Function of Atypical Protein Kinase C λ in Differentiating Photoreceptors Is Required for Proper Lamination of Mouse Retina

Temporal and spatial expression of aPKCλ in developing mouse retina

We first analyzed the expression pattern of aPKCλ in the developing mouse retina. aPKCλ mRNA was expressed broadly, relatively stronger in the outer neuroblastic layer, during embryogenesis of the mouse retina (Fig. 1A). After birth, the expression of aPKCλ mRNA was observed strongly in the inner nuclear layer and the ganglion cell layer, and weakly in the photoreceptor layer (Fig. 1B). The expression pattern of aPKCλ protein was not consistent with that of aPKCλ mRNA. Immunohistochemical analysis using an anti-aPKCλ antibody showed that aPKCλ protein was detected intensely at the apical edge of the embryonic retina during embryogenesis (Fig. 1C). Strong immunoreactivity was observed in the photoreceptor layer and the ganglion cell layer at P9 (Fig. 1D). These expression patterns of aPKCλ during retinogenesis suggest that aPKCλ plays important roles in polarity formation of retinal cells.

Expression of aPKCλ in differentiating photoreceptors is required for proper retinal lamination

To ablate aPKCλ in the photoreceptors, we used a transgenic mouse line expressing Cre under the control of the mouse Crx promoter (Crx-cre) (Nishida et al., 2003). In this transgenic mouse line, we showed previously that Cre-mediated recombination occurred in the postmitotic differentiating photoreceptors and pinealocytes of the pineal gland using CAG-CAT-Z mouse line in which lacZ expression is directed by Cre-mediated recombination (Sakai and Miyazaki, 1997; Nishida et al., 2003). In addition, we confirmed that the recombination in Crx-cre/CAG-CAT-Z mice occurred in photoreceptor-committed cells also at the embryonic stages (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). We further confirmed by immunostaining that Cre recombinase is not significantly expressed in progenitors. The Ki-67 antigen is a nuclear protein expressed in proliferating cells during all active parts of the cell cycle (Schluter et al., 1993). Immunostaining results showed that Cre recombinase was expressed at the outermost layer in the retina after embryonic day 12.5 (E12.5) and was not colocalized with Ki-67-positive progenitors (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Furthermore, β-galactosidase was not colocalized with Ki-67 in the Crx-cre/CAG-CAT-Z retina at E12.5, indicating that the recombination does not occur in the progenitors (supplemental Fig. 3, available at www.jneurosci.org as supplemental material). Next, to examine whether or not Cre recombinase is expressed in ganglion cells, we immunostained E13.5 CKO retina with anti-Cre and anti-Brn3b antibodies and confirmed that these signals do not overlap each other (data not shown). We also found that early-differentiated ganglion cells were properly located and laminar structure is properly organized in the innermost layer of the E15.5 retina; thus, it is unlikely that aPKCλ is ablated in ganglion cells (supplemental Fig. 4, available at www.jneurosci.org as supplemental material). We mated the Crx-cre mouse line with an aPKCλ^flox mouse line in which the aPKCλ allele is flanked by loxP sites (Fig. 2A) (Akimoto, Noda, and Ohno, unpublished observation). We obtained aPKCλ^flox/flox/Crx-cre (aPKCλ CKO) and analyzed phenotypes by comparing them with those of control mice with the genotype aPKCλ^flox/wild/Crx-cre, which showed no abnormal phenotype. The aPKCλ CKO mouse was viable but showed moderate microphthalmia and hydrocephalus. Although the cause of hydrocephalus is not clear, hydrocephalus is considered very unlikely to cause and affect retinal phenotypes. We analyzed details of retinal sections from the CKO and the control mice at various developmental stages. Mouse retinas begin to differentiate at E11.5, and Cre expression begins ∼E12.5 in postmitotic differentiating photoreceptors (supplemental Fig. 2, available at www.jneurosci.org as supplemental material) (Nishida et al., 2003). No obvious difference between the CKO and the control retina was observed at E11.5 (Fig. 2B,C), but slight disorganization of the retina was observed in the CKO retina at E13.5 (Fig. 2D,E), E14.5, P3, and P8 (data not shown). At P14, the retinal laminar structure was fully developed in the control retina, but the lamina of the CKO retina was remarkably disrupted and plexiform patches were observed (Fig. 2 F, G). These results show that ablation of aPKCλ function in postmitotic photoreceptors led to lamination defects not only in the photoreceptor layer but also in the total retina.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

aPKCλ ablation in differentiating photoreceptor cells causes severe retinal lamination defect. A, The Crx-cre transgene construct (top) and the aPKCλ genomic allele flanked with loxP before (middle bottom) and after (bottom) the recombination by Cre recombinase. Exon 5 is designed to be deleted, which corresponds to the pseudosubstrate region of aPKCλ (middle top). B-G, Laminar formation of retina during development in control (B, D, F) and aPKCλ CKO retina (C, E, G). By P14, retinal lamination is complete and distinguishable as RPE, photoreceptor layer composed of OS and outer nuclear layer (ONL), outer plexiform layer (OPL), inner nuclear layer (INL), inner plexiform layer (IPL), and ganglion cell layer (GCL) in control (F). Note that laminar disorganization of the neural retina is observed even at E13.5, whereas Cre expression in differentiating photoreceptors begins at ∼E12.5 (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Scale bars, 100 μm.

Lamination defect does not affect cell fate determination

To determine the distribution of various retinal cell types, we performed immunohistochemical analysis of retinal sections from the CKO and the control mice at P14, using antibodies against retinal cell type markers (Fig. 3A-J). All types of retinal cells were dispersed in the CKO retina (Fig. 3A-J) (data not shown); ganglion (anti-Brn3b antibody), photoreceptor (anti-rhodopsin antibody), amacrine (anti-Pax6 antibody for a nuclear marker; anti-HPC-1 antibody for a cytoplasmic marker), bipolar (anti-Chx10 antibody), and horizontal cells (anti-calbindin antibody), and Müller glia (anti-cyclin D3 antibody for a nuclear marker; anti-S-100β antibody for a cytoplasmic marker). Amacrine (anti-HPC-1 antibody) and Müller glia (anti-S-100β antibody) processes were also detected in patches (Fig. 3A,B,G,H). To examine whether the absence of photoreceptor cell polarity affects cell fate, we harvested and dissociated retinas from both the CKO and the control mice followed by immunostaining with the cell type-specific markers. There was no significant difference in cell numbers of each retinal cell type including the photoreceptor between the aPKCλ CKO and the control retinas (Fig. 3K). Therefore, this indicates that neither defects of photoreceptor cell polarity nor disorganized lamination affect retinal cell fate determination. Furthermore, our observation suggests that a cell-intrinsic program may play a major role in retinal cell fate determination. A similar observation implicating the importance of a cell-intrinsic property in cell fate determination was recently reported using the retinal clonal-density culture system and ganglion cell deprivation (Cayouette et al., 2003; Mu et al., 2005).

aPKCλ is required for proper polarization of photoreceptors in vivo

To investigate the effect of aPKCλ ablation on photoreceptor polarization, we used electron microscopy to examine the morphology of P14 CKO mutant photoreceptors. The mammalian photoreceptor is a highly polarized cell composed of OS, IS, CB, and ST (Fig. 4A). Each of them shows distinguishable morphology. The OS, which is thought to evolve from the ciliated dendrites (Goldstein and Yang, 2000), contains stacks of flattened double lamellas in the form of discs, and the IS contains the large number of long mitochondria (Fig. 4B,D). In the aPKCλ CKO retina, the photoreceptors, which can be distinguished by their cell nuclei with condensed heterochromatin, were randomly distributed in the retina, and no evident OS, IS, or ST structures were observed (Fig. 4C). Although these structures were not formed, the components of the OS, including rhodopsin (Fig. 4F), S-opsin (Fig. 4H), Rom1 (Fig. 4J), and rod photoreceptor CNCG (Fig. 4L) were detected in the aPKCλ CKO retina (Fig. 4E-L) by immunohistochemical analysis. The normal photoreceptor forms a complex synaptic arrangement, designated as a triad. The horizontal and bipolar cell processes invaginate into a photoreceptor terminal at the site of the synaptic ribbon structure (Missotten, 1962) (Fig. 4M,O). At the terminal of aPKCλ CKO photoreceptors, we could not detect any typical triad although synaptic ribbons were observed (Fig. 4N,P). In contrast, many synaptic termini of the retinal neurons other than the photoreceptors were found in the CKO retina by electron microscopy (data not shown). Together, all of these findings suggest that aPKCλ is necessary for proper photoreceptor polarization in vivo, although loss of cell polarity does not affect differentiation and production of photoreceptor components.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Lamination defect does not affect on cell fate determination. A, B, Distribution of rod photoreceptor cells (Rho, rhodopsin; blue), bipolar cells (Chx10, red) and amacrine processes (HPC-1, green) was detected by immunohistochemistry in control (A) and in aPKCλ CKO retina (B) at P14. C-J, Distribution of rod photoreceptors (green) (C-J) and amacrine and ganglion cells (Pax6, red) (C, D), Müller glial cells (cyclin D3, red) (E, F), Müller glial processes (S100-β, red) (G, H), and ganglion cells (Brn3b, red) (I, J) was detected by immunohistochemistry in control (C, E, G, I) and aPKCλ CKO retina (D, F, H, J) at P14. Scale bar, 100 μm. K, The percentage of cell types in control or aPKCλ CKO retina. Note that lamination defect did not affect retinal cell fate determination. Scale bars indicate SD.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

aPKCλ is essential for photoreceptor polarization but not for differentiation. A, Model illustration of control and aPKCλ-null photoreceptors. B-D, Ultrastructure of control (B, D) and aPKCλ-null photoreceptors (C). Arrow indicates ribbon synapse (RS). Scale bar, 1 μm. E-L, Failure of morphogenesis but existence of components of photoreceptors. Expression and distribution of rhodopsin (red) (E, F), S-opsin (red) (G, H), Rom1 (red) (I, J), and CNCG (red) (K, L) in control (E, G, I, K) and aPKCλ-null photoreceptors (F, H, J, L). Insets are higher-magnification images of retina. Nuclei were counterstained with DAPI (blue). Scale bar, 10 μm. M-P, Ultrastructure of synaptic terminals in control (M) and aPKCλ-null photoreceptors (N). Model illustration of synaptic terminal (O, P). Note SR is formed, but there is no penetration of horizontal (H) and bipolar (B) cells into aPKCλ-null photoreceptor (N). N, Nucleus; SR, synaptic ribbon; ONL, outer nuclear layer. Scale bars, 1 μm.

Proper polarizations of differentiating photoreceptors are required for apical junction formation

Ablation of aPKCλ function in postmitotic photoreceptors led to lamination defects not only in the photoreceptor layer but also in the entire retina. To understand the mechanism of severe retinal lamination defects, we analyzed the distribution of photoreceptor-committed cells during retinal development. In the control retina, photoreceptor-committed cells that expressed Crx were observed at the outer neuroblastic layer at E14.5 (Fig. 5A). In contrast, in the E14.5 aPKCλ CKO retina, Crx-expressing cells were already separated from the outermost layer and located randomly in the inner layer (Fig. 5B). aPKCλ, Par3, and Par6 were located at the apical edge of the embryonic control retina, but expression of these molecules was mislocalized in the CKO retina (supplemental Fig. 5A-F, available at www.jneurosci.org as supplemental material). We confirmed by Western blotting that the amount of aPKCλ protein was substantially reduced in the CKO retina, whereas the amounts of Par3 and Par6 were not affected (supplemental Fig. 5G, available at www.jneurosci.org as supplemental material). These data suggest that aPKCλ is required for proper localization of photoreceptor-committed cells at the apical edge of the retina. The aPKC complex is known to localize at the TJs in epithelial cells as well as the AJs in neuroepithelial cells (Manabe et al., 2002). We examined whether or not aPKCλ localizes to the AJ systems in the mouse retina. aPKCλ colocalized with N-cadherin, β-catenin, and afadin in E15.5 retina (supplemental Fig. 6, available at www.jneurosci.org as supplemental material). Zona occludens-1 (ZO-1), a PDZ [PSD-95 (postsynaptic density 95)/discs large/ZO-1] domain protein, localizes both at the TJs and the AJs (Itoh et al., 1993; Izumi et al., 1998; Suzuki et al., 2001). N-cadherin colocalizes with ZO-1 in the apical surface of neuroepithelial cells (Aaku-Saraste et al., 1996). We found that continuous expression of ZO-1, N-cadherin, and β-catenin, which are located at the apical edge of E13.5 control retina, were dispersed in the CKO retina (Fig. 5C-H). We confirmed by Western blotting that the protein amount of N-cadherin and β-catenin was not changed in the CKO retina. We also observed the dispersion of other AJ-associated proteins, nectin and afadin, from the apical edge of the CKO retina (Fig. 5I,J) (data not shown). We then examined the formation of apical junction complex (AJC) in E15.5 retina by electron microscopy. In the control retina, we observed that the AJs were formed between immature photoreceptors harboring distinct segment extension beyond the AJC (Fig. 5K,M). The AJC will become the outer limiting membrane (OLM), and the photoreceptors form AJs at the OLM with microvilli of Müller glial cells, which are not yet differentiated at this stage. All retinal cells other than the photoreceptors locate inner than the OLM; therefore retinal cells that extend their segments beyond the AJC are considered to be immature photoreceptors. In addition, photoreceptors are the only retinal cells that harbor segments. It is interesting that the AJ formation between the photoreceptors are detected, because only AJs between the photoreceptor and the Müller glia have been reported so far. In contrast, in the aPKCλ CKO retina, we observed the AJs neither at the retinal apical edge nor around scattered photoreceptors (Fig. 5L,N) (data not shown). These data therefore suggest that ablation of aPKCλ leads to disruption of the AJs and resulted in the dispersion of the photoreceptors. Thus, our results indicate that the expression of aPKCλ is required for formation of the AJs between the differentiating photoreceptor cells.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Differentiating photoreceptors in aPKCλ CKO retina failed to form adherens junctions. A, B, In situ hybridization of Crx on retinal sections from control (A) and aPKCλ CKO (B) at E14.5. Scale bar, 100 μm. C-J, Immunostaining of adherens junction-related molecules in control (C, E, G, I) and aPKCλ CKO retina (D, F, H, J). ZO-1 (C, D), N-cadherin (E, F), β-catenin (G, H), and afadin (I, J) were detected using respective antibodies (red). Arrows indicate apical edge. Nuclei were counterstained with TOTO-3 iodide (642/660) (blue). Scale bars, 10 μm. K-N, Ultrastructure of apical edge in control and aPKCλ CKO retina. Adherens junction was detected at the apical edge in control (K, M, arrowhead) but not in aPKCλ CKO retina (L, N). PR, Photoreceptor; Mg, melanin granule in retinal pigment epithelium indicated by arrow. Scale bars, 2 μm.

Polarized photoreceptor cells anchor progenitors at the apical edge of neural retina

The absence of the AJs in the aPKCλ CKO photoreceptors may explain why the photoreceptors scatter; however, it does not explain why all types of the retinal cells are dispersed. During retinal development, proliferating progenitor cells are bipolar in shape with a process extending toward the apical surface of the epithelium and another process that terminates at the basal surface (Saito et al., 2003). The somal position of the progenitors varies with cell cycle stage: mitotic progenitors are found near the apical surface adjacent to the developing pigment epithelium. Newly postmitotic cells leave the apical surface, migrate into the proper layers, and form synaptic connections. We investigated the possibility that loss of the AJC leads to abnormal localization of the progenitors followed by differentiation in the CKO retina. At E15.5, mitotic progenitors were localized at the apical edge in the control retina (Fig. 6A). In contrast, the CKO retina showed scattered distribution of mitotic progenitors (Fig. 6B). We also examined the distribution of S-phase retinal progenitors in E15.5 retinas. S-phase progenitors were scattered all over the retina including the apical edge in the mutant retina, whereas they are located in the inner layer in the control retina (supplemental Fig. 7, available at www.jneurosci.org as supplemental material). Thus, in the mutant retina, the abnormal distribution of mitotic progenitors is unlikely to be caused by the disruption of interkinetic nuclear migration by abnormal photoreceptors. We therefore consider that dispersion of the progenitors and after differentiation at the inappropriate location may lead to laminar disorganization.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Progenitors failed to form adherens junctions during mitosis and were distributed randomly throughout retina. A, B, Distribution of progenitors during mitosis in control (A) and aPKCλ CKO retina (B) immunostained with phospho-histone H3 antibody at E15.5. Scale bar, 100 μm. C-F, N-cadherin (red) (C, D) and β-catenin (red) (E, F) were expressed around progenitors (phospho-histone H3, green) both in control (C, E) and aPKCλ CKO retina (D, F). Arrows indicate the apical edge. Scale bars, 10 μm. Insets are higher-magnification images of progenitors. G-J, AJC at electron-microscopic level of E15.5 retina. Immature photoreceptor (PR) and dividing progenitor (P) containing distinct nuclear alignment at the apical edge (G, arrowhead). AJs were formed between progenitors in mitosis and differentiating photoreceptors (I, arrowheads). Apical junction formed at abnormal position with a dividing progenitor (H, J, arrowheads). Centrosomes were directed to AJC (arrow) (J). Scale bars, 2 μm. K, L, AJ-associated actin bundles (phalloidin, green) and centrosomes (γ-tubulin, red) localize apically in the control retina at E15.5 (K). This polarized distribution was disrupted in aPKCλ CKO retina (L). The arrowhead indicates gathering of centrosomes and AJ-associated actin bundles. Scale bar, 10 μm.

To understand the mechanisms of progenitor dispersion in the CKO retina, we examined junction formation in the progenitors. In the control retina, N-cadherin was distributed mainly at the apical surface of the mitotic progenitors as was β-catenin (Fig. 6C,E). In the aPKCλ CKO retina, randomly distributed mitotic progenitors also expressed N-cadherin and β-catenin in characteristic sites (Fig. 6D,F). At the electron-microscopic level, we found that the AJs were formed between the mitotic progenitors, which have distinct nuclear morphologies, and the photoreceptors (Fig. 6G,I). The AJs were not formed at the apical surface of the aPKCλ CKO retina (Fig. 5L,N) but were observed at dislocated progenitors (Fig. 6H,J). Abnormal formation of the AJC between mislocalized progenitors was detected by gathering of centrosomes in the aPKCλ CKO retina by electron microscopy (Fig. 6J). Immunohistochemical staining also showed that ectopic distribution of centrosomes and AJ-associated actin bundles in the CKO retina (Fig. 6L). The centrosomes and the AJ-associated actin bundles serve as reliable indicators of apical polarity of the control retina (Chenn et al., 1998; Wei and Malicki, 2002) (Fig. 6K).

Thus, the progenitors seem to maintain AJ formation ability even in the aPKCλ CKO retina. Therefore, these progenitors may disperse because of lack of the AJ formation ability in unpolarized differentiating photoreceptor cells in the aPKCλ CKO retina. In other words, photoreceptors may anchor progenitors at the apical edge of the embryonic retina.

In vivo role of aPKCλ in photoreceptor polarization

The photoreceptor cells have both the neuronal and the epithelial characteristics. They are generated and differentiated from the neuroepithelium and have axon-dendrite polarity as well as the other neurons in the CNS (Goldstein and Yang, 2000). At the same time, they also have epithelial properties, such as linear arrangement of AJs and apico-basal polarity. In the aPKCλ CKO retina, the OS and IS of the photoreceptors were absent while they expressed components of OS, including rhodopsin, cone opsin, rom1, and CNCG (Fig. 4C,F,H,J,L). Similarly, in the CKO retina, ST of the photoreceptors were also absent while they expressed components of ST such as synaptic ribbons (Fig. 4N). These results indicate that the photoreceptors cannot form OS and ST without aPKCλ function even if they express OS and ST components. Then, is the loss of photoreceptor polarity in the CKO retinas primarily caused by loss of aPKCλ or not? We cannot eliminate both possibilities, but the result that the synaptic termini of retinal neurons other than photoreceptors were formed in the CKO retinas (data not shown) supports the idea that aPKCλ is primarily required at least for synapse formation of the photoreceptors. The failure of formation of OS and IS in the CKO retina is supposed to be primarily caused by lack of aPKCλ as well; however, it awaits future analysis to draw a clear conclusion.

Several downstream targets of aPKCλ have been suggested. The evolutionally conserved Crumbs (Crb) complex is localized to the apical side and required to organize apico-basal polarity in the epithelial cells of the Drosophila embryo (Muller, 2000; Ohno, 2001; Tepass et al., 2001; Knust and Bossinger, 2002; Henrique and Schweisguth, 2003; Roh and Margolis, 2003). Bazooka, a Par3 homolog, interacts with Crb directly, and phosphorylation of Crb by DaPKC is required for epithelial polarity in Drosophila (Sotillos et al., 2004). The mouse Crumbs homolog, Crumbs1 (Crb1), actually localizes to the IS of photoreceptors, and mutations in the human CRB1 are reported to cause retinitis pigmentosa and Leber congenital amaurosis (van Soest et al., 1999; Cremers et al., 2002; den Hollander et al., 2002). However, the mouse Crb1 mutants show only mild retinal phenotypes, discontinuous OLM, and spotty rosettes of photoreceptors (Mehalow et al., 2003; van de Pavert et al., 2004). Therefore, another or unknown Crumbs homolog other than CRB1 are more likely to be involved in inducing photoreceptor polarization as a target of aPKC.

One of the other possible targets of aPKC may be the microtubule-based dynein/kinesin motor complex (Goldstein and Yang, 2000; Nishimura et al., 2004). So far, a kinesin II subunit, KIF3A, is known to be required for normal transportation in photoreceptors (Marszalek et al., 2000). Other kinesins and/or related proteins may contribute to structural establishment of photoreceptors, because significant morphological change of photoreceptors was not observed in the KIF3A mutants (Marszalek et al., 2000).

Interestingly, a recent study shows that glycogen synthase kinase-3β (GSK-3β) regulates neuronal polarity through the phosphorylation of collapsing response mediator protein-2 (CRMP-2) (Yoshimura et al., 2005). aPKC can phosphorylate GSK-3β and inactivate its kinase activity (Etienne-Manneville and Hall, 2003). We detected the expression of CRMP-2 in developing photoreceptors (data not shown). Thus, aPKC may also regulate photoreceptor polarity through the regulation of GSK-3β and CRMP-2.

Role of photoreceptor in retinal lamination

In vertebrates, the neural retina develops from a single-layered neuroepithelium. With maturation, the retina develops to form three major laminas: the photoreceptor layer, the inner nuclear layer, and the ganglion cell layer. In the aPKCλ CKO retina in this study, severe laminar disorganization was observed not only in the photoreceptor layer but also in the entire retina. What is the mechanism that leads to the entire retinal lamination defect although aPKCλ is ablated only in the photoreceptors?

There have been several reports of mouse mutants showing laminar disorganization in mammalian retina, which suggest important roles of retinal pigment epithelium (RPE) and Müller glia for retinal lamination. In the RPE-ablated mice, the retinal lamina is weakly disorganized, but each retinal layer is still distinguishable (Raymond and Jackson, 1995). shh (sonic hedgehog) signaling from retinal ganglion cells plays a role in laminar organization in the mouse retina through the Müller glia, which express Patched-1 (Wang et al., 2002). In the aPKCλ CKO retina, the boundary between the retina and the RPE is partially missing (Fig. 2E). However, we consider that abnormality of the RPE or Müller glia is unlikely to be a primary cause of the lamination defect in the CKO retina because of the following reasons. First, Cre recombinase is expressed in neither the RPE nor Müller glia (supplemental Fig. 2, available at www.jneurosci.org as supplemental material) (Nishida et al., 2003). Second, differentiating photoreceptor cells have already been detached from the apical edge at E14.5, when most of the RPE seems intact (Fig. 5B). Third, Müller glia have not yet developed until E17 at which the lamination defect has already been observed in the CKO retina.

Genetic studies of zebrafish show several alleles that play important roles in epithelial polarity of the retinal neuroepithelium. The retinal phenotypes of several zebrafish mutants, including nok (Wei and Malicki, 2002), glo (Malicki et al., 2003), pac (Masai et al., 2003), ncad (Masai et al., 2003), moe (Jensen et al., 2001; Jensen and Westerfield, 2004), and ome (Malicki et al., 1996; Malicki and Driever, 1999), are similar with those of aPKCλ CKO retina to some extent; laminar arrangement is disorganized and patches of the plexiform matter are scattered. nok and moe are considered to be necessary for TJ formation of RPE and also for the polarity formation of RPE and the neuroepithelial sheet. Glo, pac, and ncad alleles encode N-cadherin and are thought to play a role in neuroepithelial integrity. They also function in the integrity of the inner plexiform layer. We showed in this report that the expression of N-cadherin was dispersed in the aPKCλ CKO retina (Fig. 5F). Therefore, the retinal laminar disorganization of the zebrafish mutants described above may be at least partly attributable to loss of N-cadherin in postmitotic photoreceptor cells. Intriguingly, only the heart and soul (has) mutant, which encodes aPKC, shows a different phenotype from that of mouse aPKCλ CKO retina; plexiform matter is not scattered in patches (Malicki et al., 1996; Horne-Badovinac et al., 2001). The loss of plexiform matter in zebrafish has mutant retina indicate that aPKCλ contributes to process formation of all types of the retinal cells. The complete laminar disorganization observed in these zebrafish mutants has been considered to be attributable to polarity defects in retinal progenitor cells, but we speculate that polarity defects in the photoreceptors may also contribute to these phenotypes.

At the apical side of the photoreceptors, OLM, containing AJs, is formed between the photoreceptors and Müller glia cells. We reported here that AJ formation was also detected between the photoreceptors and the progenitors, in addition to AJ formation between the photoreceptors themselves (Figs. 5, 6). Because the photoreceptors have AJs and the Par3/Par6/aPKC complex locates at the apical edge, we considered that the similar mechanisms proposed for epithelial cells could apply for the retinal integrity. AJ as well as cell adhesion molecules such as N-cadherin and nectin were not detected at the apical edge in the CKO retina (Fig. 5F,L,N) (data not shown). Based on these findings, we propose the following mechanism for the striking lamination defect of the aPKCλ CKO retina (supplemental Fig. 8B, available at www.jneurosci.org as supplemental material). First, disruption of polarity and AJ formation in postmitotic differentiating photoreceptor cells leads to the absence of AJC during retinogenesis. Second, mitotic progenitors then fail to anchor to the apical surface and scatter. Third, progenitors exiting the cell cycle undergo normal differentiation even in the abnormal position. Fourth, irregular AJC formation of progenitors in abnormal positions may lead to rosette formation. Together, we consider that formation and maintenance of AJC between photoreceptor-committed cells and progenitors are required for proper lamination during development. Other retinal cells are unlikely to be involved in anchoring progenitors, because photoreceptors are the only retinal cells that deposit at the outermost layer and are supposed to maintain AJC until E17 when OLM is formed with the terminal process of Müller glia, which functions as a supporting cell.

In this study, we have shown that normal photoreceptor polarization is required for retinal lamination by deprivation of aPKCλ in postmitotic photoreceptors. How is retinal laminar properly organized during retinal development? We have shown that aPKCλ is expressed at the apical edge of the embryonic retina (Fig. 1; supplemental Fig. 9, available at www.jneurosci.org as supplemental material). During development, neural retina is generated from single neuroepithelium sheet, and neurogenesis begins at ∼E11.5. Morphological defects of the retina were detected after Cre recombinase began to be expressed in photoreceptor-committed cells. Because gliogenesis begins at ∼E17, our observations suggest that correct polarization in postmitotic photoreceptors is required for the proper laminar formation of the mammalian retina before gliogenesis begins (supplemental Fig. 8A, available at www.jneurosci.org as supplemental material).