Fibroblast growth factors (Fgfs) function as key secreted signalling molecules in many developmental events. The zebrafish is a powerful model system for the investigation of embryonic vertebrate haematopoiesis. Although the effects of Fgf signalling on haematopoiesis in vitro have been reported, the functions of Fgf signalling in haematopoiesis in vivo remain to be explained. We identified Fgf21 in zebrafish embryos. Fgf21‐knockdown zebrafish embryos lacked erythroid and myeloid cells but not blood vessels and lymphoid cells. The knockdown embryos had haemangioblasts and haematopoietic stem cells. However, the knockdown embryos had significantly fewer myeloid and erythroid progenitor cells. In contrast, Fgf21 had no significant effect on cell proliferation and apoptosis in the intermediate cell mass. These results indicate that Fgf21 is a newly identified factor essential for the determination of myelo‐erythroid progenitor cell fate in vivo.
Vertebrate haematopoietic cells are derived from self‐renewing multipotential stem cells (Thisse & Zon, 2002). Haematopoiesis in vertebrates, from zebrafish to humans, is an evolutionarily conserved programme. The zebrafish is a powerful model system for the investigation of embryonic vertebrate haematopoiesis. In zebrafish embryos, primitive haematopoiesis occurs in the intermediate cell mass (ICM) and the anterior lateral mesoderm (ALM) and gives rise to progenitors that differentiate into embryonic blood cells. Definitive haematopoiesis occurs in the kidney marrow, which remains the principal site of this process throughout adulthood. Vascular cells and haematopoietic stem cells are derived from haemangioblasts that are derived from ventral mesoderm. Erythroid progenitor and lymphoid progenitor cells are derived from haematopoietic stem cells. Myeloid progenitors are assumed to be derived from haematopoietic stem cells by analogy with amniote definitive haematopoiesis (Hsu et al, 2001; Lieschke et al, 2002; Davidson & Zon, 2004; Ransom et al, 2004).
Many key haematopoietic transcription factor genes have been identified (Davidson & Zon, 2004). Some key secreted signalling molecules in haematopoiesis have also been identified. Disruption of bone morphogenetic protein (Bmp) signalling leads to a reduction or a lack of ventral mesoderm, resulting in an absence of blood cells (Thisse & Zon, 2002). Hedgehog signalling is one of the mechanisms in the regulation of both vasculogenesis and haematopoiesis (Vokes et al, 2004; Detmer et al, 2005). Vascular endothelial growth factor (Vegf) is a crucial factor for vascular development. Vegf can stimulate endothelial cell differentiation and haematopoiesis (Liang et al, 2001). Fibroblast growth factors (Fgfs) also function as key secreted signalling molecules in many developmental events. The vertebrate Fgf family is large, comprising 22 members (Itoh & Ornitz, 2004). However, although the effects of Fgf signalling on haematopoiesis in vitro have been reported (Moroni et al, 2002), the functions of Fgf signalling in haematopoiesis in vivo remain to be explained.
The zebrafish animal model offers a unique opportunity to discover and study novel genes required for the control of normal vertebrate developmental events. Mouse Fgf21 was expressed in the liver (Nishimura et al, 2000). However, roles of Fgf21 in developmental events remain to be explained. We identified zebrafish Fgf21 and inhibited Fgf21 signalling in zebrafish embryos. Fgf21‐knockdown zebrafish embryos lacked erythroid and myeloid cells but not blood vessels and lymphoid cells, indicating that Fgf21 is essential for the determination of myelo‐erythroid progenitor cell fate.
Results and Discussion
We identified zebrafish Fgf genes by using the basic local alignment search tool (BLAST) to search for the zebrafish genomic sequences and expressed sequenced tag sequences with the amino‐acid sequences of human FGFs. Then, we isolated their full‐length complementary DNAs from zebrafish embryo cDNA. One of the cDNAs encodes a protein of 194 amino acids that is most homologous to human FGF21 among human FGFs. A sequence‐based comparison of this protein with human FGFs by phylogenetic analysis also indicates that the protein is most closely related to human FGF21 (Fig 1A). However, although human FGF21 and mouse Fgf21 are highly similar (∼75% amino‐acid identity; Nishimura et al, 2000), this zebrafish protein is less similar (∼34% amino‐acid identity) to human FGF21 (data not shown). Human FGF21 is closely associated with carbonic anhydrase XI (CA11) and dehydrogenase/reductase member 10 (DHRS10) on chromosome 19 at q13.33 (Fig 1B). Therefore, we examined the location of this gene in the zebrafish genome. This gene was also closely associated with ca11 and dhrs10 on chromosome 3 (Fig 1B). These results indicate that this gene is zebrafish Fgf21 (accession number AB209932).
Injections of antisense morpholino oligonucleotides (MOs) have been demonstrated to inhibit the corresponding gene functions in zebrafish embryos efficiently and specifically (Nasevicius & Ekker, 2000). We injected Fgf21 antisense and five‐base‐mismatched Fgf21 antisense (control) MOs (10 ng per embryo) into zebrafish two‐cell embryos. Control MO‐injected embryos at 36 h post‐fertilization (hpf) developed as well as did wild‐type embryos (Fig 2Aa,b). In contrast, Fgf21 MO‐injected embryos developed abnormally, showing bent trunks (Fig 2Ac). Control MO embryos as well as wild‐type embryos at 30 hpf had erythroid cells detected by o‐dianisidine staining (n=39/40, 98%; Fig 2Ad,e). However, Fgf21 MO embryos had essentially no erythroid cells (n=45/68, 66%; Fig 2Af). We prepared capped zebrafish Fgf21 RNA including the entire coding region but not the 5′ non‐coding region. As the Fgf21 MO sequence (25 bases) includes 13 bases of the 5′ non‐coding region, the translation of the Fgf21 RNA should be unaffected by the Fgf21 MO. We examined whether the Fgf21 RNA could rescue the phenotype of Fgf21 MO embryos. The loss of erythroid cells in Fgf21 MO embryos was definitely rescued by injection of the Fgf21 RNA (10 pg per embryo; n=27/32, 84%; Fig 2Ag). In addition, we injected splice‐site‐targeted Fgf21 MO (E2I2), the sequence (25 bases) of which corresponds to that between the second exon and intron of the coding region, into two‐cell embryos (Fig 2B). Fgf21 MO (E2I2) embryos also had essentially no erythroid cells (n=19/30, 63%; Fig 2Ah). To determine the expression of Fgf21 in embryos, RNA was isolated from wild‐type and Fgf21 MO (E2I2) embryos at 24 hpf. Fgf21 cDNA was amplified from the RNA by reverse transcription–PCR (RT–PCR; Fig 2B,C). The expression of Fgf21 in Fgf21 MO (E2I2) embryos was significantly decreased in comparison with that in wild‐type embryos (Fig 2C). Furthermore, we also examined the effect of the Fgf receptor inhibitor SU5402 (Mohammadi et al, 1997) on erythroid cells. Embryos were treated with SU5402 during 11–13 hpf. The SU5402‐treated embryos lacked erythroid cells, in a dose‐dependent manner (n=6/14, 43% at 10 μM; n=66/70, 94% at 100 μM; Fig 2Ai). These results indicate that Fgf21 is required for erythropoiesis.
Angiogenesis in the embryonic tail and trunk produces functional intersegmental vessels. With microangiography using fluorescein isothiocyanate–dextran (Weinstein et al, 1995), we visualized the functional vascular system in zebrafish embryos at 36 hpf. The blood vessels in the trunk and tail were generated in Fgf21 MO embryos (Fig 2Aj,k). flk1 is a marker gene for endothelial cells of the blood vessels (Thisse & Zon, 2002). The expression of flk1 was observed in the blood vessels of wild‐type embryos at 24 hpf. The expression was also observed in the blood vessels of Fgf21 MO embryos (Fig 2Al,m). These results indicate that Fgf21 is not essential for vasculogenesis.
The temporal expression of Fgf21 during zebrafish embryonic development was first examined by RT–PCR (Fig 3A). The expression of Fgf21 was weakly detected at 14 hpf and subsequently increased at least until 48 hpf. We examined the expression of Fgf21 in zebrafish embryos during 12–48 hpf by whole‐mount in situ hybridization using a digoxigenin‐labelled antisense Fgf21 probe. The definite expression of Fgf21 was not detected in embryos at 12 and 14 hpf (data not shown). The expression of Fgf21 was detected in the trunk adjacent to the ICM at 24, 36 and 48 hpf (Fig 3Ba–c). To determine the Fgf21 expression region, the transverse sections of the trunk at 24 hpf were examined. Fgf21 was expressed in the notochord adjacent to the ICM, where gata1 was abundantly expressed (Fig 3Bd,e). However, some notochord‐less mutants (ntl, flh and boz) do not seem to have an obvious erythroid defect (Chin et al, 2000); therefore, the roles of Fgf21 expressed in the notochord remain to be explained.
mpo and l‐plastin are marker genes for heterophil granulocytes and macrophages, respectively (Fig 4A; Crowhurst et al, 2002). The expression of mpo and l‐plastin was greatly decreased in the ICM of Fgf21 MO embryos (Fig 4Ba–d). rag1 is a marker gene for lymphoid cells (Davidson & Zon, 2004). The expression was essentially unchanged in the thymus of Fgf21 MO embryos at 4 days post‐fertilization (dpf; Fig 4Be,f). These results indicate that Fgf21 is crucial for myelopoiesis but not lymphopoiesis.
Distinct types of blood cells are derived from haematopoietic stem cells as a result of a highly conserved gene programme (Thisse & Zon, 2002). Haematopoietic stem cells are derived from haemangioblasts. scl and gata2 are marker genes for haemangioblasts and haematopoietic stem cells, respectively (Fig 4A). The expression of scl and gata2 was observed in the ICM of wild‐type embryos at 12 and 18 hpf, respectively (Fig 4Bg,i). Their expression was also observed in Fgf21 MO embryos, indicating that Fgf21 is not required for haemangioblasts and haematopoietic stem cells (Fig 4Bh,j). Erythroid progenitors are derived from haematopoietic stem cells. Myeloid progenitors are assumed to be derived from haematopoietic stem cells by analogy with amniote definitive haematopoiesis. Erythroid and myeloid cells are derived from myeloid and erythroid progenitors (Hsu et al, 2001; Lieschke et al, 2002; Davidson & Zon, 2004; Ransom et al, 2004). The interplay of pu.1 and gata1 determines the myelo‐erythroid progenitor cell fate in zebrafish. gata1 and pu.1 determine whether myelo‐erythroid progenitors become erythroid and myeloid cells, respectively (Fig 4A; Galloway et al, 2005; Rhodes et al, 2005). The expression of gata1 was greatly reduced in the ICM of Fgf21 MO embryos at 13 and 24 hpf (Fig 4Bk–n). The expression of pu.1 was also greatly reduced in the ICM of Fgf21 MO embryos at 24 hpf (Fig 4Bo,p). These results indicate that Fgf21 is required for the induction of myelo‐erythroid progenitor cells. However, as we did not examine the expression of pu.1 in the anterior ventral mesoderm at 13–14 hpf, where most of the pu.1‐positive cells were born (Lieschke et al, 2002), we could not eliminate the possibility that the migration of pu.1‐positive cells around the embryo was inhibited. We also examined the expressions of runx1 and c‐myb, marker genes for putative definitive progenitors, which were expressed in the dorsal aorta at 48 hpf (Davidson & Zon, 2004). Their levels were essentially unchanged in the dorsal aorta of Fgf21 MO embryos (data not shown).
Proliferating cells, phosphohistone H3 (H3P)‐positive cells, were examined in the ICM of zebrafish embryos by immunohistochemistry using an anti‐H3P antibody. H3P‐positive cells were detectable in the ICM of wild‐type embryos at 13 and 24 hpf (Fig 5A,C). The rates of H3P‐positive cells in the ICM were not significantly affected by the injection of Fgf21 MO (Fig 5B,D–F). We also examined apoptosis in the ICM of wild‐type and Fgf21 MO embryos at 13 and 24 hpf by TdT‐mediated dUTP nick end labelling (TUNEL) assay. A few and essentially no apoptotic cells were detectable in the ICM of embryo at 13 and 24 hpf, respectively (Fig 5G,I). The rates of apoptotic cells in the ICM were not significantly affected by injection of Fgf21 MO (Fig 5H,J). In contrast, apoptotic cells were abundantly detectable in the head region of Fgf19 MO embryos at 16 hpf as a positive control (data not shown; Miyake et al, 2005). These findings indicate that Fgf21 is not essential for cell proliferation and apoptosis in the ICM.
Four Fgf receptor genes (Fgfr1–Fgfr4) have been identified in zebrafish, and their expression profiles were compared in embryos (Tonou‐Fujimori et al, 2002). No obvious erythroid defect was reported in Fgfr1‐knockdown embryos (Scholpp et al, 2004). However, as the receptor specificity of Fgf21 is not known, it remains to be explained which Fgf receptor mediates the activity of Fgf21 in haematopoiesis.
Some key secreted signalling molecules have crucial roles in haematopoiesis. Bmp is required for ventral mesoderm, a lack of which results in an absence of blood cells (Thisse & Zon, 2002). Hedgehog signalling is crucial for both vasculogenesis and haematopoiesis (Vokes et al, 2004; Detmer et al, 2005). Vegf can stimulate endothelial cell differentiation and haematopoiesis (Liang et al, 2001). In contrast, Fgf21 is essential for myelo‐erythroid progenitor cell fates. Fgf21 is a secreted signalling molecule that has a unique role in haematopoiesis. These findings indicate for the first time that Fgf signalling functions in developmental haematopoietic events in vivo.
Fish maintenance. Zebrafish (Danio rerio) were maintained as described (Westerfield, 2000). Embryos were obtained by natural spawning and cultured at 28.5°C in zebrafish Ringer's solution (Kimmel et al, 1995).
Isolation and characterization of zebrafish Fgf21 cDNA. Using BLAST to search for zebrafish genomic DNA sequences with the amino‐acid sequence of human FGF21, zebrafish Fgf21 genomic DNA was identified. To determine the 5′ non‐coding region, zebrafish embryonic cDNA was analysed by the rapid amplification of cDNA ends method (Frohman et al, 1988). The full‐length cDNA was isolated by PCR with primers (5′‐CTTAGTCAGTCATGCTTTTTGC‐3′, 5′‐TGCTAATGTAAAGCATGTCTT‐3′) and the cDNA as a template. The apparent evolutionary relationships of zebrafish Fgf21 and members of the human FGF family were examined using Clustal W. The positions of zebrafish Fgf21, dhrs10 and ca11 on chromosome 3 were obtained from Ensembl Zebrafish Genome Browser. The positions of human FGF21, DHRS10 and CA11 on chromosome 19 were obtained from LocusLink.
Whole‐mount in situ hybridization. Digoxigenin‐labelled RNA probes were synthesized by in vitro transcription using T7 or SP6 RNA polymerase. A 0.59‐kb Fgf21 probe was synthesized using the full‐length cDNA‐containing plasmid. Other probes used were zebrafish flk1 (AF487829), scl (AF045432), gata2 (BC053131), gata1 (U18311), pu.1 (AF321099), mpo (AF34904), l‐plastin (BC062381) and rag1 (U71093). Whole‐mount in situ hybridization was performed as described (Koshida et al, 1998). To detect Fgf21 expression, whole‐mount in situ hybridization was performed using the Tyramide Signal Amplification (TSA) system (Perkin Elmer).
Fgf21 knockdown by Fgf21 morpholino oligonucleotide. MOs were synthesized by Gene‐Tools, LLC (Corvallis, OR, USA). The sequences of the MOs used are as follows: Fgf21 MO, the sequence (25 bases) of which corresponds to the 5′ non‐coding region (13 bases) and the coding region (12 bases), 5′‐GGCAAAAAGCATGACTGACTAAGCT‐3′; five‐base‐mismatched Fgf21 (control) MO, 5′‐GGaAAtAAGCATcACTGAgTAAcCT‐3′; splice‐site‐targeted Fgf21 MO (E2I2), the sequence (25 bases) of which corresponds to that between the second exon and intron of the coding region, 5′‐GCAGTGGTTGCTTACAATTTAGATT‐3′. Lowercase letters in the control Fgf21 MO indicate mismatched bases in Fgf21 MO. MOs were diluted in Danieau buffer (Nasevicius & Ekker, 2000). Fgf21 MO (14 μg/μl) and control Fgf21 MO (14 μg/μl) were injected at a volume of 0.70–0.80 nl into two‐cell embryos. To see erythroid cells, the embryos were processed for o‐dianisidine staining (Paffett‐Lugassy & Zon, 2005). For whole‐mount in situ hybridization, the embryos were fixed in 4% paraformaldehyde.
Fgf21 morpholino rescue experiments with Fgf21 RNA. Capped zebrafish Fgf21 sense RNA was synthesized using the mMESSAGE mMACHINE kit (Ambion, Austin, TX, USA) according to the manufacturer's protocol from linearized pCS2+ plasmids containing the entire coding region of zebrafish Fgf21 cDNA. For Fgf21 morpholino rescue experiments, 10 pg of Fgf21 sense RNA was injected separately, immediately after the injection of Fgf21 MO into two‐cell embryos. The embryos were processed for o‐dianisidine staining (Paffett‐Lugassy & Zon, 2005).
SU5402 treatments. SU5402 (Calbiochem, San Diego, CA, USA) was dissolved in dimethyl sulphoxide at a concentration of 20 mM. Embryos were incubated during 11–13 hpf in SU5402 diluted to a concentration of 10 or 100 μM in zebrafish Ringer's solution.
Detection of proliferating and apoptotic cells. Proliferating cells were detected by immunohistochemistry using a rabbit polyclonal anti‐phosphorylated histone H3 (H3P) antibody (Upstate Biotechnology, Lake Placid, NY, USA) as described previously (Maroon et al, 2002). Apoptotic cells were examined by TUNEL assay using the DeadEnd colorimetric detection kit (Promega, Madison, WI, USA) as described previously (Maroon et al, 2002).
This work was supported by the 21st Century COE Program from the Ministry of Education, Culture, Sports, Science and Technology, Japan (H.Y.), the Takeda Science Foundation, Japan (N.I.) and the Mitsubishi Foundation, Japan (N.I.).
- Copyright © 2006 European Molecular Biology Organization