The role of integrins in Drosophila egg chamber morphogenesis

ABSTRACT The Drosophila egg chamber comprises a germline cyst surrounded by a tightly organised epithelial monolayer, the follicular epithelium (FE). Loss of integrin function from the FE disrupts epithelial organisation at egg chamber termini, but the cause of this phenotype remains unclear. Here, we show that the β-integrin Myospheroid (Mys) is only required during early oogenesis when the pre-follicle cells form the FE. Mutation of mys disrupts both the formation of a monolayered epithelium at egg chamber termini and the morphogenesis of the stalk between adjacent egg chambers, which develops through the intercalation of two rows of cells into a single-cell-wide stalk. Secondary epithelia, like the FE, have been proposed to require adhesion to the basement membrane to polarise. However, Mys is not required for pre-follicle cell polarisation, as both follicle and stalk cells localise polarity factors correctly, despite being mispositioned. Instead, loss of integrins causes pre-follicle cells to constrict basally, detach from the basement membrane and become internalised. Thus, integrin function is dispensable for pre-follicle cell polarity but is required to maintain cellular organisation and cell shape during morphogenesis.


INTRODUCTION
Epithelial cells are the most common animal cell type, adhering to each other to form the sheets and tubes that comprise many tissues and organs. They therefore play essential roles as barriers between compartments and in regulation of the directional traffic of molecules from one side of epithelium to the other. In order to carry out these functions, all epithelial cells must be polarised, with their apical domains facing the outside or lumen and their basal domains contacting the basement membrane (Rodriguez-Boulan and Macara, 2014). The orientation of the polarity axis depends on a lateral cue from cell-cell adhesion and external cues from the apical and/or basal sides of the epithelium. In well-characterised vertebrate epithelia, polarity is oriented by integrin-mediated adhesion to the basement membrane, and disruption of integrins leads to an inverted polarity and multi-layering (Chen and Krasnow, 2012;Eaton and Simons, 1995;Yu et al., 2005;Akhtar and Streuli, 2012). By contrast, primary epithelia in Drosophila, which derive from the cellular blastoderm, do not require integrins or a basal cue for their polarity and polarise in response to apical signals (Schmidt and Grosshans, 2018). Some Drosophila epithelia form later in development from mesenchymal-to-epithelial transitions, and it has been proposed that these secondary epithelia require a basal cue to polarise (Tepass, 1997). In support of this view, the endodermal cells of the embryonic midgut must contact the basement membrane of the visceral mesoderm in order to polarise, and the enterocytes of the adult midgut require components of the integrin adhesion complex to integrate into the epithelium and polarise (Chen et al., 2018;Tepass and Hartenstein, 1994).
It is less clear whether integrin adhesion to the basement membrane is required in the other well-characterised secondary epithelium in Drosophila, which is formed by the follicle cells that surround developing female germline cysts. All cells in the Drosophila egg chamber are generated in a structure known as the germarium, which resides at the anterior tip of each ovariole (Fig. 1A). The follicle stem cells (FSCs), which produce the somatic cells in each egg chamber, lie partway along the germarium (until this point the germline cysts are surrounded by escort cells; Fig. 1A). FSC progeny migrate to surround each germline cyst as it moves through region 2 of the germarium. These progeny cells give rise to both the main follicle cells and, via a signalling relay, the polar cells and interfollicular stalk cells (Fig. 1A) (Grammont and Irvine, 2001;McGregor et al., 2002;Torres et al., 2003).
Early work suggested that the polarity of the follicular epithelium depends on a basal cue provided by contact with the basement membrane, as well as lateral adhesion between cells and an apical cue provided by contact with the germline cysts (Tanentzapf et al., 2000). Consistent with this, mutations in myospheroid (mys), the only β integrin subunit expressed in the Drosophila ovary, cause disorganisation of the follicle cell epithelium (FCE) (Delon and Brown, 2009;Devenport and Brown, 2004;Fernández-Miñán et al., 2007). This disorganisation only occurs in mutant clones at the egg chamber termini, however, and lateral clones are indistinguishable from neighbouring wild-type cells ( Fig. 1B and  Fig. 1C).
Integrins establish a connection between the cytoskeleton and the extracellular matrix (ECM), which generally lies on only one side of a cell (Maartens and Brown, 2015). Integrin signalling could therefore potentially provide a symmetry-breaking polarity cue, and it has thus been proposed that loss of integrin function in the terminal follicle cells leads to a loss of polarity either directly or through misoriented cell divisions (Fernández-Miñán et al., 2007. However, we observed that spindles are correctly oriented in mys mutant follicle cells . Furthermore, spindle misoriention in the FCE does not cause disorganisation of the tissue, as cells that are born outside of the epithelium re-integrate (Bergstralh et al., 2015). These findings prompted us to re-evaluate the role of integrins during oogenesis to investigate the cause of the mys mutant phenotype.

Myospheroid does not regulate apical-basal polarity
As previously reported, mutant clones homozygous for a null allele of mys exhibit disorganisation of the epithelium in terminal regions ( Fig. 1B) (Fernández-Miñán et al., 2008). This phenotype is also observed upon disruption of apical-basal polarity regulators [e.g. Discs large and Lethal (2) giant larvae] (Bilder et al., 2000). We therefore considered the possibility that Myospherioid regulates 2 RESEARCH ARTICLE Development (2019) 146, dev182774. doi:10.1242 follicle cell polarity. However, as previously reported, the loss of Mys function has no effect on the localisation of polarity factors in main body follicle cells (Fig. 1C) (Fernández-Miñán et al., 2008). Furthermore, the β-integrin Mys and the α-integrin Multiple edematous wings (Mew), unlike other polarity factors, do not localise in a polarised manner, although they can only function as ECM receptors on the basal side, where the ECM is normally localised (Fig. 1D,E). Disruption of Crag or Rab10, which promote the basally directed secretion of the ECM, causes epithelial disorganisation without obviously affecting the localisation of polarity determinants (Denef et al., 2008;Lerner et al., 2013). This finding raised the possibility that Mys, which binds the ECM directly, might also regulate polarised ECM secretion. We therefore examined the location of the ECM in mys mutant clones using a GFP protein trap in the viking locus (Vkg-GFP) (Collagen IV, α2 chain) (Morin et al., 2001). In the areas of disorganisation, accumulations of Vkg are present within the epithelial layer, presumably because it is secreted by the disorganised cells and then trapped between them (Fig. 1F). However, in contrast to crag mutant follicle cells, Vkg was not found apical to the mys mutant cells (i.e. between the follicle cells and the germline cells; Fig. 1F′, dashed line). This is true whether or not the mutant cells were part of a disorganised area. This result suggests that epithelial disorganisation in mys mutant clones is not a consequence of mispolarised secretion of ECM components.
Integrins are not the only transmembrane proteins capable of interacting with the ECM. Dystroglycan (Dg), like the Mys/Mew integrin heterodimer, is expressed until stage 12 of oogenesis, binds to laminin and, via intracellular binding partners, links the ECM to the actin cytoskeleton (Ibraghimov-Beskrovnaya et al., 1992). It is therefore possible that integrins and Dg act as redundant ECM receptors to provide a basal polarity cue. As previously reported, mitotic clones of the null allele Dg 086 affect neither the polarity nor the organisation of the follicle cells, regardless of whether they occur at the termini or along the sides of an egg chamber (Fig. 1G,  Fig. S1A) (Christoforou et al., 2008;Haack et al., 2013). We therefore generated mys null mutant clones marked by the loss of RFP and Dg null mutant clones marked by the loss of GFP in the same egg chambers, and examined the double-mutant clones lacking both RFP and GFP (Fig. 1H-I″). Cells mutant for both mys and Dg still polarise correctly, provided that they are in contact with the germline and therefore can receive any apical polarity cues (Fig. 1H). Furthermore, the presence of double-mutant cells at the termini of egg chambers does not make the epithelial disorganisation discernibly worse (Fig. 1I). More importantly, the lateral polarity marker Discs large (Dlg; Dlg1) is still excluded from the basal side of mys Dg double-mutant cells in the disorganised regions that lack contact with the germline, indicating that the basal domain has been specified as distinct from the lateral domain ( Fig. 1I′, arrowheads). Thus, the ability of integrin mutant cells to polarise is not due to a redundant function with Dystroglycan. This suggests that the follicle cells do not need a basal cue to polarise and that the mys phenotype is caused by some other defect.
The mys disorganisation phenotype originates in the germarium We next asked when the mys phenotype arises during oogenesis. To address this question, we depleted mys mRNA from the entire epithelium at different stages of development. GR1-GAL4 is expressed in the follicle cells of fully formed egg chambers, starting weakly at approximately stage 2/3 of oogenesis and getting stronger through development ( Fig. 2A,A′). Expressing a shRNA directed against mys under the control of GR1-GAL4 is lethal at 25 or 29°C, probably because GR1-GAL4 reduces Mys function in other tissues during larval or pupal development. However, adult flies eclose if the larvae and pupae are grown at 18°C, at which temperature the GAL4 system is inefficient. We therefore raised GR1-GAL4; UASmys-shRNA flies at 18°C, then shifted them to 29°C for 2 or 6 days before dissecting their ovaries. Under these conditions, Mys is strongly depleted from stage 3/4 of oogenesis onwards (Fig. 2B,B′). The morphology of the FCE is normal in these flies ( We used Traffic Jam-GAL4 (TJ-GAL4) to investigate a role for Mys prior to stage 2/3. TJ-Gal4 is expressed in all somatic cells in the Drosophila ovary, apart from the interfollicular stalk cells (Fig. S2B, arrowhead). Flies expressing mys-shRNA under the control of TJ-GAL4 demonstrate gross disorganisation of ovarioles, with multi-layering of the follicle cells and incomplete encapsulation and fusion of the germline cysts ( Fig. 2D-E″). An identical phenotype is produced by a second mys-shRNA that targets a different region of the mRNA, confirming that this phenotype is due to Mys knockdown (Fig. S2C). TJ-GAL4 drives expression in the germarium in both pre-follicle cells and escort cells, but these latter cells are unlikely to be involved in the phenotype, as disorganisation first appears in the region where the FSCs reside (Fig. 2D,D′, arrow). We also considered the possibility that the phenotype arises during ovariole development, because TJ-GAL4 is expressed in the pupal ovary (Vlachos et al., 2015). This was tested by allowing the flies to develop and eclose at 18°C. These flies developed morphologically wild-type germaria (only 1 of 12 germaria examined displayed mild disorganisation) and egg  , aPKC (green), F-actin (red) and DNA (blue). The flies were raised until eclosion at 18°C and transferred to 29°C as adults for 2 days before dissection. Mys staining is lost from the follicle cells from stage 3/4 onwards (B′,C″), but this has no effect on tissue organisation or apical-basal polarity (C′) (n=6). (D,D′) An ovariole expressing UAS-mys-RNAi under the control of TJ-Gal4, stained for Mys (green), aPKC (white), F-actin (red) and DNA (blue). Mys is strongly depleted from the somatic cells at all stages, resulting in gross disorganisation of the ovariole from the approximate position of the follicle stem cells (arrow). Blue shading marks the early germline cysts (n=31). (E-E″) An ovariole expressing UAS-mys-RNAi under the control of TJ-Gal4, stained for aPKC (green), FasII (red) and DNA (blue). The follicle cells that contact the germline are correctly polarised. Early germline cysts are outlined in dashed white lines (n=4). E′ and E″ show separate channels. Scale bars: 10 μm. ovary, and still develops when Mys is specifically knocked down in the adult germarium. Taken together, these findings indicate that the epithelial disorganisation in mys mutants arises in the germarium before stage 2/3 of oogenesis, and that this is the only stage when Mys function is required for the formation of a normal follicular epithelium.
Individual germline cysts still remain distinct from the somatic tissue and undergo their usual morphological changes, such as transforming from a disc of cells into a sphere ( Fig. 2E-E″). They are also capable of generating eggs, although these are generally misshapen, probably because integrins are required for the egg chamber rotation that is required for normal egg elongation ( Fig. S2H) (Lewellyn et al., 2013). Furthermore, those follicle cells contacting the germline are able to polarise, as evident by the Fig. 3. See next page for legend.

RESEARCH ARTICLE
Development (2019)  Integrins are required to anchor the FSCs to their niche, and mys mutant FSCs move into the centre of the germarium and are more frequently lost than wild-type FSCs, or become quiescent (Hartman et al., 2015;O'Reilly et al., 2008). However, the FSCs still produce an approximately normal number of pre-follicle cells when Myospheroid is knocked down. This may be because TJ-GAL4 is not expressed in the FSCs (although it appears that TJ-GAL4 is expressed in all somatic cells in the germanium; Fig. S2B) or because there is no competition between wild-type and mutant FSCs, as Mys is knocked down in all cells. Thus, in addition to this role in FSC maintenance, Mys is required in the FSCs and/or the pre-follicle cells for normal egg chamber morphogenesis.

Mys participates in stalk formation
The observations that ovarioles fail to separate into discrete egg chambers when Mys is strongly depleted in the germarium (and onwards) and that epithelial disorganisation only occurs when mitotic clones occur at the egg chamber termini, raised the possibility that disorganisation reflects a defect in interfollicular stalk formation. We tested whether the interfollicular stalk cells are able to differentiate from the other pre-follicle cells, which go on to form the follicle cells and polar cells, by staining for the stalk cell marker Lamin C (Fig. 3A). This protein is still expressed in patches of cells between germline cysts when Mys is depleted using TJ-GAL4; UASmys-shRNA (Fig. 3B). Furthermore, Eyes absent (Eya), which is expressed by the follicle cells but not the interfollicular stalk and polar cells, is expressed in cells contacting the germline but not in patches of cells between the cysts (Fig. S2I,J). Thus, the somatic cells appear to be able to undergo their typical position-dependent differentiation programmes into follicle, polar, and interfollicular stalk cells in the absence of integrins. This suggests that the lack of interfollicular stalks upon Mys depletion is due to a failure in the cell shape changes and movements that usually generate this structure. Because little is known about the morphogenesis of interfollicular stalks, we set out to characterise this process in more detail. The stalk begins as a double row of elongated cells between developing germline cysts that arrange themselves so that their apical sides meet along the midline of the germarium (Fig. S3A). As previously described, these cells then undergo an intercalation, such that the two-cell-thick column becomes a single row of cells (Godt and Laski, 1995) (Fig. S3A). In the early stages of stalk formation, aPKC localises to the apical tips of the stalk cells, facing the centre of the column (Fig. 3C-C‴′,H, bracket 1). As intercalation proceeds, aPKC does not remain associated with the tips and instead forms small dots on the lateral membrane that lie along the centre of the column (Fig. 3D). The intercalating cells eventually resolve into a single line of cells with staggered nuclei. At this point, the aPKC foci have moved to the middle of the lateral interface between adjacent cells ( Fig. 3E-E‴,H, bracket 2). Finally, the cells fully intercalate such that their nuclei are arranged in a line ( Fig. 3F-F‴). A small focus of aPKC persists at the centre of the interface between cells, but this eventually disappears and aPKC is almost undetectable (Fig. 3F-G‴). This leaves only lateral polarity factors, such as Dlg, localised along the length of the stalk cell-cell contacts (Fig. 3H, bracket 3). By contrast, lateral cell adhesion components such as FasII and Fasciclin III (Fas3), which are expressed along the lengths of cell-cell contacts while the cells are in the germarium, are lost over the course of intercalation (Fig. S3B,C).
Adherens junction components, such as E-cadherin (Shotgun), demonstrate a similar pattern of localisation to aPKC (Fig. S3D,E). A similar dynamic localisation of E-cadherin has also been described in the developing basal stalks in the Drosophila pupal ovary, which link developing ovarioles to the posterior end of the ovary (Vlachos et al., 2015;King, 1970). Like intrafollicular stalks, basal stalks are formed by a cell intercalation event that generates a single row of cells from (in this case) a disorganised area of cells (Godt and Laski, 1995). Examination of developing basal stalks reveals that aPKC also localises to the central tips of these cells as they line up two abreast prior to intercalation (Fig. 3I,I′, Fig. S3F). Once a single row of cells has been established, single accumulations of aPKC can be seen between each cell along the midline of the basal stalk (Fig. 3J,J′, Fig. S3G). These findings suggest that the intercalation events in the pupal basal stalks and adult interfollicular stalks are driven by a highly similar sequence of morphological events.
As outlined above, interfollicular stalks are made by a complex series of cell movements and shape changes (Fig. 3K). Throughout this process, integrins (both Mys and the α subunit Mew) are strongly enriched along the basal sides of the stalk cells, in contrast to their uniform distribution around the membrane later in development (Fig. 4A, Fig. S4A). To determine whether Mys is required cell-autonomously in the presumptive stalk cells for stalk morphogenesis, we used the 24B-GAL4 driver to express UAS-Flipase (UAS-Flp), in the interfollicular stalks (Fig. 4B,B′, arrows) to generate clones specifically in these cells. Although this does result in the majority of clones being found in the stalks, it also produces a few clones in the main follicle cells (Fig. 4C). This suggests that either 24B-GAL4 is also expressed at low levels in the follicle cells (which cannot be detected when 24B-GAL4 is used to drive the expression of UAS-LacZ; Fig. 4B) or that the stalk cells are still capable of becoming follicle cells when they first begin to differentiate. Nevertheless, these clones allowed the evaluation of the role of Mys in the interfollicular stalk precursors. Interfollicular stalks containing mys mutant cells are malformed compared with wild-type stalks of a similar age (Fig. 4C,D). This does not appear to be due to a failure of these cells to differentiate, as Lamin C and 24B-GAL4 are still expressed in mutant cells in the disorganised stalks (Fig. 4E,E′, Fig. S4B). Eya expression is also downregulated in both mutant and wild-type cells (Fig. 4F,F′). Instead of forming single lines of cells, the interfollicular stalks containing mys clones are a disorganised mass that can contain wild-type as well as mutant cells (Fig. 4G-H′). Fully mature wild-type stalks are made up of cells with a rectangular cross-section, whereas the mys mutant Fig. 4. See next page for legend. cells are round in appearance and usually lie in the centre of the cluster (Fig. 4G-I′). Dlg, however, is still only enriched in the regions where cells contact their neighbours (Fig. 4G′,H′), whereas aPKC expression is absent in the more mature disorganised stalks, as seen in wild-type interfollicular stalks (Fig. 4J,J′). Thus, the intercalation process does not occur correctly in the absence of integrin function, resulting in disordered cell masses several cells thick.
Our observation that Mys is required for the morphogenesis of interfollicular stalks raised the possibility that the areas of epithelial disorganisation at the ends of each egg chamber are in fact misshapen stalks. To test this hypothesis, we generated randomly positioned mys clones (using a heat shock-Flp), while also driving lacZ expression with the stalk-specific driver 24B-GAL4. The cells in the disorganised regions of the follicular epithelium at the termini did not express 24B-GAL4, indicating that they are not formed by malformed interfollicular stalks, which abut them posteriorly (Fig. 4K,K′). When younger egg chambers were examined (before interfollicular stalks have fully formed), the region of epithelial disorganisation was only partially composed of 24B-GAL4-expressing cells, which lie posteriorly (Fig. 4L,L′). Thus, mys mutants disrupt the morphogenesis of the terminal follicle cells as well as that of the interfollicular stalks. Our results also suggest that areas of disorganisation anterior to the interfollicular stalk are formed from the follicle cells, like those in Fig. 4L (dashed line), that are neither in contact with the germline (to their anterior) nor destined to become the stalk.

Mys controls somatic cell positioning in the germarium
Because both the mys RNAi and early mutant clones demonstrate that the mys phenotype arises in the germarium, we analysed the mutant phenotype at these stages in more detail. In wild-type germaria, one surface of every somatic cell contacts the ECM, whether they are follicle cell precursors or future stalk and polar cells (Fig. 5A,A′). By contrast, mys mutant cells often lose contact with the ECM and become internalised within the germarium (Fig. 5B, white asterisks). This internalisation means that cells are mispositioned when egg chamber morphogenesis begins, and this is likely to be the cause of the disorganised interfollicular stalks and follicular epithelium at the termini later in development (Fig. 5C).
Whereas F-actin and the myosin regulatory light chain Spaghetti Squash (Sqh) are primarily apical in later-stage follicle cells, they are basally enriched in pre-follicle cells (Fig. 6A-A″, Fig. S4C,D) (Alégot et al., 2018;Finegan et al., 2019). This suggests that the basal surface of a pre-follicle cell is under contractile tension. Fig. 6B-B‴′ shows three cells, each with a reduced basal surface, at various stages of cell internalisation. Mutant cells typically show an increased basal enrichment of myosin compared with the adjacent wild-type cells (Fig. 6B-C‴′). Thus, the mutant cells may respond to the loss of friction provided by integrin-ECM adhesion by undergoing a basal contraction that helps to internalise them.
Finally, we asked whether the loss of integrin disrupts the polarity of the interfollicular stalk cells, because these cells never contact the germline and therefore cannot receive the germline-dependent apical polarity cue. We found that mys mutant cells still localise aPKC to their presumptive apical sides even if they are internalised, suggesting that attachment to the ECM is not a requirement for polarity (Fig. 6D,D′). Furthermore, aPKC still localises to the apical sides of mutant cells when two mutant cells are facing each other (Fig. 6D,D′, asterisks). Thus, integrins are also not required for the establishment of apical-basal polarity in the interfollicular stalk cells, despite their lack of an apical cue.

DISCUSSION
Although it has long been known that integrins are required for the correct organisation of the follicular epithelium at the egg chamber termini, the reason for this defect has remained unclear. Here, we show that disorganisation in mys mutant tissue arises from the failure of the pre-follicle cells to remain attached to the ECM in the germarium. Integrins are not required to maintain the organisation of the epithelium once it has formed. Our analysis reveals that all pre-follicle cells remain in contact with the ECM of the basement membrane as they surround a new germline cyst to form an egg chamber. Loss of integrin function results in mutant cells that detach from the ECM and become internalised, which disrupts the arrangement of cells in a single layer (Fig. 7). As development continues, cells that contact a germline cyst are able to develop as normal, as are any mutant cells that are generated once an egg chamber has been formed. However, mutant cells that lie between germline cysts cannot undergo the normal morphogenetic movements. Signalling from the posterior germline cyst to induce interfollicular stalk cell fate occurs normally despite the altered arrangement of cells, but these cells do not rearrange to form a singlecell-wide stalk and remain as a cluster. Finally, those cells that neither become stalk nor are in contact with the germ line develop into the disorganised regions of cells seen at the termini of later-stage egg chambers, immediately adjacent to the stalk cells. Thus, the mass of disordered mys mutant cells at egg chamber termini consists of both terminal and stalk cells. This may explain some of the observations that suggested a later role for integrins in Notch signalling, in which marker expression was interpreted on the assumption that all cells were terminal cells (Gómez-Lamarca et al., 2014).
Those mutant pre-follicle cells that lose their connection to the ECM appear to undergo a basal constriction, leading to them delaminating from the monolayer into the interior of the tissue. This is similar to a proposed apical cell extrusion process, in which a cell that becomes detached from the ECM is forced out of the tissue by its neighbours, which spread to fill the gap (McCaffrey and Macara, 2011). However, due to the architecture of this particular tissue, the 'extruded cells' move apically into the tissue rather than out of it. As the mutant pre-follicle cells delaminate, the basal constriction is often associated with increased levels of basal myosin, suggesting that increased myosin contractility may facilitate invagination of the mutant cells. This raises the possibility that integrin adhesion to the ECM normally represses basal acto-myosin contractility, as has been observed in tissue culture cells (Arthur and Burridge, 2001;Arthur et al., 2000). In support of this view, increasing the levels of the myosin regulatory light chain, Sqh, significantly enhances the epithelial multi-layering phenotype of mys mutant clones at the termini (Ng et al., 2016).
The integrin adhesion complex is required for apical-basal polarisation of the other secondary epithelium in Drosophila, the midgut, as well as a number of vertebrate epithelia, where it is thought to transduce a basal polarity cue from the basement membrane (Akhtar and Streuli, 2012;Chen and Krasnow, 2012;Chen et al., 2018;Tepass and Hartenstein, 1994;Yu et al., 2005). By contrast, integrins appear to perform a purely adhesive function in the pre-follicle cells and play no discernible role in their polarisation. The follicle cells that contact the germ line polarise in response to an unknown apical cue (Tanentzapf et al., 2000). However, mys mutant follicle cells that fail to contact the germ line still manage to partially polarise by generating distinct basal and lateral domains, as revealed by Dlg localisation. Furthermore, a distinct basal domain still forms when the second major ECM receptor, Dystroglycan, is also removed. Thus, integrins and Dystroglycan do not function as redundant receptors in the follicle cells for a basal polarity cue from the ECM. One possibility is that there is another ECM receptor that transduces the basal polarity signal. Alternatively, Dlg may be targeted to lateral membranes in response to cell-cell adhesion. In this scenario, there would be no basal polarity cue and instead basal polarity would be specified by the absence of cell-cell contact.
It is even more surprising that the interfollicular stalk precursors polarise normally in the absence of integrins, because these cells never contact the germ line and therefore cannot receive an apical polarity cue, yet they still localise aPKC to the apical domain. This suggests that the cells read some basal cue that generates a polarity that targets apical factors, such as aPKC, to the opposite side of the cell, but the underlying mechanisms are entirely unknown.
The formation of the stalk provides an interesting example of how apical-basal polarity can be remodelled during morphogenesis. The stalk forms from a column of cells that is initially two cells wide, in which each cell forms an apical domain facing the cell on opposite side. As the cells move past each other to form a one-cell-diameter rod, aPKC and E-cadherin do not remain at the leading 'apical' edge of each cell, but instead form a lateral spot that remains approximately in the centre of the stalk, before they eventually disappear. Stalk formation is essentially a process of cell intercalation that resembles the initial steps in the formation of the notochord. In Ascidians, for example, presumptive notochord cells intercalate to transform a two-cell-wide rod into a line of single cells (Veeman and McDonald, 2016;Oda-Ishii et al., 2010). Furthermore, integrin-mediated adhesion to the basement membrane plays a key role in both stalk and notochord morphogenesis (Buisson et al., 2014;Pulina et al., 2014). The stalk could therefore provide a genetically tractable model for investigating how such cell intercalation processes occur.

Mutant alleles
The following Drosophila melanogaster mutant alleles were used: mys xG43 (Wieschaus et al., 1984) and Dg 086 (Christoforou et al., 2008), recombined with FRT19A and FRTG13, respectively. Both stocks were combined to allow the generation of clones mutant for both mys and Dg in the same egg chamber.

UAS lines
The UAS-Mys-shRNA lines TRiP.JF02819 (Bloomington Drosophila Stock Center, 27735) and TRiP.HMS00043 (Bloomington Drosophila Stock Center, 33642) were used to deplete Mys and were generated by the Transgenic RNAi Research Project (Perkins et al., 2015). UAS-LacZ and UAS-mCD8::GFP were used to show expression patterns of the above Gal4 lines.
Other stocks w 1118 (Bloomington 5905 and 6326) and y,w (Bloomington 1495) were both obtained from Bloomington.

Immunostaining
Ovaries were fixed for 15 min in 4% paraformaldehyde and 0.2% Tween 20 in PBS. Ovaries were then incubated in 10% bovine serum albumin (in PBS with 0.2% Tween 20) to block for at least 1 h at room temperature. Primary and secondary immunostainings lasted at least 3 h in PBS with 0.2% Tween 20. Three washes (approximately 10 min each) in PBS-0.2% Tween 20 were carried out between stainings and after the secondary staining. Primary antibodies were diluted 1:200. Secondary antibodies and Phalloidin were diluted 1:200.

Imaging
Imaging was performed using an Olympus IX81 (40×/1.3 UPlan FLN Oil or 60×/1.35 UPlanSApo Oil). Images were collected with Olympus Fluoview Ver 3.1. with the exception those in of Fig. S2H, which were imaged using a Fig. 7. A model for how mys mutant cells become internalised. The epithelial disorganisation seems to result from a failure of mutant cells (green) to withstand basal contractions, because these cells are not stably connected to the ECM. The cells are therefore unable to maintain their basal area and undergo a basal constriction (arrows in middle panel), causing them to be drawn into the tissue.
Leica MZ16F dissecting microscope fitted with a Qimaging Qiclick camera and collected using QCapture Pro 6.0. Image processing was performed using ImageJ.

Drosophila genetics
Follicle cell clones of mys XG43 and Dg 086 were induced indiscriminately both alone and simultaneously by incubating hs-FLP; FRT mutant/FRT marker larvae or pupae at 37°C for 2 h out of every 12 h over a period of at least 2 days. Apart from the heat shocks, the crosses were raised at 25°C. Adult females were dissected at least 2 days after the last heat shock. mys XG43 clone production was targeted to the interfollicular stalk cells by using 24B-Gal4 to drive the expression of UAS-FLP. Crosses were raised at 25°C and adult females were dissected at least 2 days after they had eclosed. RNAi lines against Mys were crossed with TJ-Gal4 and either raised solely at 25°C (Fig. 2E,F, Fig. 3B, Fig. S2C,H), solely at 18°C (Fig. S2D,E) or at 18°C and then adult flies were shifted to 29°C for 2 days before dissection (Fig. 2C,D, Fig. S2F,G). RNAi lines against Mys were crossed with GR1-Gal4 and had to be raised at 18°C in order to produce adults, which were then moved to 29°C for 2 (Fig. 2C,D) or 6 ( Fig. S2A) days before dissection.