A hydraulic feedback loop between mesendoderm cell migration and interstitial ﬂuid relocalization promotes embryonic axis formation in zebraﬁsh

Interstitial ﬂuid (IF) accumulation between embryonic cells is thought to be important for embryo patterning and morphogenesis. Here, we identify a positive mechanical feedback loop between cell migration and IF re-localization and ﬁnd that it promotes embryonic axis formation during zebraﬁsh gastrulation. We show that anterior axial mesendoderm (prechordal plate [ppl]) cells, moving in between the yolk cell and deep cell tissue to extend the embryonic axis, compress the overlying deep cell layer, thereby causing IF to ﬂow from the deep cell layer to the boundary between the yolk cell and the deep cell layer, directly ahead of the advancing ppl. This IF relocalization, in turn, facilitates ppl cell protrusion formation and migration by opening up the space into which the ppl moves and, thereby, the ability of the ppl to trigger IF relocalization by pushing against the overlying deep cell layer. Thus, embryonic axis formation relies on a hydraulic feedback loop between cell migration and IF relocalization.


INTRODUCTION
Interstitial fluid (IF) accumulations are commonly observed in early embryonic tissue. In some species, such as mice, these accumulations eventually coalesce into a large central lumen, the blastocoel, 1,2 during blastula stages, while in others, such as zebrafish, they remain as small interstitial gaps distributed over the embryo. 3,4 The function of IF accumulations in embryogenesis has only begun to be elucidated. Recent studies in mice have suggested that long-distance fluid exchange between hundreds of micron-sized lumina within the blastoderm leads to the formation of a single dominant lumen, the forming blastocoel, and that this process is important for initial embryo symmetry breaking. 2 Furthermore, the expansion of the blastocoel in mice, driven by an increase in luminal pressure, has been proposed to promote junction maturation between trophectoderm cells lining the blastocoel, thereby helping trophectoderm cells to sustain increased pressure within the blastocoel. 1 The mouse blastocoel has also been suggested to function as a source of morphogen signaling, influencing cell-fate specification in the surrounding cells. The release of fibroblast growth factor 4 (FGF4) into the mouse blastocoel influences epiblastto-primitive-endoderm lineage segregation within the inner cell mass during blastocyst development by promoting endoderm differentiation in cells of the inner cell mass facing the cavity. 5 Likewise, deposition of bone morphogenetic protein (BMP) into the pro-amniotic cavity, another large fluid-filled cavity forming in the early mouse embryo, together with its diffusion through narrow interstitial channels, plays a critical role in the formation of a robust BMP signaling gradient within the early mouse embryo. 6 Although these findings point to an important role of IF accumulations and their coalescence into large fluid-filled cavities for embryo patterning and differentiation, comparably little is known about whether and how IF accumulations also function in the processes by which the embryo takes shape. Gastrulation, when the different germ layers-ectoderm, mesoderm, and endoderm-are formed within the embryo, is typically accompanied by characteristic and large-scale reorganization of IF accumulations. These dynamic changes in IF accumulations are likely important for germ layer formation, as evidenced, for instance, by studies on chick and sea urchin gastrulation, where the blastocoel appears to act as a vessel into which the individual mesoderm cells ingress, thereby facilitating embryo stratification. 7,8 Furthermore, in frog gastrulation, intrablastocoelic hydrostatic pressure promotes fibronectin matrix assembly on the blastocoel roof, forming the substrate for mesoderm migration. 9 During subsequent stages of frog gastrulation, a second fluidfilled cavity, the gastrocoel or archenteron, emerges, by accumulating IF at the expense of the shrinking blastocoel. 10 This redistribution of IF between the two cavities coincides with the internalized head-mesoderm at the blastocoel roof undergoing convergence movements, leading to its compaction and stiffening, a process important for neural crest cell migration. 11 Yet, while these different findings suggest a potential function of IF accumulations in regulating cell migration during gastrulation, direct evidence supporting such function is still sparse.
In zebrafish, IF accumulations during early blastula stages are widely distributed among the deep cells (DCs) within the blastoderm. 3,4 However, unlike in other vertebrates, these accumulations do not coalesce into a large central cavity and, consequently, no clearly recognizable blastocoel is formed at the onset of gastrulation. 4 Interestingly, recent studies have shown that cultured zebrafish blastoderm explants lacking the yolk cell are capable of forming a blastocoel, suggesting that the association of the blastoderm with the yolk cell interferes with blastocoel formation within the embryo. 3,12,13 Yet, whether IF accumulations, and their dynamic redistribution, play any role in zebrafish gastrulation is still unknown.

IF relocalization coincides with axial mesendoderm internalization and migration
To investigate the potential role of IF accumulations within the blastoderm during zebrafish gastrulation, we first systematically determined how the distribution of IF changes during this period. At pre-gastrula stages (4-6 h post fertilization [hpf]), IF localization gradually changed from an initially even distribution of fluid-filled gaps between DCs to a preferential accumulation of IF at the boundary between the enveloping layer (EVL) at the surface of the blastoderm and the DCs underneath (named 'EVL-DC boundary' hereafter; 3.5-6 hpf; Figures S1A-S1D). This was followed by spatiotemporally highly coordinated relocalization of IF from the EVL-DC boundary to the boundary between the DCs and the underlying yolk syncytial layer (YSL) (named 'YSL-DC boundary' hereafter; 6-8 hpf; Figures S1A-S1D). Notably, this relocalization became first apparent at the blastoderm/germ ring margin at the time when mesendoderm progenitors began internalizing there (6 hpf, Figures 1A-1B 0 ; Video S1), and then spread from the margin to the animal pole of the blastoderm, concomitant with the animal pole-directed migration of mesendoderm progenitors ( Figures 1C and 1C 0 ; Video S1).
To test whether there might be a correlation between the relocalization of IF from the EVL-DC to the YSL-DC boundary, and the concomitant internalization and animal-directed migration of mesendoderm progenitors, we analyzed to what extent these different events are spatiotemporally coordinated. The accumulation of IF at the YSL-DC boundary closely correlated in both space and time with the disappearance of IF from the EVL-DC boundary and the appearance of mesendoderm progenitors migrating along the YSL-DC boundary toward the animal pole ( Figures 1A-1C 0 ; Video S1). Importantly, this highly orchestrated relocalization of IF was most clearly recognizable at the dorsal side of the gastrula (Figures 1A-1C 0 ; Video S1), where a compact cluster of anterior axial mesendoderm (prechordal plate [ppl]) progenitors, closely followed by highly cohesive posterior axial mesendoderm progenitors, internalizes and migrates in a straight path toward the animal pole. [14][15][16][17] In lateral and ventral regions of the gastrula, in contrast, where mesendoderm progenitors internalize and migrate as more loosely aggregated cells, 18 the relocalization of IF was much less pronounced (Figures S1E and S1E 0 ). Moreover, analyzing the total amount of IF within the blastoderm during its relocalization did not show any major changes in IF volume (Figures S1F-S1G 0 ), suggesting that this relocalization represents a redistribution of IF within the blastoderm rather than de novo accumulation at the YSL-DC boundary. Collectively, these findings point to the possibility that there might be a functional link between the relocalization of IF from the EVL-DC to YSL-DC boundary with the internalization and animal-directed migration of ppl progenitors at the onset of gastrulation.
Extracellular osmolarity controls the amount of IF accumulation within the blastoderm To address the functional relevance of the IF accumulation at the YSL-DC boundary for ppl progenitor cell migration, we sought to interfere with the overall IF accumulation within the blastoderm and analyze how this affects ppl cell migration. To this end, we asked how IF accumulation is regulated within the blastoderm. Figure 1. Interstitial fluid accumulation requires Na + /K + -ATPase activity and relocalizes during axial mesendoderm internalization and migration (A) Schematic of interstitial fluid (IF) distribution during axial mesendoderm (ME) internalization at 6 h post fertilization (hpf); D, dorsal and S, sagittal views; EVL, enveloping layer; DC, deep cell; YSL, yolk syncytial layer. (A 0 ) Schematic illustrating the principle of IF relocalization analysis at the dorsal side; AV, animal-vegetal axis; pink and blue rectangles indicate the regions of IF accumulation analysis at the YSL-DC and EVL-DC boundaries, respectively, during axial ME internalization and migration (black rectangle). (B, C, and E-J) Maximum intensity projections (D, dorsal view) and single cross-sections (S, sagittal view) of gsc::caax-EGFP embryos expressing EGFP within the axial ME and injected with fluorescent dextran to mark IF in control (B and C), ouabain-treated (E and F), ouabain-treated and mannitol-injected (G and H), and mannitol-injected (I and J) embryos at 6 hpf (B, E, G, and I), and 8 hpf (C, F, H, and J); dashed white lines outline the anterior axial mesendoderm/prechordal plate (ppl) and, due to the leaky nature of this transgenic reporter line, also the posterior axial ME (pam); dashed blue lines outline the IF at the EVL-DC boundary; dashed pink lines outline the IF at the YSL-DC boundary (S, sagittal view); scale bars, 100 mm. (B 0 , C 0 , and E 0 -J 0 ) IF distribution profiles along the AV axis relative to the position of ppl/pam, marked by gsc::caax-EGFP expression (black lines), in control (green, B 0 and C 0 ), ouabain-treated (blue, E 0 and F 0 ), ouabain-treated and mannitol-injected (yellow, G 0 and H 0 ), and mannitol-injected (red, I 0 and J 0 ) embryos at 6 hpf (B 0 , E 0 , G 0 , and I 0 ) and 8 hpf (C 0 , F 0 , H 0 , and J 0 ); multi-color curves represent average values of n = 3 independent embryo replicates of the position of ppl/pam (black lines) and IF distribution at the EVL-DC (blue lines) and YSL-DC (pink lines) boundaries; arrowheads (pink) indicate the IF accumulation ahead of the ppl. (D) Schematic illustrating the experimental design. Control and ouabain-treated (1 mM, 1-3 hpf) embryos were injected with fluorescent dextran (to mark the IF) either alone or in combination with mannitol (400 mM). See also Figures S1 and S2 and Videos S1 and S2.
The first accumulations of IF within the blastoderm became apparent at early cleavage stages (32-to 64-cell stage, Figures S2A-S2B 0 ), when cleavage orientation changes 19 and cells at the blastoderm surface polarize along their apicobasal axis ( Figure S2C). Surface cell polarization and IF accumulation were accompanied by the restricted localization of Na + /K + -AT-Pases, key regulators of IF accumulation in other organisms 20,21 to the basolateral side of surface cells ( Figure S2C 0 ), pointing at the possibility that Na + /K + -ATPases are involved in IF accumulation within the blastoderm. To test this possibility, we transiently blocked Na + /K + -ATPase activity at the onset of IF formation by exposing embryos to 1 mM of the specific Na + /K + -ATPase inhibitor, ouabain, 1,5 from 1 to 3 hpf ( Figure 1D). Treated embryos showed strongly reduced IF accumulation within the blastoderm from the beginning of IF formation in control embryos (64-to 128-cell stage, Figure S2D) to the end of gastrulation ( Figures S2J and S2K). This reduced accumulation of IF was accompanied by a transient swelling of the blastoderm cells ( Figure S2E), which, however, normalized before the onset of gastrulation movements ( Figures S2F and S2G). Reduced IF accumulations in ouabain-treated embryos displayed the same overall pattern of dynamic relocalization as found in control embryos, first to the boundary between EVL and DCs and then from there to the YSL-DC boundary, although in strongly affected embryos this was often hardly recognizable ( Figures 1E-1F 0 , S2J, and S2K; Video S2). This suggests that ouabain treatment predominantly interferes with IF accumulation per se, but not necessarily with the processes regulating its relocalization within the blastoderm.
To determine whether Na + /K + -ATPases function in IF accumulation by regulating intra-to extracellular fluid osmolarity, as found in other organisms, 20,21 we injected mannitol (36.4 ng/embryo) into the extracellular space of ouabain-treated embryos at 3.5 hpf to increase extracellular osmolarity. Mannitol injection partially rescued the overall IF accumulation in ouabain-treated embryos, thereby restoring the characteristic accumulation ahead of the ppl ( Figures 1G-1H 0 , S2L, and S2M; Video S2). This suggests that ouabain inhibits IF accumulation within the blastoderm by changing the ratio of extra-to intracellular osmolarity. Consistently, injection of mannitol into the extracellular space of 3.5-hpf control embryos increased the total amount of IF above the level typically found in control embryos but did not abolish the general pattern of IF relocalization observed in control embryos, except for some minor level of premature IF relocalization to the YSL-DC boundary in mannitol-injected embryos ( Figures 1I-1J 0 , S2N, and S2O; Video S2). Together, these findings suggest that regulating extracellular osmolarity represents an effective tool for tuning the amount of IF accumulation in early blastula and gastrula stage embryos.
IF accumulation is required for ppl cell protrusion formation and migration Being able to modulate IF accumulation within the blastoderm, we next asked whether IF accumulation in front of the advancing ppl progenitors is needed for their migration toward the animal pole of the gastrula. Embryos transiently exposed to 1 mM ouabain from 1 to 3 hpf to block IF accumulation within the blastoderm showed a pronounced thickening of the dorsal germ ring at the position where the embryonic shield forms 15   (C) Schematic illustrating the principle of ppl distance from the margin (C 0 ) and thickness (C 00 ) quantification. (C 0 ) Distance of the ppl leading edge to the germ ring margin as a function of developmental time after the onset of internalization (6 hpf, corresponding to 0 0 ) for control (green, n = 9), ouabain-treated (blue, n = 9), ouabain-treated and mannitol-injected (yellow, n = 6), and mannitol-injected embryos (red, n = 4); mean ± SEM. (C 00 ) Ppl thickness from 6 to 8 hpf in control (green, n = 4), ouabain-treated (blue, n = 4), ouabain-treated and mannitol-injected (yellow, n = 4), and mannitol-injected embryos (red, n = 4); mean ± SEM.  Figure S3 and Video S2.

OPEN ACCESS
Article IF accumulation within the blastoderm, we partially restored IF accumulation by injecting mannitol into the extracellular space in ouabain-treated embryos. Mannitol injection not only effectively restored IF accumulation ahead of the advancing ppl progenitors (Figures 1G-1H 0 ; Video S2) but also their migration away from the germ ring margin toward the animal pole (Figures 2C-2C 00 ; Video S2). In contrast, increasing IF accumulations within the blastoderm above control levels by injecting mannitol in untreated control embryos had no recognizable effect on germ ring morphology or ppl progenitor cell migration (Figures 2C-2C 00 ; Video S2). Finally, to strengthen the functional link between IF accumulation and ppl cell migration, we sought to interfere with IF accumulation within the blastoderm by blocking Na + / K + -ATPase activity through drugs other than ouabain and then determine whether such reduction in IF accumulation would also lead to defects in ppl cell migration. We found that treating embryos with 10 mM digitoxin, a specific inhibitor of Na + /K + -ATPase activity, 22 not only led to strongly reduced IF accumulation but also diminished animal-pole-directed ppl cell migration ( Figures S3F-S3J), thereby closely phenocopying the effects observed after ouabain treatment. Collectively, these findings suggest that IF accumulation in front of the internalized ppl has a permissive function, facilitating ppl cell migration toward the animal pole.
To understand why IF accumulation ahead of the advancing ppl progenitors is required for normal ppl progenitor cell migration, we analyzed ppl progenitor cell morphology and movement in ouabain-treated versus control embryos. Once internalized, cells at the leading edge of the ppl migrate toward the animal pole using the YSL as a substrate for their migration and forming various types of cell protrusions, including lamellipodia, filopodia, and blebs that adhere to the YSL by Cadherin-1 (Cdh1)-mediated  Article cell-cell adhesion 14,23 (Figures 3D and 3F). Extracellular matrix accumulation at the YSL-DC boundary, in contrast, only becomes detectable at later stages of gastrulation 24 and thus is unlikely to be directly involved in ppl cell migration. Protrusion formation in ppl cells closely coincided with the accumulation of IF in front of the internalized ppl ( Figures 3D-3G), and was strongly reduced in ouabain-treated embryos with blocked IF accumulation within the blastoderm ( Figures 3D-3G). To test whether this reduction of ppl cell protrusion formation in ouabain-treated embryos was due to ouabain blocking IF accumulations and not other potentially more direct effects of ouabain on ppl cell migration and/or protrusion formation, we plated induced anterior axial mesendoderm cells on fibronectin-coated substrates and analyzed whether ouabain would affect protrusion-driven mesendoderm cell spreading. No significant difference in cell spreading capacity could be detected between treated and control cells ( Figures S3C-S3D), suggesting that ouabain does not directly interfere with anterior axial mesendoderm cell protrusion formation and spreading. Finally, to exclude that ouabain treatment interferes with ppl cell migration by affecting ppl cell-fate specification, we analyzed the expression of hatching gland gene (hgg), a characteristic marker gene of ppl cell-fate specification, 25 in ouabain-treated and control embryos. The expression of hgg remained largely unchanged in ouabain-treated embryos ( Figure S3E), indicating that ouabain does not affect ppl cell migration by interfering with ppl cell-fate specification. Collectively, these results suggest that the accumulation of IF in front of the advancing ppl facilitates ppl cell migration by promoting cell protrusion formation.
IF relocalization requires ppl cell migration Next, we asked whether not only IF relocalization facilitates ppl cell migration but also whether ppl migration feeds back on IF relocalization. For testing this hypothesis, we determined whether IF relocalization would occur when ppl progenitor cell internalization and animal-directed migration are impaired. To this end, we first turned to maternal-zygotic one-eyed pinhead mutant embryos (MZoep), which are defective in mesendoderm specification and germ layer formation. 26 In MZoep mutants, IF largely remained at the EVL-DC boundary, and only small IF fragments appeared at the YSL-DC boundary at the time (6-7.5 hpf; Figures 3A-3C 0 ) when IF relocalization was observed in wild-type (WT) embryos or MZoep mutants rescued by injection of oep mRNA ( Figures 1A-1C 0 and S4A-S4B 0 ). At later stages of gastrulation (7-8 hpf), IF relocalization was also observed in MZoep mutants; however, different from WT embryos, this relocalization simultaneously occurred in most of the blastoderm except its margin ( Figure S4C). This suggests that mesendoderm internalization and migration are required for timely and spatially coordinated IF relocalization at the onset of gastrulation.
To further test whether the amount and migratory activity of internalized mesendoderm determine the extent of IF relocalization, we analyzed IF relocalization in cyclops (cyc) morphant embryos, where the number of ppl progenitors, and thus the size of the ppl, is reduced at varying degrees, 27,28 and in slb/wnt11 mutant embryos, where the advancement of ppl cells toward the animal pole is diminished. 25 In cyc morphant embryos, IF relocalization from the EVL-DC to the YSL-DC boundary ahead of the ppl leading edge was still detectable (6-8 hpf; Figures 3E-3F 0 ). However, the extent by which this relocalization occurred was greatly diminished, closely matching the variable reduction in the size of the ppl found in cyc morphant embryos ( Figure 3D), suggesting that the size of the ppl determines the extent of IF relocalization. Furthermore, in slb/wnt11 mutant embryos, reduced animal-pole-directed migration of ppl cells was accompanied by a reduction in the progressive relocalization of IF from the EVL-DC to the YSL-DC boundary and a less pronounced accumulation of IF ahead of the advancing ppl (Figures 3G-3I 0 ; for control see Figures 1B-1C 0 ), suggesting that the migration of ppl cells is required for progressive IF relocalization.

Ppl internalization and migration are sufficient to trigger IF relocalization
To determine whether ppl cell internalization and migration are not only required but also sufficient for IF relocalization from the EVL-DC to the YSL-DC boundary, we sought to rescue IF relocalization in MZoep mutant embryos by transplanting a cluster of ppl cells (15)(16)(17)(18)(19)(20)    Article is different from the situation in WT embryos, where IF predominantly accumulated ahead of the migrating ppl due to the presence of more posterior axial mesoderm progenitors behind the ppl in these embryos, obstructing IF accumulation there ( Figures 4E and 4E 0 ; Video S3). Notably, the transplanted ppl cell cluster, while being uniformly surrounded by IF at the YSL-DC boundary, still displayed directed migration away from the germ margin toward the animal pole ( Figure 4B; Video S3), leading to a progressive relocalization of IF from the EVL-DC to the YSL-DC boundary along the path of their migration ( Figures 4E  and 4E 0 ; Video S3). This progressive relocalization was slowed down when migratory defective slb/wnt11 ppl cells were transplanted into slb/wnt11 morpholino-injected MZoep host embryos ( Figures S4H-S4L), supporting the notion that ppl cell migration triggers this relocalization. Collectively, these findings suggest that ppl cell internalization and migration are sufficient to trigger IF relocalization from the EVL-DC to the YSL-DC boundary, but also that this relocalization does not provide positional cues required for directed ppl cell migration.
To further test whether IF accumulation around the cluster is required for animal pole-directed ppl cell migration, as found in WT embryos, we inhibited IF accumulation in host MZoep embryos by transiently exposing them to 1 mM ouabain during IF formation (1-3 hpf). When a ppl cell cluster from an untreated WT embryo was transplanted to the YSL-DC boundary of an ouabain-treated MZoep embryo, no clear IF relocalization to the YSL-DC boundary around the transplanted cell cluster was observed from shield to mid-gastrulation stages (6-8 hpf; Figures 4F-4G 0 ; Video S3). This lack of proper IF relocalization was accompanied by the transplanted ppl cells failing to efficiently migrate away from the germ margin toward the animal pole (Figure 4B; Video S3). To exclude that the inability of ppl cells to undergo directed migration in ouabain-treated MZoep embryos is not due to ouabain directly affecting the migratory capability of the transplanted ppl cells, we asked whether ppl cells can undergo directed migration when transplanted from an ouabaintreated WT donor embryo into an untreated MZoep host embryo. We found that ppl clusters from ouabain-treated donors not only properly accumulated IF at the YSL-DC boundary surrounding them ( Figures S4D-S4E 0 ) but also migrated toward the animal pole similar to non-treated transplanted ppl cells ( Figure 4B). In contrast, when ppl cells from ouabain-treated donors were transplanted into ouabain-treated MZoep hosts, IF relocalization failed ( Figures S4F-S4G 0 ) and the transplanted cells did not display directed migration toward the animal pole ( Figure 4B). Together, this suggests that IF relocalization is required for normal ppl cell migration toward the animal pole.
Finally, to determine whether the inability of transplanted ppl cells in ouabain-treated MZoep host embryos to undergo directed migration was due to defective IF relocalization to the YSL-DC boundary affecting their protrusion formation, as found in ouabain-treated WT embryos ( Figures 2E-2G), we analyzed protrusion formation of the transplanted cells. Although ppl cells transplanted into untreated MZoep host embryos formed both lamellipodia and filopodia along the YSL-DC boundary, preferentially oriented in the direction of their migration toward the animal pole, the formation of such oriented protrusions was strongly diminished in ppl cells transplanted into ouabaintreated MZoep embryos lacking IF relocalization ( Figures 4C  and 4H-4I 0 ). This suggests that, similar to the situation in ouabain-treated WT embryos, defective IF relocalization in ouabain-treated MZoep host embryos interferes with the directed migration of transplanted ppl cells by impairing protrusion formation in these cells.
Ppl triggers IF relocalization from the EVL-DC to the YSL-DC boundary by pushing against the overlying epiblast Our finding that the ppl triggers IF relocalization from the EVL-DC to the YSL-DC boundary, raises questions as to the mechanism(s) by which this highly orchestrated IF relocalization is achieved. One possibility is that the ppl, by squeezing in between the DC layer and YSL, pushes the overlying DC layer (epiblast) outward against the EVL, thereby triggering an opposite-directed inward flow of IF from the EVL-DC boundary through the DC layer to the YSL-DC boundary. To challenge this hypothesis, we first asked whether the advancing ppl indeed pushes against the overlying epiblast by analyzing the extent of DC movements along the radial axis of the epiblast adjacent to the ppl (Figures 5A-5H; Video S4). We detected a pronounced outward radial movement of DCs above the ppl, which led to DC compaction above the ppl and gradually propagated in the direction of the animal pole as the ppl was moving in the same direction ( Figures 5A-5C, 5F, and 5H; Video S4). These progressive outward radial movement and compaction of DCs were also accompanied by an indentation of the YSL/yolk cell directly below the advancing ppl ( Figures S5A-S5C), consistent with the ppl squeezing in between the DC layer and YSL, thereby deforming them. To further challenge this hypothesis, we turned to MZoep mutants lacking ppl and, consequently, timely IF accumulation at the YSL-DC boundary 26 (Figures 3A-3C 0 ). No clearly recognizable radial movement and compaction of DCs and yolk cell indentation were detected in MZoep embryos ( Figures 5D, 5E, 5G, 5H, and S5A-S5C; Video S4), further supporting the notion that these processes depend on ppl moving in between the DC layer (epiblast) and YSL toward the animal pole.
For pushing forces originating from the advancing ppl triggering an opposite-directed inward flow of IF from the EVL-DC boundary through the DC layer to the YSL-DC boundary, the DC layer ahead of the advancing ppl must remain sufficiently porous to allow for such flows to happen. To address this requirement, we analyzed DC-layer compaction ahead of the ppl during ppl cell migration (Figures S5D and S5E). We found that, while the DC layer directly above the ppl clearly underwent compaction, the DC layer ahead of the ppl remained loosely packed with many small pockets of IF accumulations (Figure S5E), compatible with the possibility of IF flowing through this layer.
To monitor whether the IF indeed flows through the DC layer from the EVL-DC to YSL-DC boundary, we injected a photoactivatable dye (DMNB [dimethoxynitrobenzyl]-caged fluorescein) into the IF and locally uncaged it with an ultraviolet (UV) laser at the EVL-DC boundary ahead of the migrating ppl at the onset of internalization ( Figure 5I). We then followed the redistribution of these locally photoactivated fluorescent tracers during ppl animal pole-directed migration (Figures 5J and S5F; Video S5). As a control, we performed the same experiment in MZoep mutant embryos lacking ppl specification/migration and proper IF relocalization 26 (Figures 3A-3C 0  and S5F 0 ). With the onset of ppl internalization and animal pole-directed migration in WT embryos, the photoactivated fluorescent tracer relocalized from its source of generation at the EVL-DC boundary to the place of IF accumulation at the YSL-DC boundary directly ahead of the migrating ppl ( Figures 5J, 5J 0 , and 5L; Video S5). This relocalization was not solely due to simple diffusion of the uncaged dye from its source of generation, as the rate of uncaged dye accumulation at the YSL-DC boundary was clearly higher in WT embryos as compared with MZoep mutants lacking ppl and IF relocalization ( Figures 5K, 5K 0 , 5L-5M 0 , S5C, and S5D; ɑ WT = 0.9367 > ɑ MZoep = 0.5116; Video S5). Together, these findings support the notion that IF flows from the EVL-DC to the YSL-DC boundary as a result of ppl internalization and migration.
Finally, we asked whether pushing by the ppl against the overlying DC layer (epiblast) would be sufficient to trigger IF relocalization. To this end, we transplanted a passivated and fluorescently labeled magnetic bead, roughly the same size as the anterior ppl (d = 40 mm), into the YSL-DC boundary of MZoep mutant embryos and determined whether pulling the bead along this boundary would trigger IF relocalization from the EVL-DC to the YSL-DC boundary ( Figure 6A). Strikingly, we found that IF accumulated at the YSL-DC boundary around the transplanted bead, both for control (unpulled) beads and beads that we subsequently pulled using a magnet, and that this IF accumulation around the transplanted bead at the YSL-DC boundary was accompanied by a reduction of IF above the bead at the EVL-DC boundary ( Figures 6B-6J 0 ; Video S6). Moreover, pulling the bead along the YSL-DC boundary led to a progressive IF relocalization from the EVL-DC to the YSL-DC boundary along its movement path, which persisted for several minutes (Figures 6E-6J 0 ; Video S6), consistent with the notion that mechanically pushing against the overlying DC layer (epiblast) is sufficient to trigger progressive and persistent IF relocalization from the EVL-DC to the YSL-DC boundary. To further challenge this model, we sought to trigger ectopic IF relocalization in relocalization-defective MZoep mutant embryos by applying local pushing forces from the outside on the EVL. Locally pushing on the EVL of MZoep mutants using a micropipette led to a partial relocalization of IF from the EVL-DC to the YSL-DC boundary ( Figures S6A-S6D), suggesting that such pushing forces are sufficient to trigger this relocalization.
Collectively, these observations suggest that the ppl, by moving in between the YSL and epiblast, pushes against the overlying epiblast, leading to IF flowing from the EVL-DC to the YSL-DC boundary just ahead of the ppl, which in turn promotes ppl cell migration.

DISCUSSION
Recent work in mouse embryos has suggested that IF accumulation within the forming blastocoel/pro-amniotic cavity can affect cell-fate specification and differentiation by exerting mechanical forces on the surrounding tissues and/or functioning as signaling hubs. 1,5,6 Our findings in zebrafish, in contrast, point to a predominantly morphogenetic function of localized IF accumulation ahead of the advancing ppl in providing sufficient space for unhindered ppl cell protrusion formation and directed migration. This not only identifies a yet unknown function of IF accumulation in embryo morphogenesis but also points to apparent differences in the developmental function of IF accumulation between zebrafish and mice. These different roles of IF accumulations in controlling cell-fate specification in mice versus cell migration in zebrafish could be due to mouse, but not zebrafish, embryos forming a large blastocoel/pro-amniotic cavity, which exposes the blastoderm-lining cells to a large fluid interface, affecting their fate specification through biochemical signaling and mechanosensation. 1,5 Further analysis of IF accumulation in other blastocoel forming and non-forming species will be needed to elucidate the origin of the functional diversification of IF accumulation in early embryogenesis.
Mechanistically, we propose that a hydraulic feedback loop between ppl cell migration and IF accumulation is achieved by the migrating ppl pushing against the overlying DC layer (epiblast), thereby building up a pressure gradient that triggers IF flow and relocalization ahead of the advancing ppl. This IF accumulation, again, promotes ppl cell migration by opening up the space required for unhindered cell migration. Ppl progenitors have previously been shown to migrate in between DCs and the YSL, using the YSL as a substrate for their migration. 29 Moreover, friction with the overlying DC layer has been shown to slow down directed ppl cell migration, 29 suggesting that the contact between ppl cells and the overlying DC layer constrains ppl cell migration. Consequently, IF accumulating at the YSLepiblast boundary will likely have an effect on ppl cell migration by simply removing physical obstacles hindering ppl cell migration. This notion is also supported by our observation that defective ppl cell migration in embryos with reduced IF accumulation can be fully rescued simply by increasing osmotic pressure and, as a result, IF accumulation, arguing against a specific function of IF in providing biochemical cues promoting ppl cell protrusion formation and migration.
An important component of the proposed positive hydraulic feedback loop is the function of ppl cell migration in triggering IF relocalization by pushing against the overlying non-internalizing DC layer (epiblast). For this to happen, pushing forces from the advancing ppl must be efficiently transmitted from the ppl to the EVL-DC boundary, and, in response to this pushing force, IF must be able to flow through the DC layer from the EVL-DC to the YSL-DC boundary. This requires the epiblast above the ppl being sufficiently rigid to allow forces being efficiently transmitted from the ppl to the YSL-DC boundary, and the DC layer ahead of the advancing ppl being sufficiently loose and porous for allowing IF to flow through. Such dual properties of the epiblast are likely to be achieved by the ppl triggering the compaction of the overlying DCs and thus enhancing DC tissue rigidity above the ppl, while ahead of the ppl the DCs would remain sufficiently loose and porous for IF to flow through ( Figures 5A-5C, 5F, 5H, S5A, and S5B; Video S4). Interestingly, the distribution of IF accumulations has previously been shown to reflect cell connectivity and, as a result, tissue rigidity within the zebrafish blastoderm. 30 It is thus conceivable that compaction of the DC layer above the ppl leads to an increase in tissue connectivity and, consequently, rigidity, which again would facilitate the transmission of forces from the advancing ppl to the EVL-DC boundary. Yet, how DC compaction translates into specific changes in tissue rigidity and porosity still needs to be investigated.
Cell migration in non-confluent tissue not only involves dynamic changes in cell-cell and cell-matrix contacts but also the reorganization of IF accumulations within the tissue. Our finding of migrating cells self-organizing the physical space required for their unhindered migration by driving large-scale IF flow and relocalization within the embryo, unravels a simple and universally applicable mechanism by which cells can migrate within the spatial confinement of non-confluent tissues.

Limitations of the study
In this study, we show that IF accumulation ahead of the advancing ppl is required for ppl cell protrusion formation and migration, suggesting that IF functions in ppl cell migration by facilitating cell protrusion formation. Although this is a likely scenario, given the functional interdependence between cell protrusion formation and migration, we have not yet addressed the potential contribution of other and/or additional mechanisms by which IF accumulation facilitates ppl cell migration, e.g., by modulating tissue rigidity through an increase in hydrostatic pressure. We also show that pushing forces from the ppl can trigger IF flow from the EVL-DC boundary to the YSL-DC boundary ahead of the ppl. Although such mechanisms would, in principle, be sufficient to explain IF accumulation ahead of the advancing ppl, there might be other potential mechanisms, such as changes in IF osmolarity due to pushing forces by the ppl modulating mechanosensitive ion channels in adjacent DCs, the contribution of which still needs to be addressed in the future by using in vivo sensors for IF osmolarity.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Materials availability
This study did not generate new unique reagents.

Data and code availability
There are no new datasets reported in this paper. Code used for image analysis and any additional information required to reanalyze the data reported in this paper are available from the lead contact upon request. d Imaging data, including raw data, reported in this paper are available from the lead contact upon request. d Code used for image analysis reported in this paper is available from the lead contact upon request. d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Fish lines and maintenance Zebrafish (Danio rerio) were housed in 28 C water (pH 7.5 and conductivity 400 mS) with a 14 h on/10 h off light cycle. All zebrafish husbandry was performed under standard conditions according to local regulations, and all procedures were approved by the Ethics Committee of IST Austria regulating animal care and usage. Wild type AB and TL strains, and Tg(gsc::EGFP-CAAX), Tg(actb2::HRAS-EGFP), MZoep, slb/wnt11f2, and MZoep;Tg(gsc::EGFP-CAAX) zebrafish lines were used in this study. Embryos were raised in either E3 medium or Danieau's buffer, kept at 28.5 or 31 C and staged according to Kimmel et al. 43 Fish were bred in the zebrafish facility at IST Austria according to local regulations, and all procedures were approved by the Ethics Committee of IST Austria regulating animal care and usage.

METHOD DETAILS
Dextran, mRNA and Morpholino injections Embryos were oriented laterally in injection molds and injected with a microinjection needle at 1-cell stage through the yolk cell into the blastoderm. 44 Microinjection needles (30-0020, Harvard Apparatus) were pulled using a P-97 needle puller (Sutter Instruments). Injections were performed using a microinjection system (PV820, World Precision Instruments). For blastoderm segmentation, the cell cytoplasm was labelled by injection of 1 nl of 1 mg/mL of Dextran-FITC (D1820, Invitrogen) or Dextran Alexa Fluor 647 (D22914, Invitrogen) in wild type (wt) and MZoep mutant embryos at the 1-cell stage. For labelling the YSL, 0.5 nL of 1 mg/mL of Dextran TMR (D1868, Invitrogen) was injected into the forming YSL at high stage (3.3 hpf). mRNA transcription was performed using the mMESSAGE mMACHINE Kit (Ambion) and all plasmids used in this study are detailed in the key resources table. The cell nuclei were labelled by injection of 50 pg H2A-mCherry mRNA into at 1-cell stage embryos. 37 For labeling cell membranes and cortical actin, 80 pg membrane-RFP, 36 50 pg of membrane-GFP, 45 and 50 pg lifeact-RFP mRNA were co-injected in 1-cell stage embryos. 38 To

Ouabain treatments
For inhibiting interstitial fluid (IF) accumulation in zebrafish embryos, embryos were exposed to 1 mM of the specific Na + /K + -ATPase inhibitor Ouabain (O3125, Sigma-Aldrich) in standard embryo medium (E3) from 0.5-3.3 hpf, after which the inhibitor was washed off, embryos were dechorionated in E3, and the remaining IF accumulations were labelled as described above (interstitial fluid labelling and manipulation of interstitial fluid amount).

Digitoxin treatments
For inhibiting interstitial fluid (IF) accumulation in zebrafish embryos, embryos were exposed to 10 mM of theNa + /K + -ATPase inhibitor Digitoxin (D5878, Sigma-Aldrich, dissolved in dimethylsulfoxide (DMSO)) in standard embryo medium from 0.5-3.3 hpf, after which the inhibitor was washed off, embryos were dechorionated in E3, and the remaining IF accumulations were labelled as described above (interstitial fluid labelling and manipulation of interstitial fluid amount). For control, 1% DMSO was added to E3.

Sample preparation for live imaging
For upright confocal live imaging, embryos were dechorionated, mounted in 0.3%-0.5% low melting point (LMP) agarose (Invitrogen) and placed in custom-made 2% agarose molds. For high-resolution confocal imaging, embryos were mounted in 0.5% LMP agarose on glass bottom dishes (MatTek). Fixed samples were mounted in 0.5%-1% LMP and placed in custom-made 2% agarose molds and imaged on an upright confocal microscope. Live imaging was performed at 28.5 C ± 1 C.

Imaging setups for live and fixed imaging
For confocal imaging of live and fixed samples, Zeiss LSM 880 and 900 upright confocal microscopes equipped with a Zeiss Plan-Apochromat 20x/1.0 water immersion objective were used. Bright-field/fluorescence time-lapse imaging was performed using a Nikon Eclipse inverted wide-field microscope equipped with CFI Plan Fluor 310/0.3 objective (Nikon) and a fluorescent light source (Lumencor). For high-magnification confocal imaging of prechordal plate progenitors, a Zeiss LSM880 inverted microscope, equipped with a Plan-Apochromat 340/NA 1.2 water-immersion objective (Zeiss), was used. In vitro cell cultures were imaged using a Nikon Ti2E inverted wide-field microscope equipped with Plan Apo l 20x/0.75 DIC air objective (Nikon) and a fluorescent light source (Lumencor).

Volumetric analysis of gastrulating zebrafish embryos Data acquisition
For quantifying the interstitial fluid (IF) volume, Tg(actb2:HRAS-EGFP) embryos, marking the plasma membrane of all cells, were dechorionated at the 64-cell stage, the yolk cell (yolk) was labelled with 60 mM Nile Red in E3 (20 min at 28 C), and the IF was labelled as described above (interstitial fluid labelling and manipulation of interstitial fluid amount). Embryos were then mounted in glass capillaries (size 4, inner diameter 2.15 mm) with corresponding Teflon-coated plungers. Matching FEP tube (OD/ID 2.15/1.58 mm) was tightly inserted into the glass capillary. For mounting the sample, an adapted strategy from Kaufmann et al. 46 was used in order to ensure uninterrupted development during imaging: first, 150 ml of 0.5% LMP agarose containing 1:2000 dilution of fluorescent beads (500 nm, red) was aspirated into the FEP tube using a plunger, after which an embryo was aspirated together with 300 mM of E3 media, resulting in final LMP agarose concentration of 0.1-0.2% within the tube. The sample was held vertically until the embryo settled at the lower portion of the FEP tube. Volumetric imaging of gastrulating zebrafish embryos, with the plasma membrane, IF and yolk labelled, was performed using a Zeiss Lightsheet Z.1 light-sheet microscope, equipped with Plan Apochromat 10x/0.5 water immersion detection and two 5x/0.1 air illumination objectives, in a heating chamber at 28.5 C. Two views (0 and 180 ) were acquired, each containing image stacks of 100-120 z-planes at 3-4 mm spacing. The frame rate of the multi-view recording was 3-5 min. For each timepoint and channel, the embryo was illuminated from both sides (dual illumination) and the emitted light was collected at a perpendicular angle. Next, the capillary containing the sample was turned by 180 and the illumination was repeated. Hence, for each timepoint and channel, the dataset consists of 4 views, representing the same embryo from different orientations. The 4 views were reconstructed using the BigStitcher plugin in Fiji. 41,47 The fluorescent beads contained in the sample capillary were then localised and used as reference points for estimating the relative shift between the 4 views. Multi-view reconstruction was performed using these reference points (precise descriptor-based method, default parameters). To exclude imaging artefacts caused by light diffraction due to the opaque yolk structure, a 2x2x2 tiling approach was implemented, resulting in 8 equally sized blocks spanning the embryo. When fusing the 4 views together into one new 3D image stack, for each view only those tiles were included that had sufficient image quality (i.e. the areas of the embryo having the best orientation relative to illumination and detection objective). By selecting tiles for fusion specifically for each view, image slices showing increased light diffraction due to an obstructed light path were excluded. Excluded image slices were then compensated for by selecting the corresponding tiles in another view, thereby including only tiles with good image quality. This approach resulted in enhanced overall image quality of the embryo for all timepoints and channels. The fused dataset comprising all channels and timepoints was down-sampled 4x (in x,y,z) and saved as a.tiff file.

Data analysis
The total embryo volume was computed by segmenting the embryo outline using the expression of HRAS-EGFP labelling the plasma and yolk membranes. Direct thresholding was not possible due to little contrast within the yolk of the embryo. Therefore, a segmentation pipeline based on supervised pixel-classification using the software ilastik 42 (ver. 1.3.3.) was established. Ilastik intensity and texture features computed at three scales (1, 1.6, and 3.5s) were used to represent each voxel. After interactive labelling of 5000 foreground (i. e. embryo) and 2000 background voxels over the entire time range, a Random Forest classifier was trained and used to predict volumetric probability maps for all analysed embryos. The resulting probability maps were smoothed with an isotropic Gaussian filter (3s) before thresholding at probability 0.5, yielding the initial embryo segmentation. In a few cases, holes in the segmentation were observed in low contrast yolk regions, which were addressed by 3D binary morphological hole-filling to compute the final segmentation. Interstitial fluid volume was computed by segmenting the Alexa647 signal, specifically labelling the interstitial fluid using pixel-classification in ilastik. Intensity and texture features were computed at a wide range of scales and thresholded the resulting maps at probability 0.5, resulting in interstitial fluid segmentation for each time point. Yolk volume was computed by segmenting the yolk cell labelled with Nile Red using the marker-based watershed algorithm. First, a boundary indicator map was computed using Gaussian gradient magnitude with s ranging from 1.7 to 2.2. This map exhibited high values at the interface of background and embryo, but also between yolk and blastoderm. Next, a small set of voxel markers for background, yolk and blastoderm was defined. Background markers were chosen at the edges of the image stacks and a yolk marker in the centre of the image stack. Markers for the blastoderm were chosen manually from inside the blastoderm with up to 3 distinct marked voxel regions. Then, the watershed algorithm growed all semi-automatically marked voxels spatially in accordance with the supplied boundary indicator map. Low values in the boundary map allowed marker growth, whereas high values in the boundary map inhibited growth. The algorithm stopped when all pixels were either assigned to background, yolk, or blastoderm. All resulting embryo region segmentations were visually inspected. Finally, the resulting segmentations were masked with the embryo segmentation from above to ensure that no voxels outside of the embryo were assigned to either yolk, blastoderm or interstitial fluid. The volume of each region was computed by counting the assigned voxels and multiplication with the according voxel size. All post-processing steps were done with custom Python scripts (ver. 3.8) using the scipy (ver. 1.7.0) and scikit-image libraries (ver. 1.18.1).

qRT-PCR (quantitative real-time PCR)
Wild type (wt) zebrafish embryos (AB background) were collected after laying and half was used as control and the other half was Ouabain-treated from 0.5-3.3 hpf, after which the inhibitor was washed off. Embryos were then dechorionated in E3, and developed until 8 hpf. Total RNA was extracted from 10 embryos per sample using Trizol (Invitrogen). To ensure complete removal of genomic DNA, DNA-free DNA Removal Kit (Thermo Fisher Scientific) was used. cDNA library was generated with the iScript Reverse Transcription Supermix (BIO-RAD) kit according to manufacturer's instructions. The same amounts of mRNA were used from different samples for the qRT-PCR experiment. Experiments were repeated 5x independently in triplicates. Absolute QPCR SYBR Green Mixes (Thermo Scientific) and Bio-Rad C1000 Thermal Cycler was used for qRT-PCR reaction. As negative control, the provided NO-RT Supermix, containing no reverse transcriptase was used. For positive control and normalisation, the housekeeping gene elongation factor1a (ef1a) was used. To generate the standard curve and determine the linear range, a series of cDNA dilutions were tested to finally use a 1:10 cDNA dilution. The following primers were used at a final concentration of 142 nM: hgg1 (5' -CTATTATGAGAGAGCAT GTAGCAG -3' and 5' -CATGTATATGTAGCCTTTGTCAC -3'), ef1a (5' -TCTCTCAATCTTGAAACTTAT -3' and 5' -AACACCCAGGCG TACTTGAA -3') Whole mount immunofluorescence (WMIF) Embryos were fixed in 4% paraformaldehyde (PFA) overnight at 4 C. After fixation, they were washed in 0.1% Triton X-100 in 1xPBS (PBT) and dechorionated, followed by permeabilization for 1 h in 0.5% PBT and incubation in blocking solution (0.5% Triton X-100, 1% dimethylsulfoxide (DMSO) and 10% goat serum in 13PBS) for 3 h at RT. Embryos were then incubated with the primary antibody embryos. Embryos were incubated at room temperature (RT) for 30 min post-implantation for wound healing, after which the IF was labelled, as described above (interstitial fluid labelling and manipulation of interstitial fluid amount). At 50% epiboly, embryos with correct localisation of the magnetic bead (at the YSL-deep cell/epiblast boundary) were selected for the experiment. The experiment was performed on an inverted Leica SP5 confocal microscope equipped with micropipettes aspiration and magnetic tweezer systems. To prevent it from moving towards the magnetic coil when the magnetic field was switched on, the embryo was held on the yolk cell (YC) side by a blunt micropipette (OD 80 mm/30 ), applying a slight aspiration without deforming the surface of the YC. The embryo, with a magnetic bead facing the objective, was oriented laterally, and the bead was pulled in parallel to the animal-vegetal (AV) axis via a locally applied magnetic field during continuous imaging. IF distribution profiles were quantified along the axis of controlled displacement of a magnetic bead following the same method as described above (analysis of interstitial fluid relocalisation at the dorsal side).

Interstitial fluid relocation by micropipette pushing
To determine whether pushing forces applied on the surface of the blastoderm can trigger relocalization of interstitial fluid (IF) from the enveloping layer (EVL)-deep cell (DC) boundary to the yolk syncytial layer (YSL)-DC boundary, a micropipette with an outer diameter of 120 mm was pressed against the EVL in MZoep embryos at 6 hpf where the IF was labelled with Alexa647 as described above (interstitial fluid labelling and manipulation of interstitial fluid amount). IF relocalization was measured around a 15mm z-section around the pipette by segmenting the IF as described above (analysis of interstitial fluid relocalisation at the dorsal side) and determining the fluid fraction. The fluid fraction was measured at the EVL-DC and YSL-DC boundaries to determine the degree of IF relocalisation in response to the applied pushing forces.

QUANTIFICATION AND STATISTICAL ANALYSIS
Graphpad Prism 7 was used for statistical analysis. All data were tested for normal distribution with D'Agostino & Pearson normality test. In case data passed normality test (alpha = 0.05), then a parametric test, such as unpaired t-test or ordinary one-way ANOVA (in the case of multiple comparisons) was used. In case the data did not pass the normality test, a non-parametric Mann-Whitney test was used. The individual statistical tests used for experiments are mentioned in the figure legends. In general, n was considered as the number of independent embryo experiments.