During human embryonic development, the fertilized egg undergoes multiple rounds of cell division and differentiation to form the pluripotent epiblast at the blastocyst stage, which ultimately gives rise to all tissues in the fetus 1 . Embryo development from the pluripotent epiblast commences upon implantation of the blastocyst into the uterine wall, following which pluripotent stem cells self-organize to form a roughly spherical structure containing a fluid-filled lumen 2 .
Lumens or fluid-filled cavities are a ubiquitous feature of metazoans and are often evolutionarily linked to the origin of body-plan complexity 3 . Proper formation of the epiblast lumen is critical for establishing morphogen gradients that drive subsequent embryonic development 4,5 . While the physical mechanism of lumen expansion in the human epiblast is unknown, established mechanisms of de novo lumenogenesis in other model systems involve apoptosis or osmotic pressure gradients 6,7 . Apoptosis drives lumenogenesis in certain mammary epithelial models where cells at the center of a cluster die, resulting in a hollow cavity 8 . Osmotic pressure gradients drive lumen growth in the mouse blastocyst [9][10][11] , MDCK (Madin-Darby canine kidney) cells 12 , bile canaliculi 13 , and zebrafish inner ear 14 . In each of these cases, apico-basally polarized cells with tight junctions, pump osmolytes into the lumen which builds osmotic pressure and drives water into the lumen, expanding its volume 15,16 . While pressure has been shown to drive lumen expansion in the mouse blastocyst [9][10][11] , mechanisms of lumen expansion in other early embryonic lumens such as the epiblast cavity are much less understood 17,18 . Importantly, mouse epiblasts cannot be used to fully understand human epiblast morphogenesis as they exhibit key morphological differences with human epiblasts -pluripotent stem cells in mouse epiblasts form a hollow cup shaped structure fused with extraembryonic cells called the egg cylinder whereas in humans epiblasts, pluripotent stem cells form a hollow roughly spherical structure 18 . Recently, the study of polarity 19 and pluripotency 20 dynamics necessary for epiblast lumenogenesis have provided key insights into the cellular processes involved, but the physics driving lumen expansion in the human epiblast remains unclear.
Human induced pluripotent stem cell (hiPSC) models of the embryo reproduce key aspects of development and serve as excellent tools to uncover mechanisms orchestrating human embryogenesis 21,22 , since human embryos cannot be studied directly due to ethical concerns.
hiPSCs have been previously used to model the human epiblast using basement membrane based matrices 20,[23][24][25][26] such as Matrigel as well as using engineered hydrogels 27 . We have previously shown that hiPSCs form lumen-containing structures that morphologically and phenotypically model the human epiblast in a highly reproducible manner, when cultured in 3D in engineered hydrogels which model the confinement experienced by the epiblast in vivo due to blastocyst cavity pressure and extraembryonic cells 27 . In this epiblast model (hereafter referred to as hiPSC epiblast), we now dissect the mechanisms regulating lumen expansion. Epiblast lumen nucleation is initiated by exit from naïve pluripotency and subsequent transition to formative and primed pluripotency 20,26,28 through polarization events mediated by cytoskeletal proteins 29 involving the formation of a specialized structure called the apicosome 30 . However, the physical mechanism of lumen expansion in the human epiblast following initial polarization is unknown. Our experiments and simulations reveal a previously undescribed mechanism of lumen expansion mediated by apical actin polymerization that drives early lumen expansion up to a critical lumen size of ~12 m radius, followed by a transition to osmotic pressure gradient driven lumen growth in lumens larger than 12 m radius.
We formed hiPSC epiblasts by culturing single hiPSCs in 3D viscoelastic alginate hydrogels. In the presence of specific biophysical cues of hydrogel stiffness, viscoelasticity, and cell-adhesion ligand (RGD) density, hiPSCs self-organize into lumen-containing structures that are reminiscent of the human epiblast 27 . While the initial elastic modulus of the hydrogels is 20 kPa, these gels exhibit fast stress relaxation, with a stress relaxation half time of ~70 s (Figure S1A). Further, the relaxation modulus over ~30 mins is ~1 kPa, which is on the same order of magnitude as that experienced by the epiblast cells in the mouse blastocyst 10 (Figure S1A). Thus, these hydrogels roughly mimic the confinement faced by the epiblast in the human blastocyst, however, the precise values of human blastocyst pressure and mechanical contribution of extraembryonic cells are unknown. In these hydrogels, hiPSCs proliferate and begin to form lumens around day 3 of culture, creating 3D monolayered cellular structures with a central, roughly spherical lumen (Figures 1A and 1B). Similar to human epiblasts 20,31,32 , these structures polarized along the apicobasal axis in response to matrix signaling 27 and maintained expression of pluripotency proteins such as Oct4, Sox2 and Nanog, as well as formative pluripotency factors such as Otx2 28,33 , over 7 days of culture (Figures 1C, S1B and S1C). hiPSC clusters also mirrored the morphological features of human epiblasts. The numbers of cells in hiPSC clusters on different days of culture were akin to human epiblasts, with day 3 and day 7 of in vitro culture corresponding to 7 to 8 days post fertilization (d.p.f.) and 11 to 12 d.p.f in human embryos respectively, suggesting similar proliferation dynamics (Figure 1D). Further, lumen volumes and growth rates of hiPSC clusters were close to those of human epiblasts (Figure 1E). Nuclear morphology metrics such as area and perimeter of hiPSC clusters were also similar to those of human epiblasts (Figure S1D). Overall, hiPSCs in engineered 3D alginate hydrogels maintained pluripotency, polarized along the apicobasal axis, and showed lumenal and nuclear morphological similarities to human epiblasts. Therefore, these structures model the human epiblast, allowing study of mechanisms driving epiblast lumen expansion.
We sought to understand the mechanisms underlying lumen expansion in hiPSC epiblasts as a model of human epiblasts. Guided by previous studies of lumenogenesis 6,7 , we first investigated known mechanisms of de novo lumenogenesis including apoptosis and osmotic pressure gradients. During lumen growth in hiPSC epiblasts, few to none apoptotic cells were detected (Figures S1E to S1G), demonstrating that apoptosis does not drive lumen expansion in hiPSC epiblasts, consistent with previous studies 34 .
We next studied the role of osmotic pressure gradients in driving lumen expansion. To build osmotic pressure in the lumen, two requirements need to be met: (i) ion flux into intercellular space or lumen at the apical surface 15 , and (ii) formation of tight junctions to prevent osmolytes from leaking 35 (Figure 1F). These requirements allow osmotic pressure to build up, which draws water into the lumen, generating force necessary for lumen growth. To test if hiPSCs formed tight junctions, cell-impermeable fluorescent dextran was added to the culture media. If mature tight junctions were present, dextran would be expected to be excluded from the lumen. Strikingly, dextran entered lumens smaller than ~12 m in radius, indicating that hiPSCs do not form mature tight junctions during early stages of lumen expansion when the lumen size is below ~12 m radius (Figures 1G to 1H and S2A to S2H). Dextran also localized to the intercellular spaces in junctions between cells and was excluded from cells themselves suggesting that dextran entered lumens through diffusion along the intercellular spaces and not via other mechanisms such as transcytosis (Figures S2I and S2J). As tight junction marker ZO-1 localized to the cell-cell boundary of smaller lumens as well (Figure S2K), it was plausible that tight junction formation was gradual with a complete seal forming at a lumen size of ~12 m radius. But this was not the case. Large macromolecular dextran, with a diameter (~12 nm; 70 kDa dextran) comparable to intercellular space due to adherens junctions (~20 nm) 36,37 , entered lumens smaller than ~12 m in radius but was excluded from lumens larger than this size, highlighting the complete lack of mature tight junctions in hiPSC epiblasts with smaller lumens (Figures 1H and S2G to S2H).
As mature tight junctions were absent in hiPSC epiblasts with smaller lumens (radius < 12 m; hereafter referred to as smaller epiblasts), any ions pumped into these lumens are expected to leak along the intercellular spaces, preventing large pressures from building up. To confirm that this was the case, diffusion dynamics in smaller lumens were measured by observing fluorescence recovery after photobleaching (FRAP) of dextran. The fluorescence signals recovered ~2 min after photobleaching, suggesting that dextran can freely diffuse along the intercellular spaces (Figures 1I to 1K; Video S1). Taken together, these results reveal a lumen size-dependent initiation of tight junction formation, with lumens below ~12 m in radius being leaky.
To further assess the role of osmotic pressure gradients in driving lumenogenesis, lumen shapes were examined. Lumen shapes are expected to be convex or bent outward if pressure was the sole driver of epiblast lumenogenesis, whereas irregularly shaped lumens that are bent inwards suggest that osmotic pressure gradients are not a dominant driver of lumen growth 38 . Lumen shapes were highly irregular for smaller lumens but transitioned to a more bulged, convex shape in larger lumens (Figure 1L to 1N). Thus, the irregularly shaped lumens in smaller epiblasts further indicate that osmotic pressure is not a major driver of early lumenogenesis in hiPSC epiblasts, whereas the regularly shaped lumens in larger epiblasts indicate that osmotic pressure gradients could drive lumen expansion in larger lumens. Finally, laser ablation through an entire cell in smaller epiblasts did not cause any drastic change in cell or lumen size or shape indicating that smaller lumens are not pressurized (Figure S2L). Taken together, the lack of tight junctions, free diffusion out of the intralumenal space, and lumen shapes together indicate that initial expansion of the lumen is not driven by osmotic pressure gradients and emphasize the existence of a pressure-independent mechanism of lumenogenesis.
As epiblast lumenogenesis mechanisms are required to produce forces necessary to overcome resistance from their environment -the surrounding hydrogel in case of the hiPSC epiblast versus extraembryonic cells and blastocyst cavity pressure in case of the human epiblast -we next examined force generation associated with lumenogenesis in order to gain insight into the underlying mechanisms driving early lumen expansion. hiPSC epiblasts of all sizes generated local matrix deformations on the order of tens of micrometers over 18 hr (Figures 2A and 2B; Video S2). As the hydrogels are viscoelastic and viscoplastic, with stresses relaxed on a timescale of minutes and the material undergoing permanent deformation, the magnitude of forces required for the measured matrix deformations depends on the timescale and dynamics of force application.
Nonetheless, as some force generation is necessary, we next probed different cellular force generating machineries to uncover the pressure-independent mechanism responsible for epiblast lumenogenesis.
We first examined the role of actomyosin contractility in lumen expansion, given the wellknown function of the actomyosin cytoskeleton network in generating contractile forces. Myosin II was mostly punctate and largely localized at the basal surface, which could not explain the pattern of forces associated with lumen expansion (Figures 2C and S3A). Further, inhibition of actomyosin contractility on day 3 of culture or in smaller epiblasts (lumen radius < 12 m) as well as on day 7 or in larger epiblasts (lumen radius > 12 m) (Figure 2D), did not significantly impact lumen formation (Figures 2E and S3B). These results demonstrate that actomyosin contractility does not drive lumen expansion in hiPSC epiblasts.
We next examined actin structures and their potential role in driving lumen expansion, as actin polymerization in bundled or branched networks produces protrusive forces that drive cellular morphogenesis in a variety of contexts 39 . In hiPSC epiblasts, F-actin was densely localized at the apical surface (Figure 2F). Super-resolution microscopy using an Airyscan system revealed that apically, F-actin formed a dense mesh-like structure with microvilli protruding from this mesh (Figure 2G). Further, actin nucleation factor N-WASP and actin branching complex Arp2/3 were enriched at the apical surface, which would be expected to promote the formation of a dendritic actin network (Figure 2H). As formation of a lumen and apical surface are intertwined, the time evolution of the apical actin mesh formation was quantified. Apical surface area per cell increased in size as lumens grew but reached an equilibrium size of ~100 m 2 at a lumen size of ~12 m radius (Figures 2I, S3C and S3D). In fact, all cells in larger epiblasts had a similar apical surface area of ~100 m 2 while cells in smaller epiblasts had a wide range of apical surface areas at any given timepoint that were close to or less than 100 m 2 , highlighting cell-cell variations in apical surface formation (Figure 2J). Distinct lumen growth dynamics were observed for smaller and larger lumens while cell volume and thickness stayed relatively constant (Figures S3E to S3G).
Overall, these data show that as lumens form, cells grow their apical surfaces up to an equilibrium value, which is achieved at a lumen size of ~12 m radius, coinciding with the timing of tight junction formation.
We next tested whether actin polymerization could drive early lumen expansion using inhibition studies. Dendritic actin network growth is driven by the Arp2/3 complex, which is nucleated via N-WASP, while linear actin polymerization is initiated via formins. Strikingly, inhibition of actin polymerization by any of these proteins -Arp2/3 complex, N-WASP, and formins -strongly reduced lumen formation in smaller epiblasts but had no impact on larger epiblasts (Figures 2K, 2L, S3H and S3I). Thus, actin polymerization is necessary for lumen expansion in smaller epiblasts. Given these observations, we hypothesized that the growth of apical actin in each cell generates force to drive epiblast lumen expansion in a pressure-independent manner until apical actin growth equilibrates at a lumen radius of ~12 m.
To examine how apical actin polymerization could drive lumen expansion, we first performed time-lapse imaging of fluorescently labelled F-actin during early lumen expansion.
Interestingly, lumen expansion correlated with apical actin polymerization of only a few cells in the epiblast and, in some cases, specifically correlated with increase in apical length of a single cell while other cells maintained relatively constant apical lengths (Figures 3A and 3B; Video S3 top row). While apical lengths are expected to increase with increasing lumen area, it was striking to see large cell-cell variations in growth dynamics for smaller epiblasts (Figures 3C and 3D). In line with these features, smaller epiblasts generated radially asymmetric matrix deformations (Figure 3E).
As individual cells polymerize actin to expand their apical surfaces, it would be expected that they resist each other's expansion, and that the hydrogel would resist overall lumen expansion.
To study these, we first perturbed the dynamics of apical actin in cells and their neighbors by performing laser ablation of apical actin in individual cells in smaller epiblasts (Figure 3F; Video S4 top row). No immediate change in apical length was observed post ablation, suggesting that the stiff apical actin mesh is not under large levels of compression or tension (Figures 3F to 3I and S4A to S4D; Video S4 top row). As the hydrogels are both viscoelastic and viscoplastic, this observation suggested that the stresses resisting the epiblast expansion were relaxed and the hydrogel was plastically deformed, so that residual stresses remaining on the epiblast are low at any given timepoint. Over a timescale of minutes following ablation, however, apical length of the ablated cell decreased, while that of neighboring cells increased, indicating active actin polymerization driving lateral apical expansion in the neighboring cells and lateral pushing forces (Figures 3G to 3I and S4A to S4D; Video S4 top row). But, as the apical actin signal in the ablated cell began to recover, the apical length of the ablated cell expanded again, suggesting that the apical actin re-growth in the ablated cell pushes back against the neighboring cells (Figures 3G, 3H and S4A to S4D). For comparison, no substantial changes in apical lengths were observed in non-ablated controls on a timescale of minutes (Figure S5). These ablation studies directly connect actin network growth to apical expansion and indicate the following interpretation. In smaller epiblasts, cells resist the growth of apical actin in neighboring cells and when such resistance is disrupted, say via ablation, actin in cells neighboring the ablated cell, can actively polymerize, increasing their apical lengths.
To directly test if the growing hiPSC epiblasts are under compression globally from the hydrogel, we dissolved the hydrogel and observed epiblast morphology (Figure S6). hiPSC epiblasts immediately expanded in size post hydrogel dissolution indicating that epiblasts were under some compression (Figures S6A to S6C). Complementarily, following cell lysis, lumens collapsed, and the hydrogel expanded into the space formerly occupied by the epiblast, confirming that the hydrogel was under compression due to epiblast growth (Figures S6D and S6E). Overall, these observations point to quasi-static actin growth where actin polymerization generated forces drive apical growth and lumen expansion, and stress relaxation and plasticity in the hydrogel prevent large compression from building up in hydrogel and thus in the apical actin mesh as well.
Under the idea that actin polymerization at the apical surface of single cells drives lumen growth, cell-cell variations in actin polymerization rate and corresponding apical actin mechanics in smaller epiblasts, should result in a wide distribution of apical curvatures (Figures 2J and 3A to 3D). Analysis of lumen curvature showed a broad range of local curvatures in smaller lumens, as cells build their apical surfaces (Figures 3J to 3L). However, once all cells reach a mature apical size with a dense apical actin mesh, which is the case in larger epiblasts (Figures 2I and 2J), apical surfaces became uniformly flatter (Figures 3J to 3L). Taken together, these results show that apical actin polymerization at the apical surface of single cells drives lumen expansion.
In order to gain insight into the physical mechanisms governing lumen expansion and to investigate the role of actin polymerization in this process, we developed a theoretical model and conducted computational simulations. Before exploring the mechanisms responsible for driving lumen expansion, we first wanted to understand cellular geometric constraints that govern lumen shapes and sizes, agnostic to the specific mechanism driving lumen expansion. In order to do so, we employed a particle-based description of cells to recapitulate lumen size and shape (Figure 4A). Our model incorporated two key geometric constraints observed in experiments as lumens grew: (i) constant cell volume (Figure S3F), and (ii) constant cell layer thickness or cleft length (Figures S3G). These observations imply that cells act effectively as incompressible objects during lumen expansion and can readily adapt their shape to form a confluent layer around the lumen.
Based on these assumptions, the predicted apical lengths and number of cells as a function of lumen size (defined in Equations 1 and 2 in Supplementary Methods) from the model were in excellent agreement with the experimental observations (Figures 4B to 4D). The model predicted rapid growth of apical surface area of individual cells in smaller epiblasts (Figure 4C). In larger epiblasts, however, growth of the apical surface in individual cells is predicted to stall (Figures 4B and 4C), in line with experimental observations (Figures 2I and 2J). These predictions independently point to rapid apical growth as a potential mechanism of lumen expansion in smaller epiblasts. As our experiments revealed apical actin polymerization to be responsible for rapid growth of the apical surface and subsequently lumen expansion, we next assessed the role of apical actin polymerization as well as other physical mechanisms in driving lumen expansion.
To better understand the physics of how apical actin polymerization can drive lumen expansion, we developed a continuum model considering a cell cluster in an elastic hydrogel with a nascent lumen at the interior of the cluster. Cells actively polymerize apical actin allowing rapid growth of the apical surface. Cells in the model also pump ions through the apical and basal surfaces. Water fluxes are driven by osmotic and hydrostatic pressure gradients that result from ion flux and actin-generated stress respectively, but paracellular leaks can dissipate osmotic gradients (Figure 4E). In smaller clusters with leaky junctions, ion concentrations did not build up in the lumen and thus there was no difference in osmotic pressure between the lumen and the hydrogel (Figure 4F). Actin polymerization on the other hand generated stresses that counteract the resistance offered by the elastic hydrogel resulting in lumen expansion (Figure 4G). Using the Laplace relation, the pressure difference across the cell layer was estimated in the scenario of constant intracellular pressure and constant curvature of the cell surface, and accounting for passive and active stresses in the apical actin network (as detailed in Equations 6-9 in Supplementary Methods). Actin stiffness determined the resulting stresses in the network when it is subject to deformation, while the undeformed length factor defines the effective change in the rest length of the apical actin network. Importantly, our model predicted that increasing stiffness of apical actin network enhances lumen growth resulting in water influx into the lumen (Figures 4H and 4I). In contrast, apical actin networks with reduced stiffness are predicted to have smaller lumens. These results highlight the importance of apical actin stiffness in mediating actin polymerization driven lumen expansion.
As the lumen expands, the enclosing cell layer becomes more susceptible to buckling due to increasing stresses in the confining hydrogel that resists the lumen expansion, so we next sought to understand the implications of buckling to lumen growth. To simplify our analysis, we assumed that the gel is elastic with a Young's modulus of 1 kPa, comparable to the relaxed modulus of the hydrogels used in experiments (Figure S1A; Supplementary Methods Table 1), and the cell layer is incompressible, with a Young's modulus of 20 kPa as measured previously in epithelial monolayers 40 . Now, as lumens expand, pressure in the hydrogel is expected to increase while the pressure that the cell layer can withstand 41 without buckling reduces (Figure 4J). Hence, there is a critical lumen size at which the pressure exerted by the hydrogel exceeds the critical buckling pressure of the cell monolayer, causing it to buckle (Figure 4J). For the Young's modulus of the cell monolayer and modulus of the gel assumed above, the critical lumen size is about 12 m. To facilitate lumen expansion beyond this critical size, development of lumenal pressure is required to counterbalance the hydrogel pressure, thereby preventing buckling of the cell layer, as buckling is not observed physiologically.
Overall, our computational model for smaller epiblasts demonstrates that apical actin polymerization, which results in growth and stiffening of the apical surface, is sufficient for lumen expansion, even without osmotic pressure gradients. However, modeling predicts that a transition to pressure driven growth is required once the lumen reaches a certain size to avoid cell layer buckling due to increased stress from the surrounding hydrogel.
As apical actin polymerization is equilibrated in larger epiblasts, and given the predicted transition to pressure-driven growth in the modeling, we next examined the mechanism driving further growth of these lumens. Concomitant with the maturation of the apical surfaces (Figures 2I to 2L) at a lumen radius of ~12 m, hiPSC epiblasts form mature tight junctions (Figures 1G, 1H, S2D to S2H and S2K), which could allow osmotic pressure to build inside the lumen. Further, these lumens have convex, bulged out shapes (Figures 1L to 1N). Overall, these characteristics are consistent with osmotic pressure driven lumen growth 16 .
To test if osmotic pressure was responsible for growth of larger lumens, we performed time-lapse imaging. Unlike actin polymerization driven lumen growth, apical lengths of most cells in larger epiblasts generally increased over time and positively correlated with increase in lumen area (Figures 5A to 5D; Video S3 bottom row). Subsequently, larger epiblasts generated radially uniform matrix deformations as they grew (Figures 5E to 5G). Pressure-driven growth is governed by Young-Laplace law, which necessitates cells to be under tension to balance the lumenal pressure 10,12 . Thus, laser ablation of apical actin was performed to examine if cells were under tension or compression. Ablated cells exhibited an immediate increase in apical length postablation revealing that cells were under tension (Figures 5H to 5K and S4E to S4H; Videos S4 bottom row and S5). With time, apical length of ablated cells increased further while no change in neighboring cells was observed (Figures 5H to 5K and S4E to S4H; Videos S4 bottom row and S5). Overall, these results provide strong evidence that osmotic pressure drives lumen growth in larger epiblasts.
To better understand the physics of osmotic pressure driven lumen growth, we applied our numerical model to larger epiblasts with tight junctions. The model enabled us to determine the equilibrium cell and lumen sizes depending on the passive and active ion transport across the cell layer, passive leak through the cleft or intercellular space, as well as the number and mechanical properties of the cells forming the layer (Section B in Supplementary Methods for detailed analysis). For simplicity, we assumed constant properties of the surrounding matrix and cell cortex, although in general they might exhibit nonlinear responses to applied stress. Other relevant parameter values are summarized in Supplementary Methods Table 1. In this case, ion pumping into the lumen increased osmotic pressure and resulted in robust lumen growth (Figures 5L and 5M). Predicted lumen sizes were in excellent agreement with experimental observations (Figure 5N). Balance between apical and basal ion pumping was found to be critical for buildup of osmotic pressure (Figure 5M). Higher basal pumping without apical pumping, and vice versa, prevented accumulation of ions in the lumen and no lumen growth occurred (Figure 5M). Overall, our experiments and model indicate that ion pumping in the presence of tight junctions builds osmotic pressure to drive the growth of larger lumens.
Finally, to investigate the in vivo relevance of the mechanisms discovered in hiPSC epiblasts, we analyzed transcriptional signatures of the human epiblast as a function of developmental time using single cell RNA sequencing data generated from peri-implantation human embryos 42,43 . Epiblast lumen forms soon after implantation at ~7 d.p.f and expands in volume up to gastrulation at ~14 d.p.f (Figures 1A and 1B) 2 . Based on analysis of epiblast cells from 7 to 14 d.p.f using KNN (k-nearest neighbor) clustering of UMAP (Uniform Manifold Approximation and Projection) dimensionality reduction plots, cells were divided into two subpopulations (Figures 6A and 6B). As the two epiblast subpopulations were roughly separated based on developmental time, we annotated these as early and late epiblasts (Figure 6B). As expected, early epiblast cells showed higher expression of naïve pluripotency markers such as DNMT3L and KLF4 and lower expression of primed pluripotency marker SFRP2, as compared to late epiblast cells (Figures 6C and 6D). Interestingly, several actin polymerization related genes including those encoding for proteins in the Arp2/3 complex such as ARPC1B, ARPC5, ARPC2 were upregulated in the early epiblast but transitioned to a lower expression level in the late epiblast (Figure 6E and 6F). This is consistent with the expectation from our hiPSC epiblast findings where cells actively build a branched apical actin network at earlier stages of lumen expansion but transition to an equilibrium apical size at later stages (Figure 2I). Similar transcriptional signatures were observed in a different single cell RNA sequencing dataset 43 of 8-12 d.p.f human embryos as well (Figure 6G to 6L). Altogether, the upregulation of actin polymerization genes early in lumenogenesis are suggestive that actin polymerization may drive early lumen expansion in the human epiblast, as we have found in hiPSC epiblasts in this study.
Taken together, our experimental and simulation results uncover the physical mechanisms of lumen expansion in hiPSC epiblasts. hiPSC epiblasts closely model the morphology and phenotype of human epiblasts. We describe two distinct lumen-size-dependent mechanisms that generate the force necessary for lumen expansion. hiPSC epiblasts with lumen smaller than ~12 m radius lack mature tight junctions, allow free diffusion of ions and macromolecules between the lumen and the hydrogel, and prevent large osmotic pressures from building up. Lumen growth in these smaller epiblasts is driven by actin polymerization into a dense network on the apical surface via N-WASP, Arp2/3 and formins. Force generation by actin polymerization, aided by rapid stress relaxation in the hydrogel, ultimately drives lumen growth and overall expansion of the hiPSC epiblast in the hydrogel (Figure 7). When apical actin mesh in individual cells reaches a defined equilibrium size at a lumen radius of ~12 m, coinciding with the formation of tight junctions, the mechanism of lumen growth switches to osmotic pressure gradient driven (Figure 7). Lastly, transcriptional expression profiles of human epiblasts suggest the existence of similar mechanisms in vivo as those discovered in hiPSC epiblasts.
Mechanisms of human epiblast lumenogenesis have been difficult to explore owing to ethical concerns, technical challenges with in vitro culture of human embryos and limitations of stem-cell models of the human epiblast 18 . Matrigel or reconstituted basement membrane dependent models of the human epiblast provide a limited window into lumen growth as cells quickly differentiate in culture 24,27 , thus only allowing the study of early polarization and lumen opening events 19,25,30 . In mice, while peri-implantation development is significantly different from humans, a few different mechanisms have been suggested to play a role in epiblast lumenogenesis: (i) electrostatic repulsion between apically deposited anti-adhesive proteins such as podocalyxin which have a high negative charge 34 and (ii) lumenal fluid transport due to osmotic gradients 44 .
However, these mechanisms do not provide complete physical explanations for sustained lumen expansion. While anti-adhesive proteins such as podocalyxin can help create a "non-stick" apical surface, electrostatic repulsive forces are negligible on a length scale of microns due to Debye-Hückel screening in electrolyte solutions 15 . While podocalyxin plays an important role in establishing cell polarity and potentially in lumen nucleation, direct contribution of podocalyxin to sustained lumen expansion is not expected. Directed fluid transport on the other hand does not in itself generate force for lumen expansion unless the fluid is pressurized or accompanied by other cellular force generating mechanisms. By using engineered hydrogels which provide a prolonged window into human epiblast lumen expansion, we rule out these suspected mechanisms and discover a force generating mechanism responsible for sustained lumen growth.
Here, we discovered a mechanism of actin-polymerization-driven lumen expansion in the human epiblast, which may be relevant to other contexts in development. Actin polymerization and assembly into branched networks via Arp2/3 complex is known to generate pushing forces and drive several cellular processes including lamellipodial protrusions during migration, vesicle trafficking and polarization 45,46 . Further, actin forces generated by actin polymerization at the apical surface have been proposed to drive apical surface and/or junction expansion in other systems, including Xenopus embryos 47 and Drosophila eye 48 . Such actin structures also play a vital role during early embryogenesis 46,49 . For example, expanding apical actin rings in a preimplantation mouse embryo push cells against each other, stabilizing cell-cell junctions 50 and allowing the formation of a pressurized blastocyst cavity 10 . While these actin rings help seal the mouse embryo before lumen formation 50 , the actin structures we discovered in hiPSC epiblasts serve to expand lumens and are distinct from those observed in the mouse blastocyst. We find that in hiPSC epiblasts, apical actin polymerization and formation of a branched actin mesh generates force to drive initial lumen expansion. Theoretical modeling confirms that actin polymerization forces and network stiffness are sufficient to drive lumen growth. Apical actin as well as actomyosin networks are a common feature of epithelia, but the structural features of apical actin mesh and how actin polymerization is physically aligned to drive lumenogenesis are yet to be explored. As this is the first time that actin polymerization forces have been considered as a driving force for lumen expansion, it is possible that this mechanism could be relevant to other contexts in development. Several tissues including the zebrafish gut 51 , Drosophila lung 52 and MDCK cysts 38 have lumenal surfaces that are bent inwards and have dense apical actin. Because osmotic-pressure driven growth is expected to result in bulged out lumens, osmotic pressure driven growth might not explain early lumenal growth in these other contexts.
The existence of two distinct mechanisms of lumen growth, dependent on a critical lumen size at which tight junction formation and apical maturation occur, highlights close crosstalk between the cell polarity machinery, growth of apical domains, tight junctions, and lumen size.
Cell geometries are also tightly controlled with cell volume and thickness staying roughly constant during both actin polymerization and osmotic pressure driven lumen growth. Further, cell-layer stretching and cycles of lumen inflation and collapse, which are characteristic of pressure driven lumen growth in other model systems 9,10,12,14 , are not observed in hiPSC epiblasts, suggesting that pressures generated in the epiblast are relatively low. While high pressures are useful for disrupting structures such as zona pellucida during blastocyst hatching 17 , they could cause tissue rupture 10,12 and compromise embryo integrity. Moreover, lumen volumes and cell numbers closely follow a power law relationship during pressure driven growth, highlighting a careful balance between pressure magnitude and cell number, thereby allowing cells to maintain a fixed volume and thickness. Such control over embryo size could be pivotal for subsequent embryonic patterning events such as amnion formation and gastrulation by establishing appropriate signaling gradients 4,5,44,53 . Notably, we did also find mechanisms of robustness in this system. While inhibition of actin polymerization disrupts early lumen formation, lumens ultimately form as cells multiply even with continuous actin inhibition, highlighting the presence of alternate mechanisms of epiblast lumenogenesis to ensure robust development, but such a delay in lumenogenesis could possibly impact other developmental processes. Overall, our results provide a quantitative understanding of the mechanisms that drive epiblast lumenogenesis and suggest that size control and robustness are inherently coded into these mechanisms.
While 3D live imaging of apical surface would improve visualization of actin polymerization driven lumen growth, live imaging of hiPSC epiblasts with high z-resolution was not possible due to phototoxicity effects upon increased laser exposure. hiPSC epiblasts replicate many features of the human epiblast, but a few key differences remain. For example, our epiblast model does not undergo amnion formation, an event that coincides with epiblast cavity growth. Moreover, our model lacks extraembryonic cells and their corresponding biochemical signals. The role of extraembryonic cells and mechanisms of amnion formation require further investigations. Also, several new questions arise about the function of the epiblast lumen. While the epiblast lumen has been suggested to shield the epiblast cells from extraembryonic signaling to ensure robust gastrulation 54 , the existence of size-dependent mechanisms of lumen growth points to a larger role for the epiblast lumen in regulating embryo size and orchestrating embryonic development. To our knowledge, the actin polymerization driven lumen growth mechanism discovered here is the only force-generating, pressure-independent lumenogenesis mechanism and could be at play in other model systems as well. Finally, our theoretical model assumes that 3D lumens retain their symmetric spherical shape during both stages of expansion, allowing us to simplify our analysis to the 2D geometry of the mid-plane cross-section to elucidate the underlying mechanisms. However, the modeling of elongated lumens will require inclusion of more sophisticated geometries and consideration of stress anisotropy. In conclusion, this study advances our understanding of human embryonic development and expands our knowledge of the biological toolkits that cells utilize to make a lumen.
Figure 1: hiPSC epiblasts model the human epiblast and osmotic pressure does not drive lumenogenesis in hiPSC epiblasts. (A) Schematic of peri-implantation human embryos and hiPSC epiblasts. (B) Immunostains of human and hiPSC epiblasts. Human epiblast images were generated in and modified with permission from ref. 32 . (C) Immunostains of Oct4, Sox2, Nanog, Otx2 (formative pluripotency), Ezrin and Podocalyxin (apical polarity) in hiPSC epiblasts. (D) Quantification of cell numbers in human 32 and hiPSC epiblasts (mean s.d.; ns: not significant p > 0.05, one-way ANOVA; n = 43 (day 3), 26 (day 5), 32 (day 7) hiPSC epiblasts; N = 3 biological replicates). Data for human epiblasts was generated in ref. 32 . (E) Correlation between lumen volume and cell numbers in human 20,24,31,32,55 (A). (C) Spearman correlation values between lumen area and cell apical lengths for smaller epiblasts. Rows indicate each epiblast's apical lengths correlated to corresponding lumen area. Spearman correlation values for each pair are listed and range between: 1 (perfect correlation), 0 (no correlation), and -1 (perfect anti-correlation). For Spearman correlation values > 0.5, p-value is < 0.05. Cells are listed in decreasing order of their respective Spearman correlation values. (D) Percent cells per epiblast whose apical lengths are positively correlated (Spearman correlation value > 0.5) with lumen area (mean s.d.). (E) Representative matrix deformations generated during lumen growth in smaller epiblasts and quantification of radial asymmetry in matrix deformation (mean s.d). Asymmetry index = magnitude of vector sum of deformations / average magnitude of deformations. (F) Apical actin ablation and recovery. Outlines indicate ablated surface. (G) Kymograph showing apical actin of ablated and neighboring cells for epiblast shown in (F). Length of colored lines = apical length; intensity of colored lines = average apical actin intensity. (H) Quantification of apical length and actin intensity of ablated and neighboring cells for smaller epiblasts. (I) Quantification of post-ablation and final apical lengths of ablated and neighboring cells, and lumen area for smaller epiblasts. Apical lengths of the two neighbors were averaged (mean s.d.; ***p < 0.001, ns: not significant p > 0.05, Mann-Whitney; n = 8; N = 4 biological replicates). (J) Representative cross-sectional immunostains of smaller and larger epiblasts. (K) 3D reconstruction of lumenal surface and quantification of Gaussian curvature for epiblasts shown in (J). (L) Frequency distribution of Gaussian curvature for epiblasts shown in (J). Scale bars: 10 m (A and K), 20 m (E, F and J), 1 m (G). Hydrogel is considered linear elastic for this model with a modulus of 1 kPa which is the relaxed modulus of viscoelastic alginate hydrogels used in experiments. 𝐿 ̃a: initial length of apical actin mesh; La: final length of apical actin mesh; a: undeformed length factor; Ka: stiffness of apical actin mesh; Ja: ion flux into lumen; Jleak: ion leak along the leaky junctions; Pgel: stress exerted by the hydrogel on the cell cluster. (F) Model shows that osmotic pressure does not build in leaky lumens. Ions pumped into the lumen diffuse out along the leaky junctions in smaller epiblasts preventing buildup of osmotic pressure. (G) Cells actively polymerize apical actin that generates stress to deform the hydrogel and drive lumen expansion. Dashed line indicates spontaneous lumen opening. Cell volume and cell thickness or cleft length are assumed to be constant. (H and I) Plots showing that increase in apical actin stiffness (Ka) and decrease in undeformed length factor (a) result in higher lumen growth rates. Lumen growth rates as a function of apical actin stiffness (Ka) for given values of undeformed length factor (a) are shown in (I). All other parameters were kept constant. Ha = hb = 0.6 m, Rcell = 6 m, La/𝐿 ̃a = 12, Lb/𝐿 ̃b = 1. (J) Buckling pressure considerations predict a transition to pressure-driven lumen growth at ~12 m lumen radius to prevent cell layer buckling. during lumen growth for epiblast shown in (A). (C) Spearman correlation values between lumen area and cell apical lengths for larger epiblasts. Rows indicate each epiblast's apical lengths correlated to corresponding lumen area. Cells are listed in decreasing order of their respective Spearman correlation values. (D) Percent cells per epiblast whose apical lengths are positively correlated (Spearman correlation value > 0.5) with lumen area (mean s.d.; **p < 0.01, Mann-Whitney; n = 5 (smaller epiblasts), 5 (larger epiblasts); N = 5 biological replicates). (E) Representative matrix deformations generated during lumen growth in larger epiblasts. (F) Quantification of maximum matrix deformation per hour (mean s.d.; **p < 0.01, Mann-Whitney; n = 8 (smaller epiblasts), 7 (larger epiblasts); N = 5 biological replicates). (G) Quantification of radial asymmetry in matrix deformation (mean s.d.; ***p < 0.001, Mann-Whitney; n = 8 (smaller epiblasts), 7 (larger epiblasts); N = 5 biological replicates). Asymmetry index = magnitude of vector sum of deformations / average magnitude of deformations. (H) Apical actin ablation and recovery. Red outline indicates ablated surface. (I) Kymograph showing apical actin of ablated and neighboring cells for epiblast shown in (H). (J) Quantification of apical length and actin intensity of ablated and neighboring cells for larger epiblasts. (K) Quantification of post-ablation and final apical lengths of ablated and neighboring cells, and lumen area for larger epiblasts. Apical lengths of the two neighbors were averaged (mean s.d.; **p < 0.01, ns: not significant p > 0.05, Mann-Whitney; n = 5; N = 4 biological replicates). (L and M) Schematic and predictions of the theoretical model of osmotic pressure driven lumen growth. Ion pumping into the lumen builds osmotic pressure as ions cannot diffuse out along tight junctions. ccell is the total concentration of ions in the cells at equilibrium and clum is the total concentration of ions in the lumen at equilibrium. (N) Model predictions closely match experimental observations of lumen radius, number of cells and cell apical surface area of larger epiblasts. Scale bars: 20 m (A, E and H), 1 m (I). STAR★Methods KEY RESOURCES
TABLE REAGENT or RESOURCE SOURCE IDENTIFIER Antibodies Rabbit Polyclonal Anti-Oct4 Cell Signaling Technology Cat#2750; Lot#5; RRID:AB_823583 Rabbit Monoclonal Anti-Sox2 Cell Signaling Technology Cat#3579; Clone D6D9; Lot#8; RRID:AB_2195767 Rabbit Monoclonal Anti-Nanog Cell Signaling Technology Cat#4903; Clone D73G4; Lot#8; RRID:AB_10559205 Goat Polyclonal Anti-Otx2 R&D Systems Cat#AF1979; Lot#KNO0922011; RRID:AB_2157172 Mouse Monoclonal Anti-Ezrin Sigma-Aldrich Cat#E8897; Clone 3C12; Lot#049M4838V; RRID:AB_476955 Mouse Monoclonal Anti-Podocalyxin R&D Systems Cat#MAB1658; Clone 222328; Lot#JKW0219121; RRID:AB_2165984 Mouse Monoclonal Anti-ZO-1 Thermo Fisher Scientific Cat#33-9100; Clone ZO1-1A12; Lot#TL277395; RRID:AB_2533147 Rabbit Polyclonal Anti-phospho-myosin light chain 2 Cell Signaling Technology Cat#3674; Lot#5; RRID:AB_2147464 Mouse Monoclonal Anti-Arp2/3 complex Sigma-Aldrich Cat#MABT95; Clone 13C9; Lot#3574550; RRID:AB_11205567 Rabbit Polyclonal Anti-N-WASP Thermo Fisher Scientific Cat#PA5-52198; Lot#WK3443175A; RRID:AB_2644914 Rabbit Polyclonal Anti-Active Caspase-3 R&D Systems Cat#AF835SP; Lot#CFZ4223041; RRID:AB_2243952; Alexa Fluor 555 Goat Anti-Rabbit IgG Thermo Fisher Scientific Cat#A21428; Lot#2192278; RRID:AB_2535849 Alexa Fluor 647 Goat Anti-Rabbit IgG Thermo Fisher Scientific Cat#A21244; Lot#2390713; RRID:AB_2535812 Alexa Fluor 555 Donkey Anti-Goat IgG Thermo Fisher Scientific Cat#A21432; Lot#1697092; RRID:AB_2535853 Alexa Fluor 555 Goat Anti-Mouse IgG1 Thermo Fisher Scientific Cat#A21127; Lot#2384708; RRID:AB_2535769 Alexa Fluor 647 Goat Anti-Mouse IgG1 Thermo Fisher Scientific Cat#A21240; Lot#2482960; RRID:AB_2535809 Alexa Fluor 555 Goat Anti-Mouse IgG2a Thermo Fisher Scientific Cat#A21137; Lot#2335727; RRID:AB_2535776 Chemicals, peptides, and recombinant proteins ProNova UP VLVG alginate NovaMatrix Cat#4200501; Batch#BP-1212-24; Batch#BP-1903-04 MES hydrate Sigma-Aldrich Cat#M8250 Sodium chloride Fisher Scientific Cat#S671 RGD peptide (GGGGRGDSP) Peptide 2.0 Custom order N-hydroxysulfosuccinimide (Sulfo-NHS) Thermo Fisher Scientific Cat#24510 N-(3-dimethylaminopropyl)-N'-ethylcarbodiimide hydrochloride (EDC) Sigma-Aldrich Cat#E6383 Hydroxylamine hydrochloride Sigma-Aldrich Cat#255580 DMEM/F-12 Thermo Fisher Scientific Cat#11330057 Calcium sulfate dihydrate Sigma-Aldrich Cat#C3771 hESC-qualified Matrigel LDEV-free Corning Cat#354277 mTeSR1 STEMCELL Technologies Cat#85850 ROCK inhibitor Y-27632 STEMCELL Technologies Cat#72304 Accutase STEMCELL Technologies Cat#07920 LookOut Mycoplasma PCR Detection Kit Sigma-Aldrich Cat#MP0035 Paraformaldehyde Alfa Aesar Cat#43368-9M Sucrose Fisher Scientific Cat#S5-3 O.C.T. Compound Tissue-Tek Cat#23-730-571 DPBS Fisher Scientific Cat#21-600-010 DPBS containing Ca 2+ and Mg 2+ Cytiva Cat#SH30264.01 Triton X-100 Sigma-Aldrich Cat#T8787 Cytochalasin D Thermo Fisher Scientific Cat#PHZ1063 Bovine serum albumin (BSA) Sigma-Aldrich Cat#A4503 Goat serum Gibco Cat#16210072 Glycine Fisher Scientific Cat#G46-1 DAPI Thermo Fisher Scientific Cat#D1306 Alexa Fluor 488 Phalloidin Invitrogen Cat#A12379 Alexa Fluor 555 Phalloidin Invitrogen Cat#A34055 ProLong Gold antifade reagent Life Technologies Cat#P36930 Ethylenediaminetetraacetic acid (EDTA) Sigma-Aldrich Cat#E9884 SYTOX Green Nucleic Acid Stain Thermo Fisher Scientific Cat#S7020 3 kDa Texas Red dextran Thermo Fisher Scientific Cat#D3328 10 kDa Texas Red dextran Thermo Fisher Scientific Cat#D1828 40 kDa Texas Red dextran Thermo Fisher Scientific Cat#D1829 70 kDa Texas Red dextran Thermo Fisher Scientific Cat#D1830 Dimethyl sulfoxide (DMSO) Fisher Scientific Cat#BP231-100 Blebbistatin (myosin II inhibitor) Abcam Cat#ab120425 ML-7 (myosin light chain kinase inhibitor) Tocris Bioscience Cat#4310 ML-141 (selective Cdc42 Rho family inhibitor) Tocris Bioscience Cat#4266 CK-666 (Arp2/3 inhibitor) Sigma-Aldrich Cat#SML0006 SMIFH2 (formin inhibitor) Sigma-Aldrich Cat#S4826 Wiskostatin (N-WASP inhibitor) Abcam Cat#ab141085 Cytochalasin D (actin polymerization inhibitor) Sigma-Aldrich Cat#C8273 SiR-actin Cytoskeleton Inc. Cat#CY-SC001 N-acetyl-L-cysteine (NAC) Sigma-Aldrich Cat#A9165 1 µm diameter fluorescent carboxylate-modified microspheres (FluoSpheres) Thermo Fisher Scientific Cat#F8816 CellEvent Caspase-3/7 GR Thermo Fisher Scientific Cat#C10432 Z-VAD-FMK (pan-caspase inhibitor) AAT Bioquest Cat#13300 Deposited data Single-cell RNAseq of human in vitro cultured embryos Xiang et al. 42 GEO: GSE136447 Single-cell RNAseq of human in vitro cultured embryos Zhou et al. 43 GEO: GSE109555 Experimental models: Cell lines Human: RiPSC.BJ iPSC line generated through synthetic mRNA reprogramming of BJ human fibroblast cells Durruthy-Durruthy et al. 57 ; Laboratory of Vittorio Sebastiano N/A Human: AICS-0024 iPSC line (WTC11) AICS AICS-0024; RRID:CVCL_JM15 Human: PODXL-EGFP hESC line Taniguchi et al. 30 ; Laboratory of Kenichiro Taniguchi N/A Software and algorithms Original code This paper https://github.com/aza kharov1/Lumen All original code has been deposited on GitHub and is publicly available as of the date of publication. URL is listed in the key resources table. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS Cell lines and culture Three hPSC lines were used in this study. First, was a hiPSC line (RiPSC.BJ) generated through synthetic mRNA reprogramming of BJ human fibroblast cells 57 (a gift from Dr. Vittorio Sebastiano (Department of Obstetrics and Gynecology, Stanford University)). Second, was a hiPSC line purchased from Coriell Institute (AICS-0024) in which MYH10 has been endogenously tagged with mEGFP using CRISPR/Cas9 technology in WTC-11 (GM25256) hiPSCs. Third, was a hPSC line expressing PODXL-EGFP generated using H9 human embryonic stem cells (hESCs) 30 was thawed, cultured as described above and encapsulated in alginate hydrogels without continued passaging. This was done to ensure high-quality of hiPSCs and to maintain hiPSCs at a low passage number and normal karyotype as characterized previously 57 . Both hiPSC lines were checked for mycoplasma contamination and tested negative (LookOut Mycoplasma PCR Detection Kit, Sigma-Aldrich MP0035).
Sodium alginate rich in guluronic acid blocks was purchased (ProNova UP VLVG; 28 kDa molecular weight; NovaMatrix). RGD (arginine-glycine-aspartate) peptides were coupled to alginate using carbodiimide chemistry 61 (Thermo Fisher Scientific 11330057). Reconstituted 3% (w/v) alginate was diluted with solution containing cells and crosslinked using calcium sulfate (Sigma-Aldrich C3771) to make hydrogels with 2% (w/v) final alginate concentration, initial elastic modulus of 20 kPa, loss tangent of ~0.08, stress relaxation half-time of ~70 s, and 1500 M RGD density as described previously 27 . Detailed protocols for RGD-conjugation of alginate, preparation of alginate hydrogels and exact recipes for alginate hydrogels used in this study have been published previously 27,62 .
Compression tests were performed using an Instron 5848 MicroTester to quantify the initial elastic modulus and stress relaxation behavior of the alginate hydrogels. Alginate disks of 2 mm thickness and 4 mm diameter were prepared and equilibrated in DMEM/F-12 for 24 hr.
Unconfined compression tests were then performed on alginate disks using a 4 mm diameter cylindrical probe. Gels were compressed from 0 to 10% compressive strain at a deformation rate of 1 mm per min. 10% compressive strain was then maintained for 1 hr and the corresponding stress was measured over time (stress relaxation test). To calculate the initial elastic modulus, a straight line was fitted to stress vs. strain data for the initial strain ramp between 5% and 10% compressive strain. The slope of this linear fit was reported as the initial elastic modulus. Next, to quantify the stress relaxation behavior at 10% compressive strain, the time at which relaxation modulus drops to half of its initial value was measured and reported as τ1/2. Final relaxed modulus during the stress relaxation test (~1 kPa) was taken as the effective hydrogel stiffness for cellular processes (including lumen growth) that were much slower than the hydrogel stress relaxation half-time (~70 s).
To make hydrogels, appropriate volumes of 3% (w/v) alginate and dissociated single hiPSCs were added to a luer lock syringe (Cole-Parmer). In a second syringe, appropriate volumes of calcium sulfate and serum-free DMEM/F-12 were added. The two syringes were connected with a coupler and the solutions were mixed by passing them back and forth six times. The mixture of cell, alginate and calcium sulfate solution were either directly deposited into an 8-well Lab-Tek chamber slide (Thermo Fisher Scientific) or onto a hydrophobic glass plate which was then covered with another glass plate with a 1 mm spacer between plates. The cell alginate mixture was then allowed to gel for 30 mins at room temperature. Hydrogels had a final alginate density of 2% (w/v), cell concentration of 1 million per mL of hydrogel and calcium concentration of 33 mM.
Hydrogels were punched out using a 6 mm diameter biopsy punch, immersed in mTeSR1 media with 10 μM ROCK inhibitor (Y-27632, STEMCELL Technologies) to prevent dissociationinduced apoptosis and transferred to an incubator at 37°C and 5% CO2. 24 hr post encapsulation, media was changed to mTeSR1, which was replenished daily.
hiPSCs cultured in hydrogels were fixed with 4% paraformaldehyde (PFA, Alfa Aesar) for 1 hr, and then washed three times with DPBS containing Ca 2+ and Mg 2+ (cPBS, Cytiva). For immunostaining of whole hiPSC epiblasts, these hydrogels were immersed in cPBS and stored at 4°C until used. For immunostaining two-dimensional sections of hydrogels, gels were placed in 30% (w/v) sucrose (Fisher Scientific) overnight and then transferred to 50% (v/v) of 30% (w/v) sucrose and OCT compound solution (Tissue-Tek) for 5 hr. The hydrogel was then embedded in OCT, frozen, and sectioned at 40 μm thickness using a cryostat (Leica CM1950). For immunostaining of whole hiPSC epiblasts, PFA-fixed hiPSC epiblasts were first recovered from alginate hydrogels by incubating the hydrogel for 5 mins at room temperature in 50 mM EDTA (ethylenediaminetetraacetic acid) solution. EDTA chelates calcium ions and dissolves the alginate hydrogel, following which mild centrifugation was performed to collect the hiPSC epiblasts. Next, for immunostaining, the same protocol as above was followed with slight modifications: (i) all steps were performed in ultra-low attachment 24-well plates (Corning), (ii) longer incubation -1 hr permeabilization, 3 hr blocking and overnight secondary antibody steps.
Images were acquired on a Leica SP8 laser scanning confocal microscope using a HC FLUOTAR L 25×/0.95 NA water immersion objective.
Caspase-3/7 staining hiPSCs in alginate hydrogels were incubated with SYTOX Green Nucleic Acid Stain (1 μM; Invitrogen, Thermo Fisher Scientific S7020) for 1 hr at 37°C and imaged on a Leica SP8 laser scanning confocal microscope with a HC FLUOTAR L 25×/0.95 NA water immersion objective. For Caspase-3/7 staining, hiPSCs in alginate hydrogels without (control) or with a pan caspase inhibitor (20 M Z-VAD-FMK; added from Day 1 of culture onwards) were incubated with CellEvent Caspase-3/7 reagent (2 μM; Invitrogen, Thermo Fisher Scientific C10432) for 30 min at 37°C and imaged on a Nikon Ti2 spinning disk confocal microscope with a CFI Plan Fluor DLL 10×/0.3 NA air objective or a CFI Apo LWD Lambda S 40×/1.15 NA water immersion objective.
Cell impermeable fluorescent dextran was used to assess tight junction permeability in hiPSC epiblasts. Dextrans with the following molecular weights were used: 3 kDa (Thermo Fisher Scientific D3328), 10 kDa (Thermo Fisher Scientific D1828), 40 kDa (Thermo Fisher Scientific D1829) and 70 kDa (Thermo Fisher Scientific D1830). All these dextrans are conjugated with Texas Red fluorophore and are zwitterionic. For permeability assay, dextran dissolved in mTeSR1 at a final concentration of 10 M was added to the hydrogel and incubated at 37°C for 1 hr. hiPSC epiblasts were then imaged using Leica SP8 laser scanning confocal microscope with a HC FLUOTAR L 25×/0.95 NA water immersion objective.
To quantify tight junction permeability, outlines were drawn manually around the lumen boundary, inside a cell and in the hydrogel using both the fluorescent dextran and brightfield images to measure dextran intensity in the lumen, cell, and hydrogel using ImageJ (NIH). Cell dextran intensity normalized to that in the hydrogel was consistently ~0.1 confirming that dextrans were cell impermeable. Lumenal dextran intensity was then normalized to that in the hydrogel, called normalized lumenal dextran intensity (NLDI), as a measure of tight junction permeability.
Lumen size and shape metrics were measured in ImageJ. Lumen radius was determined from measured lumen area assuming a perfect circle.
𝑀𝑒𝑎𝑠𝑢𝑟𝑒𝑑 𝑙𝑢𝑚𝑒𝑛 𝑎𝑟𝑒𝑎 𝜋 [1] Fluorescence recovery after photobleaching (FRAP) studies
For quantifying diffusion dynamics in smaller epiblasts which lacked tight junctions, fluorescence recovery of dextran was observed after photobleaching on a Zeiss LSM 780 laser scanning confocal microscope using a LCI PLAN NEO 25×/0.8 NA oil immersion objective at 37°C and 5% CO2. hiPSCs were incubated with 3 kDa dextran for 1 hr before imaging as described in the previous section. For photobleaching, lumen boundary was manually outlined and excited with a micro-point laser at 594 nm (as excitation peak of Texas Red-labelled-dextran is 595 nm) to bleach lumenal dextran using 100 scan iterations of 100% laser power (max power: 0.194 mW).
Images were taken before and after bleaching, at 1 min intervals.
Recovery profiles obtained after photobleaching were measured as NLDI (see previous section) and normalized such that initial NLDI was 1 and post-bleach NLDI was 0. The data was then fit to a previously described experimental recovery curve which assumes bleaching of a 2D circular spot followed by free diffusion of non-bleached molecules into the bleached spot from all directions 56 :
where 𝐼 0 and 𝐼 1 are modified Bessel functions of the first kind of zero and first order, k is the mobile fraction and 𝜏 𝐷 is the characteristic diffusion time. 1/2 (diffusion half-time) was calculated as the time at which fluorescence recovers to half the final equilibrated value. While this model accurately fit the experimental data (R 2 > 0.99), the assumptions of this model are not completely valid for lumenal bleaching as dextran can diffuse into the lumen only via intercellular spaces and not along all radial directions. Thus, the diffusion coefficient obtained from 𝜏 𝐷 cannot be directly compared to diffusion coefficients predicted by the Stokes-Einstein equation.
For all small-molecule inhibition studies, the drug was added 24 hr before imaging or continuously starting on day 2 of culture (for plots labelled continuous inhibition). The inhibitors used were Blebbistatin (10 µM; Abcam ab120425, myosin II inhibitor), ML-7 (10 µM; Tocris Bioscience 4310, myosin light chain kinase inhibitor), ML-141 (2 µM; Tocris Bioscience 4266, selective Cdc42 Rho family inhibitor), Y-27632 (10 µM; STEMCELL 72304, ROCK inhibitor), CK-666 (50-100 µM; Sigma-Aldrich SML0006, Arp2/3 inhibitor), SMIFH2 (10-20 µM; Sigma-Aldrich S4826, formin inhibitor) and Wiskostatin (5-10 µM; Abcam ab141085, N-WASP inhibitor). All drugs were dissolved in dimethyl sulfoxide (DMSO) and diluted in mTeSR1 media before adding to hiPSCs. DMSO alone was added to mTeSR1 media as a vehicle control. Percent clusters with lumen was manually quantified from brightfield images using ImageJ.
For live imaging of F-actin in hiPSC epiblasts, SiR-actin (100 nM; Cytoskeleton Inc. CY-SC001) was added to hiPSCs in hydrogels for 12 hr following manufacturer's protocol. SiR-actin did not impact F-actin or hiPSC epiblast morphogenesis (Figure S7) and has been previously
shown to not alter F-actin dynamics at a concentration of 100 nM or lower 63 . For timelapse imaging of F-actin, mTeSR1 media was supplemented with 100 nM SiR-actin (to maintain strong F-actin fluorescence) and 2.5 mM N-acetyl-L-cysteine (NAC, Sigma-Aldrich A9165; antioxidant to reduce phototoxicity). Fluorescent F-actin (excitation peak of SiR-actin is 652 nm) and brightfield images were acquired every 15 min for a total duration of 8 hr on a Leica SP8 laser scanning confocal microscope with a HC FLUOTAR L 25×/0.95 NA water immersion objective at 37°C and 5% CO2.
For quantifying timelapse data, lumen areas and apical lengths of individual cells were manually measured using ImageJ. To understand the correlation between lumen growth and increase in cell apical actin lengths for a fixed number of cells, 4 hr time windows were picked from acquired data, during which there was no change in cell number in both smaller and larger epiblasts. Correlation between lumen area and individual cell apical lengths was quantified by calculating Spearman correlation values between each pair of curves in GraphPad Prism (9.3.1).
Percent cell apical lengths positively correlated with lumen area was calculated by counting number of cells whose correlation with lumen area has a Spearman value >0.5 and dividing by total number of cells in a given epiblast. 3D timelapse imaging of hiPSC epiblasts with high zresolution was not possible due to phototoxicity effects upon increased laser exposure.
Super-resolution microscopy was performed to visualize apical actin mesh in SiR-actin stained hiPSC epiblasts, using a Zeiss Airyscan2 LSM 980 inverted confocal microscope with a LCI PLAN NEO 25×/0.8 NA oil immersion objective at 37°C and 5% CO2. Images were acquired in super-resolution mode with a voxel size of 0.0974 × 0.0974 × 0.81 µm 3 (x y z) and processed using 15 iterations of a 3D iterative joint deconvolution (jDCV) algorithm (Zeiss).
Laser ablation was performed to determine the stress state of actin in SiR-actin stained hiPSC epiblasts, using a Zeiss LSM 780 laser scanning confocal microscope with a LCI PLAN NEO 25×/0.8 NA oil immersion objective at 37°C and 5% CO2. A micro-point laser at 405 nm was used to ablate apical actin or to cut through the entire thickness of a cell in both smaller and larger epiblasts using 150 scan iterations of 100% laser power (max power: 1.85 mW). Images were taken before and after ablation, at 1 min intervals.
To measure recoil or retraction of apical actin in hiPSC epiblasts, apical lengths and apical actin intensity of ablated cell and its neighboring cells as well as lumen area were manually measured using ImageJ at each time-point: pre-ablation, immediately post-ablation and at 1 min intervals post-ablation up to 13 min. Apical lengths and apical actin intensities of ablated and neighboring cells at different timepoints were stitched together to generate kymographs of the combined apical surface of the three cells (1 ablated + 2 neighbors).
Note that laser ablation did not disrupt the integrity of the cell membrane. No indication of a punctured membrane or leakage of cytoplasmic material was observed in brightfield images during laser ablation experiments for both smaller and larger epiblasts. Ablated cells stayed alive for longer than 2 hr post-ablation and maintained their nuclear and cellular size during this period (Figures S7E and S7F). In some larger epiblasts, laser ablation resulted in drastic lumen collapse, further highlighting that lumens in larger epiblasts are pressurized (Video S5).
Description of the main hypotheses and equations of the physical models and the methods underlying the numerical simulations are provided in supplementary information.
For quantifying hydrogel deformations, 1 µm diameter fluorescent carboxylate-modified microspheres (FluoSpheres, Thermo Fisher Scientific F8816) were encapsulated in alginate hydrogels. Timelapse images of fluorescent beads were collected during lumen growth in both smaller and larger epiblasts on a Leica SP8 laser scanning confocal microscope with a HC FLUOTAR L 25×/0.95 NA water immersion objective at 37°C and 5% CO2. Acquired images were corrected for drift using an ImageJ plugin (StackReg). Next, the drift-corrected images were used to calculate matrix deformations in MATLAB by tracking beads using a particle image velocimetry algorithm (PIVlab; open source code) using three cross-correlation windows (128 × 128, 64 × 64, and 32 × 32 pixel interrogation windows). Maximum matrix deformation was selected from within ~100 µm 2 around the cells. Radial asymmetry of matrix deformations (asymmetry index) was quantified by taking the ratio of magnitude of vector sum of deformations to average magnitude of deformations. Finally, note that the direct estimation of forces from matrix deformations is challenging due to hydrogel viscoelasticity and plasticity. Thus, in this study, matrix deformations were used as a proxy for force generation, as matrix deformation only occurred as a result of cellular forces in these gels.
Lumen solidity: hiPSC epiblasts were first incubated with fluorescent dextran (see previous section on tight junction permeability studies) to visualize smaller lumens. Outlines of lumen were manually drawn in ImageJ using fluorescent dextran or brightfield images. Lumen size and shape metrics including lumen solidity were measured in ImageJ. Lumen radius was determined from measured lumen area assuming a perfect circle (see equation [1]).
Nuclear area and perimeter: Nuclear area and perimeter of hiPSC and human epiblasts as well as hiPSCs in 2D culture, were respectively measured from immunostained images acquired in-house and from previously published sources 24,31,32,55,[64][65][66] . Using ImageJ, nuclear images were thresholded, smoothened using median filter (radius of 4 pixels), and processed using a Watershed algorithm (ImageJ) to separate touching nuclei. Nuclear area and perimeter were then measured in ImageJ.
Cell thickness of hiPSC epiblasts: For quantifying average thickness of cell layer in smaller and larger epiblasts, 5 lines were manually drawn per cluster from the apical to the basal surface and their corresponding lengths were measured from immunostained images of whole hiPSC epiblasts in ImageJ. These 5 lengths were averaged and reported as the average cell thickness.
For validating use of F-actin, dextran and brightfield images for demarcating lumen boundary, lumen outlines were drawn using fluorescent images of podocalyxin (PODXL-EGFP hPSC line), F-actin (SiR-actin), dextran and brightfield images and compared (Figure S2A to S2C). As expected, dextran entered lumens in smaller epiblasts but not in larger epiblasts (Figure S2A). As dextran occupies the enter lumenal volume in smaller epiblasts, we considered dextran as the most reliable marker of lumenal surface. All lumen outlines were comparable with only minor differences, suggesting that podocalyxin, F-actin, dextran and brightfield images can all be used to reliably demarcate the lumenal surface (Figure S2B). Lumen area measured using dextran, F-actin and brightfield were all similar without any statistically significant differences (Figure S2B). However, lumen area measured using podocalyxin was slightly smaller, possibly because podocalyxin, a transmembrane protein, extends into the lumenal volume (Figure S2B).
Lumen volume, number of cells, apical surface area and cell volume quantification: High resolution z-stacks of nucleus and F-actin in immunostained whole hiPSC epiblasts were acquired (see previous section on immunofluorescence of whole hiPSC epiblasts) using a Leica SP8 laser scanning confocal microscope with a HC FLUOTAR L 25×/0.95 NA water immersion objective.
Images were acquired with a voxel size of 0.0909 × 0.0909 × 0.5691 µm 3 (x y z). Lumen volume, number of cells, apical surface area of individual cells and individual cell volumes of hiPSC epiblasts were quantified from these high-resolution z-stacks using Imaris 9.9 software (Bitplane). Lumen curvature quantification: To quantify lumen surface curvature in smaller and larger epiblasts, lumen surfaces reconstructed in Imaris from high-resolution z-stacks of immunostained whole hiPSC epiblasts were exported to ImageJ. LimeSeg plugin in ImageJ was then used to compute gaussian curvature at each point of lumen surface.
To compare morphological features of human and hiPSC epiblasts, number of cells and lumen volumes of hiPSC epiblasts on different days of culture were compared to those of human epiblasts. Average number of cells in human epiblasts on different days post fertilization were previously reported 32 . To quantify the evolution of lumen volumes as a function of number of cells in human epiblasts (Figure 1E), previously published data and images were used. Lumen volumes were estimated from previously reported radius of gyration values of human epiblasts 24 , assuming human epiblasts to be a perfect sphere and individual cell volume to be ~1500 µm 3 (which was the average cell volume in hiPSC epiblasts). For these human epiblasts, number of cells were reported 24 . Next, to obtain additional data points from human epiblasts, we analyzed published 2D immunostained images of human epiblasts 20,24,31,32,55 and measured lumen area and total cell area in the mid-plane of human epiblasts. This 2D lumen area and total cell area in mid-plane were used to estimate 3D lumen volume and number of cells in the human epiblast using the following assumptions: (i) human epiblast tissue and lumen are perfectly spherical, (ii) each cell in human epiblast has a cell volume of ~1500 µm 3 (which was the average cell volume in hiPSC epiblasts).
A key limitation of this comparison is that human embryo images are only available from in vitro cultured human embryos where embryos are attached to 2D tissue culture plates, resulting in somewhat altered lumen and cell morphology.
Human embryo derived single cell RNA sequencing datasets 42,43 were analyzed using Seurat R package (v4.2.0). Unprocessed datasets were obtained from publicly available GEO repositories: GSE136447 42 and GSE109555 43 . Datasets were filtered to only include epiblast cells based on previously annotated populations in respective studies. Default setups in Seurat were used unless noted otherwise. Cells with ≤4,500 genes detected were discarded from analysis. Gene expression was calculated by normalizing raw counts by the total count, multiplying by 10,000 and performing log-transformation. Then, principal component analysis was performed in Seurat.
Cell clusters were identified by a K-nearest neighbor (KNN) clustering approach with a resolution of 0.8. Non-linear dimensionality reduction was performed using Uniform Manifold Approximation and Projection (UMAP) algorithm (dimensions 1 to 10). UMAP showed that cell clusters obtained by KNN based clustering were roughly separated based on embryo age (days post fertilization) and were annotated accordingly. Finally, average expression level of different genes of interest were plotted for each cell cluster.
Statistical analyses were performed using GraphPad Prism 9.3.1 software. List of all statistical tests used, and corresponding exact number of samples (n values) and exact P values are provided in Table S1. All Mann-Whitney tests used were two-tailed tests. Bar plots and respective error bars are defined throughout the figures. For ANOVA tests, F values and degrees of freedom (DFn: degrees of freedom between groups; DFd: degrees of freedom within groups) are provided in Table S1. P values less than 0.05 were considered statistically significant. All tests used were two-tailed unless mentioned otherwise.
In Figure 1, hiPSC epiblast immunostaining images (Figures 1B and 1C) are representative of 3 independent biological replicates. In Figure 2, images (Figures 2C, 2F, 2G, 2H and 2K) are representative of 3 independent biological replicates. In Figure 3, images (Figure 3J) are representative of ≥3 independent biological replicates. In Figure S1, hiPSC epiblast immunostaining images (Figures S1B, S1C, S1F and S1G) are representative of ≥2 independent biological replicates. In Figure S2, hiPSC epiblast immunostaining images (Figure S2K) are representative of 3 independent biological replicates. In Figure S3, hiPSC epiblast immunostaining images (Figure S3A) are representative of 3 independent biological replicates. For all experimental plots, data are pooled from 3 biological replicates unless mentioned otherwise in the figure captions. N (biological replicates) values for all plots are listed in Table S1.
Epiblast lumen formation C o n t r o l 5 0 ! M C K -6 6 6 1 0 0 ! M C K -6 6 6 C o n t r o l 1 0 ! M S M I F H 2 2 0 ! M S M I F H 2 C o n t r o l 5 ! M W is k o s t a t in 1 0 ! M W is k o s t a t in 0 20 40 60 80 100 % clusters with lumen Day 3 (24 hr inhibition) !!! !!!! !! !!!! !!! !! A Lumen opening and growth in hiPSC epiblasts require force generation 02:00 hh:mm 05:00 08:00 11:00 14:00 17:00 01:00 04:00 07:00 10:00 13:00 16:00 hh:mm Lumen radius < 12 μm Lumen radius > 12 μm B Lumen opening requires actin polymerization E L C o n t r o l 1 0 ! M B le b b is t a t in 1 0 ! M M L -7 2 ! M M L -1 4 1 1 0 ! M Y -2 7 6 3 2 0 20 40 60 80 100 % clusters with lumen Day 3 (24 hr inhibition) ns ns ns ns C o n t r o l 1 0 ! M B le b b is t a t in 1 0 ! M M L -7 2 ! M M L -1 4 1 1 0 ! M Y -2 7 6 3 2 0 20 40 60 80 100 % clusters with lumen Day 7 (24 hr inhibition) ns ns ns ns K Day 7 Day 3 Control Arp2/3 inhibition H F Actin forms a mesh-like structure apically that attains a mature size at 12 μm lumen radius Microvilli SiR-actin Lumen center Lumen edge 3x zoom Apical actin mesh G Super-resolution SiR-actin I 0 10 20 30 40 50 60 10 100 Lumen radius (!m) Apical surface area per cell (!m 2 ) 12 µm 0 4 8 12 16 20 24 10 100 Lumen radius (!m) Individual cell apical surface area (!m 2 ) 12 µm Actomyosin contractility does not drive lumen opening Formin inhibition N-WASP inhibition Day 3 Day 5 Day 7 Arp2/3 N-WASP DAPI Phalloidin Merged C o n t r o l 5 0 ! M C K -6 6 6 1 0 0 ! M C K -6 6 6 C o n t r o l 1 0 ! M S M I F H 2 2 0 ! M S M I F H 2 C o n t r o l 5 ! M W is k o s t a t in 1 0 ! M W is k o s t a t in 0 50 100 % clusters with lumen Day 7 (24 hr inhibition) ns ns ns ns ns ns 20 μm 0 μm Lumen radius < 12 μm Matrix deformation 50 μm 0 μm Lumen radius > 12 μm D a y 3 D a y 5 D a y 7 1 10 100 Lumen radius (!m) !!!! !!!! 12 µm C D Myosin II DAPI Phalloidin Merged Day 3 Day 7 J Figure A 0 2 4 6 8 10 12 14 16 0 5 10 15 0 50 100 150 Time (min) Cell apical length (!m) Cell 1 Cell 2 Cell 3 Cell 4 Cell 5 Lumen area (µm 2 ) Lumen area B F G Time (1 min steps) Ablated cell Neighbor 1 Neighbor 2 1 μm Lumen radius = 11 μm I Pre-ablation Ablated 0 min 2 min 4 min SiR-actin 6 min 8 min 10 min 12 min H P o s t -a b la t io n F in a l 0.0 0.5 1.0 1.5 Apical length normalized by initial length Ablated cell !!! P o s t -a b la t io n F in a l 0.0 0.5 1.0 1.5 Lumen area normalized by initial area ns P o s t -a b la t io n F in a l 0.0 0.5 1.0 1.5 Apical length normalized by initial length Neighboring cells !!! 0 2 4 6 8 10 12 14 0.50 0.75 1.00 1.25 1.50 Time (min) Apical length (normalized to initial) Ablated Neighbor 1 Neighbor 2 Ablated cell 0 2 4 6 8 10 12 14 0.0 0.5 1.0 Time (min) Normalized apical actin intensity Ablated Neighbor 1 Neighbor 2 Ablated cell 4 min 6 min 8 min 10 min SiR-actin 12 min 14 min Cell 5 Cell 4 Cell 3 Cell 2 Cell 1 Lumen radius < 12 μm Matrix deformation 50 μm 0 μm C E 6x zoom L u m e n r a d iu s < 1 2 ! m 0 20 40 60 80 100 % cell apical lengths positively correlated with lumen area Significant positive correlation D L u m e n r a d iu s < 1 2 ! m 0.0 0.2 0.4 0.6 0.8 1.0 Asymmetry index Pre-ablation Ablated Significant negative correlation J Gaussian curvature Lumen 3D reconstruction Large curvature DAPI Phalloidin G + -K -0 . 0 3 0 -0 . 0 2 8 -0 . 0 2 6 -0 . 0 2 4 -0 . 0 2 2 -0 . 0 2 0 -0 . 0 1 8 -0 . 0 1 6 -0 . 0 1 4 -0 . 0 1 2 -0 . 0 1 0 -0 . 0 0 8 -0 . 0 0 6 -0 . 0 0 4 -0 . 0 0 2 0 . 0 0 0 0 . 0 0 2 0 . 0 0 4 0 . 0 0 6 0 . 0 0 8 0 . 0 1 0 0 . 0 1 2 0 . 0 1 4 0 . 0 1 6 0 . 0 1 8 0 . 0 2 0 0.01 0.1 1 10 100 Gaussian curvature Frequency (%) Lumen radius < 12!m Lumen radius > 12!m Flat (small curvature) L Lumen radius <12 μm radius Lumen radius >12 μm radius Apical surfaces become flatter as lumens grow Apical actin polymerization drives lumen expansion in smaller epiblasts Figure Transition to pressure-driven growth is necessary to avoid buckling of the cell layer
Critical buckling pressure Gel pressure Stable lumens Unstable to buckling P cr ! 1.2 kPa R cr ! 12 µm
Cell geometric constraints dictate evolution of apical surface area and lumen size
Apical actin growth rate (dA apical / dR lum )
Normalized lumen size (R lum /R cell )
Number of cells
Apical actin stiffness Figure For the lumen, the change in the number of ions then reads dtN i lum = -J i a + J i leak , where the ion leak through the cleft is assumed to be proportional to the difference in concentrations between the lumen and the external solution For the lumen, the change in the number of ions then reads dtN i lum = -J i a + J i leak , where the ion leak through the cleft is assumed to be proportional to the difference in concentrations between the lumen and the external solution -4 -2 0 2 4 -4 -2 0 2 4 UMAP (daywise)
7 d.p.f 8 d.p.f 9 d.p.f 10 d.p.f 12 d.p.f 14 d.p.f A E a r ly E p ib la s t L a te E p ib la s t 0 1 2 3 4 Expression level DNMT3L !!!! E a r ly E p ib la s t L a te E p ib la s t 0.0 0.2 0.4 0.6 0.8 1.0 Expression level KLF4 ! E a rl y E p ib la s t L a te E p ib la s t 0 1 2 3 Expression level SFRP2 !!!! E a r ly E p ib la s t L a te E p ib la s t 0 1 2 3 Expression level ACTN1 !! E a r ly E p ib la s t L a te E p ib la s t 0 1 2 3 Expression level ARPC1B !!!! E a r ly E p ib la s t L a te E p ib la s t 0 1 2 3 Expression level ARPC2 ! E a r ly E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level ARPC5 !! Naive pluripotency Primed pluripotency Actin polymerization C D E -4 -2 0 2 4 -4 -2 0 2 4 UMAP Early epiblast Late epiblast sc-RNAseq of 7-14 d.p.f human embryos B KNN based cell clustering Differences in gene expression -4 -2 0 2 4 6 -10 -5 0 5 10 UMAP (daywise) 8 d.p.f 10 d.p.f 12 d.p.f G E a r ly E p ib la s t M id E p ib la s t L a t e E p ib la s t 0 1 2 3 4 Expression level DNMT3L !!!! !!!! ns E a r ly E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level KLF4 !!!! !!!! ns Naive pluripotency Primed pluripotency Actin polymerization I J K -4 -2 0 2 4 6 -10 -5 0 5 10 UMAP Early Epiblast Mid Epiblast Late Epiblast sc-RNAseq of 8-12 d.p.f human embryos H KNN based cell clustering Differences in gene expression Analysis of scRNA-seq data generated in Zhou et al., Nature, 2019 E a r ly E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 Expression level SFRP2 !!!! ns !!!! E a r ly E p ib la s t M id E p ib la s t L a t e E p ib la s t 0 1 2 3 Expression level CD24 !!!! ns !!!! E a r ly E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level ARPC1B ns ! !!!! E a r ly E p ib la s t M id E p ib la s t L a t e E p ib la s t 0 1 2 3 Expression level ARPC2 !!!! !! !!!! E a r ly E p ib la s t M id E p ib la s t L a t e E p ib la s t 0 1 2 3 4 Expression level ARPC3 !!!! ns !!!! E a r ly E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level ARPC5 ! ns ns Other relevant genes L E a r ly E p ib la s t M id E p ib la s t L a te E p ib la s t 0 1 2 3 4 5 Expression level ACTB ns ns ns E a rl y E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level ACTN1 ns ns ns E a rl y E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 Expression level ACTR2 ns ns ns E a rl y E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level ACTR3 ns ns ns E a r ly E p ib la s t M id E p ib la s t L a te E p ib la s t 0 1 2 3 4 Expression level EZR !!!! !!!! ns E a rl y E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level MYH9 ns ! ! E a rl y E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level MYH10 ns ns ! E a r ly E p ib la s t M id E p ib la s t L a te E p ib la s t 0 1 2 3 4 Expression level PODXL ns ns ns E a rl y E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level RAC1 ns ns ns E a rl y E p ib la s t M id E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level WASL ns ns ns Analysis of scRNA-seq data generated in Xiang et al., Nature, 2020 Other relevant genes F E a r ly E p ib la s t L a te E p ib la s t 0 1 2 3 4 5 Expression level ACTB ns E a r ly E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 Expression level ACTR2 ns E a r ly E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 Expression level ACTR3 ns E a r ly E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 Expression level ARPC3 ns E a r ly E p ib la s t L a te E p ib la s t 0 1 2 3 4 Expression level EZR ns E a r ly E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 Expression level MYH9 !! E a r ly E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 2.5 Expression level MYH10 ns E a r ly E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 Expression level WASL ns E a r ly E p ib la s t L a te E p ib la s t 0.0 0.5 1.0 1.5 2.0 Expression level RAC1 !! Red outline indicates ablated surface. (F and H) Kymograph showing apical actin of ablated and neighboring cells for larger epiblasts shown in (E) and (G) respectively. Scale bars: 20 µm (A, C, E and G), 1 µm (B, D, F and H). Figure S7 [supporting Figs. 3 and 5]: SiR-actin does not impact F-actin or hiPSC epiblast morphogenesis and laser ablation of apical surface does not disrupt cell viability. (A and B) Representative brightfield images and quantification of percent clusters with lumen and projected Methods Fig. 2C), reaching a maximum in epiblasts with a lumen radius of ~3 µm, which is predicted to monotonically decrease as lumen grows. This spontaneous lumen opening can be easily accommodated with a small increase in the number of cells. The predicted number of cells in a spherical epiblast can be found as 𝑁 !"##$ ≈ 4𝜋𝑅 #'( / /𝐴 )*+!)# . The model's predictions for the number of cells at different lumen sizes is in excellent agreement with our experimental data (Supplementary Methods Fig. 2D). Furthermore, the model predicts that lumen expansion at earlier stages of epiblast growth requires a rapid increase in apical lengths, a process associated with dynamic formation and polymerization of the actin network.
As epiblasts are embedded in a gel that becomes compressed as the lumen expands, it is important to determine the critical lumen size at which the epiblast becomes unstable due to the pressure exerted by the gel. This critical size corresponds a configuration where the cell layer is prone to inward buckling to reduce deformations in the gel. The gel pressure is proportional to the Young's modulus of the gel Egel and the amount of strain εgel, and reads as
where 𝑅 where 𝜆 = 𝐸ℎ/(1 -𝜈 / ), 𝐾 = 𝐸ℎ 8 /(1 -𝜈 / ) are compressional and bending rigidities of the monolayer, respectively, and 𝑅 is the radius [S3]. Assuming an incompressible (𝜈 = 0.5) cell monolayer of thickness ℎ = 𝐿 !#"%& = 𝑅 !"## = 6 µm, apparent Young's modulus 𝐸 = 20 kPa (following Harris et al. [S4]) and constrained by gel with modulus 𝐸 3"# = 1 kPa, the buckling instability is predicted to occur when the lumen radius exceeds 𝑅 #'( /𝑅 !"## ≈ 1.95, corresponding to 𝑅 #'( ≈ 12 µm (Supplementary Methods Fig. 2E). This suggests that larger epiblasts are required to develop higher pressure in the lumen to counteract the force exerted by the gel and maintain their shape. However, lumens can maintain their symmetry when compression from the gel is small. length factor in the cortex, we expect that they are both increased but stiffness increase outcompetes any increase in undeformed length factor to drive lumen growth.
Overall, the above model predictions allow us to conclude that (i) apical actin polymerization, associated with actin network stiffening, may promote lumen growth in smaller epiblasts, (ii) cell size and cell contact or cleft length remain constant during lumen growth, which we verified experimentally, and (iii) in larger epiblasts, lumen pressure is required to balance the pressure from the gel to avoid buckling. To examine the lumen growth at later stages, where lumen pressure is necessary to avoid buckling, we employ a model for pressure-driven lumen growth that accounts for ion and water fluxes through the cell layer and the intercellular space and results in pressure gradients sufficient to drive lumen growth.
In the other limit, when ion pumping on apical side is greater than that on the basal side, the model predicts a reduction in cell size due to decreasing osmotic pressure in the cells. As a result, the cleft length decreases and ions easily leak from the lumen, preventing lumen formation (green region in Supplementary Methods Fig. 4C).
In the intermediate regime at high enough basal ion pumping that is balanced by leakage and pumping ions out of the cell through the apical side, the lumen size is shown to be strongly dependent on the ratio between active transport on both sides. Interestingly, lumens are predicted to become larger with increasing pumping through the basal side because it intensifies the passive ion leak into the lumen, and it also increases the size of cleft and cell. Whereas dependence of lumen growth on pumping across the apical side is also influenced by the pumping rate through basal side. When basal ion pumping is relatively small, lumens are predicted to become larger with increase in apical pumping. However, at higher basal pumping levels, lumens become smaller with increase in apical pumping.
In order to capture the experimental observations in large hiPSC epiblasts, our model was employed to predict lumen growth in a scenario characterized by a substantial increase in the number of cells. For that, we fixed the number of cells and allowed the system to grow to an equilibrium size while keeping the rest apical and basal lengths constant (the rest lengths in undeformed state correspond to the case with 8 cells in the cluster). In Fig. 4D of Supplementary Methods, we present the model predictions for the number of cells and cell apical area depending on the lumen radius. The simulation results are in an excellent agreement with our experimental observations and indicate that in large epiblasts, cells necessitate rapid proliferate, even though the apical area of individual cells remains relatively constant. This behavior sharply contrasts with that in smaller epiblasts where cells require intensive growth and stiffening at the apical side to facilitate lumen expansion. This again indicates that in matured epiblasts, apical growth alone is insufficient to drive lumen growth, and that there must be lumenal osmotic pressure that surpasses the pressure exerted by the surrounding gel and also promotes cell division due to the stretching in the cell layer. Physical parameters used in our model are listed in Supplementary Methods Table 1.
In our experiments, measured cell volume lies in the range of 1000 -1500 𝜇𝑚 8 , from which we ascertain lower value 𝑅 !"## = 6 𝜇𝑚 for cells of spherical shape. Assuming that epiblast development starts with a single cell, the initial epiblast size can be estimated as 𝑅 Assuming solvent permeability of the lipid membranes lies within the range [S9] of 10 AP -10 AQ 𝑚/𝑠 , the upper value for water permeability factor can be estimated as 𝛬 ( ≈ 7 × 10 AB/ .