Ocular development begins in vertebrates with the separation of the eye field into two optic primordia and the formation of bilaterally symmetrical evaginations of the ventral forebrain, the optic vesicles. The expanding optic vesicles then induce the surface ectoderm to form the lens primordia, the lens placode. After contact of the optic vesicles with the lens placode, both tissues invaginate to form a two-layered optic cup and lens vesicle, respectively. As development proceeds, the optic stalk is formed proximally and the distal optic cup surrounds the lens. The outer layer of the optic cup forms the retinal pigmented epithelium (RPE) and the inner layer forms the neural retina that consists of multipotent progenitor cells that eventually produce six neuronal cell types and one glial cell type in the adult retina. The determination of proximodistal and dorsoventral territories of the eye occurs very early at the optic vesicle stage (Chow and Lang,2001; Uemonsa et al.,2002), and eye problems at later stages, in fact, may have their origin in aberrations of this early patterning event.
The different events involved in eye development are controlled by transcription factors whose strict expression patterns and regulation within defined populations of cells are required for appropriate patterning (Chow and Lang,2001). Pax6 is widely regarded as the “master control” gene of eye development, because it is required for the formation of eye structures in animals ranging from flies to humans (Ashery-Padan and Gruss,2001; Pichaud et al.,2001; Pichaud and Desplan,2002). In the mouse, Pax6 is coexpressed early in development with Pax2, after which time they corepress each other to become distinctly expressed in the optic cup and optic stalk, respectively (Schwarz et al.,2000; Baumer et al.,2003). Pax6 null mice do not develop optic cups, and Pax2 null mice display exaggerated optic cup formation at the expense of optic stalk (Schwarz et al.,2000). The expression levels of Pax6, in particular, must be regulated over a very narrow range, because both decreased and increased levels can lead to ocular defects (Schedl et al.,1996). The Vax family of homeodomain transcription factors is also required for appropriate development of the ventral optic vesicle and optic stalk. Vax1 is normally expressed in the optic stalk and ventral optic cup, and its absence leads to optic cup coloboma and proximal expansion of the Pax6 territory (Hallonet et al.,1999). Similar to Pax2, the expression of Vax1 is necessary to repress Pax6 expression to define distinct optic cup and optic stalk territories (Barbieri et al.,1999; Mui et al.,2002; Take-uchi et al.,2003), although Pax2 and Vax1 function independently in parallel pathways. Vax2 expression in the ventral optic cup is required for closure of the optic fissure, determination of eye polarity, and maintenance of appropriate axon connections to the brain (Barbieri et al.,1999; Mui et al.,2002). In Vax1;Vax2 double null mice, all optic stalk/nerve characteristics are lost in favor of retinal cell types due to proximal expansion of the Pax6 expression territory (Mui et al.,2005). The expression of Vax2 in the ventral optic cup is believed to repress that of the T-box transcription factor Tbx5 that is expressed dorsally as a result of bone morphogenetic protein (BMP) signaling (Sasagawa et al.,2002).
The secreted morphogen Sonic hedgehog (Shh) plays a role in ocular development during two temporally and spatially distinct events. In mice, ocular morphogenesis is initiated at embryonic day 8.0 (E8.0) by Shh secreted from the prechordal plate and then the ventral forebrain (midline) (Marti et al.,1995; Li et al.,1997), which separates the eye field and forms the bilaterally symmetrical optic vesicles. Shh null mice exhibit severe holoprosencephaly characterized by cyclopia and a proboscis-like structure on the face (Chiang et al.,1996), which is a similar phenotype to that observed in some human cases where Shh mutations occur (Roessler et al.,1996; Schell-Apacik et al.,2003). In the BF-1 mouse, specific deletion of Shh from the ventral telencephalon results in optic stalk mispatterning in the presence of ventral diencephalic Shh (Huh et al.,1999). The particular importance of this pathway in ventral patterning of the visual system was discovered in humans where Shh gene mutations can cause a less-severe ocular coloboma characterized by incomplete closure of the choroid fissure (Schimmenti et al.,2003). Optimal outgrowth of the optic vesicles involves the coordinate activity of several signals, including Shh (Ohkubo et al.,2002). When Shh expression was abrogated in the developing chick, decreased cell proliferation, increased cell death, and hypoplasia was observed in the optic vesicles (Ohkubo et al.,2002). In zebrafish and Xenopus, midline Shh signaling regulates the expression of Pax and Vax genes to establish proximodistal and dorsoventral characteristics of the eye (Ekker et al.,1995; Macdonald et al.,1995; Take-uchi et al.,2003; Lupo et al.,2005). A later Shh signaling event, independent of the midline, follows the onset of neural retina development that occurs in the mouse at E12. The first-born neuronal cells in the retina, the retinal ganglion cells, express the Shh protein that, in the retina, signals to retinal precursor cells to control cell proliferation, differentiation, and organization (Jensen and Wallace,1997; Levine et al.,1997; Zhang and Yang,2001; Wang et al.,2002,2005), as well as influencing the development of the optic nerve (Wallace and Raff,1999; Dakubo et al.,2003).
The Shh signaling pathway is well conserved in organisms ranging from flies to humans (Ingham and McMahon,2001). The binding of Shh to its membrane receptor Patched (Ptch) relieves its inhibitory action on Smoothened (Smo), hence promoting intracellular signaling events. The downstream effectors of the pathway are the Gli transcription factors, of which three are known to exist in vertebrates (Gli1, Gli2, and Gli3; Hui et al.,1994). In mice, the Shh pathway is mediated primarily by Gli2 and Gli3 that can be modified post-translationally to either activate or repress target gene expression (Ruiz i Altaba,1999a,b; Sasaki et al.,1999; Aza-Blanc et al.,2000; Ruiz i Altaba et al.,2002). Whereas Gli2 is widely regarded as being a strong activator and weak repressor of target gene expression, Gli3 is a strong repressor of target genes. As a direct target of the Shh pathway, Gli1 is regulated transcriptionally in the mouse and the resulting Gli1 protein plays a complementary activator role to primary pathway activation (Bai et al.,2002). Gli1 null mice are viable and fertile as adults, indicating that Gli1 function is redundant in the presence of Gli2 and Gli3 (Bai et al.,2002). Gli2 null mice, on the other hand, display central nervous system (CNS) midline patterning defects (Ding et al.,1998; Matise et al.,1998) and Gli3 null mice (i.e., Gli3xt/xt) display several ocular defects ranging from anophthalmia to microphthalmia, and coloboma (Franz and Besecke,1991; Hui and Joyner,1993). In humans, point mutations in the Gli3 gene lead to Greig cephalopolysyndactyly, a condition often characterized by hypertelorism, which is an exaggerated separation of the eyes (Wild et al.,1997). The roles of the different Glis can vary between tissues and developmental time points, and it is likely a combination of Gli function that ultimately determines the final outcome of the initial Shh signal.
Shh pathway abnormalities are known to lead to developmental defects and tumors, although the effect of mutations of this pathway on ocular development and vision-related conditions remains poorly understood. It has been well documented that the ablation of midline Shh activity leads to serious early ocular patterning defects, but the roles of the downstream effectors of Shh have not been closely examined. The purpose of this study was to determine the roles of the three vertebrate Glis in mediating early ocular development. We examined the morphological phenotypes in the developing eyes of Shh and Gli combination mutant mice and monitored the expression of important patterning genes using in situ hybridization. We demonstrate that Shh antagonizes Gli3 repression at the midline to promote bilateral eye development, but complementary pathway mediation by all three Glis is required to pattern the proximal eye structures. Gli3 is required to specify the dorsal optic stalk and to distinguish the retina and optic nerve territories. Gli2 and Gli1 play complementary roles to Gli3 in promoting ventral and dorsal optic stalk patterning, but are not required to define the optic stalk/optic cup boundary.
Temporal and Spatial Expression of Shh Pathway Genes During Early Ocular Development
To understand the role of the Hedgehog (Hh) pathway in early murine ocular development, we first monitored by in situ hybridization the expression of the important pathway genes, i.e., Shh, Ptch1, Gli1, Gli2, and Gli3, from the optic vesicle stage at E9.5 to the optic stalk/optic cup stage at E11.5 (Fig. 1). The expression of the universal Hh target genes Ptch1 and Gli1, in particular, indicated the possible range of action of midline Shh in patterning the optic vesicle. The specific and overlapping areas of Gli1, Gli2, and Gli3 expression indicated where these transcription factors are likely influencing downstream target genes during early visual system development and served as a reference to indicate where specific Gli activity was lacking in the Gli mutant mice examined in this study.
At E9.5, Shh is expressed in the ventral forebrain where it persists throughout development (Fig. 1A,F,K,P). In response to midline Hh activity, the membrane receptor Ptch1 is expressed throughout the ventral forebrain and proximal optic vesicles (Fig. 1B). Ptch1 expression remained increased adjacent to the midline but was absent in the cup up to E11.5 (Fig. 1G,L,Q). Gli2 and Gli3 were expressed throughout the optic vesicle, ventral optic stalk, dorsal optic stalk, and optic cup during early development, but were excluded from the Shh-expressing midline cells (Fig. 1D,E,I,J,N,O,S,T). The expression of Gli2 and Gli3 are largely overlapping until E11.5, where Gli2 expression is restricted to the proximal optic stalk and optic cup (see arrows Fig. 1S,U) and Gli3 is expressed along the proximodistal axis with low expression adjacent to the midline and highest expression in the distal optic stalk near the junction with the cup (see arrows in Fig. 1T,V). Also at E11.5, Gli2 and Gli3 expression persists in the dorsal optic stalk. As a direct target of the Hh pathway, the expression of Gli1 is strictly dependent on Hh activity mediated by Gli2 and Gli3 and, thus, was a reliable indicator of those cells that were being acted upon by the morphogen. Gli1 was initially expressed throughout the optic vesicle adjacent to the midline and surface ectoderm (Fig. 1C). After contact of the optic vesicle with the surface ectoderm at E10, Gli1 was expressed adjacent to the midline in the ventral optic stalk, optic cup, and lens vesicle, but was absent in the presumptive dorsal optic stalk (Fig. 1H). As ocular genesis continued at E10.5, Gli1 expression became restricted to the optic stalk and ventral optic cup, but was absent in the lens (Fig. 1M). By E11.5, Gli1 expression was restricted to the same territory as Gli2 adjacent to the midline source of Hh and was completely absent from the optic cup and distal optic stalk (Fig. 1R). Gli1 expression in the extraocular mesenchyme, outside of the RPE, is a result of signaling from Indian hedgehog (Ihh) and not midline Shh (Wallace and Raff,1999; Dakubo et al.,2003). By E12.5, Gli1 is expressed in the developing neural retina, corresponding with the differentiation of the first-born Shh-expressing retinal ganglion cells in the inner-most part of the neural retina (Jensen and Wallace,1997; Wang et al.,2005).
Shh Alleviates Gli3 Repressor Activity at the Midline
Shh signaling at the midline is required for eye field separation, as demonstrated by the cyclopic phenotype in Shh null mice (Chiang et al.,1996). The eye structures that form in the absence of Shh are very small, pigmented, and rudimentary; therefore, no gene expression or patterning data are presented in this study. However, the removal of one Gli3 allele from Shh null mice (i.e., Shh−/−Gli3+/,xt mice) can partially rescue the cyclopic phenotype because two fused optic cups were formed, as were two fused stalks that were attached to the brain (Fig. 2C,K). Remarkably, Pax6 and Chx10 are expressed normally in the cups of these mutants (Fig. 2C,W), and they display normal dorsoventral polarity where Vax2 is expressed in the ventral cup and Tbx5 is expressed at the opposite pole of Vax2 (Fig. 2O,S). Pax2 is expressed at low levels in the optic stalks (data not shown) and in the ventral pole of the cup (Fig. 2G), whereas Vax1 expression is confined to the optic stalks (Fig. 2K). Gli3xt/xt mice exhibit several early patterning phenotypes, including anophthalmia (Franz and Besecke,1991), so only those mutants that developed eyes were included in this study. Gli3xt/xt mice have separated eye fields that exhibit proximodistal patterning defects with persistent expression of the optic stalk markers (Vax1 and Pax2) in the ventral optic cup at the expense of the cup markers (Pax6 and Chx10; Fig. 2B,F,J,V). Dorsoventral polarity, however, is established appropriately in Gli3xt/xt mice (Fig. 2N,R). Shh−/−Gli3xt/xt mice develop rudimentary eyes that have separated from each other to a greater extent than Shh−/−Gli3+/xt mice, but whose optic cups and lenses are poorly defined (Fig. 2D). Pax6 is expressed in the thin retinal tissue in the optic cups and also in the lens (Fig. 2D), but Chx10 expression was not detected (Fig. 2X). Pax2 expression was evident in the ventral optic cup, but expression of both Pax2 and Vax1 was decreased in the rudimentary optic stalks (Fig. 2H,L). Vax2 was expressed at low levels in the ventral optic cup, but no Tbx5 expression was observed in these mutants (Fig. 2P,T). Although the loss of Gli3 in Shh−/− mice can partially rescue the cyclopic phenotype, its absence is not sufficient to appropriately separate and pattern the early eye structures.
Complementary Gli Activity Patterns the Early Eye Structures
In addition to examining the specific relationship between Shh and Gli3 at the midline, we also examined the combinatorial disruption of the three known vertebrate Gli genes to dissect the roles of these pathway components during early ocular development. This was achieved through analysis of Gli2;Gli3 and Gli1;Gli2 combination mutant embryos. In these embryos, an active midline Shh signal was present, but downstream gene activation was differentially compromised. Although the expression territories of the proximodistal and dorsoventral patterning genes were appropriate in most cases (Figs. 3, 4), measuring the lengths of the ventral optic stalk (midline to cup), the dorsal optic stalk (diencephalon to cup), and the optic cup (axial length) revealed significant roles of the Glis in patterning these early transient structures (see Table 1).
Table 1. Length of Ventral and Dorsal Optic Stalk and Optic Cup in Embryonic Day 11.5 Gli Combination Mutant Embryos
Wild-type (n = 4)
Gli1+/+Gli2−/− (n = 4)
Gli1−/−Gli2+/− (n = 4)
Gli1+/−Gli2−/− (n = 3)
Gli1−/−Gli2−/− (n = 3)
Represents a significant difference from wild-type by Student's t-test (P < 0.05).
The Gli2;Gli3 combination mutant embryos were representative of the abrogation of primary Hh pathway mediation, although it must be noted that Gli1 is likely playing a role in most mutants, as a result of even minimal transcriptional activation of this pathway component. In the absence of Gli2 alone at E11.5, the bilateral separation of the eyes was attenuated compared with wild-type (Fig. 3B; Table 1), although all patterning genes were expressed normally. The Gli2−/−Gli3+/xt embryos exhibited a more severe phenotype with a shortened optic stalk and larger optic cup (Fig. 3C; Table 1). In Gli+/−Gli3xt/xt embryos, the optic cup and lens structures were elongated proximodistally but the optic cup/optic stalk boundary was well defined (Fig. 3D). Interestingly, in the absence of both Gli2 and Gli3, two optic cups formed, but the optic stalks failed to form properly (Fig. 3E). Expression of Pax2 and Vax1 persisted in the optic cup in regions where Pax6 and Chx10 expression was absent and the ventral optic stalk was shortened and rudimentary (Fig. 3E,J,O,Z1111). The dorsal optic stalk that normally extends between the diencephalon and the dorsal optic cup was attenuated in Gli2−/−Gli3+/+ and Gli2−/−Gli3+/xt but not detectable in Gli2−/−Gli3xt/xt and Gli+/−Gli3xt/xt mutants (Fig. 3; Table 1). All mutants displayed varying proximodistal patterning defects, but dorsoventral polarity was well established in all optic cups, as indicated by ventral Vax2 and dorsal Tbx5 expression (Fig. 2P–Y).
Assuming that Gli1, Gli2, and Gli3 are the sole mediators of the Hh signal in vertebrates, the role of Gli3 in early ocular patterning can be isolated in Gli1−/−Gli2−/− mutants (Fig. 4). Proximodistal and dorsoventral markers were clearly expressed in the correct domains in all Gli1;Gli2 combination mutants, but patterning of the ventral and dorsal optic stalk did not occur appropriately (Table 1). Although Gli2−/− mutants exhibit shorter ventral and dorsal optic stalks compared with wild-type, this phenotype is exaggerated as the Gli1 alleles are removed (see Fig. 4; Tables 1, 2). In Gli1+/−Gli2−/− mice, the ventral optic stalks projected laterally instead of dorsolaterally as in the wild-type (compare Fig. 4E and G), and the domain of the Pax2- and Vax1-expressing dorsal optic stalk was shortened (Fig. 2G). The Gli1−/−Gli2−/− phenotypes varied in the embryos studied, with some ventral territories displaying minimal optic stalk outgrowth from the midline. The absence of Gli1 alone did not result in any observable patterning defects (data not shown). Also, Tbx5 was expressed in the dorsal optic cup and Chx10 was observed throughout the optic cups of all Gli1;Gli2 mutants analyzed (data not shown). The shortened ventral optic stalks were manifested as shortened optic nerves, measured in the different mutants as the distance between the optic cup and the optic chiasm (Table 2). In the absence of both Gli1 and Gli2, Gli3 alone can initiate bilateral eye development and define the optic cup/optic stalk boundary, but Gli1 and Gli2 are also required in complementary roles for appropriate outgrowth and specification of the ventral and dorsal optic stalks.
Table 2. Optic Nerve Lengths in Gli1;Gli2 Combination Mutant Mice at Embryonic Day 14.5
Wild-type (n = 4)
Gli1+/+Gli2−/− (n = 4)
Gli1−/−Gli2+/− (n = 4)
Gli1+/−Gli2−/− (n = 3)
Gli1−/−Gli2−/− (n = 3)
Represents a significant difference from wild type by Student's t-test (P < 0.05).
Optic Nerve Head Fails to Develop in the Absence of Shh
By E14.5, the optic cup and optic stalk are further developed and form the neural retina and optic nerve, respectively. At this stage, the retina is surrounded by RPE to form the eye, which is distinct from the optic nerve (Fig. 5A,F). For appropriate comparison between mutants, horizontal eye sections were collected and presented from the area of the optic nerve head that is distinguished by the presence of the triangle-shaped territories consisting of Pax2-expressing astrocyte precursor cells (Fig. 5). This figure is representative of the ventral territory of the retina and may not accurately reflect gene expression in the dorsal retina, but all figures are comparable to each other. At this developmental stage, Pax6 is expressed only in the neural retina and lens epithelium, whereas Pax2 expression is restricted to astrocytes in the optic nerve and the optic nerve head (Fig. 5).
In Gli2−/−Gli3+/xt mutants, the optic nerve heads were well defined and the only consistent phenotype was a mild elongation of the retina and RPE along the proximodistal axis (Fig. 5B,G). In Gli2+/−Gli3xt/xt mutants, Pax2 expression was localized to the well-formed optic nerve head but the neural retina and RPE were expanded proximally (Fig. 5C). Of all mutants examined, the Gli3xt/xt phenotypes varied most significantly, with many lacking eye structures entirely, so only those that developed eyes were included in this analysis. The lack of definition between the optic nerve head and neural retina is clearly seen in Gli3xt/xt eyes at E14.5 (Fig. 5E,J). The ventral retinas clearly show a distal expansion of Pax2 at the expense of Pax6, and the RPE extends proximally along the optic nerve. Interestingly, the retinas and lenses of Shh−/−Gli3+/xt mutants continue to develop and grow from E11.5 to E14.5. Pax6 is appropriately expressed in the neural retinas and lens epithelia, but there was no evidence for the development of specialized Pax2-expressing astrocyte precursor cells at the optic nerve head (Fig. 5D,I). Gli2−/−Gli3xt/xt mutant embryos were not observed at E14.5, because they die earlier in development and Shh−/−Gli3xt/xt mutant eye structures were either absent or too rudimentary at this age. In all Gli1;Gli2 combination mutants examined, including Gli1−/−Gli2−/− embryos, the eye is shaped normally and the optic nerve heads are well formed (data not shown). Therefore, Gli3 alone is sufficient to pattern the optic nerve/neural retina boundary in the absence of Gli2 and Gli1.
Ocular development is a complex process that involves different signaling events and the strict spatial and temporal expression of various patterning genes. The importance of Shh from the CNS midline in controlling early ocular development has been shown in several vertebrate systems (Amato et al.,2004), but the differential roles of the pathway mediators, the Gli transcription factors, remains poorly understood. We show that Gli1, Gli2, and Gli3 are differentially expressed in the early ocular structures and mediate the Hh pathway in a nonredundant manner. The observation that unique or combinatorial disruption of Gli activity can affect different aspects of early ocular patterning (see summary of phenotypes in Table 3) clearly shows that strict regulation of this pathway is critical to the normal development of the eye.
Table 3. Summary of Ocular Phenotypes in Shh and Gli Combination Mutant Mice
RPE, retinal pigmented epithelium.
Rudimentary pigmented eye structure
Two fused retinas attached to the brain by two fused optic stalks; normal lenses; normal RPE development; no optic nerve head
Rudimentary optic cups and optic stalks
Proximodistally elongated eye cups; no dorsal optic stalk
Shortened and mispatterned ventral optic stalk; shortened dorsal optic stalk
Shortened and mispatterned ventral optic stalk; shortened dorsal optic stalk
Shh Relieves Gli3 Repression at the Midline
Shh from the midline is absolutely required for early patterning of the mammalian visual system (Chiang et al.,1996; Roessler et al.,1996). The small pigmented eye structures of Shh null mice are formed from a rudimentary optic vesicle that is deficient in the territories that give rise to the optic cup and optic stalks (Chiang et al.,1996). The significant rescue of the Shh null phenotype observed when removing one Gli3 allele showed that decreasing Gli3 repressor activity adjacent to the midline was permissive for distal ocular patterning, likely through the coordinate activities of other factors such as the BMPs, retinoic acid, fibroblast growth factors (FGFs), and Wnt pathway components (Ohkubo et al.,2002; Cavodeassi et al.,2005; Lupo et al.,2005). In the absence of a Hh signal, the full-length activator (or derepressor) form of Gli3 is proteolytically cleaved into its dominant repressor form (Wang et al.,2000), so the primary role of Shh is to antagonize Gli3-mediated target gene repression adjacent to the midline, which promotes eye field separation by suppressing Pax6 expression (Li et al.,1997) and permitting bilateral patterning of the proximal ocular structures (Rallu et al.,2002). The underlying Shh signal from the prechordal plate serves to suppress Pax6 expression medially to form the eye fields, which accounts for the normal levels of Pax6 expression in Shh−/−Gli3+/xt optic cups and retinas (Fig. 2C). Consequently, although it was beyond the scope of this particular study, the normal expression of Pax6 was sufficient for the development of most early born retinal cell types at E14.5, even in the absence of a midline Shh signal (M. Furimsky, unpublished observations).
Although bilateral separation of the optic cups was significantly rescued in Shh−/−Gli3xt/xt compared with Shh−/−Gli3+/xt embryos, the poorly formed eyes suggest that Gli3 repressor activity is required for normal patterning of the distal ocular structures (Fig. 2D). It is very interesting, however, that the expression of Chx10 was not detected in the optic cup of the Shh−/−Gli3xt/xt mutants even though it was expressed in Shh−/−Gli3+/xt cups (Fig. 2W,X). Because Pax6 is expressed in the thin optic cup, it is possible that there is a requirement for Hh pathway activation to express Chx10 downstream of Pax6. The Shh−/−Gli3xt/xt mutants show that overcoming Gli3 repressor activity at the midline is sufficient for bilateral eye development but that patterning along the proximodistal axis is aberrant and retinal precursor cells are not specified.
Complementary Gli Activity Mediates Midline Shh Signaling
Antagonizing Gli3 repressor activity at the midline is not the exclusive role of Shh, but rather one consequence of Hh pathway activation that also includes the transcriptional activation of target genes by means of all three Gli proteins. In this study, we cannot discriminate between diencephalic and telencephalic derived Shh (Huh et al.,1999), so it is likely that the phenotypes we observe arise because of aberrant signaling from one or both of these sources of the morphogen. The pattern of expression of Hh pathway genes at early stages of ocular development suggest that a gradient of Shh activity, reflected by expression of the target gene Gli1, is initially involved in patterning the optic vesicle, optic stalk, and optic cup (Fig. 1). In developing neural tissues, Hh signaling is known to establish a gradient of Gli activity that specifies different cell types (Stamataki et al.,2005), so it is possible that this graded activity is establishing the cellular characteristics along the proximodistal axis of the visual system. This active signaling gradient, however, is strongly opposed by Gli3 repressor activity in the distal territory of the ventral optic stalk. Similar to the developing neural tube where Gli3 expression is highest dorsally to restrict the ventral Shh signal (Litingtung and Chiang,2000; Stone and Rosenthal,2000; Persson et al.,2002; Bai et al.,2004; Stamataki et al.,2005), Gli3 expression in the optic stalk is highest distally to restrict the proximal Shh signal that is actively mediated by Gli1, Gli2, and Gli3.
The expression of Gli1 is strictly regulated at the transcriptional level in the mouse by Gli2 and Gli3 (Bai et al.,2004), so the Gli2−/−Gli3xt/xt mutants represented a scenario where all primary Shh signal mediation was absent (Fig. 3). A phenotype similar to that of Shh null mice would be expected if Gli2 and Gli3 were acting solely as pathway activators, but optic vesicle outgrowth clearly occurred in Gli2−/−Gli3xt/xt mutants, forming an optic cup and shortened optic stalk (Fig. 3D). Unless Shh is promoting optic vesicle outgrowth through a different mediator, these findings suggest that optic cup formation requires the inhibition of all Gli-mediated repressor activity adjacent to the midline, which would be permissive for the expression of genes involved in distal patterning. In Gli2+/−Gli3xt/xt or Gli2−/−Gli3+/xt, where minimal primary signal mediation was present, significant ventral optic stalk outgrowth occurred and the optic stalk/optic cup boundaries were well defined (Figs. 3, 5; Table 1). Therefore, overcoming repressor activity at the midline is a key role of Shh, but the activation of Gli2 and Gli3 is a requirement for proximal eye pattern formation.
Because Gli3 is playing a strong repressor role that opposes the midline Shh signal distally, Gli3xt/xt mice represent a gain of midline Shh function phenotype, which is characterized by persistence of high levels of Gli1 expression in the optic cup at E11.5 (data not shown) and extension of the optic stalk territory (Vax1 and Pax2) distally into the optic cup territory (Fig. 2). Although we did not observe any differences in ventral optic stalk length between Gli3xt/xt mutants and wild-type mice, it is not surprising that the human GLI3 mutant phenotype is characterized by hypertelorism where eyes are separated from each other to a larger degree than normal due to unopposed midline Shh activity (Wild et al.,1997). In Gli2+/−Gli3xt/xt mice, the ventral optic stalks were shorter and the optic stalk/optic cup boundary was well defined compared with Gli3xt/xt (compare Figs. 2 and 3 and Fig. 5C,E). So reducing Gli2 activity at the midline is sufficient to partially compensate for the exaggerated expansion of the optic stalk territory in the absence of distal Gli3 repression.
The Gli1−/−Gli2−/− eye phenotypes indicated that Hh pathway mediation by Gli3 alone is insufficient for normal ocular development (Fig. 4; Tables 1, 2). Although the optic stalk/optic cup and optic nerve/neural retina boundaries were well defined in all Gli1;Gli2 combination mutants (see Fig. 4), the decreased length of their respective ventral optic stalks and optic nerves indicated that Gli1 and Gli2 complement Gli3 to promote normal optic vesicle and ventral optic stalk outgrowth (Fig. 4; Tables 1, 2). We could not relate these findings to changes in cell proliferation at Ell.5, however, because no significant changes in bromodeoxyuridine (BrdU) incorporation were observed when comparing the optic stalks of the Gli1;Gli2 mutants to wild-type (data not shown), although this does not preclude the possibility that there is an earlier proliferation defect. Nonetheless, our results demonstrate that there is a requirement for the activities of all three Glis to appropriately promote outgrowth of the optic vesicle and optic stalk from the midline.
Gli Activity Specifies the Dorsal Optic Stalk
Because Shh is principally involved in patterning the ventral structures in the visual system, it was somewhat surprising to discover a significant Gli requirement for the specification of the dorsal optic stalk. This transient structure, derived from specified tissue bridging the diencephalon and the dorsal optic cup, ultimately gives rise to non-neural tissue that surrounds the optic nerve (Chow and Lang,2001). Because the presumptive dorsal optic stalk is not exposed to Hh signaling after the optic vesicle stage based on Gli1 expression in this region (see Fig. 1), it is most likely that complementary Gli activity specifies this region very early and the defect is manifested later in development. All Gli1;Gli2 and Gli2;Gli3 combination mutants exhibited shortened dorsal optic stalks, and this structure fails to develop in mice devoid of all Gli3 activity (Figs. 2, 3; Table 1). It would be useful, therefore, to examine earlier development of the presumptive dorsal optic stalk at the optic vesicle stage with a focus on specific target gene expression in this area, as well as later stages in optic nerve development in Gli mutants, because early dorsal optic stalk abnormalities can result in later myelination defects.
Hh Pathway Defines the Pax2 and Vax1 Expression Territory
Midline Shh signaling promotes the expression of several eye developmental regulatory genes, including members of the Pax and Vax families of transcription factors. In studies examining early ocular patterning in the zebrafish and Xenopus, increased Shh signaling led to expansion of the Pax2, Vax1, and Vax2 expression territories (Ekker et al.,1995; Macdonald et al.,1995; Hallonet et al.,1999; Take-uchi et al.,2003; Lupo et al.,2005). A consequence of increased expression of these genes during ocular development was a reduction in Pax6 expression in the optic cup. When the Shh signal was absent in our mutants (i.e., Shh−/−Gli3+/xt or Shh−/−Gli3xt/xt), the expression of Pax2 and Vax1 was indeed greatly reduced. Clearly, the absence of strong midline Gli3 repressor activity alone was not sufficient in the Shh−/−Gli3xt/xt mutants to rescue normal expression of Pax2 and Vax1 and optic stalk patterning (Fig. 2C,D). The potential relationship between Gli3, Vax1, and Pax2 in ventral patterning and determination of the optic cup/optic stalk barrier is highlighted by the observation that the respective mutant mice display similar patterning defects, such as coloboma, extension of the RPE along the optic nerve, and lack of definition between the optic cup and optic nerve (Bertuzzi et al.,1999; Baumer et al.,2003).
Hh Does Not Influence Dorsoventral Polarity in the Mouse
In zebrafish and Xenopus, midline Shh influences the expression of Vax2 and dorsoventral polarity of the optic cup (Take-uchi et al.,2003; Lupo et al.,2005). The ventral eye cup territory is specified by the activation of Vax2, whose expression is influenced by FGFs, retinoic acid, and Hh, with increased strength of midline signaling expanding the ventral eye territory. In our study, however, none of the Gli mutants exhibited defects in dorsoventral patterning in the optic cup, because the Vax2 and Tbx5 expression territories were well established. Even in the absence of Shh (i.e., Shh−/−Gli3+/xt and Shh−/−Gli3xt/xt) Vax2 was clearly localized to the ventral cup (Fig. 2O,P). The Vax2 null mouse shows no differences in ventral Pax2 or dorsal Tbx5 expression from the wild-type but does display ocular coloboma and extended RPE, as in our Gli3 mutants (Barbieri et al.,1999; Mui et al.,2002). The lack of dorsoventral patterning phenotype in Gli mutants means that the midline signals involved in ventralizing the eye territories do not necessarily affect dorsal patterning in the mouse as they may in species such as the frog or fish.
Optic Nerve Head Development Is Independent of Midline Hh Signaling
Although proximodistal patterning of the mouse visual system is established at an early stage of development, the Hh-dependent specification of the optic nerve head (optic disc) at the boundary between the optic nerve and retina is a later independent event. The organization of Pax2-expressing astrocyte precursor cells at the optic nerve head appears to be exclusively dependent on Shh signaling from retinal ganglion cells (Dakubo et al.,2003) and is not influenced by midline Shh signaling. Even though the boundary between the retina and optic nerve is well established in Shh−/−Gli3+/xt eyes, the lack of optic nerve head specification is consistent with that observed in conditional Shh mutant mice, where the inactivation of this gene is restricted to the retina and the disc astrocytes fail to form (Dakubo et al.,2003). It was remarkable, however, that, in the severely mispatterned eyes of Gli2+/−Gli3xt/xt mutants, the optic nerve head develops normally. Only the Gli3xt/xt mutants lacked a well-defined optic nerve head at E14.5, but the presence of this Pax2-expressing structure is likely masked by the increased expression of Pax2 in the ventral retinas of these mutants. In Gli2−/−Gli3+/xt, Gli2+/−Gli3xt/xt, and Gli1−/−Gli2−/− mutants, the optic nerve head is always well established (Fig. 5G,H, and data not shown), so the roles of the Gli proteins appear redundant in specifying this structure.
Mechanism of Gli Requirement in Early Ocular Patterning
Based on the phenotypes observed in our Gli mutant mice, we propose the following model of differential Gli activity in early development of the mouse visual system (Fig. 6). Initially, Shh relieves the strong repressor activity of Gli3 adjacent to the midline to permit eye field separation and optic vesicle outgrowth. Activation of Gli2 and Gli1 plays roles complementary to Gli3 by mediating expansion of the proximal optic vesicle that ultimately gives rise to the optic stalk. Shh from the midline actively promotes patterning of the proximal optic stalk (Pax2 and Vax1 expression territory), but increased Gli3 repressor activity in the distal optic stalk is required to define the boundary between the optic stalk and optic cup.
The Shh null allele was created by targeted disruption of exon 2 of the Shh gene (St-Jacques et al.,1998), Gli1 and Gli2 alleles contain a deletion of the zinc finger domain of the gene (Mo et al.,1997) and Gli3xt alleles contain a large deletion of the 3′ end of the gene (Hui and Joyner,1993). Genetic crosses were performed to obtain Shh+/−Gli3xt/xt mice on a mixed C57/Bl6 and C3HeB/FeJ background, Gli1+/−Gli2+/− mice on a CD1 background, and Gli2+/−Gli3+/xt mice on a mixed CD1 and C3HeB/FeJ background. The embryos were genotyped using tail tissue and a polymerase chain reaction protocol using specific primers to the different genes. Mice were coupled in the late afternoon and allowed to mate overnight. The next morning, males were removed from the cage and females were checked for vaginal plugs. The presence of the vaginal plug early the next morning was considered as E0.
Genetic and Phenotypic Analysis
Embryos were harvested and tissues fixed in 4% paraformaldehyde. Tissues were further cryoprotected in 30% sucrose in Dulbecco's phosphate buffered saline (GIBCO) before embedding in equal amounts of 30% sucrose and OCT (Tissue Tek). Upper bodies or heads were isolated, depending on the age of the embryo, and tails were collected for genotyping. Tissues were sectioned for in situ hybridization (14 μm) by using a Leica CM 1850 cryostat.
All mutant embryo phenotypes were compared with littermate controls with the same mixed background. At least three litters of each combination mutant pairing were analyzed and compared at each age (E11.5 and E14.5). The figures are representative of the phenotypes and gene expression consistently observed at the different embryonic stages. All sections were analyzed using a Zeiss Axioplan upright microscope and images were captured using an AxioVision 2.05 digital video camera. All images were processed using Adobe Photoshop 7.0.
Determination of ocular structure lengths were performed with appropriate measurement tools available in Adobe Photoshop 7.0. All measurements were made along linear portions of the tissue to accurately represent the length of the tissue. Pixel lengths were then translated to numerical values knowing the field diameter of the various objectives used. Data are presented as means ± SEM, and statistical analysis was performed using Student's t-test (P < 0.05).
In Situ Hybridization
In situ hybridization was performed according to Wallace and Raff (1999). Briefly, sections were hybridized overnight with specific digoxigenin (DIG) -labeled riboprobes at 65°C in a moist chamber. Sections were washed at high stringency, incubated with an alkaline phosphatase–conjugated anti-DIG antibody and finally stained in nitro blue tetrazolium/5-bromo-4-chloro-3-indoylphosphate. The blue color indicated the regions of specific in situ gene expression. Antisense RNA riboprobes were prepared by reverse transcription from linearized plasmids containing complete or partial sequences of the mouse genes of interest: Shh, Ptch1, Gli1, Gli2, Gli3, Pax2, Pax6, Vax1, Vax2, Tbx5, Chx10. Plasmids were generous gifts from A.L. Joyner, C.C. Hui, A. P. McMahon, and I. Skerjanc.
We thank Chantal Mazerolle and Sherry Thurig for excellent technical assistance. The Gli1 and Gli2 mutant mice were kindly provided by Alex Joyner at the Skirball Institute (New York, NY). This research was supported by a grant from the National Cancer Institute of Canada (NCIC) to V.A.W. M.F. is a recipient of the CIHR/Canadian National Institute for the Blind (CNIB) E.A. Baker Foundation Postdoctoral Fellowship award.