Response of Drosophila epithelial cell and tissue shape to external forces in vivo

Many tissues are subject to external forces in vivo. Endothelial cells are exposed to shear by blood flow, lung epithelial cells are compressed during an asthma attack, and the epithelial lining of the bladder is stretched by urine filling. Such external forces result in changes to the shape of individual cells, as well as to the tissue and the entire organ. For instance, during filling, the bladder expands and bulges while the epithelial lining reduces in height and individual cells expand to flatten. These changes are governed by cell-intrinsic mechanical properties that confer compliance or resistance to external forces (St Johnston and Sanson, 2011). Importantly, cell-intrinsic properties and external forces have to be precisely balanced. An epithelial cell that is compliant to being stretched can expand to accommodate the surface of a growing organ. In contrast, high stiffness allows the cell to resist deformation but could ultimately cause catastrophic rupture of the epithelial lining. Although the role of actomyosin and adhesion in promoting cell-intrinsic shape changes in single cells and epithelial sheets is well documented (Heisenberg and Bellaïche, 2013; Munjal and Lecuit, 2014; Harris, 2018; Röper, 2015; Takeichi, 2014), we have insufficient understanding of how cells resist or comply with external forces to balance cell, tissue and organ shape in vivo (Mao et al., 2013; Legoff et al., 2013; Mao and Baum, 2015; Ladoux and Mège, 2017; Sethi et al., 2017).

We wanted to investigate specifically how epithelial cells accommodate expansion of an internal structure in vivo and which cellular mechanisms determine three-dimensional (3D) cell shape when an epithelium is being stretched. In the developing Drosophila egg chamber, about 850 epithelial follicle cells surround 15 germline-derived nurse cells and one oocyte (Duhart et al., 2017). During 14 stages of development, the germ line grows in size and elongates to give rise to a large cylindrical egg. Throughout all stages, the apical surfaces of follicle cells face the interior germ line (Fig. S1A-A″). Thus, any change in the germline surface must be matched by a change in the apical surface of the epithelium and associated adherens junction (AJ) network. Importantly, after cell divisions cease at stage 6, about 50 cells of anterior fate undergo a cuboidal-to-squamous shape transition and come to overlie all nurse cells. The remaining cells of main body and posterior fate undergo a cuboidal-to-columnar shape transition to overlie the oocyte (Kolahi et al., 2009; Horne-Badovinac and Bilder, 2005; Xi et al., 2003) (Fig. S1A‴). The bulk of these shape transitions occur at stage 9. Previous models suggest that anterior cell flattening represents a compliant response to germline growth, whereas cuboidal and columnar shapes resist flattening through high apical stiffness (Kolahi et al., 2009; Horne-Badovinac and Bilder, 2005; Koch, 1963; Wu et al., 2008; Wang and Riechmann, 2007). Although regulators of apical–basal polarity, AJs and actomyosin are all essential to follicle cell shape (Wang and Riechmann, 2007; Zarnescu and Thomas, 1999; Baum and Perrimon, 2001; Conder et al., 2007; Ng et al., 2016; Tanentzapf et al., 2000; Grammont, 2007), theoretical considerations concerning how the post-mitotic follicle epithelium accommodates germline growth and, at the same time, resists expansion to promote cuboidal and columnar shapes have not been formally tested.

Because the apical-junctional surface of follicle cells faces the germ line, we first analyzed how AJs and the apical cortex are organized during the early phases of post-mitotic growth between stages 6 to 9. We focused our analysis on cuboidal main body cells, a cell shape common to all stages. Strikingly, we observed pronounced changes in AJ appearance. Specifically, AJs identified by localization of E-cadherin (E-cad) and β-catenin (β-cat) became increasingly corrugated (Fig. 1A-A″) (Sherrard and Fehon, 2015). We define corrugations as the ratio of observed junction length to that of a straight line. We present the difference between the corrugation value measured and the corrugation value of a straight line as the relative surplus junction length (see Materials and Methods). Specifically, the relative surplus junction length quadrupled between stages 6 and 9 (from 0.016±0.003 to 0.068±0.004). Importantly, corrugations were also observed in anterior flattening and posterior columnarizing cells at stage 9, indicating that AJ corrugations increased in the entire epithelium (Fig. S1B). Corrugations were exclusively observed at the level of AJs; basolateral membranes remained straight (Fig. 1B-C″).

These changes in AJ appearance coincided with pronounced reorganization of Myosin II (MyoII). At stage 6, the active phosphorylated form of MyoII regulatory light chain (MRLC, also known as Spaghetti squash, Sqh) and the MyoII heavy chain (MHC, also known as Zipper, Zip) localized primarily to AJs. Strikingly, at stage 9, MyoII was depleted from AJs and strongly enriched medially (Fig. 1D-G‴; Fig. S1C,C′). Medial MyoII enrichment was observed in anterior, main body and posterior cells at stage 9, indicating that the shift in MyoII localization occurred in the entire epithelium (Fig. S1D). Importantly, during live imaging of the apical cortex at stage 9, medial MyoII did not display oscillatory dynamics in the time frame published for other epithelial models (Fig. S1E) (Alégot et al., 2018; Mason et al., 2013; Martin et al., 2009; He et al., 2010). A shift in MyoII localization and an increase in AJ corrugations could also be observed in live imaging using Airyscan technology, excluding processing artifacts as a source for changes in junctional architecture (Fig. S1F-G′).

To understand better how corrugating AJs interface with the medial cortex, we visualized localization of E-cad, MyoII and Actin. Actin filaments visualized by follicle cell-specific expression of utABD-GFP under the control of tj-GAL4 were enriched at junctions and in the medial cortex after stage 6 (Fig. S1H-I″). Of note, cells positioned over the oocyte had higher levels of apical Actin, reflecting microvilli differentiation, and were not analyzed further (Fig. S1J,J′) (Schlichting, 2006). Using Airyscan imaging, we observed that medial and junctional Actin filaments interdigitated with MyoII filaments, suggesting that medial MyoII and Actin localization are coordinated (Fig. S8). Importantly, multiple examples of MyoII filaments connecting to corrugating junctions and forming intercellular cables with MyoII filaments of neighboring cells could be observed (Fig. 1H-H″). Combined, these observations describe pronounced reorganization of the medial-junctional surface during post-mitotic stages of germline growth and indicate that a contractile medial actomyosin cortex is connected to AJs. In agreement with this conclusion, we observed expansion of the apical surface upon circular laser ablation of the medial cortex in main body cuboidal cells at stage 9 (Fig. 1I-I″). This demonstrated that medial contractility acting on AJs specifically restricts apical surface area expansion.

To quantitatively understand how tension at AJs can be altered by such profound rearrangements during egg chamber development, we measured recoil velocities upon laser ablation of individual AJ bonds (Shivakumar and Lenne, 2016; Farhadifar et al., 2007; Sugimura et al., 2016) in main body cuboidal cells at stages 6 and 9. Contrary to a naive expectation that the epithelium would be stretched by the increase in germline surface, we surprisingly observed that junctional tension decreased from stages 6 to 9 (Fig. 1J,K,M-M″; Movies 2 and 3). Moreover, during nurse cell dumping at stage 11 (Duhart et al., 2017; Horne-Badovinac and Bilder, 2005), when columnar cells need to accommodate the rapid increase in oocyte surface, AJ tension dropped to undetectable levels (Fig. 1L,M,M″; Movie 4). Combined, these observations strongly support a model whereby the developmentally controlled relaxation of AJ tension decreases the internal resistance of the junctional network to external stretching forces imposed by germline growth. This would allow epithelial cells to adapt their apical area smoothly to that of the growing germline surface. We speculate that junctional depletion of MyoII and increased AJ length support the developmentally coordinated reduction in AJ tension.

If junctional tension in cuboidal main body cells decreases between stages 6 and 9, why are cuboidal cells not flattening and how can cells columnarize as the germ line expands? A mathematical model suggests that cuboidal and columnar shapes resist flattening by an anterior to posterior gradient of relatively increasing apical stiffness (Kolahi et al., 2009; Horne-Badovinac and Bilder, 2005). To test this idea, we first examined the circumstances under which cuboidal cells transition to columnar shapes. Between stages 5 and 10A, the oocyte increases in size more rapidly than nurse cells (Fig. 2A,E) (Kolahi et al., 2009). As a result, follicle cells initially positioned over nurse cells come, row by row, into contact with the growing oocyte. Specifically, whereas most main body cells, as identified by mirr expression (Xi et al., 2003), contact the nurse cells at stage 8, all main body cells contact the oocyte by stage 10A (Fig. 2B,B′,F). Of note, posterior terminal cells, as identified by pnt expression (Xi et al., 2003), contact only the oocyte during all stages (Fig. S2A).

Between stages 5 and 10A, all posterior and main body cells transition from a cuboidal to a columnar aspect ratio via a cell shape gradient observed most prominently at stage 9 (Duhart et al., 2017; Kolahi et al., 2009). At each stage, cells in contact with the oocyte were taller in height and had smaller apical surface areas than cells still contacting nurse cells (Fig. 2C-D′). Thus, posterior and main body cells in contact with the oocyte were columnar, whereas main body cells still contacting nurse cells were cuboidal. To account for this contact-correlated behavior of future columnar cells, we distinguished between nurse-cell-contacting cells (NCCs) and oocyte-contacting cells (OCCs) (Fig. 2G). A previous study revealed that all future columnar cells grow equally in volume (Kolahi et al., 2009). Thus, the differences in apical area and height of NCCs and OCCs describe different cell aspect ratios accommodating the same cell volume. As a result, apical area or cell height individually serve as proxy for follicle cell shape.

The contact-dependent behavior of cell shape suggested that a simple cell-intrinsic property gradient was not sufficient to explain the existence of cuboidal and columnar shapes, but that the germline contact surface could be crucial in defining cell shape. Importantly, oocyte-derived patterning signals such as epidermal growth factor (EGF) are insufficient to explain contact-dependent differences. Expression of a constitutively active EGF receptor in follicle cells prevents specification of anterior fates (Xi et al., 2003) and thus squamous cell flattening at stage 9 (Fig. S2B). However, despite an increase in cell size (compare Fig. S2C,D with Fig. 2C,D), the areas of cells in contact with nurse cells (NCCs) were still significantly larger and their heights smaller than for cells in contact with the oocyte (OCCs) (Fig. S2C,D). In fact, we observed a 3.9-fold change in NCC to OCC aspect ratios (12.8 versus 3.2), compared with a 2.3-fold change in wild-type cells (5.3 versus 2.0). Conversely, loss of EGF signaling does not prevent acquisition of columnar shape in cells over the oocyte (González-Reyes et al., 1995) and EGF signaling is not activated in columnarizing ventral cells (Nilson and Schüpbach, 1999). Thus, oocyte-derived EGF signaling is required for fate patterning but is neither sufficient nor necessary to promote the transition of main body cells from a cuboidal to columnar shape.

To prove that the germline contact surface is necessary for alternate cell shapes, we analyzed egg chambers with mispositioned oocytes. In egg chambers homozygous for acf1¹ (Börner et al., 2016) or fat2^N103.2 (Horne-Badovinac et al., 2012), OCCs were still taller in height than NCCs (Fig. S2E). Moreover, an analysis of cell shapes in wild-type epithelia surrounding an E-cad-RNAi-expressing germ line with a mispositioned oocyte (Godt and Tepass, 1998; González-Reyes and St Johnston, 1998) revealed that OCCs were taller in height and their apical surfaces were smaller than in NCCs (Fig. 2H-K). Thus, by genetically separating columnar shape from topological position within the chamber and from patterning by posterior pole cells, in agreement with previous studies (González-Reyes and St Johnston, 1994), we found that cells lacking anterior fate only acquire columnar shape if in contact with the oocyte and maintain cuboidal shape while still in contact with nurse cells.

Because the apical-junctional surface of follicle cell faces the germ line, we speculated that position over nurse cells or oocytes regulates the apical cortex and AJ network to control changes in apical areas and, consequently, the shift in aspect ratio from cuboidal to columnar shape. We therefore analyzed localization of apical actomyosin and AJ markers between stages 7 and 9 to understand better how germline contact regulates these cell shape transitions.

In contrast to our expectation that columnar OCCs would present with relatively higher stiffness or contractility, we observed that cuboidal NCCs had relatively higher levels of medial MyoII than OCCs (Fig. 3A-B′; Fig. S3A). Importantly, MyoII levels in the medial cortex dropped sharply at the NCC–OCC boundary. Follicle cell-specific expression of a fluorophore-tagged Sqh verified that the apical domain alone reflected these differences in fluorescence intensity (Fig. 3C-D′; Fig. S3B). In contrast, basal MyoII patterns did not correlate with nurse cell or oocyte contact (Fig. S3C) and, as previously reported for stages 8 and 9 (He et al., 2010), levels were significantly lower than apical levels (Fig. 3D-D″). Combined, these observations suggest that maintenance of high levels of medial MyoII in cuboidal NCCs is directly or indirectly promoted by nurse cell contact. Accordingly, high MyoII levels also correlated with nurse cell contact in egg chambers containing a mislocalized oocyte (Fig. 3E,E′; Fig. S3D-D″), confirming that nurse cell contact rather than main body patterning per se is sufficient for high MyoII levels. Of note, squamous-fated cells displayed reduced levels of MyoII starting at stage 7-8 (Fig. 3A,B), indicating that this nurse cell-contacting population regulates MyoII differently; thus, it was not analyzed further. We also observed higher levels of E-cad and β-cat in NCCs than in OCCs at stages 8 and 9 (Fig. S3E,F). However, levels declined gradually in OCCs, even in egg chambers with a mislocalized oocyte (Fig. S3G). Currently, we cannot imply a cellular mechanism that regulates MyoII recruitment or E-cad expression. However, we speculate that medial MyoII recruitment is promoted by direct contact of the apical NCC surface with nurse cells, whereas E-cad is downregulated, possibly transcriptionally, when cells contact the oocyte.

To understand whether high MyoII levels in NCCs and low levels in OCCs correlate with differences in junctional tension, we analyzed vertex recoil velocities upon laser ablation in NCCs and OCCs at stage 9. Indeed, AJ tension in NCCs was higher than in OCCs (Fig. 3F,F′). Importantly, AJ tension in NCCs was dependent on MyoII. RNAi-mediated knockdown of Rok, the gene encoding MRLC kinase, caused a significant decrease in the measured initial recoil velocities at stage 9 (Fig. S3H-H″). If more medial MyoII is the source of high AJ tension in NCCs, we might expect that AJs in NCCs would be more corrugated than in OCCs, as the contractile medial cortex would deflect AJs more strongly. Indeed, the relative surplus junction length in NCCs was longer than in OCCs (0.093±0.012 and 0.029±0.004, respectively). Importantly, almost 75% of this difference arose between the last NCC (0.075±0.009) and the first OCC (0.027±0.006) row, consistent with the sharp drop in medial MyoII levels over this boundary (Fig. 3G-G‴). Combined, our results demonstrate that, at stage 9, high levels of medial MyoII and strong AJ corrugations coincide with relatively higher levels of AJ tension in NCCs than in OCCs.

High levels of medial MyoII and AJ tension in NCCs at stage 9 contradicted our expectations about the regulation of cuboidal and columnar shapes. We expected that smaller apical areas upon contact with the oocyte would depend on high contractility, whereas the larger apical areas in NCCs would be associated with lower contractility. Our observations suggested that different levels of external forces could underlie differences in NCC and OCC AJ tension.

Thus, to dissect the source and relevance of high AJ tension in NCCs, we genetically manipulated regulators of actomyosin contractility. Using RNAi-mediated knockdown driven by tj-GAL4, a driver displaying highest activity after stage 5, we reduced Rok function in the entire epithelium (Fig. S3H). Strikingly, whereas OCCs displayed only minor alterations to their cell shape at stage 9, RNAi-expressing NCCs responded with an expansion of apical areas and a reduction in lateral heights (Fig. 4A-B′,D-I). Apical expansion of NCCs was not a result of cell loss or multinucleation, often observed when MyoII function is reduced in mitotic stages 2 to 5 (Wang and Riechmann, 2007). In fact, at stage 9, epithelia expressing Rok-RNAi only rarely contained binucleate cells and cell numbers were conserved (31.0±0.3 and 30.6±0.2 cells along the anterior–posterior (A/P) axis in wild-type and tj-GAL4, UAS-Rok-RNAi chambers, respectively; n=5 each). These observations were confirmed by expressing Rok-RNAi specifically in main body cells only using the mirr-GAL4 driver, demonstrating that aberrant NCC flattening does not arise because of impaired anterior cell flattening or posterior cell columnarization (Fig. 4C,C′,H-I; Fig. S4A,A′). Moreover, an analysis of hypomorphic sqh¹ mutant clones revealed that sqh¹ NCCs experienced much larger changes to their apical surface area than sqh¹ OCCs, compared with surrounding wild-type NCC and OCC areas (Fig. 4J-K). Importantly, the observed increase in apical areas of Rok-RNAi or sqh¹ NCCs occurred at the expense of AJ corrugations (Fig. 4D,E; Fig. S4B,B′). This provides firm evidence for the idea that AJ corrugations are maintained by medial MyoII, which pulls AJs into the medial cortex and thereby reduces the effective apical area. Combined, these results demonstrate a strong requirement for high medial MyoII in NCCs to prevent apical area expansion, and thus a shift towards a flatter NCC aspect ratio.

To confirm that medial MyoII connecting to AJs contributes to controlling apical NCC areas between stages 6 and 9, we genetically manipulated core AJ components. Mosaic analysis of E-cad null clones revealed minimal changes to cell shape, because of compensation by N-cad (data not shown) (Loyer et al., 2015). We thus expressed α-catenin (α-cat) RNAi in the epithelium to eliminate the catenin-mediated linkage of E-cad or N-cad to Actin and thus AJ formation (Röper, 2015; Takeichi, 2014; Brasch et al., 2012). Using this strategy, we found that NCCs expressing α-cat-RNAi were flattened by stage 9 with smooth lateral membranes, whereas OCCs exhibited less distorted cell shapes (Fig. 5A-B′,E,F,H,I; Fig. S5A). These results establish that AJs mediate the function of high medial contractility in NCCs.

However, as loss of Catenins prevents cell adhesion and therefore transmission of external and internal forces through the junctional network, we wanted to analyze further the role of AJs by modulating rather than abrogating AJ function. Strikingly, we found that overexpression of Afadin (Also known as Canoe, Cno) (Mandai et al., 2013; Bonello et al., 2018; Walther et al., 2018), an AJ regulator that co-localizes with β-cat to follicle cell junctions (Fig. S5B,B′), caused flattening of NCCs by stage 9 (Fig. 5C,C′,G,J-L). In contrast, cno-expressing OCCs acquired small apical areas but, importantly, displayed severe AJ hypercorrugation (Fig. 5G). These phenotypes were observed when Cno was overexpressed in all follicle cells using tj-GAL4 or in main body cells only using mirr-GAL4 (Fig. 5D,D′,K,L; Fig. S5C,C′). Apical expansion of cno-expressing NCCs was not a result of cell loss (31.0±0.3 and 30.4±0.4 cells in a row along the A/P axis in wild-type and tj-GAL4, UAS-cno chambers, respectively; n=5 each). Furthermore, neither medial MyoII (Fig. S5D,E) nor corrugations were altered [relative surplus junction length in cno-expressing NCCs was 0.070±0.069 compared with 0.068±0.053 for wild-type NCCs; P=0.75 in a Wilcoxon–Mann–Whitney (WMW) test] suggesting that Cno did not alter actomyosin-coupling to junctions (Sawyer et al., 2009). Instead, overexpression of a GFP-tagged Cno revealed that Cno localized with E-cad in vesicle-like structures at AJs (Fig. S5F-F″), indicating a possible function in E-cad trafficking, as reported for mammalian Afadin (Tachibana et al., 2000; Hoshino et al., 2005). Moreover, the apical areas of NCCs expressing cno were larger than for NCCs expressing Rok-RNAi, where only contractility on the junctional circumference is reduced (compare Fig. 4H with Fig. 5K). Combined, these observations strongly suggest that Cno overexpression increases absolute AJ length, thereby causing expansion of the apical NCC surface. We further deduce that total AJ length delimits the maximum apical NCC area, which medial MyoII contractility reduces to the actually observed corrugated NCC surface. Final surface areas of NCCs subject to external stretching forces thus depend on a functional ratio between medial contractility and AJ length.

Combined, our observations demonstrate that maintenance of apical NCC areas, and thereby of cuboidal shape, relies on the precise balance between AJ length and medial MyoII contractility. In contrast, apical OCC areas, and thereby acquisition of columnar shape, did not exhibit a similarly strong requirement for AJ function or MyoII contractility. The results suggest that NCCs and OCCs exhibit different molecular requirements for maintenance of their apical areas and associated cell shapes.

To assess the tissue-level consequences of the specific sensitivity of NCCs to deregulation of MyoII and AJ function, we closely analyzed egg chambers with Rok-RNAi- or cno-expressing epithelia.

By convention, egg chambers in which anterior cells initiate flattening and cuboidal NCCs are still positioned over nurse cells are scored as stage 9 (Duhart et al., 2017; Horne-Badovinac and Bilder, 2005; Koch, 1963; Spradling, 1993). Surprisingly, epithelia expressing Rok-RNAi or cno with aberrantly flattened NCCs enclosed significantly larger germ lines than wild-type epithelia at stage 9; in fact, germ lines were more similar to a size normally observed at stage 10A (Fig. 6A-D). To test whether these germ lines also displayed a nurse cell to oocyte ratio characteristic of stage 10A, we measured nurse cell and oocyte sizes. Indeed, the nurse cell to oocyte ratio in egg chambers with stage 9 epithelia expressing Rok-RNAi or cno was closer to that of stage 10A wild-type egg chambers (Fig. 6E). Combined, these results indicate that aberrant flattening of epithelia expressing Rok-RNAi or cno delays the cuboidal-to-columnar shape transition, thereby creating a developmental mismatch with the growing germ line. Ultimately, abnormal flattening of NCCs expressing Rok-RNAi or cno expands the total surface area of the main body cell population, causing a delay and even failure of all main body cells to acquire contact with the growing oocyte (Fig. 6F). Specifically, the severe expansion of cno-expressing NCC areas correlates with egg chamber degeneration at stage 9 and female sterility (Fig. S6A-C). In conclusion, restriction of apical NCC areas is crucial in facilitating contact with the expanding oocyte and for complete cuboidal-to-columnar shape transitions by stage 10A.

We were puzzled by the observation that NCCs responded more sensitively than OCCs to the manipulation of actomyosin and AJ function by expanding apical areas at stage 9. This implies that contact with nurse cells drives apical NCC expansion. Nurse cells could drive expansion by coordinating surface growth or surface shape with NCCs. A 1.8-fold increase in nurse cell surface contributes to germline growth between stages 8 and 10A (Kolahi et al., 2009). Thus, the up to fivefold increase in apical areas of cno-expressing NCCs is not sufficiently accounted for by growth alone.

We therefore asked whether nurse cell shape contributed to apical expansion of genetically manipulated NCCs. Strikingly, apical expansion in epithelia expressing Rok-RNAi or cno coincided with bulging of individual nurse cells into apical NCC surfaces at stage 9 (Fig. 7A,B). Furthermore, deformation of the apical surface coincided with significant widening of the dorsal–ventral (D/V) axis of the nurse cell cluster, concomitant with shortening of the A/P axis (Fig. 7C). Significant differences in the aspect ratio of total egg chambers containing epithelia expressing Rok-RNAi or cno were not observed at stages 7 and 8 (Fig. S7A). This excluded defects acquired during rotation-driven axis elongation up to stage 8 (Duhart et al., 2017; Haigo and Bilder, 2011; Bilder and Haigo, 2012; Cetera et al., 2014) as a cause of nurse cell cluster aspect ratio changes at stage 9. Combined, the results suggest that apical NCC surfaces constrain bulging of individual nurse cells and widening of the D/V axis of the nurse cell cluster at stage 9.

To provide additional evidence for this idea, we investigated the shape of nurse cells in the absence of external constraints. At stages 8 and 9, nurse cell shape could be externally constrained by the basement membrane and cuboidal NCCs. To separate these constraints, we first enzymatically removed the basement membrane from a stage 10A egg chamber using collagenase. At stage 10A, nurse cells are only covered by the basement membrane and ultrathin squamous cells, the latter of which are not expected to contribute significantly to the combined material properties of the interface between nurse and epithelial cells. Indeed, removal of the basement membrane caused rounding of individual nurse cells and reduction of the nurse cell cluster aspect ratio to that of a much rounder shape (Fig. 7D,E; Fig. S7B) (Chlasta et al., 2017). This demonstrates that, in the absence of any external constraints, the default shape of the nurse cell cluster is round rather than elongated.

To test whether the basement membrane is necessary for constraint of nurse cell shape at earlier stages, when cuboidal NCCs are still in contact with nurse cells, we enzymatically removed the basement membrane from stage 8 egg chambers (Fig. S7C). Importantly, the D/V aspect ratio of the nurse cell cluster did not change, demonstrating that the NCC epithelium is sufficient to constrain nurse cell shape (Fig. 7F,G). To prove that NCC contractility prevents nurse cell cluster rounding, we eliminated NCCs. RNAi-mediated knockdown of the apical determinant atypical protein kinase C (aPKC) causes extreme NCC flattening, essentially eliminating NCC function at stage 9 (Fig. 7H). In response, nurse cells bulged into the basement membrane-enclosed space. To prove that this shape was not just passively filling space but was normally constrained by the epithelium, we additionally removed the basement membrane. Strikingly, nurse cells expanded along the D/V axis, suggesting that prior to stage 10A, nurse cells also need to be actively compressed at NCC positions to prevent them from acquiring a round and compact shape (Fig. 7H). Importantly, when we removed the basement membrane from a wild-type stage 9 egg chamber, we observed that nurse cell bulging at the anterior pole coincided with constriction of egg chambers at NCC positions (Fig. 7I). This suggests that, when forces are not balanced at the anterior pole covered by thin flat cells, NCCs can squeeze nurse cells and cause them to bulge anteriorly. Taken together, these results demonstrate that the high levels of medial MyoII and AJ tension observed at NCC positions create a circumferentially contractile sleeve around nurse cells to ensure nurse cell cluster elongation along the A/P axis during egg chamber growth at stage 9 (Fig. 7J,J′).

This study explores how a junctional network accommodates expansion of an internal structure in vivo and how cells can resist deformation into a flat shape as the epithelium is being stretched. We found that remodeling of the medial-junctional cortex correlated with a decrease in AJ tension in follicle cells between stages 6 to 9, despite a naive expectation that growth of the germline surface stretches and thus increases tension in the epithelium. Furthermore, contrary to our expectation that resistance to being stretched supports the differentiation of columnar shape (Kolahi et al., 2009), we found that cuboidal cells resist being stretched by recruiting high levels of medial MyoII if in contact with nurse cells. At the level of individual cells, this prevents cell flattening. However, at the tissue level, it ensures that all main body cells make contact with the oocyte, which is crucial for further egg development. Moreover, at the organ level, medial MyoII recruitment and high AJ tension in NCCs ensure that organ growth is channeled into elongation.

In addition to a shift in expression of MyoII regulators (West et al., 2017; Gutzman and Sive, 2010), a decrease in AJ tension during stages 6 and 9 might be caused by a shift from junctional to medial MyoII localization. Medial contractility acting at an angle to AJs could reduce the effective force felt by cellular vertices, compared with contractility acting in parallel to AJs. Currently, we do not know what drives MyoII into the medial cortex. Mechanosensing of germline growth could provide external cues for MyoII and, consequently, AJ remodeling (Chanet et al., 2017; Weng and Wieschaus, 2016). However, we speculate that the developmentally controlled relaxation of AJ tension allows the junctional network to adapt the apical follicle cell area smoothly to that of the growing germline surface while, at the same time, maintaining epithelial integrity.

AJ corrugations have been described to arise by medial MyoII-induced ratchet-like cycles of apical constriction in other tissue (Röper, 2015; Mason et al., 2013; Martin et al., 2009). Although we did not detect MyoII oscillations, we found that medial MyoII connected to corrugating AJs and that reduction of Rok or sqh function expanded the apical surface as AJs straighten. This demonstrates that corrugations do not arise from surplus AJ length under low tension, which encloses a limited apical surface, but that corrugations arise by medial MyoII contractility. However, hypercorrugated junctions in cno-expressing OCCs indicate that an extreme surplus of AJ length under relatively low tension also promotes corrugations. We thus conclude that the balance between absolute junction length, medial contractility and external forces regulates apical surface size. Consequently, excessive expansion of cno-expressing NCCs occurs despite the presence of medial MyoII, whereas expansion observed upon Rok knockdown is limited by normal junction length.

We found that NCCs have relatively high levels of medial MyoII and AJ tension, and respond more sensitively to manipulation of MyoII and AJ function than OCCs. This occurs even though NCCs and many OCCs are principally of the same main body fate. This suggests that contact to nurse cells drives NCC behavior and recruits MyoII into the medial cortex to reinforce the apical surface against deformation. We tested whether E-cad-mediated adhesion of NCCs to nurse cells could drive NCC behavior. However, RNAi-mediated double knockdown of N-cad and E-cad in the germ line inhibited the migration of border cells, as reported previously (Niewiadomska et al., 1999), but did not disrupt cell shape transitions until after stage 10 (not shown). Therefore, E-cad-dependent adhesion cannot account for NCC-specific behaviors and future studies need to address these currently unknown mechanisms of NCC–nurse cell communication. In contrast to NCCs, we found that OCCs display reduced levels of MyoII and junctional tension, and lower sensitivity to loss of MyoII and AJ function. This indicates that columnarization is not driven by cell-intrinsic apical constriction. Instead, we speculate that a fate-specified columnar shape of main body cells is stretched into a cuboidal aspect ratio by contact with nurse cells at stage 9.

Previous studies suggest that elongated egg shape is determined by a molecular corset at the basal surface of the follicle epithelium channeling growth of the egg chamber into the A/P axis (He et al., 2010; Haigo and Bilder, 2011; Bilder and Haigo, 2012; Cetera et al., 2014; Chlasta et al., 2017; Andersen and Horne-Badovinac, 2016; Crest et al., 2017; Chen et al., 2016; Cetera and Horne-Badovinac, 2015). Polarized ECM properties act between stages 2 and early 9 (Haigo and Bilder, 2011; Bilder and Haigo, 2012; Chlasta et al., 2017; Crest et al., 2017). Basal actomyosin contractions ensure egg elongation from late 9 to 10B (He et al., 2010). Our study suggests a mechanism that ensures nurse cell cluster elongation during stage 9, where apical MyoII levels in main body cells are higher than at the basal side. Moreover, we demonstrated that basal MyoII enrichment does not correlate with nurse cell contact and, thus, does not correlate with the contact-dependent sensitivity of cell shapes we observed for NCCs. Our data are consistent with a model in which relatively higher levels of apical NCC contractility establishes a contractile sleeve that constrains nurse cell bulging and nurse cell cluster rounding. Accordingly, genetic reduction of apical contractility or an increase in AJ length causes nurse cell rounding as NCCs expand and flatten. Importantly, apical contractility also shapes the egg chamber prior to stage 6 (Alégot et al., 2018). Thus, the crucial importance of the apical domain prior to stage 6 and at stage 9 suggests that basal and apical constraints ensure egg elongation at different stages of development.

All experiments were performed on Drosophila melanogaster. For detailed genotypes, please refer to Table S1. Stocks and experimental crosses were maintained on standard fly food at 18°C or 25°C. For mirr-GAL4-driven expression of Rok-RNAi and cno, adult females were shifted to 30°C for 48 h to switch off tub-GAL80-mediated repression of GAL4 prior to dissection. Mosaic analysis was performed using the FLP/FRT system (del Valle Rodríguez et al., 2011). For follicle epithelium clones, Flipase (FLP) expression was induced in young adult females using heat shock for 1 h at 37°C. For germline clones, FLP expression was induced for 1 h at 37°C at 96 h and 120 h after egg lay at 25°C. Flies were fed yeast paste for 48-72 h prior to dissection.

Ovaries were dissected and fixed in 4% formaldehyde in PBS for 15 min at 22°C. Washes were performed in PBS containing 0.1% Triton X-100 (PBT). Ovaries were incubated with the following primary antibodies in PBT overnight at 4°C: guinea-pig anti-Spaghetti-squash 1P (MRLC-1P) (1:400, gift from Robert Ward, University of Kansas, Lawrence, KA, USA), mouse β-catenin (1:100; DSHB, N27A1), rat anti-E-cadherin (1:50; DSHB, DCAD2), rabbit anti-GFP (1:200, Thermo Fisher Scientific, G10362), rat anti-RFP (1:20, gift from H. Leonhardt, Ludwig Maximilians University of Munich, Germany, 5F8), mouse anti-Dlg (1:100; DSHB, 4F3), rat anti N-cadherin (1:20; DSHB, DN-EXH8), mouse anti-protein kinase C ζ (1:50; Santa Cruz Biotechnology, H-1, sc-17781), mouse anti-β-galactosidase (1:1000; Promega, Z378B). Ovaries were incubated with secondary antibodies for 2 h at 22°C. Nuclei were stained with DAPI (0.25 ng/μl; Sigma-Aldrich) and F-actin stained with phalloidin (coupled to Alexa Fluor 488 or Alexa Fluor 647, 1:100, from Molecular Probes or phalloidin-TRITC, 1:400, from Sigma-Aldrich). Egg chambers were mounted using Molecular Probes Antifade Reagents. Samples were imaged using Leica TCS SP5, SP8 or ZeissLSM880 confocal microscopes. Samples were processed in parallel and images were acquired using the same confocal settings, if fluorescence intensities were to be compared. Higher resolution images were obtained using an Airyscan detector on a Zeiss LSM880 confocal microscope and were post-processed with ZEN (Huff, 2015). Images were processed and analyzed using Fiji (ImageJ, 1.48b) (Schindelin et al., 2012).

Individual ovarioles were dissected out of the muscle sheet and mounted on a standard microscope slide with a minimal volume of Schneider's medium supplemented with FBS and insulin as described (Prasad et al., 2007). Spacers were fashioned from double-sided tape; slides were covered with a coverslip and sealed with Halocarbon oil. Super-resolution imaging was performed using an Airyscan detector on a Zeiss LSM880 confocal microscope and post-processed with ZEN (Huff, 2015). Images were acquired at 30 s time intervals.

Individual ovarioles were dissected from the surrounding muscle sheet and incubated in Schneider's medium supplemented with 1000 Units/ml collagenase (CLSPA; Worthington Biochemical Corp) for up to 30 min, rinsed in 1× PBS three times and then fixed and immunostained individually as described above in an eight-well tissue culture dish.

Laser ablation on live-egg chambers expressing shg-GFP (Huang et al., 2009) were performed on two setups: using the inverted microscope setup described previously (Farhadifar et al., 2007) (Figs 2 and 3) or an inverted Zeiss Spinning Disc (Yokogawa CSU-22) with laser ablation unit (Rapp OptoElectronic) (Fig. S3). Briefly, individual ovarioles were dissected out of the muscle sheet and mounted on a standard microscope slide with spacers fashioned from double-sided tape, covered with a coverslip and sealed with Halocarbon oil (Sigma-Aldrich). Experiments were performed on freshly dissected ovarioles prepared every 20 min. Then, 32 pulses/µm of the laser (λ=355 nm, 1000 Hz) were applied at a length of 0.22 µm for ablations of cell–cell junctions or circular cuts with radii adjusted to the size of the apical surface. Images were taken every 0.3 s or 0.5 s for up to 40 s.

All images and movies were analyzed in Fiji (ImageJ, 1.48b) (Schindelin et al., 2012), unless otherwise stated. Graphs were generated with Microsoft Excel 365 or R version 3.2.0. Statistical tests were performed in R 3.2.0. Data sets were checked for normality of distribution with Shapiro's test and homogeneity of variances by applying Bartlett's or Levene's test. Statistical tests are indicated in the figure legends. The α-value for statistical analysis was set to 0.05.

Measurement of fluorescence intensity traces for junction and cytoskeleton markers was performed using the line and profile plot tools in Fiji. The surface occupied by squamous fated cells was approximated by a line of the same length as that obtained for OCCs in the same egg chamber. The remaining segment between ‘squamous-fated’ and OCC cells was denoted as NCCs. A fit was applied to the intensities using a smoothing function in R, which automatically chose a curve-fitting method based on the group of the largest size of data points between squamous fated cells, NCCs or OCCs for each stage. Apical and basal Sqh intensity in the NCCs was measured in mid-sections of egg chambers with the line tool in Fiji and the background intensity subtracted.

Apical areas of epithelial cells were measured at the level of AJs using the polygon tool in Fiji. Cell heights were measured perpendicular to the tangent of the apical surface using the line tool in a medial cross-section of the egg chamber. To obtain fold-changes in cell areas and heights for sqh¹ clones, average cell areas and heights of NCCs and OCCs in sqh1 homozygous clones and in neighboring heterozygous and homozygous wild-type cells were obtained and a fold-change ratio calculated for OCCs and NCCs within each individual egg chamber. Junctional corrugations were quantified by calculating the ratio of (1) junction length obtained by tracing the β-cat signal between two vertices using the segmented line tool and (2) the distance between the same vertices obtained by using the straight-line tool. This value is theoretically 1 when the junction is a straight line and >1 when corrugated. The relative surplus junction length is the difference between the corrugation value measured (>1) and the corrugation value of a straight line (=1).

To measure vertex displacement after ablation of AJs, a kymograph of the AJs between the two vertices was generated in Fiji. The vertices of the ablated junction were tracked pre- and post-ablation and distances between the vertices were obtained for each time point over the period of recording. For each ablation event, the change in distance between the vertices at any post-ablation time point relative to the average distance from ten pre-ablation time points was obtained. The change in distance was normalized to the average junction length across all samples within one experimental condition. The mean relative distance was then plotted as a function of time. In a first approach, a double exponential fit d(t)=d₁(1−e^−t/T1)−d₂(e^−t/T2−e^−t/T1) (Farhadifar et al., 2007; Landsberg et al., 2009) was applied to estimate the initial velocity of the average curves, where T₁ is the slow relaxation time and T₂ is the fast relaxation time of the vertices of ablated cell bonds, d₁ is the final change of distance between vertices of ablated cell bonds at t→∞ and d₂ is the change in distance due to fast relaxation only. The fit parameters were calculated and the standard error was determined as shown in Table S3. The fit parameters d₁ and T₁ were poorly estimated for some data sets by the double exponential expression given above (Table S4). Fast time scale responses are associated with linear elastic behavior of the cytoskeleton cortex and slower ones with viscous behavior. Because T₂ or fast relaxation time ranged from 0.3 to 1 s in our measurements and was well estimated, we assumed that the recoil of the vertices in this time interval was similar to that of a linear elastic solid and, therefore, the magnitude of initial velocity was directly proportional to the tension in the junctions. Thus, instead of obtaining the initial velocity v₀ by solving the equation v₀=d₁/T₁−d₂(1/T₁−1/T₂) as described previously (Landsberg et al., 2009), we obtained initial velocities by calculating the slope of the curve between t=0 and t=0.5 or 0.6 s, which is expected to approximately cover the linear phase of the curves (Mayer et al., 2010).

Using the line tool in Fiji, we measured the maximum width (W) across posterior nurse cells and the maximum length (L) of the nurse cell compartment or the total egg chamber measured from and to basal surfaces at the maximum width and length in a medial section and then calculated the ratio of length to width.

Using the polygon tool in Fiji, traces of the nurse cell compartment and oocyte were generated in the medial section of the egg chambers. Both areas were summed for total germline area and used as a proxy for the egg chamber volume. The ratio of nurse cell to oocyte areas was calculated to give the relative size.

We thank G. Salbreux, S. Grill and the Life Imaging Center (LIC, University Freiburg) for discussions and technical help with experiments. We thank R. Ward, Y. Bellaiche, E. Knust, S. Eaton, U. Tepass, M. Grammont, A. Carmena and H. Leonhardt for sharing reagents. We thank the Bloomington Drosophila Stock Center (BDSC), Vienna Drosophila Resource Center (VDRC) and Developmental Studies Hybridoma Bank (DSHB) for providing fly stocks and antibodies. We thank the IMPRS-LS and SGBM graduate schools for supporting our students.