Drosophila melanogaster Guk-holder interacts with the Scribbled PDZ1 domain and regulates epithelial development with Scribbled and Discs Large

Epithelial cell polarity is controlled by components of the Scribble polarity module, and its regulation is critical for tissue architecture and cell proliferation and migration. In Drosophila melanogaster, the adaptor protein Guk-holder (Gukh) binds to the Scribbled (Scrib) and Discs Large (Dlg) components of the Scribble polarity module and plays an important role in the formation of neuromuscular junctions. However, Gukh's role in epithelial tissue formation and the molecular basis for the Scrib–Gukh interaction remain to be defined. We now show using isothermal titration calorimetry that the Scrib PDZ1 domain is the major site for an interaction with Gukh. Furthermore, we defined the structural basis of this interaction by determining the crystal structure of the Scrib PDZ1–Gukh complex. The C-terminal PDZ-binding motif of Gukh is located in the canonical ligand-binding groove of Scrib PDZ1 and utilizes an unusually extensive network of hydrogen bonds and ionic interactions to enable binding to PDZ1 with high affinity. We next examined the role of Gukh along with those of Scrib and Dlg in Drosophila epithelial tissues and found that Gukh is expressed in larval-wing and eye-epithelial tissues and co-localizes with Scrib and Dlg at the apical cell cortex. Importantly, we show that Gukh functions with Scrib and Dlg in the development of Drosophila epithelial tissues, with depletion of Gukh enhancing the eye- and wing-tissue defects caused by Scrib or Dlg depletion. Overall, our findings reveal that Scrib's PDZ1 domain functions in the interaction with Gukh and that the Scrib–Gukh interaction has a key role in epithelial tissue development in Drosophila.

Cell polarity is a key property of tissue development and manifests itself as the asymmetric organization of cellular components, such as proteins and lipids into distinct cellular domains. Correct establishment of cell polarity is important for tissue architecture, cell proliferation, cell migration, and cellular fate, and its dysregulation has been recognized as a cancer hallmark, with ϳ70% of epithelial cancers displaying defects in polarity regulation (1)(2)(3)(4). Cell polarity is regulated by the interplay between three key polarity modules, Scribble, PARtitioning-defective (PAR), 5 and Crumbs (5). Epithelial cell polarity (apico-basal polarity) is established and maintained by the antagonistic interactions between the Scribble module and the PAR and Crumbs complexes, resulting in the restriction of PAR and Crumbs complex components to the apical cortex and the Scribble module components to the basolateral cortex. In apico-basal cell polarity, the three polarity modules function to specify the apical and basal membrane domains and to position the adherens and tight junctions, which are required for cellcell contact, cell communication, epithelial tissue coherence, and tissue growth regulation (6).
The Scribble module comprises three tumor suppressor proteins, Scribbled (Scrib), Disc large (Dlg), and Lethal-2-giant larvae (Lgl), which are highly conserved in structure and function from the vinegar fly, Drosophila melanogaster, to humans (7). Genetic analyses in Drosophila have provided valuable insight into their in vivo function (8,9) and revealed that Scrib, Dlg, and Lgl, in addition to their role in cell polarity, are also involved in cell proliferation, differentiation, and migration (8 -12). In addition to the core Scribble module components, studies in Drosophila neuromuscular junctions revealed that the interaction between Scrib and Dlg is mediated by an adaptor protein, termed GUK-holder (Gukh) (13). However, at a molecular level the precise interactions between these three tumor suppressors, and indeed the role of Gukh, are not well-defined.
Scrib is a scaffold protein belonging to the LAP (LRR and PDZ) protein family and contains 16 Leucine-rich repeats (LRRs), two LAP-specific domains (LAPSADa and LAPSADb), and four PDZ domains. Whereas the N-terminal LRR domain is critical for Scrib's cortical localization (14), the four PDZ domains are required for cell-cell junction localization (13) and are the major mediators of Scrib interactions with other proteins (15). hScrib (human Scrib) PDZ domains have been shown to bind a diverse set of cellular interactors, including ␤-PIX, MCC, GIT1, and ␤-catenin (16 -20), enabling Scrib to integrate a range of cellular cues for the establishment of apicobasal polarity, cell migration, and cell signaling.
Gukh was identified in a yeast two-hybrid screen as a protein that bound to the Drosophila Dlg GUK domain (13). Two orthologs of Drosophila Gukh have been identified in humans, Nance-Horan syndrome (NHS) and NHSL1 (23). NHS is an uncommon X-linked disorder characterized by congenital nuclear cataracts, dental irregularities, and craniofacial dysmorphisms, with mental deficiencies also occurring in ϳ30% of the cases (24). The molecular details of how NHS mutations cause the NHS are unclear, although recent studies have implicated it in regulation of epithelial junctions (25) as well as actin remodeling (23,26).
Drosophila Gukh contains a GUK-holding domain at its C terminus, which directly binds the Dlg GUK domain (13,27). Additionally, another study showed that the binding of Gukh to the GUK domain of Dlg occurs in a mutually exclusive manner via the PDZ domain, only permitting Gukh interaction when the PDZ remains unbound (28). However, others have reported that interactions between Gukh and Dlg require the SH3-GUK domain of Dlg (29,30). This interaction is regulated via interdomain interactions of PDZ3-SH3-GUK via a PDZ3-binding motif in a linker region enabling dynamic regulation of ligand binding to Dlg PDZ3 (29,30).
Furthermore, in yeast two-hybrid assays the Gukh C-terminal region is able to engage Scrib PDZ2 but not PDZ3-4 (13). Moreover, co-immunoprecipitation analysis from Drosophila larval muscles showed that Scrib can form a complex with Dlg and Gukh, and the interaction with Dlg is reduced in a Gukh mutant (13). These genetic data suggest that Gukh is important for the formation of a ternary complex between Scrib, Dlg, and Gukh. Consistent with this notion, all three proteins co-localize at Drosophila neuromuscular junctions, and Dlg and Gukh are required for the correct localization of Scrib (13). This interaction is likely to be evolutionarily conserved, as a Zebrafish Gukh ortholog, Nhs1b, also physically interacts with Scrib and Dlg in cultured cells, and nhs1b and scrib genetically interact as shown by the observation that heterozygotes show a strong neural cell migration defect relative to single heterozygotes (31).
Although Drosophila Gukh interacts with Dlg and Scrib in neuromuscular junctions (13), its role in epithelial tissue formation and the molecular basis for the Scrib-Gukh interaction remain to be defined. Here, we identify the Scrib PDZ1 domain as the major interacting PDZ domain with the Gukh C-terminal peptide. Using X-ray crystallography, we then define the structural basis of Scrib PDZ1 interactions with the Gukh C-terminal peptide. Furthermore, our studies reveal a novel role for Gukh in epithelial development. We show that Gukh is expressed in larval-wing and eye-epithelial tissue and co-localizes with Scrib and Dlg at the apical cell cortex. Importantly, we show that Gukh functions with Scrib and Dlg in Drosophila epithelial tissues, with depletion of Gukh enhancing the eye-and wing-tissue defects caused by Scrib or Dlg depletion. These findings provide the first evidence for a role for Gukh in the Scribble module in the control of epithelial cell polarity and provide structural and mechanistic insights into the Scrib-Gukh interaction in Drosophila.

Molecular and structural basis of Scrib-Gukh interaction
To understand the molecular basis of the reported interaction between Scrib and Gukh, we performed protein-ligand interaction studies using individual recombinant PDZ domains from Scrib and C-terminal peptides derived from Gukh, and we determined binding affinities using isothermal titration calorimetry (ITC) ( Fig. 1 and Table 1). ITC experiments were conducted using purified Drosophila Scrib PDZ1, PDZ2, PDZ3, and PDZ4 domains with wildtype Gukh C-terminal peptides. Raw heats of titrations obtained for PDZ1 with Gukh WT peptide revealed a tight interaction with a calculated K D of 664 nM, whereas PDZ3 engaged Gukh WT peptide with only modest affinity with a K D of 27.8 M. In contrast, PDZ2 and PDZ4 did not show any detectable binding to Gukh WT peptide, thus rendering the Scrib PDZ1 domain as the primary functional interaction site for the Gukh C terminus.
We then examined the structural basis of the PDZ1-Gukh interaction by determining the crystal structure of a Scrib PDZ1-Gukh C-terminal peptide complex (Table 2). Drosophila Scrib PDZ1 adopts the typical PDZ fold consisting of six ␤-strands and two ␣-helices that form a ␤-sandwich structure ( Fig. 2A) and is highly similar to the previously determined human PDZ1 (hsPDZ1) structure (PDB code 2W4F). The PDZ1 domain from the PDZ1-Gukh complex superimposes with the apo hsPDZ1 domain with an r.m.s.d. of 0.8995 Å over Crystal structure of Scrib PDZ1-Gukh complex 92 C␣ atoms, indicating that the binding of the Gukh peptide to the PDZ1 domain does not substantially change the overall fold (Fig. 2B).
In the PDZ1-Gukh structure, PDZ1 features an atypical ␤5 displaying increased flexibility in its geometry. PDZ1 interacts with Gukh peptide via its canonical ligand-binding groove formed by the ␤2 strand and helix ␣2 (Fig. 2, A and C). Gukh binding to PDZ1 buries a combined total of 937 Å 2 of solventaccessible surface area, with the interface having a shape complementarity score of 0.78, indicating a very good fit. In the complex, the Gukh peptide faces the ␤2 strand in an anti-parallel manner, with its N-terminal solvent exposed and its C terminus stabilized by the PDZ1 ␤1-2 loop. The complex is achieved via an extensive network of hydrogen bonds formed by Leu-741 PDZ1 -Leu-1788 Gukh , Leu-743 PDZ1 -Leu-1788 Gukh ,
To validate the crystal structure, we performed site-directed mutagenesis to target two key interactions, Arg-765 PDZ1 -Glu-1785 Gukh and His-796 PDZ1 -Phe-1784 Gukh , as well as the floor of the canonical PBM-binding site via G747W (Table 3). Mutation of Arg-765 to an Ala resulted in a 5-fold reduction in affinity (K D ϭ 5.1 M), whereas the H796A displayed no affinity for WT Gukh, suggesting that both residues make important contributions for Scrib PDZ1-Gukh interactions. Furthermore, introduction of the large hydrophobic Trp at Gly-747 PDZ1 also completely ablates the ability of PDZ1 to bind to the PBM of Gukh (Table 3).

Gukh is expressed in wing and eye-epithelial tissue and colocalizes with Dlg and Scrib in the apical cortex
Considering the previously established role for the Scrib-Gukh interaction in neuromuscular junction development, we next examined whether Scrib and Gukh are involved in epithelial tissue development. To determine whether Gukh was expressed in Drosophila larval epithelial tissues, we stained developing eye and wing epithelia from third instar larvae with an anti-Gukh antibody (13). Staining with the anti-Gukh antibody, and using expression of a gukh RNAi line in the wing epithelium or clones of a gukh loss-of-function P transposableelement mutant in the eye epithelium as controls, revealed that Gukh was expressed in the developing wing ( Fig. 3, A, quantified in C) and eye (Fig. 3, B, quantified in D) tissues. We then examined the cellular distribution of Gukh relative to Scrib and Dlg (Fig. 4). In the larval-wing epithelium, Gukh was localized cortically and enriched apically, co-localizing with Dlg and Scrib in the apical cortex; however, cross-sections revealed that Gukh extended further apically and basolaterally than did Dlg and Scrib (Fig. 4A). A similar co-localization between Gukh and Scrib or Dlg was observed in the larval eye epithelia, although Gukh exhibited a stronger staining of the apical cortex of photoreceptor cells, which are enriched in F-actin (Fig. 4B). Thus, Gukh shows overlapping localization with Scrib and Dlg in two larval epithelial tissues, although Gukh is also distributed more generally around the cell cortex, which is consistent with Gukh possessing a WH1 F-actin-binding domain at its N terminus (13,26).

Gukh genetically interacts with Scrib and Dlg in epithelial tissues
Previous studies revealed that the Gukh GUK-holder domain binds to the Dlg GUK domain and that the C-terminal PDZbinding motif of Gukh interacted with Scrib PDZ domains (13). However, whether Gukh has a functional role with Scrib or Dlg in epithelial tissue is currently unknown. Consequently, we examined the genetic interaction between scrib and gukh in Drosophila eye and wing epithelial tissues. To manipulate the expression of several transgenes in a tissue-specific manner, the UAS/GAL4 system was used to selectively knock down scrib, dlg, or both genes using UAS-RNAi lines in the Drosophila eye or wing epithelial tissues, using the ey-GAL4 or dpp-GAL4 drivers, respectively. Genetic interactions with gukh expression/ function knockdown were then examined relative to a control transgene (UAS-lacZ or UAS-GFP, to control for UAS element number).
The scrib knockdown using two RNAi transgenes (one on the 2nd chromosome and the other on the 3rd chromosome) via the ey-GAL4 driver in the Drosophila developing eye resulted in a slightly reduced and disorganized eye phenotype (termed a rough eye phenotype, Fig. 5A) relative to control adult eyes (Fig.  5B). To verify that this phenotype was modifiable, we knocked down Dlg using RNAi, which showed, as expected, a robust enhancement of the eye roughness (Fig. 5A) and a reduction in eye size (Fig. 5F). Importantly, knockdown of Gukh using RNAi enhanced the Scrib knockdown rough eye phenotype (Fig. 5A) and led to a slight reduction in eye size (although this was not statistically significant below p Ͻ 0.05). We then examined the effect on the Scrib knockdown eye phenotype upon expression of the Gukh-C-terminal region (Gukh-C), which lacks the important F-actin-binding WH1/EVH1 domain at the N terminus, and it is expected to act in a dominant-negative manner (13). When gukh-C was overexpressed with concurrent scrib knockdown, an enhancement of the eye roughness was also observed (Fig. 5A). In contrast, individual overexpression of

Crystal structure of Scrib PDZ1-Gukh complex
gukh-C or gukh RNAi expression showed normal eye size and arrangement of ommatidia relative to the wildtype; however, dlg RNAi showed a slight decrease in eye size relative to the other genotypes (Fig. 5B). Altogether, these results reveal a genetic interaction between scrib and gukh in the eye, although not as strong as that observed between scrib and dlg. We next examined whether gukh genetically interacts with dlg. Because Dlg knockdown via the ey driver did not pro- Å resolution) bound to Gukh peptide (cyan) and human PDZ1 (green, residues 724 -819, 1.91 Å resolution). C, simulated anneal composite omit electron density map encompassing the binding groove of Scrib PDZ1 in complex with Gukh. PDZ1 is shown as light pink sticks, and Gukh is shown as cyan sticks. The electron density map is shown as a blue mesh contoured at 1.0 and was calculated by omitting the entire Gukh peptide. D, cartoon representation of Scrib PDZ1-Gukh complex (light pink and cyan) with the Erbin PDZ domain bound to a synthetic p120-like peptide (dark blue and orange, PDB code 1N7T). Residue Trp-4 from p120-like peptide engages a pocket on the ␤2-3 loop and is critical for the high-affinity interaction with Erbin. Gukh does not utilize a Trp residue to achieve a high-affinity interaction, and instead it utilizes Phe-1784 for hydrophobic interactions with Scrib PDZ1 His-796 on the opposite side of the ligand-binding groove.
Crystal structure of Scrib PDZ1-Gukh complex duce a strong phenotype (Fig. 5B), we examined whether Gukh knockdown could modify the small rough eye phenotype observed upon Scrib and Dlg co-knockdown (Fig. 5, C and G). Strikingly, both gukh RNAi and gukh-C expression resulted in a strong enhancement of the small rough eye phenotypes of eyϾscrib RNAi dlg RNAi (Fig. 5, C and G) that were statistically significant (p ϭ 0.001 and 0.0003, respectively). Thus, Gukh genetically interacts with Dlg and Scrib in the eye, suggesting that Scrib, Dlg, and Gukh function in the same genetic process in eye-epithelial development.
To extend these results, we used the dpp-GAL4 driver to knock down scrib or dlg in another epithelial tissue, the wing epithelium, and we examined the interactions with gukh RNAi and gukh-C. The Dpp driver is expressed along the anteriorposterior boundary in the developing wing, which constitutes the region between the 3rd and 4th wing vein of the adult wing. Expression of gukh-RNAi or gukh-C via the dpp driver did not affect the wing phenotype (Fig. 5D), but dppϾscrib RNAi resulted in a severe wing phenotype and less than 6% survival (data not shown); therefore, the interaction with gukh could not be examined. However, dppϾdlg RNAi resulted in a normal wing phenotype, except for the truncation of the 3rd wing vein in 25% of cases (Fig. 5E). Importantly, knockdown of gukh or overexpression of gukh-C together with dlg RNAi resulted in a truncation of the 3rd wing vein in 57 and 67% (respectively) of flies examined. This was greater than a 2-fold increase in comparison with the 25% seen in the GFP control (Fig. 5, E and H). Thus, gukh genetically interacts with dlg in the wing epithelium, suggesting that Gukh and Dlg function in a common genetic pathway in wing development.

Discussion
Epithelial tissues are highly polarized, and the Scribble polarity module is a critical regulator of epithelial tissue organization and polarization in Drosophila and in mammals. In addition to the core components of the Scribble module, Scrib, Dlg, and Lgl, an important regulatory role has been emerging for the adaptor protein Gukh. Notably, in Drosophila neuromuscular junctions (13) Gukh is a crucial adaptor protein that enables assembly of a functional ternary complex of Scrib, Dlg, and Gukh, thereby allowing correct synaptic localization of Scrib. However, neither the molecular basis for Scrib-Gukh interactions nor a role for Gukh in epithelial tissue structure or function has previously been described. We now show that the Scrib PDZ1 domain is the major high-affinity interaction site for Gukh, and we define the molecular basis for this interaction. Furthermore, we now provide the first description of Gukh expression and function in Drosophila larval epithelial tissues.
We show that Gukh is expressed in the larval-wing and eyeepithelial tissues and that Gukh is generally cortically localized and overlaps with Scrib and Dlg at the apical cell cortex. Impor-  In en-driven gukh RNAi discs down-regulation of Gukh staining was observed compared with the uniform staining seen in the control disc. B, generation of ey-FLP clones in otherwise wildtype imaginal eye discs by crossing gukh BG02660 FRT82B flies to ey-FLP FRT82B Ubi-GFP at 25°C. The absence of GFP represents patches of mutant tissue and reveals down-regulation of Gukh staining. C, quantification of the intensity ratio in the wing samples (from A), which is measured by the intensity of Gukh staining in GFP domains compared with the non-GFP domain. There is a significant decrease (by ϳ25%) in the intensity ratio when gukh RNAi is expressed (n ϭ 14 -18 wing discs). D, quantification of the intensity ratio in the eye samples (from B), showing that the intensity ratio of Gukh staining in non-GFP mutant tissue compared with GFP-marked tissue was significantly decreased by ϳ50% (n ϭ 10 -18 eye discs). Images were taken at ϫ40 (A) and ϫ20 (B) magnification. Scale bars, 50 m. Error bars represent mean Ϯ S.D. Student's t test used to test for significance. ****, p Յ 0.0001.

Crystal structure of Scrib PDZ1-Gukh complex
tantly, our studies have revealed that Gukh together with Scrib and Dlg are key mediators of epithelial tissue development in Drosophila, with loss of Gukh in combination with loss of Scrib and Dlg leading to morphological and differentiation defects in eye and wing tissues.

Expression and localization of Gukh in epithelial tissue
Our expression analysis of Gukh protein in the eye and wing epithelium revealed co-localization with Scrib and Dlg at the apical cortex; however, Gukh was also distributed more apically as well as basolaterally around the cell cortex, and in the differentiated region of the eye epithelium strong staining was observed in the apical region of the photoreceptor cells, where F-actin accumulates. Because Gukh/NHS1, via its WH1 domain, regulates the WAVE/SCAR-ARP2/3-branched F-actin pathway (26,31), Gukh's general cortical localization might be commensurate with this role in F-actin biogenesis, which conceivably might also function independently of its function with Dlg and Scrib. However, our genetic data in Drosophila shows that expression of Gukh-C (which can bind to Dlg and Scrib but lacks the F-actin regulatory WH1 domain) phenocopies knockdown of gukh in its interaction with scrib and dlg in the eye and wing. This suggests that the Gukh WH1 domain and regulation of F-actin is essential for Scribble module function in epithelial development.

Function of Gukh in larval epithelial tissues
In the eye, knockdown of Scrib or Scrib with Dlg via the ey driver gave rise to tissue growth and patterning defects, most likely due to the disruption of epithelial cell polarity and the deregulation of the Hippo-and JNK-signaling pathways (14,32). The enhancement of the Scrib and Scrib Dlg phenotypes by impairment of Gukh function (via RNAi or the dominantnegative transgene) suggests that Gukh functions in the same genetic pathway as Scrib and Dlg in epithelial cell polarity and cell signaling. The more robust enhancement observed by Gukh impairment in the background of Scrib and Dlg knockdown is consistent with the proteins functioning in a tripartite complex in epithelial tissues, as has been shown in the neuromuscular junctions (13). Indeed, Gukh and Dlg are both required for Scrib localization in the neuromuscular junctions, with greater Scrib mislocalization being observed in the double mutant (13). It remains to be determined whether this co-regulation of Scrib localization by Dlg and Gukh also occurs in epithelial tissues; however, this function is consistent with the stronger enhancement of the Scrib knockdown phenotype by dual knockdown of Dlg and Gukh.
In the wing, knockdown of Gukh together with Dlg resulted in a pronounced truncation of the 3rd wing vein. Because EGFR-Ras signaling is a key pathway involved in wing vein formation (33), and Scrib has been previously shown to repress this pathway (18,34), we first considered whether Gukh might also function together with Dlg and Scrib in regulating EGFR-Ras signaling. However, because knockdown of Scrib function would be expected to increase Ras signaling, and enhanced Ras signaling is associated with ectopic wing veins (35), which was not observed in dlg gukh impaired wing tissue, it is unlikely that the Ras pathway is up-regulated by Dlg knockdown in this context. Conversely, knockdown of Dpp (transforming growth factor-␤/BMP) signaling results in truncated wing veins (36), and therefore the generation of a similar phenotype by dlg and gukh impairment suggests Dlg and Gukh might positively regulate the Dpp pathway in this context. Consistent with this notion, Scrib via its LRR domain has been shown to bind to the BMP type I receptor, Tkv, type II receptor, Pnt, and the phosphorylated (active) Mad transcription factor in Dro- Crystal structure of Scrib PDZ1-Gukh complex sophila wing posterior cross vein development, which is thought to facilitate BMP receptor signaling (37,38). Additionally, the Scribble module protein, Lgl, has been implicated in the regulation of Dpp secretion in embryonic ectodermal cells (37,38). Further studies are required to determine whether Dlg, Scrib, and Gukh interact to affect Dpp signaling during wing development.

Molecular interaction of Gukh with Scrib
To establish a molecular basis for the observed genetic interaction between Gukh and Scrib, we performed a biochemical analysis of individual PDZ domains of Drosophila Scrib with peptides encoding for the C-terminal PDZ-binding motif in Gukh. Our binding data demonstrated that the major site of Crystal structure of Scrib PDZ1-Gukh complex interaction between Gukh and Scrib is the Scrib PDZ1 domain, thus for the first time establishing a definitive molecular basis for this interaction. These findings are in contrast to that of Mathew et al. (13), who showed that Scrib PDZ2 but not PDZ3 and PDZ4 had strong interaction with the Gukh C-terminal peptide in yeast two-hybrid assays; however, they were unable to draw conclusions regarding PDZ1 in their experiments due to high background activity with these constructs (13). Thus, our findings have revealed a previously unexplored role for Scrib PDZ1 in binding to Gukh. However, in contrast to the Mathew et al. report (13), our study did not reveal a role for Scrib PDZ2 in binding the Gukh C-terminal peptide, which might be due to differences in construct design, post-translational modification in the yeast system, or to the inherent propensity for false positives using the yeast two-hybrid system, due to heterologous protein expression and inappropriate cell localization. Importantly, our findings provide a rationale for the observation that the Scrib truncation mutant scrib 5 (39), which results in the loss of Scrib PDZ3 and PDZ4 domains, is functionally active and displays normal adherens junction and basolateral/septate junction (SJ) formation, whereas the scrib 4 mutant, which lacks active PDZ domains, results in disrupted SJ formation. These data, together with our results, indicate that the presence of an intact and active PDZ1 domain in Scrib should be necessary and sufficient for normal adherens junction and septate junction formation. Moreover, because we have shown that Gukh interacts with Scrib PDZ1, and Gukh genetically interacts with Scrib and Dlg in epithelial tissues, this suggests that Gukh could contribute to the role of Scrib-PDZ1 in epithelial structure as well as potentially in directed epithelial cell migration (39).
Interestingly, our measurements of Scrib PDZ-Gukh interactions identified the PDZ1-Gukh interaction as unusually tight for PDZ domain interactions with an endogenous ligand, with tight nanomolar interactions typically found in PDZ-synthetic ligand complexes such as those of Erbin or ZO-1 interactions with peptides derived from phage display. To better understand the structural basis for this, we determined the crystal structure of Scrib PDZ1-Gukh. Superimposition of the PDZ1 domain from Drosophila bound to Gukh over the human Scrib PDZ1 domain (PDZ 2W4F) reveals no significant structural changes in the ligand-binding groove upon Gukh binding (Fig. 2B), which has previously been observed for the GRIP1 PDZ6 -peptide complex (40). Examination of the interface of the PDZ1-Gukh complex revealed an extensive net of hydrogen bonds and ionic interactions, which supplement the insertion of the C-terminal Gukh Leu-1788 into a hydrophobic pocket. In particular, the ionic interaction between Gukh Glu-1785 and Scrib Arg-765 is reminiscent of Erbin Atg-49 -p120catenin-like peptide E4 (41). Loss of this ionic interaction in a Scrib R765A mutant leads to an ϳ8-fold loss of Gukh binding to Scrib, suggesting that the Arg-765 PDZ1 -Glu-1785 Gukh salt bridge is important. Interestingly, despite displaying nanomolar affinity, Gukh does not contain any Trp residues in its PDZbinding motif. In the case of Erbin, a Trp in the Ϫ1 position has been shown to be important for binding, with a second Trp in position Ϫ4 position being critical for the high-affinity interaction by engaging the ␤2-3 loop (Fig. 2D). A similar key role is played by a Trp in the Ϫ6 position of a synthetic peptide in complex with ZO1 (42,43), which has been shown to contribute substantially to the binding of ZO1 by inserting into the ␤2-3 loop in a similar location as the Ϫ4 Trp in the Erbin complex (43). In contrast, Gukh harbors a Phe in the Ϫ4 position rather than a Trp, which nonetheless makes significant contact with the Scrib PDZ1 domain via -stacking with His-796. Thus, the Ϫ4 position in Gukh is still able to contribute substantially to binding to PDZ1 by exploiting the opposite side of the ligand-binding groove. Indeed, mutation of His-796 to an Ala abrogates binding to Gukh, supporting the notion that the His-796 PDZ -Phe-1784 Gukh -stacking is important for the Scrib-Gukh interaction. Furthermore, Gukh only forms a single hydrogen bond with the ␤2-3 loop via the main chain carbonyl of Pro-1782, indicating that the engagement of the ␤2-3 loop is not necessary for a high affinity nanomolar interaction.
In conclusion, our study has revealed novel roles and regulatory mechanisms for Gukh in epithelial development. Our discovery of a novel role for Drosophila Scrib PDZ1 in the interaction with Gukh, and the important function of Gukh together with Scrib and Dlg in epithelial tissue morphogenesis and differentiation, increases our understanding of a previously poorly-studied protein. It will now be important to investigate whether the vertebrate Gukh orthologs also interact with Scrib and Dlg in a similar manner in epithelial tissue development.

ITC
Purified Drosophila Scrib PDZ domains were used in titration experiments against 8-mer peptides spanning the C terminus of Drosophila Gukh isoform-A (LPSFETAL, GenScript). Raw heats were measured using a Microcal NanoITC200 system (GE Healthcare) at 25°C. Because of a lack of useful aromatic amino acids in PDZ domain proteins, protein concentrations were calculated using the Scope method (45) by measuring absorbance at 205 and 280 nm using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). As controls, a non-binding Gukh mutant peptide (LPSFEAAA, Gen-Script) and a superpeptide (RSWFETWV, GenScript) engineered to harbor pan-PDZ binding activity (46) as positive control were used. Binding isotherms were analyzed using Origin 7.0E (MicroCal).

dmPDZ1-Gukh complex crystallization and data collection
The complex of dmPDZ1 with Gukh peptide was reconstituted by mixing protein and peptide at a 1:2 molar ratio. The dilute protein complex was then concentrated to 20 mg/ml using a 3-kDa molecular mass cutoff centrifugal concentrator (Millipore), flash-cooled, and stored under liquid nitrogen. Crystallization trials were carried out using 96-well sitting-drop trays (Swissci) and vapor diffusion at 20°C either in-house or at the CSIRO C3 Collaborative Crystallization Centre, Melbourne, Australia. A 0.15-l dmPDZ1-Gukh peptide complex was mixed with 0.15 l of various crystallization conditions using a Phoenix nanodispenser robot (Art Robbins). Commercially available screening kits (PACT Suite and JCSG-plus Screen) were used for the initial crystallization screening, with hit optimization performed using a 96-well plate at the CSIRO C3 Centre. Crystals of dmPDZ1 in complex with Gukh peptide were obtained at 20 mg/ml in 0.2 M zinc acetate dehydrate, 0.1 M sodium cacodylate, pH 6.5, and 10% (v/v) propanol. The crystals were cryo-protected using 30% (v/v) ethylene glycol and flash-cooled at 100 K using liquid nitrogen. Hexagonal rod crystals were obtained belonging to space group P3 2 12. Unfortunately, structural determination failed with those crystals, which were subsequently used in a cross-seeding experiment into the Shotgun screen at CSIRO C3. After 2 months, the crystals of dmPDZ1 in complex with Gukh peptide were obtained at 20 mg/ml in 30% (v/v) PEG 4000, 0.2 M sodium acetate, and 0.1 M Tris chloride, pH 8.5. The crystals were cryo-protected using 30% (v/v) ethylene glycol and flash-cooled at 100 K using liquid nitrogen. All diffraction data were collected on the MX1 beamline at the Australian Synchrotron using ADSC Quantum 315r CCD detector (Area Detector Systems Corp., Poway, CA) with an oscillation range of 1.0°per frame using a wavelength of 0.9537 Å. Diffraction data were integrated using XDSme (47) and scaled using AIMLESS (48). The structure of dmPDZ1-Gukh peptide was solved by molecular replacement using Phaser (49) with the structure of hsPDZ1 (PDB code 5VWC) as a search model. The final TFZ and LLG values were 19.1 and 471, respectively. The solution produced by Phaser was manually rebuilt over multiple cycles using Coot (50) and refined using PHENIX (51). Data collection and refinement statistics details are summarized in Table 1. MolProbity scores were obtained from the MolProbity web server (52). Shape complementarity was calculated using the program SC (53). Coordinate files have been deposited in the Protein Data Bank under the accession code 5WOU. All images were generated using the PyMOL Molecular Graphics System, Version 1.8, Schrödinger, LLC. All software was accessed using the SBGrid suite (54). All raw diffraction images were deposited on the SBGrid Data Bank (55) using accession number 5WOU. Crystal structure of Scrib PDZ1-Gukh complex these proteins has been previously confirmed (11), and we have confirmed that the scrib 101128 RNAi efficiently knocks down the Scrib protein (data not shown). We also confirmed the efficacy of the gukh-RNAi line by testing its ability to decrease Gukh levels when expressed via engrailed-GAL4 in the larvalwing epithelium, and we verified the specificity of the Gukh antibody by showing that Gukh immunoreactively was decreased in gukh BG02660 mutant clones in the eye epithelium (Fig. 3).

D. melanogaster stocks and genetic analysis
Stocks of ey-GAL4 and UAS-scrib RNAi 11663 C2V and UASscrib RNAi 11663 C3S or ey-GAL4 and UAS-dlg RNAi 4689 C2V and UAS-scrib RNAi 11663 C3S were generated and balanced over CyO and TM6B. Stocks of dpp-GAL4 UAS-GFP with UASscrib RNAi 11663 C2V or UAS-dlg RNAi 4689 C2V were generated and maintained at 18°C (where expression of the transgenes is low).
For induction of gukh clones in the eye epithelium, gukh BG02660 FRT82B flies were crossed to ey-FLP FRT82B Ubi-GFP at 25°C, and the eye-antennal discs were dissected from third instar larvae.

Analysis of genetic interactions in the Drosophila eye and wing tissues
For analysis of genetic interactions in Drosophila eyes, crosses of eyϾscrib RNAi (2nd) scrib RNAi (3rd); eyϾdlg RNAi (2nd) scrib RNAi (3rd) to UAS-gukh-C, UAS-gukh RNAi , UAS-dlg RNAi or UAS-lacZ (control) were conducted at 29°C. Crosses of ey-GAL4 to each UAS-RNAi or UAS-transgene were performed as controls. Drosophila adults were collected 7-8 days after crossing at 29°C, and at least 50 progeny were examined from each cross, and photographs were obtained for at least six samples for each genotype. Eye size was measured using Adobe Photoshop Extended tools.
For analysis of genetic interactions in Drosophila wings, dppϾscrib RNAi 11663 C2V (2nd) or dppϾdlg RNAi 4689 C2V (2nd) were crossed to UAS-gukh-C, UAS-gukh RNAi , UASdlg RNAi , or UAS-GFP (control) at 25°C. Crosses of dpp-GAL4 to each UAS-RNAi or UAS-transgene were performed as controls. Scoring was performed based on the presence or absence of the truncated wing vein phenotype. At least 50 individual Drosophila adults were scored from each cross. For imaging, wings from 10 adult flies/sample were carefully detached from the torso and mounted on glass slides using a mixture of methyl salicylate and Canada Balsam (Sigma) in ratio of 1:1 and were left to dry overnight. Microscopy images of Drosophila eyes and wings were obtained at ϫ2.5 and ϫ3.2 magnification, respectively, on an Olympus SZX7 microscope equipped with an INFINITY-1 camera and images were processed using INFIN-ITY capture software.

Immunofluorescent staining, confocal microscopy, and quantification
Primary antibodies were rabbit anti-Gukh antibody (1:500 (V. Budnik) raised to the N terminus of Gukh) and mouse anti-Dlg (Developmental Studies Hybridoma Bank, 4F3). Antibody staining was performed using similar methodology as described previously (11). The secondary antibodies were goat anti-rabbit AlexaFluor-568 (Invitrogen) and goat anti-mouse AlexaFluor-633 (Molecular Probes). Wing and/or eye imaginal discs were mounted onto glass slides in one drop of ProLong GOLD antifade mountant (Molecular Probes, catalog no. P36934) or 80% glycerol in PBS and covered with a glass coverslip. All confocal images were taken on either a confocal Leica TCS (true confocal scanner) SP5 (Leica Microsystems, Germany) or a confocal Zeiss ELYRA (Carl Zeiss, Germany) microscope. Staining intensity and cell migration were quantified using the Fiji (ImageJ) image analysis software. Quantification of Pixel intensity was determined for Gukh staining after RNAi-mediated knockdown (marked by the expression of GFP) or in gukh mutant clones (marked by the absence of GFP expression) versus the wildtype tissue using Fuji software. Adult eye size was determined by drawing a region of interest followed by area measurement (size ϭ 0-infinity pixel units, circularity ϭ 0.00 -0.10). Statistical analysis was conducted with Student's t test using Graphpad Prism, where p Ͻ 0.05 (version 6.00 for Windows, GraphPad Software, La Jolla, CA).