Zebrafish R-cadherin (Cdh4) controls visual system development and differentiation



In zebrafish, R-cadherin (cadherin-4 or Cdh4) is expressed in the retina and in retinorecipient brain regions, suggesting that Cdh4 functions during visual system development. Cdh4 function was examined during retinogenesis and retinal axon outgrowth using antisense morpholino oligonucleotides and mutant Cdh4 construct expression. In knockdowns, Cdh4 was reduced or absent, eyes were small, and retinae lacked discrete laminae. Increased cell death produced the small eye phenotype. Zn5-, Pax6-, and zpr-1–positive cells were reduced or absent in knockdown retinas but, when present, were in the correct laminae. Cdh4 knockdowns had sparse or absent retinal ganglion cell axons. When present, axons projected contralaterally but lacked fine branching and failed to reach the tectum or arborize the entire tectum. Mutant Cdh4 construct expression during retinal ganglion cell differentiation reduced or ablated neurite formation. Cdh4 is necessary for neural retina survival and differentiation, and required for normal retinotectal projection formation and tectal arborization. Developmental Dynamics 233:930–945, 2005. © 2005 Wiley-Liss, Inc.


Neural tissue differentiates from multipotent neuroepithelial cells to form various neuronal and glial cell types. These events occur in a program determined by inductive events and cellular interactions. The neural retina differentiates from a common precursor cell to form a laminated structure that consists of alternating cellular and synaptic layers, with photoreceptors located at the outermost lamina and the retinal ganglion cells located at the innermost lamina. Other cell types, including the bipolar cells, amacrine cells, horizontal cells, and Muller glia, are located in laminae between the photoreceptors and ganglion cells. In the zebrafish, the retinal ganglion cells first become postmitotic at 28 hours postfertilization (hpf) and soon afterward begin to grow axons and dendritic processes (Schmitt and Dowling, 1996). The other retinal cell types differentiate sequentially, and synaptic connections within the retina are established to produce the functional sensory organ. Retinotectal projections initiate and exit the eye together, forming the optic nerve. Ganglion cell axons migrate into the brain and intersect ganglion cell axons from the opposite eye at the midline, then migrate to the optic tectum and arborize within the neuropil (Stuermer, 1988; Burrill and Easter, 1994).

Numerous molecules that control axonogenesis and axon migration have been studied intensively (Yu and Bargmann, 2001; Hatten, 2002; Marin and Rubenstein, 2003). Cadherin cell adhesion molecules were identified as regulators of axonal migration (Clandinin and Zipursky, 2002; Tepass et al., 2002). Cadherins are cell surface glycoproteins that mediate homotypic cell-to-cell adhesion in most tissues (Tepass et al., 2000). Differential gene expression of cadherins and local regulation of cadherin activity allows binding and sorting of different cell populations (Tepass et al., 2000). Cadherins function as a migration substrate controlling developmental morphogenetic processes and neurite formation and neural process migration (Redies, 2000; Tepass et al., 2000; Yagi and Takeichi, 2000). Thus, cadherin adhesion controls many developmental processes, including neural developmental processes, but specific functions of most individual cadherin family members have not been characterized.

Cadherin-4/R-cadherin expression was characterized in chick, mouse, and zebrafish (Inuzuka et al., 1991a, b; Hutton et al., 1993; Matsunami et al., 1993; Tanihara et al., 1994; Liu et al., 1999a, b) and was found in retinal ganglion cells, amacrine cells in the retina, discrete regions of the nervous system (including the major visual targets in the brain), and epithelial cells of organs such as the pancreas and kidney. The mouse cadherin-4 gene was disrupted by gene targeting techniques, producing defects in kidney development, but no defects in neural tissues were described (Dahl et al., 2002). These R-cadherin knockout mice survived to adulthood and were fertile. Also, R-cadherin and cadherin-6 were shown to cooperate in mouse telencephalon neural cell compartmentalization (Inoue et al., 2001).

Analysis of zebrafish Cdh4 expression showed that this cadherin is expressed when the first retinal ganglion cells are born in the retina and expression proceeds in a wave of neurogenesis around the retina; Cdh4 expression also coincides with neurogenesis in brain neuromeres; and Cdh4 expression proceeds from the ventrolateral optic tectum to the dorsomedial tectum, anticipating the arrival of retinal ganglion cell axons and synaptogenesis (Liu et al., 1999a, b, 2001a). In this study, morpholino oligonucleotide (MO) knockdown technology and mutant cadherin expression were used to examine the function of Cdh4 in the zebrafish visual system.


Morphology of cdh4 Morphants

Due to the strong correlation of Cdh4 expression with neurogenesis in the retina and brain and with the trajectory of the retinotectal projection, we hypothesized that Cdh4 participates in retina formation, retinotectal axon migration, and tectal recognition events. To directly examine the function of cdh4 in visual system development, MOs (see Experimental Procedures section) were designed to inhibit Cdh4 translation. These oligonucleotides were tested and found to effectively inhibit cdh4 activity. Two antisense cdh4 MOs produced the same phenotype, and the affected structures correlate directly with the expression pattern of cdh4/Cdh4 in zebrafish (Liu et al., 1999a, b, 2001a, b). In normal 48 hpf zebrafish embryos, retinal lamina are distinct. The inner plexiform layer is clearly seen in living embryos using differential interference contrast (DIC) microscopy (Figs. 1B, 2A). Also, the lens is large and relatively transparent. The cdh4 morphants at this stage could be categorized as slightly, moderately, or severely affected. The most severe effects were small eyes, small and opaque lenses, lack of retinal lamination, brain disorganization, and short body length (Figs. 1G,H, 2D). Moderately affected 48 hpf embryos show somewhat smaller eyes, smaller lenses, evidence of brain disorganization, somewhat shorter body length, and the main distinguishing criteria for moderately affected embryos is that there is some but very little detectable inner plexiform layer formation (Figs. 1E,F, 2B,C). The slightly affected embryos (Fig. 1C,D) were largely indistinguishable from control embryos by DIC microscopy, but the inner plexiform layers were detectably thinner and disorganized.

Figure 1.

Cdh4 knockdown affects eye size, body length, and brain morphology. A–H: Zebrafish embryos were injected (1–2 nl at 0.5 mM) with either standard control morpholino oligonucleotide (MO, A,B) or cdh4 MO (C–H). These injections produced a hypomorphic series of effects. At 2 days postfertilization (dpf), living embryos were imaged using a stereo-dissecting microscope (A,C,E,G) or by differential interference contrast microscopy (B,D,F,H). A–H: The cdh4 morphants (C–H) were micro-ophthalmic, had small lenses, lacked retinal lamination, had perturbed brain organization and shorter body length compared with controls (A,B). D,F,H: The fourth ventricle (IV) was enlarged and the hindbrain (hb) was reduced in cdh4 morphants.

Figure 2.

Eyes of cdh4 morphants are small and lack discrete laminae at 2 days postfertilization (dpf). A–K: Zebrafish embryos were injected with either standard control morpholino oligonucleotide (MO, A,E,H) or cdh4 MO (B–D,F,G,I–K). A–D: At 2 days postfertilization (dpf), living embryos were imaged using differential interference contrast microscopy. Ventral views of eyes are shown (A–D). A: Rostral is up and lateral is left. In eyes of control embryos, all the major layers characteristic of the adult retina, the retinal ganglion cell layer (rgl), inner plexiform layer (ipl), the inner nuclear layer (inl), outer plexiform layer (opl) and the outer nuclear layer (onl) are visible. B–D: In cdh4 MO-injected embryos, a graded series of effects on eye development were seen. B,C: In moderately affected embryos, eyes were smaller in size, the inner plexiform layer was sparse, the distance between the inner plexiform layer and the lens was not uniform, and the outer nuclear layer was sparse or absent. D: Both the eyes and lens of severely affected embryos were small and laminae were altogether absent. E–G: BODIPY–ceramide labeled living 2 dpf embryos were imaged using confocal microscopy. E: In control embryos at 2 dpf, there was a distinct continuous inner plexiform layer separating the retinal ganglion cell layer and the inner nuclear layer. F: In moderately affected 2 dpf cdh4 MO knockdowns, the inner plexiform layer was sparse and irregular and the outer nuclear layer was absent. G: In severely affected 2 dpf cdh4 MO knockdowns, lamination was absent. H–K: Toluidine blue-stained plastic sections of 2 dpf embryos clearly show the laminar organization of cells within the control retina (H). I: In slightly affected cdh4 knockdowns, laminae are present, but cell shape, particularly within the outer nuclear layer differs from that of controls. J,K: In severe cdh4 knockdowns, laminae are absent. Scale bars = 50 μm.

These different categories of morphant phenotypes represent a hypomorphic series that likely results from variable levels of Cdh4 suppression, and the severe effects likely represent a null phenotype. This finding was supported by immunoblotting experiments that showed Cdh4 expression levels in cdh4 MO-injected embryos correlated with the severity of morphant phenotype (Fig. 3E). Within the retina, Cdh4 is normally expressed in the retinal ganglion cell layer, the inner nuclear layer, and the optic nerve (Fig. 3A; Liu et al., 1999a, 2001a). We found that Cdh4 expression was specifically suppressed in the brain (data not shown) and in all retinal layers in cdh4 morphants (Fig. 3B–D; background fluorescence from this antiserum was still detected in the epidermis and lens, which cannot be competed with excess immunizing peptide; Liu et al., 2001b). Together, results from cdh4 MO injection experiments indicate that the morphant phenotype is specific by the following criteria: independent MOs produce the same morphant phenotype; the affected structures in morphants are known to express cdh4; and, protein levels were reduced in cdh4 morphants, which correlated with the severity of phenotype in a hypomorphic series.

Figure 3.

Reduction of Cdh4 protein levels correlates with phenotypic severity of cdh4 morphants. A–D: Zebrafish embryos were injected with either standard control morpholino oligonucleotide (MO, A) or cdh4 MO (B–D). A: Confocal imaging of 3 days postfertilization (dpf) Cdh4 immunostained cryosections showed Cdh4 localization in the retinal ganglion cell layer, in the amacrine cell layer of the inner nuclear layer, in the lens, and in the optic nerve of the control embryo. B,C: In slightly and moderately affected cdh4 MO-injected embryos, staining of the retinal ganglion cell layer, the amacrine cell layer and the optic nerve was reduced. D: In severely affected embryos, Cdh4 staining was not detected in the neural retina and reduced in the lens (the apparent immunoreactivity of the ectoderm and lens is nonspecific, that is, it could not be competed with excess immunizing peptide; Liu et al., 2001b). E: Corresponding immunoblots of individual embryos of various phenotypic severities in the hypomorphic series showed that, compared with control embryos at 3 dpf (con), Cdh4 expression was reduced in slightly and moderately affected embryos (sli, mod) and absent in severely affected embryos (sev).

To examine effects of Cdh4 knockdown on retinal histology in more detail, living 48 hpf embryos were labeled using the fluorescent lipid BODIPY–ceramide and plastic sections were stained with toluidine blue (Fig. 2E–K). Using BODIPY–ceramide labeling, retinal laminae were visualized by laser scanning confocal microscopy. In normal zebrafish, retinal ganglion cell layer, inner plexiform layer (including the two synaptic sublaminae), inner nuclear layer, outer nuclear layer, and cellular structure in the lens could be distinguished (Fig. 2E). In moderate cdh4 morphants, retinas were thinner and disorganized, but there was some evidence of inner plexiform layer formation (Fig. 2F). In contrast, severe cdh4 morphants showed little or no evidence of lamination, and the retina was thin with the appearance of an undifferentiated neuroepithelium (Fig. 2G). Histological sections of control and cdh4 morphants support the findings from DIC microscopy and BODIPY–ceramide staining (Fig. 2H–K). Indeed, increasing phenotype severity correlates with evidence for reduced differentiation. The most severe cdh4 morphants show no evidence of differentiated cells, only an undifferentiated neuroepithelium (Fig. 2K).

Cdh4 Knockdown Disrupts Retinal Differentiation

The retinal morphology in Cdh4 knockdown embryos indicated that this tissue failed to differentiate normally, and retinal disorganization in Cdh4 knockdowns suggested that retinal cell types may have failed to segregate into distinct laminae. To distinguish between these different possibilities, control and cdh4 MO-injected embryos were examined by immunofluorescence staining using differentiation markers. Acetylated tubulin, a marker for neuronal processes, was used to label 48 hpf control and cdh4 morphant embryos. Whole-mount labeled embryos were visualized using two-photon microscopy. Ventral views of optical sections that included the eye and the optic nerve regions were projected; low-magnification views in Figure 4A,B are shown for orientation. In normal retinas from 48 hpf embryos, extensive neural process formation was found throughout the retina, and retinal ganglion cell axon projections form an optic nerve that exits the eye en route to the brain (Fig. 4C). In severe cdh4 morphants, neuronal processes within the retina were sometimes entirely absent (data not shown), and when present, only sparse retinal ganglion cell projections were detected (Fig. 4D). Normally, the optic nerve exits the eye and projects dorsally and forms a chiasm with the optic nerve from the opposite eye before proceeding to a ventrolateral region of the contralateral optic tectum or other visual targets along this route (Fig. 4E; Stuermer, 1988; Burrill and Easter, 1994). In cdh4 morphants with a few retinal ganglion cell axon projections, these axons similarly exit the eye and form a chiasm (Fig. 4F). Acetylated tubulin staining indicates that there was a reduction in the number of retinal ganglion cell axon projections that may be a consequence of an actual disruption in retinal ganglion cell differentiation.

Figure 4.

cdh4 knockdown results in sparse retinal ganglion cell axon outgrowth at 2 days postfertilization (dpf). A–F: Zebrafish embryos were injected with either standard control morpholino oligonucleotide (MO, A,C,E) or cdh4 MO (B,D,F). Embryos were fixed at 2 dpf, and whole-mount preparations with axon tracts immunostained using acetylated tubulin antibodies were visualized using two-photon imaging. Ventral view projections are shown of the planes containing the eyes and optic nerves. A,B: Low-magnification views are shown for orientation. A: In control embryos, axons exit the retinal ganglion cell layer to form the optic nerve, which then bends to travel more dorsally and rostrally, crosses the midline at the optic chiasm, then projects straight toward the contralateral optic tectum. B: In severe cdh4 knockdowns, both the retinal ganglion cell layer and optic nerve are sparse and the bend in the optic nerve as it travels toward the optic chiasm is absent, but the sparse optic nerve does project normally across the midline toward the contralateral optic tectum. C,D: In high-resolution views of the control eye, extensive neural processes of all the major retinal layers are evident (C), whereas in the cdh4 knockdown eye, neuronal processes are largely absent and only a sparse disorganized retinal ganglion cell layer is visible (D). Projection images of two-photon image volumes are shown. E,F: High-resolution views of the optic chiasm show the thick well fasciculated optic nerve of the control (E) and the thin loosely fasciculated optic nerve of the cdh4 knockdown (F). Scale bars = 50 μm.

To test whether Cdh4 knockdown reduces the number of retinal ganglion cells produced during retinal histogenesis, control and knockdown embryos were stained using zn5 monoclonal antibody that detects DM-GRASP/neurolin present on retinal ganglion cells and other neurons (Laessing and Stuermer, 1996; Fashena and Westerfield, 1999). In normal embryos, a layer of retinal ganglion cells of uniform thickness and a thick optic nerve were labeled (Fig. 5A). In cdh4 morphants, a hypomorphic series of defects was detected in zn5-labeled retinas. In the most mildly affected embryos with a slightly thinner inner plexiform layer, the retinal ganglion cell layer was slightly thicker than that of control embryos (Fig. 5B). Other moderately and severely affected cdh4 morphants (Fig. 5C,D) showed a progressive thinning of the retinal ganglion cell layer with an irregular border and a reduction in the thickness of the optic nerve. In some severe morphants, no detectable zn5 staining was present (data not shown). In all cases where the zn5-labeled cells were present, the retinal ganglion cells were confined to a lamina at the innermost portion of the retina and the optic nerve exited the eye in a distinct fascicle, like that of control embryos.

Figure 5.

cdh4 knockdown affects differentiation of the retina. A–T: Zebrafish embryos were injected with either standard control morpholino oligonucleotide (MO, A,E,I,M,Q) or cdh4 MO (B–D,F–H,J–L,N–P,R–T). Retinal ganglion cells were visualized at 2 days postfertilization (dpf), using confocal imaging of zn5-immunostained whole-mount embryos. Images are ventral view projections at the level of the optic nerve. Zn5 identifies retinal ganglion cells and their axon processes, among other cells in the developing zebrafish. A: In control embryos, the retinal ganglion cell layer was uniform in thickness and had smooth boundaries. B: The retinal ganglion cell layer was often expanded in cdh4 MO-injected embryos that were only slightly affected. C: In moderately affected embryos, the retinal ganglion cell layer was sparse with irregular boundaries and the optic nerve was thin. D: In severely affected cdh4 knockdowns (D), the retinal ganglion cell layer ranged from nearly absent (D), with only one or two zn5-positive cells and their projections, to altogether absent (not shown). E–T: For Pax6 and zpr-1 immunolocalization, embryos were fixed at 3 dpf (E–L) or at 4 dpf (M–T) and 8 μm horizontal cryosections were cut. Adjacent sections of the same eye (except H, L) were immunostained for Pax6-positive amacrine cells (E–H,M–P) and zpr-1–positive double cone photoreceptors (I–L,Q–T) and visualized by confocal microscopy. cdh4 knockdown images are arranged from left to right as a graded series of slightly affected (F,J,N,R), moderately affected (G,K,O,S), then severely affected (H,L,P,T) embryos. F,H,J,L,N–P,R–T: Amacrine cells and photoreceptor cells fail to differentiate in severe knockdowns (H,L,P,T), but, when present in less affected knockdowns (F,J,N,O,R,S), they were localized in appropriate retinal laminae. Scale bar = 50 μm.

Cdh4 is also expressed in amacrine cells (Fig. 3A; Liu et al., 1999b). Pax6 was used as an amacrine cell marker to determine whether Cdh4 knockdown disrupts amacrine cell differentiation or retinal tissue organization. Pax6 staining of amacrine cells in control 3 and 4 days postfertilization (dpf) embryos shows that amacrine cells were confined to the inner nuclear layer (Fig. 5E,M). In slightly and moderately affected cdh4 morphants, Pax6-positive amacrine cells were present but reduced in number, and these cells were confined to the appropriate lamina, the inner nuclear layer (Fig. 5F–G,N–O). Pax6-positive amacrine cells were not detected in severe cdh4 morphant embryos at 3 or 4 dpf (Fig. 5H,P). These data show that Cdh4 knockdown prevents normal differentiation of the retinal cell types that normally express this cadherin.

To examine whether tissue organization or differentiation of other retinal cell types was affected by Cdh4 knockdown, a double cone photoreceptor-specific marker, zpr-1 (Larison and Bremiller, 1990), was used to label retinas in control and cdh4 morphant embryos. Slight and moderate morphants displayed reduced numbers of double cone photoreceptors compared with control embryos at the same stage (Fig. 5I,Q). This effect was more pronounced in 3 dpf embryos than in 4 dpf embryos (Fig. 5J–K,R–S). In severe morphants, there was little or no detectable zpr-1 staining in both 3 and 4 dpf embryos (Fig. 5L,T). Together with histological data (Fig. 2H–K), these data indicate that Cdh4 knockdown affects the differentiation of retinal cell types that express this cadherin and those that do not. However, when these retinal cell types were present in the cdh4 morphants, they were found in the correct retinal lamina.

Differentiated retinal cell types are clearly reduced in cdh4 morphants, but this effect could be a consequence of reduced specification of retinal precursors rather than differentiation per se. crx and otx5 are transcription factors that are among the earliest known markers of photoreceptor identity. crx promotes both specification and differentiation of retinal neurons, and crx and otx5 positively regulate one another in the zebrafish retina (Shen and Raymond, 2004). In early retinal development, crx is expressed in all proliferating retinal progenitors, but at 3 dpf, crx expression is confined to the developing photoreceptors (both rods and cones) and to the outer part of the inner nuclear layer (Liu et al., 2001c; Shen and Raymond, 2004). Expression of crx can be used as a marker of photoreceptor specification and differentiation. In 78 hpf cdh4 morphants, expression of both crx and otx5 were present, but overall expression reduced (Fig. 6B,C,E,F, page 3), which may be due to cell death in the retina (see below). These data indicate that retinal precursors are specified, and loss of differentiated retinal cell types is a consequence of reduced differentiation as a consequence of Cdh4 knockdown.

Figure 6.

Expression of early markers of photoreceptor identity is reduced in cdh4 morphants. A–F: Zebrafish embryos were injected with either standard control morpholino oligonucleotide (MO, A,D) or cdh4 MO (B,C,E,F). crx (A–C) and otx5 (D–F) labeling in 78 hours postfertilization (hpf), embryos were used as markers of photoreceptor specification and differentiation. A,D: In controls, crx and otx5 labeling were confined to the photoreceptor layer and the outermost part of the inner nuclear layer. B,E: Some reduction of staining was seen in slightly affected cdh4 morphants. C,F: In severe cdh4 morphants, both crx and otx5 staining were reduced and lacked defined laminae.

As an additional test of retinal lamination, mitotic nuclei were localized using histone H3 antibodies (Wei et al., 1999) in 2 dpf embryos, which are normally confined to the basal lamina of the developing retina (Fig. 7A). In all cdh4 morphants, regardless of the severity, mitotic nuclei were still confined to the basal lamina (Fig. 7B–D).

Figure 7.

Microphthalmia is due to increased cell death in cdh4 knockdowns. A–D,F–M: Zebrafish embryos were injected with either standard control morpholino oligonucleotide (MO, A,F,H,J,L) or cdh4 MO (B–D,G,K,I,M). Mitotic cells were visualized in eyes of 48 hours postfertilization (hpf) embryos using histone-H3 immunostaining (control, n = 31; slight, n = 16; moderate, n = 15; severe n = 13). A–D: In both control (A) and in cdh4 knockdowns (B–D), dividing cells were confined to the peripheral lamina of the retina. E: Comparison of the number of H3-positive nuclei per μm2 showed no significant difference between controls and cdh4 knockdowns. Statistical comparisons of these data were made in two ways. A two-tailed unpaired Students t-test was used to compare controls and pooled data from all cdh4 morphant groups (P = 0.07). One-way analysis of variance was used to compare controls and unpooled cdh4 morphant groups (P = 0.08). E: Chart shows unpooled group data. Error bars = 1 SD. A–D: Representative confocal images are shown of control (A) and cdh4 knockdowns from the slightly affected (B), moderately affected (C), and severely affected (D) cdh4 morphant groups. Dying cells were visualized using acridine orange (AO) vital staining at 36 hpf (control, n = 9; cdh4 MO, n = 14) and at 50 hpf (control, n = 5; cdh4 MO, n = 6). F–M: Representative differential interference contrast images (F–I) and a corresponding fluorescent image (J–M) are shown for each group; control at 36 hpf (F,J), cdh4 MO injected at 36 hpf (G,K), control at 50 hpf (H,L), and cdh4 MO at 50 hpf (I,M). AO-labeled cells were counted. Cell death was increased in the retina of cdh4 knockdowns compared with controls at both 36 hpf (P < 0.01) and 50 hpf (P < 0.01). Randomly selected embryos were used for AO staining to minimize potential bias. Cell death was increased in the lens at 50 hpf (P < 0.05) but not at 36 hpf. Error bars = 1 SD.

Cell Death in Cdh4 Knockdowns Causes the Small Eye Phenotype

Mitotic nuclei were counted in 2 dpf control and Cdh4 knockdown embryos, and the number of nuclei per unit area was compared to determine whether a reduction in cell division was responsible for small eye phenotype seen in the morphants. No statistical difference was detected between control and knockdown eyes (Fig. 7E; statistical data found in figure legend). Thus, no evidence for an effect of Cdh4 knockdown on mitotic rate was detected.

Next, we examine apoptosis in control and Cdh4 knockdown embryos. Previous studies described increased apoptosis at 36 hpf in the retina and lens (Cole and Ross, 2001). In Cdh4 knockdown embryos, a dramatic increase in apoptosis over basal levels was detected in 36 hpf retina but not the lens (Fig. 7F–G,J–K,N). This increase in apoptosis continued in the 50 hpf retina in Cdh4 knockdowns compared with control embryos (Fig. 7H–I,L–N). Also, the Cdh4 knockdown lens displayed an increased number of apoptotic cells at 50 hpf compared with control embryos (Fig. 7H–I,L–N). Increased apoptosis is responsible for the reduced size of the retina and lens seen in Cdh4 knockdown embryos.

Defective Retinotectal Projection in Cdh4 Knockdown Embryos

To visualize the retinal ganglion cell projection to their primary optic target, the optic tectum, control and cdh4 morphant embryos were fixed and eyes were filled by pressure injection with DiI (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate), which labels retinal ganglion cell axons by anterograde diffusion within the cell membrane (Godement et al., 1987). In normal 3–5 dpf embryos, the retinal ganglion cell axons exclusively project to the contralateral tectum and extensively arborize the tectal neuropil (Stuermer, 1988; Burrill and Easter, 1994; Fig. 8A). In moderate cdh4 morphants, the retinal ganglion cell axons that were present projected exclusively to the contralateral tectum and no other defects in retinal ganglion cell axon pathfinding were detected (Fig. 8B,C). However, in these moderate cdh4 morphant embryos, the retinotectal projections did not completely arborize the tectum and were generally confined to the ventrolateral region of the tectum.

Figure 8.

Retinal ganglion cell axons in 5 days postfertilization (dpf) cdh4 knockdowns project to the contralateral optic tectum but do not arborize normally. A–L: Zebrafish embryos were injected with either standard control morpholino oligonucleotide (MO, A,D,G,J) or cdh4 MO (B,C,E,F,H,I,K,L). Embryos were fixed at 5 dpf and retinotectal projections were traced by injecting one eye in each embryo with the lipophilic dye DiI (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate). Projection images of two-photon volumes of the DiI-filled retinotectal projections (shown in red) were overlaid with differential interference contrast (DIC) images from the same embryo (A–C). Staining within the injected eye was digitally removed to leave the DIC image of the eye visible. A–C: In controls (A), retinal ganglion cells project contralaterally and arborize to fill the optic tectum. In cdh4 knockdowns (B,C), retinal ganglion cells project to the contralateral optic tectum but do not arborize normally. D–L: These same three fish are shown at higher resolution. A–C: The DiI-injected fish were counterstained with DAPI (4′,6-diamidine-2-phenylidole-dihydrochloride) so that the boundaries of the tectal neuropil could be seen in relationship to the retinotectal projections (shown in gold). DAPI (shown blue–violet) stains the cell nuclei. Cellular regions are filled with nuclei, but nuclei are absent from regions of neuropil such as the optic tectum (ot). D,G,J: In controls (D,G,J) from a dorsal view (D,G), the retinotectal projections arborized to fill the entire optic tectum. J: A medial view of a sagittal cut through the optic tectum of the control volume shows the deep tectal layers. E,F,H,I,K,L: In cdh4 knockdowns, both retinotectal projections and the optic tectum were affected. E,H: Optic tecta in some moderately affected cdh4 knockdowns were apparently normal in size, but the retinotectal projections were sparse, disorganized, and did not fill the tectal neuropil. K: A medial view of a sagittal cut through the optic tectum of the moderately affected cdh4 knockdown lacks the well defined deep layers. F,I,L: In severe knockdowns, the tectal neuropil was smaller (F,I) and some neurites extended beyond the neuropil boundary and into the cellular layer (arrowheads I,L). Scale bars = 50 μm.

A more detailed view of the tectal arborization pattern was obtained by imaging 3 dpf embryos using two-photon microscopy and advanced visualization techniques. A series of optical section comprising the complete volume of the tectal arborization was collected, and these images were rendered to create a three-dimensional volume that could be viewed at any orientation using voxel rendering software, Voxx (Clendenon et al., 2002). Inspection of these volumes showed that control retinotectal arborizations were extensive and finely branched (Fig. 9A–D; see Supplementary Movie S1, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat). In moderate cdh4 morphant embryos, retinotectal projections were less-extensively arborized and more fasciculated and processes showed numerous varicosities (Fig. 9E–H; see Supplementary Movie S2). In the most severe cdh4 morphant embryos, no retinal ganglion cells were present; therefore, there were not projections to observe. However, some severe cdh4 morphants had at least a few axons, and their retinotectal projections extended to the contralateral tectum but failed to arborize with the tectum (Fig. 9I–L).

Figure 9.

Retinal ganglion cells in 3 days postfertilization (dpf) cdh4 knockdowns arborize only the lateral optic tectum or terminate before reaching the tectum. A–L: Zebrafish embryos were injected with either standard control morpholino oligonucleotide (MO, A–D) or cdh4 MO (E–L). Embryos were fixed at 3 dpf, and retinotectal projections were traced by injecting eyes with the lipophilic dye DiI (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate). In both controls and cdh4 knockdowns, retinal ganglion cell axons project toward the contralateral optic tectum. In controls, upon reaching the optic tectum, retinal ganglion cells projections arborized extensively within the tectal neuropil. I–L: In severely affected cdh4 knockdowns, the optic nerve was sparse and failed to reach the optic tectum. E–H: In moderate cdh4 knockdowns, the optic nerve reached the optic tectum, but arborized only the more lateral region. In knockdowns, the nerve fibers were blunt and the fine branching seen in controls was absent. Multiple renderings are shown of each two-photon volume. A,E,I: Dorsal views of the optic tectum, with rostral down and lateral to the left. B,F,J: In the next view, the volumes have been rotated so that we are looking at a rostral view of the fish with dorsal up and lateral to the left. C,D,G,H,K,L: The upright volume renderings of B,F,J have been rotated counterclockwise.

To examine the retinotectal arborization pattern within the tectal neuropil, 5 dpf embryos were double labeled by injecting DiI into the eye and the cell nuclei were counterstained using DAPI (4′,6-diamidine-2-phenylidole-dihydrochloride; Fig. 8D–K). The entire neuropil region of these embryos was visualized using two-photon microscopy, and projections of the image volumes are presented. In control embryos, the retinotectal projections completely arborize the neuropil and are confined within this compartment (Fig. 8D,G). Also, a majority of retinal ganglion cell axons project into surface layers of the tectum but also into deeper layers of the neuropil in control embryos (Fig. 8J). Retinotectal projections in moderate Cdh4 knockdown embryos did not fill the entire space of the tectal neuropil (Fig. 8E,H). In some cases, the tectal neuropil of the moderate Cdh4 knockdown embryos was smaller (Fig. 8F,I). In a few cases, retinotectal axon projections extended beyond the tectal neuropil, though the cellular layer of the optic tectum, to the ependymal surface (Fig. 8C,I,L). Together, our data show that Cdh4 is necessary for normal tectal arborization.

Cytoplasmic Domain Deleted Cdh4 Mutant Inhibits Neuronal Process Formation and Extension

To study Cdh4 function in early generation of retinal ganglion cell neuronal processes, a cytoplasmic domain deleted dominant-negative Cdh4 construct (Cdh4ΔC) was injected into embryos along with control enhanced green fluorescent protein (eGFP) DNA at the one- to eight-cell stage. In control experiments, either the eGFP construct (eGFP) or a myc-tagged eGFP construct (eGFP/MT) was injected. The Cdh4 construct and control constructs (eGFP and eGFP/MT) were inserted in the HSP70/4 vector, allowing control of transfected genes using heat shock (Shoji et al., 1998).

The first retinal ganglion cells are born and become postmitotic around 28 hpf and extend neuronal processes soon afterward (Burrill and Easter, 1995; Schmitt and Dowling, 1996). These events coincide with the onset of cdh4/Cdh4 expression (Liu et al., 1999a, 2001b). Because the expression of exogenous gene products takes approximately 2–4 hr after heat shock (Shoji et al., 1998), Cdh4ΔC dominant-negative protein in embryos heat shocked at 24–26 hpf was first expressed at the time when the retinal ganglion cells begin to differentiate. Normally by 48 hpf, a majority of retinal ganglion cells have extensive dendrites projecting to the inner plexiform layer and axons that reach the anterior optic tectum (Stuermer, 1988; Burrill and Easter, 1995). Heat-shock treatment at 46–48 hpf allowed us to test whether activation of the Cdh4ΔC construct later during retinal ganglion cell development had any effect on the morphology of Cdh4ΔC-expressing cells.

Survival rates were similar between Cdh4ΔC-injected embryos and those embryos injected with the control constructs (Table 1). Embryos that expressed eGFP without the heat shock, and embryos that showed moderate to severe deformities in their bodies and/or tails (Table 1) were excluded from analysis. Injections of plasmid DNAs into early embryos (one to eight cells) result in random integration of the injected DNA into early blastomeres, producing a mosaic expression pattern later in development (Hyatt and Ekker, 1999). Expressing cells were detected using antibodies against either the myc-tag or against the eGFP, and these expressing cells were found throughout the embryonic structures including the brain, eyes, and muscles (Fig. 10). Approximately one third of the surviving embryos had cells in the eye that expressed the exogenous proteins, with approximately ten labeled cells detected in the retina. Most of the labeling was detected in the trunk muscle fibers (Fig. 10B). Anti-myc staining in Cdh4ΔC-expressing muscle fibers was examined because they are large and showed that the Cdh4ΔC protein was concentrated at the cell peripheries, suggesting that the mutant Cdh4 protein was localized to the cell membrane and available to interact with endogenous Cdh4 molecules (Fig. 10B). However, a previous study indicates that cadherin proteins lacking cytoplasmic domain sequences are inefficiently transported to the cell surface (Chen et al., 1999). The mechanism of action for cytoplasmic domain deleted mutant cadherin is not known.

Table 1. Summary of Plasmid DNA Injections
ConstructsInjected (n)SurvivalExpressed eGFP without HSDeformed
  1. HS, heat shocked; n, number of embryos; eGFP, enhanced green fluorescent protein; hpf, hours postfertilization.

HS at 26–28 hpf    
eGFP172109 (63.4%)9 (8.3%)13 (11.9%)
MT/eGFP8769 (79.3%)4 (5.8%)10 (14.5%)
Cdh4ΔC/eGFP361242 (67.0%)18 (7.4%)31 (12.8%)
HS at 46–48 hpf    
Cdh4ΔC/eGFP9771 (73.2%)3 (4.2%)10 (14.1%)
Figure 10.

Retinal ganglion cells (RGCs) expressing Cdh4ΔC at 24–26 hours postfertilization (hpf) have few or no processes at 56 hpf. A,B,D,F,H: Photomicrographs from retinal sections. C,E,G: Photomicrographs from whole-mount embryos. A,B: Shown are Cdh4ΔC-expressing cells in the retina (A) and in trunk skeletal muscle fibers (B) that were heat shocked at 25–26 hpf. A: A cross-section of the eye (dorsal to the right and lateral above) showing Cdh4ΔC-expressing cells throughout the thickness of the retina. B: A cross-section (dorsal above and lateral to the right) of the body trunk region showing that the anti-myc immunostaining was located mainly in the periphery of each muscle fibers (mf). C,D,F: Control RGCs expressing either MT/eGFP (C,D) or eGFP (F). These RGCs have numerous dendrites (arrow) and each has an axon (arrowhead). A,G,H: RGCs expressing the dominant-negative Cdh4ΔC (heat shocked at 25–26 hpf) have many fewer or no processes. E: A Cdh4ΔC-expressing RGC (heat shocked at 48 hpf) with normal RGC morphology. gcl, retinal ganglion cell layer; hHS, hours heat shock; inl, inner nuclear layer; onl, outer nuclear layer; sc, spinal cord; eGFP, enhanced green fluorescent protein. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Retinal cells expressing the control proteins (eGFP or eGFP/MT) after heat shock at 24–26 hpf were found throughout the retina at 48 hpf, and these cells expressing control proteins showed normal morphologies, including oval-shaped or elongated somata and extensive dendrites (Fig. 10C,D,F). Retinal cells expressing the Cdh4ΔC construct after heat shock at 24–26 hpf were detected throughout the retinal laminae of the embryonic zebrafish (Fig. 10A), suggesting that laminar migration was not affected in these cells (Riehl et al., 1996). But unlike the controls, Cdh4ΔC-expressing retinal ganglion cells displayed a rounded soma shape and had significantly fewer processes (Figs. 10A,G,H, 11). An axon was found in the vast majority of control construct expressing retinal ganglion cells (Fig. 11A), but only approximately one third of Cdh4ΔC-expressing retinal ganglion cells had an axon (Fig. 11A). Control construct expressing retinal ganglion cells had two to six dendrites with a total length of approximately 18 microns (Fig. 11B,C). In contrast, an average of approximately one dendrite per Cdh4ΔC-expressing retinal ganglion cell was significantly lower than control cells (Fig. 11B), and the average total length of their dendrites was significantly shorter than those of the control retinal ganglion cells (Fig. 11C). All control cells contained dendrites and/or an axon, but approximately one third of the Cdh4ΔC-expressing retinal ganglion cells had no axon or dendrites (Fig. 11D).

Figure 11.

Effects of Cdh4ΔC on the generation of retinal ganglion cell (RGC) axons. A–E: Cdh4ΔC expression affected the generation of RGC axons (A), number of dendrites per RGC (B), total length of dendrites (C), generation of neuronal processes (D), and size of RGCs (E). A: Of RGCs transfected with MT/eGFP, 93% had axons (n = 42); with Cdh4ΔC heat shocked at late times (46–48 hpf; Cdh4ΔCL), 100% had axons (n = 41); and with Cdh4ΔC heat shocked at early times (25–26 hpf; Cdh4ΔCE), 33.8% had axons (n = 73). B: The number of dendrites per RGC were significantly reduced in Cdh4ΔCE RGCs (1.2 ±1.1) compared with both controls (4.1 ± 0.7, P < 0.01) and Cdh4ΔCL RGCs (3.7 ± 0.9, P < 0.05). C: Total length of dendrites was significantly reduced in Cdh4ΔCE RGCs (3.1± 3.5 μm) compared with both controls (17.6 ± 3.9 μm, P < 0.001) and Cdh4ΔCL RGCs (15.3 ± 3.6 μm, P < 0.001). D: All control RGCs and late heat-shocked RGCs had at least one process, whereas 28.4% of Cdh4ΔCE RGCs had no processes. E: RGC circumference of Cdh4ΔCE transfected RGCs (15.1 ± 2.3 μm) was unchanged relative to either Cdh4ΔCL RGCs (16.1 ± 2.6 μm) or controls (15.7 ± 1.7 μm). eGFP, enhanced green fluorescent protein.

Activation of Cdh4ΔC at 46–48 hpf produced Cdh4ΔC-expressing retinal ganglion cells that had similar morphologies as control retinal ganglion cells (Fig. 10E), and these retinal ganglion cells each had an axon and several dendrites with a total number and length being similar to control construct expressing retinal ganglion cells (Fig. 11).


The consequences of blocking cdh4 gene activity in zebrafish were studied within the visual system, producing a microphthalmia phenotype with reduced differentiation of retinal cell types. The cell types affected in the retina included those that express Cdh4 (ganglion and amacrine cells) and those that do not express Cdh4, like photoreceptors. In the less severe examples of the morphant phenotype, a reduced number of retinal ganglion cells project axons along a relatively normal trajectory, to the optic tectum; they make correct pathfinding decisions at the midline; but these axons do not arborize the tectum normally. In addition, inhibiting Cdh4 function by expressing a cytoplasmic domain deleted version of Cdh4 (Cdh4ΔC) in retinal ganglion cells showed that normal Cdh4 activity is needed within an individual cell for the neurite extension, indicating that the need for Cdh4 during neurite extension is cell autonomous.

Mouse knockout of the R-cadherin gene displayed defects in the kidney, but these defects did not have functional consequences. Effects on neural structure, and in particular, the visual system in R-cadherin–deficient mice were not reported in this study. Perhaps detailed examination in the nervous system would reveal similar defects as we observed in cdh4 zebrafish morphants. Alternatively, there may be redundant or compensatory functions provided by other cadherin genes in mice that makes R-cadherin function more dispensable.

Comparison Between Effects of Cdh4 and Cdh2 Deficiency on Visual System Development

Retina development proceeds from the formation of the optic cup through subsequent growth, differentiation, and morphogenesis of the functional sensory organ. Various mutations have been identified that affect retina development in zebrafish. Different categories of mutations affecting retinal development were identified (Pujic and Malicki, 2004). One group of mutations affect cell polarity within the undifferentiated neural epithelium, which affects the organization of the retina, causing retinal cells to mix within various layers of the retina, rather than being confined to the appropriate lamina. N-cadherin gene (cdh2 or ncad) is defective in the parachute and glass onion mutations. These mutations cause retinal lamination defects and were categorized in this first group of genes (which also includes heart and soul, which encodes an atypical protein kinase C isoform).

A second group of mutations affects the neurogenic wave of retina differentiation, where undifferentiated neuroepithelial cells undergo a program of differentiation steps to form the various retinal cell types. An example of a gene that controls this neurogenic wave of retina differentiation is the sonic you (syu) gene that encodes Sonic Hedgehog (Shh; Neumann and Nuesslein-Volhard, 2000). In this study, R-cadherin gene (cdh4) activity was inhibited using MOs, producing a phenotype that resembles this second class of mutations that affects the neurogenic wave of retina differentiation.

Like the syu phenotype, the most severe phenotype caused by morpholino knockdown of the cdh4 gene was little or no neural differentiation in the retina. In syu mutants, approximately half the embryos have undifferentiated retinas, and the other embryos show reduced levels of differentiation; Retinal cells that differentiate in syu mutant eyes were confined to appropriate retinal laminae (Schauerte et al., 1998; Stenkamp et al., 2000; Shkumatava et al., 2004). In cdh4 morphants, there was a hypomorphic series of phenotypes with progressively less retinal differentiation, but the retinal cells that form were confined to the appropriated laminae. Shh expression is activated in a ventronasal region of the developing retina at 28–30 hpf (Neumann and Nuesslein-Volhard, 2000). Cdh4 expression also initiates in the ventronasal retina region, and expression proceeds in a pattern that corresponds to the neurogenic wave controlled by hedgehog expression (Liu et al., 1999b).

Ganglion cell axon retinotectal projections in zebrafish traverse a well-characterized route to the contralateral tectum (Stuermer, 1988; Burrill and Easter, 1994). In cdh2/ncad mutant embryos, retinotectal projections show pathfinding defects at the brain midline, and many retinal ganglion cell axons project to and arborize the ipsilateral optical tectum (Masai et al., 2003). In cdh4 MO-injected embryos, retinal ganglion cells project axons to the contralateral tectum, similar to normal embryos, but once these axons reach the optic tectum, they fail to properly arborize within the neuropil. Therefore, different phenotypes were found for cdh2/ncad and cdh4 deficiency with respect to retinal development and the retinotectal projection formation. The relationship between individual cadherins and specific pathfinding and axon migration regulatory molecules unclear and will require additional experimentation.

Extracellular domain deleted cadherin constructs were previously used to perturb cadherin function (Levine et al., 1994). Although the mechanism of action for the cytoplasmic domain deleted cadherin constructs is not fully understood, these dominant-negative constructs have been shown to block function of a specific cadherin family member, possibly by competing with the endogenous cadherin for homotypic extracellular domain interactions (Levine et al., 1994). Other studies suggest that cadherin cytoplasmic domain deletion would severely limit its transport to the cell surface (Chen et al., 1999), which would limit its ability to compete by binding endogenous cadherin molecules. Here, we find that cytoplasmic domain deleted zebrafish Cdh4 construct inhibited neurite formation and extension. Notably, extracellular domain deleted Cdh2 construct (NcadΔE) expression in Xenopus affected neurite formation in retinal ganglion cells (Riehl et al., 1996), but extracellular domain deleted cadherin constructs affect all cadherins expressed by a given cell. The effects of the Cdh4ΔC construct on the development of the retinal ganglion cell processes support our findings using MOs to knockdown Cdh4 expression.

It is conceivable that the Cdh4ΔC may interfere with Cdh2 function, in addition to Cdh4 function because Cdh4 and Cdh2 can form heterotypic interactions in vivo (Shan et al., 2000), but homotypic cadherin interactions are stronger (Steinberg and Takeichi, 1994). Also, we noted that there was a strong correlation between the Cdh4 expression pattern (Liu et al., 1999a, b) and the cells affected by Cdh4ΔC. If Cdh4ΔC also affected Cdh2-expressing cells, then we would expect more widespread effects, because this cadherin is far more broadly expressed (Bitzur et al., 1994), supporting the idea that Cdh4ΔC specifically inhibits Cdh4 function. This construct allowed us to examine the effect of blocking Cdh4 function in individual cells or small groups of cells within an otherwise normal retina, and Cdh4ΔC was found to inhibit neurite outgrowth. This finding indicates that Cdh4 function during neurite formation and extension is cell autonomous.


Numerous studies have implicated cadherin function in neural development (Tepass et al., 2000; Yagi and Takeichi, 2000; Pujic and Malicki, 2004), but only limited information is available about functional roles for individual cadherin molecules that control morphogenesis of neural tissues. In this study, Cdh4 function was examined in the developing visual system, and we found that Cdh4 controls retinal cell differentiation, retinal ganglion cell neurite formation, and retinotectal axon migration. Cdh4 function complements the previously described role of Cdh2 in the developing visual system, and together, these cadherins (and probably additional cadherins) coordinate morphogenesis of this complicated sensory system.



Zebrafish (Danio rerio) were raised and kept under standard laboratory conditions (Westerfield, 2000) in accordance with Indiana University and University of Akron policies on animal care and use. For some experiments, 0.2 mM phenylthiourea (PTU) was added to prevent melanization.

5′-Rapid Amplification of cDNA Ends of Zebrafish cdh4 cDNA

Published cdh4 cDNA sequences are incomplete at the 5′-end (Liu et al., 1999b), lacking sequences encoding N-terminal amino acid sequences. To identify translation initiation sequences for cdh4 required for morpholino oligonucleotide design, 5′-rapid amplification of cDNA ends (5′-RACE) was performed. Using sequence encoding the C-terminal regions of the zebrafish Cdh4 (Liu et al., 1999a), standard methods for 5′RACE (kit used according to manufacturer's instructions, Roche Molecular Biochemicals, Indianapolis, IN) were used to obtain a 1.6-kb DNA fragment containing 126 bp of 5′-untranslated region, and sequences encoding the signal sequence, presequence, extracellular domains 1 and 2 (EC1, 2), and the remaining portion of EC3 that was missing previously from the original sequence (Liu et al., 1999b). Similar to other classic type I cadherins, there was the HAV sequence in the EC1 of Cdh4. There was also a conserved endoproteolytic cleavage site (RRQKR) within the presequence of Cdh4, immediately before the EC1, which was identical to that of chick and mouse Cdh4 and was thought to be required for proper processing of proteins (Ozawa and Kemler, 1990). The first 19 amino acids in EC1 of zebrafish Cdh4 were identical to those of chick, mouse, and human Cdh4. The deduced amino acid sequence from EC1 to EC3 of zebrafish Cdh4 also shared high homologies with the chick (76%) and mouse Cdh4 (75%), and less homology (63%) with zebrafish Cdh2. These values were similar to those obtained by comparing the C-terminal half among these molecules (Liu et al., 1999b). Sequence for this 5′-RACE cDNA clone was deposited in GenBank (accession no. DQ018999).

MO Injection

Translation blocking morpholino antisense oligonucleotides (MOs; RcadMphA 5′-AAG GAG GCA GAT GTT TGT TAT TCA C-3′, RcadMphB 5′-TTC CTG TGA GAT GTG CTG TCG GTA G-3′, and standard control 5′-CCT CTT ACC TCA GTT ACA ATT TAT A-3′), purchased from Gene Tools (Philomath, OR), were used as described (Nasevicius and Ekker, 2000). MOs were designed according to Gene Tools targeting guidelines. MO sequences were compared with databases using BLAST, and no significant similarities were found to any sequences other than zebrafish cdh4. MOs were injected into one- to eight-cell stage embryos at 4.2 μg/μL (0.5 mM) in Daneau buffer (58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO4, 0.6 mM Ca(NO3)2, 5.0 mM HEPES pH 7.6). Total volume injected was 1–2 nl (4.2–8.4 ng) per embryo.

BODIPY–Ceramide labeling

BODIPY–ceramide-Fl-C5 (Molecular Probes) was dissolved in dimethyl sulfoxide (DMSO) to a stock concentration of 5 mM. Dechorionated living embryos were soaked in 100 μM BODIPY–ceramide in Embryo Medium with 10 mM HEPES for 2 hr in the dark (Cooper et al., 1999). The embryos were washed and confocal images were acquired using a Zeiss LSM-510 confocal microscope.

DiI Injections

For dye injections, zebrafish larvae, age 3–5 dpf, were fixed using 4% paraformaldehyde in phosphate buffered saline (PBS) overnight at 4°C, rinsed with PBS, then embedded in 1% low melting temperature agarose. Whole eye fills were done by pressure injection of DiI (V-22888, Molecular Probes, Eugene, OR). Some embryos were counterstained with DAPI (Molecular Probes) after DiI injection. Image volumes were acquired using a Bio-Rad MRC1024 laser scanning confocal system (Bio-Rad, Hercules, CA) equipped with a Titanium–Sapphire laser for two-photon illumination mounted on a Nikon inverted microscope. An excitation wavelength of 760 nm was used. These two-photon image volumes were rendered using Voxx, a voxel based three-dimensional near real time rendering program, developed at the Indiana Center for Biological Microscopy (Clendenon et al., 2002; http://www.nephrology.iupui.edu/imaging/voxx/index.htm).

Acridine Orange Staining

To detect apoptosis, live embryos were removed from their chorions and incubated for 2 min in Embryo Medium (Westerfield, 2000) containing 5 μg/ml of acridine orange (A-3568, Molecular Probes). After several rinses in Embryo Medium, DIC and epifluorescence images were collected using a Nikon Diaphot microscope (Nikon, Inc., Melville, NY) equipped with DIC optics and a SPOT RT camera (Diagnostic Instruments).


For immunolabeling, embryos were fixed in 4% paraformaldehyde in PBS overnight at 4°C. Embryos that were whole-mount immunolabeled were incubated for one hour in blocking solution (0.5% Triton X-100, 5% goat serum, 0.2% bovine serum albumin, 1% DMSO, 0.05 M NH4Cl, 0.025 M glycine, 0.025 M L-lysine HCl in PBS). All rinses were in PBS with 0.5% Triton X-100. Whole-mounts were incubated in primary antibody in block overnight at 4°C. Acetylated-tubulin antibody (T-6793, Sigma, St. Louis, MO) was used at 1:1,000 dilution and Zn-5 antibody (Zebrafish Resource Center, Eugene, OR) was used at 1:200 dilution. Image volumes were acquired using a Bio-Rad MRC1024 laser scanning confocal system (Bio-Rad) equipped with a Titanium–Sapphire laser for two-photon illumination mounted on a Nikon inverted microscope. Two-photon image volumes were rendered using Voxx (Clendenon et al., 2002).

For section immunolabeling, eight micron cryosections were cut horizontally through the midsection of the eyes (Barthel and Raymond, 1990). Slides were incubated 1 hr at room temperature in blocking solution (10% normal goat serum, 0.5% Triton X-100, 0.05 M NH4Cl, 0.025 M glycine, 0.025 M L-lysine HCl in PBS). Sections were incubated overnight in primary antibody diluted in a solution containing 2% normal goat serum, 0.5% Triton X-100, in PBS. Primary antibodies used were Pax6 antibody (AB5409, Chemicon International, Temecula, CA) at 1:50 dilution, phosphohistone 3 antibody (Upstate Biotechnology) at 1:200 dilution, Cdh4 affinity purified antipeptide antibody (Liu et al., 2001b) at 1:200 dilution, Zpr-1 (Zebrafish Rescource Center) at 1:200 dilution and anti-myc antibody (Sigma) at 1:600 dilution. Texas Red-conjugated anti-rabbit or anti-mouse secondary antibodies made in goat (Jackson Immunoresearch Labs, West Grove, PA) were used at 1:100 or mouse (Vector Laboratories, Burlingame, CA). Confocal images were acquired using a Zeiss LSM-510 microscope. Metamorph (Universal Imaging Corp, Downingtown, PA) and LSM Image Browser (Carl Zeiss, Inc., Thornwood, NY) were used to obtain single plane and projection images from confocal image stacks.

In Situ Hybridization

Whole-mount in situ hybridization of zebrafish embryos was performed as described (Liu et al., 1999b). Digoxigenin-labeled riboprobes for crx (Liu et al., 2001c) and otx5 (Gamse et al., 2002) were synthesized from cDNA as run-off transcripts from linearized templates by using the Genius System DIG RNA Labeling Kit (Roche, Indianapolis, IN).

Cdh4 Construct Generation

Reverse transcriptase-polymerase chain reaction (RT-PCR) was performed using total RNA isolated from 48 hpf zebrafish embryos to obtain a DNA fragment containing the cdh4 signal sequence, presequence, extracellular and transmembrane domains, but lacking the cytoplasmic domain (SalI site: 5′-GTCGACGAATAACAAACATCTGCCTCC-3′; 3′ end with AgeI site: 5′-ACCGGTTCTTTCTCCCTTCGCTTCACC-3′). This fragment was cloned into the SalI and AgeI sites of the HSP70/4 eGFP vector (gift of P. A. Raymond, University of Michigan), which had the eGFP region excised. A myc tag sequence from the CS2+MT vector (Turner and Weintraub, 1994) was then inserted, with 3′ to, and in-frame with, the cdh4 fragment in the HSP70/4 vector. The myc tag insert was obtained using PCR with specific primers with NotI restriction sites flanking both sides (5′ end: 5′-GCGGCCGCTCCCATCGATTTAAAGCTATG-3′; 3′ end: 5′-GCGGCCGCTCACTATAGTTCTAGAGGC-3′). To generate control construct, PCR was performed using the same primers as the above for the myc tag sequence, except that the restriction sites were changed to SalI (5′ end) and AgeI (3′ end). This DNA fragment was inserted into the SalI and AgeI sites in the HSP70/4eGFP. This control construct was called MT/eGFP. Construction was verified using restriction mapping and DNA sequencing.

Mutant Cadherin Expression Experiments

Embryos (one- to two-cell stage) were placed in agarose troughs. Plasmid DNA was pressure injected (1–2 nl, 50–60 ng/μl in sterilized water; phenol red was added to the injection solution as a tracer) into blastomeres or into yolk immediately below the blastomeres. The cadherin construct DNA was mixed at a ratio of 2:1 (w/w) with HSP70/4 eGFP plasmid DNA. Cdh4ΔC construct was coinjected with eGFP construct to identify embryos that did not have tight control by the heat shock promoter. Embryos expressing moderate to high levels of eGFP before the heat-shock treatment were discarded. To verify that the Cdh4ΔC construct and the eGFP construct were expressed in the same cells, eGFP construct was coinjected with Cdh4ΔC construct into one- to four-cell stage embryos; of 348 eGFP-labeled retinal cells, 317 (91.1%) were also labeled with the anti-myc antibody, showing that an overwhelming majority of retinal cells coexpressed the two proteins. Control injections of eGFP or MT/eGFP were also performed.

Injected embryos were incubated at 28.5°C. Immediately before heat shock, embryos were examined using epifluorescence microscopy to sort out any embryo that expressed eGFP, which may also express the cadherin mutant proteins without the heat shock. For heat shock, beakers containing embryos were then placed in a 38.5°C water bath for 1 hr. After heat shock, embryos were returned to the 28.5°C water baths and allowed to develop to 54–56 hpf.

Procedures for tissue processing were described in detail previously (Liu et al., 1999a, b, 2001a). Briefly, embryos were anesthetized in 0.02% methane tricaine sulfonate (Sigma), fixed in 4% paraformaldehyde, cryoprotected in 20% sucrose, and sectioned (25–30 μm). To prepare tissues for whole-mount immunocytochemistry, embryos were rinsed in 70% methanol, followed by 100% methanol, stored at −20°C in 100% methanol until use.

eGFP-tagged constructs were detected using anti-eGFP (Medical & Biological Laboratories Co., Ltd., Nagoya, Japan), and myc-tagged constructs were detected using an anti-myc antibody (Sigma). Labeled retinal cells with somata positioned within the retinal ganglion cell layer were considered retinal ganglion cells, and their dendritic processes originated from the basal side (adjacent to the inner plexiform layer) and/or from the basal lateral regions, whereas their axons originated from the vitreal surface (adjacent to the lens) of the cell and/or projected toward the optic nerve head.

The shape and size of the retinal ganglion cell soma, the presence of the axon, the number and length of the dendrites (no distinction was made between primary branches and their secondary ones) were recorded. Camera lucida drawings of the labeled cells under Nomarski optics were digitized and quantified using NIH Image software (Scion Corporation, Frederick, MD). Measurements for mutant cadherin-expressing or eGFP/MT-expressing retinal ganglion cells were made from whole-mount embryos. The morphology of eGFP-expressing retinal ganglion cells was examined on tissue sections. All data were collected only from well-labeled retinal ganglion cells whose processes (or lack of processes) could be identified with confidence.


The 5′RACE was performed with the guidance and support of Dr. Pamela Raymond, while Qin Liu was a postdoctoral fellow in her laboratory at the University of Michigan. Dr. Pamela A. Raymond generously provided the HSP70/4 eGFP vector. Zn-5 antibody was obtained from the Zebrafish Resource Center, which is supported by grant P40 RR12546 from the NIH-NCRR. J.A.M. and Q.L. were funded by the National Institutes of Health and S.G.B. received a postdoctoral fellowship from Prevent Blindness America/Fight For Sight.