Essential role for the alpha 1 chain of type VIII collagen in Zebrafish notochord formation



Several zebrafish mutants identified in large-scale forward genetic screens exhibit notochord distortion. We now report the cloning and further characterization of one such mutant, gulliverm208 (gulm208). The notochord defect in gulm208 mutants is exacerbated under conditions of copper depletion or lysyl oxidase cuproenzyme inhibition that are without a notochord effect on wild-type embryos. The gulm208 phenotype results from a missense mutation in the gene encoding Col8a1, a lysyl oxidase substrate, and morpholino knockdown of col8a1 recapitulates the notochord distortion observed in gulm208 mutants. Of interest, the amino acid mutated in gulm208 Col8a1 is highly conserved, and the equivalent substitution in a closely related human protein, COL10A1, causes Schmid metaphyseal chondrodysplasia. Taken together, the data identify a new protein essential for notochord morphogenesis, extend our understanding of gene–nutrient interactions in early development, and suggest that human mutations in COL8A1 may cause structural birth defects. Developmental Dynamics 237:3715–3726, 2008. © 2008 Wiley-Liss, Inc.


The notochord is a midline structure that functions both as an axial skeleton and as an important signaling center to other tissues during embryogenesis (Scott and Stemple,2005). It is one of the defining structures of the chordate phylum, and is the first organ to fully differentiate in vertebrates (Scott and Stemple,2005). The notochord arises from the dorsal organizer and likely represents a primitive form of cartilage, as indicated by its expression of type II collagen, type IX collagen, sox9, aggrecan, and chondromodulin (Stemple,2005). In higher chordates, the notochord plays a critical role in vertebral column formation and persists postembryonically as the nucleus pulposus of intervertebral discs (Stemple,2005).

In zebrafish, studies of notochord morphogenesis are facilitated by the genetic tractability of this model organism and by the ease with which its notochord is visualized throughout early development (Parsons et al.,2002; Coutinho et al.,2004; Stemple,2005). During late notochord morphogenesis, chordamesoderm differentiates into mature notochord, a transition that involves the down-regulation of specific marker genes and the inflation of notochord cell vacuoles which exert turgor pressure on an extracellular matrix sheath (Scott and Stemple,2005). If sheath formation is impaired, this turgor pressure results in notochord abnormalities that are readily observed, permitting insight into the mechanisms of both notochord morphogenesis and extracellular matrix formation. Indeed, several zebrafish notochord mutants have already been identified during large-scale forward genetic screens (Odenthal et al.,1996; Stemple et al.,1996), and early studies of these mutants established the importance of laminins and coatomer proteins in notochord differentiation and sheath formation (Parsons et al.,2002; Coutinho et al.,2004). Because notochord abnormalities and extracellular matrix defects have both been implicated in the development of structural birth defects (Cleaver and Krieg,2001; Royce and Steinmann,2002; Miner,2005), elucidating the mechanisms of notochord formation will not only permit a greater understanding of the biology of this important organ, but may also allow for the therapeutic intervention or prevention of structural birth defects, which are a major cause of morbidity and mortality in children (NBDPN,2002).

Previous work in our laboratory has demonstrated that copper deficiency causes notochord distortion in zebrafish embryos and that this distortion results from the inhibition of specific lysyl oxidase cuproenzymes, which crosslink collagens in the notochord sheath (Mendelsohn et al.,2006; Gansner et al.,2007). In addition, we have recently conducted a forward genetic screen for zebrafish mutants that exhibit increased notochord distortion after partial lysyl oxidase inhibition, identifying one such mutant (Gansner et al.,2008). These studies raised the intriguing possibility that notochord mutants previously identified in large-scale forward genetic screens (Odenthal et al.,1996; Stemple et al.,1996) could provide insight into the molecular mechanisms of copper-dependent notochord formation. We now report the mapping and further characterization of one such recessive lethal mutant, gulliverm208 (Stemple et al.,1996). The analysis presented here demonstrates a role for the alpha 1 chain of type VIII collagen in zebrafish notochord formation and an interaction with lysyl oxidase cuproenzymes.


gulm208 Mutants Are Sensitized to Lysyl Oxidase Inhibition

Gulliverm208 (gulm208) mutants exhibit notochord distortion similar to what is obtained under conditions of copper deficiency or direct pharmacologic lysyl oxidase cuproenzyme inhibition (Stemple et al.,1996; Mendelsohn et al.,2006; Anderson et al.,2007; Gansner et al.,2007). However, the distortion observed in gulm208 mutants is less robust, and we therefore tested these mutants for sensitivity to lysyl oxidase inhibition (Fig. 1). Clutches from gulm208/+ intercrosses were incubated in vehicle (Fig. 1A,B) or in a dose of the copper chelator neocuproine (Fig. 1C,D) that does not cause notochord distortion in wild-type embryos (Fig. 1C and Gansner et al.,2007). Under such copper-limiting conditions, gulm208 mutants exhibited increased notochord distortion (Fig. 1D vs. Fig. 1B, arrowheads), presumably due to the partial inhibition of lysyl oxidase activity (Gansner et al.,2007). This hypothesis was confirmed by incubating clutches from gulm208/+ intercrosses in a dose of the irreversible lysyl oxidase inhibitor β-aminopropionitrile that does not cause notochord distortion in wild-type embryos (Fig. 1E). Under these conditions, increased notochord distortion in gulm208 mutants was again observed (Fig. 1F vs. Fig. 1B, arrowheads).

Figure 1.

gulm208 mutants are sensitized to lysyl oxidase inhibition. A–F: Clutches from gulm208/+ intercrosses were incubated in vehicle (A,B), the copper chelator neocuproine (2 μM; C,D), or the lysyl oxidase inhibitor β-aminopropionitrile (1 mM; E,F). The notochord is normal in wild-type embryos treated with vehicle, neocuproine, or β-aminopropionitrile (A,C,E). Notochords of gulm208 mutants incubated in neocuproine and β-aminopropionitrile (D,F, arrowheads) are substantially more distorted than mutants incubated in vehicle (B, arrowheads). Insets show notochord at higher magnification. Embryos were treated with PTU to inhibit melanin pigmentation and photographed at 30 hours postfertilization (hpf).

Interestingly, the notochord distortion observed in gulm208 mutants is accompanied by a decrease in the length of the yolk sac extension (wild-type length = 3.23 ± 0.09 vs. mutant length = 2.64 ± 0.18 in vehicle), and the increased notochord distortion observed after neocuproine or β-aminopropionitrile treatment is associated with a further shortening of this structure in mutant but not wild-type embryos (wild-type length = 3.33 ± 0.15 vs. mutant length = 2.31 ± 0.09 in neocuproine; wild-type length = 3.22 ± 0.12 vs. mutant length = 2.13 ± 0.16 in β-aminopropionitrile; Fig. 1). These data suggest that force generation in the notochord is impaired in gulm208 mutants and confirms a synergy between the gulm208 locus and both copper chelation and lysyl oxidase inhibition.

Electron Microscopy Reveals Notochord Abnormalities in gulm208 Mutants

To determine whether a defect in notochord sheath formation could explain the notochord phenotype of gulm208 mutants, we imaged truncal cross-sections from gulm208 mutant and wild-type embryos by transmission electron microscopy. This revealed abnormalities in both the vacuolated notochord cells and notochord sheath of gulm208 mutants at 24 hours postfertilization (hpf; Fig. 2). In the vacuolated notochord cells of gulm208 mutants, the rough endoplasmic reticulum exhibits striking areas of engorgement compared with wild-type embryos, presumably due to retention of protein that fails to be secreted and instead forms large circular aggregates (Fig. 2B vs. Fig. 2A, arrows). These aggregates are also observed in the hypochord, which is closely apposed to the ventral aspect of the notochord (Fig. 2C, red arrows) and plays a critical role in the formation of the dorsal aorta (Cleaver and Krieg,1998; Eriksson and Lofberg,2000). While the inner (i), medial (m), and outer (o) layers of the notochord sheath are present in both wild-type embryos and gulm208 mutants, the fibrillar medial layer appears disordered in gulm208 mutants (Fig. 2E vs. Fig. 2D).

Figure 2.

Electron microscopy reveals notochord abnormalities in gulm208 mutants. A–E: Transmission electron micrographs of truncal cross-sections from embryos at 24 hpf. A: Notochord sheath and vacuolated notochord cell of a wild-type embryo, with small areas of protein accumulation visible in the rough endoplasmic reticulum (arrow). B: Notochord sheath and vacuolated notochord cell of a gulm208 mutant with large circular aggregates of protein in the rough endoplasmic reticulum (arrow). C: Notochord and hypochord of a gulm208 mutant with large circular aggregrates of protein in the hypochord (red arrows) and notochord (yellow arrow). D: Notochord sheath of a wild-type embryo with inner (i), medial (m), and outer (o) layers indicated. Small areas of protein accumulation are visible in the rough endoplasmic reticulum (red arrow). E: Notochord sheath of a gulm208 mutant where the collagen fibrils in the medial (m) layer appear disorganized. A large circular aggregate of protein is visible in the rough endoplasmic reticulum (red arrow). Not, vacuolated notochord cell; Hyp, hypochord.

The gulm208 Mutation Disrupts the Zebrafish col8a1 Gene

To determine the molecular basis of the gulm208 phenotype, we meiotically mapped the gulm208 locus to a centromeric region of chromosome 9 (Fig. 3A). This region contains 26 genes (Supp. Fig. S1, which is available online), one of which encodes the alpha 1 chain of type VIII collagen (Col8a1), a known lysyl oxidase substrate (Lee and Kim,2006; Supp. Fig. S1). Cloning and sequencing of zebrafish col8a1 revealed a missense mutation in gulm208 mutants that is absent in wild-type embryos (Fig. 3B). This mutation results in the substitution of histidine for tyrosine in the C1q-like domain of Col8a1 (Y628H; Fig. 3C). Protein sequence alignment demonstrates that the tyrosine mutated in gulm208 fish is highly conserved (Fig. 3D; see also the Discussion section), and that zebrafish Col8a1 is similar to homologues in other species (Supp. Fig. S2). Sequencing of other genes located in the region between our flanking markers did not identify any additional mutations capable of causing the gulm208 phenotype (Supp. Fig. S1).

Figure 3.

The gulm208 mutation disrupts the col8a1 gene. A: The gulm208 lesion was meiotically mapped to a centromeric region of chromosome 9 bounded by markers zK46K9 and zK229B18. The number of recombinants at each marker is noted. According to the physical genome assembly (Zv7), 26 genes are located within the critical region, including the alpha 1 chain of type VIII collagen (col8a1). B: Sequencing of wild-type and gulm208 mutant cDNAs revealed a missense mutation in gulm208 mutant col8a1 that changes a tyrosine to a histidine. C: Structure of the predicted 711-amino acid zebrafish Col8a1 protein. NC1 and NC2 are two noncollagenous domains. The Y628H substitution present in gulm208 mutants (arrow) is located within the C1q-like domain (grey) of NC1. D: Protein sequence alignment of N-terminal portions of select C1q-like domains. Conserved residues, including Y628 (arrow) are shaded yellow. Sequence numbering refers to the zebrafish Col8a1 sequence (DrCol8a1).

Expression of col8a1 Is Consistent With the gulm208 Phenotype

To determine the spatiotemporal expression of zebrafish col8a1, we performed whole-mount in situ hybridization on wild-type embryos at various stages of development (Fig. 4). Consistent with the hypothesis that the Y628H substitution disrupts notochord morphogenesis, col8a1 is robustly expressed in the notochord at the 5-, 10-, 15-, and 20-somite stages (Fig. 4A–D, arrows). Starting at the 10-somite stage and continuing through 24–30 hpf, expression is also observed in the prechordal plate region (Fig. 4B–E, asterisk), floorplate (Fig. 4B–E,G, black arrowhead), and hypochord (Fig. 4B–E,G, white arrowhead). At 24 hpf, notochord staining is diminished (Fig. 4E) and col8a1 is expressed in the caudal somites (Fig. 4F, arrowheads). By 48 hpf, col8a1 expression is restricted to the forming jaw cartilages (Fig. 4H, I, arrows). A difference in staining intensity was not detected between gulm208 mutant and wild-type embryos processed in the same basket (data not shown), suggesting that mutant col8a1 transcripts are not degraded or overexpressed.

Figure 4.

col8a1 expression is consistent with the gulm208 phenotype. A–I: Whole-mount in situ hybridization for col8a1 was performed on wild-type embryos at various developmental stages. A: Lateral view of a five-somite embryo with col8a1 expression in the vacuolated notochord cells (arrow). B–D: Lateral views of 10- (B), 15- (C), or 20- (D) somite embryos with col8a1 expression in the notochord (arrow), floorplate (black arrowhead), hypochord (white arrowhead), and prechordal plate region (asterisk). E: Lateral view of an embryo at 24 hours postfertilization (hpf) with col8a1 expression in the floorplate (black arrowhead), hypochord (white arrowhead), and prechordal plate region (asterix). F: Dorsal view of an embryo at 24 hpf with col8a1 expression in the caudal somites (arrowheads). G: Lateral view of an embryo at 30 hpf with col8a1 expression in the floorplate (black arrowhead) and hypochord (white arrowhead). H: Dorsal view of an embryo at 48 hpf with col8a1 expression in jaw cartilages (arrows). I: Lateral view of an embryo at 48 hpf with col8a1 expression in jaw cartilages (arrows). J: reverse transcriptase-polymerase chain reaction (RT-PCR) for col8a1 at the indicated developmental stages. As controls, maternally-expressed 5′-AMP-activated protein kinase catalytic subunit alpha-1 (ampka1) and zygotically expressed spadetail (spt) were amplified in parallel (Griffin et al.,1998; Wagner et al.,2004). The fold induction of col8a1 relative to ampka1 is noted. Unfert, unfertilized; NA, not applicable.

To precisely map the onset of col8a1 expression, we used a more sensitive method for transcript detection, reverse transcriptase-polymerase chain reaction (RT-PCR). This revealed that col8a1 is expressed at very low levels maternally and is up-regulated between 9 hpf and the three-somite stage (Fig. 4J). Importantly, a PCR product was not obtained if the reactions were performed in the absence of cDNA (Fig. 4J).

Y628H Substitution Prevents Col8a1 Trimerization but Does Not Activate the Unfolded Protein Response

Because C1q-like domains are important for oligomerization (Innamorati et al.,2006), we tested whether the Y628H substitution impairs zebrafish Col8a1 trimer formation using an in vitro assay previously validated in studies of a closely-related collagen, collagen X (Yamaguchi et al.,1991; Chan et al.,1995,1996). While transcription and translation from either wild-type or mutant plasmid contructs results in the production of Col8a1 monomers (Fig. 5A, lanes 1 and 2), only wild-type zebrafish Col8a1 is able to trimerize (Fig. 5A, lane 1). Cotranscription and translation of half the amounts of wild-type and mutant Col8a1 together did not prevent trimerization (Fig. 5A, lane 3), consistent with loss of function of mutant Col8a1. Bands migrating faster than monomer may represent late initiation products, as they were not present in a negative control lane and were not cleaved in reactions supplemented with canine pancreatic microsomes, which facilitate proteolytic processing of the Col8a1 signal peptide (data not shown).

Figure 5.

The Y628H substitution prevents Col8a1 trimerization. A: Constructs encoding wild-type (lane 1) or gulm208 mutant (lane 2) Col8a1 were transcribed and translated in the presence of radiolabeled methionine, and the products analyzed as described under the Experimental Procedures section. Col8a1 monomers and trimers are indicated. Cotranscription and translation of half the amounts of wild-type and mutant construct together does not prevent trimerization (lane 3). The approximate location of molecular weight markers (kDa) is noted. B: Reverse transcriptase-polymerase chain reaction (RT-PCR) to assess the splice status of xbp1 in wild-type embryos and gulm208 mutants at 24 hours postfertilization (hpf). Splicing of xbp1 is not observed in gulm208 mutants but occurs after incubation of wild-type embryos in tunicamycin. U, unspliced; S, spliced. Amplification of the gene spt was performed in parallel to confirm that equal amounts of cDNA were used for each condition.

We also considered whether the gulm208 phenotype could result from a toxic effect of mutant Col8a1 aggregates, which are observed in the endoplasmic reticulum of gulm208 mutants at 24 hpf (Fig. 2B,E). These aggregates could lead to endoplasmic reticulum stress, which results in the processing of x-box binding protein-1 (xbp1) mRNA to a splice form encoding an activator of the unfolded protein response (Ron and Walter,2007). However, xbp1 splicing is not observed in wild-type embryos or gulm208 mutants at 24 hpf (Fig. 5B). Furthermore, incubation of wild-type embryos in tunicamycin, an inhibitor of N-glycosylation that causes endoplasmic reticulum stress, results in xbp1 splicing (Fig. 5B) but does not cause the gulm208 phenotype (data not shown). Finally, 4-phenylbutyric acid, a chemical chaperone that reduces endoplasmic reticulum stress (Ozcan et al.,2006; de Almeida et al.,2007), does not rescue the gulm208 phenotype (data not shown). These data suggest that the Y628H substitution causes notochord distortion through insufficient Col8a1 deposition in the notochord sheath rather than through activation of the unfolded protein response.

Morpholino Knockdown of col8a1 Recapitulates the gulm208 Phenotype

To determine whether loss of col8a1 recapitulates the notochord phenotype observed in gulm208 mutants, we designed morpholinos to knock down col8a1 in zebrafish. While injection of standard control morpholino into wild-type embryos did not result in any visible phenotypic effect (Fig. 6A; Table 1), injection of a splice morpholino targeting col8a1 caused notochord distortion indistinguishable from what is observed in gulm208 mutants (Fig. 6B; Table 1). Importantly, this splice morpholino also caused a dose-dependent reduction in col8a1 transcript, as detected by RT-PCR (Fig. 6C). Injection of a second, distinct morpholino targeting the start site of col8a1 gave identical phenotypic results when used at a dose of 2.4 ng (data not shown), confirming that the gulm208 phenotype results from loss of Col8a1. Of interest, injection of either the start or splice morpholinos at higher doses resulted in more robust notochord distortion (Fig. 6D and data not shown). Neither morpholino altered melanin pigment formation (data not shown).

Figure 6.

Morpholino knockdown of col8a1 recapitulates the gulm208 phenotype. A: Wild-type embryos injected with 12 ng of control morpholino exhibit normal morphology. B: Embryos injected with 2.4 ng of a splice morpholino targeting col8a1 develop notochord distortion (arrowheads) characteristic of gulm208 mutants. C: Reverse transcriptase-polymerase chain reaction (RT-PCR) at 24 hours postfertilization (hpf) demonstrating a dose-dependent reduction in col8a1 transcript after morpholino injection. D: Embryos injected with 12 ng of a splice morpholino targeting col8a1 develop notochord distortion (arrowheads) that is worse than in gulm208 mutants. Insets show notochord at higher magnification. All embryos were treated with PTU to inhibit melanin pigmentation and photographed at 30 hpf.

Table 1. Morpholino Knockdown of col8a1 Recapitulates the Notochord Distortion Observed in gulm208 Mutantsa
Specific morpholinoDose of morpholino (ng)Embryos scored (no.)Dead or dysmorphic embryos (no.)Notochord distortion
  • a

    Wild-type embryos were injected with morpholino and live embryos sorted to new dishes at 10 hours postfertilization (hpf); these embryos were scored for notochord distortion at 30 hpf. Data shown are the pooled results of three independent experiments.

Control1218610172 (98%)4 (2%)
col8a1 (splice)2.417022 (1%)166 (99%)

Injection of 200 pg of mRNA encoding full-length col8a1 into clutches from gulm208/+ intercrosses did not rescue the gulm208 phenotype (data not shown), whereas injection of 800 pg of mRNA caused dysmorphism during epiboly and at 26 hpf (data not shown). This suggests that col8a1 expression must be restricted to specific cell types during early development in order for rescue to be successful.


These data reveal a previously unappreciated role for the alpha 1 chain of type VIII collagen (Col8a1) in zebrafish notochord formation. Type VIII collagen is formed through homopolymerization of either alpha 1 or alpha 2 polypeptide chains (Greenhill et al.,2000) and is a member of the short-chain collagen family that includes type X collagen. The experiments with neocuproine (Fig. 1C,D) reveal a gene–nutrient interaction between copper and col8a1 that significantly alters notochord morphogenesis. The notochord sensitivity of gulm208 mutants to lysyl oxidase inhibition (Fig. 1D,F) suggests that some mutant Col8a1 may be able to fold in vivo, and residual Col8a1 function could explain the enhanced notochord distortion observed with higher doses of col8a1 morpholino (Fig. 6D). Importantly, the mechanism by which morpholinos decrease specific gene transcription precludes the accumulation of protein aggregates in the rough endoplasmic reticulum, arguing that mutant Col8a1 aggregates do not cause the gulm208 phenotype. In addition, no mutant Col8a1 is present in the morpholino-injected wild-type embryos, and thus mutant protein cannot be interfering with normal collagen assembly in the extracellular space. Taken together, the data presented here strongly suggest that the Y628H substitution causes notochord distortion through insufficient Col8a1 deposition in the notochord sheath.

Implications for Other Notochord Mutants

Our results advance the possibility that published late notochord mutants such as crash test dummy, zickzack, wavy tail, quasimodo, and kinks (Odenthal et al.,1996; Stemple et al.,1996) may encode mutations in lysyl oxidases, proteins required for lysyl oxidase activity, or lysyl oxidase substrates. In some cases, additional phenotypes may help to identify the mutant gene. For instance, quasimodo exhibits defects in both notochord and melanin pigment formation (Odenthal et al.,1996), suggesting a mutation in a protein required for global copper homeostasis. The lesion in quasimodo has been mapped to a region near the atp7a locus (Geisler et al.,2007), and may thus be allelic to calamity (Mendelsohn et al.,2006).

Our data also suggest that force generation in the notochord can be indirectly measured by quantitating the length of the yolk sac extension, which is simple to do and could prove useful in studies of other notochord mutants. Because col8a1 is not expressed in the yolk sac extension (Fig. 4), the reduction in the length of this structure in gulm208 mutants, both before and after pharmacologic treatment (Fig. 1), provides evidence for impaired force generation in the notochord, and implicates Col8a1, copper, and lysyl oxidases in this force generation process.

Conservation of Amino Acid Y628

The Y628H substitution identified in gulm208 mutants is located within the C1q-like domain of Col8a1 (Fig. 3C). This domain is present in several other proteins, including the structurally similar collagen paralogues COL8A2 and COL10A1 (Fig. 3D; Yamaguchi et al.,1991; Bogin et al.,2002; Kvansakul et al.,2003). Alignment of all 31 human C1q-like domain-containing proteins reveals only eight invariant residues, including the tyrosine corresponding to Y628 in zebrafish Col8a1 (Tom Tang et al.,2005). Furthermore, this tyrosine is conserved in all predicted 52 zebrafish C1q-like domain-containing proteins (Mei and Gui,2008). In view of such notable evolutionary conservation, it is not surprising that the Y628H substitution present in gulm208 mutants could be deleterious. Indeed, an identical substitution at the equivalent position in human COL10A1 (Y597H) causes Schmid metaphyseal chondrodysplasia, an autosomal dominant skeletal disorder (Bonaventure et al.,1995). The Y597H mutation is predicted to prevent proper folding of the COL10A1 C-terminal noncollagenous domain (NC1) in the endoplasmic reticulum and to cause disease by haploinsufficiency (Bogin et al.,2002). This mechanism may also account for the gulm208 phenotype because the crystal structures of the COL10A1 and COL8A1 NC1 domains are very similar and in each case the histidine would destabilize the hydrophobic core of the protein (Bogin et al.,2002; Kvansakul et al.,2003).

Our data support the hypothesis that the Y628H substitution inhibits proper folding of individual Col8a1 chains. First, the large circular aggregates visualized by electron microscopy are consistent with unfolded Col8a1 monomer that is retained in the rough endoplasmic reticulum (Fig. 2B,E, arrows). Similar aggregates resulting from protein retention in the endoplasmic reticulum are observed in three zebrafish coatomer mutants, although the cause of retention is different in these mutants and leads to a more severe phenotype (Coutinho et al.,2004). Second, the in vitro trimerization assay directly demonstrates a defect in mutant Col8a1 trimerization (Fig. 5A, lane 2 vs. lane 1). This presumably results from a primary defect in NC1 domain folding rather than a defect in interchain assembly because the mutated tyrosine is buried in the NC1 domain (Kvansakul et al.,2003). We propose that the Y628H substitution in Col8a1 inhibits monomer folding in gulm208 mutants. Because the NC1 domain nucleates trimer formation (Brass et al.,1992; Zhang and Chen,1999; Illidge et al.,2001), this unfolded Col8a1 monomer would subsequently fail to trimerize and form higher-order assemblies (Fig. 7).

Figure 7.

Model of Col8a1 assembly in wild-type and gulm208 mutant embryos. Col8a1 monomer containing the Y628H mutation is unable to fold, preventing trimer formation and higher-order assembly that occurs in the absence of the mutation. The Y628H mutation is located within the NC1 domain of Col8a1 (asterisk).

Comparative Expression of type VIII Collagen

Zebrafish col8a1 expression is restricted to a few tissues during early development (Fig. 4), and the absence of col8a1 expression in zebrafish eye was initially puzzling because Type VIII collagen is a major component of Descemet's membrane in other vertebrates (Labermeier and Kenney,1983; Kapoor et al.,1986; Sawada et al.,1990; Kabosova et al.,2007). However, zebrafish encode a paralogue of col8a1 located on chromosome 22 (BAC CR956626), and an expressed sequence tag encoding part of this paralogue (DN898414) has been isolated from zebrafish eye, specifically from the anterior segment where Descemet's membrane is located. Thus, it appears that partitioning of tissue-specific patterns of expression has occurred (Lynch and Force,2000), and type VIII collagen distribution in zebrafish is expected to be broader than what is predicted from the expression data presented here (Fig. 4).

Previous work demonstrates that type VIII collagen is distributed in the cartilage matrix and perichondrium of fetal calf tissues (Kapoor et al.,1988). The zebrafish notochord is most closely related to cartilage as a tissue (Stemple,2005), and expression of col8a1 in notochord and jaw cartilages (Fig. 4A–D,H, I) suggests an association between cartilage and type VIII collagen deposition that is evolutionarily conserved during early development. This association is supported by the similar spatiotemporal expression profiles of col8a1 and cartilage-expressed col2a1 in zebrafish embryos (compare Fig. 4 with Fig. 3 of Yan et al.,1995) (Yan et al.,1995). Importantly, loss of wild-type col8a1 expression in jaw cartilages likely causes the severe head malformation noted in gulm208 mutants by 72 hpf (Stemple et al.,1996), and thus, the two principal phenotypes of gulm208 mutants (notochord distortion and head malformation) involve the major cartilage and cartilage-related tissues of zebrafish during early embryogenesis. Mice lacking both alpha 1 and alpha 2 chains of type VIII collagen exhibit anterior segment abnormalities in the eye but are not noted to have notochord or jaw malformations (Hopfer et al.,2005). The reasons for these differences are presently unknown, though Col8a1 is expressed in at least some cartilages of newborn mice (Muragaki et al.,1992).

Short-chain Collagens in Human Disease

Col8a1 belongs to a small group of highly related short-chain collagens, and the gulm208 mutant zebrafish may provide insight into the mechanism by which specific types of point mutations in these collagens cause disease. Indeed, our findings may be relevant to the molecular pathogenesis of Schmid metaphyseal chondrodysplasia, which is currently debated. Schmid metaphyseal chondrodysplasia results from mutations at multiple sites in COL10A1, and functional haploinsufficiency due to nonsense-mediated decay of mutant message appears to cause this disorder (Chan et al.,1998; Bateman et al.,2003). However, an effect of misfolded protein on chondrocyte differentiation secondary to activation of the unfolded protein response has also been invoked (Ho et al.,2007; Tsang et al.,2007). The determination of whether a missense mutation causes disease through loss of function or endoplasmic reticulum stress is potentially critical because treatment options may differ. 4-Phenylbutyric acid has been shown to reduce endoplasmic reticulum stress and restore glucose homeostasis in a mouse model of type 2 diabetes (Ozcan et al.,2006) and is already approved by the U.S. Food and Drug Administration for the chronic management of certain urea-cycle disorders (Maestri et al.,1996). Thus, a disease caused by endoplasmic reticulum stress may be amenable to treatment with this chemical chaperone. Our data suggest that mutations affecting monomer folding (Fig. 7) cannot be rescued by 4-phenylbutyric acid (Fig. 5B and data not shown).

To date, no human disease has been shown to result from mutations in COL8A1. While COL8A1 has been evaluated as a candidate gene for certain eye disorders due to the presence of type VIII collagen in Descemet's membrane and to the fact that mutations in COL8A2 cause corneal dystrophies (Gottsch et al.,2005; Liskova et al.,2007; Mok et al.,2008), the data presented here suggest that a role for COL8A1 in cartilage development should now be considered. Indeed, the expression pattern of col8a1 in zebrafish (Fig. 4) and the notochord and head phenotypes of gulm208 mutants (Stemple et al.,1996) suggest that mutations in COL8A1 could result in chondrodysplasias. Several different mutations in COL10A1 cause Schmid metaphyseal chondrodysplasia (Bateman et al.,2005), but not all patients with this disease have an identifiable mutation in COL10A1 and additional unknown loci are implicated in its pathogenesis (Bonaventure et al.,1995; Wallis et al.,1996; Ridanpaa et al.,2003; Bateman et al.,2004). COL8A1 may thus be a candidate gene for Schmid metaphyseal chondrodysplasia or another chondrodysplasia.


Zebrafish Maintenance

Gulliverm208 was obtained from the Zebrafish International Resource Center (Eugene, OR). 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. In Figures 1, 2, and the first two lanes of 5B, “wild-type” refers to either +/+ or +/− embryos.


Live embryos were anesthetized in tricaine, mounted in 2% methylcellulose, and imaged using an Olympus SZX12 zoom stereomicroscope fitted with an Olympus DP70 camera. Electron microscopy was carried out as previously described (Gansner et al.,2007) except that negatives were directly scanned using a Canon Canoscan 8400F. For the electron microscopy, sections from two wild-type and two mutant embryos were examined.

Pharmacologic Treatment and Yolk Sac Extension Measurements

Pharmacologic compounds were purchased from Sigma (St. Louis, MO). β-aminopropionitrile (A3134) and neocuproine (N1501) were prepared as 100 mM stocks in egg water (Westerfield,1993) and dimethyl sulfoxide, respectively, and diluted in egg water. Tunicamycin (T7765) was prepared as a 1 mg/ml stock in dimethyl sulfoxide (DMSO) and used at a concentration of 2 μg/ml. 4-Phenylbutyric acid (P21005) was prepared as a 2 M stock in DMSO and used at a concentration of 5 mM in buffered egg water adjusted to a pH of 6.9 (Mendelsohn et al.,2008). In all cases, embryos were placed in compound between 6 and 10 hpf. Lengths of individual yolk sac extensions were measured from photographs of 10 mutant and 10 wild-type embryos per pharmacologic treatment using ImageJ software (Rasband,1997–2007). Values are reported in arbitrary units, with standard deviations noted.

Meiotic Mapping

Gulliverm208 was mapcrossed to the polymorphic WIK strain and the progeny (AB*/WIK) raised to adulthood. The gulm208 mutation was assigned to chromosome 9 by centromeric linkage analysis (Johnson et al.,1995,1996) using simple sequence length polymorphism (SSLP) markers (Shimoda et al.,1999) and DNA from early pressure gynogenetic diploids. For fine mapping, 1,295 mutant embryos from AB*/WIK females crossed to AB*/AB males were collected and assessed for recombination along chromosome 9 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 ( Primers for the BAC markers used in Figure 3 are: zC81J7, forward 5′-TTGTTCTGCAAATTTTGTTGG-3′ and reverse 5′-GGGGCAACCCCTCTAAAGT-3′; zK46K9, forward 5′-AACACAAGCTGGGACTGGAC-3′ and reverse 5′-GATGTCTAACACAAACAC- ATTGG-3′; zC206L17, forward 5′-GAGTATCACCTCTGACAGATGGG-3′ and reverse 5′-GCAGCATTGACGGTGAAGGTCAC-3′; zK229B18, forward 5′-TGATTAATACAACATGGGCA-3′ and reverse 5′-CGCACTGTGAAATAACATGA-3′; zC184M13, forward 5′-GGAAGTGTGTGTGTGCGTTT-3′ and reverse 5′-GGCCACAAGAACCATGACA-3′ (descendents of one WIK grandparent) and also forward 5′-TGGTGTAGGGGGCTAACAAG-3′ and reverse 5′-TGAACTGAACTGGCATGAACA-3′ (descendents of the other WIK grandparent). DNA from wild-type and heterozygote embryos as well as WIK and AB grandparents was used to ensure polymorphism between AB and WIK.

Cloning and Annotation

Full-length col8a1 was amplified from wild-type and mutant cDNA and cloned into pCR-XL-TOPO (Invitrogen). cDNA for these reactions was generated with Superscript III reverse transcriptase (Invitrogen) and polymerase chain reaction (PCR) carried out using Phusion DNA polymerase and high-fidelity buffer (Finnzymes). The primers used for PCR were: forward primer 5′-GAGTGAGCCCACCAATCCTTG-3′ and reverse primer 5′-CCTTCAAATTCTTACTTATTCTTGC-3′. The full-length zebrafish col8a1 coding sequence is available at Genbank (accession no. EU781032). The protein sequence alignment of zebrafish Col8a1 with orthologues from other species was created using ClustalW2 (Larkin et al.,2007). Signal peptides were predicted using SignalP 3.0 (Bendtsen et al.,2004).

Trimerization Assay

Wild-type and mutant col8a1 sequences were subcloned into pCS2+ and in vitro transcribed and translated in the presence of TRAN35S-LABEL (MP Biomedicals) using the SP6 TNT Coupled Reticulocyte Lysate System (Promega) essentially as described (Chan et al.,1995,1996). Five microliters of each 25-μl reaction was mixed with an equal volume of 2× sample buffer containing 250 mM Tris pH 6.8, 4 M urea, 4% sodium dodecyl sulfate (SDS), 24 mM dithiothreitol, and a small amount of bromophenol blue. Samples were heated at 50°C for 5 min and subjected to SDS-polyacrylamide gel electrophoresis on a 1-mm gel with 3.5% stacking and 7.5% resolving layers. The gel was fixed for 30 min in a mixture of 20% methanol and 10% glacial acetic acid, washed in water for 5 min, and vacuum dried at 80° C for 2 hr before exposure of a low-energy phosphor screen (Amersham) and image acquisition using a Storm imaging system (Amersham).

Whole-Mount In Situ Hybridization and Frozen Sections

Embryos were manually dechorionated at the indicated developmental stages, fixed in 4% paraformaldehyde-phosphate buffered saline overnight at 4°C, and dehydrated by methanol series. The probe construct was generated by cloning the 5′ end of zebrafish col8a1 into pCRII (Invitrogen). The primers used for the PCR amplification reaction were: forward primer 5′-GAGTGAGCCCACCAATCCTTG-3′ and reverse primer 5′-CTCCTGGACTTCCAATACCC-3′. Digoxigenin (DIG) -labeled antisense RNA probes were synthesized using a DIG-labeling kit (Roche), and whole-mount in situ hybridization was performed as previously described (Thisse et al.,1993; Mendelsohn et al.,2006).


RNA was obtained from pooled embryos or unfertilized eggs using Trizol reagent (Invitrogen). For Figures 4J and 5B, cDNA was prepared from 500 ng of RNA using an oligo dT primer and Superscript III reverse transcriptase (Invitrogen). PCR amplification over 35 cycles was subsequently performed using 1 μl of cDNA and GoTaq polymerase (Promega) in a final volume 25 μl. The annealing temperature was 57°C and the extension time 1 minute. Primers to col8a1 were: forward primer 5′-GAGTGAGCCCACCAATCCTTG-3′ and reverse primer 5′-CCTCTGGGAATGGTCTCACC-3′. Primers to xbp1 were forward primer 5′- GTTCAGGTACTGGAGTCCGC-′3 (Hu et al.,2007) and reverse primer 5′-GGATGTCCAGAATACCAAGCAGG-3′. Primers to ampka1 and spt (also known as tbx16) have been reported previously (Mendelsohn and Gitlin,2008). For Figure 6C, cDNA was prepared from 1 μg of RNA and the annealing temperature was 55°C. The forward primer to col8a1 was as noted above, but the reverse primer was 5′-GACATTCCTGGCTTACCAATTCC-3′. Band quantitation was performed using ImageJ software (Rasband,1997–2007). For Figure 4J, band quantitation of col8a1 was calculated using a different exposure of the same gel where the bands were not saturated.

Morpholino and mRNA Injections

Morpholino oligonucleotides (Nasevicius and Ekker,2000) targeting start and splice sites in col8a1 were resuspended in Danieau buffer (start MO) or water (splice MO), diluted to include 0.05% phenol red, and injected into one-cell embryos. Standard control morpholino (Gene Tools, LLC) was resuspended in Danieau buffer and likewise injected. Morpholino sequences were 5′-CCGTAGGAGAAGATAATCTCAAGGA-3′ (start MO) and 5′-TAAAGTGTATCTCCTTACCTTTCCT-3′ (splice MO). The start MO was used at doses of 2.4 ng and 7.2 ng. Capped, polyadenylated mRNA for rescue experiments was generated from wild-type or mutant full-length clones of zebrafish col8a1 using the mMESSAGE mMACHINE kit (Ambion). Individual embryos from gulm208/+ intercrosses were injected with either 200 or 800 pg of mRNA.


We thank Marilyn Levy for the electron microscopy and Stephen Johnson for help with meiotic mapping. We also thank Stephen Johnson, Erik Madsen, and Robert Mecham for careful review of the manuscript. J.M.G. was supported by an NIH Medical Scientist Training Program grant and J.D.G. was funded by the NIH.