ojoplano-mediated basal constriction is essential for optic cup morphogenesis

Visual acuity in chambered eyes relies on focusing light onto a hemispheric cup in which surface photoreceptors are arranged. In vertebrates, two consecutive morphogenetic events shape the primordium into a hemispheric structure (Schmitt and Dowling,1994). The first step involves the lateral evagination of retinal progenitor cells (RPCs) to form optic vesicles. It has recently been reported that this step is driven by the individual migration of RPCs(Rembold et al., 2006), under the control of the transcription factor Rx3(Loosli et al., 2001). In a second series of morphogenetic events, the optic vesicles transform into the hemispheric bi-layered cup. Although extensive information is available regarding the initial specification and patterning of the eye anlage(Chow and Lang, 2001; Martinez-Morales et al.,2004b), the cellular and molecular mechanisms underlying optic cup folding remain largely unexplored. In particular, the dynamic movements that shape the optic cup have never been recorded in living embryos.

During organogenesis, spatial tissue rearrangements rely on a limited number of stereotypical cell behaviors that are exemplified by, at opposite extremes, individual cell migration and coordinated cell shape change(Locascio and Nieto, 2001; Pilot and Lecuit, 2005). At the molecular level, changes in the morphology of cells depend on the local organization of the cytoskeleton. In turn, cytoskeletal rearrangements depend on membrane polarization and compartmentalization, a consequence of cell-cell and cell-substrate interactions (Nelson,2003).

In epithelial tissues, morphogenetic events along the apicobasal axis have a deep impact on macroscopic rearrangements. Apical constriction in polarized epithelia is the best studied of these phenomena, and involves the conversion of columnar cells into wedge-shaped cells. It is a common mechanism employed to bend cell sheets in several tissues in different organisms; for example,during gastrulation in Drosophila(Leptin, 2005) or neural tube closure in vertebrates (Wallingford,2005). The molecular mechanism underlying apical constriction has been studied in detail in Drosophila and involves the recruitment of the acto-myosin cytoskeleton to the cellular cortex(Young et al., 1991). Both in Drosophila and in vertebrate epithelia, cell polarity depends on the local assembly of receptors and adaptor proteins. Three apical or sub-apical complexes, Crumbs/PATJ/Stardust, Par3/Par6/atypical protein kinase C (aPKC),and Scribble/Discs-Large/Lethal-Giant-Larvae, have been shown to be essential for epithelial polarization in all model organisms analyzed(Bilder, 2003; Nelson, 2003; Schock and Perrimon, 2002). The polarized distribution of some of these apical polarity determinants has also been reported in the developing neural retina(Malicki, 2004).

Cell polarization has also been linked to the interaction of cells with the extracellular matrix (ECM), which, in focal adhesions, is mainly mediated by transmembrane integrin receptors. Focal adhesions contain a number of integrin-associated scaffolding proteins (e.g. paxillin, vinculin, FAK) that act as an interface between the plasma membrane and the actin cytoskeleton(Burridge and Chrzanowska-Wodnicka,1996; Turner,2000). In contrast to the universal role of the above-mentioned apical determinants, the role of focal adhesions in cell polarity largely depends on the cellular context. Although it has been shown that integrins have no role in establishing the polarity of monostratified epithelial cells in Drosophila (Fernandez-Minan et al., 2007), they are required to place general apicobasal determinants in the stratified epithelium of the mammalian epidermis(Lechler and Fuchs, 2005). We are still far from understanding how general cell polarity and cell adhesion are interconnected. Moreover, there is little information on how these complex cellular processes are coordinated to control the morphogenesis of an entire organ.

Here, we have identified the medaka (Oryzias latipes) mutation ojoplano (opo), which affects the morphogenesis of several epithelial tissues. We employ this morphogenetic mutant as a tool to understand the cellular mechanisms underlying eye formation. We provide a detailed comparative analysis of the morphogenetic movements underlying optic cup folding, as well as the polarity and cell shape dynamics of neuroblasts,in wild-type and in opo mutants. opo encodes an uncharacterized transmembrane protein that localizes to compartments of the secretory pathway and to neuropeithelial end-feet. We provide evidence that opo function is essential to localize focal adhesion components basally and that integrin function is required for optic cup morphogenesis. A single previous report in humans identified incomplete homologous transcripts of the gene within a locus linked to orofacial clefting hereditary disease(Davies et al., 2004). The craniofacial phenotype observed in medaka mutants further supports the link of opo to one of the most prevalent birth defects in humans.

Medaka (Oryzias latipes) wild-type strains Cab and Kaga were kept as described (Koster et al.,1997). The Kaga strain was used for chromosomal assignation and positional cloning. Embryos were staged according to Iwamatsu(Iwamatsu, 1994).

Stage 23 rx2::eGFP embryos were immobilized in balanced salt solution (BSS; containing 0.01% tricaine and 3 mM heptanol) and embedded in 1%low-melting agarose. Time-lapse analysis was performed using SPIM(Huisken et al., 2004) with a Zeiss 40×/0.8W lens and a 488-nm laserline. Samples were imaged for more than 30 hours with a time resolution of 10 minutes. Several central planes,spanning 9 μm, were integrated using ImageJ (NIH). For 3D reconstructions,the information from full stacks was integrated using Imaris (Bitplane). For standard confocal microscopy, stage 23 rx2::mYFP embryos were immobilized. Time-lapse analyses were performed on a Leica TCS-SP5 microscope with a HCPlanApo20×/0.7NA objective and a 514-nm laserline. At each time point, three optical sections spanning 3 μm were integrated using Imaris. Samples were imaged with a time resolution of 2 minutes.

opo was assigned to chromosome 20 by bulk segregation analysis(Martinez-Morales et al.,2004a). The genetic distance to the locus was narrowed by testing genetic markers (Naruse et al.,2000) on 1296 mutant chromosomes. We used the markers Ola2311a(8/1296; 0.6 cM) and Ola4805 (7/1296; 0.5 cM) as references flanking the mutation. From these anchoring points, chromosomal walking defined an 85-kb interval (0/1296) containing opo. All genetically mapped positions were confirmed at Ensembl (Kasahara et al., 2007).

Rescue experiments were performed using the BACs Hd228o24, Hd195c24 and Hd219g12 (Khorasani et al.,2004). Purified BACs were injected (10-30 ng/ul) into one-cell stage embryos. For tissue-specific cDNA rescue, rx2::opo plasmid was injected (20 ng/μl) into one-cell stage embryos using the meganuclease approach (Thermes et al.,2002). Splicing morpholinos were obtained from Gene Tools. The following morpholinos were injected:

MOm-opoEx18a, 5′-TGGGTCCAGCTGGTGCTGACctgtt-3′ (250-500μM);

MOm-opoEx3d, 5′-tcatcaagcaggactcacccaTCAG-3′ (100-300 μM);and

MOz-opoEx3d, 5′-ggactcacccaTCAGAAATTCAGCC-3′ (300 μM). opo exon 18 skipping resulted in protein destabilization, as assessed by western blot (not shown).

The complete gene structure of opo was deduced by 3′ and 5′ RACE (SMART-RACE, Clontech). Eight different primers were used for the 5′ and 3′ RACE (sequences are available upon request). 5′ RACE reactions detected a single 5′ end (also confirmed using GeneRacer, Invitrogen). The 3′ race detected several splicing isoforms,further confirmed by RT-PCR. EMBL accession numbers AM920650 and AM920651 correspond to the long and short isoforms, respectively. For RT-PCRs, total RNA was extracted (TRIzol, Invitrogen) and reverse transcription reactions were performed (Super Script III, Invitrogen). A 3′ fragment of the chicken Opo homolog was also isolated by RT-PCR (EMBL accession number AM920652). The 24 exons of opo and their flanking genomic regions were sequenced from heterozygous parents and their mutant progeny. Although several polymorphisms were detected, no mutation was identified. opo cDNAs derived from wild-type and mutant embryos were also compared. The anomalous opo splicing variant in exon 4 was never observed in wild-type embryos (20 independent cDNAs were analyzed), but was detected in four out of seven mutant full-length cDNAs analyzed. Examination of flanking introns 3 and 4, both in heterozygous parents and mutant progeny,revealed an insertion of two base pairs in intron 4 that correlates with the mutation. The insertion is located 715 bp 3′ of the affected splicing acceptor in the sequence:

Wild type,5′-TTGATTTATTTATTTTTATTGCTTTTGCGCTAAAAGGTT-3′;

Mutant,5′-TTGATTTATTTATTTATTTATTGCTTTTGCGCTAAAAGGTT-3′.

This sequence contains a branch point sequence (underlined) and a polypyrimidine tract. Moreover, the 10-bp wild-type motif ATTTATTTTT disrupted by the insertion (in bold italics) is conserved in intron 4 of teleosts (Fugu,Tetraodon and Stickleback). These observations suggest that the disrupted region has a conserved regulatory role related to local assembly of the spliceosome complex.

Transmembrane predictions were performed at Sosui[http://bp.nuap.nagoya-u.ac.jp/sosui(Hirokawa et al., 1998)] and Phobius[http://phobius.sbc.su.se(Kall et al., 2004)].

Stable lines rx2::eGFP and rx2::mYFP were generated by ET-recombination (Muyrers et al.,1999). Both eGFP and mYFP (Clontech) were introduced in a cosmid containing the rx2 gene, 4 amino acids downstream of the ATG start site.

Expression constructs were cloned into I-SceI vectors, and injected (10-20 ng/μl) into one-cell stage embryos(Thermes et al., 2002). To generate the tyr::eGFP construct, a 3-kb promoter fragment of medaka tyrosinase was fused to eGFP. The vectors rx2::opoand rx2::opoGfp were cloned, using Gateway technology (Invitrogen) by inserting opo and opo:GFP into the rx2 destination vector. The rx2 destination vector was generated by inserting a 2.4-kb fragment of the medaka rx2 promoter upstream of a gateway destination cassette in the I-SceI transgenesis vector. vsx3:Intβ1tail-eGFP and vsx3::torso^D/βcyt were generated by cloning the fusions Intβ1tail:eGFP and torso^D/βcyt under the control of the retinal specific vsx3 promoter (5-kb fragment upstream of the vsx3ATG). Intβ1tail:eGFP was generated by cloning eGFP at the C terminus of the 70 C-terminal amino acids of the Drosophila integrin beta PS. The dominant-negative torso^D/βcyt construct was reported previously (Martin-Bermudo and Brown,1999). The medaka vsx3::OLint-β1Δ21 construct is equivalent to a human dominant-negative variant, previously shown to interfere with cell adhesion to laminin (Retta et al., 1998). Fusion proteins (zebrafish) ASIP/par3:eGFP(von Trotha et al., 2006) and pEGFPC1-(avian) paxillin WT1-559(West et al., 2001) were subcloned into pCS2+. Capped sense RNAs were synthesized in vitro(mMessage Machine, Ambion). Purified RNA was then injected (50-100 ng/μl)into one-cell stage embryos.

Antibodies against Opo protein were generated by immunizing animals with purified recombinant GST-fusions. opo fragments encoding either the 112 N-terminal or the 183 C-terminal amino acids of the protein were cloned into pGEX-KG for expression. Purified GST:opo-N and GST:opo-C were employed to generate both mono and polyclonal antibodies. Polyclonal anti:opo-N was employed in immunostainings at a dilution of 1:500. The following commercial primary antibodies and molecular probes were used:anti-paxillin monoclonal antibody at 1:300 (BD Biosciences); polyclonal anti-phosphorylated histone H3a at 1:500 (Upstate Biotechnology); polyclonal anti-γ-tubulin at 1:500 (Sigma T5192); polyclonal anti-laminin at 1:500(Sigma L9393); monoclonal anti-acetylated tubulin at 1:1000 (Sigma T6793);monoclonal anti-α-tubulin at 1:1000 (Sigma T9026); polyclonal anti-PKCζ (C-20) at 1:1000 (Santa Cruz Biotechnology); polyclonal anti-calnexin (Stressgen) at 1:1000; monoclonal anti-GM130 (BD Biosciences) at 1:250; monoclonal anti-integrin β/CD29 (Epitomics); phalloidin Alexa-488 (Invitrogen) at 1:10.

Embryo heads were homogenized in 5% β-mercaptoethanol, 2% SDS,1 mM PMSF (phenylmethylsulphonyl fluoride) and 50 mM Tris-HCl (pH 6.8). Supernatants were separated by SDS-PAGE (20 μg of protein per lane) and transferred to nitrocellulose membranes, blocked with 5% non-fat dry milk in PBT (PBS, 0.1% Tween) and probed with primary antibodies [anti:α-tubulin monoclonal (1:1000); anti:opo-N monoclonal supernatant (1:20) and anti:opo-C polyclonal (1 ug/ml)]. Blots were developed using peroxidase-conjugated anti-rabbit or anti-mouse antibodies (1:7000, Jackson). Antibody binding was visualized using the ECL system (Amersham).

Embryos were fixed in 4% paraformaldehyde-PBS overnight at 4°C,equilibrated in 30% sucrose in PBS, embedded in OCT compound and sectioned at 18 μm using a cryostat (Leica). Tissue was blocked for 1 hour in PBS 0.1%Tween (PBT) and 10% FCS. Primary antibodies diluted in PBT-10% FCS were added overnight at 4°C. After washing in PBT, sections were incubated with Alexa-conjugated (Molecular Probes) secondary antibodies for 1 hour. Fluorescence sections were mounted with PBS/glycerol and analyzed with a confocal imaging system (TCS SP2, Leica). For cortical actin staining,sections were washed with PBS/Triton (0.1% Triton X-100) and incubated with phalloidin Alexa-488 (diluted 1:10 in 5% DMSO) in PBS/Triton overnight at 4°C. Methylene Blue-Azure II stainings were performed as described(Loosli et al., 2004).

Cells were transfected with CMV::opoGfp (FuGENE6, Roche) and allowed to express for 24 hours at 37°C. Cells were then fixed in ice-cold(-20°C) methanol for 4 minutes, washed with PBS, then incubated with appropriate dilutions of primary antibodies for 30 minutes and secondary antibodies labelled with Alexa-568 for 30 minutes. After PBS washing, cells were mounted in Moviol and images acquired on a Zeiss Axiovert200 microscope.

In situ hybridization was performed using digoxigenin riboprobes as described (Martinez-Morales et al.,2005). A 3′ opo probe common to all isoforms was used. For vibratome sections, embryos were embedded in gelatin/albumin and sectioned at a thickness of 30 μm (Leica).

Cell transplantation was done as described(Rembold et al., 2006). Briefly, embryos were dechorionated with proteinase K (10 mg/ml in H₂O) and hatching enzyme, and maintained on dishes coated with 1%agarose/BSS. At blastula stage (stage 10/11), approximately 20-40 cells were transplanted from donors into the animal pole of host embryos (either wild-type or opo mutants). The development of both labelled donors(rx2::eGFP) and unlabelled hosts was followed for phenotypical assessment at stages 24 to 26.

Quantitative data are expressed as mean±s.e.m. Significant differences were evaluated by two-way ANOVA with Bonferroni post-hoc tests(GraphPad Prism) and are indicated when significant.

In a large-scale ENU mutagenesis screen in medaka(Loosli et al., 2004), we identified ojoplano (opo), a mutant that affects the morphogenesis of several epithelial tissues. opo is a recessive lethal mutation with full penetrance and minimal phenotypical variability. Mutant embryos first become apparent at the onset of optic cup folding (stage 23), by virtue of their abnormal retinal morphology. Later during organogenesis, additional morphogenetic defects were observed in the brain,heart, fins, craniofacial region and eye. The ocular phenotype was particularly prominent, giving rise to the name: ojoplano, `flat eye'in Spanish. As development proceeds, mutant optic vesicles fail to fold properly, resulting in a large ventral opening(Fig. 1A,B). Moreover, retinal ganglion cell axons exit the misshapen optic cup ventrally in defasciculated bundles (Fig. 1C,D). Despite the mutant eye morphology, RPC proliferation, neuronal differentiation and retinal lamination were normal at late organogenesis stages(Fig. 1E,F; see also Fig. S1A in the supplementary material). In addition, the lens, despite its abnormal morphology, was correctly specified (see Fig. S1A in the supplementary material). The abnormal shape of the optic cup is not a consequence of a failure in ventral retina specification. Known ventral markers were normally expressed in mutant retinae (see Fig. S1B in the supplementary material; data not shown), indicating that patterning is not altered in opomutants.

As with the eye phenotype, none of the other morphogenetic defects observed in opo mutants were associated with patterning failures. Despite their correct initial specification, neural crest cells failed to migrate/delaminate (see Fig. S2A in the supplementary material). Consequently,neural crest derivatives are reduced or absent in opo mutants. Similarly, despite the correct specification of fin bud mesenchyma, the apical ectodermal ridge was misshaped. As a result, pectoral fins fail to grow properly (see Fig. S2C in the supplementary material). These deficiencies indicate a general role for the mutated gene in epithelial morphogenesis. Here, we focus on optic cup folding to gain insight into the function of opo in tissue morphogenesis.

To gain insight into the cellular mechanisms shaping the vertebrate eye, we followed optic cup morphogenesis in vivo in wild-type and mutant embryos. By using an eye-specific transgenic line, rx2::eGFP, that labels all RPCs, eye folding was followed by SPIM(Huisken et al., 2004). Both wild-type and opo retinae were imaged through the transition from optic vesicle to optic cup (30 hours, stages 22 to 27). At each time point, a complete stack of images spanning the whole retina was recorded from a lateral perspective. Three-dimensional renderings show that, although epithelial folding occurs in mutant retinae along the dorsoventral axis, it is not even initiated along the anteroposterior axis(Fig. 1G,H).

To visualize the process clearly, we generated projections of several slices at the level of the lens. In wild-type retinae, the initial folding of the epithelium occurs rapidly (see Movie 1 in the supplementary material). After 6 hours, as the tissue grows and anterior and posterior borders of the sheet bend ventrally, the eye primordium was already folded into a hemisphere. Despite this, the final ventral fusion of retinal borders, sealing the optic fissure, was not complete after 40 hours of recording (stage 27). In opo retinae, the epithelial sheet, although growing along the anteroposterior axis, does not bend ventrally even after 33 hours of development (see Movie 2 in the supplementary material). The clear temporal separation of cup folding and fissure closure thus indicates that these are distinct events with independent mechanisms.

We mapped the opo locus to chromosome 20 by bulk segregation analysis (Martinez-Morales et al.,2004a) and identified the mutated gene by BAC walking(Fig. 2A). The locus sits in a 400-kb gene desert and includes a 70-kb transcription unit with homology to an uncharacterized human gene. Incomplete transcripts of this gene have been described in humans, and the locus has been proposed as being causative for orofacial clefting syndrome (Davies et al.,2004). The complete structure of the medaka gene, termed ojoplano (opo) after the mutation, was deduced by 3′and 5′ RACE (Fig. 2A). The gene comprises 24 exons (3568 bp in total), and is transcribed into several splicing isoforms. opo transcripts do not encode any known protein domain. Orthologous genes can be identified in syntenic regions of all of the vertebrate genomes available (see Fig. S3A in the supplementary material). Two conserved regions were found: a 90-amino acid N-terminal domain, which corresponds to exons 3 to 5, and a 230-amino acid C-terminal domain, corresponding to exons 18 to 22. Transmembrane topology prediction algorithms identified four putative transmembrane helices in the conserved carboxy domain, orienting both termini of the protein towards the cytoplasm(see Fig. S3B in the supplementary material). Using either monoclonal antibodies against the N terminus (Fig. 3B, Fig. 4A) or polyclonal antibodies against the C terminus (not shown), we identified two isoforms by western blot: a 130-kDa band, corresponding to a 1091-amino acid protein encoded by an mRNA spanning all exons, and a 42-kDa band,corresponding to the skipping of the intermediate exons 5 to 16.

The injection of BAC Hd228o24, covering the opo locus, into the offspring of a carrier-cross resulted in a rescue efficiency of 36%(Fig. 2B). In addition, we observed an intermediate phenotype never present in uninjected carrier-crosses: a partial rescue of the facial phenotype and/or asymmetric rescue of the fins. To unambiguously confirm the identity of the causative gene, we placed opo cDNA under the control of the eye specific promoter rx2 (rx2::opo). Injection of this construct efficiently rescued the mutant phenotype in a tissue specific manner: 52% of the opo mutant embryos identified by their fin phenotype developed wild-type eyes (Fig. 2C).

We detected an anomalous splicing variant in opo mutants that was never observed in wild type. In mutants, four out of seven opo cDNAs analyzed used an alternative splice-acceptor in exon 4, skipping the first four nucleotides of the exon. This introduces a frameshift that results in a premature stop codon and leads to a truncated protein of 40 amino acids(Fig. 3A). In mutant embryos,we detected an insertion of two base pairs in intron 4 that is linked to the atypical splicing (see Materials and methods). As a consequence, Opo protein levels are reduced to 20% of wild-type levels for the 130-kDa full-length protein, and to 60% of wild-type levels for the 42-kDa isoform, as detected by western blot (Fig. 3B). These data indicate that the mutation is hypomorphic and that the observed phenotype is due to insufficient synthesis of wild-type protein.

To understand the phenotypical consequences of a progressive decrease of opo function, we performed morpholino injection experiments. Morpholinos interfering with the splicing of exon 18 were injected into rx2::mYFP embryos to examine the optic vesicle phenotype. Injected embryos were classified either as wild type-like, opo-like or strongly affected (Fig. 3C). opo-like morphants phenocopied mutants: eye folding did not occur,and craniofacial and fin malformations developed. In strongly affected morphants, the optic vesicles did not evaginate fully and remained fused at the midline (Fig. 3C). Molecular analysis of the phenotypical groups showed a direct correlation between the severity of the phenotype and the skipping of exon 18(Fig. 3D). Taken together, our results demonstrate that the progressive loss of opo function induces an increasingly severe morphogenetic phenotype, which indicates a requirement of Opo at different levels for different morphogenetic processes.

Similar phenotypes were obtained upon injection of a second morpholino blocking the splicing between exons 3 and 4 (see Table S3 in the supplementary material). Moreover, in parallel interference experiments in zebrafish, the injection of morpholinos against the exon 3/4 junction caused a phenotype similar to that of medaka opo: a failure of optic vesicle folding(see Fig. S4 in the supplementary material).

To study Opo regulation during embryogenesis, we investigated its developmental profile by western blot. At stage 10, prior to the onset of zygotic transcription, maternal Opo protein was detected in embryo extracts. Protein levels decreased at late gastrula stages and increased during organogenesis, peaking at stage 26 (Fig. 4A). The spatial expression pattern of opo was investigated by in situ hybridization. At stage 26, both in wild type and in mutant embryos, opo was expressed in all tissues affected by the mutation, including eye, brain and neural crest precursors(Fig. 4; see also Fig. S5 in the supplementary material). Interestingly, transcripts are localized at the basal side of the pseudo-stratified epithelium of the developing retina both in wild-type and in opo mutant embryos(Fig. 4B,C). Furthermore, this localization is not exclusive to the medaka retina, as we could detect a similar distribution of the chicken homologous transcript at the basal surface of the optic vesicle epithelium (see Fig. S5C in the supplementary material). To explore the subcellular localization of the Opo protein, we expressed a GFP-tagged opo gene under the control of the eye-specific promoters rx2 and vsx3. In transient rx2::opoGfp clones,Opo-GFP decorated the nuclei and also localized to the membranes of the basal cytoplasmic feet (Fig. 4D). In vivo examination of protein distribution in the stable vsx3::opoGfpline also revealed a preferential localization of Opo-GFP to the basal side of the epithelium (Fig. 4E),particularly to the tips of the neuroepithelial feet(Fig. 4F). Opo enrichment at the basal surface was further confirmed by immunostaining using polyclonal antibodies against the N terminus of the protein (see Fig. S5 in the supplementary material). The perinuclear distribution of Opo suggests that the protein localizes to the ER. Subcellular colocalization experiments in HeLa cells transfected with CMV::opoGfp showed a perinuclear and reticular distribution of the Opo-GFP fusion protein, which colocalized with the ER marker calnexin and showed some correlation with the golgi marker GM130(Fig. 4F-K). These experiments indicate that Opo localizes to intracellular compartments of the secretory pathway, mainly to the ER.

Given the central importance of epithelial polarization to tissue morphogenesis, the question arises as to whether Opo might be involved in the control of polarity. In contrast to the more common apical constriction, optic cup morphogenesis entails the bending of an epithelial sheet around its basal side. Given that the neural retina is polarized in the apicobasal plane(Malicki, 2004), and that optic cup folding may be linked to polarity, we examined the localization of apical and basal determinants in wild-type and opo mutant retinae. At stage 23, Par3-GFP (von Trotha et al.,2006) was localized to the apical side of the retina(Fig. 5A), showing that RPCs are polarized before optic cup folding. A similar distribution was found in opo mutants (Fig. 5E). Normal segregation of apical determinants in opo mutants was further confirmed in sections of wild-type and mutant retinae stained forγ-tubulin, which labels apical centrosomes(Fig. 5B,C,F,G), and aPKC (not shown). By contrast, we observed striking differences between the basal cell surfaces of wild-type and opo mutant retinae. During optic cup folding, the feet of RPCs attach to the basal lamina of the retina and form an array of narrow parallel cytoplasmic extensions that nuclei do not enter (zone of nuclear exclusion, ZNE). Using an rx2::mYFP (membrane YFP)transgenic line, we examined this basal organization of the tissue in wild-type and opo. In contrast to the membrane array observed in wild type, the feet of opo mutant neuroblasts are wider and appear disorganized, and nuclei are found closer to the basal surface(Fig. 5E,J; see Fig. S6 in the supplementary material). Consistently, although cortical actin cytoskeleton recruitment, as visualized by phalloidin staining, was normal at the apical junctions in opo mutants, reduced actin deposition was observed at the ZNE (Fig. 5K,P). At the basal surface, cytoskeletal organization and epithelial polarization depend on integrin-mediated focal contacts (Bokel and Brown, 2002; Schock and Perrimon, 2002). Therefore, we examined the distribution of paxillin, a scaffolding protein that links integrin receptors to the cytoskeleton (Turner, 2000). In contrast to its basal localization in wild-type retinae, paxillin accumulated poorly at the basal side in the mutants, as shown by immunostaining (Fig. 5L,Q) or after paxillin:gfp mRNA injection(Fig. 5M,R). Similarly,integrin β1, an essential component of the main receptor for laminins and collagens (Hynes, 1992), was also reduced basally, as revealed both by immunostaining(Fig. 5N,S) and by the eye-specific transgenic line vsx3::integrinβ1-eGFP(Fig. 5O,T). This deficient cortical recruitment of focal adhesion components is not a consequence of impaired basal lamina deposition, as laminin distribution was normal in mutant retinae (see Fig. S7A in the supplementary material). Furthermore, basal lamina integrity in opo mutant retinae was also confirmed by electron microscopy (see Fig. S7B in the supplementary material).

To assess whether focal adhesion formation has a role in optic cup morphogenesis, we interfered with integrin function by specifically expressing a dominant-negative construct in the eye anlage. We made use of the dominant-negative chimera torso^D/βcyt, which previously has been shown to interfere with integrin-adhesive function(Martin-Bermudo and Brown,1999). Injection of this construct under the control of the eye-specific promoter vsx3(vsx3::torso^D/βcyt) in the reporter line rx2::mYFP impaired optic cup folding in 27% of the treated embryos(n=102). The optic cup morphogenetic defects observed in these embryos largely resembled those of opo mutants(Fig. 6A,B) and persisted to late stages (Fig. 6C,D). Similar defects were observed in 7% of the embryos injected (n=71)with a second dominant-negative construct, vsx3::OLint-β 1Δ21, which lacks the last 21 cytoplasmic residues of medaka integrin β1 (not shown). These experiments indicate that integrin-adhesive function is required for optic cup folding.

It is known that nuclei adapt their shape in response to the mechanical tension transmitted through the extracellular matrix(Maniotis et al., 1997). The defective basal organization of the opo mutant retina suggested that RPCs might not adhere well to the basal lamina, which would result in a reduction in the transmitted tension. In agreement with this idea, opo mutant nuclei appear round, in contrast to the elongated morphology observed in wild-type RPCs (Fig. 7A,B). We quantified changes in nuclear morphology during retinal morphogenesis by determining the ratio between the longer and shorter nuclear axes at several developmental stages (Fig. 7A-C). As the retina folds, nuclei progressively adopt an elongated shape in wild type but not in opo mutant retinae. This suggests that nuclear morphology serves as read-out of the tension exerted on RPCs. Alternatively, Opo protein could have a structural role in determining nuclear shape. To distinguish between these possibilities, the inability of individual cells to transduce tension or the inability of opo mutant nuclei to modify their shape, we performed transplantation experiments. In mosaic retinae (upon transplantation of either wild-type or opo rx2::eGFP cells) nuclear morphology was analyzed at stages 24 to 26. In control homotypic transplantations, nuclei of transplanted cells showed no changes in their morphology. By contrast, we observed an altered, elongated nuclear morphology in opo mutant clones in a wild-type context(Fig. 7E;ratio=2.34±0.11). Moreover, when wild-type cells were transplanted into an opo mutant host, they retained their elongated nuclear shape and rescued nuclear morphology in adjacent mutant cells(Fig. 7F;ratio=2.27±0.14). We can therefore rule out a cell-intrinsic function for Opo in regulating nuclear shape. Instead, these experiments suggest that,in the appropriate context, mutant cells can transduce tension provided laterally by neighboring wild-type cells. The round nuclear morphology would therefore represent an intrinsic failure of opo mutant cells to apply tension. To explore this further, we followed cell dynamics in vivo using rx2::mYFP transgenics. Cell shape changes were followed by time-lapse confocal microscopy over 3 hours at stage 23. Cells in wild-type retinae showed very active behavior at their basal side. Both filopodia and basal surface contractions were observed (see Movie 3 and Fig. S8 in the supplementary material). These dynamic contractions were not observed in the mutants (see Movie 4 and Fig. S8 in the supplementary material), suggesting that tension cannot be exerted properly at the basal surface when opofunction is impaired (Fig. 7D).

Because of its size and accessibility, the eye represents a paradigm for vertebrate organogenesis (Spemann,1901). Here, we report the identification and characterization of the mutation opo, which affects epithelial morphogenesis and eye organogenesis in particular. Using time-lapse microscopy, we have characterized the series of events involved in retina folding, and have determined when and how this morphogenetic process is impaired in opomutants. We propose a three-step model for optic cup formation:anteroposterior folding, dorsoventral folding and optic fissure formation. The first morphogenetic event, which fails in opo mutants, folds the optic vesicle epithelium along its anteroposterior axis, bending the tissue ventrally so that the anterior and posterior borders converge. In a second,temporally overlapping event, the retinal tissue folds along the dorsoventral axis. This process is not affected in opo mutants, demonstrating that the two morphogenetic movements, although cooperative, have an independent genetic control. Once the hemispheric cup has formed, the third morphogenetic event involves the fusion of the optic cup borders and the ingression of cells from the ventral optic stalk region into the choroid fissure(Holt, 1980). Our analyses demonstrate that the folding of the optic cup and the formation of the optic fissure are consecutive events clearly separated in time. Several additional lines of evidence show that the early events shaping the cup and those forming the fissure later are genetically uncoupled. Although interference with the gene network that specifies the ventral retina often results in a failure to close the optic fissure (i.e. coloboma), these manipulations do not interfere with the formation of a hemispheric cup(Peters, 2002; Torres et al., 1996). Similarly, although optic disc formation is impaired in Bmp7-null mice because of a failure in the ventral ingression of Pax2⁺ cells, optic cup development occurs normally (Morcillo et al., 2006). Reciprocally, here, we show that the identity of the ventral retina is unaffected in opo mutant embryos despite the optic cup malformation.

Here, we describe the isolation of a mutation in the opo gene. A mutant splicing isoform causes the premature truncation of the protein and subsequently reduces protein levels in opo embryos. Further reduction of protein levels by morpholino treatment causes a progressively stronger and earlier phenotype that resembles the lack of vesicle evagination observed in Rx3 mutants (Loosli et al.,2001).

During optic cup formation, both opo transcripts and protein are basally localized, suggesting a role in cell polarity. We show that, despite the morphogenetic defects, apical marker localization, as well as cell proliferation and neuronal lamination (which both depend on apical identity),occur normally in opo mutants. This is in contrast to zebrafish mutants in which the localization of apical determinants is disrupted(Horne-Badovinac et al., 2001; Jensen and Westerfield, 2004; Malicki, 2004). Although cell cycle and lamination defects are reported in the retina of these mutants, the organ shape appears unaffected. Interestingly, some zebrafish apical-polarity mutants show morphogenetic defects in the neural tube(Lowery and Sive, 2004). This is in agreement with the proposal that apical constriction is the morphogenetic mechanism involved in vertebrate neurulation(Wallingford, 2005).

In contrast to apical constriction, the basal bending of an epithelial sheet is a less studied phenomenon. Recently, basal constriction has been proposed as the driving force for the morphogenesis of the midbrain-hindbrain boundary (Gutzman et al.,2008). Here, we propose an opo-mediated basal constriction as the mechanism responsible for optic cup morphogenesis. We describe that, at the basal side of opo mutant retinae, neuroblast morphology is abnormal and cortical actin levels are reduced. Pioneering work has described optic cup folding as an active process that is dependent on ATP and Ca²⁺, and potentially on a microfilament band(Brady and Hilfer, 1982). Our results show Opo protein accumulation at the ZNE, where cellular feet attach to the basal lamina forming a tight array of cytoplasmic extensions. Moreover,we show that the focal adhesion proteins integrin β1 and paxillin are not properly recruited to the ZNE in the mutants, suggesting that Opo functions to localize these crucial factors to the basal surface of RPCs. Specific interference with integrin-adhesive function indicates that focal contacts are required for optic cup folding, potentially by transmitting basolaterally the tensions that drive the epithelial bending. By analogy to what happens apically during neurulation, we propose that an active basal constriction,dependent on focal contacts, shapes the retinal epithelium into a hemispheric optic cup.

opo is present as a single copy in all of the vertebrate genomes that are available for searching. Blast and tblastn analyses against vertebrate genomes revealed two conserved regions: a 90-amino acids N-terminal motif and a 230-amino acid C-terminal motif that contains the predicted transmembrane passes. Whereas the C-terminal motif has homology with transcripts found in other metazoans (not protostomes), the N-terminal motif of the protein can be considered a vertebrate innovation. This fact and the fact that opo is expressed embryonically indicate that it is not a housekeeping gene but rather that it plays a regulatory role during development.

In humans, the opo ortholog sits in a large gene desert at the distal region of chromosome 6p. Deletions of this region cause a complex syndrome that includes mental retardation, abnormal brain sutures, heart defects, eye abnormalities and craniofacial malformations(Palmer et al., 1991). The extent to which the distal 6p deletion syndrome relates to the broader role of opo in epithelial morphogenesis requires further investigation. However there is already an established link between the opo ortholog in humans and the orofacial clefting syndrome(Davies et al., 2004). As medaka mutants also show strong craniofacial defects, they provide a model for understanding the underlying causality behind this common birth defect.