|Home | About | Journals | Submit | Contact Us | Français|
Recent studies demonstrate that lysyl oxidase cuproenzymes are critical for zebrafish notochord formation, but the molecular mechanisms of copper-dependent notochord morphogenesis are incompletely understood. We therefore conducted a forward genetic screen for zebrafish mutants that exhibit notochord sensitivity to lysyl oxidase inhibition, yielding a mutant with defects in notochord and vascular morphogenesis, puff daddygw1 (pfdgw1). Meiotic mapping and cloning reveal that the pfdgw1 phenotype results from disruption of the gene encoding the extracellular matrix protein fibrillin-2, and the spatio-temporal expression of fibrillin-2 is consistent with the pfdgw1 phenotype. Furthermore, each aspect of the pfdgw1 phenotype is recapitulated by morpholino knockdown of fibrillin-2. Taken together, the data reveal a genetic interaction between fibrillin-2 and the lysyl oxidases in notochord formation and demonstrate the importance of fibrillin-2 in specific early developmental processes in zebrafish.
Structural birth defects are a significant cause of morbidity and mortality in humans and result from both genetic and environmental factors (Murray, 2002; Shaw et al., 2003; Detrait et al., 2005). Nutrition is a critical environmental factor affecting early development (Miles et al., 2005), and deficiencies in several nutrients during pregnancy are known to cause structural birth defects (Detrait et al., 2005; Shaw et al., 2006). As nutritional status can now be increasingly interpreted within a specific genetic context, identification of alleles that confer susceptibility to nutritional or metabolic abnormalities may allow for therapeutic intervention or prevention of birth defects. Zebrafish provide a genetically tractable model to examine gene-nutrient interactions during early vertebrate development as early morphogenesis can be directly observed and levels of specific nutrients precisely controlled in the embryonic milieu via pharmacologic manipulation (Jensen et al., 2006; Mendelsohn et al., 2006; Gansner et al., 2007).
The zebrafish notochord is well suited to the study of gene-nutrient interactions during development. This prominent organ is easily visualized, consisting of a series of notochord vacuolar cells that exert turgor pressure on an extracellular matrix sheath (Scott and Stemple, 2005). The notochord provides axial support to the growing embryo, is essential for subsequent vertebral column formation, and secretes signaling molecules required for neural patterning, muscle differentiation, and cardiac formation (Stemple, 2005). Furthermore, defects in extracellular matrix formation lead to compromised notochord sheath function and notochord abnormalities and therefore can be readily detected. Indeed, large forward genetic screens (Odenthal et al., 1996; Stemple et al., 1996) have established the importance of laminins and coatomer proteins in notochord differentiation and sheath formation (Parsons et al., 2002; Coutinho et al., 2004).
The fibrillins are a family of extracellular matrix proteins that form 10 nm-diameter microfibrils around the notochord and in other tissues during development (Gallagher et al., 1993; Wunsch et al., 1994; Quondamatteo et al., 2002; Hubmacher et al., 2006). The human genome encodes three fibrillins which are numbered sequentially, have a highly modular and conserved domain organization, and exhibit about 68% amino acid identity with each other (Corson et al., 2004). Because of this similarity, fibrillin family members have historically been distinguished by a few salient features, including a single internal unique region which is proline-rich in fibrillin-1, glycine-rich in fibrillin-2, and proline/glycine-rich in fibrillin-3 (Hubmacher et al., 2006). The composition of this domain, which may act as a hinge region (Haston et al., 2003), is thought to contribute to functional differences between individual family members (Hubmacher et al., 2006). In addition, the number and location of integrin-binding RGD-motifs and N-glycosylation sites differ between family members (Hubmacher et al., 2006). While little is known about fibrillin-3, the other fibrillins have a long history and were first studied in avian development using monoclonal antibodies to fibrillin-1 (clone 201) (Sakai et al., 1986) and fibrillin-2 (the JB3 antibody) (Wunsch et al., 1994). In particular, the role of fibrillin-2 in embryo patterning and axis formation has been extensively studied in avians (Rongish et al., 1998; Czirok et al., 2004; Czirok et al., 2006).
Our laboratory has recently delineated the phenotype of copper deficiency in a zebrafish model of Menkes disease, which includes a strikingly distorted notochord (Mendelsohn et al., 2006). This notochord distortion is due to the inhibition of two specific copper-dependent lysyl oxidases that normally crosslink collagens and other extracellular matrix proteins in the notochord sheath (Gansner et al., 2007). Partial disruption of collagen II, a lysyl oxidase substrate and notochord sheath protein, sensitizes embryos to notochord distortion after suboptimal copper nutrition or partial lysyl oxidase inhibition, demonstrating a complex interplay of gene dosage and nutrient availability critical to notochord formation (Gansner et al., 2007). To understand the molecular mechanisms of notochord formation in relation to copper nutrition, we conducted a forward genetic screen for mutants exhibiting notochord sensitivity to lysyl oxidase inhibition. In this current study, we report the identification and characterization of one such mutant with defects in notochord and vascular morphogenesis, puff daddy (pfdgw1).
To elucidate the role of copper homeostasis in development, we performed a screen that couples N-ethyl-N-nitrosourea mutagenesis to pharmacologic sensitization using a sub-threshold dose of the lysyl oxidase inhibitor 2-mercaptopyridine-N-oxide (Anderson et al., 2007). A mutant, puff daddy (pfdgw1), was identified with notochord abnormalities (Fig. 1B, black arrow) and normal melanin pigmentation (Fig. 1B, white arrowhead) characteristic of the previously identified phenotype of impaired lysyl oxidase function. pfdgw1 mutants also display a cavernous caudal vein (Figs. 1B, D, white arrows) and fin fold attenuation (Figs. 1B, D, black arrowheads) not present in wild-type (+/+ and +/-) embryos at 30 hpf (Figs. 1A, C). Skin distention secondary to edema occurs in the lateral truncal region near the yolk-sac extension of pfdgw1 mutants by 30 hpf (Fig. 1F vs. Fig. 1E, arrowheads), and blood cell extravasation is visible in this space (Fig. 1F, arrow).
To examine vascular development in pfdgw1 fish, pfdgw1 was crossed to a transgenic line that expresses enhanced green fluorescent protein in vascular endothelial cells (Lawson and Weinstein, 2002). In wild-type embryos, the caudal vein forms a venous plexus with a characteristic reticular pattern (Fig. 2A, arrowheads), and the dorsal aorta and cardinal vein are appropriately lumenized (Fig. 2C, circles). In pfdgw1 mutants, endothelial cells are disorganized around a cavernous caudal vein (Fig. 2B, arrowheads), and the diameters of the large axial vessels are reduced to a variable degree (Fig. 2D, circles). Blood cells do not circulate in mutants with particularly small-diameter axial vessels despite a pumping heart (data not shown), and increased vascular resistance due to reduced vessel diameter may contribute to the observed impaired heart contractility in pfdgw1 mutants (data not shown). The distended area in the lateral truncal region of pfdgw1 mutants (Fig. 1F, arrowheads) is not lined by fli1-expressing cells (data not shown), a finding consistent with edema. While the notochord, fin fold, and vascular abnormalities are consistently penetrant in clutches from pfdgw1 fish, by 3 dpf, the truncal edema resolves and blood flow occurs through the caudal vein in most pfdgw1 mutants. However, the swim bladder does not inflate, resulting in embryonic lethality (data not shown).
The phenotype of pfdgw1 mutants suggested a defect in notochord sheath formation, and we therefore imaged this organ in pfdgw1 mutants and wild-type embryos by transmission electron microscopy (Figs. 3A-D). Ultrastructurally, the notochord sheath consists of inner, medial, and outer layers, all of which are clearly visible in a cross-section from a wild-type embryo at 30 hpf (Figs. 3A, C). While the inner and medial sheath layers are present in pfdgw1 mutants, the outer layer is strikingly diminished in size at the region of notochord folding (Figs. 3B, D).
Interestingly, lysyl oxidase inhibition results in impaired notochord formation without an effect on the outer sheath layer (Figs. 3E, F and Gansner et al., 2007), suggesting that the product of the pfdgw1 locus has roles in addition to those related to lysyl oxidase crosslinking. To examine this, we first incubated clutches from pfdgw1/+ intercrosses in vehicle (Figs. 4A, B and Table 1) or a dose of the copper chelator neocuproine that does not cause notochord distortion in wild-type embryos (Fig. 4C, Table 1, and Gansner et al., 2007). Consistent with the results of our original screen, pfdgw1 mutants were sensitized to increased notochord distortion under copper-limiting conditions (Fig. 4D vs. Fig. 4B and Table 1), presumably due to partial inhibition of lysyl oxidase activity (Gansner et al., 2007). To confirm this finding, we incubated clutches from pfdgw1/+ intercrosses in a dose of the irreversible lysyl oxidase inhibitor β-aminopropionitrile that causes very mild notochord herniation in wild-type embryos (Fig. 4E, white arrowhead). Under these conditions, notochord sensitization in pfdgw1 mutants was again observed (Fig. 4F vs. Fig. 4B and Table 1). Importantly, pfdgw1 mutants could be distinguished from heterozygote and wild-type siblings in these experiments by the cavernous caudal vein and truncal edema present in the mutants. These additional phenotypes are never observed after lysyl oxidase inhibition with neocuproine or β-aminopropionitrile, even when higher doses of these compounds are used (Gansner et al., 2007).
We next tested for a genetic interaction between the lysyl oxidases involved in late notochord formation (Gansner et al., 2007) and the pfdgw1 product by injecting clutches from pfdgw1/+ intercrosses with control morpholino (Figs. 5A, B and Table 2) or morpholino targeting loxl5b (Figs. 5C, D and Table 2). The specificity of this lysyl oxidase morpholino has been demonstrated previously (Gansner et al., 2007), and partial morpholino knockdown of loxl5b that does not cause notochord distortion in wild-type embryos (Fig. 5C and Gansner et al., 2007) exacerbates the notochord phenotype of pfdgw1 mutants compared to control morpholino (Fig. 5D vs. Fig. 5B and Table 2). Once the pfdgw1 lesion was identified (see below), we genotyped a subset of 32 embryos from these experiments to confirm the genotype assignments in Table 2. Each directly-determined genotype matched the one assigned in Table 2 (data not shown), and testing of 17 phenotypically wild-type embryos that appeared worse after loxl5b morpholino injection (Table 2) revealed that 15 were heterozygous for the pfdgw1 lesion (data not shown), possibly due to a slight increase in the sensitivity of heterozygote embryos to lysyl oxidase inhibition.
The consistent worsening of the notochord phenotype in pfdgw1 mutants after partial lysyl oxidase inhibition demonstrates an interaction between the pfdgw1 product and the lysyl oxidases. However, the notochord distortion observed in pfdgw1 mutants in these sensitization experiments does not adopt the characteristic sine-wave pattern seen in wild-type embryos treated with high-dose neocuproine (Fig. 5E), and at this dose of neocuproine, the notochords of pfdgw1 mutants never develop this classic shape (Fig. 5F vs. Fig. 5E). These observations, taken together with the electron microscopy studies (Fig. 3), suggest that the pfdgw1 product and the lysyl oxidases have overlapping but also distinct roles in late notochord formation.
To determine the molecular basis for the notochord sensitivity to lysyl oxidase inhibition and other phenotypes observed in pfdgw1 mutants, we identified the locus mutated in pfdgw1 fish (Fig. 6). Meiotic mapping localized the pfdgw1 lesion to a telomeric region of chromosome 22, and inspection of the physical genome assembly (Zv6) for candidate genes in this region revealed sequence coding for an extracellular matrix protein of the fibrillin family (Fig. 6A). A fragment of this gene is annotated as fibrillin-3 in the current zebrafish genome assembly (Zv7), and human orthologues to the flanking genes timm44 and trh1 (Fig. 6A) flank human fibrillin-3. However, BLAST searches using the translated fibrillin EST sequences (Fig. 6A) demonstrated better hits in both cases with human fibrillin-1 and fibrillin-2 than with fibrillin-3 (data not shown), suggesting that the zebrafish gene might exhibit greater amino acid identity to a fibrillin besides fibrillin-3.
Sequencing of a 5′ segment of the zebrafish fibrillin on chromosome 22 revealed a nonsense mutation in pfdgw1 mutant embryos that was absent in adult, wild-type fish (Fig. 6B). This mutation abrogates an AvaII restriction enzyme site, allowing wild-type and heterozygote embryos to be distinguished (Fig. 6C) and permitting genotyping of fish stocks. Genotyping of stored DNA samples confirmed that the original pfdgw1 mutant identified in our screen was heterozygous for the nonsense mutation indicated in Fig. 6B, and that this mutation was absent in both the WIK grandparent used for the mapcross (see Experimental Procedures) and the AB fish used for the outcross (data not shown). Since all 7 recombinants at the flanking markers shown in Fig. 6A were homozygous for the mutation (data not shown), the recombination rate at this location is 0% (0 out of 597 meiotic events), consistent with the hypothesis that the nonsense mutation causes the pfdgw1 phenotype.
Cloning of the full-length zebrafish fibrillin on chromosome 22 revealed that it encodes a 2868 amino acid protein highly similar to human, mouse, and rat fibrillin-2 (Figs. 6D, E and Fig. S1). In particular, the glycine-rich domain and location of the RGD motifs are characteristic of fibrillin-2 but not fibrillin-1 or fibrillin-3 (Figs. 6D, E and Fig. S1) (Corson et al., 2004; Hubmacher et al., 2006). In addition, 11 of 12 potential N-glycosylation site are conserved (Fig. S1) and two calcium-binding EGF-like domains are encoded before the first TGF-β binding (TB) domain (Fig. 6D and Fig. S1), a sequence arrangement that is observed in human fibrillin-2 but not in human fibrillin-3 (Fig. 6E) (Corson et al., 2004). Overall, the zebrafish fibrillin on chromosome 22 exhibits 75% amino acid identity with human fibrillin-2 but only 67% and 68% identity with human fibrillin-1 and fibrillin-3, respectively (Fig. 6E). In view of the high amino acid sequence identity with human fibrillin-2 and the presence of functional domains that have historically been used to distinguish fibrillin-2 from other fibrillins (the glycine-rich domain, RGD motifs, and N-glycosylation sites), we name this gene fibrillin-2 (fbn2). The premature stop codon identified in pfdgw1 mutants is predicted to result in a severely truncated protein product of only 161 amino acids (Fig. 6D).
Since the current zebrafish genome assembly (Zv7) contains a gene fragment annotated as fibrillin-2 on chromosome 10 (data not shown), we cloned the majority of this gene to determine whether it is a paralogue of zebrafish fbn2 (Fig. S2). Surprisingly, the fibrillin gene on chromosome 10 encodes a protein without RGD motifs and with a novel proline/glutamine-rich domain not found in any fibrillin reported to date (Fig. 6E and Fig. S3). Consistent with previous nomenclature that differentiates fibrillins based on these attributes (Hubmacher et al., 2006), we name this gene fibrillin-4 (fbn4) (Fig. 6E). Compared to the three human fibrillins, zebrafish fibrillin-4 exhibits highest amino acid identity with human fibrillin-2 (Fig. 6E). However, this identity (68%) is relatively low, since it is comparable to the identity between individual human fibrillin family members (Corson et al., 2004). Importantly, zebrafish fibrillin-2 exhibits greater amino acid identity with human fibrillin-2 than does zebrafish fibrillin-4 (Fig. 6E). In addition, zebrafish fibrillin-2 and fibrillin-4 are only 66% identical, suggesting that these zebrafish fibrillins are not paralogues created by a genome duplication event in teleosts (Woods et al., 2000).
In order to determine the evolutionary relationship of zebrafish fibrillin-2 and fibrillin-4 to fibrillins in other species, we performed a phylogenetic analysis based on nucleotide coding sequences (Fig. 6F). This revealed that zebrafish fibrillin-2 is evolutionarily related to human fibrillin-3, and that zebrafish fibrillin-4 is distantly related to human fibrillin-2 (Fig. 6F). These findings were corroborated by examining the exon structure of the zebrafish fibrillins. In zebrafish fbn2, a single exon encodes both the sixth calcium-binding EGF-like motif and the first half of the second TB domain (data not shown), analogous to what is observed in human fibrillin-3 but not in other human fibrillins (Corson et al., 2004). By contrast, the exon structure of zebrafish fbn4 at this location is identical to that of human fibrillin-2 (data not shown). Importantly, a putative orthologue of fibrillin-2 in Xenopus laevis (Skoglund et al., 2006; Skoglund and Keller, 2007) clusters with zebrafish fbn2 and human FBN3 in the phylogenetic tree (Fig. 6F, arrow). This suggests that in fish and amphibians, a fibrillin with highest amino acid identity to human fibrillin-2 has evolved at the fibrillin-3 locus.
To determine the developmental expression of zebrafish fbn2, whole-mount in situ hybridization was performed on embryos from pfdgw1/+ intercrosses (Fig. 7). fbn2 expression is first visible during gastrulation in the hypoblast (mesendoderm) of embryos at 9 hpf (Fig. 7A). In 3-somite embryos, fbn2 expression occurs in the paraxial mesoderm (Fig. 7B, arrowheads) and notochord (Fig. 7B, arrow), consistent with a role for zebrafish fibrillin-2 in notochord morphogenesis. In 7-somite embryos, fbn2 is expressed in the somites and notochord (Fig. 7C, arrow), with foci of increased staining present near notochord-somite boundaries (Fig. 7C, arrowheads). By the 20-somite stage, fbn2 is also expressed in the developing venous plexus (Fig. 7D, arrowhead), which forms abnormally in pfdgw1 mutants, and in the eye (Fig. 7D, arrow). Frozen sections of 20-somite embryos revealed that this eye expression is restricted to the lens placode (Fig. 8A, arrowheads) and demonstrated fbn2 expression in the notochord, hypochord, floorplate, and paraxial mesoderm (Fig. 8B). At 24 hpf, fbn2 is expressed in the hypochord (Fig. 7E and Fig. 8C, arrows), which plays a critical role in axial vascular development (Cleaver and Krieg, 1998; Eriksson and Lofberg, 2000), and in the fin fold epidermis (Fig. 7E and Fig. 8C, arrowheads), where fbn2-expressing cells form two parallel lines on either side of the midline (Figs. 8C, D, arrowheads). Despite expression of fbn2 in the lens placode (Fig. 8A), no differences in lens morphology were observed by electron microscopy between pfdgw1 mutants and wild-type embryos at 30 hpf (data not shown).
Expression of fbn2 was almost completely absent in a quarter of embryos from pfdgw1/+ intercrosses at each developmental stage (Figs. 7F-J and data not shown). At 24 hpf, the mutant phenotype correlated with loss of fbn2 expression, and genotyping of embryos after in situ hybridization demonstrated an exact concordance between the homozygous mutant genotype and abrogation of fbn2 expression at three different stages of development examined (data not shown). Furthermore, heterozygote embryos exhibited an intermediate level of fbn2 expression (Fig. 9), revealing that compensatory up-regulation of the wild-type allele does not occur in heterozygous mutant embryos. The reduced amount of staining in heterozygous and homozygous mutant embryos supports the specificity of the in situ hybridization reactions. Given the distribution of the in situ probes across the entire length of the zebrafish cDNA (see Experimental Procedures), these findings mitigate against the presence of alternative splice products and suggest nonsense-mediated decay of the fbn2 transcript (Behm-Ansmant and Izaurralde, 2006).
We also examined the developmental onset of fbn2 expression by reverse transcriptase-polymerase chain reaction (RT-PCR), which is more sensitive than in situ hybridization (Fig. 7K). This revealed that fibrillin-2 mRNA is expressed at very low levels in wild-type embryos maternally and until 6-7 hpf, when it is upregulated but is not yet detectable by in situ hybridization (Fig. 7K and data not shown). At 9 hpf and the 3-somite stage, more robust fbn2 expression is apparent (Fig. 7K), and this correlates with the detection of fbn2 by in situ hybridization (Figs. 7A, B). As a control, RT-PCR was performed in parallel with primers to the gene spadetail (spt) (Fig. 7K). spt expression is relatively constant over the time period examined except that it is not present maternally (Fig. 7K) (Griffin et al., 1998). Importantly, an RT-PCR product was not obtained if the reactions were performed in the absence of RNA (Fig. 7K).
To confirm that the pfdgw1 phenotype directly results from loss of fibrillin-2, we designed morpholinos to abrogate zebrafish fbn2 expression (Fig. 10). Whereas wild-type embryos injected with control morpholino demonstrated normal formation of the notochord, caudal vein, and fin folds (Figs. 10A, C), embryos injected with a fbn2 start morpholino exhibited notochord kinks (Fig. 10B, black arrows), a cavernous caudal vein without venous plexus formation (Figs. 10B, D, white arrows), and fin fold attenuation (Figs. 10B, D, arrowheads). The fbn2 start morpholino also caused truncal skin distention characteristic of pfdgw1 mutants (Fig. 10F, arrowheads) that was not observed with control morpholino (Fig. 10E, arrowheads). These findings were consistent and were dose dependent (Table 3). The fbn2 start morpholino had no visible effect on embryos at a dose of 0.12 ng (Table 3), and at an intermediate dose of 0.48 ng, only a cavernous caudal vein with truncal edema was observed (data not shown), possibly reflecting differences in transcript abundance between notochord and caudal vein. At a dose of 7.2 ng, embryos developed non-specific findings suggestive of morpholino toxicity, such as reduced melanin pigmentation and mild necrosis (data not shown). Injection of fbn2 start morpholino into pfdgw1 mutant embryos did not alter their phenotype (Fig. 10H vs. Fig. 10G), indicating that the pfdgw1 mutation results in a null phenotype, consistent with the early location of the premature stop codon and the in situ hybridization findings (Fig. 6D and Figs. 7F-J). Importantly, it also did not result in the reduced melanin pigmentation or necrosis observed with high dose (7.2 ng) fbn2 start morpholino, providing evidence that these phenotypes are non-specific. Injection of a morpholino targeting a splice site in fbn2 also recapitulated the pfdgw1 phenotype (data not shown), further confirming the specificity of the fbn2 knockdown.
Injection of mRNA encoding full-length (9 kb) fbn2 into clutches from pfdgw1/+ intercrosses did not rescue the pfdgw1 phenotype, but expression of zebrafish fibrillin-2 in mutant embryos could not be confirmed by western immunoblotting since none of the available antibodies cross-reacted with zebrafish fibrillin-2 (data not shown).
These data reveal a previously unappreciated role for fibrillin-2 in zebrafish notochord and vascular morphogenesis. Fibrillins are large glycoproteins that polymerize to form microfibrils in a wide variety of elastic and non-elastic connective tissues (Hubmacher et al., 2006). Microfibrils associate with many different extracellular matrix proteins, bind cell-surface integrins, and have recently been shown to modulate the activity of soluble signaling factors (Ramirez and Dietz, 2007). The demonstration in this current work that the notochords of pfdgw1 mutants are sensitized to lysyl oxidase inhibition (Fig. 4 and Fig. 5) is consistent with the recent realization that microfibril structural elements are part of larger, more complex networks associated with impaired morphogenesis (Ramirez et al., 2007). Indeed, our data reveal gene-gene and gene-nutrient interactions between fibrillin-2 and the copper-dependent lysyl oxidases in notochord formation that were not previously apparent. Interestingly, our findings are consistent with the recent discovery in yeast that whereas only ~20% of gene deletions have an obvious phenotypic consequence, 97% exhibit a measurable growth phenotype under situations of specific chemical or environmental stress (Hillenmeyer et al., 2008). In addition, the findings add to previous elegant genetic studies that elucidated novel molecular pathways required for notochord sheath formation in zebrafish (Parsons et al., 2002; Coutinho et al., 2004; Stemple, 2005).
The zebrafish genome appears to encode three fibrillins, similar to what is observed in human and other vertebrate genomes (data not shown). In the current work, we present sequences for two of these fibrillins, which we name fibrillin-2 (fbn2) and fibrillin-4 (fbn4) (Figs. 6D, E). The third fibrillin, a putative orthologue of fibrillin-1 (fbn1), has not been cloned but has been targeted using morpholinos (Chen et al., 2006), and we have independently confirmed its embryonic expression (data not shown).
The two cloned zebrafish fibrillins are named based on amino acid identity to human fibrillins and on the presence of unique domains that are thought to modulate fibrillin function. Zebrafish fibrillin-2 exhibits the highest amino acid identity with human fibrillin-2 (Fig. 6E) and encodes a glycine-rich internal unique region characteristic of fibrillin-2 in other species (Fig. S1). It also possesses two RGD motifs located in the same relative positions as human fibrillin-2 (Figs. 6D, E, and Fig. S1). Fibrillin RGD motifs mediate interactions with specific cell-surface integrins, and increasing data suggest that these interactions are determined in part by the surrounding amino acid sequence (Sakamoto et al., 1996; Bax et al., 2003; Bax et al., 2007; Jovanovic et al., 2007). It is therefore not surprising that the two RGD motifs in fibrillin-2 have different receptor specificities (Sakamoto et al., 1996), and the spacing of the RGD motifs, which differs between human fibrillin-2 and fibrillin-3, may thus affect integrin binding and protein function.
In contrast to zebrafish fibrillin-2, zebrafish fibrillin-4 exhibits relatively low amino acid identity to human fibrillins, with only 68% identity to human fibrillin-2 (Fig. 6E). Furthermore, zebrafish fibrillin-4 contains a novel proline/glutamine-rich internal unique region not found in other fibrillins and also completely lacks RGD motifs. Zebrafish fbn4 is thus unlikely to function as a fibrillin-2 orthologue, although it appears to be a distant evolutionary relative of human fibrillin-2 by phylogenetic analysis (Fig. 6F) and is flanked by the zebrafish orthologue of SLC27A6 that flanks human fibrillin-2 (data not shown).
The zebrafish fibrillin nomenclature presented here has been informed by studies in amphibians (Skoglund et al., 2006; Skoglund and Keller, 2007). Xenopus laevis fibrillin contains an antigenic determinant recognized by the JB3 antibody to fibrillin-2, and the cloned 3′ fragment of this fibrillin exhibits 77% amino acid identity with human fibrillin-2 but only 68% and 71% identity with fibrillin-1 and fibrillin-3, respectively (Skoglund et al., 2006). Although the 5′ sequence of Xenopus fibrillin is unknown, it seems likely based on the published expression and functional data that this fibrillin is an orthologue of human fibrillin-2 (Skoglund et al., 2006; Skoglund and Keller, 2007). Importantly, the amino acid sequences of Xenopus fibrillin and zebrafish fbn2 are 75% identical over the region cloned, and their expression patterns are similar (see below). Since both genes cluster phylogenetically with human fibrillin-3 (Fig. 6F), the data suggest that Xenopus fibrillin and zebrafish fbn2 are functional orthologues of fibrillin-2 that have evolved at the human fibrillin-3 locus. The fact that fibrillin family members are already fairly similar (~68% amino acid identity in humans), may help to explain how this could have occurred.
The spatiotemporal expression of zebrafish fbn2 is similar to that of fibrillin-2 in other vertebrates. In chick and quail, fibrillin-2 protein localization has been directly and extensively studied using the JB3 monoclonal antibody (Wunsch et al., 1994; Sugi and Markwald, 1996; Rongish et al., 1998; Visconti et al., 2003). In chick, the earliest fibrillin-2 immunostaining is detected in the primitive streak, which gives rise to notochord and mesodermal structures that exhibit persistent fibrillin-2 staining throughout development (Wunsch et al., 1994; Visconti et al., 2003). In particular, fibrillin-2 containing microfibrils are detected extracellularly around the notochord and somites (compare Figs. 7B, C), and also in splanchnic mesoderm and segmental plate mesoderm, which is equivalent to paraxial mesoderm in zebrafish (Fig. 7B) (Wunsch et al., 1994; Visconti et al., 2003). In quail, immunostaining using the JB3 antibody revealed prominent fibrillin-2 deposition around the notochord and somites, as well as fibrillin-2 immunoreactivity in the primitive mesocardium and segmental plate mesoderm (Rongish et al., 1998). Of note, fibrillin-2 appeared to physically integrate somites and notochord (Rongish et al., 1998), which may explain the increased fbn2 expression at notochord-somite boundaries observed in zebrafish (Fig. 7C, arrowheads). Overall, the findings in avian species using the JB3 antibody are consistent with the zebrafish RT-PCR and in situ hybridization results reported above (Fig. 7 and Fig. 8), both in terms of time of onset (pre-gastrulation) and tissue distribution (compare Fig. 7B with Fig. 3 of Rongish et al., 1998).
Zebrafish fbn2 expression is also similar to that of Xenopus fibrillin, a putative orthologue of fibrillin-2 (Skoglund et al., 2006; Skoglund and Keller, 2007). RNase protection assays demonstrate that Xenopus fibrillin is markedly upregulated in mid-gastrula embryos (Skoglund et al., 2006), which correlates well with the initial increase in zebrafish fbn2 transcript at 6-7 hpf observed by RT-PCR (Fig. 7K). In addition, fibrillin-2 expression is first detectable by in situ hybridization during gastrulation in both organisms (Fig. 7A) (Skoglund et al., 2006). Exact comparison of fibrillin-2 expression between Xenopus and zebrafish at post-gastrula stages is difficult because Xenopus has a distinct neurulation period before segmentation whereas these stages overlap in zebrafish and a distinct neurulation period is absent (Kimmel et al., 1995). Nevertheless, Xenopus fibrillin is expressed in the notochord, somites, floorplate, hypochord, and eye by in situ hybridization (Skoglund et al., 2006), analogous to what is observed in zebrafish (Figs. 7B-D and Figs. 8A, B). At later stages, Xenopus fibrillin is also expressed in tail fins (Skoglund et al., 2006), mimicking zebrafish fin fold expression (Fig. 7E and Figs. 8C, D). Interestingly, while notochord staining is stronger than somite staining in Xenopus (Skoglund et al., 2006), the opposite is true in zebrafish (Figs. 7B, C and Fig. 8B), possibly due to a differential requirement for fibrillin-2 in axial extension. Future studies using antibodies to zebrafish fibrillin-2 will be required to determine the exact timing of zebrafish fibrillin-2 protein deposition in the notochord-somite boundary during gastrulation, but this may occur later in zebrafish than in Xenopus, as is the case in chick (Skoglund et al., 2006).
Fibrillin-2 expression has also been studied during human and mouse embryonic development. In humans, immunostaining demonstrates that fibrillin-2 has a wide tissue distribution and is present in fetal eye as well as in the notochordal sheath and perinotochordal mesenchyme (Zhang et al., 1994; Quondamatteo et al., 2002). In mouse, a wide tissue distribution with prominent mesenchyme staining has been noted (Zhang et al., 1995). The overall conserved expression pattern of fibrillin-2 argues for conservation of function among these different species.
The functions of specific microfibrillar proteins in various tissues during development are incompletely characterized (Ramirez and Dietz, 2007). Fibrillin-2 binds multiple other microfibrillar proteins in vitro, including microfibril-associated glycoprotein-1 and fibrillin-1 (Lin et al., 2002; Charbonneau et al., 2003; Werneck et al., 2004), and morpholino knockdown of zebrafish microfibril-associated glycoprotein-1 or the predicted zebrafish fibrillin-1 results in caudal vein dilation and altered plexus formation (Chen et al., 2006) similar to what is observed in pfdgw1 mutants (Fig. 2B). A specialized type of heterofibril composed of fibrillin-1, fibrillin-2, and microfibril-associated glycoprotein-1 may thus be required for venous plexus formation in zebrafish. By contrast, tissue-specific roles unique to fibrillin-2 are indicated by the notochord kinks, reduced axial vessel diameters, fin fold attenuation, and truncal edema observed in pfdgw1 mutants but not fibrillin-1 or microfibril-associated glycoprotein-1 morphants (Fig. 1 and Fig. 2 vs. Chen et al., 2006).
The precise role of microfibrils in notochord formation is unclear, but our results suggest that microfibrils contribute to the strength of the sheath that envelops the notochord, at least in zebrafish. The sensitivity of the notochord phenotype in pfdgw1 mutants to lysyl oxidase inhibition (Fig. 4 and Fig. 5) may result in part from an impaired ability to recruit or correctly position lysyl oxidases at their site of action in the notochord sheath, a hypothesis supported by studies demonstrating a role for microfibril-associated proteins in this process (Liu et al., 2004; Freeman et al., 2005). This model is also consistent with the observation that the heterozygote pfdgw1 mutants are largely unaffected by lysyl oxidase inhibition (Fig. 4 and Fig. 5), provided that enough fibrillin-2 is present in these fish to allow for proper targeting of lysyl oxidase.
Our data from mutant and morphant zebrafish demonstrate that loss of fibrillin-2 causes early embryonic phenotypes in zebrafish (Fig. 1B, Figs. 2B, D, and Fig. 10B). Interestingly, morpholino knockdown of Xenopus fibrillin-2 results in gastrulation arrest (Skoglund and Keller, 2007), a finding that likely reflects differences in the mechanisms of axial extension employed by Xenopus and zebrafish embryos. Alternatively, an additional fibrillin family member may compensate for the loss of fibrillin-2 in zebrafish but not in Xenopus during gastrulation. In this regard, mice homozygous for a null allele of fibrillin-2 are viable but die in utero when crossed into a fibrillin-1 deficient background, suggesting partial functional redundancy between fibrillin-1 and fibrillin-2 (Arteaga-Solis et al., 2001; Chaudhry et al., 2001; Carta et al., 2006). Fibrillin-2 knockout mice exhibit syndactyly and congenital contractures of the forelimbs (Arteaga-Solis et al., 2001) but not notochord or vascular abnormalities, possibly due to evolutionary differences between mice and zebrafish.
Microfibrils have recently been demonstrated to modulate critical morphogenetic signaling events (Arteaga-Solis et al., 2001; Neptune et al., 2003; Ng et al., 2004; Habashi et al., 2006; Cohn et al., 2007), and the phenotypes observed in pfdgw1 mutants may therefore result in part from altered signaling. While the notochord distortion observed in pfdgw1 mutants (Fig. 1B) likely results from weakness of the notochord sheath due to disruption of the outer sheath layer (Figs. 3B, D), the etiology of the endothelial cell disorganization and altered caudal vein formation in pfdgw1 mutants (Fig. 2B) is less certain. The mechanisms of venous plexus formation may be similar to those involved in epithelial branching morphogenesis, where signaling by bone morphogenetic protein-7 plays a prominent role (Dean et al., 2004; Grishina et al., 2005). A genetic interaction between fibrillin-2 and bone morphogenetic protein-7 has been demonstrated in fibrillin-2 knockout mice (Arteaga-Solis et al., 2001), and epithelial branching morphogenesis is inhibited in embryonic lung cultures after fibrillin-2 knockdown (Yang et al., 1999). Signaling by bone morphogenetic protein-7 or other morphogens may therefore be required for endothelial cell branching and venous plexus morphogenesis in pfdgw1 mutants. Since heterofibrils of fibrillin-2 and fibrillin-1 may be required for caudal vein formation, and mutations in fibrillin-1 lead to excessive transforming growth factor-β activity (Neptune et al., 2003; Ng et al., 2004; Habashi et al., 2006; Cohn et al., 2007), we considered whether increased transforming growth factor-β signaling could explain the caudal vein abnormalities in pfdgw1 mutants. However losartan, an inhibitor of transforming growth factor-β signaling (Habashi et al., 2006) does not rescue the vascular phenotype of pfdgw1 mutants (data not shown).
Mutations in human fibrillin-2 cause congenital contractural arachnodactyly, a rare disease with autosomal dominant inheritance that is characterized by congenital joint contractures, arachnodactyly, kyphoscoliosis, malformed ear helices, and vascular abnormalities (Gupta et al., 2002). The pathogenesis of congenital contractural arachnodactyly is uncertain and could result from haploinsufficiency or a gain of function mutation (antimorphic and/or neomorphic) in fibrillin-2. Our results demonstrate that heterozygosity for a null allele at the fbn2 locus in zebrafish does not cause any overt phenotype (Figs. 1A, C, E and data not shown) even in mature fish (data not shown), a finding corroborated by mice heterozygous for a null allele of fibrillin-2 (Arteaga-Solis et al., 2001; Chaudhry et al., 2001; Carta et al., 2006). Furthermore, the mutations that are known to cause congenital contractural arachnodactyly all cluster within a well-defined region near the middle of human fibrillin-2, a finding that is statistically unexpected and suggests a role for this region in mediating interactions with other proteins (Park et al., 1998; Gupta et al., 2002; Nishimura et al., 2007). These data suggest that a gain of function mutation (antimorphic and/or neomorphic) in human fibrillin-2 is required to cause congenital contractural arachnodactyly, a hypothesis now testable through transgenic expression experiments in the pfdgw1 mutant. Taken together, the data reveal a genetic interaction between fibrillin-2 and the lysyl oxidases in late notochord formation, demonstrate the importance of fibrillin-2 in specific early developmental processes in zebrafish, and provide insight into the pathogenesis of congenital contractural arachnodactyly. Furthermore, the pfdgw1 mutant described here should now permit studies to elucidate the cell biological mechanisms of fibrillin-2 in early development.
Zebrafish were reared under standard conditions at 28.5°C (Westerfield, 1993) and staged as described (Kimmel et al., 1995). Synchronous, in vitro fertilized embryos were obtained for all experiments, which were carried out in accordance with Washington University's Division of Comparative Medicine guidelines. AB stocks were used except when pfdgw1 was crossed to WIK for mapping and to Tg(fli1a:EGFP)y1 (Lawson and Weinstein, 2002) for visualization of vascular defects. In Figs. 1, ,2,2, 3A-D, ,4,4, ,5,5, 7A-J, and and8,8, “wild-type” refers to either +/+ or +/- embryos, which could not be phenotypically distinguished.
Mutagenesis was carried out with N-ethyl-N-nitrosourea according to standard methods (Solnica-Krezel et al., 1994). Briefly, male zebrafish were incubated for 1 hour in 3 mM N-ethyl-N-nitrosourea at weekly intervals for five consecutive weeks and outcrossed to generate F1 carriers (AB*/AB) after a rest period. Gynogenetic haploid offspring from F1 females were screened using a dose (50 nM) of the lysyl oxidase inhibitor 2-mercaptopyridine-N-oxide (Anderson et al., 2007) that is slightly lower than the dose required to cause notochord distortion in haploids. In addition, frank notochord mutants detected in concurrent screens were also interrogated for notochord sensitivity to partial lysyl oxidase inhibition by this method. Random intercrosses of F2 progeny confirmed transmission of the mutation.
Live embryos were anesthetized in tricaine, mounted in 2% methylcellulose, and imaged using either Olympus SZX12 (bright-field) or Olympus MVX10 (fluorescent) microscopes fitted with Olympus DP70 cameras. Frozen sections mounted on slides were imaged using an Olympus BX60 microscope fitted with a Zeiss AxioCam. Electron microscopy was carried out as previously described (Gansner et al., 2007).
All pharmacologic compounds were purchased from Sigma-Aldrich (St. Louis, MO) except losartan, which was purchased from AK Scientific, Inc. (Mountain View, CA). β-Aminopropionitrile (A3134) was prepared as a 100 mM stock in egg water (Westerfield, 1993) and diluted in egg water; neocuproine (N1501) and 2-mercaptopyridine-N-oxide (188549) were prepared as 100 mM stocks in dimethyl sulfoxide and diluted in egg water. Losartan (I497) was prepared as a 50 mM stock in dimethyl sulfoxide, diluted in water, and used at final concentrations of 5 and 50 μM. Embryos were incubated in each compound starting at 3-6 hpf. Statistical analysis was performed as previously described (Gansner et al., 2007).
F2 carriers were mapcrossed to the polymorphic WIK strain and the progeny (AB*/WIK) raised to adulthood. The pfdgw1 mutation was assigned to chromosome 22 by centromeric linkage analysis (Johnson et al., 1995; Johnson et al., 1996) using simple sequence length polymorphism (SSLP) markers (Shimoda et al., 1999) and DNA from early pressure gynogenetic diploids. For fine mapping, 597 mutant embryos from AB*/WIK females crossed to AB*/AB males were collected and assessed for recombination along chromosome 22 by SSLP analysis. Embryos were incubated in 2 μM neocuproine to accentuate the mutant phenotype and facilitate rapid mutant identification. When necessary, candidate marker primer pairs for BAC sequences were generated using the Zebrafish SSR search website of the Massachusetts General Hospital (http://danio.mgh.harvard.edu/markers/ssr.html). Primers for the BAC markers used in Fig. 6A are: zC124A3 – forward 5′-GAGTGTTGCAAGTGTAACTTGCCC-3′ and reverse 5′-ATGGGAAACTAGTTATTTGGCACAG-3′; zC239M17 – forward 5′-GAGTCTAATCAGTGGAGACTTGG-3′ and reverse 5′-CGTACAGATGCTGATCTGGG-3′; zK49M19 – forward 5′-GCATCGTTGCAACTTGCTT-3′ and reverse 5′-TGATGGCAGAATAGTTTCACACA-3′; bZ36A1 – forward 5′-GGCAATAGATTTCAAAGGTGTTTT-3′ and reverse 5′-AATCCAAGGCAATGCAGAAA-3′. DNA from wild-type and heterozygote embryos as well as WIK and AB grandparents was used to ensure polymorphism between AB and WIK. Based on available genetic maps of the zebrafish genome (Geisler et al., 1999; Shimoda et al., 1999), the pfdgw1 mutation was localized to a telomeric region of chromosome 22.
Zebrafish fibrillin-2 was cloned first in pieces and then in its entirety using Superscript III reverse transcriptase (Invitrogen) and Phusion DNA polymerase with high-fidelity buffer (Finnzymes). The pfdgw1 mutation was initially detected by cloning the 5′ end of fibrillin-2 from mutant embryo and wild-type adult fin cDNA into pCRII (Invitrogen) using forward primer 5′-AGGGTGAGGCACATTTACCG-3′ and reverse primer 5′-GTGTCTTCACACTCGTCGATG-3′. Full-length wild-type fibrillin-2 was subsequently cloned into pCR-XL-TOPO (Invitrogen) after amplification from 52 hpf embryo cDNA using forward primer 5′-AGGGTGAGGCACATTTACCG-3′ and reverse primer 5′-CTGCAGTGAAGGGCATAGGG-3′. Full-length clones from 3 separate PCR reactions were sequenced and compared, allowing mutations introduced during PCR to be identified. No alternative splice forms were detected. The start site of fibrillin-2 was confirmed by 5′ rapid amplification of cDNA ends (RACE) using wild-type RACE-ready cDNA from 15-somite embryos and the primer 5′-GACCCTCTGCTGCACCTCCCCCTCAG-3′. The full-length zebrafish fbn2 sequence is available at Genbank (accession number EU449516). The 5′ portion of zebrafish fibrillin-4 (accession number EU854565) was cloned in two overlapping pieces using the following primer sets: A) forward primer 5′-GCAGTCCTTCTGTTGTCCAGG-3′ and reverse primer 5′-CAGACAGATGATGACCAGGCGG-3′ and B) forward primer 5′-CCTTTCACTGTGCAATGGAGGC-3′ and reverse primer 5′-CCACATTATTGACACACCGACC-3′. The start site was inferred from an EST (EE319589). The 3′ portion of zebrafish fibrillin-4 (accession number EU854566) was cloned in multiple overlapping pieces using the following primer sets: A) forward primer 5′-CGAATGTGGTCAGAATCCTC-3′ and reverse primer 5′-CCCATGACCTCCATTACAGC-3′ B) forward primer 5′-GTGAATGTGGTGTTGGGTTC-3′ and reverse primer 5′-GGTGTTGTGTGTAACCCTGTGG-3′ C) forward primer 5′-CCTGGGATATGTGCTCCTGG-3′ and reverse primer 5′-CCCATGACCTCCATTACAGC-3′ D) forward primer 5′-CGTTTAACACCACCAAAGCC-3′ and reverse primer 5′-GCAGTGGCATCTGTAAGAGC-3′. An 86 bp sequence connecting the 5′ and 3′ portions was inferred from Zv7_Scaffold988 based on the modular nature of fibrillins. A 253 bp extension to the 3′ end was also inferred in this manner. The far 3′ end and stop codon could not be determined. For both fibrillins, conserved protein domains were identified using Motif Scan (Pagni et al., 2004) and by comparison with annotated orthologues (Hubmacher et al., 2006). The NetNGlyc 1.0 Server (http://www.cbs.dtu.dk/services/NetNGlyc/) was used to predict N-glycosylation sites with a potential greater than 0.32.
DNA coding sequences were aligned using ClustalW2 (Larkin et al., 2007) and a phylogenetic tree was constructed in TOPALi v2.5 by the maximum likelihood method (PhyML) using HKY substitution and gamma rate models (Guindon and Gascuel, 2003; Milne et al., 2004; Anisimova and Gascuel, 2006). 100 bootstrap runs were performed. Similar trees were obtained with MEGA4.0 using maximum parsimony or minimum evolution methods (Tamura et al., 2007), except that pDrfbn4 diverged before all other fibrillins to form its own one-member group. Accession numbers for the non-zebrafish sequences used were: HsFBN1 – NM_000138; HsFBN2 – NM_001999; HsFBN3 – NM_032447; MmFbn1 – NM_007993; MmFbn2 – NM_010181; RnFbn1 – NM_031825; RnFbn2 – NM_031826; BtFbn1 – NM_174053; pXlfbn – DQ310728; and HsLTBP1 – NM_000627.
Embryos from pfdgw1/+ intercrosses were manually dechorionated at the indicated developmental stages, fixed in 4% paraformaldehyde-PBS, and dehydrated by methanol series. Three fibrillin-2 probe constructs were generated by cloning nonoverlapping fragments of the cDNA into pCRII (Invitrogen). DIG-labeled antisense RNA probes were synthesized from these constructs using a DIG-labeling kit (Roche), and whole-mount in situ hybridization performed as previously described (Thisse et al., 1993; Mendelsohn et al., 2006). All embryos of the same developmental stage were processed in a single well to ensure identical reaction conditions. Frozen sections were prepared as described (Gansner et al., 2007).
DNA was extracted from embryos or caudal fin tissue and PCR amplified using primers flanking the mutation: forward primer 5′-GCGGTGTGCGAGAGCGGATG-3′ and reverse primer 5′-GCATTCTTGAAGGAGCCCCG-3′. The resulting product was then digested with the restriction enzyme AvaII (New England Biolabs) for 5 hrs at 37°C and a small aliquot run on a 10% TBE polyacrylamide gel for visualization. PCR product derived from the mutant allele is not cut by AvaII. For genotyping after whole-mount in situ hybridization, embryos were first rehydrated to PBS and incubated in 300 mM NaCl for 4 hours at 65°C to reverse crosslinking. They were then placed in DNA extraction buffer (80 mM KCl, 10 mM Tris pH 8, 1 mM EDTA, 0.3% Tween 20, 0.3% Igepal), boiled 10 minutes, and the DNA phenol-chloroform extracted. Genotyping was performed by restriction digest, as described above.
RNA was obtained from pooled wild-type embryos or unfertilized eggs using Trizol reagent (Invitrogen) and RT-PCR performed with the SuperScript III One-Step RT-PCR Platinum Taq HiFi kit (Invitrogen). Half a microgram of total RNA was used per 25 μL reaction. The primer annealing temperature was 57°C and 30 cycles of PCR were performed. Fibrillin-2 primers were: forward primer 5′-CACCTGTAAATGCCCCTCGG-3′ and reverse primer 5′-GAAGCCCACCCCATTGATGC-3′. Spadetail primers were: forward primer 5′-GAAGATGTTTACTGACCACTCAG-3′ and reverse primer 5′-GCCTGTTTGTTTAAGACATT-3′.
Morpholino oligonucleotides (Nasevicius and Ekker, 2000) targeting start and splice sites in fibrillin-2 were resuspended in Danieau buffer (start MO) or water (splice MO), diluted to include 0.05% phenol red, and injected into 1-cell embryos. Standard control morpholino (Gene Tools, LLC) was resuspended in Danieau buffer and likewise injected. Morpholino sequences were: 5′-GCGACTCCTGAAGCGCCGGTAAATG-3′ (start MO) and 5′-GGGATACTTACGAACTATACACTGG-3′ (splice MO). The splice MO was used at a dose of 1.7 ng. The loxl5b splice morpholino (5′-GCCTGTGGAATAAACACCAGCCTCA-3′) was prepared as previously described (Gansner et al., 2007) and was used at a dose of 5 ng. Capped, polyadenylated mRNA for rescue experiments was generated from a full-length clone of zebrafish fibrillin-2 using the mMESSAGE mMACHINE kit (Ambion). Individual embryos from pfdgw1/+ intercrosses were injected with between 100 and 1000 pg of mRNA in multiple separate experiments. Statistical analysis was performed as previously described (Gansner et al., 2007).
We thank Stephen Johnson for help with meiotic mapping, Marilyn Levy for the electron microscopy, Bryce Mendelsohn for helpful suggestions, and Stephen Johnson and David Wilson for careful review of the manuscript. This work was supported by NIH Medical Scientist Training Program grant T32 GM07200 (JMG and ECM) and by NIH grant DK44464 (JDG).
Grant Sponsor: National Institutes of Health; Grant Numbers: T32 GM07200 and DK44464.