Pax6 Interacts with cVax and Tbx5 to Establish the Dorsoventral Boundary of the Developing Eye*

Dorsoventral pattern formation of the optic cup is essential for vertebrate eye morphogenesis and retinotectal topographic mapping. Dorsal and ventral aspects of the eye are distinct at early stages of development; cVax homeodomain protein expression is confined to the ventral optic cup, whereas Tbx5 (T-box transcription factor) expression domain becomes restricted to the dorsal region. Misexpression of cVax or Tbx5 induces profound defects in eye morphology and abnormal vis-ual projections. In the Pax6 (cid:1) / (cid:1) mutant Tbx5 fails to be expressed, and Vax1 and -2 are abnormally present in the entire optic vesicle. During eye development Pax6 becomes expressed in a gradient at the optic cup stage due to the specific activation of a highly conserved intronic (cid:2) enhancer in the Pax6 locus. We observed that the highest level of Pax6 in the optic cup corresponds to the boundary between non-overlapping cVax and Tbx5 territories. To further investigate how these transcription factors control the patterning of the eye, we over-expressed Pax6 in the chick optic cup (E2) using in ovo electroporation. We observed that overexpression of Pax6 extends the Tbx5 and Bmp4 domains but reduces the cVax expression of cup 830 BTX electroporator. an stereomicroscope lin- earized RNA polymerase prepared

Formation of the vertebrate eye involves a series of morphogenetic and inductive events that begin with the evagination of the optic vesicles from the forebrain. This is followed by invagination of the optic vesicles to create the optic cups. This results in the formation of two layers giving rise to the retina pigment epithelium, the neural retina, parts of the ciliary body, and a portion of the iris (1). An early and key event in eye development is the establishment of the nasal-temporal and dorsalventral axes in the optic primordium, which is fundamental for the later formation of the retinotectal projections.
Development of the dorsoventral polarity has already been initiated in the optic vesicle stage and precedes the onset of neuronal differentiation (2,3). At this stage gene expression is restricted along the dorsoventral axis, thus dividing the retina into several domains (4). Two transcription factors expressed in a mutually exclusive manner specify the dorsal and ventral compartments of the developing retina; Tbx5, a member of the T-box gene family, is expressed in the dorsal part of the retina in mouse and chicken (5,6). Misexpression of Tbx5 in the developing eye induced dorsalization of the eye and altered projections of retinal ganglion cell axons (7). Vax homeodomain protein expression is restricted to the ventral part of the developing eye in mouse (Vax1 and -2) and chicken (cVax) (8). Vax2 inactivation in mouse revealed an important role of this gene in the specification of the ventral optic vesicle (9,10). Misexpression of cVax leads to ventralization of the retina and abnormal retinotectal projections (8). Therefore, the correct expression of Tbx5 and cVax is critical for the early determination of the dorsal and ventral compartments of the eye and for the proper development of the retinotectal projections.
Pax6 is a highly conserved transcription factor with two DNA binding domains, the paired and the homeodomain, and a transactivation domain (11,12). Pax6 encodes different proteins through alternative splicing and internal initiations (13). Three proteins of 48, 46, and 43 kDa contain a paired domain, but two proteins of 33 and 32 kDa are devoid of this DNA binding domain. Pax6 is required for the proper development of the pancreas and the nervous system, where it is known to be critical for eye development (14,15). Pax6 is expressed in presumptive eye tissues. Both the semi-dominant inheritance pattern of the Pax6 mutant phenotype (16,17) and experimental manipulations using transgenes based on Pax6 locus (18) indicate that achieving the correct level of Pax6 is important for development of a normal eye. Because in Pax6Ϫ/Ϫ mice formation of the embryonic eye is disrupted early in development when the optic vesicle fails to form an optic cup (19), the function of Pax6 at these early steps is not well understood. Pax6 is first widely expressed in the optic vesicle. Later, in the optic cup, Pax6 becomes expressed in a distal (high)-proximal (low) gradient due to the specific activity in the retina of a highly conserved intronic enhancer named ␣ (20,21). Interestingly, this gradient initiates at the timing of dorsoventral axis formation, at the onset of cVax expression and when Tbx5 becomes restricted to the dorsal part of the retina. In addition, it was recently reported that in the Pax6Ϫ/Ϫ mouse mutant small eye (Sey), Tbx5 failed to be expressed, whereas Vax1/ Vax2 expression extended over the entire optic vesicle (21). Therefore, it was of particular interest to investigate the relevance of Pax6 for cVax/Tbx5 expression and its potential role in the dorsoventral patterning of the eye.
Overexpression of Pax6 in the chick optic cup using in ovo electroporation extended Tbx5 and Bmp4, reduced cVax expression domains, and caused defects in eye morphology. In addition, we showed that Tbx5 and cVax interact with Pax6 and that they modulate in an opposite manner the activity of the Pax6 ␣ enhancer. Finally, we observed that in the optic cup the highest level of Pax6 expression corresponds to the boundary between the non-overlapping cVax and Tbx5 territories. Taken together, these data strongly suggest that Pax6, together with cVax and Tbx5, contributes to the establishment of the dorsoventral axis of the retina.

EXPERIMENTAL PROCEDURES
Plasmids-Green fluorescent protein (GFP) 1 was expressed from the pVNC3-EGFP vector. Quail cDNA encoding the Pax6 isoforms such as p46, p48, and p30 (12) and the mouse Mitf cDNA were cloned into the pSG5 expression vector (22). The p46⌬48 -211 resulted from a PstI fragment deletion, and the p46⌬170 -342 resulted from a KpnI fragment deletion of the pSG5p46 (12). Details on the G3-138 glucagon promoter have been already described (23). The Tbx5 expression vector and the GST-Tbx5 plasmids were provided by Dr. Colin Goding. A HindIII-EcoRI fragment from the pMIW-cVax provided by Dr. Constance Cepko (8) was cloned into the pV2NC7-HA. pTKCAT-EP and pTKCAT constructs were previously described (20) as well as GST-p46, paired and homeodomains (24). For the synthesis of GST-cVax, a 983-bp fragment of cVax cDNA, was generated by PCR using the pVNC7cVaxHA vector as a template and primers containing cloning sites (GGGACGTGCAGCCGAATTCCATGTTTGGGAAACAAG and CCGCTCGAGCTTTATTGTTGGTCCGGGAGTAGGG) and cloned into the EcoRI-XhoI sites of the pGEX-4T-3.
In Ovo Chick Electroporation-Fertilized white Leghorn chick embryos were incubated at 38°C in a humid atmosphere for 2 days. DNA (2 g/l) was injected into one optic cup of HH stage12-13 (E2) chick embryos. An electric current (5 ϫ 20-V pulses for 50 ms, with an interval of 500 ms between each pulse) was then applied over the optic cup using an ECM 830 BTX electroporator. The untreated optic cup served as an internal control. Expression of GFP was detected with a fluorescence stereomicroscope (Leica MZFLIII). GFP-positive embryos were processed for in situ hybridization or incubated for a longer period of time.
In Situ Hybridization-In situ hybridization was performed as previously described (25). A plasmid containing the Pax6 cDNA was linearized with XhoI and transcribed with T3 RNA polymerase to generate a RNA probe. Probes for the chick cVax, Tbx5, and Bmp4 genes were kind gifts from Drs. Paola Bovolenta and Maria Marx and were prepared as described (26,27). Simultaneous detection of cVax and Tbx5 transcripts using two-color in situ hybridization was performed by sequential detection of digoxygenin-and fluorescein-labeled probes with alkaline phosphatase-conjugated antibodies (28). After in toto hybridization treatment, embryos were rinsed with phosphate-buffered saline, 15% saccharose for 24 h at 4°C, embedded in phosphate-buffered saline, 15% saccharose, 7.5% gelatin, frozen, and cryostat-sectioned.
In Vitro Translation and Glutathione S-Transferase (GST) Pulldown Assays-The 35 S-radiolabeled proteins were translated in vitro using the TNT system (Promega). The GST chimerical proteins were extracted from bacteria after the Amersham Biosciences instructions. The pull-down assays were performed as previously described (22). Percentage of retention was measured for each condition after exposure of the protein gel to a phosphorimaging screen.
Cell Cultures and Transfection Assays-Baby hamster kidney (BHK)-21 cells were cultured in 10% fetal calf serum in Dulbecco's modified Eagle's medium. Dissociated quail neuroretina cells dissected from 8-day-old quail embryos were plated in Dulbecco's modified Eagle's medium/F-12 containing 10% fetal calf serum, 1% minimal essential medium vitamins 100ϫ, and 10 g/ml conalbumin. BHK21 cells transfections were performed with polyethyleneimine (exgen 500, Euromedex, Souffelweyersheim) reagent according to the instructions of the manufacturer. A CMV-LacZ vector was co-transfected for normalization of CAT assays by controlling the ␤-galactosidase activity. Quail neuroretina cell transfections were performed by the calcium phosphate method. CAT assays were performed as previously described (20). Levels of CAT activity were quantified after exposure of the thin-layer chromatograms to a PhosphorImager screen.

RESULTS
Overexpression of Pax6 in the Optic Cup Leads to an Abnormal Eye Phenotype-In the Pax6 mutant mouse (Sey), dorsal characteristics of the optic cup are lost at the expense of ventral ones, suggesting that Pax6 may play a role in regulating retinal dorsoventral polarity. To test this possibility we used in ovo electroporation to overexpress different Pax6 constructs in the chick optic cup. These constructs are depicted in Fig. 1A. First, a plasmid carrying the full-length cDNA coding for p46 or p48 and a plasmid carrying the green fluorescent protein (EGFP) were co-electroporated into one optic cup of E2 chick embryos (HH 12-13); the other eye of the same embryo served as a control. This enabled visualization of the electroporation target site by virtue of GFP expression. As a negative control embryos were electroporated with the plasmid carrying GFP alone. 48 h later eyes electroporated with p46 or p48 showed an abnormal phenotype (Fig. 1, B and D); oval-shaped eyes were formed instead of normal round eyes, and Pax6-electroporated eyes FIG. 1. In ovo electroporation of Pax6 caused eye morphological defects. A, representation of the different constructs used in this study; p46, p48, and p30 are Pax6 isoforms. p46 has two DNA binding domains, paired and homeo. p48 contains an additional 42-base pair exon located in the paired domain. p30 is devoid of the paired domain. The p46⌬48 -211 and the p46⌬170 -342 are truncated forms of p46. p46⌬48 -211 lacks the homeo and transactivation domains but contains an intact paired domain. p46⌬170 -342 misses both the paired and the homeodomain but contains an intact transactivation domain. The plus (ϩ) sign indicates that electroporation of the construct leads to an abnormal ventrally rotated eye phenotype (as shown in B, D, E, and F). B-G, ventral views of E4 chick embryo heads whose right eye (on the left) has been electroporated with a plasmid encoding p46 (B) or GFP (C), p48 (D), p30 (E), p46⌬170 -342 (F), and p46⌬48 -211 (G). Comparing both eyes, the electroporated eye (indicated by an arrowhead) showed morphological defects such as misshaping, rotation toward the ventral midline, and abnormal widened optic fissure. Note that the optic fissures in the electroporated sides are shortened compared with the control sides (arrowheads in B, D, E, and F). Such a phenotype was not observed after electroporation of GFP alone (C) or p46⌬48 -211 (G). Scale bar, 1 mm.
were rotated toward the ventral midline and sometimes presented a widened optic fissure (Fig. 1B). This phenotype was not observed after electroporation of GFP alone (  Table I)).
Pax6 is a transcription factor with three functional domains, two DNA binding domains (paired and homeo) and a transactivation domain (Fig. 1A). To investigate which domain of Pax6 was involved in the eye phenotype, we electroporated different constructs including the p30 Pax6 isoform and two truncated forms of Pax6 (p46⌬48 -211 and p46⌬170 -342). The same abnormal eye phenotype was observed after electroporation of p30, which is devoid of the paired domain (Fig. 1E). The p46⌬170 -342 contains an intact paired domain but is devoid of the homeodomain and part of the transactivation domain, which abolishes transactivation activity (Fig. 1A) (12). Surprisingly, the same abnormal eye morphology was observed after electroporation of the p46⌬170 -342 (Fig. 1F), indicating that the eye phenotype we observed was independent of the transactivation properties of Pax6. Electroporation of the p46⌬48 -211, which lacks both in the paired and homeodomain but contains the entire transactivation domain, did not alter eye morphology (Fig. 1G). Taken together these results suggest that, in contrast to the transactivation domain, both of the DNA binding domains in Pax6, namely the paired and the homeodomains, were involved in causing the observed abnormal eye morphology.
Overexpression of Pax6 Reduces cVax and Extends Tbx5 and Bmp4 Expression Domains-To address the relationship between this abnormal eye phenotype and the expression profiles of known early markers of dorsoventral polarity, we examined the consequence of overexpressing Pax6 in the optic cup on cVax, Tbx5, and Bmp4 expression. Embryos were sacrificed 24 h after electroporation and analyzed for gene expression by whole-mount in situ hybridization. Electroporation of GFP alone or p46⌬48 -211 did not alter the expression domains of cVax, Tbx5, or Bmp4 (data not shown). However, overexpression of the different Pax6 constructs (p30, p46, p48, and p46⌬170 -342) caused an expansion of the dorsal markers (Tbx5 and Bmp4) and a reduction of the ventral marker cVax ( Fig. 2 and data not shown). Any modification in the gene expression territory was measured using metamorph software.  2H). The cVax expression domain was reduced after misexpression of Pax6. Fig. 2, K-P, shows two examples of embryos electroporated with the p46 (Fig. 2, K-M) and the p48 (Fig. 2, N-P) isoforms. Compared with the control eyes, the injected eyes showed a reduction of the cVax expression domain (compare Fig. 2, K and N with L and O). Analyses using serial sections confirmed that cVax expression was reduced and limited to the ventral most region of the retina (Fig. 2M). In the ventral view, intensity of the cVax staining in the optic stalk was reduced in the electroporated side compared with the control side (Fig. 2P).
cVax and Tbx5 Proteins Physically Interact with Pax6 Paired and Homeodomains-Because both the Pax-6 paired domain and homeodomain were independently able to alter cVax and Tbx5 expressions patterns, we tested whether Pax6 interacts physically with cVax and Tbx5. To this end we prepared glutathione-Sepharose beads coupled with a GST fusion protein containing the full-length Pax6 protein (GST-p46; amino acids  Pax6 in Dorsoventral Patterning of the Eye 1-416), the full-length Tbx5 protein (GST-Tbx5), or the fulllength cVax protein (GST-cVax). Glutathione-Sepharose beads coupled to GST alone served as a negative control. No interaction was detected with the radiolabeled GFP protein (Fig. 3A,  lanes 1-3) or with the GST alone (Fig. 3A, lanes 1, 5, 8, 11; B,  lanes 1 and 4; C, lanes 1 and 5). However, strong proteinprotein interactions were observed between Pax6 and cVax (Fig. 3A, lane 6) as well as between Pax6 and Tbx5 (Fig. 3A,  lanes 9 and 12).
To identify which domains of Pax6 are required for physical interaction with Tbx5 and cVax, we used GST fusion proteins containing either the paired domain from the p46 (GST-paired; amino acids 3-131) or the p46 homeodomain (GST-homeo; amino acids 224 -284). Both the Pax6 paired and homeodomains were independently able to bind the radiolabeled Tbx5 (Fig. 3B, lanes 2 and 3) or cVax (Fig. 3B, lanes 5 and 6), indicating that both domains are involved in Pax6/Tbx5 and Pax6/cVax interactions. Finally, we tested these interactions with the truncated form p46⌬48 -211, which did not alter eye morphology and did not change dorsoventral marker expression patterns. Fig. 3C shows that the interactions observed with the p46 wild type protein (Fig. 3C, lanes 2 and 3) as well as with the paired-less proteins (arrowhead) produced by internal initiation (13) were almost completely lost with the p46⌬48 -211 protein (Fig. 3C, lanes 6 and 7), indicating that cVax/Pax6 or cVax/Tbx5 protein-protein interactions do not involve the transactivation domain but require the paired and/or the homeodomains.
The Highest Level of Pax6 Expression in the Chick Optic Cup Corresponds to the Boundary between cVax and Tbx5 Territories-It has been previously shown that Pax6 is expressed in a distal (high) proximal (low) gradient in the developing mouse eye. This pattern of expression appears at the optic cup stage and results from the activity of the conserved ␣ enhancer in intron 4 of the Pax6 genomic locus (21). To compare the spatial expression profiles of the early known markers of dorsoventral polarity cVax and Tbx5 with Pax6 in the chick optic cup, we performed whole-mount in situ hybridization of E3 chick embryos. As shown in Fig. 4A, Tbx5 and cVax were expressed in non-overlapping domains in the optic cup; Tbx5 was restricted to the dorsal part of the retina, whereas cVax was present in the ventral retina and the optic stalk. In addition, there was a white narrow band at the nasal/temporal level of the retina separating cVax and Tbx5 expression domains (Fig. 4A, arrows). Interestingly, this cVax/Tbx5 boundary corresponded to the highest activity of Pax6 expression in the optic cup (Fig. 4B,  arrows). In summary, at the early optic cup stage (HH18 -19), the retina is divided into four domains of restricted gene expression along the dorsoventral axis; they are a first dorsal domain where Pax6 and Tbx5 are coexpressed, a second domain where only Pax6 is strongly expressed, a third domain where Pax6 and cVax are coexpressed, and finally, a ventral domain where only cVax is present (Fig. 4C).
cVax Represses Whereas Tbx5 Increases Pax6 ␣ Enhancer Activity in Vitro-Because Pax6 activity via the intronic ␣ enhancer contributes to the establishment of Pax6 gradient expression in the mouse retina (21), we next asked whether cVax or Tbx5 could directly modulate Pax6 regulatory sequences. We used two constructs containing the thymidine kinase (TK) promoter driving the CAT reporter gene alone (pTKCAT) or linked to the Pax6 ␣ enhancer (pTKCAT-EP). E8 quail neuroretina cells were transfected with cVax or Tbx5 together with the CAT constructs. Compared with the basal activity observed with the TK promoter alone, transfection of the ␣ enhancer alone resulted in a 4-fold increase in CAT activity (Fig. 5A). The CAT activity of the ␣ enhancer was reduced in presence of cVax in a dose-dependent manner (Fig.  5A). By contrast, Tbx5 expression increased the activity of the ␣ enhancer (Fig. 5A). These data show that cVax and Tbx5 were able to modulate Pax6 ␣ enhancer activity but in an opposite way, suggesting that they might contribute to the establishment of the Pax6 gradient.
Because cVax and Tbx5 proteins were able to physically interact with Pax6, we then asked whether these interactions could modulate the Pax6 transcriptional activity in vitro. We used as a Pax6 target gene the glucagon promoter linked to the CAT reporter gene (23). cVax and Tbx5 were cotransfected with Pax6 in BHK21 cells together with the glucagon promoter. Compared with the vector control, Pax6 expression resulted in a 35-fold increase of CAT activity (Fig. 5B). Co-transfection of cVax with Pax6 resulted in a reduction of the Pax6 transcriptional effect in a dose-dependent manner (Fig. 5B). Co-transfection of Pax6 with increasing amounts of Tbx5 did not change  2 and 3). However, strong interactions were observed between cVax and p46 (lane 6) or between Tbx5 and p46 (lanes 9 and 12). Lanes 4 and 7, 10 and 13, 20% input. B, Tbx5 and cVax interact with both p46 paired and homeo DNA-binding domains. No retention was observed with GST alone (lanes 1 and 4). Both Tbx5 and cVax were able to bind the p46 paired domain (prd, lanes 2 and 5) and homeodomain (Hom, lanes 3 and 6). C, the Pax6 transactivation domain is not involved in Pax6/Tbx5 or Pax6/cVax interactions. The 35 S-labeled p46 (lanes 1-4) or p46⌬48 -211 (lanes 5-8) proteins were incubated with GST proteins as indicated. GST alone showed no interaction with p46 (lane 1) or with p46⌬48 -211 (lane 5). Although cVax and Tbx5 interact with the full-length protein p46 (lanes 2 and 3), they were not able to interact with p46⌬48 -211, indicating that the p46 transactivation domain is not required for these interactions. It is interesting to note that the p30 paired-less protein (indicated by an arrowhead), which is synthesized together with the p46 product because of an internal AUG initiation codon (12,22), also binds to the GST-cVax or to the GST-Tbx5. its transactivation level (Fig. 5B). These data show that cVax, but not Tbx5, was able to repress Pax6 transcriptional activity in vitro.

DISCUSSION
Our data provide evidence that overexpression of Pax6 at the optic cup stage caused dorsalization of the developing eye by up-regulating Tbx5 and down-regulating cVax. Tbx5 and cVax interact with Pax6 and modulate in an opposite manner the activity of the Pax6 ␣ enhancer, which is essential for its gradient expression. Most importantly, the present results suggest that Pax6 is involved in delimiting a boundary between the dorsal and the ventral compartments of the developing eye.
Molecular Dorsalization of the Developing Eye after Pax6 Overexpression-In this study we report that overexpression of Pax6 promoted dorsalization of the optic cup; extension of the dorsal markers, Tbx5 and Bmp4, and reduction of the ventral marker cVax were observed in the developing eye after Pax6 overexpression. Consistent with our data, it was previously reported that misexpression of Tbx5 (7) or Bmp4 (7, 29) caused dorsalization of the eye. Conversely, misexpression of cVax has been shown to ventralize the retina (8). Previously, BMP4 and the morphogen sonic hedgehog have been involved in patterning the dorsoventral axis of the eye by antagonistic actions (29 -31). Interestingly, we observe that Pax6 electroporation produced similar effects to those found after Bmp4 overexpression (29) or after blocking Sonic hedgehog activity by antibodies (31). In all cases, the expression of dorsal markers was expanded, whereas ventral markers were reduced, and similar morphological defects (eye deformation and rotation toward the ventral midline) were observed. Interestingly, oval-shaped eyes were also observed after overexpression of Tbx5 (7). Because of Tbx5 up-regulation and cVax down-regulation in Pax6 misexpressing eyes, the morphological defects that we observe could result from an abnormal development of the dorsal retina at the expense of the ventral retina. In the Pax6Ϫ/Ϫ mutant mouse (Sey), Tbx5 is no longer expressed, whereas Vax1/Vax2 extend to the entire optic vesicle (21). Thus, the absence of Pax6 induces ventralization of the eye, whereas overexpression of Pax6 causes dorsalization of the eye, suggesting that, during normal development, Pax6 may activate Tbx5 and repress Vax.
Pax6 Interacts with Tbx5 and cVax for Dorsoventral Patterning of the Eye-What are the molecular mechanisms by which Pax6 regulates Tbx5 and cVax activity ? Our data suggest that the observed eye phenotypes arise as a result of Pax6 protein interactions with cVax and Tbx5 rather than by a direct effect of Pax6 on cVax or Tbx5 regulatory sequences. First, we observed that all the Pax6 isoforms we electroporated (p46, p48, and p30) were able to dorsalize the eye. The p46 and p48 paired domains bind distinct DNA sequences, which are also different from the one recognized by the homeodomain (32). Therefore, it is very unlikely that these Pax6 isoforms share the same target genes. Second, the p46⌬170 -342, which is no longer able to transactivate (12), was still able to dorsalize the eye, suggesting that this phenotype was independent from Pax6 transcriptional activation abilities. Thus, we favored the hypothesis that protein-protein interactions are involved in eye dorsalization. Indeed, we demonstrated that Pax6 protein was able to interact with both Tbx5 and cVax proteins. We observed that the two Pax6 DNA binding domains (paired and homeodomain) were also able to bind both Tbx5 and cVax proteins. Consistent with this, both the p30 isoform, which contains only the homeodomain, and the p46⌬170 -342 construct, which contains only the paired domain, were able to dorsalize the eye. In addition, the p46⌬48 -211 construct, which lacks both DNA binding domains but contains the entire transactivation domain, did not interact with cVax or Tbx5. We did not observe any abnormal eye phenotypes or alterations in cVax/Tbx5 expression patterns after p46⌬48 -211 electroporation. This suggests that these protein-protein interactions occur via the paired, the homeodomain, or both. At the functional level, we showed that Pax6/ cVax interaction inhibited Pax6 transactivation properties. Because Pax6 is able to transactivate its own promoters (20), inhibition of Pax6 activity through cVax interaction contributes to Pax6 down-regulation, also mediated by the c-Vax repression of the intronic ␣ enhancer in the retina. Pax6/cVax interaction may, therefore, play an important role in the early determination of the dorsal versus ventral retina.
A Gradient of Pax6 Expression May Be Required for Establishing Dorsal-Ventral Boundary in the Developing Eye-Pax6 has been shown to be expressed in a gradient in the developing mouse eye after activation of an intronic ␣ enhancer. Before optic cup formation, Pax6 is widely expressed in the optic vesicle and the surface ectoderm. Then, at the optic cup stage, Pax6 becomes expressed in distal high, proximal low gradient (21). Interestingly, it is precisely at this stage that Tbx5 and cVax become restricted to the dorsal and ventral part of the retina, respectively. We observed the same expression patterns in the chick optic cup. In the chick eye, Tbx5 expression is first detected throughout the retina at stage 11 then becomes confined to the dorsal part from stage 14 (7). In contrast to Tbx5, cVax expression extends from the telencephalic ventral midline toward the optic vesicles, where it is exclusively expressed in the ventral retina from stage 14 (8). Thus, during normal eye development, cVax is never expressed in the dorsal part of the eye. Because in the Pax6 mutant, Vax1 and -2 expression extends over the entire optic vesicle, Pax6 may be involved in restricting Vax expression to the ventral retina, thus inhibiting Vax extension at the dorsal-ventral boundary. Our data support this hypothesis. First, we observed that the highest level of Pax6 gradient expression corresponds to the boundary between dorsal and ventral compartments. Second, Pax6 overexpression in vivo caused cVax repression and Tbx5 up-regulation. Third, cVax inhibited the activity of the ␣ enhancer, which is important for the establishment of a Pax6 gradient in the optic cup. Fourth, we observed an inhibitory interaction at the protein level between cVax and Pax6. These data taken together suggest that the Pax6 gradient of expression would be involved in separating dorsal and ventral compartments of the optic cup. The highest level of Pax6 expression may be required for forming the dorsal/ventral boundary in the developing eye, acting as a barrier between dorsal and ventral determinants by promoting dorsal fate (Tbx5, BMP4) and repressing ventral fate (cVax). Interestingly, Pax6 has been involved in the dorsoventral patterning of several neural tissues including spinal cord (33), telencephalon (34), and pituitary gland (35). For example, transient dorsal expression of Pax6 during pituitary gland development is required for establishing a sharp boundary between dorsal and ventral cell types (35).
These findings highlight the importance of Pax6 in early eye development. The role of Pax6 in morphogenesis and retinal dorsoventral axis formation underlines the fact that Pax6 acts not only as a transcription factor but also through interactions with others transcription factors.