A functional role of the extracellular domain of Crumbs in cell architecture and apicobasal polarity

Morphogenesis and organogenesis rely on the capacity of cells to coordinately change shape. Shape changes in epithelial cells, which are present in most tissues and organs, greatly contribute to tissue remodelling. Epithelial cells show a marked apicobasal polarity, which is established and maintained by the components of the subapical region (SAR) (Knust and Bossinger, 2002; Tepass et al., 2001). The evolutionarily conserved protein Crb is a major apical polarity determinant and the only known transmembrane component of the SAR (Assémat et al., 2008; Bazellieres et al., 2009; Bulgakova and Knust, 2009). Crb contains one single transmembrane domain, a short cytoplasmic domain (Intra), and a long extracellular (Extra) domain (Tepass et al., 1990). The Intra domain interacts with different proteins to organise a membrane-associated core-complex defining apicobasal polarity with Stardust (Sdt), Pals1-associated tight junction protein (Patj) and Lin-7, and establishes links with apical βH-spectrin, actin cytoskeleton and basolateral polarity complexes. In addition, Crb participates in other cellular events through interactions with, for example, the Par polarity network (aPKC/Par6) and the Hippo pathway (Bazellieres et al., 2009; Bulgakova and Knust, 2009). It has been postulated that Crb Intra is sufficient to rescue the absence of crb in several instances (Johnson et al., 2002; Klebes and Knust, 2000; Wodarz et al., 1995), suggesting that Crb Extra can be dispensable. In contrast to Crb Intra, the function of the Extra domain, which contains 29 EGF-like and four laminin A G-like repeats, is less well understood. In spite of the fact that its organisation is conserved from flies to mammals (Richard et al., 2006), and that many mutations in the Crb1 human homolog responsible for severe retinal dystrophies map at this protein domain (Bujakowska et al., 2012), no general role of Crb Extra in Crb activity has been established. Based on indirect data some reports suggested the ability of Crb molecules to interact with other Crb molecules (Chen et al., 2010; Fletcher et al., 2012; Pellikka et al., 2002; Wei et al., 2006) or other proteins (Herranz et al., 2006; Richardson and Pichaud, 2010). Moreover, a recent study describes adhesion between zebrafish Crb molecules (Zou et al., 2012).

Given the importance of Crb both in physiological and pathological conditions, it is crucial to understand Crb requirements in morphogenesis. Here we provide experimental evidence for a role of the Extra domain in protein stabilisation required for Crb control of apicobasal polarity and cell shape.

We previously documented a differential accumulation of Crb in the ectoderm, identifying groups of cells with increased Crb levels at stages 9–11, that correspond to dorsal epidermis and tracheal placodes (Fig. 1A,A′). This increased posttranscriptional accumulation in tracheal placodes correlates with the apical constriction that precedes invagination (Letizia et al., 2011). In this study we aimed to further explore the correlation between Crb levels and cell architecture.

At confocal resolution we observed that increased Crb accumulation in dorsal epidermis or tracheal placode cells correlates with a narrower apical domain compared to cells with lower Crb levels at the ventral epidermis of stage 10 embryos (Fig. 1A′,B). In addition, in dorsal cells, SAR markers labelled a longer domain along the apicolateral membrane while adherens junctions (AJs) occupied a more basal position (Fig. 1B′–B‴). High-resolution ultrastructural analyses using transmission electron microscopy (TEM) revealed different organisation of these cell types. Dorsal epidermal cells were columnar or wedge-shaped and the apical membrane facing the exterior (apical-free domain) was smooth and dome-shaped (Fig. 1C). AJs were present at the lateral membrane beneath an extended junction-free region where adjacent membranes were in contact (apicolateral domain) (Fig. 1G). We speculate that this apicolateral domain corresponds to the elongated SAR where high levels of Crb accumulate. In contrast, ventral cells were cuboidal, with a flat apical-free domain and with AJs positioned more apically, resulting in a short apicolateral domain (Fig. 1E,I). Scanning electron microscopy (SEM) confirmed differences between the apical domain of dorsal and ventral epidermal cells (Fig. 1K,M). These results indicate that, in wild-type embryos, different levels of Crb correlate with different cell morphologies.

In agreement with a role of crb in cell morphology, our analyses revealed defects in cell shape in crb mutants (Fig. 1D,H,L,F,J,N). Tracheal (Letizia et al., 2011) and dorsal cells appeared rounded (Fig. 1D) and with bulged and protruded apical domains (Fig. 1L). AJs, when present, occupied a more apical position (Fig. 1H), as compared to wild type (Fig. 1G). Interestingly, confocal and ultrastructural analyses indicated that uniform accumulation of Crb, as observed when Crb is overexpressed, also induced dorsal cells to adopt an unpolarised shape (Fig. 1O–R). Since both the absence and the mislocalisation of Crb affect cell shape it suggests that Crb needs to be properly localised to control cell architecture.

To evaluate Crb requirement in cell morphology we sought to understand its mechanism of apical localisation. Although Crb targeting and stabilisation to the apical membrane is not fully understood, both Crb Extra and Intra domains may participate in the process (Fletcher et al., 2012).

A mechanism by which Crb Extra may stabilise Crb apically is through protein interactions in trans, as previously suggested. We and others (Chen et al., 2010; Hafezi et al., 2012; Pellikka et al., 2002) observed that in crb mutant clones the accumulation of Crb and Crb-interacting proteins is lost in mutant cells but also at cell–cell contacts of wild-type cells facing mutant cells (clone border, Fig. 2C). To test for Crb Extra interactions we expressed Crb^extra (a membrane-bound GFP-tagged form of Crb that lacks the Intra domain; Fig. 2B) (Pellikka et al., 2002), in crb mutant clones and assayed for Crb accumulation at the clone border by analysing Patj accumulation [Crb antibody recognises both Crb^extra and wild-type Crb (Tepass and Knust, 1993)]. Strikingly, we found a complete rescue of Patj at the clone border (Fig. 2D; supplementary material Fig. S1A). As Crb^extra cannot rescue Patj accumulation in mutant cells, its accumulation at the border must be due to the presence of full-length Crb in wild-type cells. This demonstrates that Crb Extra is able to stabilise Crb in trans in neighbouring membranes (Fig. 2D″).

In contrast to Crb^extra, expression of Crb^intra (a membrane-bound GFP-tagged form of Crb that lacks the Extra domain; Fig. 2B) in crb mutant clones did not rescue Patj accumulation at clone border (Fig. 2E), indicating that the Extra domain is necessary for Crb apical stabilisation.

In support of these observations we find that Crb Extra, but not Crb Intra, localises apically in the presence of wild-type Crb. When Crb^extra is expressed in the embryonic epidermis we detected accumulation along the apicobasal membrane. Crb^extra also accumulated in a distinct apicolateral region that corresponds to the SAR, as visualised by Patj accumulation (Fig. 2F, arrows). Crb^extra is stabilised apically only if full-length (FL) Crb is present, as the distinct apical accumulation was lost in the absence of endogenous Crb (crb^11A22 mutants; Fig. 2G). This loss was not due to a general polarity defect as Crb^extra also localised in an unpolarised pattern in tissues where crb is not required for polarity (supplementary material Fig. S1A,B). On the other hand, Crb^intra was unable to distinctly localise apically in any condition tested. When expressed in the embryo or wing discs, Crb^intra was present around the whole cell membrane (Fig. 2H; supplementary material Fig. S1C; Fig. S2C), as observed for other Crb^intra tagged versions (Klebes and Knust, 2000; Muschalik and Knust, 2011; Wodarz et al., 1995). This mislocalisation is not due to an overexpression effect, since, in contrast, comparable levels of Crb^FL (supplementary material Fig. S2F) are able to localise properly (supplementary material Fig. S2B,E).

Altogether our data provides experimental confirmation for a role of Crb Extra in mediating interactions between Crb molecules, that are required for Crb apical stabilisation in vivo. Although both Crb Extra and Intra domains are needed for Crb apical localisation, the Extra domain can stabilise Crb apically whereas the Intra alone cannot, presumably due to the absence of its extracellular domain.

To determine whether, as suggested by our in vivo analysis, Crb Extra molecules are able to mediate cell adhesion we performed cell aggregation assays in vitro. We expressed full-length Crb [Crb^FL (Wodarz et al., 1995)], Crb^extra and Crb^intra using an actinGal4VP16 promoter. Cells expressing Crb^FL formed aggregates (Fig. 3A,A′), indicating that Crb is able to mediate cell–cell interactions. Cells expressing Crb^intra showed GFP uniformly around the cell membrane but did not consistently aggregate (Fig. 3B,B′,F). We observed on occasion the formation of small clusters of Crb^intra expressing cells (Fig. 1B,G). Nonetheless, this clustering is Crb independent since Crb^intra and Crb^FL expressing cells are not able to adhere to each other (Fig. 1H). In contrast, aggregation assays with mixed populations showed that cells expressing Crb^extra adhered to Crb^FL cells (Fig. 1E,H). Furthermore, cells expressing Crb^extra alone aggregated, and the majority of Crb^extra expressing cells were part of aggregates (Fig. 1C,C′,F,G). These results indicate that the Extra domain of Crb is required and sufficient to mediate cell adhesion.

Our results support previous reports suggesting that Drosophila Crb may engage in homophilic interactions (Fletcher et al., 2012; Pichaud and Desplan, 2002; Tepass et al., 2001), although they do not rule out indirect Crb–Crb interactions mediated by a partner. Importantly our data are in line with recent studies showing that zebrafish Crb2a and Crb2b (Zou et al., 2012) and Drosophila Crb (Röper, 2012) engage in cell–cell interactions, suggesting a universal capacity of Crb proteins to mediate adhesion.

We find that Crb Extra is critical for Crb apical stabilisation. On the other hand our analysis indicate that Crb apical accumulation plays a role in cell shape. Therefore, we expected the Extra domain to be required for cell architecture. In agreement with this we find a defective rescue of cell morphology when Crb^intra is expressed in crb mutants (Fig. 4D; supplementary material Fig. S4), in contrast to the rescue obtained with Crb^FL (Fig. 4C; supplementary material Fig. S4). In addition, confocal z-sections show that in contrast to Crb^FL, Crb^intra is unable to fully rescue AJs positioning and SAR enlargement (supplementary material Fig. S5). Although a certain rescue of cell morphology could be expected by providing Crb^extra, such rescue was very mild (Fig. 4E; supplementary material Fig. S4), probably due to the inability of this construct to localise apically in the absence of endogenous Crb.

The requirement of Crb Extra in Crb apical stabilisation predicts a role of this domain also in maintaining the general apicobasal polarity. However, it has been suggested that the Extra domain of Crb is dispensable for cell polarity (Klebes and Knust, 2000; Wodarz et al., 1995). To evaluate the sufficiency of Crb Intra in apicobasal polarity we compared the ability of Crb^intra and Crb^FL to rescue crb mutant defects in polarity, AJs formation, and embryonic patterning. Mild overexpression of Crb^FL rescued crb defects (Fig. 4H,M,R; supplementary material Fig. S4). Under these conditions we did not detect gross dominant defects and Crb was apically localised (supplementary material Fig. S2B,E). In contrast, Crb^intra could only partially rescue Crb activity (Fig. 4I,N,S; supplementary material Fig. S4). This partial rescue might reflect the activity of some Crb^intra molecules that randomly accumulate in apical regions, which would interact with partners preferentially accumulated there. This would compensate for the absence of Crb^intra apical stabilisation. Finally, the expression of Crb^extra in crb mutants did not rescue polarity and adhesion (Fig. 4J,O,T).

Our results indicate that both the Extra and Intra domains of Crb are required to fully rescue Crb requirements in embryonic tissues, highlighting the need of protein integrity. We propose a two-step mechanism in which once Crb reaches the apical membrane it becomes stabilised through interactions in trans, and also possibly in cis as proposed (Fletcher et al., 2012), mediated by Crb Extra. In a second step, the apical localisation would be reinforced or fixed by interactions with other proteins through the Intra domain, as also proposed recently (Fletcher et al., 2012). The Intra domain would also mediate downstream effects (e.g. on polarity, adhesion and cell morphology) through interactions with these partners (Bulgakova and Knust, 2009).

We report the role of Crb Extra in assisting protein stabilisation in the SAR, likely by mediating interactions between Crb molecules in neighbouring membranes. This stabilisation is functionally relevant as it is critical to maintain cell polarity. In addition, we describe a new specific Crb-dependent mechanism controlling cell shape changes. We propose that localised Crb stabilisation plays a role in cell architecture by the organisation of the apical domain and positioning of AJs. Our results are consistent with a model (Fig. 4U) in which increased levels of Crb in the SAR enlarge the apicolateral domain at the expense of the apical-free domain, leading to the narrowing of the apical area. In addition, AJs become positioned more basally, likely through interactions mediated by the Intra domain (Bilder et al., 2003; Grawe et al., 1996; Harris and Peifer, 2004; Tanentzapf and Tepass, 2003; Tepass, 1996). These morphological changes confer a columnar or constricted shape to the cell. In contrast, lower Crb levels allow an expanded apical-free domain and an apical positioning of AJs, promoting a cuboidal morphology. Crb–Crb interactions between neighbouring membranes mediated by Crb Extra, besides stabilising Crb apically, could help to stabilise cell contacts above AJs conferring to the SAR a previously undescribed junctional character. This could represent a rapid and reversible mechanism of cell shape change.

Interestingly, a similar mechanism has been recently described, where the positioning of AJs by Par1/Par3 apicobasal polarity drives epithelial folding (Wang et al., 2012). This suggests that the positioning of AJs controlled by apicobasal polarity can be a general mechanism for apical narrowing (Sumigray and Peifer, 2012). Furthermore, as Crb is also upregulated in other structures that undergo a similar apical narrowing, such as salivary glands (Myat and Andrew, 2002), posterior spiracles (Lovegrove et al., 2006) or segmental grooves (our unpublished observations) it would be tempting to speculate that in these tissues Crb accumulation also controls cell shape changes. Finally, our results are also in line with a recent report proposing a role for the Crb Extra domain in an anisotropic localisation of Crb that determines actomyosin contractility (Röper, 2012).

The following lines were used: y¹w¹¹⁸, crb^11A22, FRT82B-GFP and hsFlp; tubulin-Gal4,FRT82Btubulin-Gal80, described in FlyBase (http://flybase.org/), resilleGFP (Morin et al., 2001), UASCrb^intra (UASCrb^{intra+transmembrane}-GFP, see below), UASCrb^extra (UASCrb^{extra+tranmembrane}-GFP; a gift from U. Tepass) and UASCrb^FL (a gift from E. Knust). nulloGal4 drives ubiquitous expression in the ectoderm and salPEGal4 in the wing pouch of imaginal discs.

The FLAG tag of UAS-Crb_FLAGintra [a gift from E. Knust (Richard et al., 2009)] was replaced by a GFP sequence into a BglII restriction site. The UAS-Crb^intra construct was introduced into y w¹¹¹⁸ embryos by standard P-element transformation (Ashburner, 1989). Different lines were analysed, showing a comparable GFP accumulation.

Embryos were staged as described previously (Campos-Ortega and Hartenstein, 1985) and imaginal discs obtained by dissecting third instar larvae. We followed standard protocols for immunostainings. The primary antibodies used were: anti-Crb (Cq4) 1∶10, anti-Ecad (DCAD2) 1∶100 and alpha-Spectrin (3A9) 1∶5 from DSHB, anti-aPKC ζ C-20 1∶500 (Santa Cruz Biotechnology), anti-Patj 1∶1000 (H. Bellen), anti-pTyr 1∶1000 (Upstate), anti-GFP 1∶600 (Molecular Probes and Roche), anti-βGal 1∶600 (Abcam). Cy3-, Cy2- and Cy5-conjugated secondary antibodies (Jackson ImmunoResearch) were used at 1∶300.

Confocal images were obtained with Leica TCS-SPE or TCS-SP5 system. Images were imported into ImageJ and Adobe Photoshop and assembled using Adobe Illustrator.

Imaginal wing discs with positively or negatively crb^11A22 marked clones were generated by FLP-mediated mitotic recombination (Duffy, 2002). Second instar larvae were heat shocked for 1 hour at 37°C.

For TEM embryos were prepared following the protocol established (Moussian and Schwarz, 2010). The samples were analysed by a Philips CM10 electron microscope at 60 kV. SEM was performed as described (Moussian et al., 2006).

Drosophila S2 cells were transiently transfected using Effectene reagent (Qiagen). Constructs were co-transfected with actinGal4VP16 (gift from J. Treisman). As Crb^FL is not tagged we co-transfected Crb^FL and DsRed (gift from R. Axel) to visualise expressing cells and confirmed membrane Crb expression with Crb antibody (supplementary material Fig. S3). Cells were used for aggregation assays after 3 days of expression.

Aggregation assays were carried out on transfected cells, with comparable transfection efficiency, as described previously (Snow et al., 1989). For mixed cell aggregation, half of each individual cell population was mixed together at the time of the assay. Aggregation was imaged and quantified using a TE-200C Nikon inverted microscope coupled with a CoolSNAP camera (Photometrics) and the images were processed in ImageJ.

Reactions were carried out using SYBR Green Universal PCR Master Mix (Applied Biosystems) at an annealing temperature of 59°C. StepOnePlus Real-Time PCR System was used according to manufacturer's instructions. Gene expression levels were normalised using endogenous α-Tubulin and differences in crb gene expression were determined using the StepOne 2.2 software.

We thank N. Luque, E. Fuentes, E. Rebollo, L. Bardia and Matthias Floetenmeyer for technical assistance, E. Knust and U. Tepass for reagents and the Developmental Studies Hybridoma Bank and Bloomington Stock Centre for antibodies and fly lines. We thank Llimargas lab for discussions and J. Casanova, A. Casali, M. Furriols M. Milán and K. Campbell for critically reading the manuscript. S.R. acknowledges a contract from the ‘Ramón y Cajal’ programme.