Sustained high levels of RA signaling modify photoreceptor fate
Treatment of zebrafish embryos with 0.3 μM RA beginning at the time of retinal neurogenesis (36 hpf) resulted in increased densities of rod photoreceptors and correspondingly decreased densities of cone photoreceptors. Local and global pattern attributes (as measured by NND, DRP, and quadrat analyses) of the enhanced rod mosaic and the depleted red cone mosaic collectively suggested that the retinal progenitor population competent to generate photoreceptors, generated more rods, possibly at the expense of red cone photoreceptors. Assessment with other rod and cone-specific markers supported this hypothesis. Alternative hypotheses, such as accelerated/decelerated differentiation of specific cell types, widespread but selective cone cell death, or the generation of photoreceptors of mixed phenotype, were either discounted or poorly supported. A parsimonious explanation for these results is that sustained RA treatment beginning at the time of retinal neurogenesis influences retinal progenitors and photoreceptor precursors, favoring a rod fate over a cone fate, with the spatial positions of the 'missing' cone photoreceptors being anomalously occupied by additional rods.
Several transcription factors have been shown to alter the fates of rods vs. cones. In mouse, absence of either NRL, or NR2E3, results in a retina with photoreceptors expressing only cone markers and having ambiguous photoreceptor morphologies [61–63], and there is evidence that NRL activates NR2E3 to suppress the development of cones . Two other nuclear receptors, RXRγ and TRβ2, are required to suppress expression of the S-cone opsin in mice, favoring the production of cones expressing M-opsin [65, 66]. A compelling example of photoreceptor fate manipulation by a single transcription factor is provided by the zebrafish tbx2b mutant, in which there is a replacement of UV cones with rod photoreceptors, and the 'new' rods have an entirely unambiguous identity although they occupy the spatial positions of UV cones within the photoreceptor mosaic . Tbx2b is expressed in the developing retina well in advance of photoreceptor neurogenesis, suggesting that intrinsic factors controlling photoreceptor fate may exert their effect on progenitor cells rather than upon immediately postmitotic precursors.
The present study now provides evidence that an extracellular signal - RA - can influence the fate of retinal progenitor cells in the zebrafish. Somewhat similar to the situation with tbx2b, RA has these effects only when administered in advance of photoreceptor neurogenesis. However, sustained high levels of RA signaling, up to and including the time of photoreceptor terminal mitosis, are required. This finding, together with evidence for induced RA signaling within proliferative cells and new rod photoreceptors, suggests that the developmental trajectory of retinal progenitors can be influenced by extrinsic as well as intrinsic factors. It is interesting that only one other 'extracellular' signaling mechanism has been demonstrated to disrupt the formation of the photoreceptor mosaic in zebrafish: the Notch signal transduction pathway . These effects were obtained by treating zebrafish embryos with pharmacological inhibitors of Notch signaling at 24 hpf or earlier, well in advance of photoreceptor neurogenesis.
An unresolved question is the relationship between embryonic rod and cone progenitors . A recent study from our laboratory  found that in zebrafish embryos both rods and cones, as well as rod and cone precursors, express an identical suite of "photoreceptor-specific" transcription factors including rx1, neuroD, crx, and nr2e3, suggesting that additional factors are required to specify rod vs. cone fate. It is possible that the mechanism controlling photoreceptor fate decisions in zebrafish is stochastic, and depends in part on the relative strength of competing local extracellular (and intracellular) cues to impel a progenitor toward the rod program of development. This model is consistent with further results of the present study, demonstrating that RARαb knockdown causes a reduction in endogenous RA signaling and the number of rod photoreceptors without a significant alteration of other retinal cell types.
The local strength of an RA signal therefore may be a factor influencing rod vs. cone cell fate decisions in the zebrafish retina. The enzymes involved in RA synthesis in the vertebrate retina are expressed in ventral and dorsal domains, suggesting endogenous RA exists in a gradient across the retina [28, 31, 32, 69], with its lowest level in the central retina. The use of the RARE-YFP reporter line corroborates that in zebrafish a strong ventral domain of endogenous RA signaling exists , which is also the location of the initial patch of rod and cone photoreceptors . A smaller patch of rods forms in the dorsal retina. In contrast, the central retina (in the center between RA gradients), initially develops few rods. Treatment with exogenous RA during retinal neurogenesis may disrupt the endogenous gradients of RA, increasing the number of retinal progenitors that assume a rod photoreceptor fate and decreasing the population available for cone genesis. In RARE-YFP embryos treated with exogenous RA, some of the YFP+ cells that are mitotic are in the position of potential rod progenitor cells (Figure 7), and the numbers of YFP+ rods may correspond to the additional rods in RA-treated retinas (Figure 8). Similarly a loss of RA signaling in RARαb morphants reduces the number of photoreceptors precursors that ultimately assume a rod fate.
Pleiotropic effects of prolonged high levels of RA signaling
The RA treatment used in this study resulted in a complex retinal phenotype, with effects on photoreceptor fate as discussed above, but also on laminar position of photoreceptors, and retinal cell survival. The ectopic photoreceptors in RA-treated retinas (Table 3) may have resulted from the fate-influencing activity of RA, such that cells positioned to become inner retinal neurons instead expressed photoreceptor genes. Surprisingly, many of the ectopic photoreceptors could be labeled with cone-specific markers, suggesting that the cone-to-rod fate influencing effect of RA is limited to cells that ultimately reside in the ONL. Alternatively, RA may affect cell movements of photoreceptor precursors, causing some cells to migrate in a basal direction rather than into the ONL. Related to this speculation is that the effects of RA on retinal cell survival were most pronounced within the INL, further indicating specific effects of RA upon this cell population. Exogenous RA may result in cell death in the INL due to the generation of nonviable cells of abnormal phenotype; alternatively, RA may have some other selectively toxic effect on cells of the INL. It is also possible that the apparent effects of RA on photoreceptor fate are indirectly mediated by a tissue disorganizing outcome of RA toxicity. Indeed, we were unable to assess the effects of RA during retinal neurogenesis independently of RA's effects on retinal cell survival. However, in the RARαb morphants, exogenous RA failed to significantly alter the production of rod photoreceptors, providing compelling evidence that the effects of exogenous RA are mediated at least in part by RA signaling via specific RA receptors.
Cellular RA signaling in response to exogenous RA
Endogenous RA signaling in the vertebrate retina exists in distinct, separate dorsal and ventral domains [31, 34, 55], suggesting that any effects of endogenous RA signaling on photoreceptor development outside of these domains must be indirect. In contrast, exogenous RA leads to global changes in photoreceptor gene expression in the retina [31, 52], or to changes in the production and positioning of specific photoreceptor cell types (the present study) suggesting that the ability of retinal cells to respond to RA is more widespread. Our results with the RARE-YFP line are consistent with global, as well as direct effects of RA on photoreceptor development, as exogenous RA treatment leads to widespread upregulation of the transgene. Multiple cell types are capable of engaging in RA signaling, including mitotic cells, rod and cone photoreceptors, RPE, Müller glia, and inner retinal neurons. Increases in RA signaling occur on a rapid temporal scale that indicates no requirement for upregulation of signaling machinery. A conclusion from these results is that many cells in the retina possess receptors and coactivators capable of generating a cellular response to RA. A more speculative inference is that presumptive photoreceptors that experience prolonged RA signaling during retinal neurogenesis are driven toward a rod rather than a cone fate.
Retinoic acid receptors and endogenous RA signaling during retinal neurogenesis
The complete family of RAR and RXR genes in zebrafish has been identified [57, 58, 70]. In the present study we clarified the retinal expression patterns of RARαb and RXRγ. During retinal neurogenesis, RXRγ mRNA appears in a strong ventronasal domain, and then is expressed in cells of the INL (Figure 10). Later in development, RXRγ mRNA is transiently expressed in outer retina. The RXRγ-expressing cells of the outer retina at 48 hpf likely correspond to retinal progenitors fated to become photoreceptors or Müller glia, and the RXRγ-expressing cells of the outer retina observed at 55 hpf correspond to newly-differentiating rod and cone photoreceptors [1, 3]. After formation of the embryonic retina, RXRγ expression becomes limited to the edge of the CGZ (Figure 10D), the major source of new retinal neurons during larval and adult growth [9, 71] and within the most recently-generated cells of the ONL. RXRγ is therefore a candidate for mediating RA signaling in retinal progenitor cells and cells of the ONL. Our data are consistent with those obtained from mouse and chick models, where RXRγ is expressed in developing cone photoreceptors [23, 24, 41]. Interestingly, the apparent peak of expression of RXRγ in the outer retina (55 hpf in zebrafish) precedes photoreceptor opsin expression in the majority of photoreceptors  (and see ref. 41), making unambiguous colocalization studies in rods vs. cones unfeasible, and suggesting roles for RA signaling over the time of photoreceptor determination as well as differentiation.
RXRs act as homodimers or as heterodimers with other nuclear receptors such as RARs or thyroid hormone receptors (TRs). Similarities in retinal phenotypes of RXRγ null as compared to TRβ2 null mice led to the suggestion that these two nuclear receptors operate together to influence cone photoreceptor gene expression [41, 65]. However, the binding partner(s) for RXRγ in the developing retina are not clearly known. In addition, RARα has been shown to mediate RA signaling in the mouse retina [72, 73]. The results of the present study found a zebrafish homologue of RARα, RARαb, is expressed early in neurogenesis in the RGC layer, and at low levels throughout the retina at later stages (Figure 10H). Therefore RARαb is also considered a candidate for regulating endogenous RA signaling in the retina.
Targeted knockdown of RARαb resulted in a reduction, though not absence, of the endogenous expression of reporter in the RARE-YFP line (Figures 11 and 12). This is in agreement with knockout studies of RARα in mice, in which the absence of RARα was associated with the elimination of expression of an RA signaling reporter transgene [72, 73]. The role of RARα in retinal cell differentiation is more ambiguous. In mice, knockout of RARα has no effect on retinal morphology or retinal cell differentiation. In the present study using zebrafish, knockdown of RARαb resulted in a significant reduction of the number of rod photoreceptors in the central and dorsal retina. These distinct results in mouse and zebrafish may reflect differential subfunctionalization of the RA receptor subtypes in the two model organisms. Alternatively, the experimental endpoints available in zebrafish (number of differentiating rods) may be more sensitive for the detection of the function of RARα. We suggest that, in the zebrafish, RARαb and an unknown binding partner mediate the activation of a rod neurogenesis program in retinal progenitor cells in response to RA. This mechanism was tested further by treating RARαb morphants with exogenous RA to determine if knockdown would block an increase in the number of rods. Exogenous RA resulted in a widespread upregulation of RA signaling in RARαb morphants, but not in a significant increase in the average number of rods. We interpret this to mean that the RARαb receptor is not essential for mediating RA signaling in response to exogenous RA, but does play a role in regulating rod production in the zebrafish retina.
The presence of retinoid receptors in the developing ONL, together with the capacity of differentiating rods, cones, and progenitor cells to respond directly to exogenous RA, indicates a significant role (or roles) for endogenous RA in regulating photoreceptor development. Defining these endogenous roles has been remarkably elusive. In teleost models, reduction of RA synthesis has been accomplished by knocking down expression of β,β-carotene-15,15'-oxygenase (bcox) , or of the apc gene, which also indirectly reduces RA synthesis , or of the vax2 transcription factor gene, which alters the distribution of RA-synthesizing enzymes . In each of these models the disruption of RA synthesis results in disruption of photoreceptor morphology and of expression of photoreceptor markers [75–77]. Temporally-selective reduction of RA synthesis by treating embryos with the pharmacological inhibitor citral over the time of photoreceptor differentiation resulted in reduced rod opsin expression . In the present study we have uncovered a role for a specific RA receptor, RARαb, in mediating the effects of endogenous, as well as exogenous, RA upon photoreceptor development. However, we note that, in the RARαb morphants, RARαb is chronically depleted over a protracted developmental period, similar to the situation of chronic RA depletion in bcox and apc morphants [75, 77]. Therefore, while these studies collectively provide compelling evidence that endogenous RA signaling, and specifically RARαb, are required for normal photoreceptor development, interpretation of the phenotypes is complicated by the consideration of early roles for RA signaling in eye formation. In the present study we have used a more targeted approach by knocking down the RARαb receptor, which avoids the effects of a global reduction in RA synthesis. However, an indirect role for RARαb in regulation of rod production cannot be ruled out.
Dynamic roles for RA during vertebrate retinal development
Overwhelming evidence from several animal models supports numerous functions for RA signaling during the development of the vertebrate eye [25, 26, 78, 79]. A collection of in vitro and in vivo studies specifically demonstrates important activities of RA signaling with respect to photoreceptor development. These include: a) RA promotes the rod fate at the expense of other, non-photoreceptor retinal cell fates [36, 40]; b) RA accelerates or decelerates the rate at which differentiating photoreceptors express specific markers [31, 52, 80]; c) RA promotes photoreceptor survival [37, 81], and d) RA recovers photoreceptor differentiation in ethanol-treated embryos . The results reported here now also support a role for RA in promoting the rod photoreceptor fate at the expense of cone fates. We suggest that temporal shifts in the role of RA signaling, and by implication the functional, molecular targets of the RA signaling machinery, underlie these distinct experimental outcomes. Retinal progenitors competent to generate photoreceptors may assume a transient state of plasticity that can be influenced by extrinsic factors such as RA, or intrinsic factors such as tbx2b , resulting in altered fate of their progeny. Differentiating photoreceptors may also experience a period of sensitivity to extrinsic factors such as RA, which regulate the rate at which they express photoreceptor-specific genes [31, 68]. This model predicts that targets of RA signaling will be at least partially distinct in retinal progenitor cells as compared to differentiating rod and cone photoreceptors. This model is consistent with a molecular mechanism recently demonstrated in mouse retina, in which post-translational modifications of nuclear hormone receptors modulates their activity in a dynamic manner . Our ongoing experiments are aimed at identifying cell-type-selective molecular targets in order to further reveal mechanisms through which RA controls photoreceptor development.