Keywords

9.1 Introduction

9.1.1 GRNs Specifying the Eye Morphogenetic Field

Under the influence of signals determining the mayor embryo axes (WNTs, BMPs, Nodal and FGFs), a large group of cells in the anterior neural plate gets specified as presumptive eye tissue (Wilson and Houart 2004). Classical explant experiments in salamanders have shown that, even before this region evaginates to form the optic vesicles, it is already committed and will develop as an optic cup when cultured in vitro (Lopashov and Stroeva 1964). Early expressed in this territory, a number of transcription factor-encoding genes (known as eye field transcription factors or EFTFs) have been acknowledged as the molecular signature defining the identity of the tissue. These include homeobox genes such as Rx, Pax6, Six3, Lhx2, or Six6. Among them, especially Rx, Pax6 and Six3 are essential for eye formation in all vertebrate models analysed (Sinn and Wittbrodt 2013). The fact that eye specification in Drosophila also depends on eyeless, twin of eyeless (both homologous of Pax6) and sine oculis (homologous of Six3 and Six6) suggests a conserved “Kernel” for the development of the eye field in bilaterians (Davidson and Erwin 2006; Wagner 2007). Although it seems clear that vertebrate EFTFs constitute central nodes of a complex GRN, their precise hierarchical relationships are still poorly understood. The miss-expression of a few eye specification genes, such as Six3 and Pax6, is sufficient to induce the ectopic expression of eye tissues in vertebrates (Chow et al. 1999; Loosli et al. 1999; Zuber et al. 1999; Lagutin et al. 2001). While this fact points to a top hierarchical position for both genes in a “linear GRN model”, it is very likely that the network’s assembly is more complex and multiple steps of feedback regulation exist, as previously reported for eye specification in Drosophila (Treisman 1999; Kumar and Moses 2001). Thus, it has been shown that the miss-expression of Six3 or Pax6 mRNAs recruits other eye specification genes, and that the co-expression of EFTF cocktails acts as a much more potent inducer than that of single genes, being sufficient to induce ectopic eyes outside the nervous system (Zuber et al. 2003) and to instruct pluripotent cells into the eye developmental program (Viczian et al. 2009). Interestingly, these studies also showed that EFTF cocktails’ efficiency to induce ectopic eyes largely depended on the inclusion of Otx2 in the mixture (Zuber et al. 2003). This is in agreement with previous reports showing the important role of Otx genes in eye formation (Matsuo et al. 1995; Martinez-Morales et al. 2001), and with the observation that ectopic eye induction mediated by Pax6 or Six3 is restricted to the Otx expression domain (Chow et al. 1999; Loosli et al. 1999).

An attempt to define the regulatory relationships among nodes (i.e. the genes and their regulators) at the core of the eye field GRN has been carried out in Xenopus (Zuber et al. 2003). In this report EFTFs regulatory interactions were tested in overexpression experiments and a tentative GRN, comparable to that proposed for Drosophila eye development, was deduced (Fig. 9.1). Several predictions from this model were consistent with hierarchical relationships found in Xenopus and other vertebrate models through gain and loss of function experiments. For example, a downstream role for Six6/Optx2 in the GRN specifying the eye field was confirmed (Zuber et al. 1999, 2003; Li et al. 2002). However, although useful as a working model, it should be taken with caution as is merely based on overexpression studies and some of its assumptions have been already proved to be incorrect. This is the case for the prominent position of Rx/Rax at the top of the hierarchy in the eye GRN. In contrast to the model’s prediction, the expression in Xenopus of the EFTF (i.e. Pax6, Six3, Lhx2, and Six6) seems unaffected at early neurula stage in Rax mutant embryos (Fish et al. 2014), thus indicating a downstream role for this gene in the network. More importantly, there is strong evidence showing that the exact wiring of the eye field GRN varies in different vertebrate groups. With the possible exception of Six3, which seems to occupy a prevalent upstream position by suppressing canonical Wnt signalling anteriorly in all species analysed (Wallis et al. 1999; Carl et al. 2002; Lagutin et al. 2003; Liu et al. 2010; Nakayama et al. 2013), the regulatory weight and hierarchical position of other EFTF differs considerably among species. In some cases, phenotypic discrepancies between mutants in different species can be attributed to the existence of multiple paralogs (i.e. genes related by duplication within a genome) for a given EFTF in the teleost models. This is the case for Pax6, whose inactivation in mouse and Xenopus results in an almost complete loss of the eye territory (Hill et al. 1991; Suzuki et al. 2013), whereas only causes microphthalmia when one of the two Pax6 paralogs, pax6b, is mutated in zebrafish (Kleinjan et al. 2008). Nevertheless, gene duplication cannot always justify the observed phenotypic discrepancies. Thus, in mouse Rx mutants eye field determinants are down-regulated at very early stages and consistently eye development is impaired even before optic vesicle evagination (Mathers et al. 1997; Zhang et al. 2000; Medina-Martinez et al. 2009), but in Xenopus and teleost fish Rx function seems dispensable during eye field specification , being required for optic vesicle evagination and eye identity maintenance later on Loosli et al. (2001, 2003), Rembold et al. (2006) and Fish et al. (2014). Similarly, in Lhx2 mutant mice eye development is arrested prior to the formation of an optic cup (Porter et al. 1997; Tetreault et al. 2009) whereas the homologous mutation in zebrafish (belladonna) displays a milder phenotype affecting the patterning of the ventral forebrain and eye (Seth et al. 2006). An extreme example of functional divergence among vertebrate species is the case of the transcription factor ET/Tbx2, which appears to play a central role during eye field specification in Xenopus (Zuber et al. 2003), but its loss of function only causes a mild microphthalmia in mice (Behesti et al. 2009).

Fig. 9.1
figure 1

GRNs specifying the eye morphogenetic field: a Schematic representation of anterior neural plate domains (color-coded) during eye field specification and optic vesicle evagination stages. Relevant transcription factors, downstream genes and morphogenetic processes are indicated. b Hypothetical eye specification networks are represented for Drosophila and vertebrates. Homologous genes are indicated in similar colored boxes. di diencephalon; mb midbrain; hb hindbrain

Full size image

In summary, the existence of cooperative effects, feedback regulatory loops, and species-specific wiring hinders the definition of a precise architecture for the core GRN involved in vertebrate eye specification. Even less information is available on the structure of the downstream layer of the network. Yet, it is likely that this sub-network includes genes controlling optic vesicle evagination. In tetrapods, the anterior neural tube develops as a hollow structure and vesicle evagination occurs by lateral bulging of the neuroepithelium (Hilfer 1983; Eiraku et al. 2011). In contrast, the neural tube develops as a compact tissue in teleosts, and the formation of the optic vesicle requires the migration, rearrangement and epithelialization of individual precursors (England et al. 2006; Rembold et al. 2006; Ivanovitch et al. 2013). A few downstream targets of rx3: nlcam, cxcr4 and epha4a/b4b have been shown to control the migratory and adhesive behaviour of the eye field precursors as optic vesicle evaginate in zebrafish (Brown et al. 2010; Bielen and Houart 2012; Cavodeassi et al. 2013). In spite of these advances, systematic attempts to identify potential downstream targets of the EFTF have not been carried out until recently, either by exploring the eye field transcriptome (Viczian et al. 2009), or by interrogating the network structure upon mutation of rx3 using an RNA-seq approach (Yin et al. 2014). The emergence of the powerful next-generation sequencing technologies coupled to ChIP methods has allowed the identification of cis-regulatory modules at a genome scale (ENCODE_Project_Consortium et al. 2012). The enormous potential of these approaches to investigate the complexity of the GRNs involved in eye development has just started to be explored. A couple of studies have been carried out to systematically characterize cis-regulatory modules occupied by Otx2, during gastrulation in Xenopus and in the adult mouse retina (Samuel et al. 2014; Yasuoka et al. 2014). Additional ChIP-seq studies focused on other EFTF will be instrumental not only to clarify the wiring diagram of the core eye field GRN, but also to infer direct cis-regulatory targets of these transcription factors.

9.1.2 GRNs Specifying Eye Domains

Once the eye morphogenetic field is specified, signalling molecules derived from the retina and neighbouring tissues act to restrict the precursors’ potentiality, subdividing the optic vesicle into three regions: the neural retina , the retinal pigmented epithelium (RPE) and the optic stalk .Footnote 1 Inductive signals include SHH and nodal secreted from the CNS midline, FGFs from the retina and the presumptive lens ectoderm, and activins, Wnts, and BMPs from the extraocular mesenchyme and the dorsal ectoderm; which specify the optic stalk , the neural retina and the RPE respectively (Adler and Canto-Soler 2007; Martinez-Morales et al. 2009; Fuhrmann 2010; Steinfeld et al. 2013) (Fig. 9.2). At early stages, vertebrate eye subdivisions cannot be considered tissue compartments in the strict sense of the term, as transfer of precursor cells has been reported between different domains (Holt 1980; Picker et al. 2009; Kwan et al. 2012). Thus, limits between territories are initially dynamic, and depend on sustained signalling input that maintains tissue identity by regulating domain-specific transcription factors. By the time the optic cup has folded, ocular tissues are stabilized into genuine compartments (i.e. with defined borders and no cellular intermingling) through reciprocal transcriptional repression. Some examples of mutual transcriptional repression contributing to border definition have been reported. They include Pax2/Pax6 and Mitf/Vsx2 antagonism that participate in the definition of the optic stalk /neural retina and RPE/neural retina borders respectively (Schwarz et al. 2000; Horsford et al. 2005; Bharti et al. 2008).

Fig. 9.2
figure 2

GRNs specifying eye domains: a signaling molecules derived from the presumptive lens ectoderm (grey), the midline and the dorsal ectoderm/mesoderm pattern the optic vesicle into the neural retina (green), optic stalk (brown) and RPE (blue) domains. b Hypothetical GRNs specifying each retinal domain are depicted. Transcriptional regulators and known downstream targets are represented in black. Signaling pathways are color-coded. Repressive interactions are represented in red

Full size image

The development of the different eye tissues entails the bifurcation of the eye field specification GRN into mutually exclusive developmental programs controlled by local sub-networks. This process translates into distinctive cell morphologies within each domain: flat for the RPE, and long or short bottle-shaped for the neural retina or optic stalk respectively. These differential cell geometries, established within a few hours window, will condition the morphogenetic movements that take place during optic cup invagination. Thus RPE and neural retina epithelia fold over the lens vesicle to form a bi-layered cup and optic stalk lips converge ventrally to close the choroid fissure groove (Martinez-Morales et al. 2009; Eiraku et al. 2011; Kwan et al. 2012). Each of the ocular domains retains during embryogenesis certain potentiality for transdifferentiation into a different compartment (Coulombre and Coulombre 1965; Pittack et al. 1991; Guillemot and Cepko 1992; Turque et al. 1996; Vogel-Hopker et al. 2000; Rowan and Cepko 2004). As organ development proceeds and retinal domains progressively acquire divergent morphological and physiological features (e.g. pigmentation, glial and neuronal cell types, etc.), potentiality is lost and the competence for transdifferentiation in the adult is only maintained in amphibians (Del Rio-Tsonis and Tsonis 2003; Fuhrmann et al. 2014).

Although much has been advanced in the last years, our knowledge on the structure of the GRNs that control the developmental programs of the neural retina , RPE and optic stalk is still fragmentary and even the precise relationships among the top upstream genes of these networks are unclear. Here, we summarize the main findings for each of the three ocular domains.

Neural retina GRN: At the end of the evagination process, the optic vesicle comprises two back-to-back epithelial layers: an outer layer (dorsal in teleost fish) apposed to the presumptive lens ectoderm, and an inner layer (ventral in teleosts) surrounded by mesenchymal tissue. These two layers initially similar in size and volume will differentiate to generate a thick neural retina , and a thin RPE (Svoboda and O’Shea 1987; Li et al. 2000). The neural retina specification network pivots on the transcription factor-encoding gene Vsx2, also known as Chx10, which is the first determination gene differentially expressed in the presumptive neural retina versus the presumptive RPE (Liu et al. 1994). A number of reports have shown that Vsx2 has an essential role in the specification of the retinal domain, restraining RPE identity (i.e. RPE specific GRN) through direct repression of the transcription factor Mitf (Rowan and Cepko 2004; Horsford et al. 2005; Bharti et al. 2008; Zou and Levine 2012). Vsx2 activity seems to be required for the maintenance of a neural retina specific GRN, whose main regulators (nodes) are inherited core components of the eye field GRN, including Rx, Pax6, Six3 and Six6 (Medina-Martinez et al. 2009; Fuhrmann 2010; Bharti et al. 2012).

FGFs derived from the presumptive lens ectoderm play a fundamental role in positioning the neural retina at the expenses of the RPE territory (Guillemot and Cepko 1992; Pittack et al. 1997; Hyer et al. 1998; Vogel-Hopker et al. 2000; Cai et al. 2010). Thus, FGF signalling acts to suppress the gene encoding for RPE transcription factor Mitf while activating the neural retina determinant Vsx2, setting up the boundary between both tissues (Nguyen and Arnheiter 2000; Horsford et al. 2005). Several laboratories have dissected the signalling cascade responsible for this inductive activity, which operates through the Shp2/MEK/ERK pathway (Zhao et al. 2001; Galy et al. 2002; Cai et al. 2010). Interestingly, the well-described trans-differentiation of the RPE to neural retina by FGF does not occur in null mutant mice for Vsx2 (Horsford et al. 2005). Thus, Vsx2 seems to be a direct target of the FGF/ERK pathway and Mitf repression by FGF depends on Vsx2 function.

The precise regulatory relationships between the core components of the neural retina specification network (i.e. Vsx2, Pax6, Six3, Six6; and Rx) are currently unclear. However, some of the downstream targets of the network have been inferred by transcriptomic analyses in Vsx2 knockout models, as well as in Vsx2−/− induced pluripotent stem cells (Rowan and Cepko 2004; Phillips et al. 2014). Again, RNA-seq and ChIP-seq technologies will be instrumental to identify more components of this GRN and to investigate systematically the wiring scheme of its core components.

Retinal Pigmented Epithelium GRN: The RPE is a highly specialized monolayered epithelium essential for the correct development and homeostasis of the adjacent neural retina (Raymond and Jackson 1995; Strauss 2005). Establishment of the RPE gene regulatory network depends on the cooperative activity of two core transcriptional regulators: Mitf and the Otx family members Otx1 and Otx2 (Martinez-Morales et al. 2004; Fuhrmann et al. 2014). Mitf encodes a basic helix-loop-helix (bHLH) transcription factor that plays a key role as master regulator of pigmented cell specification, both in melanocytes and retinal neuroepithelial cells (Hodgkinson et al. 1993; Steingrimsson et al. 2004; Arnheiter 2010). Mitf loss-of-function impairs the correct specification of the presumptive epithelium, which remains un-pigmented and develops as a pseudo-stratified neuroepithelium (Mochii et al. 1998; Nakayama et al. 1998; Bumsted and Barnstable 2000; Nguyen and Arnheiter 2000). Conversely, Mitf gain of function enhances the RPE regulatory network, and in certain genetic background mediates the transdifferentiation of the neural retina into pigmented cells (Planque et al. 1999; Horsford et al. 2005). Similarly, Otx genes are early restricted to the RPE territory during optic cup stages and are required to establish the identity of this tissue (Bovolenta et al. 1997; Martinez-Morales et al. 2001; Lane and Lister 2012). The expression of Mitf and Otx genes in the presumptive RPE depends on their reciprocal activity, and both cooperate to induce a pigmented phenotype interacting directly at the protein level (Martinez-Morales et al. 2003; Lane and Lister 2012). Both transcription factors have been shown to operate directly on the direct downstream effectors of the pigmentation cascade. Thus, Mitf and Otx proteins activate the transcription of melanogenic genes such as QNR71, Tyrosinase, TRP1 and TRP2, acting synergistically through their consensus motives, CATGTG (M-box) and TAATCC/T (K50-type homeodomain), respectively (Goding 2000; Martinez-Morales et al. 2003). Interestingly, it has been shown that Pax6 activity, normally associated to the development of the neural retina , is essential for the establishment of the RPE identity in conjunction with Mitf (Baumer et al. 2003; Bharti et al. 2012). The establishment and maintenance of the RPE regulatory network depends on the inductive activity from surrounding tissues, including the extraocular mesenchyme and the surface ectoderm. Among the inductive signals, activins derived from the mesenchyme (Fuhrmann et al. 2000) as well as BMPs and Wnts from the dorsal ectoderm (Hyer et al. 2003; Muller et al. 2007; Steinfeld et al. 2013) control the differentiation of the RPE.

As previously discussed for eye field specification (see previous section), species-specific differences in the architecture of the pigmented epithelium GRN have been documented among vertebrate groups. Thus, in teleosts Mitf seems to have a less important regulatory weight in RPE determination, being the regulatory network more dependent on Otx (Lane and Lister 2012). Divergent regulation has also been reported for inductive signalling. In mice Wnt-dependent RPE specification has been characterized as a β-catenin dependent process that involves the direct activation of TCF/LEF sites in Mitf and Otx2 enhancers (Fujimura et al. 2009; Westenskow et al. 2009). In contrast, RPE induction in chicken requires the cooperative activity of Wnt and BMP signalling through a GSK3β-pSmad pathway (Steinfeld et al. 2013).

Optic stalk (OS) GRN: As eye development proceeds precursor cells from the optic vesicle differentiate in two fundamentally different populations. Those precursors located proximally to the midline will give rise to the OS, whereas more distal cells will form the optic cup, including the neural retina and RPE domains (Peters 2002). Eventually, optic stalk cells undertake the differentiation program that leads to the formation of the optic nerve. This local GRN is established under the influence of signalling molecules that emanate from the midline and pattern the optic vesicles along the proximo-distal axis. Nodal, hedgehog (Hh), and FGF signalling pathways have been identified as positive signals for the establishment and maintenance of the OS developmental program, while restricting distal BMP inducers (Peters 2002). Nodal family members, such as one-eyed pinhead (oep) and cyclops, play an essential role in patterning the central nervous system ventral midline (Rebagliati et al. 1998; Sampath et al. 1998). Mutations in genes encoding for these TGFβ related ligands result in cyclopic defects and loss of midline identity markers, particularly Hh (Macdonald et al. 1995; Rohr et al. 2001). Hh, acting as a morphogen, is necessary to induce the expression in the proximal optic vesicle of the key nodes of the OS GRN, Pax2, Vax1 and Vax2 (see below) both in mammals and teleost models (Ekker et al. 1995; Macdonald et al. 1995; Chiang et al. 1996). Modifiers of Hh proximo-distal signalling help to define the morphogen influence domain in the ventral optic vesicle (Lee et al. 2008; Cardozo et al. 2014). In addition to axial signalling, other independent pathways active in the ventral optic vesicle, such as FGFs and retinoic acid, have been shown to regulate the expression of OS specification genes (Take-uchi et al. 2003; Lupo et al. 2005; Cai et al. 2013).

The core GRN for OS identity comprises three homebox-encoding genes Pax2 (Torres et al. 1996; Macdonald et al. 1997), Vax1 and Vax2 (Barbieri et al. 1999, 2002; Bertuzzi et al. 1999; Mui et al. 2005; Kim and Lemke 2006). Their mutations result in OS impaired development and hence are associated to coloboma, choroid fissure malformations and axonal guidance defects. Although most of the downstream targets of this core network need to be identified, the segregation of the optic cup and OS domains depends on the repression of Pax6, a central node in the specification of both the neural retina and the RPE territories (Schwarz et al. 2000; Mui et al. 2005; Bharti et al. 2012).

As already mentioned, the bifurcation of the eye field specification GRN into domain-specific developmental programs has a direct impact in the acquisition of defined cell morphologies within each compartment. However, very little is known on the molecular machineries controlling these morphogenetic processes. In fact, understanding how a particular GRN unfold may require the identification of its downstream targets. These, operating under the control of the master regulators, will modify directly basic cell properties such as adhesion, shape and contractility. The ojoplano (opo) gene, which has an essential role in neural retina morphogenesis by controlling integrin polarized endocytosis, is a paradigmatic example of such type of targets (Martinez-Morales et al. 2009; Bogdanovic et al. 2012). Recent advances in whole-genome transcriptomics and epigenomics open the possibility of systematically surveying for the downstream determinants of cell geometry and epithelial morphogenesis in early eye development.

Finally, most of the important nodes of the GRNs involved in eye domains specification (e.g. Pax6, Vsx2, Rx, Otx2, etc.) have also been identified as key nodes of the “coloboma gene network” (Fig. 9.3): i.e. the network of genes that have been found mutated in human families affected by microphthalmia, anophthalmia, and coloboma (MAC) (Gregory-Evans et al. 2004, 2013). Although this group of diseases represents a significant cause of blindness in children (5–10 %) (Porges et al. 1992), its molecular causes are complex and far from being completely understood. Therefore, gaining insight into the architecture of the GRNs involved in eye development has important medical implications.

Fig. 9.3
figure 3

Coloboma gene network: genes mutated in human families affected by Microphthalmia, Anophthalmia, and Coloboma (MAC) are depicted in this network. Neural retina (green) and optic stalk (red) specific sub-networks are indicated with different colors. Adapted from Gregory-Evans et al. (2013)

Full size image