Discussion
In this study, an ultrastructural analysis of Crx-/- photoreceptors was carried out. As Crx mutations have been associated with Leber's congenital amaurosis, the findings in this study broaden our understanding of the pathology of this disease. Two prominent pathologies were characterized in the Crx-/- retina: (1) An absolute block in outer segment morphogenesis was noted, with the block occuring at the elongation stage of outer segment formation; (2) Crx-/- photoreceptors exhibited a severe perturbation in synapse formation. This represents the first report of a synaptogenesis defect in an animal model of LCA.

Crx-/- photoreceptors cannot complete outer segment morphogenesis
Mutations in Crx represent one of a collection of gene mutations that lead to an outer segment formation defect. Homozygous null mutations in the peripherin/RDS gene [36] or in rhodopsin [29] lead to a failure of outer segment formation. The deficits in peripherin-/- and Crx-/- photoreceptor morphogenesis were found to be very similar. Vesicular structures in Crx-/- photoreceptors were observed that were similar to those previously noted in the rds/peripherin-/- mouse. It was initially proposed that these vesicles were due to the breakdown of outer segment membranes that were not properly recruited to the outer segments in the absence of peripherin, or were from the result of the breakdown of the microvilli of Müller cells [30]. Strong support in favor of the former explanation was provided by Nir and colleagues who demonstrated the presence of rhodopsin protein in these vesicles using immunoelectron microscopy against a rhodopsin epitope [37]. Further, as shown here, the vesicles appear to bud from the inner segments themselves.
In developing photoreceptors, an extraordinary growth process occurs whereby the outer segment is generated from the nascent connecting cilium (see [38] and references therein). Peripherin/RDS and ROM-1 proteins (localized in disc rims) and the opsin proteins (localized throughout the discs) have important roles in the structural integrity of mature outer segments (see [39,29]). ROM-1-/- mice produce disorganized outer segments with large disks [40]. Crx, by virtue of being a transcription factor, presumably controls genes that are responsible for the building and perhaps maintenance of the outer segment structure, including rhodopsin and peripherin. Using northern blots [34], microarrays [10], and serial analysis of gene expression (SAGE) [35], we have defined a large number of genes that are altered in their expression level in Crx-/- mice. We found that rhodopsin expression was severely diminished in Crx-/- animals, and peripherin mRNA was reduced by approximately 30%. Transgenic mice with variable levels of expression of wild type rhodopsin exhibit rod degeneration [41], indicating the importance of the level of rhodopsin expression. In addition, the timing of rhodopsin expression may be very important, as indicated by studies in Drosophila.
In Drosophila, rhodopsin (ninaE) is expressed in photoreceptors R1–R6. In ninaE null mutants, the rhabdomere, a structure analogous to vertebrate outer segments, fails to develop in R1–R6 photoreceptors [42], reminiscent of the situation in rhodopsin-/- mice [29]. An intriguing experiment by Kumar et al. [43] demonstrated a temporal requirement for rhodopsin expression during rhabdomere development. In ninaE null flies, a ninaE transgene under the control of a heat shock promoter was subjected to various temperature shifts during development. Heat shock during the normal time of rhodopsin onset resulted in substantial and long-lasting rescue of photoreceptor structure and transient rescue of photoreceptor physiology. However, expression shortly before or after this critical period failed to rescue, suggesting that rhodopsin expression during a discrete window of time in development is essential for proper rhabdomere morphogenesis. This result is consistent with observations in the rat wherein rhodopsin onset occurs with strict timing in the developmental history of most rods in vivo [44]. Thus, through its regulation of rhodopsin levels, or perhaps through control of the kinetics of the up-regulation of rhodopsin beginning at about P6, Crx may be regulating outer segment morphogenesis. The similarty of the two cases may extend further. At present, the closest Crx relative in Drosophila is Otd, the founding member of the class of homeobox genes to which Crx belongs. Interestingly, in one hypomorphic allele of Drosophila otd, otduvi, photoreceptor morphogenesis is also disrupted [45].
We found that there are many other genes that are dependent upon Crx. Those that are expressed at a lower level in the Crx-/- retina, such as rhodopsin and peripherin, comprise many that are either enriched or specific to photoreceptors in their expression [35]. They include enzymes that are important in lipid metabolism, protein folding and transport, as well as in other processes that one might envision would be important in building a structure such as the outer segment. In situ hybridization using probes from this collection of genes has revealed that some of them have their RNA localized to the inner segment, a finding typical for proteins targeted to the outer segment. Future analyses of the function of these genes might reveal their role in outer segment biogenesis.
Finally, polarization of photoreceptors was found to be largely intact, as was ciliogenesis. Another LCA gene, CRB1, and a related gene CRB3, have been implicated in ciliogenesis in in vitro models [46]. The Drosophila homologue of CRB1, Crumbs, has been implicated in photoreceptor morphogenesis [47]. However, the spontaneously occurring mouse mutant in CRB1, the Rd8 mouse, develops shortened outer segments that subsequently degenerate [48], suggesting that photoreceptor polarization and synaptogenesis are intact in this mutant. While CRB1 and Crx have been both linked to LCA, further work is necessary to determine if their function is linked in retinal development.

Synaptogenesis is perturbed in Crx-/- photoreceptors
The Crx-/- mouse demonstrates the most severe abnormality of photoreceptor synapses reported to date. The peripherin-/- mouse develops a normal complement of photoreceptor terminals which then degenerate as the photoreceptors are lost [30]. Also, similarly in rhodopsin (Rho) and cyclic nucleotide-gated channel, alpha-3 (CNGA3) double knockout mice (Rho-/-, CNGA3-/-), synapses are reported to form normally [49]. These observations demonstrate that photoreceptor synaptogenesis can occur in the absence of outer segment formation. In keeping with this observation is the fact that some electroretinogram activity is present in peripherin-/- mice, suggesting that minimal phototransduction is present in these mice, enough to drive activity at the photoreceptor synapse. In vitro studies wherein synapse elements are formed in the absence of proper outer segment development and, therefore, possible absence of light-dependent photoreceptor activity, have indicated the independence of phototransduction and synapse formation, at least for the initial stages [50,51]. These data then suggest that the fact that the Crx-/- photoreceptors do not have proper synaptic endings is not due to a lack of outer segment formation. A more likely explanation is that Crx plays a role in photoreceptor synapse formation, perhaps by regulating directly, or indirectly, important genes in this process. Using immunohistochemistry, we examined the expression of common pre-synaptic terminal proteins, including KIF3a, SV2, and synaptophysin, and were unable to observe qualitative differences between Crx-/- and control tissue at P14 (data not shown). Examination of their RNA levels by SAGE showed no significant difference for all 3 genes, though very few tags were recovered from these genes and thus the analysis of RNA levels may not be significant [35]. However, since other genes expressed in photoreceptors were significantly altered in their expression level in the Crx-/- mouse, there are many candidates that could be important for photoreceptor morphogenesis. Tags from three genes from proteins expressed in photoreceptor terminals were found to be decreased in a statistically significant fashion, namely the HGF-regulated tyrosine kinase substrate, the CRIPT protein, and synaptotagmin 1 (Blackshaw and Cepko, unpublished data). An example of a gene that was increased in the Crx-/- retina is HRG4 (a homologue of the C. elegans Unc119 gene) (Blackshaw and Cepko, unpublished data) which encodes a component of the ribbon synapse [33]. The fact that it is upregulated might indicate a response to the lack of proper terminal structures. Several other genes encoding putative cytoskeletal elements also were increased (e.g. microtubule associated protein 4) or decreased (e.g. cofilin 1) in the Crx-/- retina, with P values of <.005. It is not known whether any of these genes are involved in building or regulating synaptic structures, but they are now genes that might lead to a better understanding of the construction and function of the relatively unique structure of the ribbon synapse.
Abnormal photoreceptor terminal formation was noted in a study that examined retinal development in the laminin beta2 chain-deficient mouse [52]. Several pathologies were noted in these mice. First, laminin beta2 chain-deficient mice displayed abnormal outer segment elongation, but a more mild phenotype than that of the Crx-/- mice; the outer segments were reduced by 50% in length. Also photoreceptor terminals were perturbed in laminin beta2 mutants, but again the phenotype was more subtle then that of Crx-/- mice. The outer plexiform layer of the beta2-deficient retinas demonstrated only 7% normal invaginating synapses, while the remainder had various pathologies, including floating synaptic ribbons, as seen here. The mechanistic relationship of these two molecules, if any, in photoreceptor morphogenesis is unknown to date. The mRNA for laminin beta2 was not detected in the SAGE study of the relative RNA levels in Crx-/- and Crx+/+ and thus we cannot comment on whether the levels of RNA for laminin beta2 were altered.

Crx-/- mice are a model for LCA
Crx has been implicated in three photoreceptor diseases that result in human blindness, cone-rod dystrophy2, Leber's congenital amaurosis, and retinitis pigmentosa (for review, see [53]). The cone-rod dystrophies (CRDs) are characterized by loss of cone-mediated vision in the first decade of life or later, with concomitant or subsequent loss of rod-mediated vision [54]. Conversely, RP is notable for initial loss of rod function, followed by loss of cone-mediated vision [55]. The majority of known genes responsible for human genetic blindness, encode proteins expressed almost exclusively, or exclusively, in photoreceptors, particularly in the outer segment [35]. Many of these proteins are required for phototransduction or outer segment structure. The mechanisms whereby mutations in rod-specific genes, such as rhodopsin, lead eventually to cone degeneration in RP remain obscure. Mutations in Crx were the first, and still one of a very few examples of a transcription factor mutation leading to photoreceptor disease.
LCA is a disease in which there is little or no photoreceptor function in infancy; thereby, likely developmental in etiology ([17,56] for review). The Crx-/- mouse may be an excellent model for studying the pathology of this disorder, particularly the subtype of the disorder where Crx mutations are involved. The vast majority of histopathological studies of LCA in human tissue have been derived from adult patients with LCA where secondary changes are likely to be present. Indeed in animal models of LCA, secondary reactive and/or degenerative changes occur early after the abnormal formation of retinal tissue [57]. The only study in human tissue derived from a human 33-week retina with proposed RPE65 mutations was reported to have abnormal retinae at this early stage [24]. These authors report cell loss, including thinning of the photoreceptor layer. In addition, they claim in the text to have seen aberrant synaptic and inner retinal organization, although their examination of photoreceptor synapses unfortunately are not presented in the data section of the paper. Given the scarcity of available human tissue, the characterization of the primary pathology of LCA will require animal models. In the current study, we present data that argue that, in addition to outer segment morphogenesis, synaptogenesis may also be critically impaired in at least a subset of LCA.