α-Spectrin and integrins act together to regulate actomyosin and columnarization, and to maintain a monolayered follicular epithelium

Monolayered epithelia are sheets of adherent, polarized cells that act as physical barriers and constitute structural components of organs and tissues. The formation and maintenance of the monolayered structure are crucial for both proper function of the epithelia and whole-body homeostasis. During carcinogenesis, loss of epithelial architecture leads to the formation of multilayered epithelia, disorganized cell masses and increased tumorigenic potential.

The Drosophila melanogaster ovary constitutes an excellent model system in which to study the molecular and cellular basis of epithelial morphogenesis. The adult ovary is composed of various ovarioles that contain a line of egg chambers at different developmental stages [stage1-14 (S1-14)]. Each egg chamber is composed of 16 germline cells (including the oocyte), and a layer of somatic cells (the follicle cells, FCs) forming a monolayered epithelium termed the follicular epithelium (FE) (Fig. 1A). FCs are derived from stem cells that are located in the germarium. Up to S6 of oogenesis, FCs undergo several rounds of mitotic cycles to form the FE, then exit mitosis and enter an endocycle. From S7, most FCs change their shape from cuboidal to columnar, and migrate towards the posterior (Fig. S1). The factors important for formation of a monolayered FE are not yet fully understood, but mutations in genes controlling polarity and mitosis lead to FE multilayering, such as aPKC (Abdelilah-Seyfried et al., 2003), Notch and the Hippo pathway (Meignin et al., 2007; Polesello and Tapon, 2007; Yu et al., 2008). In addition, integrins and spectrins (Spec) are also important for maintaining a monolayer.

The spectrin-based membrane skeleton (SBMS) is a scaffold made from building blocks of tetramers of two α and two β Spec subunits that line the cell membrane. The function of the SBMS has been greatly studied in erythroid cells, and a variety of erythrocyte disorders are associated with mutations in Spec genes. Members of the Spec family are conserved in all eukaryotes, with a greater conservation between Drosophila and mammalian non-erythroid Specs than between erythroid and non-erythroid mammalian forms (Baines, 2003, 2009; Salomao et al., 2006). However, in contrast to mammals, the Drosophila genome features a single form of α-Spec (human αII-like), a conventional β subunit (human βII-like) and a heavy β subunit (β_H), making it easier to characterize their function in non-erythroid cells (Byers et al., 1989; Dubreuil et al., 1990; Lee et al., 1997). The (αβ)₂ and (αβ_H)₂ tetramers are distinctively localized in the basolateral and apical domains, respectively (Dubreuil et al., 1998; Lee et al., 1997; Thomas et al., 1998; Zarnescu and Thomas, 1999).

A diversity of functions has been attributed to spectrins based on studies in both cell culture and model organisms. In invertebrates, spectrins are essential for morphogenesis and animal growth. Drosophila spectrins have been recently identified as modulators of the cell growth Hippo pathway in various tissues (Deng et al., 2015; Fletcher et al., 2015; Wong et al., 2015). In ovaries, a mutant β-Spec allele with a premature stop codon at amino-acid 1046 shows defects in FE integrity, actin organization and oocyte polarity, partially phenocopying hippo mutants (Wong et al., 2015). α-Spec mutant FCs also form a stratified epithelium, with polarity defects (Lee et al., 1997), and FCs expressing an α-Spec RNAi show hippo-like differentiation defects (Fletcher et al., 2015). By contrast, β_H-Spec does not regulate Hippo or actin in ovaries (Fletcher et al., 2015; Thomas et al., 1998; Zarnescu and Thomas, 1999). These findings confirm the idea that different tissues exhibit different dependence on the apical versus lateral spectrin cytoskeleton (Dubreuil et al., 2000; Hulsmeier et al., 2007; Thomas and Williams, 1999).

These previous studies concentrated on spectrins as Hippo modulators. We decided to broaden the analysis of SBMS function in a monolayered epithelium by studying the consequences of eliminating SBMS on the proliferation, polarization and differentiation of FCs, as well as on the architecture of the FE. As α-Spec is the major component of both the apical and lateral spectrin cytoskeletons, we decided to concentrate on α-Spec. Eliminating α-Spec in the FE results in stratification of the terminal regions, especially the posterior, supporting previous findings (Fletcher et al., 2015; Lee et al., 1997). However, in contrast to Hippo, we find that the function of the SBMS in the monolayered FE is not to control mitosis, differentiation or polarity, but to regulate the actomyosin cytoskeleton, septate junctions (SJs) and cell shape. The α-Spec mutant phenotype is similar to that of integrin [myospheroid (mys)] mutants, and α-Spec and integrins colocalize in the lateral membrane of the FCs. We propose that a functional spectrin-integrin complex is important for regulation of the actomyosin cytoskeleton and tissue architecture.

To understand the defects of eliminating α-Spec on FE morphogenesis, we studied FCs mutant for the null allele α-Spec^rg41. As in previous reports, we found that α-Spec FCs form a multilayered epithelium. These tumor-like masses are observed only at the terminal (anterior/posterior) domains of the egg chamber, and never at the mid-body (Fig. 1). This terminal requirement for α-Spec in epithelial architecture is identical to the terminal requirement for Hippo and integrins (Fernández-Miñán et al., 2007; Meignin et al., 2007; Polesello and Tapon, 2007).

Mutations that result in a multilayered FE often prevent the FCs from differentiating. FCs differentiate from S6, as revealed by differential expression of markers, such as Fasciclin 3 (Fas3), Hindsight (Hnt; also known as Pebbled) and Eyes absent (Eya). Fas3 is expressed at high levels in immature FCs, but its expression becomes gradually restricted to the polar cells as oogenesis proceeds (Bai, 2002; Muzzopappa and Wappner, 2005) (Fig. 1Ai), whereas Hnt is upregulated from S6 (Sun and Deng, 2007). Analysis of mosaic FE containing both control and α-Spec mutant FCs (also named FCCs) revealed that some S6-8 α-Spec FCCs show high levels of Fas3 (Fig. 1Ai-Aii versus Aiii) and lack Hnt expression (Fig. 1Ci,D). However, in all cases, the defects are only observed in α-Spec cells located in the ectopic layers of the multilayered epithelium (Fig. 1A,C), and not in mutant cells that are either adjacent to the oocyte (Fig. 1Aiii, yellow arrow) or forming a monolayer (Fig. 5). This differentiation phenotype is stronger at the posterior than at the anterior pole, and weaker in older egg chambers: 100% (n=45) of the S6 α-Spec FCCs show Hnt defects, but these defects are absent at S9 (n=24) (Fig. 1C,D). Similarly, fewer α-Spec FCCs have defective Fas3 levels in older egg chambers (62.5% and 25% in early S6-8 and S9 egg chambers, respectively, n=16). It is interesting to note that even though Fas3 is properly downregulated in most α-Spec cells, the remaining Fas3 does not show the apicolateral accumulation that is observed in control posterior FCs (PFCs; Fig. 1Aii′ versus Aiii″, yellow arrows; Fig. 5; Dubreuil et al., 2001). Finally, Eya, which is downregulated in cells from S6, was also properly downregulated in α-Spec mutant monolayers, and α-Spec cells adjacent to the oocyte, but had a stronger expression in the α-Spec ectopic layers (data not shown).

Thus, α-Spec is required for the FE to maintain a monolayer, and for the FCs to mature only when part of ectopic layers. These varying defects in differentiation indicate that α-Spec plays a secondary role in this aspect of oogenesis, unlike the Hippo pathway, which fully blocks FC maturation when mutated in terminal FCs (Meignin et al., 2007; Polesello and Tapon, 2007).

Recent work has shown that apical and lateral components [such as α-Spec, aPKC and Discs large (Dlg; Dlg1 – FlyBase)] localize properly in β-Spec mutant monolayers (Wong et al., 2015). By contrast, α-Spec FCs that are part of a hyperplastic region show a loss of epithelial polarity (Lee et al., 1997). However, the α-Spec ovaries analyzed by Lee et al. were also heterozygous for discs lost (dlt), a polarity gene. The dlt heterozygosity might have increased the polarity phenotype in α-Spec mutant FCs. To characterize further the role of α-Spec in epithelial polarity, we analyzed the distribution of aPKC and Dlg in FCs that lack only α-Spec. These proteins localize correctly in α-Spec FCCs that maintain a monolayered epithelium (Fig. 2A; n=15). However, mutant cells in ectopic layers show mislocalized aPKC and Dlg in 80% and 97% of the mutant epithelia, respectively (Fig. 2B; n=35). aPKC often expands into the lateral membrane (red arrows) and by S8-9 Dlg is no longer found restricted to the lateral membrane (yellow arrows), resulting in the colocalization of these two proteins in some instances (Fig. 2Bi, white arrows). However, some degree of polarity is still preserved, as Dlg and aPKC can localize correctly in the most ectopic layer.

All these observations suggest that spectrins become relevant for epithelial polarity when FCs lose contact with the germline or the basal membrane, or that an apical and basal cue can compensate for the lack of a spectrin cytoskeleton.

The repolarization of the oocyte in mid-oogenesis is induced by a signal from the PFCs (Deng and Ruohola-Baker, 2000; Frydman and Spradling, 2001; González-Reyes et al., 1995; MacDougall et al., 2001; Roth et al., 1995), a process suggested to depend on the SBMS, as oocyte polarity is often aberrant in egg chambers with β-Spec clones (Wong et al., 2015) or α-Spec, dlt/+ cells (Lee et al., 1997). As we have observed that oocyte-adjacent α-Spec cells mature and polarize properly, we wondered whether elimination of only α-Spec affects repolarization of the oocyte. In wild-type egg chambers, oocyte nucleus migration from posterior to the dorsal-anterior corner is complete by S7, and provides a read-out of oocyte polarity. This migration is also observed in 95.5% of egg chambers with large α-Spec FCCs (Fig. S2; n=45). A more sensitive assay to detect oocyte polarity defects is Staufen localization, which by S9 forms a tight crescent at the posterior of the oocyte (Fig. S2A). Of the 95.5% mutant egg chambers that showed wild-type nucleus positioning, we found that Staufen is localized properly in all of them, whereas 13% of follicles with partial posterior clones show Staufen expressed in a crescent-shaped area that is shifted towards the control cells (Fig. S2B,D; data not shown). Thus, oocyte polarity is largely unaffected in egg chambers with α-Spec mutant PFCs.

The results described above show that α-Spec FCCs show polarity and differentiation defects only in the ectopic layers of a multilayered epithelium, and induce mild oocyte polarity defects. By contrast, PFC differentiation and oocyte polarity is aberrant in all hippo mosaic egg chambers (Meignin et al., 2007; Polesello and Tapon, 2007; Yu et al., 2008). To test whether α-Spec FCCs show any hippo-like phenotype, we analyzed whether α-Spec is required for the FCs to exit mitosis, a process that is Hippo dependent (Meignin et al., 2007), by detecting phospho-histone 3/PH3. PH3 is only detected until S6, and never later, in control cells (Fig. 3A) (Deng et al., 2001; Schaeffer et al., 2004). By contrast, hippo FCs are often positive for PH3 at S7-10B (Fig. 3B,E) (Meignin et al., 2007; Polesello and Tapon, 2007). As with wild type, α-Spec FCCs never express PH3 after S6 (Fig. 3C,D). Hence, and unlike Hippo, α-Spec is not required for FCs to exit mitosis.

Our initial observation that Fas3 accumulates apicolaterally in S9 PFCs (Fig. 1Aii′), and that such localization was lost in α-Spec cells in contact with the oocyte (Fig. 1Aiii′), suggest that α-Spec and Fas3 might be involved in processes other than FC maturation and polarity (as these processes are not aberrant in oocyte-adjacent α-Spec cells). Earlier works identified Fas3 as a component of SJs, the invertebrate counterparts of tight junctions (TJs) (Brower et al., 1980; Fehon et al., 1994; Patel et al., 1987; Woods and Bryant, 1991). SJs form basal to the zonula adherens (ZA), and in addition to blocking diffusion they also aid cell adhesion (Fehon et al., 1991; Tepass and Tanentzapf, 2001; Woods and Bryant, 1993; Zak and Shilo, 1992). The aberrant distribution of Fas3 in α-Spec FCs raises the possibility that α-Spec is required for SJs formation. Thus, we characterized the localization pattern during oogenesis of the SJ components Fas3, Dlg (Woods et al., 1996), Coracle (Cora) (Fehon et al., 1994) and Fas2 (Wei et al., 2004).

All four proteins are expressed uniformly along the lateral membrane in early FE (Fig. 4A-D). From S7, an apicolateral concentration of these proteins becomes distinct in PFCs. This apicolateral localization has been previously observed for Fas3, and our findings show Dlg, Cora and Fas2 to behave similarly. These observations are consistent with transmission electron micrographs that detected incipient SJs at S6 (Müller, 2000), and suggest that the loss of apicolateral accumulation of Fas3 in α-Spec FCs might be related to defects in SJ formation. In fact, the apicolateral localization of Fas3, Dlg, Cora and Fas2 is absent in α-Spec cells (Fig. 5), even when the mutant cells form a monolayer. It is unclear at this resolution how the SJ components localize in mutant cells; the proteins sometimes extend along the lateral membrane (white arrows) or are absent altogether (yellow arrows). However, a departure from the normal apicolateral accumulation is clear, and it suggests a function for α-Spec in SJ formation.

These localization defects are specific to SJ proteins, and not to all apical junctions, as the adherens junction markers Armadillo (Arm) and E-Cadherin (Shotgun – FlyBase) are largely unaffected in α-Spec FCs (Arm 87.5% as control, n=16; E-Cadherin 93% as control n=14; Fig. S2E-I; data not shown).

We noticed that α-Spec FCCs appear to have a reduced apical-to-basal length (Fig. 5A). Regulation of epithelial cell shape, such as changes in relative sizes of apical, basal and lateral membranes, is key to morphogenesis. FCs undergo various morphogenetic changes, including the transformation from a cuboidal to a columnar epithelium at mid-oogenesis (Dobens and Raftery, 2000). PFCs increase their height four times from S4/5 to S9, but their width remains constant (Fig. S1; Fig. S3A), supporting previous conclusions that FC columnarization does not result from a decrease in apical surface, but rather from an expansion of the lateral membrane, probably driven by cellular growth (Kolahi et al., 2009). To test whether columnarization is aberrant in α-Spec FCCs, we measured the height and width of S9 α-Spec PFCs. The mean height of control S9 FCs is 18.95 µm, whereas that of α-Spec S9 FCs is 12 µm (Fig. 6; n=40), suggesting that α-Spec mutant cells fail to become columnar. By contrast, the width of α-Spec S9 cells is similar to control cells (Fig. 6). Thus, at mid oogenesis, α-Spec cells maintain a cuboidal shape, whereas control cells undergo columnarization by growing longer lateral membranes.

The failure of α-Spec FCCs to form a columnar epithelium is not a consequence of the tissue architecture being disrupted, as the above measurements were performed in mutant monolayers. Our findings show that SBMS is required for the control of cell shape not only in erythrocytes, but also in epithelial cells. Furthermore, both cell shortening and aberrant localization of SJ components are cell-autonomous defects of α-Spec FCCs within a monolayered epithelium, thus possibly preceding the onset of multilayering.

The phenotype of α-Spec mutant FCs resembles loss of integrin function. Cells mutant for myospheroid (mys), which encodes for the only β-chain in the ovary (Devenport and Brown, 2004; Fernández-Miñán et al., 2007), form multilayers, and display differentiation and polarity defects at the ectopic layers, but exit mitosis properly. In addition, we have previously shown that integrins regulate cell shape in the wing epithelium (Dominguez-Gimenez et al., 2007). Similarly, mys FCs also show shape defects, with a reduced height, but a similar width, compared with control FCs (Fig. S3B). Thus, both integrins and α-Spec are required for apical contraction-independent cell elongation during FE morphogenesis.

We noticed that α-Spec mutant FCs accumulate more F-actin (Fig. 6C, inset; Fig. S4A,B). The actomyosin cytoskeleton has been repeatedly linked to cell shape changes. Rho1 and its effector Myosin II regulate apical-basal length of wing disc cells (Widmann and Dahmann, 2009), and FCs mutant for Rok (Rho kinase) or for the regulatory chain of Myosin II [also known as MRLC or spaghetti squash (sqh)] present an abnormal shape (Wang and Riechmann, 2007). The spectrin cytoskeleton provides erythrocytes with mechanical properties, and is somehow functionally linked to the actomyosin network in the Drosophila eye (Deng et al., 2015). Clear functional interactions also exist between integrins and actomyosin (He et al., 2010), and we have shown here that α-Spec and mys FCs show similar phenotypes, including aberrant shape. These observations prompted us to test whether integrins and α-Spec interact with the Rho1-myosin pathway during FE morphogenesis.

To study whether α-Spec might interact with the actomyosin cytoskeleton in regulating FE architecture, we first analyzed the effects of increasing myosin activity in cells that lack α-Spec, using several strategies. Firstly, we overexpressed Sqh (fused to GFP or mCherry) in α-Spec FCC-containing egg chambers, and found it to increase S3-6 multilayers from 48% (n=25) to 66% (n=27) (Fig. 7). This multilayering enhancement is more dramatic at mid-oogenesis, as all S7/8 egg chambers show more than two layers when an extra copy of Sqh was expressed in α-Spec cells (n=20), compared with 50% when the FCs were only mutant for α-Spec (n=24) (Fig. 7E; Fig. S5A-C). The fact that all α-Spec S7/8 egg chambers display ectopic layers when Sqh was overexpressed prevented us from assessing whether increased myosin enhanced the cell shape phenotype.

Secondly, we increased Rho1 levels in α-Spec mutant cells. We were unable to obtain an FE that overexpressed Rho1 (GFP- or mCherry-tagged) and also had α-Spec FCCs. Occasionally, we obtained a few cells of the right genotype, but they showed clear signs of undergoing death (data not shown). As an alternative, we overexpressed Rho1 in α-Spec mutant cells by driving expression of UAS-α-SpecRNAi, UAS-Rho1 or UAS-constitutively active-(CA)-Rho1 by the FC-specific driver trafficjam(tj)-Gal4. S7/8 egg chambers expressing α-SpecRNAi in FCs show multilayers in 33% of the cases (n=12), whereas overexpression of Rho1 or CA-Rho1 in these cells increases the multilayering phenotype to 50% (n=12) and 80% (n=10), respectively.

We then tested whether a reduction in myosin activity might rescue the α-Spec multilayering. To do this, we reduced the levels of the Myosin II gene zipper (zip) by driving the expression of UAS-zipRNAi with tj-Gal4, and inducing α-Spec FCCs in the same epithelium. Similarly, we expressed a zip-dominant negative-(DN) form in cells that are also mutant for α-Spec (Fig. S6). In both cases, we observed a reduction in the S7/8 multilayering phenotype, from 90% (control, n=10) to 45% (zipRNAi, n=11), and from 50% (control, n=10) to 20% (zipDN, n=10).

As elimination of integrins affect FCs similarly to the loss of α-Spec, we next tested whether integrins might also interact with the actomyosin cytoskeleton to control FE architecture. Indeed, we find that overexpression of Sqh (fused to either GFP or mCherry) in egg chambers containing mys mutant FCs increases the percentage of S3-6 follicles with multilayers from 42% (mys alone, n=21) to 68% (sqh-GFP, n=97) or 82% (sqh-mCherry, n=56) (Fig. 8).

To investigate how specific this increased stratification effect of myosin is, we analyzed the impact that overexpressing Sqh has on the multilayering phenotype of bazooka⁸¹⁵ FCs. We observe that increasing myosin activity does not enhance bazooka⁸¹⁵ multilayers (Fig. S5D-F). All these results together suggest that spectrins and integrins are required to maintain the proper levels of actomyosin activity in the monolayered FE. This is further supported by the fact that both F-actin and Sqh are misplaced and expressed at higher levels in α-Spec mutant cells (Fig. 6C; Fig. S4).

Defects in FE tension and morphogenesis are likely to result in egg chambers with aberrant shapes (Wang and Riechmann, 2007). Young egg chambers are spherical, elongating from S5 along the anterior-posterior (AP) axis to create the elliptical shape of the egg. We noticed that α-Spec egg chambers seemed rounder than wild-type egg chambers (Figs 1, 3, 7). In fact, S8/9 mutant egg chambers with a multilayered FE either at the anterior, or both anterior and posterior poles, had a shorter AP axis (Fig. S7). This does not seem to be the case when multilayers were only present at the posterior, as the AP axis was not statistically different from controls. Even though mutant egg chambers seem rounder, the dorsoventral axis is slightly shorter in α-Spec egg chambers than in controls, suggesting that the total volume of the germline in α-Spec egg chambers is smaller.

Egg chambers with integrin mutant FCs render rounder eggs (Bateman et al., 2001). To study whether this is similar in eggs resulting from α-Spec mutants, we measured the AP axis in eggs obtained from α-SpecRNAi, tj-Gal4 egg chambers, and observed that, similar to integrins, these eggs also show a reduced AP axis compared with controls (Fig. S7C).

We find that in the germline α-Spec is not a major regulator of the Hippo pathway. Mutations in hippo, β-Spec or α-Spec result in a stratified FE, but contrary to previous interpretations, and unlike Hippo, spectrins are not required for the FCs to exit mitosis. We believe that the suggestion that Spec mutant FCs over-proliferate was an over-interpretation from the multilayering phenotype, as α-Spec cells were not checked for mitotic markers in that report (Fletcher et al., 2015). Again unlike hippo, α-Spec mutant PFCs only show defects in differentiation when they are located in the ectopic layers of the stratified FE, and oocyte polarity is largely unaffected in mutant egg chambers. It was recently shown that a β-Spec allele with a premature stop codon at amino-acid 1046 partially phenocopies hippo, with strong defects in FE integrity, actin organization and oocyte polarity (Wong et al., 2015). The null β-Spec^G113 mutant allele (Hulsmeier et al., 2007) behaves similarly to α-Spec mutants, showing Hnt defects mainly in ectopic layers (Fig. S8A-B,F), but Fas3 mislocalization in monolayers (Fig. S8C-D′). More importantly, β-Spec^G113 FCCs exit mitosis properly (Fig. S8A,A‴). The differences observed between the two β-Spec alleles are likely to be due to the fact that β-Spec^G113 is a null allele.

In conclusion, α-Spec and β-Spec FCCs do not phenocopy hippo mutants when the cells are part of a monolayer, and they seem to adopt a partial hippo-like differentiation phenotype only when positioned at ectopic layers, even though α-Spec and β-Spec cells never divide after S6. Thus, the main function of the spectrin cytoskeleton in FCs is not proliferation control or regulation of the Hippo pathway, although an interaction between spectrins and Hippo might occur once the FCs are within an aberrantly organized FE. The function of spectrins in FCs is in contrast with other tissues, where α- and β-Spec appear to regulate growth through Hippo (Deng et al., 2015; Fletcher et al., 2015; Wong et al., 2015).

Similar to Hippo, α-Spec and β-Spec are required for the FE to maintain a monolayer. There is an increase in the multilayering phenotype in egg chambers with large clones from S3/6 to S7/8 [36% (n=37) and 100% (n=34), respectively; Fig. S9A-B]. Also, the presence of control cells in α-Spec mosaic epithelia aids the mutant cells to maintain a monolayer from S6, as there is a higher percentage of S7-9 egg chambers with multilayers when the FE contains large α-Spec clones (Fig. S9C, ‘full’; 95%, n=54) than when the mutant clone is only at the posterior end (Fig. S9C, ‘post’; 53%, n=49). The control of FE architecture appears to be mediated by the lateral spectrin network. Loss of α-Spec seems to disrupt both lateral (α/β) and apical (α/β_H) SBMS in the FE, as β and β_H subunits are no longer localized laterally and apically in α-Spec cells (Fig. S8G; Lee et al., 1997), but no multilayering was reported for β_H-Spec egg chambers, in which a loss of apical α-Spec was observed (Zarnescu and Thomas, 1999), suggesting that the loss of the lateral α/β is responsible for the FE stratification. Also, β_H-Spec is mislocalized in sosie mutants, but the FE architecture is maintained (Urwyler et al., 2012).

Incipient SJs are first detected between the FCs with the completion of proliferation at S6 (Mahowald, 1972; Müller, 2000; this work). We show here that the localization of several SJ components is affected in α-Spec FCCs, suggesting that spectrins are required for proper SJ formation. This is further supported by other observations. First, Fas3 localization is affected in β-Spec FCCs. Second, Neuroglian (an SJ component) is required for maintaining the stability of the FE (Wei et al., 2004). Third, the reduction of both α- and β-Spec leads to mislocalization of Dlg, Neuroglian and Fas2 in neuromuscular junctions (Featherstone et al., 2001; Pielage et al., 2005). And fourth, it has been suggested that the SBMS and ankyrin associate with SJ components (Bennett and Chen, 2001; Dubreuil et al., 2001).

As the mislocalization of SJ components in Spec mutant FCCs is observed in monolayers, and thus prior to the onset of stratification, we speculate that Spec-dependent distribution of SJ components might contribute to the Spec function in the epithelium. This idea is supported by Crumbs overexpression, which leads to defects in SJs and ZA, and multilayering of the ectoderm cells (Klebes and Knust, 2000; Wodarz et al., 1995), and by dpak (Pak – FlyBase) FCs, which mislocalize Fas3 and show multilayering and columnarization defects (Conder et al., 2007). Furthermore, the aberrant accumulation of Fas2 at the lateral membrane of Tao FCs prevented membrane shrinking in the cuboidal-to-squamous transition (Gomez et al., 2012). However, fas3, fas2 and cora mutant cells do not show shape defects or multilayering (data not shown). Thus, if SJ components contribute to the α-Spec phenotype at all, it might be not because they are absent in α-Spec mutant cells, but because they are not properly distributed.

Transitions between squamous, cuboidal and columnar epithelial cell shapes are common during development, and contribute to the morphogenesis of tissues. Here, we demonstrate a cell-autonomous role for α-Spec in promoting the cuboidal-to-columnar shape transition of the FCs. It is important to point out that the FE undergoes lateral elongation without apical constriction (Kolahi et al., 2009; Fig. S3), which might allow phenotypes to be interpreted in a simpler manner. This morphogenetic FC behavior is similar to that of vertebrate neuroepithelia, where cell elongation precedes apical constriction (Suzuki et al., 2012), and it would be interesting to study the function of Spec in the columnarization of these cells.

Although the molecular mechanism of apical constriction-independent cell elongation is unknown, we think that a primary role for the SBMS lies in facilitating changes in cell shape, which is further supported by the cell shape defects in α-Spec gut epithelia (Lee et al., 1993), perhaps by contributing to the proper distribution of adhesion molecules. This function of the SBMS in membrane biology is conserved in other cells, as spectrins stabilize the plasma membrane during blastoderm cellularization (Pesacreta et al., 1989), and control photoreceptor morphogenesis through the modulation of membrane domains (Chen et al., 2009; Williams et al., 2004). The spectrin cytoskeleton might also impact on FE columnarization by interacting with the actomyosin cytoskeleton. It is known that apical-basal elongation in drebrin E (drebrin 1) depleted human Caco2 cells is impaired, as a possible consequence of the lack of interaction between drebrin E with spectrins and actomyosin (Bazellieres et al., 2012). Also, the elongation of neuroepithelial cells depends on the assembly of an actomyosin network in the apical junctional complex, regardless of whether cells are constricting or not (Hildebrand, 2005). In Drosophila wing discs, the Rho1-Myosin II pathway at the apicolateral membrane seem to regulate the cuboidal-to-columnar shape transition, whereas in the germline, Rok and sqh mutant FCs fail to adopt a normal shape (Wang and Riechmann, 2007). Finally, SBMS seems to modulate cortical actomyosin contractility in the eye (Deng et al., 2015), and possibly in the FE (this work; Wong et al., 2015). Together, these data suggest that Myosin II activity is aberrant in α-Spec mutant FCs, contributing to defects in columnarization and FE architecture.

Increasing Rho1 and Sqh activities enhances the Spec multilayering phenotype, whereas reducing Myosin II activity decreases it. In addition to this functional link between the SBMS and the Rho-Myosin pathway, we also show that mys cells fail to columnarize, and that an extra copy of sqh increases the mys multilayering phenotype. It has been shown that integrins regulate the Rho-Myosin pathway to induce actomyosin-generated forces (Geiger et al., 2009). Thus, as is the case for spectrins, integrins might also control cell shape and epithelia morphogenesis by modulating the actomyosin activity.

How the SBMS and integrins might modulate actomyosin is unknown, and one possible mechanism is by regulating Myosin II activity directly. However, we would like to propose an alternative mechanism. Spectrins can bind F-actin, and integrins and spectrins interact with proteins involved in the association of F-actin with the membrane (Beaty et al., 2014; Médina et al., 2002). Furthermore, α-Spec and integrins regulate the actin cytoskeleton through Rac (Bialkowska et al., 2005). Previous studies have shown that both β-Spec and mys mutant FCs display similar defects in the basal level of F-actin (Delon and Brown, 2009; Wong et al., 2015), which are recapitulated in α-Spec mutant cells (Fig. S4A). Thus, any defects in actin organization in mys and Spec mutant FCs could in turn result in defects in the activity of Myosin II.

Regardless of whether integrins and spectrins regulate F-actin or myosin, or both, spectrins and integrins might act together. The SH3 domain of α-Spec interacts with Tes (Rotter et al., 2005), a component of integrin-dependent focal adhesions (Coutts et al., 2003), and mammalian αII-Spec stabilizes β3-integrin anchorage, suggesting α-Spec as a physical link between focal adhesions and F-actin (Ponceau et al., 2015). In the FE, we observe that α-Spec and αPS1 colocalize in the lateral, and possibly apical, membrane (Fig. S10). In addition, we show that the localization of α-Spec in mys clones, and the localization of βPS in α-Spec mutant clones, is majorly unaffected (Figs S11, S12). Furthermore, we find that expression of a constitutively active integrin that reduces multilayering of mys FCCs (Fernández-Miñán et al., 2007; Meignin et al., 2007), fails to rescue α-Spec multilayers (data not shown). Thus, we would like to propose that α-Spec and integrins act independently of each other, but as part of the same functional complex regulating the actomyosin cytoskeleton and tissue architecture.

An early event following oncogenic mutations in an epithelium is the escape of the daughter cells from the monolayered epithelium, forming disorganized masses. Spindle orientation has been linked to tumor-like growth in various tissues, and we find that there is a good correlation between spindle misorientation and ‘tumor-like masses’ at the FE: hippo, mys and α-Spec FCCs show misaligned spindles and severe multilayering (Fig. S13; Fernández-Miñán et al., 2007; Meignin et al., 2007), whereas Notch FCCs, which overproliferate, do not show multilayering or spindle orientation defects (data not shown). However, perpendicular divisions alone are insufficient to promote stratification, and a mechanism, depending on lateral cell-cell adhesions, is in place to avoid multilayering as a sole consequence of spindle misorientation (Bergstralh et al., 2013, 2015). We would like to propose that spindle misorientation contributes to FE disorganization, but that this ‘safeguard’ mechanism is somehow inactive in hippo, mys and Spec mutant FCCs. What other aspect of the mutant phenotypes might then be linked to multilayering? A clue might come from the Spec mutant and mys FCCs. First, there is an increase in the α-Spec multilayers after S6, when both FCs and egg chambers undergo various morphogenetic changes. Second, the volume of the germline surrounded by large α-Spec FCCs appears smaller. And third, Myosin II activity is increased in α-Spec and mys mutant cells. In our interpretation of the results, a proper distribution of Myosin II activity in a Spec- and integrin-dependent manner allows the right amount of forces to be distributed across the membrane and the epithelium. Thus, it is possible that proper cell-cell interactions, adequate force balance and precise spindle orientation are key to maintaining a monolayered epithelium, especially upon the mechanical stress induced by morphogenesis.

FRT2A α-Spec^rg41 : α-Spec^rg41 is an allele containing a 20-base pair deletion that creates a frame-shift and premature termination near the 5′-end of the transcript, and is a null allele (Fig. S1C-D). The original study by Lee et al. (1997) created excision clones of a p[>lacZ,α-spec>] construct in flies transheterozygous for the l(3)α-spec^rg41 chromosome and the deficiency Df (3L)R-R2. This deficiency eliminates α-Spec, but also dlt, which encodes a protein required for FE formation and polarity (Bhat et al., 1999; Klebes and Knust, 2000). Because of this, heterozygosis of dlt in the cells that lack the p[>lacZ,α-spec>] construct might have influenced the described phenotypes, and these phenotypes might not be only due to the lack of α-Spec.

sqh-GFP and sqh-mCherry are fusion proteins expression of which is driven by the endogenous sqh regulatory sequences and can substitute the sqh gene (Martin et al., 2009; Royou et al., 2004). Stocks used were: FRT42Dhippo^42-47 (Wu et al., 2003,), FRT19Aβ-spec^G113 (Hulsmeier et al., 2007), FRT101mys¹¹ (also known as mys^XG43) (Bunch et al., 1992), FRT19Abazooka⁸¹⁵ (Djiane et al., 2005), FRT40Fas3^A142 (Wells et al., 2013), FRT19AFas2^G0336 (Szafranski and Goode, 2007), FRT42Dcora¹ (Laprise et al., 2009), UAS-Rho1 (Bloomington-7334), UAS-constitutively active-CA-Rho1 (Bloomington-8144), UAS-zip^RNAi (VDRC7819), UAS-zipDN (Monier et al., 2010), UAS-α-spec^RNAi (VDRC25387).

To generate most follicle cell mutant clones, we used the heat shock flipase (hs-flp) system (Chou and Perrimon, 1992). Mutant clones were marked by the absence of GFP or RFP (Xu and Rubin, 1993). The heat shock was performed at 37°C for 2 h over 3 days during third instar larval. To generate mys and baz follicle cell clones, we used the FRT/FLP technique combined with the Gal4 system. The e22c-Gal4 driver is expressed in the follicle stem cells in the germarium and was used in combination with UAS-flp.

The Sqh experiments in Figs 7 and 8 and Fig. S5 used flies that contain three copies of the sqh gene: two endogenous ones and the fluorescently tagged version, which is expressed under the sqh promotor.

All UAS constructs were expressed by the FC-specific driver traffic-jam-Gal4 (DGRC104055), except for the UAS-zipDN experiment, which was expressed by the MARCM system (mosaic analysis with a repressible cell marker). In this case, hs-flp expression was induced by heat-shocking only adults, at 37°C for 1 h, over 2 days. GFP-positive cells were either α-Spec mutant or α-Spec mutant overexpressing zipDN-GFP.

All females analyzed were between 3 and 10 days old.

Full names of genotypes shown in each figure are listed in supplementary Materials and Methods.

Primary antibodies were: rabbit anti-β-Spec (1:500; gift of Dr Klämbt, University of Münster, Germany), mouse anti-Fas2 [1:100; 1D4, Developmental Studies Hybridoma Bank (DSHB)], chicken anti-GFP (1:2000; ab13970, Abcam), mouse anti-Arm (1:200; N2 7A1, DSHB), rat anti-E-Cadherin (1:200; DCAD2, DSHB), mouse anti-α-Spec (1:250; 3A9, DSHB), rat anti-Tubulin (1:500; MAB 1864, Chemicon), rabbit anti-aPKC (1:1000; C-20, Santa Cruz Biotechnology), mouse anti-Dlg (1:500; 4F3, DSHB), mouse anti-Fas3 (1:100; 7G10, DSHB), mouse anti-Hnt (1:15; 1g9-s, DSHB), rabbit anti-PH3 (1:500; 06-570, Upstate Biotechnology), rabbit anti-Staufen (1:3000; gift of Dr St Johnston, Gurdon Institute, University of Cambridge, UK), mouse anti-integrin βPS (1:50; CF.6G11, DSHB), mouse anti-Cora (1:100; gift of Dr Gardiol, Gurdon Institute, University of Cambridge, UK). Species-appropriate AlexaFluor488-, AlexaFluor568- and AlexaFluor647-conjugated secondary antibodies were used (1:100; Molecular Probes).

For immunostaining, we followed standard procedures (Williams et al., 2014). For F-actin staining, Alexa-coupled Phalloidin (1:200; Invitrogen) was added in PBS/2% Tween-20 for 30 min prior to final washes and mounting. The samples were mounted in Vectashield (Vector)+DAPI, or incubated with the DNA dye TOPRO-3 (1:1000; Molecular Probes) for 10 min and then mounted in Vectashield (Vector).

We thank Dr Contreras-Sepulveda for the initial analysis of α-Spec mutants; Dr Gonzalez-Reyes for discussions and comments on the manuscript; Drs Gardiol, Sanson, Ropër, Klämbt, Thompson and St Johnston for reagents.