Apical Spectrin Is Essential for Epithelial Morphogenesis but Not Apicobasal Polarity in Drosophila

The karst stocks mwh ve kst¹ e/TM6B, mwh ve kst² e/TM6B, mwh kst^14.1 red e/TM6B, mwh Df(3L)1226 red e/TM6B, and kst⁰¹³¹⁸/TM6B have been previously described (Thomas et al. 1998). The recombinant stocks mwh kst^1-1151 FRT80B/TM6B and mwh kst^2-2119 FRT80B red/TM6B were derived from the above stocks by recombination with the rucuca chromosome (Lindsley and Zimm 1992) to remove ancillary lethals and to w; FRT80B (Xu and Rubin 1993) for other purposes. The slbo¹/CyO stock was a gift from Dr. D. Montell (Johns Hopkins University) and was introduced into the karst mutant background by standard genetic methods.

To detect β_H, we used serum #243 (Thomas and Kiehart 1994) at 1/200. To detect α-spectrin, we used ascites fluid #N3 at a dilution of 1/5,000 (Byers et al. 1987). To detect β-spectrin, we used rabbit serum #89 (Byers et al. 1989) at a dilution of 1/200. To detect DE-cadherin, we used the monoclonal antibody #DCAD2 (Uemura et al. 1996) at a dilution of 1/20. To detect myosin IB, we used affinity-purified rabbit antibodies (Morgan et al. 1995) at a dilution of 1/200. To detect Notch, we used the monoclonal antibody 9C6 (Wharton et al. 1985) at 1/200. To detect cytoplasmic myosin II, we used the rabbit serum #656 (Kiehart and Feghali 1986) at 1/200. To detect Fasciclin III, we used the monoclonal antibody 7G10 (Patel et al. 1987; obtained from the Developmental Studied Hybridoma Center) at 1/10.

FITC, Cy3 or Cy5 conjugated, and affinity-purified secondary antibodies used were all made in goat and were obtained from Jackson ImmunoResearch Laboratories, Inc. These antibodies were rehydrated according to the manufacturer's instructions and used at dilutions of 1/100. Alexa 488 conjugated secondary antibody was obtained from Molecular Probes and used at a dilution of 1/200.

2–4-d old females fed with yeast paste at 25°C were dissected and their ovaries were teased apart into individual ovarioles in PBS (130 mM NaCl, 7 mM Na₂HPO₄, 3 mM NaH₂PO₄, pH 7) using a tungsten needle. The tissue was fixed in buffer B (16.7 mM KH₂PO₄/K₂HPO₄, pH 6.8, 75 mM KCl, 25 mM NaCl, 3.3 mM MgCl₂, 5% paraformaldehyde; Freeman et al. 1986) for 15–20 min at room temperature. After a 30-min wash in PBT (10 mM Na₂HPO₄/NaH₂PO₄, pH7.4, 175 mM NaCl, 0.1% Triton X-100), samples were blocked in PBT-NGS (PBT, 5% normal goat serum). The samples were then incubated in PBT-NGS containing the primary antibody for 4–5 h at room temperature or overnight at 4°C. After one wash in PBT and one wash in PBT-NGS for 30 min each, the samples were incubated in PBT-NGS with the appropriate secondary antibody for 2–3 h at room temperature or overnight at 4°C. After two washes in PBS, the stained ovaries were equilibrated and mounted in mounting medium (100 mM Tris-Cl, pH 8.5, 80% glycerol, 2% n-propyl gallate). To costain for F-actin, samples were further incubated for 30 min at room temperature in 165 nM FITC- or TRITC-phalloidin (Molecular Probes) in the first wash after secondary antibody incubation. Nuclei were visualized by staining with 5 μg/ml propidium iodide in PBT for 20 min at room temperature (de Cuevas et al. 1996).

Ovaries were dissected in PBS and fixed in 1% glutaraldehyde in PBS for 20 min at room temperature. After three to four washes in PBS, the tissue was incubated in prewarmed reaction solution (10 mM sodium phosphate, pH 7.2, 150 mM NaCl, 1 mM MgCl₂, 3 mM K₃Fe(CN)₆, 3 mM K₄Fe(CN)₆, 0.1% X-Gal) for at least 30 min at 37°C. The samples were extensively washed in PBS and mounted in mounting medium.

Imaging of immunofluorescently stained ovaries was done using an MRC1024 confocal microscope (Bio-Rad Laboratories). Imaging of ovaries stained for β-galactosidase activity was done on a BX50 microscope (Olympus Corp.) equipped with a Dage/MTI CCD72T camera and DSP2000 signal processor, and were imported directly into a Power Macintosh 8100/80AV (Apple Computer) using a DT2255 frame grabber (Data Translations) controlled by the public domain program NIH Image (v1.61 available on the internet at http://rsb.info.nih.gov/nih-image). Images were contrast stretched as appropriate using Adobe Photoshop v4.0 (Adobe Systems, Inc.) and the figures assembled and annotated in Adobe Illustrator v6.0.

Images of sagittal optical sections of 81 wild-type and 318 karst mutant egg chambers at stage 9 or 10A (a comprehensive combination of kst¹, kst², kst^14.1, and kst⁰¹³¹⁸ alleles over one another and over Df(3L)1226), that had been stained for α-spectrin, DE-cadherin, actin, or Notch, were acquired. The distances described in text and Fig. 2 were then measured and analyzed using Excel 98 (Microsoft Corp.) and/or Deltagraph (DeltaPoint, Inc.).

To measure the follicle cell apical surfaces, we stained wild-type and kst⁰¹³¹⁸ egg chambers for DE-cadherin to outline the apical contours. To ensure that only cells being viewed en face were measured, the following procedure was used. The z-axis motor on the confocal was stepped gradually out from the sagittal plane of the oocyte towards the apical domain of the upper follicle cell monolayer until the oocyte had just disappeared. Measurement of the apical surface areas was done using NIH Image software. Specifically, we used the Threshold option to isolate the cell outlines, after which the Measure option was used to quantify the number of pixels per apical surface per cell. Any breaks in the DE-cadherin staining, particularly in karst mutant egg chambers, were manually closed with straight, one pixel-wide lines before area measurement. The position of the center of each cell measured relative to the nurse cell/oocyte boundary was also recorded for each cell.

Oogenesis in flies takes place in ovaries formed of 12–16 ovarioles, each of which consists of an anterior structure called the germarium and several egg chambers sequentially ordered with regard to their developmental stage (for review, see Spradling 1993; see Fig. 1 A for example). The germarium is comprised of three zones (Fig. 1 B). In zone 1, two germline stem cells divide asymmetrically to give rise to a cystoblast and a stem cell. The cystoblast then divides synchronously four times to produce 16 cell cysts interconnected by ring canals as a result of incomplete cytokinetic events. In zone 2, 16 cell cysts become surrounded by a pool of follicle cells produced by asymmetric division of two to three somatic stem cells. By this stage, 1 of the 16 germ cells has adopted an oocyte fate and becomes located at the posterior of the cyst. The remaining 15 nurse cells undergo polytenization. In zone 3, fully formed stage 1 egg chambers begin to emerge from the germarium, bounded by a well polarized follicular epithelium (Fig. 1 A). Egg chambers will continue to grow and increase in size up to stage 9, while the follicle cell monolayer accommodates this growth by a series of cell divisions.

At the onset of stage 9, after all divisions have ceased, the majority of the follicle cell monolayer undergoes a concerted migration towards the posterior of the egg chamber onto the oocyte membrane. Those follicle cells that are left behind become squamous and continue to reside on the nurse cells. During this morphogenetic event, the migratory follicle cells undergo a change in cell shape from cuboidal to columnar, a process that is instrumental in accommodating them on the growing oocyte membrane. Concomitantly, a group of 6–10 anterior follicle cells called border cells round up (Fig. 1 A, * in stage 9) and plunge in between the nurse cells to reach the anterior of the oocyte (Fig. 1 A; * in stage 10). After the follicle cell monolayer and the border cells have reached the anterior of the oocyte, a specialized subset of follicle cells called centripetal cells migrate along the nurse cell–oocyte interface such that the follicle cells now completely surround the oocyte. Communication between nurse cells and the oocyte is maintained via ring canals that remain open until the nurse cells dump their cytoplasmic contents into the oocyte at stage 11.

Although two partial descriptions of the distribution of β_H during oogenesis have been published by other groups (de Cuevas et al. 1996; Lee et al. 1997), no complete description exists. We therefore began by fully characterizing the distribution of β_H in developing wild-type ovarioles (Fig. 1). β_H expression in the germline is low to undetectable in the most anterior region of the germarium, and first appears in zone 2 (Fig. 1 B). In zone 3, fully formed stage 1 egg chambers emerge, β_H is found uniformly along the nurse cell and oocyte membranes (Fig. 1 B). Later, in mid-oogenesis, β_H is slightly enriched on the outer edge of the ring canals (Fig. 1 F).

In the soma, β_H is strongly expressed at the very anterior tip of the germarium in the terminal filament and the cap cells that contact the germline stem cells (Fig. 1A and Fig. B). In close proximity to the 16 cell cysts in region 2, we detect high levels of β_H expression in the vicinity of the somatic stem cells and in their progeny, the follicle cells (Fig. 1A and Fig. B). As individual cysts become enveloped in follicle cells, β_H is slightly enriched in the cells that move in to segregate adjacent cysts (Fig. 1 B). β_H continues to be strongly expressed here as these become the stalk cells that separate successive egg chambers along the ovariole (Fig. 1 A, arrow). β_H is apically polarized in the follicle cells (Fig. 1A and Fig. B). β_H is downregulated in the migrating border cells at stage 9 (Fig. 1 A, *), but is again expressed on the apical surface of these cells when they begin to secrete the micropyle after stage 10 (Fig. 1C and Fig. E, arrowheads). At stage 10, β_H is part of a prominent terminal web-like structure at the apical ends of the follicle cells that are secreting egg components (Fig. 1 G). This structure appears to be anchored in the ZA by fine fibers of staining around its edge. β_H is also expressed on the apical surfaces of the follicle cells secreting the dorsal appendages (Fig. 1 D, arrow) and the chorion (data not shown). β_H colocalizes with α-spectrin in the germline and soma at all these locations (Fig. 1); however, the latter is more widespread presumably through its association with the conventional β isoform.

In mid–stage 9 egg chambers, the follicle cell monolayer and the border cells migrate in a concerted fashion (Montell et al. 1992; see also Fig. 1). However, in 73% of karst egg chambers the border cells migrate ahead of the follicle cell monolayer (Fig. 2 A). The degree to which the border cells migrate ahead of the follicle cells during stage 9 is quite variable in keeping with the variable expressivity of the karst mutation (Thomas et al. 1998). We occasionally find egg chambers where the border cells are retarded relative to the follicle cells; however, this extreme situation probably arises from a combination of weak expression of the follicle cell phenotype combined with strong expression of the border cell phenotype (see Discussion).

The lack of coordination in cell migration makes it difficult to assess the progression of each chamber through stage 9. We therefore resorted to a morphometric approach. The four distances indicated in Fig. 2 B were measured in sagittal optical sections of 81 wild-type and 318 karst mutant egg chambers. Since the nurse cells do not grow or shrink at this time, all these measurements were normalized against the anterior-posterior length of this cell cluster.

Pairwise comparisons of the parameters (FC/NC), (BC/NC), and [oocyte/(oocyte+NC)] (Fig. 2, C–E) reveals the following defects in karst mutant egg chamber morphogenesis. (a) Most karst border cell clusters are migrating ahead of the follicle cell monolayer during stage 9 (Fig. 2 C). This could arise either due to faster border cell migration or slower follicle cell migration. (b) Most karst mutant oocytes occupy a larger portion of the egg chamber than in the wild-type during follicle cell migration, while the oocyte exhibits no significant overgrowth at the completion of migration (Fig. 2 D). This suggests that karst follicle cells are delayed in their migration relative to growth of the oocyte, or may respond more slowly to oocyte growth. (c) Similarly, most karst mutant oocytes occupy a larger portion of the egg chamber than in wild-type during border cell migration (Fig. 2 E). This effect is not as strong as for the follicle cells, consistent with the observation that the border cells generally migrate ahead of the follicle cells in karst mutant egg chambers, but it does suggest that there is a slight delay in border cell migration. The most parsimonious model accounting for these data suggests that the karst mutation causes a significant disruption of follicle migration onto the oocyte membrane and a slight delay in border cell delamination or migration through the nurse cells.

Consistent with the hypothesis that the primary morphogenetic defect lies in follicle cell migration, some follicle cells in karst mutant egg chambers often remain in contact with the nurse cell membranes at stage 10A. These follicle cells still attempt to make the appropriate adhesive contacts with the oocyte membrane (Goode et al. 1996), pulling the oocyte membrane towards them and grossly distorting the nurse cells/oocyte interface (Fig. 3 A). In most cases, the subsequent inward migration of the follicle cell layer at stage 10B proceeds along the nurse cell/oocyte interface in a relatively normal fashion. However, in rare, extreme cases, the centripetally migrating cells penetrate between nearby nurse cell membranes and cause one or more nurse cells to become included within the egg along with the oocyte (Fig. 3 B).

The failure of karst follicle cells to complete their migration onto the oocyte by the onset of stage 10B implies that the total apical surface area of the epithelium is greater than that of the oocyte membrane. Moreover, karst mutant follicle cells often appear to have a more cuboidal shape than in the wild-type (see Fig. 5G and Fig. H). Since there is no over-proliferation in the mutant monolayer (data not shown), this cannot arise due to an increase in cell number. However, an inability of karst follicle cells to properly change their cell shape or constrain their apical surface area at the appropriate size would explain this observation. We therefore compared the apical surface area of wild-type and mutant follicle cells during monolayer migration. This analysis reveals a sharp decrease in the apical surface area of the wild-type follicle cells as they approach and migrate onto the oocyte (Fig. 4 A). In contrast, the majority of karst mutant follicle cells fail to apically constrict (Fig. 4 A). The mean apical surface area of the mutant follicle cells is almost twice that of the wild-type (Fig. 4 B). Moreover, comparison of the apical surface areas of mutant follicle cells in chambers during migration with those where migration has been completed reveals a slight increase (Table; P < 0.001). This suggests that, in addition to the constriction defect, the monolayer cannot withstand the forces exerted by the growing oocyte.

Examination of the follicle cell apices stained for DE-cadherin also reveals conspicuous disruptions in the staining pattern of DE-cadherin in karst mutant egg chambers (Fig. 5, A–F). In the mildest cases, this staining is missing at three- or four-cell vertices, but we also see large breaks in the normally continuous belt of staining in more extreme cases. These observations are consistent with the hypothesis that the absence of β_H weakens the ZA, and that it breaks up as the apices attempt to constrict or accommodate the growth of the oocyte. However, the apicolateral polarization of the ZA is largely unaffected (Fig. 5G and Fig. H).

The border cell cluster delaminates from the follicle cell epithelium and migrates between the nurse cells to the anterior of the oocyte during stage 9. In ∼10% of karst mutant egg chambers, we observe migratory cells that are well separated from, or trail behind, the main border cell cluster (Fig. 6). The trailing cells upregulate α-spectrin (Fig. 6A and Fig. B) and DE-cadherin (data not shown) in a manner that resembles wild-type clusters, suggesting that they are true border cells. To confirm that these cells have a border cell fate, we looked for the expression of the border cell marker slow border cells (slbo) by introducing the enhancer trapped LacZ gene associated with the slbo¹ P-element allele (Montell et al. 1992) into the karst mutant background and staining for β-galactosidase activity. All of the migratory cells in slbo¹/+; kst egg chambers express β-galactosidase, although the intensity of staining exhibited by the separated cells is often lower than the main border cell cluster (Fig. 6, D–F). None of the trailing cells express Fasciclin III, indicating that they do not contain polar cells (data not shown; Patel et al. 1987), and thus do not represent second, independently organized, migratory clusters. Furthermore, the total number of β-galactosidase–positive cells ranges from 7 to 11, very close to the normal range in border cell number in wild-type chambers (6–10; Spradling 1993). Together, these data suggest that all of the migratory cells are bona fide border cells and that the normal number of border cells is specified in karst mutants, but that the cluster is unable to remain together as a unit.

β_H is no longer detectable at the apical domain of the follicle cells in any allelic combination of karst alleles that we have examined (data not shown). The localization of β_H to the apical domain has been previously shown to be dependent on α-spectrin (Lee et al. 1997). To see if α-spectrin is dependent on β_H for its localization to the apical domain, and to confirm that no apical spectrin function remains in karst mutants, we examined the distribution of α-spectrin in karst follicle cells. While the lateral α-spectrin distribution is unaffected by this mutation, apical α-spectrin is no longer detectable by immunofluorescence (Fig. 7C and Fig. E). This indicates that the stable recruitment of α-spectrin to the apical domain is dependent on β_H, and that there is thus a mutual interdependence between α-spectrin and β_H. This further suggests that αβ_H-spectrin is recruited to the apical domain as a heterodimer or tetramer, or that following separate recruitment only the dimers or tetramers remain stably associated with the apical domain.

Spectrin associated with E-cadherin has been implicated in the apicobasal cell polarization pathway (Yeaman et al. 1999), and fly α-spectrin mutations cause a breakdown in monolayer polarity including the loss of apical β_H (Lee et al. 1997). However, karst mutants form a follicle epithelium that appears to have a well polarized morphology. To further verify this observation, we examined the distribution of the apical markers Notch, myosin II, and unconventional myosin IB (data not shown; Edwards and Kiehart 1996; Morgan et al. 1995; Xu et al. 1992), as well as the apical concentration of actin in the follicle cell brush border. These and similar experiments in mutant eye and wing imaginal discs (data not shown) indicate that the distribution of several apical markers is not conspicuously affected by the lack of β_H.

Given the significant similarity between β_H and β-spectrin (Thomas et al. 1997), it is also possible that polarity is retained in karst mutant epithelia because of functional redundancy between these two proteins. Moreover, the karst mutant phenotype exhibits variable expressivity despite genetic evidence that we have null alleles (Thomas et al. 1998), and this might also result from functional redundancy. Such redundancy would predict that conventional β-spectrin would be recruited to the apical domain in the absence of β_H. However, this is not the case (Fig. 7B, Fig. D, and Fig. F), confirming that β_H is unnecessary for apicobasal polarity and indicating that we must look elsewhere for the source of variability in the karst phenotype.

β_Heavy-spectrin is a member of the spectrin family of proteins that have been implicated in cadherin-mediated cell polarization (Yeaman et al. 1999), and is associated with the ZA (Thomas et al. 1998; Thomas and Williams 1999). In this paper, we have characterized the effects of loss of function karst mutations (which eliminate β_H) on fly oogenesis. These mutations cause defects in the migration of the follicle cell monolayer and in border cell delamination, but do not compromise the apicobasal polarity of its constituent cells. The defect in monolayer migration is characterized by a failure of the follicle cells to constrict their apical surfaces and by visible breaks in the ZA.

We present here the first complete description of the distribution of β_H during Drosophila oogenesis (Fig. 1). β_H is strongly expressed in the terminal filaments and cap cells that sit adjacent to the germline stem cells at the anterior end of the germarium. In the germline, it is first seen on the membrane in region 2, in the 8–16-cell cysts. β_H is not obviously polarized at any of these locations, and remains uniformly distributed in the nurse cells and oocyte. In the soma, β_H is prominently expressed in the vicinity of the somatic stem cells and their derivative, the follicle cell epithelium, where it is restricted to the apical domain. These results corroborate the previous partial descriptions of the distribution of β_H during oogenesis (de Cuevas et al. 1996; Lee et al. 1997). We have also found that β_H is prominently expressed in the stalk cells, at a number of sites of high secretory activity late in oogenesis, and is downregulated in the border cells during their migration.

The downregulation of β_H in the anterior region of the germarium would suggest that β_H has no role in germ cell division and/or oocyte specification within the germline. However, β_H is strongly expressed in the terminal filament and cap cells that sit adjacent to the germline stem cells and early cystoblasts, the cellular activities of which are believed to regulate germline stem cell division and polarization (reviewed in Spradling et al. 1997). β_H could thus interact with one or more components in the terminal filament and cap cells to generate signals that affect polarity or proliferation in the germline.

β_H is part of a prominent terminal web subtending the follicle cell brush border that forms as they migrate onto the oocyte at stage 9 and begin to secrete yolk protein. β_H is also expressed in the border cells once these cells begin to secrete the micropyle and in the follicle cells that are secreting chorion to form the dorsal appendages. The prominence of β_H in all these locations of high secretory activity suggests that there may be a role for the apical membrane skeleton in the targeting or delivery of secretory vesicles even though it is not required for overall apicobasal polarity (see below). A similar role has been proposed for the gut-specific terminal web β-spectrin, TW260, in the chicken (Hirokawa et al. 1983).

In karst mutant ovaries, the primary defect in follicle cell morphogenesis is a failure of this monolayer to complete migration onto the oocyte membrane by the end of stage 9 (Fig. 2 and Fig. 3). Specifically, these cells fail to undergo the normal apical contraction associated with the development of a more columnar shape upon contacting the oocyte membrane (Fig. 4). Furthermore, staining for the ZA marker DE-cadherin revealed conspicuous breaks in the normally continuous belt of staining for this protein around the follicle cell apices (Fig. 5).

Apical contraction is a well established process for the generation of form in an epithelium; however, the exact mechanism by which cells achieve this phenomenon is still unclear. β_H is associated with two actin-rich structures at the apical pole, the ZA and the terminal web subtending the microvillar brush border. F-actin at the ZA lies in a circumferential band of microfilaments of mixed orientation that can be induced to undergo a purse string–like contraction in vitro that is mediated by myosin II (Burgess 1982; Hirokawa et al. 1983; Keller et al. 1985). β_H may be necessary for contraction of this bundle. Spectrin cross-linking of F-actin can stimulate myosin ATPase (Coleman and Mooseker 1985) and is necessary for cortical contraction of sea urchin eggs (Walker et al. 1994). The distribution of myosin II was not conspicuously affected in karst mutant follicle monolayers (data not shown); however, β_H may be necessary for the correct organization of the contractile actin bundle through its ability to cross-link F-actin, or for the attachment of the bundle to the membrane or the ZA.

Alternatively, β_H may play a structural role in the follicle cell terminal web (Mahowald 1972; Morgan et al. 1995). This is a region of dense actin cross-linking between bundles of microfilaments that support the overlying microvilli and is integrated into the circumferential F-actin bundle and the ZA at its margins. This structure also contains myosin II (Herman and Pollard 1981; Hirokawa et al. 1982), but is not contractile. However, a decrease in apical surface area presumably requires remodeling of the terminal web to produce a corresponding decrease in the size of this network or an increase in microvillar density. Neither a terminal web nor β_H is required to produce microvilli (Fath and Burgess 1995; Thomas et al. 1998). However, β_H in the terminal web might help stabilize the contractile process. The apical domain of all of the follicle cells must apically constrict, presumably generating significant tension in the adhesive network that binds the epithelium together. In this particular case, the terminal web may be an essential structural component of a supracellular actin network. The lack of cross-linking by β_H in the terminal web might thus result in excessive strain on the ZA and its ultimate breakage.

The fact that follicle cell migration does initiate and proceed to some extent in karst mutants indicates that not all motile forces are eliminated by this mutation. This may simply mean that not all intercellular tension generated by apical constriction is eliminated by the karst mutation. However, another possibility is that apical constriction is not the only force-generating mechanism responsible for monolayer migration. Indeed, the observation that hypomorphic mutations in the regulatory light chain of cytoplasmic myosin II do not prevent the migration of the follicle cell monolayer (Edwards and Kiehart 1996) does indicate this. At least two other forces should be considered. First, the oocyte is growing through the uptake of hemolymph and yolk during stage 9, and this inflation may contribute to migration and to tension at the ZA. Second, the follicle cells make specific adhesive contact with the oocyte membrane (Goode et al. 1996) and this might produce localized tension as each follicle cell first contacts, and then moves onto the oocyte.

The failure of karst follicle cells to complete their migration onto the oocyte by the onset of stage 10B leads to aberrant centripetal cell migration in a small number of egg chambers. In such cases, the inwardly migrating cells find their way in between nurse cell membranes enclosing one of the latter along with the oocyte. Given the high frequency of defective migration, it is perhaps surprising that this latter defect is relatively rare and that oogenesis can often proceed to completion in the absence of β_H. We suspect that the ability of the centripetally migrating cells to seek out the nurse cell/oocyte interface by specific adhesion to the oocyte membrane (Goode et al. 1996) results in a substantial compensation for the migration defect that is generated by the karst mutation. Thus, the inclusion of a nurse cell with the oocyte after centripetal cell migration is infrequent and a relatively normal egg results. Moreover, after migration is completed, the oocyte continues to grow as dumping occurs and the egg matures, during which time the follicle cell apices must again grow and presumably any slack in the karst monolayer is taken up. This would explain why karst mutant females are not completely sterile and do lay some fertilized eggs.

The border cells delaminate from the follicular epithelium at the onset of stage 9 and migrate between the nurse cells to the anterior of the oocyte (Spradling 1993). As part of this developmental program, we have shown that they downregulate β_H. While there is no detectable β_H at any of the border cell–border cell interfaces, it is difficult to say with certainty that there is no β_H in a normal cluster. This is because these cells do not fully depolarize (Niewiadomska et al. 1999), and the residual apical surface is closely juxtaposed to the surrounding nurse cell membranes that contain abundant β_H protein. Migration normally occurs as a tight cluster of 6–10 cells; however, in karst mutant egg chambers, one or more border cells are separated from the main cluster (Fig. 6).

The presence of β_H at two or possibly three locations during normal border cell delamination and migration suggests a number of possible roles for this protein in border cell morphogenesis. Its presence at the nurse cell membranes on which the border cells migrate could contribute to the rigidity of this substratum or to inter-nurse cell adhesion. In this context, we note the similarity between the patchy distribution of β_H (Fig. 1 A, arrowhead) along the border cell migration route and that of DE-cadherin, a molecule required for migration (Oda et al. 1997; Niewiadomska et al. 1999). The loss of either of these functions might impact border cell migration. β_H might also be present on the outermost apical surface of the migrating border cell cluster. Here it could contribute to the adhesion between the border cells and the nurse cells. Loss of this function might cause the border cell cluster to have difficulty attaching to the nurse cells during delamination and the subsequent migration. Finally, β_H is present in the nonmigratory follicle cells that surround the delaminating cluster, where it might play a role in border cell delamination. This latter hypothesis is the most likely explanation for the border cell phenotype. If the absence of β_H either on the apical border cell surface or on the nurse cell surface was responsible for the breakup of the cluster, we would expect to see either a more dramatic or a preferential breakup of clusters late in migration. However, there is no indication that the number of separated cells increases with the distance migrated and we see many examples of early break up.

The normal boundary between the border cells that will delaminate and the surrounding follicle cells is marked by a dramatic decrease in the level of β_H in the border cells. In addition, β_H is also prominent at the follicle cell/border cell membrane interface during delamination. This boundary must have at least two properties. It must serve to allow the border cells to detach and it must allow the follicle cell monolayer to reseal the gap left by the departing cluster. We hypothesize that the presence of β_H in the surrounding follicle cells is part of a differential adhesion system that causes the surrounding, nonmigratory follicle cells to seek out one another to reseal the gap and in doing so to sacrifice contact with the border cells. Elimination of β_H in the nonmigrating cells would affect the precise physical boundary between groups of cells with different adhesive properties preventing this rearrangement of cell contacts, and thus proper detachment of the border cells. The precise role of β_H in generating this boundary remains open. β_H is localized in part at the ZA and its presence or absence could be responsible for modulating DE-cadherin–based adhesion. Such differential adhesion is clearly part of the mechanism by which the oocyte positions itself relative to the overlying follicle cells (Godt and Tepass 1998; Gonzalez-Reyes and St. Johnston 1998). However, such sharp transitions in cell fate can be generated by lateral inhibition mechanisms (Artavanis-Tsakonas et al. 1999), and it is also possible that β_H is responsible for stabilizing some of the signaling molecules involved in such processes during border cell specification.

The initial development of the follicular epithelium is essentially unaffected by the loss of either α-spectrin (Deng et al. 1995; Lee et al. 1997) or β_H (this paper), suggesting that the spectrin membrane skeleton is not essential for the establishment of polarity in this particular case. We have also demonstrated that there is no conspicuous breakdown of apicobasal polarity in the absence of the apical SBMS later in oogenesis (Fig. 7). This contrasts with the phenotype of α-spectrin mutants in Drosophila (Deng et al. 1995; Lee et al. 1997), in which a loss of cell polarity and breakdown of the follicular epithelium is seen. Together, these two results indicate that the loss of apicobasal polarity in the α-spectrin mutants reflects a requirement for the basolateral SBMS specifically, or is a synergistic consequence of losing both the apical and basolateral membrane skeletons. Resolution of this ambiguity awaits further characterization of β-spectrin mutants. Although overall apicobasal polarity remains intact in karst mutants, it remains possible that specific proteins that are dependent on binding to the apical SBMS for delivery and/or retention in the apical membrane are depolarized by the absence of (αβ_Η)₂.

In keeping with the observation that different spectrin isoforms do not generally colocalize (Glenney and Glenney 1983; Lazarides et al. 1984), β_H and β-spectrin are found in mutually exclusive domains in epithelial tissues (Dubreuil et al. 1997, Dubreuil et al. 1998; Lee et al. 1997; Thomas et al. 1998). β-spectrin is recruited to the membrane by proteins that bind to the adapter protein ankyrin (Dubreuil et al. 1997). However, β_H lacks an ankyrin binding site and is not recruited in this manner (Dubreuil et al. 1997; Lee et al. 1997; Thomas et al. 1997). Integral membrane proteins that bind to ankyrin thus provide a polarizing influence that can specifically establish a basolateral membrane skeleton. In this paper, we have shown that in the absence of β_H, β-spectrin does not become recruited to the apical surface. This indicates that β_H is not excluding β-spectrin from potential binding sites in this domain and that β_H must therefore be specifically recruited to this domain. This result is also significant because it implies that the variable expressivity associated with the karst phenotype (Thomas et al. 1998) does not arise through redundancy of function between β_H and β-spectrin.

Current models for the origins of epithelial polarity suggest that spectrin plays a key role in establishing and/or maintaining the apicobasal axis (Yeaman et al. 1999). This model is largely based on the behavior of molecules involved in establishing the basolateral domain. The combined observations on the α-spectrin (Deng et al. 1995; Lee et al. 1997) and karst mutant phenotypes suggests that the basolateral membrane skeleton may be playing such a role; however, our results indicate that apical spectrin [i.e., (αβ_Η)₂] does not. It remains unclear what the precise mechanism is by which apical spectrin acts during morphogenetic events. β_H exhibits a close colocalization with the ZA, and its levels at this location are regulated in concert with DE-cadherin (Thomas et al. 1998; Thomas and Williams 1999). Furthermore, we have observed mild disruptions of the ZA in eye/antennal imaginal discs (Zarnescu, D.C., and G.H. Thomas, unpublished observations) that are similar to the effects reported in this paper. None of our results to date reveal how closely (αβ_Η)₂ is associated with the ZA; however, the contrasting behavior of the apical and basolateral SBMS during the emergence of apicobasal polarity (Thomas and Williams 1999) combined with the phenotypic data presented in this paper strongly suggest that these two cytoskeletal structures have somewhat distinct rather than identical roles in their respective domains, at least when it comes to the generation and/or maintenance of apicobasal polarity.

The results presented in this paper add to a growing body of evidence that apical spectrin is essential for epithelial morphogenesis. Moreover, we show that an apical SBMS is not required for establishing or maintaining apicobasal polarity, as seems to be the case for the basolateral SBMS. It is unknown at present whether or not β_H or any other β-spectrin plays a similar role in morphogenesis in vertebrates. The observations that β_H is evolutionarily old (Thomas et al. 1997), that there is a homologue in C. elegans encoded by the sma1 gene (McKeown et al. 1998), and that the chicken brush border protein TW260 has a similar size and contour length to β_H (350 kD; Dubreuil et al. 1990; Glenney et al. 1982), all strongly suggest that there is a vertebrate homologue of β_H that has yet to be cloned. The karst phenotype bears close resemblance to the phenotype exhibited by C. elegans embryos mutant at the sma1 locus that encodes the worm homologue of β_H (McKeown et al. 1998). In sma1 mutants, the worms fail to elongate at the wild-type rate during embryogenesis, resulting in a smaller embryo. Moreover, this phenotype seems to arise from a defective contraction of the embryonic epithelium (McKeown et al. 1998). Thus, while the geometry of this developmental event is quite different from follicle cell migration in the fly, β_H appears to be playing a similar role in the two organisms. In vertebrates, the apical accumulation of spectrin has been correlated with the initiation of neurulation in mice (Sadler et al. 1986), and, in the sea urchin embryo, an apical accumulation of α-spectrin is associated with the involution of tissues during embryogenesis (Wessel and Chen 1993). This suggests that apical spectrin has a general role in the morphogenesis of epithelia mediated by apical contraction.

The authors thank Spyros Artavanis-Tsakonas, Dan Branton, Dan Kiehart, John Lee, Mark Mooseker, and Tadashi Uemura for gifts of antibodies, Denise Montell for supplying the slbo¹ enhancer trap line, and Carol Gay for the use of her confocal microscope. We also thank Maggie de Cuevas and Bryce MacIver for useful discussions, as well as Susan Abmayr, Esther Siegfried, and the members of the Thomas lab for critically reading this manuscript.

This work was funded by National Institutes of Health grant GM52506 to C.M. Thomas.