dramatic decrease in the number of embryonic RBCs in circulation, which could be attributed to the defective differentiation of erythropoietic progenitors. The expression of pdgfαb was enriched in sclerotome, and the receptor genes pdgfra and pdgfrb were expressed in the adjacent lateral plate mesoderm. Their spatial localization facilitates ligand-receptor binding and signal initiation, thereby activating downstream ERK1/2 signaling to promote erythroid progenitor differentiation. Indeed, overexpression of pdgfαb in the sclerotome could effectively restore RBC formation in pdgfαa^−/−;pdgfαb^−/− mutants. Our findings strongly imply that sclerotome-derived PDGF signaling serves as a niche cue responsible for primitive hematopoiesis. Interestingly, it has been suggested that PDGFB signaling in the placental microenvironment plays a role in protecting hematopoietic stem/progenitor cells from definitive erythroid differentiation (Chhabra et al., 2012), implying that PDGF signaling might have distinct roles in primitive and definitive erythropoiesis.

The following zebrafish (Danio rerio) lines were used: wild type (Tübingen), Tg(gata1:dsRed), Tg(flk:GFP), pdgfαa^−/−, pdgfαb^−/− and pdgfαa^−/−;pdgfαb^−/− (Mao et al., 2019). Zebrafish embryos collected from controlled pair- or group-mating were kept at 28.5°C in Holtfreter’s solution (Kimmel et al., 1995). All animal experiments were performed in accordance with the guidelines approved by the Committee on Animal Experimentation of South China University of Technology (permission number 2022096).

The primitive vector foxc1b:EosFP was kindly provided by Professor Anming Meng (Tsinghua University, China). foxc1b:pdgfαb-P2A-EosFP plasmid was created with a fusion expression of pdgfαb and EosFP under the control of the foxc1b promoter (Qiu et al., 2016). Capped mRNAs were synthesized in vitro for Tol2 and caMEK1 from the corresponding linearized plasmids (Bolcome et al., 2010) using the mMessage mMachine SP6 kit (Ambion). Both plasmid and mRNA were injected into embryos at the one-cell stage. In the rescue experiments, 200 pg Tol2 mRNA and 30 pg foxc1b:pdgfαb-P2A-EosFP plasmids were co-injected into embryos. To constitutively activate ERK signaling, 20 pg caMEK1 mRNA was injected into embryos. To disrupt PDGF signaling, 7 ng pdgfra MO (5′-ATGGTCACGTAGATTGTGCTCAGCT-3′; Kartopawiro et al., 2014) and/or 10 ng pdgfrb MO (5′-ACAGGAACTGAAGTCACTGACCTTC-3′; Wiens et al., 2010) was injected into wild-type embryos.

Digoxigenin-UTP-labeled antisense RNA probes for gata1, hbbe3, npas4l, runx1, cmyb, pu.1, lcp1, scl, etv2, flk (kdrl) (Liu et al., 2008), pdgfαa, pdgfαb, pdgfra and pdgfrb (Mao et al., 2019) were synthesized using a MEGAscript Kit (Ambion, USA) according to the manufacturer's instructions. WISH using these RNA probes was performed following standard procedures (Ning et al., 2013). The stained embryos were then processed to 10 µm thick cryosections. For dual-color fluorescence in situ hybridization, anti-digoxigenin-POD (1:400; 11633716001, Roche) and anti-fluorescein-POD (1:400; 11426346910, Roche) were used as primary antibodies to detect digoxigenin-labeled pagfra or pdgfrb probes and the fluorescein-labeled gata1 probe, respectively, and the embryos were then stained with fluorescein tyramide (1:50; PerkinElmer) and cyanine 3 tyramide (1:50; PerkinElmer), respectively.

Real-time quantitative PCR was performed as previously described (Yan et al., 2019). The primer sequences are listed in Table S1. The expression levels of β-actin was used as reference to normalize each sample.

Embryos were harvested at the indicated stages, and immunostaining was performed as previously described (Ning et al., 2013). The following antibodies were used: anti-DsRed (1:200; 632496, Clontech, USA), anti-mCherry (1:200; ab125096, Abcam, UK), anti-p-PDGFRα (phospho Y849) (1:250; ab79318, Abcam, UK), anti-p-ERK1/2 (1:250; 9101, Cell Signaling Technology, USA), anti-p-AKT (1:400; 4060, Cell Signaling Technology, USA) and anti-BrdU (1:1000; B2531, Sigma, Japan). The secondary antibodies used were as follows: Alexa Fluor^® 488 AffiniPure donkey anti-rabbit IgG (1:200; 711-545-152, Jackson ImmunoResearch), Alexa Fluor^® 594 AffiniPure donkey anti-mouse IgG (1:200; 715-585-150, Jackson ImmunResearch), Alexa Fluor^® 488 AffiniPure donkey anti-mouse IgG (1:200; 715-545-150, Jackson ImmunoResearch) and Alexa Fluor^® 594 AffiniPure donkey anti-rabbit IgG (1:200; 711-585-152, Jackson ImmunoResearch). Anti-p-PDGFRα (phospho Y849) (1:1000; ab79318, Abcam, UK) and anti-β-actin (1:5000; 66009, Proteintech, USA) were used for western blotting experiments with donkey HRP-conjugated anti-rabbit IgG (1:10,000; NA934V, GE Healthcare) and sheep HRP-conjugated anti-mouse IgG (1:7000; NA931V, GE Healthcare) as the secondary antibodies, respectively.

Confocal images of immunofluorescence or live embryos were captured using a Nikon A1R+ confocal microscope (20× or 10× objective), and the fluorescence intensity of immunofluorescence images was quantified using ImageJ version 1.48.

Wild-type embryos were treated with 0.25 μM PDGFR tyrosine kinase inhibitor V (521234, Calbiochem, Germany) from shield stage to the indicated stages and then harvested for in situ hybridization. To block ERK1/2 signaling, embryos were exposed to 25 μM U0126 (U120-1MG, Sigma, Japan) from shield stage and harvested in the indicated stages for further experiments.

Embryos were harvested at 48 hpf. Staining of hemoglobin by o-dianisidine was performed as previously described (Ransom et al., 1996).

Embryos at specific stages were subjected to 10 mM BrdU (B5002, Sigma) on ice for 20 min. The incorporated BrdU and DsRed protein were detected using anti-BrdU (1:1000; B2531, Sigma) and anti-DsRed (1:200; 632496, Clontech, USA), respectively, for whole-mount immunostaining. TUNEL assays were carried out using an In Situ Cell Death Detection Kit, TMR red (12156792910, Roche) following the manufacturer's instructions.

We are grateful to Dr Anming Meng (Tsinghua University, China) for providing the plasmid foxc1b:EosFP to construct the system of specifically overexpression of pdgfαb in sclerotomal cells.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201807.reviewer-comments.pdf