|Home | About | Journals | Submit | Contact Us | Français|
Organogenesis of the eye is a multi-step process that starts with the formation of optic vesicles followed by invagination of the distal domain of the vesicles and the overlying lens placode resulting in morphogenesis of the optic cup. The late optic vesicle becomes patterned into distinct ocular tissues; the neural retina, retinal pigment epithelium (RPE) and optic stalk. Multiple congenital eye disorders, including anophthalmia or microphthalmia, aniridia, coloboma and retinal dysplasia, stem from disruptions in embryonic eye development. Thus, it is critical to understand the mechanisms that lead to initial specification and differentiation of ocular tissues. An accumulating number of studies demonstrate that a complex interplay between inductive signals provided by tissue-tissue interactions and cell-intrinsic factors is critical to ensure proper specification of ocular tissues as well as maintenance of RPE cell fate. While several of the extrinsic and intrinsic determinants have been identified, we are just at the beginning to understand how these signals are integrated. In addition, we know very little about the actual output of these interactions. In this chapter, we provide an update of the mechanisms controlling early steps of eye development in vertebrates, with emphasis on optic vesicle evagination, specification of neural retina and RPE at the optic vesicle stage, the process of invagination during morphogenesis of the optic cup and maintenance of the RPE cell fate.
The vertebrate eye is formed through coordinated interactions between neuroepithelium, surface ectoderm, and extraocular mesenchyme, which originates from two sources: neural crest and mesoderm. Following eye field formation, the neuroepithelium of the ventral forebrain evaginates, resulting in the formation of bilateral optic vesicles (Fig.3.1A). The distal portion of the vesicle makes contact with the overlying surface ectoderm (lens ectoderm), which is then induced to form the lens placode. This interaction results in invagination of the lens placode and distal optic vesicle leading to formation of a bilayered optic cup (Fig.3.1B). The neural retina develops from the inner layer of the optic cup, and the retinal pigment epithelium (RPE) is derived from the outer layer. The margin between the two layers gives rise to peripheral structures, the iris epithelium and ciliary body. The most proximal part of the optic vesicle, the optic stalk, “narrows” to become the optic fissure. The lens vesicle eventually separates from the surface ectoderm and differentiates into the mature lens. Tissue-tissue interactions, mediated by extracellular factors and intrinsic signals such as transcription factors, control differentiation of ocular tissues starting at the optic vesicle stage.
Much progress has been made in recent years elucidating the mechanisms involved in evagination and proximodistal patterning of the optic vesicle, and morphogenesis of the optic cup, therefore these topics will be the focus of this chapter. The reader is referred to many excellent reviews for discussions on other aspects of early eye development, such as lens and optic stalk formation, dorsoventral and nasotemporal patterning of the optic vesicle, as well as differentiation of ciliary body and iris epithelium (Adler and Canto-Soler, 2007; Davis-Silberman and Ashery-Padan, 2008; Donner et al., 2006; Hyer, 2004; Lang, 2004; Morcillo et al., 2006; Takahashi et al., 2009; Yang, 2004; Zhao et al., 2010).
The first morphological sign of eye morphogenesis is evagination of the optic vesicles, which occurs in the ventral forebrain during the final stages of neural tube formation (Fig.3.1A). Detailed analyses have revealed changes in cell behavior that take place during the evagination process. In mouse, the cellular shape of optic vesicle cells changes dramatically, accompanied by transient alterations in basal lamina composition (Svoboda and O’Shea, 1987). In fish and frogs, ocular cells undergo extensive movements that are essential for organogenesis of the eye. Fish retinal progenitor cells first display a directed movement toward the midline, followed by an outward turn into the evaginating optic vesicles (Rembold et al., 2006). It still needs to be determined whether these directed migrations occur in other vertebrates such as chick and mouse. Taken together, these findings are evidence that coordinated changes in cell shape and cellular behavior are required for evagination of the optic vesicles.
At the molecular level, it has been shown that the retinal homeodomain transcription factor Rx/RAX mediates some of these cell behaviors. Null or hypomorphic Rx alleles display anophthalmia in humans (RAX), mouse (Rx), frog (Rx1), zebrafish (Rx3/chokh) and medaka (eyeless) (Andreazzoli et al., 1999; Furukawa et al., 1997; Kennedy et al., 2004; Loosli et al., 2003; Mathers et al., 1997; Tucker et al., 2001; Voronina et al., 2004), indicating that Rx genes are essential for early eye development. Work in zebrafish and medaka indicates why early eye develoment fails. In Rx3 null mutants, optic vesicle evagination is disrupted in a cell autonomous manner; specifically, the outward directed movement of cells appears to be disturbed (Loosli et al., 2003; Rembold et al., 2006; Stigloher et al., 2006; Winkler et al., 2000). These data suggest that Rx/RAX is involved in the extensive cell movements that are integral to evagination.
How does Rx/RAX work? Information about its role can be gleaned through identification of target genes. Three lines of evidence demonstrate that the Ig-domain cell adhesion molecule, Nlcam, is a potential direct target of Rx3 in zebrafish. First, predicted Rx3 binding sites are found in the Nlcam locus (Brown et al., 2010). Second, in Rx3/chk mutants, Nlcam is ectopically upregulated in the eye field suggesting that Rx3 represses NIcam expression. Third, similar to Rx3 null mutants, overexpression of Nlcam results in smaller eyes, and cell tracking experiments revealed that mutant cells display increased convergence at the midline (Brown et al., 2010). Thus, downregulation of Nlcam expression by Rx3 may be necessary to enable retinal progenitor cells in the eye field to move away from the midline and outward to contribute to the evaginating optic vesicles. It was also recently shown that Rx-deficient cells are excluded from optic vesicle domains in embryonic mouse chimeras consisting of wild type and mutant Rx cells (Medina-Martinez et al., 2009). Together with the zebrafish data, these findings are consistent with the idea that Rx controls segregative behavior of retinal progenitor cells.
There are other possible mechanisms by which Rx/RAX may act during early eye development. Work has shown that that Rx participates in suppressing the canonical Wnt pathway to prevent the induction of posterior fates of the anterior neural plate, and it promotes non-canonical Wnt signaling that control morphogenetic movements of ocular cells (Martinez-Morales and Wittbrodt, 2009). Another potential function of Rx is the regulation of proliferation (Stigloher et al., 2006). Optx2, which controls proliferation in the eye field, is dependent on Rx function (Zuber et al., 2003; Zuber et al., 1999). Finally, Rx is essential for the expression of other key regulators of early eye formation such as Lhx2, Pax6, Mab21l2, Six3 to control, directly or indirectly, specification of retinal progenitor cells in the optic vesicle. In summary, while Rx is clearly essential during optic vesicle formation, additional potential mechanisms of Rx function need further examination.
During evagination of the optic vesicle, the neural retina and RPE domains are specified (Fig.3.2). The neural retina develops from the distal/ventral portion of the optic vesicle, while the RPE emerges from the dorsal region (Hirashima et al., 2008; Kagiyama et al., 2005). At the optic vesicle stage, the neuroepithelium is bipotential; the presumptive retina is competent to develop into RPE (Araki and Okada, 1977; Clayton et al., 1977; Horsford et al., 2005; Itoh et al., 1975; Opas et al., 2001; Rowan et al., 2004; Westenskow et al., 2010) and, conversely, the presumptive RPE can differentiate into retina (Coulombre and Coulombre, 1965; Reh and Pittack, 1995; Stroeva, 1960; Stroeva and Mitashov, 1983). Interestingly, studies in chick reveal that the dorsal and ventral portions of the optic vesicle have distinct developmental potency; after removal of the dorsal optic vesicle, the anterior ventral domain can regenerate both retina and RPE while the dorsal portion can only develop into a pigmented, RPE-like vesicle that does not invaginate (Hirashima et al., 2008). Thus, the anterior ventral domain of the optic vesicle may be the driving force for morphogenesis of the eye and proper specification of ocular tissues.
The earliest genes that show domain-specific expression for the retina and RPE are the homeobox gene Vsx2 (formerly Chx10) and the bHLH transcription factor Mitf, respectively (Fig.3.2; Burmeister et al., 1996; Green et al., 2003; Hodgkinson et al., 1993; Nguyen and Arnheiter, 2000). In mouse, Mitf is initially expressed throughout the optic vesicle and is subsequently downregulated in the distal domain when Vsx2 expression is initiated (Nguyen and Arnheiter, 2000). Both transcription factors are essential for early patterning and maintenance of cell fate in the optic vesicle (see below), and Vsx2 also controls other aspects of retinal development such as proliferation (Green et al., 2003; Sigulinsky et al., 2008).
The earliest known patterning gene is the LIM homeobox transcription factor Lhx2, which is first expressed in the eye field and is required for expression of Mitf and for retinal determinants in the optic vesicle (Zuber et al., 2003; Yun et al., 2009). In Lhx2 mouse mutants, expression of other eye field transcription factors initiates normally, but eye development arrests at the optic vesicle stage, and the lens fails to form (Porter et al., 1997; Tetreault et al., 2009; Yun et al., 2009). Recently, a more detailed analysis revealed that expression of optic vesicle regional patterning markers is severely disturbed (Yun et al., 2009). For example, the expression of Mitf, Chx10/Vsx2 and Tbx5 is never initiated, while expression of Pax2, Vax2 and Rx is initiated but not maintained. Interestingly, mosaic analysis of conditionally inactivated Lhx revealed that gene functions cell autonomously to promote Chx10 and Mitf expression (Yun et al., 2009). Thus, Lhx2 is uniquely required in the early optic vesicle for specification into both neural retina and RPE and to regulate optic cup formation (see below).
The RPE is required for growth of the eye, it controls proper lamination of the retina, and it regulates differentiation of the photoreceptors (Bharti et al., 2006; Martinez-Morales et al., 2004; Strauss, 2005). Genetic ablation of the RPE or disruption of RPE specification genes result in microphthalmia, RPE-to-retina transdifferentiation and coloboma during murine eye development (Bumsted and Barnstable, 2000; Martinez-Morales et al., 2001; Nguyen and Arnheiter, 2000; Raymond and Jackson, 1995; Scholtz and Chan, 1987).
The RPE is specified at the early optic vesicle stage, long before pigmentation becomes obvious (Fig.3.2A). Two key players in RPE specification are the transcription factors Mitf and orthodenticle homeobox 2 (Otx2). Mitf is the first gene that is specifically expressed in the presumptive RPE in the optic vesicle (for reviews, see Bharti et al., 2006; Martinez-Morales et al., 2004)). Mitf is a key regulator of pigment cell development in the RPE and neural crest; it transactivates crucial genes for terminal pigment differentiation (e.g. Dct, Tyrp1 and tyrosinase). Otx2 is expressed in the eye field and expression appears to persist until the late optic vesicle stage when it is downregulated in the presumptive retina, similar to Mitf. Otx2 is required for Mitf expression and transactivates expression of pigment genes in cooperation with Mitf (Martinez-Morales et al., 2003; Martinez-Morales et al., 2004; Martinez-Morales et al., 2001). Recent studies demonstrate that RPE specification requires interaction with extraocular tissues, however, the exact mechanism is not resolved (Buse and de Groot, 1991; Fuhrmann et al., 2000; Lopashov, 1963; Muller et al., 2007; Stroeva, 1960).
In chick, some progress has been made in clarifying the role of extraocular tissues and signaling pathways regulating RPE development, however, some of the results are controversial. Robust Mitf expression in chick is detectable at the optic vesicle stage, however, in contrast to mouse, expression is restricted to the presumptive RPE domain (Fuhrmann et al., 2000; Ishii et al., 2009; Mochii et al., 1998; Muller et al., 2007). Previous studies, including our own, indicate that the adjacent extraocular mesenchyme is required for expression of RPE-specific genes such as Mitf, the Mitf target melanosomal matrix protein MMP115 and Wnt13, in explant cultures of chick optic vesicles (Fuhrmann et al., 2000; Kagiyama et al., 2005). In the absence of extraocular mesenchyme, the TGFβ family member activin can restore RPE marker expression (Fuhrmann et al., 2000). Since the explants were prepared before Mitf is robustly expressed in the presumptive RPE domain, we conclude that the extraocular mesenchyme is essential for induction of Mitf expression/RPE fate in the chick optic vesicle (Fuhrmann et al., 2000; Mochii et al., 1998).
Some of our findings are at odds with other published works. Previously, it was shown that Mitf can be expressed earlier at low levels in the entire evaginating chick optic vesicle when the neuroepthelium is in close contact with the overlying ectoderm and mesenchyme is still absent (Muller et al., 2007). The surface ectoderm expresses BMPs and BMP-coated beads can induce ectopic Mitf expression, when implanted adjacent to the optic vesicle (Hyer et al., 2003; Muller et al., 2007). Therefore, it was proposed that BMP secreted by the surface ectoderm acts an inducer of RPE fate in the optic vesicle (Muller et al., 2007). However, optic cup morphogenesis is disturbed in BMP-treated eyes and BMP can induce apoptosis in the optic vesicle, which may confound these results (Hyer et al., 2003; Muller et al., 2007; Trousse et al., 2001). In addition, loss of function-studies show that BMP signaling is required for RPE differentiation in the ventral optic cup but these experiments do not address a role during earlier stages, when the RPE is specified (Adler and Belecky-Adams, 2002b; Muller et al., 2007). Thus, further experiments in chick are required to elucidate the true nature of interacting tissue and which actual RPE-inducing signal mediates this interaction.
In contrast to chick, the mouse extraocular mesenchyme surrounds the budding optic vesicle at very early stages when Mitf starts to become expressed (Bassett et al., 2010; our own unpublished observations). Consistent with our previous work suggesting a role for extraocular mesenchyme in RPE specification in chick, we observed in mouse optic vesicle explant cultures that removal of extraocular mesenchyme interferes with Mitf expression (unpublished observations). Further, other studies demonstrate that mutations in genes critical for extraocular mesenchyme development are accompanied by ocular malformations and abnormal development of the mouse RPE (Bassett et al., 2007; Evans and Gage, 2005; Gage et al., 1999; Grondona et al., 1996; Kastner et al., 1994; Kitamura et al., 1999; Kume et al., 1998; Matt et al., 2008; Mori et al., 2004; Moser et al., 1997; West-Mays et al., 1999). However, in mouse, a direct role for extraocular mesenchyme in induction of the RPE has not been shown so far. Interestingly, interference with BMP signaling does not interfere with RPE development in mouse suggesting that, while the utilization of a common signaling pathway (TGFβ) may be conserved, the actual signal can be different depending on the species (Furuta and Hogan, 1998; Morcillo et al., 2006; Wawersik et al., 1999).
In addition to extracellular signaling, a few intrinsic determinants are known to regulate early aspects of RPE development. In zebrafish, certain Rx3 alleles can interfere specifically with RPE development and it was proposed that Rx confers competence on the presumptive RPE to respond to inducing signals from the mesenchyme (Rojas-Munoz et al., 2005). In Pax2/Pax6 compound mutant mice, Mitf is not expressed in the optic vesicle, the RPE transdifferentiates into retina and the optic vesicle does not invaginate to form an optic cup (Baumer et al., 2003). Interestingly, Otx2 expression persists but is not sufficient to promote RPE formation. Furthermore, both Pax2 and Pax6 bind to and activate the Mitf-A enhancer that controls expression of Mitf in the RPE (Baumer et al., 2003). These results suggest that Pax2 and Pax6 are redundantly required for RPE specification in mouse, however, it is not clear whether they act upstream or downstream of a potential signal from the mesenchyme. In summary, these studies indicate that several upstream regulators, intrinsic and extrinsic, ensure that Mitf is sufficiently expressed to promote RPE cell fate in the optic vesicle (Fig.3.2A).
The MAP kinase FGF signaling pathway is important for different steps of neural retina development (Fig.3.2B). First, it is essential for patterning of the retina in the distal optic vesicle, and, second, for initiation of retinal neurogenesis. FGF ligands and receptors are abundantly expressed in ocular and extraocular tissues, and specifically, FGF1 and FGF2 show strong expression in the lens ectoderm (de Longh and McAvoy, 1993; Pittack et al., 1997; Vogel-Hopker et al., 2000). Removal of the surface ectoderm in chick embryos interferes with neuronal marker expression in the distal domain (Hyer et al., 1998; Pittack et al., 1997). In addition, eyes develop microphthalmic as pigmented vesicles with a few neuronal cells intermingled suggesting that proximodistal patterning of the optic vesicle into retina and RPE is disturbed. However, retinal differentiation capacity appears to be preserved since expression of neuronal markers is rescued when the pre-placodal lens ectoderm is replaced by a source of FGF (Hyer et al., 1998).
This finding was interpreted to mean that FGF derived from the lens ectoderm is necessary to maintain the retina domain in the distal optic vesicle (Hyer et al., 1998). However, it is also possible that an unknown signal from the lens ectoderm activates FGF signaling in the presumptive retina. Moreover, studies in zebrafish and chick suggest that the ocular neuroepithelium itself may be an FGF signaling center that controls the onset and progression of retinal neurogenesis in the optic cup (Martinez-Morales et al., 2005; McCabe et al., 1999; Picker and Brand, 2005; Vinothkumar et al., 2008).
In contrast to chick, FGF/tyrosine receptor activation is not required for the progression of neurogenesis in the mouse central retina (Cai et al., 2010). Thus, the signals driving neurogenesis may be distinct in different species.
MAPK FGF signaling also regulates retina specification, a function conserved in all vertebrates (Fig.3.2B). Previously, it was shown that in FGF9 mouse mutants, the RPE expands into the presumptive retina suggesting that FGF9 helps to define the boundary between retina and RPE (Zhao et al., 2001). While this result is informative, it is possible that redundant FGFs may compensate for the loss of FGF9. Furthermore, the precise ligands and receptors that regulate these processes are not known. To address this concern, Cai et al. analyzed the role of general FGF pathway activation during optic vesicle and optic cup morphogenesis by manipulating tyrosine phosphatase Src homology 2 (Shp2), which associates with FGF receptor tyrosine kinases and is required for complete activation of FGF signaling (Cai et al., 2010). Conditional inactivation of Shp2 at the early optic vesicle stage in mouse (before retinal specification) results in loss of Vsx2 expression in the distal portion. Instead, Mitf expression persists, and the affected part of the optic cup acquires RPE-like morphology and becomes pigmented (Cai et al., 2010). Further analysis shows that Shp2 acts downstream of FGF; ectopic Ras activation can genetically rescue retinal development in Shp2 mutant eyes. This is the first evidence for a direct requirement of FGF signaling in retina specification in the optic vesicle.
This role for FGF was also shown using a different approach. Strikingly, gain of functions studies in frog, chick and rodents demonstrate that activation of the MAPK FGF pathway can cause the presumptive RPE to transdifferentiate into retina with fully differentiated cell types, such as ganglion cells and photoreceptors (Galy et al., 2002; Guillemot and Cepko, 1992; Hyer et al., 1998; Mochii et al., 1998; Nguyen and Arnheiter, 2000; Park and Hollenberg, 1989; Pittack et al., 1997; Pittack et al., 1991; Reh et al., 1987; Sakaguchi et al., 1997; Spence et al., 2007; Vergara and Del Rio-Tsonis, 2009; Vogel-Hopker et al., 2000; Yoshii et al., 2007; Zhao et al., 2001; Zhao et al., 1995). The finding that FGF cannot induce transdifferentiation of RPE into retina in optic vesicle cultures of Vsx2 null mutant mice shows that Vsx2 is required to mediate the effect of FGF either directly or indirectly (Horsford et al., 2005). Consistent with this, loss of Vsx2 gene activity mimics FGF loss of function, resulting in a retina-to-RPE transdifferentiation (Horsford et al., 2005; Rowan et al., 2004). Vsx2 may act by directly suppressing transactivation of the Mitf gene in the distal optic vesicle (Bharti et al., 2008). This would put FGF in a single pathway upstream of Vsx2, which then acts to repress Mitf expression, allowing the distal optic vesicle to develop into retina (Fig.3.2B).
BMP signaling may also participate in early steps of retina development. This hypothesis is supported by the finding that BMP7 null mice display varying incidents of microphthalmia or anophthalmia, depending on the genetic background. In anophthalmic BMP null mice, the expression of retina-specific genes are downregulated in the optic cup, with concomitant ectopic expression of RPE genes such as Mitf (Morcillo et al., 2006). While this phenotype could be due to a failure of lens induction (see below), the possibility that BMP signaling cell autonomously induces or maintains expression of retina genes in the distal optic vesicle cannot be excluded. Bolstering this idea, Murali et al showed that in compound mutant mice with homozygous inactivation of BMP receptor types Ia and Ib, Vsx2 expression is downregulated in the optic cup and retinal neurogenesis fails to initiate (Murali et al., 2005). While Vsx2 expression was not analyzed at earlier stages, retinal specification may be already disturbed in the distal optic vesicle since, for example, the retina-specific marker FGF15 is not expressed in the mutant eye. Therefore, BMP signaling may be involved in retina specification or maintenance of the retina domain in the mouse optic vesicle.
The distal portion of the optic vesicle makes contact with the overlying surface ectoderm, resulting in the specification of the lens ectoderm (pre-placodal stage). This interaction leads to invagination of the lens placode and distal optic vesicle resulting in formation of a bilayered optic cup (Fig.3.1B, 3.3). The neural retina and RPE develop from the inner and outer layer of the optic cup, respectively. The lens vesicle eventually separates from the surface ectoderm and differentiates into the mature lens. In this section, we will discuss recent progress that has been made with respect to the process of optic vesicle invagination and the role of tissue-tissue interactions mediated by extracellular factors.
Following optic vesicle evagination, the distal optic vesicle and lens ectoderm (lens pit) invaginate, forming the optic cup. The mechanics behind the morphogenesis of invagination are just beginning to be understood. Work in different organisms suggests that there are various ways to generate force and tension that are integral to invagination. A study conducted in chick indicates that invagination is a Ca2+-dependent process. The authors of this work suggest that “apical bands of microfilament” exist in the retinal cells that contract, thus generating the force to enable optic cup formation (Brady and Hilfer, 1982). In Medaka fish, a mutant, ojoplano, was recently identified that exhibits several morphogenesis defects, including improper invagination or folding of the optic cup (Martinez-Morales et al., 2009). Ojoplano encodes a novel transmembrane protein with partial homology to a candidate gene for orofacial clefting syndrome, which is also associated with some eye abnormalities (Mertes et al., 2009). In ojoplano mutants, the expression of focal adhesion proteins such as the integrin beta1 receptor in the basal surface of the retina appeared reduced, which may cause reduced tension and a change in cellular shape in retinal progenitors (Martinez-Morales et al., 2009). Since a partial optic cup is formed in these mutants, additional mechanisms may exist that regulate the process of invagination. Future studies will need to address what kind of signal initiates invagination and how conserved the particular mechanism directed by ojoplano protein is across vertebrates.
Filopodia also provide mechanical force necessary for invagination. During the coordinated invagination of the optic vesicle and lens pit, both tissues are in tight apposition. Recent studies revealed that basal filopodia that mostly originate from the lens ectoderm transiently tether the presumptive lens and retina to coordinate invagination of the lens pit (Chauhan et al., 2009). Production of these filopodia is dependent on the Rho family member GTPase cdc42 and the cdc42 effector IRSp53, and a failure of filopdia formation leads to defects in lens pit invagination (Chauhan et al., 2009). Further studies will likely discover additional mechanisms that aid in this dramatic morphogenetic process.
The process of invagination is dependent upon tissue-tissue interactions. When pre-placodal ectoderm is ablated from the optic vesicle, distal optic vesicle invagination is perturbed, but the retina is specified (Hyer et al., 2003). What’s more, Vsx2 expression remains normal, indicating that patterning of the optic vesicle and invagination are independent processes (Hyer et al., 2003). Conversely, if surface ectoderm ablation occurs later, at the lens placode stage, an optic cup forms without a properly formed lens (Hyer et al., 2003; Smith et al., 2009). These experiments show that invagination requires specific, precisely timed interactions between retina and surface/lens ectoderm.
The discovery that pre-placodal lens specification is critical for invagination of the distal optic vesicle is confirmed by genetic studies in mouse. Similar to the pre-placodal lens ablation experiments, disruption of expression of the homeobox transcription factors Six3 or Pax6 or the HMG trancription factor Sox2 during the pre-placodal stage, results in failure of thickening of the lens placode and lens formation, and disrupted invagination that leads to arrest at the optic vesicle stage (Ashery-Padan et al., 2000; Grindley et al., 1995; Kamachi et al., 1998; Smith et al., 2009). Recent studies revealed that Six3 is expressed in the surface ectoderm before Pax6, and without Six3, Pax6 is downregulated and Sox2 never expressed (Liu et al., 2006). Furthermore, using Chromatin immunoprecipitation, EMSA and luceriferase reporter assays, it was demonstrated that Six3 directly activates Pax6 and Sox2 expression (Liu et al., 2006). These and other findings indicate that Six3, Pax6 and Sox2 act in a complex regulatory network to regulate each other during lens induction and specification (Fig.3.3; for review, see Donner et al., 2006; Lang, 2004; Liu et al., 2006).
Induction of Sox2 in lens ectoderm is indirectly controlled by the TGFβ family member BMP7, which signals either upstream or downstream of Pax6 (Fig.3.3; Faber et al., 2001; Furuta and Hogan, 1998; Gotoh et al., 2004; Wawersik et al., 1999). Different lines of mice with a null mutation in the BMP7 gene can exhibit severe eye defects such as arrest at the optic vesicle stage and failure of lens formation (Dudley and Robertson, 1997; Morcillo et al., 2006; Wawersik et al., 1999). While some regional patterning of the optic vesicle appears normal (e.g. expression of Lhx2 and the RPE marker Dct), expression of Sox2 and Pax6 are not maintained in the pre-placodal lens ectoderm (Morcillo et al., 2006; Wawersik et al., 1999).
A role for BMPs in lens induction is also supported by the phenotype of Lhx2 mouse mutants. In these animals, eye development arrests at the optic vesicle stage and the lens never forms (Porter et al., 1997; Yun et al., 2009). However, expression of Lhx2 in the surface ectoderm is not required, since lens-specific disruption of Lhx2 has no obvious effect (Yun et al., 2009). What, then, is the source of the defect? Analysis of several pathways implicated in lens induction revealed that signaling downstream of BMP is specifically disrupted in the optic vesicle and in lens ectoderm. Further, in Lhx2 mutants, some but not all aspects of the eye phenotype can be rescued by treatment of Lhx2 mutant explants with exogenously added BMPs (Yun et al., 2009). These results imply not only that BMP signaling regulates lens induction, but also indicates that other, unknown factors are involved.
A potential downstream target of BMPs during lens induction is MAB21l2. MAB21l2 is similar to the C. elegans MAB-21 cell fate-determining gene, a downstream target of TGFβ signaling. In the developing eye, it is first expressed in the dorsal optic vesicle and expression extends later, in the optic cup, to the RPE, retina and optic stalk (Yamada et al., 2004). In Mab21l2 null mice, the area of contact between the surface ectoderm and optic vesicle is reduced, and an optic cup and lens never forms, resulting in an eye rudiment. This phenotype is due in part to the fact that the optic cup displays no Vsx2 expression resulting in a severe proliferation defect (Yamada et al., 2004). Other direct or indirect downstream targets of BMP signaling are not known and have yet to be determined.
FGF signaling is also required for Pax6 expression in the lens placode (Faber et al., 2001; Gotoh et al., 2004). This is demonstrated by the finding that disruption of the major docking protein FGF receptor substrate 2 alpha (FRS2alpha), that links FGFR2 with several downstream targets, can result in defective invagination of the optic vesicle. The concomitant loss of BMP expression in the optic vesicle may enhance the eye phenotype (Gotoh et al., 2004). Defective invagination may be due to reduced FGF signaling in the lens ectoderm, since a significant reduction of Shp2 activation in the distal optic vesicle does not interfere with lens induction (Cai et al., 2010). Interestingly, it has been suggested that retina-derived N-cadherin could act as an alternative ligand for FGF receptor signaling in the lens (Smith et al., 2010). However, the source of the actual FGF receptor ligand(s) is not clear and it is possible that a high degree of redundancy exists.
In addition, Retinoic acid signaling is required for optic cup morphogenesis (for review, see Duester, 2009). Retinaldehyde dehydrogenases (Raldh1, 2, 3) mediate the final step of retinoic acid synthesis. Raldh2 is present in the mesenchyme and Raldh3 is expressed in the RPE (Molotkov et al., 2006). Both enzymes synthesize retinoic acid, providing an essential signal to the neural retina required for morphogenetic movements that lead to ventral invagination of the optic cup (Molotkov et al., 2006). Retinoic acid is further required to induce apoptosis in the extraocular mesenchyme, and one target gene in the extraocular mesenchyme is the transcription factor Pitx2, which is also required for RPE differentiation (Gage et al., 1999; Gage and Zacharias, 2009; Matt et al., 2005). These and other findings indicate that retinoic acid synthesis and signaling is complex during development of the eye (Duester, 2009).
Subsequent to initial establishment of the RPE in the optic vesicle, proliferation in the presumptive RPE ceases, leading to the formation of a single layer of cuboidal cells that become pigmented. As development proceeds, a period of differentiation and further maturation follows that results in dramatic morphological, structural and functional changes of the RPE tissue such as formation of tight junctions, expansion of apical microvilli and invagination of the basal membrane, establishment of polarity and retinoid recycling machinery (Burke and Hjelmeland, 2005; Finnemann, 2003; Marmorstein, 2001; Marmorstein et al., 1998; Rizzolo and Kwang, 2007; Strauss, 2005). The RPE fate is reversible for several days following the initial activation of differentiation as evidenced by a propensity to hyperproliferate and to differentiate into retina, for example by treatment with FGF (Stroeva, 1960; Zhao et al., 1997). These results suggest that maintenance of the RPE fate is controlled by the concerted effort of multiple factors during this prolonged period (Fig.3.4).
The sonic hedgehog signaling (shh) pathway is required for maintenance of RPE fate in the ventral optic cup. In chick and mouse, reduced shh signaling may not affect RPE specification but results subsequently in loss of RPE marker expression, increased proliferation of the RPE and transdifferentiation into retina (Huh et al., 1999; Zhang and Yang, 2001). RPE differentiation defects are also observed in frog, when shh signaling is downregulated (Perron et al., 2003). Growth arrest specific 1 (Gas1), a GPI-anchored cell surface protein that binds shh, may be a positive co-regulator of shh signaling (Allen et al., 2007; Lee et al., 2001b; Martinelli and Fan, 2007). Disruption of Gas1 results in RPE defects that are very similar to effects caused by defective shh signaling as described above, e.g. ectopic proliferation and transdifferention into retina (Lee et al., 2001a). In Gas1 mutants, RPE specification occurs normally in the optic vesicle, but subsequently in the optic cup, the ventral RPE fails to slow down proliferation. This proliferation defect precedes transdifferentiation into retina suggesting that Gas1 is required for downregulation of proliferation in the ventral RPE (Lee et al., 2001a). The dorsal RPE in Gas1 mutants is not affected; therefore, differentiation in the dorsal and ventral RPE may be controlled by distinct mechanisms.
Other pathways involved in maintenance of the RPE in the ventral optic cup are retinoic acid and BMP signaling (Fig.3.4). Retinoic acid signal transduction occurs via the retinoic acid receptors (RARα,β,γ) that bind to RXR receptors and form heterodimers when bound to the RA response element in target genes. RAR mutants display a range of severe eye defects, including microphthalmia and transdifferentiation of RPE into retina (Lohnes et al., 1994; Matt et al., 2008). Furthermore, disruption of BMP signaling by overexpression of the BMP antagonist noggin caused transdifferentiation of the ventral RPE (Adler and Belecky-Adams, 2002a). These findings further support the notion that dorsal and ventral RPE development is regulated by distinct mechanism.
Recent work demonstrates that the Wnt/β-catenin pathway also controls differentiation of the RPE in the optic cup. In brief, activation of the Wnt/β-catenin pathway results in cytoplasmic stabilization of β-catenin, ultimately converting TCF/LEF transcription factors from repressors into activators (Nusse, 2009). In the differentiating zebrafish, chick, and mouse RPE, TCF/LEF-responsive reporters are activated, and several pathway components are expressed during RPE development (Chang et al., 1999; Cho and Cepko, 2006; Dorsky et al., 2002; Lee et al., 2006; Liu et al., 2003; Wang et al., 2001; Yasumoto et al., 2002). We and others observed that interference with Wnt/β-catenin signaling in chick and mouse RPE of the optic cup causes loss of TCF/LEF reporter activation and severe eye defects such as microphthalmia and transdifferentiation of the RPE into retina (Fujimura et al., 2009; Westenskow et al., 2009; Westenskow et al., 2010).
Further analysis using chromatin immunoprecipitation and transactivation assays revealed that a complex of β-catenin/TCF/LEF binds to and activates those Mitf and Otx2 enhancers that regulate expression in the RPE (Fujimura et al., 2009; Westenskow et al., 2009; Westenskow et al., 2010). It was in this context that it was recently shown that the orphan receptors COUPTFI (Nr2f1) and perhaps COUPTFII (Nr2f2) of the steroid/thyroid hormone receptor superfamily are also directly regulated by β-catenin/TCF/LEF (Fujimura et al., 2009). In agreement, double mutant COUPTFI/II receptor mice show severe developmental abnormalities at the optic cup stage, including transdifferentiation of the RPE into retina (Tang et al., 2010). COUPTFs can directly regulate expression of Otx2. Thus, expression of the RPE key determinants Otx2 and Mitf is regulated by several mechanisms that may act in parallel and may reinforce each other (Fig.3.4).
Perhaps surprisingly, ectopic activation of the Wnt/β-catenin pathway in the presumptive retina is not sufficient to induce RPE cell fate. Instead it acts in the peripheral eye to maintain undifferentiated progenitor cells and to promote differentiation into ciliary body and iris epithelium (Cho and Cepko, 2006; Kubo et al., 2003; Kubo et al., 2005; Liu et al., 2007). The reason behind the insufficiency of Wnt/β-catenin to induce RPE may lie in our observation that overexpression of Otx2 and Wnt/β-catenin signaling can promote ectopic Mitf expression in the developing chick retina (Westenskow et al., 2010). Thus, Otx2 may be a required co-factor that confers competence to respond to RPE promoting signals such as Wnt/β-catenin signaling.
In summary, several intrinsic and extracellular factors control different aspect of eye organogenesis. Though a lot of progress has been made demonstrating how these signals are connected in a network, in many cases we do not know whether this regulation is direct, or how these interactions work on the molecular level. We also have still little insight into the actual output of these interactions. Finally, some processes during early eye development are still a mystery. For example, it is not clear what kind of mechanism ultimately drives invagination of the distal optic vesicle, nor what is the complete repertoire of factors that ensure specification of the retina in the optic vesicle. Thus, future work will need to address these and other questions.
Special thanks to Julie Kiefer for critical reading and helpful suggestions on the manuscript. My apologies to those authors whose work is not cited due to space limitations. Supported by NIH/NEI (EY014954 to S.F., Core Grant EY014800) and by an unrestricted grant from Research to Prevent Blindness, Inc., to the Department of Ophthalmology, University of Utah.