Crumbs stabilises epithelial polarity during tissue remodelling

Polarity and adhesion are hallmark features of epithelial tissues. In the last decade, much has been learnt about the genetic and molecular mechanisms controlling cell polarity in epithelial cells. However, epithelia are extensively remodelled during morphogenesis, when cell rearrangements require the modulation of adhesion and impose a challenge on the stability of cell polarity. In order to unravel the relationship between cell polarity and morphogenesis, we have chosen a simple epithelial system, the renal tubules of the Drosophila embryo. This epithelium has the great advantage that the elaboration of apicobasal polarity and morphogenetic cell rearrangements are clearly separated in time, whereas these two processes occur concurrently in the ectoderm during gastrulation and early germ-band extension.

In the Drosophila embryo, apicobasal polarity is evident in the blastoderm through the apical localisation of Par-6, which depends on the activation of the small GTPase, Cdc42 (Hutterer et al., 2004). At the same time Bazooka (Baz)/Par3 colocalises with the adherens junction component Drosophila E-cadherin, and later acts during gastrulation to stabilise aPKC and Par-6 apically (Harris and Peifer, 2004; Harris and Peifer, 2005). As a consequence, Scribble (Scrib), Lethal giant larvae (Lgl) and Discs large (Dlg) concentrate basolaterally (Bilder et al., 2003; Hutterer et al., 2004; Tanentzapf and Tepass, 2003). Polarity very soon becomes dependent on the activity of Crumbs (Crb) and Stardust (Sdt), which are recruited to the sub-apical domain by Baz. In the epidermis the Crb complex acts to maintain Baz apically and to restrict Scrib, Lgl and Dlg to the basolateral domain (Bilder et al., 2003; Tanentzapf and Tepass, 2003). Embryos mutant for crb or sdt fail to maintain epidermal cell polarity or to establish a zonula adherens (ZA). This results in the loss of epithelial integrity and the epidermis undergoes widespread cell death (Grawe et al., 1996; Tepass et al., 1990; Tepass, 1996). However, despite the expression of Crb and Sdt in all ectodermally derived epithelia, these tissues are differentially affected by the loss of crb or sdt (Tepass et al., 1990), suggesting that distinct features of epithelia underlie differences in their dependence on crb or sdt.

In this paper we show that, whereas the apicobasal polarity of renal tubule cells depends on Baz and Scrib throughout development, Crb and Sdt are required only later as the tubule cells reorganise during extensive morphogenetic movements. If these movements are arrested tubule cell polarity is maintained in the absence of Crb. Furthermore, we provide evidence that the phase when the maintenance of cell polarity depends on Crb coincides with the time when the epithelium shows an enhanced requirement for exocyst function. This result underscores the importance of targeted membrane trafficking for cell polarity during tissue remodelling. Similar observations in other epithelia of the Drosophila embryo indicate a general function for Crb in cell polarity during morphogenesis.

We first sought to establish the pattern of expression of cell polarity and adhesion proteins in the renal tubules. During stage 10 they bud out from the polarised epithelium of the hindgut primordium, forming short, fat tubes, which elongate dramatically during stages 13-15 by convergent-extension cell rearrangements, to produce the extended mature tubules with just two cells lining the lumen (Fig. 1A,E,I) (Ainsworth et al., 2000). Thus the establishment of cell polarity (before stage 10) and morphogenetic cell rearrangement (stages 13-16) do not overlap in time. This differs from the epidermis where these two processes occur concurrently during gastrulation and early germ-band extension.

From the time of tubule eversion, during stage 10, the localisation of cell polarity and junctional proteins in the tubules was found to be identical to the epidermal pattern in wild-type embryos (Fig. 1 and data not shown). Significantly, apical and basolateral proteins, as well as ZA-associated proteins, were expressed in defined, non-overlapping domains throughout tubule reorganisation (Fig. 1B-D,F-H,J-L). Thus the process of cell rearrangement does not disrupt apicobasal polarity or cell adhesion during wild-type development.

We assessed the requirement for the activity of cell polarity proteins for tubule development by analysing loss-of-function mutations, where necessary removing both the maternal and zygotic contributions. Tubule cell polarity was lost or severely disrupted already in stage 11 embryos lacking both maternal and zygotic baz or scrib (Fig. 2). By contrast, tubule cells in embryos mutant for crb (or sdt, data not shown) appeared normal at this stage (Fig. 3A-F); Baz and Dlg remained localised (Fig. 3C,D) and electron microscopy revealed that ZAs develop (Fig. 3E,F). Accordingly, in flies carrying the amorphic allele crb^8F105, which encodes a truncated protein (Wodarz et al., 1993), Crb protein was still apically localised in tubule cells (Fig. 3B), although it was no longer localised in the epidermis at this stage. Thus the Crb complex is not required for the establishment of polarity and formation of ZAs in the renal tubules or for the maintenance of cell polarity early during their development.

Crb is a potent apical determinant, and strong overexpression in epidermal cells causes expansion of the apical membranes at the expense of the basolateral domain (Wodarz et al., 1995). We overexpressed full-length Crb using a ubiquitous (data not shown) or a tubule-specific driver but found that, despite high levels of Crb (Fig. 3G-K), tubule cell morphology remained unaffected in embryos from stage 11 to 13 (Fig. 3G-K) and the localisation of Baz was unaltered (Fig. 3J,K). Taken together with our loss-of-function data, this shows that the tubules are relatively insensitive to the level of Crb early in their development.

We next asked whether there is a requirement for Crb when the tubules are remodelled by cell rearrangements later in development. Tubules of wild-type and crb^8F105 mutant embryos at stage 16 show remarkable differences in their morphology. Wild-type tubules have extended, coursing through the body cavity to take up their characteristic positions, whereas the tubules of crb mutant embryos failed to elongate, presenting the appearance of disorganised clusters of cells (Fig. 4A,B). Strikingly, apical and basolateral proteins were delocalised in the tubule cells of stage 16 embryos [Baz, aPKC, Par-6 and Dlg shown in Fig. 4C-F, Lgl, Scrib (as Dlg) not shown]. Ultrastructural analysis revealed that the ZAs have dispersed by this stage (Fig. 4G,H).

Since crb is essential for cell polarity and epithelial integrity during remodelling of the tubules, we analysed whether the overexpression of Crb affects this process as well. Crb overexpression resulted in the delocalisation of Baz (Fig. 4I,J) and Dlg (Fig. 4K,L) in stage 16 tubules. This behaviour is in contrast to the insensitivity of Baz localisation to the cutB-Gal4-mediated overexpression of Crb in tubule cells of stage 11-13 embryos (Fig. 3K). Staining for Armadillo, normally localised to ZAs (Peifer, 1993), reveals a loss of junctional integrity in stage 16 embryos (Fig. 4M,N). Consistent with the evidence that tubule cell polarity is compromised, tubules became multilayered cords of cells lacking a clear lumen (Fig. 4K,L). Together, the loss- and gain-of-function phenotypes for crb indicate two stages in the maintenance of renal tubule cell polarity; the first is independent of Crb whereas the second requires its activity.

In order to discover the underlying cause of this change in the regulation of tubule cell polarity, we sought to pinpoint more precisely the onset of mutant phenotypes in embryos either lacking or overexpressing Crb. The delocalisation of cell polarity markers and changes in tubule morphology that result from the loss of Crb function became apparent only during stage 13 (Fig. 5A-C). This was mirrored by the onset of sensitivity to the overexpression of Crb; Baz remained apically localised at the beginning of stage 13 but became delocalised in late stage 13 tubule cells (Fig. 5D,E). Stage 13 marks the beginning of two major morphogenetic movements in the tubules; elongation by convergent-extension movements and integration of a population of mesodermal cells, which later differentiate as a distinct cell type, the stellate cells (Denholm et al., 2003). These morphogenetic activities occur concurrently and their timing suggests a correlation between Crb-dependent polarity and these events.

To test the hypothesis that Crb is required to maintain tubule cell polarity during cell rearrangement, we assessed cell polarity in the absence of Crb function and under conditions in which these morphogenetic movements are arrested. Cut is expressed in tubule cells (Fig. 1) and is required for the development of this tissue. In embryos carrying mutations in cut, tubules bud out from the hindgut primordium but never elongate, remaining as epithelial blisters close to the hindgut-midgut junction (Hatton-Ellis et al., 2007). In addition, stellate cells, marked by Teashirt, did not integrate into the resulting tubule blisters (Fig. 5F,G,H,I). In these embryos tubule cell polarity was unaffected, as revealed by apical Baz and basolateral Fasciclin II (Fas II, Fig. 5J; also known as Fas2). Tubules in cut;crb double-mutant embryos of the same stage resembled those of cut mutants; they formed epithelial blisters in which Baz and Fas II were normally localised (Fig. 5K,L). The delocalisation of Baz and of basolateral proteins, seen in the tubule cells of stage 16 crb mutant embryos (Fig. 4D), did not occur. These data indicate that Crb function is only required for cell polarity during morphogenetic activity.

Crb and Sdt have been ascribed a conserved role in ectodermal epithelia of Drosophila for the maintenance of polarity, namely to restrict the distribution of Baz, aPKC and Par-6 (apical) and Scrib, Dlg and Lgl (basolateral) and, as a consequence, to allow spot adherens junctions to coalesce to form continuous ZA belts (Bachmann et al., 2001; Grawe et al., 1996; Hong et al., 2001; Tepass and Knust, 1993; Wodarz et al., 1993; Tepass, 1996). It has been noted, however, that different epithelia in the embryo show differing sensitivities to the loss of crb (Tepass and Knust, 1993; Wodarz et al., 1993), suggesting that distinct features underlie differences in their dependence on the activity of the Crb complex. Our analysis of the renal tubules now provides an explanation for this apparent phenotypic variation. In early tubules Crb is not required to localise either apical or basolateral cell polarity proteins, ZAs form normally in its absence and tubule cell polarity is maintained. However, as soon as the tubule cells start to reorganise, changing their neighbours and remodelling their junctional contacts, Crb becomes indispensable, both to maintain the polarised distribution of apical and basolateral proteins and for the maintenance of ZAs (Fig. 4H).

To find out whether the maintenance of epithelial integrity during tissue remodelling is a general function of crb, we analysed epidermal cells, which undergo major rearrangements during germ-band extension. The timing and degree of cell movement varies between trunk and head epidermis. The onset of rearrangement occurs later in the head, where stomodeal invagination starts about an hour after the onset of germ-band elongation in the trunk (Campos-Ortega and Hartenstein, 1997). In the germ-band of crb mutant embryos, loss of cell polarity, as revealed by defects in the localisation of junctional proteins, is first apparent during stage 8 as cells rearrange (Grawe et al., 1996; Tepass, 1996), whereas in the stomodeal region loss of cell polarity, as well as cell movements, was found to occur during stage 10 (Fig. 6A-C). Thus, loss of polarity in crb mutant embryos correlates with the timing of cell rearrangement.

We next sought to stall the morphogenetic movements associated with germ-band elongation. In our hands drug-induced stalling (Bertet et al., 2004) is unreliable but we find that germ-band elongation arrests prematurely in Krüppel (Kr) mutant embryos (Fig. 6E) (Bertet et al., 2004; Irvine and Wieschaus, 1994). In these embryos the polarity of epidermal cells was normal (Fig. 6D,E, compare with defects evident in the trunk and tail epidermis of crb mutant embryos in Fig. 6F). In contrast to crb mutant embryos, the epidermis of embryos doubly mutant for crb and Kr exhibited relatively normal distribution of cell polarity and junctional proteins (Fig. 6F-K) and cuticle preparations showed that there is some suppression of the crb phenotype; a continuous, though reduced, cuticle sheet was formed (Fig. 6L,M). We found a similar rescue using a different Kr allele (Kr⁹, data not shown), confirming the specificity of the effects of Kr loss of function. Together these data show that there is substantial rescue of cell polarity in the epidermis of crb mutant embryos if cell movements are strongly reduced. These data reveal a striking similarity between the behaviour of the epidermis and the tubules; cell polarity in both tissues becomes dependent on Crb when cells rearrange. We therefore suggest that Crb has a conserved role in the maintenance of cell polarity during tissue reorganisation and that the differing sensitivities of tissues to the loss of crb are related to the time and degree of tissue morphogenesis that they undergo.

The maintenance of cell polarity during cell movement and rearrangement depends on a number of processes, including targeting of transmembrane proteins to specific cell surface domains. This process involves, among others, the function of the exocyst complex to tether secretory vesicles to specific sites (Lipschutz and Mostov, 2002; Whyte and Munro, 2002). The Sec6/8 (exocyst) complex, originally described in S. cerevisiae, is made up of eight subunits. Drosophila homologues of exocyst subunits are required for the targeted delivery of junctional proteins (Langevin et al., 2005), apical proteins, including Crb (Blankenship et al., 2007), Rhodopsin 1 (Beronja et al., 2005), as well as membrane and secreted proteins required for cell growth (Murthy et al., 2003; Murthy et al., 2005; Murthy and Schwarz, 2004). In order to explore further the early resistance of tubule cell polarity to the loss of crb activity, we analysed the effects of the loss of exocyst function in mutants for exo84, by using an allelic combination that removes all zygotic and the majority of maternal Exo84 (Blankenship et al., 2007). Strikingly, loss of epidermal cell polarity at stage 11 was not mirrored in the tubules, where cell polarity and tissue integrity were maintained, manifested by the unaltered distribution of Crb, Baz, Sas and Fas II (Fig. 7A-D). However, stage 16 mutant tubules were abnormally shaped having failed to complete tissue extension (Fig. 7E-H). Apical proteins spread into the lateral membrane domain of the tubule cells (Fig. 7F) or were partially lost from apical domains (Fig. 7H), whereas basolateral proteins remained localised (Fig. 7H).

Blankenship et al. (Blankenship et al., 2007) showed that there is a similar phenotype in the epidermis, in that cell polarity, as well as the ZA, is established in exo84 mutants but during stage 9, first Crumbs, and subsequently other apical and junctional proteins mislocalise. The timing of the onset of cell polarity phenotypes in both the epidermis (Blankenship et al., 2007) and the tubules (this study) of exo84 mutant embryos is remarkably similar to the time of appearance of similar defects in these tissues in crb mutant embryos and coincides with the onset of tissue morphogenesis. Thus our data using an allelic combination that strongly reduces exo84 function suggest that early tubule cell polarity is less dependent on the targeted secretion of membrane proteins, through activity of the exocyst, than later in development when exocyst-mediated, targeted membrane delivery underpins tubule cell polarity. These findings support the conclusions of Blankenship et al. (Blankenship et al., 2007), and lead us to suggest further that in the epidermis the initial localisation of Crumbs is largely independent of Exo84 activity but that as cell junctions and cell membranes are remodelled during germ-band extension, there is a dual requirement for both these proteins to maintain cell polarity.

Epithelial cell rearrangements are accompanied, and may be driven by, changes in the spatial arrangement of their intercellular contacts (Bertet et al., 2004; Classen et al., 2005; Zallen and Wieschaus, 2004). Contacts with new neighbours are engineered by the reorganisation of intercellular junctions through vesicle recycling; ZAs shrink as the interface disappears between cells that lose contact and expand as borders develop between new neighbours. These changes involve dynamin-dependent endocytosis of ZA components, rab-mediated vesicle trafficking and the targeted recycling of vesicles to specific membrane domains (Classen et al., 2005; Langevin et al., 2005). Here we show that Crb is specifically required to maintain apicobasal polarity and the integrity of ZAs as these cell activities occur. If cell rearrangements are halted or reduced in crb mutants, there is extensive rescue of epithelial integrity. This leads us to propose a model in which the Crb complex is dispensable for the establishment of cell polarity in embryonic epithelia but that as soon as morphogenetic cell rearrangements start, the Crb complex acts both to stabilise apical proteins and to restrict the spread of basolateral proteins (Fig. 8).

The requirement for E-cadherin in different tissues shows a similar dependence on the degree of morphogenetic activity (Tepass et al., 1996; Uemura et al., 1996). It was shown that the zygotic Drosophila E-cadherin mutant phenotype can be rescued in dynamic tissues, for example in the neurectoderm and Malpighian tubules, by suppressing morphogenetic cell movements (Tepass et al., 1996).

It is tempting to speculate that Crb acts by targeting recycling vesicles of ZA components in order to maintain junctional integrity in the elongating renal tubules. Without Crb ZAs are lost and membrane domains no longer remain distinct, leading to the collapse of cell polarity. Alternatively, lack of Crb could result in loss of cell polarity in morphogenetically active tissues and, as a consequence, ZAs cannot be maintained. In this case the primary requirement for Crb during cell movement would be to maintain the apical localisation of Baz/Par-6/aPKC, thereby also ensuring the normal distribution of basolateral proteins (Bilder et al., 2003; Hutterer et al., 2004; Tanentzapf and Tepass, 2003).

The exo84 mutant phenotype in remodelling tissues could be explained by a depletion of apical Crb caused by reduced exocyst-mediated delivery, resulting in a phenotype reminiscent of crb mutant embryos. This view is supported by recent findings showing that expression of dominant-negative Cdc42 in stage 9-11 embryos results in loss of Crb and other proteins from the apical membrane and disruption of cadherin-based adhesion in the morphogenetically active ventral neuroectoderm. Furthermore, mutations in crb enhance the phenotype induced by dominant-negative Cdc42 expression in this tissue (Harris and Tepass, 2008). The authors suggest that Cdc42 normally acts to repress endocytosis of apical proteins including Crb, so that inactivation of Cdc42 results in defects in cell polarity, leading to the loss of ZAs and tissue disruption. Whether Crb ensures ZA plasticity during cell rearrangements by restricting excessive endocytosis of apical proteins indirectly via stabilisation of Baz and/or Par-6 and Cdc42, or in a more direct way by regulating the recycling of junctional proteins remains to be determined. Similar defects in ZA integrity were also observed in the epithelium of the developing adult dorsal thorax upon loss of Cdc42 function. Here, it was shown that Cdc42–Par-6–aPKC control endocytosis by the cytoskeletal regulators Arp2/3 (Georgiou et al., 2008) and Cip4 and WASP (Leibfried et al., 2008). These results complement and support recent data showing that Cdc42, together with PAR-3, PAR-6 and PKC, are required for membrane trafficking in C. elegans coelomocytes and human HeLa cells (Balklava et al., 2007).

Members of the Crb complex also play a critical role in ZA stability and apical membrane delivery or stabilisation during photoreceptor development, when the ZAs enlarge and the apical domain selectively expands as the rhabdomere forms (Berger et al., 2007; Hong et al., 2003; Izaddoost et al., 2002; Pellikka et al., 2002; Richard et al., 2006). These morphogenetic events require the targeted delivery and retention of large amounts of membrane. Here too, it is not yet clear whether Crb acts directly on the stability of ZA components or indirectly, by controlling other polarity proteins. Although Drosophila Par-6 is delocalised in crb mutant photoreceptor cells (Berger et al., 2007), other data suggest that the Crb complex regulates ZA integrity and trafficking of apical membrane via stabilisation of the membrane-associated cytoskeleton, including β_H-spectrin (Pellikka et al., 2002) (Mélisande Richard, Nadine Muschalik and E.K., unpublished data).

Vertebrate homologues of members of the fly Crb complex appear to have conserved roles in the control of epithelial integrity. Loss of oko meduzy, one of the five zebrafish crb orthologues, of the sdt orthologue nagie oko or of mosaic eyes, a regulator of Crb, perturbs polarity and morphogenesis of the retinal neuroepithelium and the heart (Malicki and Driever, 1999; Rohr et al., 2006; Wei and Malicki, 2002). In addition, RNAi-mediated knock down of the mammalian Sdt orthologue, Pals1, in Madin Darby canine kidney (MDCK) cells in culture prevents proper delivery of E-cadherin to the cell surface (Wang et al., 2007), a phenotype strikingly similar to that of crb or sdt mutant epithelia in Drosophila.

These observations suggest that, as in the fly, vertebrate epithelia affected by the loss of Crb are those that undergo morphogenetic reorganisation, including cell shape change. Loss of human CRB1 is associated with retinitis pigmentosa and Leber congenital amaurosis, resulting in retinal degeneration and blindness (den Hollander et al., 1999), a phenotype with striking similarity to flies with crb mutant photoreceptor cells, which exhibit light-dependent retinal degeneration (Johnson et al., 2002). In mice mutant for Crb1, photoreceptor cells develop normally but later their ZAs degenerate so that photoreceptors in certain areas of the retina are displaced, followed by cell death (Mehalow et al., 2003; van de Pavert et al., 2004). Further experiments will elucidate whether the defects observed in morphogenetically active epithelia and photoreceptor cells are based on a common cell biological function of the Crb complex in these two cell types.

The mutant strains used were crb^11A22, crb^8F105, baz^xi106, scrib², onr^142-5, cut^c145, Kr¹ and Kr⁹. Maternal and zygotic mutants of baz^xi106 and scrib² were made by the FLP-DFS dominant female sterile method as previously described (Chou and Perrimon, 1996). exo84 mutant embryos were generated using onr^142-5 [as described by Blankenship et al. (Blankenship et al., 2007)]. For ectopic expression of Crb, UAS-crb^30.12e was expressed (Wodarz et al., 1995) using the Gal4 drivers da-GAL4 and cutB-GAL4.

The following primary antibodies were used: mouse anti-Arm (1:100; Developmental Studies Hybridoma Bank); rabbit anti-Baz (1:1000; a gift from A. Wodarz); rat anti-Crb (1:100); mouse anti-Cut (1:200; a gift from I. Rebay); mouse anti-Dlg (1:500; Developmental Studies Hybridoma Bank); mouse anti-Fas II (1:10; Developmental Studies Hybridoma Bank); mouse anti-Nrt (1:10; Developmental Studies Hybridoma Bank); rabbit anti-SAS (1:500; a gift from D. Cavener); rabbit anti-Tsh (1:300; a gift from S. Cohen). Heat fixation was used for the labelling of Arm, as previously described (Tepass, 1996). For all other staining, embryos were fixed in 4% paraformaldehyde, using standard techniques. Primary incubations were performed overnight, followed by incubation with appropriate biotinylated secondary antibodies and amplification using the Vector Elite ABC Kit (Vector Laboratories). For fluorescent double labelling, the appropriate secondary antibodies were used conjugated with FITC or Cy3 (Jackson Immunoresearch). When required, an additional amplification step using streptavidin-conjugated Cy3 or FITC (Jackson Immunoresearch) was performed.

For transmitted-light microscopy, embryos were mounted in 1,3-diethyl-8-phenylxanthine (DPX) and viewed using a Zeiss Axioplan compound microscope at room temperature with a ×20 (Pan-NeoFluar; NA 0.50; Zeiss) objective, a JVC-KY55B digital camera and Neotech Image Grabber. For confocal microscopy, embryos were mounted in Vectashield fluorescent mounting medium (Vector Laboratories) and viewed at room temperature with a ×40 oil (Leica PL APO; NA 1.25) objective on a Leica SP1 or SP5 scanning laser microscope, and LSM software. Images were processed in Adobe Photoshop 7.0.

Transmission electron microscopy was carried out using standard techniques (Grawe et al., 1996). Ultrathin sections were cut on a Reichert OM-U2 ultramicrotome and were analysed using a Philips electron microscope.