Constriction imposed by basement membrane regulates developmental cell migration

The basement membrane (BM) is a specialized extracellular matrix (ECM), which underlies or encases developing tissues. Mechanical properties of encasing BMs have been shown to profoundly influence the shaping of associated tissues. Here, we use the migration of the border cells (BCs) of the Drosophila egg chamber to unravel a new role of encasing BMs in cell migration. BCs move between a group of cells, the nurse cells (NCs), that are enclosed by a monolayer of follicle cells (FCs), which is, in turn, surrounded by a BM, the follicle BM. We show that increasing or reducing the stiffness of the follicle BM, by altering laminins or type IV collagen levels, conversely affects BC migration speed and alters migration mode and dynamics. Follicle BM stiffness also controls pairwise NC and FC cortical tension. We propose that constraints imposed by the follicle BM influence NC and FC cortical tension, which, in turn, regulate BC migration. Encasing BMs emerge as key players in the regulation of collective cell migration during morphogenesis.


Introduction
The basement membrane (BM) is a dense, sheet-like form of an extracellular matrix (ECM) that underlies the basal side of epithelial and endothelial tissues and enwraps most developing organs [1,2]. BMs are highly conserved across evolution and are composed of a core set of proteins, including the secreted glycoproteins laminin, type IV collagen (Coll IV), entactin/nidogen, and the heparan sulfate proteoglycan Perlecan [3]. Originally viewed as a static support ECM, BMs have recently been proven to act as active regulators of tissue shape and homeostasis.
Encasing BMs regulate the morphogenesis of the tissues they enwrap both molecularly, by modulating signaling to cells through surface receptors, and mechanically, by imposing  Furthermore, our results show that follicle BM stiffness controls pairwise FC and NC cortical tension. Finally, we show that a direct manipulation of FC tension affects inversely BC migration speed. Based on these results, we propose that constraints imposed by the follicle BM can affect FCs and NCs cortical tension, which, in turn, influences BC migration. Our results unravel a new role for the mechanical properties of the BMs enclosing developing organs, the regulation of the intercellular forces controlling collective cell migration. In addition, this work provides new insights towards our understanding of the remarkable and assorted ways whereby BMs regulate cell migration during morphogenesis.

The stiffness of the follicle BM influences BC migration
To test the role of the follicle BM on BC migration, we reduced the levels of 2 of its main components, Laminins or Coll IV. Laminins are large heterotrimeric glycoproteins that contain 1 copy each of an α, β, and γ chain. Drosophila contains 2 different Laminin trimers, each composed of one of the two α chains, encoded by the wing blister (wb) and Laminin A (LanA) genes, and the shared β and γ chains, encoded by the Laminin B1 (LanB1) and Laminin B2 (LanB2) genes, respectively [28][29][30][31]. To reduce Laminin levels in follicles, we either expressed a LanB1 RNAi in all FCs by means of the traffic jam (tj) Gal4 line (tj>LanB1 RNAi ) or used a hypomorphic viable condition for the locus, LanB1 28a /l (2)K05404 (hereafter referred to as LanB1 hyp , Table 1), as both have been shown to drastically reduce LanB1 levels up to S8 [11,32]. We have previously shown by atomic force microscopy measurements that follicle BM stiffness was significantly diminished in these mutant conditions, from the germarium up to S8 [32]. As here we found that LanB1 levels in the mutant follicles were kept constant and significantly inferior to controls until the end of oogenesis (S1 Fig), we expected that the reduction in BM stiffness would sustain up to S9, when BCs are migrating. To confirm this, we utilized atomic force microscopy and measured the apparent elastic modulus, K, as a proxy for BM stiffness [32]. Force-distance curves were analyzed for an indentation depth of 200 nm, because egg chambers are composite materials with several layers, and structures away from the surface farther than approximately 1/10 of the indentation depth do not contribute to K [33]. As expected, we found that the K values of laminin-depleted follicles (n = 16 from 7 different egg chambers) were significantly smaller compared with controls (n = 12 from 6 different egg chambers, S2A and S2B Fig).
Another key component of BMs is type IV collagen, a unique member of the collagen superfamily, which is composed of different a1 chains that form distinct heterotrimers. The Drosophila Coll type IV trimer consists of two α1 chains and one α2 chain, encoded by the genes Collagen at 25C (Cg25C) and Viking (Vkg), respectively [34][35][36]. To decrease Coll IV levels, we expressed a Vkg RNAi in all FCs using the tjGal4 driver (tj>Coll IV RNAi , Table 1), an approach previously shown to efficiently reduce both Coll IV levels and BM stiffness up to S8 [7,37]. In addition, as it was the case for Laminins, we found that the K values of tj>Coll IV RNAi follicles were kept significantly smaller than those of controls at S9 (S2A and S2B Fig).
These results demonstrated that the approach here used to decrease Laminins and Col IV levels in follicles caused a reduction in BM stiffness at S9, which is when BCs are migrating. This allowed us to test the effects of lessening BM stiffness in BC migration.
To quantify BC migration in the different conditions in which the BM stiffness was altered, we performed live imaging of control and Laminin or Coll IV-depleted egg chambers. To visualize BCs, we generated transgenic flies that expressed the green fluorescent protein (GFP) driven by an enhancer of the torso-like (tsl) gene, which is active in BCs and in a group of posterior FCs (tslGFP, see Materials and methods, [38]). Quantification of BC migration speed from live imaging analysis revealed that BCs from tslGFP; tj>LanB1 RNAi (n = 7), tslGFP; LanB1 hyp (n = 5), or tslGFP; tj>Vkg RNAi (n = 5) egg chambers moved faster than BCs from control follicles (tslGFP; tj>Gal4, n = 8) (Fig 1D-1H, S1 Movie, see Materials and methods). In all experimental egg chambers, BCs completed their migration and reached the oocyte as in controls.
BC migration to the oocyte has been divided in 2 phases. In an early phase-from detachment until halfway to the oocyte-movement is more streamlined and sliding. In the late phase-from midway to contacting the oocyte-movement is slower with clusters rotating in place [39,40]. We found that reducing Laminins or Col IV levels increased the migration speed of both phases (Fig 1I and 1J). Since we found that BC migration was similarly increased when reducing either Laminins or Coll IV levels and as Laminins were required for proper Coll IV deposition in the ovary [32], we concentrated our analysis on laminin-depleted egg chambers.
These results suggest the existence of a direct correlation between BM stiffness and BC migration speed. However, the general FC driver tjGal4 directs expression in all FCs, including BCs. Even though Laminins are not expressed in BCs during their migration [41] (S1A Fig), to completely rule out the possibility that the faster BC migration phenotype observed in tj>LanB1 RNAi follicles was due to a possible requirement of this BM component in BCs, we analyzed the effects of expressing a LanB1 RNAi specifically in BCs. To achieve efficient RNAi knockdown in BCs, we combined the BC Gal4 line slboGal4 [24] with 2 other Gal4 lines, C306Gal4 and torso-likeGal4 (tslGal4), which are expressed before BCs detach from the FE ( [24,38], Table 1). As expected, we could not detect any LanB1 signal in the BCs of these follicles triple>LanB1 RNAi (S3A and S3A' Fig). In addition, live imaging of egg chambers coexpressing the LanB1 RNAi and a GFP, to visualize the BCs, showed that BCs from the tri-ple>GFP; LanB1 RNAi egg chambers (n = 6) moved at similar speed than controls (triple>GFP; n = 5, S3 Fig and S2 Movie). This result indicated that Laminins do not have an essential function in BCs, a finding that is consistent with the reported absence of Laminins expression in BCs during their migration [41].
To further test the idea that the stiffness of the BM affects BC migration, we analyzed the effect of increasing BM stiffness. To this end, we used the tjGal4 to overexpress an mCherry tagged version of EHBP1(EHBP1mCh) in all FCs, an approach known to increase Coll IV fibril deposition and BM stiffness ( [8], Table 1). In fact, in agreement with those studies, we found that the K values of tj>EHBP1mCh follicles obtained by AFM were significantly higher than those of controls (S2B Fig). Live time-lapse imaging showed that the overexpression of EHBP1mCh in all FCs delayed BC migration (tslGFP; tj> EHBP1mCh, S4A-S4B' Fig, S3 Movie, n = 6). However, as mentioned above, the general driver tjGal4 drives expression in all FCs, including BCs (S4 Fig). In this context, the BC migration defects observed in tslGFP; tj>EHBP1mCh follicles could also be due to an increase in Coll IV levels in BCs, rather than just a rise in follicle BM stiffness [37]. Thus, to directly assess the impact of BM stiffness in BC migration, we used the mirrorGal4 (mirrGal4) driver, whose expression is restricted to a central region of the FE [42]. A previous study showed that by S8, the BM at the central region was stiffer than that at terminal regions [37]. Here, we found that this difference persisted into S9 egg chambers (S2D Fig). In addition, we found that this difference was higher when EHBP1mCh was expressed in the central region of the follicle using the mirrGal4 line compared to controls (mirr>EHBP1mCh, S2D and S2E Fig). This result demonstrated, in agreement with previous results [37], that overexpressing EHBP1mCh locally increased BM stiffness. Next, we analyzed BC migration in these experimental conditions and found that, while BCs migrated normally anterior to the mirror region in the tslGFP; mirr>EHBP1mCh egg chambers (Fig 2A-2C, n = 8), migration speed was reduced once BCs reached the mirror region ( Fig 2D, n = 8, S4 Movie). Furthermore, while migration was always completed in controls follicles (n = 6), migration in the mirror region was halted in 20% of the experimental cases analyzed (Fig 2E, n = 9).

Effect of the BM stiffness on cellular protrusions dynamics and migration mode
As mentioned in the Introduction, BCs extend actin-rich protrusions while migrating between the NCs. As these protrusions extend between NCs, it is logical to think that any constraint(s) could influence the formation and dynamics of these protrusions. Thus, we next analyzed by live imaging the behavior of BC cellular protrusions in control and experimental clusters throughout the migratory process. In agreement with previous results [39,43], we found that BC protrusions in control egg chambers showed an overall front bias and that the protrusion at the front of the cluster was longer than protrusions from other areas of the cluster (n = 7, Fig 3A, 3E, and 3K and S1 Movie). This is maintained when manipulating BM stiffness (Fig 3E  and 3F). However, BCs from laminin-depleted egg chambers (tslGFP; tj>LanB1 RNAi ) showed a significant increase in the number of protrusions extended laterally (n = 6, Fig 3B and 3E). Moreover, BCs from tslGFP; mirr>EHBP1mCh egg chambers showed a decrease in the number of these lateral protrusions (n = 8) compared with controls (n = 6), specifically in the mirr region (Fig 3C, 3D, and 3F). Previous studies have shown that reducing E-cadherin (Ecad) or myosin function reduces directional persistence as a consequence of generating ectopic long side protrusions, resembling the large leading protrusion that initiate and guide BC migration [44,45]. In contrast, here we show that BCs from laminin-depleted egg chambers migrate faster towards the oocyte, despite showing an increase in the number of lateral protrusions. However, we found that the length of the ectopic side protrusions and that from the front and rear of clusters from laminin-depleted egg chambers was similar to that of controls ( Fig 3K). This could explain why BCs from laminin-depleted egg chambers do not loose directionality besides having increase ectopic lateral protrusions.
As mentioned above, BCs show efficient forward movement during the early phase and increased tumbling during the late phase [39]. Next, we analyzed whether modifying the BM would affect the rotation mode. Using manual tracking (see Material and methods), we found that BC clusters from laminin-depleted egg chambers rotated slower than those from controls (n = 8, Fig 3G and 3G', S5 Movie). In addition, we found that while control BC clusters always rotated (n = 8, Fig 3K), BC clusters in laminin-depleted egg chambers did not rotate in 42.8% of the cases analyzed (n = 7, Fig 3K). In contrast, the slow-moving BC clusters of EHBP1  Quantification of BC cluster rotation. The statistical significance of differences was assessed with a t test, * P overexpressing follicles were still able to rotate (Fig 3I-3J' and 3L, n = 10, S6 Movie). This result demonstrates that the stiffness of the follicle BM could alter the mode of BC migration.

Direct correlation between BM stiffness and NC cortical tension
The current model proposes that NCs exert compressive forces over BCs that are perpendicular to the migration pathway and that mechanically influence their migration [26]. In this scenario, and since NCs are externally constrained by both the FCs and the follicle BM, the increase in migration speed we observe upon reduction of BM stiffness may arise from a drop in NC cortical tension. To test this possibility, we first analyzed to what extent stiffness of the BM could influence NC cortical tension. In wild-type egg chambers, BM stiffness over the NCs has been shown to gradually decrease from early S9 to S10 in the anterior region of the egg chamber, while a group of around 50 anterior FCs, the so-called stretched FCs (StFCs), progressively flatten in an anterior to posterior wave [46][47][48]. To test whether these changes in BM stiffness affected NC cortical tension, we performed laser ablation experiments in NCs under the StFCs throughout the flattening process of these cells (S5A-S5F Fig, [49]; see Materials and methods, S7 Movie). All cuts were made perpendicular to the AP axis (S5 Fig) and the behavior of cell membranes, visualized with Resille-GFP [50], was monitored up to 10 s after ablation (S7 Movie). We found that, although some differences in BM stiffness were reported for mid and late S9 egg chambers relative to early S9 [46], NC cortical tension did not change significantly throughout these stages, which is when BCs migrate (n = 16, S5A-S5F Fig). However, we did detect a significant decrease in NC cortical tension at S10 (n = 15, S5F Fig), when anterior FCs were totally flattened, BC migration was completed and BM stiffness was much lower than that found at earlier stages [46]. This result suggested that reducing BM stiffness below a certain threshold diminished NC cortical tension. Based on this result, we next tested whether the decrease in BM stiffness detected in the laminin-depleted egg chambers could affect NC cortical tension. BM stiffness over the NCs was found to be homogenous along the A/P axis in early S9 control egg chambers [46,51]. In line with these results, we found, by laser ablation experiments, that NC cortical tension was also homogenous along the A/P axis at this stage (S5G Fig). Thus, we decided to perform the next set of laser ablation experiments in NCs located in the central region of early S9 follicles, ahead of migrating BCs, as this will be the environment BCs will encounter during their movement towards the oocyte (Fig 4A). We found that the cortical tension of NCs in laminin-depleted follicles (Resille-GFP-tj>LanB1 RNAi , n = 16) was lower than in controls (Resille-GFP-tjGal, n = 15, Fig 4B-4D, S8 Movie). Accordingly, the levels of the phosphorylated form of the regulatory chain of non-muscle Myosin II (encoded by the gene spaghetti squash, pSqh) at the NC cortex were lower in laminin-depleted follicles (n = 10) compared with controls (n = 10) (S6A, S6B and S6E Fig). An increase in NC contraction was shown to increase the levels of pSqh at the periphery of the BC cluster [26]. Here, we found that pSqh levels were reduced in the BCs of laminin-depleted egg chambers (n = 9) compared with controls (n = 8) (S6G, S6H and S6K Fig).
Altogether, our results support a model in which stiffness of the BM impacts on compressive forces exerted by NCs over the BCs, which is important for their movement. To further test this idea, we analyzed whether increasing BM stiffness could augment NC cortical tension. To do this, we expressed EHBP1mCh in the central region of the FE and measured cortical tension by laser ablation and pSqh levels. Indeed, we found that the cortical tension and pSqh value < 0.05. Horizontal and vertical lines indicate mean and SD, respectively. Scale bars in A-D and G-J, 10 μm. The raw data underlying panels E, F and K-M are available in S1 Data. BC, border cell; BM, basement membrane.
https://doi.org/10.1371/journal.pbio.3002172.g003 levels at the cortex of NCs located under the mirror region in mirr>EHBP1mCh egg chambers (n = 14 and n = 7) was higher than that found in control (n = 13 and n = 7, Figs 4F-4H, S6C, S6D and S6F), while no statistically significant difference was observed between experimental (n = 14 and n = 7) and control (n = 13 and n = 7) NCs located anterior to the mirror region (Figs 4I-4K and S6F). Furthermore, we found that the levels of pSqh were higher in BCs from mirr>EHBP1mCh (n = 8, S6I, S6J, and S6L Fig), being specifically significant at the lateral cluster periphery, where contraction forces from NCs could be higher in this experimental situation.
Finally, to further test our hypothesis that BM stiffness could influence BC migration via regulating NC cortical tension, we analyzed whether a direct reduction of NC cortical tension would speed up BC migration. In order to do this, we expressed an RNAi against the Abelson interacting protein (Abi) in NCs, as its expression in FCs has been shown to cause a reduction in their cortical tension [52]. We found that the expression of an Abi RNAi RNAi in NCs (n = 5, nos>Abi RNAi ) increased BC migration speed compared with controls (n = 5) (S7 Fig and S9 Movie). In addition, we found that rotation of the BC cluster was affected. Thus, while BCs rotated in 100% of control nosGal4 egg chambers (n = 5), rotation was hampered in 20% of nos>AbiRNAi egg chambers analyzed (n = 5, S7D Fig). This result supports our idea that less tension in NCs can prevent BC migratory rotation, which we propose contributes to increase speed.

Contribution of cytoplasmic pressure to BC migration
Cell mechanical properties are not only dictated by cortical tension but also by cytoplasmic pressure. Collagenase treatment decreases NC cytoplasmic pressure, leading to suggest that BM stiffness could influence NC cytoplasmic pressure [51]. In this scenario, the effects on BC migration we observed when reducing laminin levels could also be due to alterations in NC cytoplasmic pressure. To test this possibility, we measured NC cytoplasmic pressure in early S9 control and tj>LanB1 RNAi egg chambers. The extent of curvature of the posterior-facing NC membrane (the radius of the circle fitting this membrane) has been proposed to reflect differences in the cytoplasmic pressure of adjacent NCs [51]. Using this method, a gradient of cytoplasmic pressure along the A/P axis, with pressure decreasing from anterior to posterior, was reported. Using this approach, we found that this gradient was maintained in both laminin-depleted and EHBP1mCh overexpressing follicles, as we did not detect any significant difference between the curvature of NCs from tj>LanB1 RNAi or mirr>EHBP1mCh follicles and that of controls (S8A-S8F Fig, n > 10 per position). This is consistent with previous results showing that altering the BM stiffness, by collagenase treatment, did not affect the gradient of NC cytoplasmic pressure [51]. However, even though absolute values of pressure were found lower in collagenase-treated follicles compared to untreated follicles (Lamire and colleagues) [51], we did not detect any significant difference between the curvature of NCs from tj>LanB1 RNAi or mirr>EHBP1mCh follicles and that of controls (S8E and S8F Fig). In any case, to more directly address a possible role of NC cytoplasmic pressure on BC migration, we analyzed BC migration in dicephalic (dic 1 ) and kelch (Kel ED1 ) mutant egg chambers, as they were shown to display reduced and increased NC cytoplasmic pressure, respectively [51]. Our mirr expressing region of S9 control (mirrGal4) and mirr>EHBP1mCh egg chambers expressing Resille-GFP before and after NC bonds are ablated. Blue bars indicate points of ablation. (G, J) Quantification over time of initial velocity of vertex displacement and (H, K) vertex displacement over time of the indicated ablated bonds. The statistical significance of differences was assessed with a t test, * P value < 0.05 and *** P value < 0.001. Horizontal and vertical lines indicate mean and SD, respectively. Scale bars in B, F, and I, 10 μm. The raw data underlying panels C, D, G, H, J, and K are available in S1 Data. BC, border cell; BM, basement membrane; FC, follicle cell; GFP, green fluorescent protein; NC, nurse cell. https://doi.org/10.1371/journal.pbio.3002172.g004 in vivo analysis showed that the speed at which BCs moved in dic 1 and Kel ED1 follicles was similar to that found in controls (S8G- S8J Fig and S10 Movie). All together, these results suggest that NC cytoplasmic pressure on its own does not seem to strongly contribute to the forces regulating BC migration.

Direct correlation between BM stiffness and apical and basal FC cortical tension
As variations in BM mechanics were shown to impact on FC shape [46], we wished to determine if follicle BM stiffness affected FC cortical tension. We first checked whether reducing BM stiffness altered basal FC tension by measuring cortical tension on the basal side of FCs in S9 control and laminin-depleted FCs. Laser cuts were made in main body cuboidal FCs contacting NCs just ahead of migrating BCs and perpendicular to the dorso/ventral axis (Fig 5A  and 5A'). We observed that the tension at cell-cell contacts on the basal side of FCs in laminin-depleted egg chambers (tj-Resille-GFP>LanB1 RNAi , n = 15, S11 Movie) was reduced compared to controls (tj-ResilleGFP, n = 15, Fig 5B-5D, S11 Movie). These results demonstrated the existence of a direct correlation between follicle BM stiffness and FC basal cortical tension. To further test this hypothesis, we assessed whether increasing BM stiffness could cause the opposite effect, a rise in basal FC tension. Indeed, we found that FCs expressing EHBP1mCh (Resille-GFP; mirr>EHBP1mCh, n = 17) showed increased tension at basal FC-FC contacts compared to controls (Resille-GFP; mirr, n = 16, Fig 5E-5G).
We next tested whether manipulating BM stiffness could also affect tension at cell-cell contacts on the apical side of FCs (S12 Movie). Recent studies have shown that decreasing Col IV levels reduced apical FC surface and caused defects in adherent junctions remodeling [46], suggesting a link between BM stiffness and FC apical tension. Similar to what happened on the basal side, our results showed that decreasing ((tj-Resille-GFP>LanB1 RNAi , n = 15) or increasing (mirr>EHBP1mCh, n = 15) BM stiffness respectively reduced or augmented tension at apical FC-FC contacts (Fig 5H-5M and S12 Movie).
All together, these results suggested that the mechanical properties of the BM could also influence FC cortical tension. Because FCs are in direct contact with NCs, FC tissue mechanics could affect BC migration. To further test this hypothesis, we analyzed whether a direct manipulation of FC tension would affect BC migration. Expression of an Abi RNAi in all FCs led to elimination of actin-containing networks [23,53]. Furthermore, we recently showed that a reduction in Abi levels rescued the increase in FC cortical tension due to integrin elimination, suggesting that Abi might regulate cortical tension [52]. We thus analyzed BC migration in follicles expressing an Abi RNAi in all FCs (tslGFP; tj>AbiRNAi). We found that BCs moved faster in tslGFP; tj>AbiRNAi follicles (n = 6) compared to controls (tslGFP; tj, S9 Fig and S13 Movie, n = 8). While these results supported the idea that FC tension could influence BC migration, the fact that tjGal4 drives gene expression in all FCs, including BCs, and that Abi depletion can block other collective migratory processes such as egg chamber rotation [23], the defects observed in tj>AbiRNAi follicles could be, in part or totally, due to a specific role of Abi in BCs, rather than an effect on FC tension. Thus, we used mirrGal4 to deplete Abi only in central FCs (mirr>Abi RNAi ). We first tested whether this approach altered FC tension of cellcell contacts on the basal side of FCs. To this end, we performed laser ablation on the basal side of FCs located in the mirror area ( Fig 6A). Indeed, we found that tension at FC-FC basal contacts was reduced in the Resille-GFP; mirr>Abi RNAi follicles (n = 10) compared to controls (Resille-GFP; mirr, Fig 6B and 6C, n = 13). Furthermore, laser ablation in NCs (Fig 6D) revealed that NC cortical tension was also reduced in the experimental egg chambers (Resille-GFP; mirr>Abi RNAi , n = 15) compared to controls (Fig 6D-6F, n = 14). Finally, in vivo analysis  (Fig 6G-6I and S14 Movie).

Discussion
BMs are thin, dense sheet-like types of ECM that underlie or surround virtually all animal tissues. By a combination of molecular and mechanical activities, BMs play multiple roles during tissue morphogenesis, maintenance, and remodeling [3]. One of the most prominent developmental functions attributed to BMs is cell migration. The conventional role of BMs during developmental migration is to provide an underlying substratum for moving cells. However, there are examples during morphogenesis where cells do not use BMs as a direct substratum, yet, the 3D environment through which they move is itself encased by a BM. Here, our results show that the stiffness of these encasing BMs can also influence collective cell migration, unraveling a new mechanical role for BMs during development.
The Drosophila egg chamber has served as a paradigm to study the role of encasing BMs in morphogenesis. Studies so far have revealed that the mechanical properties of the BM surrounding the egg chamber influence the behavior of cells in direct contact with it, the FCs. Constrictive forces applied by the egg chamber BM regulate changes in the shape of FCs that underlie tissue elongation [7,8]. BM stiffness was also shown to affect the onset and speed of the collective movement of all FCs over the inner surface of the BM, during the process of "global tissue rotation" [32]. Here, we show that the mechanical properties of the egg chamber BM can also regulate the behavior of cells that are not in direct contact with it, the BCs. BCs move using as a substratum NCs that are surrounded by FCs, encased, in turn, by a BM. Previous studies have shown that reciprocal mechanical interactions between the migrating BCs and the NCs substratum facilitate the movement of BCs [26,27]. Our results reveal that constraints exerted by the BM also influence BC migration, as stiffness of the BM evenly affects NC cortical tension. Recent studies have shown that BCs migrate along the center of the egg chamber, between the junctures where several NCs meet, as the energy cost of unzipping these junctures is most favorable [25]. In this context, we could speculate that the reduction in NC cortical tension detected when reducing BM stiffness could weaken NC-NC junctures, allowing for faster migration. In this scenario, we also propose that, in contrast to what happens in controls where the number of side protrusions is low compared to the front one, the reduced NC cortical tension we observed in laminin-depleted follicles could lead to a lessening of cellular junctions, thus explaining the increased number of side protrusions we observe in this experimental situation. However, our results also show that even though NC cortical tension is reduced in laminin-depleted egg chambers, it seems to be maintained above the threshold necessary to preserve migration along the central path. Several studies in the field have led to propose a grapple and pull model for BC migration, by forward-directed protrusions. In this scenario, BCs extend a long protrusion at the front, grip the substrate at the tip and the cell body pulls itself toward the tip, absorbing the protrusion into the cell body [54]. Furthermore, the ectopic formation of long protrusions in BCs in different experimental conditions slows BC migration [26,55]. Here, we show that BCs from laminin-depleted egg chambers show increased number of side protrusions, yet they migrate faster. However, we show that these ectopic side protrusions are not long. In fact, they have the same length as side protrusions indicated ablated bonds. The statistical significance of differences was assessed with a t test, * P value < 0.05, ** P value < 0.01, and *** P value < 0.001. Horizontal and vertical lines indicate mean and SD, respectively. Scale bars in B, E, H, and K, 5 μm. The raw data underlying panels C, D, F, G, I, J, L, and M are available in S1 Data. BC, border cell; BM, basement membrane; FC, follicle cell; NC, nurse cell.
https://doi.org/10.1371/journal.pbio.3002172.g005 PLOS BIOLOGY from controls. This could explain why BCs from laminin-depleted egg chambers do not loose directionality besides having increase ectopic lateral protrusions. Finally, even though the grapple and pull model is a well-accepted general model for BC motility, an alternative model was proposed [45], in which rather than functioning as effective grapples or attachment sites, the tip of the protrusion functions as sensory organs. This was based on 3 key observations: (1) 80% of protrusions retract; (2) protruding and retracting clusters move with similar speed; and (3) BCs retract the great majority of protrusions before crawling into the space vacated by them [45]. Furthermore, BCs continue to crawl forward even though the protrusion retracts [45]. Then, they propose that "cryptic" lamellipodia observed between BCs [56] might serve a crawling function. In this context, orientation and number of protrusions do not necessarily have to correlate with migration speed, sustaining our results showing that BCs can migrate faster in laminin-depleted egg chambers, despite the increase in lateral protrusions.
Besides cortical tension, cell mechanical properties are also dictated by cytoplasmic pressure. A gradient of cytoplasmic pressure within a defined group of NCs induces TGFβ signaling in the surrounding epithelial cells. This, in turn, has been proposed to induce local softening of the BM and cell shape changes that promote elongation of the follicle, thus supporting a role for cytoplasmic pressure in shaping cells and organs [51]. However, here we show that cytoplasmic pressure on its own does not seem to compellingly influence BC migration. These results suggest that even though the different mechanical properties of cells are interconnected, each of these properties on their own might impact the diverse cell biological processes in distinct ways. In fact, cytoplasmic pressure, and not cortical tension, has been proposed to regulate Drosophila nurse cell shape [51]. Similarly, during the process of mitotic rounding in the columnar epithelium of the Drosophila wing disc, cell area expansion was shown to be largely driven by cytoplasmic pressure, while roundness was primarily driven by cell-cell adhesion and cortical stiffness [57]. In the future, it will be interesting to understand the specific contribution of the different mechanical properties to the diverse morphogenetic processes.
Altogether, we would like to propose a model of the forces regulating BC migration in which, in addition to the already described forces that NCs apply directly over BCs and vice versa, more distant compressive forces exerted by the egg chamber BM emerge as another key regulator of this migratory process, by adjusting the mechanical properties of NCs and FCs (Fig 7). Although it would be interesting to test in the future if the stiffness of the follicle BM could influence BC migration in other ways, like regulating the diffusion of chemo-attractants cues coming from the oocyte, our results prove that the 3D environment mechanically affecting developmental cell migration goes beyond adjacent cells or tissues and suggest that forces exerted by all components of this complex 3D environment are most likely coordinated to allow proper cell migration.
We believe that the new mechanical role of the BM in cell migration proposed in this work is not just restricted to BCs, as, often, during morphogenesis, migrating cells move through tissues encased by a BM. During avian cranial neural crest cell migration, the BMs from the ectoderm and neural tube separate, yet remain in close proximity forming a laminin-lined "channel" through which the cranial neural crest cells migrate [15]. Defective formation of this channel affects cranial neural crest migration [15]. We foresee that changes in the stiffness of the laminin-containing matrix lining the channel may also affect cranial neural crest migration. In the mouse cerebral cortex, BMs are found in the pia and around blood vessels [58]. Radial glia cells located near the ventricle extend long processes that attach to the pial BM. These cellular extensions serve as scaffold for the migration of Cajal-Retzius cells and neuroblasts, from the ventricular layer to the pial surface [59]. A major function of the pial BM is to provide an attachment site for the radial glia cell endfeet [59]. Damage to the pial BM results in defects in radial glial cell extension and consequently neuronal migration [59]. In this context, we believe that changes in the mechanical properties of the pial BM could also affect radial glial cell extension, which would result in neuronal migration defects.
The mechanical cues from the BMs have also been shown to contribute to the progression of several diseases including cancer. Epithelial cancer cells are bound within a BM and surrounded by a healthy stroma composed of collagen fibers and stromal cells. During the evolution of epithelial cancer invasion, cancer cells signal to the stromal cells to remodel the ECM at the tumor-stroma interface, loosen their cell-cell adhesion, and breach the BM around the tumor and invade (reviewed in [60]). Experiments over the last years have focused on the impact of the stiffness of the ECM present at the tumor-stromal interface in tumor invasion. Thus, cell culture experiments have demonstrated that the rigidity of this ECM strongly affects the migration of glioma cells [61]. Based on the results here presented, we would like to propose that the stiffness of the BM enclosing the tumor could also play an important role on the initiation of tumor invasion.
Physical constraints imposed by BMs dictate many cellular events during tissue development. This is the case of branching morphogenesis in many mammalian organs where direct cell-matrix interactions, mainly mediated by integrins, orchestrate cellular rearrangements that sculpt the emerging tissue architecture [62]. However, not all cells undergoing developmental cellular rearrangements are in direct contact with a BM. Instead, there are many examples in which a proportion of these cells contact other cells, which themselves are enclosed by a BM. This happens during the process of budding of the mammary gland, where the BM covers multilayered terminal ends growing dynamically by cellular rearrangements [63]. Similarly, in the stratified tip of pancreas and salivary glands, both outer, in contact with the BM, and inner tip cells rearrange and change position during the process of clefting [64,65]. Based on the results we present in this work, we foresee that the mechanical properties of the BM would also influence the rearrangement of cells located in the inner layers. Investigating this will help understanding the complexity of morphogenesis in the context of multicellular developing tissues.
The results of this work unravel that, besides acting as a substratum, BMs can also regulate developmental cell migration by imposing constriction forces over the 3D environment through which cells move. A recent study, mining the Cancer Genome Atlas, has found that copy number alterations and mutations are more frequent in matrisome genes than in the rest of the genome, and that these mutations are statistically likely to have a functional impact [66]. Thus, identifying new roles for BMs in cell migration is helpful to understand not only morphogenesis but also cancer progression.

Resource availability
Lead contact. Detailed information and requests for resources generated in this study should be addressed to and will be fulfilled by the lead contact María D Martín Bermudo (mdmarber@upo.es).
Materials availability. All reagents generated in this study are available upon request to the lead contact.

Experimental model and subject details
Drosophila husbandry. Drosophila melanogaster strains were raised at 25˚C on standard medium. Laminin hypomorphs were trans-heterozygous for lanB1 28a [11] and l(2)k05404 [67], neither of which affect the coding region. To knock down laminin, Coll IV or Abi levels or to increase EHBP1 expression in the FE, the traffic jam (tj) [68] and the mirror (mirr) [42] Gal4 drivers were used in combination with the following lines: UAS-LanB1 RNAi (VDRC 23119), UAS-Vkg RNAi (BDSC 50895), UAS-Abi RNAi (DGRC-Kyoto 9749R), and UAS-EHBP1mch (BDSC 67145). To visualize cell membranes in vivo, using the membrane marker Resille-GFP, and, at the same time, use the tj Gal4 driver to express a desired UAS construct in all FCs, we generated flies with a recombinant second chromosome carrying both transgenes, Resille-GFP-tj. Stocks used in this study are described in Table 1.

Generation of Drosophila transgenic lines expressing GFP in border cells.
To generate the tslGFP construct, an EcoR1/BamH1 fragment from C4PLZ-tsl(G), previously shown to drive expression in BCs [38], was subcloned into the pCasper-GFP-PH vector [69]. The plasmid was introduced into the germ line of w 1118 flies by the company BestGene following standard methods and several independent transgenic lines were isolated.

Methods details
Immunohistochemistry. Flies were grown at 25˚C and yeasted for 2 days prior to dissection. Ovaries were dissected at room temperature in Schneider's medium (Sigma Aldrich).
After that, fixation was performed incubating egg chambers for 20 min in 4% paraformaldehyde in PBS (ChemCruz). Samples were then permeabilized with a phosphate-buffered saline +1% Tween20 solution (PBT) and incubated in a blocking solution (PBT10) for 1 h. Ovarioles were then incubated with the primary antibodies, rabbit anti-phosphomyosin (Cell Signalling) or anti-LanB1 antibody (Abcam, EPR3189) overnight, followed by a washing step and incubation with secondary antibodyies Cy3 and Cy2 (Jackson ImmunoReseach Laboratories). To visualize F-acting and DNA, ovarioles were incubated for 30 min with Rhodamine Phalloidin (Molecular Probes, 1:40) and 10 min with Hoechst (Molecular Probes, 1:100), respectively. Experimental and control ovarioles were always mixed and stained together. Samples were mounted in Vectashield (Vector Laboratories) and imaged on a Leica SPE confocal microscope (DM 2500).
Time-lapse imaging. For time-lapse imaging, 1 to 2 days old flies were fattened on yeast for 24 to 48 h before dissection. Ovaries were dissected in Schneider's medium. Culture conditions were performed as described in [43]. Movies were acquired on a Leica SP5 MP-AOBS confocal microscope with a 40 × 1, 3 PL APO objective and a Nikon confocal microscope with a 40× oil objective. Z stacks of 30 to 40 slices (0.75 to 1.5 mm interval) were taken every 3 to 4 min.
Atomic force microscopy measurements. Atomic force microscopy measurements were performed as in [32]. Ovarioles were dissected out of the muscle sheath to make sure that the AFM cantilever was in direct contact with the BM. Stiffness of ex vivo ovarian tissues immobilized on Petri dishes using Cell-Tak (BD Biosciences, Oxford, United Kingdom) and containing Drosophila culture media was tested by AFM within 30 min of dissection. Monodisperse polystyrene beads (diameter 5.46 ± 0.12 μm, microParticles GmbH, Berlin, Germany) were glued to a silicon cantilever with a nominal spring constant of 0.03 N/m (Arrow TL1, Nano-World, Neuchatel, Switzerland). Cantilevers were mounted on a JPK CellHesion 200 (Bruker Nano, Berlin, Germany) set up on an inverted optical microscope. Exact cantilever spring constant was determined using the thermal noise method included in the AFM software. The cantilever was positioned at the desired position by brightfield microscopy. Force-distance-curves were taken at an approach speed of 10 μm/s and a maximum force F = 5 nN. Force-distance curves were analyzed using the JPK Data Processing software, which fits the Hertz model to the data. The apparent elastic modulus, K = E/1-ν 2 , which is a measure of elastic stiffness, was determined using F = 4/3 K p Rδ 3/2 , where R is the radius of the indenter, δ the indentation depth, and F the applied force, with E being the Young's modulus and ν the Poisson's ratio. Normal distribution of AFM measurements was confirmed using the Kolmogorov-Smirnov test. The statistical significance of the differences between experimental and control values was evaluated using two-tailed t tests, with p value thresh-holds of *p < 0.05, **p < 0.01, and ***p < 0.001.14 measurements from 6 control egg chambers, 16 from 8 laminin-depleted, 15 from 7 Col IV-depleted, and 14 from 7 EHPB1mCh overexpressing ovarioles were collected. In addition, between 4 to 6 measurements in both the anterior and mirror region from 4 mirr and mirr> EHPB1mCh different ovarioles were collected.
Laser ablation. Laser ablation experiments were performed in an Olympus IX-81 inverted microscope equipped with a spinning disk confocal unit (Yokogawa CSU-X1), a 100× oil objective, a 355 nm pulsed laser, third-harmonic, solid-state UV laser, and an Evolve 512 EMCCD digital camera (Photometrics). To analyze tension in NCs and FCs, a pulse of 1,000 mJ energy and 20 msec. and 75 mJ energy and 4 msec. duration, respectively, was applied to sever plasma membranes of cells. In all cases, cell surfaces were visualized with the membrane marker Resille-GFP and a Cobolt Calypso state laser (l = 491 nm 50 mW) was used for excitation of the GFP. To minimize potential effects due to anisotropic distribution of forces in NCs and FCs, cuts were made perpendicular to the AP axis and to the dorsal ventral axis of the egg chamber, respectively. Images were taken 3 s before and 10 s after laser pulse, every 0,5 s. To analyze the vertex displacements of ablated cell bonds, the vertex distance increase from different ablation experiments (DL) was averaged using as L 0 the average of distance of the vertexes 3 s before ablation. The initial velocity was estimated as the velocity at the first time point (t1 = 0,5 sec). Standard deviation (SD) was determined.
Image processing and data analysis. Measurements of cluster velocity, cluster rotation, and protrusion dynamics were done manually as described in [70]. Briefly, once the body of the cluster was defined according to [40], forward directed speed was calculated using the distance between the center of the cluster at 1 time point and the next in the X axis only. When needed, drift of egg chambers was corrected using the plugin StackReg from FIJI-Image J [71]. To study BC cluster rotation, the position of an individual BC was manually tracked in z stacks of GFP confocal sections time point by time point for an interval of 20 min during the late phase of migration. To analyze protrusion orientation, a line was drawn from the center of the cluster to the tip of the extension. The angle of this line relative to the x-axis was calculated. Extensions were classified as front (45˚to 315˚), side (255˚to 315˚and 45˚to 135˚), and back (135˚to 225˚). In all cases, an average of 150 protrusions was analyzed.
Analysis of the radius of curvature of NCs was carried out by measuring the radius of manually drawn circles that fit the curvature of the posterior facing membranes of NCs. This quantification was done only on 1 NC per row (anterior, central anterior, central posterior, and posterior) in at least 10 independent follicles.
For the quantification of pSqh fluorescent intensities in NCs, square regions of interest (ROIs) of 28.94 μm 2 (20 × 20 pixels) were applied to the membranes of central-posterior and posterior NCs. The mean gray value was then measured using the FIJI-Image J measure tool. This analysis was performed in individual Z-confocal sections corresponding to focal planes in which BCs were visible. For the quantification of pSqh fluorescence intensity in BCs, maximum projections of 5 Z-confocal sections were taken every 0.75 μm. Then, 28.94 μm 2 square ROIs were applied to the front, rear, and both sides of BCs and the mean gray value was then measured with the FIJI-Image J measure tool.
Statistical analysis of significant differences between control and experimental samples was done using Student's t test, * P value < 0.05, ** P value < 0.01, and *** P value < 0.001. Outliers from data sets were identified by the ROUT method with Q = 1% and were not included in the analysis and graphs. The statistical significance of differences was assessed with a t test, *** P value < 0.001. Horizontal and vertical lines indicate mean and SD, respectively. Scale bars in A' and B', 20 μm. The raw data underlying panel C are available in S1 Data. (JPG) S1 Data. Raw data behind all graphs. (XLSX)