We tested whether injecting moe mRNA into moe^- embryos could rescue embryonic and early larval defects. We found that injection of wild-type moe mRNA into moe^- mutant 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 moe^- mutants 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 moe^- mutants, but we also confirmed that embryos were moe^- mutants by labeling with anti-Moe antibodies at 60 hpf and comparing labeling to wildtypes and uninjected moe^- mutants (Figure 1G, H, I). In moe mRNA injected moe^- mutants very little anti-Moe labeling was observed except background (Figure 1I, double arrowheads). The weak anti-Moe labeling in moe mRNA injected moe^- mutants 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 2122. We examined whether injection of moe mRNA into moe^- mutants 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 panCrb antibody. In wild-type embryos at 60 hpf, anti-panCrb 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 moe^- mutants, 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 moe^- mutants leads to the apical relocalization of Crumbs proteins and ZO-1 in the retina and brain (Figure 1L, Q, R).

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; 2425). 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 molecular weight protein may represent a post-translationally modified form of Epb4.1l5^short or more likely is non-specific reactivity (Figure 2D).

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 moe^- mutant defects. We injected epb4.1l5^long mRNA into 1–4 cell moe^- embryos 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 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 moe^- mutants localized cortically like endogenous Moe in wild-type 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 moe^- mutant 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 moe^- mutants (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 myc-tagged at its N' terminus was still able to rescue moe^- mutant defects as well as untagged Epb4.1l5^long (data not shown)

We next tested whether injection of mRNA encoding Epb4.1l5^short could rescue moe^- mutant 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 moe^- mutants at the 1–4 cell stage failed to rescue moe^- mutant 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 moe^- mutants (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_ΔPBD 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 moe^- defects in these cases may be due to a rapid loss of the protein generated by rescue constructs.

At 60 hpf, mutant moe^- embryos 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 moe^- mutants, 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 compared to uninjected moe^- mutants and their detection required very careful examination (Table 1). The severity of RPE defects varies in uninjected moe^- mutants: an examination of 35 uninjected moe^- mutants 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 moe^- mutants to nearly normal in injected moe^- mutants, 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 moe^- mutants 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 moe^- mutants, GPF⁺ rods are ectopically localized throughout the retina, regardless of the severity of RPE defects (Figure 4H, J, L). In moe^- mutants 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.

Because mRNA injection could rescue early moe^- mutant 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, moe^- mutants, and moe^- mutants 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 retinas, 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 moe^- mutants, 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 moe^- mutants 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 moe^- mutants 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 moe^- mutants (i.e. injected with epb4.1l5^long mRNA) were significantly larger than rods in wild-type retinas at 6 dpf, whereas, rods in moe^- mutants 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 moe^- mutants 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, moe^- mutants, and epb4.1l5^long mRNA injected moe^- mutant larvae immobilized in methylcellulose at 5 dpf. We found that wild-type larvae exhibited an average of 7.8 (+/- 0.6) TM/minute, moe^- mutants completely lacked TM, and epb4.1l5 mRNA injected moe^- mutant 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 moe^- mutants many of these larvae have functional vision as measured by the optokinetic response.

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 421 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 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 moe^- mutant 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, moe^- mutant, and epb4.1l5^long mRNA injected moe^- mutant 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 moe^- mutants, very little Crb2a/b protein is detected in the area surrounding GFP⁺ rod photoreceptors, ZO-1 labeling is highly disorganized suggesting 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 moe^- mutants, Crb2a/b labeling is evident, but reduced compared to wild-type retinas, and like moe^- mutants, 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 moe^- mutant and epb4.1l5^long mRNA injected moe^- mutant larvae suggest that outer segments form (Figure 4D, F, G, I), we wanted to determine whether these outer 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, moe^- mutant and epb4.1l5^long mRNA injected moe^- mutant larvae. We found that disk morphology and packing appeared relatively normal in photoreceptors in moe^- mutants and epb4.1l5^long mRNA injected moe^- mutants (Figure 6D-F).

AB wild-type strain and the moe^b781 allele were maintained and staged as described previously 732. 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.

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_ΔPBD(epb4.1l5^long with the four C'terminal TTEL deleted), epb4.1l51^{short+long_PBD} (short C-terminal 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 rescued, and in all cases higher amounts failed. Concurrent with immunohistochemical analysis, we probed with anti-Moe on alternate sections to confirm the moe^- genotype. All constructs were sequenced prior to mRNA synthesis.

We generated rabbit polyclonal antibodies against the C-terminal 20 amino acids of mammalian Crb2 (N' AGARLEMDSVLKVPPEERLI C'; 95% identity with D. rerio Crb2a, 90% identity with Crb2b) conjugated to keyhole limpet hemocyanin. This anti-Crb antibody recognizes GST-tagged recombinant purified intracellular domains of D. rerio Crb1, Crb2a, Crb2b, Crb3a, and Crb3b by western blot but not GST alone (data not shown). Rabbit antibodies were generated to be specific for either Epb4.1l5^long or Epb4.1l5^short by immunizing rabbits with purified recombinant His-Epb4.1l5^{long_445–731} and His-Epb4.1l5^{short_445–504} protein that was purified with HisLink resin (Promega). All rabbits were immunized and boosted with about 500 μg protein in PBS (University of Massachusetts-Amherst antibody facility). Anti-Epb4.1l5^short serum was affinity purified against 2 mg GST-Epb4.1l5^{short_445–504} (cDNA cloned into pGEX-4T1 vector (Amhersham Biosciences) that was purified with Glutathione Sepharose (Amersham Biosciences), cross-linked to a NHS-activated Sepharose (Amersham Biosciences) column, and eluted in 100 mM glycine pH 2.5. Affinity purified antibodies were dialyzed against PBS, and BSA was added to 1mg/ml and glycerol to 50%.

Protein was extracted from mouse tissues by homogenization in PBS with protease inhibitors (Complete Mini, Roche), 1 mM AEBSF, 0.2 mM Na₂SO₄, 1% Trition X-100. 60 μg of total protein from each tissue was resolved on a 10% SDS-PAGE gel. Zebrafish protein was prepared by homogenizing zebrafish embryos or larvae in 2 μl reducing sample buffer (plus protease inhibitors) per individual. To reduce interference from yolk proteins, embryos were deyolked according to Link et al., 2006 34. Samples were denatured at 100°C for 5 min, vortexed, and insoluble debris pelleted by centrifugation. Western blotting was conducted according to standard procedures. Primary antibodies: rabbit anti-Epb4.1l5^long 1:750; affinity purified rabbit anti-Epb4.1l5^short 1:15, 5 μg/mL final; rabbit anti-Myc 1:10,000 (Bethyl); mouse anti-α-Tubulin 1:2000 (Developmental Studies Hybridoma Bank).

Zebrafish embryos and larvae were fixed in 4% paraformaldehyde/0.1M NaPO₄ pH 7.2 for 1 hr at room temperature, washed in 0.1M NaPO₄ pH 7.2, embedded in agar, and equilibrated in 30% sucrose, and frozen sections cut (18 μm or 30 μm). Sections were incubated in blocking solution (20% goat serum, 10 mg/mL BSA, 145 mM NaCl, 10 mM lysine, 50 mM Tris pH 7.4, 0.1% Tween-20), then in primary and secondary antibody, overnight at 4°C and 3 hours at room temperature respectively (rabbit anti-Moe 1:500, rabbit anti-panCrb 1:1000, mouse anti-ZO-1 1:10, rabbit anti-CAII 1:500, mouse anti-ZPR1 1:100, mouse anti-Rhodopsin 1:100 (B6–30), goat anti-mouse and goat anti-rabbit 1:100 (Molecular Probes and Jackson ImmunoResearch), and TO-PRO-3 1:500 (Invitrogen)). Slides were cover-slipped with Prolong anti-fade reagent (Invitrogen). Images were acquired on a Zeiss 510C META confocal microscope and processed with ImageJ 1.37v 35. Volume determination of outer segments was performed on albino larvae by quantifying the voxels represented by Rhodopsin (B6–30) labeling in 3D reconstructions with the ImageJ plugin Sync Measure 3D 36.

Larvae were fixed at 6 days in 4% PFA/0.5% gluaraldehyde/0.1M NaPO₄ 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.

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).