Expression of crumbs confers apical character on plasma membrane domains of ectodermal epithelia of drosophila

Cell, Vol. 62, 67-76,

July 14, 1995, Copyright

0 1995 by Cell Press

Expression of Crumbs Confers Apical Character on Plasma Membrane Domains of Ectodermal Epithelia of Drosophila Andreas Wodan; Uwe Hinz, Martin Engelbert, and Elisabeth Knust lnstitut fiir Entwicklungsbiologie Universittit zu Win 50923 KLiln Federal Republic of Germany

Summary The crumbs protein of Drosophila is an integral membrane protein, with 30 EGF-like and 4 lamlnin A G domain-like repeats in its extracellular segment, which is expressed on the apical plasma membrane of all ectodermally derived epithelia. Here, we’ present evidence to show that the insertion of crumbs into the plasma membrane is necessary and sufficient to confer apical character on a membrane domain. Overexpression of crumbs results in an enormous expansion of the apical plasma membrane and the concomitant reduction of the basolateral domain. This is followed by the redistribution of fiHuYY-spectrin, a component of the membrane cytoskeleton, and by the ectopic deposition of cuticle and other apical components into these areas. Strlkingly, overexpression of the membrane-bound cytoplasmic portion of crumbs alone is sufficient to produce this dominant phenotype. Our results suggest that crumbs plays a key role in speclfying the apical plasma membrane domain of ectodermal epithellal cells of Drosophila.

Epithelial tissues in multicellular organisms are characterized by a pronounced apicobasal polarity. This polarity is manifested not only in the asymmetric distribution of organelles or cytoplasmic proteins, but also in the differentiation of the plasma yembrane into two distinct domains, the apical and the basolateral domains, which, though physically continuous, differ in their protein and lipid compositions and serve distinct functions. The polarized distribution of cell surface components is generated by differential sorting of proteins and lipids in the trans-Golgi network and endosomes. Once localized in the appropriate domain, the diffusion of proteins within the membrane is restricted byspecializedjunctions, byattachment of proteins to the underlying cytoskeleton, by cell-cell and cell-substrate interaction, or by a combination of these (for reviews see Rodriguez-Boulan and Nelson, 1989; Simons and Wandinger-Ness, 1990; Nelson, 1992; Matter and Mellman, 1994). How cell surface polarity is established during develop*Present address: of Developmental 943053426.

Howard Biology,

Hughes Medical Institute and Department Stanford University, Stanford, California

ment is not well understood, and the mechanisms involved may vary depending on the organism considered. In the preimplantation mouse embryo, a polarized phenotype of the initially apolar blastomeres can be observed for the first time at the 8-cell stage, upon compaction, and is likely to be initiated by asymmetric cell contacts, which are thought to induce an asymmetric distribution of ions across the blastomeres (for reviews see Wiley et al., 1990; Fleming, 1992). Stabilization of polarity is subsequently achieved by the redistribution of a variety of proteins, including ion channels and cell adhesion molecules, to the apical or basolateral plasma membrane domains. One of the proteins that is of crucial importance for the establishment of cell polarity in the early mouse embryo is the homophilic cell adhesion molecule uvomorulin (E-cadherin). Mouse embryos homozygous mutant for the E-cadherin gene fail to establish a trophectoderm epithelium at the blastocyst stage (Larue et al., 1994; Riethmacher et al., 1995). In the Drosophila embryo, the first cleavages occur in the absence of cytokinesis, giving rise to the syncytial blastoderm. Cellularization takes place about 3 hr after fertilization by the invagination of plasma membranes from the periphery, which is accomplished by the addition of new material as well as unfolding of the highly pleated membrane of the egg (for reviews see Warn and Robert-Nicoud, 1992; Foe et al., 1993). Nothing is known so far about the mechanisms that lead to the initial establishment of polarity and to the differentiation of apical and basal membrane domains of cells at the blastoderm stage. Recent work in Drosophila has not only revealed that at least some of the basic mechanisms and proteins controlling epithelial cell polarity are conserved between invertebrates and vertebrates, but has also uncovered novel gene products (for reviews see Knust et al., 1993; Knust, 1994). Phenotypic analyses of two genes of Drosophila, crumbs (crb) and stardust (sdt), suggest that they act to control cell polarity of ectodermally derived embryonic epithelia. Loss-offunction mutations in either crb or SC%lead to embryonic lethality, and the mutants display nearly identical phenotypes, characterized by the loss of cell polarity in most ectodermally derived epithelia, followed by disintegration of epithelial structure and extensive cell death in some tissues. Strikingly, correct localization of proteins destined for the apical plasma membrane is already perturbed in both mutants before any morphological defects are detectable (Tepat3 et al., 1990; Tepa6 and Knust, 1990, 1993; Wodarz et al., 1993). crb encodes a large transmembrane protein (crb), which is expressed exclusively on the apical plasma membranes of all ectodermally derived epithelia from gastrulation onward(lepa6et al., 1990). The proteincontains30EGF-like repeats and 4 laminin A G domain-like repeats in its extracellular domain (Tepa6 et al., 1990; Patthy, 1991). The small cytoplasmic domain consists of only 37 amino acids, but nevertheless is of crucial importance for the function of crb: truncation of this domain leads to a complete loss

of function (Wodarz et al., 1993). The data presented in this paper suggest that crb, particularly its cytoplasmic portion, plays a key role in the control of cell polarity by specifying the apical plasma membrane domain of ectodermal epithelial ceils in Drosophila. Results Since loss of crb function results in the loss of cell polarity (Tepal3 et al., 1990; Tepa6 and Knust, 1990) we were interested in studying the effects on epithelial development of overexpression of the wild-type protein or modified versions thereof. For this purpose, we made use of the two component GAL4 system described by Brand and Perrimon (1993). Several effector strains were established that carry transgenescomprising different parts of the coding region of crb placed downstream of five GAM-binding sites (UASG; see Figure 1). Reproducible differences were observed with respect to the level of crb overexpression attained in the progeny of various effector strains crossed to the same activator strain, presumably owing to the chromosomal position of the effector construct (see below). Hence, we were able not only to activate crb in a variety of spatially and temporally restricted patterns using different activator strains, but also to manipulate the degree of overexpression by using different effector strains, thus producing a phenotypic series, which, in turn, facilitated understanding of the phenotype induced. Overexpression of Full-Length Crb Rescues the Mutant crb Phenotype To ensure its functional sufficiency, we tested the minigene encoding crbM for its ability to rescue the cuticle mutations. In embryos phenotype of crb loss-of-function homozygous mutant for crb, no contiguous cuticle is

crb

crb

wt

TM

SP

elm

Figure

1. Constructs

Used

in This Study

Schematic representation of the proteins encoded by the four constructs used in this study. Crb* encodes the full-length wild-type crb protein. EGF-like repeats are shown as stippled boxes and laminin A G domain-like repeatsas hatched boxes, respectively. Crbotimcontainsa premature stop codon immediately N-terminal of the transmembrane domain and, thus, encodes a secreted version of crb that lacks the transmembrane and cytoplasmic domains. Crb’*rm and crb’““lack the complete extracellular domain (dashed line), except for the N-terminal region including the signal peptide. Abbreviations: SP, signal peptide; TM, transmembrane region: triangle, MYC epitope (not to scale).

Figure 2. Expression of Crb” and Crb’“vs in crb Mutant Partially Rescue the Loss-of-Function Phenotype

Embryos

Can

(A) Cuticle preparation of an embryo homozygous for the amorphic allele crf~“*~~. The only remains of the cuticle are small grains that are presumably secreted by surviving tracheal cells, whereas most epidermal cells die. (B) Upon expression of crb” in c~W~~~ mutants using a GAL4 activator line that gives rise to GAL4 expression in a broad stripe of cells in each segment, major parts of the cuticle shield are restored, including the denticle belts. (C) Overexpression of crbln’*mr results in the same degree of rescue as using the construct encoding the full-length protein. Genotypes: ~rb”~~/crb”~~ (A); crb”9GAL43863; ~rW”~~/crb’~~~~ (B); crb’nr+9GAL43863; crb’u22/crb11A22 (C). Anterior is to the left, dorsal up.

formed (Figure 2A; Tepal3 and Knust, 1990). Expression of crbYL in embryos that lack a functional endogenous crb gene leads to a restoration of major parts of the cuticle, visible as a contiguous cuticular shield with welldeveloped denticle belts (Figure 28). We were not able to obtain complete rescue, since we did not use the crb promoter to drive GAL4 expression and the expression patterns of the activator lines available were not sufficiently similar to the expression pattern of the wild-type cfb gene. Nonetheless, the observed rescue convincingly shows that the protein encoded by crb” has all the properties of the endogenous protein and is functionally equivalent to it. The amount and quality of the rescued tissue depends on the expression level of the respective effector line: using aweak effector line results in a less pronounced rescue than effector lines with moderate expression levels, whereas the rescued embryos exhibit traits typical of the dominant phenotype upon use of a strong effector line (see below). Overexpression of Crb Leads to Structural Disruption of Ectodermally Derived Epithelia Overexpression of crb”’ in the wild-type epidermis leads to severe defects, the severity of which is strictly correlated with the level of crb overexpression as judged from the

Trmbs

and Epithelial

Cell Surface

Polarity

Figure 4. Morphology of the after Crb Overexpression Figure 3. Phenotypes pression

Caused

by Different

Degrees

of Crb Overex-

(A) a-Crb antibody staining of a wild-type embryo at early utage 12 (stages according to Campos-Ortega and Hartenstein [1965)), focus is on the lateral surface of the embryo. Prominent staining can be seen on the apical surface of the tracheal pits (arrowheads). (6) a-Crb antibody staining of an embryo of approximately the same age as in (A), exhibiting weak overexpression of crb driven by the patched activator, GAL4=‘. No effects on embryonic development are detectable. (C) Cuticle preparation of a wild-type first instar larva. Note the regular array of ventral denticle belts. (D) a-Crb staining and (E) cuticle preparation of embryos with a moderate level of crb overexpression. The denticle belts are reduced in size, fail to differentiate properly, and exhibit irregular spacing. The dorsal hole is the result of defective dorsal closure owing to overexpression of crb in the amnioserosa. (F) a-Crb staining and (G) cuticle preparation of embryos with a high level of crb overexpression. The denticle belts are reduced to tiny, rudimentary structures (arrowheads). The overall cuticle structure is extremely irregular. Genotypes: wild type (A); crb”-‘d/GAL45591 (6); wild type (C); crb”“Y GAL4=’ (D and E); crb”+zVGAL4 55g1 (F and G). Anterior is to the left, dorsal up.

level of expression i&whole-mount antibody stainings, and can easily be scored by the analysis of the cuticle of first instar larvae (Figure 3). Levels of crb overexpression in the epidermis that are only slightly higher than in wild-type (Figures 3A and 38) do not cause any visible cuticle defects and allow the development of fully viable larvae (Figure 3C), which develop into fertile flies. At higher levels of crb overexpression (Figures 3D and 3F), the appearance of the cuticle changes dramatically; the denticle belts, characteristic cuticular structures indicative of normal epidermal development, are reduced in size (Figure 3E) or are hardly detectable at all (Figure 3G). The structure of the cuticle in general is extremely irregular, and large numbers of compact cuticular grains are visible. This phenotype is the result of a massive disturbance of the epithelial tissue structure of the epidermis, as can be seen in histological sections. Instead of forming a regular monolayered sheet of epithelial cells with a pronounced apico-

Epidermis

in Wild-Type

Embryos

and

Parasagittal sections of awild-type embryo (A) and an embryo expressing high levels of crb ubiquitously (6) at stage 16, stained with toluidine blue. In the wild type, the epidermal cells have flattened surfaces and form a very regular, monolayered epithelial tissue. The underlying somatic muscles are connected to the epidermis via a segmentally repeated structure known as the apodeme (arrowheads). Upon overexpression of crb, the epithelial morphology is severely affected. Apodemes are present, but display a highly disorganized morphology. Genotypes: wild type (A); crb”-Y+; GAL4d’WZ/+ (6).

basal polarity (Figure 4A), epidermal cells are rounded, exhibit no obvious polarity, and occasionally lie on top of each other (Figure 46). The dominant phenotypes observed in the progeny of a cross between an effector line and an activator line is clearly reduced, if one copy of the endogenous cfb gene is removed (data not shown). We interpret this finding to mean that the transgene encodes a functional protein and that endogenous and overexpressed crb contribute additively to the dominant phenotype. Removal of one copy of the endogenous crb gene lowers the overall dosage of crb present in the cells, which leads to the same modification of the phenotype as the use of a weaker effector line. The degree to which the tissue morphology in various ectodermally derived epithelia is affected upon overexpression of crb varies, depending on the tissue considered. Strikingly, the tissues that are most sensitive to crb overexpression also show the most severe defects in crb and sdf loss-of-function mutants (TepaR and Knust, 1990, 1993; Wodarz et al., 1993). Particularly in the amnioserosa, ceil shape changes upon overexpression of crb are very prominent. The cells appear round, instead of flat as in wild type, and are no longer able to form a continuous epithelial sheet. Consequently, the amnioserosa does not invaginate, but remains on the dorsal surface of the embryo as an amorphic clump of cells (Figures 5A and 5B). In addition, dorsal closure of the epidermis does not occur. In contrast, overexpression of crb in the hindgut or in the salivary glands has only minor effects on epithelial morphology, such as the appearance of small indentations in their apical surfaces, giving the lumen a roughened appearance (Figures 5C and 5D). The results suggest that

Ceil 70

Figure 5. Phenotypic Consequences of Crb Gverexpression in the Amnioserosa and the salivary Glands (A) The amnioserosa of a stage 14 wild-type embryo appears as a squamous epitheiium that covers the dorsal side of the embryo. The embryo shown here carries a GAL4 enhancer trap insertion that gives rise to GAL4 expression in the amnioserosa and a UAS-/ecZ insert as reporter, the expression of which was monitored by staining with an anti+gaiactosidase antibody. (B) In addition to the GAL4 enhancer trap insertion and the UAB-/acZ transgene, this embryo carries also an UAS-orb transgene. Thus, ceils showing 5gatactosidase staining do also express high kvels of crb. Ceils of the amnioserosa have adopted a rounded shape and fail to form a contiguous epithelium. against the protein encoded by sas. Note the smooth

(C) The lumen of a salivary glend of a stege 13 wild-type embryo, stained with an anttbody appearance of the lumen and the restrict’ion of sas immunoreacttvity to the apical surface. (D) Upon strong overexpression of crb, the lumen of the salivary glands acquires a rough appearance that is characterized by the presence of numerous, sas-immunoreactive indentations in the apical surface. Low levels of sas staining can be seen on the basolaterai membranes of the salivary gland cells. (A) and (B) are views from the dorsal side, and (C)and (D) are from the lateral side. Genotypes: UAS-/acZ/+; GAL4=-Y+ (A); UAS-/acZ/+; GAL4%.V crbWa (B); wild type (C); crb*/+; GAL4 -I+ (D). Scale bar for (C) and (D) = 10um.

some ectodermal tissues strictly depend on crb function for maintenance of their epithelial tissue structure, whereas others can use alternative mechanisms that are functionally independent of crb. Inspection of the epidermis at the ultrastructural level was performed to understand further the cellular basis of the mutant phenotype produced by overexpression of crb. The wild-type epidermis of embryos at stage 16 is a monolayered epithelium, the apical, outer, face of which is covered by abundant microvilli, which secrete cuticle, visible as two electron dense layers surrounding a less dense layer and deposited at the tip of each microvillus, as previously described (Poodry, 1960; Figures 6A and 6E). After overexpression of crb, ectopic deposition of cuticle can be observed in areas of the cell surface that in the wild-type epidermis would form contacts with neighboring cells or with the extracellular matrix (Figure 6B). Very frequently, cuticle is deposited in deep clefts between adjacent cells. As a consequence, the cells separate from each other, leading to the isolation of small groups of cells or even single cells, which are completely surrounded by cuticle (Figure6B). Thedevelopment of junctions at the remaining contact sites between these cells is severely impaired, and in most cases, zonulae adherentes and septate junctionsare missing (Figures 6F and 6G). The abnormal deposition of cuticle also affects the attachment of muscles to the epidermis. Normally, the somatic muscles connect to the epidermis at the basalmost point of a specialized epidermal cell, the tendon cell, which, like the adjacent epidermal cells, shows the typical polarity of an epithelial cell (Figure 6C). Upon overexpression of crb, the tendon cell secrets cuticle even at the basal tip, immediately adjacent to the muscle insertion site (Figure 6D). Owing to the loss of the monolayered tissue structure, the tendon cell does not reach the surface of the embryo, but is buried by overlying epidermal cells.

Insertion of Crb Confers Apical Character on a Plasma Membrane To understand better the basis of the dominant phenotype caused by crb overexpression, we looked at the localization of crb protein and other apical and basolateral markers with the confocal laser scanning microscope. In contrast with previous reports (Tepa6 and Knust, 1993; Wodarz et al., 1993) at this level of resolution, the endogenous crb is highly enriched in the apicolateral region of the plasma membrane, close to the position of the zonula adherens (Figure 7A), and not uniformly distributed on the surface as suggested by examinations of histological sections. When overexpressed at low levels, increased crb staining is visible in the same locations as in wild type (data not shown). At high levels of overexpression, however, the proportion of the plasma membrane containing crb is much higher (Figure 78). This misdistribution isvisible before cell shape changes become manifest. To test whether the misdistribution of crb leads to altered localization of other apical markers as well, we performed double immunofluorescence stainings using an antibody against the fibronectinlike transmembrane protein encoded by the stranded at second (sas) gene (Schonbaum et al., 1992). In wild-type embryos, the expression of sas is restricted to the apical plasma membrane. Unlike crb, sas is evenly distributed on the apical membrane (Figure 7A). As a result of crb overexpression, the localization of sas is dramatically altered. Sas can now be seen on a much greater fraction of the cell surface and colocalizes to a high degree with the overexpressed crb (Figure 78). Very similar observations were made for the apically secreted protein encoded by the yellow locus (Kornezos and Chia, 1992; data not shown). This suggests that, upon association of crb with a plasma membrane, this membrane acquires apical characteristics and attracts other apical proteins. To determine whether the expansion of the cm-positive apical mem-

~,umbs

and Epithelial

Cell Surface

Polarity

Figure 6. Ultrastructural Analysis of the Epidermis in Wild-Type Embryosand after Overexpression of Crb (A) Electron micrograph of the epidermis of a wild-type embryo. At this stage, zonulae adherentes (arrows) and septate junctions (arrowhead) are present in all epidermal cells (TepaB and Hartenstein, 1994). (B) Electron micrograph of the epidermis of an embryo overexpressing crb at a high level. The cells fail to organize a normal epithelium and exhibit an altered apicobasal polarity. Many cells possess two or more surface areas covered by cuticle (asterisk) owing to ectopic secretion of cuticle (arrows). (C) Electron micrograph of a wild-type apodeme. The tendon cell (asterisk) shows the typical polarity of an epithelial cell. (D)The apodeme of an embryooverexpressing crb at a high level. Note the deposition of cuticle at the basal tip of the tendon cell, immediately adjacent to the muscle insertion site (arrow). (E) Epidermis of a wild-type embryo at stage 16 at higher magnification. The most obvious marker for the apical face of the epidermis is the electron-dense layer of cuticle, which is in contact with microvillus-like protrusions of the apical plasma membrane (inset). Zonula adherens (arrow) and septate junction (arrowhead) are indicated. (F and G) Epidermal cells of embryos overexpressing crb at high levels occasionally develop abnormal junctions (F; arrowhead) or, more often, lack junctions completely (G). Note that the cells secrete cuticle not only on the apical, but also on the basal surface. All sections shown in this figure are from embryos at stage 16 (15-22 hr AEL). Genotypes: wildtype (A, C, and E); crb”Y+; GAL4daGz/+ (8, D, F, and G). Scale bars in (A)-(D) = 5 urn. in (E)-(G) = 1 urn.

. brane domain is associated with a reduction of the basolatera1 membrane, we monitored the subcellular localization of Fasciclin Ill (Faslll), a basolaterally expressed integral membrane adhesion molecule (Pate1 et al., 1987; Snow et al., 1989) and compared it with the expression of sas in wild-type and mutant embryos. In the wild-type epidermis, Faslll is expressed all over the basolateral plasma membrane domain and is excluded from the apical plasma membrane. The expression domains of Faslll and sas are therefore mutually exclusive, except for a very narrow region of colocalization at the most apical part of the lateral membrane (Figure 7C). In embryos overexpressing crb, the area that shows expression of Faslll is clearly reduced, but most importantly, the expression domains of Faslll and sas are still mutually exclusive, except for the very narrow region of colocalization that is also seen in the wild-type epidermis (Figure 7D). This finding clearly shows that the dominant phenotype associated with crb overex-

pression is not simply caused by a loss of cell polarity and random delivery of proteins to the plasma membrane, but is the result of the enlargement of the apical plasma membrane domain associated with a reduction of the basolateral plasma membrane domain. Hence, overexpression of crb leads to a clear gain-of-function phenotype, which is precisely the opposite of that caused by loss-of-function mutations, which lead to a complete loss of apical character. Overexpression of Crb Leads to a Relocalization of fin,,-Spectrin to Basolateral Membranes is a component of the membrane cyPt+eaV(P+spectrin toskeleton and shows a highly polarized localization in most ectodermal tissues, e.g., in the epidermis, the foreand hindgut, and the tracheal system (Thomas and Kiehart, 1994). In the epidermis, f3++-spectrin is particularly concentrated in apicolateral positions close to the zonula

Cdl 72

Figure 7. Overexpression of Crb Changes the Distribution of Other Apical and Basolateral Markers (A) Double label immunofluorescence image of the epidermis of a wild-type embryo at stage 14. The green label marks the crb protein and the red label the sas protein. Sas is evenly distributed on the apical plasma membrane, whereas crb is concentrated in apicolateraf positions close to the zonula adherens. The yellow color indicates cofocafiration of crb and sas. The bright& staining structures betow the epidermis are branchesof the tracheal system, which express both antigens. (B) The epidermis of an embryo at the same age as the embryo in (A) overexpressing high levels of crb. Crb can be seen not only on the apical surface of the epiderrnal cells, but also on the basolateral membrane. Thesame is true for sas, which shows a high degree of colocalization with the overexpresaed crb. (C) The epidermis of a wild-type embryo at stage 16, stained wtth antibodies egatnst sas (red) and Faslll (green). As in (A), the sas protein is evenly distributed on the apical membrane, whereas Faalll is present on the baaofateral plesma membrane. The expression domains of sas and Faslll are mutually exclusive. (0) The epidermis of an embryo at approximately the same age as the embryo in (C) overexpressing high levels of crb. &topic expression of sas (red) can be seen on the basal side of many cells, while the portion of the plasma membrane that shows expression of Faslll is clearly reduced. As in (C), the expression domains of sas and Faslll show very little overlap. Genotypes: wild type (A and C); crb@V+; GAL4 -I+ (B and D). Scale bar = 10 urn.

adherens (Figure SA) and thus colocalizes with armadillo, Notch, and crb (Thomas and Kiehart, 1994; our own observations, compare Figure 7A). Upon overexpression of crb, the localization of j3H-spectrin changessignificantly (Figure

Figure 8. Aftered LoW&fon of t&SgecWn in tfre Epfdermis of Crb Overexpressing Embryos (A) The epidermis of a wild-type embryo at stage 14, stained with an antibody against &epectdn. similarly to crb, f3n-spectrin is localized at the apicolateral plasma membrane, ckxsely adjacent to the zonula adherens. (8) In the epidermis of embryos overexpressing crb, bk-spectrin is detectableat basolateral membranes, indkzatlngasignifrcant modification of the cytoskefetaf organization in these cells. Genotypes: wild type (A): crb-/+; GAL4-/+ (B). Scale bar = 10 urn.

86). The protein is no longer concentrated at apicolateral positions, but is now also present at the basolateral membrane. PH-spectrin, like other members of the spectrin superfamily, contains several binding sites for other cytoskeletal proteins, particularly actin, as judged from the cDNA sequence (Dubreuil et al., 1990). Thus, it is likely that the observed redistribution of j3H-spectrin in crb-overexpressing embryos is a manifestation of a profound reorganization of the membrane cytoskeleton in these cells. Overexpression of a C-Terminal Deletion Mutant of Crb Does Not Affect Epithelial Polarity To test which part of the crb molecule is responsible for the dominant phenotype, we generated two complementary deletion mutants of crb, one (crbwa) containing only the extracellular part of the protein and the other (crbim) only the signal peptide, transmembrane domain, and the cytoplasmic domain (see Figure 1). Cverexpression of crbem, even at very high levels, had no obvious consequences for epithelial development, nor for viability or fertility of embryos, larvae, and adult flies. Conversely, we did not observe any rescue of the crib loss-of-function cuticle phenotype upon expression of crbema in homozygous mutant crb embryos. This finding is rather surprising, since the extracellular domain of crb comprises 97% of the fulllength crb protein (2082 amino acids). We can exclude the possibility that the truncated crbprotein was retained in the cells rather than being secreted, since we observed proper secretion into the lumen of the salivary glands (Figures 9C and 9D), as well as deposition of secreted crbema protein on the apical surface of the epidermis (data not shown). At high levels of overexpression, the secretion of crbexha, like the misdistributed full-length crb, was not

y;rnbs

and Epithelial

Cell Surface

Polarity

Figure 9. Properties Crb Proteins

of Cverexpressed

Mutant

Sagittal optical sections of a wild-type embryo at stage 12 (A) and of an embryo overexpressing high levels of crb’“‘T (B), slightly older than the one shown in (A), stained with anti-crb antibody. While the staining in the wild-type embryo is restricted to the apical surface of the epidermis and the lumen of the fore- and hindgut, crb staining of an embryo overexpressing crb”rrm*c is not restricted to the apical surface in the epidermis. (C) Salivary gland of a wild-type embryo at stage 16, stained with the anticrb antibody. Endogenous crb is expressed on the apical membrane of the salivary gland cells, thus outlining the lumen. (D) Salivary gland of an embryo overexpressing crbmra. Since both the endogenous crb and crbl”“” are recognized by the crbantibody, the staining is a superimposition of the expression of both proteins. However, most of the staining seen in (D) is due to expression of crbwr”, since the staining is much more intense than in a wild-type salivary gland. Note that strong staining at the apical surface of the salivary gland cells. can only be seen in the lumen, indicating that crb enra is efficiently secreted Abbreviations used: ep, epidermis; fg, foregut; hg, hindgut. Genotypes: wild type (A and C); crb mtfsmyc38‘da,+; GA,&M332,+ (B); &-,e”W4 “‘S,GAL4559 1 (D). Scale bars for (A) and (B) = 100 Mm, for (C) and (D) = 10 pm.

restricted to the apical plasma membrane not shown).

domain

(data

Overexpression of the Cytoplasmic Domain of Crb Causes the Same Dominant Phenotype as Overexpression of the Full-Length Protein Cverexpression of both crbintra and crbintrs’“yc (see Figure 1) mutant proteins, which completely lack the extracellular segment, caused exactly the same dominant phenotype as overexpression of the full-length crb protein. At high levels of overexpression, the crbintra-mycprotein is detectable in all plasma membranes, without any significant preference for the apical membrane domain (data not shown). In this situation, we observed a misdistribution of the endogenous crb protein that was dependent on the level of overexpression of crbintra-myc.Instead of being expressed exclusively on the outer face of epithelial cells (Figure 9A), the endogenous &b was also detectable on membranes that normally contact neighboring cells or the basal lamina (Figure 96). These findings clearly indicate that overexpression of the membrane-bound cytoplasmic tail of crb alone is sufficient to modify epithelial cell surface polarity and to increase the proportion of the plasma membrane that exhibits apical characteristics. These results can be explained in two ways: one possibility is that the expression of the cytoplasmic part of crb on the plasma membrane is itself sufficient to specify apical character, or alternatively, that it is the accompanying distribution of the endogenous crb that is responsible for apical quality. To discriminate between these two possibilities, crbintra-mycwas expressed in a crb mutant background devoid of any endogenous crb protein. Strikingly, moderate levels of crbi”rra-mv expression in the epidermis of embryos lacking any endogenous crb protein not only rescued the cell lethality associated with

crb loss-of-function mutations, but also was sufficient to restore a nearly wild-type cuticle pattern including denticle belts (see Figure 2C). In fact, we were not able to find any difference in the capability to rescue the crb mutant phenotype between crb intra-myc and crb* (compare Figures 28 and 2C). This finding clearly shows that the cytoplasmic domain of crb is necessary and sufficient to induce the dominant phenotype and, thus, represents the functional part of the crb protein that is involved in specifying the apical characteristics of the plasma membrane. Discussion Data presented in this paper show that the insertion of the transmembrane crb protein is necessary and sufficient to confer apical character on the plasma membrane of embryonic epidermal cells of Drosophila: loss of crb function leads to the loss of cell polarity in many ectodermally derived epithelia, which becomes manifest in an inability to localize apical proteins properly in crb mutants (Wodarz et al., 1993). This may be explained by a loss of apical identity, since overexpression of crb, on the other hand, results in an expansion of the apical membrane domain, which, in the most extreme case, leads to a complete “apicalization” of the plasma membrane. This phenotype is characterized by the ectopic expression of several apical markers (e.g., sas and yellow), the ectopic formation of microvilli, and secretion of cuticle. Expansion of the apical membrane domain is accompanied by a reduction of the basolateral membrane domain, since expression of the basolateral marker Fad// is virtually excluded from membrane domains that show expression of apical markers. There are several possibilities to explain the misdistribution of crb to membranes that normally have basolateral characteristics. First, the phenotype could result simply

Cell 74

from a tremendous expansion of the apical surface due to the insertion of abundant crb protein. Second, if one assumes that the apical protein sorting machinery in the Drosophila epidermis has a limited capacity, overexpression of crb may be sufficient to cause its diversion into an abnormal sorting pathway. A comparable observation has recently been made for the low density lipoprotein (LDL) receptor: overexpression of two mutant versions in polarized Madin-Darby canine kidney (MDCK) cells results in missorting of these proteins to the apical plasma membrane, suggesting that the basolateral sorting machinery of these cells can be saturated to some extent (Matter et al., 1992). We can exclude the possibility that the observed dominant phenotypes are caused simply by random delivery of all kinds of proteins due to overloading of the secretory machinery, since neither GAM-mediated overexpression of crbema, nor of Serrate, another transmembrane protein with EGF-like repeats, has any such effect (this work; Speicher et al., 1994). Finally, the misdistribution of crb could be explained if one assumes that crb is normally delivered to the apical and the basolateral plasma membrane, but is specifically retained on the apical membrane, mediated by association with either cytoplasmic factors (e.g., cytoskeletal proteins) or with extracellular components. Overexpression of crb could then lead to the saturation of available ligands at the apical plasma membrane, the saturation of the apparatus required for endocytosis on the basolateral membrane, or both. Specific retention at the apical plasma membrane by components of the cytoskeleton has recently been repotted for the epithelial Na+ channel (arENaC) in primary rat alveolar epithelial cells (Rotin et al., 1994). In that study, it was shown that a proline-rich sequence at the C-terminus of the poreforming unit of this channel binds specifically to the SH3 domain of a-spectrin, thus anchoring the channel in the apical membrane cytoskeleton. How crb confers apical character on a membrane domain is not yet understood. Its predominant apicolateral localization close to the zonula adherens may suggest a function in the deV8lOpment of these junctions. In fact, both loss of crb and its overexpression lead to defects in junction biogenesis (F. GraW8, A. W., B. Lee, E. K., and H. Skaer, unpublished data). Strikingly, the small cytoplasmic portion of the crb protein is not only necessary for this process, since removal of the C-terminal 23 amino acids completely abolishes cfb function (Wodarz et al., 1993), but is also sufficient to perform this function as shown here. One likely possibility is that this region may interact with components of the cytoskeleton, thereby modifying the cytoarchitecture of the cell, including microtubules, which contribute fundamentally to the COntrOl of motor-mediated vesicle trafficking of organelles and vesicles in the cytoplasm and, hence, to the organization of epithelial cells (Gilbert et al., 1991; Lafont et al., 1994; Mays et al., 1994). The latter is assumed to be modified by cell-cell and cell-substrate interactions mediated by adhesion molecules, such as cadherins and integrins (McNeil1 and Nelson, 1992). Thus, expression of the adhesion molecule uvomorulin in nonpolarized fibroblasts in-

duces the redistribution of Na+/K+-ATPase, ankyrin, and fodrin to sites of cell-cell contacts (McNeil1 et al., 1990; Nelson et al., 1990). This process is blocked by truncation of the cytoplasmic portion of uvomorulin, which is linked to the cytoskeleton by a family of cytoplasmic proteins, the catenins (for review s8e Kemler, 1993). In the experiments described here, the relocalization of Bt+-spectrin upon crb overexpression indicates altered organization of the membrane cytoskeleton. Since in the wild-type embryo both crb and p,+sp8ctrin are colocalized at the apicolateral plasma membrane, the altered localization Of BH-spectrin in crb overexpressing embryos could be caused by adirect physical interaction between these two proteins. Indirect sup port for this assumption comes from the observation that the cytoplasmic tail of crb contains a cluster of three consecutive proline residues that are deleted in the amorphic allele crbeFIOs (Wodarz et al., 1993). On8 may speculate that this region might be a potential binding site for the SH3 domain of &spectrin, although we cannot exclude the possibility that the relocalization of PH-spectrin is only an indirect consequence of crb overexpression. One intriguing question remains with respect to the initial establishment of cell surface polarity and the mechanism(s) that induces and maintains the apical localization of crb during early Drosophila embryogenesis. During cellularization of the embryo, only part of the forming basolatera1 plasma membrane is synthesized de novo, while part of it is derived by unfolding of the initially highly pleated membrane at the apex of the forming cells (Foe et al., 1993), implying that at this early stage no separation into two distinct membrane domains has yet occurred. crb transcripts are already present in the Cellular blastoderm, Where they are apically localized (Tepal3 et al., 1990), but the crb protein present at this stag8 is not associated with th8 plasmamembrane(A. W. and E. K., unpublished data). Only from the beginning of gastrulation onward can the crb protein be detected in association with the apical membrane, raising the question as to the mechanisms that initiate and maintain the apical localization of the crb protein. So far, we have no information concerning the sorting signal(s) that controls the apical delivery of the crb protein. Both apical and basolateral proteins have a common biosynthetic pathway, leading from the endoplasmic reticulum through the Golgi complex, and are in many instances differentially sorted in the trans-Golgi network. There is increasing evidence that, at least in MDCK cells, sorting of proteins to the basolateral membrane domain is controlled by signals that reside in their cytoplasmic domains and that removal of these sequences may result in their default delivery to the apical membrane. Less information is available on signals controlling apical localization of proteins. Only in one case, the influenza virus hemagglutinin, is the signal known to reside in the protein itself, while in other cases, a glycosylphosphatidylinositol (GPI) anchor or carbohydrate moieties may act as apical sorting signal (for review see Matter and Mellman, 1994). Diffusion of the crb protein from the apical to the basolateral membrane could be prevented either by a barrier within the membrane, not visible at the ultrastructural level, or alter-

;r;Jmbs

and Epithelial

Cell Surface

Polarity

natively, by interactions with other components, mediated by either the cytoplasmic region of crb or its large extracelMar segment containing EGF-like and laminin A G domain-like repeats. So far, none of the putative partners nor the part of the crb protein necessary for these interactions is known. Further experiments aimed at identifying the various interacting components will give us further insight into the control of a process that is crucial for the maintenance of cell surface polarity. Experimental

Procedures

Generation of Crb Effector Constructs, the daG Activator Construct, and Germline Transformation For all crb effector constructs, inserts were cloned between the EcoRl and Xbal restriction sites of the pUAST polycloning site (Brand and Perrimon, 1993). For the wild-type construct crb”, the full-length crb minigene described in Wodarz et al. (1993) was cloned into pUAST. To construct crb’“‘” and crb’“L-R, which lack the: entire extracellular segment except the N-terminus including the signal peptide, a PCR fragment was generated containing base pairs l-500. As template for this reaction, we used thecrbcDNAc2-2(Tepa6etal., 199O)subcloned into the Bluescript vector, with the 73 primer of the Bluescript vector as sense primer and primer 627 (5”ATCAGCAGATCTGGAGCCATTAAAGTAC-37 containing a newly introduced Bglll site not present in cDNA ~2-2 as antisense primer. The PCR product was cleaved with EcoRl and Bglll and ligated into EcoRI-Bglll-cut pUAST to generate ~36.1. A second PCR fragment was generated using a genomic KpnlEcoRl fragment as template, which covers the 3’part of the crb coding region. With primers 620 (!j’-CATlCGAGATClTCGACCACAGACAlTG9’, sense), which also contains a newly introduced Eglll site, and primer 622 (5”CGATACTCTAGAGCAATGT-3’, antisense), which contains the genomic Xbal site 3’of the polyadenylation signal, the region between position 6444 and the Xbal site was amplified. After cleavage with Bglll and Xbal, the PCR product was ligated into Bglll- and Xbal-cut ~36.1 to generate crb’*“. To construct crb’“r”m*, a double-stranded oligonucleotide encoding an epitope of the human c-MYC protein, which is recognized by the monoclonal antibody 9E-10 (Evan et al., 1965), was inserted into the Bglll site of crb’““. For constructing crbefifa, a secreted version of the crb protein, a premature stop codon was introduced immediately N-terminal to the transmembrane domain by changing a G at position 6434 into T by site-directed mutagenesis using primer 542 (5’-CCCGTCCTACTTGGCCGT-3’), which results in a change of the triplet GAG coding for glutamic acid to a termination codon (TAG). Numbering of bases is according to the updated version of the crib cDNA sequence (EMBL accession number M33753), which differs in several positions from the originally reported sequence (lepa6 et af., 1990). To construct pdaG, a 3.2 kb Kpnl-Bglll fragment encompassing the da promoter, and first intron was isolated from pBluescript-da1 ABam (provided by H. V&sin, M. Brand, and Y. N. Jan) and ligated into pGATB (Brand and Perrimon, 1993), linearized with Kpnl and BamHI. From the resulting plasmid, a 6.3 kb Kpnl-Notl fragment was excised and ligated into pW5N, cut with Kpnl and Notl (provided by M. Haenlin). P element transformation of the constructs was done according to the procedure described by Spradling (1966). We used w”” as recipient strain. At least eight independent transgenic lines were established for each construct.

Activation of Effector Constructs by GAL4 Activator Lines The properties of the GAL4 system have been described by Brand and Perrimon (1993). Several activator strains were used in this work, most of which were generated using the enhancer trap technique: GAL4-’ expresses GAL4 under the control of the patched promoter (Hinzet al., 1994); GAL43323 expresses GAL4 in the amnioserosa and the salivary glands (U. H., unpublished data); GAL4daW2, described above, gives a more or less uniform expression pattern; GAL4JBS3 expresses GAL4 in a broad stripe of cells per segment.

Immunocytochemistry, Cuticle Preparations, and Transmlnslon Electron Mlcroscopy Whole-mount stainings of embryos with monoclonal ant&b antibody Cq4 (Tepa6 and Knust, 1993) and anti-g-galactosidase (Boehringer) were done as described in Wodarz et al. (1993). Cuticle preparations were made as described by Wieschaus and Niisslein-Volhard (1966). Preparation of specimens for transmission electron microscopy was done as described in Tepa6 and Hartenstein (1994). Confocal Laser Scanning Microscopy Embryos were essentially fixed and washed as described in Wodarz et al. (1993), except that 4% paraformaldehyde in PBS was used as fixative. Dilutions of primary antibodies were as follows: anti-crb Cq4 1:2; rabbit anti-sas 1:500 (E. Organ and D. Cavener, unpublished data); anti-Faslll 7GlO 1:4 (Pate1 et al., 1967); rabbit anti-&-spectrin I:500 (Thomas and Kiehart, 1994). For single label immunofluorescence stainings, either FITC-conjugated goat anti-mouse (Sigma) or goat anti-rabbit (Vector) secondary antibodies were used at 1 :lOO. In double label experiments, theadditional secondary antibody waseither biotinylated sheep anti-mouse (Sigma) or goat anti-rabbit (Vector) at 1 :lOO, followed by incubation in streptavidin-Texas red (Boehringer) at 1:50. Embryos were mounted in Vectashield (Vector) mounting medium and viewed on a Bio-Rad MRC 1000 confocal laser scanning microscope equipped with COMOS software (Bio-Rad). Acknowledgments Correspondence should be addressed to E. K. The authors would like to thank the following people who have contributed to making this work possible. We are particularly indebted to E. Organ and D. Cavener, who allowed us to use their antibody against sas prior to publication, to I. Stiittem for teaching us how to inject embryos, and to F. Graweforan introduction toelectron microscopytechniques. Wethank A. Brand, N. Perrimon, and M. Levine for providing us with fly stocks prior to publication, M. Haenlin for pWBN, H. V&sin, M. Brand, and Y. N. Jan for the da construct, and M. Wasser and 8. Giebel for its initial modification. We would like to thank D. Fristrom for giving us valuable information about antigens with polarized expression patterns and W. Chia, C. Goodman, G. Thomas, D. Brower, D. Fristrom, and D. Cavener for sending antibodies. Special thanks to R. Nusse, in whose laboratory the final experiments for this paper were performed. We thank J. A. Campos-Ortega and P. Hardy for critical comments on the manuscript. This work was supported by a doctoral fellowship from Boehringer lngelheim Fonds and a postdoctoral fellowship from the Deutsche Forschungsgemeinschaft to A. W. and a grant from the Deutsche Forschungsgemeinschaft to E. K. Received

September

23, 1994; revised

May 19, 1995.

References Brand, A. H., and Perrimon, N. (1993). Targeted gene expression as a means of altering cell fates and generating dominant phenotypes. Development 178, 401-415. Campos-Ortega, J. A., and Hartenstein, Development of Drosophila melanogaster

V. (1965). The Embryonic (Berlin: Springer-Verlag).

Dubreuil, R. R., Byers, T. J., Stewart, C. T., and Kiehart, D. P. (1990). A 6-spectrin isoform from Drosophila (&) is similar in size to vertebrate dystrophin. J. Cell Biol. 717, 1649-1656. Evan, G. L., Lewis, G. K., Ramsay, G., and Bishop, Isolation of monoclonal antibodies specific for human oncogene product. Mol. Cell. Biol. 5, 3610-3616.

J. M. (1965). c-myc proto-

Fleming, T. P. (1992). Trophectoderm biogenesis in the preimplantation mouse embryo. In Epithelial Organization and Development, T. P. Fleming, ed. (London: Chapman and Hall), pp. 11 l-136. Foe, V. E., Odell, G. M., and Edgar, B. A. (1993). Mitosis and morphogenesis in the Drosophila embryo. In The Development of Drosophila melanogaster, M. Bate and A. Martinez-Arias, eds. (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press), pp. 149-300. Gilbert,

T., Le Bivic, A., Quaroni,

A., and Rodriguez-Boulan,

E. (1991).

Microtubular organization and its involvement in the biogenetic pathwaysof plasma membrane proteins in Caco-2 intestinal epithelial cells. J. Cell Biol. 113, 275-288. Hinz, U., Giebel, B., and Campos-Ortega, J. A. (1994). The basichelix-loophelixdomainof Drosophilalethalofscuteprotein issufficient for proneural function and activates neurogenic genes. Cell 76, 7787. Kemler, R. (1993). From cadherins to catenins: cytoplasmic interactions and regulation of cell adhesion. Trends Genet. 321. Knust, Trends

E. (1994). Control of epithelial Genet. IO, 275-280.

development

protein 9, 317-

in Drosophile.

Knust, E., Tepa8, U., and Wodarz, A. (1993). crumbs and stardust, two genes of Drosophila required for the development of epithelial cell polarity. Development (Suppl.) 119, 261-268. Kornezos, A., and Chia, W. (1992). Apical secretion and association of the Dro.sopni/a yellow gene product with developing larval cuticle structures during embryogenesis. Mol. Gen. Genet. 235, 397405. Lafont, F., Burkhardt, J. K., and Simons, K. (1994). microtubule motors in basolateral and apical transport Nature 372, 801-803.

Involvement of in kidney cells.

Larue, L., Ohsugi, M., Hirchehain, J., and Kemler, R. (1994). E-cadherin null mutant embryos fail to form a trophectoderm epithelium. Proc. Natl. Acad. Sci. USA 91, 8263-8267. Matter, K., and Mellman, I. (1994). Mechanismsof cell polarity: sorting and transport in epithelial cells. Curr. Opin. Cell Biol. 6, 545-554. Matter, K., Hunziker, W., and Mellman, I. (1992). Basolateral sorting of LDL receptor in MDCK cells: the cytoplasmic domain contains two tyrosine-dependent targeting determinants. Cell 71, 741-753. Mays, R. W., Beck, K. A., and Nelson, W. J. (1994). Organization and function of the cytoskeleton in polarized epithelial cells: a component of the protein sorting machinery. Curr. Opin. Cell Eiol. 6, 16-24. McNeill, H., and Nelson, W. J. (1992). Shaping an epithelial cell: the role of cell adhesion molecules in the reorganization of the membrane cytoskeleton. In Epithelial Organization and Development, T. P. Fleming, ed. (London: Chapman and Hall), pp. 53-88. McNeill, H., Ozawa, M., Kemler, R., and Nelson, W. J. (1996). Novel function of cell adhesion molecule uvomorulin as an inducer of cell surface polarity. Cell 62, 309-316. Nelson, W. J. (1992). Regulation of cell surface to mammals. Science 258, 948-955.

polarity

from bacteria

Nelson, W. J., Shore, E. M., Wang, A. Z., and Hammerton, R. W. (1990). Identification of a membrane cytoskeleton complex containing the cell adhesion molecule uvomorulin (E-cadherin), ankyrin and fodrin in Madin-Darby canine kidney epithelial cells. J. Cell Biol. 170, 349357. Patel, N., Snow, P. M., and Goodman, C. S. (1987). Characterization and cloning of fasciclin Ill: a glycoprotein expressed on a subset neurons and axon pathways in Drosophila. Cell 48, 975-988. Patthy, L. (1991). sophila and their 289, 99-101.

Laminin possible

of

A related domains in CRB protein of Drorole in epithelial polarization. FEBS Lett.

Poodry, C. A. (1980). Epidermis: morphology and development. In The Genetics and Biology of Drosophila, Volume 2, (London: Academic Press), pp. 443497. Riethmacher, D., Brinkmann, V., and Birchmeier, C. (1995). Atargeted mutation in the mouse E-cadherin gene results in defective preimplantation development. Proc. Natl. Acad. Sci. USA 92, 855-859. Rodriguez-Boulan, E., and Nelson, W. J. (1989). Morphogenesis the polarized epithelial cell phenotype. Science 245, 718-725.

of

Rotin, D., Bar-Sagi, D., O’Brodovich, H. O., Merilainen, J., Lehto, V. P., Canessa, C. hf., Rossier, 8. C., and Downey, G. P. (1994). An SH3 binding region in the epithelial Na+ channel (arENaC) mediates its localization at the apical membrane. EMBO J. 13, 44484450. Schonbaum, C. P., Organ, E., Qu, S., and Cavener, D. R. (1992). The Drosophile melanogaster stranded at second (sas) gene encodes a putative epidermal cell surface receptor required for larval develop ment. Dev. Biol. 157. 431445.

Simon% K., and Wandinger-Ness, thelia. Cell 62, 207-210.

A. (1990).

Polarized

sorting

in epi-

Snow, P. M., Bieber, A. J., and Goodman, C. S. (1989). Fasciclin Ill: a novel homophilic adhesion molecule in Drosophila. Cell 59, 313323. Speicher, S. A., Thomas, U., Hinz, U., and Knust, E. (1994). The Serrate locus of Dfosophi/a and its role in morphogenesis of wing imaginal discs: control of cell proliferation. Development 120, 535-544. Spradling, A. C. (1966). P element mediated sophile, a Practical Approach, D. 8. Roberts, pp. 175-197. TepaE, U., and Hartenstein, junctions in the Drosophi/a

transformation. In Droed. (Oxford: IRL Press),

V. (1994). The development of cellular embryo. Dev. Biol. 167, 563-596.

Tepa8, U., and Knust, E. (1990). Phenotypicand developmental analysis of mutations at the crumbs locus, a gene required for the development of epithelia in Drosophile melanogaster. Roux’s Arch. Dev. Biol. 199, 189-206. TepaE, U., and Knust, E. (1993). crumbs and stardust act in a genetic pathway that controls the organization of epithelia in Drosophila me/anogaster. Dev. Biol. 759, 311-326. Tepa8, U., There% C., and Knust, E. (1990). crumbs encodes an EGFlike protein expressed on apical membranes of Drosophila epithelial cells and required for organization of epithelia. Cell 67, 787-799. Thomas, G. H., and Kiehart. D. P. (1994). EiH-,-spectrin tissue and subcellular distribution during Drosophila Development 120,2039-2050.

hasa restricted embryogenesis.

Warn, R. M., and Robert-Nicoud, M. (1992). The formation and fate of the blastoderm epithelium of the Drosophi/a embryo. In Epithelial Organization and Development, T. P. Fleming, ed. (London: Chapman and Hall), pp. 136-l 61. Wieschaus, E., and NOsslein-Volhard, In Drosophila, a Practical Approach. Press), pp. 199-227.

C. (1988). Looking at embryos. D. B. Roberts, ed. (Oxford: IRL

Wiley, L. M., Kidder, G. M., and Watson, A. J. (1998). Cell polarity and development of the first epithelium. Bioessays 12, 67-73. Wodarz, A., Grawe. F., and Knust, E. (1993). CRUMBS is involved in the control of apical protein targeting during Drosophila epithelial development. Mech. Dev. 44, 175-187.