Fig. 1.2  When Hans Spemann performed his elegant optic vesicle ablation studies over a century ago, it was known that the lens derives from the ectoderm (pink), while the optic vesicle gives rise to the retina. However, what triggers the lens to form at its correct location over the retina? (a) The micro-dissection studies were technically demanding especially given the small size of the amphibian embryos. In this photograph, the optic vesicle can be appreciated budging under the surface ectoderm (arrows). The broad area of head ectoderm that is competent to form a lens shaded pink. (b) Cross-sectional diagram in the plane represented

by the white frame in panel A. After pealing back a flap of ectoderm, Hans Spemann used a hot needle to selectively destroy the optic vesicle on one side while preserving its overlying ectoderm. (c) The remarkable finding was that ablating the optic vesicle completely prevented formation of the lens (green) during subsequent stages of development. These observations, which indicate that signals from the optic vesicle are able to trigger lens differentiation in the adjacent competent ectoderm, provided the first experimental evidence of tissue induction [62, 328]). This figure is an unpublished diagram from the author’s laboratory

are thought to trigger the expression of the transcription factor otx2 in the late gastrula ectoderm. Inducers from the neural folds from the anterior neural plate then induce pax6 expression in the anterior ectoderm. It is thought that the expression of these transcription factors is important to making the surface ectoderm competent to respond to the optic vesicle during the late neural stage. In turn, the optic vesicle secretes factors that induce the synthesis of sox transcription factors. These factors initiate the production of the lens. The inner layer of the optic cup becomes the neural retina and the outer layer the retinal pigmented epithelium (RPE). In Hans Spemann’s own words, “Though the results in detail may need to be retested and supplemented in many cases, this seems to be certain: that in several if not all species of the Amphibia in the neural stage or shortly after the closing of the medullary fields, the rudiment of the lens is more or less firmly determined; that the epidermis possesses the potency for lens formation in different degrees; and, finally, that the optic cup possesses the ability to activate this potency for lens formation” [62].

1.4  Men are born with two eyes, but only one tongue, in order that they should see twice as much as they say4

The formation of two eyes has provided vertebrates with stereopsis as well as a broader view of the world. How does the developing embryo form two separate and identical eyes? The solution to this mystery is integral to the larger question of how the central nervous system accomplishes bilateral development of most of its structures [63]. Our current understanding, which is emerging through the combination of clinical observations, classical embryology, and molecular biology, is that a single eye field is bisected by signals from the underlying prechordal mesoderm. The early eye field therefore represents a larger area of neural ectoderm that is competent to respond to the inducer signal to form an eye. If the competence of the central eye field is not inhibited

4Charles Caleb Colton (1780–1832)

when induction begins, the entire field will respond with the formation of a single eye. Typically, the inhibitory signals to the central eye field are not entirely absent, and fused globes result. In fact, true cyclopia, i.e., a single eye, is vanishingly rare, with most cases showing two fused eyes (synophthalmia). A case of synophthalmia in Patau’s syndrome is shown in Fig. 1.3 [64]. A histological study of the eyes of this case is shown in Fig. 1.4 including immunohistochemical characterization of the dysplastic rosettes of the retina [64].

The problem of cyclopia should be considered in the context of the larger problem of creating bilaterality in

the CNS and dorsal-ventral patterning [65, 66]. Indeed, cyclopia is part of the spectrum of the devel­opmental abnormalities seen in holoprosencephaly (a signal cerebral hemisphere). The phenotype of holoprosencephaly is quite variable consisting of a spectrum from severe manifestations with major brain and face anomalies to clinically normal individuals with only a single fused central incisor to clue in the observer (reviewed in [67–71]). Holoprosencephaly is the most common developmental defect of the forebrain in humans with an incidence as high as 1:250 during embryogenesis. However, due to intrauterine lethality, the liveborn

Fig. 1.3  Synophthalmia in Patau’s syndrome. (a) Thirty-­ two-week fetus showing partial midline fusion with single proboscis. Autopsy revealed holoproencephaly and multiple developmental abnormalities including transposition of the great vessels with pulmonary artery hypoplasia; interventricular and interatrial septal defects; fused cerebral hemispheres with a common ventricle; absence of the olfactory bulbs and tracts, and

basal ganglia. One thalamus was present with partial medullar and cerebellar fusion. (b) Closer view of the partially fused eye lids inferior to a single centrally placed proboscis. (c) Partial karyotype illustrating triploidy of chromosome 13. Compare with Fig. 1.4. Adapted from Chan et al. [64] with permission of BioMed Central

Fig. 1.4  Histological analysis of synophthalmic eyes from case presented in Fig. 1.3. (a) Horizontal section showing that the eyes consist of two partially fused globes. Although the lens is separate, the posterior chamber is common, and there is a single optic nerve. (b) The neural retina in many areas does not show the usual laminar histology but is composed of collections of neurons arranged in apparent cylinders often with a central lumen. The lumen of the cylinders is usually rimmed by a definite line reminiscent of the external limiting membrane (ELM) of the normal retina. (c) Immunohistochemical localization of rod

opsin. (d) Immunohistochemical localization of interphotoreceptor retinoid-binding protein (IRBP). (e) Immunohistochemical localization of cellular retinaldehyde binding protein (CRALBP) demonstrates Müller cell differentiation consisting of radial process extending between the rosette neuronal cells. The processes abruptly end at the ELM-like structure. Taken together, these findings indicate that the rosette structures seen in this case of trisomy 13 represent a dysplastic process rather than a differentiation of neoplastic cells as in retinoblastoma. Modified from Chan et al. [64]

prevalence is 1:16:000. Holoprosencephaly is a malformation sequence in which impaired midline cleavage of the embryonic forebrain is a fundamental feature [66, 72–76]. The prosencephalon fails to cleave sagitally into cerebral hemispheres, transversely into telencephalon and diencephalon, and horizontally

into olfactory and optic bulbs. Given the number and complexity of the cellular interactions that must occur in the developing forebrain, it is not surprising that a variety of genes (at least 12 different loci) and a variety of teratogens have been implicated in the pathogenesis of holoprosencephaly [66, 68, 69, 72, 76–87].

1.5  Sonic Hedgehog and Revisiting

Homer’s Odyssey

Significant advances have been made in recent years toward understanding the pathogenesis of cyclopia and holoprosencephaly. In particular, the “Sonic Hedgehog pathway” has emerged as having a central role in both of these pathological states [88–93]. “Sonic the Hedgehog” was a popular Sega Genesis video game character. The Sonic Hedgehog gene (Shh) was named after this hero when it was noted that mutations of this gene cause spiny backs in fruit flies (Mr. Sonic Hedgehog has blue spines down his back!). Sonic Hedgehog is a secreted protein that acts as a “cell fate switch.” Shh is the most extensively characterized vertebrate homolog and is involved in a wide variety of embryonic events. It can act as both a short-range, contact-dependent factor and a long-range, diffusible morphogen. Shh genes are highly conserved. In the human embryo, shh is expressed in the notochord, the floorplate of the neural tube, the gut, and the developing limbs. Interestingly, Hedgehog proteins undergo autocatalytic processing and modification that are critical for signaling activity [89, 94–97]. Autoprocessing of Hedgehog includes covalent attachment of cholesterol onto the carboxy terminus of its N-terminal domain. The N-terminal domain contains all known signaling capabilities, while the C-terminal domain is responsible for the intramolecular precursor processing, acting as a cholesterol transferase [98]. The cholesterol moiety is thought to direct Hedgehog protein traffic in the secretory cell [99]. Furthermore, binding of cholesterol to Sonic Hedgehog enhances its solubility, allowing it to diffuse as a paracrine factor [100– 105]. This ability to act at distance is critical to its function as a morphogen signal during embryogenesis (reviewed in [106]).

The importance of cholesterol to the Hedgehog signal transduction pathway appears to explain the teratogenic effects of the steroidal-alkaloid compounds, jervine and cyclopamine, the later deriving its name from its tendency to induce cyclopia. These compounds are now known to cause cyclopia by inhibiting Hedgehog signaling [99, 107, 108]. The specific mechanism of cyclopamine’s action is through binding the product of the Smoothen gene (see below) [99, 107, 108, 137]. Interestingly, both jervine and cyclopamine are found in the Veratum plant family. Pregnant cattle, goats, or sheep that graze on the corn lily plant Veratrum californicum early during pregnancy can

give birth to deformed offspring with cyclopia. Since Veratrum plants are found in the Mediterranean regions, it is plausible that the legendary Cyclops of Odysseus was not completely an invention of Homer’s imagination­ but may have been based on the occasional observation of cyclopic ewes in ancient times.

Sonic Hedgehog exerts its morphogenic effects by diffusing to cells that express the cell surface receptor Ptch, which is homologous to the Drosophila segment polarity gene Patched-1 (reviewed in [109]). Ptch controls Hedgehog-responsive genes through the transcription regulatory molecule Cubitus Interruptus (Ci) (reviewed in [110]). Key to this pathway is Smoothen, a membrane protein that binds to Ptch. In the absence of Hedgehog binding to Ptch, Smoothened is inactive, and Ci is tethered to cytoplasmic microtubules. While parked on the microtubles, Ci is cleaved, and a portion of Ci diffuses into the nucleus where it represses transcription. However, Hedgehog binding to ptch blocks the ability of Ptch to inhibit Smoothened. As a result, intact Ci can enter the nucleus to activate Hedgehog response genes.

In view of the complexity of the Hedgehog pathway, it is not surprising that blocking cholesterol modification of shh is not the only road to holoproscencephaly. In humans, mutational activation of Ptch and Shh can also result in holoproscencephaly with cyclopia [72, 74, 77, 78, 80, 83, 87, 91, 111–113]. How does disruption of the Hedgehog pathway lead to cyclopia? The answer to this question comes through a series of elegant classical embryological and modern molecular studies.

The debate over how the cyclopic eye forms has been going on for more than a century. The arguments are closely connected with the mechanisms normally leading to the formation of the two separate retinal primordia. The various models for formation of the retinal primordia are reviewed in [114]. In theory, cyclopia could result from the fusion of two originally separate eyes, or from the failure of a single primordium to separate during development. Recent work has established­

that the latter possibility is correct. Microdissection studies have shown that removal of the prechordal mesoderm leads to the formation of a single retina in chick embryos and Xenopus explants [49, 114]. For example, Li et al. [114] noted that removal of the prechordal plate resulted in fusion of the forebrain as well as the retina. This result is illustrated in Fig. 1.5. The future retinas were identified by in situ hybridization with a chicken Pax-6 probe. In ~27% of embryos without the prechordal plate, a single retina, continuous from one side of the embryo to the other, was

Fig. 1.5  Elegant microdissection studies such as this one performed by Li et al. [114] in the chick embryo have shown that removal of the prechordal mesoderm leads to the formation of a single retina. The major conclusion from these types of studies is that there is a single retina morphogenetic field that resolves into two retina primordial. This is accomplished by suppression of retina formation in the median region of the field. The signal for this repression, which presumably is Sonic Hedgehog, comes from the prechordal plate. The data depicted here show the effects of prechordal plate removal on retina formation in chick embryos. (a–d) are ventral views. (a) A diagram of a stage 5 chick embryo showing the location of presumptive retina primordial (indicated by two gray circles) relative to the prechordal plate (marked as a blue line). The red dot symbolizes Hensen’s node. (b) A diagram of the region removed from the prechordal mesoderm (indicated by a box superimposed on the blue line). (c) A stage 13 chick embryo showing Pax-6 expression in the eyes. (d) Pax-6 expression in a stage 13 chick embryo from which the prechordal plate was removed at stage 5. (e) A transverse section of a stage 13 control chick embryo after in situ hybridization with the Pax-6 probe. Ventral is up. (f ) A transverse section of a stage 13 embryo which lacked the prechordal plate. The level of section is similar to that of the control embryo shown in (e). Ventral is up. Reproduced with permission from Development (Li et al, [114])

formed. That only a fraction of embryos had cyclopia was probably due to incomplete removal of the prechordal plate. The same investigators went on to show that the prechordal plate expresses shh and was able to rescue the cyclopic phenotype in transplantation

studies. Such studies provide strong support for a role of the prechordal plate in the formation of two retinas. Interestingly, transplantation of the prechordal plate to the vicinity of the optic cup was able to suppress the expression of pax6 in the retina. The key role of pax in

development is discussed further below. The important conclusion from studies such as these is that there is a single retinal morphogenetic field that resolves into two retina primordia. “This is accomplished by suppression of retina formation in the median region of the field. The signal for this repression comes from the prechordal plate.” As pointed out by Li et al. [114], this conclusion is consistent with the model suggested more than 75 years ago by Adelmann [115].

New concepts in development often go hand-in- hand with new insights into the mechanisms of evolution. This is exemplified by intriguing studies of the Hedgehog pathway in blind cave fish by Yamamato et al. [116]. These investigators found that the embryonic midline controls eye degeneration in blind cavefish by overactivation of the Hedgehog pathway. Key to their experiments was comparing ocular development in a species of the teleost Astyanax mexicanus that has normal vision and lives at the water surface with that of the blind species that lives in caves. Eye primordia are formed during cavefish embryogenesis. However, these primordia arrest in development, degenerate, and sink into the orbits. Remarkably, transplanting a surface fish embryonic lens into a cavefish optic cup can restore a complete eye. Compelling evidence is provided that in the cavefish, expression of shh and the related tiggy-winkle Hedgehog gene (twhh) is expressed in an expanded area along the anterior embryonic midline. This expanded Hedgehog signaling results in hyperactivation of downstream genes, lens apoptosis, and arrested eye growth and development. These features can be mimicked in surface fish by twhh and/or shh overexpression. The observations require that we modify our thinking regarding the evolution of eye regression. It had been generally assumed that the regression was caused by the accumulation of function mutations in eye genes, which accumulate without penalty due to conditions of relaxed selection for eyesight. The findings of Yamamato et al. [116] raise the alternative paradigm that control of eye regression is achieved by a gain of function in Hedgehog or related midline-signaling. Eye regression should therefore be viewed as being driven by natural selection for an adaptive trait [116]. The use of new experimental systems such as the zebrafish model promises to shed more light onto the mechanism of cyclopia [75, 76, 82, 83, 117–122].

The elegant complexity of the Hedgehog signal pathway reflects its central importance as a regulatory

system. This is underscored by the realization that mutations of ptch cause Gorlin syndrome or nevoid basal cell carcinoma syndrome [123–127]. This syndrome, which is inherited in an autosomal dominant manner, is characterized by dental, skeletal, and ­radiographic abnormalities including falx calcification, bifid/fused ribs and altered vertebral segmentation, and a predisposition to tumor development including early-onset basal cell carcinomas [128]. In fact, ptch mutations are common in sporadic basal carcinomas.

Ocular abnormalities are often present in Gorlin syndrome, including the first patient with this syndrome examined by Dr Gorlin [123]. However, of the various developmental abnormalities seen in the syndrome, the ocular findings, which are present in 15–25% of patients, are less well characterized [129, 130]. The ocular findings that may be present in Gorlin syndrome include defects of organogenesis (microphthalmia, coloboma, ocular hypoplasia), cataract, and posterior segment abnormalities (inappropriate retinal myelination, retinoschisis) [131–136]. It is therefore intriguing that the Hedgehog pathway appears to play an important role not only in establishing separate eyes, but also in the normal development the retinal structures themselves. Of particular importance is the observation that shh is expressed in retinal-ganglion cells. This could set the stage for an influence of the ganglion cells on normal organization of the remainder of the retina. Interestingly, ptch and gli are expressed in retinal neuroblasts, and astrocyte precursor cells in the optic nerve [137, 138]. These relationships are summarized diagrammatically in Fig. 1.6. In this model, retinal ganglion cell-derived shh expression is required for Hedgehog target gene induction in the retina and optic nerve. This induction plays a role in precursor cell proliferation, photoreceptor differentiation, and normal cellular organization [139–141]. To address the question of the role of the Hedgehog pathway and the ptch receptor in particular on the development of the mammalian retina, Black et al. [142] studied ocular development in mice heterozygotic for disruption of the ptch gene [143–146]. The retinas of PtchlacZ(+/-) mice exhibit abnormal cell cycle regulation, culminating in photoreceptor dysplasia and Müller cell gliosis. Interestingly, the PtchlacZ(+/-) mice also show vitreoretinal abnormalities resembling those found in patients with Gorlin syndrome. In these patients, an intraretinal glial response results in epiretinal membrane

Fig. 1.6  Clues to the histogenesis of retinal dysplasia are coming from studies of the effect of disruption of the Hedgehog pathway on the lamination of the neural retina. Of particular relevance are studies of retinal dysplasia in Gorlin syndrome. This syndrome results from mutations of the PTCH gene on chromosome 9q23.1, the human homologue of the Drosophila patched gene. PTCH is a transmembrane protein that functions as the receptor for members of the Hedgehog family of intercellular signaling molecules (see text). Black et al. propose that disruption of the ability of ptch receptor to respond to Sonic Hedgehog elaborated by the ganglion cells prevents the normal lamination of the retina with the formation of retinal rosettes [142]. Their model is shown here in a set of diagrams of the developing and adult mouse retina. At late stages of embryogenesis, the retina consists of two layers, the retinal ganglion cell (RGC) layer and the neuroblast (NB) layer, which contains proliferating precursor cells. RGC axons are located on the surface of the retina and exit the eye at the optic disc to form the optic nerve. The adult retina is organized into: RGC, inner nuclear layer (INL), and the rod and cone-containing outer nuclear layer (ONL). The nuclear layers are separated by the inner and outer plexiform layers (IPL, OPL), which contain neuronal processes. Müller cells span the width of the retina. In the embryonic and adult retina, Shh is expressed in RGCs, and Ptch is expressed neuorblastic layer (embryonic) and Müller cells (adult). The dysplastic retina (far right) shows rosette formation in the ONL, and an epiretinal membrane (ERM) at the vitreoretinal interface. Here, Muller cell processes have recruited contractile cells leading to retinal traction. Adapted from Black et al. [142] with permission of Oxford University Press

formation. These membranes, due to their proliferative and contractile nature, cause significant visual loss, especially in older patients with the syndrome. The investigators hypothesize that alteration of Müller cell/ Hedgehog signaling may play a role in the pathogenesis of the idiopathic epiretinal membranes and rosette formation in the retina of these patients. The role of the Hedgehog pathway in retina disease will be further uncovered by ongoing research in several laboratories aimed at clarifying its role in the normal patterning and differentiation of the retina [138–140, 147–152].

1.6  The Homeotic Genes:

Master and Commander5

Orchestrating the formation of a complex structure such as the eye requires turning on some, and suppressing other defined sets of genes at specific times during development. Gene regulatory proteins that do just that by recognizing short DNA segments were first appreciated in the 1950s. One of the first regulatory proteins to be recognized was the lambda repressor. Encoded by the bacterial virus, bacteriophage lambda, this repressor shuts off the viral genes that encode for viral coat particles. Turning off the synthesis of the coat proteins allows the virus to multiply silently within the cell.

Understanding how gene regulatory proteins work would have to wait for X-ray crystal structures of higher resolution than that which lead to the original discovery of the DNA double helix structure. Indeed, for 20 years after its discovery, the helix was thought to have a monotonous structure of ten nucleotides spaced exactly at 36° helical twists completing each spiral turn. However, better X-ray structures and solution NMR replaced this idealized structure with a nonuniform spiraling DNA helix consisting of “major” and “minor” groves. The groves can be appreciated in Figs. 1.7 and 1.8. The larger domain provided by the major grove allows a new set of proteins to interact with the DNA. These regulatory proteins were first identified in bacteria and were called helix-turn-helix DNA-binding proteins because they consist of two a helices bent at an angle. The C-terminal helix or recognition helix interacts with the DNA by fitting into the major groove. Remarkably, the recognition helix is able to specifically identify the different base pairs from their edges without the need to open the double helix! Variations in the amino acid residues that make up the recognition domain dictate the specificity of the homeodomain for the particular DNA sequence. Outside of the helix-turn- helix, the remainder of the transcription factor can vary greatly, allowing various helix-turn-helix motifs to present their DNA binding motifs in unique ways. Furthermore, the regions of the protein outside of the homeodomain may modify the DNA-homeodomain interaction by making their own contacts with the DNA, thus fine-tuning the interaction. Finally, it should be mentioned that each homeodomain transcription factor

5Patrick O’Brian (1969)

Fig. 1.7  The significance of the X-ray crystal structure of the human Pax6 paired domain-DNA complex is that it provides a general model for Pax protein-DNA binding. This image, based on the structure reported by Xu et al (1999, Genes & Development. 13:1263-75) [153], is from the RCSB Protein Data Bank (www.rcsb.org/pdb; NDB ID: PD0050).

binds as a dimer to the DNA recognition sequences [153]. These DNA sequences are therefore arranged as symmetrically half-sites. An important point is that this allows each protein monomer to make a nearly identical set of contacts, increasing binding affinity and allow­ ing cooperative interaction with other factors.

It was not long after these genes were recognized in bacteria that their counter parts in the fruit fly were found. This led to the discovery of a class of genes called homeotic selector genes. Homeotic genes were found to play an important role in orchestrating fly body-plan development. When the sequences of several homeotic genes were compared, a striking feature was noted: each contained a nearly identical stretch of 60 amino acid residues. The conserved region defines this class of proteins and is termed the homeodomain. Interestingly, when the three-dimensional structure of the homeodomain protein was determined, it was found to have a helix-turn-helix structure related to that already known in bacteria. This was the first indication that principles of gene regulation in bacteria may also apply to higher organisms. Indeed, homeodomain proteins are now well recognized from bacteria to man.

To appreciate the significance of homeotic genes to the development of complex structures such as the human eye, it may be helpful to digress momentarily to the early history of the field. In 1859, Charles Darwin, motivated by curiosity of the origin of diversity, noticed that repetition of elements along the length of the body was a common feature of many animals and that the variation of related structures contained in these elements contributes to diversity. It is now well recognized that serial homology is a feature of metameric (segmented animals), whose structures such as legs, nerve ganglia, appendages, blood vessels, and so on occur within each segment. Darwin considered that through natural selection, structures gradually change from one form to another. However, the zoologist William Bateson, who coined the term “genetics,” pointed out that evidence for intermediate forms was often lacking. Bateson was therefore particularly interested in the phenomenon where the structures on a particular segment were transformed to the structures normally present on another segment [154, 155]. In 1894, he defined the term homeosis as the process whereby one segment is transformed into the likeness of another. Examples of homeosis are present in both

plants and animals. For example, Cuenot (1921) noted that amputation of the antenna of the stick insect caused regeneration of a leg instead of an antenna in its place [338]. This concept is particularly interesting in that it went against Darwin’s hypothesis that structures change gradually through evolution.

Almost 30 years would go by before Thomas Morgan in 1923 working on fruit flies in his crowded lab at Columbia University made the observation that homeosis is inherited and that the responsible genes appear to reside on the fly’s third chromosome [156]. Fifty more years would pass before the discovery of a complex of genes whose role in development is to define the appendages that characterize each of the three segments that make up the thorax [157, 158]. This insight into the

molecular basis of homeosis came from a set of mutations in Drosophila that cause strange changes in the body of the adult fly. For example, in the antennapedia mutant, legs replace antennae on the head. In the bithorax mutant, an extra pair of wings appears where normally there should be small appendages known as halteres. These mutations fit Bateson’s definition of homeosis as the transformation of parts of the body into structures appropriate for another position.

The homeotic selector genes are all part of a multigene family and lie in two gene clusters, the bithorax and the antennapedia complexes. The bithorax complex controls differences in abdominal and thoracic segments, while those in the antennapedia complex control the difference in the thoracic and head segments. Together,