Zebrafish N-Cadherin Is Essential for Retinal Lamination
To provide insight into the role of N-cad during retinal development, we analyzed different stages of the N-cad loss-of-function mutant parachutepaR2.10 (pacpaR2.10; Lele et al., 2002) in toluidine blue–stained thin sections (Fig. 1).
Figure 1. Eye development in the N-cadherin pacpaR2.10 mutant. A,B: Wild-type (WT) and mutant eyes in cross-section at 24 hours postfertilization (hpf). The WT eye cup is well formed and contains a lens (L) primordium and a pseudostratified retinal epithelium. Compared with the elongated and highly organized cells in the WT retina (arrowheads), mutant retinal cells are more disorganized and more rounded (arrowheads). C–E: WT and mutant eyes at 48 hpf. At this stage, the retinal lamination defect, including rosette formation, is clearly visible in the mutant (examples of rosettes are indicated by dotted lines). Arrows point to corresponding cell types in the WT and mutant: yellow arrows, retinal ganglion cells; white arrows, inner and outer plexiform layer; orange arrows, inner nuclear layer; green arrows, photoreceptors; red arrows, lens epithelium. The pigment epithelium can be easily seen because of its black pigment. E: In the mutant, it often does not fully surround the eye cup. Mutant cell bodies lack a uniform orientation like in WT. Unlike WT, the mutant optic nerve is often bordered by retinal ganglion cells (E, yellow arrows, compare with WT optic nerve in F). F–H: WT and mutant eyes of 3-day-old fish. Arrows are as above. In most mutants, eyes and brain are not fully separated (G, also J,L). The mutant lens epithelium in G (sectioned peripherally) contains some rounded-looking cells, whereas the mutant lens epithelium in H (sectioned centrally) is completely normal. I–M: WT and mutant eyes of 5-day-old fish. Arrows are as above. The lens is sometimes located off-center in the mutant, resulting in retinal sections without lens (L) and sections, where the lens (sectioned slightly peripherally) is not surrounded by much retina (M). Arrowheads in L indicate that, in the brain, rosettes are formed like in the retinae. The mutant photoreceptor cells appear more rounded and are sometimes moved away from the pigment epithelium, forming rosettes (green arrows). The mutant lens epithelium of the centrally sectioned lens in J shows clear abnormalities (blow up: K), in that cells are rounded and multilayered.
Download figure to PowerPoint
At 24 hpf, the wild-type (WT) zebrafish eye cup is well formed and contains a lens primordium (Fig. 1A; Schmitt and Dowling, 1994). WT cells are organized within a pseudostratified epithelium in which each cell is elongated (Fig. 1A, arrowheads) and contacts the vitreal surface at the lens with its basal side and the ventricular surface with its apical side. pacpaR2.10 mutant retinae show an overall more disorganized structure (Fig. 1B). Retinal cells are often more rounded (arrowheads) and are not arranged in a uniform orientation.
Between 24 and 48 hpf, lamination occurs in WT, first in the central and slightly later in the peripheral retina and, although most synapses have not formed yet, all major cell types are present (Fig. 1C; Schmitt and Dowling, 1999; Malicki, 2000). In the same developmental interval, rosettes form in the pacpaR2.10 mutant retina (Figs. 1D, 2A). They typically contain plexiform layer tissue, surrounded by inner nuclear layer cells or photoreceptor cells. Mutant ganglion cells are often excluded from these rosettes and form distinct clusters (yellow arrows Fig. 1D,G,H,J) or line the optic nerve (yellow arrows in Fig. 1E). Photoreceptor cells sometimes form rosettes of their own (green arrows in Figs. 1D, 2A).
Figure 2. Rosette formation in the pac mutant. Arrows are as in Figure 1. A: Larger magnification of Figure 1D, showing three types of retinal rosettes at the 3-day stage: rosettes containing islands of inner plexiform layer tissue (IPL), rosettes containing islands of outer plexiform layer tissue (OPL), and rosettes of photoreceptors (green arrows). B: Larger magnification of Figure 1J, showing the same types of rosettes as in A at the 5-day stage. C: Cell cluster or rosette formed by retinal ganglion cells at day 5, magnified from Figure 1J. D: Rosette-like structures of the pigment epithelium, larger magnification of Figure 1L. E–G: Electron photomicrographs of 5-day-old retinae. E: WT outer plexiform layer and surrounding tissue. INL, inner nuclear layer; ONL, outer nuclear layer containing photoreceptor nuclei; N, nucleus. F: Mutant inner plexiform layer island and surrounding nuclei of the inner nuclear layer. Some mutant retinae contain small pigment epithelium (PE) islands within other retinal tissue. G: Mutant outer plexiform layer island, surrounded by nuclei of the inner or outer nuclear layer. Original magnifications = ×2,200 in E–G.
Download figure to PowerPoint
All cell types appear to be present as judged from histology; however, there is a clear lamination defect in the mutant (Fig. 1D,E). Although we could not differentiate horizontal, bipolar, and amacrine cells, there is no relative reduction of inner nuclear layer cells within the mutant, and we consider it likely that these cells are present. Resulting from the lamination defect, it is sometimes not possible to distinguish between inner and outer plexiform layer tissue in the mutant. However, the smaller plexiform islands close to the pigment epithelium, sometimes surrounded by photoreceptor cells, are most likely outer plexiform islands, whereas the larger ones closer to the lens are most likely inner plexifom islands (Fig. 2A).
At 72 hpf (3 days), WT retinal differentiation is basically complete, including synaptogenesis (Fig. 1F; Schmitt and Dowling, 1999). Corresponding to the growth of the plexiform layers in WT, the islands of plexiform layer tissue enclosed within the rosettes have enlarged in the mutant retinae (Fig. 1G,H).
The pacpaR2.10 mutation is lethal at day 5 of development. The eyes of still-alive mutants of this stage are of approximately WT size or slightly smaller but still display a severe lamination defect (Figs. 1J,L,M, 2B).
Our results demonstrate that N-cad is crucial for the development or maintenance of retinal lamination throughout the maturation process of the growing zebrafish eye. The disturbance of neuroepithelial integrity and the loss of cell polarity can be explained by weakened cell adhesion in the pacpaR2.10 mutant and is in accordance with the results presented by Matsunaga et al. (1988): the retinal rosettes of the pacpaR2.10 mutant resemble those caused by antibody blockage of N-cad in chick retinae.
The expected function of N-cad as mediator of cell adhesion in adherens junctions (Tepass et al., 2000) is reminiscent of the postulated role of zebrafish N-cad in the maintenance of neuronal organization within the neural tube (Lele et al., 2002). Indeed, the retinal rosettes formed by inner nuclear layer cells and plexiform layer tissue appear indistinguishable from those formed within brain neuroepithelia (Fig. 1L, arrowheads). Rosette formation also occurs at the same stage in both structures.
Rosette formation within brain tissue upon N-cadherin blockage was previously described by Gänzler-Odenthal and Redies (1998). The application of function blocking antibody to intact developing neuroepithelia causes an invagination of the ventricular layer, resulting in rosettes that contain cells of the ventricular layer plus remnants of ependymal lining.
In the pacpaR2.10 mutant the pigment epithelium (ventricular side of the retina) is also disturbed to some degree: The edges of the pigment epithelium often form cell clusters (Fig. 1H, bottom; Fig. 1J, bottom left) or the edges curl around forming rosette like structures (Figs. 1L, 2D). Sometimes small islands of pigment epithelium cells can be found within other retinal tissue (Fig. 2F). Photoreceptor rosettes appear to form as invagination from the ventricular side (Fig. 1D, green arrow; Fig. 2A,B, green arrows), whereas retinal ganglion cells form clusters that could be interpreted as invaginations from the vitreal side (Fig. 2C). However, most retinal and brain rosettes in the pac mutant have a synapse containing a plexiform island in their center (Figs. 2 A–C,F,G, 3E).
Figure 3. Ultrastructure of the eye region in 5-day-old wild-type (WT; A,D) and the N-cadherin pacpaR2.10 mutant (B,C,E) zebrafish. A: WT photoreceptor cells (example outlined) form membranous discs (MD) in the outer segment, partially surrounded by the pigment epithelium (PE). The inner segment or ellipsoid (EL) contains numerous mitochondria, and the nuclei (N) form the outer nuclear layer. B: The electron photomicrograph depicts a disturbed area of the outer mutant retina, including part of the pigment epithelium. In the mutant, photoreceptor cells are often not properly arranged in parallel like in WT. Photoreceptor orientation is sometimes even inverted, so that the nucleus lies closer to the pigment epithelium than the rest of the cell (cell labeled with N and EL; the MD segment is not in plane with the section). The (outer) plexiform layer (PL) comes close to the pigment epithelium, only separated by one row of nuclei. The mutant pigment epithelium cells are indistinguishable from WT. C: This electron photomicrograph depicts part of a rosette formed by photoreceptors similar to the one shown in Figure 2B. Note that although some photoreceptor cells are deformed (see also leftmost photoreceptor in B), a membranous disc segment and a segment containing mitochondria are formed like in WT. D: WT inner plexiform layer. E: Mutant inner plexiform layer island (of a rosette like shown in Fig. 2F). The number and morphology of synapses (arrows and insets) is indistinguishable between WT and mutant. Original magnifications = ×4,000 in A–C, ×30,000 in D,E, ×56,000 in insets.
Download figure to PowerPoint
Thus, rosette formation in the pacpaR2.10 mutant affects all retinal layers. The rosette architecture described here appears more profound than that of the study of Gänzler-Odenthal and Redies (1998). The explanation for this finding may be that, in the pac mutant, N-cad is absent from the beginning on.
Rosette formation in the retina has also previously been described for mice lacking p56lck (Omri et al., 1998), and as a result of agents inducing retinal damage such as cisplatin and hyperoxia (Penn et al., 1992; Yang et al., 2000). A retinal lamination defect in zebrafish resembling the phenotype of the pacpaR2.10 mutant has been described by Wei and Malicki (2002) for zebrafish mutated in the guanylate kinase-family scaffolding protein nagie oko. An even more severe retinal lamination defect is described for the zebrafish glass onion mutant (Pujic and Malicki, 2001). Together, these results suggest that a complex network of nonredundant activities is necessary to ensure a proper formation and maintenance of retinal lamination and that N-cad is a crucial element of this complex.
Zebrafish N-Cadherin Participates in the Maintenance of Lens Epithelium Integrity
Generally, the pacpaR2.10 mutant eyes are not completely separated from the brain, and the pigment epithelium does not completely surround the eyeball (Fig. 1G,J,L). Because some mutant lenses are located off-center in the eyeball, sections through the middle of such eyes often show no lens at all (Fig. 1L), whereas other sections show the lens with very little retina around (Fig. 1E,M). In contrast to the study by Ferreira-Cornwell et al. (2000), demonstrating the requirement of N-cad for lentoid formation in culture, pacpaR2.10 mutant fish develop lenses in all cases. This finding indicates that in vivo N-cad is not required for lens formation.
Whereas mutant lens fibers always appear normal, the lens epithelial cells round up and appear to loose orientation in some cases (20%, n = 50), resulting in the formation of multilayered cell clusters and a reduced lens size (Fig. 1K). Although this phenotype has only partial penetrance (see Fig. 1H for mutant lens with normal morphology), it points out that N-cad, most probably by means of its mediation of cell adhesion, is involved in the maintenance of lens epithelium integrity. Other cell adhesion molecules, such as NrCAM (Moré et al., 2001) or R-cadherin (cadherin-4; Liu et al., 2001), also expressed in the lens at developmental stages, may partially compensate the N-cad defect in the pacpaR2.10 mutant lens fibers.
N-Cadherin Function Is not Essential for the Formation of Photoreceptor Ultrastructure or for Synaptogenesis
The severely disorganized structure of pacpaR2.10 mutant retinae suggests that cellular ultrastructure and/or connectivity might be correlatively perturbed. To analyze synapse formation and to characterize subcellular fine structure, we performed electron microscopy.
Pigment epithelial cells of pacpaR2.10 mutant retinae contain pigment granules and are indistinguishable from WT (Fig. 3A,B). Unlike WT (Fig. 3A), mutant photoreceptor cells are sometimes inverted in their orientation (Fig. 3B). Sometimes they form rosettes of their own (Figs. 1J, 2B) resulting in a positioning nonadjacent to the pigment epithelium (Fig. 3C). Of interest, mutant photoreceptors contain the typical fine structure observed in WT, including membranous discs for rhodopsin storage and an apical enrichment of mitochondria. However, photoreceptors are sometimes deformed, having broader and shorter membranous discs segments than in WT (Fig. 3B,C). Inner nuclear layer cells and ganglion cells, although often severely misplaced, appear indistinguishable from WT cells at an ultrastructural level at day 5 (data not shown).
Electron microscopic studies have shown that the catenin/cadherin adhesion system is present at the synaptic cleft, bordering the transmitter release zone (Uchida et al., 1996). Cadherin expression in synapses has been studied during development, regeneration, and activity-dependent plasticity (Benson and Tanaka, 1998; Huntley and Benson, 1999; Bozdagi et al., 2000; Shan et al., 2002), and there is evidence that cadherins, including N-cad, may be involved in synaptogenesis and synapse maintenance in the vertebrate central nervous system (Obst-Pernberg and Redies, 1999).
Surprisingly, the inner and outer plexiform layer islands (Fig. 2F,G) in pacpaR2.10 mutant retinae contain approximately as many synapses (per area of plexiform layer tissue in sections) as in the corresponding WT layers (Fig. 3D,E). Synapse morphology, including the formation of postsynaptic densities and synaptic vesicles, shows no abnormalities (Fig. 3D,E, insets). Thus, although N-cad function is essential for retinal lamination, the generation of cellular ultrastructure as well as embryonic and larval synaptogenesis before day 5 is independent from N-cad function. However, it remains to be determined whether the mutant synapses formed are made with correct target specificity: Inoue and Sanes (1997) perturbed laminar specificity in the chick brain by using a plant lectin or antibodies against N-cad, and target specificity errors occurred in N-cad–deficient axons of the Drosophila eye (Lee at al., 2001).
Although retinal pacpaR2.10 mutant cells often appear more rounded, residual tissue coherence prevents cells from drifting apart completely. In addition, we do observe some residual lamination in some mutant retinae, especially concerning the laminar arrangement of pigment epithelium, photoreceptors, and outer plexiform layer (Fig. 1H, left side of the retina). Other cell adhesion molecules (other cadherins or members of the Ig superfamily of adhesion molecules like L1 and NrCAM) are likely involved here. Alternatively, the pacpaR2.10 mutation might not be a full N-cad null, because the mutation results in incorrect splicing of the precursor mRNA (Lele et al., 2002), and residual correct splicing leading to very low N-cad levels undetectable by immunoblot could be imagined. The pacpaR2.10 mutation nevertheless provides excellent in vivo evidence for N-cad function, as described elsewhere for the neural tube (Lele et al., 2002), and as described here for the retina.