Divergent polarization mechanisms during vertebrate epithelial development mediated by the Crumbs complex protein Nagie oko

Cell polarity and maintenance of epithelial integrity are tightly linked to many developmental processes including the establishment of transepithelial diffusion barriers within gut and epidermis, as well as the morphogenesis of complex organs including the heart, retina and neural tube. Many studies have implicated proteins containing PDZ domains in cell polarization (Sheng and Sala, 2001; Bilder et al., 2003; Tanentzapf and Tepass, 2003; Hurd et al., 2003). The membrane-associated guanylate kinase (MAGUK) proteins represent an important class of these proteins and are characterized by PDZ, SH3 and the non-catalytical GUK domains. Several members of this family also contain a 4.1 or HOOK domain, which functions in intramolecular interactions between SH3 and GUK domains (Nix et al., 2000; Tavares et al., 2001) as well as binding to other members of the protein 4.1 superfamily (Lue et al., 1994; Cohen et al., 1998). The multi-domain organization of MAGUK scaffolding proteins facilitates the assembly of alternative protein complexes with temporal or tissue specificity. One well-studied example is the Drosophila discs large protein (Dlg) (Hough et al., 1997; Mendoza et al., 2003; Bachmann et al., 2004). Currently the functional relevance of different protein-protein interactions in vivo remains to be determined for many other MAGUK proteins.

Zebrafish Nagie oko (Nok) is a member of the conserved Stardust (Sdt)/protein associated with Lin7 subfamily member-1 (Pals1)/Mpp5 family of MAGUK proteins (Tepass and Knust, 1993; Kamberov et al., 2000; Hong et al., 2001; Bachmann et al., 2001). In addition to PDZ, SH3, 4.1 and GUK domains, Nok contains the evolutionarily conserved region 1 (ECR1) domain and a bipartite L27 domain (Fig. 1). Several of these domains have been shown to mediate direct binding to other cell-polarity proteins. Association with the transmembrane protein Crumbs (Crb) at the subapical region of Drosophila embryonic epithelia or at tight junctions of vertebrate epithelia is mediated through the PDZ domains of Sdt and Pals1 proteins (Bachmann et al., 2001; Hong et al., 2001; Roh et al., 2002; Hurd et al., 2003) (Fig. 1). The more N-terminal L27 domain (L27N) has been shown to mediate binding to Patj, another protein that associates with murine Pals1 at the tight junction (Roh et al., 2002), whereas the more C-terminal L27 (L27C) domain mediates binding to Lin-7 (Doerks et al., 2000). The Pals1 ECR1 domain is required for interaction with the PDZ domain of Par6 (Wang et al., 2004). However, interaction between the Pals1 ECR1 domain with Par6 is not essential for the maintenance of epithelial polarity within mammalian Madin Darby canine kidney (MDCK) epithelial cells (Wang et al., 2007).

In zebrafish, the nok, heart and soul (has)/prkci, and oko meduzy (ome)/crb2a mutations affect maintenance of cell polarity and epithelial integrity within the neural retina and the retinal pigmented epithelium (RPE) (Malicki and Driever, 1999; Wei and Malicki, 2002; Horne-Badovinac et al., 2001; Peterson et al., 2001; Omori and Malicki, 2006). In addition, nok and has/prkci regulate myocardial cell polarity, adhesion and cell-shape changes, which contribute to heart morphogenesis (Yelon et al., 1999; Horne-Badovinac et al., 2001; Peterson et al., 2001; Trinh et al., 2005; Rohr et al., 2006; Rohr et al., 2008). Phenotypic similarities between these mutants support the genetic and biochemical evidence from other systems for a direct interaction between the Crb-Nok and Par6-aPKC protein complexes.

Here, we functionally test the role of specific protein-protein interaction domains of Nok during zebrafish embryonic development. Tissue-specific differences in the ability of different Nok deletion proteins to rescue epithelial polarity defects suggest varying compositions and changing protein-protein interactions between components of the Crb-Nok and the Par6-aPKC complex during the embryonic development of zebrafish.

The complex epithelial phenotypes of nok mutants prompted us to systematically test the requirement of individual Nok protein-protein interaction domains in mediating the correct polarization and maintenance of different developing epithelia, including the squamous monolayered RPE surrounding the neural retina, the myocardium and the neural retina. To this end, we generated a series of deletion constructs that encode truncated and Myc-tagged versions of the wild-type (WT) protein (Fig. 1A; see Materials and Methods). Genetic rescue experiments were performed in the nok^s305 or nok^m520 null mutant backgrounds, or in nok morpholino antisense oligonucleotide (MO^nok) injected embryos that were co-injected with MO^p53 to suppress the major off-target effects of morpholinos mediated through p53 activation (Nasevicius and Ekker, 2000; Langheinrich et al., 2002; Robu et al., 2007). The nok^s305 or nok^m520 mutations cause a complete loss or a C-terminal truncation of the GUK domain, respectively, and only severely reduced levels of mutant protein can be detected (Wei and Malicki, 2002; Rohr et al., 2006) (Xiangyun Wei, University of Pittsburgh, PA, personal communication). To control for efficient expression of Nok deletion proteins at 32 hours post fertilization (hpf), protein levels were detected by western blotting of embryonic extracts and probing with an anti-Myc antibody (Fig. 1B).

To compare the phenotypic consequences of expressing different Nok deletion proteins in nok^s305 or nok^m520 mutants compared with their expression in nok morphants, we analyzed the different mutant/morphant phenotypes and could not detect differences (see below). Therefore, neither nok^s305 nor nok^m520 provide rescue activity in the context of complementation assays when injected with mRNA encoding Nok deletions.

Consistent with a function in apical-basal cell polarity, complete loss of nok causes a severe curvature of the body and disruption of central portions of the RPE which leaves most of the retina without pigmented cells (Fig. 2D). Injection of His-Myc-tagged nok^wt mRNA into nok^m520 mutant embryos complemented the severely disrupted and patchy RPE phenotype (n=31/31 rescued embryos) and curved body form (n=24/31 rescued embryos; Fig. 2B,H). For evaluating the efficacy of the RPE rescue, we only defined embryos as rescued in which the retina was entirely covered by pigment cells as observed in the WT.

Heart tube elongation in nok mutants and morphants is impaired, in accordance with a disruption of myocardial cell cohesion and a failure of cardiomyocytes to correctly expand in size (Fig. 3B,G) (Rohr et al., 2006). To visualize the rescue efficiency of Nok^wt protein on heart development, we introduced a transgene that expresses green fluorescent protein (GFP) under the control of the cmlc2 promoter [Tg(cmlc2:GFP)] (Huang et al., 2003) into the nok^m520 mutant background. Injection of mRNA encoding His-Myc-tagged Nok^wt could completely rescue heart tube elongation in a subset of mutant embryos (n=6/31 rescued embryos). The nok^m520 mutant genotype was verified by PCR (see Materials and Methods). Similarly, co-injection of mRNA encoding His-Myc-tagged Nok^wt together with MO^nok resulted in the rescue of heart tube elongation as assessed using the atrium-specific antibody S46 (n=14/39 rescued embryos; Fig. 3C,G).

To assess the organization and polarity of the neural retina and to visualize apical junctions at 34 hpf, we performed immunohistochemical stainings of retinal tissue using an antibody against ZO-1 (Fig. 4) and Rhodamine phalloidin, which has a high affinity for submembranous F-actin (not shown). The neural retinal tissue was counterstained with an antibody against Nok to monitor expression of His-Myc-tagged Nok protein. In comparison to the nok^s305 mutant phenotype (Fig. 4A), embryos injected with His-Myc-tagged nok^wt mRNA showed no ectopic ZO-1-positive bundles within the neural retina, and the integrity of the tissue was maintained (n=6/6 embryos rescued). In addition, no ectopic F-actin staining was seen. Therefore, similarly to the rescue of the RPE phenotype and myocardial development, Nok^wt protein completely rescues the neural retinal polarity and integrity defects (Fig. 4B).

To evaluate the relevance of different protein-protein interaction domains of Nok for maintenance of epithelial polarity and cardiac morphogenesis, we first tested whether the interaction domain of Nok that is required for association with Par6 is essential. Interactions between murine Pals1 and the scaffolding factor Par6 are mediated via the ECR1 domain (Wang et al., 2004). We thus tested the functionality of the corresponding Nok^ΔPar6 mutant, which lacks the ECR1 domain, in complementing different epithelial defects. Whereas the His-Myc-tagged Nok deletion protein significantly complemented the nok^m520 mutant RPE and body curvature phenotypes (Fig. 2C,H), it failed to complement nok^m520 mutant (n=0/26 embryos rescued) and nok morphant heart tube elongation defects (n=0/122 embryos rescued; Fig. 3D,G), as well as neural retinal organization and polarity defects (n=0/6 embryos rescued; Fig. 4D). This finding suggests that Nok does not require Par6 as a direct scaffolding partner within the RPE or for correct body form, whereas direct protein-protein interactions between Nok and Par6 are required for the maintenance of polarity within the myocardium and the neural retina.

We next tested the requirement of the Nok interaction domain that is predicted to mediate association with Patj or Lin7 in the polarization and maintenance of zebrafish embryonic epithelia. The murine Pals1 N-terminal bipartite L27 domains (L27N and L27C) mediate binding to Patj or Lin-7, respectively (Doerks et al., 2000; Roh et al., 2002). In zebrafish, there are three Lin-7 homologs (Wei et al., 2006), but only a single Patj homolog (N.B.-A. and S.A.-S., unpublished results), none of which has been characterized functionally. We performed a functional complementation analysis by overexpressing His-Myc-tagged Nok^ΔPatjΔLin7 or Nok^ΔPatj in nok mutants and morphants. We found that loss of the L27 domain does not affect RPE integrity (Fig. 2E,H; and data not shown) or early heart morphogenesis (Fig. 3E,G; and data not shown). By contrast, the L27 domains that may mediate Nok binding to Patj or Lin-7 appeared to be essential for correct body form (Fig. 2E,H) and for neural retinal polarity and integrity (n=0/6 embryos rescued; Fig. 4C).

Association of Drosophila Sdt and murine Pals1 with the intracellular tail of the transmembrane protein Crb is mediated via their PDZ domains (Bachmann et al., 2001; Hong et al., 2001; Roh et al., 2002). We assayed the functional importance of the Nok PDZ domain by testing whether nok^ΔCrb mRNA, which encodes a protein lacking the PDZ domain, could complement nok mutants or morphants. Although the truncated His-Myc-tagged protein could be detected on western blot (Fig. 1B), as well as on tissue sections of the neural retina (Fig. 4F), none of the mutant phenotypes was rescued (Fig 2F,H, Fig. 3F,G; for the retina phenotype n=0/6 embryos were rescued, Fig. 4F). Therefore, the PDZ-domain-mediated association of Nok with partner proteins is essential for the correct polarity and integrity of all epithelia analyzed.

Similarly, injection of nok^ΔC mutant mRNA encoding a His-Myc-tagged C-terminal truncation of Nok that deletes the GUK, HOOK and parts of the SH3 domain, failed to rescue any of the phenotypes assayed, although Nok^ΔC protein could be detected on western blot and stained tissues (Fig. 1B, Fig. 2G,H; for the retina phenotype n=0/6 embryos were rescued, Fig. 4E; and data not shown). This finding confirms that C-terminal truncation renders Nok inactive. Together, these results demonstrate that different embryonic epithelia utilize divergent Nok-mediated mechanisms for tissue polarization. Whereas correct polarization of the neural retina requires each of the Nok protein-protein interaction domains, myocardial polarity is not dependent on the two Nok L27 interaction domains that are thought to mediate association with Patj or Lin7. Moreover, integrity of the RPE does not require the Nok ECR1 domain, which is thought to mediate the association with Par6 or the two Nok L27 domains that mediate association with Patj or Lin7. Similarly, correct body form is not dependent on the expected direct interactions between Par6 and Nok via the ECR1 domain. By contrast, the Nok PDZ domain, which is thought to mediate association with Crb, is essential for polarization of all embryonic epithelia tested.

To define domains responsible for tethering Nok to apical junctions, we analyzed the subcellular distribution of the different HisMyc-tagged Nok deletion proteins at gastrula stages (6 hpf) and within myocardial cell junctions (34 hpf) by fluorescence immunohistochemistry. During late gastrula stages, the different Nok deletion proteins colocalized with aPKC at apical junctions of the superficial enveloping layer (EVL) of WT embryos (supplemental material Fig. S1). In contrast to our finding that the PDZ-domain-mediated association of Nok with Crb is functionally essential for the polarization of epithelia, Nok^Δcrb correctly associated with the apical membrane, which suggests that the presence of WT Nok protein stabilizes protein complexes containing this deletion protein. Therefore, in the WT, apical association is not completely disrupted in any of the deletion mutants tested at this stage.

To determine the localization of Nok deletion proteins within myocardial tissue, hearts of 34 hpf transgenic [Tg(cmlc2:GFP)];nok^s305 mutant embryos were stained with anti-Nok antibody (raised against the N-terminal 200 residues, which include the ECR1 domain and parts of the L27N domain) (Bit-Avragim et al., 2008) and anti-ZO-1 antibodies. In nok^s305 mutant myocardial cells, ZO-1-positive adhesion junctions were lost and endogenous Nok protein could not be detected (Fig. 5A). As expected for Nok^wt and Nok^ΔPatjΔLin7 proteins, which partially complement nok mutant or morphant cardiac morphogenesis defects, myocardial cells had a WT appearance and were correctly polarized, exhibiting ZO-1-positive and Nok-positive adhesion junction belts throughout the myocardial layer (Fig. 5B,C arrowheads). Therefore, association of Nok with Patj or Lin-7, which is thought to be mediated via the two L27 domains, is not necessary for apical localization of Nok and for the maintenance of apical junctions within myocardial cells. This finding is consistent with our observation that cardiac morphogenesis does not require the two L27 domains of Nok.

By contrast, and corresponding with their failure to complement heart morphogenesis defects, His-Myc-tagged Nok^ΔPar6, Nok^ΔCrb and Nok^ΔC proteins did not localize to myocardial cell membranes but were enriched within the cytoplasm of myocardial cells (Fig. 5D-F). Similarly, ZO-1-positive junction belts were lost throughout the myocardial layer in these mutant tissues. Together, these observations suggest that stabilization of ZO-1 junctional belts within myocardial cells and localization of Nok to myocardial cell junctions requires ECR1, PDZ or C-terminal-domain-mediated protein-protein interactions.

In order to determine whether apical localization of Nok requires Crb, we analyzed the distribution of endogenous Nok protein in ome/crb2a^m289 mutants. Whereas endogenous Nok and ZO-1 localized to apical junction belts within WT myocardial tissue (Fig. 5G), both proteins were mislocalized and not distributed in a polarized fashion in ome/crb2a^m289 mutants (Fig. 5H). Together, these results suggest that the Nok PDZ-domain-mediated interaction with Crb proteins is essential for the correct polarization of myocardial cells and that Ome/Crb2a is required for the correct tethering or localization of Nok to apical junctions within this tissue at 34 hpf.

Our observation that the Nok ECR1 domain, which is thought to mediate association with Par6 proteins, is functionally dispensable both within the RPE and for straight body form, suggests that Par6 proteins are not essential scaffolding partners for Nok within these tissues. Par6 proteins can bind murine Pals1 via the PDZ domain (Hurd et al., 2003; Wang et al., 2004). Binding of Par6 to Crb proteins is also mediated via the PDZ domain (Hurd et al., 2003; Lemmers et al., 2004; Kempkens et al., 2006).

Mutation of zebrafish pard6γb causes phenotypes similar to nok and has/prkci mutants within myocardial cells, the RPE and the neural tube (C.M. and D.Y.S.S., unpublished results). To test whether the Pard6γb PDZ domain is essential for apical localization, we generated a mutant Pard6γb construct that lacks several essential residues within the PDZ-binding domain that have been shown to be essential for Par6 PDZ-domain-mediated association with other proteins (Pard6γb^ΔKPLG) (Joberty et al., 2000; Hurd et al., 2003). The subcellular localization of Pard6γb^ΔKPLG was determined by expressing N-terminal fusions of eGFP with Pard6γb^wt or Pard6γb^ΔKPLG. H2B-mRFP mRNA co-injection was used to label nuclei (Megason and Fraser, 2003). Similarly to Pard6γb^wt-GFP, Pard6γb^ΔKPLG-GFP primarily localized to the apical surface of the neural tube in WT or pard6γb^s441 mutant embryos (Fig. 6A-C). Therefore, apical localization of Pard6γb within the neural tube does not require direct association with Crb or Nok via its PDZ domain. This finding suggests that Pard6γb is not an essential scaffolding partner for Nok within this tissue.

To test whether apical localization of Pard6γb within the neural tube is indeed independent of Nok or Ome/Crb2a, we expressed Pard6γb^wt-GFP in nok^m520 (Fig. 6E) or ome/crb2a^m289 mutants (Fig. 6F) and found that it failed to localize to the apical surface of the neural tube. Similarly, loss of Ome/Crb2a caused the mislocalization of Nok within the neural tube (Fig. 6H). Therefore, apical localization of Pard6γb within the neural tube is independent of PDZ-domain-mediated interactions but requires the presence of Nok or Ome/Crb2a. This observation suggests that apical tethering of Pard6γb within the neural tube is mediated by alternative binding partners.

Our results provide the first systematic structure-function characterization of protein-protein interaction domains of the Crumbs complex protein Nok during development. Our analysis suggests that distinct multi-protein assemblies of Crb-Nok and Par6-aPKC function in the polarization of different zebrafish embryonic epithelia.

We have shown that apical junctional localization of Nok within the neural tube and myocardium requires Ome/Crb2a. Localization of the respective Nok deletion protein to apical myocardial cell junctions is impaired, which is consistent with an essential requirement for the Nok PDZ domain to mediate interactions with Crb proteins. Moreover, maintenance of the epithelial integrity of all embryonic tissues assayed in this study depends on the presence of the Nok PDZ domain. Our findings add further support to the idea that PDZ-domain-mediated interactions between Drosophila Sdt or murine Pals1 and Crb are generally essential for epithelial maintenance (Bachmann et al., 2001; Hong et al., 2001; Roh et al., 2002).

We have shown that zebrafish Nok^ΔC localizes apically during gastrula stages, but fails to associate with the apical membrane of myocardial cells at later stages and to complement epithelial polarity defects within this tissue. Since Nok^ΔC is not functional in any of the tissues assayed, the C-terminus may be required to stabilize the protein. In Drosophila, the mutation in sdt^XP96 is thought to result in a premature stop codon that produces a truncated protein lacking the GUK domain, similarly to the Nok^ΔC deletion mutant used in our study (Hong et al., 2001). The truncated Drosophila Sdt protein localizes correctly within embryonic photoreceptor cells but fails to associate with the membrane in pupal and adult photoreceptor cells. In line with these findings, the C-terminus of mammalian Pals1 has been shown to be essential for tight junction formation and E-cadherin trafficking in a calcium-switch assay (Wang et al., 2007). Intramolecular conformational rearrangement caused by binding of GUK-4.1/Hook-SH3 has been predicted to be a common feature of MAGUK proteins (Nix et al., 2000; Tavares et al., 2001). Additional members of the membrane palmitoylated protein family (Mpp) have been shown to interact with Mpp5 through the GUK domain. Mpp5 can recruit Mpp4 into the apical Crb1 complex via physical interaction between its GUK domain and the SH3 domain of Mpp4 (Kantardzhieva et al., 2005). Mpp3, another Mpp family member with potentially redundant epithelial function, is recruited to the retinal outer-limiting membrane by Mpp5 (Kantardzhieva et al., 2006). Taking these observations into account, our results suggest that the Nok GUK domain is essential for epithelial integrity.

Our results demonstrate that the localization of zebrafish Pard6γb within the neural tube depends on Ome/Crb2a and Nok. However, apical Pard6γb membrane association within this tissue is not critically dependent on its PDZ motif, which has been shown to mediate direct interactions with Crb or murine Pals1 (Hurd et al., 2003; Lemmers et al., 2004; Wang et al., 2004; Kempkens et al., 2006). This finding underlines the importance of alternative protein-protein interactions of this multi-domain linker protein. The N-terminus of Drosophila Par6 has been shown to interact with one PDZ domain of Drosophila Patj, and this interaction might also be involved in stabilizing zebrafish Pard6γb at the apical membrane of neural tube cells (Nam and Choi, 2003). Since the Pard6γb PDZ domain is not required for apical localization of the protein and because the Nok L27(N) domain, which presumably interacts with Patj, is required for neural tube integrity (as indicated by the severe body curvature in mutants), Pard6γb may directly associate with Patj within neural tube cells. In turn, Patj, which thus functions as a scaffold for Par6-aPKC complex proteins, attaches to the L27(N) domain of Nok. The Nok PDZ-domain-mediated interaction with the transmembrane protein Ome/Crb2a may tether the entire multi-protein complex to the apical membrane (Fig. 7B).

In myocardial cells, deletion of the Nok bipartite L27 domains complements the nok loss-of-function phenotype and stabilizes ZO-1 junctional belts to which the respective Nok deletion protein correctly localizes. Therefore, within myocardial cells, unlike in neural tube cells, Patj (and Lin7) may not be essential to stabilize protein complexes containing Nok. This suggests that, either the attachment of Patj to Crb proteins is via a distinct physical interaction or Patj is not functional in myocardial cells. Consistent with this finding and supporting the idea of an alternative mode of protein-protein interactions, the Nok ECR1 domain, predicted to mediate direct interactions with Par6 proteins (Wang et al., 2004), is essential for myocardial tissue integrity. Thus, our results suggest that within myocardial cells, direct interactions between the Par6-aPKC complex via the Nok ECR1 domain are essential and stabilize the entire multi-protein complex at the apical membrane through the Nok PDZ-domain-mediated attachment to Ome/Crb2a (Fig. 7C). Consequently, our finding that not only Nok, but also Pard6γb utilizes alternative protein-protein interactions in a modular and highly tissue-specific manner reveals yet another level of complexity in the assembly of multi-protein complexes at the apical membrane.

Members of the MAGUK protein family are ideally suited for the modulation of membrane-associated multi-protein assemblies because of the presence of multiple protein-protein interaction domains. Our structure-function analysis of Nok protein-protein interaction domains has revealed that the ECR1 domain, which is thought to mediate interactions with Par6 proteins, and the L27 domain, which may mediate interactions with Patj and Lin7, are both required for correct retinal tissue polarity. Therefore, direct interactions between Nok and Par6 proteins as well as interactions between Nok and Patj appear to be essential within this tissue. The PDZ domains of Drosophila Sdt and murine Pals1, as well as that of Par6, have been shown to directly associate with the ERLI motif of Crb in vitro (Bachmann et al., 2001; Hong et al., 2001; Roh et al., 2002; Hurd et al., 2003; Lemmers et al., 2004). Moreover, similar calculated binding affinities between Crumbs and Sdt or Drosophila Par6 have recently been demonstrated (Kempkens et al., 2006). Since the small size of the Crb ERLI motif does not permit a simultaneous interaction with both Nok and Par6 PDZ domains, it is likely that Nok ECR1- and L27-domain-mediated interactions are both used by the Crb-Nok-Patj and the Par6-aPKC complexes for the assembly of a highly conserved `canonical' apical multi-protein complex within neural retinal cells (Fig. 7A).

In summary, our study suggests that alternative tissue-specific scaffolding interactions are in place between Crb-Nok, Patj, Lin-7 and Par6-aPKC (Fig. 7). Based on the results shown here, we suggest that two of the scaffolding proteins, Nok and Pard6γb, utilize alternative modes of apical localization within different epithelia. Currently, it is not known whether differential splicing of the vertebrate orthologs pals1/mpp5 or nok occurs. However, it is conceivable that the modular nature of protein-protein interactions among changing binding partners of the Crb and Par6 protein complexes increases the complexity of multi-protein assemblies at the apical membrane. This additional level of complexity in the assembly of epithelial junctions could foster the adaptation and specialization of epithelial tissues during development.

Zebrafish were maintained under standard conditions (Westerfield, 1994). Embryos were staged by hpf at 28.5°C (Kimmel et al., 1995). The following fish strains were used: wild-type AB, Tg(cmlc2:GFP) (Huang et al., 2003), nok^m520, nok^s305 (Rohr et al., 2006), ome^m289 (Omori and Malicki, 2006) and pard6γb^s441 (C.M. and D.Y.S.S., unpublished results). The nok^s305 mutation causes a premature truncation of the protein that deletes the entire GUK domain (X. Wei, personal communication). Genetically, nok^s305 behaves like a complete, or almost complete, loss-of-function allele (Rohr et al., 2006).

The Nok^wt (a.a. 1-703), Nok^ΔC (a.a. 1-429) and Nok^L27 (a.a. 151-265) domain were produced by PCR amplification from a full-length cDNA template, introducing XhoI and XbaI restriction sites at the 5′ and 3′, respectively, and constructs were subsequently subcloned into the pCS2+His-Myc expression vector (Rohr et al., 2006). For the other Nok-truncated proteins, the following residues were deleted: Nok^ΔCrb (Δ293-364), Nok^ΔPar6 (Δ54-64), Nok^ΔPatjΔLin7 (Δ151-265), Nok^ΔPatj (Δ152-209). Primer sequences are available upon request. The h2b:mrfp construct was kindly provided by S. G. Megason (Megason and Fraser, 2003).

Based on the published sequence of pard6γb (accession number NM_212563), we designed PCR primers to amplify full-length cDNA. Additional PCR primers with incorporated restriction enzyme sequences allowed us to place these full-length cDNA sequences into the pCS2⁺ vector. The KPLG mutation was introduced by site-directed mutagenesis (primer sequences are available upon request). For fusion proteins, GFP followed by a short amino acid sequence (S-G-G-G-G-S) was placed into the existing expression constructs at the 5′ end of the coding sequence.

Constructs were transcribed using the SP6 MessageMachine kit (Ambion). For rescue experiments, nok/Tg(cmlc2:GFP) embryos were injected with 50-75 pg mRNA. For overexpression (to determine the subcellular localization patterns), 75-100 pg mRNA were used. For MO injections, the following concentrations were used: 200 μM (MO^nok) and 150 μM (MO^p53). MO^nok: 5′-TCAGCAGCGGCTCCAGACACTGA-3′; MO^p53: 5′-GCGCCATTGCTTTGCAAGAATTG-3′.

Antibodies were raised against amino acids 1-200 of Nok. Appropriate primers were used to amplify the cDNA encoding this portion of the protein and to clone the fragment into pET23/T7, a modified pET23a vector (Obermann et al., 1998). The His-tagged protein was expressed in E. coli BL21-CodonPlus(DE3)-RIL (Novagen), solubilized using a buffer containing 6 M guanidine-HCl and purified as described (Chumpia et al., 2003). Rabbits were immunized four times with the purified recombinant protein according to a standard protocol (Biogenes, Berlin, Germany). Specific antibodies were affinity purified using Ni-NTA agarose beads (Sigma) essentially as described (Chumpia et al., 2003).

Immunohistochemistry was performed as previously described (Horne-Badovinac et al., 2001). The following antibodies were used: rabbit anti-aPKCζ (1:100, Santa Cruz Biotechnology), mouse anti-Myc (1:200, Invitrogen), S46 (1:20, Developmental Studies Hybridoma Bank), mouse anti-ZO-1 (Zymed Laboratories, Invitrogen), goat anti-rabbit Cy5 (1:200), anti-mouse FITC (1:200) (Jackson ImmunoResearch). For sectioning, stained embryos were postfixed overnight at 4°C in 2% paraformaldehyde, 0.3 M sucrose. Embryos were embedded in 4% low melting agarose and sectioned on a Leica VT1000 Vibratome. Confocal images were obtained using the Leica TCS SP2 confocal microscope with a 40× oil lens and 2× zoom. Whole embryos were documented under a Leica MZFLIII stereomicroscope using the 1× and 4× objectives with 5-10× zoom and Leica IM50 software package. Photos were processed using Photoshop (Adobe).

Protein extracts of 24 hpf zebrafish embryos were prepared according to standard protocols (Westerfield, 1994). Membranes were probed with mouse anti-Myc antibody (1:1000, Invitrogen). For loading control, membranes were stripped and tested for acetylated-tubulin (mouse anti-acetylated tubulin, 1:1000, Sigma).

For genotyping of rescued nok^m520 embryos, we made use of a SalI restriction site that is deleted by the mutation and performed PCR on tail tissue followed by a SalI digest. DNA primers used for genotyping are available upon request.

Similar to our findings, Bulgakova et al. recently showed that, in the Drosophila eye, different domains of the Nagie oko orthologue Stardust are differentially required for its apical localization in pupal and adult photoreceptor cells, and that the assembly and stability of the Crb-Sdt complex relies on different regulatory processes at different developmental stages (Bulgakova et al., 2008).

We are indebted to S. Fraser, J. Malicki, H.-J. Tsai, E. Wanker and X. Wei for sharing reagents and tools and to Robert Fechner for expert technical assistance with the fish facility. We would like to thank Thomas Willnow, Ulrike Ziebold and Elisabeth Knust for comments on the manuscript. We would also like to thank Natalia Bulgakova and Elisabeth Knust for sharing unpublished observations and Xiangyun Wei for providing information on the molecular nature of the nok^s305 allele prior to publication. This work was supported, in part, by a presidential grant from the Helmholtz Society.