A dynamic and mosaic basement membrane controls cell intercalation in Drosophila ovaries

Cell and tissue morphogenesis require proper assembly and coordination between the cytoskeleton and the extracellular matrix (ECM). The basement membrane (BM) is an essential, sheet-like extracellular matrix that outlines and supports epithelial tissues, and is made of highly conserved components, including Collagen IV (ColIV). Fast-growing evidence shows that BM is an active signalling platform involved in many distinct developmental processes, such as cell differentiation, proliferation, survival, polarisation and migration (Daley and Yamada, 2013; Morrissey and Sherwood, 2015; Ramos-Lewis and Page-McCaw, 2019; Yurchenco, 2011). But how exactly BM is assembled dynamically and what the functional role of its multiple components is during morphogenesis remain largely unknown.

Drosophila oogenesis is a powerful, genetically tractable model for studying the role of BMs in epithelial morphogenesis. Ovaries are made of strings of egg chambers (ovarioles) at different stages of their development, which are covered with BMs on their external (basal) side (Fig. 1A) (Duhart et al., 2017; McLaughlin and Bratu, 2015; Spradling, 1993a). Egg chambers are steadily produced in the germarium, a structure located at the tip of the ovariole (Fig. 1A, left side), and containing somatic and germline stem cells. A whole spectrum of morphogenetic events occurs during oogenesis, including cell migration, proliferation, elongation and intercalation; these processes take place in a very short period of time and involve a limited number of cells that are well characterized. Drosophila ColIV consists of two α chains, Cg25c (α1) and Viking (Vkg) (α2), forming hetero-trimers (Blumberg et al., 1988; Fessler and Fessler, 1989; Lunstrum et al., 1988; Ricard-Blum and Ruggiero, 2005; Van De Bor et al., 2015). The ColIV network self-assembles in the extracellular space and is maintained at the cell surface through its binding to the integrin trans-membrane receptors. We and other labs have previously shown that three distinct tissues produce and target ColIV to specific regions of the ovariole. First, follicular cells deposit ColIV at the basal surface of the egg chambers through a Crag/Rab10/phosphatidyl-inositol-dependent mechanism (Bunt et al., 2010; Denef et al., 2008; Devergne et al., 2014; Horne-Badovinac et al., 2012; Lerner et al., 2013; Medioni and Noselli, 2005; Van De Bor et al., 2015); second, larval and embryonic hemocytes deposit ColIV around the germarium in a juxtacrine manner (Bunt et al., 2010; Martinek et al., 2011; Van De Bor et al., 2015); and third, fat body ColIV secreted in the hemolymph is targeted around the entire ovariole (Matsubayashi et al., 2017; Pastor-Pareja and Xu, 2011; Van De Bor et al., 2015). However, mechanisms through which fat body ColIV incorporates into the mature ovariole BM remains unknown.

The egg chamber model also helped elucidate how BM acts on organ shaping. Initially spherical, egg chambers lengthen along their antero-posterior axis to form elongated embryos. It was found that follicular cells encapsulating the egg chambers produce ColIV fibrils that exert mechanical constraint perpendicular to the long axis of the egg chambers (Bilder and Haigo, 2012; Daley and Yamada, 2013; Haigo and Bilder, 2011; Isabella and Horne-Badovinac, 2016), inducing a symmetrical, decreasing gradient of stiffness from the central to the terminal regions, thus promoting terminal extension of the tissue (Crest et al., 2017).

During their complex morphogenesis, egg chambers of successive stages are separated from each other through a short stalk, composed on average of seven cells (Fig. 1A,B). Stalks are essential for maintaining ovariole organization and preventing egg chamber fusion (Besse et al., 2002; Chang et al., 2013; Kirilly and Xie, 2007; Morris and Spradling, 2011; Spradling, 1993b). Stalk and follicle cells both originate from the germarium follicle stem cells (FSCs). Hh signalling maintains the pool of FSCs, which produce follicular cells expressing Eye Absent (Eya) and stalk/polar cells precursors expressing the Castor (Cas) nuclear factor. Some of the precursors of the stalk/polar cells gradually differentiate into stalk cells and start to express Lamin C (LamC), a marker that is specific to cells subject to strong mechanical stress (Chang et al., 2013; Donohoe et al., 2018; Mellerick et al., 1992; Osmanagic-Myers et al., 2015; Wang et al., 2015; Zhang and Kalderon, 2000). Stalks are assembled through a mediolateral cell intercalation process, during which two rows of Cas-positive cells eventually intercalate to form one string of seven cells separating two adjacent egg chambers (Chang et al., 2013; Vlachos et al., 2015). Stalks serve as linkers, and one important feature is their strong resistance to mechanical stress, raising the question of the role of the BM in stalk morphogenesis and physical integrity.

In this study, we have characterised the architecture and function of BM in developing stalks. We describe five distinct stages in the formation of a mature stalk, during which the BM shows a dynamic and regulated organisation. Results show that ColIV of the stalk has multiple origins, including the stalk itself, and follicle and fat body cells. During this mosaic assembly, deposition of ColIV from the fat body is found to be dependent on initial deposition of ColIV of follicular origin; hence, pools of ColIV from different origins appear to be coordinated to assemble into a mature BM. Functional analysis shows that reduction of ColIV levels is sufficient to induce follicle fusion, a phenotype similar to the one induced by the ablation of stalk cells or a loss of hh or castor function. Additionally, ColIV and its natural receptors, integrins, are shown to be required for cell intercalation important for stalk formation but also for maintaining cell-cell attachment within the stalk. Our study reveals that dynamic assembly of ColIV from multi-tissue origin is essential for cell intercalation, for the making of functional mechanically resistant ovarioles and for female fertility.

In order to characterize the ECM architecture during stalk formation, we concomitantly analysed stalk cell fate [using the fate markers Castor (Cas) and Lamin C (LamC)] and endogenous ColIV expression (using Vkg::GFP, a protein-trap GFP insertion in the N-terminal part of the endogenous Vkg protein) (Medioni and Noselli, 2005; Morin et al., 2001). In parallel, we also analysed matrix ultrastructure and thickness by electron microscopy (EM) and immuno-fluorescence (IF) (see Materials and Methods).

Detailed analysis of the process of stalk formation allowed us to define five relevant phases (I to V) associated with distinct states of matrix assembly, as detailed thereafter (Fig. 1B-F). Phase I, initiation: between region 2a and 2b of the germarium, a slight constriction appears in which precursor stalk cells start to express the Cas marker (Fig. 1A,B-B″,G). The matrix in contact with the Cas-positive cells is thin, regular and indistinguishable from the adjacent matrix (Fig. 1B′′′, H). Phase II, convergence: between region 2b and stage 1 egg chambers, a clear constriction is now visible, in which the number of Cas-positive cells decreases with few Cas cells starting to express the LamC marker (Fig. 1A,C-C″,G). These cells flatten, show elongated nuclei and form two rows facing each other (Fig. 1C-C″,G). The matrix contacting the Cas and LamC cells exhibits significant shape modification and irregular thickening (Fig. 1C′′′,H). Phase III, intercalation: between stage 2 and stage 4 egg chambers, Cas- and LamC-positive cells have completed their intercalation process and form one string of cells separating adjacent egg chambers. From then on, stalks are composed on average of seven cells, strongly expressing the LamC marker (Fig. 1D-D″,G). The matrix around stalk cells has significantly thickened (Fig. 1D″′,H). Phase IV, posterior closing: between stage 4 and stage 6 egg chambers, an important stalk matrix modification occurs, forming an alveolar structure by expanding in between each stalk cell. In addition, a layer of matrix is deposited between the posterior extremity of the stalk and its contacting, stage 6 egg chamber (Fig. 1E-E′′′,G,H). Interestingly, at this stage, stalk cells are no longer aligned with polar cells of stage 6 egg chambers. Phase V, anterior closing: between stage 6 and stage 9 egg chambers, while stalk cells show a decreased Cas expression, matrix deposition takes place between the most anterior cell of the stalk and the adjacent stage 6 egg chamber. As previously noted, this is concomitant with the misalignment of the stalk cells and the polar cells (Fig. 1F-F′′′,G,H). In addition, both ultrastructure analysis (data not shown) and Vkg::GFP expression revealed a thin layer of matrix between the terminal cells and the inner cells of the stalk (Fig. 1F″, arrowhead). Interestingly, shortly after the anterior and posterior closing, stalk terminal cells in contact with the egg chamber matrix acquire an elongated shape compared with the inner cells of the stalks. Quantification of the thickness of ColIV-containing BM from both EM and confocal images (see Materials and Methods for details) reveals that, from stage I-to-V, the BM increases its thickness by nearly threefold (Fig. 1H).

Altogether, results show that ECM undergoes dramatic modifications during the entire process of stalk formation. Stalk cells undergo cell intercalation to acquire a chain-like organisation, which is accompanied by a dynamic, singular organisation of the ECM in between stalk cells and at the boundary with egg chambers. This architecture, together with the thickening of the matrix, likely contributes to the formation of strong cell-cell attachments within the stalk and in connection with egg chambers; hence, providing the appropriate mechanical properties of stalks.

Previous work, including ours, has shown that three distinct tissues produce and target ColIV to specific regions of the ovariole: follicular cells deposit ColIV at the basal side of the egg chambers, fat body ColIV is targeted around the entire ovariole, and embryonic and larval hemocytes deposit ColIV around the germarium (Denef et al., 2008; Devergne et al., 2014; Horne-Badovinac et al., 2012; Isabella and Horne-Badovinac, 2016; Lerner et al., 2013; Pastor-Pareja and Xu, 2011; Van De Bor et al., 2015) (Fig. 2A). To determine the origin of ColIV found in stalks, we first analysed the pattern of vkg mRNA by in situ hybridisation on wild-type ovaries together with Cas and LamC staining (Fig. 2B). Results show that stalk cells strongly express vkg mRNA during the initiation (I) and convergence (II) phases (Fig. 2C,C′). The signal then decreases during the intercalation (III) and posterior closing (IV) phases (Fig. 2D-E′). At the anterior closing (V) stage, vkg mRNA expression becomes undetectable (Fig. 2F,F′). Hence, stalk cells produce ColIV before and during the intercalation process with a dynamic and fading pattern. Interestingly, in situ hybridisation performed on 30-day-old flies revealed the presence of vkg mRNA in the fat body, suggesting that the abdominal fat body is constantly producing ColIV and contributes to the ovariole matrix for the whole adult life (Fig. 2G,G′).

To further determine the origin of ColIV in the stalk, we expressed in a Vkg::GFP background (showing endogenous expression) an RFP-tagged form of ColIV (UAS-in-RFP-Vkg) in the fat body, the stalk cells or the follicular cells using tissue-specific GAL4 drivers (Fig. S1). Results show that RFP-Vkg expressed in the fat body is deposited around the entire ovariole, including stalks (Fig. S1A-C) (Van De Bor et al., 2015), indicating that ColIV secreted into the hemolymph by fat body can reach all structures of the ovariole non-autonomously and non-specifically. However, unlike ColIV produced by the follicular cells, ColIV originating from the fat body does not show fibrillar organisation around the egg chambers (Fig. S1D-F′). Finally, ColIV produced by stalk or follicular cells is deposited in an autocrine manner and does not diffuse away from the producing cells (Fig. S1G-L).

Given the mosaic origin of stalk BM, we next wanted to determine the respective contribution of each tissue to the mature matrix. To achieve this, we carried out clonal analysis with two combined fluorescent markers (Vkg::GFP and RFP clonal marker) to differentiate ColIV-expressing cells from receiving ones (see Materials and Methods) (Haigo and Bilder, 2011). In this scheme, cells expressing both markers (GFP+RFP) are ColIV producers, while RFP-negative clones indicate receiving cells. When a RFP-negative clone encompasses the entire stalk (cells within a clone do not express Vkg::GFP), we could observe a GFP signal in the matrix of the stalk (Fig. 2H,H′), suggesting that this pool of Vkg::GFP has an external fat body origin. In addition, we still observed posterior and anterior thin layers of Vkg::GFP that separated stalk cells and the contacting egg chambers (Fig. 2H′-H″), indicating that this Vkg::GFP is not produced by stalk cells but rather by the follicular cells or the fat body. In an RFP-negative clone covering the entire stalk as well as adjacent follicular cells, we still observed GFP signal in the stalk matrix (Fig. 2I,I′), confirming the fat body origin of this signal. However, no signal was observed at the junction between stalk cells and follicular cells, (Fig. 2I-I‴), indicating that follicular cells, rather than fat cells, deposit ColIV on each side of the stalk that insulates the egg chambers. In a complementary experiment, we also observed that, in contrast to stalk and follicular cells, RFP-Vkg produced by the fat body is not present at the interface between the stalks and the egg chambers or in the matrix present in between the inner cells of the stalk (compare Fig. 2J-J″ and Fig. 2K-K″). These data indicate that the tissue of origin and/or mode of secretion (autonomous versus non-autonomous) determine ColIV organisation and pattern of deposition.

Altogether, results show that ColIV surrounding the stalks has three distinct origins. The fat body continuously produces ColIV, which is always available in the hemolymph and is deposited in the matrix around the stalk at all stages (Fig. 2L). Stalk cells produce ColIV from the initiation to the intercalation phase, which is deposited around and in between stalk cells. Finally, ColIV from follicular cells is deposited in an autocrine manner at the basal side of the follicular cells, leading to the closing and separation of the follicular cells from the stalk cells.

ColIV present around the ovariole has a mixed origin (Van De Bor et al., 2015; this work). In order to precisely determine the fraction of ColIV originating from the fat body, we used the same clonal analysis as above. Results show that ovarioles, which are not producing Vkg::GFP show a GFP signal (Fig. 3A,B), suggesting a fat body origin. In support of this view, no fibrils were observed around these egg chambers (Fig. 3A′-B′). We next compared the GFP signal present in the matrix of these ovarioles with the GFP signal present in the matrix of Vkg::GFP-producing ovarioles (Fig. 3C-J). Results show that, as the egg chambers grow, the fat body contribution decreases. Indeed, the fat body contributes up to 60% of the ColIV present in the matrix of stage 1 egg chambers, but only 10% of the ColIV of stage 10 egg chambers (Fig. 3K). In older egg chambers, synthesis of ColIV by follicular cells substitutes for the fat body ColIV. In contrast, fat body is a constant source of ColIV found around the stalk at different stages, with a contribution of 60% of the ColIV before the intercalation process and 40% at later stages (Fig. 3L).

These results show a dynamic contribution from different tissues to the matrix of follicles and stalks. They also reveal that the fat body is one of the main ColIV providers for the stalk matrix, suggesting a specific role of this distant tissue to stalk formation and/or homeostasis (see below).

The heterogeneous contribution of the fat body ColIV to the ovariole BM raises the question of the mechanisms controlling its distribution in these distant tissues. In particular, we wondered whether and how the contribution from one tissue might affect the contribution from another. A possible mechanism involves competition of ColIV from different origins for the same ECM receptors, leading one type of collagen to simply substitute another type when receptors become vacant (passive deposition). To start addressing the question of a potential redundancy of ColIV from different sources, we engineered new genetic tools allowing the expression of RFP-Vkg in the fat body (using Lpp-lexA and LexAop-in-RFP-Vkg) while generating loss-of-function clones of Vkg::GFP in follicular and stalk cells (using the GAL4/UAS system). We first verified that the pattern of Vkg::RFP was unchanged when driven by Lpp-lexA compared with Lpp- Gal4 (compare Fig. 4A and Fig. S1B). In this scheme, quantifying the RFP signal in Vkg::GFP loss-of-function clones allows the assessment of ColIV substitution by a different source (see Materials and Methods).

In control, RFP signal is homogeneous in the BM (Fig. 4B). When clones are located in stalk cells, the levels of Vkg::RFP inside (GFP-negative cells) and outside (GFP-positive cells) the clones are identical (Fig. 4C,D), suggesting that an equal amount of fat body ColIV accumulates in the stalk BM in presence or in absence of stalk ColIV. Similarly, in follicular cell clones of stage 5 egg chambers onwards, no difference could be detected inside and outside the clones (Fig. 4E). However, from stage 5 to 9 egg chambers, we observed that Vkg::RFP is preferentially deposited in the BM of wild-type cells (Fig. 4E,F). In this condition, the RFP signal was 50% lower in mutant compared with wild-type cells. Interestingly, this result suggests that efficient fat body ColIV deposition requires previous deposition of follicular ColIV.

To assess whether this coupling has any functional relevance, we next tested whether ColIV originating from the fat body is able to rescue the phenotype induced by loss of ColIV in the follicular cells. Loss of function of ColIV in follicle cells (using cb41-gal4 driver) leads to round eggs with an aspect ratio of 2.3, compared with 2.9 for control embryos (Fig. 4G,H) (see Materials and Methods). Interestingly, expressing Vkg::RFP in the fat body (Lpp-lexA, lexAop-in-RFP-Vkg) did not rescue the loss of ColIV from follicular cells (same aspect ratio of 2.2) (Fig. 4G,H). Altogether, these results suggest a sequential and coordinated deposition of ColIV from follicular cells and fat body, revealing an unexpected, regulated pattern of basement membrane assembly.

To test the functional role of ColIV in stalk formation, ColIV expression was silenced in stalk and follicular cells located at the pole (using c306-gal4; slbo-gal4 drivers; Fig. 5A,B). In this condition, the level of Vkg::GFP was dramatically reduced at the poles of egg chambers (Fig. 5D), leading 83% of the ovariole to develop a leaking phenotype, in which nurse cells and/or follicular cells (identified through their nuclear size) leaked out and were retained in the muscular sheath (Fig. 5C-E,G; Fig. S2A). We interpret this phenotype as being the result of matrix weakening at the pole, leading to rupture upon mechanical stress imposed by the muscular sheath. Additionally, 30% of the ovarioles displayed a fusion phenotype in which several egg chambers coalesced into one large fusion structure (Fig. 5C,C′,G; Fig. S2A) characterised by the absence of stalks. Finally, 30% of the ovarioles exhibited a pairing phenotype, in which adjacent egg chambers were juxtaposed (Fig. 5C,C′,F-F′,G; Fig. S2A). In some intermediate cases, isolated LamC-positive stalk cells could be found in between egg chambers, (Fig. 5F,F′). Interestingly, early ablation of stalk cells using expression of the pro-apoptotic gene hid induced the same syndrome, suggesting that impaired stalk structure/function accounts for these defects (Fig. S2A,B,C,C′,D,D′). These phenotypes lead to female sterility, demonstrating that proper ColIV assembly and BM integrity of stalk cells is essential to maintain ovariole structure and sustain female fertility.

We next characterised the role of ColIV in stalk morphogenesis, by silencing ColIV production specifically in stalk cells using both VkgRNAi and Cg25cRNAi constructs (designated ColIV RNAi) in a Vkg::GFP background. RNAi constructs were targeted using cas-gal4 or bab-gal4, tub-gal80^ts drivers (Fig. S1G-J) (see Materials and Methods). In both conditions, ovarioles seemed to develop normally and females were fertile. However, we observed a significant reduction of Vkg::GFP signal around the stalks. At stage II, the signal decreased at least by 50% compared with the control and by 20-30% at later stages (Fig. 6A). Analysis of matrix structure through EM showed a massive decrease in matrix thickness from stage II to V compared with the control (Fig. S3A-E). Hence, reducing ColIV deposition from stalk cells alters their matrix architecture at all stages.

We next tested the role of ColIV in stalk cell differentiation by quantifying the number of Cas- and LamC-positive cells. With both drivers, the average number of cells is identical to the control at stage IV, suggesting that stalk cells differentiate normally (Fig. 6B; data not shown). However, we observed a delay of about 12 h in the intercalation process, as well as in the expression of Cas and LamC markers (Fig. 6B-E). Hence, proper ColIV assembly is important for normal stalk development and morphogenesis.

In addition to these early defects, we also found that loss of function of ColIV in stalk cells (cas-gal4; ColIV^RNAi) leads to defective stalk structure. First, cells do not maintain their alignment and form two clusters on each side of the stalk (Fig. 6D″-D′′′), suggesting that ColIV is important to promote cell rearrangement, adhesion between stalk cells through the intercalating ECM and proper alignment within the stalks. Second, lack of ColIV in both stalk and follicular cells (bab-gal4, tub-gal80ts; ColIV^RNAi) precludes anterior and posterior closing. In this condition, 30% of ovarioles show a leaking phenotype characterised by the presence of nurse cells and/or follicular cells within the stalk (Fig. 6E′), suggesting that ColIV deposited by follicular cells at the boundary between egg chambers and stalks prevents cells coalescing from the two structures. Finally, 20% of ovarioles display posterior detachment of stalks from the next egg chamber, inducing dislocation of the ovariole. EM analysis further revealed that the morphology of mutant terminal stalk cells was dramatically affected, with the stalk terminal cell not elongating properly compared with the control (Fig. S3B-E). Altered cell elongation could possibly modify stalk-egg chamber adhesion and favour stalk detachment from the egg chamber (Fig. 6E″). Altogether, our functional data show that ColIV is important for stalks to acquire the appropriate matrix architecture, organisation and adhesive properties that are important for the whole ovariole structure and fertility.

We next determined the role of ColIV produced by the fat body, using fat cell-specific driver (lpp-gal4) to silence ColIV expression in a Vkg::GFP background. Although we could not completely abolish ColIV expression from the fat body, we nonetheless observed a strong decrease in Vkg::GFP around the stalk compared with the control (Fig. 7A). Cas and LamC cells differentiate as in the control, albeit with a delay for the Cas marker (Fig. 7B). Interestingly, in 30% of ovariole, we observed that the intercalation process is impaired (Fig. 7C-C″′,D-D″′). Cells within the stalk have lost their alignment, as for epithelial ColIV loss-of-function conditions, showing that ColIV originating from fat body is also playing an important role in stalk morphogenesis.

Integrin receptors bind ECM and associate with the cytoskeleton to mechanically connect extracellular and intracellular cell structures. The role of integrins in stalk morphogenesis has not been established, so we first investigated their pattern of expression during ovariole development. Integrins are αβ heterodimers sharing a common β subunit (βPS), Myospheroid (Mys), which we found to be strongly expressed in stalk cells from stages I to V (Fig. 8A). Detailed analysis reveals that βPS starts to accumulate at the basal side of stalk cells during the convergence and intercalation phases (Fig. 8B-E′). At stage IV, βPS is also visible in between stalk cells as for ColIV (Fig. 8D,D′), suggesting that integrins might be important for stalk formation and cell-cell interactions within the stalk.

Interestingly, we found that absence of βPS induced an egg chamber fusion phenotype, as observed in ColIV mutants (Fig. 8F, Fig. S2A). Furthermore, stalk morphogenesis was delayed, the intercalation process was affected and structure of the mature stalk was impaired (Fig. 8G-G″). The majority of stalks are formed by two juxtaposed rows of cells with ColIV being mislocalised and found in between cells. Altogether, these phenotypes indicate a clear role for integrins in the formation and function of stalks, and suggest that absence of integrins may alter the ColIV-dependent mechanical properties of stalk cells.

To further assess whether integrins function is linked to ColIV, we checked the pattern of integrin expression in the absence of ColIV. In this condition, we observed a dramatic decrease in βPS signal from stage I to III (Fig. S4, compare A-D′ with Fig. 8B-E′), suggesting that ColIV is important for timely βPS localisation and/or stability, which in turn affects cell adhesion and intercalation. A possible pathway through which integrin expression could be regulated involves a mechanical coupling between the cell-ECM and the nucleus through LamC and the LINC complex (Alam et al., 2016; Donohoe et al., 2018).

The proper assembly of basement membranes is essential for morphogenesis, organ function and physiology (Sekiguchi and Yamada, 2018). BMs are very diverse in their structure and composition, and show dynamic remodelling throughout development. However, it is still poorly understood how such diversity is built and how it affects the many functions of BMs. In this study, we developed specific tagged forms of Drosophila ColIV to unravel the composition, origin and functional role of a dynamic BM involved in cell intercalation during oogenesis. Results revealed the identity of the tissues contributing ColIV to developing stalks, with both autonomous (stalk and follicle cells) and non-autonomous (fat body) origins. By determining the relative contribution from each individual tissue (ColIV tissue ratios), we could reveal a mode of sequential deposition, with assembly of exogenous ColIV from fat body depending on prior ColIV from follicular cells. These results reveal a regulated assembly of stalk ColIV, rather than a passive building based on simple ColIV availability. This mode of assembly also suggests the existence of collagens with different biochemical properties, depending on their source, and the need to mix these forms to build BM with appropriate bio-mechanical features. Finally, we found that ColIV of different origins and integrins are all essential to build functional stalks through cell intercalation, with key functional roles in organ morphogenesis and integrity.

How is basement membrane assembled in developing tissues? In the simplest scenario, BM components would be produced by diverse sources and would reach vacant cell surface receptors and self-assemble into a matrix network. In this model, integration of ColIV molecules would just depend on their local availability, i.e. ColIV from different origins would be strictly interchangeable and would contribute equally to global and local ColIV levels. However, our results provide evidence that the cellular/organ origin of ColIV determines where the molecules can go and how they are assembled. In this emerging scenario, ColIV molecules are not all equivalent, and their assembly is regulated according to their origin, biochemical properties and/or range (autonomous versus non-autonomous). We hypothesise that discrimination among ColIV molecules could be determined by specific modifications or modes of delivery.

To support this view, our tracing experiments using tagged forms of ColIV showed that the BM of the stalk is assembled from three sources: the fat body, stalk and follicle cells (Fig. 2). Results show that ColIV made from these tissues have different diffusion capabilities: while stalk and follicle ColIV are integrated cell autonomously, fat body ColIV is diffusible, reaching distant tissues and being incorporated non-autonomously. Our rescue experiments revealed another interesting feature, namely that the loss of ColIV from one tissue cannot simply be replaced by ColIV from the other two tissue sources (Fig. 4), ruling out a model in which ColIVs would be interchangeable. Indeed, we show that fat body ColIV is able to integrate ovariole BM only if follicular ColIV is already present, suggesting a sequential deposition of ColIV in the process of BM assembly. A possible mechanism explaining this behaviour involves the level of integrin receptors present in follicle cells. This view is supported by our data showing that the absence of follicular ColIV causes a reduction of βPS signal at the cell surface (Fig. 8), hence reducing the ability of this tissue to bind ColIV. Finally, our work provides direct evidence that fat body ColIV and follicular ColIV have distinct organisation into the BM. Indeed, we demonstrate that fat body ColIV does not form fibrils in the BM, contrary to follicular ColIV (Fig. 3). Therefore, fibrillar ColIV already present around the egg chamber is not sufficient to force fat body ColIV to form fibrils. In agreement with previous work (Isabella and Horne-Badovinac, 2016), we propose that the follicular cells specifically form ColIV fibrils, with the structure difference between follicular and fat body ColIV being possibly due to modifications and/or mode of secretion.

Animal and organ development require a number of well-characterised morphogenetic movements. Among these, cell intercalation is a universal cellular remodelling process leading to the lengthening of a tissue along one axis without change in overall cell number. For example, cell intercalation plays a crucial role in convergent extension during gastrulation, dorsal closure, wound healing and other processes (Gettings et al., 2010; Razzell et al., 2014; Roszko et al., 2009). Most studies have focused on the role of adhesive changes, cytoskeleton or planar cell polarity in the process (Walck-Shannon and Hardin, 2014; Xie et al., 2018). In Xenopus, Fibronectin binding to Integrin has been shown to regulate cell adhesion and cell polarity involved in cellular convergent extension (Huang and Winklbauer, 2018). However, the exact role of the ECM itself in cell intercalation is poorly understood, in part because of the lack of specific tools to visualize matrix components in vivo during morphogenesis.

Using the stalk cell model coupled with the engineering of new-tagged BM components, we have been able to demonstrate the role of ColIV during cell intercalation. Stalks assemble a particular BM, making an alveolar structure that recruits Integrins and is tightly linked to follicle cells (Figs 1 and 8). Furthermore, the BM of stalks shows a dramatic thickening throughout its development. Blocking ColIV function is sufficient to disrupt the intercalation process, leading to misshapen stalks, suggesting that inter-stalk cell BM is essential to drive intercalation and maintain the string-like structure of the stalks; hence, their function as egg chamber linkers. A possible mechanism driving cell intercalation could involve the formation of a stiffness pattern, which promotes cell migration and reorganization through durotaxis (the migration of cells up a gradient of rigidity). Together with the thickening of BM, stalks express LamC, a marker of cells undergoing mechanical constraint (Donohoe et al., 2018; González-Cruz et al., 2018; Lammerding et al., 2004; Swift and Discher, 2014). It is striking that absence of ColIV delays the appearance of LamC as well as Cas markers in stalk cells, which is reminiscent of situations in which modification of matrix stiffness affects cell differentiation (Buxboim et al., 2010; Daley and Yamada, 2013; Engler et al., 2006; Handorf et al., 2015; Smith et al., 2018). Consistently, loss of ColIV or Integrins leads to dramatic leakage of stalks and egg chambers. Altogether, these results suggest a crucial role of stalk BM for building a strong, mechanically resistant structure necessary for ovariole integrity and female fertility. Further work will help establish the ColIV delivery code underlying the assembly of specific BMs and aid better understanding of the role of the extracellular matrix in complex tissue morphogenesis.

Flies were raised on standard cornmeal-agar medium at 25°C. The following strains were used: W1118 (Bloomington Stock Center), Vkg::GFP (Morin et al., 2001), Ubi-nls-RFP ,Vkg ::GFP and FRT40A/FRT40A ; T155-gal4, UAS-FLP (gifts from Adam J. Isabella and Sally Horne-Badovinac, University of Chicago, IL, USA), Cas-gal4, Lpp-gal4, Bab-gal4, cb41-gal4, Vkg::GFP, Vkg^RNAiKK; Cg25c^RNAiGD (Bloomington Stock Center), and Vkg^RNAiKK, Cg25c^RNAiGD, UAS-Mys^RNAiKK (VDRC Stock Center). Fly lines generated through this work were: C306gal4; slbo-gal4, UAS-GFP, Tub-gal80ts; Bab-gal4, Vkg^RNAiKK; Cg25c^RNAiGD, UAS-in-RFP-Vkg^4M, Vkg::GFP, yw, hsFLP; LpplexA/Act<<gal4, UASGFP; UAS Vkg^RNAiGD/lexAop-in-RFP-Vkg^M1 and Vkg::GFP;UAS-Hid, tub-gal80^ts/cas-gal4. Unless otherwise noted, 2-day-old flies were analysed. bab-gal4>VkgRNAi is lethal at early stages, hence flies were raised at restrictive temperature (18°C) and switched at permissive temperature (25°C) after pupal eclosion.

Ovaries were dissected in 1×PBS and fixed in 1×PBS 3.7% formaldehyde. Samples were blocked in 1×PBS containing 0.3% Triton X-100 and 2% BSA for 30 min prior to incubation with antibodies and then mounted in Vectashield-DAPI (Vector Laboratories). The following primary antibodies were used: rabbit anti-Castor (1:500; a gift from D. Montell, University of California Santa Barbara, USA), mouse-Integrin βPS (myospheroid) CF6G11(DSHB) (1:100), mouse anti-LaminC LC28.26 (DSHB; 1:100) and rabbit anti-Viking (1:500; this work). Alexa Fluor 546- and Alexa Fluor 488-conjugated secondary antibodies (Molecular Probes, Jackson ImmunoResearch, 1:500) were used. Imaging was performed on a confocal microscope Zeiss LSM 780.

vkg RNA probes were synthesized by in vitro transcription. Samples were fixed in 1×PBS 3.7% formaldehyde for 10 min and rinsed twice in PBT (1×PBS 0.1% Tween20). Samples were subsequently pre-hybridised for 10 min in 1:1 hybridisation solution:PBT (50% deionized formamide, 5×SSC, 0.1%Tween 20, 50 µg/ml heparin and 100 µg/ml herring sperm DNA) followed by 1 h in hybridisation solution at 65°C. Hybridisation was carried out at 65°C. After three 20 min washed in hybridisation solution and one wash in 1:1 hybridisation solution:PBTw (1×PBS, 0.1% Tween 20) at 65°C, samples were washed three times at room temperature in PBTw and blocked for 30 min in PBTw 2% BSA. Samples were then incubated with anti-digoxigenin antibodies (Fab fragment, Roche, 11214667001, 1:1000) at room temperature. Samples were washed three times for 20 min in PBTw and incubated with a FITC-coupled anti-rabbit secondary antibody (1:500, Jackson ImmunoResearch, Molecular Probes, 111-095-144) before proceeding with the TSA-based detection of viking mRNA.

Polyclonal antibodies were raised against Viking using the Eurogentec DoubleX protocol. The following peptides were used for rabbit immunization: NPSGSLKTDGNYRA (amino acids 1764-1777) and NTRQYNRRPREDTTAP (amino acids 1925-1940).

The GFP insertion site of the Vkg::GFP flies was verified by sequencing and the same insertion was used for the UAS-in-RFP-Viking (see below). For UAS-in-RFP-Viking transgenic lines, the full-length coding sequence of Viking, in which the RFP was inserted after nucleotide 93 was cloned into pUAST-C5 (EcoRI-NheI). Full-length constructs (6587 nucleotides, NM_001273142.1) were synthesised (Epoch Life Science) allowing a fusion with the RFP without the insertion of restriction site or additional amino acids.

A Lipophorin-Lexa (Lpp-LexA):Lipophorin:Gal4 construct (kindly provided by Suzanne Eaton) was used to amplify the 5.5 kbp promoter and the 3 kbp 3′UTR sequence of lipophorin. The fragments were cloned into pBPnlsLexA (Addgene). A 5.5 kpb promoter sequence was inserted into the attR1 gateway site, whereas the 3kpb 3′UTR sequence was inserted in the HindIII restriction site. Injections were made by BestGene. LexAop-in RFP-Vkg : the in-RFP-Viking sequence was amplified and cloned into the pJFRC-19-13XLexAop2 as an XhoI-XbaI fragment. Injection were made by BestGene.

Ovaries were dissected in 1×PBS and directly transferred in fixative for 1 h at room temperature (1.5% glutaraldehyde in 0.075 M cacodylate buffer). After embedding in resin (Epon mix), longitudinal ultra-thin sections of ovaries were prepared following standard procedures. Sections were stained in uranyl acetate and lead citrate following standard protocols. Ultrathin sections were observed with a Phillips CM120 electron microscope equipped with a Gatan digital CCD camera at the Centre Technologique des Microstructures (UCBL, Villeurbanne, France; Lyon Bio-image platform), Lyon Multiscale Imaging Center (LyMIC). Two ovarioles per genotype were analysed.

We measured the matrix thickness on both electron microscopy (EM) and immunofluorescence (IF) pictures using Fiji. The average thickness was measured on EM pictures from different regions and slices (n>10), perpendicular to the cell membrane. To measure the matrix thickness on an IF picture, a line crossing perpendicularly the matrix was drawn and the intensity along the line was plotted. The thickness of the intensity gradient was taken as a value for the thickness of the matrix. To avoid background noise, we determined that the extremities of the picture correspond to half of the max intensity.

Signal quantification was carried out using Fiji. A line of two pixels depth, which is much smaller than the matrix thickness, was drawn inside the stalk matrix, and the intensity along the line was plotted. The values along the line were average. For each stage (n>10).

For Ubi-nls-RFP ,Vkg::GFP, FRT40A/FRT40A; T155-gal4, UAS-FLP, heat shocks were induced at 37°C for 20 min on 12 h-old flies. Flies were dissected 36 h later. Only RFP-positive cells express Vkg::GFP.

For hsFLP; LpplexA, Act<<gal4, UASGFP; Vkg::GFP, UAS Vkgⁱ, lexAopVkgRFP, heat shocks were induced at 37°C for 20 min on 12 h-old flies. Flies were dissected 36 h later.

Eggs from 4-day-old flies raised at 29°C were collected after 12 h of egg laying. The aspect ratio (AP/DV length) of eggs was measured from pictures using Fiji. n=6 collections of at least 100 eggs.

We thank S. Horne-Badovinac and D. Montell for reagents and fruitful discussions; all members of the laboratory for fruitful discussions and comments on the manuscript; the Bloomington Drosophila Stock Center, Vienna Drosophila RNAi Center (VDRC) for providing Drosophila fly lines; and the iBV PRISM imaging platform, the Centre Technologique des Microstructures (Lyon) for providing state of the art imaging resources.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.195511.reviewer-comments.pdf