The role of the extracellular domain in vertebrate Crb1 and Crb2 proteins is unknown and no molecules have been identified that interact with this region. In Drosophila, the extracellular domain plays a critical role in determining the length of the stalk region in photoreceptors 5 . The structure of zebrafish Crb2a protein is similar to Drosophila Crumbs and has a large extracellular domain with 19 EGF-like domains and 3 Laminin-like domains, a single pass transmembrane domain, and a small intracellular domain with a FERM binding domain (FBD) and a PDZ-binding domain (PBD; Fig. 1A). We asked whether the role of Crb2a extracellular domain is conserved across species and sought to determine the role of the extracellular domain of Crb2a in protein localization and rod morphogenesis. We generated transgenics expressing the full-length Crb2a protein (Crb2a^FL), the extracellular domain of Crb2a containing the transmembrane domain (Crb2a^Extra_TM), and the extracellular domain of Crb2a that lacks the transmembrane domain and, thus, is a secreted Crb2a extracellular domain (Crb2a^Extra_Secr; Fig. 1B),

The rod photoreceptor has four morphologically and functionally distinct compartments. There are two basal compartments, the proximal cell body and the distal synaptic region, and two apical compartments, the proximal inner segment and the distal outer segment that is filled with membranous discs packed with photon-capturing Rhodopsin molecules (Fig. 1C). The apical and basal compartments are separated by a specialized adherens junction called the outer limiting membrane (OLM). The inner segment can be further subdivided into the proximal myoid region and the distal ellipsoid region (Fig. 1C). All the transgenic lines were made in or crossed into the Tg(Xop:EGFP) line 20 for analysis of rod morphology. The four compartments are visible in rods in the Tg(Xop:EGFP) background when labeled with anti-Rhodopsin antibodies 21 , the OLM is also visible as a small gap or constriction in GFP fluorescence (Fig. 1D, arrow).

At 6 d, which is about 3 days after the first rod birthdates, Crb2a^FL protein localized primarily to the inner segment, intensively in the proximal myoid and at a lower level in the ellipsoid, and very little was found on the plasma membrane of the cell body and none was found in the outer segment (Fig. 1E-E"'). The myoid region of the inner segment of Crb2a^FL expressing rods was enlarged (Fig. 1E-E"'). This morphology was so unique and distinct that we could easily distinguish Crb2a^FL rods from wild-type or other transgenic lines by observing GFP fluorescence alone. We also observed that ectopic processes often projected from the inner segments of Crb2a^FL expressing rods (Fig. 1E"').

Crb2a^Extra_TM was very strong in the outer segment where we observed very fine stripes of Crb2a^Extra_TM that appear more concentrated on one side of disks in the outer segment, perhaps near the axoneme, and there was weak plasma membrane labeling in the cell body and inner segment (Fig. 1F-F" and Additional file 1B, B'). Crb2a^Extra_Secr localized throughout the entire expressing rod except the outer segment (Fig. 1G-G"). We observed Crb2a^Extra_Secr in the region where Crumbs (predominantly Crb2a) proteins normally localize, just apical to the OLM (Fig. 1G", white bracket) and surrounding the inner segments of cones (Fig 1G", yellow bracket and Additional file 2). We observed no Crb2a^Extra_Secr below the OLM except within the cell bodies of the expressing rods. Single z-sections of Crb2a^FL-, Crb2a^Extra_TM- and Crb2a^Extra_Secr-expressing rods are shown in Additional file 1A-C'.

In order to investigate the role of the intracellular domain, consisting of the FERM-binding and PDZ-binding domains, we over-expressed several Crb2a constructs that lack the extracellular domain (Fig. 2A). We found that the localization of the Crb2a constructs that lack the extracellular domain is very different from those that retain the extracellular domain. At 6 d, Crb2a^IntraWT localized to the cell body plasma membrane, the proximal inner segment including the Golgi apparatus (see Additional file 3 for Golgi labeling in Crb2a transgenics), and outer segment (Fig. 2C-C"). Stripes of Crb2a^IntraWT were often observed in the outer segments (Fig. 2C') and the number of stripes correlated with the number of light cycles to which the cells had been exposed. The intracellular construct that lacks a functional FERM binding domain, Crb2a^IntraΔFBD, localized similarly to Crb2a^IntraWT, with labeling in the cell body plasma membrane, proximal inner segment (presumptive Golgi apparatus), and outer segment (Fig. 2D-D"). The amount of Crb2a^IntraΔFBD protein in the cell body plasma membrane appeared lower than Crb2a^IntraWT. We also observed stripe patterns of Crb2a^IntraΔFBD in the outer segments (Fig. 2D'), suggesting that this construct may be regulated similarly to Crb2a^IntraWT. The intracellular construct that lacks the PDZ binding domain, Crb2a^IntraΔPBD, localized mostly to the plasma membrane of cell body and the inner segment and there was very little labeling in the outer segment (Fig. 2E-E"). We also made a control construct that lacks a functional FERM binding domain and PDZ binding domain, Crb2a^{IntraΔFBDΔPBD}, and found this construct localized to the plasma membrane of the cell body, proximal inner segment (presumptive Golgi apparatus), and outer segment (Fig. 2F-F"), similar to Crb2a^IntraWT and Crb2a^IntraΔFBD. Single z-sections of Crb2a^IntraWT-, Crb2a^IntraΔFBD-, Crb2a^IntraΔPBD- and Crb2a^{IntraΔFBDΔPBD}-expressing rods are shown in Additional file 1D-G'. We note that the levels of Crb2a^IntraWT, Crb2a^IntraΔFBD and Crb2a^{IntraΔFBDΔFBD} in the cell bodies are much lower than Crb2a^IntraΔPBD, which may not be clear from the confocal images (Fig. 2). The only intracellular construct that did not localize to the outer segment was Crb2a^IntraΔPBD, suggesting that this construct may be actively retained in the inner segment and cell body by the remaining FBD or that this construct is trafficked differently than the other constructs.

Studies in Drosophila assumed that substitution of two key residues in the FBD would compromise the binding between the FERM-binding domain in Crumbs and the FERM protein, but this idea was not directly tested 14 . In this study we tested whether the two mutations, Tyr10 to Ala and Glu16 to Ala (Fig. 2A), in the FBD are sufficient to abolish the interaction between Crb2a and the FERM protein Moe. We generated GST tagged Crb2a^IntraWT and Crb2a^IntraΔFBD and tested their ability to bind the His-tagged FERM domain of Moe. We found that recombinant HIS-Moe_FERM bound to GST-Crb2a^IntraWT but very poorly to GST-Crb2a^IntraΔFBD, at a level close to the negative control of GST alone (Fig. 3). This result suggested that the two mutations we made to generate Crb2a^IntraΔFBD were sufficient to interfere with Moe-Crb2a interactions. This suggests that overexpressed Crb2a^IntraΔFBD may lose its binding to Moe and thus becomes mislocalized to the outer segment (Fig. 2D').

Since many of our Crb2a transgene products are mislocalized to the outer segment and cell body, we asked whether over-expression of any of the Crb2a constructs interferes with Rhodopsin localization in the outer segment. To recognize Rhodopsin we used a Rhodopsin-specific monoclonal antibody (clone R6-5 21 ). In wild-type rods at 6 d, Rhodopsin localized only to the outer segment (Fig. 1B, 2B), and we found that Rhodopsin localization is normal in all our transgenic lines (Figs. 1 and 2 and data not shown), indicating that over-expression of either the intracellular or extracellular domains of Crb2a does not interfere with Rhodopsin transport or targeting even though many transgene products localized to the outer segment.

Our previous work suggested Moe is a negative regulator of the Crumbs function in photoreceptors, and loss-of-moe function resulted in a larger outer segment and affected photoreceptor morphology 8 . We sought to determine whether overexpression of any of the Crb2a constructs affected rod size or morphology. We examined rod morphology at 6 d in the lines that express the intracellular domains of Crb2a. Gross morphology of rods as visualized by GFP in combination with anti-Rhodopsin antibody labeling was normal in Crb2a^IntraWT, Crb2a^IntraΔFBD, Crb2a^IntraΔPBD and Crb2a^{IntraΔFBDΔPBD} transgenics (Fig. 2A-F). However, when we measured the volume or size of rod outer segments at 6 d, we found that the volume of outer segments of Crb2a^IntraWT, Crb2a^IntraΔFBD, Crb2a^IntraΔPBD expressing rods increased significantly compared to wildtypes (Fig. 4A, red circles), whereas, the size of the inner segment plus cell body was not significantly different from wildtypes (Fig. 4A, black circles). These results suggest that over-expression of any defined domain of the Crb2a intracellular domain increases the size of outer segments in rods without affecting gross morphology.

We measured the size of Crb2a^FL, Crb2a^Extra_TM and Crb2a^Extra_Secr expressing rods at 6 d. The outer segments of Crb2a^FL rods were not significantly larger than wild-type rods but outer segments of Crb2a^Extra_TM and Crb2a^Extra_Secr rods are larger than those in wildtypes (Fig. 4B). The most dramatic morphological change we observed was in Crb2a^FL rods, where the myoid region of the inner segment was enlarged (Fig. 1E-E"). The enlargement of the myoid in Crb2a^FL rods was most obvious in the area immediately above the OLM, therefore we measured the width the myoid just apical to the OLM in confocal z-projections as indicated in Fig. 4C. The width of the myoid in Crb2a^FL expressing rods was nearly 50% wider than wildtypes (Fig. 4C, C'). The widths of the myoid in Crb2a^Extra_TM or Crb2a^Extra_Secr rods were not significantly different from wildtypes (Fig. 4C, C'). We also quantified the size of the inner segment by measuring the area of the entire myoid in confocal z-projections as indicated in Fig. 4D. This measurement showed that Crb2a^FL, Crb2a^Extra_TM, Crb2a^Extra_Secr rods had a significant increase in the size of myoid (Fig. 4D').

Our next goal was to determine whether expression of any of our transgenes affected the levels or localization of endogenously expressed Crb2a/b proteins. The panCrb antibodies that we used previously and the one we raised and used in this study were generated against a C' terminal peptide that includes the PBD that is highly conserved in all members of the Crumbs family of proteins and these antibodies recognize all zebrafish Crumbs proteins 8 (and data not shown). Thus, the panCrb antibody recognizes all constructs that retain the C' terminal peptide and cannot be used to distinguish between those constructs and endogenous Crumbs proteins. In order to distinguish between endogenous Crb2a/b proteins and intracellular transgene products, we required an antibody that recognizes the extracellular domain of Crb2a.

In a search for such an antibody, we tested the monoclonal antibody zs-4 and found that its labeling pattern is remarkably similar to panCrb antibody labeling in the photoreceptor layer. We performed double labeling with zs-4 and panCrb antibodies and found a nearly identical localization of the two antibodies (Fig. 5A-A"). They do not colocalize perfectly (Fig. 5A", arrow), suggesting that zs-4 labels only Crb2a or only Crb2b 8 . Since crb2b is rapidly down-regulated and most of the panCrb labeling is also labeled by zs-4, it was more likely that zs-4 recognizes Crb2a. Unfortunately, zs-4 does not recognize protein in western blot (data not shown). However, using the Crb2a^Extra_TM and Crb2a^Extra_Secr transgenics that we produced, we show that zs-4 recognizes the extracellular domain of Crb2a. To show that zs-4 recognizes the extracellular domain of Crb2a, we labeled retinal sections from two of our transgenic lines, Crb2a^Extra_TM and Crb2a^Extra_Secr, with the zs-4 antibody and an anti-HA antibody. Since zs-4 and the anti-HA (clone 16B12) are both IgG₁ isotype, we tested several other anti-HA polyclonals and monoclonals that might be compatible for zs-4 double-labeling. The best anti-HA antibody we tested was an IgG₃ isotype but it was weaker than clone 16B12 and the background is higher (Fig. 5B, C). Nonetheless, we found that zs-4 labeling resembled the anti-HA labeling except that, in addition, it also labels the endogenous Crumbs proteins localized to the inner segment region (Fig. 5B'-C' and data not shown). The zs-4 antibody also recognizes Crb2a^Extra_Secr that localizes around the inner segments of double cones (Fig. 5E-F' and Additional file 2). We also observed that zs-4 labeling was lost in ome (crb2a) mutant embryos (Fig. 5G, H) and zs-4 did not label tissues that express only crb2b, such as kidney podocytes (YCH and AMJ, unpublished observation). Our observations, taken together, support the conclusion that zs-4 recognizes the extracellular domain of Crb2a, and, thus, this antibody was a useful tool in this study to localize the endogenous Crb2a proteins in Crb2a^Intra transgenics.

Constructs similar to some of ours were tested for dominant affects in Drosophila, and the over-expression of some Crumbs constructs caused mislocalization of other apical polarity proteins, but it was not reported whether over-expression of Crumbs constructs affected the localization of endogenous Crumbs proteins 5 14 . To determine whether endogenous Crb2a/b proteins were mislocalized in the Crb2a^FL transgenics, we double labeled sections with anti-HA (clone 16B12) and anti-panCrb antibodies and examined whether the anti-HA and panCrb labeling appeared 100% coincident. This is an imperfect experiment because the relative affinities of the zs-4 monoclonal and the panCrb polyclonal are likely to be different but it is the best approximation available. In wildtypes, anti-panCrb labels primarily the inner segment region of photoreceptors with additional weak labeling in the outer segment region, which could be background (Fig. 6A). We observed that in Crb2a^FL expressing rods, anti-panCrb antibody labeling seems to extend beyond anti-HA labeling, and labels the distal myoid and ellipsoid regions of the rod inner segments (Fig. 6B', brackets). The inset in Fig. 6B shows the ectopic inner segment processes in Crb2a^FL transgenic rods. Because both Crb2a^Extra_TM and Crb2a^Extra_Secr transgene products lack the intracellular domain, we used anti-HA to localize transgene products and anti-panCrb to localize endogenous Crb2a/b proteins. We did not observe mislocalization of anti-panCrb antibody labeling in Crb2a^Extra_TM or Crb2a^Extra_Secr transgenics (Fig. 6C-C"' and 6D-D"').

To address whether expression of the intracellular constructs caused mislocalization of endogenous Crb2a, we used zs-4 antibody to localize endogenous Crb2a in Crb2a^IntraWT, Crb2a^IntraΔFBD, Crb2a^IntraΔPBD and Crb2a^{IntraΔFBDΔPBD} transgenic lines (Fig. 7). We found that zs-4 labeling in Crb2a^IntraWT (Fig. 7B), Crb2a^IntraΔFBD (Fig. 7C), and Crb2a^IntraΔPBD (Fig. 7D) and Crb2a^{IntraΔFBDΔnΔPBD} (Fig. 7E) was similar to that in wildtypes (Fig. 7A), indicating that overexpression of these constructs did not alter endogenous Crb2a localization.

We asked whether overexpression of Crb2a constructs alters the localization of other Crumbs complex components such as Moe or those that interact indirectly, such as Prkci (also known as aPKCΔλ) 22 23 24 . We used antibodies against Moe and Prkci to localize these two proteins in our transgenics. We found that the localization of Moe and Prkci was unaffected in all transgenic lines (Additional file 4 and Additional file 5, respectively).

AB wild-type strain, Tg(Xop:EGFP);alb ^-/+, the Tg(Xop:Crb2a), ome ^m98 fish lines were maintained and staged as previously described according to Westerfield 38 . All experiments involving animals were performed with approval by and in accordance with the University of Massachusetts-Amherst Institutional Animal Care and Use Committee (IACUC).

Diagrams of the transgene constructs are shown in Figs. 1B and 2A. The constructs were cloned behind 0.8 kB of the Xenopus rod opsin promoter (Xop; 17 ). All constructs have an N'terminal signal peptide (SP) taken from zebrafish Crb2b (SignalP 3.0 Server) and an influenza hemagglutinin (HA) tag (YPYDVPDYA) just after the SP. To make Crb2a^FL, the SP of Crb2a was substituted with the Crb2b SP and the HA tag was engineered immediately after the predicted SP cleavage site. To make Crb2a^Extra_Secr and Crb2a^Extra_TM, we used site-directed mutagenesis to place a stop codon just before and after the transmembrane domain, respectively. To make Crb2a^IntraWT, the SP from zebrafish Crb2b was introduced with PCR using a pair of primers in which the HA tag sequence was engineered immediately after SP within the reverse primer and the sequence containing the Crb2a transmembrane domain plus the intracellular domain was amplified by PCR and cloned in-framed following the HA sequence. Crb2a^IntraΔPBD was made using Crb2a^IntraWT as a template and site-directed mutagenesis was performed to introduce a stop codon after the last glutamic acid residue (E16) in the FERM-binding domain. To make Crb2a^IntraΔFBD, the conserved amino acids tyrosine (T10) and glutamic acid (E16) of Crb2a^IntraWT were substituted with alanines to compromise the FERM-binding domain as described by Izaddoost and colleagues 14 . Site-directed mutagenesis was performed on Crb2a^IntraΔPBD to introduce a stop codon after E16 to make Crb2a^{IntraΔFBDΔPBD}. All constructs contain a poly-adenylation sequence at the 3'-end. The constructs were cloned into pTol 18 19 .

Transgenic zebrafish lines were generated using pTol system 18 19 . We coinjected 40 ng/mL of pTol-transgene construct plasmid with 40 ng/mL transposase mRNA into one-cell stage Tg(Xop:EGFP);alb ^-/+ embryos. Injected embryos were grown to adulthood and out-crossed with Tg(Xop:EGFP);alb ^-/+ to produce offspring. We used PCR to identify transgenic offspring; forward primer GGCATGCCGTCCCTAAAAG designed within the promoter region, and the reverse primer AGCGTAATCTGGAACATCGTAT within the HA tag sequence. We identified three germline transgenic founders (F0) for each construct and generated F1 lines. We confirmed transgene expression by anti-HA immunohistochemistry. Transgenic F1s and subsequent generations were identified by PCR on fin clip DNA. F1 carriers were out-crossed with Tg(Xop:EGFP);alb ^-/+ line to produce F2s.

Construction and expression of Crb2a fusion proteins and HIS-Moe_FERM were described previously 8 . GST-Crb2a^IntraΔFBD was made using site-directed mutagenesis on GST-Crb2a^IntraWT (as described above) to introduce mutations in two residues of FBD (E10 and T16). Protein interactions were performed as described previously 8 , using 10 mg of His-Moe_FERM incubated with 10 mg of GST, GST- Crb2a^IntraWT or GST-Crb2a^IntraΔFBD.

Production and levels of transgene products were assessed on 6 d retinal sections by anti-HA antibody labeling. We fixed 6 d zebrafish in the afternoon in 4% paraformaldehyde. Cryostat sections (20-30 μm) were treated with 0.1% SDS for 15 min, washed in PBS with 0.1% Tween (PBS-Tw), incubated in 10% goat serum in PBS-Tw, rinsed briefly in PBS-Tw, and incubated overnight at 4°C in primary antibody (monoclonal anti-HA IgG₁, 1:1000 (Covance); monoclonal anti-HA IgG₃, 1:500 (Upstate); rabbit anti-Moe, 1:1000; rabbit anti-panCrb (we raised against the synthetic peptide, AGARLEMDSVLKVPPEERLI), 1:500; anti-aPKCζ, 1:1000 (Santa Cruz); rabbit anti-GFP 1:200 (Molecular Probes), anti-Rhodopsin monoclonal R6-5, 1:50 21 ; zs-4 antibody, 1:10 (University of Oregon Monoclonal Antibody Facility); rabbit anti-GOLGA5 1:500 (Sigma HPA000992). Sections were washed, incubated with the appropriate secondary antibodies (-FITC/-TRITC (Molecular Probes) 1:100, -CY5 1:100 (Jackson) goat anti-mouse IgG3 rhodamine red-conjugated, 1:100; goat anti-mouse IgG₁ Cy-5-conjugated, 1:250; goat anti-mouse IgG₁ rhodamine red-conjugated, 1:100; goat anti-mouse IgG_2a Cy-5-conjugated, 1:100 (Jackson Laboratory)), and analyzed with a Zeiss LSM 510 Meta Confocal System. We primary analyzed the retinas in alb ^-/- individuals to ensure that the entire outer segment was visible and not obscured by the RPE; the localization of transgene products was the same in pigmented siblings (data not shown). Confocal images are a single scan (averaged 4 times) at about 1 μm optical thickness. Volume of wild-type and transgenic rods was measured using Sync Measure 3D function of Image J. Only cells that were completely captured in the confocal stacks were measured. Outer segments were outlined and measured by an overlap of anti-GFP and R6-5 labeling; cell compartments that are free of R6-5 labeling were outlined and measured as inner segment plus cell body. The width of inner segment (Fig. 4C) and the area of myoid regions (Fig. 4D) were outlined and measured using Image J.