generation from RPCs is known and certain early morphological and molecular markers exist, postmitotic cells differentiating along the rod photoreceptor pathway can be defined as rod precursors, which distinguish them from RPCs. Rod precursors are also distinguishable from mature rod photoreceptors on the basis that rod inner and outer segment morphogenesis and synaptogenesis are barely under way, and the genes specific for rod phototransduction are not yet expressed.

Rods have been traditionally identified as such when they express the rhodopsin gene because of its exquisite cell type specificity. Perhaps because of this, many studies of rod development have focused on the cellular parameters and molecular factors associated with rhodopsin expression. For example, the timing of activation of rhodopsin expression in the rat retina varies between rod precursors born before and those born after E19 [11, 12] and the timing of rhodopsin expression in other species has been correlated with the timing of cone opsin expression in neighboring cells (see Chapter 1). There are now several signaling pathways and transcription factors identified that regulate these parameters as well as control the expression of other factors important for rod function and survival. As another example, the exquisite cell type specificity of rhodopsin expression prompted efforts to identify the cis-acting DNA sequences and associated transcription factors that regulate rhodopsin expression. Many of those factors, perhaps not surprisingly in retrospect, play important roles in the pathway leading to mature rods. In the next sections, we describe those factors that in recent work were strongly implicated in controlling the important early steps in rod photoreceptor differentiation. Table 1 summarizes extracellular and transcription factors implicated in different aspects of rod photoreceptor development. Included in the table are those factors we are not able to describe due to space constraints.

TRANSCRIPTION FACTORS

Basic Helix-Loop-Helix Genes

Basic helix-loop-helix (bHLH) genes constitute a large class of transcription factors that contain a stretch of positively charged amino acids termed the basic motif, which is important for DNA binding, and a helix-loop-helix domain, which is important for dimerization with other bHLH proteins. The bHLH genes have important roles in the nervous system, from the earliest stages of embryogenesis to the survival and function of mature neurons. In the Drosophila retina, the bHLH gene atonal has an essential role in specifying the R8 photoreceptor, which sets in motion the wave of neurogenesis in each presumptive ommatidial unit [35]. No fewer than ten bHLH genes have been identified in the mammalian retina, and most have some role either in RPC maintenance or in promoting cell fate and differentiation of neurons and Müller glia [36–38].

Among the retinal expressed bHLH genes, Neuro D may have the most direct role in photoreceptor development because of its abundant, although not necessarily exclusive, expression in photoreceptor precursors and because its overor misexpression is sufficient to promote photoreceptor differentiation in RPCs or in nonneural ocular cells such as RPE and iris epithelial cells [39–47]. Neuro D expression is also correlated with de novo photoreceptor differentiation in damage-induced regeneration in the chick retina [48], and Neuro D is expressed in the rod precursor lineage in the adult goldfish retina [49], a

ECM extracellular matrix, TF transcription factor

specialized neurogenic population that gives rise to new rods throughout the life of the organism.

Resolving the requirement of Neuro D, and for that matter, other bHLH genes, in rod photoreceptor differentiation has been more difficult. Although work done in the chick retina suggests that Neuro D is required for photoreceptor development in that organism [50], the predominant photoreceptor phenotype in Neuro D knockout mice is decreased survival [42, 43]. One reason why Neuro D may not be required for rod photoreceptor differentiation in mice is because of redundancy or functional compensation by other bHLH proteins in its absence.

To address this, Akagi et al. (2004) [51] developed multiple bHLH triple-knock- out mice, and the most severe deficit in photoreceptor number was observed in the Mash1−/−, Math3−/−, Neuro D−/− triple-knockout mouse. However, rod photoreceptor development is still evident even in this combinatorial mutant. Furthermore, the high levels of photoreceptor apoptosis observed in this mutant leaves it unresolved whether elimination of these three genes causes a direct problem in photoreceptor specification or differentiation. Compounding the problem further, other bHLH genes are upregulated when certain bHLH genes are genetically inactivated, and there is ample evidence for functional redundancy among these genes.

Otx2

Otx2 is a homeobox gene that is widely expressed in early embryos and becomes progressively restricted to the anterior portion of the neural tube. Consistent with its expression pattern, the Otx2 knockout mouse is embryonic lethal and lacks forebrain structures [52–54]. In the developing eye, Otx2 is expressed in the developing RPE and retina, and genetic studies in compound mutants of Otx2 and its paralog Otx1 showed that these genes are essential for RPE development [55–57].

Retinal Otx2 expression has been examined in several vertebrate species, and some differences in timing of activation have been reported (i.e., proliferating RPC vs. postmitotic precursors) [55–60]. In mice, Otx2 is expressed early in retinal development, but in postmitotic cells and in a pattern consistent with the generation of retinal neurons, most notably photoreceptors. Otx2 expression in the photoreceptors is transient and is ultimately expressed in the inner nuclear layer (INL) in presumptive bipolar cells, horizontal cells, and Müller glia [61]. Baas et al. (2000) [62] showed that Otx2 subcellular localization (nuclear vs. cytoplasmic) is developmentally regulated and cell type specific.

To investigate the role of Otx2 in neural retinal development, Nishida et al. (2003) [61] generated conditional knockouts by crossing mice with a floxed Otx2 allele to mice with Cre recombinase expression controlled by the Crx promoter. Although some Otx2expressing cells persisted in these mice, the changes in retinal cell differentiation were dramatic. There was an almost-complete absence of rhodopsin-expressing cells and a major increase in the number of cells expressing markers of amacrine cells. The magnitude of these changes suggests that loss of Otx2 culminated in a change in fate from rods to amacrine cells, and this is consistent with Otx2 overexpression experiments. However, there is a large increase in apoptotic cells, and this could account in part for the loss in rods. Thus, although the effects of Otx2 inactivation on the retina are likely

to be complex, these experiments demonstrated the essential nature of Otx2 for rod photoreceptor development or survival.

Crx

Crx is a homeobox gene that is closely related to Otx2. Initial reports showed that mutations in the human CRX gene cause autosomal dominant cone-rod dystrophy (CORD2) and late-onset retinitis pigmentosa [63, 64]. Subsequent studies showed that both dominant and recessive CRX mutations are also linked to Leber’s congenital amaurosis (LCA7), suggesting that CRX function is important for photoreceptor development in humans [65–71]. As described next, work done in animal models and in biochemical studies strongly supports this idea.

Crx was discovered by its sequence similarity to other known homeobox genes [63, 72] and in a DNA–protein interaction screen (one-hybrid screen) using a retinal complementary DNA expression library and the Ret4 DNA sequence found in the rhodopsin promoter as bait [73]. Like Otx2, Crx expression begins early in retinal development [58, 60, 72, 74–78]. However, Crx follows Otx2 activation and is likely to be a direct transcriptional target of Otx2 in postmitotic rod and cone precursors [61]. Whereas Otx2 expression is downregulated in photoreceptors and remains strong in the INL in the mature retina, Crx remains highly expressed in the outer nuclear layer (ONL) and in a smaller subset of cells in the INL than Otx2.

The importance of Crx in mammalian photoreceptor formation is now well established. Overexpression of Crx is sufficient to bias cells toward a photoreceptor fate in the embryonic retina and in adult iris epithelial cells [72, 79], and overexpression of a dominant negative form of Crx interferes with outer segment formation [72]. Consistent with these findings, genetic inactivation of Crx in mice causes developmental disruptions in outer segment formation and synaptogenesis as well as decreased expression of many phototransduction genes [80–83].

Crx acts as a direct transcriptional activator, and many studies showed Crx to be an integral component of multiprotein complexes that occupy the regulatory regions of several rod and cone photoreceptor genes [60, 72, 73, 75, 84–94]. Several interacting proteins have also been found that interfere with Crx-mediated transactivation, and these interactions may be important for the timing of photoreceptor gene activation during differentiation. These interactions may also function to restrict the transcriptional activation function of Crx to rods or bipolar cells [95–99]. Crx has also been shown to directly interact with Ataxin-7, a protein that undergoes polyglutamine expansion and is responsible for spinocerebellar ataxia type 7 (SCA7). While the interaction with wild-type Ataxin-7 is important for the assembly of a Crx transactivation complex, the polyglutamine-expanded, neurodegenerative form of Ataxin-7 interferes with this interaction, resulting in a downregulation of photoreceptor gene expression, and this appears to be a causative factor for the cone-rod dystrophy found in SCA7 [100–103].

Interestingly, Crx is not essential for the initial specification and differentiation of photoreceptors, as evidenced by their appearance in the Crx−/− mouse retina. This is somewhat surprising given its early onset of expression. However, Bibb et al. (2001) [74] showed that, in the fetal human retina, the expression of several photoreceptor genes that

are targets of Crx regulation in the mouse precedes the onset of CRX expression, which further suggests that Crx is not required for early photoreceptor development.

Why cell fate specification and early differentiation of photoreceptors occur in the absence of Crx is not known. One possibility is that Otx2 partially compensates or is redundant with Crx. Consistent with this idea, Crx and Otx2 are both in the Otx gene family [58, 76, 104], both are expressed early in photoreceptor differentiation, Crx expression is not maintained in the Otx2−/− retina, and Otx2−/− mice show no evidence of photoreceptor formation [61]. Furthermore, Crx can interact with Mitf and promote pigment cell differentiation in vitro to an extent comparable to Otx1 and Otx2 [58]. However, experiments in Xenopus suggest that Otx5b (Crx) and Otx2 are not redundant in retinal cell type specification: Otx5b and Otx2 promote photoreceptor and bipolar cell fates, respectively [60]. Zebrafish contain two Crx-related genes (Crx and Otx5), and each seems to have unique functions in regulating gene expression in the retina and pineal gland, which in mammals appear to be relegated to Crx alone (mammals contain a single Crx/Otx5 gene) [59, 75, 105]. In contrast to other species, zebrafish Crx is also expressed in RPCs, and morpholino-mediated RNA knockdown disrupts proximo-distal patterning of the retina and causes a general delay in neuronal differentiation [59]. To ultimately address the issue of whether redundancy or compensation accounts for the nonessential nature of Crx in photoreceptor cell fate specification in the mouse retina, it may be necessary to create a dominant negative Crx allele that interferes with the functions of both Otx2 and Crx.

QRX/Rax-L/Rx-L

QRX is a homeobox gene in the Rx family. Rx proteins contain paired-like homeodomains and are distantly related to other important retinal transcription factors, such as Chx10, Vsx1, Pax6, and the Otx cohort. QRX is found in the human and bovine genomes [106], and putative orthologs are identified in chick (Rax-L) [107, 108] and Xenopus (Rx-L) [109]. The role of QRX/Rax-L/Rx-L is not as well established as the other genes described in this chapter for two reasons: An ortholog has not been identified in the mouse, and a congenital photoreceptor disease has not been found in humans. Nevertheless, QRX, Rax-L, and Rx-L have all been shown to bind to conserved DNA sequences in the rhodopsin promoter, interact with Nrl and Crx, and promote transcription [106, 107, 109]. Furthermore, dominant negative experiments in chick and gene knockdown experiments in Xenopus both supported a role in the maturation or survival of photoreceptors [107, 109].

NRL

The Nrl gene is a member of the Maf transcription factor family. Maf proteins contain a basic motif linked to a leucine zipper domain (bZIP), which is required for DNA binding and protein dimerization, respectively [110]. Several Maf genes have important roles in lens development [110–114] and Nrl has emerged as a key player in rod photoreceptor development. Genetic inactivation of Nrl (Nrl−/−) leads to a complete loss of rods, but interestingly, their absence is not due to a failure in photoreceptor cell production or in decreased survival. Rather, the ONL in the Nrl−/− retina is well

Fig. 1. Interaction between extracellular signals and transcription factors that results in a stepwise progression of rod photoreceptor development. See Table 1 for definitions of abbreviations.

populated, and these cells have the characteristic expression profiles and physiological and morphological properties of cones, in particular S cones [115–117]. These observations strongly suggest that Nrl is an essential transcription factor for executing the rod differentiation program and, at the same time, for keeping the cone differentiation suppressed. Consistent with this, many studies showed that Nrl is a direct positive regulator of rhodopsin gene transcription.

A prediction of this model is that Nrl should be expressed in postmitotic photoreceptor precursors. While true, some studies suggested that Nrl is expressed in the inner retina, and its onset of expression is earlier than the period of rod genesis [118, 119]. However, more recent studies examining Nrl protein expression in human and mouse embryos [120, 121] and in a transgenic mouse line expressing green fluorescent protein (GFP) under the control of the Nrl promoter [122] suggested that Nrl may be specific to postmitotic rod precursors.

Although the discrepancies in the reported Nrl expression characteristics are not completely resolved, Oh et al. (2007) expressed Nrl under the control of the Crx promoter in both Nrl+/+ and Nrl−/− mice [123]. In both genetic backgrounds, the expression of Nrl drove rod photoreceptor development and did so at the expense of cones. In addition, transgenic Nrl expression in the Nrl−/− retina rescued the dysmorphic architecture of the ONL. These dramatic results showed not only that Nrl is sufficient to drive rod photoreceptor development when expressed in early postmitotic photoreceptor precursors, but also that cone precursors are competent to execute a rod photoreceptor program. Thus, “photoreceptor precursor” may be a more accurate description of earlystage rod and cone precursors, and onset of Nrl expression may signify a transition to a more restricted rod precursor state (see Fig. 1 and Conclusions).