Scribbled Optimizes BMP Signaling through Its Receptor Internalization to the Rab5 Endosome and Promote Robust Epithelial Morphogenesis

Epithelial cells are characterized by apical-basal polarity. Intrinsic factors underlying apical-basal polarity are crucial for tissue homeostasis and have often been identified to be tumor suppressors. Patterning and differentiation of epithelia are key processes of epithelial morphogenesis and are frequently regulated by highly conserved extrinsic factors. However, due to the complexity of morphogenesis, the mechanisms of precise interpretation of signal transduction as well as spatiotemporal control of extrinsic cues during dynamic morphogenesis remain poorly understood. Wing posterior crossvein (PCV) formation in Drosophila serves as a unique model to address how epithelial morphogenesis is regulated by secreted growth factors. Decapentaplegic (Dpp), a conserved bone morphogenetic protein (BMP)-type ligand, is directionally trafficked from longitudinal veins (LVs) into the PCV region for patterning and differentiation. Our data reveal that the basolateral determinant Scribbled (Scrib) is required for PCV formation through optimizing BMP signaling. Scrib regulates BMP-type I receptor Thickveins (Tkv) localization at the basolateral region of PCV cells and subsequently facilitates Tkv internalization to Rab5 endosomes, where Tkv is active. BMP signaling also up-regulates scrib transcription in the pupal wing to form a positive feedback loop. Our data reveal a unique mechanism in which intrinsic polarity genes and extrinsic cues are coupled to promote robust morphogenesis.


Introduction
Epithelial cells are characteristically polarized with apical-basal plasma membrane polarity, which is crucial for tissue homeostasis [1,2]. Many of intrinsic factors underlying apical-basal polarity have been identified to be tumour suppressors [1,2]. Multiple distinct but interacting groups of proteins, including the partitioning defective (PAR) proteins, the Crumb (Crb) complex and the Scribbled (Scrib) complex, play crucial roles in regulating epithelial polarity [1]. Among these, the Scrib complex, composed of Scrib, Discs large 1 (DLG1) and Lethal giant larvae (LGL) [3], has previously been identified as a basolateral determinant of epithelial cells by instructing epithelial-specific cytoskeletal rearrangements, the distribution of apical proteins, and polarized trafficking machinery [3][4][5]. Strong loss-of-function (LOF) of the Scrib complex causes tumor-like growths [4,5].
Decapentaplegic (Dpp), a conserved Drosophila bone morphogenetic protein (BMP)-type ligand, plays crucial roles in various developmental contexts [6]. The Dpp ligands bind to the type I receptor Thickveins (Tkv) and type II receptor Punt. This stimulates Tkv to phosphorylate the transcriptional factor Mad. Upon phosphorylation, Mad binds co-Smad Medea and then translocates to the nucleus for transcriptional regulation of target genes. During pupal stages, Dpp is produced in the longitudinal veins (LVs) of the pupal wing and trafficked into the posterior crossvein (PCV) region from around 18 hr after pupariation (AP) to establish long-range signaling [7]. Prior to dpp transcription in the PCV region at around 26 hr AP [8], long-range BMP/Dpp signaling appears to instruct PCV patterning and differentiation, since either loss-or gain-of-function of Dpp trafficking is sufficient for causing crossveinless or ectopic crossvein phenotypes [9,10]. Dpp trafficking is supported by the BMP binding proteins short-gastrulation (Sog) and crossveinless (Cv) to direct Dpp ligand localization in the PCV region [7]. Since the Dpp trafficking mechanism appears to limit ligand available in the PCV region [7,9], systems that optimize the signaling response to limiting external ligands must be crucial. Thus PCV formation serves as a unique model to address molecular mechanisms underlying the optimization of BMP signaling [6,[11][12][13][14][15].
Epithelial morphogenesis is one of the key processes in animal development and is frequently regulated by highly conserved signaling molecules. To promote robust morphogenesis, ensuring precise interpretation of signal transduction as well as spatiotemporal distribution of extracellular ligands are important. Here, using PCV formation in Drosophila as a model, we demonstrate that the basolateral determinant Scrib complex is required for PCV formation through optimizing BMP signaling. Scrib regulates Tkv localization by facilitating its meeting extracellular ligands, and subsequently is involved in Tkv internalization to Rab5 endosomes. Our data suggest a mechanism that optimizes the signaling response to limiting external ligands for patterning and differentiation of epithelia.

Scrib is required for PCV formation
In a screen for novel components involved in PCV formation, we identified the components of the Scrib complex as candidates. To investigate how Scrib regulates PCV formation, we analyzed the Scrib complex in the Drosophila pupal wing using a conditional knockdown approach [16]. We observed that knockdown of scrib, dlg1 or lgl in the pupal wing blade disrupted PCV formation (Fig 1A-1D and S1A Fig). Notably, knocking down Scrib complex components in this manner did not disrupt cell polarity, as shown by the apical marker aPKC and the basolateral marker DLG1 (Fig 1E and 1F). Thus a partial reduction in Scrib complex gene expression is sufficient to disrupt PCV morphogenesis but not to disrupt apical-basal polarity, indicating that loss of the Scrib complex disrupts PCV formation by another mechanism. To understand whether the PCV-less phenotypes resulted from reduced BMP signaling, phosphorylated Mad (pMad) staining (a readout of BMP signaling [17]) was used to examine the effects of scrib knock down in the PCV region. Knockdown of scrib in the wing caused a loss of BMP signaling in the PCV region (Fig 1G and 1H). scrib or dlg1 mutant clone analysis further suggested that Scrib and DLG1 are required for BMP signaling in the PCV region in a cellautonomous manner (Fig 1I and S1B Fig).
Loss of BMP signal in scrib clones in the PCV region appears not be caused by perturbed ligand trafficking. GFP-Dpp ligands expressed in scrib clones are rather diffusible (S1C-S1F Fig). These results support that Scrib is required for optimizing BMP signal in the PCV region but not for facilitating BMP ligand trafficking. Interestingly, Scrib proteins are enriched in the vein primordial cells at 24 hr AP, where BMP signaling is positive (Fig 2A). These observations indicate that the Scrib complex may be up-regulated by BMP signaling and may constitute a feedback loop. To test this idea, we assessed pMad signaling and Scrib protein staining in pupal wings at different time points. At 18 h AP, the pMad signal is widely observed in the prospective LV regions, but faint in the PCV region, in line with previous reports that PCV morphogenesis initiates around 18h AP (Fig 2A) [7,8]. pMad staining becomes refined in LVs, and intensifies in the PCV around 20-24 hr AP (Fig 2A-2E). Scrib is uniformly distributed throughout wing tissues at 18 h AP, when BMP signaling is low. Scrib then gradually accumulates in the vein progenitor cells while BMP signaling becomes refined at 20-24 hr AP (Fig 2A-2E). Based on these data, we postulate that the BMP/Dpp signal may regulate scrib transcription.
To study whether transcriptional levels of scrib and dlg1 are also susceptible to BMP/Dpp signal activity, we performed quantitative PCR (qPCR) and found that scrib and dlg1 are transcriptionally regulated by BMP/DPP signaling ( Fig 2F). Our data also indicate that scrib and dlg1 transcription are up-regulated by BMP signal in future intervein cells as well as wing vein progenitor cells, when the constitutively active form of Tkv (caTkv) are ectopically induced (Fig 2G and 2H). To understand whether Scrib levels are up-regulated as vein material or specifically up-regulated by BMP signal, we compared Scrib and DE-cadherin (DE-Cad), a key molecule for epithelial morphogenesis [18], by producing ectopic clones of caTkv. Interestingly, up-regulation of Scrib but not DE-Cad become visible already at 20 h AP (S1G Fig). DE-Cad then becomes up-regulated at 24 h AP (S1H Fig), suggesting that both Scrib and DE-Cad are positively regulated by BMP signaling but in a distinct manner. Together, our data suggest that feedback through Scrib and DLG1 is crucial for long-range BMP/Dpp signaling during PCV development.

Scrib regulates Tkv distribution in the PCV region
How does Scrib affect long-range BMP/Dpp signaling in the PCV region? During the course of Dpp trafficking into the PCV region, Dpp ligands appear to move through the basal side of the pupal wing epithelia [9]. Since the Scrib complex has been proposed to regulate the apicalbasal protein distribution of epithelial cells [4,19], we hypothesize that Scrib may affect BMP receptor localization. To test this hypothesis, we investigated the distribution of the BMP type I  receptor Tkv in the PCV region. In wild-type pupal wings, Tkv is located at the basolateral region of PCV cells, however Tkv localization was observed more apically in scrib mutants ( Fig  3A-3C and S2A and S2B Fig). To investigate whether Tkv localization is regulated by Scrib or modulated as a secondary affect of changing cell polarity, we used an RNAi approach. Our data reveal that Tkv is enriched apically but reduced basally in scrib RNAi wings (S2C and S2D Fig). These results suggest that Scrib ensures enhanced basal Tkv localization where BMP ligand trafficking takes place. We also found that Tkv was frequently observed in puncta at the basal side of the PCV region (Fig 3D and 3E). We wondered whether the observed amounts of Tkv reflect internalized receptors. At the basal side of the PCV cells, Tkv often co-localizes with Scrib and DLG1 at the Rab5-positive early endosomes ( Fig 3D and S2E Fig). A key process in canonical BMP signal transduction is production of pMad by activated receptors and its accumulation in the nucleus. In wild type cells, substantial amounts of pMad co-localized with Tkv and Scrib as puncta in early endosomes ( Fig 3E). Since pMad subsequently accumulates in the nucleus, endosome-localized pMad appears to be transient. Our data also reveal that the number of co-localization of Tkv with Rab5 is decreased in scrib RNAi wings due to less number of Tkv puncta ( Fig  3F-3H and S2F Fig). Taken together, these data suggest that Scrib-dependent Tkv localization at Rab5 endosomes is key for producing robust BMP signaling in the PCV region.

Clathrin-dependent endocytosis is important for BMP signaling in the PCV region
If Tkv localization at Rab5 endosomes is key for BMP signaling in the PCV region, ablation of receptor internalization may reduce the BMP signal. Speculating that Scrib facilitates Tkv internalization through a clathrin-dependent mechanism, we investigated whether clathrin-dependent endocytosis affects BMP signaling in the PCV region. Adaptor complex-2 (AP-2), which functions together with clathrin to initiate endocytosis in the plasma membrane, is composed of four subunits: α, β, μ and σ [20]. Due to its importance in various cellular functions, complete inhibition of endocytosis causes severe tissue damage. Therefore, the relevance of endocytosis was analyzed by investigating genetic interactions and weak loss-of-function analysis. First, we analyzed genetic interactions of dpp alleles with AP-2α (ada), and found that ada interacted genetically with dpp in the PCV region (Fig 4A-4D). Ectopic PCV phenotypes are occasionally observed when Dpp signaling in the PCV region is partially disturbed [21]. This might be caused by incomplete maintenance of positive feedback mechanisms, resulting in ectopic ligand trafficking. Next we found that RNAi of AP2 complex subunits shows crossveinless phenotypes (Fig 4E-4G and S3A and S3B Fig). The PCV-less phenotypes were confirmed to be due to loss of BMP signaling in the PCV region, since the pMad signal was largely ablated in the prospective PCV region of wings expressing AP-2μ RNAi ( Fig 4H). Importantly, knocking down AP-2μ in this condition did not disrupt epithelial polarity, as shown by the apical localization of aPKC and the basolateral localization of DLG1 ( Fig 4I). We then investigated whether the early endosome protein Rab5 is required for BMP signaling in the PCV region. Ectopic expression of a dominant negative form of Rab5 significantly reduced BMP signaling in the PCV region ( Fig 4J). These results suggest that clathrin-mediated internalization of the BMP receptor to Rab5 endosomes is a crucial step for BMP signaling in the PCV region. compared to control yw flies. Relative mRNA expression level of genes of interest was calculated with normalization to gapdh mRNA level. ventral veinless (vvl) was used as a positive control. Error bar, SEM. ****P<0.001, **P<0.01, two-paired t-test. (G, H) In situ hybridization analysis of scrib (G) and dlg1 (H) in pupal wings at 24 h AP (upper panels). caTkv expression clones were generated using MARCM. Clones were marked by GFP (lower panels). Upregulated scrib or dlg1 corresponding to caTkv clones are indicated by arrows. Sense probe was applied to the wings expressing caTkv clones as a negative control. doi:10.1371/journal.pgen.1006424.g002

Scrib interacts with Tkv
How then does Scrib play a role in Tkv internalization and signaling? First, we addressed whether physical interactions between Scrib and Tkv are key for Scrib function. Scrib belongs to the LAP protein family, containing Leucine-rich repeats (LRR), PDZ and C-terminal (CT) domains ( Fig 5A). To investigate whether Tkv and Scrib are physically associated, we performed co-immunoprecipitation (co-IP) analysis. GFP-tagged Tkv and MYC-tagged fulllength or truncated Scrib were co-transfected into Drosophila S2 cells. Our data suggest that full-length Scrib weakly associates with Tkv (S4A Fig). In contrast, when different domains of Scrib were co-expressed, LRR regions showed strong interactions with Tkv (Fig 5B), suggesting that the LRR domain is sufficient for interaction with Tkv. Previous studies suggest that the LRR domain is mainly responsible for cell polarity maintenance and proliferation control, while the PDZ domain mediates physical interactions with a variety of proteins [22]. A Scrib fragment including the LRR and PDZ domains, but not the CT domain, showed strong interactions with Tkv ( Fig 5A and 5C). Since the CT domain is capable of binding to LRR (S4B Fig), this domain may function as a regulator of Scrib-Tkv interactions. Next, we examined whether the conformational change conferred by activation affects the association of Tkv with Scrib. Although caTkv associates with full-length Scrib less efficiently (S4A Fig), LRR interacts with caTkv and pMad (Fig 5D), suggesting that Tkv complexed with Scrib is able to phosphorylate Mad. We then tested whether LRR is sufficient for maintaining BMP signaling in the PCV region in vivo. BMP signaling in scrib mutant clones in the PCV region was sufficiently restored by LRR as well as by full-length Scrib, but not by Scrib without an LRR domain ( Fig  5E and 5F). Consistently, pMad signaling in the PCV region remains intact in mutant cells of scrib 5 allele that lacks the PDZ3/4 and CT domains (Fig 5G) [22]. We further confirmed that the LRR, but not the PDZ domain substantially restores cell polarity in scrib mutant cells of the pupal wings (S4C Fig). Therefore, restoration of BMP signaling in scrib mutant clones LRR might be mediated through biochemical interaction of Scrib, Tkv and pMad in endosomes or through restoration of tissue polarity. However, our RNAi data in which Scrib regulates BMP signaling in the PCV region in a distinct manner from regulation of tissue polarity (Fig 1) sufficiently support that the former scenario is more likely. Taken together, these results indicate that the LRR domain not only sustains epithelial polarity, but also interacts with the BMP receptor to maintain BMP signaling.

Rab5 and Scrib form a complex with Tkv
How is internalized Tkv regulated to effect BMP signaling? One simple hypothesis is that Tkv preferentially localizes to Rab5 endosomes to optimize BMP signaling. In that case, Tkv may associate with Rab5 as well as with Scrib to form a complex in the early endosome. To test this hypothesis, Tkv, Rab5 and LRR were co-expressed in S2 cells. We found that Tkv, Rab5 and LRR form a protein complex (Fig 6A). When caTkv and Rab5 are co-expressed, they not only interact with each other, but also with pMad (Fig 6B and 6C and S5A Fig), suggesting that Tkv in the Rab5 endosome is capable of phosphorylating Mad. Tkv is able to associate with various forms of Rab5, including wild-type, constitutively active (Q88L) and dominant negative (S43N) (S5B Fig) [23]. Furthermore, the BMP type-II receptor Punt was also detected in the complex with Rab5 and LRR (S5C Fig). These results are consistent with our hypothesis that  BMP receptors are preferentially localized in the Rab5 endosome to optimize BMP signaling. We further found that the constitutively active form of Rab5 partially rescued loss of BMP signaling in scrib mutants (Fig 6D and 6E), suggesting that the Rab5 endosomes play significant roles in BMP signaling during PCV formation.

Discussion
This study shows that the Scrib complex, a basolateral determinant, is a novel feedback component that optimizes BMP signaling in the PCV region of the Drosophila pupal wing.
During PCV development, limited amounts of Dpp ligands are provided by the Dpp trafficking mechanism [7,9]. Furthermore, amounts of receptors appear to be limited since tkv transcription is down-regulated in the cells in which the BMP signal is positive [7,24], a mechanism that serves to facilitate ligand diffusion and sustain long-range signaling in the larval wing imaginal disc [25,26]. To provide robust signal under conditions in which both ligands and receptors are limiting, additional molecular mechanisms are needed. Previous studies suggest that two molecules play such roles. Crossveinless-2 (Cv-2), which is highly expressed in the PCV region, serves to promote BMP signaling through facilitating receptor-ligand binding [12,27]. Additionally, the RhoGAP protein Crossveinless-c (Cv-c) provides an optimal extracellular environment to maintain ligand trafficking from LVs into PCV through down-regulating integrin distribution at the basal side of epithelia [9]. Importantly, both cv-2 and cv-c are transcriptionally regulated by BMP signaling to form a feedback or feed-forward loop for PCV formation.
Scrib, a third component, sustains BMP signaling in the PCV region in a different manner. First, to utilize Tkv efficiently, Scrib regulates Tkv localization at the basolateral region in the PCV cells, where ligand trafficking takes place. Regulation of receptor localization could be a means of spatiotemporal regulation of signaling molecules during the dynamic process of morphogenesis. Second, to optimize the signal transduction after receptor-ligand binding, Scrib facilitates Tkv localization in the Rab5 endosomes. Localization of internalized Tkv is abundant at Rab5 endosomes in the PCV region of wild-type, but not scrib RNAi cells. While the physical interaction between Scrib, Tkv and Rab5 in the pupal wing remains to be addressed, our data in S2 cells suggest that physical interactions between these proteins are key for preferential localization of Tkv at the Rab5 endosomes. Recently, Scrib has been implicated in regulating NMDA receptor localization through an internalization-recycling pathway to sustain neural activity [28]. Therefore, Scrib may be involved in receptor trafficking in a context-specific manner, the molecular mechanisms of which, however, remain to be characterized. Third, BMP/Dpp signaling upregulates scrib transcription in the pupal wing. Moreover, knockdown of scrib leads to loss of BMP signaling in PCV region but not loss of apical-basal polarity (Fig 1). These facts suggest that upregulation of Scrib is key for optimizing BMP signaling by forming a positive feedback loop.
Previous studies indicate that cell competition takes place between scrib clones and the surrounding wild-type tissues in the larval wing imaginal disc [29]. Moreover, cell competition has been documented between loss-of-Dpp signal and the surrounding wild-type tissues [30]. We presume that the mechanisms proposed in this study are independent of cell competition for the following reasons. First, scrib RNAi and AP-2μ RNAi data reveal that loss of BMP signal in the PCV region is produced without affecting cell polarity (Figs 1 and 3). Thus, cell Previous studies established that receptor trafficking plays crucial roles in signal transduction of conserved growth factors, including BMP signaling. Several co-factors have been identified as regulators of BMP receptor trafficking [31][32][33][34][35][36][37][38]. Some of them down-regulate BMP signaling [31][32][33][34][35]38], while others help maintain it [36,37]. We propose that the Scrib-Rab5 system is a flexible module for receptor trafficking and can be utilized for optimizing a signal. During larval wing imaginal disc development, BMP ligands are trafficked through extracellular spaces to form a morphogen gradient. Although previous studies indicate that regulation of receptor trafficking impacts BMP signaling in wing imaginal discs [23,39], BMP signaling persists in scrib or dlg1 mutant cells in wing discs (S6 Fig). Wing disc cells interpret signaling intensities of a morphogen gradient. In this developmental context, an optimizing mechanism might not be beneficial to the system. In contrast, cells in the PCV region use the system to ensure robust BMP signaling.
Taken together, our data reveal that a feedback loop through BMP and Scrib promotes Rab5 endosome-based BMP/Dpp signaling during PCV morphogenesis. Since the components (BMP signaling, the Scrib complex, and Rab5 endosomes) discussed in this work are highly conserved, similar mechanisms may be widely utilized throughout Animalia.

Antibodies, chemicals and immunohistochemistry
Pupal wings were fixed in 3.7% formaldehyde (Sigma-Aldrich) at 4°C overnight. Wing imaginal discs were fixed in 3.7% formaldehyde at room temperature (RT) for 20 minutes. All immunostaining and in situ hybridizations were performed as described previously [7,9].

Imaging and image analysis
Fluorescent images were obtained with a Zeiss LSM700 upright confocal microscope. Images of in situ hybridization and adult wings were obtained with a Nikon Eclipse 90i microscope. All confocal immunofluorescent images were processed and analyzed with ImageJ (NIH). The images were the composite of a stack with projection of max intensity unless specified in detailed.
in vivo RNAi nub-GAL4 was used to drive the RNAi in the wing blade in combination with the temperaturesensitive GAL80 [16]. scrib, dlg1 or lgl RNAi flies were cultured at 25, 27 or 25°C, respectively. AP-2 subunit RNAi flies were maintained at 18°C for 5 days and then cultured at 27°C.

Cell culture and production of recombinant proteins
Drosophila S2 cells were used for producing recombinant proteins as previously described [44]. Cells were transfected with HiFugene transfection reagent (Promega) according to the manufacturer's protocol. S2 cells were transfected with the plasmids expressing indicated cDNA and tub-GAL4. Three days after transfection, cells were collected and cell lysates were subjected to immunoprecipitation using the GFP-Nanotrap A Kit (Chromotek) according to the manufacturer's instructions. S2 cells were subjected to the IP lysis buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 1 mM EDTA, 5% glycerol) on ice for 30 min. The supernatants obtained were subjected to the WB as an input and subsequent IP. Western blotting was conducted as previously described [44]. All biochemical data shown are representative of no less than three independent assays.
Reverse transcription-quantitative real-time polymerase chain reaction (RT-qPCR) Pupal wings were dissected without fixation, and subjected to RNA purification using TRIzol reagent (Invitrogen

Statistics
All experiments were carried out independently at least three times. Error bars indicate s.e.m. Statistical significance was calculated by the two paired t-test method.