Abstract

A novel zebrafish gene bloody fingers (blf) encoding a 478 amino acid protein containing fifteen C₂H₂ type zinc fingers was identified by expression screening. As determined by in situ hybridization, blf RNA displays strong ubiquitous early zygotic expression, while during late gastrulation and early somitogenesis, blf expression becomes transiently restricted to the posterior dorsal and lateral mesoderm. During later somitogenesis, blf expression appears only in hematopoietic cells. It is completely eliminated in cloche, moonshine but not in vlad tepes (gata1) mutant embryos. Morpholino (MO) knockdown of the Blf protein results in the defects of morphogenetic movements. Blf-MO-injected embryos (morphants) display shortened and widened axial tissues due to defective convergent extension. Unlike other convergent extension mutants, blf morphants display a split neural tube, resulting in a phenotype similar to the human open neural tube defect spina bifida. In addition, dorsal ectodermal cells delaminate in blf morphants during late somitogenesis. We propose a model explaining the role of blf in convergent extension and neurulation. We conclude that blf plays an important role in regulating morphogenetic movements during gastrulation and neurulation while its role in hematopoiesis may be redundant.

Keywords: Zebrafish; Convergent extension; Neurulation; Zinc finger; Gastrulation; Neural tube; Spina bifida; Blood; Hematopoietic

Article Outline

Introduction

Materials and methods

Blf clone isolation

In situ hybridization

Northern blotting

MO injection

Cell-tracing

Real-time RT-PCR

Results

Isolation of the bloody fingers gene and its expression pattern

Blf expression in hematopoietic mutants

Morpholino knockdown of Blf

Extension and convergence of axial tissues are perturbed in Blf morphants

Neurulation and cell–cell adhesion defects in Blf morphants

Discussion

Acknowledgements

References

Introduction

Morphogenetic movements accompanied by coordinated cell migration and cell shape changes lead to the establishment of the body axis during vertebrate development. During zebrafish gastrulation, prospective mesendodermal cells involute and ingress. This is followed by convergence and extension movements in which mesendoderm and ectoderm undergo cell intercalations along the medial–lateral axis that narrow the tissues and consequently extend them along the anterior–posterior axis (reviewed in Heisenberg and Tada, 2002). Gastrulation is followed by neurulation during which the neural plate is specified and transformed into neural tube. In contrast to other vertebrates, neural folds are not evident in zebrafish; instead, a neural keel is formed (reviewed in Strahle and Blader, 1994). The neural tube is then shaped by oriented cell divisions and convergent extension (CE) movements. It has been proposed that the mechanism of teleost neurulation is similar to the secondary neurulation seen in the tailbud of higher vertebrates. However, recent studies have demonstrated the epithelial origin of the zebrafish neural keel, arguing that zebrafish neural tube formation may be just a variant of primary neurulation (Papan and Campos-Ortega, 1994 and Reichenbach et al., 1990; reviewed in Lowery and Sive, 2004).

Several zebrafish mutants have been identified which exhibit defects in convergent extension. Among them are pipetail (ppt), knypek (kny), and trilobite (tri) mutant embryos which exhibit a shortened body axis from late gastrulation stages onwards (Hammerschmidt et al., 1996, Marlow et al., 1998 and Solnica-Krezel et al., 1996). These loci have been shown to encode wnt5, glypican knypek, and strabismus, respectively, components or modulators of the wnt non-canonical signaling pathway (Jessen et al., 2002, Rauch et al., 1997 and Topczewski et al., 2001). This pathway is analogous to the planar cell polarity (PCP) pathway in Drosophila and is thought to regulate convergent extension movements in vertebrates (Darken et al., 2002, Heisenberg et al., 2000, Park and Moon, 2002, Shulman et al., 1998 and Wallingford et al., 2000). Some other transducers or modulators of the PCP pathway have also been shown to affect convergent extension in zebrafish including Frizzled-2 (Fz2) and Prickle (Pk) proteins (Carreira-Barbosa et al., 2003, Kilian et al., 2003, Sumanas et al., 2001 and Veeman et al., 2003).

Defects in convergent extension and neurulation are known to cause embryo abnormalities in many vertebrates, including humans. Incomplete closure of the neural tube, spina bifida, is among the most common birth defects contributing to infant mortality and serious disability (Copp et al., 2003). Mutations in the mouse orthologs of the Drosophila PCP genes disheveled, strabismus, scribble, and flamingo, all result in failure to initiate neural tube closure (Ueno and Greene, 2003). Loss of function of wnt, frizzled-7, disheveled, strabismus, and prickle has been shown to perturb gastrulation movements and result in open neural tube in Xenopus (Hoppler et al., 1996, Goto and Keller, 2002, Sumanas et al., 2000, Takeuchi et al., 2003, Wallingford and Harland, 2002 and Winklbauer et al., 2001). These data demonstrate the involvement of the PCP pathway in regulating both convergent extension and neurulation and illustrates the interdependence of both fundamental processes of morphogenesis in different vertebrates. However, the open neural tube phenotype has not been described in the zebrafish as yet.

In the current study, we have identified a gene encoding a novel zebrafish multiple C₂H₂ zinc finger protein family member, Bloody fingers (Blf). Zinc finger proteins contain a small peptide domain with a special secondary structure stabilized by a zinc ion bound to Cys and His residues of the finger (reviewed in Iuchi, 2001). The most common type of zinc fingers, C₂H₂, is primarily involved in DNA binding and transcriptional regulation. There are hundreds of different C₂H₂ proteins in vertebrate genomes. One of the subclasses of C₂H₂ proteins includes multiple-adjacent C₂H₂ zinc finger proteins. Members of this subclass can have as many as 29 adjacent zinc fingers such as in the protein Roaz (Tsai and Reed, 1997). In addition to DNA binding, zinc fingers in these proteins can also engage in protein–protein interactions to promote homo- or heterodimerization such as the one observed between the blood-specific Ikaros protein and two of its homologues, Alios and Helios (Kelley et al., 1998 and Morgan et al., 1997). Some of the multiple C₂H₂ domain proteins are also known to bind single-stranded or double-stranded RNA, e.g., TFIIIA and dsRBP-Zfa proteins (Finerty and Bass, 1997 and Friesen and Darby, 1998).

Blf protein is predicted to contain 15 sequential zinc fingers spanning almost the entire length of the protein with no other motives detectable. Blf RNA is expressed ubiquitously in early zebrafish embryos and is progressively restricted to the posterior dorsal and lateral mesoderm during gastrulation while later expressed in the hematopoietic progenitor cells. Knockdown of Blf protein using antisense morpholino oligonucleotides (MOs) revealed its critical role in convergent extension and neurulation movements. Blf-MO-injected embryos display shortened body axis, split neural tube similar to the spina bifida condition in humans, and delaminating dorsal ectodermal cells. This is the first time that a phenotype similar to spina bifida has been demonstrated in zebrafish suggesting that zebrafish can be used to model this human birth defect. We also propose a model explaining how defective convergent extension results in the open neural tube of blf morphants.

Materials and methods

Blf clone isolation

A cDNA clone of 1.92 kb encompassing a short stretch of the 5′UTR, the complete putative ORF sequence of blf and the 3′UTR was isolated from the zebrafish embryonic blood cell-specific cDNA library (Long et al., 2000). Using the 5′RACE kit (Promega), an additional 5′UTR sequence was isolated from synthesized cDNA derived from total purified RNA of 24 hpf zebrafish embryos. The combined size of the isolated blf cDNA and a predicted poly-A tail of 200–300 nucleotides is in close agreement with the experimentally observed blf size of 2.4 kb, as determined by Northern blotting, suggesting that the complete blf cDNA sequence was isolated.

In situ hybridization

In situ hybridization was performed as described (Jowett, 1999). To synthesize DIG-labeled antisense blf probe, a blf-pTriplEx (Clontech) construct was linearized with KpnI and transcribed with T7 RNA polymerase (Promega). Other probes used include: no tail (ntl) (Schulte-Merker et al., 1994); sonic hedgehog (shh) (Ekker et al., 1995); spondin-1b (spon1b) (Higashijima et al., 1997); myod (Weinberg et al., 1996); crestin (Luo et al., 2001); cb497 cDNA, corresponding to hairy-related 4 (her4) (Thisse et al., 2001); cb112 cDNA, similar to cytokeratin E7 (Thisse et al., 2001); gata1 (Detrich et al., 1995); gata2 (Detrich et al., 1995); tal1 (scl) (Liao et al., 1998); cmyb (Thompson et al., 1998); and ikaros (Willett et al., 2001). To synthesize a probe for the forkhead box A1 (foxa1), the ORF of foxa1 was amplified by PCR from the post-somitogenesis stage cDNA library (kindly donated by S. C. Ekker), subcloned into the 4-TOPO vector (Invitrogen), linearized with SpeI, and transcribed using T7 RNA polymerase.

Northern blotting

Northern hybridization was performed as described (Hopwood et al., 1989). Total RNA purified from 10 embryos was loaded on a denaturing agarose gel for each stage analyzed. To synthesize blf probe, the ORF of blf was amplified by PCR, subcloned into the SpeI site of pT3TS (Hyatt and Ekker, 1999). The blf-T3TS/SpeI fragment was then labeled with ³²P and used as a probe. Ethidium bromide staining of ribosomal RNA was used as a loading control.

MO injection

Three blf-specific MOs, MO1 (TGGATCACTCATTTTCTCTGTGTTC), MO2 (TTTCTCTGTGTTCTCCCTCCACTTG), MO3 (TTATGCTGGTTTCTGGATCACTCAT), and the 5-base MO2 mismatch (TTTCACTGTCTTCTGCCTGCACTAG) were used to study the function of Blf protein. 2–4 ng of 1 mg/ml MO solution in Danieau buffer (Nasevicius and Ekker, 2000) supplemented with 15 mM Tris–Cl (pH 7.5) was microinjected into zebrafish embryos at the one-cell to two-cell stage as described (Hyatt and Ekker, 1999).

To inhibit the function of draculin, 5–10 ng of dra1 and dra2 MOs (CAGGGTTTTGTTGTATTCTTCATCT and TCACGCTTTTCTCTATTTCCAAGTG) was injected. Also, the following combinations of blf and dra MOs were injected: 2–4 nl blf MO2/dra1 MO, 1:2 ratio, 1 mg/ml total MO concentration; 2–4 nl blf MO2/dra2 MO, 1:4 ratio, 1 mg/ml total; and 1–4 nl blf MO2/blf MO3/dra1 MO/dra2 MO, 1:1:1:1 ratio, 1 mg/ml total. The ratios were chosen to minimize non-specific MO effects observed at higher doses.

Cell-tracing

Cell labeling with DMNB-caged fluorescein dextran (Mw 10,000, Molecular Probes) was performed as described (Yamashita et al., 2002).

Real-time RT-PCR

To analyze gene expression level, batches of ten 2.5 ng blf MO2, 7.5 ng blf MO3-injected and uninjected control embryos were frozen on dry ice at the 10-somite stage. Total RNA was purified using the RNAquous-4PCR kit (Ambion). cDNA was synthesized using a mix of random hexamers (Roche) and the Powerscript reverse transcriptase (BD Biosciences). Real-time PCR was performed using the iCycler iQ Real-Time PCR Detection System (Bio-Rad) and the iQ SYBR Green Supermix (Biorad). Relative cDNA amounts were calculated using the iCycler program (Bio-Rad) and normalized to the expression of elongation factor 1α (EF1α). The following primer pairs were used: her4 AAGACACACAGCAATGACTCC, ACTAAAGCCGTGTGGTCATCG; ngn1 AGACTCTGCGCTTCGCTCAC, TTCTGGAGATTCTATACCGAC; foxb1.2 CGCCATCGAGAACATCATCG, CAGAGATTGCGGAGAGTTCG; crestin TTAGGATTGACTCCTTTCACC, TCTAAGCATGCGCGAGATGTG; gata3 TCAACCTTGAAGCCTCGCAC, GAGTCATTCCACCTTGCAGG; cytokeratin E7 ATCTGCGCAAGAAAATCTCC, ACTGCTGTCGCATCTCCTCC; and EF1α TCACCCTGGGAGTGAAACAGC, ACTTGCAGGCGATGTGAGCAG. The following PCR profile was used: 95°C 3 min; 95°C 10 s, 58°C 30 s, 72°C 30 s (data acquisition point), repeat 40 times. In the case of her4, data acquisition was performed at 84°C. Specific amplification of each product was confirmed by gel and melting curve analysis.

Results

Isolation of the bloody fingers gene and its expression pattern

A zebrafish cDNA library was prepared from the purified hematopoietic progenitor cells using gata1-GFP transgenic zebrafish (Long et al., 2000). One of the isolated cDNA clones encoded a novel 478 amino acid protein, Bloody Fingers (Blf), containing fifteen sequential C₂H₂ type zinc finger domains which comprised almost the entire length of the protein (Fig. 1). No other homology domains were found using BLAST searches of protein databases. All of the 15 zinc fingers are highly conserved; each of them is 21 amino acid (a.a.) long linked by the 7 a.a. long linker region. Although these zinc finger domains are highly similar to the ones in multiple zinc finger proteins from other vertebrates, we were unable to find a protein that has a similar structure of 15 zinc fingers from any other organism. The most structurally similar protein to Blf is the zebrafish protein Draculin, containing 13 sequential zinc finger domains (Herbomel et al., 1999) (Fig. 1). The two proteins share 42% similarity and 36% identity, displaying high homology in the zinc finger and most of the linker regions.

We investigated the expression pattern of blf using Northern blotting and in situ hybridization. No maternal expression was detected. A single band of 2.4 kb was observed during most of the zygotic stages analyzed (Fig. 2A). The strongest expression of blf was noted immediately after the start of the zygotic transcription, at the sphere, 30% and 45% epiboly stages (Fig. 2A, and data not shown). Expression decreased dramatically afterwards. blf mRNA was distributed uniformly throughout the embryo from the sphere to the shield stages, as analyzed by in situ hybridization (Fig. 2B, and data not shown). During later epiboly stages, blf RNA gradually became excluded from the animal pole and the most dorsal region of the embryo (Figs. 2C, D). Between the tailbud and the 4 somite stages, blf mRNA became localized to the posterior dorsolateral and lateral mesoderm, surrounding the yolk plug in a semicircular pattern (Figs. 2E, F). Expression at those stages was restricted to the mesodermal cell layer as determined by section analysis (data not shown). The expression region became tighter around the yolk plug and gradually disappeared as embryos showed very little expression of blf during the 5–7 somite stages. blf expression reappeared at the 8-somite stage in two bilaterally symmetrical stripes within the posterior lateral mesoderm, the region where hematopoietic precursor cells are known to originate. This expression domain was spatially distinct from the earlier blf expression close to the tailbud region. Expression of blf in presumptive hematopoietic cell precursors grew stronger during later somitogenesis (Figs. 2G–L). As blood circulation began at 24 hpf, blf expression was observed in the intermediate cell mass as well as in the circulating blood cells (Fig. 2M). blf expression became much weaker shortly after that and was not detectable after 30 hpf (data not shown).

Blf expression in hematopoietic mutants

Based on the blf expression pattern in hematopoietic mesoderm, we analyzed blf expression in four different hematopoietic mutants. Cloche (clo) mutants form no blood cells or blood vessels, and are thought to affect one of the earliest steps in hemangioblast formation (Stainier et al., 1995 and Thompson et al., 1998). blf-expressing cells were almost completely absent in 1/4 of the progeny (16 out of 68 embryos) from clo^m39 heterozygous carriers as analyzed by in situ hybridization at the 12–26 somite stages (Figs. 3A, B). Moonshine/vampire (mon) mutants are defective in the transcriptional intermediary factor 1γ (TIF1γ) and are bloodless at the onset of circulation (Ransom et al., 1996, Ransom et al., 2004 and Stainier et al., 1996). blf expression was almost completely absent in 1/4 of the progeny (4 out of 15 embryos) from mon^tu244b heterozygous carriers (Figs. 3C, D). blf expression was not affected in vlad tepes (vlt^m651) mutants (>40 embryos from heterozygous carriers analyzed) which are defective in the gata1 transcription factor and form very few circulating blood cells (Lyons et al., 2002 and Weinstein et al., 1996) (Fig. 3E). blf expression was not affected in merlot/chablis mutant embryos defective in the erythrocyte protein 4.1 (Shafizadeh et al., 2002; data not shown). These data suggest that Blf functions downstream of clo and TIF1γ but upstream of or independently from GATA1 and protein 4.1 in zebrafish hematopoiesis.

Morpholino knockdown of Blf

We used antisense morpholino oligonucleotides (MOs) to analyze the function of blf in embryogenesis (Nasevicius and Ekker, 2000). Injection of blf MOs resulted in undulations and curvature of axial structures, most notably the notochord (Fig. 4). Blf-MO-injected embryos (morphants) were much shorter than control uninjected embryos. Blf morphants also displayed delaminating cells from the dorsal ectoderm during mid-somitogenesis (Figs. 4E, H); most of these embryos died shortly after that. Three different blf-specific MOs caused the same phenotype, while a 5-base blf-mismatch MO and multiple blf-unrelated MOs never caused any of the described defects, demonstrating specificity of the blf knockdown phenotype (Fig. 4; data not shown).

Extension and convergence of axial tissues are perturbed in Blf morphants

To better understand the nature of the defects observed in blf morphants, we performed molecular analysis of different axial markers. The notochord in blf morphants was thicker and severely undulating as evident from the expression of no tail (ntl) and sonic hedgehog (shh) (Ekker et al., 1995 and Schulte-Merker et al., 1994) (Figs. 5A–D). Analysis of myod expression (Weinberg et al., 1996) revealed that the somites were broader, more closely spaced and disorganized while the paraxial mesoderm was curved, following the shape of the notochord in blf morphants (Figs. 5E, F). The floorplate was expanded and undulating, similarly to the notochord, in blf morphants as analyzed by the expression of shh, spondin1b (spon1b) (Higashijima et al., 1997), and foxA1 (Odenthal and Nusslein-Volhard, 1998) (Figs. 5C, D, G–J). In addition, the hypochord showed axial undulations as well (Figs. 5I, J). These results show that the axial tissues, including the axial and paraxial mesoderm, fail to extend and converge in blf morphants.

To assay convergent extension in blf morphants, we labeled cells by uncaging fluorescent dye at the shield stage. In control embryos (only injected with DMNB-caged fluorescein dextran), labeled lateral mesendodermal cells converged at the midline by the 2-somite stage (Figs. 6A, B). In blf morphants, lateral mesendodermal cells failed to converge at the midline and remained positioned laterally (Figs. 6C, D). These results confirm the defective convergence in blf morphants.

Neurulation and cell–cell adhesion defects in Blf morphants

To understand delamination of ectodermal cells observed in blf morphants, we analyzed the formation of neural tissue in greater detail. We used hairy-related 4 (her4) as a marker for neuroectoderm (Takke et al., 1999 and Thisse et al., 2001) and gata3 as a marker for non-neural ectoderm (Neave et al., 1995 and Thisse et al., 2001). During early somitogenesis, the neuroectoderm was expanded in blf morphants, as evident from a wider expression area of her4 in the neural plate (Figs. 7A, B, E, F) and a wider gap from which gata3 expression was excluded (Figs. 7C, D, G, H). Also, during mid-somitogenesis, the expanded neural plate became split into two across the midline (Figs. 7I, J, M, N). The dorsal cells along the midline located inside of the split neural tube expressed the non-neural ectodermal marker gata3 and the epidermal marker cytokeratin E7 (Thisse et al., 2001) (Figs. 7G, H, K, L, O, P). The delaminating cells showed a very strong expression of cytokeratin E7 demonstrating their epidermal nature (Figs. 7K, L, O, P). The expression of crestin in pre-migratory neural crest cells (Luo et al., 2001) was split into two stripes, following the shape of the split neural tube (Figs. 7Q, R). No or very little migration of the neural crest cells was observed in the blf morphants (Figs. 7S, T).

The expanded neural plate observed in blf morphants during early somitogenesis may form due to either the failure of the neuroectodermal cells to converge or to increased neural induction. We distinguished between the two models by assaying a number of neural and ectodermal markers using real-time RT-PCR (Table 1). Expression of the neural markers neurogenin (ngn1) (Korzh et al., 1998) and foxb1.2 (Odenthal and Nusslein-Volhard, 1998 and Thisse et al., 2001) was not significantly affected in blf morphants (Table 1). Expression of the neural marker her4 and neural crest-specific crestin was somewhat downregulated in blf morphants, while non-neural ectoderm-specific gata3 and epidermal cytokeratin E7 were slightly upregulated. The lack of increase in neural marker expression in blf morphants supports the hypothesis that the observed neurulation defects in blf morphants are caused by defective morphogenetic movements and not by inductive events. Downregulation of certain neural markers and upregulation of genes specific to non-neural ectoderm is consistent with a subset of neural cells adopting epidermal fate and delaminating in blf morphants (see Discussion).

We investigated formation of dorsal mesoderm during gastrulation by analyzing goosecoid (gsc) expression (Stachel et al., 1993). No significant difference in gsc expression was found between wild-type embryos and blf morphants at the 60% epiboly stage (data not shown).

Given the expression of blf in hematopoietic cells at later stages, we performed extensive analysis of blood cell formation in the blf morphants. We used moderate doses of MO2 and MO3 morpholinos which allowed survival of the injected embryos until 36 hpf or longer. Both MO2- and MO3-injected embryos formed approximately normal numbers of blood cells in circulation as determined by visual observation and o-dianisidine heme staining (Detrich et al., 1995) (data not shown). At 20–24 hpf, expression of the early hematopoietic markers gata1, gata2, tal1/scl, cmyb, and ikaros was not significantly affected as analyz