Wnt2b/β-catenin-mediated canonical Wnt signaling determines the peripheral fates of the chick eye

Organ formation requires the establishment of pattern in relatively broad domains, leading to specific fate choices by individual cells, along with coordinated proliferation. The vertebrate eye is an excellent tissue in which to investigate these processes (Chow and Lang, 2001; Livesey and Cepko,2001; Zaghloul et al.,2005). The eye field comprises groups of cells biased to develop into various eye structures. It is specified in the anterior neural plate, and evaginates to form the OV, which in turn invaginates to form the OC. The OC is composed of two epithelial layers, with the outer layer becoming the retinal pigmented epithelium (RPE) and the inner layer becoming the neural retina (NR)(Chow and Lang, 2001; Martinez-Morales et al.,2004). A third region, the peripheral rim of the OC, forms two of the peripheral support tissues, the ciliary body and the iris(Beebe, 1986). These tissues are composed of a non-pigmented inner layer, which is continuous with the NR,and a pigmented outer layer, which is continuous with the RPE. The ciliary body is divided into two regions, a non-folded, pars plana, and a folded, pars plicata. In addition to the cells originating from the OC, and thus the neural tube, neural crest-derived mesenchymal cells contribute to the formation of the ciliary folds, forming connective tissues and ciliary muscles(Beebe, 1986). The ciliary body produces aqueous humor, which is crucial for maintaining the intraocular pressure of the anterior chamber, and regulates the shape of the lens,allowing it to focus on objects at various distances. In addition, the pigmented layer of the ciliary margin (PCM) appears to harbor adult retinal stem cells (Ahmad et al., 2000; Moshiri et al., 2004; Tropepe et al., 2000). Cells in the pupillary margin have another interesting feature. These cells undergo a unique form of transdifferentiation when they lose their pigment and differentiate into the smooth muscle cells of the pupillary sphincter(Link and Nishi, 1998; Volpe et al., 1993). Several genes, encoding secreted proteins, transcription factors and extracellular matrix proteins, have been shown to play a role in the development of the ciliary body and iris (Dhawan and Beebe,1994; Hsieh et al.,2002; Kubota et al.,2004; Pressman et al.,2000; Thut et al.,2001). However, the fundamental molecular mechanisms and timing of the specification of these tissues are unknown.

Wnt signaling has been implicated in cell cycle control in some regions of the central nervous system, through control of the cell cycle machinery(Megason and McMahon, 2002). In other tissues, Wnts also provide instructive signals for cell fate specification. For example, Wnt1 in neural crest stem cells determines sensory neural fate in a β-catenin-dependent manner(Lee et al., 2004). Transient expression of the stabilized or constitutively active (CA) β-catenin in cortical progenitor cells using the enhancer element of Nestinproduced a folded cortex phenotype, resulting from excess cell proliferation(Chenn and Walsh, 2002). However, recent studies using Nestin::Cre lines to activate or inactivateβ-catenin showed that canonical Wnt signaling is required for the establishment of the dorsal identity of the pallium during forebrain development (Backman et al.,2005). If a primary role for canonical Wnt signaling in the developing forebrain is patterning, proliferation might be a secondary effect.

The expression of Wnt signaling genes in the OC has been characterized. Wnt2b, Fz4 and Lef1, among others, are expressed in the peripheral part of the OC and lens (Jasoni et al.,1999; Kubo et al.,2003). Based upon the expression pattern, and on studies of function, several roles have been ascribed to Wnt2b in the eye. Wnt2b signaling was proposed to regulate proliferation in the lens, based upon its expression in the proliferative lens epithelium(Jasoni et al., 1999). In addition, Wnt2b was shown to have a role in the formation of laminae in the retina (Nakagawa et al.,2003). In Xenopus, Frizzled 5 is expressed within the ciliary marginal zone (CMZ), flanking the region containing stem cells, and regulates the transition of progenitor cells to neural precursors via the activation of Sox2 (Van Raay et al.,2005). Early morphogenesis, including eye-field formation, also requires the coordination of Wnt/β-catenin signaling, mediated by multiple Wnt ligands and receptors(Cavodeassi et al., 2005). Using a clonal assay in retinal reaggregation cultures, along with overexpression studies, it was proposed that Wnt2b plays a role in the maintenance of retinal progenitor cells (RPCs) in the CMZ of the chick retina(Kubo et al., 2003). However,three observations suggest that the canonical Wnt signaling pathway is involved in other aspects of eye development as well. One observation is that in situ hybridization (ISH) for Wnt2b shows a dynamic expression pattern,suggesting an involvement in several aspects of eye development. Wnt2b is exclusively expressed within the dorsal surface ectoderm (SE) at the OV stage,and in both the SE and RPE at the early OC stage. At later stages, Wnt2b is expressed solely in the iris epithelium, both in the pigmented and non-pigmented layers, but not in the RPE. The second is that activation ofβ-catenin did not promote retinal progenitor cell proliferation in an explant experiment (Ouchi et al.,2005). The third observation is that reporter assays indicate that the highest level of signaling through the canonical Wnt pathway occurs in the RPE and in the tip of the OC (Liu et al.,2003). These two regions have relatively low proliferation rates compared with the retina (Kubota et al.,2004), which has a high proliferation rate, but low Wnt reporter activity, in keeping with Beebe's observation of a low rate of proliferation in the cells at the anterior rim of the OC(Beebe, 1986). Similarly, wg or constitutively active armadillo, the fly homolog ofβ-catenin, do not induce excess proliferation when overexpressed in the developing Drosophila eye. Rather, they induce the formation of peripheral tissues, such as the head cuticle or pigment rim, surrounding the eye (Baonza and Freeman, 2002; Tomlinson, 2003; Treisman and Rubin, 1995).

In this study, we investigate the function of Wnt2b/β-catenin in the developing chick eye. In contrast to two previous reports of Wnt2b showing an increase in the proliferation of retinal cells in vitro(Kubo et al., 2003; Kubo et al., 2005), we do not find an increase in retinal cell proliferation when Wnt signaling is activated in the retina in vivo. Instead, we found that CA-β-catenin(Funayama et al., 1995) or Wnt2b interferes with the maintenance of retinal progenitor identity, and leads to the conversion of retinal cells into the peripheral fates of the ciliary body/iris. Furthermore, loss-of-function studies involving the expression of DN-Lef1 or a fusion of Lef1 with the engrailed repressor(Lef1-En) showed an inhibition of peripheral marker expression and iris hypoplasia without affecting retinal tissues. Together, our results suggest that the Wnt2b/β-catenin pathway plays a crucial role in specifying the peripheral fates of the eye, in keeping with the role of wg in the Drosophila eye.

The alkaline phosphatase (AP) coding sequence, followed by the SV40 poly(A)signal, was removed from pCAG:AP (T. Matsuda and C.L.C., unpublished) with HindIII, blunt-ended with Klenow and digested with EcoRI. The luciferase-coding sequences from M50 and M51(Veeman et al., 2003) were removed with HpaI and EcoRI, and the remaining vector containing eight tandem repeats of Lef1/TCF-binding sites or mutated Lef1/TCF-binding sites was ligated to a DNA fragment encoding AP. The resulting constructs are named SuperTopAP and SuperFopAP, respectively.

Fertilized eggs were obtained from SPAFAS (CT) and incubated at 37°C to Hamilton-Hamburg (HH) stage 10 (Hamburger and Hamilton, 1951). Viral injection (RCAS:CA-β-catenin and RCAS:Wnt2b) and plasmid electroporation were carried out as described previously (Schulte et al.,1999). Retinas with morphological phenotypes (CA-β-catenin induced a thin/folded phenotype and Wnt2b a minor thinning) at E5.5 and E7.5 were further analyzed for gene expression changes by in situ hybridization,followed by anti-viral gag antibody staining to visualize the areas of viral infection. The number of retinas showing gene expression changes (see Results)was more than 80% of RCAS:CA-β-catenin- and RCAS:Wnt2b-injected embryos manifesting the morphological phenotypes described above. Gene expression changes were correlated with viral infection in an individual animal, which was possible as infection was not complete, and this was considered the best internal control for effects of introduced genes. In addition, retinas infected with RCAS or RCAS:GFP, or untreated retinas, were used as controls. For electroporation (pMIW III:GFP+SuperTopAP±RCAS:CA-β-catenin,RCAS:DN-Lef1and RCAS:Lef1-En), the DNA concentration was ∼400 to 800 ng/μl in PBS; ∼400 ng/μl of pMIW III:GFP plasmid was included to visualize the area of transduction. Embryos were examined 24-48 hours after electroporation for the presence of the GFP in or around the retina, and gene expression changes (Collagen IX) and ciliary body/iris phenotype were analyzed at E4.5 and E14-E16, respectively, using in situ hybridization, followed by antibody staining (anti-GFP or anti-viral gag) to visualize the areas of viral infection. Experimental protocols with embryos were approved by the Institutional Animal Care and Use Committee at Harvard University.

E5 chick retinas were electroporated in a chamber containing 400 ng/μl of pCAG:CA-β-catenin DNA and 400 ng/μl of pCAG:GFP, and cultured for 16-18 hours in explant culture medium (10% FCS, 45% HAMS F12 nutrient, 45%DME, 200 mM L-Glutamine, 10 mM Penicillin/Streptomycin, 10 mM Hepes). To label dividing retinal progenitors in S-phase, 5 μl of [³H]thymidine(NEN) was added to 1 ml of culture medium (final concentration of 5 μCi/ml)6 hours prior to harvesting. The areas with GFP⁺ cells were cut out under a fluorescence microscope (Leica Fluo III) and dissociated by trypsin or papain. Autoradiography was carried out as described previously(Morrow et al., 1998). For in ovo labeling, 3 or 6 days after in ovo viral injection into the OV at stage 10, 0.3 ml of 5 mg/ml BrdU was injected to the yolk 30 minutes prior to harvesting. Centrally located thin and folded areas were examined.

Section ISH was performed as previously described(Murtaugh et al., 2001). Probes were labeled with digoxigenin (DIG), and AP-conjugated anti-DIG antibody (1:2500, Roche) was used for detection with NBT/BCIP (5 mg/ml). Probes used were Collagen IX (Dhawan and Beebe, 1994), Bmp7 (Oh et al.,1996), WFDC1 [J. Trimarchi and C.L.C., unpublished; ChEST719l20(MRC geneservice)], Chx10 (Chen and Cepko,2000), Notch1 (Austin et al.,1995), Wnt2b (Jasoni et al.,1999) and Lef1 (Kengaku et al., 1998). For the Wnt reporter assay, staining for AP activity was performed in NTMT, as described(Murtaugh et al., 2001).

IHC was performed as described previously(Dyer and Cepko, 2001). The primary antibodies used were α-BrdU (Sigma) at 1:200, α-pH3(Upstate Biotechnology) at 1:200, α-β-Tubulin III (Upstate Biotechnology) at 1:250, 3C2 mAb (Developmental Studies Hybridoma Bank; DSHB)at 1:50, α-p27 (DSHB) at 1:200, α-Collagen IX (DSHB) at 1:50,α-Pax6 (DSHB) at 1:30, α-GFP (Molecular Probes) at 1:200,α-Visinin mAb (DSHB) at 1:30 and α-SMA (Sigma) at 1:200. Secondary antibodies used were Cy2- or Cy3-conjugated anti-mouse or anti-rabbit antibodies (Jackson Immunoresearch Laboratories) at 1:250. For BrdU labeling,sections were separately treated with 0.1% trypsin for 10 minutes and,subsequently, with 2N HCl for 30 minutes, which destroys the fluorescence of GFP. Also, IHC of α-GFP was performed to detect the area of SuperTopAP electroporation after AP reaction, which obscures or destroys the GFP fluorescence. For nuclear counterstaining, DAPI (blue) was added to the final wash solution at 0.0005%.

After antibody staining, retinal sections were subjected to fluorescent detection using an in situ cell death detection fluorescence kit (Roche),according to the manufacturer's instructions. Incubation was performed for 3 to 10 hours.

Clonal analysis of retinal reaggregation cultures treated with Wnt2b-conditioned medium showed that Wnt2b can stimulate the proliferation of E5 RPCs in vitro (Kubo et al.,2003). A recent observation that Wnt2b inhibits the expression of Notch and proneural genes in the retina led to the proposal that Wnt2b keeps RPCs undifferentiated in the peripheral OC(Kubo et al., 2005).

In order to assay whether canonical Wnt signaling promotes the proliferation of RPCs in an intact tissue, the proliferation of retinal cells in an explant, in the presence or absence of CA-β-catenin, was assessed. Retinas were electroporated with a control GFP plasmid with the ubiquitous CAG promoter/enhancer (pCAG:GFP) only, or with a mixture of plasmids encoding CA-β-catenin and GFP (pCAG:CA-β-catenin+pCAG:GFP). When CA-β-catenin was expressed in embryonic day 5.5 (E5.5) explants, the fraction of mitotic RPCs decreased compared with that of the GFP-expressing(or non-electroporated) control cells (25.2±2.4% to 10.7±3.5%, n=5; P=0.0001; Fig. 1A).

To assess mitotic activity in vivo, an avian replication-competent retroviral vector encoding N terminus-deleted CA-β-catenin[RCAS:CA-β-catenin (Funayama et al.,1995)] was used to infect the OV at HH stage 10. Viral particles were injected or viral DNA was electroporated into the OV, whereupon viral expression and replication would initiate in infected or electroporated cells,allowing the virus to spread throughout the eye by further infecting dividing cells. Viral infection could be tracked by immunohistochemical (IHC) detection of a viral gag antigen using antibodies such as 3C2 or α-p27 gag(Chen and Cepko, 2002). Injection of bromodeoxyuridine (BrdU) was then carried out at later stages, 1 hour prior to harvest. The effects on cell proliferation were initially assayed at E4.5, when patches of infected cells were relatively small,indicating that viral infection occurred locally, and had not spread throughout the eye. We did not observe any significant effect on retinal cell proliferation, except in rare infected patches of cells that showed a partial inhibition of retinal cell proliferation when compared with uninfected retinas(see Fig. S1 in the supplementary material). The effects on proliferation were also assayed later. At E7.5, a significant reduction in the number of BrdU-labeled cells was observed in extensively infected regions. These infected areas were thinner and folded, compared with RCAS:GFP-infected control retinas (Fig. 1D-H, Fig. 3). The percentage of cells that incorporated BrdU in severely folded central retinas resulting from CA-β-catenin infection was 9.7±3.9% (n=4, Fig. 1B,H), which was significantly reduced relative to the central and peripheral parts of the control retina [19.2±2.9% (n=4, P=0.01) and 30.3±2.8% (n=4, P=0.004), respectively; Fig. 1B,D,E,G], but similar to that of ciliary/iris epithelia from control retina (9.8±1.1%, n=4, P=0.863; Fig. 1B,F,G). The fraction of cells in M-phase of the cell cycle also was assessed using an α-pH3 antibody. A similar reduction of the fraction of cells labeled with α-pH3 was seen in vivo in highly folded areas of the RCAS:CA-β-catenin-infected retina. The percentage ofα-pH3-stained cells in the RCAS:CA-β-catenin-infected, thin and folded epithelium was 0.6±0.7% (n=4, Fig. 1C,H), which is significantly less than was observed in the central and peripheral parts of the control retina [3.1±1.4% (n=4, P=0.017) and 4.5±0.6% (n=4, P=0.001), respectively; Fig. 1C-E,G], but similar to that of ciliary/iris epithelia from control retina (0.5±0.3%, n=4, P=0.309; Fig. 1C,F,G). Consistent with the reduction of BrdU incorporation and pH3 staining in vivo, the level of cyclin D1 RNA was decreased in RCAS:CA-β-catenin-infected retinas at E5.5 (see Fig. S1 in the supplementary material).

The expression of Wnt signaling genes was compared with those of RPGs. Wnt2b, the only identified Wnt in the peripheral OC(Liu et al., 2003), was expressed in the SE apposed specifically to the dorsal region of the OV(Fig. 2A,B). In the early OC,Wnt2b expression was seen in the presumptive RPE and peripheral rim(Fig. 2C,D). This pattern was maintained throughout the OC stages and later expression was confined to the iris epithelium (see Fig. S1 in the supplementary material)(Jasoni et al., 1999). Similarly, Lef1 and Fz4 were weakly expressed in the dorsal region of the invaginating OV (Fig. 2E; data not shown). Later, they were expressed in the periphery of the OC, including in the precursors of ciliary and iris epithelia(Fig. 2G; Fig. S1 in the supplementary material). Expression of Chx10(Chen and Cepko, 2000) was limited to the central and ventral areas of the invaginating OV, as its expression was gradually reduced in the dorsal OV(Fig. 2F) where Wnt signaling genes, Lef1 and Fz4, were observed. At the OC stage and later, Chx10 was highly expressed in mitotic RPCs in the more central regions, but weakly in the periphery (Fig. 2H; Fig. S1 in the supplementary material). Other RPGs, such as Notch1(Bao and Cepko, 1997) and Cyclin D1, are expressed in mitotic RPCs at the OC stage (data not shown). These data reveal complementary, but partially overlapping, expression of RPGs and Wnt signaling genes at both the OV and OC stages.

A previous report using a TCF/Lef-lacZ transgene reporter in the mouse eye showed that a canonical Wnt signal is present in the retinal margin and distal RPE at E14, and is further limited to ciliary/iris epithelia at E17.5(Liu et al., 2003). However,this study did not address when the Wnt signaling could first be detected. A Wnt signaling reporter, SuperTopAP (Fig. 2I), was constructed to investigate this point in chicks. SuperTopAP contains eight tandem TCF/Lef1-binding sites, and has alkaline phosphatase (AP) as a reporter in place of the original luciferase gene of SuperTopFlash (Veeman et al.,2003). SuperTopAP was electroporated into the early OV at HH stage(st.) 10 and the retinas were examined at a later OV stage (st. 13), or at an early OC stage (st. 15). When examined at st. 13, strong Wnt reporter activity was detected in the dorsoposterior OV, which will develop into the RPE and the ciliary body/iris epithelia (n=9, Fig. 2J). By contrast, the portion of the anterior OV that will give rise to the NR, did not exhibit Wnt signal activity. Similarly, 24 hours post-electroporation, Wnt reporter activity was high in both the RPE and the tip of the peripheral OC, but was very low in the adjacent central and inner OC (n=15, Fig. 2K). However, there was very weak reporter activity in the central OC, in some cells on the scleral side of the NR where it abuts the RPE (Fig. 2K′), consistent with a previous report(Liu et al., 2003). There was an absence of Wnt activity in the ventral OV and OC, which is likely to be due to the delayed development of the ventral retina relative to the dorsal counterpart (Fig. 2J,K). As a negative control, SuperFopAP (Fig. 2I) was made, in which the TCF/Lef1-binding sites were mutated, as in SuperFopFlash (Veeman et al.,2003). SuperFopAP gave no detectable signal in the developing retina and RPE (n=12, Fig. 2L).

In summary, the pattern of Wnt reporter activity is similar to the expression of Wnt2b and Lef1 at the OC stage. Both Wnt activity and Wnt2b expression were evident in the peripheral OC and RPE, whereas they were undetectable or very low in the NR-forming areas of the OV and OC.

Reduced proliferation of RPCs upon Wnt signal activation might result from an inhibition of the cell cycle in RPCs. Alternatively, a cell fate change,from highly mitotic retina to that of less mitotic peripheral tissue(Beebe, 1986; Kubota et al., 2004), could lead to a reduction in cell proliferation.

To further explore the in vivo role of Wnt signaling, a gain-of-function study was carried out by expressing CA-β-catenin or Wnt2b using a replication-competent retrovirus (RCAS:CA-β-catenin or RCAS:Wnt2b) in the developing OV. The efficacy of CA-β-catenin in the developing retina was demonstrated through its ability to ectopically activate a SuperTopAP Wnt reporter in the central retina, where normal Wnt activity is low or absent(see Fig. S2 in the supplementary material; compare with Fig. 2K). Eyes infected with RCAS:CA-β-catenin did not show any obvious superficial phenotypes at E4.5. Retinal sections showed patches of infected cells, but the thickness of the retinal epithelium was not severely affected (see Fig. S1 in the supplementary material). A rare morphological change was the formation of a kink at the scleral side of an infected patch. When the tissues were harvested at E7.5 (after which animals die as a result of the spread of the virus), a disorganized head with protruding eyes was observed (12/17; ∼71% of all infected retinas, Fig. 3A,B). The overall size of the affected eyes varied: sometimes they were slightly bigger, but other times, they were similar or smaller than control eyes. Examination of the infected eyes revealed small folds on the vitreal side of the retina (Fig. 3C,D). Cross sections revealed unusually thin areas, and folds, which are similar to the normal morphology of the ciliary/iris epithelia(Fig. 3E-I). In severely affected eyes, the folded area of the retina had only one-cell layer, and cells had a cuboidal shape, a characteristic of normal ciliary/iris epithelial cells (Fig. 3J; Fig. S2 in the supplementary material). Weaker phenotypes were observed when eyes were harvested at E5.5 (Fig. 4A,B). Expression of Wnt2b also caused a minor thinning of the retina at E5.5 and E7.5 [Fig. 4C,D; the average thickness of the control- and RCAS:Wnt2b-infected thinner retinas at E7.5 were 150 μm (n=7) and 82.4 μm (n=7), respectively; P=0.007], but both expressivity and penetrance was always much lower than that induced by CA-β-catenin.

To test whether the identity of the cells in the thin and folded area of the retina had changed, the expression of RPGs, such as Chx10 and Notch 1(Austin et al., 1995; Chen and Cepko, 2000), was examined. Infection with RCAS:CA-β-catenin or RCAS:Wnt2b caused RPCs to partially lose retinal progenitor characteristics in that both viruses led to a partial to complete loss of the RPGs Chx10 (n=10, Fig. 4A,C and Fig. S2 in the supplementary material) and Notch1 (n=9, Fig. 4B,D) at E5.5 and E7.5(data not shown).

The decrease in mitotic activity in explants and in ovo, and the loss/reduction in RPG expression upon Wnt activation, could be due to premature cell cycle withdrawal and subsequent differentiation. To examine this possibility, staining with anti-β-Tubulin III, a marker of ganglion and amacrine cells (Fig. 4I),was carried out at E7.5. Infection with RCAS:CA-β-catenin or RCAS:Wnt2b led to a complete or partial loss of β-Tubulin III staining (n=7 and 3, respectively; 4J,K). Residual β-Tubulin III staining was seen in the border between infected and uninfected areas (arrowhead in Fig. 4J), which is likely to result from the labeling of axons originating from uninfected ganglion cells. The differentiation of photoreceptor cells was assessed using a Visinin antibody. The partial to complete reduction of Visinin⁺photoreceptor cells at the scleral side of the retina is evident in the affected retina (Fig. 4L,M). Amacrine cells were assayed using anti-Pax6, which is normally expressed in RPCs, amacrine, ganglion and horizontal cells at E7.5(Fig. 4N). Additionally, Pax6 is expressed in the pigmented and non-pigmented layers of ciliary/iris epithelia (see Fig. S3 in the supplementary material). Absence or reduction of Pax6⁺ amacrine cells in the inner side of the retina was clearly observed in the RCAS:CA-β-catenin-infected eye (n=9, Fig. 4O). The effects on Pax6 expression in RPC and horizontal cells were not clear because of the disorganized epithelial structure in RCAS:CA-β-catenin-infected retinas.

To investigate whether Wnt signaling can switch the NR to peripheral fates,the expression of genes whose endogenous expression is enriched in the peripheral OC and ciliary/iris epithelia was investigated. Collagen Type IX is a marker of both the developing ciliary and iris epithelia at E7.5, although the expression is weaker in the iris epithelium. Its expression starts around E4, and is confined to the periphery of the eye at later stages[Fig. 5A,B; see also Fig. S4 in the supplementary material (Dhawan and Beebe, 1994; Kubo et al.,2003)]. Similarly, Bmp7, an essential factor for ciliary body formation (Zhao et al., 2002),is exclusively expressed in the periphery of the OC beginning at E6.5(Fig. 5C,D).

The expression of peripheral markers in RCAS:CA-β-catenin-infected retinas was examined at E5.5 and E7.5. In the E5.5 central retina where extensive folding had not yet occurred, activation of Wnt signaling led to ectopic induction of Collagen IX and Bmp7 (n=8 and 5, respectively, Fig. 5E-H), and a recently discovered peripheral marker, WFDC1 (n=5, Fig. 5I,J; J. Trimarchi and C.L.C., unpublished). Induction of ectopic peripheral markers was confined to the infected areas, but was not manifest in all infected cells of all infected areas. At E7.5, when the thinning and folding of the central retina was apparent, Collagen IX protein and RNA were ectopically induced in the thin and folded regions of the affected retinas (data not shown). In addition, Wnt2b and Lef1, genes expressed in precursor cells of the iris and ciliary body,respectively, were induced in the thin or folded areas (n=3 and 3,respectively, Fig. 5K,L). Furthermore, the severely folded, thin central tissue expressed smooth muscle specific actin (SMA; n=6, Fig. 5M,N), which is normally expressed in invaginating and migrating iris epithelial cells (Zhao et al.,2002).

In order to determine whether Wnt signaling is necessary for the development of the peripheral eye, loss-of-function experiments were carried out in vivo. Reduction of Wnt signaling was achieved by expression of inhibitory molecules, Lef1-En, a fusion of Lef1 with the En repressor (G. Kardon and C. Tabin, unpublished) or a dominant-negative allele of Lef1,DN-Lef1, a deletion of the β-catenin-binding domain(Kengaku et al., 1998) (data not shown). First, the effects of RCAS:Lef1-En on peripheral marker expression were examined after in vivo electroporation. Collagen IX expression was occasionally reduced at the periphery of the E4.5 retina when Lef1-En or DN-Lef1 was expressed (2/8; 25%, Fig. 6A,B). However, the expression of Collagen IX was only significantly reduced when viral infection was maximal, as judged by the signal intensity of IHC for a viral gag protein(Fig. 6A′,B′).

The effects of viral expression of Lef1-En on the maturing peripheral tissues were examined at E14-E16, when the structures of the ciliary body and iris are morphologically distinguishable. The majority of the electroporated/infected eyes (13/17; 76% affected) exhibited a partial loss of ithe ris (Fig. 6C). Sections of the affected eyes revealed a localized hypoplasia of the iris and, to a lesser extent, the ciliary body, compared with unaffected regions in the same eye(Fig. 6D-F), which are indistinguishable from RCAS:GFP-infected retinas (0/26; 0% affected, data not shown). To further examine this phenotype, the development of iris-derived muscles in the severely affected eyes was examined using an α-SMA antibody. Consistent with the morphological defects seen in RCAS:Lef1-En-infected eyes at E14, a severe reduction of α-SMA-positive cells in the stromal layer of the developing iris was observed in severely affected eyes (4/10; 40%, Fig. 6G) compared with the unaffected side (0/10; 0%, Fig. 6H). RCAS:Lef1-En-infected animals also had defects in RPE pigmentation. As shown in Fig. 6I,J,RCAS:Lef1-En-infected areas of the RPE exhibited hypopigmentation.

Gain- and loss-of-function analyses performed in vivo led us to propose that the Wnt2b/CA-β-catenin signal is sufficient for the determination of the identity of the peripheral eye. Four lines of evidence support this proposal. First, the expression of the RPGs, Chx10, Notch1 and Cyclin D1, as well as markers of differentiated retinal cell types (Visinin, Pax6 andβ-Tubulin III) was lost or reduced following the introduction of CA-β-catenin and Wnt2b. These results support a loss of retinal identity upon Wnt signal activation. Kubo et al.(Kubo et al., 2003) also reported a loss of retinal differentiation markers and a loss of Notch1 following the introduction of Wnt2b, but observed the induction of Chx10, Pax6 and Rx1 at E3.5.

The second line of evidence is the observation of expression of Collagen IX, Bmp7 and WFDC1 following exposure to Wnt2b/CA-β-catenin, as was also seen by Kubo et al. (Kubo et al.,2003). It is clear that Collagen IX is a marker of the ciliary body, as it is expressed in the peripheral OC, in the ciliary epithelium during mid-embryonic stages, and in the pars plana at later embryonic stages(see Figs S4, S5 in the supplementary material). It is not clear whether Collagen IX is also a marker of retinal CMZ stem cells, which is the interpretation favored by Kubo et al.(Kubo et al., 2003). Induction of Bmp7 is consistent with a fate change, as it has been shown that Bmp signaling is essential for ciliary body formation(Zhao et al., 2002). We also found induction of Wnt2b and Lef1, two additional markers of the periphery,following the introduction of CA-β-catenin. We also demonstrated that the thin and folded tissue infected with CA-β-catenin expresses smooth muscle actin, which is normally expressed in invaginating iris epithelial cells that transdifferentiate into smooth muscle cells in the stroma of the iris(Fig. 5M). The expression of all of these peripheral markers, including a distinctive iris marker, suggests that the cells exposed to a high level of Wnt signal are induced to form the ciliary body (expressing Collagen IX and Bmp7) and iris (expressing Wnt2b and SMA). The third line of evidence is the morphology of retinal tissue infected with CA-β-catenin. It is thin and highly folded, resembling the morphology of normal ciliary body/iris tissue. Although folding could be secondary to overproliferation, we did not find evidence of overproliferation,as discussed below. The fourth line of evidence is that abrogation of Lef1-mediated transcription led to a loss of iris tissue in vivo, some disorganization of the ciliary body, and a reduction of peripheral marker expression. The partial decrease of Collagen IX expression by Lef1-En or DN-Lef1 (Fig. 6A) is likely to be due to the partial, not complete, loss of Lef1 activity in the peripheral retina. Alternatively, Lef1-En-mediated partial inhibition of the transcription of Collagen IX might only occur within the area most dependent upon high Lef1 activity. This is the area where Wnt2b is maximal, at the tip of the OC, in the region containing precursors of iris cells. The fact that the loss-of-function experiments yielded data that are complementary to the gain-of-function data strongly supports a model in which high Wnt signaling induces the most peripheral eye fate(s).

Retinal cells expressing CA-β-catenin or Wnt2b appear to first slowly lose retinal identity and then gradually acquire peripheral identity. RPCs would be expected to receive conflicting signals: retinal fate promoting activity by RPGs and peripheral fate promoting activity by a Wnt signal. Probably as a consequence, some retinal cells with Wnt signal activation in the central retina undergo apoptotic cell death (see Fig. S3 in the supplementary material). However, cell death cannot create the phenotype of the acquisition of peripheral markers by retinal cells. It is also worth noting that the progressive loss of retinal identity and the acquisition of RPE identity occurred slowly in the Orj mouse mutant, which has a null allele of Chx10 (Horsford et al.,2005; Rowan and Cepko,2004).

Not all infected areas showed a robust phenotype when assayed for gene expression changes and/or morphological changes. Instead, or in addition to the slow kinetics noted above, the lack of a phenotype might be due to the fact that there are Wnt antagonists, Sfrp1 and Sfrp2, in the retina(Blackshaw et al., 2004; Esteve et al., 2003). A high level of Wnt signaling early when patterning is likely to occur might be needed to cause a fate switch and this might not have been achieved in all infected areas.

Kubo et al. (Kubo et al.,2003; Kubo et al.,2005) performed two studies in which they investigated the role of Wnt2b signaling in the early chick retina. Cultures of dissociated and reaggregated cells treated with Wnt2b displayed an increase in proliferation. Similarly, using a Wnt2b-expressing retrovirus and in vitro cultures of infected tissue, the authors found a dramatic increase in proliferation in the presence of EGF and/or bFGF. As mentioned above, they also found a gain of retinal progenitor markers, a loss of neuronal markers, and a gain of Collagen IX expression. Although some of the results reported here confirm and extend their findings, they also provide an alternative interpretation for some of their observations. Kubo et al. (Kubo et al., 2003) argue that expression of Collagen IX and hyperproliferation are due to the acquisition of stem cell behavior. Cells at the far periphery of the retina of amphibians and fish exhibit stem cell characteristics in vivo, and cells isolated from the PCM of mammals can generate neurospheres in vitro (Ahmad et al., 2000; Moshiri et al.,2004; Tropepe et al.,2000). As Wnt2b-expressing cultured retinas demonstrate a high degree of proliferation in vitro, it is possible that retinal cells exposed to a high level of Wnt2b normally in the periphery in vivo are induced to form stem cells. However, Kubo et al. (Kubo et al., 2003; Kubo et al.,2005) did not demonstrate the induction of proliferation in vivo,so it could be that additional factors are needed for such an induction of proliferation in vivo, and perhaps for an induction of stem cell properties. We also examined the effect on proliferation at E4.5 and E7.5 after infection with retroviruses expressing CA-β-catenin and Wnt2b at the OV stage, and did not find an increase in proliferation. The level of proliferation in vivo in the infected central retinal areas was almost identical to that of the normal ciliary/iris epithelia, which is lower than that normally found in uninfected, central retina (Beebe,1986; Kubota et al.,2004). As Kubo et al. (Kubo et al., 2005) clearly demonstrated a dramatic hyperproliferation in vitro in response to Wnt2b, there might be an inhibition of this role of Wnt in vivo, perhaps provided by surrounding ocular tissues. The Wnt antagonists Sfrp1and Sfrp2 are expressed in the central retina(Blackshaw et al., 2004; Esteve et al., 2003). Alternatively, it is possible that some other aspect of the in vitro culture environment enhances proliferation in response to Wnt signaling.

A difference between our studies and those of Kubo et al.(Kubo et al., 2003; Kubo et al., 2005) that may account for our observations of additional phenotypes is our use of CA-β-catenin. We were able to produce phenotypes when Wnt2b was overexpressed, and these were similar to, although weaker and less penetrant than, those we obtained with CA-β-catenin. This was expected for several reasons. First, ectopic CA-β-catenin does not require intracellular signal transducers, as it functions at the last step of the signal transduction cascade. Second, CA-β-catenin can bypass the presence of Wnt antagonists in the central retina. Third, because Wnt2b would take time to be produced, secreted, received and transduced in order to affect transcriptional targets, it is possible that high enough levels of Wnt signaling were not achieved quickly enough to change fate following Wnt misexpression. By contrast, β-catenin would act immediately on its targets.

CA-β-catenin is known to have effects on cell-cell adhesion. Although we cannot completely exclude the possible involvement of CA-β-catenin in producing thin and folded phenotypes by affecting the adhesion(Ouchi et al., 2005), several lines of evidence support the interpretation that the observed phenotypes are mainly caused by Wnt signal activation. First, we observed similar phenotypes with both CA-β-catenin and Wnt2b, as discussed above. Second,gain-of-function phenotypes obtained with CA-β-catenin and Wnt2b were complementary to the loss-of-function phenotypes obtained with Lef1-En or DN-Lef1. Third, the expression of CA-β-catenin was sufficient to activate a canonical Wnt reporter (see Fig. S2 in the supplementary material).

A recent study of the Hdac1 mutant of zebrafish found that introduction of CA-β-catenin in vivo did lead to hyperproliferation in the central retina (Yamaguchi et al.,2005). This finding might indicate that zebrafish, which have a continuously proliferating ciliary marginal zone (CMZ), use Wnt signaling to drive normal CMZ proliferation. Similarly, Xenopus frizzled 5 is expressed in the domain-flanking stem cells in the CMZ, and is involved in regulating proliferation and differentiation(Van Raay et al., 2005). The fact that there is rapid and ongoing conversion of CMZ cells to RPCs in fish and amphibians throughout their lives might call for a regulation of this conversion that is not needed in animals with only a transient, or no, CMZ,i.e. in chick and mouse. If Wnt signaling is the regulator of this conversion in fish and amphibians, perhaps some aspect of this role of Wnt is also present in chicks, and it is this role that was revealed by the in vitro behavior of chick retinal cells exposed to high Wnt2b levels(Kubo et al., 2003; Kubo et al., 2005).

Previous models of the patterning of the eye recognized the early division of the OV into domains that would give rise to the outer OC and inner OC, and thus the RPE and retina, respectively(Chow and Lang, 2001; Graw, 2003; Martinez-Morales et al.,2004). We are now suggesting that there is a third domain of the OV, determined by high Wnt signaling, which will give rise to the peripheral OC, and thus the ciliary body and iris. Expression of Wnt2b/Lef1/Fz4 and canonical Wnt reporter activity had been previously observed at the OC stages,but not at the OV stage (Jasoni et al.,1999; Kubo et al.,2003; Liu et al.,2003). The sensitive reporter SuperTopAP allowed the observation of Wnt activity at the OV stage. Both Wnt reporter activity and Lef1 expression were observed as early as the stage when the OV invaginates, when Wnt2b is not expressed in the OV (Fig. 2A,J). Furthermore, within the invaginating OV, the expression pattern of Lef1 is complementary to that of Chx10, an early retinal marker(Fig. 2E,F). This observation indicates that the initial blueprint for the patterning of the ciliary body/iris via Wnt2b/β-catenin signaling is present as early as the OV stage. In addition to the source of Wnt2b in the SE and periphery of the OC,Wnt2b is expressed in the early lens [Fig. 2D; Fig. S1 in the supplementary material(Jasoni et al., 1999)]. Induction of the peripheral fates of the OC had been proposed to be dependent upon signals from the lens (Thut et al.,2001). Thus this previous model, and data presented here, are consistent with the model that Wnt2b produced by the lens, the SE and the peripheral OC coordinate to produce high Wnt signaling, and thus induce the peripheral fates of the OC. Other factors, including unidentified Wnts expressed in overlying mesoderm or neuroectoderm, might also contribute to the Wnt signal activation in the OV. The latter possibility is supported by the observation that extraocular mesenchyme possesses an activity that induces Wnt2b in the RPE and represses RPGs in the NR(Fuhrmann et al., 2000). However, because Wnt2b has an expression pattern that is almost identical to the Wnt reporter activity, it is likely to plays a major role in canonical Wnt signaling at the OC stage (Fig. 2C,D,K).

A role for Wnt2b in RPE and extraocular development is suggested by the Wnt2b expression pattern, which expands to the dorsal RPE and the tip of the developing OC (Fig. 2C,D)(Jasoni et al., 1999; Kubo et al., 2003). A recent study found that Wnt signal in periocular SE was required to suppress lens formation in these cells, providing a mechanism for the division of the lens from non-lens in the early SE (Smith et al., 2005). Wnt2b may be the Wnt that is upstream ofβ-catenin for this pattern formation. In support of a role for Wnt signaling in RPE development was our finding that RCAS:Lef1-En and RCAS:DN-Lef1 led to a lack of pigmentation in the RPE at later stages of development (Fig. 6I,J and data not shown). This was of interest because retinal tissue was observed to convert into RPE in a mouse mutant lacking Chx10(Horsford et al., 2005; Rowan and Cepko, 2004), and we found that CA-β-catenin introduction into the retina led to loss of Chx10 in the chick. In addition, it is known that ectopic expression of Chx10 in the developing RPE causes a lack of pigmentation, and thus the hypopigmentation we observed following the introduction of RCAS:Lef1-En and RCAS:DN-Lef1 might have been caused by a de-repression of Chx10 in the RPE. It will be worth examining RPE cells expressing DN-Lef1 at later stages of development to see whether they express other retinal markers and eventually differentiate into retinal cells.

Our findings provide an additional link between the development of the vertebrate and invertebrate eye. In Drosophila, photoreceptor cells are surrounded at the periphery with a non-neural cuticular structure. wg, the Drosophila homolog of the Wnt genes, is expressed in the margin of the eye imaginal disc, which is the anlage of peripheral eye tissues. Activation of wg, or armadillo, the Drosophila β-catenin, in the eye imaginal disc promotes head cuticle formation at the expense of ommatidia(Baonza and Freeman, 2002), and has been proposed to act as a morphogen to pattern the peripheral structures(Tomlinson, 2003; Treisman and Rubin, 1995).

Wnt signaling thus promotes the development of the non-neural, peripheral support structures in both Drosophila and chicks. The similarity of wg/Wnt expression and function in eye development provides an additional line of evidence that strengthens the proposed evolutionary conservation of the vertebrate and invertebrate eyes(Fig. 7). The modern version of this model originated with the observation of a conserved expression and activity for the eyeless/Pax6 gene(Gehring, 2002; Gehring, 2004; Halder et al., 1995; Onuma et al., 2002). The fact that wg/Wnt appears to play a role in patterning the central and peripheral eye structures suggests that the visual structure of the last common ancestor of flies and vertebrates had not only a photoreceptive component, but a support structure as well. A conserved unit of neural and non-neural eye tissues has also been suggested by the observation of a single-celled dinoflagellate that has several of the support structures of an eye, including pigment, a lens, a cornea and a photoreceptor(Gehring, 2004). The fact that Pax6 plays a role in the development of not only the NR, but also the supporting tissues, such as the lens, cornea, iris and RPE, might also be seen as being in keeping with this model.

An alternative interpretation of the use of Wnt/Wg for both vertebrate and invertebrate eye patterning needs to be considered. Because Wnt pathway genes are commonly used for many developmental processes, Wnt signaling in eye development may only reflect the utility of Wnts in development, rather than a homologous process (Fernald,2004). The identification of additional genes that play a key role in the patterning of these early eye domains, and establishment of a conserved role for them, will be required to strengthen the model of a homologous process for the establishment of neural and non-neural division of the early eye.

We are grateful to Drs R. Moon and D. Beebe for SuperTopFlash,SuperFopFlash and the anti-Collagen IX probe. We thank members of the Cepko and Tabin laboratories for helpful comments and discussion, particularly Dr J. Mansfield for critical reading of the manuscript. This research was supported by the Howard Hughes Medical Institute.