Core fucosylation is required for midline patterning during zebrafish development

Zebrafish fucosyltransferases

The zebrafish database was analyzed for sequences homologous to the mammalian FucTs families. As reported by others, the zebrafish database does not contain α1,2 FucT sequences (Kageyama et al., 1999). On the other hand, there appear to be five zebrafish α1,3/4 FucTs, similar to the presence of multiple α1,3/4 FucTs in other species, and only one α1,6 FucT, as is the case in mammals (Ma et al., 2006) (Fig. 1B). As noted above, the loss of FucT8 leads to growth retardation and neonatal lethality in mice (Wang et al., 2005), although the consequences on embryonic development and patterning have not been addressed. Therefore, in this study we focused on the role of α1,6 FucT (FucT8) during zebrafish development; a study that is facilitated by the apparent lack of redundant family members. The zebrafish fuct8 gene is 1746bp long with 9 exons and predicts a protein with 78% identity to mammalian FucT8 (Fig.1C).

FucT8 morphants resemble midline pathway mutants

To determine the possible roles of FucT8 during zebrafish development, 1-2 cell stage embryos were injected with antisense morpholino oligonucleotides (MO) designed against a FucT8 splice donor site. RT-PCR analysis confirmed the efficacy of the splice blocking MO to prevent FucT8 splicing (Fig. S1). Parallel studies using a translation blocker antisense MO produced similar phenotypes to that seen with splice blocker MO. For controls, embryos were injected with identical levels of control oligonucleotides as indicated in the Experimental Procedures.

Out of more than 500 embryos injected with FucT8 MOs, 85-90% showed a similar morphant phenotype. At 24 hours post fertilization (hpf), the FucT8 MO-injected embryos have smaller eyes as compared to controls. This phenotype is more apparent at 48 hpf when the embryos show mild cyclopia (i.e., close set eyes), small forebrains, and a hindbrain ‘bump’ that appears to reflect a swollen hindbrain ventricle (Fig. 2A-C). Furthermore, FucT8 MO-injected embryos have U-shaped somites and a highly curved body axis (Fig. 2B). Most of these phenotypes, including mild cyclopia, U-shaped somites and a curved body axis, are similar to those seen in previously described zebrafish midline pathway mutants (Brand et al., 1996). These observations suggest that FucT8 might be required for midline signaling.

We confirmed that the MO injection reduced FucT8 activity by staining embryos with AAL, a lectin that specifically recognizes α1,6-linked fucose moieties. Control-injected embryos retain strong AAL labeling, which is reduced to background levels following MO injection, thus confirming the loss of FucT8 activity (eye sections shown in Fig. 2D).

Loss of differentiated retinal cells in FucT8 morphants

Since eye development appeared to be a clear target of FucT8 MO injection, we characterized the extent of retinal cell differentiation in control and FucT8 MO-injected embryos. DAPI staining shows that although the FucT8 morphant eye is much smaller than controls (Fig. 3A), the retina contains numerous cells that are loosely stratified into layers, similar to that seen in some patterning mutants, such as smu (Masai et al., 2005). Retinal ganglion cells (RGCs), Müller glial cells and photoreceptors were identified within the retina using zn-5, anti-carbonic anhydrase II (CA-II), and zpr-1 antibodies, respectively. RGCs are the first cells to differentiate in the retina and appear around 28 hpf (Laessing and Stuermer, 1996). Even by 48 hpf, FucT8 morphants have few to no RGCs (Fig. 3B). As a result, the optic nerve and optic chiasm are not seen in morphant embryos (Fig. 3C). Müller glia and photoreceptor cells are also extremely reduced in the morphants (Fig. 3D,E). The reduced eye size in FucT8 morphants is not a consequence of decreased cell proliferation as assayed by BrdU incorporation (Fig. 3F); rather, BrdU incorporation appears higher in FucT8 morphant eyes, similar to that reported in smu mutant eyes (Masai et al., 2005). Furthermore, FucT8 morphants show a slight overall increase in apoptotic cells throughout the embryo as assayed by Acridine Orange (Fig. S2), similar to that reported in the smo mutant (Chen et al., 2001), and which might contribute to the overall reduced size of the FucT8 morphant embryos, relative to controls. Regardless, the absence of multiple cell types suggests that core fucosylation may play some role in patterning of the retina.

Earlier patterning defects in the eye

To determine if the lack of differentiated cell types in the retina results from defective patterning, we analyzed the expression of well-established patterning markers by in situ hybridization. Islet-1 is a lim homeodomain transcription factor that is expressed in the neural retina and elsewhere in the embryo (Inoue et al., 1994). Although present in other domains, islet-1 expression is lost in the developing eye of FucT8 morphants (27 hpf, Fig. 4A). ath5 is a homolog of mammalian atonal and is expressed in RGC precursors (Masai et al., 2000). ath5 expression is lost in FucT8 morphants, consistent with the loss of RGCs seen at 48 hpf (Fig. 4B). Finally, pax 2.1 labels the astrocytes in the optic stalk (Krauss et al., 1991), and at 27 hpf is mis-expressed at the midline of FucT8 morphants, which illustrate a shorter, or reduced, optic stalk and expanded pax 2.1 expression in eyes (Fig. 4C). Collectively, these results implicate FucT8 as being essential for proper specification of the retina.

Other cell types and tissues affected in FucT8 morphants

As discussed above, FucT8 morphant embryos display patterning defects similar to those seen in midline pathway mutants. We therefore examined for the presence of midline patterning defects by assessing the expression of various markers. In the spinal cord, islet-1 labels the primary motor neurons at 24 hpf and all motor neurons by 48 hpf. When stained with islet-1 at 48 hours, the motor neurons in FucT8 morphants appear disorganized, being scattered along the dorsal-ventral axis (Fig. 5A). This is accompanied by a loss of axonal projections from motor neurons, as viewed by anti-acetylated tubulin antibody (Fig. 5B). Dorsal root ganglion cells are also significantly reduced in the morphants and lack any axonal projections as judged by zn-5 labeling (Fig. 5C).

The spinal cord defects seen in FucT8 morphants are associated with U-shaped somites that have fewer and shorter muscle fibers than control somites, as assessed by anti-myosin heavy chain (F-59) antibody (Fig. 5D). Similarly, the ventral migration of neural crest cells into the somitic mesoderm is greatly reduced (Fig. 5E), as are the jaw cartilages that derive from neural crest precursors (Fig. 5F). Other midline defects include a lack of hindbrain commissures (Fig. 5G). In summary, FucT8 morphants show a broad range of midline patterning defects indicating loss of proper midline signaling in the absence of α1,6 fucosylation.

Identification of the FucT8 substrate

In order to identify the target proteins that are α1,6 fucosylated, we first compared the AAL-reactive glycoproteins in lysates of control and FucT8 MO-injected embryos by SDS-PAGE and lectin blotting. AAL specifically recognizes α1,6 fucosyl residues, and as expected, lectin staining was inhibited by an excess of free fucose (bracket, Fig. 6A). FucT8 morphants have greatly reduced AAL reactivity in the high molecular weight area, suggesting that one of the major FucT8 target proteins is ≥ 200 kDa. The overall protein profile showed no difference in overall staining intensity (data not shown), indicating that the reduced AAL staining was not a result of reduced protein load for morpholino-injected embryos.

AAL-conjugated agarose beads were used to precipitate the target protein from control and morphant lysates. Similar to what was found by AAL lectin blotting, AAL lectin precipitated a complex of two-three large molecular weight glycoproteins in control lysates that was absent in FucT8 morphants (bracket, Fig. 6B). The corresponding AAL-reactive bands were excised from Coomassie-stained gels and sequenced by Mass Spectrometry (MS). Of the 21 candidate matches, the “hypothetical” protein LOC321166, molecular weight of 412 kDa, was represented by 96 distinct peptide sequences (score = 5385), whereas all others were represented by four or fewer peptides (scores range from 31-149). BLAST searches reveal that LOC321166 has high homology to human and mouse Apolipoprotein B 100 (ApoB), and is predicted to be zebrafish ApoB.

In retrospect, it is not surprising that the AAL-reactive proteins migrated on SDS-polyacrylamide gels as a complex of large molecular mass proteins, since others have reported that ApoB 100 migrates as a family of related polypeptides, some of which are constitutively seen when resolved on SDS-PAGE, with others being the result of posttranscriptional editing of the ApoB mRNA (Chen et al., 1986; Young et al., 1986). Nevertheless, sequence analysis verified that each of the individual high molecular weight AAL-reactive bands were derived from the ApoB polypeptide. Furthermore, when AAL-reactive proteins were collected by AAL-agarose beads and resolved by SDS-PAGE, LOC321166 (zfApoB) was again identified by MS sequence analysis.

ApoB is α1,6 fucosylated

Human ApoB is a large molecular weight glycoprotein with a well-defined role in lipid metabolism and transport of cholesterol (Fisher and Ginsberg, 2002). To confirm that ApoB is indeed α1,6 fucosylated, we used the human hepatocyte cell line, HepG2, since there are no known antibodies that recognize zebrafish ApoB. HepG2 lysates were incubated with either AAL-agarose to identify α1,6 fucosylated glycoproteins, or with anti-ApoB antibodies; the precipitates were collected, washed and resolved by SDS-PAGE. The proteins were transferred to PVDF and stained with anti-ApoB antibody. Anti-ApoB antibody detected a high molecular weight polypeptide in Hep2G lysates that was of similar size to the ApoB-reactive polypeptide precipitated by AAL beads (Fig. 6C). As expected, the immunoprecipitated ApoB was reactive with AAL by lectin blot analysis (data not shown). These results confirm that ApoB is α1,6 fucosylated, and are consistent with the loss of AAL reactivity following FucT8 morpholino injection.

ApoB morpholino injections phenocopy FucT8 morphant phenotypes

If the loss of ApoB core fucosylation is indeed one of the primary targets of FucT8 morpholinos, then ApoB knockdown should phenocopy the FucT8 morphants. 1-2 cell stage embryos were injected with 5-8 ng of either one of two ApoB translation blocking morpholinos and probed with markers that were used to characterize the FucT8 morphants. Both ApoB morpholino oligonucleotides produced similar phenotypes (~ 400 embryos injected), at ~90% penetrance. Consistent with ApoB being a target of FucT8 morpholino knockdown, ApoB morphants display similar or more severe phenotypes as the FucT8 morphant embryos. For example, islet-1 expression in the eye is greatly reduced at both 27 and 48 hpf (arrowheads, Fig. 7A,B), similar to that seen in FucT8 morphants (Fig. 4A). Similarly, the occurrence of RGC precursors, as assayed by ath-5 expression at 27 hpf, is greatly reduced in both ApoB (arrowheads, Fig. 7C) and FucT8 morphants (Fig. 4B). The optic stalk is also reduced, or absent, in both ApoB (arrow, Fig. 7D) and FucT8 morphants (Fig. 4C), as assayed by pax 2.1 expression. Finally, immunolabeling with zn-5 antibodies illustrates the loss of RGCs (arrowheads) and optic chiasma (arrow) in 48 hpf ApoB morphants, similar to that seen in FucT8 morphants (Fig. 3C). In summary, the ApoB and FucT8 morphants display similar phenotypes as assessed by these markers, and further suggest that the lack of ApoB α1,6 fucosylation leads to early midline patterning defects.

Knockdown of FucT8 leads to defective midline patterning

Knocking down FucT8 during zebrafish development results in phenotypes that are very similar to those seen in midline patterning mutants, in general, and in Shh pathway mutants and morphants, in particular. For example, FucT8 morphants have reduced or disorganized motor neurons and lack axonal projections from motor neurons and dorsal root ganglia, similar to Shh pathway mutants sonic you (syu), smoothened (smo) and detour (dtr) (Lewis and Eisen, 2001; Ungos et al., 2003). Shh signaling is also required for slow muscle formation and proper somite architecture, as evidenced by the smu and yot mutations, and which are also sensitive to FucT8 knockdown (Barresi et al., 2000; Lewis et al., 1999). Similarly, neural crest migration is disrupted in FucT8 morphants, leading to jaw defects and other anomalies, as seen in the syu and smu mutants (Barresi et al., 2000; Honjo and Eisen, 2005). Shh pathway mutants and morphants are also characterized by severely reduced numbers of RGCs, Müller glia and photoreceptors in the developing retina, which is seen in FucT8 morphants as well (Shkumatava et al., 2004). Similarly, the increased cell proliferation seen in FucT8 morphant eyes is similar to that reported by Masai et al. (2005) in smu mutants, consistent with their suggestion that Shh signaling is required for exit from the cell cycle. All these results strongly indicate that core fucosylation is necessary for proper eye development and midline patterning during zebrafish development. Initial attempts to rescue the FucT8 morphant phenotypes by co-injection of FucT8 mRNA were unsuccessful, but we were unable to determine if the RNA was properly translated and/or stable. Nevertheless, the lack of AAL reactivity clearly indicates the loss of FucT8 activity in FucT8 morphant embryos.

Apolipoprotein B function and core fucosylation

Analysis of the core fucosylated glycoproteins during zebrafish development suggests that ApoB serves as one of the major FucT8 substrates, and that failure of ApoB core fucosylation leads to defective midline patterning. This is supported by, among other results, the ability of ApoB morpholinos to phenocopy the FucT8 morphant phenotypes.

ApoB forms the protein component of low-density lipoprotein particles (LDL), which transport plasma lipids and carry free or esterified cholesterol. In mice, the loss of ApoB leads to hypercholesterolemia in heterozygous animals, whereas most ApoB homozygous-null mice die during embryonic development between days 9-11.5 and show severe neurological defects, including exencephalus. Heterozygous mice survive further into development, and are characterized by incomplete neural tube closure and hydrocephalus (Homanics et al., 1993; Huang et al., 1995; Farese et al., 1995). Similarly, functional ApoB appears necessary for viability in humans, as the only known ApoB mutations are single point substitutions in the receptor-binding domain that are maintained in heterozygous form and lead to hypobetalipoproteinemia and hypercholesterolemia (Soutar and Naoumova, 2007). All of these findings suggest that ApoB plays critical roles during early development in both mice and humans.

Although ApoB is heavily glycosylated, possessing at least 16 N-linked oligosaccharide chains, the function of ApoB glycans is not yet fully understood. It is known that N-linked chains are required for ApoB secretion and assembly into LDL particles, since tunicamycin treatment or mutagenesis of the glycosylation motifs leads to greatly reduced ApoB secretion and LDL assembly, as well as increased intracellular degradation of ApoB (Taniguchi et al., 1989; Liao and Chan, 2001; Vukmirica et al., 2002; Harazono et al., 2005). Interestingly, seven of the 16 N-glycosylation sites are located within the ApoB domain responsible for LDL-receptor binding as well as interactions with heparan sulfate proteoglycans. However, it is not yet clear how N-linked glycans influence receptor binding and/or interactions with the extracellular matrix (Yang et al., 1989; Vukmirica et al., 2002; Harazono et al., 2005). Nevertheless, these studies indicate that ApoB activity is greatly influenced by the extent and quality of its N-glycosylation. Results presented here support and extend this view, in that the loss of a terminal fucosyl residue, rather than the global loss of all N-linked chains, also leads to a loss of ApoB activity.

Studies from other model systems indicate that core fucosylation of cell surface receptors influences ligand binding and/or ligand induced intracellular signaling (Li et al., 2006; Wang et al., 2006a; Wang et al., 2006b; Zhao et al., 2006). In this light, core fucosylation of ApoB N-linked chains that reside within the LDL receptor-binding motif may influence ApoB binding to its receptor, such as megalin (Lpl-2), and/or its ability to elicit LDL receptor signaling. Alternatively, the loss of core fucosylation may prevent ApoB secretion, as is the case when the entire N-linked chains are eliminated (Taniguchi et al., 1989; Liao and Chan, 2001; Vukmirica et al., 2002; Harazono et al., 2005). In this regard, FucT8 morphants appear to have somewhat reduced levels of ApoB protein as judged by protein-stained gels (Fig. 6D). Nevertheless, further studies in better defined model systems are required to distinguish between these, as well as other, possible modes of action. In any event, the loss of core fucosylation leads to a loss of ApoB activity and a consequent reduction in midline signaling.

Sonic hedgehog signaling in FucT8 and ApoB morphant embryos

Since most of the phenotypes observed in the FucT8 and ApoB morphants are similar to those seen in Shh pathway mutants and Shh morphant embryos (Barresi et al., 2000; Honjo and Eisen, 2005; Lewis and Eisen, 2001; Shkumatava et al., 2004; Ungos et al., 2003), we are obviously interested in whether Shh signaling is altered in these morphant embryos. As an initial approach to this possibility, we examined the expression of the Shh receptor, patched-1 (ptc-1), a well-established marker of Shh levels (Chen and Struhl, 1996). Interestingly, ptc-1 expression is upregulated around the notochord and neural tube in 24 hpf FucT8 and ApoB morphant embryos (Fig. S3). The increased expression of ptc-1 close to the source of Shh release suggests that Shh may not be transported appropriately and accumulates around its source in the absence of FucT8/ApoB leading to increased ptc-1 expression. Although these results are preliminary and there are likely other factors contributing to the morphant phenotypes, these data are consistent with the possibility that loss of core fucosylation leads to a nonfunctional ApoB that is normally required for Shh transport and long-range signaling.

Recent studies in model organisms show that lipoproteins can serve as direct carriers of signaling proteins, such as Shh and Wnt, thus establishing and/or maintaining the morphogen gradient (Panakova et al., 2005; Willnow et al., 2007). Shh is one of the most extensively studied signaling molecules in vertebrate development, the activity of which requires post-translational processing, including N-terminal palmitoylation and the addition of a cholesterol moiety at the C-terminus (Jeong and McMahon, 2002; Guerrero and Chiang, 2007). The cholesterol component has been implicated in both long-range Shh signaling as well as restricting the spread of Shh (Lewis et al., 2001; Li et al., 2006). Similarly, the precise role of the palmitate modification is undefined, but recent studies suggest that it is required for the production of large molecular weight hedgehog signaling complexes and long-range signaling (Chen et al., 2004). In this light, it is particularly interesting that the loss of functional ApoB, resulting from FucT8 knockdown, leads to phenotypes similar to those seen in loss of Shh signaling, consistent with the view that ApoB-containing lipoproteins facilitate Shh signaling in vertebrates, as they do in lower organisms (Panakova et al., 2005).

In summary, we show here for the first time that core fucosylation of ApoB is essential for proper midline patterning, presumably by facilitating the transport of lipid-modified signaling molecules, such as Shh. Further studies are required to explore this possibility in more detail.

Experimental Procedures