Research Article

Hydraulic fracturing and active coarsening position the lumen of the mouse blastocyst

See allHide authors and affiliations

Science  02 Aug 2019:
Vol. 365, Issue 6452, pp. 465-468
DOI: 10.1126/science.aaw7709

The making of a lumen

During the early days of mammalian development, the formation and positioning of a fluid-filled lumen, the blastocoel, defines the first axis of embryo symmetry. Dumortier et al. describe how a blastocoel lumen arises from the hydraulic fracturing of cell-cell contacts into hundreds of micrometer-size water pockets that then form a single large lumen (see the Perspective by Arroyo and Trepat). The authors characterized the process of lumen formation mechanically and molecularly and were able to manipulate the first axis of symmetry of the mammalian embryo experimentally. Thus, fluid dynamics plays a key role in the embryo and may play a similar role in the formation of other biological cavities.

Science, this issue p. 465; see also p. 442


During mouse pre-implantation development, the formation of the blastocoel, a fluid-filled lumen, breaks the radial symmetry of the blastocyst. The factors that control the formation and positioning of this basolateral lumen remain obscure. We found that accumulation of pressurized fluid fractures cell-cell contacts into hundreds of micrometer-size lumens. These microlumens eventually discharge their volumes into a single dominant lumen, which we model as a process akin to Ostwald ripening, underlying the coarsening of foams. Using chimeric mutant embryos, we tuned the hydraulic fracturing of cell-cell contacts and steered the coarsening of microlumens, allowing us to successfully manipulate the final position of the lumen. We conclude that hydraulic fracturing of cell-cell contacts followed by contractility-directed coarsening of microlumens sets the first axis of symmetry of the mouse embryo.

During pre-implantation development, the mammalian embryo forms the blastocyst, which consists of a squamous epithelium, the trophectoderm (TE), enveloping the inner cell mass (ICM) and a fluid-filled lumen, the blastocoel (13). This lumen forms invariably at the interface between ICM and TE cells and segregates the ICM toward one side of the embryo, hence breaking its radial symmetry. This defines the first symmetry axis in the embryo, which guides the position of the main axes of the mammalian body (4). External constraints, such as the one provided by the zona pellucida, are thought to guide the orientation of this axis (5). However, little is known of the mechanisms that are intrinsic to the embryo. Lumen formation and positioning have been studied in the context of apical lumens (6), such as the ones formed in vitro (7, 8) or in vivo within the epiblast upon implantation of the blastocyst (9). Apicobasal polarity allows building an osmotic gradient that draws water from the outside medium into the apical compartment sealed by tight junctions (6). The apical membrane is depleted of adhesion molecules and can contain proteins that act as a contact repellent (10). This makes the apical membrane weakly adhesive and a favorable interface for the collection of fluid between cells. However, blastocysts are akin to “inverted cysts” (11, 12) and form their lumen on the basolateral side of cells (13, 13), where adhesion molecules concentrate (fig. S1), making this interface mechanically less favorable for fluid to accumulate from the outside medium (fig. S2). Therefore, it is unclear how the blastocoel forms within an adhesive interface and acquires its position between TE and ICM cells.

The blastocoel follows from swelling and discharge of microlumens

One possibility would be to separate one or a few neighboring TE-ICM contacts and expand the blastocoel from this initial gap (14). Alternatively, multiple intercellular gaps could appear between cells, as can be observed between TE cells in electron micrographs of blastocysts from mice (15) and nonhuman primates (16, 17), which would merge into a single lumen via unknown mechanisms (3). To investigate this, we used mouse embryos expressing a membrane marker (18) to perform high-resolution imaging at the onset of lumen formation, when blastomeres have completed their fifth cleavage (movie S1). We observed hundreds of micrometer-size lumens (microlumens) forming simultaneously throughout the embryo between cell-cell contacts and at multicellular interfaces (Fig. 1A). Notably, this included contacts between ICM cells, where the blastocoel never forms.

Fig. 1 The blastocoel forms by swelling and discharge of microlumens.

(A) Snapshots of blastocoel formation in an embryo with membrane marker (mTmG) from movie 1. Microlumens form transiently at cell-cell contacts (red arrowheads) and multicellular junctions (blue arrowheads). They first swell and then shrink as the blastocoel (purple star) expands. The lower panels are 3× magnifications of the green squares in the upper panels. Scale bar, 10 μm. (B) Blastocoel (purple) and microlumens growth dynamics at bicellular (red) or multicellular (blue) junctions. For bicellular microlumens, means ± SEM of all microlumens at a cell-cell contact are shown.

We found that the size of microlumens evolves rather synchronously throughout the embryo, showing an initial phase of growth followed by a shrinking period (figs. S3 to S5). Although eventually most microlumens shrink, one continues expanding and becomes the blastocoel (Fig. 1B). From image and data analysis, we characterized two types of microlumens, at either bicellular or multicellular interfaces, and determined stereotypic parameters describing their evolution: For all microlumens except the blastocoel, we identified a swelling phase followed by a discharge phase (Fig. 1B, fig. S3, and movie S1). After 87 ± 10 min of steady growth, microlumens between two contacting cells shrank within 68 ± 8 min (mean ± SEM of 35 contacts from seven embryos). Multicellular microlumens grew to 10 times the size of bicellular microlumens over a similar duration (swelling for 75 ± 21 min and discharge for 65 ± 21 min, mean ± SEM from 21 multicellular junctions from seven embryos) as a result of higher swelling and discharge rates (fig. S3). The phenomenon is rather synchronous with bicellular microlumens appearing within 24 ± 4 min from one another and peaking 25 ± 12 min earlier than multicellular microlumens (figs. S4 and S5). From these observations and analysis, we conclude that the formation of the blastocoel does not result from localized separation of contacts between TE and ICM cells specifically. Instead, transient ruptures of cell-cell contacts result in the formation of myriad lumens throughout the embryo that subsequently disappear, leaving only one remaining.

Microlumens form by hydraulic fracturing of cell-cell contacts 

To understand what controls the initiation of the blastocoel, we first investigated how microlumens form at cell-cell contacts. Contacts are enriched in Cdh1, the main cell-cell adhesion molecule of the blastocyst (19), which provides them mechanical stability. To visualize the localization of Cdh1 during lumen formation, we generated a transgenic mouse line expressing Cdh1 fused to green fluorescent protein (Cdh1-GFP) under its endogenous promoter (20) in addition to a membrane marker (18). When microlumens formed, Cdh1-GFP localization was reorganized, seemingly accumulating at the edges of microlumens (Fig. 2A and movie S2). This made the distribution of Cdh1-GFP more heterogeneous (Fig. 2B). The reorganization of Cdh1 could directly result from the accumulation of fluid detaching cell-cell contacts. Such hydraulic fracturing of cell-cell contacts has been described in vitro when fluid, pressurized at a few hundreds of pascals, is flushed through the basolateral side of mature epithelial monolayers (21). In the pre-implantation embryo, an osmotic gradient forces fluid into the intercellular space from the outside medium rather than from the cells, which seem to keep their volume constant during blastocoel formation (fig. S2).

Fig. 2 Hydraulic fracturing can be directed by cell adhesion.

(A) Cdh1-GFP (green) and membrane (magenta) reorganize during microlumen formation. Top: Snapshots of the whole embryo 90 min before and 90 min after microlumen appearance. Scale bar, 10 μm. Bottom: Magnifications of the dashed rectangles. Scale bar, 1 μm. (B) Coefficient of variation of Cdh1-GFP intensity along cell-cell contacts (337 and 333 contacts from 24 embryos at –90 and +90 min compared to the time of microlumen appearance, respectively). *P < 10–4 (Student’s t test). (C) Schematic diagram of chimera experiments. WT, wild type. (D) Chimeric embryos composed of control mTmG (cyan) and control mG (magenta) cells or of control mG (cyan) and mCdh1 (red) cells. Scale bars, 10 μm.

Hydraulic fracturing requires a rise in hydrostatic pressure between cells, first within microlumens and eventually in the blastocoel. To evaluate the pressure in the blastocoel, we used micropipette aspiration, which allows noninvasive measurement of the surface tension and hydrostatic pressure of liquid droplet–like objects (22). We measured pressures of 296 ± 114 Pa (mean ± SD) for 25 blastocysts (fig. S6), which is about 10 times the pressure for individual blastomeres (23, 24). This reveals that the hydrostatic pressure in the blastocyst is large and of magnitude comparable to pressures capable of inducing hydraulic fracturing in vitro (21). Therefore, fluid accumulation may be responsible for detaching cell-cell contacts and reorganizing Cdh1. To block fluid accumulation, we inhibited the Na/K–adenosine triphosphatase pump to prevent the building of an osmotic gradient drawing water into the embryo (25), inhibited the Rho kinase to disrupt epithelial polarized transport (13), or added sucrose to override the osmotic gradient (26) (fig. S7). All three complementary treatments prevented microlumen formation, and Cdh1-GFP remained homogeneously distributed at cell-cell contacts (fig. S7 and movies S3 to S5). This indicates that fluid accumulation is required for detaching cell-cell contacts and Cdh1 reorganization.

Although fluid accumulation may displace Cdh1, conversely, Cdh1 could provide mechanical resistance to fluid accumulation. Therefore, we patterned cell-cell adhesion to test whether this could affect lumen expansion. To achieve this, we took advantage of embryos in which Cdh1 is knocked out maternally (mCdh1), which form a blastocyst despite their initially lower adhesion (23, 27). We generated chimeric embryos made either from two differently labeled control embryos or from control and mCdh1 embryos (Fig. 2C). Control chimeras formed their blastocoel on either half of the embryo (Fig. 2D). On the other hand, in mCdh1 chimeras, the blastocoel formed preferentially on the mCdh1 half (Fig. 2D). This affected the allocation of cells to the different tissues of the blastocyst (fig. S8), with mCdh1 blastomeres contributing almost exclusively to the mural part of the TE, which is the TE that lines the blastocoel (fig. S8). Therefore, patterning of Cdh1 levels is sufficient to direct the accumulation of fluid. Together, these findings indicate that pressurized fluid collects along the path of lowest mechanical resistance, separates cell-cell contacts, and reorganizes Cdh1 adhesion molecules. We therefore propose that microlumens form throughout the embryo by hydraulic fracturing of cell-cell contacts.

Microlumens coarsen in a process akin to Ostwald ripening

We then proceeded to investigate how the embryo resolves the hundreds of microlumens into one single blastocoel during the discharge phase. One mechanism would be for microlumens to coalesce and fuse when in close proximity. On time scales ranging from tens of minutes to hours, we could not observe frequent fusion events or movements of microlumens toward the final position of the blastocoel (movies S1 and S2). We used light-sheet microscopy to image microlumens at high temporal and spatial resolution, and observed that microlumens seem to drain their content on time scales of a few minutes (movie S6). To visualize the fluid at the origin of the blastocoel, we immersed 16-cell–stage embryos in medium containing a fluorescently labeled dextran, which becomes trapped by tight junctions within the intercellular space (28). Dextran transiently accumulated at all cell-cell contacts before being flushed into the blastocoel, which suggests that microlumens are physically connected through the intercellular space (movie S7). In fact, any difference in hydrostatic pressure between two connected microlumens should lead to a flow of luminal fluid from the more pressurized microlumen to its counterpart (Fig. 3A). The pressure in microlumens is related to the tension and curvature at the lumen interface via the Young-Laplace equation (supplementary text). At the same microlumen tension, a pressure difference is due to differences in microlumen sizes. This leads smaller microlumens to discharge into larger ones (Fig. 3A and movie S8) and may explain why small bicellular microlumens shrink earlier than larger multicellular microlumens (Fig. 1B and fig. S3). This process is reminiscent of Ostwald ripening (29), which drives the slow coarsening of foams. In the embryo, however, the exchange is not limited by diffusion but rather by fluid flow through the intercellular space. In addition, surface tension may not be homogeneous and, in fact, the direction of the flow may be reversed for a given asymmetry in tension between two lumens (Fig. 3A).

Fig. 3 Physical model of lumen coarsening and localization.

(A) Phase diagram for the fluid flow between two lumens as a function of their tension asymmetry δ and area asymmetry β. Adhesive contact tension γc is 1 (fig. S9). (B) Mean time evolution of the area of the winning lumen (purple squares) and losing lumens (orange triangles) resulting in the coarsening of the lumen network as depicted in the inset (50 simulations with γTE = 0.9, γICM = 1, γc = 1, pumping rate λp = 10–3 a.u.). (C) Winning probability for lumens at TE-ICM interfaces as a function of tension asymmetry δTE-ICM. Each point results from averaging of 1000 simulations (λp = 10–3 a.u.). Schematic diagrams show the mean localization of winning lumens.

To study microlumen dynamics, we modeled the embryo as a two-dimensional network of pressurized microlumens and performed numerical simulations to predict the position of the final lumen (Fig. 3, B and C). The size of a microlumen evolves by direct swelling and via fluid exchange through the network (Fig. 3B and supplementary text). As observed experimentally (Fig. 1B and fig. S3), lumens show a swelling phase followed by a discharge of all lumens but one, which siphons all the fluid (Fig. 3, B and C, and movie S8). This biphasic signature is characteristic of a coarsening mechanism akin to Ostwald ripening (supplementary text) (29).

Blastocoel fluid is guided by differences in cell contractility

In networks with homogeneous lumen tension, the lower connectivity of microlumens at the border favors the formation of the final lumen in the interior (Fig. 3C and supplementary text). In agreement with mCdh1 chimera experiments (Fig. 2D), the model predicts that a pattern of adhesive contact tension could be sufficient to position the blastocoel at the TE-ICM interface (fig. S9). Alternatively, when a small excess in tension is imposed on microlumens in the interior, the final lumen ends up invariably at the margin (Fig. 3C). The model predicts that higher lumen tension at the ICM-ICM interface is sufficient to position the blastocoel between TE and ICM cells, as observed in vivo.

To examine whether differences in surface tension between blastomeres could explain how coarsening directs luminal fluid toward the TE-ICM interface, we first investigated the shape of microlumens. When measuring the radius of curvature of microlumens, we detected asymmetries, which were more pronounced at heterotypic TE-ICM interfaces than at homotypic TE-TE interfaces [mean ± SEM, curvature ratio of 1.01 ± 0.02 for 71 TE-TE microlumens and 1.25 ± 0.04 for 58 TE-ICM microlumens from seven embryos, P < 10–2 (Student’s t test); Fig. 4, A and B, and movie S1]. Moreover, in most cases, microlumens at TE-ICM interfaces bulge into the TE cell rather than into ICM cells (79% of 58 microlumens from seven embryos, χ2 test P < 10–3). This suggests differences in the hydrostatic pressure and/or surface tension between TE and ICM cells. Such differences are supported by the cell-scale curvature of TE-ICM interfaces, where ICM cells bulge into TE cells (fig. S10). This disparity is likely due to higher contractility of ICM cells relative to TE cells, which is responsible for their sorting at the 16-cell stage (24). Indeed, inhibiting cell contractility decreases the curvature at TE-ICM interfaces (fig. S10) and the surface tension of the blastocyst (fig. S6). Our model predicts that this pattern of contractility within the embryo would result in microlumens preferentially discharging to the TE-ICM interface (Fig. 3C and movie S9).

Fig. 4 Microlumen coarsening is controlled by cell contractility.

(A) The curvature of microlumens is symmetric at TE-TE interfaces (left, arrowheads) and more asymmetric at TE-ICM interfaces (right image shows a confocal slice 16 μm deeper than the one above). The lower panels are 3× magnifications of the green rectangles in the upper panels. Scale bar, 10 μm. (B) Radius of curvature at microlumens facing TE-TE (71) and TE-ICM (58) interfaces from seven embryos during the discharge phase, 60 min after inversion. The ratio of curvature is color-coded with TE cells ordered randomly for TE-TE ratios. *P < 10–7 (paired Student’s t test); n.s., not significant (P > 10–2). (C) Schematic diagram of chimera experiments. (D) Chimeric embryos composed of control mTmG (cyan) and control mG (magenta) cells or of control mG (magenta) and mCdh1 (green) cells. Scale bars, 10 μm.

To experimentally test the ability for heterogeneous contractility to direct the offloading of microlumens, we generated chimeras (Fig. 4C) using embryos in which Myh9 is maternally knocked out (mMyh9), which are viable despite their lower levels of Myh9 and lower surface tension (24). As previously shown (Fig. 2, C and D), control chimeras form their blastocoel on either side of the embryo (Fig. 4D). On the other hand, in mMyh9 chimeras, the blastocoel forms preferentially on the mMyh9 half (Fig. 4D). This affects the allocation of cells to the different tissues of the blastocyst (fig. S8). In agreement with previous analogous experiments (24), mMyh9 blastomeres are depleted from the ICM and instead contribute mostly to the mural TE (fig. S8). Therefore, differences in Myh9 levels are sufficient to control the position of the blastocoel, in agreement with corresponding simulations (movie S10). On the basis of these findings, we propose that a mechanism akin to Ostwald ripening directs the offloading of luminal fluid into the blastocoel.


We have established that both patterned cell-cell adhesion and patterned cell contractility can position the blastocoel (Figs. 2D and 4D). This suggests that both localized hydraulic fracturing of cell-cell contacts and directed fluid offloading could set the first axis of symmetry of the mouse embryo. These two phenomena constitute complementary mechanisms explaining the formation and positioning of lumens in general, and of basolateral lumens in particular, which can be found in healthy (30) and pathological situations (12). As far as the blastocyst is concerned, although both hydraulic fracturing and active coarsening can position the blastocoel, we are unable to detect any conclusive relation between the final position of the blastocoel and the initial localization of Cdh1 or contact fracture (figs. S4, S5, and S11). On the other hand, differences in contractility are undoubtedly at play (24, 31, 32) (Fig. 4 and fig. S10). Therefore, we propose that the first axis of symmetry of the mouse embryo is positioned by contractility-mediated ripening of microlumens formed after hydraulic fracturing of cell-cell contacts.

Determining the molecular and mechanical events that control the formation of microlumens and the fracture of cell-cell contacts constitutes an exciting research avenue that will greatly benefit from previous studies on contact mechanical stability (21, 33). Efforts to build a comprehensive model of blastocoel formation will need to integrate explicitly individual cell mechanics and to evaluate the contributions of other cellular processes such as vesicular trafficking (15) and ion exchange (8).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S11

Movies S1 to S10

References (3444)

References and Notes

Acknowledgments: We thank the imaging platform of the Genetics and Developmental Biology unit at the Institut Curie (PICT-IBiSA@BDD) for their outstanding support; the animal facility of the Institut Curie for their invaluable help; the Buchholz laboratory for providing the Cdh1-GFP BAC construct; Y. Petersen from the transgenesis platform at the European Molecular Biology Laboratory; T. Hiiragi and his lab for their generous help in making the Cdh1-GFP mouse line; and the Maître and Turlier labs, Y. Bellaïche, L. Noiret, A. Coulon, A. Clark, and G. Charras for fruitful discussions and comments on the manuscript. Funding: Research in the lab of J.-L.M., who is supported by the Institut Curie, the CNRS and the INSERM, is funded by the ATIP-Avenir program, an ERC-2017-StG 757557, a PSL “nouvelle équipe” grant and Labex DEEP (ANR-11-LBX-0044) which are part of the IDEX PSL (ANR-10-IDEX-0001-02 PSL). The lab of H.T. acknowledges support from the Fondation Bettencourt-Schueller, the CNRS-INSERM ATIP-Avenir program, and the Collège de France. Author contributions: H.T. and J.-L.M. conceptualized the project and acquired funding. J.G.D. and J.-L.M. designed experiments. J.G.D., A.-F.T., and L.d.P. performed experiments. J.G.D., A.-F.T., L.d.P., and J.-L.M. analyzed the data. M.L.V.-S., A.M., and H.T. designed the theoretical model and performed the numerical simulations. Competing interests: Authors declare no competing interests. Data and materials availability: The microscopy data, ROI, and analyses are available on the following repository under a CC BY-NC-SA license: The code of the simulations is available on the following repository under a MIT license: The Cdh1-GFP transgenic mouse line is available upon request.

Stay Connected to Science

Navigate This Article