Multiple domains of Stardust differentially mediate localisation of the Crumbs-Stardust complex during photoreceptor development in Drosophila.

Drosophila Stardust (Sdt), a member of the MAGUK family of scaffolding proteins, is a constituent of the evolutionarily conserved Crumbs-Stardust (Crb-Sdt) complex that controls epithelial cell polarity in the embryo and morphogenesis of photoreceptor cells. Although apical localisation is a hallmark of the complex in all cell types and in all organisms analysed, only little is known about how individual components are targeted to the apical membrane. We have performed a structure-function analysis of Sdt by constructing transgenic flies that express altered forms of Sdt to determine the roles of individual domains for localisation and function in photoreceptor cells. The results corroborate the observation that the organisation of the Crb-Sdt complex is differentially regulated in pupal and adult photoreceptors. In pupal photoreceptors, only the PDZ domain of Sdt – the binding site of Crb – is required for apical targeting. In adult photoreceptors, by contrast, targeting of Sdt to the stalk membrane, a distinct compartment of the apical membrane between the rhabdomere and the zonula adherens, depends on several domains, and seems to be a two-step process. The N-terminus, including the two ECR domains and a divergent N-terminal L27 domain that binds the multi-PDZ domain protein PATJ in vitro, is necessary for targeting the protein to the apical pole of the cell. The PDZ-, the SH3- and the GUK-domains are required to restrict the protein to the stalk membrane. Drosophila PATJ or Drosophila Lin-7 are stabilised whenever a Sdt variant that contains the respective binding site is present, independently of where the variant is localised. By contrast, only full-length Sdt, confined to the stalk membrane, stabilises and localises Crb, although only in reduced amounts. The amount of Crumbs recruited to the stalk membrane correlates with its length. Our results highlight the importance of the different Sdt domains and point to a more intricate regulation of the Crb-Sdt complex in adult photoreceptor cells.


Introduction
Polarisation of cells largely depends on the asymmetric distribution of multiprotein complexes that define spatially and functionally distinct membrane-associated compartments within cells. Formation of these complexes brings proteins involved in the same process into close proximity, thereby facilitating their interaction and ensuring efficiency and specificity of function, such as adhesion or signalling. The core constituents of membrane-associated complexes are often scaffolding proteins and one or more transmembrane protein(s). The latter can be receptors or adhesion molecules and often serve to anchor the complex at the membrane. In several cases, multiprotein complexes provide the basis for formation of cell-typespecific junctions. Although precise localisation of these complexes is crucial for their function, little is known about the mechanisms that control targeting, assembly and maintenance of individual components at a particular site within a cell.
Central components of many membrane-associated protein complexes are scaffolding proteins, each of which possesses several protein-protein interaction domains that, together, allow a wide variety of interactions to occur. Membrane-associated guanylate kinases (MAGUKs) form a family of scaffolding proteins engaged in organising multiprotein complexes that are often associated with cellular junctions and signalling complexes, e.g. the vertebrate tight junction (TJ) and Drosophila septate junction (SJ) in epithelial cells, and the neuromuscular junction (NMJ). Their capacity to serve as a platform for recruiting larger protein assemblies results from the presence of multiple protein-protein interaction domains: one to three PSD-95/Discs large/zonula occludens 1 (PDZ) domains, a Src homology-3 (SH3)-domain and a guanylate kinase (GUK) domain. Some members additionally contain one or two Lin-2-Lin-7 (L27)domains in their N-terminus and/or a Hook domain between the SH3-and the GUK-domain. This modular structure facilitates the recruitment of various components into supramolecular protein complexes, the composition of which often depends on the cell type and/or the developmental stage. Strikingly, genes that encode MAGUK-proteins often give rise to tissue-and stage-specific protein isoforms through alternative splicing, thus increasing their versatility and the possibility for cell-type-specific interactions, localisation and/or function. For example, Drosophila discs large (dlg) expresses an epithelia-specific isoform (Dlg-A) that lacks the L27 domain and is therefore unable to bind to Drosophila Lin-7 (also known as Veli), whereas an isoform containing the L27 domain (Dlg-S97) can associate with Drosophila Lin-7 in the neuromuscular junction (Bachmann et al., 2004;Mendoza et al., 2003). Drosophila polychaetoid, an orthologue of mammalian zonula occludens 1 (ZO-1, also known as TJP1), encodes two isoforms, one localising Drosophila Stardust (Sdt), a member of the MAGUK family of scaffolding proteins, is a constituent of the evolutionarily conserved Crumbs-Stardust (Crb-Sdt) complex that controls epithelial cell polarity in the embryo and morphogenesis of photoreceptor cells. Although apical localisation is a hallmark of the complex in all cell types and in all organisms analysed, only little is known about how individual components are targeted to the apical membrane. We have performed a structure-function analysis of Sdt by constructing transgenic flies that express altered forms of Sdt to determine the roles of individual domains for localisation and function in photoreceptor cells. The results corroborate the observation that the organisation of the Crb-Sdt complex is differentially regulated in pupal and adult photoreceptors. In pupal photoreceptors, only the PDZ domain of Sdt -the binding site of Crb -is required for apical targeting. In adult photoreceptors, by contrast, targeting of Sdt to the stalk membrane, a distinct compartment of the apical membrane between the rhabdomere apically and the other distributing more broadly along the lateral membrane of epithelial cells in the wing imaginal disc (Wei and Ellis, 2001).
Dissection of mechanisms that regulate protein complex localisation is complicated by the fact that protein complexes are highly dynamic structures whose composition can undergo rapid modifications. In addition, some proteins localise in several steps and mechanisms might differ depending on the type of tissue or the developmental stage. Localisation of Drosophila Dlg to the septate junctions in epithelial cells, for example, requires the combined action of its PDZ2 and Hook domain (Hough et al., 1997). Localisation of Drosophila Scribble (Scrib), a member of the LAP protein family, is a two-step process that first requires a region containing 16 leucine-rich repeats (LRR) to target the protein cortically, then relies on the four PDZ domains to restrict it to the baso-lateral membrane (Albertson et al., 2004).
One of the scaffolding proteins of the MAGUK family is Drosophila Stardust (Sdt), the PDZ-domain of which interacts with the four C-terminal amino acids of the transmembrane protein Crumbs (Crb). In addition to a SH3-, Hook-and GUK-domain, it contains evolutionary conserved region 1 and 2 (ECR1 and ECR2, respectively) in the N-terminus, which are required to interact with Drosophila Par-6 (Bachmann et al., 2001;Berger et al., 2007;Hong et al., 2001;Wang et al., 2004) and a canonical L27 domain that interacts with Drosophila Lin-7. Drosophila sdt encodes several protein isoforms that result from differential splicing and/or transcription initiation. Two of them, Sdt-A and Sdt-B2 (Fig. 1A), differ with respect to the presence or absence of a large exon (exon 3), which encodes an N-terminal 433 amino acid region that has no obvious domain structure. All proteins expressed in the retina lack this N-terminal extension (Berger et al., 2007).
Besides maintaining epithelial tissues in the Drosophila embryo (Bachmann et al., 2001;Hong et al., 2001;Tepass and Knust, 1993), Sdt (as well as its binding partners Crb and PATJ) is also required for morphogenesis and function of the photoreceptor cells (PRCs). During development, PRCs are specified from epithelial cells of the eye imaginal discs in third instar larvae. In pupae, PRCs undergo complex morphogenetic changes that are associated with a 90° shift of the apical membrane that finally adopts a lateral position. Subsequently, the apical membrane is subdivided into the stalk, a supporting membrane immediately apical to the zonula adherens (ZA), and the rhabdomere, a central region consisting of a highly pleated array of microvilli that harbours most of the phototransduction signalling cascade proteins. Conspicuous cell elongation accompanies rhabdomere formation (Longley and Ready, 1995). In contrast to embryonic epithelia that lack sdt or crb, mutations in sdt, crb or PATJ do not disrupt PRC apico-basal polarity. However, lack of any of these genes affects PRC morphogenesis. This is manifested by a failure to correctly expand the apical membrane, and by a reduction in the length of the stalk membrane (Berger et al., 2007;Hong et al., 2003;Izaddoost et al., 2002;Johnson et al., 2002;Nam and Choi, 2003;Nam and Choi, 2006;Pellikka et al., 2002;Richard et al., 2006a). In addition, mutations in all three genes lead to progressive light-induced retinal degeneration (Berger et al., 2007;Johnson et al., 2002;Richard et al., 2006a) (reviewed in Knust, 2007). During the first half of pupal development, members of the Crb complex are localised throughout the apical membrane, but are restricted to the stalk membrane at later stages and in PRCs of adult flies. The absence of any of these proteins in adult PRCs results in delocalisation of all the others (Berger et al., 2007;Richard et al., 2006a). Recently, Drosophila Lin-7 was shown to be a component of the Crb complex in pupal and adult PRCs. In contrast to PRCs mutant for sdt, crb or PATJ, those mutant for Lin-7 do not display morphogenetic defects and maintain Crb, Sdt and PATJ proteins at the stalk membrane. However, similar to the others, Lin-7-mutant PRCs degenerate in a light-dependent manner (Bachmann et al., 2008).
The core components of the Drosophila Crb-Sdt complex are conserved throughout evolution. In mammals, one orthologue of Sdt (MPP5, also known as Pals1), three of Crb (CRB1, CRB2 and CRB3) and two of Drosophila PATJ (INADL and MPDZ) were identified (reviewed in Assémat et al., 2008;Richard et al., 2006b). With the exception of CRB1, whose expression is restricted in mouse and human to retina and brain (den Hollander et al., 2002;den Hollander et al., 1999), all are expressed in multiple tissues throughout development. Here, they form apical membraneassociated protein complexes in association with cell-type-specific proteins. In cultured epithelial cells, these complexes are important for tight-junction stability (Michel et (Bachmann et al., 2001)], and Sdt-B2 (Berger et al., 2007). (B) List of Sdt variants used in this study. All transgene-encoded proteins in A and B carry a FLAG-tag at their N-terminus. (C) Western blot analysis of Sdt-FLAG proteins from the eyes of flies that carry large clones mutant for sdt K85 and express the respective transgene using Rh1GAL4. Anti-FLAG antibody was used to detect Sdt-FLAG proteins. Eyes of flies with large clones mutant for sdt K85 but not expressing Sdt-transgenes were used as control. Straight et al., 2004;Wang et al., 2004). In addition, mutations in human CRB1 are associated with Retinitis pigmentosa 12 (RP12) and Leber congenital amaurosis (LCA) (reviewed in Assémat et al., 2008;Richard et al., 2006b). Mutation of crb and sdt orthologues in the zebrafish (Danio rerio) (oko meduzy and nagie oko, respectively) affects cellular patterning of the retina and apical membrane differentiation of some cell types, including the PRC (Malicki and Driever, 1999;Omori and Malicki, 2006;Wei and Malicki, 2002;Wei et al., 2006). Thus, the Crb-Sdt complex performs conserved functions during cell polarisation and retinal morphogenesis in different species.
Here, we describe a structure-function analysis of the Sdt protein(s) aimed at understanding the role of individual proteinprotein interaction domains during PRC morphogenesis. We show that the function of individual domains in targeting Sdt to the apical or stalk membrane, and in recruiting and stabilising other complex members, exhibits striking differences in Drosophila pupal and adult stages. In addition, we provide evidence for the importance of specific domains in rescuing morphological defects caused by sdtnull mutation.

Generation and characterisation of sdt transgenes
To study the function of Sdt and its different domains, we used two full-length transgenes that encode the two known isoforms Sdt-A and Sdt-B2 (Fig. 1A). These two isoforms differ in the presence or absence of exon 3, which encodes a 433 amino acid N-terminal region that has no obvious domain structure (Bachmann et al., 2001;Berger et al., 2007). Sdt-A was used as a template to produce multiple sdt transgenes carrying different deletions (Fig. 1B). Proteins encoded by the transgenes contain an N-terminal FLAG epitope tag, and were expressed using the UAS/GAL4 system (Brand and Perrimon, 1993). Since zygotic sdt mutants die during embryogenesis, we used a clonal approach to obtain sdt mutant photoreceptor cells (PRCs). We generated large eye clones (Newsome et al., 2000) mutant for sdt K85 and expressed the transgenes with Rh1-GAL4, which activates moderate levels of Gal4 in the outer photoreceptors, R1-R6, starting from the last third of pupal development. Alternatively, we used the mosaic analysis with a repressible cell marker (MARCM) technique that, additionally, allows mutant cells to be marked (Lee and Luo, 2001). sdt K85 was chosen because it is a complete null allele in PRCs, on the basis of protein expression and phenotype analysis (Berger et al., 2007). To make sure that the Sdt-FLAG proteins were produced, western blots of retinal extracts from Rh1GAL4/UAS-Sdt-Flag flies that contained no endogenous Sdt were probed with antibodies against FLAG or Sdt-PDZ (Fig. 1C, and data not shown).

The PDZ domain of Sdt is necessary for apical localisation at pupal stage
PRCs develop from epithelial cells of Drosophila larval eye imaginal discs. During the first half of wild-type pupal PRC development, endogenous Sdt localises, together with Crb, PATJ, Lin-7, Par-6 and F-actin, at the apical membrane (reviewed in Knust, 2007;Richard et al., 2006b). To determine the domain(s) of Sdt that are responsible for its apical targeting at this stage, we assayed the localisation of different transgene-encoded Sdt proteins in sdt K85mutant PRCs that do not express any endogenous Sdt protein ( Fig.  2A). Of all Sdt protein domains tested, only the PDZ domain was necessary for apical localisation of Sdt, wheras Sdt variants that lack the PDZ domain (Sdt-ΔPDZ and Sdt 1-L27C ) showed a low level Journal of Cell Science 121 (12) of uniform cortical association ( Fig. 2B-B''' and data not shown). All other proteins exhibit apical localisation ( Fig. 2C-C''' and data not shown). In sdt K85 -mutant PRCs of the same stage, localisation of Crb and PATJ is known to be unaffected (Berger et al., 2007). Consistent with this, Crb and PATJ localisation was unaltered in the presence of any Sdt variant, independently of whether the transgene-encoded protein localised apically or not ( Fig. 2D-D''' and data not shown). This result confirms that localisation of Crb and PATJ is independent of Sdt at this stage of Drosophila development (Berger et al., 2007).
ECR1 and ECR2 domains of Sdt are necessary to localise Par-6 apically at pupal stage Drosophila Par-6, a member of the Par-protein network, is restricted to the apical pole in PRCs at 40-50% pupal development (p.d.) and requires sdt for its correct localisation (Berger et al., 2007;Hong et al., 2003). In the absence of sdt, apical Par-6 was strongly reduced and some Par-6 protein was detected basal to E-cadherin, a marker of the ZA (Fig. 3A-A'''). Par-6 has been shown to interact with Sdt in vitro, and this interaction depends on the presence of both ECR domains -ECR1 and ECR2 (Wang et al., 2004) (Ö.K. and E.K. unpublished data). To further analyse whether these Sdt domains are also required to recruit Par-6 in developing PRCs, we studied Par-6 localisation in sdt K85 -mutant pupal PRCs that expressed different sdt-transgenes. Sdt-ΔN, which lacks both ECR motifs, failed to restore apical DPar-6 protein localisation (data not shown), supporting the importance of these motifs for in vivo recruitment of Par-6. Of all tested proteins that contained ECR1 and ECR2, only Sdt-B2 completely rescued the apical localisation of Par-6 (Fig. 3B-B'''). Sdt-A and other variants, in which the distance between the two ECR domains is larger than in Sdt-B2, partially restored the apical localisation of Par-6 (Fig. 3C-C''' and data not shown), with the exception of Sdt-ΔPDZ ( Fig. 3D-D"'). This is striking, because all of them, except Sdt-ΔPDZ, localise apically and contain both ECR motifs. We conclude from these results that apical localisation of Drosophila Par-6 is restored only by the expression of a Sdt protein in which the two ECR motifs are closely together, as SdtB2.

Multiple Sdt domains regulate its localisation at the stalk membrane in adult photoreceptor cells
In PRCs of adult wild-type eyes, Sdt is restricted to the stalk membrane, a defined region of the apical membrane between the rhabdomere and the ZA (Fig. 4A) (Berger et al., 2007;Hong et al., 2003). sdt K85 -mutant PRCs completely lack Sdt protein (Fig. 4B). To understand which Sdt domain(s) are required to target it to this restricted site, we expressed different sdt-transgenes in the absence of endogenous Sdt using Rh1GAL4. Transgenes encoding either full-length Sdt-A or Sdt-B2, or a protein that lacks the C-terminal L27 domain (Sdt-ΔL27C), completely restored normal Sdt localisation in otherwise sdt mutant PRCs (Fig. 4C-C'' and data not shown). The inner PRCs R7 and R8 that did not express Gal4, did also not express Sdt and served as an internal control. These data indicate that the C-terminal L27 domain, the binding site for Drosophila Lin-7, is dispensable for correct Sdt localisation at the stalk membrane. By contrast, Sdt proteins that lack the PDZ-, the SH3-or the GUK-domain were not detected at the stalk membrane, but found mainly to accumulate at the rhabdomere base ( Fig. 4D-D'' and data not shown). Finally, in the absence of the N-terminus (Sdt-ΔN), the transgene-encoded protein was uniformly distributed throughout the cell in a sdt K85 -mutant background (Fig. 4E-E''). Taken together, these data demonstrate that, unlike in pupal PRCs, Sdt localisation to the stalk membrane in adult PRCs depends on the presence of several domains.
The L27N and L27C domains are required to stabilise and localise Drosophila PATJ and Drosophila Lin-7, respectively Vertebrate Pals1 contains two L27 domains in its N-terminal part. The one located closer to the N-terminus, L27N, can interact with sdt K85 mutant MARCM clones expressing different sdt transgenes, stained with anti-Par-6 (red) and anti-E-cadherin to stain the zonula adherens (blue). Mutant cells expressing a transgene are visualised using expression of GFP (green). Cells not expressing GFP are wild-type and serve as a control to demonstrate that Par-6 localises apically in PRCs at this stage. In sdt K85 mutant PRCs (A-A'''), Par-6 is reduced in amount and delocalised from the apical membrane (arrows). In sdt K85 mutant PRCs that express Sdt-B2 (B-B'''), Par-6 localises apically. In most sdt K85 -mutant PRCs that express Sdt-A (C-C'''), Par-6 localises apically but, occasionally, also basolateral (arrows). Expression of Sdt-ΔPDZ in sdt K85 mutant PRCs does not restore apical localisation of Par-6 (D-D'''). the L27 domain of the multi-PDZ domain protein PATJ, whereas the other, L27C, located closer to the N-terminus, allows interaction with the single L27 domain of Lin-7 in vitro Roh et al., 2002;Sheng and Sala, 2001). Unlike vertebrate Pals1, however, Drosophila Sdt contains only one well-conserved L27 domain [as predicted by the databases SMART or PROSITE (Bachmann et al., 2001;Hong et al., 2001)] that binds to Drosophila Lin-7 (Bachmann et al., 2004). However, the N-terminus of Sdt can also interact with PATJ in vitro (Roh et al., 2002). This poses the question of whether Lin-7 and PATJ compete to bind the same L27 domain on Sdt, or whether PATJ is recruited into the complex using a different interaction domain, similar to the vertebrate situation. As pointed out previously (Bachmann et al., 2004), a stretch of 70 amino acids with sequence homology to the PATJbinding L27N domain of Pals1 precedes the canonical L27 domain of Sdt. Sequence comparison with other L27 domains (Doerks et al., 2000) revealed that this stretch carries several amino acid residues that are conserved at corresponding positions in related Drosophila species and in Anopheles gambiense (Fig. 5A). In yeasttwo-hybrid interactions, the N-terminus of Sdt, including the cryptic L27N domain (amino acids 1-659 of Sdt-A), interacted with the Nterminal 87 amino acids of Drosophila PATJ, which contains an L27 domain. Fragments lacking this L27 domain (Sdt 1-566 ), or containing only part of it (Sdt 646-760 ), failed to interact with the Nterminus of Drosophila PATJ in yeast-two-hybrid assays (Fig. 5B). By contrast, a protein containing only L27N (Sdt 640-766 ) bound Drosophila PATJ. This suggests, that the divergent L27 domain of Sdt is functionally equivalent to L27N of Pals1 and can recruit Drosophila PATJ into the Crb-Sdt complex, whereas the canonical Journal of Cell Science 121 (12) C-terminal L27 domain mediates interaction with Drosophila Lin-7 (Bachmann et al., 2004).
Drosophila PATJ and Drosophila Lin-7 are restricted to the stalk membrane in adult wild-type PRCs (Bachmann et al., 2008;Richard et al., 2006a), but are not localised in adult PRCs mutant for sdt K85 (Berger et al., 2007) (data not shown). Whereas the total amount of PATJ was strongly reduced in sdt K85 adult PRCs, the amount of Lin-7 was only slightly diminished (Fig. 6A). This discrepancy can be explained by the fact that Lin-7 is also localised in a Crb-Stdcomplex-independent way at the synapses between the PRCs and the first optic ganglion, the lamina (Bachmann et al., 2008). Thus, Sdt is crucial for PATJ and Lin-7 localisation at the stalk membrane, and their stabilisation in adult PRCs.
To further examine the relationship between these three proteins, and to confirm their interactions in vivo, we expressed different sdt transgenes and analysed the effects on PATJ and Lin-7 localisation and stability in the adult eye. Sdt-A and Sdt-B2 restored localisation of both PATJ and Lin-7 at the stalk membrane when expressed in sdt K85 adult PRCs (Fig. 7A-A'' and data not shown). Expression of these transgenes also restored the amount of both proteins to wildtype levels (Fig. 6A). Sdt proteins lacking the PDZ-, the SH3-or the GUK-domain, and thus localised at the rhabdomere base, recruited both PATJ and Lin-7 to this ectopic site ( Fig. 7B-B'' and data not shown). They also rescued the amount of both proteins to wild-type levels (Fig. 6A).
Sdt protein lacking the L27C domain (the binding site for Drosophila Lin-7) localised at the stalk membrane. It restored PATJ localisation (data not shown) and wild-type amount (Fig. 6A). As expected, Lin-7 did not localise with Sdt ( Fig. 7C-C''), and its levels remained reduced (Fig. 6A). This result is consistent with in-vitro interactions between Sdt and Lin-7 (Bachmann et al., 2004) and implies that such an interaction is essential for Lin-7 localisation and stability in PRCs of adult flies. Sdt-ΔN, which lacks the Nterminal L27-domain (and, hence, the binding site for PATJ) and is distributed throughout the cell (see Fig. 4D,D'), failed to localise  The amount of Lin-7 in these PRCs equalled that of wild-type PRCs, whereas PATJ levels were the same as in sdt K85 mutant PRCs without transgene expression (Fig. 6A). These results suggest that, in adult PRCs (1) binding of PATJ and Lin-7 to Sdt requires the Sdt L27Nand L27C-domain, respectively; (2) localisation of both proteins depends on Sdt; (3) both proteins are stabilised upon interaction with Sdt. For stabilisation to occur, the proteins need not to be localised at the stalk membrane.

Sdt stabilises and localises Crb only when present at the stalk membrane
In adult eyes mutant for sdt K85 , no Crb was discovered at the stalk membrane ( Fig. 8A-A'') and no Crb protein could be detected on western blots (Fig. 6B). Expression of Sdt-A, Sdt-B2 and Sdt-ΔL27C brought some Crb protein back to the stalk membrane ( Fig.  8B-B'' and data not shown). In PRCs that express full-length Sdt proteins, Crb protein was detected by western blots, although at reduced amounts compared with wild type (Fig. 6B). Upon Sdt-ΔL27C expression, very low levels of Crb were detected (Fig. 6B). In response to Sdt-ΔN expression, Crb was detected at the stalk membrane in only those few cases where the transgene-encoded Sdt was enriched at the stalk membrane ( Fig. 8C-C'', arrows), but the level of Crb detected by western blotting was very low (Fig.  6B). All other transgene-encoded proteins were unable to restore Crb protein accumulation in sdt K85 mutant PRCs, as revealed by immunofluorescence (Fig. 8D-D'' and data not shown) and western blot analysis (Fig. 6B).
In summary, these results demonstrate that the amount of Crb depends on the correct localisation, quantity and quality of Sdt protein. Sdt is not able to recruit Crb to ectopic positions within the cell, even if it contains the PDZ domain, the binding site for Crb. In addition, Crb is only stable when it is localised -together with Sdt -at the stalk membrane. In the assay used here, Sdt was not sufficient to recruit normal amounts of Crb to the stalk membrane. This suggests that additional, yet unknown factors contribute to Crb recruitment and/or stabilisation at the stalk membrane.
Rescuing stalk-membrane length of the sdt genetic null mutants One intriguing phenotype caused by mutations in sdt is the reduction of stalk-membrane length in adult PRCs (Berger et al., 2007;Hong et al., 2003). Similarly, stalk membranes in PRCs mutant for PATJ or crb are reduced in length (Johnson et al., 2002;Nam and Choi, 2006;Pellikka et al., 2002;Richard et al., 2006a). At present, not  much is known about the molecular basis of this phenotype. To address which part of Sdt is required for normal length, different sdt transgenes were expressed in sdt K85 mutant PRCs using Rh1-GAL4. The resulting eyes were analysed by electron microscopy. Around 20 ommatidia from two to four eyes of different individuals with the same genotype were photographed, and stalk-membrane length of R1-R6 was measured using ImageJ software (http://rsb.info.nih.gov/ij/). In PRCs mutant for sdt K85 , stalkmembrane length was reduced by more than 50% compared with wild type. Sdt-A or Sdt-B2 expression partially restored stalkmembrane length to a reduction of only 30% compared with wild type (Fig. 9). This stalk was significantly longer than that in PRCs mutant for sdt K85 , but significantly shorter than stalk-membrane length in control PRCs. All other sdt transgenes tested failed to rescue stalk-membrane length (Fig. 9). These data suggest that stalkmembrane length depends on the amount of Crb at this site, and is independent of PATJ or Lin-7 levels. High amounts of Crb (as in wild type) result in a long stalk membrane. No Crb (crb, sdt, PATJ mutants), or very low amounts of Crb (expression of Sdt-ΔL27C on a sdt mutant background, Fig. 6B), reduce stalk-membrane length. And, medium amounts of Crb at the stalk membrane (expression of Sdt-A and Sdt-B2 in a sdt mutant background, Fig.  6B, Fig. 8B-B'') result in stalk membranes of medium length.

Subcellular targeting of Sdt is differentially regulated in pupal and adult PRCs
Data presented here corroborate the view that distinct mechanisms control localisation of the Crb-Sdt complex in PRCs at different developmental stages (summarised in Fig. 10). This conclusion is further supported by the observation that a truncated PATJ protein, consisting of only L27 and the first PDZ domain, is localised correctly during the first half of pupal development, but is delocalised in adult PRCs (Nam and Choi, 2006; Journal of Cell Science 121 (12) 2006a). The stability of the complex at pupal stages seems to depend only on Crb. In pupae, all core components of the complex are mislocalised in crb-mutant PRCs (Bachmann et al., 2008;Richard et al., 2006a) (and data not shown), whereas the absence of sdt, PATJ or Lin-7 does not affect apical localisation of the others. Accordingly, Sdt localisation at this stage only depends on its PDZ domain that binds the cytoplasmic tail of Crb. Neither the noncanonical L27N domain of Sdt, which is responsible for binding PATJ as shown here, nor the other protein-protein interaction domains are required for Sdt localisation in pupal PRCs.
In the adult Drosophila eye, localisation of Crb-Sdt-complex core proteins to the stalk membrane is mutually dependent, with the exception of Lin-7, which is not required to localise other components (Bachmann et al., 2008;Berger et al., 2007;Richard et al., 2006a). Similarly, in zebrafish the levels of the Crb orthologous proteins require the function of the Sdt orthologue Nagie oko (Hsu et al., 2006). In the fly eye, changes observed at different developmental stages point to a transition in the mechanisms regulating the building and stability of the complex. As previously pointed out, this transition occurs gradually in the second half of pupal development (Richard et al., 2006a). At the same time, Bazooka, which is associated with the adherens junctions in the first half of pupal development, accumulates in the cytoplasm (Hong et al., 2003). The transition also correlates with the formation of stalk membrane, which initiates around 55% pupal development and ultimately separates the apical plasma domain into two distinct compartments (Longley and Ready, 1995). This process seems to require additional, more complex control mechanisms, as reflected by the fact that several Sdt domains are required for its proper localisation at later stages. It is very possible that other, yet unknown components contribute to the stability and/or restriction of Sdt at the stalk membrane.
Results presented here also suggest that in the adult Drosophila eye, localisation of Sdt occurs in several steps that rely on different domains. In the first step, Sdt is brought close to the apical membrane. This function is mediated by the N-terminus, including the two ECR domains and the N-terminal L27 domain. Since Par-6, a known binding partner of the ECR motifs, is localised basolaterally in adult PRCs (N.A.B., unpublished data), PATJ binding is more likely to be crucial for apical recruitment of Sdt. In Fig. 9. Rescue of stalk-membrane length by Sdt-encoding transgenes. Stalkmembrane length (mean ± s.e.m.) in sdt K85 (n=115), sdt K85 + Sdt-ΔSH3 (n=123), sdt K85 + Sdt-ΔPDZ (n=107), sdt K85 + Sdt-ΔGUK (n=118), sdt K85 + Sdt-ΔL27C (n=127), sdt K85 + Sdt-B2 (n=118), sdt K85 + Sdt-A (n=124) and wild-type (n=117) PRCs. One unit represents 1 μm. *P<0.0001 or + P<0.0001 (two-tailed test), significantly different compared with wild type or sdt K85 , respectively. fact, no localised Sdt is detected in PATJ-mutant adult PRCs (Richard et al., 2006a). In the absence of all other domains besides the N-terminus (with the exception of L27C), Sdt proteins accumulate at the rhabdomere base, a specialised region that seems to have an important role in PRCs. Many proteins involved in morphogenesis, phototransduction or endocytosis, such as Drosophila moesin, TRPL (transient receptor potential-like) and Rab11 (Cronin et al., 2006;Karagiosis and Ready, 2004;Satoh et al., 2005), to mention just a few, are enriched there. The final step, recruitment of Sdt to the stalk membrane, requires the PDZ-, the SH3-and the GUK-domain. Whereas the PDZ-domain binds Crb, no binding partners for the SH3-and the GUK-domain are known. It was shown that these two domains can bind each other in vitro (Ö.K., unpublished data). Similar interactions between corresponding domains of the human MAGUK CASK were reported to occur either intramolecularly or intermolecularly between the GUK domain of human CASK and the SH3 domain of hDLG (Nix et al., 2000). In the MAGUK PSD-93, binding of a ligand to the PDZ domain releases intramolecular inhibition of the GUK domain by the SH3 domain (Brenman et al., 1998). This possible complexity currently does not distinguish whether the failure to recruit Sdt to the stalk membrane upon removal of one of these domains is due to either the lack of binding additional partner(s) or the lack of intramolecular interactions, or both.

Par-6 apical localisation in pupal PRCs requires the Nterminus of Sdt
Whereas Sdt is not required to restrict components of the Crb-Sdt complex to the apical membrane in pupal PRCs, the apical localisation of Par-6, a member of the Par-protein network, depends on Sdt at this developmental stage (Berger et al., 2007;Nam and Choi, 2003) (this work). Recently, several studies suggested a direct interaction between the Crb-Sdt and the PAR complex, but the proposed interactions differ with respect to the partners mediating the link. Results obtained from in vitro analysis have suggested a number of interactions: aPKC with both PATJ and the intracellular domain of Crb (Sotillos et al., 2004); the PDZ domain of Par-6 with either the N-terminus of Sdt and/or Pals1 or the C-terminus of CRB1 or CRB3 (Hurd et al., 2003;Kempkens et al., 2006;Lemmers et al., 2004;Wang et al., 2004); and the N-terminus of Par-6 with the third PDZ domain of PATJ (Nam and Choi, 2003). The observations that neither Crb nor PATJ localisation is affected in sdt-mutant pupal PRCs (Berger et al., 2007) (this work) and that expression of Sdt-B2 in sdt-mutant PRCs completely restores Par-6 apical localisation, strongly suggests that in pupal PRCs the interaction between the Crb complex and Par-6 is mediated by the ECR motifs of Sdt. Sdt-A, which carries an additional 433 amino-acid-long stretch between ECR1 and ECR2, only partially restored apical recruitment of Par-6, suggesting that separation of ECR1 from ECR2 interferes with efficient interactions between the two proteins.
The role of Sdt in organising and stabilising the Crb complex and stalk-membrane length in adult photoreceptor cells Our results show that in adult PRCs, sdt controls localisation and stability of Crb, PATJ and Lin-7 but the mechanisms differ. Whenever a Sdt protein is expressed that contains binding domains for PATJ or Lin-7, the amount of the latter is, independently of localisation, restored to wild-type levels. By contrast, Crb protein is stabilised only when Sdt is associated with the stalk membrane (expression of Sdt-A, Sdt-B2, Sdt-ΔL27C and Sdt-ΔN). Interestingly, none of the constructs used, including the two fulllength variants, rescued Crb protein to wild-type levels. One possible explanation is that other, yet uncharacterised Sdt isoforms are expressed in the eye, which, together with Sdt-B2 and/or unknown interaction partners of the Crb-Sdt complex, regulate the amount of Crb at the stalk membrane. Additional Sdt isoforms are predicted by Flybase (http://flybase.bio.indiana.edu) to exist. They mainly differ from the known forms in their N-termini, which suggests alternative interaction partners.
One striking phenotype observed in PRCs mutant for crb, sdt or PATJ is the reduction of stalk-membrane length (Berger et al., 2007;Johnson et al., 2002;Nam and Choi, 2003;Pellikka et al., 2002;Richard et al., 2006a). This raises questions about how the Crb-Sdt complex regulates the size of this distinct apical membrane compartment. Our results provide evidence that the amount of Crb protein is a crucial determinant of stalk-membrane length. This agrees with the observation that Crb overexpression increases stalkmembrane length (Pellikka et al., 2002). Interestingly, these authors showed that overexpression of a Crb protein that lacks the cytoplasmic domain and, hence, the binding site for Sdt, is sufficient to cause this increase. This suggests that either the transmembrane and/or extracellular domain of Crb regulates stalk-membrane growth. Sdt contributes to the stabilisation of Crb at the stalk and, hence, is indirectly involved in the control of stalk-membrane length. It will be interesting to explore the mechanism by which Crb regulates stalk-membrane length.
We thank Andreas Wodarz and the Developmental Studies Hybridoma Bank for antibodies, Michaela Rentsch for sections and Laurel Rohde for the critical reading of the manuscript. The work was supported by grants from the Deutsche Forschungsgemeinschaft (Kn250/21-1) and the EC (QLG3-CT-2002-01266).

Note added in proof
Similar to our findings, Bit-Avragim and colleagues have evidence that Nagie oko, the zebrafish Stardust ortholog, mediates the assembly of alternative multi-protein compositions of the Crumbs-Nagie oko and Par6-aPKC protein complexes in a highly tissuespecific manner (Bit-Avragim et al., 2008).