Tissue-specific requirements for specific domains in the FERM protein Moe/Epb4.1l5 during early zebrafish development

Background The FERM domain containing protein Mosaic Eyes (Moe) interacts with Crumbs proteins, which are important regulators of apical identity and size. In zebrafish, loss-of-function mutations in moe result in defects in brain ventricle formation, retinal pigmented epithelium and neural retinal development, pericardial edema, and tail curvature. In humans and mice, there are two major alternately spliced isoforms of the Moe orthologue, Erythrocyte Protein Band 4.1-Like 5 (Epb4.1l5), which we have named Epb4.1l5long and Epb4.1l5short, that differ after the FERM domain. Interestingly, Moe and both Epb4.1l5 isoforms have a putative C' terminal Type-I PDZ-Binding Domain (PBD). We previously showed that the N' terminal FERM domain in Moe directly mediates interactions with Crumbs proteins and Nagie oko (Nok) in zebrafish, but the function of the C'terminal half of Moe/Epb4.1l5 has not yet been examined. Results To define functionally important domains in zebrafish Moe and murine Epb4.1l5, we tested whether injection of mRNAs encoding these proteins could rescue defects in zebrafish moe- embryos. Injection of either moe or epb4.1l5long mRNA, but not epb4.1l5short mRNA, could rescue moe- embryonic defects. We also tested whether mRNA encoding C' terminal truncations of Epb4.1l5long or chimeric constructs with reciprocal swaps of the isoform-specific PBDs could rescue moe- defects. We found that injection of the Epb4.1l5short chimera (Epb4.1l5short+long_PBD), containing the PBD from Epb4.1l5long, could rescue retinal and RPE defects in moe- mutants, but not brain ventricle formation. Injection of the Epb4.1l5long chimera (Epb4.1l5long+short_PBD), containing the PBD from Epb4.1l5short, rescued retinal defects, and to a large extent rescued RPE integrity. The only construct that caused a dominant phenotype in wild-type embryos, was Epb4.1l5long+short_PBD, which caused brain ventricle defects and edema that were similar to those observed in moe- mutants. Lastly, the morphology of rod photoreceptors in moe- mutants where embryonic defects were rescued by moe or epb4.1l5long mRNA injection is abnormal and their outer segments are larger than normal. Conclusion Taken together, the data reveal tissue specificity for the function of the PBD in Epb4.1l5long, and suggest that additional C' terminal sequences are important for zebrafish retinal development. Additionally, our data provide further evidence that Moe is a negative regulator of rod outer segment size.


Background
The mechanisms underlying the acquisition and maintenance of apical cell polarity are beginning to be understood and the importance of cell polarity in development is now widely appreciated. Drosophila Crumbs (Crb) and vertebrate Crumbs orthologues are important determinants of apical polarity and are critical for epithelial morphology [1][2][3][4]. The establishment of cell polarity within the developing retinal neuroepithelium is crucial for normal retinal development, as zebrafish with loss-of-function mutations in the polarity determinants aPKC /heart and soul (has), pals1/mpp5/nagie oko (nok), crb2a/oko meduzy (ome), and mosaic eyes (moe), fail to properly form cell-specific laminae [4][5][6][7][8][9]. In addition, ablation of aPKC in differentiating photoreceptors in a conditional knockout mouse results in a loss of retinal lamination [10].
The moe mutant was discovered in a zebrafish mutagenesis screen, and the moe mutations affect retinal lamination, brain ventricle formation, and heart and body morphology [7,22]. Orthologues have been identified in Drosophila (Yurt) and mammals (Erythrocyte Protein Band 4.1-Like 5, Epb4.1l5) [22,23]. The yurt and epb4.1l5 locus encode four and two isoforms respectively [20,24]. We and our colleagues have shown that Moe and Moe orthologues form a complex with Crumbs proteins that is mediated by the FERM domain, and this interaction is important for Crumbs protein function [20,21]. The mouse mutant lulu has a null-allele mutation in epb4.1l5, and has defects in the epithelial-mesenchymal transition in cells at the primitive streak and abnormal neural plate morphology that is accompanied by defects in the actincytoskeleton [24]. In this study we use a comparative genomic and proteomic approach to identify functionally important sequences within Moe and Epb4.1l5 by testing whether injection of mRNA encoding the long and short isoforms of Epb4.1l5 (Epb4.1l5 long and Epb4.1l5 short ) can functionally substitute for moe function in zebrafish. We further investigate the role of Moe within different tissues by defining what Epb4.1l5 domains are necessary to rescue distinct moedefects. Lastly, we report the histological and morphological consequences of losing Epb4.1l5 long protein in the rescued zebrafish retina after the depletion of rescue construct.

Injection of moe mRNA rescues embryonic defects in moemutants
We tested whether injecting moe mRNA into moeembryos could rescue embryonic and early larval defects. We found that injection of wild-type moe mRNA into moemutant embryos at the 1-4 cell-stage rescued brain ventricle formation, retinal pigmented epithelial (RPE) integrity, and retinal neural epithelial integrity and straightened the tail at 60 hours post fertilization (hpf) (Figure 1). We observed no abnormalities in injected wild-type embryos (or larvae). Pericardial edema in moemutants was only partially rescued by moe mRNA injection ( Figure 1D, F, arrow). The remaining pericardial edema was a convenient marker, however, and made it possible to easily distinguish between wild-type larvae and rescued moemutants, but we also confirmed that embryos were moemutants by labeling with anti-Moe antibodies at 60 hpf and comparing labeling to wildtypes and uninjected moemutants ( Figure 1G, H, I). In moe mRNA injected moemutants very little anti-Moe labeling was observed except background ( Figure 1I, double arrowheads). The weak anti-Moe labeling in moe mRNA injected moemutants suggests very little Moe protein encoded by the injected mRNA remains at 60 hpf.
We have shown that Moe interacts with Crumbs proteins, which are important apical polarity determinants and that moe loss-of-function results in a failure to localize Crb2a and the junctional protein ZO-1 at the apical surface of the retina and brain [21,22]. We examined whether injection of moe mRNA into moemutants could rescue the apical localization of Crumbs proteins and ZO-1 in the retina and brain. In order to examine Crumbs proteins in zebrafish, we used an antibody we raised against the highly conserved C' terminal peptide and because this antibody recognizes all zebrafish Crumbs proteins by western blot (data not shown) we call this antibody a pan-Crb antibody. In wild-type embryos at 60 hpf, anti-pan-Crb and anti-ZO-1 labeling localize to the apical/ventricle surface in the brain, and the apical surface and the newly forming outer limiting membrane in the retina ( Figure 1J, M, N). In moemutants, brain ventricles fail to form properly, and panCrb and ZO-1 fail to localize to the apical surfaces in the brain and retina ( Figure 1K, O, P). Injection of moe mRNA into moemutants leads to the apical relocalization of Crumbs proteins and ZO-1 in the retina and brain ( Figure 1L, Q, R).

Conservation between Moe and mouse Epb4.1l5
To help identify functionally important domains in the Moe and Epb4.1l5 proteins, we first compared their sequences ( Figure 2). The mammalian epb4.1l5 locus encodes two major splice isoforms that are represented by ESTs in both the human and mouse databases, which we term Epb4.1l5 short and Epb4.1l5 long . We provide the exon/ intron structure of the mouse epb4.1l5 locus that has 25 exons: Epb4.1l5 short is encoded by exons 1-16 and Epb4.1l5 long by exons 1-15 and 17-25 ( Figure 2A). We have not found a zebrafish transcript that encodes a protein similar to Epb4.1l5 short .
A comparison between Moe and mouse Epb4.1l5 long shows very strong homology in the FERM domain (88% identity), and there is also strong conservation flanking the FERM domain (44 amino acids preceding the FERM domain and about 35 amino acids after) as well as additional islands of strong conservation, notably a class I PDZ-binding domain (PBD) at the C' terminus of both proteins ( Figure 2B; [24,25]). Homology between Moe and Epb4.1l5 short ends at amino acid 444 ( Figure 2C, blue arrow), after which Epb4.1l5 short has 60 unique amino acids ( Figure 2C). Epb4.1l5 short is predicted to be 56 kDa, and interestingly also has a predicted binding motif for a class I PDZ domain at its C' terminus ( Figure 2C).
We raised isoform-specific antibodies against the unique C' terminal sequences of the long and short isoform of Epb4.1l5 and used them for western analysis of mouse tissue. We observed two bands that were immunoreactive with Epb4.15 long anti-sera. A protein that migrated at approximately 100 kDa was present in eye, brain, heart, lung, kidney, and testis tissue ( Figure 2D). There was an additional protein recognized in all tissues at 75 kDa, which is probably non-specific reactivity. Two proteins were detected with anti-Epb4.1l5 short affinity-purified antibody. One protein migrated at the expected molecular weight of 56 kDa and was present in brain, liver, lung, kidney, pancreas, and gut. A second protein migrating at approximately 75 kDa was broadly expressed. This higher Injection of moe mRNA rescues defects in moeembryos Figure 1 Injection of moe mRNA rescues defects in moeembryos. (A-B) At 60 hpf, in wild-type embryos, the floor of the diencephalic ventricle is visible (A, white arrow head), the RPE is uniform, and the body axis is straight. (C, D) In moeembryos, the ventricles are small or absent, the RPE is patchy, the tail curves and there is pericardial edema (D, arrow). (E, F) In moeembryos injected with moe mRNA, the floor of the diencephalic ventricle is visible (E, white arrow head), the RPE is uniform, and the body axis is straight but mild pericardial edema persists (F, arrow). Anti-Moe labeling of 60 hpf wild-type embryos (G), moeembryos (H), and moeembryos injected with moe mRNA (I): the plexiform labeling in moeembryos injected with moe mRNA (I, double arrowheads) is largely background. Adherens junctions (ZO-1, green) and panCrb labeling (red) are apically localized at the retina (arrow) and brain ventricle surface (asterix) in wild-type (J) and moeembryos injected with moe mRNA (L), but are ectopically localized within the developing eye (arrow) and at the presumptive brain midline in moemutants with abnormal ventricle formation (asterix, K). High magnification confocal z-projections of TO-PRO-3 nuclear staining, and ZO-1 and panCrb labeling in the retina and brain in wild-type (M, N), moe -(O, P), and moeembryos injected with moe mRNA (Q, R) at 60 hpf. (G, J, H, K, I and L are all single confocal z-sections). Scale bars, 50 μm (G-L), 10 μm (M-R).
Genomic structure of the mouse epb4.1l5 locus and expression of its two major splice isoforms Figure 2 Genomic structure of the mouse epb4.1l5 locus and expression of its two major splice isoforms. (A) Diagram of the inton/exon structure of the Mus musculus epb4.1l5 locus. Exons that are common to both isoforms are black, the exons unique to epb4.1l5 long are indicated in red and the unique exon in epb4.1l5 short is indicated in green. Bars represent 1 kb and 10 kb scales for exon and intron lengths, respectively. (B) ClustalX alignment of mouse Epb4.1l5 long and zebrafish Moe. Ymo1 long and Moe share a high degree of homology within the FERM domain (black). Ymo1 long and Ymo1 short are identical up to Lysine 444 (blue arrow) and then alternately spliced into the long (red) and short (green) isoforms. Moe and Epb4.1l5 long , and Epb4.1l5 short have predicted C'terminal PDZ-binding domains (Pink [TTEL]) and (light green [MTEI]). (*) identical, (:) highly conserved, (.) moderately conserved. (C) Western analysis of mouse tissues with antibodies raised against the unique C'terminal sequences of Epb4.1l5 long and Epb4.1l5 short . Two bands are immunoreactive with the anti-Epb4.1l5 long antibody, one migrates at the expected molecular weight of Epb4.1l5 long (100 kDa) and is present in the eye, brain, heart, lung, kidney, and testis. An additional band migrates at 75 kDa, which is probably non-specific. Two bands are recognized by the affinity purified Epb4.1l5 short antibody. A lower band migrates at the predicted molecular weight of 56 kDa and is present in brain, liver, lung, kidney, pancreas, and gut. A second band with a broad expression pattern migrates at approximately 75 kDa. Anti-α-Tubulin was used as a loading control. molecular weight protein may represent a post-translationally modified form of Epb4.1l5 short or more likely is non-specific reactivity ( Figure 2D).

Functional comparative genomics reveals important domains in Moe/Epb4.1l5
To identify functionally important sequences in the Moe and orthologous protein Epb41.l5, we tested whether injection of mouse epb4.1l5 long mRNA could substitute for moe and rescue moemutant defects. We injected epb4.1l5 long mRNA into 1-4 cell moeembryos and found that it rescued brain ventricle formation, retinal pigmented epithelial integrity, and retinal lamination and straightened the tail like injection of moe mRNA (Table 1 and data not shown). Injection of epb4.1l5 long also rescued apical localization of ZO-1 and anti-panCrb labeling in Both the PDZ-binding domain and unique sequences in Epb4.1l5 long required for rescue of moemutant defects Figure 3 Both the PDZ-binding domain and unique sequences in Epb4.1l5 long required for rescue of moemutant defects.
(A-C) At 30 hpf in wild-type embryos, Moe localizes cortically in brain and retinal neuroepithelial cells and is concentrated at the apical surface (A) and ZO-1 (green) and panCrb (red) localize to the apical surface of the retina (B) and brain (C). (D-F) At 30 hpf in moeembryos, there is no Moe labeling (D) and ZO-1 (green) and panCrb (red) fail to localize to the apical surface of the retina (E) and brain (F). (G) At 30 hpf in moeembryos injected with epb4.1l5 long mRNA, Epb4.1l5 long immunoreactivity is cortically localized in most retinal and brain neuroepithelial cells and ZO-1 (green) and panCrb (red) localize to the apical surface in the retina (H) and brain (I). Upper inset (G), magnified section of anti-Epb4.1l5 long labeling and lower inset, uninjected moeembryos shows no labeling with anti-Epb4.1l5 long . (J) At 30 hpf in moemutants injected with epb4.1l5 short mRNA, anti-Epb4.1l5 short is cytoplasmically localized and ZO-1 (green) and panCrb (red) do not localize to the apical surface in moeretina (K) and brain (L). Upper inset (G), magnified section of anti-Epb4.1l5 short labeling and lower inset, uninjected moeembryos shows no labeling with anti-Epb4.1l5 short . (M, N) At 30 hpf in moemutants injected with myc-epb4.1l5 FERM mRNA, ZO-1 (green) and panCrb (red) do not localize to the apical surface in moeretina (M) and brain (N). (O, P) At 30 hpf in moemutants injected with myc-epb4.1l5 long_ PBD mRNA, ZO-1 (green) and panCrb (red) do not localize to the apical surface in moeretina (O) and brain (P). (Q, R) At 30 hpf in moemutants injected with myc-epb4.1l5 short+long_PBD mRNA, ZO-1 (green) and panCrb (red) localize to the apical surface in moeretina (Q) and brain (R). (S, T) At 30 hpf in moemutants injected with myc-epb4.1l5 long+short_PBD mRNA, ZO-1 (green) and panCrb (red) localize to the apical surface in moeretina (S) and brain (T). (U) Western analysis of zebrafish embryos injected with epb4.1l5 long and epb4.1l5 short mRNA were tested for expression of protein product with isoform-specific antibodies at time points from 6 hpf to 96 hpf. (V) Western analysis of zebrafish embryos injected with epb4.1l5 FERM (anti-Myc), epb4.1l5 long_ PBD (anti-Epb4.1l5 long ), epb4.1l5 short+long_PBD (anti-Myc), epb4.1l5 long+short_PBD (anti-Epb4.1l5 long ). Blots stripped and reprobed with Anti-α-Tubulin as a loading control. Scale bars, 10 μm (A, D, G, J), 50 μm (lower insets in G, J), 10 μm (remaining panels). (A-T), single confocal z-sections. the retina and brain ( Figure 3H, I). Using antibodies we raised against the unique sequence in Epb4.1l5 long , we found that Epb4.1l5 long protein in mRNA injected moemutants localized cortically like endogenous Moe in wildtype embryos ( Figure 3A, G). Because all the proteins shown to interact with Moe do so through its FERM domain [21], we tested whether injection of epb4.1l5 mRNA encoding a myc-tagged FERM domain (amino acids 1-346, Epb4.1l5 FERM ) could rescue moemutant defects like full length moe and epb4.1l5 long mRNA injection. Injection of epb4.1l5 FERM mRNA at the 1-4 cell stage failed to rescue the defects in moemutants (data not shown) and did not lead to apical relocalization of ZO-1 and anti-panCrb labeling ( Figure 3M, N). We also ruled out the possibility that the N' terminal myc-tag interfered with protein function by showing that Epb4.1l5 long myctagged at its N' terminus was still able to rescue moemutant defects as well as untagged Epb4.1l5 long (data not shown) We next tested whether injection of mRNA encoding Epb4.1l5 short could rescue moemutant defects since it is identical to Epb4.1l5 long until amino acid 444 after which there are ~60 unique amino acids that end with a predicted PDZ binding domain that is the same class as that in Moe and Epb4.1l5 long . Injection of mRNA encoding Epb4.1l5 short into moemutants at the 1-4 cell stage failed to rescue moemutant embryonic defects and failed to apically relocalize ZO-1 and anti-panCrb labeling in the retina or brain ( Figure 3K, L). Epb4.1l5 short labeling with the Epb4.1l5 short specific antibodies appears more cytoplasmic in injected moemutants ( Figure 3J). Because many proteins with PDZ domains have been implicated in the establishment of cell polarity, we asked whether the PDZ-binding domain in Epb4.1l5 long is necessary for its function. We injected mRNA constructs encoding Epb4.1l5 long with the PBD deleted (Epb4.1l5 long_ΔPBD ), Epb4.1l5 long where its PBD is replaced by PBD from Epb4.1l5 short (Epb4.1l5 long+short_PBD ) and Epb4.1l5 short where its PBD is replaced by the PBD from Epb4.1l5 long (Epb4.1l5 short+long_PBD ). Injection of epb4.1l5 long_ PBD mRNA failed to rescue Crb2a and ZO-1 apical localization in the retina or rescue brain ventricle formation ( Figure  3O, P). However, injection of epb4.1l5 long+short_PBD or epb4.1l5 short+long_PBD mRNAs did lead to apical relocalization of ZO-1 and anti-panCrb labeling ( Figure 3Q-T).
We confirmed that protein was expressed from each of the template rescue mRNAs by western analysis of injected zebrafish. Epb4.1l5 long and Epb4.15 short were not detectable at time points beyond 72 and 48 hpf respectively (Figure  3U). Both Epb4.1l5 long+short_PBD and Epb4.1l5 short+long_PBD were expressed at 6 hpf and faint signal was visible at 24 hpf. Neither Epb4.1l5 FERM nor Epb4.1l5 long_ΔPBD were detectable at time points beyond 6 hpf ( Figure 3V), therefore a failure to rescue any moedefects in these cases may be due to a rapid loss of the protein generated by rescue constructs.

Analysis of rescue with chimeric Epb4.1l5 PBD isoforms
At 60 hpf, mutant moeembryos exhibit reduced or absent brain ventricles, pericardial edema, and RPE defects (Figure 4A, B). Injection of epb4.1l5 short+long_PBD mRNA restored RPE integrity ( Figure 4C) and retinal lamination (data not shown) in moemutants, but did not rescue the edema, brain ventricles were small or absent, and the tail curved ( Figure 4C, data not shown).
In a heterozygous moe +/incross, a roughly Mendelian inheritance (20/88; 23%) of individuals injected with epb4.1l5 long+short_PBD mRNA exhibited RPE defects; however, in all but one case, those defects were minor com- Wild-type embryos or embryos from heterozygous moe +/incrosses were injected with the indicated mRNA constructs and scored based on pericardial edema, brain ventricle defects, RPE defects, apical localization of ZO-1 or Crb2a at 30 hpf, and retinal lamination at 5 dpf. Phenotypes that were not analyzed are indicated with an NA. * See Figure 4M, N and text.
Morphological rescue of some moeembryonic defects by injection of the epb4.1l5 constructs Figure 4 Morphological rescue of some moeembryonic defects by injection of the epb4.1l5 constructs. (A) In wild-type embryos at 60 hpf, brain ventricles are visible, and the RPE is uniform (inset). (B) In moeembryos, brain ventricles are reduced in size or absent, pericaridal edema is pronounced, and the RPE is patchy (inset). (C) In moeembryos injected with epb4.1l5 short+long_PBD mRNA, brain ventricles are reduced in size or absent, pericaridal edema is pronounced, but the RPE is normal (inset). (D) moeembryos injected with epb4.1l5 long+short_PBD mRNA, brain ventricles are absent or absent and pericardial edema is pronounced, and RPE defects are milder than those in uninjected moeembryos. (E) A magnified view of the RPE of a 60 hpf wild-type embryo shows that it is uniform and the cells are confluent. (F) In a wild-type retina at 4 dpf, GFP + rods localize next to the RPE and lamination is apparent. In 60 hpf moemutants, the integrity of the RPE varies from mild (G), to moderate (I), to severe (H). However, GFP + rods are mislocalized in all moemutants 4 dpf (H, J, L). The integrity of the RPE is improved and nearly normal in a 60 hpf moemutant injected with epb4.1l5 long+short_PBD mRNA (M) and most GFP + rods are adjacent to the RPE (N). (O) A wild-type embryo injected with epb4.1l5 long+short_PBD showing brain ventricles that are reduced or absent. (P) At 30 hpf Epb4.1l5 long+short_PBD is cortically localized, upper inset is a 2× magnification of Epb4.1l5 long+short_PBD localization. Scale bars, 10 μm (F), 50 μm (lower insets in F).
pared to uninjected moemutants and their detection required very careful examination ( Table 1). The severity of RPE defects varies in uninjected moemutants: an examination of 35 uninjected moemutants showed that 8 had mild RPE defects (23%, Figure 4G), 26 had moderate RPE defects (74%, Figure 4I), and 1 had severe RPE defects (3%, Figure 4K). The shift in severity of RPE defects from mostly moderate in moemutants to nearly normal in injected moemutants, suggests that injection of epb4.1l5 long+short_PBD mRNA largely restores RPE integrity.
Because Epb4.1l5 long+short_PBD largely restored RPE integrity in moemutants we examined whether retinal lamination was also restored in these individuals. At 4 dpf, in wildtypes, GPF + rods are localized adjacent to the RPE ( Figure 4F). In moemutants, GPF + rods are ectopically localized throughout the retina, regardless of the severity of RPE defects ( Figure 4H, J, L). In moemutants injected with of epb4.1l5 long+short_PBD mRNA, most GPF + rods localize normally and are adjacent to the RPE. ( Figure 4N).
When we injected epb4.1l5 long+short_PBD mRNA into embryos resulting from an incross of moe -/+ individuals, we observed that more than 25% exhibited edema, suggesting that injection of epb4.1l5 long+short_PBD mRNA has a tissue-dependent dominant negative effect. We determined that 61% of embryos had pericardial edema and 96% had small or missing brain ventricles (Table 1). We next injected epb4.1l5 long+short_PBD mRNA into embryos from a wildtype incross and found a large proportion of these individuals exhibited pericardial edema and brain ventricle defects, but the RPE was normal, confirming that the dominant negative effects of Epb4.1l5 long+short_PBD are limited to edema and brain ventricle formation (Table 1 and Figure 4O). Lastly, Epb4.1l5 long+short_PBD retains immunoreactivity with the anti-Epb4.1l5 long sera, and we show that the protein localizes cortically like Epb4.1l5 long ( Figure 4F), suggesting that the Epb4.1l5 long PBD is not required for cortical localization.

Early Epb4.1l5 function rescues later retinal lamination and function
Because mRNA injection could rescue early moemutant defects, we investigated whether later defects of retinal development were also rescued, in particular, whether differentiated cells acquired their correct laminar position and photoreceptors their normal morphology. We compared the retinas of 6 dpf wildtypes, moemutants, and moemutants injected with moe or epb4.1l5 long mRNA (Figure 5, data not shown). The western blot experiments and immunohistochemistry (data not shown) showed that there is very little, if any, remaining Epb4.1l5 long protein after 3 dpf ( Figure 3U), the time at which most photoreceptors begin to undergo morphogenesis. In wild-type ret-inas, nuclei are arranged in distinct layers, Müller glia are radially oriented and project to and contribute to the inner (basal) and outer (apical) limiting membranes of the retina, and rods and double cones display a polarized morphology with their outer segments projecting into the RPE (Figure 5A, D, G). In moemutants, nuclei do not form distinct layers and the numbers of Müller glia, rods and double cones are reduced and their morphology is abnormal, although interestingly rods do make outer segments ( Figure 5B, E, H).
Whereas Müller glial morphology and retinal lamination are rescued in moemutants by injection of either moe or epb4.15 long mRNA, the morphology of photoreceptors (rods and double cones) is not; instead of standing perpendicular to the normal RPE (Fig. 5A, D, G and Fig. 6A), those in moemutants injected with either moe or epb4.15 long mRNA, lie collapsed in a twisted heap adjacent to the RPE (Fig. 5C, F, I and Fig. 6C, and data not shown). The failure to rescue photoreceptor morphology is likely because Moe or Epb4.15 long protein from injected mRNA is lost by the time photoreceptors undergo morphogenesis.
Previously, we showed in genetic mosaics that the outer segments of rods lacking moe function are larger than those in wild-type rods [21]. We sought to determine whether outer segments are also larger in rods in larvae where all cells have lost moe function by about 3 dpf. We measured the size of rod outer segments using anti-Rhodopsin labeling. We found that rods in rescued moemutants (i.e. injected with epb4.1l5 long mRNA) were significantly larger than rods in wild-type retinas at 6 dpf, whereas, rods in moemutants were significantly smaller than those in wild-type retinas ( Figure 5J).
We also tested whether injection of epb4.1l5 long mRNA could restore vision to moemutants as measure by the optokinetic response (OKR). The OKR measures the tracking of the eyes to a moving stimulus [26]. In our study the stimulus consisted of alternating white and black vertical bars moving to the right on a projection screen and we measured tracked eye movements (TM) in response to the stimulus over a minute time period. We measured TM in wild-type, moemutants, and epb4.1l5 long mRNA injected moemutant larvae immobilized in methylcellulose at 5 dpf. We found that wild-type larvae exhibited an average of 7.8 (+/-0.6) TM/minute, moemutants completely lacked TM, and epb4.1l5 mRNA injected moemutant larvae had an average of 3.6 (+/-0.7) TM/minute ( Figure  5K). Thus, even though photoreceptors have morphological defects in epb4.1l5 long mRNA injection into moemutants many of these larvae have functional vision as measured by the optokinetic response.

Crb2a/b protein localization and outer limiting membrane integrity requires moe function, but rod outer segment disc stacking does not
We recently showed that Moe is an important regulator of Crumbs protein localization in the embryo [21], so we were interested in whether moe function is also required at later stages in the retina for Crumbs protein localization.
Crb2a and Crb2b are the only Crumbs proteins shown to be expressed by zebrafish photoreceptors [4,21] and the antibodies we use recognize both proteins ( [21], and data not shown). The ability to rescue the embryonic defects and retinal lamination with moe or epb4.1l5 long injection allows us to ask whether the localization of Crb2a/b in photoreceptors also requires moe function. The western Injection of epb4.1l5 long mRNA into moemutants restores retinal lamination but not normal photoreceptor morphology at 6 dpf blot analysis indicates very little, if any, Epb4.1l5 long protein remains after 3 dpf in moe -mRNA injected individuals, so we can examine retinas in moemutant larvae where early defects have been rescued but where there is no endogenous Moe and little, if any, exogenous Epb4.1l5 long protein after 3 dpf. We examined the localization of Crb2a/b in wild-type, moemutant, and epb4.1l5 long mRNA injected moemutant larvae at 6 dpf. In the wild-type retina, Crb2a/b localizes just apical to the outer limiting membrane (OLM), which is labeled by anti-ZO-1 antibodies ( Figure 6A). In moemutants, very little Crb2a/b protein is detected in the area surrounding GFP + rod photoreceptors, ZO-1 labeling is highly disorganized suggest-ing the OLM has not formed, and there is no spatial relationship between Crb2a/b and ZO-1 ( Figure 6B). In epb4.1l5 long mRNA injected moemutants, Crb2a/b labeling is evident, but reduced compared to wild-type retinas, and like moemutants, ZO-1 labeling is disorganized indicating the absence of the OLM, and we also observed no spatial relationship between Crb2a/b and ZO-1 ( Figure  6C).
Although labeling of rods with anti-Rhodopsin antibodies in moemutant and epb4.1l5 long mRNA injected moemutant larvae suggest that outer segments form ( Figure  4D, F, G, I), we wanted to determine whether these outer The outer limiting membrane (OLM) is not restored in moemutants by injection of epb4.1l5 long mRNA segments were normal ultrastructurally, in particular, whether disk morphology and packing is normal. We examined outer segments by transmission electron microscopy (TEM) at 6 dpf in wild-type, moemutant and epb4.1l5 long mRNA injected moemutant larvae. We found that disk morphology and packing appeared relatively normal in photoreceptors in moemutants and epb4.1l5 long mRNA injected moemutants ( Figure 6D-F).

Discussion
In this study we sought to identify functionally important domains in the orthologous FERM proteins, zebrafish Moe and mouse Epb4.1l5. Our strategy was to use evolution, comparative genomics, and protein engineering to discover regions and sequences necessary for rescue of embryonic and early larval defects in moe deficient zebrafish. We first established that injection of wild-type moe mRNA into moeembryos rescued all embryonic defects with the exception that mild pericardial edema persisted. In mammalian EST databases there are two major splice isoforms of Epb4.1l5, Epb4.1l5 long that is 731 amino acids in length and is similar in length and shares sequence identity with Moe beyond the FERM domain ( Figure 1) and Epb4.1l5 short that is 504 amino acids. These two isoforms are identical until amino acid 444 and interestingly both contain a predicted PDZ-binding domain at their C' terminus. We raised isoform-specific antibodies and examined expression of the two orthologues by western blot of different mouse tissues. Both antibodies recognized a protein of about 75 kDa in all tissues, which is likely to be non-specific reactivity. Anti-Epb4.1l5 long recognized a protein of the expected molecular weight of about 100 kDa in eye, brain, heart, lung, kidney and testis; this expression profile agrees well with recently published immunohistochemical and mRNA expression (in situ hybridization) data in mammalian retinal, brain, and kidney tissues [24,27]. Anti-Epb4.1l5 short recognized protein of the expected molecular weight of about 56 kDa in brain, liver, lung, kidney, pancreas, spleen and gut. Not all tissues expressed both isoforms, for instance Epb4.1l5 long but not Epb4.1l5 short is found in the eye and Epb4.1l5 short but not Epb4.1l5 long is found in the gut, suggesting that the two isoforms may have non-overlapping functions.
We tested whether either of these Epb4.1l5 isoforms could functionally substitute for Moe during embryonic and early larval development. Injection of mRNA encoding Epb4.1l5 long into moemutants rescues moemutant defects and leads to the restoration of retinal lamination, RPE integrity, normal brain ventricle morphology, and the apical localization of Crumbs proteins and ZO-1 in the developing retina and brain, similar to injection of moe mRNA. Like injection of moe mRNA into moemutants, mild pericardial edema persisted in these epb4.1l5 long mRNA injected moemutants. We found that injection of epb4.1l5 short mRNA into moemutants failed to rescue any phenotypic defects and also failed to relocalize Crumbs proteins and ZO-1 to the apical surface of the retina and brain. Interestingly, while exogenous Epb4.1l5 long protein is cortically localized in neuroepithelial cells similar to endogenous Moe protein, Epb4.1l5 short is localized cytoplasmically, suggesting that either the PBDs or sequences unique to Epb4.1l5 long underlie Epb4.1l5 long protein localization.
Although edema was reduced in moemutants by injection of moe or epb4.1l5 long mRNA, it was never abolished by mRNA injection. We presume that the edema in moemutants is caused by kidney dysfunction. There are several possible reasons for failure of mRNA injection to rescue kidney function. It is possible that Moe/Epb4.1l5 long function is required for a longer period of time or perhaps continually in the kidney and protein from injected mRNA is not around long enough completely restore kidney function. A second possibility is that the Moe and Epb4.1l5 long constructs we used lack the sequence(s) needed to rescue kidney function. If this is the case, then those specific sequence(s) do not seem to reside in Epb4.1l5 short , since injection of epb4.1l5 short mRNA failed to rescue any pericardial edema, and the severity of edema was as severe as uninjected moemutants. There are, however, many minor splice variants of Moe and Epb4.1l5 long in the zebrafish and mammalian EST databases.
Because all the proteins so far identified that interact with Moe and Epb4.1l5 (also known as Ymo1) do so via the FERM domain [20,21], we tested whether expression of a Epb4.1l5 construct encoding the first 346 amino acids (Epb4.1l5 FERM ), which includes the FERM domain, could rescue apical Crumbs proteins and ZO-1 localization or any defects in moemutants. This construct failed to rescue apical Crumbs protein and ZO-1 localization and any moedefects, however, we could not detect Epb4.1l5 FERM protein after 6 hpf, suggesting that this mRNA or protein is unstable. The same result was observed when the PDZbinding domain in Epb4.1l5 long was deleted (Epb4.1l5 long_ΔPBD ); there was no phenotypic rescue in moemutants and no apical Crumbs protein and ZO-1, and we did not detect Epb4.1l5 long_ΔPBD protein after 6 hpf. These observations suggest that the PDZ-binding domain might be important for stability of Moe/Epb4.1l5 protein.
PDZ domains are important mediators of protein interactions and have been shown to be important during the establishment and maintenance of cell polarity [28,29]. We sought to identify the importance of the PDZ-binding domain in Epb4.1l5 long by replacing it with the PDZdomain from Epb4.1l5 short to generate the chimeric protein Epb4.1l5 long+short_PBD and by replacing the PDZ-domain in Epb4.1l5 short with the PDZ-domain from Epb4.1l5 long to generate the Epb4.1l5 short+long_PBD chimera. We found that injection of epb4.1l5 long+short_PBD or epb4.1l5 short+long_PBD mRNA into moemutants rescued apical localization of Crumbs proteins and ZO-1 in the retinal and brain neuroepithelium, RPE integrity (epb4.1l5 long+short_PBD almost complete rescue), and retinal lamination, but did not rescue brain ventricle morphology. Furthermore, injection of epb4.1l5 long+short_PBD mRNA into wild-type embryos caused a dominant negative phenotype-brain ventricles were small or failed to form. Moe and other cell polarity determinants, Crb2a/Ome, aPKCλ, and Nok, are required for proper lamination of the zebrafish retina [4][5][6][7][8]22]. The time at which these proteins are needed for lamination has not been determined.
Since we see very little Moe or Epb4.1l5 long protein after 60 hpf, we suggest that early Moe or Epb4.1l5 long function is sufficient to rescue retinal lamination and function. In moe or epb4.1l5 long injected moemutants at 6 dpf, Müller glial cell processes are properly oriented and span the thickness of the retina, and the retina has distinct nuclear layers. Furthermore, the vast majority of rod and double cone photoreceptors localize correctly to outer most portion of the retina to form an outer nuclear layer.
Immunohistochemical and ultrastructural analysis of Epb4.1l5 long rescued moemutant retinas, revealed that rescued rods form outer segments, but they were not always oriented with their outer segments toward the RPE ( Figure  5D-I, 6D-F). This may be a consequence of the failure of the rescued individuals to establish or maintain the OLM ( Figure 6C). Despite the morphological defects of rods and cones in the moemutants injected with epb4.1l5 long mRNA, many of these larvae are visually competent as tested by optokinetic response. Interestingly, our ultrastructural analysis revealed that moerods formed outer segments, complete with organized membranous discs, suggesting that mechanisms that dictate apical opsin transport and disc formation do not require moe function.
Previously, we showed that Moe function is required for the localization of Crb2a and ZO-1 protein at the apical surface of the developing retina in zebrafish embryos [7,21]. We show here that in the wild type retina at 6 dpf, anti-panCrb labeling localizes just above the OLM in the subapical region. In moephotoreceptors, anti-panCrb labeling is not detectible, and ZO-1 appears disorganized. When we examined photoreceptors in epb4.1l5 long mRNA rescued moe-mutants at 6 dpf, which is several days after detectable Epb4.1l5 long protein, we observed that both ZO-1 and Crumbs proteins are present in the photoreceptor region ( Figure 6C). The design of our rescue experiment allowed us to analyze the localization of Crumbs proteins and ZO-1 in rescued moe-photoreceptors several days after exogenous Epb4.1l5 long was gone. In epb4.1l5 long mRNA injected moe-mutants, ZO-1 and panCrb labeling is not normal and mislocalized ectopic plagues of ZO-1 and panCrb labeling appears to be at the interface of photoreceptors, and/or photoreceptors and Müller glia and there is no clear relationship between ZO-1 and anti-pan-Crb labeling. Thus, Epb4.1l5/More function is required to maintain, or establish, the OLM and the localization of Crumbs proteins relative to it.
We measured the size of rod outer segments in moemutants that had been injected with epb4.1l5 long mRNA but that lack measurable Moe protein during photoreceptor morphogenesis, and found that these genetically moedeficient rods were nearly twice the normal size (362 μm 3 compared to wild-type outer segments 197 μm 3 ). This observation is in agreement with previous data from our lab and others implicating Moe and the Drosophila orthologue, Yurt, as negative regulators of apical membrane size in photoreceptors [20,21]. We observed that in uninjected moe-mutants, rod outer segments are smaller than wild-types (90.9 μm 3 compared to 362 μm 3 ), this could be a consequence of the general ill health of moe mutants at 6 dpf, and/or the isolation of photoreceptors from factors secreted by the RPE and Müller glia [30,31].

Conclusion
Our strategy to use comparative genomics and protein engineering has revealed that the function of Moe/ Epb4.1l5 protein is modular and that particular regions can be assigned particular functions. We also show that the C' terminal domain that encodes the PDZ-binding domain in Moe and Epb4.1l5 long is important but not sufficient to confer full protein function to the FERM domain and that other sequences in Moe and Epb4.1l5 long are important. The next challenge will be to identify the PDZcontaining protein that interacts with the PDZ-binding domain in Moe/Epb4.1l5 long and the additional protein(s) that interact with the unique sequences in Epb4.1l5 long . Although the role of the Crumbs complex in epithelial morphogenesis has been much studied, the molecular mechanism of the Crumbs complex function is still unknown. The identification of additional proteins that interact with Moe/Epb4.1l5 long may help to determine the mechanistic function of the Crumbs complex.

Animals
AB wild-type strain and the moe b781 allele were maintained and staged as described previously [7,32]. For analysis requiring EGFP-expressing rods, the moe b781 allele was crossed into the Tg(Xop:EGFP) transgenic line [33]. To block pigmentation, embryos were treated with 2.5 μg/mL phenylthiourea (PTU) beginning at about 20 hpf.

RNA injections
For mRNA transcription, PCRTopoII or pBSII vectors containing the cDNAs of full-length moe, epb4.1l5 long , epb4.1l5 short , or myc-tagged fusions of the following constructs; epb4.1l5 long , epb4.1l5 FERM (1-346 N-terminal amino acids), epb4.1l51 long_ PBD (epb4.1l5 long with the four C'terminal TTEL deleted), epb4.1l51 short+long_PBD (short Cterminal AAMTEI replaced with LLTTEL) or epb4.1l5 long+short_PBD (long C-terminal LLTTEL replaced with AAMTEI), were linearized by NotI or PspOMI restriction digest, and transcribed with the Sp6 or T7 Message Machine transcription kit (Ambion). Roughly 100-250 pg (amount varied per construct for a final 0.2 fM) of mRNA was injected into yolk of 1-4 cell embryos obtained from an incross of moe b781 or moe b781 /Xop-GFP heterozygotes. For those mRNAs that failed to rescue at the above molarity, we tested whether injecting more res-cued, and in all cases higher amounts failed. Concurrent with immunohistochemical analysis, we probed with anti-Moe on alternate sections to confirm the moegenotype. All constructs were sequenced prior to mRNA synthesis.

Transmission electron microscopy
Larvae were fixed at 6 days in 4% PFA/0.5% gluaraldehyde/0.1M NaPO 4 pH 7.2 overnight at 4°C. Samples were rinsed, incubated in 1% osmium tetroxide in PBS at room temperature, dehydrated in an ethanol series, and then rotated in a 1:1 ratio of 80% ethanol:LR White resin (Electron Microscopy Sciences) for one hour at room temperature, and in a change of the same solution overnight at room temperature. Samples were then equilibrated in 100% LR White 2X for 2 hours, and polymerized. Ultrathin sections (100 nm) were cut then stained with 2% uranyl acetate for 30 min and 0.5% lead citrate for 12 min at room temperature. Sections were visualized on a JOEL100S Transmission Electron Microscope.

Optokinetic response
Briefly, Optokinetic Response (OKR) was measured according to protocol modified from Rinner et al., 2005 [37], at 5 dpf in WT, moe b781 , and epb4.1l5 long mRNA injected moe b781 mutants that were immobilized in a drop of 2% Methyl Cellulose. A dissecting scope was used to visualize Tracked eye Movements (TM) per minute in response to a moving visual stimulus. Cycling vertical lines were generated with Optomotor 1.2 software (f, 1Hz; λ, 120 pxls; speed, 5), and rear projected with an inFocus LCD digital projector on a 180° curved screen approximately 4.5 cm from larvae (Optomotor 1.2 software was generously provided by Harold Burgess).