Otx genes are required for tissue specification in the developing eye

Vertebrate eye development proceeds through the co-ordinated activation of morphogenetic programs based on inductive interactions among tissues of different embryonic origin. The eye primordium comprises the optic vesicle (OV), a lateral protrusion of the diencephalic wall, and the lens placode, a thickening of the ectoderm overlying the vesicle. The progressive determination of this primordium further originates functionally specialised eye tissues: the optic stalk (OS), the neural retina (NR), the retinal pigment epithelium (RPE) and the crystalline. Recent studies have demonstrated that all these eye developmental processes are controlled by a complex network of regulatory genes many of which have been highly conserved throughout evolution (Jean et al., 1998; Gerhing and Ikeo, 1999). In particular, the specification of eye compartments appears to be controlled by the mutual regulation and the progressive restriction of the expression domains of a number of transcriptional regulators initially coexpressed within the eye anlage (Macdonald et al., 1995; Schwarz et al., 2000; Nguyen and Arnheiter, 2000). This is the case for instance of Pax2, Pax6, or Mitf which are initially expressed throughout the OV and later become restricted to the OS, NR and RPE, respectively (Nornes et al., 1990; Walter and Gruss, 1991; Torres et al., 1996; Nakayama et al., 1998; Schwarz et al., 2000). Analysis of Pax2 and Pax6 null mutant mice demonstrated that the activity of both genes is necessary not only to establish the identity but also the border between OS and optic cup (OC; Hill et al., 1991; Torres et al., 1996; Schwarz et al., 2000). Furthermore, in mice with mutation in the Mitf locus, patches of the RPE lose their characteristics, and acquire features of NR (Nakayama et al., 1998; Bumsted and Barnstable, 2000; Nguyen and Arnheiter, 2000).

Expression analysis of Otx2, one of the two vertebrate homologous of the Drosophila cephalic gap gene orthodenticle (otd; Finkelstein and Perrimon, 1990; Simeone et al., 1992; Simeone et al., 1993) had lead to the proposal that Otx2 might also participate in the establishment of the eye territories (Bovolenta et al., 1997). Even though little is known about the expression of Otx1 during vertebrate eye development, both Otx1 and Otx2 are initially expressed throughout the OV (Simeone et al., 1993). Later Otx2 becomes specifically restricted to the dorsal portion of the vesicle, the presumptive RPE territory (Bovolenta et al., 1997). After that regional specification of the eye is achieved, a second wave of Otx2 expression was found also in the NR. In particular, OTX2 was localised to postmitotic neuroblasts committed to both neuronal and glia cell types (Bovolenta et al., 1997). In spite of the critical distribution of Otx genes in the developing eye, analysis of Otx1 and Otx2 mutant mice has only given limited information about how these genes function during eye formation. Otx1^−/− mice lack ciliary processes and lachrymal glands (Acampora et al., 1996), whereas Otx2^+/− mice are either normal (Acampora et al., 1995) or lack lens, cornea and iris or display anophthalmia or microphthalmia (Matsuo et al., 1995), depending on the genetic background. A complete analysis of Otx2 activity in the eye is impaired by the high embryo lethality due to gastrulation defects and by the severe head malformation present in Otx2 null homozygotes (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996).

To understand Otx functions in general it is important to consider the particular functional interaction between the two genes. Although Otx genes are responsible for different aspects of anterior vertebrate brain morphogenesis, a different transcriptional control rather than divergent protein activities appears to be responsible for the distinct defects found in Otx1 and Otx2 null mice (Simeone, 1998; Acampora and Simeone, 1999). Thus, the proteins encoded by the two genes have, at least in part, similar biochemical activities, since replacement of Otx1 gene with Otx2 cDNA and vice versa could partially rescue the respective knock-out phenotype (Acampora et al., 1998; Acampora et al., 1999; Suda et al., 1999). In agreement with this idea, the two Otx genes appear to cooperate in some aspects of vertebrate brain morphogenesis since a minimum level of OTX protein is required to specify the isthmic organizer position and, therefore, the further patterning of the anterior brain (Acampora et al., 1997; Suda et al., 1997). Because a similar synergistic action of Otx genes is necessary for the normal development of the mouse inner ear (Morsli et al., 1999), we have postulated that cooperation of these genes might be also necessary for eye development.

Here we have analysed the eye phenotype of Otx-deficient mice. We show that in all Otx1^−/−; Otx2^+/− and in 30% of Otx1^+/−; Otx2^+/− genotypes, the development of all the tissues derived from the OC were severely affected. The failure of the RPE differentiation was the major phenotypic malformation observed in these mutant embryos. Lack of RPE was associated with an expansion of the prospective NR and OS territories. Based on these findings we propose a model in which Otx proteins are required for the determination of the RPE, cooperating with other eye patterning genes in the specification of the different tissues derived from the vertebrate OV.

Otx-deficient mice were generated by crossing Otx1^+/−; Otx2^+/− males with Otx1^+/−; Otx2^+/+ females in a BL6/DBA2 background and the offspring were further genotyped as described previously (Acampora et al., 1995; Acampora et al., 1996; Acampora et al., 1997). Wild-type BALB/c mice were used in Otx2-PAX2 double labelling experiments. The day of the appearance of the vaginal plug was considered as embryonic day (E)0.5.

Digoxigenin-labelled antisense RNA probes for the mouse tyrosinase, Mitf, Prox1 and Otx2 genes were prepared as described (Beerman et al., 1992; Hodgkinson et al., 1993; Oliver et al., 1993; Bovolenta et al., 1997). The ³⁵S-labelled RNA probes for Otx1, Otx2, Pax2, Pax6 and Six3 genes were generated as reported (Acampora et al., 1997).

Experiments using ³⁵S-labelled RNA probes were performed as previously described (Wilkinson, 1992). Digoxigenin-labelled hybridisations were carried out on tissue sections as described (Bovolenta et al., 1997). Briefly, embryos were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB; pH 7.3) at 4°C overnight, and then cryoprotected by immersion in 30% sucrose solution in PB. Cryostat sections (18 μm) were mounted on 2% 3-aminopropyltriethoxy-silane-coated slides, air dried, and treated with methanol for 10 minutes at room temperature. After rinses in phosphate-buffered saline (PBS), sections were prehybridised for 1 hour at 65°C in 50% formamide buffer, incubated with digoxigenin probes for 16 hours at 65°C, and further washed at the same temperature. After incubation in 0.5% Boehringer blocking reagent solution for 1 hour, sections were incubated with AP-coupled anti-digoxigenin antibody (Roche) diluted 1:5000. After the AP reaction was revealed, sections were coverslipped and analysed in a Leica DMR compound microscope.

Polyclonal antibodies against Islet1 (generated in guinea-pig), phosphorylated-histone H3 (P-H3; Upstate) PAX2 (Zymed) and α-A-crystallin were used at 1:5000, 1:1000, 1:2000 and 1:3000 dilution respectively. Polyclonal antiserum against the mouse OTX2 protein (Mallamaci et al., 1996) was used at 1:2000. Monoclonal antibodies against phosphorylated β-tubulin (Tuj1) and MITF (Neomarkers) were used at 1:4000 and 1:1000 dilution, respectively.

Embryo heads were fixed by immersion in 4% paraformaldehyde in PBS overnight at 4°C. After washing in PBS, the tissue was equilibrated in 30% sucrose in PBS, embedded in OCT compound and sectioned on a cryostat at 16 μm. Alternatively, embryos were dehydrated, embedded in Paraplast and sectioned on a microtome (Leica) at 16 μm. Tissue was permeabilised in methanol containing 10% DMSO for 10 minutes, rinsed in PBS and blocked for 1 hour in 10% FCS in PBS containing 0.1% Tween (PBTF). Sections were incubated with each primary antibody diluted in PBTF overnight at 4°C. Primary antibodies were localised with either goat anti-mouse, rabbit (Jackson ImmunoResearch) or guinea-pig (Chemicon International) biotin-conjugated IgG, followed by either Cy3 (Amersham-Pharmacia) or peroxidase-conjugated streptavidin (Jackson ImmunoResearch). Fluorescently immunolabelled sections were mounted with PBS/glycerol and analysed with a laser scanning confocal imaging system (TCS 4D; Leica). Peroxidase-coupled samples were revealed using NovaRed substrate kit (Vector Laboratories), dehydrated, coverslipped and analysed in a Leica DMR compound microscope.

Quantitative data are expressed as mean ± s.e.m. Significant differences among groups were evaluated by an unpaired t test (GraphPad Prism).

The detailed expression of Otx2 during eye development has been reported previously (Simeone et al., 1993; Bovolenta et al., 1997), in contrast to that of Otx1 which has been poorly described. Because this was a necessary piece of information to understand the phenotype of Otx-deficient mice, we have analysed Otx1 distribution by situ hybridisation in mouse embryos. Otx1 transcript localisation was compared on consecutive sections with that of Otx2 and Pax2, dorsal and ventral markers of the OV, respectively (Nornes et al., 1990; Bovolenta et al., 1997). At E9.5-E10 Otx1 mRNAs were restricted to the dorsal portion of the OV. This distribution was overlapping with that of Otx2 and complementary to that of Pax2, which was confined to the vesicle ventral half (Fig. 1A-F). Later (E10.5-E16), Otx1 was clearly localised to the presumptive RPE (Fig. 1J,M,P,S), as was Otx2. At these stages Pax2 expression was progressively limited to the ventral OS (Fig. 1; Nornes et al., 1990). Otx1, but not Otx2 expression was also evident in the peripheral NR and in the dorsal half of the developing optic nerve (Fig. 1J-U). A similar Otx1 expression pattern was observed in the developing chick eye (not shown).

Mice carrying different functional copies of Otx1 and Otx2 genes were generated by mating Otx1^+/−; Otx2^+/− double heterozygous males with Otx1^+/−; Otx2^+/+ females, as described (Acampora et al., 1997). A total of 68 embryos were analysed at different developmental stages (Table 1). Among these, all Otx1^−/−; Otx2^+/− embryos (n=24) presented gross eye malformations that ranged from unilateral anophthalmia to severe abnormalities in the lens, NR and RPE (Table 1; Fig. 2C,D). Among the other genotypes, 30% of the Otx1^+/−; Otx2^+/− embryos (n=13) showed similar but less severe malformations (Table 1). The remaining viable genotypes did not presented gross eye malformations (Table 1), but as previously reported (Acampora et al., 1996), mild ciliary defects were observed in Otx1^−/− mice from E15 onwards.

Fig. 2 illustrates the typical ocular phenotype of Otx1^−/−; Otx2^+/− neonatal mice (mice die soon after birth), as determined by histological analysis in Cresyl Violet stained sections. As compared to wild type (Fig. 2A,B), mutant mice consistently presented alterations in four different eye structures (Fig. 2C,D). The lens was either absent (Fig. 8B; Table 1) or reduced in size (Fig. 7) and was usually mislocated with a 90° rotation. The NR was dysplastic and folded. Retinal ganglion cell (RGC) axons that normally coalesce in the optic nerve head exited the eye in disorganised bundles following aberrant trajectories around the external surface of the eye (Fig. 2D). The last and most dramatic alteration was, however, the almost complete absence of a morphologically recognisable RPE. This tissue was replaced by a thickened neuroepithelium with a NR-like morphology (Fig. 8B). In several embryos, a dorsal patch of pigmented tissue (Fig. 2D) constituted the only remnant of the RPE.

Therefore, mice with the minimal dosage of OTX proteins compatible with viability presented a clear failure in the proper patterning of the eye.

To understand the morphological and molecular mechanisms responsible for the abnormal ocular phenotype described above, we have analysed the expression of different genes at critical stages of eye development. In mice, OV are well developed at E9.5. At this stage, OV outpocketing appeared normal in all Otx mutants analysed, including those with the lower Otx dosage (Fig. 3). We next tested for possible alterations in the expression of genes that may contribute to the regionalisation of the OV, such as Pax2, Pax6 and Six3 (Hill et al., 1991; Torres et al., 1996; Oliver et al., 1995; Bovolenta et al., 1998; Schwarz et al., 2000). The distribution of these genes, together with that of the remaining Otx2 allele(s), was analysed in the different Otx-deficient genotypes by in situ hybridisation on consecutive sections (Fig. 3). In wild type and Otx1^−/− mice (Figs 3A,B, 1C,F), the expression of Otx2 was limited to the dorsal half of the vesicle in the presumptive RPE territory, as previously reported (Simeone et al., 1993; Bovolenta et al., 1997). In contrast, in Otx1^+/−; Otx2^+/− and Otx1^−/−; Otx2^+/− embryos Otx2 mRNA was uniformly distributed at low level throughout the entire vesicle, including its ventral portion (Fig. 3C-E). Similarly, Pax2 transcripts, which normally become restricted to the ventral half of the OV at this stage (Figs 1B,E, 3F,G; Nornes et al., 1990), were still expressed in more dorsal locations in embryos carrying a single functional copy of Otx2 (Fig. 3H-J). None of the Otx-deficient genotypes presented significant differences in Pax6 or Six3 ocular expression as compared to wild-type embryos (Fig. 3K-T).

Mitf, a helix-loop-helix-zipper transcription factor expressed in the eye field (Hodgkinson et al., 1993), becomes specifically localised to the presumptive RPE early in eye development (Nguyen and Arnheiter, 2000; Fig. 4D). As shown by immunocytochemical detection of both proteins, MITF and OTX2 distribution overlaps in the dorsal OV (Fig. 4A,D). In Otx1^+/−; Otx2^+/− and Otx1^−/−; Otx2^+/− embryos instead, MITF localisation was not restricted to the outer layer of the OV (Fig. 4E,F), following the abnormal expression pattern of Otx2 (Figs 4B,C, 3C-E).

Altogether these data indicate that a single copy of Otx2 is sufficient for both the specification of the eye field in the anterior neural plate and the evagination of the OV from the diencephalic wall. A low level of OTX proteins is however insufficient to achieve the proper segregation of the expression domain of vesicle patterning genes, such as Pax2 or Mitf.

Soon after evagination, the OV begins to fold over the lens vesicle. The OC, which is fully formed in mice at E12.5, is the result of these morphogenetic movements. At this stage the outer layer of the cup will start to acquire characteristics of the future RPE forming a monolayer of cuboid cells that present the restricted expression of both Otx1 and Otx2 (Figs 1, 4G). In addition, the outer layer of the cup express other RPE-specific genes such as Mitf (Fig. 4J) and tyrosinase, a melanin synthesing enzyme (Fig. 4M; Beerman et al., 1992).

The folding of the OV into OC was severely altered in all Otx1^−/−; Otx2^+/− and in 33% of the Otx1^+/−; Otx2^+/− embryos analysed. In these mutants, most of the outer layer of the developing OC appeared as a thickened neuroepithelium (Figs 4K,L,N,O, 5B,D,F,H) similar to that of the prospective NR. In this abnormal layer of the cup, OTX2 as well as Mitf and tyrosinase mRNAs were completely undetectable (Fig. 4). In both mutants only small patches of the outer layer acquired RPE characteristics. Even though cells did not form a cuboid epithelium (insets in Fig. 4G-I), co-expression of OTX2, Mitf and tyrosinase was always found in these patches (Fig. 4).

In wild-type mice, the transition from OV to OC results in the restriction of the expression of Pax6, Six3 and Pax2 to specific compartments. Thus, Pax6 and Six3 transcripts become localised to the presumptive NR (Fig. 5A,C; Walther and Gruss, 1991; Oliver et al., 1995; Bovolenta et al., 1998), while those of Pax2 to the OS territory (Fig. 5E; Nornes et al., 1990; Torres et al., 1996). In Otx1^−/−; Otx2^+/− embryos the expression of Pax6 and to a lower level that of Six3 was abnormally extended to the entire OC (Fig. 5B,D). Similarly, the expression of Pax2 and that of its product are limited to the entire ventral region of the developing OC (Fig. 6A-C) and later only to the stalk region of the fully formed cup (Figs 5E, 6B,D), both Otx2-negative territories (Fig. 6A,B). In contrast, as already suggested by analysis at E9.5, Otx1^−/−; Otx2^+/− embryos presented an enlargement of the Pax2 domain which was first expanded dorsally in E10.5 OC (Fig. 6E) and subsequently extended over the RPE territory at E12.5 (Fig. 6F).

Altogether the expression profiles of these genes suggest that in Otx-deficient mice RPE specification does not occur. The outer layer of the OC acquires, instead, features of the prospective NR from one side and that of the OS from the other.

Lens formation proceeds through successive phases of specification, which involves the interaction of the presumptive lens ectoderm (PLE) with the newly evaginated OV (Grainger 1996; Grainger et al., 1997; Jean et al., 1998). Both Otx1 and Otx2 genes are expressed in the vertebrate PLE as well as in the OV (Fig. 1; Simeone et al., 1993; Bovolenta et al., 1997; Zygar et al., 1998). In Otx1^−/−; Otx2^+/− mice, lens development was clearly abnormal, as better appreciated by the detection of the expression of α-A-crystallin and Prox1 (Fig. 7), two lens specific markers (Oliver et al., 1993; Oguni et al., 1994). At E10.5 the lens pit was formed in wild-type embryos and the elongated ectodermal cells expressed α-A-crystallin (Fig. 7A). In contrast, in Otx1^+/−; Otx2^+/− and Otx1^−/−; Otx2^+/− embryos the lens vesicle was either poorly invaginated, expressing low levels of α-A-crystallin, or remained as an ectodermic α-A-crystallin-negative placode (Fig. 7B,C). From E12.5 onward, when the lens vesicle is formed, all the wild type and Otx1^+/−; Otx2^+/− and 88% of the Otx1^−/−; Otx2^+/− mice presented lens tissue, expressing both α-A-crystallin and Prox1. In the mutants, lens tissue was however reduced (Fig. 7D-I). These results suggest that a delay or a failure in the progression of lens development could account for its absence or reduced size observed in Otx-deficient mice.

We next asked whether in mutant mice the outer layer of the OC could differentiate according to its new identity of ‘prospective NR’. In wild-type embryos, many tubulin β3 and Islet1-positive RGC and amacrine cells have exited the cell cycle and have migrated to their proper layer at E17.5 (Fig. 8A,G; Young, 1985; Austin et al., 1995). Staining of equivalent sections of E17.5 Otx1^−/−; Otx2^+/− embryos demonstrated that tubulin β3- and Islet1-positive cells accumulated not only on the vitreal surface of the folded mutant NR, but also on the sclera surface of the ‘ectopic’ NR (Fig. 8B,C,H,I). In addition, these ectopically differentiated neurones appear to extend neurofilament- (not shown) and tubulin β3-positive fibres, which accumulated in aberrant positions at the sclera surface of the eye (Fig. 8B,C). This inverted configuration of the ‘newly generated retina’ was confirmed by staining with anti-Ph-H3, a mitotic marker (Fig. 8D-F). As a general rule in the CNS, neuroepithelial cells undergo cell division in the ventricular layer, whereas differentiating neurones migrate towards the pial surface (Sauer, 1935; Jacobson, 1968). Therefore, the mirror image organisation of the two ‘neural retinas’ in the mutant mice reflects the inverted orientation of the outer in relation to the inner neuroepithelial layer of the OC.

In a second wave of expression, both Otx1 and Otx2 are expressed in the NR when its differentiation begins (Fig. 1; Simeone et al., 1993; Bovolenta et al., 1997). In particular, Otx2 expression was observed in postmitotic neuroblasts committed to both neuronal and the glia cell types (Bovolenta et al., 1997). We therefore asked whether retina neurogenesis was affected in Otx1^−/−; Otx2^+/− mice.

In all developing structures, neurogenesis is controlled by the balance between cell proliferation, differentiation and apoptosis. In retinas from E18.5 wild-type embryos, Ph-H3-positive mitotic cells were located, as expected, in the ventricular layer (Fig. 9A). In the corresponding retina sections of mutant mice the number of P-H3-positive cells appeared greatly increased throughout the folded NR (Fig. 9B,C). Apoptotic cell death, identified by Cresyl Violet staining (Sengelaub and Finlay, 1982), was not a prominent feature in wild-type retina at this stage but was strikingly frequent in both the RGC and inner nuclear layers of the mutants (Fig. 9D-F). In mice, RGC and amacrine cells represent the majority of differentiated cells in E18.5 retinas (Young, 1985). These cell types, labelled by antibodies against islet-1 were quite abundant in their respective layers in wild-type embryos (Fig. 9G), while a clear decrease in positive cells was evident in mutant retinas (Fig. 9H-I). These defects were detected in 25% of the Otx1^+/−; Otx2^+/− and in all the Otx1^−/−; Otx2^+/− animals analysed at the stages from E17.5 to P1.

Statistical analysis of the data described above was performed on equivalent retinal sections from wild-type and Otx1^−/−; Otx2^+/− mice (n=4). As clearly shown in Fig. 9C,F,I, in the presence of a single Otx2 allele, retina cell proliferation and cell death were dramatically increased, while cell differentiation was severely impaired. An easy interpretation of these data is however obscured by the indications that NR folding and stratification depends on the presence of the lens (Ashrey-Padan et al., 2000; Yamamoto and Jeffery, 2000) while its differentiation and survival relies on signals generated from the RPE (Raymond and Jackson, 1995; Bovolenta et al., 1996; Frade et al., 1996). Therefore, the precise role of OTX proteins during NR differentiation needs to be established in a different experimental model where lens and RPE will be not affected.

Our analysis of the ocular phenotype of mice carrying different functional copies of Otx1 and Otx2 provide evidences that a minimal dose of Otx genes is required for the proper development of the eye. Otx genes are necessary for OV folding and for the establishment of the RPE identity. In the presence of a minimal level of OTX protein, the RPE-specific genes Mitf and tyrosinase are not activated in the outer layer of the OC. Instead, this region will go on to differentiate as an ‘ectopic’ NR. Additional defects include the alterations in size and position of the lens and an abnormal rate of cell proliferation, differentiation and death in the NR proper.

We have shown here that eye development in the mouse depends on the number of functional copies of Otx. Thus, 100% of the embryos carrying the minimum Otx dosage compatible with viability (Otx1^−/−; Otx2^+/−) presented gross eye malformations, whereas only 30% of the double heterozygous (Otx1^+/−; Otx2^+/−) embryos showed similar phenotypes. Otx genes cooperate in the specification of the mesencephalic-metencephalic territories and in the development of the inner ear. However, the threshold of OTX proteins required for proper eye development appears to be higher than those essential for the specification of these regions, given that no inner ear or brain defects were found in double (Otx1^+/−; Otx2^+/−) heterozygous embryos (Acampora et al., 1997; Suda et al., 1997; Morsli et al., 1999). In addition, our data indicate that eye development depends on the identity of the functional Otx copies present in the mice. Thus, early gross eye malformations, including lens defects, are present in Otx1^+/−; Otx2^+/− but not in Otx1^−/− embryos. In both cases, only two Otx alleles are functional but the lack of one of the Otx2 alleles seems more relevant to eye patterning than that of Otx1. In agreement with this, about 70% of the Otx2 heterozygous mice, generated on a different genetic background from those analysed in this study, presented microphthalmia or anophthalmia and agenesis of the lens (Matsuo et al., 1995). In the case of the lens, the preponderant role of Otx2 activity is further sustained by the observation that in Xenopus, Otx2 expression in the presumptive lens ectoderm is a prerequisite for the expression of Pax6 and Sox, genes fundamental for lens determination (Zygar et al., 1998). However, and independently of the relative importance of Otx1 or Otx2, it is unclear whether lens defects in Otx1^−/−; Otx2^+/− are a totally autonomous process or occur as a secondary event due to the improper patterning of the neuroepithelium, a potential source of lens-inductive signals (Karkinnen-Jaaskelainen, 1978; Porter et al., 1997; Furuta and Hogan, 1998).

Evidence from several sources support the idea that Otx genes might be involved at different steps of vertebrate eye formation, including the determination of a field permissive for eye formation, the morphogenesis of the lens and OC, and the generation of retina-specific neurones (Bovolenta et al., 1997; Simeone, 1998; Acampora and Simeone, 1999; Chow et al., 1999; Loosli et al., 1999; this study).

In all bilaterian species so far examined, the formation of the anterior neuroectoderm and, therefore, that of the eye field, which lies within the Otx expression domain (; Simeone et al., 1992; Li et al., 1994; Loosli et al., 1998, Bruce and Shankland, 1998; Mitsunaga-Nakatsubo et al., 1998; Stornaiuolo et al., 1998; Wada and Saiga, 1999), depends upon Otx2 function (Acampora et al., 1995). Only in this Otx2-positive anterior neuroectoderm, can the overexpression of Pax6 and Six3 induce the formation of ectopic eyes in vertebrates (Chow et al., 1999; Loosli et al., 1999), indicating that Otx activity provides the anterior neuroectoderm with the necessary competence for ocular specification. We show here that, despite of the subsequent gross eye malformations observed in Otx1^−/−; Otx2^+/− mice, OV outpocketing occurs normally, suggesting that a single copy of Otx2 is sufficient to establish the eye morphogenetic field and to allow its evagination.

Subsequent to eye field specification, the morphogenesis of the vertebrate eye takes place by the infolding of the undifferentiated eye primordia into a bi-layered cup. In this structure, both the basic architecture of the adult organ and the identity of the different tissues become established by the coordinated action of autonomously expressed transcription factors and inductive signals. In the presence of low levels of Otx activity, folding of the OV is severely affected. In Otx mutants, the delayed or abnormal development of the lens placode and/or the low levels of OTX protein in the lens ectoderm (Hyer et al., 1998; Nguyen and Arnheither, 2000) could be in part responsible for the morphological defect of the OV. Similar abnormal OV invagination has indeed been observed in a lens-specific Pax6 conditional mouse mutation, where the absence of the lens placode has permitted the analysis of its effects on vesicle morphogenesis (Ashrey-Padan et al., 2000). Alternatively, OTX protein activity could be directly implicated in the control of morphogenetic movements of the cells. Association between cell migration and Otx2 expression has been proposed in the mouse olfactory system (Mallamaci et al., 1996) and gastrulation movements are severely affected in Otx2^−/− mice (Simeone, 1998). Furthermore, Otx2 appears to control the expression of molecules implicated in cytoskeleton organisation and cell-cell interaction, such as calponins, R-cadherin or ephrinA2 (Rhinn et al., 1999; Morgan et al., 1999) and the ectopic expression of Otx1 and otd results in strong cell aggregation in zebrafish (Bellipanni et al., 2000). Therefore, it is possible that alterations of similar molecular pathways are responsible for aberrant morphogenesis of the OC in Otx1^−/−; Otx2^+/− and Otx1^+/−; Otx2^+/− embryos.

In addition to morphogenetic defects, Otx-deficient mice show altered expression of genes involved in the proper subdivision of the OV territory into RPE, NR and OS. Soluble molecules derived from adjacent tissues (the surface ectoderm and the surrounding mesenchyme) appear to influence NR versus RPE fate decision. Fibroblast growth factors, normally expressed in the PLE, can determine neural retina identity even in tissue fated to give rise to RPE (Pittack et al., 1991; Pittack et al., 1997; Guillemot and Cepko, 1992; Hyer et al., 1998; Zhao and Overbeek, 1999; Nguyen and Arnheither, 2000). In contrast, extraocular mesenchyme is the source of molecules, possibly activin-like signals, capable of repressing NR while activating RPE markers (Fuhrmann et al., 2000). The regulation of tissue-specific genes might therefore be the results of these (positive and negative) inductive signals. However, the suppression of PLE development does not affect the formation and the separation of NR and RPE progenitor cell (Ashery-Padan et al., 2000), giving strong support to our finding that the loss of functional alleles in Otx mutants is the direct cause of the failure of RPE determination. Interestingly, in Otx mutants the expression of Mitf and tyrosinase is also largely absent and maintained only in little patches of tissue where OTX2 is also localised. In Mitf mutants the expression of Otx2 is specifically down regulated in those areas where RPE has not differentiated (Nguyen and Arnheiter, 2000). This suggests that both transcription factors cooperate to determine the identity and possibly maintain the function of the RPE. However, we believe that Otx genes have a prevalent role in this process, because the expression of Otx precedes that of Mitf (Nguyen and Arnheiter, 2000). In addition, in most cases the entire RPE territory is affected in Otx-deficient mice, whereas only patches of the dorsal RPE is affected when the entire Mitf gene is lost (Mitf^vga-9/vga-9; Nguyen and Arnheiter, 2000).

The enlargement of the expression domains of OS- (Pax2) and NR- (Pax6 and Six3) specific genes further suggests the importance of Otx activity in the proper subdivision of the OV. Whether OTX proteins repress the expression of Pax2, Six3 or Pax6 directly or indirectly needs to be determined. However, a recent report shows that the identity of the OS and OC depends on the specific expression and reciprocal transcriptional repression of Pax2 and Pax6 (Schwarz et al., 2000). Similar mechanisms may be acting in the segregation of other genes involved in the establishment of RPE/OS and RPE/NR boundaries. A hypothetical model that takes into account all the data described above is schematically illustrated in Fig. 10. In this simplified scheme, the expression, regulated respectively by TGFβ-like and Shh signalling (McDonald et al., 1995), and the reciprocal repression of Otx and Pax2 will determine, in coordination with other genes as, for instance, Vax1 and Vax2 (Barbieri et al., 1999; Bertuzzi et al., 1999; Hallonet et al., 1999), the initial dorsal and ventral patterning of the OV (Fig. 10A,C). In this respect, we show that initially only the expression of Pax2, but not that of either Six3 or Pax6 is significantly altered in Otx-deficient mice. In addition, in Pax2 null mice, pigmented cells invade the optic stalk (Torres et al., 1996). In support of a possible antagonistic interaction between Otx and Pax2, it is also worth mentioning that the two genes may be acting in a similar antagonistic manner to establish the pattern of the inner ear primordium. Otx and Pax2 genes are expressed in the otic vesicle in the presumptive vestibular and cochlear domains, respectively (Morsli et al., 1999; Hidalgo-Sanchez et al., 2000). Furthermore, Pax2^−/− and Otx mutants (Otx1^−/−, and Otx1^−/−; Otx2^+/−) show divergent otic phenotypes, lacking the auditory (cochlea and spiral ganglion) and the vestibular (semicircular canal, lateral ampulla, utricosaccular and cochleosaccular duct) parts of the organ, respectively (Torres et al., 1996; Morsli et al., 1999).

Later, as the RPE, NR and OS progenitor cells acquire their identity, the mutual positive interaction between Otx and Mitf in the dorsal portion of the eye will induce and probably maintain the network of genes necessary for the establishment RPE identity (Fig. 10B,D; Bentley et al., 1994). Complete segregation of eye territories would also imply the mutual regulation of different OS, NR and RPE specific genes. Thus, regulation of Otx2 and Pax2 expression in cells located at the RPE/OS border will finally establish the separation of the two territories with a mechanism similar to that proposed for Pax2 and Pax6 interaction at the OS/OC boundary (Schwarz et al., 2000). In Otx mutants, Pax6 expression appeared unaffected in spite of the expansion of the Pax2-positive territory. Whether this is due to an independent autoregulation of Pax6 expression (Schwarz et al., 2000) or whether OTX proteins could also contribute to the direct regulation of Pax6 expression is not clear. Indeed, the precise hierarchical interactions among all the genes required for the OV subdivision are just beginning to be elucidated.

We are grateful to Drs H. Arnheiter, L. Montoliu and G. Oliver for gifts of the mouse Mitf, tyrosinase and Prox1 probes, respectively. Antiserum to OTX2 and anti-α-A-crystallin were kindly provided by Profs. G. Corte and H. Kondoh. We are in debt with MariCarmen Romero for help with sectioning and with Dr E. Martí for critical reading of the manuscript. We wish to thank Dr A. Nieto for the use of her Olimpus Camera. This study was supported by grants from the Spanish Dirección General de Enseñanza (PM97-0019) and the EU Biotech Program (BIO4-CT98-0399) to P. B. and the MRC (G8900955), the Italian Association for Cancer Research (AIRC), the EU Biotech Program (BIO4-CT98-0309) and the Murst-CNR Biotechnology Program Legge 95/95 to A. S. J. R. M. M. holds a postdoctoral fellowship from the Comunidad Autónoma de Madrid (02/0198/99).